Core Data for WOSP 2020 3C shared task.

collect_core_data.py

API_KEY="" # Get from https://core.ac.uk/services/api/
CORE_REQUEST_URL=f"https://core.ac.uk:443/api-v2/articles/get?metadata=true&fulltext=true&citations=true&similar=false&duplicate=false&urls=false&faithfulMetadata=false&apiKey={API_KEY}"
def get_paper_data(core_ids, batch_size=10):
    core_ids = list(core_ids)
    for i in range(0, len(core_ids), batch_size):
        batch = core_ids[i:i+batch_size]
        resp = requests.post(CORE_REQUEST_URL, json=batch)
        batch_resp_json = resp.json()
        yield batch_resp_json
        print(f"Found {len(batch_resp_json)} responses. Sleeping for 2 seconds.")
        time.sleep(2)

CORE_DATA_PATH=Path("./core_data.jsonl")
core_ids = [] # Get from data, see notebook: df.core_id.unique().tolist()
if not CORE_DATA_PATH.exists():
    print(f"File {CORE_DATA_PATH} does not exists.")
    print(f"Collecting data via the CORE API")
    core_data = sum(get_paper_data(core_ids, batch_size=10), [])
    print(f"Found {len(core_data)} records. Writing to {CORE_DATA_PATH}")
    with open(CORE_DATA_PATH, "w+") as fp:
        for d in core_data:
            print(json.dumps(d), file=fp)

core_data.jsonl

{"status": "OK", "data": {"id": "158977742", "authors": ["Thanapalasingam, Thiviyan", "Osborne, Francesco", "Birukou, Aliaksandr", "Motta, Enrico"], "citations": [], "contributors": [], "datePublished": "2018-10-28", "description": "Major academic publishers need to be able to analyse their vast catalogue of products and select the best items to be marketed in scientific venues. This is a complex exercise that requires characterising with a high precision the topics of thousands of books and matching them with the interests of the relevant communities. In Springer Nature, this task has been traditionally handled manually by publishing editors. However, the rapid growth in the number of scientific publications and the dynamic nature of the Computer Science landscape has made this solution increasingly inefficient. We have addressed this issue by creating Smart Book Recommender (SBR), an ontology-based recommender system developed by The Open University (OU) in collaboration with Springer Nature, which supports their Computer Science editorial team in selecting the products to market at specific venues. SBR recommends books, journals, and conference proceedings relevant to a conference by taking advantage of a semantically enhanced representation of about 27K editorial products. This is based on the Computer Science Ontology, a very large-scale, automatically generated taxonomy of research areas. SBR also allows users to investigate why a certain publication was suggested by the system. It does so by means of an interactive graph view that displays the topic taxonomy of the recommended editorial product and compares it with the topic-centric characterization of the input conference. An evaluation carried out with seven Springer Nature editors and seven OU researchers has confirmed the effectiveness of the solution", "fullText": "Ontology-Based Recommendation of Editorial Products \nThiviyan Thanapalasingam1, Francesco Osborne1, \nAliaksandr Birukou2, Enrico Motta1 \n \n1 Knowledge Media Institute, The Open University, MK7 6AA, Milton Keynes, UK \n{thiviyan.thanapalasingam,francesco.osborne,enrico.motta}@open.ac.uk  \n2Springer-Verlag GmbH, Tiergartenstrasse 17, 69121 Heidelberg, Germany  \naliaksandr.birukou@springer.com \nAbstract. Major academic publishers need to be able to analyse their vast \ncatalogue of products and select the best items to be marketed in scientific \nvenues. This is a complex exercise that requires characterising with a high \nprecision the topics of thousands of books and matching them with the interests \nof the relevant communities. In Springer Nature, this task has been traditionally \nhandled manually by publishing editors. However, the rapid growth in the \nnumber of scientific publications and the dynamic nature of the Computer \nScience landscape has made this solution increasingly inefficient. We have \naddressed this issue by creating Smart Book Recommender (SBR), an ontology-\nbased recommender system developed by The Open University (OU) in \ncollaboration with Springer Nature, which supports their Computer Science \neditorial team in selecting the products to market at specific venues. SBR \nrecommends books, journals, and conference proceedings relevant to a \nconference by taking advantage of a semantically enhanced representation of \nabout 27K editorial products. This is based on the Computer Science Ontology, \na very large-scale, automatically generated taxonomy of research areas. SBR also \nallows users to investigate why a certain publication was suggested by the \nsystem. It does so by means of an interactive graph view that displays the topic \ntaxonomy of the recommended editorial product and compares it with the topic-\ncentric characterization of the input conference. An evaluation carried out with \nseven Springer Nature editors and seven OU researchers has confirmed the \neffectiveness of the solution. \nKeywords: Recommender Systems, Ontology, User Interface, Scholarly \nOntology, Scholarly Data. \n1 Introduction \nMajor academic publishers need to be able to analyse their vast catalogue of editorial \nproducts and make data-driven decisions to ensure they are showcasing the right \nproducts to the right target market. This is a complex exercise that requires \ncharacterising with a high precision the topics of thousands of books and matching them \nwith the interests of the relevant scientific communities.  \nIn Springer Nature, this task has traditionally been handled manually by publishing \neditors, who tend to rely on their domain knowledge and their personal experience for \nselecting the books to be marketed at scientific venues. In addition to this, they typically \nuse Springer.com1 for searching publications associated with keywords relevant to the \nconferences in question and find additional information by querying their internal \n                                                        \n1 http://www.springer.com/  \n database of editorial products. This approach lacks a user-friendly interface and can be \nvery time-consuming, since it requires editors to manually browse a large and fast-\ngrowing catalogue of publications.  For example, in order to select books for the \nInternational Semantic Web Conference one might want to search for all the \npublications produced in the last three years that have been authored by well-known \nresearchers who are likely to attend the event. While the editorial products are tagged \nwith product market codes characterizing their topics, these are only limited to high-\nlevel research fields, such as \u201cArtificial Intelligence\u201d and \u201cDatabase Systems\u201d. The \nresults of the editor queries may thus include hundreds of items. Another issue is that \nkeyword-based queries do not take in consideration the relationships between topics \nand may miss pertinent publications that do not contain specific strings. For instance, \nsearching all books about \u201contology matching\u201d may miss publications about \u201contology \nalignment\u201d. \nIn this paper, we present Smart Book Recommender (SBR)2, an ontology-based \nrecommender system developed by The Open University (OU) in collaboration with \nSpringer Nature (SN) for supporting their Computer Science editorial team in selecting \nproducts to market at specific venues. SBR recommends books, journals, and \nproceedings by taking advantage of a semantically enhanced representation of about \n27K editorial products. In order to do so, we characterized all SN publications \naccording to their associated research topics by exploiting the Computer Science \nOntology (CSO), a large-scale automatically generated taxonomy of research areas [1]. \nFurthermore, SBR allows users to investigate why a certain publication was suggested \nby means of an interactive graph view that compares the topics of the suggested \npublication with those characterizing the input conference. \nThe rest of the paper is organized as follows. In Section 2, we discuss Smart Book \nRecommender in terms of its knowledge base, its architecture, and its user interface. In \nSection 3, we present the results of the user study. In Section 4, we discuss the steps \nrequired for large-scale deployment of the technology within the company. In Section \n5, we review the state of the art and in Section 6 we conclude outlining future directions \nof research and development. \n2 Smart Book Recommender \nSmart Book Recommender takes as input a conference series and returns a list of \neditorial products that may be of interest for the attendees of the conference. This is \nachieved by representing SN books as a set of research topics drawn from a large-scale \nComputer Science ontology, and ranking them according to their similarity with a \ntopic-centric characterization of the conference. For instance, given the conference \nseries \u201cInternational Semantic Web Conference\u201d (ISWC), SBR will return the books, \njournals, and conference proceedings that are characterized by a set of research topics \nsimilar to the one of ISWC, e.g.,  the \"Handbook of Semantic Web Technologies\u201d and \n\u201cProceedings of the European Semantic Web Conference\u201d. The primary purpose of \nSBR is to provide a concise and relevant list of publications that editors can quickly \nreview to decide which books to market during a conference. However, it can also be \nused by researchers for finding publications relevant to a certain venue of interest. \n                                                        \n2 A demo of SBR is available at http://rexplore.kmi.open.ac.uk/SBR-demo. \n SBR provides the web interface shown in Figure 1. It works according to three main \nsteps: \n1) It represents journals, books, and conferences according to the metadata of their \nchapters/articles and uses the Smart Topic API [2] to characterize each of them \nwith a semantically enhanced topic vector. \n2) It computes the similarity between conferences and other editorial products and \nsaves the results in a database.  \n3) For a given input conference, it returns a list of relevant editorial products, \nranked by their topic-centric similarity with the conference in question and \nfiltered in accordance with a number of user preferences.  \nIn order to make it easier for users to understand why a certain item was suggested, \nSBR offers also an interactive graph view that displays the topic taxonomy of the \nsuggested editorial product and compares it with the input conference. \nIn the next sections, we will discuss the system in detail. In Section 3.1, we describe \nthe knowledge bases used by SBR. In Section 3.2, we discuss the Smart Topic API, a \nservice for tagging books with a set of relevant topics. In Section 3.3, we describe how \nwe compute the similarity scores. Finally, in Section 3.4, we present the user interface.  \n \n \nFigure 1. The main interface of SBR.  \n2.1 Background data \nSBR relies on two background knowledge bases: a large database of metadata \ndescribing publications and the Computer Science Ontology 3.  \nThe database of metadata contains titles, abstracts, keywords and other information \ndescribing the chapters of about 27K books and 320 journals published by SN in the \nfield of Computer Science. In the case of conference proceedings, journals, and edited \nbooks, each chapter is usually a research paper. Since we represent conferences \n                                                        \n3 http://skm.kmi.open.ac.uk/cso \n according to their proceedings, SBR can only take as input conferences published by \nSpringer Nature.   \nThe Computer Science Ontology (CSO) [3] is a large-scale and granular ontology \nof research topics that was created automatically by running the Klink-2 algorithm [1] \non the Rexplore dataset [4]. This consists of about 16 million publications, primarily in \nthe field of Computer Science. The Klink-2 algorithm combines semantic technologies, \nmachine learning, and background knowledge from a number of web sources, including \nDBpedia, calls for papers, and web pages, to identify research topics and their \nrelationships from a given corpus of publications. CSO uses the Klink data model4, \nwhich is an extension of the BIBO ontology 5, which in turn builds on SKOS6. This \nmodel includes three classes of semantic relations: relatedEquivalent, which indicates \nthat two topics can be treated as equivalent for the purpose of exploring research data; \nskos:broaderGeneric/skos:narrowerGeneric, which indicate that a topic is a super-\narea/sub-area of another one; and contributesTo, which indicates that the research \noutputs of one topic significantly contribute to the research work within another. The \nversion of CSO used in the current prototype consists of approximately 15K semantic \ntopics linked by 70K relationships.  \n \n \nFigure 2. The Smart Topic API architecture.  \n2.2 The Smart Topic API  \nThe ongoing collaboration between The Open University and Springer Nature has \nproduced several semantic solutions for supporting the SN editorial team. These include \nthe Smart Topic API  [2, 5], an online service for automatically tagging publications \nwith a set of relevant topics from CSO. This API supports a number of applications, \nincluding Smart Book Recommender,  Smart Topic Miner [5], the Technology-Topic \nFramework [6], a system that forecasts the propagation of technologies across research \ncommunities, and the Pragmatic Ontology Evolution Framework [7], an approach to \n                                                        \n4 http://technologies.kmi.open.ac.uk/rexplore/ontologies/BiboExtension.owl  \n5 http://purl.org/ontology/bibo/  \n6 http://www.w3.org/2004/02/skos/  \n ontology evolution that is able to select new concepts on the basis of their contribution \nto specific computational tasks. \nFigure 2 shows the architecture of the system. The Smart Topic API takes as input \na JSON containing the metadata of a book and returns its description in terms of a \ntaxonomy (or optionally a list) of topics, in which each topic is associated with the \nnumber of chapters in which it appeared. It works as following: \n1) For each topic in CSO (e.g., Semantic Web), it associates all the chapters that \ncontain the label of the topic or the label of any relatedEquivalent or \nskos:narrowerGeneric (e.g., Linked Data) in the title, the abstract, or the \nkeyword field.  \n2) It reduces the list of topics associated with a book to a user-friendly number by \nmeans of set covering algorithms [5]. \n3) It infers from the topics the product market codes (PMCs) used by SN as \ninternal classification. It then returns a taxonomy of research topics and PMCs \nassociated with the (number of) chapters in which they were detected.  \n \nThe Smart Topic API powers Smart Topic Miner (STM) [5], a web interface that \nsupports SN editors in classifying proceedings. STM allows editors to submit one or \nmore proceedings, uses the API to annotate them, and then displays them as a taxonomy \nof research topics. It also offers a number of other options, such as the ability of \nexplaining why a certain topic is relevant by showing the full set of sub-topics that were \nused to infer it. STM halves the time needed for classifying proceedings from 20-30 to \n10-15 minutes and allows this task to be performed also by assistant editors, thus \ndistributing the load and reducing costs [5] . \n \n \nAlgorithm 1. The SBR algorithm   \n2.3 Similarity Computation \nIn order to characterize specific journals, books, and conferences we group the \npublications as following: 1) for books, chapters are grouped by the book DOI; 2) for \njournals, the articles are grouped using the journal DOI and their publication year (e.g., \n Journal of Intelligent Information Systems in 2016), and 3) for conferences, papers are \ngrouped using unique conference identifiers and considering only articles from the last \nfive years. We use the persistent identifiers for conferences and conference series \nintroduced in the Linked Open Data Conference Portal [8] and recently migrated to \nSciGraph [9]. Such identifiers make sure that the conference series links all relevant \nconferences, regardless of name changes (e.g., after a few years the \u201cEuropean \nSemantic Web Conference\u201d became the \u201cExtended Semantic Web Conference\u201d) and \nacronyms. \nIn an earlier version of SBR, we considered specific editions of conferences \u2013e.g., \nISWC 2013. However, on the basis of feedback from the editors, it was decided to \nconsider full conferences series rather than individual editions. This solution simplifies \nthe interface and allows us to reduce possible bias from specific conference editions, \nwhich may be affected by trendy topics exhibiting a transient burst of popularity.  \nWe employ the Smart Topic API to associate each item with a vector in which the \nelements represent research topics and their value is the number of chapters/papers in \nwhich the topic was detected. Henceforth, this value will be referred to as topic weight.   \nWe exploit this vector representation for computing the similarity between the \nconference series and the editorial products, as described in Algorithm 1. Since, the \nSmart Topic API associates publications containing topic T also with the \nrelatedEquivalent and skos:broaderGeneric of T (see S2.2), the resulting vectors allow \nus to match publications that refer to the same concepts (e.g., \u201cDeep Learning\u201d) with \ndifferent key phrases (\u201cDeep Neural Networks\u201d), at different granularity levels \n(\u201cMachine Learning\u201d, \u201cAI\u201d). \nWe assess the similarity of two semantic vectors using the cosine similarity [10], \nsince this measure relies on the orientation but not the magnitude of the topic weights \nin the vector space, allowing us to compare editorial products associated with a different \nnumber of chapters. The similarity computation is carried out offline. \nSince it is computationally infeasible to calculate the cosine similarity between each \nbook in the SN dataset, we first prune the number of candidate pairs by calculating their \nJaccard index, which is a more lightweight similarity metric, and selecting only the \nones that yield a value higher than a threshold. A data analysis revealed that by applying \na threshold of 0.125 we halve the number of candidate pairs while still producing very \ngood results. Finally, we save the cosine similarity of a pair in the database if it is greater \nthan 0.5, since according to the editors, recommendations with similarity < 0.5 are \nunhelpful. \n2.4 The web interface  \nFigure 1 shows the user interface of SBR. The user can select a conference by typing \neither the conference name (e.g., \u201cInternational Semantic Web Conference\u201d) or its \nabbreviated form (e.g., \u201cISWC\u201d). In Figure 1 the user has selected ISWC and SBR is \nshowing the top fifteen topics that characterize this venue.   \nWhen the user selects a conference, the corresponding conference ID along with the \nother user preferences (e.g., publication type, year, maximum number of results) are \nsent as JSON file to the backend via a GET request. The backend is a REST API, which \nretrieves all relevant publications that meet the criteria and returns the results as a JSON \nfile, which is then visualized by the web interface. The API was developed in Python \n and the data are pulled from a MariaDB database, while the frontend uses HTML5 and \nJavascript. \nHere, we briefly describe the settings available to the users, to allow them to \ncustomise the behaviour of the system.  \n\u2022 Types of publication \u2013 Users can specify which types of editorial products should \nbe included in the results. Currently, these include books, journals, and (other) \nconference proceedings. \n\u2022 Publication year \u2013Users can filter results to include only the ones published in a \nspecified time interval. By default, this interval is set to the last three years. \n\u2022 Maximum number of results \u2013 Users can set the maximum amount of results \naccording to their needs. This functionality is provided as normally editors can \nonly select a limited number of books to market during a conference.  \n\u2022 Filter publications by authors and editors \u2013 Users can narrow down the \nrecommendations to books authored or edited by an individual or a group of \nacademics using this free text field. This functionality is provided as editors often \nfocus on marketing editorial products produced by key researchers with high \nvisibility in the research fields relevant to a conference.  \n\u2022 Exporting data \u2013 Once a list of recommendations has been generated, it is possible \nfor the user to export the results as a CSV or JSON file. These files are typically \nused by publishing editors to submit an order to the Exhibit Department, which \ntakes care of dispatching the selected products to the conference. \n \n \nFigure 3. Recommended SN books for ISWC. \n \nFigure 3 shows the recommendation list that is loaded via an AJAX request after the \nuser has selected a conference. The results are shown as cards and sorted in descending \norder of similarity. Each publication is summarized with respect to its key elements. \nThese include title, publication year, the fifteen most significant topics, and the overall \nsimilarity score with the input conference. We display the authors of a book wherever \nthere are less than five authors, otherwise we display editors.  \nThe users can interact with each card by: \n \u2022 Examining the publication on SpringerLink 7 \u2013 A hyperlink on the publication title \nredirects users to the relevant SpringerLink page. This enables editors to collect \nadditional information regarding the publication, such as the authors of individual \npapers and the abstracts.  \n\u2022 Providing feedback for a specific card \u2013 We provide a binary feedback system that \nuses emoticon radio buttons to allow users to express their view on a \nrecommendation. The feedback is used to improve the recommender engine.  \n\u2022 Opening the graph view interface \u2013 By clicking on the \u201cvisualize topic taxonomy\u201d \nbutton, users can access a graph view, shown in Figure 4, which makes it easier for \nthem to make sense of the relationship between the selected output and the input \nconference.  \n \n \nFigure 4. Portion of the graph view showing the taxonomies of the topics \nassociated with the input conference and one of the recommended editorial products. \n \nThe graph view 8  visualizes editorial products according to their taxonomy of \nresearch topics derived from the Computer Science Ontology. The purpose is allowing \nusers to understand why a certain product was recommended and how its associated \ntopics intersects with the ones characterizing the input conference. As an example, in \nFigure 4 we show the comparison between the topic-centric characterization of the \n\u201cHandbook of Semantic Web Technologies\u201d9 and the one of the International Semantic \nWeb Conference.  The user can choose whether to visualise only the topics of the \nconference, those of the recommended publication, their intersection, or all topics of \nthe two items. Hovering over a topic shows the number of chapters/papers within the \n                                                        \n7 https://link.springer.com/  \n8 The graph view was realised in JavaScript, using the D3.js library (https://d3js.org/). \n9 https://link.springer.com/referencework/10.1007%2F978-3-540-92913-0  \n publication that are associated with the topic. A slider above the interface allows users \nto filter topics according to their weight.  \n3 Evaluation \nWe evaluated Smart Book Recommender by means of a user study involving seven SN \neditors and seven OU researchers. The goal of the study was to assess both the usability \nof SBR and also the quality of its recommendations. We structured the user study in \nthree phases. First, we provided each subject with a 10 minute introduction to SBR. \nThen we asked them to try the system for approximately 45 minutes and rate its \nrecommendations. Finally, each subject filled a questionnaire about their experience \nwith SBR. \n \nOption Applies to Definition \nBring it Editor only The item is relevant to the conference and the editor would bring this item to the conference and market it. \nRead it Researcher only The item is relevant to the conference and the researcher would want to read it. \nRelevant Both \nThe item is relevant to the conference, but the editor does not \nconsider it suitable to be marketed or the researcher does not \ndesire to read it. This could be for a variety of marketing or \npersonal reasons. \nDebatable Both \nWhether the recommended item is relevant to the conference is \nopen to discussion and different people may have different \nopinions. \nIrrelevant Both The recommended item is not relevant to the conference and should not be recommended. \nTable 1. Options available to editors and researchers for rating recommendations.  \n \nWhile editors are the main users of the system, we also evaluated SBR with a number \nof researchers, given that the whole point of the application is to assist editors in \nselecting editorial products that researchers are likely to be interested in. The expertise \nof the evaluators covered a variety of Computer Science topics, including but not \nlimited to Robotics, Semantic Web, Software Engineering, HCI, AI, Computational \nBiology, and Wireless Networks. \nWe assessed the quality of the results by considering the \u201cbring it\u201d, \u201cread it\u201d, and \n\u201crelevant\u201d books as relevant instances and computing the Precision @10, a standard \nmetric for evaluating ranked lists of items.   \nThe material produced for this evaluation is publically available at \nhttp://rexplore.kmi.open.ac.uk/SBR_eval_data and on FigShare 10. \n3.1 Quantitative Analysis  \nWe assessed the performance of SBR in suggesting relevant publications, by asking \nusers to choose two conferences in their fields of expertise and then rate SBR \nrecommendations. For each conference, SBR suggested 20 books and 10 conference \nproceedings. To keep recommendations consistent, we considered all books and \nproceedings published between 2005 and 2018, regardless of the authors and editors. \n                                                        \n10 https://doi.org/10.6084/m9.figshare.6087032.v2   \n We asked the users to rate each item by selecting one of the four options presented in \nTable 1. The sessions were video-recorded to allow further analysis.  \nThe Precision @10 is 76.8% for the books and 75.4% for the proceedings. It thus \nseems that there is not much difference in the quality of the recommendations regarding \nthese two editorial products. \nFigure 5 shows the percentage of recommendations that were tagged as \u201cbring it/read \nit\u201d, \u201crelevant\u201d, \u201cdebatable\u201d, or \u201cirrelevant\u201d by the users. Editors rated \u201cbring it\u201d or \n\u201crelevant\u201d 72.9% of the recommendations while researchers rated \u201cread it\u201d or \n\u201crelevant\u201d the 66.8% of them. In total, only 10.5% of the recommendations were rated \nas irrelevant by the editors and 7.2% by the researchers.  \n \n \nFigure 5. SBR performance as rated by the evaluators (SN editors labelled 1-7 and \nOU researchers 8-14). The results of the 28 test cases were aggregated by user. \n \nEditors would bring to the conference 31.9% of the recommended publications, \nconsidering the others not marketable for a variety of reasons, even when they were \nrelevant. On the other hand, researchers would read 14.5% of it. This discrepancy may \nbe explained by the fact that editors and researchers apply different decision-making \nstrategies when choosing whether to \u201cbring\u201d or \u201cread\u201d a publication. Researchers are \nmainly interested in publications that address their specific needs and they consider also \nthe time invested in reading it and the price. Conversely, editors take into account the \npreferences of a large group of people and consider a variety of other dimensions, such \nas how much the book sold in previous years, the popularity of the authors within the \ncommunity, the potential audience size, and so on.  \n 3.2 Qualitative Analysis  \nThe questionnaire consisted of three sections: i) an assessment of the evaluators\u2019 \nbackground and expertise, ii) five open questions, and iii) a standard System Usability \nScale (SUS) questionnaire to assess the usability of SBR. On average, the editors had \n15 years of experience in their role and extensive experience in selecting books for \nconferences. Three of them had more than 20 years of experience in their field. The OU \nresearchers had an average seniority of about 5 years. \n \n \nFigure 6. SUS questionnaire results (editors labelled 1-7 and researchers 8-14). \n \nWe will first summarize the answers to the open questions.  \n \nQ1. How do you find the interaction with the SBR interface? Both groups found \nthe user interface very intuitive. Most attributed this to the \u201csimple\u201d and \u201cwell-\norganised\u201d layout of SBR and the ability to perform queries with little user input. One \nresearcher mentioned that there was a learning curve but it was \u201ceasy to pick up\u201d, and \none editor suggested to make the text input field for searching conference series more \nnoticeable. \nQ2. How effectively does SBR support you in selecting relevant publications? \nSome editors placed the accuracy of the recommended conference proceedings higher \nthan that of the books. One editor felt that some recommended titles were generic, \npossibly due to the \u201clarge margin of error associated with vast selections of \nconferences\u201d and two pointed out that it would be beneficial to be able to select \nparticular book types, such as handbooks, textbooks and monographs. \nQ3. What are the most useful features of SBR? Five researchers found the visual \nanalytics of taxonomies useful for understanding similarities.  Three editors appreciated \nthe hyperlinks to the Springer product page. Two researchers and one editor found \nparticularly helpful the option of viewing books and conferences independently.  \n Q4. What are the main weaknesses of SBR?  There was general agreement \nbetween editors and researchers that supporting only Springer published conferences is \na significant drawback. Three editors indicated that some of the book titles relevant to \ntheir conference were not recommended. Another two mentioned that when searching \nfor books, the system returned also some proceedings (i.e., books from the LNCS \nseries). \nQ5. Can you think of any additional features to be included in SBR? Two \nresearchers and two editors would like to have the ability of modifying the automatic \nrepresentation of the input conference by adding or removing some topics. Some editors \nwould like to have direct links to conference pages and additional information about \npublications, e.g., the main subject discipline and whether they are open access or not. \n \nThe last part of the user survey consisted of a SUS questionnaire, a standard tool for \nassessing the usability of an application. The SUS questionnaire includes 10 questions \non a 1 to 5 scale, where 1 is the most negative assessment and 5 the most positive. The \naverage system is expected to score 68 out of 100. The editors and the researchers \nyielded respectively an average SUS score of 77.1\u00b115.2 and 80.3\u00b111.3, which converts \nin a percentile rank of about 75%.  \nFigure 6 shows the answers of the users to four SUS questions. The users believed \nthat SBR was easy to use (with an average score of 4.4\u00b10.7) and its functions where \nwell integrated (3.9\u00b10.6). They did not think that it was complex to use (1.4\u00b10.8) or \nthat they would need the help of a technical person to use it in the future (1.5\u00b10.5). \n3.3 Informal feedback beyond Computer Science editorial team \nIn addition to the formal evaluation reported in this section, we have also presented the \nSBR tool to a wider group of publishing editors and editorial assistants at Springer \nNature. The fifteen participants (3 sessions with 5 participants each) first saw a short 3-\nminute demo of the tool and then took part in a 10-12 minute session where they were \nencouraged to ask questions and suggest improvements. \nThe participants saw strong potential for the SBR tool over current practices, which \ninclude \u201cask colleagues for relevant books and journals via e-mail, hoping they have \ntime to reply and are in the office\u201d and \u201ccompile a list of relevant titles using various \nsystems, actually developed for other purposes\u201d. In particular, they appreciated the time \nrange and type of product filters and the support for searching for books by keynote \nspeakers. They also suggested areas for further improvements, such as the ability of i) \ndirectly querying the system with a list of research topics; ii) looking up people on the \neditorial board of a journal; iii) expand the scope of the system to other disciplines (e.g., \nMathematics).  \n3.4 Discussion and Limitations \nSBR obtained a more than satisfactory performance in recommending relevant editorial \nproducts and received a high score in term of usability. Nonetheless, the evaluation \nhighlighted some issues that we intend to address in future versions.  \nA first concern that was mentioned by a number of users is that SBR currently \nprovides recommendations only for conferences which proceedings are published by \n Springer Nature11, thus not providing support for marketing activities outside these \nconferences. In order to include more conferences, we need to also access to the \nconference proceedings published by other editors. We are thus exploring the option of \nusing datasets such as CrossRef 12, Dimensions 13, OpenCitations [11], and Core [12]. \nAnother issue arising from the evaluation is that sometimes the topic characterization \nof books with few chapters is quite sparse. In these cases, considering only title, abstract \nand keywords may not allow to identify enough topics to allow a fair comparison with \nthe other editorial products. We are thus considering using also the full text.  \nA third issue that emerged during the evaluation is the coverage of multidisciplinary \npublications. SBR represent topics by means of the Computer Science Ontology, and \ntherefore scarcely covers other fields such as Biology, Engineering, Mathematics, or \nEconomics. Therefore, publications which include other fields in addition to Computer \nScience are sometimes misrepresented, lowering the overall quality of the \nrecommendations. We plan to address this issue by applying the ontology learning \ntechniques utilized to produce the Computer Science Ontology also on other domains \nof science. \nFinally, some users mentioned that they would like the option of modifying the set \nof topics that get extracted from the conference proceedings and is used to produce the \nrecommendations. A further step in this direction would be to allow users to input \ndirectly a set of topics as a query. This would naturally require some significant changes \nto the backend, since currently all the similarity values are precomputed, but it would \nalso allow for more flexibility. Indeed, this solution may also enable us to associate \nusers with a representation of their research interests and automatically produce tailored \nrecommendations. \n4 Next steps for large scale deployment  \nSBR was well received by Springer Nature editors, but we must take some additional \nsteps to fully integrate it into their workflow.  \nIn the first instance, we intend to automatize the process for importing and \nprocessing the most recent editorial items. Currently, we renew our database every four \nmonths by importing a new dump of metadata and recalculating the similarity values. \nThis solution suffers from two limitations: it requires human intervention and the \nsystem is updated only every four-months. We plan to fully automatize this process by \ndeveloping a system for importing new metadata on a daily basis and recomputing \nseamlessly the relevant similarity values. \nIn the second instance, we plan to develop a new version of SBR that will address \nthe most important requests that came up during the user study, as discussed in previous \nsection. \nFinally, we are exploring the ability of SBR to produce collections of documents \nrelevant to certain topics, e.g., all recent publications in the field of Ontology \nEngineering. This has broader implications beyond selecting books for conferences, \nand can help compiling ad-hoc packages for industry or academic institutions in the \n                                                        \n11  Actually, since Springer Nature is one of the largest publishers of Computer Science \nconferences, its coverage of the conferences in this field is very extensive. \n12 http://crossref.org  \n13 https://www.dimensions.ai  \n developing countries. Some initial experiments in this direction have already yielded \npromising results. \n5 Related Work \nRecommender systems are software tools and methods which provide suggestions for \nitems to users, according to their preferences and needs [13]. They are typically \nclassified as collaborative filtering approaches, content-based filtering approaches and \nhybrid approaches [14].  \nContent-based recommender systems [15] rely on a pre-existing domain knowledge \nto suggest items more similar to the ones that the user seems to like. They usually \ngenerate user models that describe user interests according to a set of features [16]. \nWith the advent of the Semantic Web, several recommender systems started to adopt \nontologies for representing both user interests and items [17]. Often these systems use \nan ontology so that, given user interest in an item represented in the ontology, they can \nthen propagate such interest to relevant items and concepts. For example, given a \npositive feedback on \u201cbeagles\u201d, a system may infer (correctly or not) that a user is \ninterested in \u201cdogs\u201d, and more generally in \u201cpets\u201d. SBR exploits a similar mechanism \nwhen it infers that a publication explicitly linked to a topic (e.g., Linked Data) is also \nabout its skos:broaderGeneric concepts in CSO (e.g., Semantic Web). The main \nadvantages of these solutions are i) the ability to exploit the domain knowledge for \nimproving the user modelling process, ii) the ability to share and reuse system \nknowledge, and iii)  the alleviation of the cold-start and data sparsity problems [16, 18]. \nWe will now discuss some of these ontology-based approaches. Sieg et al. [16] \npresent an ontology-based recommender to improve personalised Web searching in \nwhich the user profiles are instances of a reference domain ontology and are \nincrementally updated based on the user interaction with the system. Middleton et al. \n[18] describe a hybrid recommender system that exploit ontologies for increasing the \naccuracy of the profiling process and hence the usefulness of the recommendations. \nThiagarajan et al. [19] use a different strategy by representing user profiles as bags-of-\nwords and weighing each term according to the user interests derived from a domain \nontology. Razmerita et al. [20] describe OntobUM, an ontology-based recommender \nthat integrates three ontologies: i) the user ontology, which structures the characteristics \nof users and their relationships, ii) the domain ontology, which defines the domain \nconcepts and their relationships, and iii) the log ontology, which defines the semantics \nof the user interactions with the system. Birukou et al [21] present an agent-based \nsystem that learns the preferences of experienced researchers and provides specific \nsuggestions to support search for scientific publications.  Colombo-Mendoza et al [22] \npropose RecomMetz, a context-aware mobile recommender system based on Semantic \nWeb technologies. This system introduced some unique features, such as the composite \nstructure of the items and the integration of temporal and crowd factors into a context-\naware model. Finally, Cantador et al. [23] propose a hybrid recommendation model in \nwhich user preferences are described in terms of semantic concepts defined in domain \nontologies. Similar to all these systems SBR builds a semantic representation of the \nitems and exploits the ontology for inferring additional concepts. However, rather than \ncreating a representation of a single user, it characterizes the overall interests of the \nresearch community associated with the proceedings of a conference series. \n SBR builds on the Smart Topic API to represent publications as vectors of research \ntopics. This is a useful representation that is used in a variety of systems for exploring \nthe research landscape [4, 6]. In recent years, we have seen the emergence of several \napproaches to annotating research articles. For instance, DBpedia Spotlight [24] is often \nused for automatically annotating papers with DBpedia concepts. Gabor et al. [25] \nintroduce an approach for annotating scientific corpora with domain-relevant concepts \nand semantic relations. The Dr. Inventor Framework [26] focuses instead on extracting  \nstructured textual contents, discursive characterization of sentences, and graph based \nrepresentations of text excerpts. \n6 Conclusions \nIn this paper, we presented Smart Book Recommender, a semantic recommender \nsystem developed in collaboration with Springer Nature which suggests editorial \nproducts to market at academic venues.  \nA user study involving seven SN editors and seven OU researchers showed that SBR \nwas able to suggest relevant materials and scored high in usability. In particular, \nSpringer Nature editors considered as relevant 72.9% of the SBR recommendations and \nassessed the system as very user friendly, yielding an average SUS score of 77.1. \nWe are now planning to further integrate the SBR tool into the process workflows at \nSpringer Nature. To this purpose, we are going to develop a new version of the system, \nwhich will take into account a variety of suggestions which arose from the user study.  \nAcknowledgements \nWe would like to thank publishing editors at Springer Nature for assisting us in the \nevaluation of SBR and allowing us to access their large repositories of scholarly data. \nReferences \n1. Osborne, F., Motta, E.: Klink-2: Integrating Multiple Web Sources to Generate \nSemantic Topic Networks. In: International Semantic Web Conference. pp. 408\u2013424. \nSpringer (2015). \n2. Osborne, F., Salatino, A., Birukou, A., Thanapalasingam, T., Motta, E.: Supporting \nSpringer Nature Editors by means of Semantic Technologies. In: International Semantic \nWeb Conference. Springer (2017). \n3. Salatino, A.A., Thanapalasingam, T., Mannocci, A., Osborne, F., Motta, E.: The \nComputer Science Ontology\u2009: A Large-Scale Taxonomy of Research Areas. In: \nInternational Semantic Web Conference 2018 , Monterey, CA (USA) (2018). \n4. Osborne, F., Motta, E., Mulholland, P.: Exploring Scholarly Data with Rexplore. In: \nInternational Semantic Web Conference pp. 460\u2013477. Springer, Berlin, Heidelberg \n(2013). \n5. Osborne, F., Salatino, A., Birukou, A., Motta, E.: Automatic Classification of Springer \nNature Proceedings with Smart Topic Miner. In: International Semantic Web \nConference. pp. 383\u2013399. Springer (2016). \n6. Osborne, F., Mannocci, A., Motta, E.: Forecasting the Spreading of Technologies in \nResearch Communities. In: Proceedings of the Knowledge Capture Conference (2017). \n 7. Osborne, F., Motta, E.: Pragmatic Ontology Evolution: Reconciling User Requirements \nand Application Performance. In: International Semantic Web Conference 2018 , \nMonterey, CA (USA). (2018). \n8. Birukou, A., Bryl, V., Eckert, K., Gromyko, A., Kaindl, M.: Springer LOD Conference \nPortal. Demo paper. In: International Semantic Web Conference. Springer, Vienna \nAustria (2017). \n9. Hammond, T., Pasin, M., Theodoridis, E.: Data integration and disintegration: \nManaging Springer Nature SciGraph with SHACL and OWL. In: Proceedings of the \nISWC 2017 Posters & Demonstrations and Industry Tracks. Springer, Vienna, Austria \n(2017). \n10. Salton, G., Buckley, C.: Term-weighting approaches in automatic text retrieval. Inf. \nProcess. Manag. 24, 513\u2013523 (1988). \n11. Peroni, S., Dutton, A., Gray, T., Shotton, D.: Setting our bibliographic references free: \ntowards open citation data. J. Doc. 71, 253\u2013277 (2015). \n12. Knoth, P., Zdrahal, Z.: CORE: Three Access Levels to Underpin Open Access. D-Lib \nMag. 18, (2012). \n13. Ricci, F., Rokach, L., Shapira, B. eds: Recommender Systems Handbook. Springer US, \nBoston, MA (2015). \n14. Burke, R.: Hybrid Web Recommender Systems. In: The Adaptive Web. pp. 377\u2013408. \nSpringer, Berlin, Heidelberg (2007). \n15. Lops, P., de Gemmis, M., Semeraro, G.: Content-based Recommender Systems: State \nof the Art and Trends. In: Recommender Systems Handbook. pp. 73\u2013105. Springer US, \nBoston, MA (2011). \n16. Sieg, A., Mobasher, B., Burke, R.: Web search personalization with ontological user \nprofiles. In: Conference on Information and Knowledge Management. p. 525. ACM \nPress, New York, New York, USA (2007). \n17. de Gemmis, M., Lops, P., Musto, C., Narducci, F., Semeraro, G.: Semantics-Aware \nContent-Based Recommender Systems. In: Recommender Systems Handbook. pp. 119\u2013\n159. Springer US, Boston, MA (2015). \n18. Middleton, S.E., Shadbolt, N.R., De Roure, D.C.: Ontological user profiling in \nrecommender systems. ACM Trans. Inf. Syst. 22, 54\u201388 (2004). \n19. Thiagarajan, R., Manjunath, G., Stumptner, M.: Finding Experts By Semantic Matching \nof User Profiles. In: International Semantic Web Conference Personal Identification and \nCollaborations: Knowledge Mediation and Extraction (PICKME 2008). pp. 7\u201318 \n(2008). \n20. Razmerita, L., Angehrn, A., Maedche, A.: Ontology-Based User Modeling for \nKnowledge Management Systems. In: International Conference on User Modeling. pp. \n213\u2013217. Springer, Berlin, Heidelberg (2003). \n21. Birukou, A., Blanzieri, E., Giorgini, P.: A Multi-Agent System that Facilitates Scientific \nPublications Search. In: International Joint Conference on Autonomous Agents and \nMultiagent Systems. pp. 265\u2013272. ACM Press (2006). \n22. Colombo-Mendoza, L.O., Valencia-Garc\u00eda, R., Rodr\u00edguez-Gonz\u00e1lez, A., Alor-\nHern\u00e1ndez, G., Samper-Zapater, J.J.: RecomMetz: A context-aware knowledge-based \nmobile recommender system for movie showtimes. Expert Syst. Appl. 42, 1202\u20131222 \n(2015). \n23. Cantador, I., Bellog\u00edn, A., Castells, P.: A multilayer ontology-based hybrid \nrecommendation model. AI Commun. 21, 203\u2013210 (2008). \n24. Mendes, P.N., Jakob, M., Garc\u00eda-Silva, A., Bizer, C.: DBpedia Spotlight: Shedding \nLight on the Web of Documents. \n25. G\u00e1bor, K., Zargayouna, H., Buscaldi, D., Tellier, I., Charnois, T.: Semantic Annotation \nof the ACL Anthology Corpus for the Automatic Analysis of Scientific Literature. \n26. Ronzano, F., Saggion, H.: Dr. Inventor Framework: Extracting Structured Information \nfrom Scientific Publications. Presented at the (2015). \n", "identifiers": ["oai:oro.open.ac.uk:55516", "10.1007/978-3-030-00668-6_21"], "publisher": "Springer", "relations": ["http://iswc2018.semanticweb.org/"], "repositories": [{"id": "86", "openDoarId": 0, "name": "Open Research Online", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "baseId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 0, "metadataAdded": 1529987167000, "metadataUpdated": 1591929819000, "depositedDate": 1529881200000}, "subjects": [], "title": "Ontology-Based Recommendation of Editorial Products", "topics": [], "types": [], "year": 2018, "doi": "10.1007/978-3-030-00668-6_21", "oai": "oai:oro.open.ac.uk:55516", "downloadUrl": ""}}
{"status": "OK", "data": {"id": "211273841", "authors": ["Myers, Christina", "Piccolo, Lara", "Collins, Trevor"], "citations": [], "contributors": [], "datePublished": "2018", "description": "This paper presents a participatory methodology to design cards on social issues with the purpose to democratise knowledge among co-designers on the learning content of educational games. Situated on the topic of everyday sexism, the methodology has been developed through an iterative process involving two collaborative workshops, two iterations of card design and a feedback survey. Extracting findings from the workshops and the feedback gathered on the co- designed cards, this paper presents insights that could be used to inform similar studies using cards to inspire and foster reflection on social issues", "fullText": "Open Research Online\nThe Open University\u2019s repository of research publications\nand other research outputs\nCo-designing Cards on Social Issues for Creating\nEducational Games\nConference or Workshop Item\nHow to cite:\nMyers, Christina; Piccolo, Lara and Collins, Trevor (2018). Co-designing Cards on Social Issues for Creating\nEducational Games. In: Proceedings of the 32nd International BCS Human Computer Interaction Conference (HCI\n2018), 4-6 Jul 2018, Belfast, UK.\nFor guidance on citations see FAQs.\nc\u00a9 2018 The Authors\nVersion: Version of Record\nLink(s) to article on publisher\u2019s website:\nhttp://dx.doi.org/doi:10.14236/ewic/HCI2018.164\nCopyright and Moral Rights for the articles on this site are retained by the individual authors and/or other copyright\nowners. For more information on Open Research Online\u2019s data policy on reuse of materials please consult the policies\npage.\noro.open.ac.uk\n\u00a9 Myers et al. Published by  \nBCS Learning and Development Ltd.  \nProceedings of British HCI 2018. Belfast, UK. \nhttp://dx.doi.org/10.14236/ewic/HCI2018.164 \n1 \nCo-designing Cards on Social Issues for \nCreating Educational Games \n \n \nChristina Myers  \nThe Open University \nMilton Keynes, UK \nchristina.myers@open.ac.uk \nLara S. G. Piccolo \nThe Open University  \nMilton Keynes, UK \nlara.piccolo@open.ac.uk \nTrevor Collins \nThe Open University \nMilton Keynes, UK \ntrevor.collins@open.ac.uk \n  \nThis paper presents a participatory methodology to design cards on social issues with the purpose \nto democratise knowledge among co-designers on the learning content of educational games. \nSituated on the topic of everyday sexism, the methodology has been developed through an \niterative process involving two collaborative workshops, two iterations of card design and a \nfeedback survey. Extracting findings from the workshops and the feedback gathered on the co-\ndesigned cards, this paper presents insights that could be used to inform similar studies using \ncards to inspire and foster reflection on social issues.   \nEducational game design. Design cards. Everyday Sexism. Participatory Design. Democratisation of Knowledge.  \n \n1. INTRODUCTION \nApplying co-design to educational games has the \npotential to generate more effective educational \noutcomes. Yet, it represents a challenge as best \npractices to co-design such games, which require \ntechnical, conceptual and educational skills, still \nhave to be defined (Carvalho et al., 2015; DeSmet et \nal., 2016). Most of the participatory models to design \neducational games are founded on educational \ntheories and game design (see for example: Amory, \n2007; Arnab et al., 2015). Some more recent \nmodels, though, have also included domain experts \nto define the learning content of the educational \ngame (De Jans et al., 2017). In scenarios where the \nco-designers are not equally experienced with the \neducational game domain or with the design \nprocess, tools to facilitate the design and maximize \neducational outcomes may be required. \nWithin the Human-Computer Interaction (HCI) \nliterature, cards have been pointed out as a tool with \nthe potential to democratise knowledge and to \nsupport collaborative design of realistic and \ninnovative projects. Chow et al. (2016) have argued \nthat cards should \u201ccapture a wide range of situations \nand personalities\u201d (p.91) to inspire designers with \nconcepts that are familiar or new to them. For \nsupporting the design of cards, Kensing and \nGreenbaum (2012) proposed some participatory \ndesign principles for transcribing diversified \nexperiences and knowledge to the cards.  \nUsing participatory approaches to design cards has \nbeen described as having the potential to enable \nmore engaging and effective design experiences \nand results (Chow et al., 2016; Golembewski and \nSelby, 2010; Halskov and Dalsg\u00e4rd, 2006). The \nliterature, however, shows limited number of studies \npresenting methodologies to this end.  \nThis work-in-progress focuses on the development \nof a methodology to co-design cards on a social \nissue. In the context of this research, such cards \nrefer to the learning content of educational games \nthat will also be co-designed in the future. The cards \naim at democratising knowledge, inspiring and \ntriggering reflection among co-designers with \ndifferent levels of understanding on the social issue.  \nThe methodology has been developed through an \niterative process consisting of collaborative \nworkshops, sequential reviews and iterations of card \ndesigns, and a summative feedback survey. \nAlthough instantiated on the social issue of everyday \nsexism, it is expected that the proposed \nmethodology can be applied to similar studies \naiming at co-designing cards for creating \neducational games on different social topics that \naffect people\u2019s everyday lives.  \nIn the next section, setting the scene for the study, \nsome key points from the literature on everyday \nsexism are introduced, followed by a review of card-\nbased tools to support design. Section 3 describes \nthe methodology to co-design a set of domain cards \non a social topic, which leads to further discussions \non the results and future work in section 4.  \n2. BACKGROUND \nThis section starts by contextualising this research \nwith an overview of everyday sexism, then turns to \npresent a literature review on card-based tools. \n  \nCo-designing Cards on Social Issues for Creating Educational Games  \nChristina Myers \u25cf Lara Piccolo \u25cf Trevor Collins \n \n2 \n2.1 Everyday sexism as a social issue \nEveryday sexism rests on the concept that sexism is \neither faced or reproduced on a daily basis but is \nrarely noticed (Barreto et al., 2009; Becker and \nSwim, 2011). Identifying everyday sexism and its \ndifferent forms, empathising with other people\u2019s \nexperiences, and reflecting on them are seen as \npossible ways to contribute to gender equality \n(Becker and Swim, 2011; Zawadzki et al., 2013). \nMore particularly, using everyday experiences to \nreflect on them is an important stepping stone in the \nhistory of feminism (Freedman, 2014; Sarachild, \n1968). \n2.2 Literature review on card-based tools \nIn the HCI domain, cards have been applied in a \nrange of contexts, with considerable distinction in \nterms of purpose, content, and target audience\u2019s \nexpertise. Naming a few, Vines et al. (2012) showed \npositive results in using cards to design banking \ntechnologies for and with people with limited \nfamiliarity with the topic. Chow et al. (2016) applied \ncards to create innovative ways to teach \nmathematics; Lucero and Arrasvuori (2013) to \nsupport the design for playfulness; both \nGolembewski and Selby (2010) and Roubira (2008) \nused cards as an inspiration tool to define and \nexplore preliminary design ideas; and Halskov and \nDalsg\u00e4rd (2006) to explore how to apply technology \nto concepts from domain studies.  \nIn Deng et al. (2014), the cards were intended to \nmake knowledge on tangible learning games \naccessible by translating \u201clengthy, dense, and jargon \nladen (body of literature) to design practice\u201d (p.1). \nWetzel et al. (2016) presented sets of cards \nthroughout the design process to create mixed \nreality games. Distinct sets of cards covered \ndifferent stages of the design process, such as cards \nwith abstract images applied at the ideation phase. \nIn the study presented by Flanagan et al. (2011), the \ncards were aimed at engaging diverse audiences to \nfacilitate the creation of game ideas prioritising \nhuman values. One of their Challenge cards (i.e. the \nsocial issue to be solved) targeted sexism and was \nillustrated as, \u201cDescription: Stereotype of a \ndiscrimination based on sexual roles. Strategy: \nEducation, awareness, legislation\u201d (Flanagan, 2010, \np.1).  \nThe structure and content of the cards are subjects \nof discussion in the literature, including textual \nelements as descriptions, examples, reflective \nquestions or lived experiences and illustrations. \nDeng et al. (2014) evaluated different cards\u2019 \nelements with groups of both experienced and \ninexperienced designers on the topics presented on \nthe cards and found out that the inexperienced ones \ntended to ignore the cards and requested more \ntextual information to understand and be able to use \nthe cards.  \nThe impact of reflective questions in the cards has \nalso been debated. Chow et al. (2016) used \nquestions to encourage reflection on everyday \nexperiences related to mathematics, arguing that it \ncould increase engagement among users. Vines et \nal. (2012) included questions on their cards for the \nusers to answer and found that the responses \nreceived were superior in quality when compared \nwith typical responses to written questionnaires. \nReports of lived experiences were also found in \nsome of the cards. The evaluation by Lucero and \nArrasvuori (2013) revealed that using real-life \nexperiences on the cards could be more \nunderstandable and inspiring than a definition of \nconcept. The study of Vines et al. (2012) also used \nquotations based on real experiences and found \npositive outcomes in terms of creating engagement, \nstating that quotations invited the users of the cards \nto learn and empathise with other people\u2019s \nexperiences.  \nIllustrations are arguably a source of inspiration and \nan important element in the aesthetics of cards \n(Lucero and Arrasvuori, 2013; Wetzel et al., 2016). A \nstudy compared PLEX Cards, which contain a \ndefinition and an informative illustration; and DIXIT \nCards, which represent abstract illustrations, and \nfound similar results in terms of numbers of ideas \ngenerated (Kwiatkowska et al., 2014). However, \nboth Chow et al. (2016) and Deng et al. (2014) \nfound different results in favour of illustrations \nrepresenting straight-forward and informative \nconcepts.  \nAmongst the literature reviewed, Golembewski and \nSelby (2010) was the only study that presented a \nmethodology for co-designing cards. The cards \nconsisted of a single word and a descriptive icon \nclassified accordingly to the domain of the project \nundertaken. This study considered the formulation of \ninitial ideas and avenues for exploration a \nfundamental part of the design process and invited \nco-designers to define the conceptual space of their \nown project by co-designing cards.  \n3. METHODOLOGY  \nThe methodology proposed differs considerably in \nits form, users and outcomes from the methodology \npresented by Golembewski and Selby (2010), in \nwhich the cards were co-designed and used by the \nsame group of designers. Instead, the cards co-\ndesigned in this study intend to be interpretable and \nusable by other groups of designers. In line with \nVines et al. (2012) and Deng et al. (2014), the set of \nco-designed cards aims to democratise knowledge \nin an easy-to-understand and inspiring manner \namong people with different background.  \nCo-designing Cards on Social Issues for Creating Educational Games  \nChristina Myers \u25cf Lara Piccolo \u25cf Trevor Collins \n \n3 \nReferring to the structure, Golembewski and Selby \n(2010) co-created cards composed by singular \nwords and icons. As seen in the study of Deng et al. \n(2014), such a minimalist structure could limit the \npotential to promote reflection among the targeted \naudience of this study. Similarly, the cards created \nby Flanagan et al. (2011) risk to overgeneralise \ncomplex issues such as sexism, by illustrating it with \na single definition.  \n3.1 Overview  \nTo co-design the cards, we organised two \nworkshops and an online survey. The first workshop \ninvolved 23 participants, 8 females and 15 males, \nwhile the second one had 4 males and 6 females. \nThe workshops were held at a university lab \ninvolving researchers related or not with the topic, \nPhD students, and administration staff. As further \ndescribed, some aspects of the first workshop were \nreviewed and informed a second collaborative \nworkshop. The workshops\u2019 participants were \norganised in groups of 3 to 4 people, each group \naddressing different categories of everyday sexism. \nThe seven categories of everyday sexism were: \n\u2018Benevolent Sexism\u2019, \u2018Sexist Language\u2019, \u2018Gender-\nbased Harassment\u2019, \u2018Gender Stereotypes\u2019, \u2018Online \nGender Discrimination\u2019, \u2018Feminism\u2019 and \n\u2018Downplaying Gender Discrimination\u2019. \nThe workshops started with a 45 minutes exercise \nwhere the groups were asked to define keywords, \nillustrations and reflective questions on a category of \neveryday sexism. The participants were also asked \nto read, individually, 8 written lived experiences \nrelated to the same category of everyday sexism \nand select the 3 most representative ones in groups. \nThe lived experiences were extracted from a website \ncalled everydaysexism.com where people shared \ntheir experiences on everyday sexism (Melville et al., \n2017). Reflective questions aimed at triggering \nreflection by future card users who would not \nnecessarily be aware on the issues. This exercise \nwas supported by a template in an A3 paper that \nguided participants towards completing the tasks, \nand also invited them to add any sort of suggestion. \nAfter this exercise, each group had 10 minutes to \nprovide feedback on the information generated by \nanother group. The workshops were concluded with \neach group presenting their results to all participants \nin 5 minutes. \nThe outcome of the two collaborative workshops, \nincluding the filled A3 page, the feedback received \nand the transcription of the groups\u2019 presentation, \ninformed the design of 7 cards, one for each \ncategory of everyday sexism. The cards were \ncomposed by a title, keywords, 3 lived experiences, \nan abstract image and 4 to 8 reflective questions.  \nOnce the cards were designed, an online feedback \nquestionnaire was developed aiming at collecting \npeople's perception on the cards\u2019 potential to inspire \nand trigger reflection on the specific category of \neveryday sexism. A total of 58 people, including \nworkshop participants, provided feedback on a set of \n3 or 4 cards each, resulting in at least 26 responses \nfor each card. Based on this feedback, a second \niteration of cards was created. \n3.2 Results  \nThe results are presented in two subsections; one \non the workshop structure (3.2.1) and the findings of \nthe feedback process (3.2.2). \n3.2.1 Revisions of collaborative workshops \nThe template on the A3 page was refined between \nthe workshops. The main difference was situated in \nthe reflective question section. Several participants \nexpressed that the task \u2018elaborate questions that \nwould raise awareness on the category of everyday \nsexism to people not necessary aware of it\u2019, was \ndifficult and that they were not satisfied with their \nresults. Following the process of reflection presented \nby  Wood Daudelin (1996), in the new version the \nfirst word of 4 questions were added as a visual \nprompt. The instruction was then, for example, to \ncreate a \u2018what question\u2019 targeting the problem \nidentification. The second workshop showed \nimproved results as the questions were more \nconnected to the lived experiences and none of the \n10 participants reported any difficulty to create \nreflective questions as a feedback. Figure 1 \nillustrates the final version of the template on the A3 \npage used in the second workshop.  \n \nFigure 1: Final version of A3 Page template \n \nThe keyword, lived experiences and illustration \nsections were evaluated as being constructive \nexercises as the participants were engaged in \ncritical discussions and the group responses were \nconsidered clear, informative and diversified. \nSeveral participants reported enjoying reading the \n\u201ceye-opening\u201d lived experiences. Figure 2 is an \nexample of an generated illustration which \ndescribes a group discussions on the use of the \nword \u2018just\u2019 to identify scenarios where people could \nbe downplaying gender discrimination. \n \nFigure 2: Example of the results in the illustration section \nCo-designing Cards on Social Issues for Creating Educational Games  \nChristina Myers \u25cf Lara Piccolo \u25cf Trevor Collins \n \n4 \nTo inform the cards design with the workshops \noutcomes, some lived experiences were slightly \nedited in order to make them shorter, and keywords \nwere used with a hashtag to reinforce their symbolic \nmeaning. In addition, based on the literature \npresented (Kwiatkowska et al., 2014; Wetzel et al., \n2016), abstract illustrations were added to the cards \nwhile the content generated on the illustration \nsection was used to complement the keywords and \nthe reflective questions.  \n3.2.2 Feedback survey \nA total of 58 people, 33 females and 25 males, \nresponded to the survey of whom, 5 people (8.6%) \nconsidered themselves experts on the topic of \neveryday sexism while 33 people (56.9%) aware, 17 \npeople (29,3%) learners, 2 people (3,4%) unaware \nand one person (1,7%) indifferent. All the workshop \nparticipants were invited to respond to this survey. \nFirstly, the responses as a 6-point Likert scale to the \nquestion \u2018I find this card very clear and \nunderstandable\u2019 showed an average rating of 4.92 \nout of 6, where the number 1 corresponded to \n\u2018Strongly disagree\u2019 and 6 to, \u2018Strongly agree\u2019. \nSecondly, the responses on \u2018I find this card very \ninspiring and lead me to reflection on [category of \neveryday sexism]\u2019 was 5.57 out of 6, on the same \nscale. The standard deviation for the first question \nwas 1.13 and 1.14 for the second question, \nsuggesting that the responses were not significantly \ndispersed over a wide range of values in any of \nthese questions.  \nThe next question was \u2018I find [card element] very \nuseful to trigger my reflection on [category of \neveryday sexism]\u2019 and the potential responses \nranked from \u2018Strongly agree\u2019 to \u2018Strongly disagree\u2019. \nThe average results on the cards were that 40% of \nthe respondent agreed or strongly agreed on the \nkeywords being very helpful; 91% on the lived \nexperiences; 38% on the illustrations; 82% on the \nreflective questions and 77% on all elements \ntogether.  \nAn open question requested feedback to improve \neach of the card separately, resulting in 147 \nsuggestions. Most of the feedback aimed at making \nthe images less abstract (21 entries), to simplify the \nreflective questions (20 entries), give more context \nto the keywords (14 entries), edit and shorten the \nlived experiences (11 entries) and limit the number \nof question to 4 per card (8 entries). \nBased on the evaluation results, a new iteration of \ncard was created, as depicted in Figure 3. The \nnumber of reflective questions was limited to 4, the \nquestions were made more concise, the lived \nexperiences re-phrased towards shortening them to \na maximum of 50 words and the keywords were \nmade more explicit (e.g. #ItsJust instead of #Just). \nThe abstract illustrations were also replaced by \nmore informative illustrations reflecting the \ninformation gathered from the workshops and \ncomplemented by the transcription of the \npresentation of each group, as illustrated in Figure 3. \nThe back of the card presented in Figure 3 contains \nfour sequential questions; the first one being \u201cWhat \nis the problem with downplaying sexism using words \nsuch as \u2018just\u2019 or \u2018only\u2019 (e.g. it\u2019s just a compliment)?\u201d.  \n \nFigure 3: Second iteration of the front of the card category \ncalled \u2018Downplaying gender discrimination\u2019  \n4. CONCLUSIONS AND FUTURE STUDY \nThis paper introduced an on-going work towards \nbuilding a participative methodology for designing \ncards on social issues, which in the context of this \nresearch, will be used to define the learning content \nrelated to everyday sexism when co-designing \neducational games. Some preliminary results based \non collected feedback are also presented, \nsupporting the proposed methodology. The \nmethodology provides guidance on how to structure \ncollaborative workshops to define various cards \nelements, namely keywords, lived experiences, an \nillustration, and reflective questions. The design of \nthese cards elements is supported by a template. \nWe recommend to other studies aiming at similar \nobjectives to use and adapt the template (Figure 1) \nto other social issue. Based on the feedback \ngathered, it is also recommended to limit the number \nof reflective questions to 4 while making them as \nconcise as possible, to avoid using abstract \nillustrations, restraint the lived experiences to \napproximately 50 words and contextualise the \nkeywords as much as possible. Future studies will \naddress the application of the cards the co-design \neducational games on everyday sexism. \n5. REFERENCES  \nAmory, A., 2007. Game Object Model Version II: A \nTheoretical Framework for Educational \nGame Development. Educ. Technol. Res. \nDev. 55, 51\u201377. \nCo-designing Cards on Social Issues for Creating Educational Games  \nChristina Myers \u25cf Lara Piccolo \u25cf Trevor Collins \n \n5 \nArnab, S., Lim, T., Carvalho, M.B., Bellotti, F., \nFreitas, S., Louchart, S., Suttie, N., Berta, \nR., De Gloria, A., 2015. Mapping learning \nand game mechanics for serious games \nanalysis. Br. J. Educ. Technol. 46, 391\u2013\n411. \nBarreto, M., Ryan, M., Schmitt, M., 2009. The glass \nceiling in the 21st century: Understanding \nbarriers to gender equality. \nBecker, J.C., Swim, J.K., 2011. Attention to Daily \nEncounters With Sexism as Way to \nReduce Sexist Beliefs. Psychol. Women Q. \n35, 227\u2013242. \nCarvalho, M.B., Bellotti, F., Berta, R., De Gloria, A., \nSedano, C.I., Hauge, J.B., Hu, J., \nRauterberg, M., 2015. An activity theory-\nbased model for serious games analysis \nand conceptual design. Comput. Educ. 87, \n166\u2013181. \nChow, J., Kusoffsky, M., Kurti, A., 2016. From \nAbstract to Concrete: Telling Math Stories \nwith Cards, in: Proceedings of the 14th \nParticipatory Design Conference: Short \nPapers, Interactive Exhibitions, Workshops \n- Volume 2, PDC \u201916. ACM, New York, NY, \nUSA, pp. 90\u201391. \nDe Jans, S., Van Geit, K., Cauberghe, V., Hudders, \nL., De Veirman, M., 2017. Using games to \nraise awareness: How to co-design serious \nmini-games?. Comput. Educ. 110, 77\u201387. \nDeng, Y., Antle, A.N., Neustaedter, C., 2014. \nTango cards. URL https://dl-acm-\norg.libezproxy.open.ac.uk/citation.cfm?doid\n=2598510.2598601. \nDeSmet, A., Thompson, D., Baranowski, T., \nPalmeira, A., Verloigne, M., Bourdeaudhuij, \nI.D., 2016. Is Participatory Design \nAssociated with the Effectiveness of \nSerious Digital Games for Healthy Lifestyle \nPromotion? A Meta-Analysis. J. Med. \nInternet Res. 18, e94. \nFlanagan, M., 2010. Tiltfactor | Grow-A-Game. URL \nhttp://www.tiltfactor.org/game/grow-a-\ngame/. \nFlanagan, M., Belman, J., Helen Nissenbaum, \nDiamond, J., 2011. Grow-A-Game: A Tool \nfor Values Conscious Design and Analysis \nof Digital Games \nFreedman, J., 2014. Reclaiming the Feminist \nVision: Consciousness-Raising and Small \nGroup Practice. McFarland & Co, \nJefferson, North Carolina. \nGolembewski, M., Selby, M., 2010. Ideation Decks: \nA Card-Based Design Ideation Tool.  \nHalskov, K., Dalsg\u00e4rd, P., 2006. Inspiration Card \nWorkshops, in: Proceedings of the 6th \nConference on Designing Interactive \nSystems, DIS \u201906. ACM, New York, NY, \nUSA, pp. 2\u201311. \nKensing, F., Greenbaum, J., 2012. Heritage: \nHaving a say. Routledge Int. Handb. \nParticip. Des. P21-36. \nKwiatkowska, J., Sz\u00f3stek, A., Lamas, D., 2014. \n(Un)Structured Sources of Inspiration: \nComparing the Effects of Game-like Cards \nand Design Cards on Creativity in Co-\ndesign Process, in: Proceedings of the \n13th Participatory Design Conference: \nResearch Papers - Volume 1, PDC \u201914. \nACM, New York, NY, USA, pp. 31\u201339. \nLucero, A., Arrasvuori, J., 2013. The PLEX Cards \nand its techniques as sources of inspiration \nwhen designing for playfulness (PDF \nDownload Available). URL \nhttps://www.researchgate.net/publication/2\n64824212_The_PLEX_Cards_and_its_tec\nhniques_as_sources_of_inspiration_when_\ndesigning_for_playfulness. \nMelville, S., Eccles, K., Yasseri, T., 2017. Semantic \nMap of Sexism: Topic Modelling of \nEveryday Sexism Project Entries.  \nRoubira, J.-L., 2008. Dixit. BoardGameGeek. URL \nhttps://boardgamegeek.com/boardgame/39\n856/dixit. \nSarachild, 1968. Consciousness-Raising: A Radical \nWeapon\u2019 \nVines, J., Blythe, M., Lindsay, S., Dunphy, P., \nMonk, A., Olivier, P., 2012. Questionable \nConcepts: Critique As Resource for \nDesigning with Eighty Somethings, in: \nProceedings of the SIGCHI Conference on \nHuman Factors in Computing Systems, \nCHI \u201912. ACM, New York, NY, USA, pp. \n1169\u20131178. \nWetzel, R., Rodden, T., Benford, S., 2016. \nDeveloping Ideation Cards for Mixed \nReality Game Design. \nWood Daudelin, M., 1996. Learning from \nexperience through reflection. Organ. Dyn. \n24, 36\u201348. \nZawadzki, M.J., Shields, S.A., Danube, C.L., Swim, \nJ.K., 2013. Reducing the Endorsement of \nSexism Using Experiential Learning: The \nWorkshop Activity for Gender Equity \nSimulation (WAGES). Psychol. Women Q. \n38, 75\u201392. \n \n \n \n", "identifiers": ["oai:oro.open.ac.uk:62141", "10.14236/ewic/HCI2018.164"], "relations": [], "repositories": [{"id": "86", "openDoarId": 0, "name": "Open Research Online", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "baseId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 1, "metadataAdded": 1562643950000, "metadataUpdated": 1591929847000, "depositedDate": 1561563672000}, "subjects": [], "title": "Co-designing Cards on Social Issues for Creating Educational Games", "topics": [], "types": [], "year": 2018, "fulltextIdentifier": "https://core.ac.uk/download/pdf/211273841.pdf", "doi": "10.14236/ewic/HCI2018.164", "oai": "oai:oro.open.ac.uk:62141", "downloadUrl": "https://core.ac.uk/download/pdf/211273841.pdf"}}
{"status": "OK", "data": {"id": "159766439", "authors": ["Pride, David", "Knoth, Petr"], "citations": [], "contributors": [], "datePublished": "2018-09-30", "description": "Most Performance-based Research Funding Systems (PRFS) draw on peer review and bibliometric indicators, two different method- ologies which are sometimes combined. A common argument against the use of indicators in such research evaluation exercises is their low corre- lation at the article level with peer review judgments. In this study, we analyse 191,000 papers from 154 higher education institutes which were peer reviewed in a national research evaluation exercise. We combine these data with 6.95 million citations to the original papers. We show that when citation-based indicators are applied at the institutional or departmental level, rather than at the level of individual papers, surpris- ingly large correlations with peer review judgments can be observed, up to r <= 0.802, n = 37, p < 0.001 for some disciplines. In our evaluation of ranking prediction performance based on citation data, we show we can reduce the mean rank prediction error by 25% compared to previous work. This suggests that citation-based indicators are sufficiently aligned with peer review results at the institutional level to be used to lessen the overall burden of peer review on national evaluation exercises leading to considerable cost savings", "fullText": "Open Research Online\nThe Open University\u2019s repository of research publications\nand other research outputs\nPeer review and citation data in predicting university\nrankings, a large-scale analysis\nConference or Workshop Item\nHow to cite:\nPride, David and Knoth, Petr (2018). Peer review and citation data in predicting university rankings, a large-\nscale analysis. In: Theory and Practice of Digital Libraries (TPDL) 2018, 10-13 Sep 2018, University of Porto,\nPortugal.\nFor guidance on citations see FAQs.\nc\u00a9 2018 Springer Nature\nVersion: Accepted Manuscript\nLink(s) to article on publisher\u2019s website:\nhttp://dx.doi.org/doi:10.1007/978-3-030-00066-017\nCopyright and Moral Rights for the articles on this site are retained by the individual authors and/or other copyright\nowners. For more information on Open Research Online\u2019s data policy on reuse of materials please consult the policies\npage.\noro.open.ac.uk\nPeer review and citation data in predicting\nuniversity rankings, a large-scale analysis\nDavid Pride and Petr Knoth\nThe Knowledge Media Institute, The Open University, Milton Keynes, UK.\n{david.pride, petr.knoth}@open.ac.uk\nAbstract. Most Performance-based Research Funding Systems (PRFS)\ndraw on peer review and bibliometric indicators, two di\ufb00erent method-\nologies which are sometimes combined. A common argument against the\nuse of indicators in such research evaluation exercises is their low corre-\nlation at the article level with peer review judgments. In this study, we\nanalyse 191,000 papers from 154 higher education institutes which were\npeer reviewed in a national research evaluation exercise. We combine\nthese data with 6.95 million citations to the original papers. We show\nthat when citation-based indicators are applied at the institutional or\ndepartmental level, rather than at the level of individual papers, surpris-\ningly large correlations with peer review judgments can be observed, up\nto r <= 0.802, n = 37, p < 0.001 for some disciplines. In our evaluation\nof ranking prediction performance based on citation data, we show we\ncan reduce the mean rank prediction error by 25% compared to previous\nwork. This suggests that citation-based indicators are su\ufb03ciently aligned\nwith peer review results at the institutional level to be used to lessen the\noverall burden of peer review on national evaluation exercises leading to\nconsiderable cost savings.\n1 Introduction\nSince the late 20th century there has been a seismic shift in many countries in\nhow research is funded. In addition to traditional grant or patronage funding,\nthere is growing use of Performance-based Research Funding Systems (PRFS) in\nmany countries. These systems fall largely into two categories; those that focus\non peer review judgments for evaluation and those that use a bibliometric ap-\nproach. The UK and New Zealand both have systems heavily weighted towards\npeer review. Northern European countries other than the UK tend to favour\nbibliometric methodologies whereas Italy and Spain consider both peer review\njudgments and bibliometrics. Research Evaluation Systems overall have dual\nand potentially dichotomous ends, firstly identifying the best quality research\nbut also, in many cases, the distribution of research funds. There is, however, a\nlarge variance in the level of institutional funding granted based on the results\nof these exercises. The UK\u2019s Research Councils distribute \u00a31.6 billion annually\nentirely on the basis of the results of the Research Excellence Framework (REF)\nwhich is the largest single component of university funding. At the other end\nof the scale, the distribution of funds based on the results of the Finnish PRFS\nis just 3% of the total research budget. Furthermore, the PRFS in Norway and\nAustralia are both used for research evaluation but are not used for funding\ndistribution [1]. Peer-review based PRFS are hugely time-consuming and costly\nto conduct. In this investigation we ask how well do the results of peer-review\nbased PRFS correlate with bibliometric indicators at the institutional or disci-\nplinary level. A strong correlation would indicate that metrics, where available,\ncan lessen the burden of peer review on national PRFS leading to considerable\ncost savings, while a weak correlation would suggest each methodology provides\ndi\ufb00erent insights.\nTo our knowledge, this is the first large-scale study exploring the relation-\nship between peer-review judgments and citation data at the institutional level.\nOur study is based on a new dataset compiled from 190,628 academic papers in\n36 disciplines submitted to UK REF 2014, article level bibliometric indicators\n(6.95m citations) and institutional / discipline level peer-review judgments. This\nstudy demonstrates that there is a surprisingly strong correlation between an in-\nstitutions\u2019 Grade Point Average (GPA) ranking for its outputs submitted to the\nUK Research Excellence Framework for many Units of Assessment (UoAs) and\ncitation data. We also shows that this makes it possible to predict institutional\nrankings with a degree of accuracy in highly cited disciplines.\n2 Related work\nThere has long been wide ranging and often contentious discussion regarding\nthe e\ufb03cacy of both peer review and bibliometrics and whether one or other, or\nboth should be used for Research Evaluation. Several other studies have specif-\nically investigated the correlation between the results of di\ufb00erent nations\u2019 peer\nreview focused Performance-based Research Funding Systems and bibliometric\nindicators. Anderson [2] finds only weak to moderate correlation with results\nfrom the New Zealand PRFS and a range of traditional journal rankings. The\nhighest correlation is r = 0.48 with the Thomson Reuters Journal Citation Re-\nport. However Anderson states that this may be due to the much broader scope\nof research considered by PRFS processes and the additional quality-related in-\nformation available to panels. Contrary to Anderson, Smith [3] used citations\nfrom Google Scholar (GS) and correlated these against the results from the New\nZealand PRFS in 2008. He found strong correlation, r = 0.85 for overall PRFS\nresults against Google Scholar citation count.\nA comprehensive global PRFS analysis was conducted by Hicks in 2012.\nHicks states there is convincing evidence that when PRFS are used to define\nleague tables this creates powerful incentives for institutions to attempt to \u2019game\u2019\nthe process, whether in regards to submission selection or sta\ufb00 retention and\nrecruitment policies [1]. A UK government funded report, The Metric Tide,\nwas published in 2015 and gave a range of recommendations for the use of\nmetrics in research evaluation exercises. The Metric Tide study had access to the\nanonymised scores for the individual submissions to the REF and was therefore\ndirectly able to compare on a paper by paper basis the accuracy of a range of\nbibliometric indicators. This study tested correlations with a range of di\ufb00erent\nbibliometric measures and found correlation with rankings for REF 4* and 3*\noutputs for some UoAs. Metrics found to have moderately strong correlations\nwith REF scores for a wide range of UoAs included: number of tweets; number of\nGoogle Scholar citations; source normalised impact per paper; SCImago journal\nrank and citation count [4].\nHowever, The Metric Tide study used di\ufb00erent citation metrics and citation\ndata sources from our approach. It is at the institutional UoA level that our\nstudy reveals some of the strongest correlations, higher than previously shown.\nIn a related study, Mryglod et al. [5] used departmental h-index aggregation to\npredict REF rankings. Their work was completed before December 2014 when\nthe REF results were published and contained ranking predictions based on their\nmodel with some degree of success. They also experimented by normalising the\nh-index for each year between 2008 and 2014 but surprisingly found little evi-\ndence that timescale played a part in the strength of the correlations they found.\nAn ad hoc study by Bishop [6] also found a moderate to strong correlation be-\ntween departmental research funding based on the results of the UK\u2019s Research\nAssessment and Evaluation (RAE) exercise conducted in 2008, and departmen-\ntal h-index. Mingers [7] recently completed an investigation that collected total\ncitation counts from Google Scholar (GS) for the top 50 academics1 from each\nUK institute and he found strong correlations with overall REF rankings. To\nour knowledge, ours is the first large-scale in-depth study that investigates the\ncorrelation between citation data and peer review rankings by discipline at the\ninstitutional level, taking into account all papers submitted to REF.\n3 Results\nFor this study we used data from the UK\u2019s Research Excellence Framework\n(REF). The last REF exercise undertaken in the UK in 2014 was the largest over-\nall assessment of universities\u2019 research output ever undertaken globally. These\nexperiments focus on the academic outputs (research papers) component of the\nREF, for which the metadata are available for download from the REF web-\nsite. The REF 2014 exercise peer reviewed and graded approximately 191,000\noutputs from 154 institutions and in 36 Units of Assessment (UoAs) from zero\nto four stars. The grading for each submission was determined according to\noriginality, significance and rigour. The peer review grades for the individual\nsubmissions were aggregated for each UoA to produce a Grade Point Average\nfor each institute. The rankings are of critical importance to the institutes as\napproximately \u00a31.6 billion in QR funding from central government is distributed\nannually entirely on the basis of the REF results [8].\nEach of the REF peer review panels individually chose whether or not to use\ncitation data to inform their decisions. Eleven out of 36 selected to do so and were\n1 If there were not 50 academics then the total number of academics on GS for that\ninstitute was used.\nFig. 1. Citation enrichment workflow used in dataset creation.\nprovided with citation data from Elsevier Scopus to assist their decision making.\nFor each area and age of publication they were given the number of citations\nrequired to put the paper in the top 1%, 5%, 10% or 25% of papers within\nits area. This gave REF reviewers a subject-level benchmark against which to\nconsider the citation data.[9]\nWhereas the aggregate GPA ranking for all UoAs and all institutes is publicly\navailable, it is now not possible to obtain a direct comparison between citation\ndata and the individual rankings for each submission as HEFCE state that these\ndata were destroyed. The rationale behind this was to preempt any requests for\nthis data under the Freedom of Information Act. [10].\n3.1 Dataset\nThe dataset creation procedure is depicted in Figure 1. We first downloaded the\nREF 2014 submission list [9]. For each output, the list contains; publication title,\npublication year, publication venue, name of institute and UoA. These fields were\nfully populated for 190,628 out of 190,963 submissions to the \u2019outputs\u2019 category\nof the REF process.\nWe decided to utilise the Microsoft Academic Graph (MAG) to enrich the\nREF submission list with citation information. At the time of the experiment\nMAG contained approximately 168m individual papers and 1.15 billion citation\npairs. This decision was motivated by the fact that while Scopus, operated by\nElsevier, was used to provide citation data to the REF process, the free version\nof the Scopus API service is limited to 20,000 requests per week. It would have\ntherefore taken almost two months to gather the required data which was not\npractical as this was more than 10 times slower than using MAG. Additionally,\nstudies by [11] and [12] have recently confirmed how comprehensive the MAG\nUoA / Subject Outputs % in MAG Citations MCPP\nPublic Health 4,881 94.61% 505,950 109.56\nClinical Medicine 13,394 90.78% 1,278,810 105.17\nPhysics 6,446 84.51% 491,151 90.15\nBiological Sciences 8,608 92.20% 620,009 78.12\nEarth Systems / Environment 5,249 91.64% 315,429 65.58\nChemistry 4,698 87.71% 246,361 59.78\nAllied Health Professions 10,358 89.35% 402,033 43.43\nAg. Vet. and Food Science 3,919 90.76% 150,959 42.44\nComp. Science and Informatics 7,645 89.22% 284,815 41.76\nEconomics and Econometrics 2,600 88.81% 95,591 41.4\nTable 1. UoAs with the highest mean citations per paper (MCPP).\nNumber of Units of Assessment (UoAs) 36\nNumber of institutes 154\nNumber of UoAs/institution pairs 1,911\nNumber of submissions (papers) 190,628\nNumber of submissions (papers) in MAG 145,415\nNumber of citations 6,959,629\nTable 2. Dataset statistics\ncitation data are. We could not utilise Google Scholar as it does not o\ufb00er an API\nand prohibits \u2019scraping\u2019 of data.\nWe systematically queried the MAG Evaluate API for each submission using\na normalised version of the publication\u2019s title (lower case, diacritics removed).\nThis returned a set of MAG IDs which which were potential matches of the ar-\nticle. We subsequently queried the MAG Graph Search API to validate each of\nthe potential matches. We accepted as a match the most similar publication title\nthat had at least 0.9 cosine similarity. This threshold was set by manually ob-\nserving about one hundred matches. Using this process we successfully matched\n145,415 REF submissions with 6.95 million citations, corresponding to a recall\nof 76% of the total initial REF submission list.\nTable 1 is ordered by the mean citations per paper (MCPP) and shows total\nnumber of submissions, percentage of these submissions available in MAG and\nthe total citations of these submissions.\nAdditionally, as described in Figure 1, we downloaded the Assessment Data\nfrom the REF 2014 website. These data contain the GPA, calculated by aggregat-\ning the peer review assessment results of individual papers for each given institu-\ntion per UoA. We then joined these data with the enriched REF submission list\nby institution name and UoA. By doing so, we obtained 1,911 UoA/institution\npairs together with their peer assessment information (GPA) and corresponding\nlists of submissions and their citation data (Table 2).\nThe full dataset used in our experiments and all results can be downloaded\nfrom Figshare. 2\n2 https://figshare.com/s/69199811238dcb4ca987\nUoA mn2017 med2017 mn2014 med2014 cd\n1 Chemistry 0.663 0.802 0.637 0.738 Y\n2 Biological Sciences 0.188 0.797 0.288 0.785 Y\n3 Aero. Mech. Chem. Engineering 0.771 0.758 0.745 0.760 N\n4 Social Work and Policy 0.697 0.752 0.629 0.635 N\n5 Comp. Sci. and Informatics 0.715 0.743 0.720 0.678 Y\n6 Economics 0.750 0.737 0.760 0.770 Y\n7 Earth Systems and Enviro. Sciences 0.472 0.707 0.512 0.686 Y\n8 Clinical Medicine 0.654 0.677 0.666 0.662 Y\n9 Public Health and Primary Care 0.535 0.674 0.607 0.653 Y\n10 Physics 0.600 0.666 0.627 0.605 Y\n. . .\n27 Comm. Cultural and Media Studies 0.369 0.355 0.334 0.267 N\n28 Philosophy 0.352 0.353 0.268 0.270 N\n29 Law 0.318 0.159 0.365 0.136 N\n30 Theology and Religious Studies 0.404 0.154 0.439 0.153 N\n31 English Language and Literature -0.168 0.102 -0.192 0.094 N\n32 Art and Design 0.157 0.075 0.187 0.118 N\n33 Anthropology and Dev. Studies 0.062 -0.009 0.222 0.145 N\n34 Modern Languages and Linguistics 0.141 -0.069 0.182 0.188 N\n35 Classics 0.155 -0.07 0.079 0.285 N\n36 Music Drama Dance & Perf. Arts 0.046 -0.094 0.051 0.039 N\nTable 3. Correlation between REF GPA output rankings and citation data\n3.2 How well do peer review judgments correlate with citation data\nat the institutional level?\nOnce we assembled the full dataset, we extracted the following overall citation\nstatistics: mean citations in December 2017 (mn2017), median citations in De-\ncember 2017 (med2017), mean citations at the time of the REF exercise (mn2014),\nand median citations at the same point (med2014). These data were then used to\ntest the correlation between citation data and REF GPA rankings for outputs\nfor every institute in every UOA. The ten highest and ten lowest measured cor-\nrelations by UoA are shown in Table 3. The citation data (cd) column denotes\nwhether the REF judging panels considered citation data in their deliberations.\nWhile we attempted to run correlations with other similar aggregate functions,\nthese are not shown in this table as they have far lower correlations with GPA.\nStrong positive correlations can be observed at the discipline level for a large\nproportion of the UoAs, particularly for median citation count in 2017. Whilst\nthe correlation was most often stronger for those UoAs that had used citation\ndata in the REF peer review process, this was not always the case. Aeronautical\nand Mechanical engineering and Social work & Policy are two disciplines, which\ndid not use citation data yet, show very strong correlations with GPA results.\nAt the lower end of the scale, there was little correlation between GPA ranking\nand citation data, notably for those subjects covered by REF panels C and\nD.[4]. Lack of coverage in many of these areas is, however, understandable as\nthese are disciplines which do not always produce journal articles, conference\nFig. 2. Correlation between med2017 citations per UoA and GPA against the coverage\nof REF submissions in MAG for all UoAs. An \u2019o\u2019 represents a non-citation based UoA\nwhilst and \u2019x\u2019 denotes a UoA that used citations.\nproceedings and other digitally published and highly citable artifacts as their\nmain type of output. There is, however, clear delineation between the more\nhighly correlated UoAs and those less correlated. The UoAs with the lowest are\ndistinct from the rest, they are having a very weak or no correlation (r <=\n0.159, n = 37, p < 0.001). Those above this level have a medium to strong\ncorrelation (r > 0.353, n = 37, p < 0.001). The low correlation for mean citations\nfor Biological Sciences is explained by a single paper which was the most highly\ncited paper in the UoA. This paper received 4,626 citations, 58% more than next\ncited paper and nine times as many citations as all other submissions for that\ninstitute combined. Furthermore, this paper came from second lowest ranked (by\nGPA) of 44 institutes. Had this paper been discounted from the correlations, the\nprediction results would have been far more clearly aligned with the other UoAs\n(mn2017=0.782, mn2014=0.766).\nThe variance of citation data coverage across UoAs led us to explore whether\nthere could be a relationship between the strength of the correlations GPA and\ncitation data correlation with the coverage of citation in a given UoA. Figure\n2 plots this for both the UoAs that used citation data and those that did not.\nWhile the graph confirms that the highly cited UoAs in MAG are those UoAs\nthat used citation data, it indicates that a few UoAs that did not also exhibit\nstrong correlations. Unsurprisingly, the plot suggests that there might be a small\nbias exhibited by extra correlation strength in UoAs that utilised citation data.\nHowever, given the small number of UoAs, this is not statistically significant.\nREF UoA / Rank GPA med2017 mc2017 rdi\ufb00 med2014 mc2014 rdi\ufb00\nChemistry\nLiverpool 3.44 Liverpool 64 0 Liverpool 26 0\nCambridge 3.42 Cambridge 54 0 Lancaster 25 +8\nOxford 3.32 Warwick 53 +3 Oxford 22 0\nUEA 3.29 Bath 51 +12 Cambridge 22 -2\nBristol 3.26 Oxford 50 -2 Queen Mary 20 +2\nBio Sciences\nICR 3.44 ICR 77 0 ICR 31 0\nNewcastle 3.33 Queen Mary 66 +15 She\ufb03eld 26 +5\nDundee 3.3 Imperial 59 +1 Imperial 25 +1\nImperial 3.26 She\ufb03eld 56 +3 Leeds 24 +27\nOxford 3.26 Edinburgh 55 +4 Edinburgh 23 +4\nAero. Mech.\nCambridge 3.34 Cambridge 25 0 Cambridge 9 0\nImperial 3.12 Imperial 23 0 Imperial 8 0\nUCL 3.06 She\ufb03eld 19 +2 Brighton 7 +13\nCranfield 3.01 Brighton 18 +12 Manchester 6 +4\nShe\ufb03eld 3.01 Manchester 17 +3 She\ufb03eld 6 0\nTable 4. Rankings by GPA and predictions produced using med2017 and med2014\nrespectively for the three most highly correlated UoAs.\n3.3 How well can citation data predict peer review based\ninstitutional rankings?\nTables 4 shows the top five institutions for Chemistry, Biological Sciences and\nAeronautical and Mechanical Engineering as ranked in the REF by GPA and pre-\ndictions of ranking usingmed2017 andmed2014 respectively. mc2017 and mc2014\nshow the median citation count for that institute. Rdi\ufb00 shows the rank di\ufb00er-\nence when ranked by a particular citation metric. The prediction performance\nindicated in these tables is not unique, in four of the five top UoAs by correlation\nstrength the highest ranked institute is predicted correctly by both med2014 and\nmed2017.\nTable 5 demonstrates the e\ufb00ectiveness of predicting based on med2014 for the\n10 most highly cited UoAs. To compare the prediction error, expressed by rdi\ufb00,\nacross UoAs, we calculated the mean rank di\ufb00erence normalised by number\nof institutions (nrdi\ufb00 ). To express overall prediction accuracy, we used Mean\nAverage Precision (MAP). The HEI column denotes the number of institutes\nsubmitting to that UoA. The parameter rt denotes the prediction rank toler-\nance. For example, rt = 3 indicates that a prediction within 3 positions of the\noriginal assessment result will be considered as correct. Given the simplicity of\nthe prediction method, this is a strong indication of the power of citation data in\nthis task. One could reasonably expect that further improvements can be made\nby employing more sophisticated indicators. However, as the predictions are not\nas good for UoAs that have lower than average mean citations per paper , we\nwould restrain from recommending the use of citation data unaccompanied by\npeer review assessments in those UoAs.\nUoA HEIs rdi\ufb00 nrdi\ufb00 MAP MAP MAP MAP MAP MAP\nrt=3 rt=5 rt=10 rt=10% rt=20% rt=30%\nComp Sci. 89 12.39 0.139 0.19 0.32 0.50 0.46 0.75 0.87\nAg. Vet. 29 4.02 0.139 0.45 0.65 0.86 0.45 0.68 0.86\nClinical Med. 31 4.38 0.141 0.51 0.70 0.93 0.51 0.77 0.93\nAllied H. 83 12.03 0.145 0.20 0.30 0.55 0.43 0.72 0.86\nEconomics 28 4.07 0.145 0.57 0.71 0.92 0.57 0.78 0.92\nChemistry 37 5.51 0.149 0.54 0.56 0.83 0.54 0.78 0.86\nEarth Systems 45 7.24 0.161 0.40 0.51 0.77 0.51 0.68 0.84\nPublic Health 32 5.18 0.162 0.50 0.62 0.84 0.50 0.68 0.84\nBio. Science 44 7.59 0.173 0.34 0.52 0.72 0.52 0.66 0.79\nPhysics 41 7.36 0.180 0.36 0.53 0.78 0.43 0.73 0.80\nAll (mean) 45 6.98 0.153 0.41 0.54 0.77 0.49 0.72 0.86\nTable 5. Rank prediction quality for top 10 UoAs with the highest mean citations per\npaper.\nUoA HEIs rdi\ufb00 nrdi\ufb00 MAP MAP MAP MAP MAP MAP\nrt=3 rt=5 rt=10 rt=10% rt=20% rt=30%\nMryglod [5]\nChemistry 29 4.89 0.169 0.37 0.82 0.82 0.37 0.82 0.82\nPhysics 32 8.63 0.270 0.28 0.40 0.65 0.28 0.46 0.65\nBio Science 31 8.38 0.270 0.22 0.38 0.70 0.22 0.51 0.64\nAll (mean) 31 7.30 0.24 0.29 0.53 0.72 0.29 0.60 0.70\nPride & Knoth (this study)\nChemistry 29 4.00 0.138 0.68 0.72 0.89 0.68 0.72 0.86\nPhysics 32 5.68 0.178 0.34 0.59 0.90 0.34 0.75 0.90\nBio Science 31 7.16 0.231 0.35 0.45 0.74 0.35 0.51 0.71\nAll (mean) 31 5.61 0.18 0.46 0.59 0.84 0.46 0.66 0.82\nImprovement 23% 25% 59% 11% 17% 59% 10% 17%\nTable 6. Comparison of the prediction performance of our study with Mryglod et al.[5]\nWe wanted to compare our prediction performance to the study of Mryglod\net al. [5]. In order to conduct a fair and exact comparison, it was necessary to\nparse a number of institutions from our input data. Mryglod et al. reported they\nwere unable to obtain citation indicators for all institutions in a given UoA.\nTheir study covered three of the top ten highly cited UoAs, we show in Table 6\nthat our predictions are significantly better across all categories.\n4 Discussion\nIt has been shown in [4], [13] and that many bibliometric indicators show lit-\ntle correlation with peer review judgments at the article level. This study, and\nthose by [7], [2] and [3], demonstrate that some bibliometric measures can o\ufb00er\na surprisingly high degree of accuracy when used at the institutional or depart-\nmental level. Our work has been conducted on a significantly larger dataset and\nour prediction accuracy is higher than shown in previous studies, despite delib-\nerately using fairly simplistic indicators. Several studies including The Metric\nTide [4], The Stern Report [14] and the HEFCE pilot study [15] all state that\nmetrics should be used as an additional component in research evaluation, with\npeer review remaining as the central pillar. Yet, peer review has been shown\nby [16], [17] and [18] amongst others to exhibit many forms of bias including\ninstitutional bias, gender / age related bias and bias against interdisciplinary\nresearch. In an examination of one of the most critical forms of bias, that of\npublication bias, Emerson [19] noted that reviewers were much more likely to\nrecommend papers demonstrating positive results over those that demonstrated\nnull or negative results.\nAll of the above biases exist even when peer review is carried out to the\nhighest international standards. There were close to 1,000 peer review experts\nrecruited by the REF, however the sheer volume of outputs requiring review\ncalls into question the exactitude of the whole process. As an example the REF\npanel for UoA 9, Physics, consisted of 20 members. The total number of out-\nputs submitted for this UoA was 6,446. Each paper is required to be read by\ntwo referees. This increases the overall total requirement to read 12,892 paper\ninstances. Therefore each panel member was required to review, to international\nstandards, an average of 644 papers in a little over ten months. If every panel\nmember, worked every day for ten months, each member would need to read and\nreview 2.14 papers per day to complete the work on time. This is, of course, in\naddition to the panelist\u2019s usual full-time work load. Moreover, Physics is not an\nunusual example and many other UoAs tell a similar story in terms of the aver-\nage number of papers each panel member was expected to review; Business and\nManagement Studies (1,017 papers), General Engineering (868 papers), Clinical\nMedicine (765 papers). The burden placed on the expert reviewers during the\nREF process was onerous in the extreme. Coles [20] calculated a very similar\nfigure of 2 papers per day, based on an estimate before the data we now have was\navailable. \u2019It is blindingly obvious,\u2019 he concluded, \u2019that whatever the panels do\nwill not be a thorough peer review of each paper, equivalent to refereeing it for\npublication in a journal\u2019. Sayer [21] is equally disparaging in regards to the vol-\nume of papers each reviewer was required to read and also expresses significant\ndoubts about the level of expertise within the review panels themselves.\nIn addition to the potential pitfalls in the current methodologies, there is\nalso the enormous cost to be considered. This was estimated to be \u00a366m for\nthe UK\u2019s original PRFS, the Research Assessment Exercise (RAE) in 2008. This\nrose markedly to \u00a3246m for the 2014 Research Excellence Framework. This is\ncomprised of \u00a3232M in costs to the higher education institutes and around \u00a314M\nin costs for the four UK higher education funding bodies. The cost to the in-\nstitutions was approximately \u00a3212M for preparing the REF submissions for the\nthree areas; outputs, impact and environment, with the cost for preparing the\noutputs being the majority share of this amount. Additionally, there were costs\nof around \u00a319M for panelists\u2019 time. [22]. If bibliometric indicators can in any\nway lessen the financial burden of these exercises on the institutions this is a\nstrong argument in favour of their usage.\n5 Conclusion\nThis work constitutes the largest quantitative analysis of the relationship be-\ntween peer reviews (190,628 paper submissions) and citation data (6.9m citation\npairs) at an institutional level. Firstly, our results show that citation data exhibit\nstrong correlations with peer review judgments when considered at the institu-\ntional level and within a given discipline. These correlations tend to be higher in\ndisciplines with high mean citations per paper. Secondly, we demonstrate that\nwe can utilise citation data to predict top ranked institutions with a surprisingly\nhigh precision. In the ten UoAs with the highest number of mean citations per\npaper we achieve 0.77 MAP with prediction rank tolerance 10 with respect to\nthe REF 2014 results. In four out of five top UoAs by correlation strength, the\nhighest ranked institute in the REF results was predicted correctly. It is impor-\ntant to note that these predictions are based on citation data that were available\nat the time of the REF exercise.\nWhile our analysis does not answer whether using citation-based indicators\nwe can predict institutional rankings better than by relying on a peer review\nsystem, our results evidence that the REF peer review process led to highly\nsimilar results as those that could have been predicted automatically using ci-\ntation data. The 11 REF UoAs with the highest mean citations per paper in\nMAG are the identical UoAs in which the peer review panels used citation data\nto inform their decisions. We argue that if peer-review is conducted in the way\nit was conducted in the REF, then it would have been more cost e\ufb00ective to\nsave a significant proportion of the \u00a3246m spent on organising the peer review\nprocess [22] and carry out the institutional evaluation purely using citation data,\nparticularly in UoAs with high mean citations per paper.\nThis has wide implication for PRFS globally. The countries whose PRFS still\nhave a peer review component should carefully consider the way in which the\npeer review process is conducted. Thus ensuring that the peer review results\nadd a new dimension to the information over that which can be obtained by\npredictions based on citation data alone. However, this advice only applies when\nthe goal of the PRFS is to rank institutions, as it is the case in the UK REF,\nrather than individual papers or researchers.\n6 Acknowledgements\nThis work has been funded by Jisc and has also received support from the\nscholarly communications use case of the EU OpenMinTeD project under the\nH2020-EINFRA-2014-2 call, Project ID: 654021\nReferences\n1. Hicks D. Performance-based university research funding systems. Research policy.\n2012;41(2):251\u2013261.\n2. Anderson DL, Smart W, Tressler J. Evaluating research\u2013peer review team as-\nsessment and journal based bibliographic measures: New Zealand PBRF research\noutput scores in 2006. New Zealand Economic Papers. 2013;47(2):140\u2013157.\n3. Smith AG. Benchmarking Google Scholar with the New Zealand PBRF research\nassessment exercise. Scientometrics. 2008;74(2):309\u2013316.\n4. HEFCE. The Metric Tide: Report of the Independent Review of the Role of Metrics\nin Research Assessment and Management; 2015. Available from: http://www.\nhefce.ac.uk/pubs/rereports/year/2015/metrictide/.\n5. Mryglod O, Kenna R, Holovatch Y, Berche B. Predicting results of the\nResearch Excellence Framework using departmental h-index. Scientomet-\nrics. 2015 Mar;102(3):2165\u20132180. Available from: https://doi.org/10.1007/\ns11192-014-1512-3.\n6. Bishop D. An alternative to REF2014?; 2013. Available from: http://deevybee.\nblogspot.co.uk/2013/01/an-alternative-to-ref2014.html.\n7. Mingers J, O\u2019Hanley JR, Okunola M. Using Google Scholar institutional level data\nto evaluate the quality of university research. Scientometrics. 2017;113(3):1627\u2013\n1643.\n8. HEFCE. Research Excellence Framework 2014: Overview report by Main Panel A\nand Sub-panels 1 to 6; 2015. Available from: http://www.ref.ac.uk/2014/media/\nref/content/expanel/member/Main%20Panel%20A%20overview%20report.pdf.\n9. HEFCE. Research Excellence Framework - Results and Submissions; 2014. Avail-\nable from: http://results.ref.ac.uk/Results.\n10. HEFCE. Annex A - Summary of additional information about outputs;\n2014. Available from: http://www.ref.ac.uk/2014/media/ref/content/pub/\npanelcriteriaandworkingmethods/01_12a.pdf.\n11. Herrmannova D, Knoth P. An analysis of the microsoft academic graph. D-Lib\nMagazine. 2016;22(9/10).\n12. Hug SE, Br\u00e4ndle MP. The coverage of Microsoft Academic: Analyzing the publi-\ncation output of a university. Scientometrics. 2017;113(3):1551\u20131571.\n13. Baccini A, De Nicolao G. Do they agree? Bibliometric evaluation versus in-\nformed peer review in the Italian research assessment exercise. Scientometrics.\n2016;108(3):1651\u20131671.\n14. Stern N, et al. Building on success and learning from experience: an independent\nreview of the Research Excellence Framework. London: UK Government, Ministry\nof Universities and Science. 2016;.\n15. HEFCE. Report on the pilot exercise to develop bibliometric indicators for the\nResearch Excellence Framework; 2016.\n16. Hojat M, Gonnella JS, Caelleigh AS. Impartial judgment by the \u201cgatekeepers\u201d\nof science: fallibility and accountability in the peer review process. Advances in\nHealth Sciences Education. 2003;8(1):75\u201396.\n17. Lee CJ, Sugimoto CR, Zhang G, Cronin B. Bias in peer review. Journal of the\nAssociation for Information Science and Technology. 2013;64(1):2\u201317.\n18. Smith R. Peer review: a flawed process at the heart of science and journals. Journal\nof the royal society of medicine. 2006;99(4):178\u2013182.\n19. Emerson GB, Warme WJ, Wolf FM, Heckman JD, Brand RA, Leopold SS. Testing\nfor the presence of positive-outcome bias in peer review: a randomized controlled\ntrial. Archives of internal medicine. 2010;170(21):1934\u20131939.\n20. Coles P. The apparatus of research assessment is driven by the academic publishing\nindustry; 2013. Available from: https://bit.ly/2EfNMeV.\n21. Sayer D. Rank hypocrisies: The insult of the REF. Sage; 2014.\n22. Technopolis. REF Accountability Review: Costs, benefits and burden; 2015.\n", "identifiers": ["oai:oro.open.ac.uk:55743", "10.1007/978-3-030-00066-0_17"], "relations": [], "repositories": [{"id": "86", "openDoarId": 0, "name": "Open Research Online", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "baseId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 1, "metadataAdded": 1532341932000, "metadataUpdated": 1591929820000, "depositedDate": 1531868400000}, "subjects": [], "title": "Peer review and citation data in predicting university rankings, a large-scale analysis", "topics": [], "types": [], "year": 2018, "fulltextIdentifier": "https://core.ac.uk/download/pdf/159766439.pdf", "doi": "10.1007/978-3-030-00066-0_17", "oai": "oai:oro.open.ac.uk:55743", "downloadUrl": "https://core.ac.uk/download/pdf/159766439.pdf"}}
{"status": "OK", "data": {"id": "42702124", "authors": ["Escobar, Vanessa M.", "Delgado-Arias, Sabrina", "Hendrick, Oscar", "Brown, Molly E.", "Griffith, Peter", "Ihli, Monica"], "citations": [], "contributors": [], "datePublished": "January 2015", "description": "The North American Carbon Program (NACP) was formed to further the scientific understanding of sources, sinks, and stocks of carbon in Earth's environment. Carbon cycle science integrates multidisciplinary research, providing decision-support information for managing climate and carbon-related change across multiple sectors of society. This investigation uses the conceptual framework of com-munities of practice (CoP) to explore the role that the NACP has played in connecting researchers into a carbon cycle knowledge network, and in enabling them to conduct physical science that includes ideas from social science. A CoP describes the communities formed when people consistently engage in shared communication and activities toward a common passion or learning goal. We apply the CoP model by using keyword analysis of abstracts from scientific publications to analyze the research outputs of the NACP in terms of its knowledge domain. We also construct a co-authorship network from the publications of core NACP members, describe the structure and social pathways within the community. Results of the content analysis indicate that the NACP community of practice has substantially expanded its research on human and social impacts on the carbon cycle, contributing to a better understanding of how human and physical processes interact with one another. Results of the co-authorship social network analysis demonstrate that the NACP has formed a tightly connected community with many social pathways through which knowledge may flow, and that it has also expanded its network of institutions involved in carbon cycle research over the past seven years", "fullText": "S\nP\nM\nV\na\nb\nc\na\nK\nN\nC\nK\nK\nC\nC\n1\ne\nc\nn\nt\nf\ns\ns\nc\ng\nc\nc\nh\n0\n0ZSocial Networks 44 (2016) 226\u2013237\nContents lists available at ScienceDirect\nSocial  Networks\njo u r n al hom ep age: www.elsev ier .com/ locat e/socnet\nocial  network  and  content  analysis  of  the  North  American  Carbon\nrogram  as  a  scienti\ufb01c  community  of  practice\nolly  E.  Browna,\u2217,  Monica  Ihli b, Oscar  Hendrickc,  Sabrina  Delgado-Ariasc,\nanessa  M.  Escobarc, Peter  Grif\ufb01thc\nDepartment of Geographical Sciences, University of Maryland, College Park, MD 20740, United States\nUniversity of Tennessee, Knoxville, TN 37996, United States\nScience Systems and Applications Inc., NASA Goddard Space Flight Center, Code 618, Greenbelt, MD 20771, United States\n r  t  i  c  l e  i  n  f  o\neywords:\north American Carbon Program\narbon cycle\nnowledge domain\nnowledge visualization\nommunities of practice\no-authorship network analysis\na  b  s  t  r  a  c  t\nThe  North  American  Carbon  Program  (NACP)  was formed  to  further  the  scienti\ufb01c  understanding  of\nsources,  sinks,  and  stocks  of  carbon  in  Earth\u2019s  environment.  Carbon  cycle  science  integrates  multidis-\nciplinary  research,  providing  decision-support  information  for managing  climate  and  carbon-related\nchange  across  multiple  sectors  of  society.  This  investigation  uses  the  conceptual  framework  of com-\nmunities  of practice  (CoP)  to explore  the  role that  the  NACP  has  played  in  connecting  researchers  into  a\ncarbon cycle  knowledge  network,  and  in  enabling  them  to conduct  physical  science  that  includes  ideas\nfrom  social  science.  A  CoP  describes  the  communities  formed  when  people  consistently  engage  in shared\ncommunication  and  activities  toward  a common  passion  or learning  goal.  We  apply  the  CoP  model  by\nusing keyword  analysis  of  abstracts  from  scienti\ufb01c  publications  to analyze  the  research  outputs  of  the\nNACP in  terms  of  its  knowledge  domain.  We  also  construct  a  co-authorship  network  from  the publications\nof  core NACP  members,  describe  the  structure  and  social  pathways  within  the community.  Results  of the\ncontent  analysis  indicate  that  the  NACP  community  of  practice  has  substantially  expanded  its research\nhttps://ntrs.nasa.gov/search.jsp?R=20150023370 2019-08-31T05:13:49+00:0on human  and  social  impacts  on  the  carbon  cycle,  contributing  to  a better  understanding  of how  human\nand  physical  processes  interact  with  one  another.  Results  of  the  co-authorship  social  network  analy-\nsis  demonstrate  that  the  NACP  has  formed  a tightly  connected  community  with  many  social  pathways\nthrough  which  knowledge  may  \ufb02ow,  and  that  it has  also  expanded  its  network  of institutions  involved\nin  carbon  cycle  research  over the past seven  years.\n\u00a9 2015  Elsevier  B.V.  All  rights  reserved.. Introduction\nClimate change has emerged as a signi\ufb01cant scienti\ufb01c, social and\nconomic challenge to society (IPCC, 2014). Understanding how\nlimate change may  evolve over the coming decades requires sig-\ni\ufb01cant investment in research about carbon and how it cycles,\nhrough both living and nonliving states (Smil, 1996). Scientists\nrequently study these biogeochemical cycles in the context of\nubsystems such as the terrestrial biosphere (land-based living\nystems), oceanic systems (both organic and inorganic forms of\narbon), and the atmosphere (Falkowski et al., 2000). These investi-\nations may  also include the speci\ufb01c role humans play in the carbon\nycle, such as the impact of human-generated emissions or the\nonsequences of climate change to agriculture and food systems\n\u2217 Corresponding author.\nE-mail address: mbrown52@umd.edu (M.E. Brown).\nttp://dx.doi.org/10.1016/j.socnet.2015.10.002\n378-8733/\u00a9 2015 Elsevier B.V. All rights reserved.(Berthelot et al., 2002; Bradbear and Friel, 2013; Dempewolf et al.,\n2014; Shindell et al., 2012). Carbon cycle science is relevant to a\ngreat many aspects of life as we know it: the condition of our envi-\nronment, the quality of air we breathe, water resources, the food\nthat we  eat, and the energy we consume.\nEngaging across the social and physical sciences to embrace all\naspects of the carbon cycle is very challenging, particularly when\nthe implications of the research are both political and economic.\nThe North American Carbon Program (NACP) is one of the few pro-\ngrams on this topic to host collaborative activities cutting across\nall carbon cycle science disciplines, and promoting opportunities to\nfoster interdisciplinary and intramural collaboration whose objec-\ntive it is to do interdisciplinary research that results in information\nthat can be directly relevant to critical social decision making\n(Michalak et al., 2011). Central to the program\u2019s science agenda is\nthe engagement of social, economic and policy-relevant research in\norder to improve how carbon cycle science is conducted to ensure\npolicy-relevant \ufb01ndings. This paper uses communities of practice\n Netw\na\na\nc\nr\nc\n1\nu\nm\np\nt\ni\ns\nv\na\nt\nf\ns\ns\nc\ne\nc\nf\ns\nu\na\na\n(\nb\nr\ne\na\nt\ne\np\nh\no\nc\nr\no\nt\na\nm\nh\ns\ns\ne\ns\nc\ne\nb\na\no\nc\np\ne\nt\no\nb\nl\nd\ntM.E. Brown et al. / Social\ns a conceptual model for exploring how the NACP has fared in cre-\nting such a knowledge network, both in terms of measuring the\nonnectivity among program participants, and in terms of incorpo-\nating measuring social, economic and policy-relevant topics into\narbon cycle science research.\n.1. History of the NACP\nThe NACP was formally recognized by the United States in 2002\nnder the mantle of the nation\u2019s overall climate change manage-\nent strategy (Wofsy and Harriss, 2002). The \ufb01rst implementation\nlan for the NACP put forward a research agenda that was cen-\nered on quantifying and understanding carbon sinks and sources\nn North America and surrounding oceans, and the integration of\nuch information into socially, economically, and politically rele-\nant decision-support systems (Sarmiento and Wofsy, 1999; Wofsy\nnd Harriss, 2002). The State of the Carbon Cycle Report established\nhat North America is a net source of CO2 to the atmosphere, due to\nossil-fuel emissions and that there are globally important carbon\ninks whose future is highly uncertain (King et al., 2007). Under-\ntanding how humans both experience and in\ufb02uence the carbon\nycle and climate change is critical to the interests of decision mak-\nrs (Bernabo, 1995; Feldman and Ingram, 2009), such as those who\nonfer support upon the member agencies of the NACP through\nunding and other resources.\nIn 2007, The US North American Carbon Program (NACP)\nponsored its \ufb01rst \u201call-scientist\u201d meeting to review progress in\nnderstanding the dynamics of the carbon cycle of North America\nnd adjacent oceans, and to chart a course for improved integration\ncross scienti\ufb01c disciplines, scales, and Earth system boundaries\nBirdsey et al., 2007). Following this meeting, a 2011 US Car-\non Cycle Science Plan was published that set forth priorities for\nesearch in carbon cycle science for the coming decade (Michalak\nt al., 2011). In addition to reaf\ufb01rming the need for basic research\nnd for continuing traditional research in carbon cycle science,\nhe plan recommended substantial expansion in research on the\nf\ufb01cacy and environmental consequences of carbon management\nolicies, strategies, and technologies; prioritization of research on\numan elements of the carbon cycle; an increased exploration\nf the direct impact of rising greenhouse gas concentrations and\narbon-management decisions on ecosystems, species, and natural\nesources; and research on how to express uncertainty in all aspects\nf the global carbon cycle as well as improved ways of conveying\nhose uncertainties to policy and decision makers, as well as society\nt large. To achieve these objectives, the report authors recom-\nended a substantial focus on conducting research that integrates\numan dimensions with the biologic, atmospheric, and oceanic\nciences. Social processes that drive land use and fossil fuel emis-\nions should be quantitatively integrated into land use/cover and\nmissions modeling to promote the integrated carbon, climate, and\nocial modeling needed to provide science and analytical tools for\nlimate action programs at various levels of government (Michalak\nt al., 2011).\nThe challenges facing the North American Carbon Program\nring to light the larger issue as to how organizations, agencies,\nnd nations at any level can cultivate the development of inter-\nrganizational and interdisciplinary networks targeted toward\nreation of speci\ufb01c kinds of knowledge resources. To that effect, this\naper seeks to apply a systemic approach for assessing the knowl-\ndge creation that takes place within a research program such as\nhe NACP. How might we compare the professed knowledge goals\nf the NACP, or similar programs, to the actual knowledge created\ny participants? We also consider how to describe the state of col-\naborations between participants within such a community. How\no collaborations amongst core participants grow and change over\nime? Are there changes in researchers\u2019 tendencies to collaborateorks 44 (2016) 226\u2013237 227\nacross institutional boundaries over the same period of time? These\nare the questions we seek to answer in analyzing the NACP as a\ncommunity of practice.\n2. Theory and rationale\nA community of practice is de\ufb01ned as \u201ca group of people who\nshare a common set of problems, or a passion about a topic, and who\ndeepen their knowledge and expertise in this area by interacting\non an ongoing basis\u201d. Wenger et al. (2002) describe three struc-\ntural elements to the CoP model: domain, community, and practice.\nDomain refers to the knowledge concerns and issues around which\nthe CoP is structured. A well-de\ufb01ned knowledge domain translates\nto a strong sense of purpose, guiding the activities of members. It\nalso implies a shared competence and commitment to the subject\nmatter. Domain manifests as the speci\ufb01c knowledge the com-\nmunity develops, shares, and maintains. The community element\nreferences the social environment itself: the people and relation-\nships through which learning, knowledge transfer, and knowledge\ncreation takes place. Practice concerns all of their rituals, systems\nof meaning, and channels of communication.\nThe CoP model provides a theoretical foundation upon which\nto base discussion and analysis of the scienti\ufb01c community and\nits research. Structural elements of the model aid us in commu-\nnicating fundamental assumptions as well as limitations of the\nstudy (Wenger et al., 2002). Additionally, it aids us in understand-\ning and expressing the relationships and distinctions between a\ncommunity, individual members, and separate but participatory\ninstitutions that provide support to scientists. Other examples\nof the CoP model being employed to study knowledge networks\nvia shared resources and sustained interaction, including an\nethnographic study of climate change adaptation projects at the\nscience-policy interface (Iyalomhe et al., 2013).\n2.1. Community of the NACP\nThere is no a single form of social structure which quali\ufb01es as a\ncommunity of practice, and membership is not by virtue of tradi-\ntional organizational or departmental boundaries. The size of the\ncommunity could be less than ten, or it could number into the\nhundreds. It might be community of individuals who all live or\nwork in close proximity to one another, or it could be distributed\nacross a wide range of geographical locations and organizational\nboundaries. The CoP model does, however, predict an approximate\ndistribution of member participation which corresponds to three\nbroad levels of investment within the community: (1) a small group\nof core members who both attend meetings regularly and who\nalso oversee functional tasks, (2) active members who regularly\nattend meetings, and (3) peripheral members who only occasion-\nally participate in the community (Wenger et al., 2002). As part\nof conceptualizing the NACP as a community of practice, we  will\nconsider if the distribution of participation frequency in meetings\nshows any agreement to the distribution suggested by the model.\nWe will describe and analyze the community in terms of rela-\ntionships between core participants using social network analysis\nmethods. As the NACP also seeks to increase collaboration between\ndifferent institutions studying the carbon cycle, we will further-\nmore look at how the relationships between individuals translate\nto connections between the institutions they represent.\n2.2. Knowledge domain of the NACPA domain is \u201ca statement of what knowledge the community will\nsteward\u201d and \u201ca commitment to take responsibility for an area of\nexpertise\u201d (Wenger et al., 2002). We  have noted that the knowledge\ndomain of the NACP is codi\ufb01ed within the US Carbon Cycle Science\n2  Netw\nP\nf\na\ni\nw\na\ns\nt\nt\nm\n(\np\n2\ng\nc\nc\nt\n\ufb01\nt\n(\ni\ni\na\no\ni\nr\n2\nd\nt\nb\no\no\ns\nt\ns\na\nk\nt\nt\ns\ne\nt\ne\no\ne\nb\nt\nh\nr\na\na\np\ne\no\ns\nC\nd\nW\nn\ni28 M.E. Brown et al. / Social\nlan (Michalak et al., 2011), the Science Implementation Strategy\nor the North American Carbon Program of 2005 (Denning, 2005),\nnd similar supporting documents. The domain of the North Amer-\ncan Carbon Program is to study the sources and sinks of carbon\nith the expectation that resulting knowledge should ultimately be\nccessible and salient to stakeholders at a variety of levels. Although\ncienti\ufb01c research does not have a simple cause-and-effect rela-\nionship with improved societal outcomes, some research has led\no reduced exposure to extreme events, improved forest manage-\nent, and effective policies to ensure agricultural sustainability\nRosenzweig et al., 2014). Rather, science is one among many com-\nlex factors that in\ufb02uence improved societal outcomes (Pielke et al.,\n010). It is inherent to the domain that natural sciences be inte-\nrated with the study of human processes.\nHowever, the integration of social and human aspects in to\narbon science is challenged by the need for translation and\nooperation between different kinds of stakeholders. Researchers\nend to interact more closely with other researchers in their own\nelds, which can frustrate interdisciplinary cooperation amongst\nhose who study natural sciences, social sciences, and economics\nFeldman and Ingram, 2009). This is what is called for accord-\nng to the NACP strategic plan, but are different disciplines truly\nntegrated into the knowledge created by the NACP? The content\nnalysis section of our research methods will compare the domain\nf the NACP against the actual knowledge produced in practice,\nn order that we may  evaluate if the human processes are truly\nepresented as necessary to ful\ufb01ll the program\u2019s goals.\n.3. Practice of the NACP\nA CoP is best characterized by consistent engagement in a\nomain-driven practice over time. Shared practice refers not only\no the actions and channels of communication between members,\nut also to the actual codi\ufb01ed knowledge that is created as a result\nf the interactions between members. For this reason, the existence\nf interpersonal relationships is a key feature \u2013 research does not\nimply happen to co-occur when people are physically present at\nhe same meetings. Interactions occur through and in the course of\nhared practices over time and space.\nIf the NACP is truly a community, then members should be inter-\ncting with one another, and these interactions should manifest as\nnowledge outputs. We  should be able to observe something about\nhe state of the community, as well as its evolution, by describing\nhese relationships. Knowledge products of relationships between\ncientists typically occur in the form of published research (Wenger\nt al., 2002) and coauthored projects. NACP members also con-\nribute to knowledge outputs in the form of of\ufb01cial reports (King\nt al., 2007; Michalak et al., 2011).\nThe continued growth and development of a community rests\nn the ability to gauge its effectiveness in cultivating the knowl-\ndge domain by which it is bound (Wenger et al., 2002). It would\ne therefore of interest to the NACP to have a method to assess\nhe community\u2019s progress of integrating carbon cycle science with\numan processes to provide support for decision making. Peer-\neviewed publications and funded projects provide a consistent\nnd accessible form of knowledge output. Here we will use content\nnalysis and social network analysis to describe the knowledge out-\nuts and member-to-member relationships, respectively. To that\nffect, a keyword analysis is performed on the abstracts of NACP\nutputs in order to derive the topics of research being produced\no that it can be compared to the knowledge goals of the program.\no-authorship relationships are also derived from the bibliographic\nata of articles and of projects as part of our social network analysis.\ne will demonstrate the value of these tools for helping a commu-\nity of practice such as the NACP to continually assess and re-orient\ntself toward its knowledge creation goals.orks 44 (2016) 226\u2013237\n3. Data\nWe  used two  datasets to explore the degree to which the NACP\nis a community of practice that has fostered the goals to expand the\nprogram\u2019s disciplinary focus, as set out in the 2011 NACP plan: the\nNACP project database and a bibliographic dataset from the Web  of\nScience. These are described below.\n3.1. NACP project database\nWe  used NACP\u2019s own  author database (found at www.nacarbon.\norg) to identify 1070 individuals who  are engaged in 408 sci-\nenti\ufb01c projects registered with the NACP, including author, title,\nabstract and keywords and used these to analyze the connectiv-\nity between NACP members. We  narrow down to 1007 members,\nincluding only members who have worked with at least one other\nperson in a project. Scientists and projects are included in the\ndatabase if a project is funded through extramural awards (grants,\ncooperative agreements, contracts, and interagency transfers) and\ndesignated as part of the NACP at the time of selection and award,\nor are designated as part of the NACP subsequent to the award\nbeing made. These projects address the goals and objectives of the\nNACP and contribute relevant research products within the time-\nframe of the program. Core projects also include standing agency\nactivities, including those that may  have another primary mission\nbut agency program leads commit to participate in NACP or to\nprovide speci\ufb01c measurements, data sets, or other products. Af\ufb01l-\niated projects are those that are identi\ufb01ed after selection or after\ncompletion by agency program managers because they are investi-\ngations, research projects, or operational programs or projects that\nare relevant or provide specialized data sets or services important\nto the NACP community. We used the project database to identify\nconnectivity between NACP scientists that participated on projects\ntogether, and in determining the topics represented by their studies\nused the abstracts provided in the NACP database.\n3.2. Bibliographic information from the Web  of Science\nBased on meeting registration records exported from the NACP\u2019s\nonline database, we found that 808 unique participants had\nattended the four NACP meetings between 2007 and 2013. Forty\nof these (approximately 5%) attended all four meetings, 79 people\nattended three meetings (10%), 184 people (23%) participated in\ntwo meetings, and 505 people (63% of participants) attended one\nmeeting. To limit the bibliographic analysis to a manageable sam-\nple, we used the 15% (113 individuals) who attended between three\nand four of the NACP meetings in the past six years, obtaining the\nfamily name, \ufb01rst initial, and organizational af\ufb01liation from NACP\nrecords. Bibliographic data and abstracts for articles were collected\nby searching the Multidisciplinary ISI Web  of Science (accessed\nbetween 12-06-2013 and 01-03-2014). We located papers writ-\nten by all individuals, resulting in a dataset of 2447 peer-reviewed\narticles published in 511 journals between 2007 and 2013. Both the\nregistration records of individuals taken from the www.nacarbon.\norg website and the bibliographic article data were imported into\nour project database and linked in a relational schema. Additional\npreprocessing included the standardization of organizational af\ufb01l-\niation data in the registration records.\nThe network analysis portion of our research required several\nadditional stages of preprocessing of the article data. For reasons\nfurther discussed in the appropriate section of our methods, we\n\ufb01rst reduced the dataset to only include those articles that had\nbeen published by greater than one NACP core member from our\nsample. This left us with 363 articles representing the work of 99\npersons. For an article to be included in our sample, each article\nin the network must connect one member of our NACP sample to\nM.E. Brown et al. / Social Networks 44 (2016) 226\u2013237 229\nTable  1\nKeyword lists by category.\nEconomy Words signifying topics of information on economy-related issues, which connect to the evolution of human\ncarbon system drivers and also relate to issues important to decision makers.\nCommand economy Food Food source\nConsumer Food availability Food sources\nCost  bene\ufb01t Food demand Food trade\nEconomic Food demands GDP\nEconomic activity Food distribution Goods and services\nEconomic forecasting Food insecurity Gross national product\nEconomic model Food policies Industrial\nEconomic models Food policy Industry\nEconomic projections Food producing In\ufb02ation\nEconomic sectors Food production Integrated assessment\nEconomical Food security Investment\nEconomies Food shortage Markets\nEconomy Food shortages Socioeconomic\nFiscal policy Waste management\nEnergy Words associated with information needed in the energy sector, such as sources of energy, energy\nmanagement concerns, and sustainability of sources.\nBioenergy Energy policies Energy uses\nBiofuel Energy policy Fossil fuel\nBiofuels Energy portfolio Fuel\nDam Energy production Fuels\nElectrical Energy resources Hydroelectric\nElectricity Energy source Oil and gas\nEnergy budget Energy sources Renewable energy\nEnergy consumption Energy supplies Sustainable energy\nEnergy ef\ufb01ciency Energy supply Transportation\nEnergy forecasts Energy sustainability Wind \ufb01eld\nEnergy harvesting Energy use Wind \ufb01elds\nEnergy management Wind power\nLand  Use & Water Words associated with human factors that determine carbon emissions from land use & water cycle.\nAgricultural Forest management Land uses\nAgriculturally Forest policies Maize\nAgriculture Forest policy Ownership\nAgroforestry Grazing Pasture\nCorn Harvest Soybean\nCrop Harvested Urban\nCropland Harvests Water availability\nCropping Land management Water management\nCrops  Land ownership Water quality\nFarming Land productivity Water supplies\nFisheries Land use Water supply\nFishing Water treatment\nHuman Impacts Words signifying content discussing how humans interact with and in\ufb02uence the carbon cycle.\nAnthropogenic Emissions Human activities\nDeforestation GHG Human activity\nDisturbance Greenhouse gas Pollute\nDisturbances Greenhouse gases Polluting\nEmission Pollution\nSociety & Decision Making Words which also general signi\ufb01ers of topics that make carbon cycle science relevant to decision makers and\nmanagers who  must consider the values and needs of their constituents.\nAdaptation Operations REGULATION\nAttitudes Planning REGULATIONS\nCultural Policies RESOURCES\nCulture Policy RISK\nDecision Political SOCIAL\nDecisions Practice SOCIOPOLITICAL\nInternational policy Practices STAKEHOLDERS\nManage Regulating STRATEGIES\nManagement STRATEGY\nManagers ZONING\nStrategies Words signifying speci\ufb01c responses and strategies for managing human impacts to the carbon cycle.\nCarbon accounting Conserve Offsetting\nCarbon budget Emissions trading Renewable\nCarbon offset Mitigation Sequestered\nCarbon offsets Monitoring, accounting, and Sequestration\nCarbon program Reporting Sustainability\nCarbon trade MRV  Sustainable\nConservation\n2  Networks 44 (2016) 226\u2013237\na\nn\nw\na\nt\nt\n4\n4\na\nt\ne\na\nk\nC\ne\ni\n2\nd\na\na\np\na\ne\nt\nw\nt\nt\nt\na\na\ni\ni\no\nt\nd\no\np\nw\nt\ni\no\ne\ni\nw\nM\nc\na\nm\n4\na\nf\np\ni\na\nt\nd\nTable 2\nTop thirty journals for all articles in the Web  of Science article dataset and for the sub-\nset of articles coded with human process keywords in the content analysis portion\nof  the study.\nRank Journals Articles % of total\nTop 30 journals in articles dataset\n1 Remote Sens. Environ. 148 6\n2  J. Geophys. Res. Biogeosci. 123 5\n3  Agric. For. Meteorol. 101 4\n4  Glob. Change Biol. 89 4\n5  J. Geophys. Res. Atmos. 85 3\n6  Biogeosciences 71 3\n7  Atmos. Chem. Phys. 54 2\n8  Geophys. Res. Lett. 45 2\n9  For. Ecol. Manage. 41 2\n10  Glob. Biogeochem. Cycle 39 2\n11  Environ. Res. Lett. 38 2\n12  Ecol. Model. 37 2\n13  Ecol. Appl. 35 1\n14  Proc. Natl. Acad. Sci. U. S. A. 31 1\n15  Int. J. Remote Sens. 29 1\n16  Agron. J. 28 1\n17  Tellus Ser. B: Chem. Phys. Meteorol. 26 1\n18  Can. J. For. Res. Rev. Can. Rech. For. 24 1\n19  Ecosystems 23 1\n20  J. Geophys. Res. Oceans 22 1\n21  PLoS One 20 1\n22  Atmos. Meas. Tech. 19 1\n23  J. Environ. Manage. 18 1\n24  Atmos. Environ. 18 1\n25  IEEE Trans. Geosci. Remote Sensing 16 1\n26  Agric. Ecosyst. Environ. 16 1\n27  Soil Sci. Soc. Am.  J. 16 1\n28  Clim. Change 15 1\n29  Bull. Amer. Meteorol. Soc. 15 1\n30  Biogeochemistry 15 1\nTop  30 journals among articles coded w/human processes keywords (n = 1397)\n1 Remote Sens. Environ. 87 6.2\n2  J. Geophys. Res. Biogeosci. 62 4.4\n3  Agric. For. Meteorol. 53 3.8\n4  J. Geophys. Res. Atmos. 52 3.7\n5  Glob. Change Biol. 49 3.5\n6  Biogeosciences 47 3.4\n7  Atmos. Chem. Phys. 36 2.6\n8  For. Ecol. Manage. 35 2.5\n9  Geophys. Res. Lett. 25 1.8\n10  Ecol. Appl. 25 1.8\n11  Environ. Res. Lett. 25 1.8\n12  Agron. J. 24 1.7\n13  Glob. Biogeochem. Cycle 23 1.6\n14  Proc. Natl. Acad. Sci. U. S. A. 22 1.6\n15  Ecol. Model. 22 1.6\n16  Tellus Ser. B: Chem. Phys. Meteorol. 17 1.2\n17  Can. J. For. Res. Rev. Can. Rech. For. 16 1.1\n18  Agric. Ecosyst. Environ. 15 1.1\n19  PLoS One 15 1.1\n20  Clim. Change 14 1.0\n21  Ecosystems 12 0.9\n22  Soil Sci. Soc. Am.  J. 12 0.9\n23  Environ. Sci. Technol. 12 0.9\n24  Bioscience 11 0.8\n25  GCB Bioenergy 10 0.7\n26  Int. J. Remote Sens. 10 0.7\n27  Atmos. Environ. 9 0.630 M.E. Brown et al. / Social\nnother through inclusion of two names in the author list. Con-\nections to people outside our sample were not represented. Next,\ne programmatically restructured the data so that an export of\nll records relating people to articles would be converted into a\nable representing all possible combinations of people connected\no other people, where they had authored the same document.\n. Methods\n.1. Content analysis of publications and projects\nScientometric mapping and analysis using bibliographic data is\n well-established methodology, for which a variety of approaches,\nechniques, and automated tools have been developed (Cobo\nt al., 2011). Existing research has demonstrated that analysis of\nbstracts and titles in bibliographic data can yield insight into the\nnowledge domains represented within the dataset (Chen, 2006).\nontent analysis may  be approached either inductively, in a purely\nxploratory context, or deductively when seeking to test known\ndeas or compare changes in content over time (Elo and Kyng\u00e4s,\n008). Our keyword analysis was conducted on both the NACP\natabase and the bibliographic data by importing the author, title\nnd abstract information from both articles and projects into a text\nnalysis software. Because we desired to compare the known and\nublished knowledge goals of the program to research outputs, we\nssembled a dictionary of terms and phrases (Table 1) that refer-\nnce different aspects of human processes and experiences relevant\no integration with carbon cycle science. The initial source of key-\nords were terms and ideas used in Michalak et al. (2011), and\nhese were then evaluated in the context of abstract texts to con\ufb01rm\nheir appropriateness for reporting on in a keyword analysis. The\nop thirty journals for all articles (2447) versus articles returning\nt least one keyword from the human processes categories (1397)\nre summarized in Table 2.\nSome words, such as \u201cbiofuel\u201d, are very explicit in their mean-\nng and relevance. Other words, such as \u201cpolicy\u201d or \u201cdecision\u201d, are\nnherently associated in the English language with human agency\nr perspective. We  have little reason to be concerned for false posi-\nives because we do not speak of trees \u201cdeciding\u201d things or clouds\neveloping a \u201cpolicy\u201d. However, these words can still be paired with\nther words to create more explicit meaning, such as with \u201cenergy\nolicy\u201d. On the other hand, \u201cenergy\u201d alone would risk confusion\nith any number of scienti\ufb01c processes described in our abstracts\nhat have nothing to do with humans. Therefore, we could not\nnclude a word such as \u201cenergy\u201d without further quali\ufb01cation in\nrder to ensure our results are meaningful.\nThe categories are neither mutually exclusive nor collectively\nxhaustive. Some keywords would have caused an article to fall\nnto more than one category. For example, the text \u201cenergy policy\u201d\nill \ufb02ag an abstract into the general category of \u201cSocial & Decision\naking\u201d category due to the word policy, as well as the more spe-\ni\ufb01c Energy category. This was deliberate; it was determined our\nnalysis would be more informative if we did not enforce a rule of\nutual exclusivity.\n.2. Social network analysis of co-authorship ties\nWe  conducted a network analysis on both the NACP database\nnd the bibliographic data, and present network graphs resulting\nrom the analysis. Network analysis is a research approach that\nrioritizes relationships between social units as opposed to focus-\nng on attributes of the individual units themselves (Wasserman\nnd Faust, 1994). Here we report on co-authorship networks\nhat describe relationships between people rather than ideas. A\nocument \u201cis co-authored if it has more than one author. It is28  J. Clim. 9 0.6\n29  Nature 9 0.6\n30  J. Environ. Manage. 9 0.6\ninstitutionally co-authored if it has more than one author address,\nsuggesting that the authors come from various institutions, depart-\nments, or other kinds of units\u201d (Melin and Persson, 1996).\nMany studies have used the relationships between authors of\na document to study broad domains such as physics (Newman,\n2001a) and mathematics (Barab\u00e1si et al., 2002). More recently,\nstudies have also employed narrower sampling approaches, such\nas only including data from a speci\ufb01c journal (Martin et al., 2013)\n Networks 44 (2016) 226\u2013237 231\no\nb\nt\ns\ng\ni\n(\nv\na\ne\nw\nW\nd\nw\na\nw\nw\ns\nN\ns\nm\nb\no\no\nw\np\nt\nn\nm\nm\na\nn\no\ni\nl\ni\ns\ns\na\nn\nt\nt\nc\ni\nc\nc\nc\nn\nt\nn\na\nt\nt\na\nt\ng\nt\nc\nn\nt\nc\nTable 3\nTheoretical and observed participation distributions. The theoretical model for CoP\n(a)  proposes three levels of community involvement and suggests distribution\nranges for the proportion of the community which will fall into each level; (b)\nsummarizes the attendance frequency distribution for individuals (n = 808) who\nattended the most recent four NACP meetings between 2007 and 2013.\n(a) CoP model proposed distribution (b) Observed attendance distribution\n(n = 808)\nCore 10\u201315% 3\u20134 Meetings 14.7%M.E. Brown et al. / Social\nr based on attendance records of participation for the active mem-\ners of a community of business (Morlacchi et al., 2008). Here we\nransform both our datasets into co-authorship network to repre-\nent interpersonal relationships present in the NACP. People are\nenerally expected to communicate and interact with one another\nn order to publish a document or conduct a project together\nNewman, 2001a).\nWe  took bibliographic data from the Web  of Science and con-\nerted into a network format by producing all possible pairs of two\nuthors for each article. The number of network ties produced by\nach article may  be expressed in terms of combinations as C(n,r),\nhere r = 2 and n is the number of authors for a given article.\ne then created a binary code for each row of network data pro-\nuced by this method to identify whether or not the connection\nas between two authors of the same institution. A second co-\nuthorship network was also created from this data, but one in\nhich we represented authors\u2019 institutions as the nodes, and in\nhich connections within the same institutions were ignored as\nelf-loops.\nWe invoked a similar strategy for the data extracted from the\nACP project database. Data here was represented as a relation-\nhip between a project and one or more people. We  used the same\nethod to convert this information to all possible connections\netween pairs of researchers who were said to have participated\nn a given project. Cross-institutional coding was not performed\nn the project dataset.\nNetworks can be analyzed to look for information about the net-\nork as a whole, or to look for individuals of interest, such as which\nerson is most central or most connected in a network. We  tended\no evaluate our graphs for information about average behavior of\nodes as a whole, or at the network level of analysis. For example,\nodularity is the fraction of the connections within given groups\ninus the expected such fraction if connections were distributed\nt random (Newman, 2006). The modularity score given to the\networks can be interpreted as describing how well a network, in\nur case the NACP, can be seen as a collection of sub-networks, or\ns a score of how well organized the network is. Since the modu-\narity score measures the quality of the partitioning of the graph\nnto smaller communities, the fact that the networks have close to\nimilar scores means that they share a similar internal community\ntructure.\nThe average clustering coef\ufb01cient is a measure of how connected\n neighborhood of nodes are. For a particular node (scientist), a\neighborhood is all nodes connected to this particular node. If all\nhe nodes in the neighborhood of a scientist-node are connected\no other scientist-nodes of that neighborhood, then the clustering\noef\ufb01cient of that neighborhood will be 1; if none of the other nodes\nn the neighborhood are connected to each other then the clustering\noef\ufb01cient will be zero (Latapy et al., 2008). In other words, the\nlustering coef\ufb01cient of a node that has at least one other node\nonnected to it, is the probability that any two randomly chosen\neighbors are connected. This probability is calculated by dividing\nhe number of triangles containing our particular scientist by the\number of possible connections between his neighbors (31). The\nverage clustering coef\ufb01cient is the average of these scores over all\nhe neighborhoods.\nIn order to measure the connectedness of the participants of\nhe NACP, we take our two datasets, the NACP project database\nnd journal articles written by the members of the NACP, and use\nhe modularity framework to understand networks within the pro-\nram. If a community is connected by multiple co-authorships, then\nhe community is interpreted to have multiple levels of communi-\nation, interaction and scienti\ufb01c engagement. If a community has\networks that are isolated from one another and are sparse, then\nhe community is assumed to be less engaged and to have less\nollaboration.Active 15\u201320% 2 Meetings 22.8%\nPeriphery 65\u201375% 1 Meeting 62.5%\nOne of the properties we look for in the network of NACP core\nco-authors is the existence of a giant component. Components refer\nto groups of nodes, which are highly inter-connected to each other.\nWhere few connections exist, a network will seem to be a scat-\ntered collection of isolated groups of scientists. This corresponds\nto a lack of social pathways through which ideas can \ufb02ow. But\ngiven enough connections, isolated groups may converge into a sin-\ngle mass, which encompasses the majority of nodes, and in which\neveryone is potentially connected to everyone else by some path\n(Newman, 2001b).\n5. Results\nTable 3 shows the three levels of involvement and their distri-\nbutions, as proposed by the CoP model, along with the summary\ndistribution of meeting attendance frequencies into three corre-\nsponding classes. The nearly two-thirds of the 808 individuals who\nonly attended one meeting between 2007 and 2013 are compared\nto the peripheral component of the CoP model. The almost 23% of\nindividuals who  attended half the meetings are compared to the\nactive component. The nearly 15% who  attended the majority or all\nmeetings are compared to the primary component of a community\nof practice. This distribution of members between the core mem-\nbers, active members, and peripheral members re\ufb02ects the theory\nof a Community of Practice.\nWe identi\ufb01ed 215 institutions, organizations, and agencies\nacross the 808 participants. Although many of the institutions did\nnot account for more than a single participant, 31% of the insti-\ntutions could account for 74% of the people. Among the subset of\n113 core members, we counted 79 institutions, 25% of which could\naccount for 58% of the people in the subset.\n5.1. Results of content analysis\nOf the 2447 articles written by at least one NACP core member\nfrom 2007 to 2013, 57% or 1397 contained at least one human-\nrelated carbon cycle science keyword. Approximately 63% (228) of\nthe 363 articles co-authored by at least two NACP core members\ncontained at least one keyword. We  found a general agreement in\nthe human-related carbon cycle science issues discussed between\n2007 and 2013 in the research written by at least one NACP core\nmember and the research co-authored by at least two NACP core\nmembers. Overall, topics on Society & Decision Making, Land &\nWater Use, and Human Impacts were more prevalent than those\ntopics dealing with human-related issues pertaining to Energy,\nStrategies and Economy (Fig. 1a and b).\nThe same prevalence was found in the analysis of the project\ndatabase. Of the 408 total project abstracts analyzed for human-\nrelated carbon cycle science keywords, 66% or 268 projects\ncontained at least one keyword. Keywords associated with Soci-\nety and Decision Making were found in 60% of the projects, those\nrelated to Land & Water Use in 46% of projects, and those associ-\nated with Human Impacts in 56% of the projects. Contained less\n232 M.E. Brown et al. / Social Networks 44 (2016) 226\u2013237\nFig. 1. (a) Results from content assessment by year and category for the NACP Core Su\nkeyword from Table 1. (b) Results by year and category for all articles (N = 2447).\nF\nl\ni\na\nc\nF\nl\ne\nmig. 2. Results showing the percent of articles from both samples referencing at\neast one keyword and the percent projects referencing at least one keyword.\nn the projects were human-related carbon cycle science keywords\nssociated with Energy (24%), Strategies (23%), and Economy (10%).\nAn overall increase in engagement with human-related carbon\nycle science issues is noted for both NACP core article samples.\nig. 2 also shows that, after 2010, those articles co-authored by at\neast two NACP core members referenced more human-related sci-\nnce topics than those articles written by at least one NACP core\nember. This trend peaks in 2011 at 73% and may  be explained\nFig. 3. Results from content assessment for articles and projects bset articles (N = 363) containing at least one human-related carbon cycle science\nby the release of the 2011 US Carbon Cycle Science Plan (14). The\nplan provides long-term direction for guiding carbon cycle research\nbased on three overarching questions, which include the effects of\nhumans on carbon cycling and the consequences of carbon man-\nagement decisions.\nSince the project database analysis does not provide information\non the evolution through time because there are no dates asso-\nciated with projects in the database, we compare the percentage\nprojects containing at least one keyword (66%) to the percentage\nof total articles from the bibliographic database from 2007 to 2013\ncontaining at least one keyword for both article samples for each\nof the human-related carbon cycle science categories (Fig. 3). The\nNACP projects show a higher focus on Society & Decision Making\n(60%), followed by Human Impacts (56%) and Land & Water Use\n(46%). Articles co-authored by at least two  NACP members focus\nmore on Human Impacts (58%), Land & Water Use (56%), followed\nby Society and Decision Making (48%). Articles written by at least\none NACP core member show a greater tendency for containing\nwords related to Society & Decision Making (53%), Land & Water\nUse (51%) and Human Impacts (44%).\nOverall, there was  a 29% increase in the bibliographic journal\narticles referencing a human-related carbon cycle science keyword\nfor articles co-authored by more than two  NACP core members\nfrom 2007 to 2013 Since these articles were from physical science\nand interdisciplinary journals only, the 29% increase in publications\nby human-related carbon cycle science keyword category.\nM.E. Brown et al. / Social Networks 44 (2016) 226\u2013237 233\nTable  4\nNACP core article authorship and network connection statistics by year. This table describes the connections derived from the 363 articles, which connect 99 of the most\nactive  participants in the NACP biannual meetings. The proportion of ties which crossed institutional boundaries is compared to the quantity of all co-authorship ties for\neach  year.\nPublication year Articles published Co-authorship\nties\nTies which also connected\ndifferent institutions\nRatio of ties crossing\ninstitutions to all ties\n2007 24 43 29 67%\n2008  37 70 59 84%\n2009  39 100 78 78%\n2010  57 330 304 92%\n2011  62 257 226 88%\n2012  78 563 524 93%\n2013  66 192 162 84%\nc\nl\nt\nt\na\no\np\na\nh\na\n5\no\ni\ns\nH\ns\nb\nr\nt\ni\nc\ni\nt\nc\nt\nt\nt\nwTotal  363 1555 \nonsidering social and human aspects of social science is quite\narge, particularly considering the short period of time and the fact\nhat the group of researchers did not change. Our results show that\nhe individuals who consistently participate in the NACP meetings\nre integrating, either intentionally or not, social and human aspects\nf the carbon cycle science into their work, and have been able to\nublish articles that consider these elements in the physical science\nnd interdisciplinary journals. The knowledge domain of the NACP\nas grown over the past seven years to include social and human\nspects of the carbon cycle.\n.2. Results for the social network analysis\nTable 4 reports annual article statistics for the average number\nf core NACP authors per article, as well as the average number of\nnstitutions and nations they represented. Fig. 4 charts the same,\nhowing brief spikes of collaborative activity in 2010 and 2012.\nowever, these spikes of activity are not maintained over time and\no the overall increase in NACP core collaborations per article for\noth individuals and institutions is more moderate. Table 4 also\neports the size of the article network in terms of co-authorship\nies and also quanti\ufb01es the number of these that constitute cross-\nnstitutional co-authorships, broken down annually. The ratio of\nross-institutional ties to all ties is also represented, showing an\nncrease from 67% in 2007 to an overall average of 89% by 2013.\nFig. 5 shows a graph visualizing the co-authorship network for\nhe NACP project database that displays 1007 people with 16,518\nonnections and 15,716 unique edges. The \ufb01gure was created using\nhe Fruchterman\u2013Reingold layout in Gephi. The algorithm repels\nhe nodes from each other and uses the connections between nodes\no pull them toward one another. The result is a network graph\nhere one can visually see well-connected groups and how they\nFig. 4. NACP core authorship statistics by year.1382 89%\ncompare to other groups in the graph. In this graph, each node\nrepresents a scientist; where the darker the blue and the larger\nthe node the greater connected the nodes are in relation to each\nother. Fig. 5 also shows the periphery and the core of the project\nnetwork \u2013 the small circles on the outside are people who work in\nvery small groups. At the center of the graph we  can \ufb01nd the large\nnodes with solid blue colors \u2013 these are scientists who  participate\nin projects and committees identi\ufb01ed by the NACP to accelerate the\nresearch and work together toward a common goal. Near the center\nof the graph we  can \ufb01nd the smaller blue nodes which do not appear\nto be as connected as the larger circles, but still more connected\nthan the scattered nodes in the periphery. These nodes belong to\nthose participating in more than one type of project or are in a core\nproject funded directly by NASA. The colors of the edges in the graph\nrepresent the keyword or keywords that highlight the relationship\nbetween the work of two  scientists. If the most frequently found\nword, within the project abstracts of two scientists, was biofuels,\nthen the keyword associated between these would be energy. If\nties occur, then we  assign the relationship as having more than one\nkeyword. The graph shows \u2018Human Impacts\u2019 as the most frequent\nkeyword showing up in relationships between scientists.\nFinally, the analysis shows that there are 67 communities in\nthe graph with a modularity score of 0.42 out of 1. The modula-\nrity score describes how the network can be seen as a collection\nof sub-networks. The score indicates that the project network is\nmoderately organized, which re\ufb02ects the average clustering coef-\n\ufb01cient of 0.896, where the best score is 1. The analysis shows that\nthe project network has many well-integrated neighborhoods.\nFig. 6 shows a similar network analysis graph for the bib-\nliographic data, with 1555 connections resulting in 630 unique\nconnections between 99 nodes. The article network has a similar\nstructure to the project network in that there is a center and a visi-\nble periphery. The nodes at the center represent the scientists who\ncollaborate together on different ideas, more so than the rest of\nthe people in the graph. The analysis shows that there are 6 com-\nmunities in the graph with a modularity score of 0.329 out of 1,\nwhich indicates that the article network is moderately organized.\nThe average clustering coef\ufb01cient is 0.594 where the best score\nis 1 for this network. Although the score is lower than the project\nnetwork, the average clustering coef\ufb01cient score for the article net-\nwork is still quite high. This shows that the article network and the\nproject network have similar degree of integrated neighborhoods.\nThe Fruchterman\u2013Reingold layout was  used to visualize this net-\nwork in Gephi, where the color shade and size of the nodes indicate\nhow connected they are. The edges are colored by the keyword,\nhowever \u2018Strategies\u2019 was  the most frequent keyword, with \u2018Human\nImpacts\u2019 being a distant second.\nWe can evaluate the bibliographic data to tell a story about\nthe \ufb01nal outcome of the network shown in Figs. 5 and 6. In 2007,\nonly roughly a third of the NACP scientists (35) were connected\n234 M.E. Brown et al. / Social Networks 44 (2016) 226\u2013237\nFig. 5. Co-authorship network using project data from all NACP members. This network had 1007 people with 16,518 connections and 15,716 unique edges. The nodes of\nthe  graph are colored by the degree of connection, so that a darker blue represents the degree of connectedness of nodes. The node sizes are also categorized by the degree\nof  connections. The edges are colored by the keywords associated between two  pairs of nodes, the meaning of which is provided in the legend.\nFig. 6. Co-authorship network using article database of 1555 connections resulting in 630 unique edges between 99 core NACP scientists. The nodes of the graph are colored\nby  the degree of connection, so that a darker blue represents the degree of connectedness of nodes. The node sizes are also categorized by the degree of connections. The\nedges  are colored by the keywords associated between two pairs of nodes, the meaning of which is provided in the legend.\nTable 5\nStatistics for NACP core co-authorship network in which people are represented as nodes.\nEvolution of co-authorship network between 2007 and 2013\nPoint of analysis Years in analysis Nodes Edges Components Avg. number of neighbors\n2007 1 35 43 9 2.11\n2008  2 53 113 5 3.17\n2009  3 64 213 3 4.47\n2010  4 76 543 4 8.16\n2011  5 83 800 2 9.18\n2012  6 93 1363 2 12.41\n2013  7 99 1555 2 12.72\n Netw\nt\nn\nn\nc\no\nf\nt\ni\nt\nd\nn\ng\ni\no\nt\nb\ni\na\na\nn\nt\nN\ni\nw\nb\nd\nt\n6\na\no\nc\no\ns\nt\nw\nt\nm\nr\nc\nw\nn\ne\nv\nr\ne\nl\nw\np\na\nc\nn\nt\nh\ns\nc\nt\nb\ncM.E. Brown et al. / Social\no each other, and their connections formed a fairly fragmented\network (9 fragments/components), with the average number of\neighbors (how many people, on average, each person is directly\nonnected to) at 2.11. Over time, however, we see the emergence\nf the highly interconnected network shown in Fig. 6. We  observe\nrom the increasing number of nodes that, as the years pass, more of\nhese highly active individuals within the community begin form-\nng collaborative relationships with each other. It is also interesting\no observe that, as new nodes enter the network, the number of\nisconnected fragments decreases from 9 to 2, and the average\number of neighbors increases from 2 to 12 (Table 5).\nThe simple fact that the size of the network, in terms of nodes,\nrew over time is not suf\ufb01cient to characterize its evolution. If NACP\nnvestigators had only begun forming collaborations with each\nther in small groups such as pairs or triplets, then we  would expect\nhat the average number of neighbors would remain relatively sta-\nle while the number of components would have substantially\nncreased. The fact that we see the network in its current form,\ns shown in Fig. 6, indicates that those individuals who were most\nctive in attending NACP meetings have formed a well-integrated\network of collaborations.\nThe summary statistics characterizing the NACP community and\nhe network analysis suggest that different institutions within the\nACP are becoming more interconnected. The data showed a steady\nncrease in the ratio of institutional co-authorship ties over time,\nhich may  suggest that continued participation in shared practices\nene\ufb01ts the opportunity for a collection of researchers representing\nifferent institutions to engage in cross-organizational collabora-\nion.\n. Discussion\nWe  observed the existence of relationships encompassing\npproximately 98% of the co-authorship network. Only a single\nther small component of two nodes existed outside of the main\nluster. If the data had described a scattering of islands of collab-\nration, each group unconnected to the others, then the lack of\nocial pathways would be expected to impede the \ufb02ow of ideas\nhroughout the community. However, the tightly clustered net-\nork of core NACP co-authorships supports the hypothesis that\nhe NACP is a community both in name and in practice, and that\nyriad social pathways exist for knowledge sharing and collabo-\nation across the majority of members. It may  also mean that the\nommunity is fairly insular, working more with each other than\nith those on the periphery. The same holds true even when the\network is represented as a graph of institutions.\nAlthough the community of practice framework tends to\nmphasize the role of interpersonal relationships between indi-\niduals, other studies have focused more on understanding the\nole of cross-institutional collaboration in knowledge creation. For\nxample, previous studies have found that cross-institutional col-\naboration supports the diffusion of innovations and new ideas\nithin a \ufb01eld (Zucker and Darby, 1996). Institutionally co-authored\napers have also been found to be more highly cited than papers\nuthored within a single institution.\nOur results suggest that the North American Carbon Program has\nultivated an increasingly connected community of practice whose\networks of collaboration span a wide variety of institutions and\nopics (Table 5). It is comprised of a knowledge network, which\nas gradually extended its topics beyond traditional carbon cycle\ncience. It is clear from the success of the NACP that encouraging\nollaborations to connect isolated fragments and cultivating long-\nerm collaborative relationships across international and cultural\noundaries is important for the improved functioning of a scienti\ufb01c\nommunity of practice.orks 44 (2016) 226\u2013237 235\nHowever, simply writing papers on relevant topics and forming\na tighter research network are not suf\ufb01cient to produce policy-\nrelevant research. The process of connecting science and decision\nmakers must be undertaken. It is not clear from our analysis that\nsuch work is being done yet in the NACP (Dilling and Lemos, 2011;\nLemos et al., 2002). Previous research has shown that just because\nresearch is policy relevant, does not mean that policy makers will\nuse it (Cash et al., 2006). Science has to be done in a way that directly\nengages stakeholders and involves them in setting the research\nagenda, working with them iteratively over long periods of time\nto ensure that the research can be more usable in the end. The\nNACP abstract database does not really provide evidence one way\nor another that that is happening.\nOur research does demonstrate the value of content analysis and\nsocial network analysis using publication data for assessing a CoP\u2019s\nknowledge production against its professed knowledge domain.\nThis has important implications for the ability of the community\nto realign itself with its goals. The community gains the opportu-\nnity to further its development, integrating new members whose\nresearch specialties will contribute to the overall knowledge goals\nof the group and focusing its efforts on gaps in practice and repre-\nsentation.\nOur research has similar \ufb01ndings to other studies that investi-\ngate how to cultivate collaboration in a community of scientists.\nFor example, one study explored the introduction of scientists\nwho were previously unacquainted and from different research\nbackgrounds using brief meetings during a conference (Vaggi\net al., 2014). According to the study, many scientist reported pos-\nitive experiences and potential new collaborations as a result of\nparticipation. The NACP could employ a similar approach by char-\nacterizing the research of members using content analysis and\ncreating new opportunities for fruitful collaborations. The social\nnetwork perspective also enables the community to further sta-\nbilize and expands its connections, by encouraging collaborations\nthat would connect isolated fragments of researcher groups to the\noverall network.\n6.1. Limitations\nThe outcome of our analysis is representative of the community\nthat was  studied and is derived from data on the of\ufb01cial members\nof the community in the NACP database. Without additional study,\nwe cannot know if the results could be generalized for describ-\ning the larger population of all carbon cycle scientists including\nthose from outside the program. Also, as previously mentioned,\nour keyword analysis results are only as thorough as the ability\nof our selected keywords to capture the intended meanings. It is\npossible that other researchers may  have categorized keywords\ndifferently than we  have chosen to do so here. Other researchers\ncould \ufb01nd different results depending on what keywords they use\nin de\ufb01ning categories. Additional limitations are inherent to the\nbibliographic dataset, which was  analyzed. Articles published by\nthe same authors but using different institutional af\ufb01liations would\nhave probably been excluded, as well as any material not archived\nin Web  of Science.\n6.2. Future research\nThis research focused primarily on observing changes occurring\nin the overall network of individuals who either attended NACP\nmeetings regularly, or who  participated in projects related to the\nprogram. Future research should focus on analyzing more closely\nthe attributes and behaviors of individuals within the community.\nA combination of network analysis with more qualitative methods\ncould yield more conclusive insight into in\ufb02uences and motivations\n2  Netw\nf\ni\nt\np\ni\na\no\n7\nt\nt\nh\nN\ni\nt\nt\nr\nb\nm\nW\na\nc\nv\nc\nn\nm\nn\ni\np\nm\ni\nr\np\nf\nm\ne\nA\nA\nt\nt\nA\ni\n0\nR\nA\nB\nB\nB36 M.E. Brown et al. / Social\nor change within a community, as well as understanding of which\nndividuals tend to drive change and how.\nAs the NACP evolves, a program of understanding the impact of\nhe program\u2019s collaborative approach on increasing its impact on\nolicy should be of greater interest. Because this would involve\nnterviewing stakeholders, it is outside of the methodological\npproach of the current article, but should be attempted by the\nrganization in the coming years (Adams et al., 2013).\n. Conclusions\nThis research found an increase in the use of social and economic\nopics in interdisciplinary carbon cycle science research from 2007\no 2013 associated with the NACP members and the papers they\nave written. One conclusion that this result could mean is that the\nACP community is actively working to incorporate human factor\nnto their research, or that the members in the NACP have increased\nhe framing of their research to include these topics to improve\nhe policy relevance of their research, although the research itself\nemains fairly similar. Although our analysis cannot distinguish\netween these two hypotheses, the NACP community is paying\nore attention to the social and economic relevance of their work.\ne  found that core NACP members are well connected with one\nnother, forming a tightly clustered network. Cross-institutional\nollaboration has increased, but more needs to be done to culti-\nate long-term collaborative relationships across international and\nultural boundaries.\nIt is dif\ufb01cult to clearly connect cause and effect with regard to\networks, and note that the analysis presented here cannot deter-\nine why the NACP network has changed. It might be that the\network and institutional diversity has grown and connectivity has\nncreased, but is it because of the development of a community of\nractice or because of the availability of funding? We  cannot deter-\nine this from the analysis presented, but believe that the NACP\ns working hard to increase both the relevance and quality of the\nesearch it does. The NACP has successfully fostered a community of\nractice, and is working toward increasing the inclusion of societal\nactors into its research. The organization can improve its engage-\nent with the international community, and its consideration of\nconomics in its research.\ncknowledgements\nThis research was funded by the National Aeronautics and Space\ndministration through a grant under the Carbon Monitoring Sys-\nem Science Area, Earth Science Directorate. This study was part of\nhe North American Carbon Program.\nppendix A. Supplementary data\nSupplementary data associated with this article can be found,\nn the online version, at http://dx.doi.org/10.1016/j.socnet.2015.10.\n02.\neferences\ndams, V., Blankenship, T., Burgess-Herbert, S., Corley, W.,  Coughlan, J., Gelso, B.,\nHinds, E., Hurley, E., Hutson, M.,  Li, J., Wilson, D., 2013. Measuring Socioeconomic\nImpacts of Earth Observations. National Aeronautics and Space Administration,\nWashington, DC.\narab\u00e1si, A.L., Jeong, H., Neda, Z., Ravasz, E., Schubert, A., Vicsek, T., 2002. Evolution\nof  the social network of scienti\ufb01c collaborations. Physica A 311, 590\u2013614.ernabo, J.C., 1995. Communication among scientists, decision makers and society:\ndeveloping policy-relevant global climate change research. Clim. Change Res.:\nEval. Policy Implic. 65, 103\u2013117.\nerthelot, M., Friedlingstein, P., Ciais, P., Monfray, P., Dufresne, J.L., Le Treut, H., Fair-\nhead, L., 2002. Global response of the terrestrial biosphere to CO2 and climateorks 44 (2016) 226\u2013237\nchange using a coupled climate\u2013carbon cycle model. Glob. Biogeochem. Cycles,\n16.\nBirdsey, R.A., Cook, R., Denning, S., Grif\ufb01th, P., Law, B., Masek, J., Michalak, A., Ogle, S.,\nOjima, D., Pan, Y., Sabin, C., Sheffner, E., Sundquist, E., 2007. Investigators share\nimproved understanding of the North American Carbon Cycle. EOS  Trans. Am.\nGeophys. Union 88, 255.\nBradbear, C., Friel, S., 2013. Integrating climate change, food prices and population\nhealth. Food Policy 43, 56\u201366.\nCash, D.W., Adger, W.N., Berkes, F., Garden, P., Lebel, L., Olsson, P., Pritchard, L.,\nYoung, O., 2006. Scale and cross-scale dynamics: governance and information\nin  a multilevel world. Ecol. Soc. 11, 1\u20138.\nChen, C., 2006. CiteSpace II: detecting and visualizing emerging trends and transient\npatterns in scienti\ufb01c literature. J. Am.  Soc. Inf. Sci. Technol. 571, 359\u2013377.\nCobo, M.J., L\u00f3pez-Herrera, A.G., Herrera-Viedma, E., Herrera, F., 2011. Science map-\nping software tools: review, analysis, and cooperative study among tools. J.\nAssoc. Inf. Sci. Technol. 62, 1382\u20131402.\nDempewolf, H., Eastwood, R.J., Guarino, L., Khoury, C.K., M\u00fcller, J.V., Toll, J., 2014.\nAdapting agriculture to climate change: a global initiative to collect, conserve,\nand use crop wild relatives. Agroecol. Sustain. Food Syst., 4.\nDenning, S.A., 2005. Implementation Strategy for the North American Carbon Pro-\ngram: Report of the NACP Implementation Strategy Group of the U.S. Carbon\nCycle Interagency Working Group. U.S. Carbon Cycle Science Program, Wash-\nington, DC.\nDilling, L., Lemos, M.C., 2011. Creating usable science: opportunities and constraints\nfor climate knowledge use and their implications for science policy. Glob. Envi-\nron.  Change 21, 680\u2013689.\nElo, S., Kyng\u00e4s, H., 2008. The qualitative content analysis process. J. Adv. Nurs. 62,\n107\u2013115.\nFalkowski, P., Scholes, R.J., Boyle, E., Canadell, J., Can\ufb01eld, D., Elser, J., Gruber, N., Hib-\nbard, K., Hogberg, P., Linder, S., Mackenzie, F.T., Moore, B., Pedersen, T., Rosenthal,\nY., Seitzinger, S., Smetacek, V., Steffen, W.,  2000. The global carbon cycle: a test\nof  our knowledge of earth as a system. Science 290, 291\u2013296.\nFeldman, D., Ingram, H., 2009. Making science useful to decision makers: climate\nforecasts, water management, and knowledge networks weather. Clim. Soc. 1,\n9\u201321.\nIPCC., 2014. Working Group II Contribution to the Fifth Assessment Report of the\nIntergovernmental Panel on Climate Change: Impacts, Adaptation and Vulner-\nability. Cambridge University Press, Cambridge, UK.\nIyalomhe, F., Jensen, A., Critto, A., Marcomini, A., 2013. The science\u2013policy inter-\nface for climate change adaptation: the contribution of communities of practice\ntheory. EPG 23, 368\u2013380.\nKing, A.W., Dilling, L., Zimmerman, G.P., Fairman, D.M., Houghton, R.A., Marland,\nG., Rose, A.Z., Wilbanks, T.J., 2007. The First State of the Carbon Cycle Report\n(SOCCR): The North American Carbon Budget and Implications for the Global\nCarbon Cycle. National Oceanic and Atmospheric Administration, National Cli-\nmatic Data Center, Asheville, NC, pp. p242.\nLatapy, M.,  Magnien, C., Vecchio, N.D., 2008. Basic notions for the analysis of large\ntwo-mode networks. Soc. Netw. 30, 31\u201348.\nLemos, M.C., Finan, T., Fox, R., Nelson, D., Tucker, J., 2002. The use of seasonal climate\nforecasting in policymaking: lessons from Ceara, northern Brazil. Clim. Change,\n55.\nMartin, T., Ball, B., Karrer, B., Newman, M.E.J., 2013. Coauthorship and citation pat-\nterns in the physical review. Phys. Rev. E 88, 012814.\nMelin, G., Persson, O., 1996. Studying research collaboration using co-authorships.\nScientometrics 36, 363\u2013377.\nMichalak, A.M., Jackson, R.B., Gregg Marland Sabine, C.L., 2011. A U.S. Carbon Cycle\nScience Plan. University Corporation for Atmospheric Research, Washington, DC,\npp. 81\u201389.\nMorlacchi, P., Wilkinson, I.F., Young, L.C., 2008. Social networks of researchers in B2B\nmarketing: a case study of the IMP  Group, 1984\u20131999. J. Bus.-to-Bus. Market.\n12, 3\u201334.\nNewman, M.E.J., 2001a. Scienti\ufb01c collaboration networks. I. Network construction\nand fundamental results. Phys. Rev. E 64, 016131.\nNewman, M.E.J., 2001b. The structure of scienti\ufb01c collaboration networks. Proc. Natl.\nAcad. Sci. U. S. A. 98, 404\u2013409.\nNewman, M.E.J., 2006. Modularity and community structure in networks. Proc. Natl.\nAcad. Sci. U. S. A. 103, 8577\u20138696.\nPielke Jr., R., Sarewitz, D., Dilling, L., 2010. Science Policy Assessment and Research\non  Climate. Center for Science and Technology Policy Research, Boulder, CO, pp.\np24.\nRosenzweig, C., Horton, R., Bader, D., Brown, M.E., DeYoung, R., Fellows, M., Friedl,\nL.,  Gotwals, K., Graham, W.,  Hall, C., Higuchi, S., Hudson, C., Jedlovec, G., Kaye, J.,\nLowenstein, M., Mace, T., Milesi, C., Patzert, W.,  Stackhouse Jr., P.W., Toufectis,\nK.,  2014. Enhancing climate resilience at NASA centers: the Climate Adaptation\nScience Investigator workgroup (CASI) for science and stewardship. Bull. Am.\nMeteorol. Soc. 95, 1351\u20131363.\nSarmiento, J.L., Wofsy, S., 1999. A U.S. Carbon Cycle Science Plan. U.S. Global Change\nResearch Program, Washington, DC, pp. 69\u201379.\nShindell, D., Kuylenstierna, J.C.I., Vignati, E., Dingenen, R.V., Amann, M.,  Klimont,\nZ.,  Anenberg, S.C., Muller, N., Janssens-Maenhou, G., Raes, F., Schwartz, J.,\nFaluvegi, G., Pozzoli, L., Kupiainen, K., H\u00f6glund-Isaksson, L., Emberson, L.,\nStreets, D., Ramanathan, V., Hicks], K., Oanh, N.T.K., Milly, G.,  Williams,\nM.,  Demkine, V., Fowler, D., 2012. Simultaneously mitigating near-term cli-\nmate change and improving human health and food security. Science 335,\n183\u2013189.\nSmil, V., 1996. Cycles of Life: Civilization and the Biosphere. New York.\n Netw\nV\nW\nW\nthe NACP Committee of the U.S. Carbon Cycle Science Steering Group. UniversityM.E. Brown et al. / Social\naggi, F., Schiavinotto, T., Lawson, J.L., Chessel, A., Dodgson, J., Geymonat, M.,  Sato,\nM.,  Salas, R.E., Csik\u00e1sz-Nagy, A., 2014. A network approach to mixing delegates\nat  meetings. eLife 3, e02273.\nasserman, S., Faust, K., 1994. Social Network Analysis: Methods and Applications.\nCambridge University Press, Cambridge, UK.\nenger, E., MacDermott, R., Snyder, W.,  2002. Cultivating Communities of Practice:\nA  Guide to Managing Knowledge. Harvard Business Press, Boston.orks 44 (2016) 226\u2013237 237\nWofsy, S., Harriss, R., 2002. The North American Carbon Program (NACP). A Report ofCorporation for Atmospheric Research, Boulder, CO.\nZucker, L.G., Darby, M.R., 1996. Star scientists and institutional transformation: pat-\nterns of invention and innovation in the formation of the biotechnology industry.\nProc. Natl. Acad. Sci. U. S. A. 93, 12709\u201312716.\n", "identifiers": ["oai:casi.ntrs.nasa.gov:20150023370", null], "relations": [], "repositories": [{"id": "151", "openDoarId": 0, "name": "NASA Technical Reports Server", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "baseId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 1, "metadataAdded": 1470189271000, "metadataUpdated": 1590281909000, "depositedDate": 1528239600000}, "subjects": ["GSFC-E-DAA-TN28159", "Social Networks (ISSN 0378-8733); 44; 226-237"], "title": "Social Network and Content Analysis of the North American Carbon Program as a Scientific Community of Practice", "topics": ["Environment Pollution"], "types": [], "year": 2015, "fulltextIdentifier": "https://core.ac.uk/download/pdf/42702124.pdf", "oai": "oai:casi.ntrs.nasa.gov:20150023370", "downloadUrl": "https://core.ac.uk/download/pdf/42702124.pdf"}}
{"status": "OK", "data": {"id": "81765890", "authors": ["Henriette Golcher", "Thomas B. Brunner", "Helmut Witzigmann", "Lukas Marti", "Wolf-Otto Bechstein", "Christiane Bruns", "Henry Jungnickel", "Stefan Schreiber", "Gerhard G. Grabenbauer", "Thomas Meyer", "Susanne Merkel", "Rainer Fietkau", "Werner Hohenberger"], "citations": [], "contributors": [], "datePublished": "2014", "fullText": "1 3\nOriginal article\nReceived: 11 January 2014 / Accepted: 23 July 2014 / Published online: 25 September 2014\n\u00a9 The Author(s) 2014. This article is published with open access at Springerlink.com\nNeoadjuvant chemoradiation therapy with gemcitabine/cisplatin \nand surgery versus immediate surgery in resectable pancreatic \ncancer\nResults of the first prospective randomized phase II trial.\nHenriette Golcher \u00b7 Thomas B. Brunner \u00b7 Helmut Witzigmann \u00b7 Lukas Marti \u00b7 \nWolf-Otto Bechstein \u00b7 Christiane Bruns \u00b7 Henry Jungnickel \u00b7 Stefan Schreiber \u00b7 \nGerhard G. Grabenbauer \u00b7 Thomas Meyer \u00b7 Susanne Merkel \u00b7 Rainer Fietkau \u00b7 \nWerner Hohenberger\nStrahlenther Onkol (2015) 191:7\u201316\nDOI 10.1007/s00066-014-0737-7\nfor the treatment of rectal cancer we examined the value \nof neoadjuvant chemoradiotherapy in pancreatic cancer in \na randomized phase II trial. Radiological staging defining \nresectability was basic information prior to randomization \nAbstract\nBackground In nonrandomized trials, neoadjuvant treat-\nment was reported to prolong survival in patients with pan-\ncreatic cancer. As neoadjuvant chemoradiation is established \nElectronic supplementary material The online version \nof this Article (doi: 10.1007/s00066-014-0737-7) contains \nsupplementary material, which is available to authorized users.\nHenriette Golcher and Thomas B. Brunner contributed equally to \nthe manuscript.\nDr. H. Golcher, M.D. (\uf02a) \u00b7 Prof. T. Meyer \u00b7 Prof. S. Merkel \u00b7 \nProf. W. Hohenberger\nDepartment of Surgery, University Hospital Erlangen,\nKrankenhausstr. 12,\n91054 Erlangen, Germany\ne-mail: henriette.golcher@uk-erlangen.de\nT. B. Brunner, M.D. \u00b7 Prof. G. G. Grabenbauer \u00b7 Prof. R. Fietkau\nDepartment of Radiation Oncology, University Hospital \nErlangen,\nErlangen, Germany\nT. B. Brunner, M.D.\nDepartment of Radiation Oncology, University Hospital Freiburg,\nFreiburg, Germany\nProf. H. Witzigmann \u00b7 S. Schreiber M.D.\nDepartment of Surgery, University Hospital Leipzig,\nLeipzig, Germany\nProf. H. Witzigmann \u00b7 H. Jungnickel, M.D.\nGeneral Surgery, Hospital Dresden-Friedrichstadt,\nDresden, Germany\nL. Marti, M.D.\nGeneral Surgery, Hospital of Kanton St. Gallen,\nSt. Gallen, Switzerland\nProf. W.-O. Bechstein\nDepartment of Surgery, University Hospital Frankfurt,\nFrankfurt/Main, Germany\nProf. C. Bruns\nDepartment of Surgery \u2013 Hospital Campus Gro\u00dfhadern, \nUniversity Hospital Munich,\nMunich, Germany\nProf. C. Bruns\nDepartment of Surgery, University Hospital Magdeburg,\nMagdeburg, Germany\nProf. G. G. Grabenbauer\nDepartment of Radiation Oncology, Hospital Coburg,\nCoburg, Germany\nProf. T. Meyer\nGeneral Surgery, Hospital Ansbach,\nAnsbach, Germany\n81 3\nH. Golcher et al.\nin contrast to adjuvant therapy trials resting on pathological \nstaging.\nPatients and methods Patients with resectable adenocar-\ncinoma of the pancreatic head were randomized to primary \nsurgery (Arm A) or neoadjuvant chemoradiotherapy fol-\nlowed by surgery (Arm B), which was followed by adjuvant \nchemotherapy in both arms. A total of 254 patients were re-\nquired to detect a 4.33-month improvement in median over-\nall survival (mOS).\nResults The trial was stopped after 73 patients; 66 pa-\ntients were eligible for analysis. Twenty nine of 33 allocated \npatients received chemoradiotherapy. Radiotherapy was \ncompleted in all patients. Chemotherapy was changed in \n3 patients due to toxicity. Tumor resection was performed \nin 23 vs. 19 patients (A vs. B). The R0 resection rate was \n48\u2009% (A) and 52\u2009% (B, P\u2009=\u20090.81) and (y)pN0 was 30\u2009% (A) \nvs. 39\u2009% (B, P\u2009=\u20090.44), respectively. Postoperative complica-\ntions were comparable in both groups. mOS was 14.4 vs. \n17.4 months (A vs. B; intention-to-treat analysis; P\u2009=\u20090.96). \nAfter tumor resection, mOS was 18.9 vs. 25.0 months (A \nvs. B; P\u2009=\u20090.79).\nConclusion This worldwide first randomized trial \nfor neoadjuvant chemoradiotherapy in pancreatic cancer \nshowed that neoadjuvant chemoradiation is safe with respect \nto toxicity, perioperative morbidity, and mortality. Never-\ntheless, the trial was terminated early due to slow recruit-\ning and the results were not significant. ISRCTN78805636; \nNCT00335543.\nKeywords Adenocarcinoma \u00b7 Chemoradiation \u00b7 \nPancreas \u00b7 Surgical procedures \u00b7 Operative \u00b7 Survival\nNeoadjuvante Radiochemotherapie mit Gemcitabin/\nCisplatin gefolgt von Resektion versus prim\u00e4rer \nResektion bei resektablem Pankreaskopfkarzinom\nErgebnisse der ersten prospektiven randomisierten Phase-\nII-Studie\nZusammenfassung\nHintergrund Mehrere nichtrandomisierte Studien zeigten, \ndass eine neoadjuvante Therapie das \u00dcberleben bei Pati-\nenten mit Pankreaskarzinom verl\u00e4ngert. Beim lokal fort-\ngeschrittenen Rektumkarzinom geh\u00f6rt die neoadjuvante \nRadiochemotherapie bereits zum Therapiestandard. Analog \nwurde der Stellenwert einer Radiochemotherapie beim Pan-\nkreaskarzinom in einer randomisierten Phase-II-Studie un-\ntersucht. Das pr\u00e4therapeutische radiologische Staging war \nGrundlage dieser Studie im Gegensatz zu adjuvanten The-\nrapiestudien, die auf pathohistologischem Staging basieren.\nPatienten und Methoden Patienten mit resektablem Pan-\nkreaskopfkarzinom wurden randomisiert in prim\u00e4re Opera-\ntion (Arm A) versus neoadjuvante Radiochemotherapie ge-\nfolgt von einer Operation (Arm B). Beide Gruppen erhielten \neine adjuvante Chemotherapie. Es waren 254 Patienten er-\nforderlich, um eine Verbesserung des medianen Gesamt-\n\u00fcberlebens von 4,33 Monaten zu erfassen.\nErgebnisse Die Studie wurde wegen z\u00f6gerlicher Rekru-\ntierung nach Einschluss von 73 Patienten beendet. Insge-\nsamt konnten 66 Patienten ausgewertet werden. Die ihnen \nzugeordnete Radiochemotherapie erhielten 29 von 33 Pa-\ntienten. Alle Patienten bekamen die vollst\u00e4ndige Bestrah-\nlungstherapie. Wegen der Toxizit\u00e4t wurde bei 3 Patienten \ndie Chemotherapie reduziert. Eine Pankreastumorresektion \nerhielten 23 vs. 19 Patienten (A vs. B). Die R0-Resektions-\nrate betrug 48\u2009% (A) und 52\u2009% (B, P\u2009=\u20090,81). Bei 30\u2009% (A) \nversus 39\u2009% (B, P\u2009=\u20090,44) der resezierten Patienten waren \nkeine Lymphknotenmetastasen vorhanden. Die postoperati-\nven Komplikationen waren in beiden Gruppen vergleichbar. \nDas mediane Gesamt\u00fcberleben betrug 14,4 vs. 17,4 Monate \n(A vs. B; \u201eIntention-to-treat\u201c-Analyse; P\u2009=\u20090,96). Nach Pan-\nkreastumorresektion stieg das Gesamt\u00fcberleben auf 18,9 vs. \n25,0 Monate (A vs. B; P\u2009=\u20090,79).\nSchlussfolgerung Diese weltweit erste randomisierte \nStudie zur neoadjuvanten Radiochemotherapie beim Pank-\nreaskopfkarzinom war in Bezug auf Toxizit\u00e4t sowie peri-\noperative Morbidit\u00e4t und Mortalit\u00e4t gut durchf\u00fchrbar. Die \nErgebnisse sind jedoch nicht signifikant, da diese randomi-\nsierte Studie vorzeitig wegen mangelnder Rekrutierung be-\nendet werden musste. ISRCTN78805636; NCT00335543.\nSchl\u00fcsselw\u00f6rter Adenokarzinom \u00b7 Radiochemotherapie \u00b7 \nPankreas \u00b7 Operative chirurgische Verfahren \u00b7 \u00dcberleben\nSurvival rates of patients with pancreatic cancer have \nimproved only marginally during the last 30 years with a \n5-year survival rate of only 6\u2009% [1]. In contrast, the prog-\nnosis of patients with rectal carcinoma has improved sub-\nstantially during the same timeframe [2]. This progress was \ndue to standardizing surgical therapy [3] worldwide and by \nthe implementation of multimodal therapy [4\u20136]. Moreover, \nin rectal cancer it was found early that a clear circumferen-\ntial margin is important and that even margins below 1 mm \ncause a significant increase in the rate of local recurrence \n[7]. All these measures caused a decline in local recurrence \nfrom 50\u2009% to about 10\u2009% and an increase of 5-year survival \nrates up to more than 50\u2009% worldwide. This progress led \nto the hypothesis that the much poorer prognosis of ductal \nadenocarcinoma of the pancreas might be improved in an \nanalogous manner.\nAdjuvant therapy has been tested in a series of RCT \nphase III trials, the most important of these are ESPAC-\n1, CONKO-001, RTOG 97\u201304, and ESPAC-3 [8\u201311]. But \nthese trials were still running or results were not yet avail-\n91 3\nNeoadjuvant chemoradiation therapy with gemcitabine/cisplatin\nvessels \u2264\u2009180\u00b0 confirmed by high resolution CT [20]. All \ninclusion criteria are completely enlisted in Table S1.\nThe protocol was reviewed and funded by Deutsche \nKrebshilfe, approved \u201cG\u00fctesiegel A\u201d by Deutsche Kreb-\nsgesellschaft and approved by the ethics committees of \nthe participating institutions. All patients provided written \ninformed consent.\nTreatment\nChemoradiation\nChemoradiation and surgery were described in detail in the \ntrial protocol. Briefly patients in Arm B received 300 mg/\nm2 gemcitabine and 30 mg/m2 cisplatin on days 1, 8, 22, \nand 29 of radiotherapy. Three-dimensional treatment plan-\nning was mandatory for radiotherapy at 1.8 Gy to 55.8 Gy \n(tumor) or 50.4 Gy [regional lymph nodes, planning target \nvolume (PTV \u2264\u2009800 ml)] [21]. Dosis modifications in case \nof toxicity of chemotherapy were specified separately for \ngemcitabine and cisplatin. Criteria for patient withdrawal \nwere also defined. Six weeks after chemoradiation, a restag-\ning CT scan was scheduled.\nSurgery\nThe surgical procedure was divided into the three steps: \nexploration, tumor resection, and lymph node dissection. At \nexploration, distant metastases had to be ruled out. Local \nresectability was assessed and in case of vascular tumor \ninfiltration the decision to resect the tumor with adjacent \nvessels was completely left to the surgeon and the individ-\nual situation.\nAdjuvant chemotherapy\nIn both arms, adjuvant chemotherapy according to the \nCONKO-001 study protocol was recommended in an \namendment from 2005 [9].\nAssessment and follow-up\nResection specimens were graded and classified accord-\ning to the sixth UICC TNM system [22]. Assessment of \nresponse to neoadjuvant therapy was based on contrast-\nenhanced restaging CT scans 6 weeks after completion \nof chemoradiation. RECIST criteria were used to classify \nresponse [23].\nAcute toxicity and adverse effects were reported using \nthe NCI common toxicity criteria v2.0 and RTOG/EORTC \nrecommendations for classifying late toxic effects of radio-\ntherapy [24, 25]. Perioperative complications were graded \nby Dindo\u2019s classification [26].\nable when the present trial was planned and conducted. The \nresults of these trials led to a change in standard treatment \nrecommending adjuvant treatment with chemotherapy since \n2007 in Germany [12].\nThe concept of neoadjuvant rather than adjuvant treat-\nment in pancreatic cancer appears attractive for several rea-\nsons. First, up to 30\u2009% of the tumors staged as resectable \ncannot be resected due to undetected metastatic disease or \nunderestimated tumor contact to peripancreatic vessels [13]. \nSecond, up to 30\u2009% of the patients cannot receive adjuvant \ntherapy because of poor post-operative performance status \n[14]. Both groups of patients are not included into adjuvant \ntrials, though improving overall survival in both arms (adju-\nvant therapy vs. no adjuvant therapy) by simple patient selec-\ntion. Neoadjuvant treatment is thought to be better tolerated \nthan adjuvant treatment and avoids postsurgical morbidity \nin patients with rapidly metastasizing tumors. Nonrandom-\nized trials using the neoadjuvant approach support this \nrationale: median OS beyond 30 months for patients after \nneoadjuvant treatment and tumor resection were described \nin several retrospective data analyses [15\u201318].\nTherefore, in 1999 we started to plan this multicenter \nrandomized phase II study in patients with locally resect-\nable cancer or probably locally resectable cancer of the pan-\ncreatic head with strict imaging eligibility criteria defining \nvascular involvement. To our knowledge, this is the first \nRCT for patients with primary and borderline (meanwhile \nevolved technical term for \u201cprobably\u201d) resectable cancer \nof the pancreatic head comparing primary surgery with \nneoadjuvant treatment followed by surgery, starting with \nrandomization in 2003. Here, we report the full results of \nthis trial, which was not picked up by the majority of the \nresearch community at the time the trial was conducted. As \na consequence, the trial could not be completed and there-\nfore shows a lack of statistical significance due to the poor \nrecruiting rate. On the other hand, the reporting of nega-\ntive trials (i.e., a trial with no clear interpretable results) is \nimportant to improve future trials.\nAn extended version of this manuscript including a \ndetailed description of all methods employed in this study is \nprovided as supplementary material.\nPatients and methods\nStudy design and inclusion criteria\nPatients with resectable, histology or cytology proven \nadenocarcinoma of the pancreatic head were randomized \nbetween surgery alone (Arm A) and neoadjuvant chemora-\ndiation followed by surgery (Arm B; Fig. 1) [19].\nResectability was defined as no organ infiltration except \nthe duodenum and maximal involvement of peripancreatic \n10\n1 3\nH. Golcher et al.\nThe statistical analysis was performed on all randomly \nassigned patients with pancreatic carcinoma and sufficient \ndata. An intention-to-treat analysis calculated overall sur-\nvival for all patients from random assignment. The Kaplan\u2013\nMeier technique was used defining death by any cause as \nan event for estimating observed survival and the two-sided \nlog-rank test to measure levels of significance. Time to pro-\ngression was defined as time to first diagnosis of progres-\nsion or recurrence or death of any cause and was analyzed \nfor all patients. Comparisons between frequencies were \nperformed using the \u201cchi-square\u201d oder \u201c\u03c72\u201d2 test or, when \nPatients were followed up for at least 36 months at \n3-month intervals until 2 years and 6-month intervals \nthereafter.\nEnd points, sample size, and statistical analysis\nThe primary endpoint of this trial was overall survival. In \n2001, the study was planned in detail and the design was \nmade to detect a change in mOS from 9.15 months in Arm A \nto 13.48 months in Arm B. A power of 80\u2009% at the two-sided \nsignificance level of 5\u2009% was chosen. It was estimated that \n127 patients per arm would be required.\nFig. 1 CONSORT diagram [36] \n11\n1 3\nNeoadjuvant chemoradiation therapy with gemcitabine/cisplatin\nmentation in the case report form sometimes changes first \nimpressions. Due to this low number of patients, the power \nfor the formal statistical analysis was limited. All eligible \npatients were evaluable for survival. Patients\u2019 characteris-\ntics are listed in Table 1.\nTreatment\nIn Arm B, 29 of 33 patients received chemoradiotherapy. \nA total of 3 patients refused and 1 patient was not fit for \nchemoradiation, but all 4 patients underwent surgery. All \n29 patients who underwent chemoradiation completed radio-\ntherapy and were treated with a median of 55.8 Gy (range \n45.0\u201357.6 Gy). Three patients had changes in chemotherapy \non day 29 due to leukopenia. One patient received 5-fluo-\nrouracil/cisplatin instead of gemcitabine/cisplatin (local \ninvestigator judgment). All other patients received chemo-\ntherapy as planned. Toxicity of chemoradiation (Arm B) \nis shown in Table 2. During chemoradiotherapy and until \nappropriate, the Fisher\u2019s exact test. P-values <\u20090.05 were \nconsidered significant.\nResults\nPatients\nBetween June 2003 and December 2009, 73 patients were \nrecruited in eight university hospitals and tertiary referral \ncenters in Germany and Switzerland. In December 2009, \nenrollment was terminated because of the poor recruitment \nrate. Seven patients (4 Arm A; 3 Arm B) were deemed ineli-\ngible because of withdrawal of consent, lack of data, and \nother tumor entity (Fig. 1). Two patients had metastases at \nrandomization (n\u2009=\u20091 distant lymph nodes, n\u2009=\u20091 liver), both \nin Arm B. These patients were not excluded, as it reflects \nreal life, where reviewing of initial data at the time of docu-\nTable 1 Patients\u2019 demographic and baseline characteristics\nCharacteristics Variable Total\nn\u2009=\u200966 (%)\nSurgery alone\nn\u2009=\u200933 (%)\nCRT and surgery\nn\u2009=\u200933 (%)\nP value\nPatient variables\nAge (years) Median (range) 63.9 (33\u201376) 65.1 (46\u201373) 62.5 (33\u201376) 0.62\nGender Male 35 (53) 17 (52) 18 (55) 0.81\nFemale 31 (47) 16 (48) 15 (45)\nKPS 100 13 (20) 7 (21) 6 (18) 0.36\n90 36 (54) 15 (46) 21 (64)\n80 12 (18) 7 (21) 5 (15)\n70 5 (8) 4 (12) 1 (3)\nClinical tumor staging\nClinical T categorya cT1 2 (3) 1 (3) 1 (3) 0.79\ncT2 30 (45) 15 (45) 15 (45)\ncT3 33 (50) 17 (52) 16 (49)\ncT4 1 (2) 0 (0) 1 (3)\nClinical N categorya cN0 52 (79) 30 (91) 22 (67) 0.03\ncN1 14 (21) 3 (9) 11 (33)\nClinical M categorya cM0 64 (97) 33 (100) 31 (94) 0.49\ncM1 2 (3) 0 (0) 2 (6)\nClinical UICC stagea I 29 (44) 16 (48) 13 (39) 0.31\nII 35 (53) 17 (52) 18 (55)\nIII 0 (0) 0 (0) 0 (0)\nIV 2 (3) 0 (0) 2 (6)\nProcedures before randomization\nExplorative surgery \nbefore randomization\nExploratory surgery 36 (55) 17 (52) 19 (58) 0.62\nLaparoscopy 28 (42) 15 (46) 13 (39)\nLaparotomy 8 (12) 2 (6) 6 (18)\nNot done 30 (45) 16 (48) 14 (42)\nBiliary stent before \nrandomization\nYes 57 (86) 28 (85) 29 (88) 1.0\nNo 9 (14) 5 (15) 4 (12)\nCRT chemoradiation; KPS Karnofsky performance status\naAccording to UICC 2002\n12\n1 3\nH. Golcher et al.\nOutcome\nThe median follow-up for all living patients was 61 months \n(range 37\u201379 months). There were 29 deaths in Arm A and \n31 deaths in Arm B. At intention-to-treat analysis mOS \nbetween the two arms was not significantly different for \nall patients irrespective of resection status (Arm A, 14.4 \nmonths; Arm B 17.4 months; P\u2009=\u20090.96; Fig. 2a).\nAfter resection, mOS was 18.9 months (Arm A) versus \n25.0 months (Arm B; P\u2009=\u20090.79; intention-to-treat analy-\nsis). Time to progression measured 8.7 versus 8.4 months \n(Arm A versus Arm B; P\u2009=\u20090.95; Fig. 2b).\nPathohistological diagnosis of pancreatic adenocarci-\nnoma at biopsy was confirmed in 42 of 44 resection speci-\nmens. One distal choledochal adenocarcinoma (Arm B) and \n1 duodenal adenocarcinoma (Arm A) were excluded from all \nanalyses. R0 resections were achieved in 16 of 33 patients \nversus 17 of 33 patients (Arm A versus Arm B; P\u2009=\u20090.81), \nand mOS was 18.9 months (Arm A) versus 25.9 months \n(Arm B; P\u2009=\u20090.75; Fig. 2c). Nodal status was (y)pN0 in 10 \nof 33 patients and 13 of 33 patients in Arm A and Arm B, \nrespectively (P\u2009=\u20090.44). (y)pN0-status resulted in signifi-\ncantly longer mOS in Arm A (Fig. 2d). Four patients had \npathologically proven distant metastases resected [Arm A \nn\u2009=\u20092 (lymph node, duodenum); Arm B n\u2009=\u20092 (lymph node)]. \nPathological results for resected patients are listed in \nTable 4.\nsurgery 15 severe adverse events were reported, mostly \ncholangitis requiring a change of stent (n\u2009=\u20099). Radiological \nresponse on restaging CT scan was rarely seen (n\u2009=\u20094 partial \nresponse), whereas most patients had no change (n\u2009=\u20098) or \nprogression (n\u2009=\u200912; missing data n\u2009=\u20095).\nIn the intention-to-treat analysis, in Arm A, 23 of \n33 patients had tumor resection and 5 patients had vas-\ncular resections to achieve clinical R0 resection. Ten of \n33 patients had an explorative laparotomy. In Arm B, 19/33 \npatients had tumor resection and 4 patients had extended \nsurgery to achieve R0 resection. Ten of 33 patients had an \nexplorative laparotomy. Four patients had no surgery due to \nprogressive disease. Resection rates between the arms were \nnot different (P\u2009=\u20090.31). In Arm B, 3 of 4 patients without \nchemoradiation had tumor resection; 1 patient had liver \nmetastases at exploration.\nOf importance, patients in Arm B did not have elevated \nrates of high-grade post-operative complications (Table 3).\nOne patient died as the result of an intraoperative myocar-\ndial infarction after tumor resection (Arm A) and 1 patient \ndied due to sepsis possibly due to cholangitis after explor-\native laparotomy (Arm A). One patient had insufficiency of \nthe pancreaticojejunal anastomosis followed by multiple \norgan dysfunction (grade 4b, Arm A; none in Arm B). The \nmost severe post-operative complications after chemora-\ndiation were grade 3b (intervention under general anesthe-\nsia) due to intraabdominal abscess/fluid retention (n\u2009=\u20094) or \ninsufficiency of the choledochojejunal anastomosis (n\u2009=\u20091).\nIn Arm A, 10 of 23 patients had adjuvant chemotherapy \nand in Arm B 7 of 19 patients.\nTable 3 Postoperative complications\nDindo\u2019s grade [36]\nAll (1\u20135) 1\u20132 3a/3b 4a/4b 5\nSurgery alone \n(Arm A; n\u2009=\u200933)\n32 17 9 4 2\nAs treated (n\u2009=\u200937) Resection \n(n\u2009=\u200926)\n23 12 6 4 1\nExploration \n(n\u2009=\u200911)\n9 5 3 0 1\nCRT and surgery \n(Arm B; n\u2009=\u200933)\n22 16 6 0 0\nAs treated (n\u2009=\u200929) Resection \n(n\u2009=\u200916)\n14 9 5 0 0\nExploration \n(n\u2009=\u20099)\n8 7 1 0 0\n(no surgery \nn\u2009=\u20094)\n\u2013 \u2013 \u2013 \u2013 \u2013\nTotal (n\u2009=\u200966) 54 33 15 4 2\nResection \n(n\u2009=\u200942)\n37 21 11 4 1\nExploration \n(n\u2009=\u200920)\n17 12 4 0 1\nNo surgery \n(n\u2009=\u20094)\n\u2013 \u2013 \u2013 \u2013\nCRT chemoradiotherapy\nTable 2 Acute toxicitya of chemoradiotherapy\nParameter N Grade\n0\u20132 3 (%) 4 (%)\nLeukopenia 29 20 7 (24) 2 (7)\nThrombopenia 29 18 10 (35) 1 (3)\nAnemia 29 27 1 (3) 1 (3)\nNausea/vomiting 28 18 10 (36) \u2013\nGastrointestinal bleeding 28 28b \u2013 \u2013\nDiarrhea 29 28 1 (3) \u2013\nElevated transaminases 29 23 5 (17)c 1 (3)\nElevated bilirubin 28 26 1 (4) 1 (4)d\nElevated alkaline phosphatase 29 24 5 (17) \u2013\nInfection 29 24e 5 (17)f \u2013\naToxicity was defined according to the National Cancer Institute \nCommon Toxicity Criteria v2.0 [34]b1 of 28 patients grade 2\nc4 of 5 patients due to cholangitis\ndDue to cholangitis\neGrade1 and 2: n\u2009=\u20097 (5 patients cholangitis, 1 patient noro virus, \n1 patient localization not known)\nf4/5 cholangitis, 1 of 5 patients urinary tract infection\n13\n1 3\nNeoadjuvant chemoradiation therapy with gemcitabine/cisplatin\nnumbers this is a negative trial and no clear conclusion can \nbe drawn from underpowered data and whether there is an \nadvantage for one therapy strategy or not.\nThe following issues of a randomized controlled trial for \nresectable pancreatic cancer have to be addressed in future \ntrial protocols: working in interdisciplinary teams, pre-\ndicting resectability, definition of vascular resection aims, \ndefinition of criteria for cancelling tumor resection during \nexplorative laparotomy, and adjuvant chemotherapy. One of \nthe main problems remains how to predict resectable tumor \nstage at diagnosis as 20\u2009% of tumors without contact to the \nperipancreatic vessels at diagnosis were not resected with \nand without neoadjuvant chemoradiation (data not shown). \nClearly, the new definition of borderline resectable pan-\ncreatic cancer is helpful, but has to be evaluated in future \ntrials. A further point of discussion is the different judg-\nment between centers with reference to cancelling surgery, \nDiscussion\nThe planning of this trial was started in 1999 with activation \nin 2003 before neoadjuvant treatment had become standard \nfor other diseases (e.g., rectal carcinoma) and therefore had \nto overcome resistance by physicians and patients likewise \nagainst the idea of neoadjuvant treatment as such. Addition-\nally, competing adjuvant trials (CONKO-001 [9], ESPAC-3 \n[11]) resulted in lower participation. Another issue was \nhistological or cytological proof of disease before random-\nization. To overcome this obstacle to recruitment, the pro-\ntocol allowed randomization after histological proof during \nexplorative laparotomy. However, to our knowledge this \nremains the first planned and evaluated multicenter RCT \ncomparing immediate surgery with surgery after neoadju-\nvant therapy in resectable pancreatic cancer, defined as vas-\ncular abutment of less than 180\u00b0. But due to low patient \nFig. 2 Kaplan\u2013Meier curves (intention to treat analysis) for a overall \nsurvival, b time to progression, c overall survival after R0 resection, \nand d overall survival according to (y)pN status. CRT chemoradiation; \nO events [a, c, and d deaths or b progression of disease] observed; N \noverall number; pNx no tumor resection (d)\n \n14\n1 3\nH. Golcher et al.\nwith the highest prognostic value of margin status. There-\nfore, higher R0 resection rates after neoadjuvant treatment \nare expected to have an impact on survival [17, 18, 28\u201330].\nNeoadjuvant treatment did not show an effect in this \nstrongly underpowered trial due to underrecruitment, but \non the other hand was a suitable instrument for selecting \npatients for surgery. Patients with initially unknown distant \nmetastases might be unmasked by preoperative therapy and \nhence spared from surgery [16]. In this trial, all patients with \nneoadjuvant treatment survived at least 3 months, whereas \nafter primary surgery 3 of 34 patients died within this time-\nframe. Additionally, less severe complications were seen after \nchemoradiation therapy, probably due to induction of fibrosis, \nwhich improves the suitability of pancreatic tissue for anas-\ntomosis. A recent meta-analysis also found similar periopera-\ntive morbidity with and without neoadjuvant treatment [18].\nToxicity of chemoradiotherapy was well manageable in \nthis trial. The well-known risk of biliary stent dysfunction \nwas managed by prompt stent replacement, but was the \nmost frequent reason for severe adverse events. Hemato-\nlogic toxicity of gemcitabine-based CRT is directly related \nto radiotherapy volume and, therefore, volumes were strictly \nlimited [31\u201333]. Additionally, consequent supportive ther-\napy may explain the improved tolerability of treatment in \nthis trial compared to others avoiding loss of weight which \nwas described to be a negative prognostic factor after neo-\nadjuvant chemoradiotherapy [34]. The patients in this trial \nwere treated with 3D-conformal plans which have recently \nbeen shown to be equally effective and not significantly \nmore toxic as IMRT plans in the neoadjuvant setting [35].\nFurthermore, predicting resectability based on CT scans \nwas difficult. Thus, the CONKO-007 (NCT01827553) trial \nwill study the role of chemoradiation in borderline resect-\nable and nonresectable pancreatic cancer. A panel of highly \nexperienced surgeons will review all CT scans before reg-\nistering to the trial and at restaging and give their statement \nabout resectability. With the experience of such a trial, the \ncriteria of R0 resectability will be evaluated and adjusted. \nThen after knowing the significance of chemoradiation for \nlocally advanced and borderline resectable pancreatic can-\ncer, the next step might be a phase II trial testing the R0 \nresectability with neoadjuvant therapy.\nConclusion\nPresented in this article are the results of a RCT implicating \nthe strategy of multimodal therapy for (borderline) resectable \npancreatic cancer which was visionary at the time of plan-\nning and conduction of the trial; it was nearly 15 years ahead \nof its time before this approach was again implemented into \nprospective trials in Europe. In the meantime, the conditions \nfor conducting interdisciplinary trials have improved much \nas only one center abandoned resection of the tumor after \ndetection of distant lymph node metastasis (2 patients) or \ndid not proceed to surgery when progression (locally, dis-\ntant, clinically) at restaging after chemoradiation was seen \n(data not shown).\nThe initially mandatory laparoscopy was reclassified as \noptional due to objections of potential trial participants in \na 2004amendment. Altogether, surgical staging was con-\nducted only in 54\u2009% of all patients and should be considered \nin further trials on preoperative treatment strategies [15].\nThe closest possible comparison of this trial is with \nadjuvant treatment, especially with the CONKO-001 trial \nconducted in the same population and with an observation \narm [9, 27]. However, the fundamental difference between \nthe reported trial here and adjuvant treatment is that the \nlatter only includes patients after resection and pathologi-\ncal staging, whereas in this study 24 of 68 patients (35\u2009%) \nhad reasons preventing curative resection despite the sug-\ngested resectability at staging. Median overall survival in \nthe CONKO-001 trial was 20.2 and 22.1 months (control \nversus adjuvant gemcitabine, P\u2009=\u20090.06). This compares well \nwith the mOS of patients with resections in this trial (18 and \n25 months; Arm A versus Arm B). In CONKO-001, resec-\ntion margin status was a negative prognostic marker in the \nobservation arm (mOS 20.8 and 14.1 months R0 versus R1). \nRecent reports about the lack of prognostic significance of \nmargins might be related to frequent underreporting of R1 \nstatus because series with high R1 resection rates correlated \nTable 4 Pathological staging\nCharacteristic Variable Surgery \nalone \n(Arm A)\nCRT and \nsurgery \n(Arm B)\nN\u2009=\u200923 N\u2009=\u200919\nPathological T categorya (y)pT1 0 2\n(y)pT2 2 2\n(y)pT3 20 15\n(y)pT4 1 0\nPathological N categorya (y)pN0 10 13\n(y)pN1 13 6\nPathological M categorya (y)pM0 21 17\n(y)pM1 2 2\nPathological UICC stagea (y)pI 1 4\n(y)pII 19 13\n(y)pIII 1 0\n(y)pIV 2 2\nGrading G1 0 0\nG2 11 9\nG3 10 8\nG4 2 1\nNot specified 0 1\nResection margin R0 16 17\nR1 7 2\nCRT chemoradiotherapy\naAccording to UICC 2002\n15\n1 3\nNeoadjuvant chemoradiation therapy with gemcitabine/cisplatin\nReferences\n 1. Siegel R, Naishadham D, Jemal A (2012) Cancer statistics, 2012. \nCA Cancer J Clin 62:10\u201329\n 2. Cunningham D, Atkin W, Lenz HJ et al (2010) Colorectal cancer. \nLancet 375:1030\u20131047\n 3. Heald RJ, Moran BJ, Ryall RD et al (1998) Rectal cancer: the Bas-\ningstoke experience of total mesorectal excision, 1978\u20131997. Arch \nSurg 133:894\u2013899\n 4. Glimelius B, Pahlman L, Cervantes A (2010) Rectal cancer: \nESMO Clinical Practice guidelines for diagnosis, treatment and \nfollow-up. Ann Oncol 21 Suppl 5:v82\u201386\n 5. Sauer R, Becker H, Hohenberger W et al (2004) Preoperative ver-\nsus postoperative chemoradiotherapy for rectal cancer. N Engl J \nMed 351:1731\u20131740\n 6. Rodel C, Liersch T, Hermann RM et al (2007) Multicenter phase \nII trial of chemoradiation with oxaliplatin for rectal cancer. J Clin \nOncol 25:110\u2013117\n 7. Quirke P, Durdey P, Dixon MF et al (1986) Local recurrence of \nrectal adenocarcinoma due to inadequate surgical resection. Histo-\npathological study of lateral tumour spread and surgical excision. \nLancet 2:996\u2013999\n 8. Neoptolemos JP, Stocken DD, Friess H et al (2004) A randomized \ntrial of chemoradiotherapy and chemotherapy after resection of \npancreatic cancer. N Engl J Med 350:1200\u20131210\n 9. Oettle H, Post S, Neuhaus P et al (2007) Adjuvant chemotherapy \nwith gemcitabine vs observation in patients undergoing curative-\nintent resection of pancreatic cancer: a randomized controlled trial. \nJAMA 297:267\u2013277\n10. Regine WF, Winter KA, Abrams RA et al (2008) Fluorouracil vs \ngemcitabine chemotherapy before and after fluorouracil-based \nchemoradiation following resection of pancreatic adenocarcino-\nma: a randomized controlled trial. JAMA 299:1019\u20131026\n11. Neoptolemos JP, Stocken DD, Bassi C et al (2010) Adjuvant che-\nmotherapy with fluorouracil plus folinic acid vs gemcitabine fol-\nlowing pancreatic cancer resection: a randomized controlled trial. \nJAMA 304:1073\u20131081\n12. Adler G, Seufferlein T, Bischoff SC et al (2007) S3-Guidelines \n\u201cExocrine pancreatic cancer\u201d 2007. Z Gastroenterol 45:487\u2013523\n13. Papalezova KT, Tyler DS, Blazer DG, 3rd et al (2012) Does pre-\noperative therapy optimize outcomes in patients with resectable \npancreatic cancer? J Surg Oncol 106:111\u2013118\n14. Wayne JD, Abdalla EK, Wolff RA et al (2002) Localized adeno-\ncarcinoma of the pancreas: the rationale for preoperative chemora-\ndiation. Oncologist 7:34\u201345\n15. Snady H, Bruckner H, Cooperman A et al (2000) Survival advan-\ntage of combined chemoradiotherapy compared with resection as \nthe initial treatment of patients with regional pancreatic carcino-\nma. An outcomes trial. Cancer 89:314\u2013327\n16. Evans DB, Varadhachary GR, Crane CH et al (2008) Preoperative \ngemcitabine-based chemoradiation for patients with resectable ad-\nenocarcinoma of the pancreatic head. J Clin Oncol 26:3496\u20133502\n17. Golcher H, Brunner T, Grabenbauer G et al (2008) Preoperative \nchemoradiation in adenocarcinoma of the pancreas. A single cen-\ntre experience advocating a new treatment strategy. Eur J Surg \nOncol 34:756\u2013764\n18. Gillen S, Schuster T, Meyer Zum Buschenfelde C et al (2010) Pre-\noperative/neoadjuvant therapy in pancreatic cancer: a systematic \nreview and meta-analysis of response and resection percentages. \nPLoS Med 7:e1000267\ndue to governmental regulations and nationwide implemen-\ntation of certified cancer centers with interdisciplinary tumor \nboards. The improvement of interdisciplinary study structures \nand the lack of better therapies evolving in the meantime led \nto boycotting this trial to copying the treatment strategy of \nneoadjuvant chemoradiation with starting a nearly identical \nstudy protocol in August 2013 (NCT01900327). Prediction \nof resectability preoperatively is still an unresolved problem \nand the long-term results of treatment for pancreatic cancer \nare still frustrating even after complete tumor resection. Thus, \nat the moment we do not have a better choice but to investi-\ngate new treatment strategies suitable for as many patients \nwith pancreatic cancer as possible.\nAcknowledgments We are grateful to the patients who participated \nin this study. We thank all of the investigators who participated in this \nstudy: N. Christen, S. Ki\u00dfenk\u00f6tter, H. Lauer, V. L\u00fcck, Krankenhaus \nDresden-Friedrichstadt, Dresden; B. Adamietz, A. Agaimy, U. Baum, \nR. Croner, A. Dimmler, M. Geiger, S. Herold, R. Janka, A. Kerga\u00dfner, \nP. Klein, S. Kr\u00fcger, M. Lindenberg, W. Melzner, T. Papadopoulos, J. \nPelz, A. Schlabrakowski, U. von Linden, C. Wei\u00df, M. Zeilinger, Uni-\nversit\u00e4tsklinikum Erlangen, Erlangen; C. Gog, D. Imhoff, C. Wull-\nstein, Universit\u00e4tsklinikum Frankfurt, Frankfurt/Main; H. Bockhorn, \nA. Hildebrand, Krankenhaus Nordwest, Frankfurt/Main; J. Behrbohm, \nK. Gumpp, J. Hauss, A. Liebmann, Universit\u00e4tsklinikum Leipzig, \nLeipzig; H. J. Schlitt, C. Z\u00fchlke, Universit\u00e4tsklinikum Regensburg, \nRegensburg; T. Horbach, Krankenhaus Schwabach, Schwabach; S. \nBischofberger, P. Folie, D. Hausmann, D. K\u00f6berle, J. Lange, C. Mey-\nenberger, I. Neuweiler, G. Ries, M. Z\u00fcnd, Kantonsspital St. Gallen, \nSt. Gallen.\nFunding This work was supported by Deutsche Krebshilfe (70\u20133046-\nHo 2 to W.H.) and Verein zur F\u00f6rderung des Tumorzentrums der Uni-\nversit\u00e4t Erlangen-N\u00fcrnberg e. V.\nCompliance with ethical guidelines \nConflict of interest R. Fietkau received honoraria from Eli Lilly and \nCompany (\u201cLilly\u201d), Fresenius SE & Co. KGaA, and F. Hoffmann-La \nRoche Ltd. H. Golcher, T.B. Brunner, H. Witzigmann, L. Marti, W.-O. \nBechstein, C. Bruns, H. Jungnickel, S. Schreiber, G.G. Grabenbauer, \nT. Meyer, S. Merkel, and W. Hohenberger state that there are no con-\nflicts of interest.\nThe study has been presented in part during the following meet-\nings by oral presentation (op) or discussed poster (dp): World Con-\ngress GI-Cancer 2012, Barcelona (dp); Annual Meeting ESTRO 2012, \nBarcelona (op); 21. Chirurgisches Friedrichst\u00e4dter Symposion May \n2012, Dresden (op); Deutscher Chirurgenkongress 2012, Berlin (op); \nDeutscher Krebskongress 2012, Berlin (dp); Viszeralmedizin 2011, \nLeipzig (op); Viszeralmedizin 2010, Stuttgart (op).\nOpen Access This article is distributed under the terms of the Creative \nCommons Attribution License which permits any use, distribution, and \nreproduction in any medium, provided the original author(s) and the \nsource are credited.\n16\n1 3\nH. Golcher et al.\n28. Esposito I, Kleeff J, Bergmann F et al (2008) Most pancreatic can-\ncer resections are R1 resections. Ann Surg Oncol 15:1651\u20131660\n29. Menon KV, Gomez D, Smith AM et al (2009) Impact of margin \nstatus on survival following pancreatoduodenectomy for cancer: \nthe Leeds Pathology Protocol (LEEPP). HPB (Oxford) 11:18\u201324\n30. Zhang Y, Frampton AE, Cohen P et al (2012) Tumor infiltration in \nthe medial resection margin predicts survival after pancreaticodu-\nodenectomy for pancreatic ductal adenocarcinoma. J Gastrointest \nSurg 16:1875\u20131882\n31. Brunner TB, Merkel S, Grabenbauer GG et al (2005) Definition of \nelective lymphatic target volume in ductal carcinoma of the pan-\ncreatic head based on histopathologic analysis. Int J Radiat Oncol \nBiol Phys 62:1021\u20131029\n32. Zhu CP, Shi J, Chen YX et al (2011) Gemcitabine in the chemora-\ndiotherapy for locally advanced pancreatic cancer: a meta-analy-\nsis. Radiother Oncol 99:108\u2013113\n33. Crane CH, Wolff RA, Abbruzzese JL et al (2001) Combining gem-\ncitabine with radiation in pancreatic cancer: understanding im-\nportant variables influencing the therapeutic index. Semin Oncol \n28:25\u201333\n34. Naumann P, Habermehl D, Welzel T et al (2013) Outcome after \nneoadjuvant chemoradiation and correlation with nutritional status \nin patients with locally advanced pancreatic cancer. Strahlenther \nOnkol 189:745\u2013752\n35. Combs SE, Habermehl D, Kessel K et al (2013) Intensity modulat-\ned radiotherapy as neoadjuvant chemoradiation for the treatment \nof patients with locally advanced pancreatic cancer. Outcome anal-\nysis and comparison with a 3D-treated patient cohort. Strahlenther \nOnkol 189:738\u2013744\n36. Schulz KF, Altman DG, Moher D (2010) CONSORT 2010 state-\nment: updated guidelines for reporting parallel group randomised \ntrials. BMJ 340:c332\n19. Brunner TB, Grabenbauer GG, Meyer T et al (2007) Primary re-\nsection versus neoadjuvant chemoradiation followed by resection \nfor locally resectable or potentially resectable pancreatic carci-\nnoma without distant metastasis. A multi-centre prospectively \nrandomised phase II-study of the Interdisciplinary Working Group \nGastrointestinal Tumours (AIO, ARO, and CAO). BMC Cancer \n7:41\n20. Lu DS, Reber HA, Krasny RM et al (1997) Local staging of \npancreatic cancer: criteria for unresectability of major vessels as \nrevealed by pancreatic-phase, thin-section helical CT. Am J Roent-\ngenol 168:1439\u20131443\n21. Fokas E, Eccles C, Patel N et al (2013) Comparison of four tar-\nget volume definitions for pancreatic cancer. Guidelines for treat-\nment of the lymphatics and the primary tumor. Strahlenther Onkol \n189:407\u2013416\n22. Sobin LH, Wittekind C (eds) (2002) TNM classification of malig-\nnant tumours. 6th edn. Wiley, New York\n23. Therasse P, Arbuck SG, Eisenhauer EA et al (2000) New guide-\nlines to evaluate the response to treatment in solid tumors. Euro-\npean Organization for Research and Treatment of Cancer, National \nCancer Institute of the United States, National Cancer Institute of \nCanada. J Natl Cancer Inst 92:205\u2013216\n24. National Cancer Institute Common Toxicity Criteria, Version 2.0: \nCancer Therapy Evaluation Program. 1998 http://ctep.cancer.gov/\nprotocolDevelopment/electronic_applications/docs/ctcv20_4-30-\n992.pdf. Accessed 1 Sept 2014\n25. Cox JD, Stetz J, Pajak TF (1995) Toxicity criteria of the Radiation \nTherapy Oncology Group (RTOG) and the European Organiza-\ntion for Research and Treatment of Cancer (EORTC). Int J Radiat \nOncol Biol Phys 31:1341\u20131346\n26. Dindo D, Demartines N, Clavien PA (2004) Classification of surgi-\ncal complications: a new proposal with evaluation in a cohort of \n6336 patients and results of a survey. Ann Surg 240:205\u2013213\n27. Oettle H, Neuhaus P, Hochhaus A et al (2013) Adjuvant chemo-\ntherapy with gemcitabine and long-term outcomes among patients \nwith resected pancreatic cancer: the CONKO-001 randomized \ntrial. JAMA 310:1473\u20131481\n", "identifiers": ["10.1007/s00066-014-0737-7"], "publisher": "Springer Nature", "relations": ["http://dx.doi.org/10.1007/s00066-014-0737-7"], "repositories": [{"id": "2612", "openDoarId": 0, "name": "Springer - Publisher Connector", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 1, "metadataAdded": 1493837349000, "metadataUpdated": 1494858357000, "depositedDate": null}, "subjects": ["journal-article"], "title": "Neoadjuvant chemoradiation therapy with gemcitabine/cisplatin and surgery versus immediate surgery in resectable pancreatic cancer", "topics": [], "types": [], "year": 2014, "fulltextIdentifier": "https://core.ac.uk/download/pdf/81765890.pdf", "doi": "10.1007/s00066-014-0737-7", "downloadUrl": "https://core.ac.uk/download/pdf/81765890.pdf"}}
{"status": "OK", "data": {"id": "82864389", "authors": ["David Papo"], "citations": [], "contributors": [], "datePublished": "2013", "fullText": "HYPOTHESIS AND THEORY ARTICLE\npublished: 19 April 2013\ndoi: 10.3389/fphys.2013.00086\nTime scales in cognitive neuroscience\nDavid Papo*\nCenter for Biomedical Technology, Universidad Polit\u00e9cnica de Madrid, Madrid, Spain\nEdited by:\nPaolo Allegrini, Consiglio Nazionale\ndelle Ricerche, Italy\nReviewed by:\nPaolo Allegrini, Consiglio Nazionale\ndelle Ricerche, Italy\nPaolo Paradisi, Consiglio Nazionale\ndelle Ricerche, Italy\nPaolo Grigolini, University of North\nTexas, USA\n*Correspondence:\nDavid Papo, Center for Biomedical\nTechnology, Universidad Polit\u00e9cnica\nde Madrid, Campus Montegancedo,\n28223 Pozuelo de Alarc\u00f3n, Madrid,\nSpain.\ne-mail: papodav@gmail.com\nCognitive neuroscience boils down to describing the ways in which cognitive function\nresults from brain activity. In turn, brain activity shows complex fluctuations, with structure\nat many spatio-temporal scales. Exactly how cognitive function inherits the physical\ndimensions of neural activity, though, is highly non-trivial, and so are generally the\ncorresponding dimensions of cognitive phenomena. As for any physical phenomenon,\nwhen studying cognitive function, the first conceptual step should be that of establishing\nits dimensions. Here, we provide a systematic presentation of the temporal aspects\nof task-related brain activity, from the smallest scale of the brain imaging technique\u2019s\nresolution, to the observation time of a given experiment, through the characteristic time\nscales of the process under study. We first review some standard assumptions on the\ntemporal scales of cognitive function. In spite of their general use, these assumptions hold\ntrue to a high degree of approximation for many cognitive (viz. fast perceptual) processes,\nbut have their limitations for other ones (e.g., thinking or reasoning). We define in a\nrigorous way the temporal quantifiers of cognition at all scales, and illustrate how they\nqualitatively vary as a function of the properties of the cognitive process under study. We\npropose that each phenomenon should be approached with its own set of theoretical,\nmethodological and analytical tools. In particular, we show that when treating cognitive\nprocesses such as thinking or reasoning, complex properties of ongoing brain activity,\nwhich can be drastically simplified when considering fast (e.g., perceptual) processes,\nstart playing a major role, and not only characterize the temporal properties of task-related\nbrain activity, but also determine the conditions for proper observation of the phenomena.\nFinally, some implications on the design of experiments, data analyses, and the choice of\nrecording parameters are discussed.\nKeywords: cognitive neuroscience, characteristic time, relaxation time, observation time, non-Gaussianity, scaling,\nfluctuation-dissipation theorem, non-self-averaging\nINTRODUCTION\nWhat\u2019s the temporal dimension of cognition? Hardly ever is this\nquestion addressed by cognitive neuroscientists, possibly because\nit appears as either trivial or meaningless.\nCognitive psychologists have long recognized that behavior\nmay often present changes over many time scales (Newell et al.,\n2001). However, experimental neuroscientists generally make\nmore or less covert assumptions on the dimensions of the phe-\nnomena they investigate, some of which become visible when\nconsidering the structure of experiments devised to investigate\ncognitive function. Experiments are typically divided into trials\nof essentially equal duration. On one hand, duration is implic-\nitly equated to the characteristic length of cognitive phenomena,\nwith the nested assumptions that such a characteristic length\ndoes exist, that it is unique (and therefore stationary modulo\nlearning-related trends), and known a priori. On the other hand,\nduration is equated to the sample mean duration, fluctuations\naround which are supposed to average out for a sufficiently high\nnumber of trials. This set of intertwined covert assumptions also\nensures that the phenomenon under investigation is appropriately\nobserved. Furthermore, since each trial is supposed to sample\nthe same aspects of a unique underlying state space, the mere\nnumber of trials guarantees that the phenomenon is appropriately\nobserved.\nThese assumptions hold true to a high degree of approxi-\nmation for many cognitive (viz. perceptual) processes, but have\ntheir limitations for other ones, e.g., thinking or reasoning. For\ninstance, a neuroscientist studying face perception knows the\napproximate duration of her phenomenon (which indeed has a\nmeaningful mean duration) can estimate the number of trials\nneeded to explore the expected cognitive and associated neural\nrange of the phenomenon, and yielding reasonable signal-to-\nnoise ratios. On the contrary, a neuroscientist studying subjects\nattempting to solve a complicated mathematical problem can-\nnot know what sort of durations (nor, in general, what average\nduration) experiments may yield, what task-related time scales,\nor quasi-periodicities brain activity may express, and as a result,\nwhether the phenomenon is appropriately observed.\nIn spite of this unique set of challenging properties, studies\nof this type of phenomena often apply the same experimental\ndesigns, techniques of data analysis, and models of brain activ-\nity as those of processes with very different characteristics, e.g.,\nperceptual processes. This, however, supposes some restrictive\nassumptions, including smoothing response times, to achieve\nwww.frontiersin.org April 2013 | Volume 4 | Article 86 | 1\nPapo Time scales in cognitive neuroscience\ntrials of even duration and extract time averages (which are dif-\nficult to interpret, given inherent non-stationarities), or using\nvery specific and constrained forms of reasoning (e.g., Goel et al.,\n1997, 1998; Osherson et al., 1998; Parsons and Osherson, 2001;\nBonnefond and Van der Henst, 2009), or limiting the analysis to\na very short time window (<1 s), e.g., in the temporal vicinity of\nan answer to a given problem (e.g., Jung-Beeman et al., 2004; Mai\net al., 2004; Lang et al., 2006; Qiu et al., 2008; Pijnacker et al.,\n2011).\nThat establishing the temporal scales of cognitive phenomena\nis both meaningful and fundamental, though, should also become\nmanifest as soon as cognitive function is understood to originate\nfrom brain activity, and is quantitatively characterized in terms of\nthe brain properties associated with the execution of given cog-\nnitive tasks. Perhaps not so intuitively, in fact, once cognition is\nmeasured in units of brain activity, the cognitive processes under\ninvestigation are de facto treated as physical phenomena. This has\ntwo main implications. Firstly, cognition is endowed with the\nphysical dimensions of the associated brain activity. Secondly, the\nstudy of cognitive phenomena becomes subject to the same prin-\nciples and experimental constraints presiding the experimental\nstudy of ordinary physical phenomena: dimensions need to be\nevaluated, through observations of limited time resolution and\nperformed over a finite, often relatively brief, time.\nIn the following, we systematically review time scale-related\naspects of cognition ranging from the resolution of the neu-\nroimaging technique, to observation time of experiments, and\nincluding the characteristic times of the processes under inves-\ntigation. We discuss some conceptual and experimental impli-\ncations which should be taken into account when designing\nexperiments in cognitive neuroscience.\nTEMPORAL RESOLUTION\nCortical activity helps psychologists refining the space used to\ndescribe cognitive function, by adding not only spatial dimen-\nsions, but also a much finer graining of the temporal axis, and\ncan therefore help describing cognitive processes even when there\nis no behaviorally observable event. Thus, it is only logical that\nexperiments should strive to maximize the resolution of the\ninstrument chosen to record brain activity.\nTemporal resolution is typically sought for so as to optimize\nthe chances of detecting, and accurately localizing, the onset of\nevents, often lacking a corresponding behaviorally observable\none. These landmarks may for instance identify the boundaries\nbetween microstates, i.e., quasi-stationary segments of duration\nl \u2264 150ms (Koenig et al., 2002) where the activity field remains\nstable, punctuated by abrupt changes to new configurations\n(Fingelkurts and Fingelkurts, 2004; Kaplan et al., 2005). These\nstable segments were proposed to be \u201catoms of thought\u201d (Koukou\nand Lehmann, 1987), supposedly corresponding to different\ninformation processing steps.\nHow well these transitions can be detected does not solely\nhinge on the instrument used to record brain activity. The instru-\nment\u2019s sampling rate sets the lower bound to detectable scales,\nvia the corresponding Nyquist frequency, i.e., half the sam-\npling frequency of a discrete signal processing system. Ultimately,\nthough, the effective temporal resolution of a given experiment\nis determined by the analyses carried out on the data and, more\nspecifically, by the size of the smallest temporal fluctuations that\nthese analyses allow resolving. So, for instance, standard trial-\naveraging in time-locked evoked potential extraction does not\npossess millisecond precision, even for a sampling rate of that\norder, and typically not even that corresponding to frequencies\nlower than the Nyquist frequency.\nThe extent to which the boundaries of quasi-stationary seg-\nments per se can identify a given cognitive process depends on the\nproperties of the phenomenon to be studied.\nEvent-related perceptual processes are typicallymodeled as fast\nphenomena of known characteristic duration L \u223c 1 s with fluctu-\nations in duration negligible with respect to the mean character-\nistic duration (\u03b4L/\u3008L\u3009 \u0006 1). For these processes, quasi-stationary\nsegments can be used to partition a given epoch into discrete tem-\nporal units, with identifiable temporal landmarks (Kaplan et al.,\n2005). One can then hope to map the identified landmarks onto\nidentifiable cognitive steps, whose temporal location averaging\nacross repetitions of the same task would then help refining.\nA clear partitioning into cognitive steps is more complex for\nslow processes, such as some forms of conceptual learning, and\nessentially stationary phenomena, such as memory processes,\nwhich often come in episodes with durations various orders of\nmagnitude larger than perceptual processes. A static or compar-\native statics approach is typically used (e.g., Karni et al., 1995),\nwherein statical measures of brain activity at different learning\nsteps are compared, ultimately drastically flattening the inherently\ndynamical aspect of cognition.\nPartitioning into cognitive steps processes such as reasoning\nand thinking is even more difficult, when equipped solely with\ntemporal resolution. Not only do these processes lack a trivial\ntemporal duration, but they also come in long episodes (L \u223c\nminutes or more), where a large number of cognitive processes\ninteract in a wide range of temporal scales, with unconstrained\ninner structure and no behavioral correlates most of the time.\nThe associated brain activity is not event-related in the classi-\ncal sense of the term, as its duration and profile are independent\nof the physical and statistical characteristics of the stimulus elic-\niting a given reasoning episode. These processes are also highly\nnon-stationary, rendering the meaning of static descriptions (viz.\ntime averages over long time windows, spanning entire reasoning\nepisodes) problematic.\nThus, for this sort of activity, both the temporal scales of\nepisodes and a partition into cognitive steps become highly\nnon-trivial. The former needs to be evaluated for the underly-\ning physical phenomenon to be properly described, while the\nmeaning of each particular quasi-stationary segment in isolation\nbecomes unclear. As a consequence, extracting meaningful infor-\nmation from a set of different trials of these processes becomes\narduous.\nIn summary, we illustrated the concept of temporal resolution,\nand showed that, rather than by the neuroimaging technique\u2019s\nsampling rate, the effective resolution of a given experiment\nis ultimately determined by the method used to analyse the\ndata. Temporal resolution is often thought to be necessary and\nsufficient to identify cognitively relevant quasi-stationary seg-\nments of brain activity. This statement holds to a high degree of\nFrontiers in Physiology | Fractal Physiology April 2013 | Volume 4 | Article 86 | 2\nPapo Time scales in cognitive neuroscience\napproximation for fast processes with a typical overall duration,\nbut not for a vast class of cognitive phenomena lacking a typical\ntemporal duration. For this latter class, additional quantifiers of\ntemporal scales need be taken into account.\nCHARACTERISTIC TIME(S)\nThe brain is a non-equilibrium system, exhibiting fluctuations in\na wide range of temporal (and spatial) scales, most of which are\nnot observable from behavior alone. Cognition should inherit,\nthough possibly in rather complex ways, a subset at least of the\nscales of brain activity, which become distinctive properties of\ncognitive processes.\nFor short duration perceptual processes, the brain can bemod-\neled as an excitable medium, and perception itself is understood\nas a process whereby the brain relaxes to equilibrium, follow-\ning exposure to discrete exogenous perturbations or stimuli. For\nperturbations smaller than a characteristic threshold, a stable sta-\ntionary solution exists, where the brain is assumed to be quiet\non average. Perturbations above the threshold make a dynami-\ncal cycle observable, after which the system goes back to its initial\nresting state, with a characteristic relaxation time of the order of\nthe average duration (\u03c4relax \u2248 \u3008L\u3009).\nOn much longer time scales, cognitive processes such as\nlearning may sometimes still be conceptualized as relaxational\nprocesses, but the cycle generally takes a non-trivial form. On\nthe other hand, processes such as long-term memory, reason-\ning and thinking, are respectively stationary and non-stationary\nnon-relaxational processes, and the corresponding neural activ-\nity is dominated not by a forcing stimulus, but by endogenous\nfluctuations (Buice and Cowan, 2009).\nThese processes have no typical duration L, and are generally\ntemporally extended, i.e., their duration is much larger than the\nsize of a typical fluctuation (L \b \u03c4fluct). This supposes examining\nthe time scales between the temporal resolution and the largest\nobservable one.\nA straightforward way to show how the time scales of cognitive\ntasks arise is to model the associated brain activity as a dissipa-\ntive dynamic process X(t), describing the motion of a diffusing\nmacroscopic particle in a complex configuration space. The par-\nticle is subject to a viscous friction, changing with a time scale \u03c4m,\nand to an additive random force \u03b7(t) drive having a characteristic\ntime \u03c4\u03b7 (Hsu and Hsu, 2009). The relationship between \u03c4m and\n\u03c4\u03b7 determines how microscopic fluctuations renormalize to give\nrise to observable macroscopic dynamical and statistical prop-\nerties, and ultimately defines temporal scales of the underlying\nphenomenon.\nCognitive neuroscience studies typically implicitly adopt a\nMarkovian approximation, where the noise has fast vanishing\nGaussian \u03b4-correlated fluctuations, and \u03c4\u03b7 \u0006 \u03c4m. The corre-\nsponding process hops without memory from a given config-\nuration to some other, and in the long time limit (t \b \u03c4m),\nthe temporal autocorrelation of macroscopic velocity fluctuations\nCii(\u03c4) = \u3008vi(x(\u03c4))vi(x(0))\u3009 (where \u3008\u00b7\u3009 designates an average over\nall times \u03c4) decays as \u223c exp(\u2212t/\u03c4m).\n\u03c4m is unique, and is called characteristic time. The Central\nlimit theorem (CLT) ensures that, in the long time limit, the\nprobability density function of the particle\u2019s velocity converges\nto a Gaussian; the mean-square distance (MSD) travelled by the\nparticle increases linearly in time:\n\u2329|x(t) \u2212 x(0)|2\u232a \u223c t2\u03bd with \u03bd = 1\n2\n(1)\na characteristic of normal diffusion.\nNote that the relationship between\nMSD and autocorrelation function is given by:\n\u2329|x(t) \u2212 x(0)|2\u232a =\n\u222b t\n0\n\u222b t\n0\n\u2329\n\u03c5i(xt1)\u03c5i(xt2)\n\u232a\ndt1dt2\n\u2248 2t\n\u222b t\n0\nCii(\u03c4)d\u03c4 (2)\nThe characteristic time \u03c4m, the corresponding correlation time\n\u03c4C , i.e., the integral of the autocorrelation function over its sup-\nport, and the correlation length, i.e., the value \u03be such that C(t =\n\u03be) = 0, (with \u03c4C =\n\u222b \u03be\n0 C(t)dt \u2264 Cmax\u03be), endow a given process\nwith a temporal scale, independently of its having a characteristic\nduration, and naturally segment it into separable parts, although\nin a statistical way.\nThese scales can be seen as a measure of how fast the system\nlooses memory of its own past. If the system forgets its own past\ninfinitely fast, this time window is infinitely short, and a moment\nis a point with no dimension. If the system has memory, at each\ngiven time, there is an active time window spanning its memory\nlength. This active window can be seen as an estimate of what the\nsystem considers as a moment.\nFurthermore, under the Markovian approximation, sponta-\nneous activity is an unstructured attractor state, and perturba-\ntions (e.g., sensory stimuli) produce a temporally local effect, so\nthat what is observed is a temporally disordered (L \b \u03be) perfectly\nelastic fast relaxation process.\nIn general, the Markovian approximation does not hold when\nlong time scales of brain activity are considered. The friction\nforce becomes retarded or frequency-dependent, and the statis-\ntics of the driving noise is neither Gaussian (Freyer et al., 2009)\nnor \u03b4-correlated. The dynamics incorporates extended memory\nand temporal non-locality. Ordinary exponential relaxation is\nreplaced by complex, viz. Mittag-Leffler (Bianco et al., 2007a)\nscaling, with stretched exponential relaxation \u223c exp [\u2212(t/\u03c4)\u03b1] at\nmicroscopic scales (t < \u03c4), and asymptotical convergence to an\ninverse power law \u223c [\u2212(t/\u03c4)\u03b1] for t \b \u03c4, 0 < \u03b1 < 1 (Novikov\net al., 1997; Linkenkaer-Hansen et al., 2001; Gong et al., 2002;\nFreeman et al., 2003; Stam and de Bruin, 2004; Buiatti et al., 2007;\nvan de Ville et al., 2010; Freyer et al., 2012). For \u03b1 \u2264 1, the correla-\ntion time \u03c4C and the corresponding correlation length \u03be diverge.\nNote that correlation time \u03c4C and characteristic time \u03c4m, which\ncoincide for exponential functional forms of the autocorrelation\nfunction, do not coincide for power-law ones. The CLT is vio-\nlated and scale separation is lost, so that microscopic stochasticity\nbecomes detectable at macroscopic scales (Grigolini et al., 1999).\nActivity undergoes anomalous diffusion with the MSD travelled\nby the particle no longer a linear function of time:\n\u2329|x(t) \u2212 x(0)|2\u232a \u223c t2\u03bd with \u03bd \t= 1\n2\n(3)\nwww.frontiersin.org April 2013 | Volume 4 | Article 86 | 3\nPapo Time scales in cognitive neuroscience\nThe logarithmic slope of the time-dependent MSD provides an\nindication of the type of motion: subdiffusion and superdiffusion\ncorrespond to 0 \u2264 \u03bd < 12 , and 12 \u2264 \u03bd < 1 respectively.\nOne fundamental implication for the temporal scales of cogni-\ntion is that activity lacks a characteristic scale, and is described by\nsome relationship defined on a set {\u03c4i}, rather than by \u03c4i-s them-\nselves. R can take the form of a scaling exponent \u03bd relating some\nmeasure of brain activity to its frequency. Exact self-similarity\nimplies that fluctuations at a given scale are similar to fluctua-\ntions at all other scales, so that the probability distribution that\nthe particle has travelled a distance x at time t is given by:\nP(x, t) = t\u2212\u03bdF (x/t\u03bd) (4)\nwhere the scaling exponent \u03bd is unique for all scales, and F is a\nscaling function. When dealing with real data, self-similarity is not\nexact but has a statistical sense, and lower and upper cut-offs nec-\nessarily appear. Without the scaling range, the scaling properties\nmay be distinctly different from those of Equations 3 and 4 (Latka\net al., 2004; Buiatti et al., 2007).\nWhen they exist, the moments of self-similar processes behave\nas power laws with respect to time (Abry, 2003), and the time\ndependence of the qth moment of the displacement is defined by:\n\u2329|x(t) \u2212 x(0)|q\u232a \u223c tq\u03bd\u2200q > 0 (5)\nSmall and large values of q sample the central part and the tails of\nthe distribution, corresponding to small and exceptionally large\ndisplacements respectively. In real data, the scaling exponent need\nnot be unique for all moments of the distribution, and entire\nspectrum of scaling exponents may exist. For instance, in the pres-\nence of multiplicative interactions and interdependencies among\ntemporal scales (Ihlen and Vereijken, 2010), the moments q of the\ntravelled distance may take the form:\n\u2329|x(t) \u2212 x(0)|q\u232a \u223c tq\u03bd(q)\u03bd(q) \t= const (6)\nand the underlying diffusion process is weakly self-similar\n(Ferrari et al., 2001) or strongly anomalous (Castiglione et al.,\n1999). The motion representing the size of neural events, exhibits\nsteps of all sizes, from local confined motion to extremely long\njumps. Furthermore, the scaling need not be of a power-law form\n(Chainais et al., 2005). The probability density P (x, t) is no longer\nspecified by a unique scaling exponent but by a spectrum of scal-\ning exponents. One way to represent the distribution P\u03bb(x, t) at\nany given scale \u03bb within the scaling range is as the convolution\nP\u03bb(x, t) = G(q,R) \u2297 P\u0003(x, t) (7)\nof the distribution P\u0003(x, t) at the highest scale and G(q,R),\nthe probability distribution of the relation across temporal scales\n(Castaing et al., 1990). More generally, can be seen as a random\nprocess in time scale \u03c4 relating, e.g., the shape of the velocity incre-\nment distribution of the dynamic process X(t) at one time-scale\nto that at other time-scales (Friedrich et al., 2000; Bacry et al.,\n2001), or the behavior through scales of the relationship between\ndifferent models in a renormalization group approach (Longo\net al., 2012). For instance, methods have been developed to derive\nan explicit equation for the changes in the variable X(t) over a\nseries of nested time-scales of decreasing durations from the data\n(Friedrich et al., 2011).\nThe spatially-extended nature of the brain gives rise to addi-\ntional time scales. Spatial extension introduces not only dynami-\ncal heterogeneity, i.e., a spatial distribution of time scales, but also\ntime scales emerging from cooperative or competitive phenom-\nena, which typically display regimes and unfold at scales different\nfrom those of spatially local components (Bianco et al., 2007b;\nAllegrini et al., 2011). Space-induced time scales can also stem\nfrom temporal non-localities induced by transmission delays, and\nfrom transients, whose duration typically scales with the size of\nthe system (T\u00e9l and Lai, 2008).\nThe presence of correlated driving noise and cross-scale\nrelationships produces temporally ordered structures (L \u223c \u03be),\nand dilates a moment, from the essentially pointwise extension\ninduced by \u03b4-correlated noise to a temporally non-local one, so\nthat in general the meaning of activity at a given time point is not\neasily divorced from activity occurring within the scaling range.\nThe size of a moment need not be stationary in time. In fact,\nthe breakdown of exact self-similarity for non-constant values of\nq in Equation 6 implies that there no longer is a unique dilation\nfactor, but a collection of factors with a given distribution. Insofar\nas temporal scale invariance is a continuous symmetry linking a\ntranslation in time to a translation in space, brain activity can\nbe seen as stemming from a system moving at constant velocity,\ngiven by the scaling exponent (Sornette, 2004). The break-down\nof scale invariance (Ciuciu et al., 2011, 2012; Zilber et al., 2012) is\ntantamount to velocity changes.\nOne possible way to represent this is to assume that brain\ndynamics makes steps between given locations of its phase space,\nin which it dwells for a certain time, and that times between two\nconsecutive steps are characterized by a given waiting-time distri-\nbution (WTD). The steps mark a sort of internal operational time,\nwhich can grow sub- or superlinearly with the physical time. While\nin the absence of multiplicative interactions, the function G of\nEquation 7 collapses into a single point and operational and phys-\nical time coincide,multiplicative interactions across scales bias the\ndistribution of waiting times between successive steps. As a result,\nlocal probability densities are time-dependent and intermittent,\nwith laminar periods interrupted by bursts of large and irregular\nactivity of different sizes (Gong et al., 2007; Freyer et al., 2012).\nIntermittency allows reducing brain phenomena evolving in\ncontinuous time to event-based point processes and therefore\ndefining the boundaries of individual microstates in a temporally\nlocal way. Point processes generate events distinct from back-\nground activity, their value being non-zero only when an \u201cevent\u201d\noccurs, and are thus particularly suitable to describe and model\nsystem dynamics characterized by the occurrence of events. The\ntemporal scales of cognitive processes are expressed in terms of\nthe scaling properties of the density distribution function f (l) of\nmicrostate durations l (Allegrini et al., 2010). In fine, if task-\nrelated brain activity is considered as a particle evolving in a\ncomplex metastable state, the global temporal dimension is of\nthe order of the Kramers escape time from or mean first passage\ntimes in a potential, i.e., the average time elapsed until a stochastic\nFrontiers in Physiology | Fractal Physiology April 2013 | Volume 4 | Article 86 | 4\nPapo Time scales in cognitive neuroscience\nprocess starting at a given point leaves a prescribed domain for the\nfirst time, and typically much longer than the dynamic time scales\ncharacterizing states of local stability (H\u00e4nggi et al., 1990).\nWhile complex scaling appears as a generic property of sponta-\nneous brain activity, the relevance of the temporal scales and their\nstructure to cognition can be appreciated from two interrelated\nperspectives.\nOn the one hand, the stimulus-related dynamic range of a\nneural network can be related to its spontaneous activity (Shew\net al., 2009). This is a consequence of the fluctuation-dissipation\ntheorem (FDT), which establishes a general relationship between\nthe (equilibrium) internal autocorrelation C(t) of fluctuations of\nsome observable of the system in the absence of the disturbance\nand the (non-equilibrium) response R(t) of a system to small\nexternal perturbations (Kubo, 1966). Suitably modified versions\nof the FDT also hold for out-of-equilibrium systems (Cugliandolo\net al., 1997; Crisanti and Ritort, 2003; Pottier and Mauger, 2004;\nAllegrini et al., 2007; Aquino et al., 2007).\nOn the other hand, the FDT allows conceptualizing cogni-\ntive processes as fields acting upon brain activity (Papo, 2013),\nwhose response and associated time scales can be studied using\nresponse theory (Kubo, 1966). This conceptualization is particu-\nlarly intuitive for perceptual stimulus-dependent brain activity.\nFor a generic stimulus-independent process, the effect of cog-\nnition can be identified with the noise driving the endogenous\ndynamics, and asymptotic regimes depend on the ratio between\ncorrelation time and intensity of the driving noise, as well as on\nits statistics (H\u00e4nggi and Jung, 1995).\nThe response to a perturbation depends on the relative time\nscale of the perturbation and the characteristic times of the\nsystem\u2019s dynamics. Suppose, for instance, that the brain is a har-\nmonic oscillator of bare frequency \u03c90, and that a given stimulus\nan applied perturbation asin(\u03c9t). Then, additive perturbations\nwould average out over a vanishingly small time interval for \u03c9 \b\n\u03c90, would follow the perturbation in a frequency-specific way for\n\u03c9 \u0006 \u03c90, and would non-generically show resonance for\u03c9 \u2248 \u03c90.\nHowever, both ecologic stimuli and brain activity are generally\nmore complex than this simple picture. Accumulating evidence\nshows that the brain generically responds to changing external\nfields with a series of avalanches spanning a broad range of scales\n(Fairhall et al., 2001; Gilboa et al., 2005; Drew and Abbott, 2006;\nLundstrom et al., 2008). Fluctuations in ongoing brain activity\ndisplay weak ergodicity breaking (Bianco et al., 2007a; West et al.,\n2008), a condition wherein the state space is still entirely accessi-\nble to the system, but where the Carlson depth \u03c4Cd, i.e., the time\nthat the system requires to visit the whole state space, may be\nvery long, as the system dwells for very long times in some of its\naccessible microstates.\nCognition may affect cross-scale relations R in a number of\ndifferent ways: bymodulating temporal correlations (Linkenkaer-\nHansen et al., 2004; Buice and Cowan, 2009), thereby inducing\nphase transitions in mean first-passage time regimes (Carretero-\nCampos et al., 2012); by inducing transitions between non-scaling\nand scaling regimes or between different asymptotic scaling\nregimes (Linkenkaer-Hansen et al., 2004; Buiatti et al., 2007;\nHe et al., 2010; Ciuciu et al., 2011, 2012; Zilber et al., 2012),\nwhich reflect the cooperative nature of brain activity (Bianco\net al., 2007b), and may correspond to dynamical transitions in\nthe system\u2019s behavior (Burov and Barkai, 2008); by modulating\nthe degree of non-ergodicity, corresponding to different ways of\nvisiting the state space, a conjecture that has not yet received\nempirical support. External perturbations may particularly influ-\nence the scale at which the WTD is expected to show scaling,\nwhile endogenous activity likely affects the scaling properties of\nthe events\u2019 waiting time distribution (Aquino et al., 2011). More\ngenerally, cognitive function should acquire the temporal scales\nof the order parameter used to describe it.\nAll these transitions should ultimately result in operational\ntime and corresponding time scale modulations. For example,\nplausible testable hypotheses are that cognition maymodulate the\nMarkov time \u03c4M , i.e., the minimum length interval over which\nthe data can be considered as a Markov process, even when the\nprocess itself is not Markovian, or the time scale of the transi-\ntion from microscopic to macroscopic dynamics (Aquino et al.,\n2007), or the time scales at which fluctuations start converging to\na Gaussian distribution (Mantegna and Stanley, 1994; Manshour\net al., 2009).\nIn summary, the perceptual neuroscientist deals with an\nexcitable medium, and what is typically studied is a fast lin-\near response to stimulations. It can be treated as an essen-\ntially exponential world, with very short memory and a unique\ncharacteristic time scale, without losing too much information.\nThe reasoning neuroscientist, on the other hand, faces long\ntime scales of ongoing brain activity, where properties such as\nnon-Gaussianity, non-ergodicity, scale-freeness and long-term\nmemory play a prominent role which cannot be overlooked. A\nfundamental physical theorem, the FDT, explains why these rest-\ning brain properties can be inherited by activity associated with\nthe execution of cognitive task. As a consequence, not only are\nthe time scales of task-related brain activity no longer unique but\nthey also possess complex relationships among them. The time\nscales and their relationships can be quantified from real data in\nterms of scaling exponents (e.g., of autocorrelation or waiting time\ndistribution functions), scaling functions and the typical shape\nof fluctuations, or even of explicit evolution equations for the\nposition or velocity and for the probability distribution of fluc-\ntuations (Fokker\u2013Planck equation). Ways in which time scales of\ntask-independent brain activity can be modulated by task-related\nbrain activity have been suggested.\nOBSERVATION TIME\nCognitive neuroscientists observe phenomena through experi-\nments in which subjects typically carry out a given task a large\nnumber of times, which are meant to adequately sample (a mean-\ningful portion of) the phase space of task-related brain activity.\nFatigue and other factors limit the typical observation time within\na given experimental session to the order of tens of minutes,\nwhile experiments (e.g., those assessing sleep-related learning)\nmay consist of more than one session, each of which comprising\na given number of trials.\nProper observation of a given process requires that the obser-\nvation time \u03c4Obs (in this case the entire experiment) be much\nlarger than any scale in the system. Thus, \u03c4Obs should be pro-\nfoundly different for processes with dissimilar time scales. A\nwww.frontiersin.org April 2013 | Volume 4 | Article 86 | 5\nPapo Time scales in cognitive neuroscience\nmeasure of observability is given by the Deborah number De :=\n\u03c4rel/\u03c4Obs, i.e., the ratio between the characteristic time of the vari-\nable used to describe brain activity, and the length of the time\nseries made available by an experiment (Reiner, 1964). The sys-\ntem is ergodic for small values of De but weakly non-ergodic for\nDe \u2192 \u221e (Rebenshtok and Barkai, 2007).\nVarious factors at different temporal scales, both within and\nacross trials, may complicate the observability of cognitive phe-\nnomena.\nThe observation time should ideally be much larger than the\ntime needed to visit the phase space of task-related brain activ-\nity (\u03c4Obs \b \u03c4Cd). Ensuring sufficient phase space sampling may\nrepresent an impervious task when dealing with complex cogni-\ntive processes. For instance, if we take the example of a generic\nopen-ended reasoning task, it is difficult to decide whether a\nreasoning episode, or indeed even an ensemble of episodes, suf-\nficiently sample the repertoire available to a subject. The way\nthe space is sampled by separate trials of an experiment may\npresent some complexity even for fast relaxational processes,\nwhere \u03c4Cd is approximated by the Poincar\u00e9 recurrence time \u03c4P.\nFor instance, observed inter-trial fluctuations can be thought of\nas a sign of the system\u2019s exploring its dynamic repertoire (Ghosh\net al., 2007; Deco et al., 2011). Contrary to the case of com-\nplex forms of reasoning, for which trials can typically be very\nlong, most perceptual tasks are of relative short average dura-\ntion, and this allows increasing the number of experimental\ntrials.\nAnother important factor to take into account is stationarity,\ni.e., invariance under time shift of the joint probability distribu-\ntion, or at least of some moments of the variables quantified in a\ngiven experiment.\nIn behavioral studies (e.g., Ihlen and Vereijken, 2010), sin-\ngle trials are often mapped into a scalar, e.g., a response time,\nand the process that is considered is given by the sequence of\nthese scalars in the experiment. The underlying cross-scale pro-\ncess is generally considered stationary over the entire experiment\nencompassing a large number of trials, independently of whether\nthe local probabilities show or not time-dependence.\nIn cognitive neuroscience, each trial is mapped onto brain\nactivity, and this forces into dealing with within-trial as well as\ninter-trial (and occasionally inter-session) stationarity. Complex\ncognitive processes such as reasoning present an inherent\ndilemma between two opposing needs: to ensure that the tortu-\nous phase space be explored on the one hand, and that the signal\nbe stationary on the other. The former may require extremely\nlong trials. While, in principle, for the inverse power law regime\nto be observable, trial length needs to be much longer than the\ntime scale of stretched exponential relaxation (Grigolini, 2008),\nthe asymptotic regimemay nonetheless be inferred through finite-\nsize scaling analysis (Fisher and Barber, 1972). This method takes\ninto account the finite size of experimental data by conveniently\nrescaling the correlation length \u03be by the system\u2019s linear size L and\nobserving how measured quantities vary for different values of L.\nThe values for the critical exponents are then recovered by tak-\ning the infinite limit L \u2192 \u221e. On the other hand, in the long\ntime regime, brain activity associated with the execution of com-\nplex cognitive tasks is characterized by the presence of long-term\nmemory, weak ergodicity breaking and aging, i.e., temporal cor-\nrelations and WTDs become dependent on the observation time\n(West et al., 2008). As a result, De may diverge and \u03c4Cd \b \u3008L\u3009.\nWhile stationarity may be assured stricto sensu only for time\nscales that are much shorter than the typical duration of, e.g.,\na reasoning episode, a quasi-stationary regime for the suscep-\ntibility of the system can be identified over time scales \u03c4\u03bc \u0006\n\u03c4QStat \u0006 \u03c4rel for a given waiting timemuch larger than the slowest\nfrequency in the data (t \b \u03c9\u22121) (Crisanti and Ritort, 2003).\nA standard assumption in cognitive neuroscience is that sep-\narate trials of a given experiment are identically distributed,\nindependent samples, each accounting for the same part of an\nunderlying attractor. The CLT ensures that trial averaged quanti-\nties approach a Gaussian distribution. For instance, event-related\nbrain potential studies assume that an observed response xj(t)\ncan be decomposed into a stereotyped evoked response ej occur-\nring with a constant delay after a given stimulus, and additive\nGaussian noise \u03bej. Averaging then improves the signal-to-noise\nratio by a factor\n\u221a\nN , where the number of responses N typically\nequals 20\u2013300, and\n\u2329\nxj(t)\n\u232a \u2192 ej(t) for N \u2192 \u221e. However, under\nconditions that are explained below, trials may be neither identi-\ncally distributed, nor independent in practice, so that adding trials\nmay not increase the signal-to-noise ratio as the square root of the\nnumber of trials.\nFor trials to represent a basic periodicity, not only should\n{\u03c4i,R} be equal for all trials, in a statistical sense, but it is also\nnecessary that \u03c4Cd \u2264 \u3008L\u3009. If, on the contrary, \u03c4Cd > \u3008L\u3009, each\ntrial corresponds to different parts of a vast attractor with trial-\nspecific subparts (rather than sampling the same part of the state\nspace). Trial repetition could then improve phase space explo-\nration rather than the signal-to-noise ratio. This may for instance\nbe the case of different reasoning trials of a given experiment, as\nthere is typically more than one way to carry out a given task, and\nthe repertoire of corresponding brain activity may be vast.\nTrial independence requires that the inter-trial interval be\nITI \b \u03c4relax. For example, in perceptual tasks, ensuring inter-trial\nindependence requires that the brain response to a given stimulus\nvanishes before the following is presented. This condition may be\nharder to fulfil than is often assumed in standard event-related\npotential studies, given that the brain response to stimuli does\nnot vanish exponentially fast, but rather in a history-dependent\nbroad-band fashion (Gilboa et al., 2005; Drew and Abbott, 2006;\nLundstrom et al., 2008).\nThe presence of learning means that the dynamic repertoire of\nbrain activity may change during the course of the experiment. In\nother words, the phase space landscape itself may evolve, as a con-\nsequence of learning. Landscape fluctuations are usually thought\nto be much slower than those of the system; however, in many\ncontexts, changes may occur on time scales \u03c4L \u2264 \u03c4Obs so that\nthe landscape dynamics cannot be neglected. For \u03c4L \u2248 \u3008L\u3009, the\nstatistics of within-trial landscape dynamics can be considered as\nquasi-stationary. When, instead, the landscape shows significant\nfluctuations at the single trial level (\u03c4L \u2264 \u3008L\u3009), the phase space\nitself may not be well-defined.\nWithin-trial statistics, landscape dynamics, path dependence,\nand essential transience all affect the way trials of a given experi-\nment can be treated in an aggregate way.\nFrontiers in Physiology | Fractal Physiology April 2013 | Volume 4 | Article 86 | 6\nPapo Time scales in cognitive neuroscience\nIn the presence of scale-free distributions, long-range correla-\ntions and inter-trial path-dependence, trials may show weak or\nno self-averaging, i.e., letting the sample become larger does not\nimprove statistics, as dispersion may remain even when the num-\nber of trials goes to infinity (Aharony and Harris, 1996). This in\nturn warns that care should be taken when averaging across trials,\nand indicates that, at least up to certain scales, the inter-trial tem-\nporal scales of a given experiment might better be accounted for\nin alternative, conceptually different ways, e.g., by characterizing\nthe scaling properties, data collapse and universality of probability\ndistributions (Bramwell et al., 1998; Friedman et al., 2012).\nFurthermore, fixed landscape and disorder-averaged prop-\nerties, respectively corresponding to a distribution of walkers\ninitially concentrated or spread out over the sample (Bouchaud\nand Georges, 1990) should be treated differently when calculating\nexperiment-wide statistics.\nFinally, while relaxation can be slow enough as to de facto rule\nout the existence of an underlying attractor dynamics, different\ntransients can be used to reveal parts of the vector field associated\nwith the cognitive space; estimates of conditional probabilities\nand of the corresponding stochastic dynamical systems can be\nderived frommultiple, small data sets rather than from of a single\nlong one (van Mourik et al., 2006).\nIn summary, a number of factors may affect observability of\ncognitive phenomena: sheer complexity of the phase space of\ntask-related brain activity, long-term memory, aging and weak\nergodicity breaking of single trial trajectories, path dependence,\nphase space evolution, and essential transience. These factors\ninfluence the way experiments ought to be designed to ensure that\nthe phenomena under study are appropriately observed, the sta-\ntus of single trials of a given experiment, and the corresponding\nanalyses used to aggregate them. Experiments (viz. through their\noverall length, their design into trials), and analyses (e.g., through\nthresholds and windowing procedures) both introduce spurious\nscales of which experimenters should be aware.\nCONCLUDING REMARKS\nA proper characterization of temporal scales provides on the one\nhand principled guidance for basic experimental steps and, on the\nother hand, a means to correctly identify the neural correlates of\nimportant cognitive phenomena.\nMethods of data analysis ranging from the level of inter-\ntrial aggregation procedures down to fundamental single-trial\nlevel ones crucially depend on the knowledge of the temporal\nscales of the phenomenon at hand. Experimental designs should\ncorrespondingly hinge on a general understanding of the phe-\nnomenon\u2019s temporal scales. For example, at the inter-trial level,\nwhen studying generic open-ended forms of reasoning, few long\ntrials may ensure better exploration of the state space than mul-\ntiple time-constrained ones. Short reasoning episodes may have\na drastically simpler neural phase space than a single longer one.\nOn the other hand, the longer the considered reasoning episode,\nthe more probable that it accurately visits the complex phase\nspace of task-related brain activity, and the better its time average\napproximates an ensemble average.\nAt the single-trial level, measures of brain activity (e.g.,\nlocal amplitudes, or correlations or synchronization between two\nrecording sites) are often time-averaged within windows in which\nthey are stationary.While the size of time windows is often chosen\nonce and for all for entire data sets, searching for true microstates,\nwhich can be thought of as stationary segments, implies a data-\ndriven segmentation (van de Ville et al., 2010). In the presence of\nlong-range temporal correlations, however, defining the bound-\naries of stationary segments is an inherently arduous task, and\neven more so when evaluating global microstates of whole-brain\nactivity. On the other hand, ensuring stationarity by considering\nshort time-windows may come at the price of sacrificing tem-\nporal scales, as typical procedures de facto constitute high-pass\nfilters on the data, and would thus lead to missing key aspects\nof the dynamics. As a consequence, an appropriate strategy may\nrequire working with quantities evaluated over extremely short\ntime-windows (e.g., Bianco et al., 2007a).\nA similar conclusion holds for the choice of recording param-\neters, viz. sampling rates. For instance, surprising though it may\nread, temporal resolution is in general not sought to increase\nthe ability to understand the complex relationship between pro-\ncesses unfolding at different temporal scales. Indeed, it is fair to\nsay that the very concept of temporal scale is essentially alien to\nstandard cognitive neuroscience. A telltale of this is that, curi-\nously, the need for temporal precision is often thought to be\ninversely proportional to the typical temporal duration of the\ncognitive process to be investigated, rather than to its complex-\nity. In other words, temporal precision is often thought to be\nneeded more when studying short (e.g., perceptual) than long\n(e.g., in reasoning) processes. The idea is that temporal precision\nis essentially required to capture fast, fleeting processes, which\nwould otherwise go unnoticed. It is however important to see\nthat, in the case of a complex process such as reasoning, the\nreconstruction of the tortuous phase space trajectory improves\nwith the amount of available (non-spuriously correlated) time\npoints.\nThe explicit handling of the temporal dimension of cognition\nallows framing the neural correlates of given cognitive processes\nin terms of temporal scales, and their set of relationships. In\nfact, describing cognitive function may sometimes essentially boil\ndown to specifying {\u03c4i,R}, which may in turn constitute a neces-\nsary condition for the statistical treatment of experimental data. A\nparticularly clear illustration is represented by complex phenom-\nena such as conceptual learning or reasoning and thinking, which,\nas we have seen, lack both an average temporal duration and\ninner segmentation. Deriving generic characteristics of such pro-\ncesses in spite of their inherent variability, or evaluating whether\na given observation time is sufficient cannot be done without\ncharacterizing time scales and describing task-modulated changes\nin the relationship among processes unfolding at different time\nscales.\nOverall, some general experimental indications for cognitive\nscientists can be derived. Accounting for the temporal scales of\ncomplex cognitive processes may involve changes in the ways\nexperiments are designed, data analysed and brain recording\ninstruments and the relative parameters are chosen. Which of\nthese aspects needs to be changed with respect to standard proce-\ndures depends on the type of cognitive process under study.While\nstudying fast perceptual processes may only involve changes in\nwww.frontiersin.org April 2013 | Volume 4 | Article 86 | 7\nPapo Time scales in cognitive neuroscience\ndata analysis methods, studying phenomena such as thinking or\nreasoningmay require a more global change of all of these aspects.\nIndeed, these fundamental changes are necessary to tackle the\nmost challenging aspects of these extremely complex cognitive\nprocesses.\nSpecifically, we highlight four main experimental implications.\n(1) With phenomena lacking a clear characteristic duration, e.g.,\nopen-ended forms of reasoning, it may be interesting to build\nexperiments with few long trials with rich dynamical repertoire.\n(2) Different trials of a given experiment (be they short-lived\nperceptual or long duration reasoning tasks) should not automat-\nically be treated as independent samples of the same part of an\nunderlying attractor, and may not self-average as their number\nis increased. This may bear important consequences both on the\ntype of metric that can be used to quantify brain activity and on\nthe statistical tests that may highlight differences between exper-\nimental conditions. (3) At shorter time scales, there is a mutual\ninfluence between the time scales of the process and the choice of\nthe window size in which relevant quantities (the choice of which\nis not the goal of the present article, and depend of the inves-\ntigator\u2019s goals) are calculated. Decisions must then be made as\nto the size of these windows. (4) While an appropriate temporal\nsampling rate is necessary to detect abrupt short-lived transi-\ntions, irrespective of the considered cognitive phenomenon under\nstudy, it is also necessary to reconstruct the underlying dynam-\nics, contrary to the common stance whereby processes such as\nreasoning or thinking may just need a poor time-resolution\nrecording device.\nFinally, at a more conceptual level, the time scales of a given\nphenomenon and, in particular, the relationship R among them\ncontain information on the mechanisms through which cognitive\nprocesses interact to give rise to cognitive performance (Holden\net al., 2009). The long time limit loss of time-scale separation\nprovides a possible mechanism through which changes at faster\ntime scales (e.g., fast perceptual and attentional processes) may\nbe embedded into activity (e.g., memory processes) unfolding\nat times scales several orders of magnitude larger (Fujimoto and\nKaneko, 2003). The presence of scaling in a broad range of time\nscales suggests that observable behavior is but the uppermost\nlevel of a scale-invariant phenomenon and may be understood\nas a macroscopic coarse-grained aspect of ongoing brain activ-\nity (Fingelkurts et al., 2010), resulting from the renormalization\nof its fluctuations, and that cognitive function may not always\nbe decomposable into less complex component processes. Time\ncan also be regarded as a variable revealing the constructive\nrules underlying the spatial structure of task-related brain activ-\nity at various scales (P\u00e9rez-Mercader, 2004; Itzkovitz et al., 2005;\nDelvenne et al., 2010).\nACKNOWLEDGMENTS\nThe author thanks Dr. Ricardo Guti\u00e9rrez for extremely fruitful\ndiscussions and comments.\nREFERENCES\nAbry, P. (2003). \u201cScaling and wavelets:\nan introductory walk?,\u201d in Processes\nwith Long Range Correlations:\nTheory and Applications, eds G.\nRangarajan and M. Ding (Berlin,\nHeidelberg:Springer-Verlag),34\u201360.\nAharony, A., and Harris, A. B. (1996).\nAbsence of self-averaging and uni-\nversal fluctuations in random sys-\ntems near critical points. Phys. Rev.\nLett. 77, 3700\u20133703.\nAllegrini, P., Bologna, M., Grigolini,\nP., and West, B. J. (2007).\nFluctuation-dissipation theo-\nrem for event-dominated processes.\nPhys. Rev. Lett. 99:010603. doi:\n10.1103/PhysRevLett.99.010603\nAllegrini, P., Menicucci, D., Paradisi,\nP., and Gemignani, A. (2010).\nFractal complexity in spontaneous\nEEG metastable-state transitions:\nnew vistas on integrated neural\ndynamics. Front. Physiol. 1:128. doi:\n10.3389/fphys.2010.00128\nAllegrini, P., Paradisi, P., Menicucci,\nD., Bedini, R., Gemignani, A.,\nand Fronzoni, L. (2011). Noisy\ncooperative intermittent pro-\ncesses: from blinking quantum\ndots to human consciousness.\nJ. Phys. Conf. Ser. 306:012027. doi:\n10.1088/1742-6596/306/1/012027\nAquino, G., Bologna, M., West,\nB. J., and Grigolini, P. (2011).\nTransmission of information\nbetween complex systems: 1/f res-\nonance. Phys. Rev. E Stat. Nonlin.\nSoft Matter Phys. 83:051130. doi:\n10.1103/PhysRevE.83.051130\nAquino, G., Grigolini, P., and West,\nB. J. (2007). Linear response and\nfluctuation-dissipation theorem for\nnon-Poissonian renewal processes.\nEurophys. Lett. 80:10002. doi:\n10.1209/0295-5075/80/10002\nBacry, E., Delour, J., and Muzy, J.\nF. (2001). Multifractal random\nwalk. Phys. Rev. E Stat. Nonlin.\nSoft Matter Phys. 64:026103. doi:\n10.1103/PhysRevE.64.026103\nBianco, S., Ignaccolo, M., Rider, M.\nS., Ross, M. J., Winsor, P., and\nGrigolini, P. (2007a). Brain, music,\nand non-Poisson renewal pro-\ncesses. Phys. Rev. E Stat. Nonlin.\nSoft Matter Phys. 75:061911. doi:\n10.1103/PhysRevE.75.061911\nBianco, S., Grigolini, P., and Paradisi, P.\n(2007b). A fluctuating environment\nas a source of periodic modulation.\nChem. Phys. Lett. 438, 336\u2013340.\nBonnefond, M., and Van der Henst, J.\nB. (2009). What\u2019s behind an infer-\nence? An EEG study with condi-\ntional arguments. Neuropsychologia\n47, 3125\u20133133.\nBouchaud, J.-P., and Georges, A.\n(1990). Anomalous diffusion\nin disordered media: statistical\nmechanisms, models and physi-\ncal applications. Phys. Rep. 195,\n127\u2013293.\nBramwell, S. T., Holdsworth, P. C.\nW., and Pinton, J.-F. (1998).\nUniversality of rare fluctuations in\nturbulence and critical phenomena.\nNature 396, 552\u2013554.\nBuiatti, M., Papo, D., Baudonni\u00e8re, P.\nM., and van Vreeswijk, C. (2007).\nFeedback modulates the tempo-\nral scale-free dynamics of brain\nelectrical activity in a hypothe-\nsis testing task. Neuroscience 146,\n1400\u20131412.\nBuice, M. A., and Cowan, J. D. (2009).\nStatistical mechanics of the neocor-\ntex. Prog. Biophys. Mol. Biol. 99,\n53\u201386.\nBurov, S., and Barkai, E. (2008).\nCritical exponent of the frac-\ntional Langevin equation. Phys.\nRev. Lett. 100:070601. doi:\n10.1103/PhysRevLett.100.070601\nCarretero-Campos, C., Bernaola-\nGalv\u00e1n, P., Ivanov, P. Ch., and\nCarpena, P. (2012). Phase tran-\nsitions in the first-passage time\nof scale-invariant correlated pro-\ncesses. Phys. Rev. E Stat. Nonlin.\nSoft Matter Phys. 85:011139. doi:\n10.1103/PhysRevE.85.011139\nCastaing, B., Gagne, Y., and Hopfinger,\nE. J. (1990). Velocity probability\ndensity functions of high Reynolds\nnumber turbulence. Physica D 46,\n177\u2013200.\nCastiglione, P., Mazzino, A., Muratore-\nGinanneschi, P., and Vulpiani, A.\n(1999). On strong anomalous diffu-\nsion. Physica D 134, 75\u201393.\nChainais, P., Riedi, R., and Abry, P.\n(2005). On non scale invariant\ninfinitely divisible cascades. IEEE\nTrans. Inf. Theory 51, 1063\u20131083.\nCiuciu, P., Abry, P., Rabrait, C., and\nWendt, H. (2011). Log wavelet\nleaders cumulant based multifrac-\ntal analysis of EVI fMRI time series:\nevidence of scaling in ongoing and\nevoked brain activity. IEEE J. Sel.\nTop. Appl. 2, 929\u2013943.\nCiuciu, P., Varoquaux, G., Abry, P.,\nSadaghiani, S., and Kleinschmidt,\nA. (2012). Scale-free and mul-\ntifractal time dynamics of\nfMRI signals during rest and\ntask. Front. Physiol. 3:186. doi:\n10.3389/fphys.2012.00186\nCrisanti, A., and Ritort, F. (2003).\nViolation of the fluctuation\u2013\ndissipation theorem in glassy\nsystems: basic notions and the\nnumerical evidence. J. Phys. A\nMath. Gen. 36, R181\u2013R290.\nCugliandolo, L. F., Dean, D. S., and\nKurchan, J. (1997). Fluctuation-\ndissipation theorems and entropy\nproduction in relaxational systems.\nPhys. Rev. Lett. 79, 2168\u20132171.\nFrontiers in Physiology | Fractal Physiology April 2013 | Volume 4 | Article 86 | 8\nPapo Time scales in cognitive neuroscience\nDeco, G., Jirsa, V. K., and McIntosh,\nA. R. (2011). Emerging concepts\nfor the dynamical organization of\nresting-state activity in the brain.\nNat. Rev. Neurosci. 12, 43\u201356.\nDelvenne, J. C., Yaliraki, S. N., and\nBarahona, M. (2010). Stability of\ngraph communities across time\nscales. Proc. Natl. Acad. Sci. U.S.A.\n107, 12755\u201312760.\nDrew, P. J., and Abbott, L. F. (2006).\nModels and properties of power-\nlaw adaptation in neural systems.\nJ. Neurophysiol. 96, 826\u2013833.\nFairhall, A. L., Lewen, G. D., Bialek,\nW., and de Ruyter van Steveninck,\nR. (2001). \u201cMultiple timescales of\nadaptation in a neural code,\u201d in\nAdvances in Neural Information\nProcessing Systems 13, eds T. K.\nLeen, T. G. Dietterich, and V. Tresp\n(Cambridge, MA: MIT Press),\n124\u2013130.\nFerrari, R., Manfroi, A. J., and Young,\nW. R. (2001). Strongly and weakly\nself-similar diffusion. Physica D 154,\n111\u2013137.\nFingelkurts, A. A., and Fingelkurts, A.\nA. (2004). Making complexity sim-\npler: multivariability and metasta-\nbility in the brain. Int. J. Neurosci.\n114, 843\u2013862.\nFingelkurts, A. A., Fingelkurts, A.\nA., and Neves, C. F. H. (2010).\nNatural world physical, brain\noperational, and mind phenome-\nnal space\u2013time. Phys. Life Rev. 7,\n195\u2013249.\nFisher, M. E., and Barber, M. N. (1972).\nScaling theory for finite-size effects\nin the critical region. Phys. Rev. Lett.\n28, 1516\u20131519.\nFreeman, W. J., Holmes, M. D., Burke,\nB. C., and Vanhatalo, S. (2003).\nSpatial spectra of scalp EEG and\nEMG from awake humans. Clin.\nNeurophysiol. 114, 1053\u20131068.\nFreyer, F., Aquino, K., Robinson,\nP. A., Ritter, P., and Breakspear,\nM. (2009). Bistability and\nNon-Gaussian fluctuations in\nspontaneous cortical activity.\nJ. Neurosci. 29, 8512\u20138524.\nFreyer, F., Roberts, J. A., Ritter, P., and\nBreakspear, M. (2012). A canonical\nmodel of multistability and scale-\ninvariance in biological systems.\nPLoS Comput. Biol. 8:e1002634.\ndoi:10.1371/journal.pcbi.1002634\nFriedman, N., Ito, S., Brinkman, B.\nA., Shimono, M., Deville, R. E.,\nDahmen, K. A., et al. (2012).\nUniversal critical dynamics in high\nresolution neuronal avalanche data.\nPhys. Rev. Lett. 108:208102. doi:\n10.1103/PhysRevLett.108.208102\nFriedrich, R., Peinke, J., and Renner,\nCh. (2000). How to quantify deter-\nministic and random influences\non the statistics of the foreign\nexchange market. Phys. Rev. Lett. 84,\n5224\u20135227.\nFriedrich, R., Peinke, J., Sahimi,\nM., and Rahimi Tabar, M. R.\n(2011). Approaching complexity by\nstochastic methods: from biological\nsystems to turbulence. Phys. Rep.\n506, 87\u2013162.\nFujimoto, K., and Kaneko, K. (2003).\nBifurcation cascade as chaotic itin-\nerancy with multiple time scales.\nChaos 13, 1041\u20131056.\nGhosh, A., Rho, Y., McIntosh, A.\nR., K\u00f6tter, R., and Jirsa, V.\nK. (2007). Noise during rest\nenables the exploration of the\nbrain\u2019s dynamic repertoire. PLoS\nComput. Biol. 4:e1000196. doi:\n10.1371/journal.pcbi.1000196\nGilboa, G., Chen, R., and Brenner,\nN. (2005). History-dependent\nmultiple-timescale dynamics in a\nsingle-neuron model. J. Neurosci.\n25, 6479\u20136489.\nGoel, V., Gold, B., Kapur, S., and Houle,\nS. (1997). The seats of reason? An\nimaging study of deductive and\ninductive reasoning. Neuroreport 8\n1305\u20131310.\nGoel, V., Gold, B., Kapur, S., and Houle,\nS. (1998). Neuroanatomical corre-\nlates of human reasoning. J. Cogn.\nNeurosci. 10, 293\u2013302.\nGong, P., Nikolaev, A. R., and van\nLeeuwen, C. (2002). Scale-invariant\nfluctuations of the dynamical syn-\nchronization in human brain elec-\ntrical activity. Neurosci. Lett. 336,\n33\u201336.\nGong, P., Nikolaev, A. R., and van\nLeeuwen, C. (2007). Intermittent\ndynamics underlying the intrinsic\nfluctuations of the collective syn-\nchronization patterns in electro-\ncortical activity. Phys. Rev. E Stat.\nNonlin. Soft Matter Phys. 76:011904.\ndoi: 10.1103/PhysRevE.76.011904\nGrigolini, P. (2008). Complexity and\ngeneralized exponential relaxation:\nmemory versus renewal. Int. J.\nBifurcat. Chaos 9, 2709\u20132716.\nGrigolini, P., Rocco, A., and West, B.\nJ. (1999). Fractional calculus as a\nmacroscopic manifestation of ran-\ndomness. Phys. Rev. E Stat. Nonlin.\nSoft Matter Phys. 59, 2603\u20132613.\nH\u00e4nggi, H., and Jung, J. (1995).\nColored noise in dynamical systems.\nAdv. Chem. Phys. 89, 239\u2013326.\nH\u00e4nggi, H., Talkner, P., and Borkovec,\nM. (1990). Reaction-rate theory:\nfifty years after Kramers. Rev. Mod.\nPhys. 62, 251\u2013341.\nHe, B. J., Zempel, J. M., Snyder, A. Z.,\nand Raichle, M. E. (2010). The tem-\nporal structures and functional sig-\nnificance of scale-free brain activity.\nNeuron 66, 353\u2013369.\nHolden, J. G., Van Orden, G. C., and\nTurvey, M. T. (2009). Dispersion\nof response times reveals cogni-\ntive dynamics. Psychol. Rev. 116,\n318\u2013342.\nHsu, D., and Hsu, M. (2009). Zwanzig-\nMori projection operators and EEG\ndynamics: deriving a simple equa-\ntion of motion. PMC Biophys. 2:6.\ndoi: 10.1186/1757-5036-2-6\nIhlen, E. A. F., and Vereijken, B. (2010).\nInteraction-dominant dynamics in\nhuman cognition: beyond 1/f\u03b1 fluc-\ntuation. J. Exp. Psychol. Gen. 139,\n436\u2013463.\nItzkovitz, S., Levitt, R., Kashtan, N.,\nMilo, R., Itzkovitz, M., and Alon,\nU. (2005). Coarse-graining and\nself-dissimilarity of complex net-\nworks. Phys. Rev. E Stat. Nonlin.\nSoft Matter Phys. 71:016127. doi:\n10.1103/PhysRevE.71.016127\nJung-Beeman, M., Bowden, E. M.,\nHaberman, J., Frymiare, J. L.,\nArambel-Liu, S., Greenblatt, R.,\net al. (2004). Neural activity when\npeople solve verbal problems with\ninsight. PLoS Biol. 2:e97. doi:\n10.1371/journal.pbio.0020097\nKaplan, A. Y., Fingelkurts, A. A.,\nFingelkurts, A. A., Borisov, B. S.,\nand Darkhovsky, B. S. (2005).\nNonstationary nature of the brain\nactivity as revealed by EEG/MEG:\nmethodological, practical and con-\nceptual challenges. Signal Process.\n85, 2190\u20132212.\nKarni, A., Meyer, G., Jezzard, P., Adams,\nM., Turner, R., and Ungerleider, L.\nG. (1995). Functional MRI evidence\nfor adult motor cortex plasticity\nduring motor skill learning, Nature\n377, 155\u2013158.\nKoenig, T., Prichep, L., and Lehmann,\nD. (2002). Millisecond by mil-\nlisecond, year by year: normative\nEEF microstates and developmental\nstages. Neuroimage 16, 41\u201348.\nKoukou, M., and Lehmann, D. (1987).\nAn information processing per-\nspective of psychophysiological\nmeasurements. J. Psychophysiol. 1,\n109\u2013112.\nKubo, R. (1966). The fluctuation-\ndissipation theorem. Rep. Prog.\nPhys. 29, 255\u2013284.\nLang, S., Kanngieser, N., Jas\u00b4kowski, P.,\nHaider, H., Rose, M., and Verleger,\nR. (2006). Precursors of insight\nin event-related brain potentials.\nJ. Cogn. Neurosci. 18, 2052\u20132066.\nLatka, M., Glaubic-Latka, M., Latka,\nD., and West, B. J. (2004). Fractal\nrigidity in migraine. Chaos Solitons\nFractals 20, 165\u2013170.\nLinkenkaer-Hansen, K., Nikouline, V.\nV., Palva, J. M., and Ilmoniemi, R.\n(2001). Long-range temporal cor-\nrelations and scaling behavior in\nhuman oscillations. J. Neurosci. 15,\n1370\u20131377.\nLinkenkaer-Hansen, K., Nikouline,\nV. V., Palva, J. M., Kaila, K., and\nIlmoniemi, R. (2004). Stimulus-\ninduced change in long-range\ntemporal correlations and scal-\ning behaviour of sensorimotor\noscillations. Eur. J. Neurosci. 19,\n203\u2013211.\nLongo, G., Mont\u00e9vil, M., and\nPocheville, A. (2012). From\nbottom-up approaches to levels of\norganization and extended critical\nsystems. Front. Physiol. 3:232. doi:\n10.3389/fphys.2012.00232\nLundstrom, B. N., Higgs, M. H., Spain,\nW. J., and Fairhall, A. L. (2008).\nFractional differentiation by neo-\ncortical pyramidal neurons. Nat.\nNeurosci. 11, 1335\u201313342.\nMai, X.Q., Luo, J.,Wu, J. H., and Luo, Y.\nJ. (2004). \u201cAha!\u201d effects in a guessing\nriddle task: an event-related poten-\ntial study. Hum. Brain Mapp. 22,\n261\u2013270.\nManshour, P., Saberi, S., Sahimi,\nM., Peinke, J., Pacheco, A.\nF., and Rahimi Tabar, M. R.\n(2009). Turbulencelike behav-\nior of seismic time series. Phys.\nRev. Lett. 102:014101. doi:\n10.1103/PhysRevLett.102.014101\nMantegna, R. N., and Stanley, H.\nE. (1994). Stochastic process\nwith ultraslow convergence to a\nGaussian: the truncated L\u00e9vy flight.\nPhys. Rev. Lett. 73, 2946\u20132949.\nNewell, K. M., Liu, Y.-M., and Mayer-\nKress, G. (2001). Time scales in\nmotor learning. Psychol. Rev. 108,\n57\u201382.\nNovikov, E., Novikov, A., Shannahoff-\nKhalsa, D., Schwartz, B., and\nWright, J. (1997). Scale-similar\nactivity in the brain. Phys. Rev. E\nStat. Nonlin. Soft Matter Phys. 56,\nR2387\u2013R2389.\nOsherson, D., Perani, D., Cappa, S.,\nSchnur, T., Grassi, F., and Fazio,\nF., et al. (1998). Distinct brain\nfoci in deductive versus probabilis-\ntic reasoning. Neuropsychologia 36,\n369\u2013376.\nPapo, D. (2013). Why should cognitive\nneuroscientists study the brain\u2019s\nresting state? Front. Hum. Neurosci.\n7:45. doi: 10.3389/fnhum.2013.\n00045\nParsons, L. M., and Osherson, D.\n(2001). New evidence for distinct\nright and left brain systems for\ndeductive versus probabilistic rea-\nsoning. Cereb. Cortex 11, 954\u2013965.\nP\u00e9rez-Mercader, J., (2004). \u201cCoarse-\ngraining, scaling and hierarchies,\u201d\nin Nonextensive Entropy -\nInterdisciplinary Applications,\neds. M. Gell-Mann and C. Tsallis\nwww.frontiersin.org April 2013 | Volume 4 | Article 86 | 9\nPapo Time scales in cognitive neuroscience\n(Oxford: Oxford University Press),\n357\u2013376.\nPijnacker, J., Geurts, B., van Lambalgen,\nM., Buitelaar, J., and Hagoort,\nP. (2011). Reasoning with excep-\ntions: an event-related brain poten-\ntials study. J. Cogn. Neurosci. 23,\n471\u2013480.\nPottier, N., and Mauger, A. (2004).\nAnomalous diffusion of a particle\nin an aging medium Physica A 332,\n15\u201328.\nQiu, J., Li, H., Yang, D., Luo, Y.,\nLi, Y., Wu, Z., et al. (2008). The\nneural basis of insight problem\nsolving: an event-related poten-\ntial study. (2008). Brain Cogn. 68,\n100\u2013106.\nRebenshtok, A., and Barkai, E.\n(2007). Distribution of time-\naveraged observables forweak\nergodicity breaking. Phys.\nRev. Lett. 99:210601. doi:\n10.1103/PhysRevLett.99.210601\nReiner, M. (1964). The Deborah num-\nber. Phys. Today 17, 62.\nShew, W. L., Yang, H., Petermann,\nT., Roy, R., and Plenz, D. (2009).\nNeuronal avalanches imply max-\nimum dynamic range in cortical\nnetworks at criticality. J. Neurosci.\n29, 15595\u201315600.\nSornette, D., (2004). Critical\nPhenomena in Natural Sciences,\nChaos, Fractals, Self-Organization\nand Disorder: Concepts And Tools. 2\nEdn. Heidelberg: Springer Series in\nSynergetics.\nStam, C. J., and de Bruin, E. A. (2004).\nScale-free dynamics of global\nfunctional connectivity in the\nhuman brain. Hum. Brain Mapp.\n22, 97\u2013109.\nT\u00e9l, T., and Lai, Y.-C. (2008). Chaotic\ntransients in spatially extended sys-\ntems. Phys. Rep. 460, 245\u2013275.\nvan de Ville, D., Britz, J., and Michel,\nC. M. (2010). EEG microstate\nsequences in healthy humans at\nrest reveal scale-free dynamics.\nProc. Natl. Acad. Sci. U.S.A. 107,\n18179\u201318184.\nvan Mourik, A. M., Daffertshofer, A.,\nand Beek, P. J. (2006). Estimating\nKramers\u2013Moyal coefficients in short\nand non-stationary data sets. Phys.\nLett. A 351, 13\u201317.\nWest, B. J., Geneston, E. L., and\nGrigolini, P. (2008). Maximizing\ninformation exchange between\ncomplex networks. Phys. Rep. 468,\n1\u201399.\nZilber, N., Ciuciu, P., Abry, P., and\nvan Wassenhove, V. (2012).\n\u201cModulation of scale-free proper-\nties of brain activity in MEG,\u201d in\nIEEE International Symposium on\nBiomedical Imaging (Barcelona),\n1531\u20131534.\nConflict of Interest Statement: The\nauthor declares that the research\nwas conducted in the absence of any\ncommercial or financial relationships\nthat could be construed as a potential\nconflict of interest.\nReceived: 16 November 2012; accepted:\n03 April 2013; published online: 19 April\n2013.\nCitation: Papo D (2013) Time scales\nin cognitive neuroscience. Front. Physiol.\n4:86. doi: 10.3389/fphys.2013.00086\nThis article was submitted to Frontiers\nin Fractal Physiology, a specialty of\nFrontiers in Physiology.\nCopyright \u00a9 2013 Papo. This is an\nopen-access article distributed under\nthe terms of the Creative Commons\nAttribution License, which permits use,\ndistribution and reproduction in other\nforums, provided the original authors\nand source are credited and subject to any\ncopyright notices concerning any third-\nparty graphics etc.\nFrontiers in Physiology | Fractal Physiology April 2013 | Volume 4 | Article 86 | 10\n", "identifiers": ["10.3389/fphys.2013.00086"], "publisher": "Frontiers Media SA", "relations": ["http://dx.doi.org/10.3389/fphys.2013.00086"], "repositories": [{"id": "2614", "openDoarId": 0, "name": "Frontiers - Publisher Connector", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 1, "metadataAdded": 1494863507000, "metadataUpdated": 1494863507000, "depositedDate": null}, "subjects": ["journal-article"], "title": "Time scales in cognitive neuroscience", "topics": [], "types": [], "year": 2013, "fulltextIdentifier": "https://core.ac.uk/download/pdf/82864389.pdf", "doi": "10.3389/fphys.2013.00086", "downloadUrl": "https://core.ac.uk/download/pdf/82864389.pdf"}}
{"status": "Not found", "data": null}
{"status": "Not found", "data": null}
{"status": "OK", "data": {"id": "41241868", "authors": ["Smith, Andrea L", "Carter, SM", "Dunlop, SM", "Freeman, B", "Chapman, S"], "citations": [], "contributors": [], "datePublished": "2015-05-26", "description": "Background\r\n\r\nUnassisted cessation \u2013 quitting without pharmacological or professional support \u2013 is an enduring phenomenon. Unassisted cessation persists even in nations advanced in tobacco control where cessation assistance such as nicotine replacement therapy, the stop-smoking medications bupropion and varenicline, and behavioural assistance are readily available. We review the qualitative literature on the views and experiences of smokers who quit unassisted.\r\nMethod\r\n\r\nWe systematically searched for peer-reviewed qualitative studies reporting on smokers who quit unassisted. We identified 11 studies and used a technique based on Thomas and Harden\u2019s method of thematic synthesis to discern key themes relating to unassisted cessation, and to then group related themes into overarching concepts.\r\nFindings\r\n\r\nThe three concepts identified as important to smokers who quit unassisted were: motivation, willpower and commitment. Motivation, although widely reported, had only one clear meaning, that is \u2018the reason for quitting\u2019. Willpower was perceived to be a method of quitting, a strategy to counteract cravings or urges, or a personal quality or trait fundamental to quitting success. Commitment was equated to seriousness or resoluteness, was perceived as key to successful quitting, and was often used to distinguish earlier failed quit attempts from the final successful quit attempt. Commitment had different dimensions. It appeared that commitment could be tentative or provisional, and also cumulative, that is, commitment could be built upon as the quit attempt progressed.\r\nConclusion\r\n\r\nA better understanding of what motivation, willpower and commitment mean from the smoker\u2019s perspective may provide new insights and direction for smoking cessation research and practice.Funding was provided by the National Health and Medical Research Council (NHMRC), Australia, grant number 1024459. SMC was supported by NHMRC Career Development Fellowship 1032963", "fullText": "RESEARCH ARTICLE\nThe Views and Experiences of Smokers Who\nQuit Smoking Unassisted. A Systematic\nReview of the Qualitative Evidence\nAndrea L. Smith1*, Stacy M. Carter1, Sally M. Dunlop2,3, Becky Freeman4,\nSimon Chapman2\n1 Centre for Values, Ethics and the Law in Medicine, School of Public Health, University of Sydney, Sydney,\nNew South Wales, Australia, 2 School of Public Health, University of Sydney, Sydney, New South Wales,\nAustralia, 3 Cancer Screening and Prevention, Cancer Institute NSW, Sydney, New South Wales, Australia,\n4 Prevention Research Collaboration, School of Public Health, University of Sydney, Sydney, New South\nWales, Australia\n* andrea.smith@sydney.edu.au\nAbstract\nBackground\nUnassisted cessation \u2013 quitting without pharmacological or professional support \u2013 is an en-\nduring phenomenon. Unassisted cessation persists even in nations advanced in tobacco\ncontrol where cessation assistance such as nicotine replacement therapy, the stop-smok-\ning medications bupropion and varenicline, and behavioural assistance are readily avail-\nable. We review the qualitative literature on the views and experiences of smokers who\nquit unassisted.\nMethod\nWe systematically searched for peer-reviewed qualitative studies reporting on smokers\nwho quit unassisted. We identified 11 studies and used a technique based on Thomas and\nHarden\u2019s method of thematic synthesis to discern key themes relating to unassisted cessa-\ntion, and to then group related themes into overarching concepts.\nFindings\nThe three concepts identified as important to smokers who quit unassisted were: motivation,\nwillpower and commitment. Motivation, although widely reported, had only one clear mean-\ning, that is \u2018the reason for quitting\u2019. Willpower was perceived to be a method of quitting, a\nstrategy to counteract cravings or urges, or a personal quality or trait fundamental to quitting\nsuccess. Commitment was equated to seriousness or resoluteness, was perceived as key\nto successful quitting, and was often used to distinguish earlier failed quit attempts from the\nfinal successful quit attempt. Commitment had different dimensions. It appeared that com-\nmitment could be tentative or provisional, and also cumulative, that is, commitment could be\nbuilt upon as the quit attempt progressed.\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 1 / 18\nOPEN ACCESS\nCitation: Smith AL, Carter SM, Dunlop SM, Freeman\nB, Chapman S (2015) The Views and Experiences of\nSmokers Who Quit Smoking Unassisted. A\nSystematic Review of the Qualitative Evidence. PLoS\nONE 10(5): e0127144. doi:10.1371/journal.\npone.0127144\nAcademic Editor: Ruth Jepson, University of\nEdinburgh, UNITED KINGDOM\nReceived: August 18, 2014\nAccepted: April 13, 2015\nPublished: May 26, 2015\nCopyright: \u00a9 2015 Smith et al. This is an open\naccess article distributed under the terms of the\nCreative Commons Attribution License, which permits\nunrestricted use, distribution, and reproduction in any\nmedium, provided the original author and source are\ncredited.\nData Availability Statement: All relevant data are\nwithin the paper.\nFunding: Funding was provided by the National\nHealth and Medical Research Council (NHMRC),\nAustralia, grant number 1024459. SMC is supported\nby NHMRC Career Development Fellowship\n1032963. www.nhmrc.gov.au.\nCompeting Interests: The authors have declared\nthat no competing interests exist.\nConclusion\nA better understanding of what motivation, willpower and commitment mean from the smok-\ner\u2019s perspective may provide new insights and direction for smoking cessation research\nand practice.\nIntroduction\nResearch into smoking cessation has achieved much. Researchers have identified numerous\nvariables related to smoking cessation and relapse, including heaviness-of-smoking, quitting\nhistory, quit intentions, quit attempts, use of assistance, socio-economic status, gender, age,\nand exposure to mass-reach interventions such as mass media campaigns, price increases or re-\ntail regulation.[1] Behavioural scientists have developed a range of health behaviour models\nand constructs relevant to smoking cessation, such as the theory of planned behaviour, social\ncognitive theory, the transtheoretical model and the health belief model.[2\u20135] These theories\nhave provided constructs to smoking cessation research such as perceived behavioural control,\nsubjective norms,[2] outcome expectations, self-regulation,[3] decisional balance,[4] perceived\nbenefits, perceived barriers and self-efficacy.[5] The knowledge generated has informed the de-\nvelopment of a range of pharmacological and behavioural smoking cessation interventions.\nYet, although these interventions are efficacious,[6\u20138] the majority of smokers who quit suc-\ncessfully do so without using them, choosing instead to quit unassisted, that is without pharma-\ncological or professional support.[9,10] Many smokers also appear to quit unplanned as a\nconsequence of serendipitous events,[11] throwing into question the predictive validity of\nsome of these cognitive models.\nThe enduring popularity of unassisted cessation persists even in nations advanced in tobac-\nco control where cessation assistance such as nicotine replacement therapy (NRT) and the\nstop-smoking medications, bupropion and varenicline, are readily available and widely pro-\nmoted.[9,10] Yet little appears to be known about this population or this self-guided route to\ncessation success. In contrast, the phenomenon of self-change (also known as natural recovery)\nis comparatively well documented in the fields of drug and alcohol addiction,[12,13] and health\nbehaviour change (for example, eating disorders, obesity and gambling).[14]\nWe recently published a systematic review of unassisted cessation in Australia.[9] We, like\nothers,[15] established that the majority of contemporary cessation research is quantitative and\nintervention focused.[16] While completing that review we determined that the available qualita-\ntive research was concerned primarily with evaluating smoker and ex-smoker perceptions of\nmass-reach interventions such as marketing or retail regulations, tax increases, graphic health\nwarnings, smoke-free legislation or intervention acceptability from the perspective of the GP,\ncurrent smoker, or third parties likely to be impacted by mass-reach interventions. Australian\nsmoking cessation research provided few insights into quitting from the perspective of the smok-\ner who quits unassisted. However our systematic review highlighted that 54% to 69% of ex-smok-\ners quit unassisted and 41% to 58% of current smokers had attempted to quit unassisted.[9]\nWe consequently became interested in what the qualitative cessation literature had to say\nabout smokers who quit unassisted. Qualitative approaches offer an opportunity to explain un-\nexpected or anomalous findings from quantitative research and to clarify relationships identi-\nfied in these studies.[17,18] By integrating individual qualitative research studies into a\nqualitative synthesis, new insights and understandings can be generated and a cumulative body\nof empirical work produced.[19] Such syntheses have proven useful to health policy and prac-\ntice.[20,21] By focusing our review on the views of smokers (i.e. on the people to whom the in-\nterventions are directed), we might start to better understand why many smokers continue to\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 2 / 18\nquit unassisted instead of using the assistance available to them. Such an understanding might\nhelp us to decide whether we should be developing better approaches to unassisted cessation or\nfocusing our attention on directing more smokers to use the efficacious pharmacological and\nprofessional behavioural support that already exists.\nIn this review, we examined the qualitative literature on smokers who quit unassisted in\norder to answer the following research questions: (1) How much and what kind of qualitative\nresearch has explored unassisted cessation? (2) What are the views and experiences of smokers\nwho quit unassisted?\nMethods\nOur qualitative synthesis is based largely on Thomas and Harden\u2019s method of thematic synthe-\nsis.[20]\nIdentification of articles for review\nWe searched MEDLINE via OvidSP, PsycINFO via OvidSP, CINAHL via EBSCO, EMBASE\nand Sociological Abstracts in September 2013 for articles reporting on views about or experi-\nences of quitting without assistance. Current best practice for identifying qualitative research\nrecommends comprehensive searches of multiple sources, balancing sensitivity against speci-\nficity to maximise number of records retrieved while reducing retrieval of records that are not\nrelevant.[22] We used empirically derived qualitative research filters where available (MED-\nLINE,[23] CINAHL[24] and PsycINFO[25]) (Table 1). We complemented this search strategy\nby conducting \u2018berry picking\u2019,[26] including grey literature searching, reference checking and\nauthor searching to uncover articles that are difficult to locate by modifying search terms and\nshifting searching strategies (Fig 1).\nRecords were included if: (1) the article reported on the views or experiences of smokers or\nex-smokers who quit; (2) the data collection and analysis methods were reported as qualitative\nby the authors; and (3) the article was in English. We set no date limits believing that early re-\nsearch was as likely to provide insightful data as more contemporary research and anticipating\nour search would produce relatively few studies. Screening was multi-levelled and moved from\nliberal to more specific. At the first level of screening (title and abstract), the focus was primari-\nly on whether the study reported on unassisted cessation as abstracts often provided limited,\nincomplete or insufficient detail to make good decisions about inclusion based on methodolog-\nical requirements.[27] We identified 3845 reports of which 11 met the inclusion criteria for the\nsynthesis (Fig 1 and Table 2).\nQuality appraisal\nWe are aware that structured approaches to quality appraisal (such as guidelines and check-\nlists) do not necessarily produce greater consistency of judgements about which papers to in-\nclude in a qualitative synthesis.[41] Concern with procedural correctness can unduly focus\nattention on the reporting of the research process and divert attention away from the analytical\ncontent of the research.[42] Authors of previous qualitative syntheses have reported a discon-\nnect between papers they believed intuitively to be well conducted research and those that\n\u2018passed\u2019 when assessed against structured quality assessment criteria.[18] We knew that many\nof the articles identified in our searches would have been published prior to the development of\nquality appraisal checklists. The retrospective application of a tool developed many years after\na study\u2019s publication appeared inappropriate given the changing norms around the reporting\nof qualitative research. We decided, like Thomas and Harden[20] and others (e.g., Atkins 2008\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 3 / 18\n[18] and Lipworth 2011[43]), to err on the side of inclusion and to judge quality on the basis of\nconceptual contribution as much as methodological rigor.\nExtracting data\nOur research questions were deliberately broad: (1) How much and what kind of qualitative re-\nsearch has explored unassisted cessation? (2) What are the views and experiences of smokers\nwho quit unassisted? We were interested not only in what the data had to say about smokers\nwho quit unassisted but also in gaining an understanding of the breadth of themes related to\nunassisted cessation. We treated as data anything reported in the results or findings sections\n(usually key concepts or findings, but also direct quotations) and, if relevant, the researchers\u2019\ninterpretations of smokers\u2019 and ex-smokers\u2019 views and experiences, as reported in the discus-\nsion or conclusion of the article.\nThematic synthesis\nData analysis involved three overlapping stages: (1) line-by-line coding of the results from the\n11 primary studies followed by (2) grouping of the line-by-line codes into descriptive themes\nthat related to (3) broader, overarching concepts.\nTable 1. Search terms.\nDatabase Search perioda Search strategy Retrieved\nMEDLINE via\nOvidSP\n1946\u2013Wk 3 Sept\n2013\n1. Smoking Cessation/ 2318\n2. (interview: or experience:).mp. or qualitative.tw. or qualitative/\n3. 1 and 2\n4. remove duplicates from 3\n5. limit 4 to (english language)\nPsycINFO via\nOvidSP\n1806\u2013Wk 3 Sept\n2013\n1. Smoking Cessation/ 1058\n2. (experiences or interview: or qualitative).tw.\n3. 1 and 2\n4. remove duplicates from 3\n5. limit 4 to (english language)\nEMBASE 1966\u2013Wk 3 Sept\n2013\n1. \u2018smoking cessation\u2019/exp OR \u2018smoking cessation program\u2019/exp 225\n2. \u2018qualitative research\u2019/exp OR \u2018qualitative research\u2019\n3. 1 and 2\n4. 1 and 2 and [English]/lim\nCINAHL via\nEBSCO\n1982\u2013Wk 3 Sept\n2013\n1. ((ti interview or ab interview) or (mh \"audiorecording\" not mm \"audiorecording\") or (ti qualitative\nstud* or ab qualitative stud*))\n176\n2. (MH \"Smoking Cessation\") OR (MH \"Smoking Cessation Programs\") OR (MH \"Smoking\nCessation Assistance (Iowa NIC)\")\n3. 1 and 2\nSociological\nAbstracts\n1952\u2013Wk 3 Sept\n2013\n1. smok* and (quit* or cessation) 62\n2. qualitative\n3. 1 and 2\nTotal retrieved 3839\na All databases were searched on 24 September 2013.\ndoi:10.1371/journal.pone.0127144.t001\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 4 / 18\nDuring the initial line-by-line coding we read and closely examined fragments of data\n(words, lines, segments and incidents) for their analytical importance. Line-by-line codes were\ncreated to reflect what was happening in these \u2018meaning units\u2019, and to show actions; for exam-\nple, what the participants were thinking, feeling or doing.[44] Next, the line-by-line codes that\nwere conceptually similar were grouped into descriptive themes and then these descriptive\nthemes were grouped into overarching concepts. Once all of the descriptive themes had been\nFig 1. Identification and screening of eligible papers for inclusion in the synthesis.\ndoi:10.1371/journal.pone.0127144.g001\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 5 / 18\nTable 2. Details of 11 studies included in the synthesis.\nSource paper Study year;\ncountry; setting\nStudy focus Participants and\nparticipant characteristics\nParticipants who quit\nUAa\nData collection and\nanalysis\nBaer et al. 1977\n[28]\nYear not stated;\nUS; community\nQuitting without\nassistance\nN = 51 (29 men, 20 women,\n2 unknown; aged 29\u201375\nyears); ex-smokers who had\nsmoked 2+ packs/day for 5\n+ years, who quit without\nprofessional direction or\nhelp, and had been an ex-\nsmoker for 2+ years\nAll quit UA Letters; convenience\nsample; content analysis\nSolheim 1989\n[29]\nYear not stated;\nUS; community\nQuitting without\nassistance\nN = 13 (7 men, 6 women;\naged 25\u201349 years); ex-\nsmokers who quit >6 months\n<2 years without assistance\nof a formalised intervention\nprogram; had previously\nsmoked 0.5 pack/per day for\n>1 year prior to quitting\nAll quit UA Semi-structured interviews;\nconvenience sample; data\nanalysis method not\nexplicitly stated\u2014included\ncoding according to\ncategories based on\ntheoretical framework and\ninterview guide\nThompson\n1995 [30]\nYear not stated;\nUS; students and\nchurch groups\nSuccessful cessation in\nwomen\nN = 10 (all women; aged 28\u2013\n48 years); women who had\nsuccessfully quit smoking;\nsmoked 10+CPD for 1+ year,\nand had quit >6 months but\n<3 years\n8/10 (80%) quit UA (1\nNRT gum; 1 hypnosis);\n5/10 had previously\nused NRT patches\nunsuccessfully\nSemi-structured interviews;\npurposive sample; data\nanalysis based on Miles\nand Huberman 1994\nMariezcurrena\n1996 [31]\nYear not stated;\nSweden;\ncommunity\nRecovery from addictions\n(tobacco, snus, drug,\nalcohol) without formal\ntreatment\nTotal N = 58; ex-\nsmokers = 38 (8 women, 30\nmen; aged 24\u201375 years); ex-\nsmokers; ceased smoking 2\n+ years with no treatment or\nintervention (including prior\ntreatment)\nAll quit UA Semi-structured interviews;\nconvenience sample; data\nanalysis based on thematic\nanalysis\nStewart 1999\n[32]b\nYear not stated;\nUS; community\nSpontaneous recovery\nfrom smoking\nN = 40 (21 females, 19\nmales; aged 30\u201380 years);\nex-smokers; tobacco free for\n5+ years without the aid of\nany self-help or formal\ntreatment programs\nAll quit UA Semi-structured interviews\n(each participant was\ninterviewed 2\u20134 times);\nconvenience sample; data\nanalysis used Spradley\u2019s\nDevelopment Research\nModel\nAbdullah and\nHo 2006 [33]\n2002; Hong Kong;\nsecondary school\nAdolescents\u2019 attitudes to\nsmoking, quitting and\nsmoking cessation\nprograms\nN = 32 (all male students in\nforms 2\u20134 equivalent to US\ngrades 8\u201310); current\nsmokers (n = 23) and ex-\nsmokers (n = 9); 26/32 had\nattempted to quit, of whom\n25/26 had attempted to quit\nUA\n25/32 (78%) quit UA 5 focus groups;\nconvenience sample;\nmodi\ufb01ed grounded theory\nNichter et al.\n2007 [34]\n2000\u20132002; US;\nWomen, Infants &\nChildren\u2019s clinics,\nfamily practice\nof\ufb01ces,\ncommunity\nFactors contributing or\nundermining quit\nattempts/harm reduction\nat onset of pregnancy\nN = 53 (all women); includes\n2 case studies of women\nwho quit UA or with minimal\nsupport (aged 20 and 31\nyears); low income women\nwho were daily smokers at\nonset of pregnancy; (16 quit;\n23 cut down; 14 continued\nsmoking)\n2 case studies of\nwomen who quit UA or\nwith minimal support\nSemi-structured interviews\n(each participant\ninterviewed x3);\nconvenience sample;\ndiscourse analysis.\n(Continued)\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 6 / 18\nsorted and grouped into concepts, the analysis became more focused. We report here only on\nthe most significant themes and concepts: those which were most original and about which the\nliterature had most to say.\nInput from teammembers\nThe first author coded the primary studies and developed the descriptive themes and concepts.\nThese were then discussed with the team as a whole and with team members individually. The\nteam members brought to the analysis a range of professional experiences and perspectives rel-\nevant to smoking cessation (including qualitative health research, tobacco control and health\nbehaviour change).\nEthics statement\nAs this was a systematic review of existing studies no ethics approval was required.\nResults\nOur findings are reported in two parts: (1) how much and what kind of qualitative research has\nexplored unassisted cessation (Tables 3 and 4); (2) what are the views and experiences of smok-\ners who quit unassisted? (Table 5).\nTable 2. (Continued)\nSource paper Study year;\ncountry; setting\nStudy focus Participants and\nparticipant characteristics\nParticipants who quit\nUAa\nData collection and\nanalysis\nOgden and Hills\n2008 [35]\nYear not stated;\nUK; community\nMechanisms in sustained\nchanges in behaviour\n(including those who lost\nweight through diet or\nexercise; and those who\nstopped smoking)\nTotal N = 34; ex-\nsmokers = 10 (6 men, 4\nwomen; aged 25\u201353 years);\nex-smokers who had been\nquit for 3+ years; (8 quit UA;\n2 with A\u2014smoking course);\nquit 3\u201320 years ago\n8/10 (80%) quit UA Semi-structured interviews;\nconvenience sample/\npossibly purposive;\nthematic analysis based on\nHuberman and Miles 1994\nBottorff et al.\n2009 [36]\n2006\u20132007;\nCanada; hospital\nantenatal units\nHow new fathers talk\nabout the experience of\ntobacco reduction or\ncessation\nN = 29 new fathers; ex-\nsmokers who quit prior to\nbirth of baby; (4/29 quit)\n4/4 quit UA (remaining\nparticipants did not\nquit)\n2x semi-structured\ninterviews with each\nparticipant; convenience\nsample; narrative analysis\nMurray et al.\n2010 [37]c\n2008; UK; general\npractice\nThe process of unplanned\nquit attempts and use of\nsupport in these attempts\nN = 20 (11 male, 4 female);\ncurrent and ex-smokers (15\nex-smokers, 5 current\nsmokers); 7/10 spontaneous\nquitters quit UA; 1/10\ndelayed quitters quit UA\n7/10 (70%) of\nspontaneous quitters\nquit UA\nSemi-structured interviews;\nconvenience sample;\nthematic analysis based on\nRitchie 1994\nMedb\u00f8 et al.\n2011 [38]\n2008; Norway;\ncommunity\nWhy older people smoke,\nwhy they quit and remain\nquit\nN = 18 elderly persons (aged\n58\u201380 years); ex-smokers\n(n = 13) and relapsed\nsmokers (n = 5) all of whom\nhad made temporary stops\n\u201cMajority of quitters\nhad stopped\nthemselves without\nmedication\u201d\nSemi-structured interviews;\nconvenience sample;\ncontent analysis (using a\nnarrative perspective)\na Studies were only included if the majority of participants quit unassisted (UA) and it was clear that the data analysis/\ufb01ndings were reporting on smokers\nwho quit unassisted.\nb Stewart\u2019s 1998 doctoral thesis,[39] on which this paper is based, was also checked for additional data.\nc Murray\u2019s 2009 doctoral thesis,[40] on which this paper is based, was also checked for additional data.\ndoi:10.1371/journal.pone.0127144.t002\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 7 / 18\nResearch question 1: How much and what kind of qualitative research\nhas explored unassisted cessation?\nThe earliest study identified was a 1977 US study investigating why smokers seeking treatment\n(psychotherapy) often fared no better than smokers who quit unassisted.[28] This was followed\nin the late 1980s and 1990s by three in-depth sociological studies (from the US and Sweden) in-\nvestigating unassisted cessation as a phenomenon in its own right,[29,31,32] and one US socio-\nlogical study in which unassisted cessation data were reported but this was not the primary\nfocus of the study.[30] Subsequent to this, no qualitative studies were identified that focused on\nunassisted cessation per se: the six post-2000 studies (from Hong Kong, US, UK, Canada and\nNorway) had as their primary focus either cessation in general[33,34,36\u201338] or health behav-\niour change.[35]\nResearch question 2: What are the views and experiences of smokers\nwho quit unassisted?\nThe full set of concepts derived from the qualitative literature is shown in Fig 2. Concepts were\ngrouped into those that included descriptive themes that have already been covered in the liter-\nature (Fig 2B, below the line) and those concepts that included descriptive themes that provid-\ned potentially new insights into unassisted cessation (Fig 2A, above the line). The existing\nquantitative smoking cessation literature has, for example, already reported on attitudes to as-\nsistance, reasons for quitting, strategies used to quit and reasons for relapsing (Fig 2B). While\nencouraged by the consistency between the qualitative and quantitative studies, our aim was to\nfocus on what the qualitative literature could report from the smokers\u2019 perspective about quit-\nting unassisted that had the potential to offer new or alternative insights into the process or ex-\nperience of unassisted cessation (Fig 2A). From this perspective the most interesting themes\nwere those that related to three concepts: (1) willpower; (2) motivation; and (3) commitment.\nFour further concepts (timing, decision-making, ownership and the perception that quitting\nunassisted was a positive phenomenon) were of interest but insufficient data were available on\nwhich to base an analysis.\nWe detail the three concepts that appeared central to smokers who self-quit (motivation,\nwillpower and commitment) in Table 5. Although these concepts appear in the scientific and\nlay literature on smoking cessation, in the following section, we explore the meaning of these\nTable 3. Overview of 11 studies synthesised.\n1970s 1980s\u20131990s 2000 onwards\nNumber of studies 1 [28] 4 [29\u201332] 6 [33\u201338]\nDisciplines Psychiatry Sociology, nursing Medicine, psychology, nursing, public health, community\nmedicine\nPopulation General, mainstream General, mainstream Speci\ufb01c populations (e.g adolescents, the elderly, new\nparents)b\nSmoking and\nquitting status\nEx-smokers, all unassisted Ex-smokers; all unassisteda Smokers, ex-smokers, relapsed smokers\nPrimary focus Unassisted cessation Unassisted cessation Cessation in general, health behaviour change in general\nAims Inform psychotherapy\nintervention design\nUnderstand the phenomenon of\nunassisted cessation\nUnderstand attitudes to cessation, reasons for quitting,\nreasons for relapse; inform intervention design\na Thompson 1995 included 1 smoker who used NRT gum and 1 who used hypnosis to quit (out of a total of 10 ex-smokers); the focus was cessation in\ngeneral rather than unassisted cessation in particular but as most participants quit unassisted the data were included in this synthesis.\nb Ogden and Hills 2008 and Murray 2010 reported on general rather than speci\ufb01c populations.\ndoi:10.1371/journal.pone.0127144.t003\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 8 / 18\nTable 4. Main themes and conclusions relating to UA quitting in the 11 studies and their contribution to the themes and concepts reported in this\nreview.\nSource paper Main themes relating to UA quitting\nreported in this paper\nMain conclusions relating to UA\nquitting reported in this paper\nConcepts and themes reported\nin this review a\nConceptual\ncontribution to\nthis review b\nBaer et al. 1977\n[28]\n(1) Pronounced differences in the\ntechniques used by participants to\nquit; (2) Challenge to self and\nmotivation appeared as a common\ncombination of techniques, as did\nmotivation and self-derogation\nMost respondents used multiple\ntechniques to quit, but there was no\nsystematic clustering of these\nmethods\nMotivation\u2014equivalent to one\u2019s\nreason for quitting; Willpower\u2014\ntautologous, ambiguous;\nWillpower\u2014a personal quality or\ntrait; Commitment\u2014being\nserious or resolute\nMedium\nSolheim 1989\n[29]\n(1) Socio-environmental factors affect\ncessation (e.g. interactions with\nfamily, friends and health\nprofessionals); (2) Thoughts pre-\nquitting primarily negative (e.g.\nassessing bene\ufb01ts and\nconsequences of smoking, or\nprocess of quitting). Thoughts post-\nquitting primarily positive; (3)\nEmotions pre-quitting included guilt,\nfear, anger, and disquiet. Emotions\npost-quitting are positive, but also\nincluded loss and resentment; (4)\nMotivational response included\ndecision-making, self-determinism,\ntaking action, messages to oneself\nSmoking cessation is a process that\nbegins before an individual stops\nsmoking; characteristic thought\nprocesses and emotions occur before\nand after cessation; actions to aid\ncessation are unique to each\nindividual; family and friends are\nin\ufb02uential; factors may be interactive,\noccur simultaneously and may be\ncumulative in their effect on the\ncessation process\nMotivation\u2014equivalent to one\u2019s\nreason for quitting\nLow\nThompson 1995\n[30]\n(1) Evolving commitment to health\nand personal growth; (2) The effect of\na smoke-free environment; (3) The\nimpact of anti-smoking education; (4)\nChanging conceptualisation of\nsmoking\nAnti-smoking education, coupled with\nsmoke-free environment, augments\nthe awareness of the effects of\nsmoking and directly impacts on\none\u2019s conceptualisation of smoking\nWillpower\u2014a strategy;\nCommitment\u2014being serious or\nresolute; Commitment\u2014 can be\ncumulative\nMedium\nMariezcurrena\n1996 [31]\n(1) Triggers precipitating or helping\nquitting; (2) Coping strategies used to\nquit; (3) Advice given by ex-smokers\nabout quitting (successful quitting\nrequired decision-making, wanting to\nstop, being determined, and belief in\noneself)\nParticipants attributed their change to\ntheir own effort; making the decision\nto stop was the most frequent trigger\nto stopping; it was related to fear,\nhealth concerns and feeling of loss of\ncontrol\nMotivation\u2014equivalent to one\u2019s\nreason for quitting; Commitment\n\u2014being serious or resolute\nLow\nStewart 1999\n[32]\n(1) Contemplation: allows for goal\nsetting; mental preparation;\nknowledge of addiction; (2) Decision\nto quit: unique decision; allows for no\nexcuses; willpower; no desire to\nsmoke; (3) Relapse: creates\nknowledge of pitfalls; less\ncommitment in previous attempt; life\nevents cannot overwhelm willpower;\nno moderation; (4) Environment:\ncontributed to smoking; motivation;\nattitude towards other smokers; (5)\nProcess of cessation: multiple\ntechniques; point of no return;\ndreams\nParticipants used multiple techniques\nto quit; most had relapsed and used\nthis as motivation to continue trying\nto quit\nMotivation\u2014equivalent to one\u2019s\nreasons for quitting; Motivation\u2014\nnot a prerequisite for quitting;\nWillpower\u2014tautologous,\nambiguous; Willpower\u2014a\nmethod; Willpower\u2014a personal\nquality or trait; Commitment\n\u2014 being serious or resolute;\nCommitment\u2014can be tentative\nor provisional; Commitment\u2014can\nbe cumulative\nHigh\nAbdullah and Ho\n2006 [33]\nThemes (importance of quitting,\nperceived barriers to quitting,\nperceived bene\ufb01ts of quitting,\nreasons to quit) were general and\nreported little speci\ufb01cally about UA\nquitting\nDecision to quit smoking was not an\nurgent or important decision; belief\nthat they could quit at any time with\nlittle dif\ufb01culty; willpower and\ndetermination can help quitting\nWillpower\u2014tautologous,\nambiguous; Willpower\u2014a\npersonal quality or trait\nLow\n(Continued)\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 9 / 18\nconcepts as defined by smokers and ex-smokers who have quit unassisted, as well as the re-\nsearchers who studied them.\nTable 4. (Continued)\nSource paper Main themes relating to UA quitting\nreported in this paper\nMain conclusions relating to UA\nquitting reported in this paper\nConcepts and themes reported\nin this review a\nConceptual\ncontribution to\nthis review b\nNichter et al.\n2007 [34]\n(1) Reasons for quitting (for the baby,\nsocial pressure, fear, appeasing\nfamily); (2) Moral authority to control\nenvironments in which smoking is\nnormative; (3) Smoking is a personal\nresponsibility and quitting is a matter\nof personal choice\nSuccessful quitters had a strong\nsense of moral identity as a mother;\nconcern for effect of smoking on\nfoetus; social networks had an\nimportant impact on woman\u2019s ability\nto quit; lack of control of environment\naffected quitting success\nCommitment\u2014being serious or\nresolute\nLow\nOgden and Hills\n2008 [35]\n(1) The role of life crises as speci\ufb01c\ntriggers to initial behaviour change;\n(2) Key sustaining conditions (a\ndisruption of function; a reduction in\nchoice; behavioural model of causes\nand solutions) which allowed the\ninitial change in behaviour to be\ntranslated into a longer term change\nin lifestyle\nIf a person no longer bene\ufb01ts from\nthe behaviour, \ufb01nds that they are\nfewer opportunities to carry out the\nunhealthy behaviour and believes\nthat the behaviour was the cause of\nhis or her problems, then an initial\nchange in behaviour is more likely to\nbe translated into a behaviour\nchange in the longer term; central to\nall themes was a process of\nreinvention and a shift toward a new\nhealthier individual\nWillpower\u2014 a method Low\nBottorff et al.\n2009 [36]\n(1) Cold turkey storyline framed\nquitting smoking as a snap decision\nwith no need for support; (2) The\n\u2018baby as the patch\u2019 storyline\ndramatised how the baby displaced\nthe need to smoke, increased\nmotivation for cessation and\nenhanced success\nCommon to all storylines was the\nmen\u2019s reluctance to rely on smoking\ncessation resources; instead self-\nreliance, willpower, autonomy \ufb01gured\nmore prominently in the narratives\nMotivation\u2014equivalent to one\u2019s\nreasons for quitting; Willpower\u2014\ntautologous, ambiguous;\nWillpower\u2014a personal quality or\ntrait\nMedium\nMurray et al.\n2010 [37]\n(1) The majority of spontaneous\nquitters had not used any support; (2)\nReasons for not using support\nincluded lack of time to access\nsupport, lack of knowledge about\nsupport available, belief general\npractitioner would not be receptive to\noffering smoking cessation support, a\nbelief they should quit on own\nThe majority of spontaneous quit\nattempts were made without the use\nof support\nCommitment\u2014being serious or\nresolute\nLow\nMedb\u00f8 et al.\n2011 [38]\n(1) Approaching a decision to stop:\nre\ufb02ection on the consequences of\nsmoking; ambivalence hardens into\nresolution and the smoker waited for\nan appropriate opportunity to quit; (2)\nThe actual stopping: many stopped\nsuddenly and unplanned as a result\nof accidental circumstances; no clear\ndecision-making, stopping without\nvisible internal struggle or resolution;\n(3) Quitting was easier than expected\nPatient preferences for quitting\nshould be explored; some smokers\nmay stop unplanned with little\nmotivation; GPs interest in the\nsmoking narrative may sometimes be\nenough to encourage cessation\nMotivation\u2014not a prerequisite for\nquitting; Commitment\u2014can be\ntentative or provisional\nLow\na Includes only the themes and concepts reported in this review, not all of the themes and concepts that were coded and mapped (see Fig 2 for full range\nof concepts).\nb Conceptual contribution to review: low: contributed to <3 themes; medium: contributed to \u00013\u20135 themes; high: contributed to \u00016 themes (see Table 5 for\nmore detail on how individual studies contributed conceptually to the review).\ndoi:10.1371/journal.pone.0127144.t004\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 10 / 18\nMotivation. Although motivation was widely reported it was difficult to discern exactly\nwhat motivation meant to the smokers as opposed to the researchers. Smokers rarely talked di-\nrectly about motivation or used the word motivation to describe their quit attempt. Yet motiva-\ntion was frequently included in the accounts researchers gave of how and why smokers quit.\nThat is, there appeared to be a disjunct between the way that researchers talked about motiva-\ntion and the way that ex-smokers understood it. On looking at the data related to motivation it\nbecame clear that when researchers talked about motivation they were in fact talking almost ex-\nclusively about reasons for quitting. Typical reasons included cost,[28] a sense of duty,\nTable 5. Concepts and descriptive themes derived from the 11 studies, with illustrative quotes.\nConcepts Descriptive themes Reported in Illustrative quotesa\nMotivation Equivalent to one\u2019s reasons for\nquitting\nBaer; Bottorff; Mariezcurrena;\nSolheim; Stewart\n\u2018I got to thinking about how much money I spend on the habit\u2019\nInformant quote in Baer; \u2018Guilt was experienced in relation to\nchildren\u2019s health\u2019 Solheim; \u2018Motivational responses are derived\nfrom the individual\u2019s need to feel competent and self-determining\nabout his or her life, and stopping the habit of smoking meets this\nneed\u2019 Solheim; \u2018I didn\u2019t like the fact that cigarettes had so much\ncontrol over where I went and what I did and who I went with . . . I\nwanted to be in control of my life\u2019 Informant 7 in Stewart\nNot a prerequisite for quitting Medb\u00f8; Stewart \u2018Our \ufb01ndings indicate that it is possible to stop smoking even at\nvery low levels of motivation\u2019 Medb\u00f8\nWillpower Tautologous, ambiguous Abdullah; Baer; Bottorff;\nStewart\n\u2018One is successful if one has willpower, one has willpower if one\nsuccessfully quits\u2019 Stewart; \u2018Willpower is the answer\u2019 Baer\nA method Ogden; Stewart \u2018The smoking group had stopped smoking either through will power\nor a smoking course\u2019 Ogden\nA strategy Thompson \u2018Six of the women in the study used sheer will-power to overcome\nthe strong urges to smoke they experienced\u2019 Thompson\nA personal quality or trait Abdullah; Baer; Bottorff;\nStewart\n\u2018Common to all the storylines was the men\u2019s reluctance to rely on\nsmoking cessation resources; instead self-reliance, willpower and\nautonomy \ufb01gured more prominently\u2019 Bottorff\nCommitment Being serious or resolute Baer; Mariezcurrena; Murray;\nNichter; Stewart; Thompson\n\u2018One of the factors that did seem to differentiate this decision was\nthat it was a \ufb01rm decision. It was often described as the \ufb01rmest\ncommitment they had ever made.\u2019 Stewart; \u2018I was thinking too that\nbefore I actually quit, that the times before, subconsciously, I really\ndidn't want to, or I wasn't taking the task seriously enough\u2019\nInformant 4 in Stewart\nCan be tentative or provisional Medb\u00f8; Stewart \u2018I always felt like it would be . . . OK, I\u2019m going to give this a valiant\nattempt and if it\u2019s not going to work, then I\u2019ll go back to smoking\nand it will be OK\u2019 Informant 12 in Stewart; \u2018I had been working on\nmy decision, you can say. I did not dread the stop because if it\nturned out to be too hard I would start smoking again\u2019 Informant\nquote in Medb\u00f8; \u2018I don\u2019t think that in previous attempts that I ever\ndecided that I would quit because I wanted to. I guess I never really\nwanted to stop\u2019 Informant 16 in Stewart\nCan be cumulative (commitment\nbuilds as the quit attempt\nprogresses)\nStewart; Thompson \u2018You can\u2019t quit [relapse] now you only have a little bit left\u2019 Informant\n17 in Stewart; \u2018\u2018I knew that if I stopped [relapsed] it would have\nkilled me. I had put too much time into this\u2019 Informant 38 in Stewart;\n\u2018The evolving commitment was also evident in words that echoed\nrepeatedly a personal determination and desire to achieve the goal\nof quitting smoking. Declarations such as \"I knew I could not turn\nback once I made my mind up\"\u2018 Thompson; \u201cIt gets to the point\nwhere you know you can do it. You\u2019ve got so much invested that if\nyou [relapsed] it\u2019d be really hard. At that point you just can\u2019t\n[relapse]\u201d Informant 17 in Stewart\na The majority of the quotes report the study authors\u2019 conclusions; the remainder are direct quotes from participants.\ndoi:10.1371/journal.pone.0127144.t005\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 11 / 18\n[28,29,31,32,36] health concerns,[28,29,31,32] feeling out of control, feeling diminished by\nbeing a smoker,[28,29,31,32] deciding the disadvantages of smoking outweighed the benefits,\n[29] or expectations that life would be better once quit.[29] We concluded the data on motiva-\ntion reported in these 11 qualitative studies added no new insights to the data on reasons for\nquitting already reported in the quantitative literature.\nSmokers used the word motivation differently: not to describe the reason they quit, but to\ndescribe what sustained them through their quit attempt. We have included these data under\nthe concept of commitment (see below). Our main conclusion about motivation is that smok-\ners and researchers appear to be using the word to denote different concepts.\nWillpower. The concept of willpower was clearly important to smokers and often used by\nresearchers to account for smokers\u2019 success or failure, but rarely examined or unpacked. Will-\npower was reported to be a method of quitting, a strategy to counteract cravings or urges\n(much as NRT or counselling is regarded as a method of quitting or a way of dealing with an\nurge to smoke)[30,32,35] or a personal quality or trait fundamental to quitting success.\n[28,32,33,36] For example, although Ogden and Hill (2008) classified their participants accord-\ning to whether they had \u2018stopped smoking through willpower or a smoking course\u2019, they gave\nno definition or explanation of what willpower was. Similarly, Thompson (1995) reported\nFig 2. Themes and concepts derived from the 11 primary studies.UA, unassisted.\ndoi:10.1371/journal.pone.0127144.g002\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 12 / 18\nmany participants used \u2018sheer willpower to overcome the strong urges to smoke\u2019, and Abdullah\nand Ho (2006) reported relapsed smokers cited \u2018willpower and determination\u2019 as key factors\nfor quitting success, but did not elaborate on what was meant by willpower. Stewart\u2019s 1999 so-\nciological study of smokers who quit unassisted[32] attempted to understand willpower from\nthe smokers\u2019 perspective, yet despite directly questioning smokers about willpower, Stewart\ncould find no agreement among smokers as to what willpower was. In summing up, Stewart\nconcluded: \u2018it is difficult to connect a successful cessation attempt with the use of willpower\nwithout creating a tautology: one is successful if one has willpower, and one has willpower if\none is successful,\u2019 capturing what is arguably still an issue in contemporary smoking cessation\nresearch.\nCommitment. Smokers\u2019 talk about commitment was nuanced and multilayered. In con-\ntrast to motivation and willpower we did not need to rely upon the researchers\u2019 interpretations\nto gain an insight into what commitment might mean to smokers. Smokers talked directly\nabout being committed. To them it meant being determined, serious or resolute. Being com-\nmitted was essential to their quitting success.[28,30\u201332,34,37] Commitment was what differen-\ntiated a serious quit attempt from previous unsuccessful quit attempts,[32] and was the\nhallmark of their final successful quit attempt.\nCommitment could also be tentative or provisional.[32,38] Medb\u00f8 (2011) reported smokers\nwho appeared keen to try to quit but were not necessarily committed to seeing the quit attempt\nthrough. It is possible a further level of commitment was being withheld, contingent perhaps\non how difficult quitting turned out to be or on how the smoker felt about being quit once they\ngot there. One of Stewart\u2019s participants illustrates the difference, \u2018OK I\u2019m going to give this a\nvaliant attempt and if it\u2019s not going to work then I\u2019ll go back to smoking and it will be OK.\u2019[32]\nThe smoker is committed to trying but not necessary committed to quitting.\nCommitment could also be cumulative. Smokers talked about a point of no return, which de-\nscribed a point in the cessation process when they had made a firm commitment to quit, they\nhad made a decision and they would not change their mind.[30,32] Smokers described this as the\npoint in time at which they believed there was too much invested to relapse now.[32]\nDiscussion\nIn this review we have synthesised the qualitative data reporting on the views and experiences\nof smokers who successfully quit unassisted (without pharmacological or professional beha-\nvioural support). The existence of only a handful of studies over more than 50 years, with no\nstudy specifically addressing unassisted cessation post-2000, indicates that up until now little\nresearch attention has been given to the lived experiences and understandings of smokers who\nsuccessfully quit unassisted. As a consequence relatively little is known about smokers\u2019 perspec-\ntives on what is the most frequently used means of quitting[10] and the way described by the\nmajority of ex-smokers as being the most \u2018helpful\u2019.[45,46] It is widely accepted that searching\nthe qualitative literature is difficult.[21,47] Although it is possible that relevant studies were\nmissed, given the comprehensiveness of our search strategy, the comparative lack of studies\nfound through searching seems likely to reflect an evidence gap, and therefore an important\narea for future research.\nThis lack of qualitative research was unexpected for two reasons. First, we were aware of a\nsmall but not unsubstantial body of quantitative evidence on smokers who quit unassisted;[48\u2013\n52] and second, in the course of our literature search we had identified a considerable number\nof qualitative studies on smoking cessation. On closer examination it became clear that few of\nthese reported specifically on smokers who quit unassisted. This supports what Kluge found in\n2009, that is, the qualitative smoking cessation research that does exist is concerned primarily\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 13 / 18\nwith evaluating the success or acceptability of smoking cessation interventions, particularly in\nvulnerable populations such as adolescents or the socially disadvantaged.[16]\nConcepts central to self-quitting\nMotivation was identified as a central concept in this review, but analysis of the studies showed\nthat motivation appeared primarily in the researchers\u2019 accounts of quitting rather than in the\nsmokers\u2019 accounts of quitting. On closer examination, the data related to motivation consisted\nalmost entirely of reasons for quitting. Within the quantitative literature on smoking cessation,\nmotivation is an established psychological construct which has been operationalised in numer-\nous studies designed to determine the role of motivation in quitting success.[53,54] Motivation\nhas been identified as critical to explaining cessation success.[55] The lack of explicit discussion\nabout motivation by smokers who quit unassisted in the studies included in this review is there-\nfore interesting. Though motivation could be inferred from the smokers\u2019 accounts; it had to be\ndone by using the variables that comprise motivation, such as reasons (motives) for quitting or\nthe pros and cons of smoking versus quitting. Given the relative lack of data, it is difficult to\nconclude whether this is (1) because smokers do not talk directly about motivation, or (2)\nwhether from the participants\u2019 perspective motivation is not the driving force behind successful\nunassisted cessation (either because another concept is more important or because too much\ntime has passed since their quit attempt for them and they have forgotten how important moti-\nvation was to them).\nFrom the studies included in this review, it appears that\u2014at least in smokers\u2019 self-under-\nstanding\u2014commitment might be more important than motivation as an explanation of suc-\ncessful unassisted cessation. The enthusiastic and explicit talk about being determined,\ncommitted, or serious suggests that this concept resonates more with smokers than the concept\nof motivation. The overlapping and at times contradictory natures of commitment and motiva-\ntion have been highlighted recently by Balmford and Borland who concluded that it may be\npossible to quit successfully while ambivalent, as long as the smoker remains committed in the\nface of ebbs and flows in motivation.[56] Further complicating the relationship, some regard\ncommitment as a component of motivation,[57] operationalizing motivation as, for example,\n\u2018determination to quit\u2019[58] or \u2018commitment to quit\u2019.[59]\nThe greater research interest in reasons for quitting or pros and cons of quitting (i.e., moti-\nvation) as opposed to commitment may be because motivation is simpler to measure, for ex-\nample by asking people to rate or rank reasons, costs or benefits. From a policy and practice\nperspective, it may also be easier to draw attention to these reasons, costs and benefits, rather\nthan engage with commitment. For example, mass media campaigns can remind smokers of\nwhy they should quit by pointing out the benefits to short-term and long-term health. However\nthis review draws attention to the importance of commitment for sustained quitting, at least\nfrom the point of view of smokers and quitters. The UK\u2019s annual Stoptober campaign in which\nsmokers committed to being smoke-free for 28 days indicates that creative approaches to ad-\ndressing commitment can be successful.[60]\nThe final concept identified, willpower, was described in terms of multiple constructs (a per-\nsonal quality or trait, a method of quitting, a strategy to counteract cravings or urges), suggest-\ning smokers and researchers may use it as a convenient or shorthand heuristic when talking\nabout or reporting on quit success. Despite this lack of clarity, the word has persisted in the\nqualitative and quantitative smoking cessation literature. It could be fruitful for future research\nto further examine the meaning of willpower, and particularly its relationship to other more\ntightly defined concepts such as self-efficacy,[61] self-regulation[62] and self-determination,\n[63] from the perspective of both researchers and smokers.\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 14 / 18\nNo matter how widely available and affordable smoking cessation assistance becomes, it is\nlikely there will always be a significant proportion of smokers who choose to quit unassisted.\n[9] It is important to understand what drives these smokers to quit this way and to better un-\nderstand their route to success. Orford and colleagues working on the UK Alcohol Treatment\nTrial made a strong case for including the client\u2019s perspective, arguing that it is wrong to as-\nsume that clients have no perspective into their own change processes, and that we should re-\nsist the dominant \u2018drug metaphor\u2019 which has adopted the model of an active professional\napplying a technique to a passive recipient.[64] McDermott and Graham also advocated for\nthe need for contemporary public health policy to ground itself in the experiences of those\nwhose lifestyles it seeks to change.[65] As the vast majority of smokers who quit successfully\ncontinue to do so without formal help, it is likely that a better understanding of this experience,\nfrom the perspective of the smokers and ex-smokers themselves, could inform more nuanced\nand effective communication and support for quitting.\nLimitations\nA potential limitation was the quality of the original articles. Our a priori decision not to assess\narticles based on the overall quality of the studies using standard guidelines or checklists meant\nthat some of the included studies failed to report, for example, the theoretical framework, the\nsampling strategy, the procedures for data analysis, or the theoretical justification for their data\nanalysis. In addition, several of the studies, especially those post-2000 reported data that were\ndescriptive rather than analytic.\nIt is possible that we, the authors of the review, were sensitised to some of the themes identi-\nfied as important due to our involvement in an ongoing grounded theory study into how and\nwhy smokers quit unassisted. Some of our interviews with ex-smokers took place at the same\ntime as this qualitative review. The influence of this overlap, however, is likely to be minimised\nby the fact that not all authors had been involved in the data collection or analysis of the\ngrounded theory study at the time of the review, and the whole authorship team were involved\nin the development and refinement of the analysis.\nAnd finally, the three concepts identified are not discrete, and are likely to overlap in many\nways. The studies identified confirmed the importance of these concepts, but did not always\nanalyse them in sufficient detail to allow us to draw firm and transferable conclusions about\ntheir meaning. We have identified the importance of these concepts: further research is needed\nto strengthen our understanding of how smokers understand and use them.\nConclusion\nOur review identifies three key concepts, motivation, willpower and commitment, circulating\nin smokers\u2019 and ex-smokers\u2019 accounts of quitting unassisted. Insufficient qualitative evidence\ncurrently exists to fully understand these concepts, but they do appear to be important in\nsmokers\u2019 and ex-smokers\u2019 accounts and so worthy of research attention. A more detailed quali-\ntative investigation of what motivation, willpower and commitment mean to smokers and ex-\nsmokers would complement the existing body of behavioural science knowledge in tobacco\ncontrol. A better understanding of these concepts from the smokers\u2019 perspective may help to\nexplain the often-puzzling popularity of quitting unassisted rather than opting to use the effica-\ncious pharmacological or professional assistance that is available. Health practitioners could\npotentially use such knowledge, in combination with what we already know from population-\nbased research into smoking cessation, to better support all smokers to quit, whether or not\nthey wish to use assistance.\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 15 / 18\nSupporting Information\nS1 PRISMA Checklist. PRISMA checklist\n(DOC)\nAcknowledgments\nAS would like to thank Dr Wendy Lipworth for advice on planning a qualitative synthesis.\nFunding for this study was provided by the National Health and Medical Research Council\n(NHMRC), Australia, grant number 1024459, www.nhmrc.gov.au. SMC is supported by\nNHMRC Career Development Fellowship 1032963.\nAuthor Contributions\nConceived and designed the experiments: SMD SC SMC BF. Performed the experiments: ALS.\nAnalyzed the data: ALS SMC SMD SC BF. Wrote the paper: ALS SMC SMD SC BF. Conceived\nand planned the review: ALS. Conducted the literature searches, identified studies to include in\nthe review, categorised the studies and extracted the data: ALS.\nReferences\n1. Scollo MM,Winstanley MH (2012) Tobacco in Australia: Facts and Issues. Melbourne: Cancer Council\nVictoria.\n2. Ajzen I (1991) The theory of planned behavior. Organizational Behavior and Human Decision Process-\nes 50: 179\u2013211.\n3. Bandura A (1986) Social foundations of thought and action: a social-cognitive theory. Englewood Cliffs,\nNew Jersey: Prentice Hall.\n4. Prochaska JO, DiClemente CC (1983) Stages and processes of self-change of smoking: toward an in-\ntegrative model of change. J Consult Clin Psychol 51: 390. PMID: 6863699\n5. Becker MH (1974) The health belief model and personal health behaviour. Thorofare, New Jersey:\nCharles B Slack.\n6. Cahill K, Stevens S, Lancaster T (2014) Pharmacological treatments for smoking cessation. JAMA\n311: 193\u2013194. doi: 10.1001/jama.2013.283787 PMID: 24399558\n7. Stead LF, Hartmann-Boyce J, Perera R, Lancaster T (2013) Telephone counselling for smoking cessa-\ntion. Cochrane Database of Systematic Reviews Issue 8. CD002850.\n8. Stead LF, Buitrago D, Preciado N, Sanchez G, Hartmann-Boyce J, Lancaster T (2013) Physician ad-\nvice for smoking cessation. Cochrane Database of Systematic Reviews Issue 5. CD000165.\n9. Smith AL, Chapman S, Dunlop SM (2013) What do we know about unassisted smoking cessation in\nAustralia? A systematic review, 2005\u20132012. Tob Control tobaccocontrol-2013-051019.\n10. Edwards SA, Bondy SJ, Callaghan RC, Mann RE (2014) Prevalence of unassisted quit attempts in pop-\nulation-based studies: a systematic review of the literature. Addictive Behaviors 39: 512\u2013519. doi: 10.\n1016/j.addbeh.2013.10.036 PMID: 24333037\n11. West R, Sohal T (2006) \u201cCatastrophic\u201d pathways to smoking cessation: findings from national survey.\nBMJ 332: 458\u2013460. PMID: 16443610\n12. SlutskeWS (2010) Why is natural recovery so common for addictive disorders? Addiction 105: 1520\u2013\n1521. doi: 10.1111/j.1360-0443.2010.03035.x PMID: 20707774\n13. Miller PM, Smith JP (2010) What do marshmallows and golf tell us about natural recovery research?\nAddiction 105: 1521\u20131522. doi: 10.1111/j.1360-0443.2010.03034.x PMID: 20707775\n14. Sobell LC (2010) The phenomenon of self-change: overview and key issues. In: Klingemann H, Sobell\nLC, editors. Promoting Self-Change From Addictive Behaviors New York: Springer. pp. 1\u201330.\n15. Chapman S, MacKenzie R (2010) The global research neglect of unassisted smoking cessation:\ncauses and consequences. PLoS Medicine 7: e1000216. doi: 10.1371/journal.pmed.1000216 PMID:\n20161722\n16. Kluge A (2009) A qualitative inquiry into smoking cessation: lessons learned from smokers. Graduate\nSchool of Emory University. Accessed 2 December 2013.\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 16 / 18\n17. Dixon-Woods M, Agarwal S, Young B, Jones D, Sutton A (2004) Integrative approaches to qualitative\nand quantitative evidence. London: Health Development Agency.\n18. Atkins S, Lewin S, Smith H, Engel M, Fretheim A, Volmink J (2008) Conducting a meta-ethnography of\nqualitative literature: lessons learnt. BMCMed Res Methodol 8: doi: 10.1186/1471-2288-8-21\n19. Barnett-Page E, Thomas J (2009) Methods for the synthesis of qualitative research: a critical review.\nBMCMed Res Methodol 9: doi: 10.1186/1471-2288-9-59\n20. Thomas J, Harden A (2008) Methods for the thematic synthesis of qualitative research in systematic re-\nviews. BMCMed Res Methodol 8: doi: 10.1186/1471-2288-8-45\n21. McDermott E, Graham H, Hamilton V (2004) Experiences of being a teenage mother in the UK: a report\nof a systematic review of qualitative studies. Glasgow: The Social and Public Health Services Unit,\nUniversity of Glasgow.\n22. Noyes J, Popay J, Pearson A, Hannes K, Booth A (2008) Cochrane handbook for systematic reviews\nof interventions. Chichester: Wiley-Blackwell.\n23. Wong SS, Wilczynski NL, Haynes RB (2004) Developing optimal search strategies for detecting clini-\ncally relevant qualitative studies in MEDLINE. Medinfo 11: 311\u2013316.\n24. Wilczynski NL, Marks S, Haynes RB (2007) Search strategies for identifying qualitative studies in\nCINAHL. Qual Health Res 17: 705\u2013710. PMID: 17478652\n25. McKibbon KA, Wilczynski NL, Haynes RB (2006) Developing optimal search strategies for retrieving\nqualitative studies in PsycINFO. Evaluation & the Health Professions 29: 440\u2013454.\n26. Sandelowski M, Barroso J (2007) Handbook for synthesizing qualitative research. New York: Springer\nPublishing Company.\n27. Saini M, Shlonsky A (2012) Systematic synthesis of qualitative research. Oxford: Oxford University\nPress.\n28. Baer PE, Foreyt JP, Wright S (1977) Self-directed termination of excessive cigarette use among un-\ntreated smokers. Journal of Behavior Therapy and Experimental Psychiatry 8: 71\u201374.\n29. Solheim K (1989) The smoking cessation process. Journal of Holistic Nursing 7: 26\u201333. PMID:\n11898217\n30. Thompson EM (1995) A descriptive study of women who have successfully quit smoking. PhD Thesis,\nGeorgia State University. Accessed 16 January 2014.\n31. Mariezcurrena R (1996) Recovery from addictions without treatment: an interview study. Cognitive Be-\nhaviour Therapy 25: 57\u201385.\n32. Stewart C (1999) Investigation of cigarette smokers who quit without treatment. Journal of Drug Issues\n29: 167\u2013186.\n33. Abdullah AS, HoWW (2006) What Chinese adolescents think about quitting smoking: a qualitative\nstudy. Substance Use & Misuse 41: 1735\u20131743.\n34. Nichter M, Nichter M, Muramoto M, Adrian S, Goldade K, Tesler L, et al. (2007) Smoking among low-in-\ncome pregnant women: an ethnographic analysis. Health Education & Behavior 34: 748\u2013764.\n35. Ogden J, Hills L (2008) Understanding sustained behavior change: The role of life crises and the pro-\ncess of reinvention. Health: An Interdisciplinary Journal for the Social Study of Health, Illness and Medi-\ncine 12: 419\u2013437.\n36. Bottorff JL, Radsma J, Kelly M, Oliffe JL (2009) Fathers' narratives of reducing and quitting smoking.\nSociology of Health & Illness 31: 185\u2013200.\n37. Murray RL, McNeill A, Lewis S, Britton J, Coleman T (2010) Unplanned attempts to quit smoking: a\nqualitative exploration. Addiction 105: 1299\u20131302. doi: 10.1111/j.1360-0443.2010.02980.x PMID:\n20642512\n38. Medb\u00f8 A, Melbye H, Rudebeck CE (2011) \"I did not intend to stop. I just could not stand cigarettes any\nmore.\" A qualitative interview study of smoking cessation among the elderly. BMC Family Practice\n12:42: doi: 10.1186/1471-2296-12-42 PMID: 21627833\n39. Stewart JC (1998) A qualitative investigation of informants who quit cigarette smoking without treat-\nment. PhD Thesis, Florida State University. Accessed 9 February 2014.\n40. Murray RL (2009) Investigating and increasing smokers' use of effective cessation support. PhD The-\nsis, University of Nottingham. Available: http://ethos.bl.uk/OrderDetails.do?did=1&uin = uk.bl.ethos.\n514878. Accessed 2013 Dec 5.\n41. Dixon-Woods M, Sutton A, Shaw R, Miller T, Smith J, Young B, et al. (2007) Appraising qualitative re-\nsearch for inclusion in systematic reviews: a quantitative and qualitative comparison of three methods.\nJ Health Serv Res Policy 12: 42\u201347. PMID: 17244397\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 17 / 18\n42. Eakin JM, Mykhalovskiy E (2003) Reframing the evaluation of qualitative health research: reflections\non a review of appraisal guidelines in the health sciences. Journal of Evaluation in Clinical Practice 9:\n187\u2013194. PMID: 12787182\n43. Lipworth WL, Hooker C, Carter SM (2011) Balance, balancing, and health. Qual Health Res 21: 714\u2013\n725. doi: 10.1177/1049732311399781 PMID: 21343435\n44. Charmaz K (2006) Constructing grounded theory: a practical guide through qualitative analysis. Thou-\nsand Oaks, California: Sage Publications.\n45. HungWT, Dunlop SM, Perez D, Cotter T (2011) Use and perceived helpfulness of smoking cessation\nmethods: results from a population survey of recent quitters. BMC Public Health 11: 592\u2013600. doi: 10.\n1186/1471-2458-11-592 PMID: 21791111\n46. GALLUP Poll Social Series: Consumption habits. Available http://www.gallup.com/poll/163763/\nsmokers-quit-tried-multiple-times.aspx. Accessed 10 October 2013.\n47. Shaw RL, Booth A, Sutton AJ, Miller T, Smith JA, Young B, et al. (2004) Finding qualitative research: an\nevaluation of search strategies. BMCMed Res Methodol 4: 5. PMID: 15070427\n48. Carey MP, Snel DL, Carey KB, Richards CS (1989) Self-initiated smoking cessation: a review of the em-\npirical literature from a stress and coping perspective. Cognitive Therapy and Research 13: 323\u2013341.\n49. Cohen S, Lichtenstein E, Prochaska JO, Rossi JS, Gritz ER, Carr CR, et al. (1989) Debunking myths\nabout self-quitting: evidence from 10 prospective studies of persons who attempt to quit smoking by\nthemselves. American Psychologist 44: 1355\u20131365. PMID: 2589730\n50. Ockene JK, Mermelstein RJ, Bonollo DS, Emmons KM, Perkins KA, Voorhees CC, et al. (2000) Re-\nlapse and maintenance issues for smoking cessation. Health Psychology 19: 17\u201331. PMID: 10709945\n51. Prochaska JO, DiClemente CC (1982) Self change processes, self efficacy and decisional balance\nacross five stages of smoking cessation. Progress in Clinical and Biological Research 156: 131\u2013140.\n52. DiClemente CC, Prochaska JO (1982) Self-change and therapy change of smoking behavior: a com-\nparison of processes of change in cessation and maintenance. Addict Behav 7: 133\u2013142. PMID:\n7102444\n53. Borland R, Yong H- H, Balmford J, Cooper J, Cummings KM, O\u2019Connor RJ, et al. (2010) Motivational\nfactors predict quit attempts but not maintenance of smoking cessation: findings from the International\nTobacco Control Four Country project. Nicotine Tob Res 12 Suppl: S4\u201311. doi: 10.1093/ntr/ntq050\nPMID: 20889479\n54. Smit ES, Fidler JA, West R (2011) The role of desire, duty and intention in predicting attempts to quit\nsmoking. Addiction 106: 844\u2013851. doi: 10.1111/j.1360-0443.2010.03317.x PMID: 21205048\n55. Nezami E, Sussman S, Pentz MA (2003) Motivation in tobacco use cessation research. Substance\nUse & Misuse 38: 25\u201350. PMID: 12602805\n56. Balmford J, Borland R, Hammond D, Cummings KM (2011) Adherence to and reasons for premature\ndiscontinuation from stop-smoking medications: data from the ITC Four-Country Survey. Nicotine Tob\nRes 13: 94\u2013102. doi: 10.1093/ntr/ntq215 PMID: 21147894\n57. Prochaska J, DiClemente C (1994) Stages of change in the modification of problem behaviors. In: Her-\nsen M, Eisler R, Miller P, editors. Progress in behavior modification Sycamore, IL: Academic\nPress. pp. 184\u2013214.\n58. Segan CJ, Borland R, Greenwood KM (2002) Do transtheoretical model measures predict the transition\nfrom preparation to action in smoking cessation? Psychol Health 17: 417\u2013435.\n59. Kahler CW, Lachance HR, Strong DR, Ramsey SE, Monti PM, Brown RA (2007) The commitment to\nquitting smoking scale: initial validation in a smoking cessation trial for heavy social drinkers. Addict\nBehav 32: 2420\u20132424. PMID: 17478057\n60. Stoptober challenge reaches new high as country\u2019s biggest mass quit attempt. Available https://www.\ngov.uk/government/news/stoptober-challenge-reaches-new-high-as-countrys-biggest-mass-quit-\nattempt Accessed 2014 Jun 26.\n61. Etter J-F, Bergman MM, Humair J-P, Perneger TV (2000) Development and validation of a scale mea-\nsuring self-efficacy of current and former smokers. Addiction 95: 901\u2013913. PMID: 10946439\n62. (2004) Handbook of self-regulation: research, theory, and applications. New York: Guilford Press.\n63. Deci EL, Ryan RM (2010) Self-determination. The Corsini Encyclopedia of Psychology Hoboken, NJ,\nUSA: JohnWiley & Sons, Inc. pp. 1\u20132.\n64. Orford J, Hodgson R, Copello A, John B, Smith M, Black R, et al. (2006) The clients\u2019 perspective on\nchange during treatment for an alcohol problem: qualitative analysis of follow-up interviews in the UK\nAlcohol Treatment Trial. Addiction 101: 60\u201368. PMID: 16393192\n65. McDermott E, GrahamH (2006) Young mothers and smoking: evidence of an evidence gap. Social Sci-\nence & Medicine 63: 1546\u20131549.\nThe Views and Experiences of SmokersWho Quit Unassisted\nPLOS ONE | DOI:10.1371/journal.pone.0127144 May 26, 2015 18 / 18\n", "identifiers": ["oai:ses.library.usyd.edu.au:2123/14523", "10.1371/journal.pone.0127144."], "language": {"code": "en", "id": 9, "name": "English"}, "publisher": "Public Library of Science", "relations": [], "repositories": [{"id": "1002", "openDoarId": 0, "name": "Sydney eScholarship", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "baseId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 1, "metadataAdded": 1466076383000, "metadataUpdated": 1591339661000, "depositedDate": 1457913600000}, "subjects": ["Article", "Publisher version"], "title": "The Views and Experiences of Smokers Who Quit Smoking Unassisted. A Systematic Review of the Qualitative Evidence.", "topics": [], "types": [], "year": 2015, "fulltextIdentifier": "https://core.ac.uk/download/pdf/41241868.pdf", "doi": "10.1371/journal.pone.0127144.", "oai": "oai:ses.library.usyd.edu.au:2123/14523", "downloadUrl": "https://core.ac.uk/download/pdf/41241868.pdf"}}
{"status": "OK", "data": {"id": "81719097", "authors": ["MingSheng Ying", "Yuan Feng", "RunYao Duan", "YangJia Li", "NengKun Yu"], "citations": [], "contributors": [], "datePublished": "2012", "fullText": "SPECIAL TOPIC\nQuantum Information\nReview\nJune 2012 Vol. 57 No. 16: 1903\u20131909\ndoi: 10.1007/s11434-012-5147-6\nc\u00a9 The Author(s) 2012. This article is published with open access at Springerlink.com csb.scichina.com www.springer.com/scp\nQuantum programming: From theories to\nimplementations\nYING MingSheng1,2*, FENG Yuan1,2, DUAN RunYao1,2, LI YangJia1,2 & YU NengKun1,2\n1Center for Quantum Computation and Intelligent Systems, Faculty of Engineering and Information Technology, University of Technology,\nSydney, NSW 2007, Australia;\n2State Key Laboratory of Intelligent Technology and Systems, Tsinghua National Laboratory for Information Science and Technology,\nDepartment of Computer Science and Technology, Tsinghua University, Beijing 100084, China\nReceived November 14, 2011; accepted February 9, 2012\nThis paper surveys the new field of programming methodology and techniques for future quantum computers, including design of\nsequential and concurrent quantum programming languages, their semantics and implementations. Several verification methods for\nquantum programs and communication protocols are also reviewed. The potential applications of programming techniques and related\nformal methods in quantum engineering are pointed out.\nquantum computation, programming languages, semantics, verification, engineered quantum systems\nCitation: Ying M S, Feng Y, Duan R Y, et al. Quantum programming: From theories to implementations. Chin Sci Bull, 2012, 57: 1903\u20131909, doi: 10.1007/s11434-012-\n5147-6\nEven though quantum hardware is still in its infancy, peo-\nple widely believe that building a large-scale and functional\nquantum computer is merely a matter of time and concen-\ntrated e\ufb00ort. As the techniques of quantum devices progress,\narchitectural studies will become critical for designing and\nimplementing bigger quantum computing systems. Indeed, a\nresearch group from top US universities, including MIT and\nUC Berkeley, has conducted their research on quantum com-\nputing architecture, with support from the DARPA Quantum\nInformation Program [1]. On the other hand, once quan-\ntum computers come into being, quantum software will play\nthe key role in exploiting the power of quantum computers.\nHowever, today\u2019s software development methodologies and\ntechniques are purely classical, and they are not suited to\nquantum computers. In the last 15 years, researchers began\nto realize the importance of quantum software. In fact, quan-\ntum software engineering was listed as a grand challenge\nin UK Grand Challenges in Computing Research, and the\ngoal of research on quantum software is explained clearly by\n*Corresponding author (email: Mingsheng.Ying@uts.edu.au; yingmsh@tsinghua.edu.cn)\nthe following excerpt from the Grand Challenges Report [2]:\nThe challenge is to rework and extend the whole of classical\nsoftware engineering into the quantum domain so that pro-\ngrammers can manipulate quantum programs with the same\nease and confidence that they manipulate today\u2019s classical\nprograms.\nIn US, EU and Canada there are already several research\nteams devoting to theoretical studies of quantum software,\nin particular quantum programming. More recently in 2010,\nIARPA in US initiated a Quantum Computer Science Pro-\ngram, and one of its major research issues is high-level quan-\ntum programming languages and quantum programming en-\nvironments, including quantum compilers. Two excellent\nsurvey papers of quantum programming are [3,4], two newer\nsurveys are [5] by one of the authors and [6], and [7, 8] also\ncontain a brief survey of quantum programming. This pa-\nper will further review the field of quantum programming,\nwith emphasis on the authors\u2019 recent work conducted at Ts-\ninghua (China) and UTS (Australia). Another quantum pro-\ngramming research group in China is led by Profs. Jiafu Xu\nand Fangmin Song at Nanjing University [9].\n1904 Ying M S, et al. Chin Sci Bull June (2012) Vol. 57 No. 16\n1 Quantum programming languages\nCurrently, quantum algorithms are expressed mainly in the\nvery low level of quantum circuits. In the history of classi-\ncal computation, however, it was realized long time ago that\nprogramming languages provide a technique that allows us to\nthink about a problem that we intend to solve in a high-level,\nconceptual way, rather than the details of implementation.\nIn order to o\ufb00er a similar technique in quantum computa-\ntion, people began to study the principles, design and seman-\ntics of quantum programming languages and to understand\nquantum computation at a higher-level [3, 4, 10]. The earli-\nest proposal for quantum programming language was made\nby Knill [11], who introduced the Quantum Random Access\nMachine (QRAM) model for quantum computing and pro-\nposed a set of conventions for writing quantum pseudocode.\nSince then, several high-level quantum programming lan-\nguages have been defined. The first quantum programming\nlanguage, QCL, was designed by O\u00a8mer [12]; he also imple-\nmented a simulator for this language. A quantum program-\nming language in the style of Dijkstra\u2019s guarded-command\nlanguage, qGCL, was proposed by Sanders and Zuliani [13].\nA quantum extension of C++ was proposed by Bettelli et\nal. [14], and implemented in the form of a C++ library. The\nfirst and now influential quantum language of the functional\nprogramming paradigm, QPL, was defined by Selinger [15]\nbased on the idea of classical control and quantum data. A\nquantum functional programming language QML with quan-\ntum control was introduced by Altenkirch and Grattage [16].\nTafliovich and Hehner [17, 18] defined a quantum extension\nof predicative programming language that supports the pro-\ngram development technique in which each programming\nstep is proven correct when it is made.\n2 Understanding quantum loop and recursive\nprograms\nIt is crucial for the design and implementation of quantum\nprogramming languages to thoroughly understand computa-\ntional mechanism of various complex quantum program con-\nstructs. Loop and recursion are powerful program constructs\nin classical computation, and understanding their behaviors\nand exploiting their powers are one of the major challenges\nin classical programming methodology. In quantum compu-\ntation, looping technique has also attracted a few researchers\u2019\nattention. For example, Bernstein and Vazirani [19] intro-\nduced some programming primitives including looping in the\ncontext of quantum Turing machines; some high-level control\nfeatures such as loop and recursion are provided in Selinger\u2019s\nQPL [15]. However, the full power of quantum loop pro-\ngrams is still to be exploited. Recently the authors initiated\na systematic study of quantum loop programs [20]. They in-\ntroduced a general scheme of quantum loop programs. The\ncomputational process of a quantum loop was carefully de-\nscribed, and then the essential di\ufb00erence between quantum\nloops and classical loops was analyzed, which mainly comes\nfrom quantum measurements in the loop guards. Further-\nmore, they introduced the notions of termination and almost\ntermination (according to termination probability) of a quan-\ntum loop. In particular, in a fixed finite-dimensional state\nspace, they found a necessary and su\ufb03cient condition under\nwhich a quantum loop program terminates on a given (mixed)\ninput state by employing Jordan normal form of complex ma-\ntrices. A similar condition is also given for almost termina-\ntion. Moreover, they proved that a small disturbance either\non the unitary transformation in the loop body or on the mea-\nsurement in the loop guard can make any quantum loop to\n(almost) terminate, provided that some obvious dimension re-\nstriction is satisfied.\nTermination analysis of quantum loops was continued\nin [21], where nondeterministic quantum programs are con-\nsidered. A nondeterministic quantum program is modelled\nby a quantum Markov decision process, which is essentially\na finite set of quantum loops with the same loop guard. A\ncharacterization of reachable space and diverging states of\na nondeterministic quantum program is presented. A zero-\none law for termination probability of the states in the reach-\nable space is proved, and an algorithm is found for checking\ntermination of nondeterministic quantum programs within a\nfixed finite-dimensional state space. Furthermore, a class of\nconcurrent quantum programs were introduced in our unpub-\nlished paper1), which are also a set of quantum loops with\nthe same loop guard, but in which these loops are executed\naccording to schedules satisfying certain fairness conditions.\nThis notion of concurrent quantum program is a natural quan-\ntum generalisation of probabilistic concurrent program de-\nfined by Hart et al. [22].\nFor further studies in this area, we still have not a\nclear understanding of quantum loops in (countably) infinite-\ndimensional state spaces. The most important problems are\nthose distinguishing quantum loops and recursions from the\nclassical ones. A popular view of quantum programming,\nproposed by Selinger [15], is often summarized by the slogan\n\u201cquantum data and classical control\u201d. This view means that\ndata manipulated by programs are quantum and thus they can\nbe in quantum superpositions. However, the control states of\nprograms are still classical in the sense that it is not allowed\nto execute a quantum superposition of two statements. The\ninvestigation of quantum loop and recursive programs men-\ntioned above is carried out in the paradigm of quantum data\nand classical control. Recently, some researchers introduced\nquantum control structures into quantum programs and initi-\nated the studies of quantum programming in the much more\ncomplicated paradigm of quantum data and quantum control\n[16]. Thereofore, an interesting issue is to extend the\n1) Yu N K, Ying M S. Termination of concurrent quantum programs. 2012\nYing M S, et al. Chin Sci Bull June (2012) Vol. 57 No. 16 1905\ninvestigation of quantum loops and recursions into the\nparadigm of quantum data and quantum control which adds a\nnew dimension of complexity in understanding behaviors of\nquantum loop and recursive programs.\n3 Semantics of quantum programming\nlanguages\nFormal semantics of a programming language gives a rig-\norous mathematical description of the meaning of this lan-\nguage, to enable a precise and deep understanding of the\nessence of the language beneath its syntax. The fact that hu-\nman intuition is much better adapted to the classical world\nthan the quantum world is one of the major reasons it is di\ufb03-\ncult to find e\ufb03cient quantum algorithms. It should also imply\nthat programmers will commit many more faults in designing\nprograms for quantum computers than programming classi-\ncal computers. Thus, it seems that giving clear and formal\nsemantics to quantum programming languages is even more\ncritical than in classical computing.\nVarious semantics of quantum programming languages\nhave been introduced. The operational or denotational se-\nmantics of some quantum programming languages were al-\nready provided when they were defined; for example qGCL,\nQPL and QML. Semantic techniques for quantum computa-\ntion have also been investigated in some abstract, language-\nindependent ways. For example, Abramsky and Coeck [23]\nproposed a category-theoretic formulation of the basic pos-\ntulates of quantum mechanics, which can be used to give\nan elegant description of quantum programs including quan-\ntum protocols such as teleportation, logic-gate teleportation,\nand entanglement swapping. An edited volume collecting the\nmain semantic techniques in quantum computation was re-\ncently published [24].\nSince it provides a goal-directed program development\nstrategy, predicate transformer semantics has a very wide\ninfluence in classical programming methodology. Two ap-\nproaches to predicate transformer semantics of quantum pro-\ngrams have been proposed in the literature. The first was\nproposed by Sanders and Zuliani [13] in designing qGCL.\nIn their approach, quantum computation is reduced to prob-\nabilistic computation by the observation (measurement) pro-\ncedure. Thus, predicate transformer semantics developed for\nprobabilistic programs can be conveniently applied to quan-\ntum programs. The second was proposed by D\u2019Hondt and\nPanagaden [25], where the notion of predicate is directly\ntaken from quantum mechanics, that is, a quantum predi-\ncate is defined to be an observable (a Hermitian operator)\nwith eigenvalues within the unit interval. In this approach,\nforward operational semantics of quantum programs are de-\nscribed by super-operators (completely positive operators),\nand a beautiful duality between state-transformer (forwards)\nand predicate-transformer (backwards) semantics is observed\nby employing the Kraus representation theorem for super-\noperators. One of the advantages of the second approach is\nthat it provides a very natural framework to model and rea-\nson about quantum programs. However, in order to further\ndevelop this approach, a major obstacle has to be overcome,\nnamely non-commutativity of quantum weakest precondi-\ntions represented by Hermitian operators. The significance\nof this problem comes from the following two observations:\n(1) Physical simultaneous verifiability of quantum weakest\npreconditions depends on commutativity between them ac-\ncording to the Heisenberg uncertainty principle; (2) Various\nlogical operations of quantum weakest preconditions will be\nneeded in reasoning about complicated quantum programs,\nbut defining these operations requires commutativity between\nthe involved quantum predicates [26]. The authors [27]\nfound several conditions under which quantum weakest pre-\nconditions commute. They [28] further systematically devel-\noped quantum predicate transformer semantics with a special\nclass of quantum predicates, namely projection operators.\nFocusing on projection predicates allowed them to use rich\nmathematical methods developed in Birkho\ufb00-von Neumann\nquantum logic [29]. In particular, the notion of commutator\nintroduced by Takeuti [30] in quantum set theory (and widely\nused in automata theory based on quantum logic [31, 32])\nhelped them to establish various healthiness conditions of\nquantum programs, e.g. termination law, conjunctivity, dis-\njunctivity and continuity. An interesting topic for further\nstudies in this direction is to establish link between the exist-\ning two approaches to quantum predicate transformer seman-\ntics, viz. probabilistic and Hermitian operator approaches,\nemploying Gleason Theorem [33].\n4 Verification of quantum programs\n4.1 Proof systems\nTo help reason about the correctness of (sequential) quan-\ntum programs, several proof systems for verification of quan-\ntum programs have been developed. Baltag and Smets [34]\npresented a dynamic logic formalism of information flows in\nquantum systems. Brunet and Jorrand [35] introduced a way\nof applying Birkho\ufb00 and von Neumann\u2019s quantum logic in\nreasoning about quantum programs. Chadha et al. [36] pro-\nposed a Hoare-style proof system for reasoning about imper-\native quantum programs in which only bounded iterations are\nallowed. Some useful proof rules were proposed in [37] for\npurely quantum programs. Finally, a full-fledged Hoare logic\nfor both partial and total correctness of quantum programs\nwas developed in [38].\n4.2 Quantum Sharir-Pnueli-Hart method\nSharir, Pnueli and Hart [39] presented a general method for\nproving properties of probabilistic programs, in which a prob-\nabilistic program is modelled by a Markov chain and an as-\nsertion on the output distribution is extended into an invari-\nant assertion on all intermediate distributions. Their method\nis essentially a probabilistic generalisation of the classical\n1906 Ying M S, et al. Chin Sci Bull June (2012) Vol. 57 No. 16\nFloyd inductive assertion method. In [40], the authors consid-\nered quantum programs modelled by quantum Markov chains\nwhich are defined by super-operators. It is shown that the\nSharir-Pnueli-Hart method can be elegantly generalised to\nquantum programs by exploiting the Schro\u00a8dinger-Heisenberg\nduality between quantum states and observables. In particu-\nlar, a completeness theorem for the Sharir-Pnueli-Hart veri-\nfication method of quantum programs is established. On the\nother hand, the Sharir-Pnueli-Hart method is in principle ef-\nfective for verifying all properties of quantum programs that\ncan be expressed in terms of Hermitian operators (observ-\nables), but it is not feasible for many practical applications\nbecause of the complicated calculation involved in the veri-\nfication. For the case of finite-dimensional state spaces, the\nauthors found a method for verification of quantum programs\nmuch simpler than the Sharir-Pnueli-Hart method by employ-\ning the matrix representation of super-operators and Jordan\ndecomposition of matrices. In particular, this method enables\nus to compute easily the average running time and even to\nanalyze some interesting long-run behaviors of quantum pro-\ngrams in a finite-dimensional state space.\n5 Implementations of quantum programming\nlanguages: Quantum compilers\nThe majority of previous research on quantum programming\nhas been devoted to designing high-level languages. From a\npractical viewpoint of implementing high-level quantum pro-\ngramming languages and designing quantum compilers, we\nshould also start with the lower levels in the hierarchy of\nprogramming languages and work upward. Some research\nin this direction has already been reported in the literature;\nfor example, Svore et al. [41] proposed a layered quantum\nsoftware architecture which is indeed a four-phase design\nflow mapping a high-level language quantum program onto\na quantum device through an intermediate quantum assem-\nbly language. Zuliani [42] designed a compiler for qGCL\nin which compilation is realized by algebraic transformation\nfrom a qGCL program into a normal form that can be di-\nrectly executed by a target machine. Nagarajan et al. [43]\ndefined a hybrid classical-quantum computer architecture \u2013\nthe Sequential Quantum Random Access Memory machine\n(SQRAM) \u2013 based on Knill\u2019s QRAM, presented a set of tem-\nplates for quantum assembly code, and developed a complier\nfor a subset of Selinger\u2019s QPL. All of these designs are based\non the popular circuit model of quantum computation. Never-\ntheless, Danos et al. [44] presented an elegant low-level lan-\nguage based on a novel and promising physical implementa-\ntion model of quantum computation, namely measurement-\nbased one-way quantum computer. One advantage of their\nlanguage is that a standardization theorem was established,\nwhich enables us to transform all programs into a standard\nform executable in a novel physical architecture allowing per-\nforming all the entanglement in the beginning.\nThe authors [45] defined a quantum extension of a classi-\ncal flowchart language. At the current stage, quantum com-\nputer hardware is still very limited and its architecture is un-\ncertain, so it seems appropriate to choose a flowchart lan-\nguage as a low-level language. The language defined in [45]\npossesses two classes of program variables: classical vari-\nables and quantum variables. These classes of variables are\nseparated, and a program state consists of a state of classi-\ncal variables and a state of quantum variables. The language\nis obtained from the classical flowchart language by adding\ntwo kinds of commands for quantum operations, namely uni-\ntary transformations and measurements. The outcome of a\nmeasurement is assigned to a classical variable. Therefore,\nmeasurement command provides a way of connecting clas-\nsical and quantum variables. Only classical variables are al-\nlowed to occur in the test condition of a conditional jump.\nThis embodies the idea of classical control and quantum data,\nand thus it is consistent with the QRAM model in [11]. To\ncompare the expressibility of the quantum flowchart language\nwith that of higher-level quantum programming languages,\nwe introduced a quantum extension of while-language. In\nparticular, we presented a structured quantum programming\ntheorem which gives a translation from quantum flowchart\nprograms to quantum while-programs. This can be seen as\na theoretical support to the structural programming paradigm\nfor quantum computers pursued by O\u00a8mer [12].\n6 Quantum process algebras\nProcess algebras (or process calculi) are popular formal mod-\nels of classical concurrent systems. They provide mathemat-\nical tools for the description of interactions, communications\nand synchronization between processes, and they also pro-\nvide formal methods for reasoning about behavior equiva-\nlence between processes by proving various algebraic laws.\nQuantum generalization of process algebras has been recently\nproposed by some researchers. To provide formal techniques\nfor modelling, analysis and verification of quantum commu-\nnication protocols, Gay and Nagarajan [46] defined the CQP\nlanguage by adding primitives for measurements and trans-\nformations of quantum states and allowing transmission of\nquantum data in the pi-calculus. To model concurrent quan-\ntum computation, Jorrand and Lalire [47] defined the QPAlg\nlanguage by adding primitives expressing unitary transforma-\ntions and quantum measurements, as well as communications\nof quantum states, to a classical process algebra, which is\nsimilar to CCS. In a series of papers [48\u201351], the authors\nproposed a model qCCS for concurrent quantum computa-\ntion, which is a natural quantum extension of classical value-\npassing CCS and can deal with input and output of quan-\ntum states, and unitary transformations and measurements\non quantum systems. In particular, the notion of probabilis-\ntic bisimulation between quantum processes was introduced,\nand their congruence properties were established. Also, a\nYing M S, et al. Chin Sci Bull June (2012) Vol. 57 No. 16 1907\ntheory of approximate strong bisimulations (strong bisimu-\nlation metrics) for quantum process algebras is proposed. As\nis well-known, a set of classical gates is universal if it can\nbe used to compute exactly an arbitrary Boolean function.\nHowever, exact universality does not make sense in quan-\ntum computation because all quantum gates form a contin-\nuum which cannot be generated by a finite set of quantum\ngates. Instead, a set of quantum gates is said to be universal\nprovided any quantum gate can be approximated to arbitrary\naccuracy by a circuit constructed from the gates in this set.\nApproximate bisimulation can be used to describe approx-\nimation between quantum processes and, in particular, im-\nplementation of a quantum process by some (usually finitely\nmany) special quantum gates. The most spectacular result in\nfault-tolerant quantum computation is the threshold theorem\nwhich means it is possible to e\ufb03ciently perform an arbitrar-\nily large quantum computation provided the noise in individ-\nual quantum gates is below a certain constant. This theorem\nconsiders only the case of quantum sequential computation.\nIts generalization in quantum concurrent computation would\nbe a great challenge. The notions of approximate bisimula-\ntion and noisy channel [52] will provide us with a formal tool\nfor observing robustness of quantum concurrent computation\nagainst inaccuracy in the implementation of its elementary\ngates, and we guess that it can be used to establish a concur-\nrent generalization of the (fault-tolerance) threshold theorem.\nThe role of entanglement in quantum computation and in-\nformation is a very interesting and important problem. It has\nbeen carefully analyzed by several researchers in the frame-\nwork of sequential computation (see for example [53]). It\nseems that entanglement is more essential in quantum con-\ncurrent computation than in sequential quantum computation.\nThe algebra of quantum processes developed in [48, 49, 51]\nshould provide us with a natural and cleaned up formal frame-\nwork for analyzing the role of entanglement in concurrent\nquantum computation.\nQuantum information science is usually divided into two\nsubareas: quantum computation and quantum communica-\ntion. Quantum computation o\ufb00ers the possibility of consider-\nable speed over classical computation by exploring the power\nof superposition of quantum states, and communication pro-\ntocols are proposed by employing quantum mechanical prin-\nciples (in particular, no-cloning property and entanglement),\nwhich are provable secure. Studies of quantum process al-\ngebras and distributed quantum computation [50] allow us to\nglue the two subareas of quantum information science.\n7 Applications of programming methodology\nto quantum engineering\nAs pointed out by Dowling and Milburn [54], we are cur-\nrently in the midst of a second quantum revolution: tran-\nsition from quantum theory to quantum engineering. The\naim of quantum theory is to find fundamental rules that gov-\nern the physical systems already existing in the nature. In-\nstead, quantum engineering intends to design and implement\nnew systems (machines, devices, etc.) that do not exist be-\nfore to accomplish some desirable tasks, based on quantum\ntheory. Experiences in today\u2019s engineering indicate that it\nis not guaranteed that a human designer completely under-\nstands the behaviours of the systems she/he designed, and a\nbug in her/his design may cause some serious problems and\neven disasters. Therefore, correctness, safety and reliability\nof complex engineering systems have attracted wide attention\nand have been systematically studied in various engineering\nfields. As pointed out in Section 1, human intuition is much\nbetter adapted to the classical world than the quantum world.\nThis also implies that human engineers will commit many\nmore faults in designing and implementing complex quantum\nengineering systems. Thus, correctness, safety and reliability\nproblem will be even more serious in quantum engineering\nthan in today\u2019s engineering.\nIn the last four decades, computer scientists have systemat-\nically developed theories of correctness and safety as well as\nmethodologies, techniques and even automatic tools for cor-\nrectness and safety verification of computer systems. In par-\nticular, model-checking [55] is an e\ufb00ective automated tech-\nnique that checks whether a formal (temporal logic) property\nis satisfied in a formal model of a system. It has become\none of the dominant techniques for verification of computer\nsystems nearly 30 years after its inception. Many industrial-\nstrength systems have been verified by employing model-\nchecking techniques.\nA question then naturally arises: is it possible and how\nto use model-checking techniques to verify correctness and\nsafety of quantum engineering systems? It seems that the cur-\nrent model-checking techniques cannot be directly applied to\nquantum systems because of some essential di\ufb00erences be-\ntween the classical world and the quantum world. To develop\nmodel-checking techniques for quantum systems, at least the\nfollowing two problems must be addressed:\n(1) The classical system modelling method cannot be used\nto describe the behaviors of quantum systems, and the classi-\ncal specification language is not suited to formalize the prop-\nerties of quantum systems to be checked. Therefore, we need\nto carefully and clearly define a conceptual framework in\nwhich we can properly reason about quantum systems, in-\ncluding formal models of quantum systems and formal de-\nscription of temporal properties of quantum systems.\n(2) The state spaces of the classical systems that model-\nchecking techniques can be applied to are usually finite or\ncountably infinite. However, the state spaces of quantum\nsystems are inherently continuous even when they are finite-\ndimensional. In order to check quantum systems, we have\nto exploit some deep mathematical properties so that it suf-\nfices to examine only a finite number of (or at most countably\ninfinitely many) representative elements, e.g. those in an or-\nthonormal basis, of their state spaces.\nThere have been quite a few papers devoted to model-\n1908 Ying M S, et al. Chin Sci Bull June (2012) Vol. 57 No. 16\nchecking quantum systems, with the target of checking quan-\ntum communication protocols. For example, Gay et al. [56]\nused the probabilistic model-checker PRISM [57] to ver-\nify the correctness of several quantum protocols including\nBB84 [58]. Furthermore, they [59, 60] developed an auto-\nmatic tool QMC (Quantum Model-Checker). QMC uses the\nstabilizer formalism [61] for the modelling of systems, and\nthe properties to be checked by QMC are expressed in Bal-\ntazar, Chadha, Mateus and Sernadas quantum computation\ntree logic [62, 63]. The purpose of [64] is to develop model-\nchecking techniques that can be used not only for quantum\ncommunication protocols but also for other quantum engi-\nneering systems. The authors defined a mathematical frame-\nwork in which we can examine various linear-time properties\nof quantum systems, such as safety and liveness properties.\nAlso, they discovered an algorithm for checking invariants of\nquantum systems.\n8 Conclusions\nProgramming research for classical computers is a huge\narea, including numerous subareas such as programming lan-\nguages and semantics, programming systems and tools, spec-\nification, testing and verification, programming paradigms,\nto name just a few. A full-fledged discipline of quantum pro-\ngramming will be at least as large as classical programming.\nIt should be emphasised that the subject of quantum program-\nming methodology is not a simple and straightforward gen-\neralisation of its classical counterpart. Some completely new\nphenomena arise in the quantum case. So we have to ad-\ndress some major problems that would not arise in the realm\nof classical and probabilistic programming. These problems\ncome from the w eird nature of quantum systems. For ex-\nample, no-cloning of quantum data requires us to distinguish\nmuch more carefully call-by-value and call-by-name seman-\ntics of programming languages. It also means that the typ-\ning systems of quantum programming languages are essen-\ntially di\ufb00erent from those of classical computing [46]. Non-\ncommutativity of observables (not simultaneous verifiability\naccording to the Heisenberg uncertainty principle) is another\ntypical feature of quantum systems. It was observed by the\nauthors [27] that non-commutativity is a major obstacle in\ndeveloping predicate transformer semantics of quantum pro-\ngramming languages.\nQuantum programming methodology is still at the very\nearly stage, and its knowledge base is highly fragmentary and\ndisconnected. Certainly, much of the major problems is left\nunsolved. In particular, the studies of quantum programming\nreviewed in this paper is solely based on the circuit model of\nquantum computation. It is interesting to investigate princi-\nples and semantics of programming languages for other mod-\nels of quantum computation, such as topological, adiabatic,\nand measurement-based quantum computation. For example,\nthe mathematical description of topological quantum compu-\ntation [65] is given in terms of topological quantum field the-\nory, knot theory and lower-dimensional topology. The studies\nof programming methodology for topological quantum com-\nputers (e.g. fixed point semantics of recursive programs) will\neven bring novel and exciting open problems to these main-\nstream areas of mathematics.\nThis work was partly supported by the Australian Research Council\n(DP110103473) and the National Natural Science Foundation of China\n(60736011).\n1 Copsey D, Oskin M, Impens F, et al. Toward a scalable, silicon-based\nquantum computing architecture. IEEE J Select Topics Quant Electron,\n2003, 9: 1552\u20131569\n2 Hoare T, Milner R. Grand Challenges in Computing Research. Swin-\ndon, UK: The British Computer Society, 2004\n3 Gay S J. Quantum programming languages: Survey and bibliography.\nMath Struct Comput Sci, 2006, 16: 581\u2013600\n4 Selinger P. A brief survey of quantum programming languages. In: Pro-\nceedings of the 7th International Symposium on Functional and Logic\nProgramming, LNCS 2998, 2004. 1\u20136\n5 Ying M S. Foundations of quantum programming. In: Proceedings of\nthe 8th Asian Symposium on Programming Languages and Systems\n(APLAS 2010), LNCS 6461, 2010. 16\u201320\n6 Ru\u00a8diger R. Quantum programming languages: An introdutionary\noverview. Comput J, 2007, 50: 134\u2013150\n7 Miszczak J A. Models of quantum computation and quantum program-\nming languages. Bull Polish Acad Sci: Tech Sci, 2011, 59: 305\u2013324\n8 Ying M S. Quantum computation, quantum theory and AI. Artif Intell,\n2010, 174: 162\u2013176\n9 Xu J F, Song F M. Quantum programming languages: A tentative study.\nSci China Ser F-Inf Sci, 2005, 51: 623\u2013637\n10 Abramsky S. High-level methods for quantum computation and infor-\nmation. In: Proceedings of the 19th Annual IEEE Symposium on Logic\nin Computer Science (LICS 2004), 2004. 410\u2013414\n11 Knill E H. Conventions for quantum pseudocode. Technical Report\nLAUR-96-2724, Los Alamos National Laboratory, 1996\n12 O\u00a8mer B. Structural quantum programming. Ph.D. Thesis. Vienna: Vi-\nenna University of Technology, 2003\n13 Sanders J W, Zuliani P. Quantum programming. In: Proceedings of\nMathematics of Program Construction, LNCS 1837, 2000. 88\u201399\n14 Bettelli S, Calarco T, Serafini L. Toward an architecture for quantum\nprogramming. Eur Phys J D, 2003, 25: 181\u2013200\n15 Selinger P. Towards a quantum programming language. Math Struct\nComput Sci, 2004, 14: 527\u2013586\n16 Altenkirch T, Grattage J. A functional quantum programming language.\nIn: Proceedings of the 20th Annual IEEE Symposium on Logic in Com-\nputer Science (LICS), 2005. 249\u2013258\n17 Tafliovich A, Hehner E C R. Quantum predicative programming. In:\nProceedings of the 8th International Conference on Mathematics of Pro-\ngram Construction (MPC), LNCS 4014, 2008. 433\u2013454\n18 Tafliovich A, Hehner E C R. Programming with quantum communica-\ntion. Electron Notes Theoret Comput Sci, 2009, 253: 99\u2013118\n19 Bernstein E, Vazirani U. Quantum complexity theory. SIAM J Comput,\n1997, 26: 1411\u20131473\n20 Ying M S, Feng Y. Quantum loop programs. Acta Inform, 2010, 47:\n221\u2013250\n21 Li Y J, Yu N K, Ying M S. Termination of nondeterministic quantum\nprograms. arXiv: 1201.0891v1\n22 Hart S, Sharir M, Pnueli A. Termination of probabilistic concurrent pro-\ngrams. ACM Trans Progr Lang Syst, 1983, 5: 356\u2013380\nYing M S, et al. Chin Sci Bull June (2012) Vol. 57 No. 16 1909\n23 Abramsky S, Coecke B. A categorical semantics of quantum protocols.\nIn: Proceedings of the 19th IEEE Symposium on Logic in Computer\nScience (LICS), 2004. 415\u2013425\n24 Mackie I, Gay S. Semantic Techniques in Quantum Computation. New\nYork: Cambridge University Press, 2010\n25 D\u2019Hondt E, Panangaden P. Quantum weakest preconditions. Math\nStruct Comput Sci, 2006, 16: 429\u2013451\n26 Varadarajan V S. Geometry of Quantum Theory. New York: Springer-\nVerlag, 1985\n27 Ying M S, Chen J X, Feng Y, et al. Commutativity of quantum weakest\npreconditions. Inform Proc Lett, 2007, 104: 152\u2013158\n28 Ying M S, Duan R Y, Feng Y, et al. Predicate Transformer Semantics\nof Quantum Programs. Semantic Techniques in Quantum Computation.\nNew York: Cambridge University Press, 2010. 311\u2013360\n29 Birkho\ufb00 G, von Neumann J. The logic of quantum mechanics. Ann\nMath, 1936, 37: 823\u2013843\n30 Takeuti G. Quantum Set Theory. Current Issues in Quantum Logics.\nNew York: Plenum Press, 1981. 303\u2013322\n31 Ying M S. A theory of computation based on quantum logic (I). Theor\nComput Sci, 2005, 344: 134\u2013207\n32 Ying M S. Quantum Logic and Automata Theory. Handbook of Quan-\ntum Logic and Quantum Structures. London: Elsevier, 2007. 619\u2013\n754\n33 Gleason A G. Measures on the closed subspaces of a Hilbert space. J\nMath Mech, 1957, 6: 885\u2013893\n34 Baltag A, Smets S. LQP: The dynamic logic of quantum information.\nMath Struct Comput Sci, 2006, 16: 491\u2013525\n35 Brunet O, Jorrand P. Dynamic quantum logic for quantum programs.\nInt J Quant Inform, 2004, 2: 45\u201354\n36 Chadha R, Mateus P, Sernadas A. Reasoning about imperative quantum\nprograms. Elect Notes Theor Comput Sci, 2006, 158: 19\u201339\n37 Feng Y, Duan R Y, Ji Z F, et al. Proof rules for the correctness of quan-\ntum programs. Theor Comput Sci, 2007, 386: 151\u2013166\n38 Ying M S. Floyd-Hoare logic for quantum programs. ACM Trans Progr\nLang Syst, 2011, 33: 19\n39 Sharir M, Pnueli A, Hart S. Verification of probabilistic programs.\nSIAM J Comput, 1984, 13: 292\u2013314\n40 Ying M S, Yu N K, Feng Y, et al. Verification of quantum programs.\narXiv: 1106.4063, 2011\n41 Svore K M, Aho A V, Cross A W, et al. A layered software architecture\nfor quantum computing design tools. IEEE Comput, 2006, 39: 74\u201383\n42 Zuliani P. Compiling quantum programs. Acta Inform, 2005, 41: 435\u2013\n473\n43 Nagarajan R, Papanikolaou N, Williams D. Simulating and compiling\ncode for the sequential quantum random access machine. Elect Notes\nTheor Comput Sci, 2007, 107: 101124\n44 Danos V, Kashefi E, Panangaden P. The measurement calculus. J ACM,\n2007, 54: 8\n45 Ying M S, Feng Y. A flowchart language for quantum programming.\nIEEE Trans Software Eng, 2011, 37: 466\u2013485\n46 Gay S J, Nagarajan R. Communicating quantum processes. In: Pro-\nceedings of the 32nd ACM Symposium on Principles of Programming\nLanguages (POPL05), 2005\n47 Jorrand P, Lalire M. Toward a quantum process algebra. In: Proceed-\nings of the 1st ACM Conference on Computing Frontiers, 2004\n48 Feng Y, Duan R Y, Ji Z F, et al. Probabilistic bisimulations for quantum\nprocesses. Inform Comput, 2007, 205: 1608\u20131639\n49 Ying M S, Feng Y, Duan R Y, et al. An algebra of quantum processes.\nACM Trans Comput Logic, 2009, 10: 19\n50 Ying M S, Feng Y. An algebraic language for distributed quantum com-\nputing. IEEE Trans Comput, 2009, 58: 728\u2013743\n51 Feng F, Duan R Y, Ying M S. Bisimulations for quantum processes. In:\nProceedings of the 38th ACM Symposium on Principles of Program-\nming Languages (POPL11), 2011. 523\u2013534\n52 Ying M S. Pi-calculus with noisy channels. Acta Inform, 2005, 41:\n525\u2013593\n53 Jozsa R, Linden N. On the role of entanglement in quantum-\ncomputational speed-up. Proc Royal Soc London Ser A-Math Phys Eng\nSci, 2003, 459: 2011\u20132032\n54 Dowling J P, Milburn G J. Quantum technology: The second quantum\nrevolution. Phil Trans Royal Soc London A, 2003, 361: 1655\u20131674\n55 Baier C, Katoen J P. Principles of Model Checking. Cambridge, Mas-\nsachusetts: MIT Press, 2008\n56 Gay S J, Nagarajan R, Papanikolaou N. Probabilistic model-checking of\nquantum protocols. In: Proceedings of the 2nd International Workshop\non Developments in Computational Models (DCM06), 2006\n57 Kwiatkowska M, Norman G, Parker P. Probabilistic symbolic model-\nchecking with PRISM: A hybrid approach. Int J Software Tools Tech\nTransfer, 2004, 6: 128\u2013142\n58 Bennett C H, Brassard G. Quantum cryptography: Public key distribu-\ntion and coin tossing. In: Proceedings of International Conference on\nComputers, Systems and Signal Processing, 1984\n59 Gay S J, Nagarajan R, Panaikolaou N. QMC: A model checker for quan-\ntum systems. In: Proceedings of the 20th International Conference on\nAutomated Verification (CAV08), LNCS 5123, 2008. 543\u2013547\n60 Papanikolaou N K. Model checking quantum protocols. Ph.D. Thesis.\nCoventry, UK: University of Warwick, 2008\n61 Nielsen M A, Chuang I L. Quantum Computation and Quantum Infor-\nmation. Cambridge: Cambridge University Press, 2000\n62 Baltazar P, Chadha R, Mateus P. Quantum computation tree logic\u2014\nmodel checking and complete calculus. Int J Quant Inform, 2008, 6:\n219\u2013236\n63 Baltazar P, Chadha R, Mateus P, et al. Towards model-checking quan-\ntum security protocols. In: Proceedings of the 1st Workshop on Quan-\ntum Security (QSec07), 2007\n64 Ying M S, Li Y J, Yu N K, et al. Model-checking linear-time properties\nof quantum systems. arXiv: 1101.0303, 2011\n65 Preskill J. Topological quantum computation. In: Lecture Notes for\nPhysics 219: Quantum Computation. Pasadena: California Institute of\nTechnology, 2004\nOpen Access This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction\nin any medium, provided the original author(s) and source are credited.\n", "identifiers": ["10.1007/s11434-012-5147-6"], "publisher": "Springer Nature", "relations": ["http://dx.doi.org/10.1007/s11434-012-5147-6"], "repositories": [{"id": "2612", "openDoarId": 0, "name": "Springer - Publisher Connector", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 1, "metadataAdded": 1493832269000, "metadataUpdated": 1494856206000, "depositedDate": null}, "subjects": ["journal-article"], "title": "Quantum programming: From theories to implementations", "topics": [], "types": [], "year": 2012, "fulltextIdentifier": "https://core.ac.uk/download/pdf/81719097.pdf", "doi": "10.1007/s11434-012-5147-6", "downloadUrl": "https://core.ac.uk/download/pdf/81719097.pdf"}}
{"status": "Not found", "data": null}
{"status": "OK", "data": {"id": "81091491", "authors": ["Elaine Kingwell", "Feng Zhu", "Ruth Ann Marrie", "John D. Fisk", "Christina Wolfson", "Sharon Warren", "Joanne Profetto-McGrath", "Lawrence W. Svenson", "Nathalie Jette", "Virender Bhan", "B. Nancy Yu", "Lawrence Elliott", "Helen Tremlett"], "citations": [], "contributors": [], "datePublished": "2015", "fullText": "ORIGINAL COMMUNICATION\nHigh incidence and increasing prevalence of multiple sclerosis\nin British Columbia, Canada: findings from over two decades\n(1991\u20132010)\nElaine Kingwell1 \u2022 Feng Zhu1 \u2022 Ruth Ann Marrie2 \u2022 John D. Fisk3 \u2022\nChristina Wolfson4 \u2022 Sharon Warren5 \u2022 Joanne Profetto-McGrath6 \u2022\nLawrence W. Svenson7 \u2022 Nathalie Jette8 \u2022 Virender Bhan9 \u2022 B. Nancy Yu10 \u2022\nLawrence Elliott11 \u2022 Helen Tremlett1\nReceived: 23 April 2015 / Revised: 25 June 2015 / Accepted: 29 June 2015 / Published online: 24 July 2015\n\u0002 The Author(s) 2015. This article is published with open access at Springerlink.com\nAbstract Province-wide population-based administrative\nhealth data from British Columbia (BC), Canada (popula-\ntion: approximately 4.5 million) were used to estimate the\nincidence and prevalence of multiple sclerosis (MS) and\nexamine potential trends over time. All BC residents\nmeeting validated health administrative case definitions for\nMS were identified using hospital, physician, death, and\nhealth registration files. Estimates of annual prevalence\n(1991\u20132008), and incidence (1996\u20132008; allowing a 5-year\ndisease-free run-in period) were age and sex standardized\nto the 2001 Canadian population. Changes over time in\nincidence, prevalence and sex ratios were examined using\nPoisson and log-binomial regression. The incidence rate\nwas stable [average: 7.8/100,000 (95 % CI 7.6, 8.1)], while\nthe female: male ratio decreased (p = 0.045) but remained\nat or above 2 for all years (average 2.8:1). From\n1991\u20132008, MS prevalence increased by 4.7 % on average\nper year (p\\ 0.001) from 78.8/100,000 (95 % CI 75.7,\n82.0) to 179.9/100,000 (95 % CI 176.0, 183.8), the sex\nprevalence ratio increased from 2.27 to 2.78 (p\\ 0.001)\nand the peak prevalence age range increased from 45\u201349 to\n55\u201359 years. MS incidence and prevalence in BC are\namong the highest in the world. Neither the incidence nor\nthe incidence sex ratio increased over time. However, the\nprevalence and prevalence sex ratio increased significantly\nduring the 18-year period, which may be explained by the\nincreased peak prevalence age of MS, longer survival with\nMS and the greater life expectancy of women compared to\nmen.\nKeywords Multiple sclerosis \u0002 Incidence \u0002 Prevalence \u0002\nEpidemiology \u0002 Sex ratio \u0002 Administrative health data\nFor the CIHR Team in the Epidemiology and Impact of Comorbidity\non MS. Members in this team is listed in acknowledgments.\n& Elaine Kingwell\nelainejk@mail.ubc.ca\n1 Faculty of Medicine (Neurology), UBC Hospital, University\nof British Columbia, 2211 Wesbrook Mall, Vancouver,\nBC V6T 2B5, Canada\n2 Departments of Internal Medicine & Community Health\nSciences, University of Manitoba, Winnipeg, MB, Canada\n3 Departments of Psychiatry and Medicine, Dalhousie\nUniversity, Halifax, NS, Canada\n4 Department of Medicine and of Epidemiology, Biostatistics\nand Occupational Health, McGill University, Montreal, QC,\nCanada\n5 Faculty of Rehabilitation Medicine, University of Alberta,\nEdmonton, AB, Canada\n6 Faculty of Nursing, University of Alberta, Edmonton, AB,\nCanada\n7 Surveillance and Assessment Branch, Alberta Health,\nGovernment of Alberta, Edmonton, AB, Canada\n8 Department of Clinical Neurosciences and Community\nHealth Sciences, Hotchkiss Brain Institute and O\u2019Brien\nInstitute for Population Health, University of Calgary,\nCalgary, AB, Canada\n9 Department of Medicine (Neurology), Dalhousie University,\nHalifax, NS, Canada\n10 Department of Community Health Sciences, University of\nManitoba, Winnipeg, MB, Canada\n11 Departments of Community Health Sciences and Medical\nMicrobiology, University of Manitoba, Winnipeg, MB,\nCanada\n123\nJ Neurol (2015) 262:2352\u20132363\nDOI 10.1007/s00415-015-7842-0\nBackground\nMultiple sclerosis (MS), a chronic, debilitating disease of\nthe central nervous system, is the leading cause of non-\ntraumatic disability in young adults [1]. It is estimated that\nmore than two million people live with this disease world-\nwide [1], although the incidence and prevalence vary geo-\ngraphically [2\u20134]. Furthermore, reports of recent increases\nin the incidence and prevalence, and in the ratio of women to\nmen with MS, have been inconsistent across regions [5\u20137].\nThe need for current and reliable estimates of MS\nincidence and prevalence has been highlighted as a public\nhealth and research priority, essential to support the plan-\nning and prioritization of health care services and to reduce\nthe overall burden of chronic disease [1, 8, 9].\nValid and reliable methods are required when estimating\nincidence and prevalence so that regional estimates can be\ncompared. Validated case definitions that use population-\nbased administrative data offer this opportunity where such\ndata are available. Canada has a universal publicly funded\nhealth care system and, in its western-most province British\nColumbia (BC), health claims data for the entire population\nare captured. No estimates of the prevalence of MS in BC\nusing population-based linked health administrative data\nhave been reported, and the incidence of MS in BC has not\nbeen previously estimated by any method.\nWe aimed to estimate the incidence and prevalence of\nMS in BC, Canada using previously validated case defi-\nnitions of MS [10, 11] based on health administrative data.\nAlso, we described the demographics of the incident and\nprevalent cases and temporal changes in their characteris-\ntics including the sex ratio.\nMethods\nBritish Columbia is situated on the west coast of Canada.\nIts population of 4.5 million people represents 13 % of the\nCanadian population. The publicly funded provincial\nhealth care programme is compulsory for residents; a\nlifelong unique personal health care number is assigned\nand is linked through provincial administrative databases to\nall hospital admissions, physician visits, prescription dis-\npensations, births, deaths, and health care plan registration\nand cancellation dates.\nAnonymized linked BC health administrative data files\nused in this study included the Hospital Admission and\nDischarge database (hospital admission dates and diagnosis\ncodes) [12], and the Medical Services Plan Billing\n(physician visits and billing diagnosis codes) [13], These\ndatabases store data on physician billing or medical ser-\nvices claims (\u2018claims\u2019) that have been submitted for pay-\nment, including the type of service provided, when and to\nwhom the service was provided, and the diagnoses related\nto the physician visit or hospital admission (coded via\nInternational Classification of Disease (ICD-9 or ICD-10-\nCA)). PharmaNet (dispensed prescriptions coded by Health\nCanada\u2019s Drug Identification Numbers) [14], and Vital\nStatistics (death dates) [15] were also accessed. Registra-\ntion Premium and Billing files [16] provided demographic\ndata: registration dates in the provincial health care plan\nconfirmed residency in BC; socioeconomic status (SES)\nwas expressed as quintiles of average neighbourhood\nincome based on regional income levels (prepared by\nStatistics Canada using postal codes [17]).\nWe utilized a previously linked data platform which\nincluded all residents of BC with C3 MS-related claims.\nLinked data were available from 1986, apart from ICD\ncodes from physician visits which were available starting\nin 1991. Prescription data were accessed for descriptive\npurposes only, and were available starting in 1996. Follow-\nup continued to the end of 2010 for the majority of those in\nthe dataset, with the remainder followed to the end of 2008.\nMS cases were identified using administrative case\ndefinitions, which have previously been validated in two\nCanadian provinces (Manitoba and Nova Scotia) [10, 11],\nand are based on hospital and physician-derived diagnostic\ncodes. The primary case definition used was C7 hospital or\nphysician claims specifically for MS for people who were\nresident in BC for[3 years following their first demyeli-\nnating disease (\u2018MS-related\u2019) claim (i.e. a claim for MS,\noptic neuritis, acute transverse myelitis, acute disseminated\nencephalomyelitis, demyelinating disease of the CNS\nunspecified, other acute disseminated demyelination, or\nneuromyelitis optica), and C3 MS claims for those\nwith B3 years of residency. When validated against the\nclinical MS definition, this algorithm was found to provide\nthe best balance of sensitivity and specificity compared to a\nseries of alternative administrative case definitions. For the\nvalidation population (Nova Scotia, Canada), all of whom\nhad at least one claim for a demyelinating disease, this\ndefinition had a sensitivity of 88 % and specificity of 68 %;\nthe specificity would, however, be substantially higher in\nthe general population given that[99.9 % of individuals\nhave no demyelinating claims [11, 18]. A second case\ndefinition was also used which required C3 MS claims\nirrespective of the cumulative residency in BC, for which\nprevious validation has demonstrated greater sensitivity\n(95 %) but less specificity (48 %) among those with at\nleast one demyelinating claim [11].\nPoint prevalence was estimated annually on July 1st,\nfrom 1991 to 2008, and incidence estimates were generated\nfrom 1996 to 2008, with inclusion of claims up to 2010.\nBoth incidence and prevalence were calculated per 100,000\npeople using the BC mid-year population and were age and\nsex standardized to the 2001 Canadian population, for\nJ Neurol (2015) 262:2352\u20132363 2353\n123\nconsistency to prior Canadian work [10, 11, 19]. The 95 %\nconfidence intervals (CI) were calculated based on the\nGamma distribution [20]. Incidence estimates began in\n1996 because at least 5 years residency with no MS-related\nclaim was required to meet the incident case definition.\nOnce this definition was met, the date of the first MS-related\nclaim was considered the incidence date of MS diagnosis.\nIndividuals who immigrated to the province after the study\nstart were followed from the date of their first registration in\nthe universal BC health care plan; as with cases that were\nresident from study start, a 5-year residency with no MS-\nrelated claim was required to be counted as an incident case.\nDescription of the incident cases included sex, age, SES,\ntime to meet the case definition, and dispensation of a MS\ndisease-modifying drug (DMD) within 3 years of the\nincident claim. This time window was chosen because the\ndiagnosis date falls within 3 years of the incident claim for\napproximately 75 % of MS cases [10]. Cases that were\nprevalent on July 1st 2008 were described by sex and age,\nSES and history of a DMD prescription (including beta\ninterferon-1a, beta interferon-1b, glatiramer acetate and\nnatalizumab). The distribution of cases across the SES\nquintiles was compared to the expected (even) distribution.\nChanges in the incidence rate and prevalence over the\nobservation period were investigated using Poisson (with\nthe BC population included as an offset) and log-binomial\nregression, respectively. The models included an interac-\ntion term between year and sex. Potential differences in the\ndistribution over socioeconomic quintiles were assessed\nusing a Chi-Squared test of homogeneity.\nFollow-up data for the years 2009 and 2010 were\nunavailable for some individuals who were alive and res-\nident in BC at the end of 2008 but had not yet met the MS\ncase definition. To assess the potential impact of this\nmissing 2 years of follow-up data on the findings, the\nnumbers of potentially missed incident and prevalent cases\nwere calculated by assuming every individual with missing\nfollow-up data would have met the case definition and the\nestimates and comparisons were repeated.\nStatistical analyses were performed using R: A Language\nand Environment for Statistical Computing v.2.15 (R Foun-\ndation for Statistical Computing, Vienna, Austria; 2012).\nThis study was approved by the University of British\nColumbia\u2019s Clinical Research Ethics Board (approval # H10-\n01361).BCMinistry ofHealth,BCVital StatisticsAgency and\nBC PharmaNet approved access to administrative health data.\nResults\nBetween 1996 and 2008, 4,222 BC residents met the\nincident case definition of C7 MS claims (or C3 MS\nclaims for those with B3 years of residency in BC) and at\nleast 5 years of residency before their first MS-related\nclaim. The standardized annual incidence rate was 7.8\n(95 % CI 7.6, 8.1) per 100,000 people; 11.5 (95 % CI 11.1,\n11.9) for women and 4.1 (95 % CI 3.8, 4.3) for men. The\nmore sensitive case definition of C3 MS claims identified\n5876 incident cases for a standardized annual incidence\nestimate of 10.9 (95 % CI 10.6, 11.2) per 100,000.\nCharacteristics of the incident cases are summarized in\nTable 1; the sex, age and SES distributions were similar\nregardless of the case definition used. Women made up\napproximately three quarters of the incident cases. The\noverall mean age at the first MS-related claim was lower\nfor women (44 years) than for men (46 years) (p\\ 0.001).\nThe distribution of cases across the SES quintiles differed\nat the time of the incident claim, with more cases in the\nmiddle and higher SES quintiles, but the absolute differ-\nences were small. Among all incident cases (using the\nprimary definition), 27 % filled a prescription for a DMD\nwithin 3 years of their incident claim; this proportion\nincreased between 1996 and 2000 from 10 to 32 %; the\nproportion with prescriptions within 3 years then remained\nstable at 31\u201333 % from 2000 to 2007 (the last calendar\nyear with 3 years of follow-up).\nThe median number of years between the first MS-re-\nlated claim and fulfilling criterion for the primary case\ndefinition was 1.3 years overall; 1.2 years for women and\n1.0 years for men. As the follow-up time decreased over\nthe observation period, the median time to reach criterion\nnaturally decreased; the longest was 2.1 years (interquartile\nrange: 0.8, 5.3) in 1996 when up to 15 years of follow-up\ndata were available. Using the more sensitive case defini-\ntion, 75 % of cases reached criterion in 1.8 years (median\n0.4 years) when up to 15 years of data were available.\nAlthough there were small fluctuations in the annual\nincidence rate (Fig. 1a, b), there was no evidence of an\nincrease in incidence between 1996 and 2008 regardless of\nthe case definition used. The average female to male\nincidence ratio across all years was 2.8 (95 % CI 2.6, 3.0).\nThis ratio varied by calendar year; the interaction between\nsex and year was statistically significant (p = 0.045) with a\nsmall decrease in the incidence rate in women, while the\nrate remained stable in men (Table 2).\nOn July 1st 2008, there were 8546 people with MS\nliving in BC and the standardized prevalence per 100,000\npeople was 179.9 (95 % CI 176.0, 183.8). The prevalence\nestimates for each year by sex and the sex ratio for\n1991\u20132008 are shown in Table 3. With the more sensitive\ncase definition, the standardized prevalence on July 1st\n2008 was 235.8 (95 % CI 231.4, 240.3), and an estimated\n11,184 cases were living in BC on the point prevalence\ndate.\nThe average age of the 8546 prevalent cases was\n52 years and 74 % were women. As observed for the\n2354 J Neurol (2015) 262:2352\u20132363\n123\nincident cases, a comparison across the SES groups\nrevealed an uneven distribution of prevalent MS cases with\nmore prevalent MS cases in the higher quintiles of SES\nthan in the lower quintiles, but small absolute differences.\nSex, age and SES distributions were similar for the\nprevalent cases identified by the alternative administrative\ncase definitions (Table 1). At least 29 % of the prevalent\ncases had received a prescription for a DMD at some point\nduring their follow-up (or 22 % of cases identified by the\nmore sensitive definition).\nThe prevalence of MS increased over the 18-year obser-\nvation period by an average of 4.7 % per year (p\\ 0.001),\nand the sex prevalence ratio increased from 2.27 in 1991 to\n2.78 in 2008 (p\\ 0.001) (Fig. 2a, b). Overall, the peak age\nTable 1 Characteristics of the\nincident (1996\u20132008) and\nprevalent (2008) multiple\nsclerosis cases in British\nColumbia, Canada\nIncident cases (1996\u20132008) Primary definition\nn = 4222\n3 Claims definition\nn = 5876\nSex, n (%)\nWomen 3124 (74) 4315 (73)\nMen 1098 (26) 1561 (27)\nAge at incidence, years\nMean (SD) 44.3 (13.2) 44.7 (13.4)\nMedian (1st quartile, 3rd quartile) 43.4 (35.2, 51.7) 43.8 (35.6, 52.4)\nTime to meet case definition, years\nMean (SD) 2.1 (2.2) 1.0 (1.7)\nMedian (1st quartile, 3rd quartile) 1.3 (0.5, 2.9) 0.4 (0.1, 1.1)\nPrescription for a DMD, n (%)\nEver 1411 (33) 1432 (24)\nWithin 3 years of incident claim 1143 (27) 1156 (20)\nSES quintile at incident claim, n (%)a\nLowest 759 (18)* 1050 (18)*\nSecond lowest 766 (18) 1091 (19)\nMiddle 902 (21) 1228 (21)\nSecond highest 857 (20) 1209 (21)\nHighest 842 (20) 1163 (20)\nUnknown 96 (2) 135 (2)\nPrevalent cases (July 1st 2008) Primary definition\nn = 8546\n3 Claims definition\nn = 11,184\nSex, n (%)\nWomen 6313 (74) 8206 (73)\nMen 2233 (26) 2978 (27)\nAge, years\nMean (SD) 52.3 (12.6) 52.3 (12.9)\nMedian (1st quartile, 3rd quartile) 52.3 (43.8, 60.5) 52.1 (43.7, 60.5)\nPrescription of a DMD, n (%)\nEver 2485 (29) 2516 (22)\nSES quintile, n (%)a\nLowest 1541 (18)* 2028 (18)*\nSecond lowest 1612 (19) 2090 (19)\nMiddle 1745 (20) 2257 (20)\nSecond highest 1782 (21) 2352 (21)\nHighest 1727 (20) 2272 (20)\nMissing 139 (2) 185 (2)\nSD standard deviation, DMD disease-modifying drug, SES socioeconomic status\n* p B 0.001\na Percentages may not sum to 100 due to rounding. Chi-squared test of homogeneity was used to compare\nSES quintiles to an expected equal distribution across the quintiles (cases with missing SES were excluded)\nJ Neurol (2015) 262:2352\u20132363 2355\n123\nof prevalentMS cases increased over time from 45\u201349 years\nin the early 1990s to 55\u201359 years in 2008 (Fig. 3).\nAmong those with missing follow-up information for\n2009\u20132010, there were 254 individuals with C1 demyeli-\nnating claim by the end of 2008 that had not yet met the\nprimary case definition and 74 who had not yet met the\nmore sensitive definition. These extra cases could poten-\ntially have increased the MS prevalence estimate for 2008,\nhad complete follow-up data to 2010 been available. The\nmaximum potential impact is an underestimate of the 2008\nprevalence by up to 5.3/100,000 (or 1.6/100,000 using\nthe C3 claim definition). The average annual 1996\u20132008\nincidence may have been underestimated by up to 0.2 cases\nper 100,000, while the estimated incidence using the\nalternative case definition would not have been affected.\nWhen all of these potential cases were assumed to have met\ndefinition and included in the estimates and comparisons,\nall findings related to changes over time and characteristics\nof the incident and prevalent cases remained the same.\nDiscussion\nThe prevalence of MS in BC, Canada has risen steadily and\nsubstantially, from 78.8/100,000 in 1991 to 179.9/100,000\nin 2008. The incidence rate in BC remained stable over the\nstudy period, averaging 7.8 per 100,000 new cases of MS\nper year between 1996 and 2008, which is high relative to\nFig. 1 Age-standardized\nannual incidence rates\n(1996\u20132008) of multiple\nsclerosis cases identified by the\nprimary case definition (a) and\nthe more sensitive but less\nspecific case definition (b) in\nBritish Columbia, Canada.\nNote: The apparent increased\nincidence in the final year\nobserved in a is a result of the\ncase criterion; all potential\nincident cases for that year\nhad B3 years of follow-up\navailable to study end and\ntherefore required only 3 claims\nto meet case definition\n2356 J Neurol (2015) 262:2352\u20132363\n123\nother parts of the world [1\u20134]. The prevalence sex ratio\nincreased over time; however, the incidence sex ratio,\nwhich averaged 2.8:1, did not increase.\nMethodological differences make direct comparisons\nwith earlier studies of MS prevalence in BC difficult. Other\nthan a study based on self-reported MS [21], the last pro-\nvince-wide estimate of MS prevalence, which used clini-\ncally confirmed definitions, was 93.3/100,000 in 1982 [22].\nOur estimates for 1991 (78.8/100,000) and 1992 (101.1/\n100,000) are compatible with this estimate from 10 years\nearlier. There are no previous estimates of MS incidence in\nBC with which to compare our observations. However,\nusing the same validated administrative MS case defini-\ntions age standardized to the same population, a similar\nannual incidence rate was recently found in central Canada\n[10] (11.4/100,000 in Manitoba using the more sensitive\ncase definition). While a somewhat higher annual incidence\nwas found in eastern Canada [11] (9.8/100,000 in Nova\nScotia using our primary case definition), the 95 % confi-\ndence intervals overlapped with those in our study. BC,\nManitoba and Nova Scotia are home to approximately\n19 % of the Canadian population; extrapolating the com-\nbined estimate from these provinces (weighted by their\nrelative population) to Canada, which had a population of\n35.5 million in 2014, would mean that approximately 3000\nnew MS cases are diagnosed each year, or 8 new cases per\nday. Furthermore, extrapolating the combined prevalence\nestimate (200/100,000) from the three provinces would\nmean that approximately 71,000 people are living with MS\nin Canada.\nTable 2 Number of incident cases and incidence rate of multiple sclerosis per 100,000 population per year (1996\u20132008) in British Columbia,\nCanada by sex and calendar year\nYear Women Men All Incidence sex\nratio (95 % CI)a\nCases/popul. Crude IR (95 % CI)\n[standardized IR]\nCases/popul. Crude IR (95 % CI)\n[standardized IR]\nCases/popul. Crude IR (95 % CI)\n[standardized IR]\n1996 231/\n1,944,984\n11.9 (10.4, 13.5)\n[12.0 (10.5, 13.6)]\n57/\n1,929,333\n3.0 (2.2, 3.8)\n[3.1 (2.3, 4.0)]\n288/\n3,874,317\n7.4 (6.6, 8.3)\n[7.5 (6.7, 8.5)]\n4.0 (3.0, 5.4)\n1997 246/\n1,983,109\n12.4 (10.9, 14.1)\n[12.4 (10.9, 14.1)]\n79/\n1,965,474\n4.0 (3.2, 5.0)\n[4.1 (3.2, 5.1)]\n325/\n3,948,583\n8.2 (7.4, 9.2)\n[8.3 (7.4, 9.3)]\n3.1 (2.4, 4.0)\n1998 238/\n2,002,595\n11.9 (10.4, 13.5)\n[11.9 (10.5, 13.5)]\n87/\n1,980,518\n4.4 (3.5, 5.4)\n[4.4 (3.6, 5.5)]\n325/\n3,983,113\n8.2 (7.3, 9.1)\n[8.2 (7.4, 9.2)]\n2.7 (2.1, 3.5)\n1999 269/\n2,018,796\n13.3 (11.8, 15.0)\n[13.2 (11.6, 14.8)]\n79/\n1,992,579\n4.0 (3.1, 4.9)\n[4.0 (3.1, 4.9)]\n348/\n4,011,375\n8.7 (7.8, 9.6)\n[8.6 (7.7, 9.6)]\n3.4 (2.6, 4.3)\n2000 246/\n2,033,853\n12.1 (10.6, 13.7)\n[11.9 (10.5, 13.5)]\n94/\n2,005,377\n4.7 (3.8, 5.7)\n[4.7 (3.8, 5.7)]\n340/\n4,039,230\n8.4 (7.6, 9.4)\n[8.3 (7.5, 9.3)]\n2.6 (2.0, 3.3)\n2001 255/\n2,053,217\n12.4 (10.9, 14.0)\n[12.2 (10.8, 13.8)]\n95/\n2,023,047\n4.7 (3.8, 5.7)\n[4.7 (3.8, 5.7)]\n350/\n4,076,264\n8.6 (7.7, 9.5)\n[8.5 (7.6, 9.4)]\n2.6 (2.1, 3.4)\n2002 243/\n2,066,320\n11.8 (10.3, 13.3)\n[11.6 (10.2, 13.1)]\n89/\n2,031,858\n4.4 (3.5, 5.4)\n[4.3 (3.5, 5.3)]\n332/\n4,098,178\n8.1 (7.3, 9.0)\n[8.0 (7.2, 8.9)]\n2.7 (2.1, 3.4)\n2003 228/\n2,079,214\n11.0 (9.6, 12.5)\n[10.8 (9.4, 12.3)]\n85/\n2,043,182\n4.2 (3.3, 5.1)\n[4.1 (3.3, 5.1)]\n313/\n4,122,396\n7.6 (6.8, 8.5)\n[7.5 (6.7, 8.4)]\n2.6 (2.1, 3.4)\n2004 235/\n2,096,756\n11.2 (9.8, 12.7)\n[11.1 (9.7, 12.6)]\n76/\n2,058,414\n3.7 (2.9, 4.6)\n[3.7 (2.9, 4.6)]\n311/\n4,155,170\n7.5 (6.7, 8.4)\n[7.4 (6.6, 8.3)]\n3.0 (2.3, 3.9)\n2005 194/\n2,117,446\n9.2 (7.9, 10.6)\n[9.0 (7.7, 10.3)]\n96/\n2,079,342\n4.6 (3.7, 5.6)\n[4.5 (3.6, 5.5)]\n290/\n4,196,788\n6.9 (6.1, 7.8)\n[6.8 (6.0, 7.6)]\n2.0 (1.6, 2.5)\n2006 235/\n2,141,450\n11.0 (9.6, 12.5)\n[10.8 (9.4, 12.3)]\n81/\n2,102,130\n3.9 (3.1, 4.8)\n[3.7 (2.9, 4.6)]\n316/\n4,243,580\n7.5 (6.7, 8.3)\n[7.3 (6.5, 8.2)]\n2.9 (2.2, 3.7)\n2007 193/\n2,173,994\n8.9 (7.7, 10.2)\n[8.6 (7.4, 9.9)]\n73/\n2,135,530\n3.4 (2.7, 4.3)\n[3.4 (2.6, 4.3)]\n266/\n4,309,524\n6.2 (5.5, 7.0)\n[6.0 (5.3, 6.8)]\n2.6 (2.0, 3.4)\n2008b 311/\n2,210,657\n14.1 (12.6, 15.7)\n[13.8 (12.3, 15.4)]\n107/\n2,173,653\n4.9 (4.0, 6.0)\n[4.6 (3.8, 5.6)]\n418/\n4,384,310\n9.5 (8.6, 10.5)\n[9.3 (8.4, 10.2)]\n2.9 (2.3, 3.6)\n1996\u20132008 3124/\n26,922,391\n11.6 (11.2, 12.0)\n[11.5 (11.1, 11.9)]\n1098/\n26,520,437\n4.1(3.9, 4.4)\n[4.1 (3.8, 4.3)]\n4222/\n53,442,828\n7.9 (7.7, 8.1)\n[7.8 (7.6, 8.1)]\n2.8 (2.6, 3.0)\nIR incidence rate, Popul. population (denominator), CI confidence interval\na Crude incidence sex ratio (incidence rate in women: incidence rate in men)\nb The apparent increase in incidence in 2008 is due to the case criterion; as of 2008 all potential cases required only 3 claims, unlike in previous\nyears, because B3 years of follow-up remained\nJ Neurol (2015) 262:2352\u20132363 2357\n123\nOur observations that the prevalence of MS has been\nincreasing by approximately 4.7 % per year in BC, and that\nthe age of the prevalent population has risen have impor-\ntant implications for broader society, including govern-\nments and health care planners. We also found a gradual\nincrease in the proportion of women to men living with\nMS; this is compatible with recent observations from\nelsewhere in Canada [10, 11] and the UK [23], and is likely\ndue to the changing demographics (older mean age) of the\ngeneral population and the greater life expectancy of\nwomen compared to men.\nThe rising prevalence in BC cannot be explained by\nincreasing numbers of new MS cases; our incidence rates\nremained relatively stable over the 13-year period despite\nTable 3 Number of prevalent cases and prevalence of multiple sclerosis per 100,000 population on July 1st (1991\u20132008) in British Columbia,\nCanada by sex and calendar year\nYear Women Men All Prevalence sex\nratio (95 % CI)a\nCases/\npopul.\nCrude PP (95 % CI)\n[standardized PP]\nCases/\npopul.\nCrude PP (95 % CI)\n[standardized PP]\nCases/\npopul.\nCrude PP (95 % CI)\n[standardized PP]\n1991 1731/\n1,692,156\n102.3 (97.5, 107.2)\n[108.8 (103.7, 114.2)]\n757/\n1,681,631\n45.0 (41.9, 48.3)\n[48.7 (45.3, 52.4)]\n2488/\n3,373,787\n73.8 (70.9, 76.7)\n[78.8 (75.7. 82.0)]\n2.3 (2.1, 2.5)\n1992 2336/\n1,741,163\n134.2 (128.8, 139.7)\n[142.3 (136.5, 148.2)]\n964/\n1,727,639\n55.8 (52.3, 59.4)\n[59.8 (56.1, 63.7)]\n3300/\n3,468,802\n95.1 (91.9, 98.4)\n[101.1 (97.7, 104.7)]\n2.4 (2.2, 2.6)\n1993 2701/\n1,790,843\n150.8 (145.2, 156.6)\n[158.7 (152.8, 164.9)]\n1092/\n1,776,929\n64.5 (57.9, 65.2)\n[65.3 (61.4, 69.3)]\n3793/\n3,567,772\n106.3 (103.0, 109.8)\n[112.1 (108.6, 115.8)]\n2.5 (2.3, 2.6)\n1994 3011/\n1,843,834\n163.3 (157.5, 169.2)\n[170.3 (164.3, 176.6)]\n1195/\n1,832,241\n65.2 (61.6, 69.0)\n[68.8 (65.0, 72.9)]\n4206/\n3,676,075\n114.4 (111.0, 117.9)\n[119.8 (116.1, 123.5)]\n2.5 (2.3, 2.7)\n1995 3288/\n1,893,063\n173.7 (167.8, 179.7)\n[179.8 (173.6, 186.1)]\n1286/\n1,884,327\n68.3 (64.6, 72.1)\n[71.4 (67.5, 75.4)]\n4574/\n3,777,390\n121.1 (117.6, 124.7)\n[125.7 (122.1, 129.4)]\n2.5 (2.4, 2.7)\n1996 3535/\n1,944,984\n181.8 (175.8, 187.8)\n[187.1 (180.9, 193.4)]\n1359/\n1,929,333\n70.4 (66.7, 74.3)\n[73.2 (69.3, 77.2)]\n4894/\n3,874,317\n126.3 (122.8, 129.9)\n[130.4 (126.8, 134.1)]\n2.6 (2.4, 2.8)\n1997 3841/\n1,983,109\n193.7 (187.6, 199.9)\n[197.8 (191.6, 204.2)]\n1432/\n1,965,474\n72.9 (69.1, 76.7)\n[74.9 (71.1, 78.9)]\n5273/\n3,948,583\n133.5 (130.0, 137.2)\n[136.6 (133.0, 140.4)]\n2.7 (2.5, 2.8)\n1998 4048/\n2,002,595\n202.1 (196.0, 208.5)\n[204.5 (198.2, 210.9)]\n1514/\n1,980,518\n76.4 (72.6, 80.4)\n[77.5 (73.6, 81.5)]\n5562/\n3,983,113\n139.6 (136.0, 143.4)\n[141.3 (137.6, 145.1)]\n2.6 (2.5, 2.8)\n1999 4325/\n2,018,796\n214.2 (207.9, 220.7)\n[214.4 (208.0, 220.9)]\n1596/\n1,992,579\n80.1 (76.2, 84.1)\n[80.1 (76.2, 84.1)]\n5921/\n4,011,375\n147.6 (143.9, 151.4)\n[147.7 (144.0, 151.5)]\n2.7 (2.5, 2.8)\n2000 4563/\n2,033,853\n224.4 (217.9, 231.00)\n[222.3 (215.9, 228.9)]\n1672/\n2,005,377\n83.4 (79.4, 87.5)\n[82.3 (78.4, 86.4)]\n6235/\n4,039,230\n154.4 (150.6, 158.2)\n[152.9 (149.1, 156.7)]\n2.7 (2.5, 2.9)\n2001 4806/\n2,053,217\n234.1 (227.5, 240.8)\n[230.0 (223.5, 236.6)]\n1761/\n2,023,047\n87.1 (83.0, 91.2)\n[85.1 (81.2, 89.2)]\n6567/\n4,076,264\n161.1 (157.2, 165.0)\n[158.2 (154.4, 162.0)]\n2.7 (2.6, 2.8)\n2002 5086/\n2,066,320\n246.1 (239.4, 253.0)\n[239.6 (233.1, 246.3)]\n1846/\n2,031,858\n90.9 (86.8, 95.1)\n[87.8 (83.8, 91.9)]\n6932/\n4,098,178\n169.2 (165.2, 173.2)\n[164.4 (160.6, 168.3)]\n2.7 (2.6, 2.9)\n2003 5285/\n2,079,214\n254.2 (247.4, 261.1)\n[244.9 (238.4, 251.6)]\n1925/\n2,043,182\n94.2 (90.1, 98.5)\n[90.1 (86.1, 94.2)]\n7210/\n4,122,396\n174.9 (170.9, 179.0)\n[168.3 (164.4, 172.3)]\n2.7 (2.6, 2.8)\n2004 5490/\n2,096,756\n261.8 (255.0, 268.8)\n[249.8 (243.2, 256.5)]\n1962/\n2,058,414\n95.3 (91.2, 99.6)\n[90.2 (86.2, 94.3)]\n7452/\n4,155,170\n179.3 (175.3, 183.5)\n[170.9 (167.0, 174.8)]\n2.8 (2.6, 2.9)\n2005 5705/\n2,117,446\n269.4 (262.5, 276.5)\n[254.3 (247.7, 261.0)]\n2044/\n2,079,342\n98.3 (94.1, 102.7)\n[92.0 (88.1, 96.1)]\n7749/\n4,196,788\n184.6 (180.6, 188.8)\n[174.1 (170.2, 178.0)]\n2.7 (2.6, 2.9)\n2006 5891/\n2,141,450\n275.1 (268.1, 282.2)\n[257.5 (250.9, 264.2)]\n2097/\n2,102,130\n99.8 (95.5, 104.1)\n[92.5 (88.6, 96.6)]\n7988/\n4,243,580\n188.2 (184.1, 192.4)\n[176.0 (172.1, 179.9)]\n2.8 (2.6, 2.9)\n2007 6073/\n2,173,994\n279.4 (272.4, 286.5)\n[259.9 (253.3, 266.6)]\n2143/\n2,135,530\n100.4 (96.2, 104.7)\n[92.3 (88.3, 96.3)]\n8216/\n4,309,524\n190.7 (186.6, 194.8)\n[177.0 (173.2, 180.9)]\n2.8 (2.7, 2.9)\n2008 6313/\n2,210,657\n285.6 (278.6, 292.7)\n[264.0 (257.5, 270.7)]\n2233/\n2,173,653\n102.7 (98.5, 107.1)\n[93.9 (90.0, 97.9)]\n8546/\n4,384,310\n194.9 (190.8, 199.1)\n[179.9 (176.0, 183.8)]\n2.8 (2.7, 2.9)\nPP point prevalence, Popul. population (denominator), CI confidence interval\na Crude prevalence sex ratio (prevalence proportion in women: prevalence proportion in men)\n2358 J Neurol (2015) 262:2352\u20132363\n123\nchanges in MS diagnostic criteria [24] and increasing\navailability of disease-modifying drugs. While this seems\nin contrast to some other regions of the world where recent\nincreases in incidence rates have been reported [5], a stable\nincidence rate has also been found over a similar time\nperiod in other Canadian provinces [10, 11, 19, 25, 26] and\nthe UK [23]. Taken together with our findings, this sug-\ngests that the incidence of MS has stabilised in some areas\nover recent years. In the absence of increasing incidence,\nthe rising prevalence may reflect longer disease duration\ndue to earlier diagnosis, improved survival with MS or\nboth. Survival has improved for both the BC general\npopulation and for people with MS in BC over the past\n30 years [27]. Similarly, improved survival has been found\nin other MS populations, including those from Denmark\n[28] and Norway [29]. Immigration of prevalent cases can\nalso influence prevalence trends and the population of\nBritish Columbia increased by nearly 30 % between 1991\nand 2008, mostly due to immigration from other Canadian\nprovinces and other countries. The prevalence estimates\ninclude MS cases that were resident throughout, as well as\nthose that immigrated to BC, during the observation period.\nNewly immigrant prevalent cases would have contributed\nto the increasing prevalence over time if there was a greater\nproportion of MS cases among those immigrating to BC.\nThe average age at the time of the incident MS-related\nclaim was 44 years. This age is comparable to that recorded\nin other Canadian provinces [10, 11] and was found to be\nwithin 3 years of the MS diagnosis date from medical charts\nor by personal report for 74 and 76 % of cases, respectively\n[10]. It is also equivalent to that identified as the first date of\ndiagnosis in the General Practice Research Database for\nFig. 2 Age-standardized\nprevalence (1991\u20132008) of\nmultiple sclerosis cases\nidentified by the primary case\ndefinition (a) and the more\nsensitive but less specific case\ndefinition (b) in British\nColumbia, Canada\nJ Neurol (2015) 262:2352\u20132363 2359\n123\ncases of MS in the UK [23]. While the date of diagnosis, or\nof first medical recognition, is frequently used to measure\nMS incidence [2\u20134], symptom onset can often occur several\nyears before the disease is first recognized or diagnosed.\nNotably, while others have reported increases in the\nincidence ratio of women to men with MS [6], we found no\nevidence of such a trend in BC. Rather, we observed a\ndecrease in this ratio over time, although the absolute\ndifferences were small. Nonetheless, the more consistently\nobserved sex differences for MS were evident; nearly three\nquarters of incident cases in BC were women and men\nwere approximately 3 years older than women at the time\nof the first MS-related claim reflecting typical differences\nin onset age between sexes.\nWe observed a small difference in the distribution of cases\nacross the socioeconomic quintiles, with a greater proportion\nof both incident and prevalent cases in the upper quintiles and\nfewer in the lower. Similar observations have been made in\nthe past [30\u201334], although others have reported either no\nrelationship or a negative association with SES [35].\nApproximately, one-third of incident or prevalent MS\ncases filled a prescription for a DMD during the study\nperiod, stabilizing from the year 2000 onwards. This pro-\nportion may seem low, particularly when compared to a\nprevious estimate (73\u201385 %) derived from a volunteer\nsample of patients recruited from Canadian MS treatment\ncentres [36]. However, our proportion was derived from\npopulation-based rather than clinic-based data, the first\nDMD (interferon beta-1b) was only approved for use in\nCanada in 1995, and not all individuals with MS would\nhave been eligible for treatment (including those unable to\nwalk, those without relapses and those with a primary\nprogressive disease course). Thus, it is likely that our data\nprovide a realistic representation of DMD use in the British\nColumbian MS population over the study period.\nThe strengths of our study include the use of adminis-\ntrative health data, which represents the entire BC popu-\nlation, and spans nearly two decades allowing us to assess\ntemporal trends. Furthermore, we used two previously\nvalidated MS case definitions to generate these estimates\n[10, 11]. The primary definition was identified as the best\nin terms of balance between specificity and sensitivity\namong candidate validated MS case definitions [11]. The\nsecondary definition generated higher estimates due to its\nFig. 3 Age-specific prevalence of multiple sclerosis identified by the primary case definition per 100,000 population by select years (1992, 1996,\n2000, 2004 and 2008) in British Columbia, Canada\n2360 J Neurol (2015) 262:2352\u20132363\n123\ngreater sensitivity, but may have included a greater pro-\nportion of false positives. Estimates generated by the pri-\nmary definition could be more useful when it is important\nto minimize the risk of including people without MS,\nwhereas the estimates from the more sensitive definition\nare likely more useful for estimating burden of disease and\nfor health care planning. Although these MS case defini-\ntions were not validated specifically using the BC admin-\nistrative data, the algorithms have been validated in Nova\nScotia and Manitoba [10, 11]. Furthermore, similar 7-claim\nadministrative case definitions of MS derived from\nadministrative data in Ontario, Canada were validated in a\nprimary care dataset in that province and found to have\nexcellent performance [18]. The structure of the Canadian\npublic health care system and the methods for coding\nphysician and hospital visits in administrative health data\nare consistent between these three provinces and BC,\nwhich suggests that the case definitions would perform well\nand can be reliably applied in BC. The BC administrative\nhealth databases have been used, both independently and\ncombined with equivalent data from other Canadian pro-\nvinces, to identify and study other chronic diseases such as\ndiabetes and hypertension [37\u201339].\nHealth administrative data have limitations. Although\nwe allowed a five-year claim-free run-in period to capture\nincident cases, it remains possible that prevalent benign\nMS cases that rarely interacted with the medical system\nwere misclassified as incident. Ascertainment is a common\nproblem with MS incidence studies due to the inevitable\ntime-lag between symptom onset and recognition of the\ndisease; the estimated incidence may be affected by\nincomplete ascertainment towards the end of follow-up.\nSimilarly, the prevalence estimates for the earliest years\n(1991 and 1992) should be treated cautiously; atypical\nprevalent cases with infrequent contact with the medical\nsystem could have been missed in these years. On the other\nhand, although we had missing follow-up data for up to 254\npotential MS cases and could not confirm that they met\ncriteria with follow-up to 2010, the estimates were not\nnotably impacted.\nWe were unable to consider ethnicity, or country of\norigin. Although most BC residents are of European\nancestry, BC has a higher proportion of residents of non-\nEuropean origin than other Canadian provinces. The pro-\nportion with European ancestry has declined over time,\nfrom 82 % of the BC population in 1996 [40] to 75 % in\n2006 [41]; people of Asian ancestry represent the largest\nminority group. The somewhat lower MS incidence rate in\nBC compared to that in Eastern Canada [11] might be\nexplained by differences in the ethnic composition of the\nsource populations [41].\nIn summary, BC has a high incidence of MS which has\nremained stable for more than a decade. However, as\nelsewhere in Canada, the prevalence and the peak age of\nthe MS population have increased significantly. Popula-\ntion-based administrative health databases and validated\ncase definitions of MS using health claims data provide a\nreliable, accessible and cost effective means of monitoring\nthe incidence and prevalence of MS.\nAcknowledgments This study was supported by the Canadian\nInstitutes of Health Research (CIBG 101829); the Rx&D Health\nResearch Foundation; and a Don Paty Career Development Award\nfrom the MS Society of Canada (to RAM). We thank the BC Ministry\nof Health, BC Vital Statistics Agency and BC PharmaNet for\napproval and support with accessing provincial data; and Population\nData BC for facilitating approval and use of the data.\nCIHR Team in the Epidemiology and Impact of Comorbidity on\nMultiple Sclerosis Ruth Ann Marrie, MD, Ph.D., FRCPC\n(University of Manitoba, Principal Investigator); John D. Fisk, Ph.D.\n(Dalhousie University, Co-principal Investigator and Site Investiga-\ntor); Sharon Warren, Ph.D. (University of Alberta, Co-principal\nInvestigator and Site Investigator); Christina Wolfson, Ph.D. (McGill\nUniversity, Co-principal Investigator and Site Investigator); Helen\nTremlett, Ph.D. (University of British Columbia, Co-principal\nInvestigator and Site Investigator); James Blanchard, MD, MPH,\nPh.D. (University of Manitoba, Co-investigator); Lawrence Elliott,\nMD, M.Sc. (University of Manitoba, Co-investigator); Bo Nancy Yu,\nMD, Ph.D. (University of Manitoba, Co-investigator); Virender Bhan,\nMBBS, FRCPC (Dalhousie University, Co-investigator); Joanne\nProfetto-McGrath, Ph.D., RN (University of Alberta, Co-investiga-\ntor); Scott Patten, MD, Ph.D. (University of Calgary, Co-investiga-\ntor); Lawrence W. Svenson, B.Sc. (University of Alberta,\nCollaborator); Patricia Caetano, Ph.D. (University of Manitoba,\nCollaborator); Nathalie Jette, MD, M.Sc., FRCPC (University of\nCalgary, Site Investigator).\nCompliance with ethical standards\nConflicts of interest Dr. Kingwell, Mr. Zhu, Dr. Profetto-McGrath\nand Mr. Svenson report no disclosures.\nDr. Marrie receives research funding from: Canadian Institutes of\nHealth Research, Public Health Agency of Canada, Manitoba Health\nResearch Council, Health Sciences Centre Foundation, Multiple\nSclerosis Society of Canada, Multiple Sclerosis Scientific Foundation,\nRx&D Health Research Foundation, and has conducted clinical trials\nfunded by sanofi-aventis.\nDr. Fisk receives research funding from: Canadian Institutes of Health\nResearch, Multiple Sclerosis Society of Canada and the National MS\nSociety (USA).\nDr. Wolfson receives funding from: Canadian Institutes of Health\nResearch, Canada Foundation for Innovation, Multiple Sclerosis\nSociety of Canada, National Multiple Sclerosis Society, Research\nInstitute of the McGill University Health Centre.\nDr. Warren receives funding from Canadian Institutes of Health\nResearch and Fort Assiniboine Equine Endeavour Foundation.\nDr. Jette is the holder of a Canada Research Chair in Neurological\nHealth Sciences and an Alberta Innovates Health Solutions (AI-HS)\nPopulation Health Investigator Award. She receives research funding\nfrom the Canadian Institutes of Health Research, AI-HS, Hotchkiss\nBrain Institute, Cumming School of Medicine, University of Calgary,\nAlberta Health Services and Alberta Health.\nDr. Bhan has received honoraria and consulting fees from Biogen\nIdec, EMD Serono, Genzyme, Novartis, Roche and Teva Neuro-\nsciences.\nDr. Yu receives research funding from: Health Science Centre\nJ Neurol (2015) 262:2352\u20132363 2361\n123\nFoundation, Canadian Institutes of Health Research, Gilead Sciences,\nCanadian International Development Agency, Multiple Sclerosis\nSociety of Canada, and National Multiple Sclerosis Society.\nDr. Elliott receives research funding from the Canadian Institutes of\nHealth Research and the Multiple Sclerosis Society of Canada.\nDr. Tremlett is funded by the Multiple Sclerosis Society of Canada\n(Don Paty Career Development Award); is a Michael Smith Foun-\ndation for Health Research Scholar and the Canada Research Chair\nfor Neuroepidemiology and Multiple Sclerosis. She has received:\nresearch support from the National Multiple Sclerosis Society, the\nCanadian Institutes of Health Research, and the UK MS Trust;\nspeaker honoraria and/or travel expenses to attend conferences from\nthe Consortium of MS Centres (2013), the National MS Society\n(2012, 2014), Bayer Pharmaceuticals (2010), Teva Pharmaceuticals\n(2011), ECTRIMS (2011, 2012, 2013, 2014), UK MS Trust (2011),\nthe Chesapeake Health Education Program, US Veterans Affairs\n(2012), Novartis Canada (2012), Biogen Idec (2014), American\nAcademy of Neurologists (2013, 2014). Unless otherwise stated, all\nspeaker honoraria are either donated to an MS charity or to an\nunrestricted grant for use by her research group.\nOn behalf of all coauthors, the corresponding author states that there\nis no conflict of interest specific to this study.\nOpen Access This article is distributed under the terms of the\nCreative Commons Attribution 4.0 International License (http://crea\ntivecommons.org/licenses/by/4.0/), which permits unrestricted use,\ndistribution, and reproduction in any medium, provided you give\nappropriate credit to the original author(s) and the source, provide a\nlink to the Creative Commons license, and indicate if changes were\nmade.\nReferences\n1. Multiple Sclerosis International Federation (2013) Atlas of MS.\nhttp://www.msif.org/wp-content/uploads/2014/09/Atlas-of-MS.\npdf. Accessed 8 December 2014\n2. Evans C, Beland SG, Kulaga S, Wolfson C, Kingwell E, Marriott\nJ et al (2013) Incidence and prevalence of multiple sclerosis in\nthe Americas: a systematic review. Neuroepidemiology\n40(3):195\u2013210\n3. Kingwell E, Marriott JJ, Jette N, Pringsheim T, Makhani N,\nMorrow SA et al (2013) Incidence and prevalence of multiple\nsclerosis in Europe: a systematic review. BMC Neurol 13(1):128\n4. Makhani N, Morrow SA, Fisk JD, Evans C, Beland SG, Kulaga S\net al (2014) MS incidence and prevalence in Africa, Asia, Aus-\ntralia and New Zealand: a systematic review. Mult Scler Relat\nDisord 3:48\u201360\n5. Alonso A, Hernan MA (2008) Temporal trends in the incidence\nof multiple sclerosis: a systematic review. Neurology\n71(2):129\u2013135\n6. Koch-Henriksen N, Sorensen PS (2010) The changing demo-\ngraphic pattern of multiple sclerosis epidemiology. Lancet Neurol\n9(5):520\u2013532\n7. Trojano M, Lucchese G, Graziano G, Taylor BV, Simpson S Jr,\nLepore V et al (2012) Geographical variations in sex ratio trends\nover time in multiple sclerosis. PLoS One 7(10):e48078\n8. Public Health Agency of Canada (2014) Preventing Chronic\nDisease Strategic Plan 2013-2016. http://www.phac-aspc.gc.ca/\ncd-mc/diabetes-diabete/strategy_plan-plan_strategique-eng.php.\nAccessed 29 September 2014\n9. The National MS Society (2014) MS Prevalence. http://www.\nnationalmssociety.org/About-the-Society/MS-Prevalence. Acces-\nsed 29 September 2014\n10. Marrie RA, Yu N, Blanchard J, Leung S, Elliott L (2010) The\nrising prevalence and changing age distribution of multiple\nsclerosis in Manitoba. Neurology 74(6):465\u2013471\n11. Marrie RA, Fisk JD, Stadnyk KJ, Yu BN, Tremlett H, Wolfson C\net al (2013) The incidence and prevalence of multiple sclerosis in\nnova scotia, Canada. Can J Neurol Sci 40(6):824\u2013831\n12. British Columbia Ministry of Health [creator] (2012) Discharge\nAbstract Database (Hospital Separations). Population Data BC\n[publisher]. Data Extract. MOH (2012). http://www.popdata.bc.\nca/data\n13. British Columbia Ministry of Health [creator] (2012) Medical\nServices Plan (MSP) Payment Information File Population Data\nBC [publisher]. Data Extract. MOH (2012). http://www.popdata.\nbc.ca/data\n14. British Columbia Ministry of Health [creator] (2012). Phar-\nmaNet: BC Ministry of Health [publisher]; 2012. Data Extract\nMOH (2012). http://www.popdata.bc.ca/data\n15. BC Vital Statistics Agency [creator] (2012). Vital Statistics\nDeaths: Population Data BC [publisher]. Data Extract. MOH\n(2012). http://www.popdata.bc.ca/data\n16. British Columbia Ministry of Health [creator] (2012) Consoli-\ndation File (MSP Registration & Premium Billing). Population\nData BC [publisher]. Data Extract. MOH (2012). http://www.\npopdata.bc.ca/data\n17. Statistics Canada (2014) Annual Income Estimates for Census\nFamilies and Individuals (T1 Family File). http://www23.statcan.\ngc.ca/imdb/p2SV.pl?Function=getSurvey&SDDS=4105&lang=\nen&db=IMDB&dbg=f&adm=8&dis=2. Accessed 11 August\n2014\n18. Widdifield J, Ivers NM, Young J, Green D, Jaakkimainen L, Butt\nDA et al (2014) Development and validation of an administrative\ndata algorithm to estimate the disease burden and epidemiology\nof multiple sclerosis in Ontario, Canada. Mult Scler. doi:10.1177/\n1352458514556303\n19. Hader WJ, Yee IM (2007) Incidence and prevalence of multiple\nsclerosis in Saskatoon, Saskatchewan. Neurology\n69(12):1224\u20131229\n20. Fay MP, Feuer EJ (1997) Confidence intervals for directly stan-\ndardized rates: a method based on the gamma distribution. Stat\nMed 16(7):791\u2013801\n21. Beck CA, Metz LM, Svenson LW, Patten SB (2005) Regional\nvariation of multiple sclerosis prevalence in Canada. Mult Scler.\n11(5):516\u2013519\n22. Sweeney VP, Sadovnick AD, Brandejs V (1986) Prevalence of\nmultiple sclerosis in British Columbia. Can J Neurol Sci\n13(1):47\u201351\n23. Mackenzie IS, Morant SV, Bloomfield GA, MacDonald TM,\nO\u2019Riordan J (2014) Incidence and prevalence of multiple scle-\nrosis in the UK 1990\u20132010: a descriptive study in the General\nPractice Research Database. J Neurol Neurosurg Psychiatry\n85(1):76\u201384\n24. McDonald WI, Compston A, Edan G, Goodkin D, Hartung HP,\nLublin FD et al (2001) Recommended diagnostic criteria for\nmultiple sclerosis: guidelines from the International Panel on the\ndiagnosis of multiple sclerosis. Ann Neurol. 50(1):121\u2013127\n25. Warren SA, Svenson LW, Warren KG (2008) Contribution of\nincidence to increasing prevalence of multiple sclerosis in\nAlberta. Canada. Mult Scler 14(7):872\u2013879\n26. Sloka JS, Pryse-Phillips WE, Stefanelli M (2005) Incidence and\nprevalence of multiple sclerosis in Newfoundland and Labrador.\nCan J Neurol Sci 32(1):37\u201342\n27. Kingwell E, van der Kop M, Zhao Y, Shirani A, Zhu F, Oger J\net al (2012) Relative mortality and survival in multiple sclerosis:\nfindings from British Columbia, Canada. J Neurol Neurosurg\nPsychiatry 83(1):61\u201366\n2362 J Neurol (2015) 262:2352\u20132363\n123\n28. Bronnum-Hansen H, Koch-Henriksen N, Stenager E (2004)\nTrends in survival and cause of death in Danish patients with\nmultiple sclerosis. Brain 127(Pt 4):844\u2013850\n29. Grytten Torkildsen N, Lie S, Aarseth J, Nyland H, Myhr K (2008)\nSurvival and cause of death in multiple sclerosis: results from a\n50-year follow-up in Western Norway. Mult Scler\n14(9):1191\u20131198\n30. Lowis GW (1990) The social epidemiology of multiple sclerosis.\nSci Total Environ 90:163\u2013190\n31. Kurtzke JF, Page WF (1997) Epidemiology of multiple sclerosis\nin US veterans: VII. Risk factors for MS. Neurology\n48(1):204\u2013213\n32. Zilber N, Kahana E (1996) Risk factors for multiple sclerosis: a\ncase-control study in Israel. Acta Neurol Scand 94(6):395\u2013403\n33. Hammond SR, McLeod JG, Macaskill P, English DR (1996)\nMultiple sclerosis in Australia: socioeconomic factors. J Neurol\nNeurosurg Psychiatry 61(3):311\u2013313\n34. Green C, Yu BN, Marrie RA (2013) Exploring the implications of\nsmall-area variation in the incidence of multiple sclerosis. Am J\nEpidemiol 178(7):1059\u20131066\n35. Goulden R, Ibrahim T, Wolfson C (2015) Is high socioeconomic\nstatus a risk factor for multiple sclerosis? A systematic review.\nEur J Neurol 22(6):899\u2013911\n36. Karampampa K, Gustavsson A, Miltenburger C, Kindundu CM,\nSelchen DH (2012) Treatment experience, burden, and unmet\nneeds (TRIBUNE) in multiple sclerosis: the costs and utilities of\nMS patients in Canada. J Popul Ther Clin Pharmacol 19(1):e11\u2013\ne25\n37. Amed S, Vanderloo SE, Metzger D, Collet JP, Reimer K, McCrea\nP et al (2011) Validation of diabetes case definitions using\nadministrative claims data. Diabet Med 28(4):424\u2013427\n38. Quan H, Khan N, Hemmelgarn BR, Tu K, Chen G, Campbell N\net al (2009) Validation of a case definition to define hypertension\nusing administrative data. Hypertension 54(6):1423\u20131428\n39. Government of Canada (2015) Canadian Chronic Disease\nSurveillance System 1999/2000-2010/2011 Ottawa 2014. http://\ndata.gc.ca/data/en/dataset/9525c8c0-554a-461b-a763-\nf1657acb9c9d. Accessed 23 June 2015\n40. Statistics Canada (1996) Census: Ethnic origin, visible minorities.\nOttawa 1998. http://www.statcan.gc.ca/daily-quotidien/980217/\ndq980217-eng.htm. Accessed 9 December 2014\n41. Statistics Canada (2006) Canada\u2019s Ethnocultural Mosaic, 2006\nCensus: Provinces and territories. Ottawa. http://www12.statcan.\nca/census-recensement/2006/as-sa/97-562/index-eng.cfm?CFID=\n124733&CFTOKEN=71329601. Accessed 9 December 2014\nJ Neurol (2015) 262:2352\u20132363 2363\n123\n", "identifiers": ["10.1007/s00415-015-7842-0"], "publisher": "Springer Nature", "relations": ["http://dx.doi.org/10.1007/s00415-015-7842-0"], "repositories": [{"id": "2612", "openDoarId": 0, "name": "Springer - Publisher Connector", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 1, "metadataAdded": 1493382149000, "metadataUpdated": 1494847305000, "depositedDate": null}, "subjects": ["journal-article"], "title": "High incidence and increasing prevalence of multiple sclerosis in British Columbia, Canada: findings from over two decades (1991\u20132010)", "topics": [], "types": [], "year": 2015, "fulltextIdentifier": "https://core.ac.uk/download/pdf/81091491.pdf", "doi": "10.1007/s00415-015-7842-0", "downloadUrl": "https://core.ac.uk/download/pdf/81091491.pdf"}}
{"status": "OK", "data": {"id": "11466311", "authors": ["Niculescu, Andreea"], "citations": [{"raw": "P. Luo, E. A. Eikman, W. Kealy, and W. Qian. Analysis of a mammography teaching program based on an aordance design model. In Academic Radiology, 13(12), pages 1542{1552, 2006.", "title": "Analysis of a mammography teaching program based on an aordance design model.", "date": "2006", "doi": "10.1016/j.acra.2006.08.016"}, {"raw": "J.G. Sheridan and G. Kortuem. Aordance-based design of physical interfaces for ubiquitous environments. In LNCS, Berlin, Heidelberg, 2006. Springer.", "title": "Aordance-based design of physical interfaces for ubiquitous environments.", "date": "2006", "doi": "10.1007/11890348_15"}, {"raw": "D.A. Norman. Aordances, conventions and design. In Interactions, pages 38{43,", "title": "Aordances, conventions and design."}, {"raw": "J. McGrenere and W. Ho. Aordances: Clarifying and evolving a concept. In Proc. of Graphic Interface, pages 179{186, Montreal, Canada, 2000.", "title": "Aordances: Clarifying and evolving a concept.", "date": "2000"}, {"raw": "H. R. Hartson. Cognitive, physical, sensory and functional aordances in interaction design. In Behaviour & Information Technology, 22(5), pages 315{338, 2003.", "title": "Cognitive, physical, sensory and functional aordances in interaction design.", "date": "2003", "doi": "10.1080/01449290310001592587"}, {"raw": "A. Holzinger, M. Kickmeier-Rust, and D. Albert. Dynamic media in computer science education; content complexity and learning performance: Is less more? In Educational Technology & Society, 11(1), pages 279{290, 2008.", "title": "Dynamic media in computer science education; content complexity and learning performance: Is less more?", "date": "2008"}, {"raw": "R. op den Akker, H. Bunt, S. Keizer, and B. van Schooten. From question answering to spoken dialogue: Towards an information search assistant for interactive multimodal information extraction. In Proc. of Interspeech, Lissbon, Portugal, 2005.", "title": "From question answering to spoken dialogue: Towards an information search assistant for interactive multimodal information extraction.", "date": "2005", "doi": "10.1007/978-3-642-17525-1_2"}, {"raw": "H. Bumann. Lexikon der Sprachwissenschaft. Kroener Verlag, Stuttgart, 3rd. edition, 2002.", "title": "Lexikon der Sprachwissenschaft.", "date": "2002"}, {"raw": "D.A. Norman. Memory and Attention: an introduction to human information processing. Wiley & Sons., London, 1969.", "title": "Memory and Attention: an introduction to human information processing.", "date": "1969", "doi": "10.1177/002194366900600407"}, {"raw": "J. Cassell, T. Bickmore, H. Vilhjalmsson, and H. Yan. More than just a pretty face: Aordances of embodiment. In Proceedings of 2000 International Conference on Intelligent User Interfaces, New Orleans, 2000.", "title": "More than just a pretty face: Aordances of embodiment.", "date": "2000", "doi": "10.1145/325737.325781"}, {"raw": "R.E. Mayer. Multimedia learning. Multimedia learning, Cambridge (MA), 2001.", "title": "Multimedia learning. Multimedia learning,", "date": "2001", "doi": "10.1017/cbo9781139164603"}, {"raw": "L. Vainio, E. Symes, R. Ellis, M. Tucker, and G. Ottoboni. On the relations between action planning,object identication, and motor representations of observed actions and objects. In Cognition, 108(2), page 444465. Elsevier, 2008.", "title": "On the relations between action planning,object identi and motor representations of observed actions and objects.", "date": "2008", "doi": "10.1016/j.cognition.2008.03.007"}, {"raw": "E.A. Scheglo and H. Sacks. Opening up closings. In Semiotica 8: 4, pages 289{327, New Orleans,U.S., 1973.", "title": "Opening up closings.", "date": "1973", "doi": "10.1515/semi.1973.8.4.289"}, {"raw": "M. Theune, B. van Schooten, R. op den Akker, W. Bosma, D. Hofs, A. Nijholt, E. Krahmer, C. van Hooijdonk, and E. Marsi. Questions, pictures, answers: introducing pictures in question-answering systems. In Proc. of the 10th. International Symposium on Social Communication, Santiago de Cuba, Cuba, 2007.", "title": "Questions, pictures, answers: introducing pictures in question-answering systems.", "date": "2007"}, {"raw": "W.W. Gaver. Technology aordances. In Proc. of CHI, New Orleans,U.S., 1991.", "title": "Technology aordances.", "date": "1991"}, {"raw": "C. Wharton, J. Rieman, C. Lewis, and P. Polson. The cognitive walkthrough method: A practitioner's guide. In J. Nielsen and R. Mack, editors, Usability Inspection Methods, pages 105{140. New York: Wiley, 1994.", "title": "The cognitive walkthrough method: A practitioner's guide.", "date": "1994"}, {"raw": "J.J. Gibson. The Ecological Approach to Visual Perception. Houghton Miin, Boston, 1979.", "title": "The Ecological Approach to Visual Perception.", "date": "1979", "doi": "10.1002/bs.3830260313"}, {"raw": "M. Oliver. The problem with aordance. In E-Learning, 2(4), pages 402{413, 2005.", "title": "The problem with aordance.", "date": "2005"}, {"raw": "D.A. Norman. The Psychology of Everyday Things. Basic Book Inc., New York,", "title": "The Psychology of Everyday Things.", "doi": "10.2307/3106094"}, {"raw": "J.J. Gibson. The theory of aordances. In R. Shaw and Bransford J., editors, Perceiving, Acting, and Knowing. Hillsdale, New York: Erlbaum, 1977.", "title": "The theory of aordances.", "date": "1977"}, {"raw": "A. Holzinger. Usability engineering for software developers. In Communications of the ACM, 48(1), pages 71{74, 2005.", "title": "Usability engineering for software developers.", "date": "2005", "doi": "10.1145/1039539.1039541"}], "contributors": ["Holzinger, Andreas"], "datePublished": "2008", "description": "Implementation of adequate conversational structures is a key issue in developing successful interactive user interfaces. A way of testing the adequacy of the structures is to prove the correct orientation of each communicative action towards a preceding action. We refer to this orientation leading to a certain response as the affordance of the communicative action. In this paper we present a case study where affordances of implemented conversational structures (including verbal and graphical elements) in a multimodal medical QA system are identified applying Conversation Analysis (CA) tools and tested using the Cognitive Walkthrough (CW) method. The CW method was modified to fit the conversational approach and tested with five expert evaluators. Results showed that the affordance analysis helps detecting inefficient constructions leading to disruptions in the dialog flow, spots unnecessary functions and provides important insights on systems easy-of-use", "fullText": "Affordances in conversational interactions with\nmultimodal QA systems\nAndreea Niculescu\nHMI Group, University of Twente\nniculescuai@ewi.utwente.nl\nAbstract. Implementation of adequate conversational structures is a\nkey issue in developing successful interactive user interfaces. A way of\ntesting the adequacy of the structures is to prove the correct orienta-\ntion of each communicative action towards a preceding action. We refer\nto this orientation leading to a certain response as the affordance of the\ncommunicative action. In this paper we present a case study where affor-\ndances of implemented conversational structures (including verbal and\ngraphical elements) in a multimodal medical QA system are identified\napplying Conversation Analysis (CA) tools and tested using the Cog-\nnitive Walkthrough (CW) method. The CW method was modified to\nfit the conversational approach and tested with five expert evaluators.\nResults showed that the affordance analysis helps detecting inefficient\nconstructions leading to disruptions in the dialog flow, spots unneces-\nsary functions and provides important insights on systems easy-of-use.\nKeywords: Affordances, Conversational Interactions, Multimodal QA systems,\nUsability Inspection, Cognitive Walkthrough\n1 Introduction\nThis paper discusses the design evaluation of a multimodal QA system from the\nperspective of the affordance concept.\nWe claim that many problems associated with natural language based interac-\ntions are originated from the lack of deeper understanding of underlying conver-\nsational structures. Typical Graphical User Interfaces (GUI) use preponderant\nvisual interaction elements. Therefore, the affordance design focuses mostly on\nvisual elements, regarding verbal units as simple cognitive support for the graph-\nics.\nMultimodal QA systems are a special case of GUIs in which graphical elements\nare combined with large text units. The rapport between verbal and graphical\nelements is here reversed: the text units are the one supported by graphical el-\nements being the quintessence of the QA interaction. Consequently, we believe\nthat interaction designers should include more extensive analyses of verbal affor-\ndances when designing interactive systems based on conversational structures,\nlike QA or dialog systems.\n2Since the concept of affordance has been often subject of intense controversies\nin HCI debates once introduced by Donald A. Norman in [1], many researchers\nsuch as Gaver, Hartson and Noman himself struggled with several definitions\nand categorizations in an effort to clarify the concept making it operational for\nevaluations. Section 2 comments in details these theoretical considerations. In\nsection 3 some short examples of practical applications of affordances are pre-\nsented followed by section 4, where the affordance concept is integrated in the\nframework of the conversational analysis protocols. The case study, including\nmethodology, short system overview, questionnaire and scenario design are pre-\nsented in section 5. The results are largely discussed in section 6. This paper\nends with conclusions containing the result summarization and improvement\nsuggestions for the QA interface.\n2 What are affordances?\nThe concept of affordance was developed by the American psychologist J.J. Gib-\nson in [2] and [3] it is a significant part of his ecological theory of direct percep-\ntion. Gibson defined the term \u201daffordance\u201d as latent action possibilities existing\nin an environment independent of the individual\u2019s ability to perceive them. These\naction possibilities are in relation with the actor\u2019s capabilities of action being\nindependent of his culture, prior knowledge or expectations. Any substance, any\nsurface, any layout has some affordances with respect to a certain actor. Accord-\ning to his theory, the action possibilities indicated by affordances are perceived\nvisually in a direct way that does not require mental information-processing ac-\ntivity, i.e. the immediate perception of the environment will inevitably lead to a\ncertain action.\nGibson\u2019s theory of non-conscious information pick-up was criticized as it fails to\nexplain how actors assign meaning to what they see deciding whether to per-\nform an action. Even if the existence of affordances is independent of the actor\u2019s\nexperience and culture, the ability to perceive such affordances may be depen-\ndent on them. Therefore, the actor may need to learn first to discriminate the\ninformation he gets from the environment in order to perceive it directly [4].\nAlthough Gibson rejected the involvement of mental activities in the process of\ndirect perception, he conceded that even for the most \u201dbasic\u201d affordance per-\nception might need to somehow develop - a clear suggestion that learning could\nbe involved.\nOther criticisms to Gibson\u2019s theory refer to the fact that affordances are defined\nas being relational but the nature of the relation between actors and environ-\nment are not further discussed. Even though the theory presents some illustrative\nexamples of affordances, it doesn\u2019t provide an analytic way of identifying affor-\ndances[5].\nDespite all the criticism, Gibson\u2019s theory of affordance brought radical changes in\nthe field of perceptual psychology and was successfully adopted as a key concept\nin other fields, such as cognitive science, robotics, artificial intelligence, design\netc.\n3In the HCI field the term \u201daffordance\u201d was introduced by David A. Norman in\nhis book \u201dThe Psychology of Everyday Things\u201d [1]. Norman adopted Gibson\u2019s\nconcept and used it to address design aspects of artifacts considering technol-\nogy as part of the environment. He defined affordances as physical, \u201dperceived\nand actual properties of the thing, primarily those fundamental properties that\ndetermine just how the thing could possibly be used. [...] Affordances provide\nstrong clues to the operations of things. Plates are for pushing. Knobs are for\nturning. [..]. When affordances are taken advantage of, the user knows what to\ndo just by looking: no picture, label, or instruction needed\u201d [1].\nHowever, even though Norman borrowed the term from Gibson he disagreed\nwith his theory about whether the mind processed or simply \u201dpicked up\u201d infor-\nmation (see [1], [6] and [7]). Consequently, Norman departs from Gibson\u2019s theory\nand considers affordances perceived properties of objects that may or may not\nexist. The perception is determined not by the action capabilities of the actor\nas Gibson says, but by his mental and perceptual capabilities. Moreover, past\nknowledge and experience are tidily coupled with affordances in Norman\u2019s view\n[4].\nFor Norman, the notion of affordance becomes a mixture between actual (shape,\nmaterial, color) and perceived properties of the object where the perceived prop-\nerties are in fact the suggestion of how the object might be used.\nThe lack of separation between physical properties and perceptual information\nabout their use created a rather ambiguous definition of affordance and gener-\nated large discussions about the meaning of the term in the HCI community.\nA substantial contribution to clarify the concept bringing the Gibsonian think-\ning back in the HCI world was made by the interaction designer William W.\nGaver. Similar to Gibson, Gaver argued that affordances do exist independently\nof their perception, being ontologically but not epistemologically relevant. He\ndefined the affordance concept as the property \u201d [...] of the environment relevant\nfor action systems\u201d [8] and proposed a taxonomy where the affordance concept\nis separated from the perceptual information available about it.\nIn Gaver\u2019s framework affordances (aff.) and their perceptual information (per.inf.)\nare defined as entities taking binary values such as \u201dyes\u201d and \u201dno\u201d. Their com-\nbinations result in four types of affordances called: perceptible, hidden, false and\ncorrect rejection.\na. Perceptible affordances: (aff.: yes, per.inf.: yes) offer a link between perception\nand action by signalizing a possible action in a visible way.\nb. Hidden affordances: (aff.: yes, per.inf.: no) offer no link between perception\nand action; actions are possible but there is no signal acknowledging their exis-\ntence (e.g. a hidden door).\nc. False affordances: (aff: no, per.inf.: yes) offer a link between perception and a\nnon-existing action possibility; actions are mistakenly signalized as being possi-\nble (e.g. a door might appear to afford opening, but it won\u2019t afford if it\u2019s locked).\nd. Correct rejection: (aff: no, per.inf: yes) refers to the situation where no action\nis afforded or signalized (i.e. no affordance).\nWhile conveying the categorization to interaction design field designers should\n4avoid false and hidden affordances as being a sign of weak design: false affor-\ndances bring users on a wrong path while hidden affordances waist resources, as\nusers will probably encounter difficulties on detecting their existence. Instead,\ndesigners should concentrate on making affordances perceptible or creating sit-\nuations where the lack of affordance is correctly rejected.\nAnother attempt to extend and refine Norman\u2019s concept of real and perceived\naffordance came from Hartson [9]. Hartson proposed four complementary types\nof affordance in the context of interaction design and evaluation: cognitive, phys-\nical, sensory and functional affordance.\nCognitive affordance -corresponding to Norman\u2019s perceived affordance- is asso-\nciated with the semantics of the interfaces and refers to design features that help\nusers knowing something (e.g. the label of a button indicating what will happen\nif a user clicks on it).\nPhysical affordance -corresponding to Norman\u2019s real affordance- is associated\nwith characteristics concerning the \u201doperability\u201d of the interface, and refers to\ndesign features that help users to accomplish accurately a physical action in the\ninterface (e.g. the size of a button that is large enough to allow users to click on\nit).\nSensory affordance is related to \u201dsense-ability\u201d characteristics of the interface\nand targets design features that help users perceiving (e.g. seeing, hearing, feel-\ning) something (e.g. the font size of a label). Sensory affordance plays a critical\nsupporting role to cognitive and physical affordances.\nThe last category, functional affordance addresses design features that help users\nto accomplish work (e.g. the internal system ability to sort numbers invoked by\na user who clicked the \u2019sort\u2019 button [9]).\nThere are several other interesting interpretations and formulations of the affor-\ndance concept but due to obvious space limitation we discussed in details only\na few of them that we considered as being the most relevant to our analysis.\n3 Practical dimensions of affordances\nThe study of affordance goes beyond theoretical speculation; authors like Vainio\net al. [10] validated the affordance concept as part of an interesting empirical\nstudy. They showed that participants during several tests could identify objects\nfaster if they were congruent with an observed action prime (e.g. power grasp\n- power grasp compatible object) rather than incongruent (e.g. power grasp -\nprecision grasp). They concluded that motor knowledge plays an important role\nin object identification and, consequently, action-related information associated\nwith an (graspable) object is an inseparable element of that object\u2019s representa-\ntion.\nIn the HCI field the affordance concept has found its practical settings as design\nmodel and analysis tool for physical and graphical user interfaces.\nSheridan & Kortuem [11] proposed an affordance-based design model of physical\ninterfaces for ubiquitous environment. They proposed an experimental method\nto study object affordances showing how the method can be applied to the de-\n5sign of concrete physical interface artifacts.\nPing et al. [12] used the affordance design model as theoretical basis and method-\nological underpinning to evaluate an e-learning program on mammogram read-\ning.\nHartson [9] explored the relationship between the affordance types associated\nwith usability problems and provided examples and a methodology scheme for\npractitioners on how to identify affordances issues involved in flawed design cases.\nHowever, the theory of affordance and its practical applications concern mainly\nthe visual perception of environment objects analyzing verbal element only\nmarginally, e.g. only when they are meant to support visual elements (see Hart-\nson\u2019s cognitive affordance). Since new media technologies such as interactive\ninformation systems like QA or dialog systems are design artifacts that use elab-\norated conversational structures with preponderate verbal text elements there is\na strong need to consider such elements as integral parts of the interaction de-\nsign. Therefore, we propose a design evaluation that uses the affordance concept\nto analyze text units and graphical elements.\n4 Affordances in conversational interactions\nAn important characteristic of conversational interactions in general is the fact\nthat they are deeply anchored in the cultural context of use; that means they are\nbased on conventions and constrains of the socio-cultural environment following\na rigorous protocol course.\nContrary to Norman who argued that the socio-cultural world is placed outside\nthe domain of affordance [6], Gaver emphasized the role of the culture together\nwith other factors such as experience and learning involved in the process of per-\nception of affordances [8]. We also believe these factors are not to be considered\naffordances but have an important function in making affordances visible.\nA way of detecting these factors and implicitly affordances in conversational in-\nteractions is to apply conversational analysis (CA). In CA, conversations can\nbe considered \u201denvironments\u201d where certain types of actions such as gestures,\nmimics, verbal statements are afforded in certain circumstances - in terms of non-\nviolating coherence principles and cultural constrains. Interlocutors can express\ntheir communicative intentions both verbally (through speech) and non-verbally\n(through gesture and mimic). Analyzing the organization of each conversational\nsequence it can be determined what kind of action possibilities (affordances)\nhave the participants in a certain moment of the conversation. We consider\nthese affordances from a pragmatic perspective, i.e. action possibilities oriented\nto achieve a certain goal.\nConversations are by nature interactive and follow a relatively strict turn-based\nprotocol. We address in this paper only the case of closed-domain question-\nanswer interactions since the system analyzed in our case deals with this type\nof interaction.\nIn general, question-answer conversations are fully structured in adjacent pairs,\n6meaning that all exchanged turns are functionally related to each other in such\nmanner that the first turn requires a certain type (or range of types) of second\nturn [13]. The adjacent pairs are grouped in three separate categories, corre-\nsponding to a conversation initialization, termination (both containing greeting-\ngreeting pairs) and body sequence (containing question-answer pairs). These\nthree categories form together what we call conversation protocol.\n4.1 Conversation initialization and termination\nStarting and ending a conversation are levels of phatic communication with social\nfunctions: they are responsible for establishing rapport or quitting the interac-\ntion \u201dcircle\u201d in a polite way.\nA conversation usually starts with a signal showing the readiness to engage in\na conversation. Such signals are salutation forms, self-presentation (if the inter-\nlocutors haven\u2019t met before), non-verbal gestures (hand shake, hugging, kissing,\nhand waving, gazing), mimic (smiling), changing the corporal position towards\nthe interlocutor, etc.\nA similar protocol for ending the conversation includes farewells and thanks ex-\nchanging, waving arms, hugs, kisses, glancing away, re-orienting body posture\naway from interlocutor etc.\nEach performed action affords in principle a similar one in return: a greeting,\na smile, a self-introduction affords symmetrical responses from the interlocutor.\nHowever, the realization of conversation initialization might differ across culture\ntaking into account participants\u2019 gender, age, social position and degree of ac-\nquaintance. For example, in Western cultures a stretched out hand will afford\nhand-shaking; in Muslim countries such a gesture will afford hand-shaking usu-\nally if both interlocutors are men; women instead will press their hand upon\nthe chest to signalize salutation response avoiding at the same time the direct\ncontact with the opposite gender.\n4.2 Conversation body\nThe conversation body contains the essential part of the interaction, namely the\ninformation exchange (also called informational communication).\nThe information exchange may start with a short explanation of the intended\nnature of the conversation-to-be preceding (the first question-answer exchange).\nAt this point a common ground (a set of propositions that make up the contex-\ntual background for the utterances to follow) can be established.\nFrom a pragmatic point of view a question may afford following responses: a\nmatching answer, acknowledgment of ignorance, suggestion for asking someone\nelse (re-routing), intermediary questions to clarify a previous question, post-\nponement, refusal to provide an answer, feedback showing that the question was\nunderstood or a request for time to process the question, etc.\n7In case of miscommunication repair strategies occur in form of explanatory ad-\njacent question-answer pairs.\nThe turns can be accompanied by non-verbal cues such as gestures and mimic\nused to emphasize the content; e.g. gaze signalizes attention and readiness for\ninteraction, rising eye-brow shows surprise, smile acknowledges agreement, etc.\nQuestion and answer pairs must respect the coherence principle by being se-\nmantically and meaningfully related to each other. In order to achieve coherent\ninformation exchange, syntactical features such as anaphoric, cataphoric and\ndeictic elements may be used. Also logical tense structure, as well as presuppo-\nsitions and implications connected to general world knowledge are deployed to\ncoherently connect answers to questions[14].\nIn common practice interlocutors do not perform their utterances at the same\ntime. Speakers usually take turns to talk. Overlapping and simultaneous talk is\ngenerally seen in Western cultures as unpleasant. The turn-taking usually occurs\nat the utterance ends, often signalized by silence.\nInterruption might be allowed if one of the interlocutors signalized verbally or\nby gesture the wish to take the turn[15].\n5 Method\nEven though conversational interactions with multimodal systems differ in many\naspects from their human counterpart, they generally follow the same conversa-\ntion protocol consisting of initialization, body sequence and termination. This\nsimilarity is intentionally simulated by designers in order to increase the system\u2019s\neasy-of-use and to make users\u2019 answers predictable.\nAs theoretical framework we adopted Gaver\u2019s taxonomy to identify affordance\nvalues and Hartson\u2019s scheme to establish affordance types. The analysis followed\nthe conversational protocol steps described above.\nThe test was carried out using the Cognitive Walkthrough (CW) method. The\nmethod is a usability engineering tool meant to help designer teams to quickly\nevaluate interaction systems from early stages of development. It does not re-\nquire a fully functioning prototype (as it is the case with the system on which\nthe test was performed) nor users involvement. CW emphasizes cognitive as-\npects, such as learnability by analyzing users\u2019 mental processes required for each\nstep[16]. Design experts perform the test taking into account the potential user\nperspective with the purpose of identifying problems that might arise during the\ninteraction. After each scenario the experts are asked to answer questions stated\nin a questionnaire.\n5.1 The IMIX system\nOur study was performed on IMIX, a multimodal interactive QA system for\nmedical queries with extended follow-up questions functionality.\nIMIX was developed for educational purposes and can deal with medical ency-\nclopedic questions, i.e. general questions that do not require expert knowledge\n8like diagnostic questions or complex medical analysis [17]. The system\u2019s users\nare expected to be people with no professional knowledge of the medical domain.\nThey will probably make use of such services only occasionally. No special train-\ning is required to interact with IMIX.\nFig. 1. Screenshot of the IMIX system\nThe system is primary text-based but allows users to use optionally speech1.\nAttached to IMIX is a talking head called Ruth.\nCognitive research into multimedia has shown that the use of text combined\nwith pictures may substantially contribute to the user\u2019s learning process of the\npresented material [18],[19]. Therefore, the IMIX answers contain text combined\nwith suitable static images. The answers are made-up by matching the query to\ndocument fragments from the data base. The text answers may be both spoken\nby Ruth or displayed on the screen. Optionally, follow-up questions can be for-\nmulated as text, speech or drawing [20].\n5.2 Prior issues to the test\nBefore starting the test the evaluators got paper sheets containing preliminary\ninformation about the test goal, a short description of the term \u201daffordance\u201d, a\n1\nDuring the test only the text-based modality was deployed.\n9detailed explanation of the human conversation protocol and a general descrip-\ntion of the IMIX system.\nAfterwards, the evaluators received the scenarios containing specific tasks to ac-\ncomplish, a list of correct actions required to complete each of these tasks and\na separate questionnaire for each scenario.\n5.3 Scenarios and user profiles\nThree scenarios covering the conversational structures implemented in the IMIX\nsystem were developed. We tried to design the scenarios as pleasant and hu-\nmorous as possible in order to achieve an enjoyable interaction. Each scenario\nfocuses on a specific task.\nIn the first scenario evaluators were asked to put a single question and analyze\nthe corresponding conversation protocol. The scenario identifies the situation of\na naive user with little medical expertise who uses IMIX to find out what means\nRepetitive Strain Injury (RSI).\nIn the second scenario the evaluators had to concentrate on the special case of\nfollow-up questions using drawing and/or typing options. The user profile of this\nscenario corresponds to a subject with search engine expertise and interests in\nthe medical domain. He uses IMIX to find information related to liver functions.\nThe third scenario addresses repair and meta-communication strategies when\nthe answer to a question is not found. The user profile addresses an expert user\nwho uses IMIX for entertaining purposes. He is seeking for information about\nthe SARS virus.\nThe evaluators were asked to keep in mind the aim of the test: to look at the way\nthe user is invited to interact with the system and NOT at the answer quality\nhe/she might get back. They also had the possibility to repeat a scenario several\ntimes if they wish so.\n5.4 Questionnaire design\nFor each scenario a separate questionnaire was developed. The questionnaires\nwere designed in accordance to Cognitive Walkthrough (CW) method[21]. How-\never, the questions were adapted to fit the special case of verbal interactions\nand grouped in three units, each one corresponding to a separate conversation\nprotocol category.\nThe purpose of the questionnaires is to detect affordances of the conversational\nsequences implemented in the protocol. The structure of the questions is similar\nfor each unit. Evaluators have first to detect elements signalizing the current\nprotocol. Then they have to anticipate users\u2019 re-action given a certain conver-\nsational sequence. Furthermore the evaluators have to determine whether the\nusers\u2019 responses are acknowledged by the system and how they will perceive this\nfeedback. Eventually each question unit ends with a question about potential vi-\nolations of the conversation protocol. This last question is meant to catch issues\nthat might have \u201descaped\u201d the evaluator\u2019s observation.\n10\nClassical CW Adapted CW\n1. Will the users be trying to produce 1.0 What elements are used to signalize [a certain action]?\nwhatever effect the action has? 1.1 Does the word [..] suggest [a certain action]?\n1.2 What kind of statements affords the question [..]?\n2. Will users be able to notice that 2.0 Will the users see there is [a certain] option ?\nthe correct action is available? 2.1 Will the users understand this signal as\nan invitation to [do a certain action]?\n3. Once users find the correct action 3.0 Will the users know how to use [a certain] option?\nwill they know that it is the right one?\n4. After the action is taken, will users 4.0 Will there be a feed-back to acknowledge the action\nunderstand the feedback they get? performed by the users?\n4.1 Will the users understand these kind of feed-back?\n5.0 Does the conversational protocol get in any\nway violated?\nTable 1. Classical CW versus Adapted CW\n5.5 Pilot study\nBefore starting the experimental run a first pilot study with one expert evalu-\nator was accomplished. From the pilot study three main observations could be\ngathered:\n1) The relatively high difficulty degree of the question demanded the presence\nof experienced evaluators having some affinity with the CW method.\n2) Typical CW questions like \u201dWill the user notice the conversational starting\nsignals(as the correct action available)?\u201d are too general. Precise formulations\nsimilar to \u201dWill the user understand this signal as an invitation to start a conver-\nsation?\u201d seemed to be more appropriated even if the questionnaire size increases,\ne.g. for each signal one separate question.\n3) Since the answers to the CW questions are often not straightforward, requir-\ning some deliberation time it seemed wise to record the testing session. In this\nway a considerable amount of time could be saved and no observation could get\nlost.\n6 Results\nThe test was completed by five evaluators with design expertise recruited from\nour department. All evaluators except one were novice using the system. The\nresults of their evaluation are summarized below following the conversational\nprotocol categorization:\n6.1 Conversation initialization\nThe conversation initialization implemented in IMIX doesn\u2019t afford symmetrical\nresponse. The conversation\u2019s start is signalized by a textual welcome message, a\nshort system presentation, a \u2019start\u2019 button and a talking head emerging from the\nbackground gazing and rising eyebrows. At this point the only afforded action is\nthe pressing of the \u2019start\u2019 button to begin the \u201dconversation\u201d. No other actions\nlike greetings or salutation gestures in return are afforded, even though according\nto the conversation protocol a greeting affords another greeting. The occurrence\n11\nof a signal (greeting message) combined with the lack of an adequate response\n(no greeting in return) indicates the presence of a false cognitive affordance.\nMost of the evaluators agreed that this stage of the interaction has less resem-\nblance to what is normally called a \u201dconversation\u201d. The talking head appearance\nis not a very convincing invitation to talk even if its blinking eyes indicate a\nwaiting behavior. One of the evaluator argued that the presence of speech, e.g.\na welcome message read by the talking head would increase the users\u2019 feeling of\nbeing involved in a conversation.\nAn adjustment in the head mimic would be beneficial as well: a gazing behaviour\ncombined with smiling is a more appropriate way to start a conversation.\nThe short system presentation was criticized as being too technical: especially\nless experienced users would have difficulties in understanding the meaning of\nhaving a \u201dmultimodal dialog\u201d with the system.\nAlso the text color should be uniform -one criticism addressed the presence of\ncolored words in the text message, a fact that could mislead internet experienced\nusers to click on it, as it is common on websites with embedded links. This would\nbe a false sensory affordance.\nAffordance type Affordance value Action\nPhysical & Cognitive Perceptible Press \u2019start\u2019 button\nSensory & Physical False Click on highlighted words\nCognitive False Give symmetrical response\nTable 2. Affordances in conversation initialization\n6.2 Conversation body\nThe conversation body includes single question-answer sequences, follow-up ques-\ntions, meta-communication and repair strategies.\nSingle question-answer sequences\nThe conversation body begins once the \u2019start\u2019 button is pressed. The users arrive\non a new screen where they receive some brief instructions on how to interact\nwith the system.\nAt this stage of the conversation the users have to choose between two input\noptions: speech or typing. All evaluators agreed that input selection modalities\nseem to be afforded in a proper manner: the buttons are intuitively labeled and\nit was estimated that all user categories won\u2019t have difficulties to select an input\noption. A suggestion was made to use additionally explanatory icons like a pen\nfor the typing option and a microphone for the speech option.\nIt was criticized the presence at the same level of two other buttons: one for the\nstop option and the other one for the new dialog. The \u2019stop\u2019 button should be\nplaced in a corner -in order to be congruent with the typical design of closing\nbuttons while the \u2019new dialog\u2019 button should be removed as being at the moment\nfunctionless and indicating a false physical affordance.\n12\nAnother remark was to adapt the head position towards the typing input field\nwhile users are typing a question in order to increase the interactive feeling and\nto give a certain feedback. Due to the lack of adequate mimic reactions and\nsynchronization with the current conversation stage the talking head gives the\nimpression that it doesn\u2019t belong to the system.\nBy selecting the option \u201dtyping\u201d an input field appears on the same window.\nThe input field affords sentence-like questions as well as keywords. Most of the\nevaluators -excepting one- considered that the full sentence capability won\u2019t be\neasily perceived by more experienced users; they would probably associate the\nsystem\u2019s functionality with the one of a typical search engine and consequently\nwould use keywords. The presence of a relatively extended input field is not a\nclear indication of the expected input and if sentence-like input is desired a short\nhow-to-ask example should be provided. It can be concluded that the input field\nhas hidden physical affordances.\nThe input field is introduced by the question \u201dWhat would you like to ask or\nsay?\u201d. The designer\u2019s intention was to let users know the system is able to han-\ndle different types of statements like full-sentences questions, greetings or even\ntransition formulations (\u201dok\u201d or \u201dthank you\u201d). But being rather too open, the\nquestion suggests it can deal with any kind of statements which certainly is a\nfalse cognitive affordance.\nOn the other hand, most of the evaluators -excepting one- concluded that iron-\nically, especially experienced users would not be aware of what exactly they\ncan utter, e.g. greetings and transition statements affordances remain hidden, as\nnothing clearly indicates their possible usages.\nAffordance type Affordance value Action\nCognitive Perceptible Chose input options\nPhysical Perceptible Type in the input field\nPhysical & Cognitive False Click \u2019new dialog\u2019 button (first dialog screen)\nCognitive False Perform whatever question\nPhysical Hidden Put sentence-like question\nCognitive Hidden Perform greeting\nCognitive Hidden Perform transition statement\nPhysical & Cognitive Hidden Press \u2019new dialog\u2019 button to type\nPhysical & Cognitive Hidden Press \u2019follow-up question\u2019 button to type\nTable 3. Affordances in single question-answer sequences\nWe continue the analysis considering the case where a naive user will enter a\ngreeting. The system will logically respond repeating the same question (\u201dHello!\nWhat would you like to ask or say?\u201d), but won\u2019t indicate how to continue the\ndialog as the input field disappears. So far a direct answer is not afforded. The\nusers need to press the button for either new-dialog or follow-up question in or-\nder to get to the input field to type in, a fact that complicates the conversational\nflow.\nNone of the labeled buttons suits semantically the actual conversational situa-\ntion -the \u2019new dialog\u2019 button should be used in situations where a dialog session\nre-initialization is wanted, while the \u2019follow-up\u2019 button refers to situations where\n13\nusers are looking for more detailed information in the medical answer. Therefore,\nit can be concluded that both buttons afford in this conversational sequence a\nhidden action, namely to allow users to get back to the typing field.\nFollow-up questions\nAfter receiving the first answer the users have the option to continue the infor-\nmation exchange on the same topic by selecting a \u2019follow-up question\u2019 button.\nThe button is labeled with a text indicating that users can type or point at\nsomething in the answer. However, the pointing option is not intuitive and has\nnot a specific usage indication. Besides, not only pointing but also drawing is\nsupported, a fact that the label doesn\u2019t specify. All evaluators agreed on the fact\nthat all user categories wouldn\u2019t know what the option does and how to use it.\nMoreover, it is not clear which advantages it has compared to the typing option.\nTherefore, we identify here a hidden physical affordance.\nIt is also not very clear the way a \u201ddrawn\u201d follow-up question is entered for\nfurther processing. Since the \u2019ok\u2019 button located in the proximity of the input\nfield can be also used for this purpose, the evaluators concluded the button has\nhidden physical and cognitive affordances.\nThere is a feedback to acknowledge the waiting pause and the users\u2019 query but\nthe feedback is not specially meant for follow-up questions, fact that should not\ndisturb the communication flow.\nAffordance type Affordance value Action\nPhysical Perceptible Type in the input field\nPhysical Hidden Use the mouse to \u201ddrawn\u201d a question\nPhysical & Cognitive Hidden Use the \u2019ok\u2019 button to enter a \u201ddrawn\u201d question\nTable 4. Affordances in follow-up questions\nMeta-communication and repair strategies\nWhen the answer of a question is not found the system displays a message re-\nquesting rephrasing. The function of the rephrasing request should additionally\nhelp users to become more successfully in finding the desired information.\nAll evaluators found the request not supportive at all; according to their esti-\nmation even expert users would experience problems rephrasing their question.\nAfter the rephrasing request users can choose between the follow-up question (in\nthe form of typing or pointing) or the new dialog option in order to get cumber-\nsomely back to the typing field. Both options were considered inadequate for this\nparticular stage of the conversation. Just like in the follow-up paragraph these\ntwo buttons indicate the presence of hidden physical and cognitive affordances.\nConversation termination\nThe interaction can be interrupted by clicking the \u2019stop\u2019 button. A real conver-\nsational termination is not afforded. Users don\u2019t have the possibility to verbally\nexpress the intention to leave the conversation, as no typing field was designed\nat this stage of the conversation. They could click on the \u2019new dialog\u2019 button and\n14\nAffordance type Affordance value Action\nCognitive Perceptible Rephrase question\nPhysical & Cognitive Hidden Press \u2019new dialog\u2019 button to rephrase\nPhysical & Cognitive Hidden \u2019Follow-up question\u2019 button to rephrase\nTable 5. Affordances in meta-communication\ntype a farewell greeting as the system affords such statements. But this option\nseems rather a less logical as nobody will probably think to start a new dialog\nwhen in fact he/she wishes to stop it. Besides, even if the system replies logically\nto the farewell greeting it doesn\u2019t allow a verbal termination of the conversation.\nThere is no feedback to acknowledge the end of the conversation and users get the\ngeneral impression of a system crash by clicking the \u2019stop\u2019 button. We certainly\nface a conversation protocol violation.\nAffordance type Affordance value Action\nPhysical & Cognitive Perceptible Press the \u2019stop\u2019 button\nCognitive False No symmetrical response\nTable 6. Affordances in conversation termination\n7 Conclusions\nExtrapolating the affordance definition given by Gaver we considered in this pa-\nper interactive information systems as artificial environments where verbal and\ngraphical elements are artifacts leading users to perform certain actions. There-\nfore we proposed a design evaluation in which not only graphical but also verbal\nelements can be analyzed under the framework of the affordance concept.\nThe results of our experiment revealed several inefficient structures that could be\nidentified analyzing affordances of conversational structures. The conversation\ninitialization and termination implemented in IMIX do not afford a symmetri-\ncal response from the user, perturbing the natural dialog flow. Systems question\nformulations are too open, a fact that might generate false expectations or disori-\nentation. The labeling of buttons should reflect the actions induced by the but-\ntons. A question should automatically generate a response environment avoiding\nunnecessary pressing of additional buttons. The system reactions should be con-\nsistent at all appearance level, i.e. verbal, mimic, gestures.\nThe study of affordances also showed unnecessary functionalities that might be\nremoved or adapted in order to become useful. For example, the presence of\nbuttons leading to certain actions should be in accordance with the conversa-\ntional sequence they are designed for: it makes no sense to start a \u201dnew dia-\nlog\u201d when no other dialog had been started before. The affordance of certain\nconversational structures like greetings in the middle of an interaction shows a\ncooperative behavior. However, it is unlikely that someone would use greetings\nat that particular conversational stage. Special features like pointing or drawing\n15\non a virtual surface should be briefly introduced to users. It is rather unexpected\nthat someone should use an unfamiliar option to ask questions when he/she has\nmore natural choices like typing or speaking.\nThe affordance analysis also provided important observations about the system\u2019s\neasy-of-use. Users may not understand the system\u2019s description, as it seems to\nbe too technical, may not be aware of its full sentence capabilities, may not know\nwhether other transition statements are allowed, may experience difficulties us-\ning the pointing/drawing option or rephrasing their questions and may probably\nfeel annoyed when they expect to be able to type a question and no input field\nis provided.\nLast but not least the affordance analysis of verbal elements proved to be benefi-\ncial and confirmed our initial claim that many problems associated with natural\nlanguage based interaction are originated from the lack of deeper understanding\nof communicative structure: most of false and hidden affordances identified were\ncognitive nature (6 pure cognitive and 6 physical-cognitive out of a total of 14).\nWe concluded that understanding affordance of verbal and graphical elements\nand being aware of their roles in conversational interaction design can help prac-\ntitioners in diagnosing usability problems from early stage of development, since\nthe affordance analysis using CW methods provides a useful and informational\nrich perspective for qualitative evaluations of prototypes without implying costly\nuser studies.\n8 Acknowledgments\nThe author is grateful to Boris van Schooten and Rieks op den Akker for in-\nteresting hints and discussions regarding this research, to Betsy van Dijk for\nuseful comments during the pilot study, to Ronald Poppe, Olga Kulyk, Bart van\nStraalen and Dennis Reidsma for participating in this experiment and to Hendri\nHondorp and Dennis Hops for helping with technical setup. Many thanks to Egon\nL. van den Broek for several valuable theoretical suggestions, to Blasimir Villa\nRodriguez for practical discussions about concepts and careful proof-reading and\nto anonymous reviewers for helpful improvement suggestions.\nReferences\n1. D.A. Norman. The Psychology of Everyday Things. Basic Book Inc., New York,\n1988.\n2. J.J. Gibson. The theory of affordances. In R. Shaw and Bransford J., editors,\nPerceiving, Acting, and Knowing. Hillsdale, New York: Erlbaum, 1977.\n3. J.J. Gibson. The Ecological Approach to Visual Perception. Houghton Mifflin,\nBoston, 1979.\n4. J. McGrenere and W. Ho. Affordances: Clarifying and evolving a concept. In Proc.\nof Graphic Interface, pages 179\u2013186, Montreal, Canada, 2000.\n5. M. Oliver. The problem with affordance. In E-Learning, 2(4), pages 402\u2013413, 2005.\n6. D.A. Norman. Affordances, conventions and design. In Interactions, pages 38\u201343,\n1999.\n16\n7. D.A. Norman. Memory and Attention: an introduction to human information\nprocessing. Wiley & Sons., London, 1969.\n8. W.W. Gaver. Technology affordances. In Proc. of CHI, New Orleans,U.S., 1991.\n9. H. R. Hartson. Cognitive, physical, sensory and functional affordances in interac-\ntion design. In Behaviour & Information Technology, 22(5), pages 315\u2013338, 2003.\n10. L. Vainio, E. Symes, R. Ellis, M. Tucker, and G. Ottoboni. On the relations between\naction planning,object identification, and motor representations of observed actions\nand objects. In Cognition, 108(2), page 444465. Elsevier, 2008.\n11. J.G. Sheridan and G. Kortuem. Affordance-based design of physical interfaces for\nubiquitous environments. In LNCS, Berlin, Heidelberg, 2006. Springer.\n12. P. Luo, E. A. Eikman, W. Kealy, and W. Qian. Analysis of a mammography\nteaching program based on an affordance design model. In Academic Radiology,\n13(12), pages 1542\u20131552, 2006.\n13. E.A. Schegloff and H. Sacks. Opening up closings. In Semiotica 8: 4, pages 289\u2013327,\nNew Orleans,U.S., 1973.\n14. H. Bumann. Lexikon der Sprachwissenschaft. Kroener Verlag, Stuttgart, 3rd.\nedition, 2002.\n15. J. Cassell, T. Bickmore, H. Vilhjalmsson, and H. Yan. More than just a pretty\nface: Affordances of embodiment. In Proceedings of 2000 International Conference\non Intelligent User Interfaces, New Orleans, 2000.\n16. A. Holzinger. Usability engineering for software developers. In Communications\nof the ACM, 48(1), pages 71\u201374, 2005.\n17. M. Theune, B. van Schooten, R. op den Akker, W. Bosma, D. Hofs, A. Nijholt,\nE. Krahmer, C. van Hooijdonk, and E. Marsi. Questions, pictures, answers: intro-\nducing pictures in question-answering systems. In Proc. of the 10th. International\nSymposium on Social Communication, Santiago de Cuba, Cuba, 2007.\n18. R.E. Mayer. Multimedia learning. Multimedia learning, Cambridge (MA), 2001.\n19. A. Holzinger, M. Kickmeier-Rust, and D. Albert. Dynamic media in computer\nscience education; content complexity and learning performance: Is less more? In\nEducational Technology & Society, 11(1), pages 279\u2013290, 2008.\n20. R. op den Akker, H. Bunt, S. Keizer, and B. van Schooten. From question answering\nto spoken dialogue: Towards an information search assistant for interactive multi-\nmodal information extraction. In Proc. of Interspeech, Lissbon, Portugal, 2005.\n21. C. Wharton, J. Rieman, C. Lewis, and P. Polson. The cognitive walkthrough\nmethod: A practitioner\u2019s guide. In J. Nielsen and R. Mack, editors, Usability\nInspection Methods, pages 105\u2013140. New York: Wiley, 1994.\n", "identifiers": ["oai:doc.utwente.nl:62635", null], "language": {"code": "en", "id": 9, "name": "English"}, "publisher": "Springer Verlag", "relations": ["http://doc.utwente.nl/62635/1/Niculescu08affordances.pdf"], "repositories": [{"id": "363", "openDoarId": 0, "name": "Universiteit Twente Repository", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 1, "metadataAdded": 1373473687000, "metadataUpdated": 1488061728000, "depositedDate": 1477350000000}, "subjects": ["Article in monograph or in proceedings"], "title": "Affordances in conversational interactions with multimodal QA systems", "topics": [], "types": [], "year": 2008, "fulltextIdentifier": "https://core.ac.uk/download/pdf/11466311.pdf", "doi": "10.1007/978-3-540-89350-9_16", "oai": "oai:doc.utwente.nl:62635", "downloadUrl": "https://core.ac.uk/download/pdf/11466311.pdf"}}
{"status": "OK", "data": {"id": "81280802", "authors": ["Friederike E. L. Otto", "Suzanne M. Rosier", "Myles R. Allen", "Neil R. Massey", "Cameron J. Rye", "Jara Imbers Quintana"], "citations": [], "contributors": [], "datePublished": "2014", "fullText": "Climatic Change (2015) 132:77\u201391\nDOI 10.1007/s10584-014-1095-2\nAttribution analysis of high precipitation events\nin summer in England and Wales over the last decade\nFriederike E. L. Otto \u00b7Suzanne M. Rosier \u00b7\nMyles R. Allen \u00b7Neil R. Massey \u00b7Cameron J. Rye \u00b7\nJara Imbers Quintana\nReceived: 15 January 2013 / Accepted: 17 February 2014 / Published online: 11 October 2014\n\u00a9 Springer Science+Business Media Dordrecht 2014\nAbstract The crucial question in the public debate of extreme events is increasingly\nwhether and to what extent the event has been caused by anthropogenic warming. In this\nstudy we investigate this question using extreme summer precipitation events in England\nand Wales as an example for probabilistic event attribution using very large ensembles\nof regional climate model (RCM) simulations within the weather@home.net project. This\nallows us to analyse the statistics of high precipitation events in England andWales, a region\nwith a high quality precipitation observational dataset. Validating the model simulations\nagainst observations shows a credible shape of the distribution of 5-day precipitation, and\nthus confidence in the results. While the risk of extreme July precipitation events has at\nleast doubled due to anthropogenic climate change in the modelling framework, no signifi-\ncant changes can be detected for the other two summer months. This study thus highlights\nthe challenges of probabilistic event attribution of complex weather events and identifies\nPart of the EQUIP special issue of Climatic Change\nThis article is part of a Special Issue on \u201cManaging Uncertainty in Predictions of Climate and Its\nImpacts\u201d edited by Andrew Challinor and Chris Ferro.\nF. E. L. Otto (\u0002)\nEnvironmental Change Institute, University of Oxford, South Parks Road, Oxford, OX1 3QY, UK\ne-mail: friederike.otto@ouce.ox.ac.uk\nM. R. Allen \u00b7 N. R. Massey\nEnvironmental Change Institute, University of Oxford, Oxford University Centre for the Environment,\nSouth Parks Road, Oxford, OX1 3QY, UK\nM. R. Allen \u00b7 C. J. Rye \u00b7 J. I. Quintana\nAtmospheric, Oceanic and Planetary Physics, University of Oxford,\nParks Road, Oxford, OX1 3PU, UK\nPresent Address:\nS. M. Rosier\nNational Institute of Water and Atmospheric Research Ltd.,\n301, Evans Bay Parade, Wellington 6021, New Zealand\n78 Climatic Change (2015) 132:77\u201391\nthe need to further decompose atmospheric features responsible for an event to occur for\nquantitative attribution analysis.\n1 Introduction\nIn recent years wet British summers have been reported frequently, with record-breaking\nrainfall in June in 2012 and a wet July following a wet spring, and the exceptionally rainy\nJuly in 2007, not record breaking in the monthly totals but with very intense shorter rain\nevents within the month, which resulted in heavy summer floods in middle England and\nWales. Although large scale flooding held off in 2012, the extreme rain caused damage to\nthe harvest and put strain on infrastructure systems. However the heavy flooding in 2007,\nfollowing very high daily and 5-day totals (with 3 top 12 5-day totals in July compared to\nthe whole observed period, and the 5th highest total daily precipitation), was classified as a\nnational disaster by the Environment Agency causing economic losses of approximately \u00a33\nbillion (Report Environment Agency UK, 2010). There is no significant trend in July rain-\nfall extremes,however, as extremes are by definition rare, a trend would only be detectable if\nthe risk of extremes changed dramatically. Trend analysis will thus not reveal subtle changes\nin extreme precipitation. In times of anthropogenic climate change with increasing global\nmean temperatures the question arises if the observed extreme events were just bad luck or\nwhether heavy rainfall in summer, especially in July, is something we have reasons to expect\nin the future and should begin adapting to in terms of infrastructure planning? Even before\na trend might be detectable, attribution analysis can reveal the influence of anthropogenic\nclimate change on very rare events, and such is the aim of this paper. July precipitation on\na timescale relevant for flooding will be the focus of this study. It is important to highlight\nhere that precipitation in the relevant month, although important, is only one factor deter-\nmining flood risk (Kay et al. 2011). We attempt to answer the question of how the odds of\nextreme July precipitation relevant for flooding have changed, not the odds of floods occur-\nring in July. Increasingly, politicians, decision-makers and non-governmental organisations\n(NGOs) want to know whether and to what extent global human-influenced climate change\nis affecting localised extreme weather events, highlighted by the fact that the Conference of\nthe Party (COP) meeting in Doha in November 2012 established a work programme specif-\nically on loss and damage due to climate change (UNFCCC report FCCC/SBI/2012/29\n2012), further developed into the \u201cWarsaw mechanism\u201d at the COP in 2013. Studies into\nthe European heat wave of 2003 (Stott et al. 2004), the England and Wales floods of 2000\n(Pall et al. 2011), and the Russian heat wave of 2010 (Dole et al. 2011; Rahmstorf and\nCoumou 2011; Otto et al. 2012) have sought to determine to what extent the risks of\nthese events occurring have changed because of anthropogenic greenhouse gas emissions,\nmany of them using the emerging method of probabilistic event attribution (PEA). Under\nthe assumption of an unchanging relationship between hazard and resulting damage, event\nprobability can be seen as a proxy for risk (Pall et al. 2011). One of the most recent series\nof studies in this field (Peterson et al. 2012; Peterson et al. 2013) attempts to answer this\nquestion for several extreme events that occurred in 2011 and 2012, with one of the studies\nbeing an attribution study of exceptionally warm Novembers in central England (Massey\net al. 2012), which explicitly applies PEA. It finds a 40 times increase in the risk of such\nwarm Novembers occurring, when comparing simulated temperatures of the 2000s with the\n1960s. Pall et al. (2011) compare river runoff derived from daily precipitation in the autumn\nof 2000, with an autumn 2000 in a world that might have been without anthropogenic green-\nhouse gas forcing and associated sea surface temperature warming. In our study we use a\nClimatic Change (2015) 132:77\u201391 79\nPEA approach combining both methods, as outlined in the following Section 2. In Section 3\nthe results are presented and analysed with emphasis on the quantification of uncertainty\n(see Section 2.4) and concluding remarks are made in Section 4.\n2 Methodology\nThe aim of this study is to use an approach of probabilistic event attribution (PEA) to\ninvestigate whether the odds of extreme precipitation in July, given the boundary con-\nditions of observed forcings and sea surface temperatures (SSTs), have changed due to\nanthropogenic influences, making use of very large ensembles of regional climate models.\n2.1 Probabilistic event attribution (PEA)\nTo achieve our aim of a probabilistic event attribution study of extreme July precipitation in\nEngland and Wales (such as led to severe flooding in large parts of central England, North-\neast England and Wales), at least two climate model ensembles are required. The first one\nis a historical ensemble using observed forcings and sea surface temperatures (SSTs), while\nthe second is a counterfactual ensemble for the \u201cworld that might have been\u201d without anthro-\npogenic forcing. Both ensembles must be sufficiently large to ensure that the statistical\nanalysis of changes in extreme events is robust (e.g., Pall et al. 2011; Kay et al. 2011). Some\nauthors have also compared the change in event occurrence in model simulations between a\nrecent decade (e.g., 2000s) and an earlier decade when the anthropogenic forcing was not as\nstrong (e.g., 1960s) (e.g., Otto et al. 2012; Massey et al. 2012). Both approaches have their\nadvantages and disadvantages. A comparison of the 1960s with the 2000s does not allow for\nthe changes in probability of occurrence of extreme events to be attributed to anthropogenic\nclimate change alone as other climate conditions have been different in the two decades\nas well. It does, however, allow for a validation of modelling results against observations.\nAdditionally comparing whole decades instead of ensembles of single years minimises the\ninfluence of large scale teleconnection patterns, e.g. the North Atlantic Oscillation (NAO),\nas decadal ensembles smooth interannual variability of such oscillations. However, this does\nnot apply for the Atlantic Multidecadal Oscillation (AMO) pattern which switches modes\nonly on decadal and longer timescales (Sutton and Dong 2012). Especially with respect to\nprecipitation, which is resolvable for large scales but largely either not understood or not\nresolvable even in a regional climate model on local scales that are important for flooding,\nboth aspects - the ability to validate the model and the reduced influence of specific SST\npatterns - are important.\n2.2 Modelling analysis\nAssessing the influence of external drivers (e.g., increased greenhouse gas concentrations\nin the atmosphere) on extreme weather is challenging because the most important events\nare typically rare, so their observed frequency is dominated by chance. In order to com-\npile robust statistics of extreme weather events, large ensembles of model simulations at\nrelatively high resolution are required. This project makes use of the large-ensemble capabil-\nity provided by the on-going climateprediction.net \u2018weather@home\u2019 volunteer computing\nnetwork (Allen 1999; Massey et al. 2006), where members of the public are producing\nmulti-thousand-member ensembles of weather simulations using regional climate models\n(RCMs) of different parts of the world. We have applied the model set-up described in\n80 Climatic Change (2015) 132:77\u201391\nMassey et al. (2012) with the regional climate model (RCM) embedded within a gen-\neral circulation model (GCM). The increased resolution of the RCM results in a more\nrealistic simulation of localised weather events, including high and low temperatures and\nextreme precipitation over a relatively small area (Jones et al. 2004). The standard ensemble\ndescribed below is an initial condition ensemble hindcast experiment over Europe which\nsimulates the historical period 1960\u20132010 including all observed forcings (AF)(we will only\nanalyse the 1960s (AF1960s) and 2000s (AF2000s) in this study) as well as the decade\nfrom 2000 to 2010 with the anthropogenic climate change signal removed (NAT2000s) as\ndescribed in Section 2.3. For the results used in this study, the RCMs run by volunteers\nare at 50 km resolution over Europe driven by a global atmospheric model. This is a rel-\natively low resolution for an RCM but given that natural variability is the largest source\nof uncertainty (Section 2.4) the best methodology to account for this uncertainty (given\nresources are not unlimited) is to trade accuracy for precision and employ large ensembles\nof relatively low resolution. The models used are HadAM3P, an atmosphere only GCM at\n1.25 \u00d7 1.875 degrees resolution, forced with observed SSTs from the HadISST data set\n(Rayner et al. 2003) and the RCM HadRM3P. Both models have been developed by the UK\nMet Office and are based upon the atmospheric component of HadCM3 (Pope et al. 2000;\nGordon et al. 2000) with some improvements in the model physics described in Massey\net al. (2012). Both models are run many hundreds of times with varied initial conditions.\nIn this way, very large ensembles of RCM simulations can be computed, of the order of\nthousands, which in turn allows greater confidence when examining the statistics of rare\nevents. We follow a similar methodology to Massey et al. (2012), which uses very large\nensembles of general circulation models (GCMs) to assess the change in risk of very warm\nNovembers in central England under two different climate scenarios: observed July in the\ndecade between 1960 and 1970 and observed July between 2000 and 2010. In addition\nto that we analyse the same decade, 2000-2010, in a representation of a world that might\nhave been without anthropogenic climate change.\nAlthough our method of comparing the 1960s with the 2000s does not permit clean\ndecoupling of anthropogenic and natural drivers, using decadal long scenarios of precip-\nitation reduces some of the effects of natural variability and allows both scenarios to be\nvalidated against observed data. The main observational dataset used here is the HadEWP\ndata which are part of HadUKP, the UK regional precipitation series provided by the\nMet Office Hadley Centre (Alexander and Jones 2001). As well as the analysis of large\nensembles of at least two scenarios showing different climate conditions, the other crucial\ncomponent of an event attribution study is the validation of the model, investigating whether\nit is capable of representing the extreme event.\nFigure 1 shows quantile-quantile plots as a measure of the model\u2019s ability to repre-\nsent the observed distribution of precipitation. Both model (HadRM3P) and observational\n(HadEWP) data are shown in the two different decades,daily averages as well as the 5-\nday average computed as a running 5-day mean are shown. Pall et al. (2011) showed that\ndaily means are the important timescales to derive river run-off,however, if river run-off is\nnot being computed, 5-day means are a better proxy for flood risk than daily means, giv-\ning some indication of ground saturation and maintaining sub-monthly variability. Fowler\nand Kilsby (2003) demonstrated that multi-day precipitation is an important causal factor\nin floods, and found no significant changes in 1-2 day precipitation in the period 1991-\n2000 compared to earlier decades but significant changes in 5-day events. Furthermore\na relatively low resolution RCM is probably more reliable on 5-day timescales as very\nlocal mechanisms accounting for extreme changes in daily precipitation are generally not\nwell represented by these types of models, especially in the summer months. Fowler et al.\nClimatic Change (2015) 132:77\u201391 81\n0 5 10\n0\n2\n4\n6\n8\n10\nprecipitation mm/day HadEWP\np\nre\nci\np\nit\nat\nio\nn\n m\nm\n/d\nay\n H\nad\nR\nM\n3P 2000s daily average\n0 5 10\n0\n2\n4\n6\n8\n10\nprecipitation mm/day HadEWP\np\nre\nci\np\nit\nat\nio\nn\n m\nm\n/d\nay\n H\nad\nR\nM\n3P 1960s daily average\n0 5 10\n0\n2\n4\n6\n8\n10\nprecipitation mm/day HadEWP\np\nre\nci\np\nit\nat\nio\nn\n m\nm\n/d\nay\n H\nad\nR\nM\n3P 2000s 5\u2212daily running mean\n0 5 10\n0\n2\n4\n6\n8\n10\nprecipitation mm/day HadEWP\np\nre\nci\np\nit\nat\nio\nn\n m\nm\n/d\nay\n H\nad\nR\nM\n3P 1960s 5\u2212daily running mean\nFig. 1 Quantile-quantile plots of observed (HadEWP) and modelled (HadRM3P) daily (left) and 5-day\n(right) July precipitation in England andWales in the decade from 2001\u20132010 (top) and 1961\u20131970 (bottom).\nThe red dots represent, from left to right, the 5, 25, 50, 75 and 95 %-quantiles\n(2007) demonstrated that RCMs are in general more likely to reproduce the observed dis-\ntribution of 5-day events compared to daily precipitation, an analysis our model validation\nbelow supports. The observed data set used for validation gives the area average over Eng-\nland and Wales encompassing roughly a region from 51.3\u201353.2N and 356.2\u20130.3E. This\nregion (51.3-53.2N, 356.2-0.3E) both encompasses the regions of highest flooding in 2007\nin Central and North East England and is large enough to be expected to be represented\nwell in an RCM of this resolution. From this figure and complementary reliability diagrams\n(not shown) it becomes evident that the model underestimates the observed precipitation,\nand that the distribution is quite well modelled within the lowest 50 %-quantile, but allo-\ncates insufficient probability to the upper percentiles of the distribution. The straight line in\nFig. 1 is the linear fit between both distributions, if both lay on this line the distributions\nwould be identical. The fact that this line is not the 1-1 line shows that the magnitude of\nprecipitation in the model does not match observations. Comparing 5-day means of precip-\nitation as shown on the right-hand side of Fig. 1 reveals a much improved goodness-of-fit\neven at the wet tail of the distribution which is important with respect to extreme precip-\nitation events, but also highlights that the magnitude of precipitation is underestimated in\nthe model. Attempting to counteract the bias in magnitude by simply multiplying the mean\n82 Climatic Change (2015) 132:77\u201391\ndifference leads to better reliability of the overall mean but does not improve the fit\nat the tails of the distribution. We thus refrain from applying a bias correction as we\nare only comparing the model with the model so the actual magnitude is irrelevant for\nthe analysis. We concentrate our analysis on 5-day means which give the better dis-\ntribution in the model and are furthermore an important indicator for flood risk (e.g,\nFowler and Kilsby 2003; Fowler et al. 2010). Additionally, the underestimation of the\nobservations is consistent for the 1960s and the 2000s so any changes in frequency and\nmagnitude between the two decades represents a change in the relative risk of extreme\nprecipitation.\n2.3 Modelling \u201cthe world that might have been\u201d\nThe natural-only forcing used to simulate a counterfactual ensemble of the decade 2001-\n2010 is a climate forcing that is identical to the one we observed in the last decade but\nwith greenhouse gas emissions, ozone, SO2 and DMS held at preindustrial levels, and with\nthe corresponding lower SSTs prescribed. It thus simulates a world that might have been\nwithout anthropogenic influences.\nThe attributable twentieth-century greenhouse gas warming in SSTs cannot be found\ndirectly from observations, because observations also contain the signal of both natural\n(e.g., solar and volcanic) and other anthropogenic (e.g. sulphate aerosol) drivers, and inter-\nnal variability. Prior studies (Stott et al. 2006) derived this warming from estimates that\nused established \u2018optimal fingerprinting\u2019 analysis (Hegerl et al. 2007; Stott et al. 2010),\nwhich uses multiple linear regression to compare observed surface temperature change and\nis described in the supplementary text of Pall et al. (2011).\nWe use a new method based on Pall et al. (2011) in which we subtract warming patterns\nfrom observed SSTs. The Met Office state-of-the-art coupled climate model HadGEM2 is\nused to compute delta values (SST difference fields) by subtracting SSTs of a HadGEM2\n\u2018natural\u2019 run over the last decade without anthropogenic greenhouse gas, ozone, SO2 and\nDMS forcing from the same model runs with \u2018all forcing\u2019. These delta SSTs are then\nsubtracted from HadISST SSTs. To reduce noise, the deltas are produced using decadal\naverages. The counterfactual SSTs were then used to estimate the sea ice concentration\nfor the \u2018world that might have been\u2019. As current state of the art GCM sea ice projec-\ntions are inconsistent with observations and with respect to other models (Eisenmann et al.\n2007), we apply an empirically-based method (Rayner et al. 2003) to provide sea ice\nfields for the natural model runs. This method to generate possible sea ice fields is inde-\npendent of GCMs, using a statistical method to fit a quadratic equation to SSTs. Using\nobservational records from HadISST SSTs (SST) and sea ice extent (SIE) the first order\napproximation:\nSST = a \u00d7 SIE2 + b \u00d7 SIE + c\nis derived. Under the assumption that this approximation holds on the time scale of interest\nso that the parameters a, b, and c are time invariant, the sea ice extent (SIE) for given SSTs\ncan be calculated. To test the algorithm this method is used to calculate the observed SIE\nfor the period 1999-2010. Figure 2 shows the result of this validation as well as the sea ice\nextent for the natural climate simulations for the same period and the algorithm is seen to\nperform well.\nClimatic Change (2015) 132:77\u201391 83\nFig. 2 Sea ice extent from 1999-2010 in observed HadISST data (black), from the sea ice algorithm (ALG,\nred), and in the \u201cworld that might have been\u201d (blue)\n2.4 Uncertainty assessment\nThe main source of uncertainty in the regional precipitation in present day climate is the\nnatural variability (Hawkins and Sutton 2011). The modelling framework employed in this\nstudy is ideal to quantify this uncertainty by using a very large ensemble of simulations of\nthe very same climate conditions with slightly varied initial conditions. However, another\nlarge uncertainty, the modelling uncertainty, cannot be addressed as only one model is used\nfor the study. Validating the model against observations gives insights into the shortcom-\nings of the model but cannot quantify this uncertainty. Introducing observations into the\nexperiment reveals another source of uncertainty - the uncertainty within the observations.\nHowever, as observed data sets in this part of the world are fairly consistent, we assume\nthis to be a small part of the overall uncertainty. To validate the model\u2019s ability to reproduce\nobservations, we compare model precipitation with the observed timeseries from HadEWP,\nmaking use of quantile-quantile plots and reliability diagrams (not shown) as statistical val-\nidation tools, described in Section 2.2. England and Wales have long records of observed\nprecipitation, and a very dense network of weather stations. Therefore we might expect the\nuncertainty in this observed dataset to be somewhat smaller than in most other regions.\nFor a qualitative assessment of uncertainty in the observed data we additionally compute\nquantile-quantile plots for the decades of 2001-2010 and 1961-1970 using NCEP reanalyses\nand our model\u2019s data. NCEP is certainly not the ideal data set for regional precipitation but\n84 Climatic Change (2015) 132:77\u201391\nis one of the few spanning the considered timeframe 1960-2010. Furthermore it is derived\nvery differently from HadEWP and is thus arguably suitable to check whether validation\nwith very different data sets gives very different results, which turns out not to be the case.\nThe picture (not shown) represents good agreement but with the 5-day mean data slightly\noverestimating the NCEP data in the higher quantiles. We made no attempt to quantify the\nuncertainty in the observed data sets, but assessments can be found in (Alexander and Jones\n2001; Kalnay et al. 1996). It is important to note that we are not so interested in the model\u2019s\nability to simulate the correct precipitation at the correct time but more to represent the right\ndistribution of precipitation within the relevant decade. The validation analysis is thus done\nseparately for the two analysed decades.\nThe model set-up we use for the first part of this study has been especially designed to\naccount for uncertainty in the initial conditions input to the model. In order to produce a\nrobust statistical analysis, we use an ensemble size of 1741 initial conditions. By current\nstandards in this area of science this is a large ensemble - thus we believe we are able to\nsample the initial condition uncertainty exceptionally well. Using this standard ensemble\nwe focus purely on the initial condition uncertainty - all other aspects, of model structure\nand external forcing, have not been varied at all. Therefore the uncertainties in other input\nvariables cannot be quantified in this approach.\nIn the later part of the study, we examine differences between models representing condi-\ntions of the 1960s and models representing conditions of the 2000s with the anthropogenic\nsignal in the SST forcing removed. The application of this method can give an initial bound\non the uncertainty originating in the fact that we do not know how present day climate\nwithout anthropogenic emissions would be, but is likely to underestimate the full range. To\nmake more comprehensive statements about the true range of uncertainty in this area, fur-\nther studies should apply different warming patterns from different GCMs to remove the\nanthropogenic signal from the SSTs (such as was done by Pall et al 2011). The use of a\nregional climate model driven by a global model, and the consistency of the two models,\nallows us to gain confidence in the model physics on different scales, even though it does\nnot allow us to quantify the uncertainty in these methods explicitly.\nIn this study we have not attempted to quantify any other type of modelling uncertainty,\ne.g. we have not varied any internal model parameters. All other sources of uncertainty\naffecting these results, e.g. stemming from stochastic variability and our own expert\njudgment and interpretation of results, cannot be accounted for in this approach.\n3 Attribution analysis\nTo analyse the results for precipitation from the regional modelling experiment, first two\nseparate ensembles are formed from the data replicating the spatial distribution of central\nEngland and Wales (51.3-53.2N, 356.2-0.3E). The ensembles are: all the Julys occurring in\nthe 1960s and all Julys in the 2000s. Our initial analysis involves comparing the two large\nmodel ensembles that represent simulations of the 1960s (AF1960s) and 2000s (AF2000s)\nrespectively. For each of these decades our ensemble is effectively an initial condition\nensemble across all the years of that decade. Hence for each of the decades we examine\na very large ensemble that represents many hundreds of different realisations of possible\nweather situations in that decade, given the climatological forcing conditions.\nAdditionally a third ensemble simulating the decade of 2001\u20132010 with the anthro-\npogenic greenhouse gas signal removed (NAT2000s) is created. Figure 3 shows all three\nensembles - 5-day average July precipitation in England and Wales is shown, for the 1960s\nClimatic Change (2015) 132:77\u201391 85\n1 10 100 1000\n0\n2\n4\n6\n8\n10\n12\nReturn time (years)\np\nre\nci\np\nit\nat\nio\nn\n in\n m\nm\n/d\nay\nReturn periods of 5\u2212daily average July precipitation in England and Wales \n1961\u22121970\n2001N\u22122010N\n2001\u22122010\nFig. 3 Return times of 5-day mean precipitation in July in the 1960s (blue) and 2000s (red) as well as the\nnatural 2000s (green). The error bars represent the 5-95 % confidence interval after resampling the ensemble,\napplying straightforward bootstrapping\n(blue), the 2000s (red) and the counterfactual 2000s (green). A similar comparison is made\nfor June and August (Fig. 4), which gives a different result which will be used to assess the\nreliability of the result, but our main focus is the July 5-day precipitation shown in Fig. 3.\nThe curves in Fig. 3 reveal very little difference between the 1960s and the 2000s over\nmuch of the lower (drier) end of the ensembles, but a small separation of these two curves\ntowards the higher (wetter) end. Where the curves separate, it is the case that the extreme\nwet events become more likely (shorter return time) under the 2000s (AF2000s, red) condi-\ntions than in the 1960s (AF1960s, blue) . This general result is in keeping with expectations\nfrom Clausius-Clapeyron constraints that, in a warmer climate (such as the 2000s here),\nJune August \nFig. 4 Return times of 5-day mean precipitation in June (left) and August (right) in the 1960s (blue) and\n2000s (red) as well as the natural 2000s (green). The error bars represent the 5-95 % confidence interval after\nresampling the ensemble, applying straightforward bootstrapping\n86 Climatic Change (2015) 132:77\u201391\nprecipitation is expected to occur in higher-volume events. The counterfactual ensemble\nfor the 2000s (NAT2000s, green) shows similar behaviour to the 1960s with the simulation\nof preindustrial precipitation being below the values of present day precipitation through-\nout. In all three ensembles the separation is small but significant especially comparing the\nNAT2000s ensemble with the all forcings simulation AF2000s.\nThis result is repeated but weaker, with less separation of the three curves, in identical\nensembles for August (Fig. 4, right) but cannot be seen in return times of 5-day precipitation\nin June in the three ensembles (Fig. 4, left).\n3.1 Attributable risk\nOnly by explicitly simulating climates including all possible weather states consistent with\nthe climate forcing conditions for the period of interest can the uncertainties be addressed\nassociated with the question: what fraction of the event probability is attributable to the\nanthropogenic drivers? The climate forcing conditions are simulated using observed green-\nhouse gas and aerosol forcings and observed SSTs. Thus not all possible climate states are\nsimulated but possible weather given the historical SSTs. The large initial condition ensem-\nble used addresses the uncertainty in the simulated weather to a high degree and under\nthe assumption of an unchanging relationship between hazard and resulting damage, event\nprobability can be seen as a proxy for risk (Pall et al. 2011). The return times of the 5-day\nmeans in Figs. 3 and 4 are almost straight, parallel, lines comparable to the return times\nof river runoff (Pall et al. 2011) which is in contrast to, e.g., monthly means or shorter\ntimescales (Fowler et al. 2010). The straight parallel lines corroborate the finding of, e.g.,\nStott et al. (2004) and Allen et al. (2007) that for the fraction of attributable risk (FAR) cal-\nculation (Allen 2003), the actual value of the threshold does not matter. Therefore return\ntimes of precipitation based on 5-day means are a good measure for event attribution studies\nas a bias in magnitudes does not matter if the fractional change in return time is independent\nof the threshold. Our choice to concentrate on 5-day means is therefore not only adequate\nin terms of flood risk assessment but also for the independence of the threshold for risk\nattribution.\nResults from the main analysis (Fig. 3) do show a significant change in the return times\nfor a high precipitation event like that which caused the floods in 2007. The highest 5-day\nmean in July 2007 was about 9 mm per day in the observations. However, as the model is\nbiased this value is not directly applicable to the modelling results. In July, a one in 100\nyear event in the natural ensemble (NAT2000s) is a one in approximately 35 year event in\nthe 2000s in present day conditions (AF2000s). At the same time a one in a 100 year event\nin the 1960s (AF1960s) has become approximately a one in 50 year event under present\nday climate conditions (AF2000s). Both ensembles, the historical and the counterfactual\none (AF1960s, NAT2000s), represent possible ways to simulate a world without anthro-\npogenic climate forcing; thus the spread between these two ensembles gives an estimate of\nthe uncertainty in the anthropogenic climate change signal, but probably not the full range\nof uncertainty. The 1960s simulations already include some amount of anthropogenic green-\nhouse gases and aerosols - both types of emissions probably influence precipitation over the\nUK. Sparrow et al. (2013) show that the SST pattern obtained from the GCM HadGEM2\nwhich is used in this experiment (NAT2000s) shows an SST pattern in the North Atlantic\nwhich is in contrast to similar GCM simulations with identical counterfactual climate forc-\nings. The North Atlantic is an important region influencing the weather systems that bring\nrainfall over the British Isles. Further studies using a whole ensemble of SST patterns from\ndifferent climate models will reveal a more comprehensive quantification of uncertainty in\nClimatic Change (2015) 132:77\u201391 87\nthe anthropogenic signal on local weather events. The analysis in this study reveals that the\nrisk of 2007 type July extreme precipitation has more than doubled due to anthropogenic\nclimate change. The same cannot be said for summer precipitation as a whole as results for\nAugust and especially June are less clear (Fig. 4). For comparing our analysis with recently\npublished analysis on very wet summers in Northern Europe (Sutton and Dong 2012) we\nadditionally calculate the surface air temperature and precipitation anomalies for the two\nanalysed decades (Fig. 5).\nRecent analysis of several European extreme weather events (Coumou and Rahmstorf\n2012), including the England and Wales floods of 2007, show that a general increase in\nsuch extreme weather events can be expected in a warming climate. This study, based on\na model, is, for July, consistent with this and another recent study by Sutton and Dong\n(2012), who show an increase in wet summers over Northern Europe, including England\nandWales, comparing two periods of high Atlantic Multidecadal Oscillation (AMO) (1931\u2013\n1960, 1996\u20132009) with a period of low AMO index (1964\u20131993) in observations. Their\ninterpretation suggests a causal relationship between the AMO index, surface air temper-\nature (SAT) anomalies and precipitation over Europe, as well as mean sea level pressure\n(MSLP) over Europe and the North Atlantic. Since the two periods we investigate fall\nalmost (we also analysed the correct phases) within the two contrasting AMO phases and\nour model is driven with observed SSTs, providing thus a realistic AMO, we calculate SAT\nand precipitation anomalies for the decade of the 2000s and the 1960s as well as in the two\nAMO periods (1964\u20131993,1996\u20132009) identified by Sutton and Dong. Figure 5 shows the\nresulting anomalies in the compared periods of the 1960s and 2000s. The anomalies for the\nperiods 1964\u20131993 subtracted from 1996\u20132009 look very similar to the anomalies shown\nin Fig. 5. As causalities are clearer in a model, given that we understand the model mecha-\nnisms and feedbacks, similar results to Sutton and Dong would corroborate the assumption\nthat AMO phases have an influence on wet British summers.\nWhile the SAT anomalies are very similar to the findings of Sutton and Dong (2012) the\nanomalies for precipitation differ whereas the MSLP anomalies (not shown) are difficult to\ncompare because the region analysed in Sutton and Dong (2012) does not match the region\nour regional model comprises. Positive MSLP anomalies are found over the British Isles in\nthe model contrasting with negative anomalies in Sutton and Dong (2012). The precipitation\nanomalies in the model are much smaller than the observed anomalies in Sutton and Dong\nand while the pattern of increased precipitation over Scandinavia is comparable, there is no\nFig. 5 Left: mean summer (JJA) surface air temperature anomalies (in degrees Celsius) over Europe, sub-\ntracting the decadal mean of the 1960s from the decadal mean of the 2000s. Right: mean summer (JJA)\nprecipitation anomalies (in percent) over Europe, subtracting the decadal mean of the 1960s from the decadal\nmean of the 2000s\n88 Climatic Change (2015) 132:77\u201391\nchange in precipitation over the British Isles in the model which is again in contrast to Sutton\nand Dong (2012). Redoing this analysis for the identified AMO phases (not shown) gives\nvery similar results, although the early 1960s are not representing an especially low AMO\nphase. Also comparing the 95 %-quantiles of precipitation rather than mean precipitation\ndoes not give qualitatively different results. A reason for this could be that the mean sea level\npressure in the model is slightly overestimating the Azores High compared to Sutton and\nDong (2012), which could result in an eastward shift of the pattern of increased precipitation\nin the model compared to the observed data. However, the overestimation is rather small\ncompared to the ERA-interim reanalysis (not shown). The fact that we do not reproduce\nSutton and Dong\u2019s results could mean that our model is inaccurate, or that the AMO is not\nthe most important trigger of increased summer precipitation over England and Wales.\n4 Discussion\nIn a similar manner to precipitation, return times for monthly temperatures (not shown)\nshow an increase of temperatures for July in central England andWales comparing the 2000s\nensemble with the decade of 1961-1970 (e.g., Massey et al. 2012). The increase in the odds\nof a wet July is a credible result given the thermodynamic fact that a warmer atmosphere can\ncontain more water vapour; assuming precipitation extremes are constrained to change with\nthe water vapour-holding capacity of the atmosphere, the Clausius-Clapeyron relation leads\nto an expectation of increased precipitation, under conditions of constant relative humidity.\nThis argument is typically invoked in the aftermath of floods as an explanation for possible\nincreases in such severe wet events under an anthropogenically warming climate, and is\ngiven in more detail in Pall et al. (2011).\nOur findings for July corroborate this argument but, since the results for the other two\nsummer months are less clear, we infer that atmospheric dynamics also play an important\nrole. The fact that our analysis of the AMO does not match the observed findings of Sut-\nton and Dong (2012) could be an indication that important processes in the model are not\nwell represented, leading to an inaccurate precipitation signal in the model. However, the\nmismatch could also indicate that the AMO is not the most important trigger (see also,\ne.g., Dieppois et al. 2013) of the increased precipitation but that other climatological drivers\nsuch as, e.g., the location of the jet-stream, are driving the observed increased precipitation\nand MSLP anomalies. The less clear, if any, signal for June and August 5-day precipitation\nindicates that more complex mechanisms than warmer SSTs alone are triggers of changes\nin summer precipitation in England and Wales. Thus, a logical further study should inves-\ntigate whether the change in return time of summer precipitation can be associated with a\ndisplacement of the jet stream.\n4.1 Implications of uncertainty assessment\nThe design of the modelling approach in much of this study aims to quantify the\nuncertainty in natural variability by simulating large initial condition ensembles very\ncomprehensively; thus our conclusions are unlikely to underestimate this source of\nuncertainty. However, given the current general lack of skill in simulating precipitation\n(as opposed to, for example, temperature), the uncertainty in the model structure and\nparameters have been shown (Fowler et al. 2010) to be of importance for the attribu-\ntion of extreme precipitation events. The addition of a perturbed parameter ensemble\nwould enable us to quantify more accurately the uncertainty in extreme precipitation\nClimatic Change (2015) 132:77\u201391 89\nevents, and this represents a promising avenue for further studies which is especially\nachievable with the climateprediction.net project.\nThe uncertainty arising from the fact that we do not know how the world might have been\nwithout anthropogenic climate change has been addressed by using two different ensembles,\nan ensemble of the 1960s including all observed forcings (AF1960s) and a counterfac-\ntual ensemble of the decade 2000\u20132010 excluding anthropogenic greenhouse gas forcing\nby using SSTs with the anthropogenic warming pattern removed (NAT2000s). These two\nensembles give some quantification of this uncertainty but most likely this is not com-\nprehensive, as discussed above (Section 2.4). This study could also be furthered, and the\nuncertainty in its conclusions potentially reduced, if not only the extreme event itself were\nanalysed, but also the weather conditions leading up to the event, and the larger-scale circu-\nlation patterns as well. A more detailed description of the flood event in the context of the\nweather of the spring and summer of 2007, therefore, could well lead to a more accurate\nanalysis of the change in risk of such an event occurring.\n4.1.1 Outcome variable\nThe variable quantified in our study is 5-day July precipitation in England and Wales, in the\nsense that we compare the estimated return period of specific levels of rainfall events in two\ndecades separated by 40 years, and a counterfactual decade. We are able to estimate return\nperiods because of having an extremely large ensemble of different realisations of these two\ndecades (initial condition ensemble).\n4.1.2 Consequence statement\nResults from our study indicate that in the warmer-climate world of the 2000s we might\nexpect 5-day high rainfall events in July to be heavier and occur more frequently, as com-\npared to the 1960s and a counterfactual simulation of the last decade. A change in the pattern\nof how heavy rainfall events occur has far-reaching consequences for many aspects of soci-\nety, such as the altered risks of flooding, and changes in water availability due to altered\nreservoir management needs. In this study we have not made any quantitative estimate of\nthe impacts of such an altered pattern of rainfall events since we have not used our weather\nevent data as input to an impacts model of any kind. Our spread of results could, however,\nbe supplied to river catchment/flood models to indicate the anticipated spread of results in\nthe likelihood of various changes to flood and water availability risks. It is however impor-\ntant to decompose drivers of precipitation to get a clear view of the chain of causality for\nprecipitation changes in England and Wales.\nAcknowledgments The authors would like to thank the climateprediction.net participants whose generous\ndonation of their spare computer processing power has enabled such large model ensembles to be created. The\nauthors are also indebted to the paper reviewers for helpful comments that greatly improved the manuscript.\nReferences\nAlexander LV, Jones PD (2001) Updated precipitation series for the U.K. and discussion of recent events.\nAtmos Sci Lett 1:1530\nAllen MR (1999) Do-it-yourself climate prediction. Nature 401:642\nAllen MR (2003) Liability for climate change. Nature 421:891\u2013892\n90 Climatic Change (2015) 132:77\u201391\nAllen M, Pall P, Stone D, Stott P, Frame D, Min S-K, Nozawa T, Yukimoto S (2007) Scientific challenges\nin the Attribution of harm to human influence on climate, vol 155. The University of Pennsilvania Law\nReview, pp 1353\u20131400\nChatterton J, Viavattene C, Morris J, Penning-Rowsell E, Tapsell S (2010) The costs of the summer 2007\nfloods in England. Environment Agency, Rio House, Waterside Drive, Aztec West, Almondsbury,\nBristol, BS32 4UD. ISBN: 978-1-84911-146-1\nCoumou, Rahmstorf (2012) A decade of weather extremes. Nat Clim Chang 2:491\u2013496\nDieppois B, Durand A, Fournier M, Massei N (2013) Links between multidecadal and interdecadal climatic\noscillations in the North Atlantic and regional climate variability of northern France and England since\nthe 17th century. J Geophys Res Atmos 118:4359\u20134372. doi:10.1002/jgrd.50392\nDole R, Hoerling M, Perlwitz J, Eischeid J, Pegion P, Zhang T, Quan X-W, Xu T, Murray D (2011) Was\nthere a basis for anticipating the 2010 Russian heat wave Geophys Res Lett 38:L06702\nEisenmann I, Untersteiner N, Wettlaufer JS (2007) On the reliability of simulated Arctic sea ice in global\nclimate models. Geophys Res Lett 34:L10501\nFowler HJ, Cooley D, Sain SR, Thurston M (2010) Detecting change in UK extreme precipitation using\nresults from the climateprediction.net BBC climate change experiment. Extremes 13:241\u2013267\nFowler HJ, Ekstro\u00a8mM, Blenkinsop S, Smith AP (2007) Estimating change in extreme European precipitation\nusing a multi-model ensemble. J Geophys Res - Atmos 112:D18104. doi:10.1029/2007JD008619\nFowler HJ, Ekstro\u00a8m M (2009) Multi-model ensemble estimates of climate change impacts on UK seasonal\nprecipitation extremes. Int J Climatol 29(3):385\u2013416\nFowler HJ, Kilsby CG (2003) A regional frequency analysis of United Kingdom extreme rainfall from 1961\nto 2000. Int J Climatol 23(11):1313\u20131334\nGordon C, Cooper C, Senior CA, Banks H, Gregory J, Johns T, Mitchell J, Wood R (2000) The simulation of\nSST, sea ice extents and ocean heat transports in a version of the Hadley centre coupled model without\nflux adjustments. Climate Dyn 16:147\u2013168\nHawkins E, Sutton R (2011) The potential to narrow uncertainty in projections of regional precipitation\nchange. Clim Dyn 37:407\u2013418\nHegerl G, Zwiers F, Braconnot P, Gillet N, Luo Y, Marengo J, Nicholls N, Penner J, Stott P (2007)\nUnderstanding and attributing climate change. Cambridge University Press\nJones RG, Noguer M, Hassell D, Hudson D, Wilson S, Jenkins G, Mitchell J (2004) Generating high\nresolution climate change scenarios using PRECIS. Met Office Hadley Centre, Exeter, UK\nKalnay E et al. (1996) The NCEP/NCAR 40-year reanalysis project. Bull Am Meteorol Soc 77:437-471\nKay AL, Crooks SM, Pall P, Stone DA (2011) Attribution of Autumn/Winter 2000 flood risk in England to\nanthropogenic climate change: a catchment-based study. J Hydrol 406:97\u2013112\nManley G. (1974) Central England temperatures: monthly means 1659 to 1973. Q J Royal Meteorol Soc\n100:389\u2013405\nMassey N, Aina T, Allen M, Christensen C, Frame D, Goodman D, Kettleborough J, Martin A,\nPascoe S, Stainforth D (2006) Data access and analysis with distributed federated data servers\nin climateprediction.net. Adv Geosci 8:49\u201356\nMassey N, Aina T, Rye C, Otto FEL, Wilson S, Jones RG, Allen MR (2012) Have the odds of warm Novem-\nber temperatures and of cold December temperatures in central England changed Bull AmMeteorol Soc\n93:1057\u20131059\nOtto FEL, Massey N, van Oldenborgh GJ, Jones RG, Allen MR (2012) Reconciling two approaches to\nattribution of the 2010 Russian heat wave. Geophys Res Lett 39:L04702\nPall P, Aina T, Stone DA, Stott PA, Nozawa T, Hilberts AGJ, Lohmann D, Allen MR (2011) Anthropogenic\ngreenhouse gas contribution to flood risk in England and Wales in autumn 2000. Nature 470:382\u2013386\nPeterson TC, Hoerling MP, Stott PA, Herring S (eds) (2013) Explaining extreme events of 2012 from a\nclimate perspective, vol 94\nPeterson TC, Stott PA, Herring S (eds) (2012) Explaining extreme events of 2011 from a climate perspective,\nvol 93\nPope VD, Gallani ML, Rowntree PR, Stratton RA (2000) The impact of new physical parameterizations in\nthe Hadley centre model: Hadam3. Climate Dyn 16:123\u2013146\nRahmstorf S, Coumou D (2011) Increase of extreme events in a warming world. PNAS 108:12905\u2013\n17909\nRayner NA, Parker DE, Horton EB, Folland CK, Alexander LV, Rowell DP, Kent EC, Kaplan A (2003)\nGlobal analyses of sea surface temperature, sea ice, and night marine air temperature since the late\nnineteenth century. J Geophys Res 108:4407\nSparrow S, Huntingford C, Massey N, Allen MR (2013) The use of very large atmospheric model ensemble\nto assess potential anthropogenic influence on the UK summer 2012 high rainfall totals[in Explaining\nextreme events of 2012 from a climate perspective]. Bull Am Meteorol Soc 94:S36\u2014S41\nClimatic Change (2015) 132:77\u201391 91\nStott PA, Gillett NP, Hegerl GC, Karoly DJ, Stone DA, Zhang X, Zwiers F (2010) Detection and attribution\nof climate change: a regional perspective. WIREs Clim Chg 1:192\u2013211\nStott PA, Kettleborough JA, Allen MR (2006) Uncertainty in continental-scale temperature predictions.\nGeophys Res Lett 33:L02708\nStott PA, Stone DA, Allen MR (2004) Human contribution to the European heatwave of 2003. Nature\n432:610\u2013614\nSutton RT, Dong B (2012) Atlantic Ocean influence on a shift in European climate in the 1990s. Nature\nGeosci 5:788\u2013792\nUNFCCC report FCCC/SBI/2012/29 (2012) Report on the regional expert meetings on a range of approaches\nto address loss and damage associated with the adverse effects of climate change, including impacts\nrelated to extreme weather events and slow onset events. Subsidiary Body for Implementation Thirty-\nseventh session Doha, 26 November to 1 December 2012\n", "identifiers": ["10.1007/s10584-014-1095-2"], "publisher": "Springer Nature", "relations": ["http://dx.doi.org/10.1007/s10584-014-1095-2"], "repositories": [{"id": "2612", "openDoarId": 0, "name": "Springer - Publisher Connector", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 1, "metadataAdded": 1493497394000, "metadataUpdated": 1494850560000, "depositedDate": null}, "subjects": ["journal-article"], "title": "Attribution analysis of high precipitation events in summer in England and Wales over the last decade", "topics": [], "types": [], "year": 2014, "fulltextIdentifier": "https://core.ac.uk/download/pdf/81280802.pdf", "doi": "10.1007/s10584-014-1095-2", "downloadUrl": "https://core.ac.uk/download/pdf/81280802.pdf"}}
{"status": "OK", "data": {"id": "41236113", "authors": ["Lupton, Deborah"], "citations": [], "contributors": [], "datePublished": "2012-04-05", "description": "The new mobile wireless computer technologies and social media applications using Web 2.0 platforms have recently received attention from those working in health promotion as a promising new way of achieving their goals of preventing ill-health and promoting healthy behaviours at the population level. There is very little critical examination in this literature of how the use of these digital technologies may affect the targeted groups, in terms of the implications for how individuals experience embodiment, selfhood and social relationships. This article addresses these issues, employing a range of social and cultural theories to do so. It is argued that m-health technologies produce a digital cyborg body. They are able to act not only as prostheses but also as interpreters of the body. The subject produced through the use of m-health technologies is constructed as both an object of surveillance and persuasion and as a responsible citizen who is willing and able to act on the health imperatives issuing forth from the technologies and to present their body/self as open to continual measurement and assessment. The implications of this new way of surveilling the body\u2019s health are discussed", "fullText": "1 \n \n \n \n \n \nM-HEALTH AND HEALTH PROMOTION: THE DIGITAL CYBORG AND \nSURVEILLANCE SOCIETY \n \n \nDeborah Lupton, Department of Sociology and Social Policy, University of Sydney \n \nEmail: deborah.lupton@gmail.com \n \nThis article is currently unpublished and has been submitted to Social Theory & Health. Its \ndate of publication is 2011.  \n2 \n \nM-HEALTH AND HEALTH PROMOTION: THE DIGITAL CYBORG AND \nSURVEILLANCE SOCIETY \n \nAbstract \nThe new mobile wireless computer technologies and social media applications using Web 2.0 \nplatforms have recently received attention from those working in health promotion as a \npromising new way of achieving their goals of preventing ill-health and promoting healthy \nbehaviours at the population level. There is very little critical examination in this literature of \nhow the use of these digital technologies may affect the targeted groups, in terms of the \nimplications for how individuals experience embodiment, selfhood and social relationships. \nThis article addresses these issues, employing a range of social and cultural theories to do so. \nIt is argued that m-health technologies produce a digital cyborg body. They are able to act not \nonly as prostheses but also as interpreters of the body. The subject produced through the use \nof m-health technologies is constructed as both an object of surveillance and persuasion and \nas a responsible citizen who is willing and able to act on the health imperatives issuing forth \nfrom the technologies and to present their body/self as open to continual measurement and \nassessment. The implications of this new way of surveilling the body\u2019s health are discussed.    \n \n \nKey words: m-health, digital technologies, cyborgs, health promotion, social theory, the body  \n3 \n \nIntroduction \nI recently attended a one-day symposium on the topic of e-health and social media. There I \nheard presentations from academics working in medicine and public health about the \npossibilities of using social media such as Facebook, YouTube, Twitter, blogs and wikis and \nmobile wireless computer technologies such as smartphones and tablet computers to promote \nhealth. Presenters discussed how integrating social media apps with mobile wireless \ncomputers allowed for the \u2018personalising\u2019 of health messages, \u2018reaching into people\u2019s \neveryday lives\u2019 by sending them messages tailored to their individual health concerns, \nconditions and problems. Thus, for example, automated SMS or emails could be individually \ntargeted and personalised: doctors could contact patients directly to remind them to adhere to \ntheir treatment programs, health promoters could encourage people daily or hourly to take \nmore exercise, avoid excessive alcohol consumption or smoking or eat healthy foods. One \ndiabetes expert spoke of \u2018smart pillboxes\u2019, which were wireless devices that could not only \nremind patients to take their medication but also alert a patient\u2019s doctor from their home if the \npatient had failed to conform to their medication regimen. A health promotion academic \nexcitedly described the potential offered by programs within mobile wireless technologies \nsuch as accelerometers and GPS systems, which could locate individuals spatially and inform \nthem they were near a park, for example, and thus could take the opportunity to have a walk, \nor note that they had not moved much in the past hour and therefore needed to spring into \naction in the interests of their health.  \nVariously referred to as \u2018e-health\u2019 or \u2018m-health\u2019 (\u2018m\u2019 as an abbreviation of \u2018mobile\u2019) \ntools, such devices can be taken almost anywhere and can connect wirelessly to the internet \nfrom most locations. Their users, therefore, are potentially always digitally connected and \ntherefore always reachable in some form. As noted above, even their bodily movements and \ngeographical location can be identified and recorded remotely. \n4 \n \nHealth promotion journals are also beginning to report upon the importance of using \nthe new social media and mobile devices to promote health (see, for example, an editorial by \nCatford, 2011). Health promoters have described the use of \u2018real-time feedback\u2019 of users\u2019 \nhealth status and \u2018prompts\u2019 and \u2018motivation\u2019 messages to \u2018change unhealthy lifestyle habits\u2019 \nvia social media platforms and mobile devices , with reference to controlling such behaviours \nas smoking, alcohol consumption, exercise, diet and sexual behaviour  (Mays et al., 2010; \nLaakso et al., 2011). One study, for example, reported the use of mobile devices to collect \ndaily information about alcohol consumption among a group of American college students, \nreferring to the devices\u2019 ability to administer \u2018just in time\u2019 interventions to intercept \nunhealthy behaviours as they happen in real-time (Mays et al., 2010). Such researchers \nfrequently make reference to linking health-preventive strategies using m-health devices with \n\u2018acceptance of greater personal consumer responsibility for healthy lifestyles\u2019, as Mays et al. \n(2010: 311) put it.  \nThe use of mobile devices in health promotion endeavours represents a significant \nshift in the methods of health promotion. Health promotion has traditionally been a low tech \narea of public health in comparison with the vast array of medical technologies used in the \nclinical setting. The primary use of technology in health promotion has tended to be in \nemploying communication media to disseminate illness-prevention messages to a wide \naudience. Health promotion has borrowed extensively from commercially-oriented social \nmarketing, advertising and public relations approaches and methods to do so. These \nindustries are now embracing social media and mobile devices as part of their publicising \nefforts. Here again, therefore, health promotion can be seen to be taking the lead from \ncommercial enterprises which are directed at marking and selling commodities.  \nBoth health promotion and commercial social marketing have used internet websites \nextensively as part of their publicity campaigns. Recent health promotion campaigns have \n5 \n \nincluded the opportunity to interact in an online support or discussion group, or to post and \nreceive messages on Facebook or Twitter about a health-related issue. For example, the \nAustralian \u2018Swap It, Don\u2019t Swap It\u2019 and the American \u2018Let\u2019s Move\u2019 campaigns, both of \nwhich are directed at weight reduction and increased exercise, provide online support, blogs, \nopportunities for participants to log personal information and links to Facebook pages and \nTwitter. What the new social media and mobile devices provide which differs from older uses \nof the internet \u2013 that is, Web 2.0 technologies compared with those offered by Web 1.0 \u2013 is \nthe opportunity to directly tailor and target health messages on an individual level, to \nintensify the pervasiveness of these messages and to surveill aspects of embodiment of users \nof mobile devices.  \nA vast number of commercial apps have been generated since the advent of \nsmartphones and tablet computers, many of which are directed at consumers who wish to \nmonitor their exercise, alcohol consumption or eating habits in the interests of improving \ntheir health or losing weight. Running programs, for example, can be downloaded to one\u2019s \nsmartphone or tablet computer which are able to record the number of kilometres run each \nsession, the route taken, automatically report these details to one\u2019s followers on social media \nsites, suggest new routes and remind the user that she or he has not run for a few days. Other \napps allow users to enter details of their meals or even take photographs of the food and then \nanalyse the meals for their nutritional value and kilojoule content. The development of \nsimilar apps by government health bodies as part of health promotion campaigns, therefore, is \nan attempt to build on the popularity of such apps and to exploit their potential for recording \ninformation about an individual\u2019s exercise or dietary habits and providing constant reminders \nto engage in health-promoting behaviours. \nWriters from medical and health promotion backgrounds about the new social media \nand mobile devices tend to confine themselves in their discussions to describing how these \n6 \n \ntechnologies could be most effectively used as tools in their efforts to help people deal with \nmedical conditions or improve their general health and wellbeing. From a sociological \nperspective, a more critical analysis may be undertaken of how these technologies may \noperate to construct various forms of subjectivities and embodiments, including identifying \nthe kinds of assumptions that are made about the target of these technologies and what the \nmoral and ethical ramifications of using them may be. The remainder of this article addresses \nthese issues, drawing upon a range of social and cultural theory to do so. \nTechnologies, health and the body/self \nMedical and health promotion discourses represent technologies as inert devices, fixed in \ntheir meaning. From the perspective of sociocultural studies of science and technology, \nhowever, technologies, including those used for medical purposes or health promotion, are \ndynamic and heterogeneous, constantly shifting in their meanings according to the context in \nwhich they are used. Such devices are viewed as \u2018actants\u2019 in a network of configuration in \nwhich non-human objects are viewed as equally as agential as are humans. Technologies \nbestow meaning and subjectivity upon their users, just as users shape the technologies and \ngive them meaning as they incorporate them into their everyday practices. Technologies \nassume certain kinds of capacities, desires and embodiments; they also construct and \nconfigure them. Further, technologies are never politically neutral, but rather are always \nimplicated in complex power relationships (Hadders, 2009; Mort et al., 2009; Mort and \nSmith, 2009; Casper and Morrison, 2010; Mansell, 2010). \nThe relationship between the human body and computerised technologies began to \nreceive attention from social and cultural theorists in the 1980s. The concept of the cyborg \nhas particularly inspired cultural theorists who have written about the implications of \ncomputerised technologies for human embodiment and subjectivity. One of the most \ninfluential scholars on this topic, Donna Haraway (1988) argued that individuals in \n7 \n \ncontemporary western societies had become cyborgs in their interaction with technologies, \nblurring the distinction between human and machine. Human bodies now interact with a vast \nnumber of technologies on a daily basis, ranging from spectacles, hearing aids and telephones \nto bicycles, aeroplanes and cars, all of which change, extend or enhance human\u2019s physical \ncapacities and capabilities. For cultural theorists writing on cyborg bodies, the human-\nmachine hybrid is complex and shifting, calling into question taken-for-granted assumptions \nabout the oppositions between organic/inorganic, natural/artificial and self/Other (Haraway, \n1988; Lupton, 1995a; Tomas, 1995; Freund, 2004; Shildrick, 2010). \nThe concept of the cyborg itself draws from the metaphor of the human body which \ndepicts it as a machine. This metaphor has been dominant in thinking about the body since \nthe industrial revolution, when machines began to have a major influence on people\u2019s \nworking and living habits. Since the age of the computer, bodies have frequently been \nunderstood as computerised systems and the human brain, in particular, is represented as an \norganic computer, with hardware, memory networks and filing systems and so on (Lupton, \n2003). So too, the immune system is frequently portrayed as a mechanical system, and \ndisease or illness are viewed as the result of an information system malfunction (Haraway, \n1989; Martin, 2000). Given the prevailing portrayal of the body as a complex information \nnetwork and disease as a communication breakdown, medicine itself has become represented \nas a system of information gathering and synthesis, to the extent that \u2018mechanical medicine\u2019 \nis being replaced by \u2018information medicine\u2019.  \nAs part of this change in representations of the body and the growing use of \ncomputerised information systems in medicine, the internal organs and workings of the body \nhave moved from being exclusively the preserve of medical students and surgeons to being \nopen to the gaze of all. Online technologies now allow anyone with access to a computer to \nview highly detailed visual images of the inside of the body. So too, the notion of patients \n8 \n \nplacing themselves under the care of a doctor and seeking their expert advice has moved to \nthe concept of patients as producing health knowledges and as acquiring expert knowledge so \nas to manage their illness themselves (Nettleton and Burrows, 2003; Nettleton, 2004; Mort et \nal., 2009; Mort and Smith, 2009). Nettleton and Burrows (2003; Nettleton, 2004) use the \nterm \u2018e-scaped medicine\u2019 to denote the recent shift in the location of medical knowledge and \npractice from the medical school and the clinic to diffuse digital information technologies \nsuch as the internet and telemedicine devices.  \nThese shifts in representation, knowledge and practice have led to the body not only \nbeing thought about and visualised in different ways, but experienced differently. The \nconcepts of the cyborg and \u2018e-scaped\u2019 or \u2018information medicine\u2019 have clear resonances for m-\nhealth initiatives. There have been claims that regular use of computerised devices shapes \nphysical aspects of human embodiment, including changing brain structure and functioning, \nor consciousness, modes of seeing and operating within the world (Lupton, 1995a; De Mul, \n1999; Kapitan, 2009). Mobile wireless devices also have implications for how bodies may \noperate and function. As observed above, these technologies, particularly smartphones which \ntend to be carried on or very close to one\u2019s person throughout the day, are able to monitor and \nmeasure their users\u2019 behaviours, including their bodily movements. Data may be collected on  \nusers\u2019 bodies, fed to the Web 2.0 platform for processing and interpreting, and then given \nback to users to allow them to monitor themselves. Others, including not only health \nprofessionals but also friends and contacts on social sites, may be informed of these data. \nThese technologies thus have a \u2018feedback\u2019 or cybernetic mechanism in that they are reactive \nand active in their relationships with their users as opposed to passively providing \ninformation. Such technologies become prostheses, or technological extensions of the body.  \nM-health and the surveillance society \n9 \n \nSurveillance used for medical or public health purposes operates on different levels, from the \nindividual, interpersonal clinical level to the national or global population level. Thus, at the \nglobal or national level, health surveillance systems are used to record and monitor cases of \nillness, conditions such as obesity or infection to maintain records of epidemiological \nchanges in disease or illness patterning. In the individual medical encounter, doctors practice \na type of personalised surveillance over each of their patients, testing, measuring and \ninvestigating features of patients\u2019 bodies, constructing and maintaining health records, noting \npatients\u2019 adherence to their advice and so on. Medical technologies have for centuries been \nemployed to render the body more visible, to exert a biopolitical gaze upon bodily structure \nand function (Foucault, 1975; Armstrong, 1995; Casper and Morrison, 2010). Mobile \nwireless devices are the contemporary end of a long line of such surveilling devices. \nTelemedicine and telecare technologies now enable health care professionals to examine and \ndiagnose patients\u2019 bodies remotely. Mobile devices allow for many biometric readings to take \nplace from any location. Devices implanted into the body have increasingly used software \nwhich allows them to communicate wirelessly with medical professionals, irrespective of the \npatient\u2019s physical location.  \nThe use of m-health in health promotion extends the temporal nature of health \nsurveillance, and allows for further refinements of the categorising and identifying of \u2018risk \nfactors\u2019 and \u2018at risk groups\u2019 that are then deemed eligible for targeting. Health-related data \nmay easily and frequently be collected from users\u2019 mobile devices each time they log on to \nthe relevant app. Such devices thus offer an unprecedented opportunity to monitor and \nsurveill individuals\u2019 health-related habits. These technologies are now becoming used not \nonly to facilitate medical supervision and monitoring of ill bodies, but now are being \nextended into the realm of well bodies in the attempt to prevent illness and disease. \n10 \n \nAs this suggests, central to a critical analysis of the use of the new social media and mobile \ndevices to promote health is a recognition of these technologies as part of \u2018surveillance \nsociety\u2019, a term used by some writers to denote the increasing ubiquity of surveillance \ntechnologies in everyday life which are used to record, survey, monitor and discipline people \n(for example,  Haggerty and Ericson, 2000; Lyon, 2007; Lyon, 2010; Bennett, 2011). It has \nbeen argued by these writers that surveillance is a condition of modernity, essential to the \ndevelopment of the capitalist economy and the contemporary nation state and central to forms \nof disciplinary power and the maintenance of social order. The fastest growing and most \ncontroversial specific type of surveillance is that using the processing of personal data \ngathered from computerised devices \u2018for the purposes of care or control, to influence or \nmanage persons and populations\u2019. These include loyalty cards offered by businesses to their \ncustomers, PINs, information gathered by websites when they are accessed by users and \nticketing systems at airports. The digital data produced by these forms of surveillance serve \nto individuate users, distinguished from others and identified by a series of criteria and then \nbehaviour analysed, to produced \u2018surveillance knowledge\u2019 (Lyon, 2010). \nVarious kinds of social relations and interactions, including power relations, are \ncreated in and through surveillance technologies. These technologies may be considered part \nof the production of the citizen in neoliberal societies. Through the sorting and typing of \nindividuals, allowing the development of profiles and risk categories, policies and strategies \nof inclusion and exclusion operate. Various types of individuals are identified as requiring \ngreater forms of disciplinary control. Not only is personal information gathered via the use of \nsurveillance technologies, but individuals can easily be grouped or sorted into discrete \ncategories and classes based on this information and then subjected to assessments based on \nprior assumptions (Lyon, 2010; Bennett, 2011).  \n11 \n \nThe Foucauldian concept of the panopticon is often employed in writings on \nsurveillance societies. The panopticon is a literal architectural structure, a prison first \nproposed by eighteenth-century reformer Jeremy Bentham. The concept of the panopticon is \nused metaphorically by Foucault in his well-known work Discipline and Punish: the Birth of \nthe Prison (1977) (Lupton, 1997)to suggest the operations of power in contemporary \nsocieties. The panopticon prison was a structure designed so that the monitoring gaze of those \nin power could operate centrally to observe inmates in their separate cells, who were unaware \nof when exactly they were being watched. This design allowed a small number of those in \nauthority to observe a large number of individuals. The concept included the idea not only \nthat prisoners should be observed by those in authority, but also that they should ideally \ndevelop self-surveillance and disciplining strategies in the effort to improve themselves. This \napproach to management of problematic populations was also taken up in relation to other \ninstitutions, such as the hospital and the school. For Foucault, the panopticon was \nrepresentative of a new form of power, one in which central surveillance and monitoring of \nindividuals was combined with those individuals developing voluntary self-management \ntechniques. The panopticon metaphor emphasises the role played by \u2018the gaze\u2019, surveillance \nand visibility in the new forms of power relations emerging in the eighteenth and nineteenth \ncenturies, but has clear resonances for surveillance society today (Haggerty and Ericson, \n2000; Brignall, 2002; Elmer, 2003; Caluya, 2010).  \nThe emergence of m-health potentially reconfigures the subject of surveillance and \ncomplicates the concept of the panoptic gaze. While there still may be an expert exerting the \npanoptic gaze upon the individual, such as in the case of the health promoter making \ndecisions about who should be part of a targeted \u2018at risk group\u2019 and thus encouraged to \nreceive health-promoting messages on their mobile devices or to subscribe to Twitter \nupdates, these technologies also encourage users to turn the gaze upon themselves or to \n12 \n \nactually invite others to do so. Such media platforms as Facebook and Twitter allow people to \nshare personal information with hundreds or more of their friends or followers, including \nregular automated updates on their exercise and dietary habits or body weight via the kinds of \napps described above. Here friends and followers are invited to contribute by the user to \nmonitor their bodily habits: the net of surveillance is thus expanded around the user\u2019s body. \nThe panoptic gaze in this case becomes inverted, so that instead of the few watching the \nmany, the many are watching the few.  \nOne example of this phenomenon is an American woman with diabetes described in a \nhealth magazine article. This woman uses social media to help her manage her condition, as \nwell as home-based technologies to monitor her blood glucose levels. She regularly reports, \non her blog and via Twitter, her daily activities and symptoms: what she ate for breakfast, \nwhat her blood readings are or how much exercise she has engaged in that day. This woman\u2019s \nmotivation for providing these details to her readers and followers is the support they in turn \ngive her in managing her condition. As she is quoted as saying:  \nBecause I have people who follow me in Twitter \u2026 it means I have some kind of \naudience that is caring for me in the background. It\u2019s helpful if I\u2019m having a rough \nday, if things are not going so well with my blood sugar. I find support there, and it \nkeeps me in line too. (cited in Hawn, 2009) \nAs these comments suggest, the use of m-health technologies blur the spatial \nboundaries between public and private surveillance, bringing public surveillance into the \ndomestic sphere. They construct users as personally responsible for their own health care and \nmanagement but also as part of a heterogeneous network of actants, which include the \nvarious technologies employed but also friends and contacts. Perhaps more useful than \ntheories of the panoptic gaze, therefore, is that which employs the Deleuze and Guattari term \n13 \n \n\u2018assemblages\u2019 to describe \u2018surveillant assemblages\u2019, or the complex interaction of \ntechnologies, data and bodies in producing defined subjectivities (Haggerty and Ericson, \n2000). The concept of assemblages recognises the heterogeneity of objects which combine to \nform certain types of bodies/selves as well as their constantly shifting and dynamic nature. It \nalso acknowledges the role played by non-human actants such as technologies in producing \nbodies/selves. In relation to surveillance technologies, assemblages are created when \ninformation about individuals is derived via surveillance technologies and then reassembled \nfor various purposes, creating a certain type of subject, or \u2018data doubles\u2019 or \u2018data selves\u2019 \nwhich can then be scrutinised, monitored and used for various purposes, including \nintervention (Haggerty and Ericson, 2000; Elmer, 2003).  \nWriting about \u2018surveillance assemblages\u2019 Elmer (2003) has contended that: \nThe observed body is of a distinctly hybrid composition. First it is broken down by \nbeing abstracted from its territorial setting. It is then reassembled in different settings \nthrough a series of data flows. The result is a decorporealized body, a \u2018data-double\u2019 of \npure virtuality.  \nYet it may also be argued that the body as it is produced via m-health technologies is far from \nbeing \u2018decorporealised\u2019. While the abstracted \u2018data-double\u2019 produced through biometric \nmeasurements and health surveillance technologies which are able to identify \u2018at-risk\u2019 \nindividuals may be categorised as a virtual cyberbody, this data-double feeds back \ninformation to the user in ways that are intended to encourage the user\u2019s body to act in certain \nways. The flow of information, therefore, is not one-way or static: it is part of a continual \nloop of the production of health-related data and response to these data. This assemblage also \nchallenges previous representations of the cyborg, in which utopian ideas about the use of \ntechnology in transcending the imperatives and constraints of the fleshly body were often \n14 \n \ndominant (Buse, 2010). The cyborg body configured by m-health technologies, in contrast, \nsupport a reflexive, self-monitoring awareness of the body, bringing the body to the fore in \nways which challenge the non-reflexive, absent body (Leder, 1990). The body is hardly able \nto disappear when its functions, movements and habits are constantly monitored and the user \nof m-health technologies is made continually made aware, via feedback, of these dispositions.  \nPrivacy, intimacy and ethical issues \nWhere once public health promotion campaigns used public space and public media to \nconvey their messages, mobile wireless devices allow the incursion of messages into the most \nintimate and personal spaces. Privacy issues have often been raised in discussions about new \ndigital media, including social media platforms and mobile devices. Some writers have \nquestioned whether the current era of personalised computerised technology use, social media \nand wide-spread surveillance has meant \u2018the end of privacy\u2019 (Lyon, 2010)? Have concepts of \nprivacy narrowed to liberal assumptions about subjectivity, are they too culturally relative or \noverly reliant on rights-based discourses, neglectful of new ways of living and being? Can the \nspatial meanings of privacy, which represent privacy as a kind of personal zone from which \nothers are excluded unless given permission to enter, remain meaningful in a context in \nwhich wired consumers are available for surveillance and data gathering for much of their \nwaking day (Bennett, 2011)? Does the concept of privacy now have any meaning in such a \ncontext? As Ericson and Haggerty (Haggerty and Ericson, 2000) note, in an era in which \npersonal information may be effectively sold, privacy is now something that may be traded \nfor services or commodities, and perhaps has lost some of its value. \nWhat has been termed \u2018the politics of gazing\u2019 (Ibrahim, 2010) is relevant to the \ndiscussion here. Ibrahim notes that the personal space has become \u2018a coveted commodity \nwhere new technologies, innovative designs and convergence occur and coalesce\u2019 (2010). \nThe politics of gazing presents new challenges and ethical and moral dilemmas. These \n15 \n \ndilemmas are located within the space of the body because of the mobility of the new mobile \ndevices. Because these devices are always around one\u2019s person and in one\u2019s personal sphere, \neffectively as prostheses of the cyborg body, it can be difficult to \u2018switch off\u2019 (Agger, 2011; \nMatusik and Mickel, 2011). Many users of social networking platforms are grappling with \ncoming to terms with new ways of defining privacy in a context in which concepts of \u2018the \npublic\u2019 and \u2018the private\u2019 are no longer confined to a spatial dimension. Notions of intimacy, \nsolitude, the personal, the secret and the hidden are challenged by the confessional of social \nmedia sites such as Facebook and Twitter, in which participants\u2019 inner thoughts and private \nbehaviours are often revealed to a large number of friends or followers, and frequently \nseveral times throughout the day. This phenomenon has been referred to as \u2018the privatization \nof the public and publicization of the public\u2019 (van Mamen, 2010).  \nThe sense of intimacy and social support that many users derive from using social \nmedia may readily translate to uploading information about their bodies using the kinds of m-\nhealth apps described above. Such information may be regarded as contributing to the \npersona that is constructed via social media sites: sharing attempts to reduce smoking or \ndrinking, or to engage regularly in exercise, for example, and receiving supportive messages \nin response, as well as commiseration for those times when the user fails to achieve her or his \ngoals of self-improvement and discipline.  However, users may still feel uncomfortable about \nwhat they perceive as exposure and invasion of personal space. They may also feel \u2018invaded\u2019 \nby the sheer overload of data that may be generated by membership of social networking sites \nand the difficulty of switching off mobile devices and taking time out from using them \n(Boyd, 2008).  \nThere are other moral and ethical issues associated with the use of such monitoring \ndevices. Accounts of using m-health technologies in the medical literature dealing with \npatient follow-up and care tends to focus on shifting responsibility for care from the clinician \n16 \n \nto the patient, placing new expectations upon the patient to manage their health in ways that \nwere traditionally viewed as the preserve of health care professionals. The rhetoric of such \naccounts uses such terms as \u2018patient empowerment\u2019 as well as cost-efficiency as positive \noutcomes of this shift of responsibility. The patient is represented as ideally willing to take on \nsuch responsibility, as active and agential in engaging in participating in the monitoring of \nher or his own body (Mort et al., 2009; Andreassen, 2011). What is often glossed over or \nignored in this discourse of patient responsibility for self-surveillance are the inherent \ninequalities that are reproduced in the use of medical information and monitoring \ntechnologies, including issues of access to such technologies (Nettleton and Burrows, 2003). \nNeither are the emotional dimensions of using such technologies acknowledged: some \npatients\u2019 emotional need to devolve responsibility to their health professionals, for example. \nNot all patients wish to become reflexive, agential and \u2018empowered consumers\u2019 of health care \n(Lupton, 1997; 2003; Andreassen and Trondsen, 2010).  \nSo too, the m-health discourse in health promotion represents people as ideally \nwilling to take on responsibility for promoting their health using these latest technologies, to \nthe point that they are happy to receive regular messages on their smartphone or to have their \nhealth habits and behaviours continuously monitored and assessed. This is a body/self \nconfigured as requiring, and in fact desiring, of constant digital surveillance. \nHealth promotion models of behaviour tend to be dominated by social psychology \ntheories which focus on individuals\u2019 behaviour to the exclusion of the socio-political context \nin which individuals live. Such use of social media tools build upon the \u2018magic bullet\u2019 \napproach to conveying health-related messages to members of the general public which has \nbeen a mainstay of health promotion models of behaviour for several decades and the \ncontinuing subject of critique from sociologists for equally long (see, for example, Lupton, \n1994; Bunton et al., 1995; Lupton, 1995b; Petersen and Lupton, 1996; Bunton, 2006; \n17 \n \nPetersen, 2010). As these critics have argued, the individualistic, targeted approach that \nappears such an enticing aspect of social media is also its disturbing property. By focusing on \nthe individual, sending regular messages to encourage that person to exercise or eat well, \nthese technologies reduce health problems to the micro, individual level. The discourses on \nm-health, like those in health promotion generally, tend to suggest that if an individual fails \nto adopt the suggestions of the texts or emails they received, that person is fully responsible \nfor the health problems they may experience. Such approaches do little, therefore, to identify \nthe broader social, cultural and political dimensions of ill-health and the reasons why people \nmay find it difficult to respond to such messages. They also do not address the possibilities \nthat the continual use of the devices may create unintended consequences such as stress, \nunwanted distraction from other activities or the infringement upon intimate relationships. \nThe new m-health technologies may be viewed as potentially contributing to feelings \nof shame and guilt that their target users may feel if they do not adopt health-related \nsuggestions, an additional stressor in an individual\u2019s day when they may well be juggling \nmultiple demands and responsibilities. There is certainly a patronising, \u2018we know better\u2019 \ndiscursive representation of the relationship between the health promoter and the mobile \ndevice user in the public health literature, in which the health promoter attempts to find new \nand better ways of encouraging members of the public to change their ways to achieve a \n\u2018healthy lifestyle\u2019 as it is defined in official public health arenas.  \nIt would be misleading, however, to represent the use of m-health technologies as \nsimply oppressive, coercive or in other ways limiting of individuals\u2019 agency and freedom. As \nthe theoretical perspectives drawn from both Foucault\u2019s writings and those of Deleuze and \nGuattari would argue, the power relations implicit in surveillance technologies are not \nnecessarily coercive or repressive. Power is diffuse, spread over many networks, operating \nnot only from state agencies but also manifold non-government organisations such as those in \n18 \n \ncommodity culture and the mass media. Power may also be viewed as productive, bringing \ncertain kinds of subjectivities and embodiments into being. Individuals are not coerced into \nproviding information or downloading health-related apps which remind them to exercise, eat \nwell or take their medications. They do so voluntarily and willingly in their efforts to improve \ntheir health or physical fitness, reduce their consumption of alcohol, give up smoking or lose \nweight. As part of presenting the self and disciplining and shaping one\u2019s body in certain \nways, citizens adopt public health injunctions or warnings in their own best interests, to \nproduce their \u2018best selves\u2019. In the term used by Deleuze and Guattari, \u2018desire\u2019 impels their \nuse of m-health platforms and devices. Nonetheless it is also important to point out that there \nare compulsions associated with choice (Petersen, 2010). Individuals make choices not in a \nsocial vacuum, but in a context in which certain kinds of subjects and bodies are privileged \nover others and there are obligations and commitments involved: the responsible, self-\ndisciplined body/self, for example, who is interested in and motivated to improve their health.  \nConclusion \nTo conclude, the new forms of computerised technologies offered by Web 2. platforms and \nmobile wireless devices offer new forms of capacities, embodiment and subjectivities. They \nconfigure the assemblage of an idealised consumer who is amenable to monitoring, \nsurveillance and disciplining of their body by way of individualised automated messages and \nthe feedback and sharing of biometric data. They also configure the professional figure of the \nhealth promoter in a different way. The health promoter becomes an individual who is \nconducting surveillance in an ever-more refined and diffuse manner on members of the target \npopulation, using technologies in unprecedented way. Via m-health technologies, the health \npromoter is able to insert her- or himself even more insistently into the private world of \nothers, accessing them in any location in which their mobile device accompanies them.  \n19 \n \nA space is opened up here for researchers to identify and explore the experiences of \nindividuals as they take up (or indeed, resist) the potentialities of mobile devices and the new \nsocial media as they are adopted into the \u2018toolbox\u2019 of health promotion. Questions that have \nyet to be answered include: what are the implications for subjectivities and embodiment in \nthe world of m-health \u2013 how are the assemblages of technologies/practices/flesh enacted and \nlived? What are the political dimensions and power relations inherent in the use of m-health \ntechnologies? How will privacy (or loss of privacy) be defined and experienced in the context \nof these media? What are the implications for how people conduct their everyday lives and \nintimate relationships?  \nThere is much talk in health promotion circles about \u2018health literacy\u2019, or knowledge \nand understanding about health and preventing ill-health that certain social groups develop. \nPerhaps it also needs to be acknowledged that \u2018digital literacy\u2019 should become a part of \nhealth literacy, and that indeed, such digital literacy might include a response on the part of \ntargeted audiences to forms of health promotion messages conveyed via mobile devices and \nsocial media platforms that is critical and contesting of them. An integral aspect of Web 2.0 \ntechnologies is the space they provide for audiences and consumers to engage with each \nother, to resist attempts to position them in certain ways, to challenge power relations: in \nshort, to \u2018talk back\u2019 to those who may be attempting to change their behaviours, both \nindividually and collectively. Will the \u2018nagging voices\u2019 of the health promoting messages \nautomatically issuing forth from a person\u2019s mobile device be eventually ignored by its user? \nOr will these messages incite even greater feelings of guilt and shame at one\u2019s lack of self-\ncontrol and self-discipline? Alternatively, will m-health technologies produce a cyborg, post-\nhuman self in which the routine collection of data about bodily actions and functions is \nsimply incorporated unproblematically into the user\u2019s sense of selfhood and embodiment? \nHow will concepts of \u2018health\u2019 itself be shaped and understood in a context in which one\u2019s \n20 \n \nbiometric indicators may be constantly measured, analysed and displayed publicly on \nFacebook or Twitter? Will the \u2018objective\u2019 measurements offered by mobile devices take \nprecedence over the \u2018subjective\u2019 assessments offered by the senses of the fleshly body? \nAddressing these questions, and many more, offers a rich seam of inquiry for social \nresearchers and theorists interested in exploring the implications of the emergence of m-\nhealth. \n  \n21 \n \n \nReferences \nAgger, B. (2011) iTime: labour and life in a smartphone era. Time & Society 20(1): 119--136. \nAndreassen, H. (2011) What does an e-mail address add? - Doing health and technology at home. \nSocial Science & Medicine 72: 521--528. \nAndreassen, H., and Trondsen M. (2010) The empowered patient and the sociologist. Social Theory & \nHealth 8(3): 280-287. \nArmstrong, D. (1995) The rise of surveillance medicine. Sociology of Health & Illness 17(3): 393--\n404. \nBennett, C.J. (2011) In defence of privacy: The concept and the regime. Surveillance & Society 8(4): \n485--496. \nBoyd, D. (2008) Facebook's privacy trainwreck: exposure, invasion, and social convergence. \nConvergence 14(1): 13--20. \nBrignall, T. (2002) The new panopticon: the internet viewed as a structure of social control. Theory \nand Science 3(1). \nBunton, R. (2006) Critical health psychology: Julie Hepworth. Journal of Health Psychology 11(3): \n343--345. \nBunton, R., Nettleton S., and Burrows R. (eds.) (1995) The Sociology of Health Promotion: Critical \nAnalyses of Consumption, Lifestyle and Risk: Routledge. \nBuse, C. (2010) E-scaping the ageing body? Computer technologies and embodiment in later life. \nAgeing and Society 30(6): 987-1009. \nCaluya, G. (2010) The post-panoptic society? Reassessing Foucault in surveillance studies. Social \nIdentities 16(5): 621--633. \nCasper, M., and Morrison D. (2010) Medical sociology and technology: critical engagements. Journal \nof Health and Social Behavior 51(1): S12---32. \nCatford, J. (2011) The new social learning: connect better for better health. Health Promotion \nInternational 26(2): 133--135. \nDe Mul, J. (1999) The informization of the worldview. Information, Communication and Society \n2(1): 69--94. \nElmer, G. (2003) A diagram of panoptic surveillance. New Media & Society 5(2): 231--247. \nFoucault, M. (1975)  The Birth of the Clinic: An Archaeology of Medical Perception. New York: \nVintage Books. \nFoucault, M. (1977)  Discipline and Punish: the Birth of the Prison. London: Penguin. \nFreund, P. (2004) Civilised bodies redux: seams in the cyborg. Social Theory & Health 2(3): 273-289. \nHadders, H. (2009) Enacting death in the intensive care unit: medical technology and the multiple \nontologies of death. health 13(6): 571--587. \nHaggerty, K., and Ericson R. (2000) The surveillant assemblage. British Journal of Sociology 51(4): \n605--622. \nHaraway, D. (1988) A manifesto for cyborgs: science, technology and socialist feminism in the \n1980s. In: E. Weed (eds.) Coming to Terms: Feminism, Theory and Practice. New York: \nRoutledge, p. 173--204. \nHaraway, D. (1989) The biopolitics of postmodern bodies: determinations of self in immune system \ndiscourse. Differences 1(1): 3--44. \nHawn, C. (2009) Take two asprin and tweet me in the morning: how Twitter, Facebook, and other \nsocial media are shaping health care. Health Affairs 28(2): 361--368. \nIbrahim, Y. (2010) The wired body and event construction: mobile technologies and the technological \ngaze. In: P. Kalantzis-Cope and K. Gherab-Martin (eds.) Emerging Digital Spaces in \nContemporary Society. Houndsmills, Basingstoke: Palgrave Macmillan, p. 123--125. \nKapitan, L. (2009) Introduction to the special issue on art therapy's response to techno-digital culture. \nArt Therapy 26(2): 50--51. \nLaakso, E.-L., Armstrong K., and User W. (2011) Cyber-management of people with chronic disease: \na potential solution to eHealth challenges. Health Education Journal 5(3): 1--8. \n22 \n \nLeder, D. (1990)  The Absent Body. Chicago, IL: University of Chicago Press. \nLupton, D. (1994) Consumerism, commodity culture and health promotion. Health Promotion \nInternational 9(2): 111-118. \nLupton, D. (1995a) The embodied computer/user. Body & Society 1(3-4): 97-112. \nLupton, D. (1995b)  The Imperative of Health: Public Health and the Regulated Body. London: Sage. \nLupton, D. (1997) Foucault and the medicalisation critique. In: A. Petersen and R. Bunton (eds.) \nFoucault, Health and Medicine. London: Routledge, p. 94-110. \nLupton, D. (2003)  Medicine as Culture: Illness, Disease and the Body. London: Sage. \nLyon, D. (2007)  Surveillance Studies: An Overview. Cambridge: Polity Press. \nLyon, D. (2010) Surveillance, power and everyday life. In: P. Kalantzis-Cope and K. Gherab-Martin \n(eds.) Emerging Digital Spaces in Contemporary Society. Houndsmills, Basingstoke: \nPalgrave Macmillan, p. 107--120. \nMansell, R. (2010) Technology, innovation, power, and social consequence. In: P. Kalantzis-Cope \nand K. Gherab-Martin (eds.) Emerging Digital Spaces in Contemporary Society. \nHoundsmills, Basingstoke: Palgrave Macmillan, p. 13--25. \nMartin, E. (2000) Flexible bodies: science and a new culture of health in the US. In: S. Williams, \nGabe, J. and Calnan, M. (eds.) Health, Medicine and Society: Key Theories, Future Agendas. \nLondon: Routledge, p. 123--145. \nMatusik, S., and Mickel A. (2011) Embracing or embattled by converged mobile devices? Users' \nexperiences with a contemporary connectivity technology. Human Relations 64(8): 1001--\n1030. \nMays, D., Cremeens J., Usdan S., Martin R., Arriola K., and Bernhardt J. (2010) The feasibility of \nassessing alcohol use among college students using wireless mobile devices: implications for \nhealth education and behavioural research. Health Education Journal 69(3): 311--320. \nMort, M., Finch T., and May C. (2009) Making and unmaking telepatients: identity and governance in \nnew health technologies. Science, Technology & Human Values 34(1): 9--33. \nMort, M., and Smith A. (2009) Beyond information: intimate relations in sociotechnical practice. \nSociology 43(2): 215--231. \nNettleton, S. (2004) The emergence of e-scaped medicine? Sociology 38(4): 661-679. \nNettleton, S., and Burrows R. (2003) E-scaped medicine? Information, reflexivity and health. Critical \nSocial Policy 23(2): 165-185. \nPetersen, A., Davis, M., Fraser, S. and Lindsay, J. (2010) Healthy living and citizenship: an overview. \nCritical Public Health 20(4): 391--400. \nPetersen, A., and Lupton D. (1996)  The New Public Health: Health and Self in the Age of Risk. \nLondon: Sage. \nShildrick, M. (2010) Some reflections on the socio-cultural and bioscientific limits of bodily integrity. \nBody & Society 16(3): 11--22. \nTomas, D. (1995) Feedback and cybernetics: reimaging the body in the age of the cyborg. Body & \nSociety 1(3-4): 21-43. \nvan Mamen, M. (2010) The pedagogy of Momus technologies: Facebook, privacy and online \nintimacy. Qualitative Health Research 20(8): 1023--1032. \n \n \nWord count: 6,787 \nDate of manuscript: 16 November 2011 \n", "identifiers": ["oai:ses.library.usyd.edu.au:2123/8202", null], "language": {"code": "en", "id": 9, "name": "English"}, "relations": [], "repositories": [{"id": "1002", "openDoarId": 0, "name": "Sydney eScholarship", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "baseId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 1, "metadataAdded": 1466076332000, "metadataUpdated": 1591339664000, "depositedDate": 1333580400000}, "subjects": ["Article", "Pre-print"], "title": "M-health and health promotion: the digital cyborg and surveillance society", "topics": ["m-health", "e-health", "health promotion", "the body", "cyborg", "sociology", "surveillance society", "digital technologies", "social theory"], "types": [], "year": 2012, "fulltextIdentifier": "https://core.ac.uk/download/pdf/41236113.pdf", "doi": "10.1057/sth.2012.6", "oai": "oai:ses.library.usyd.edu.au:2123/8202", "downloadUrl": "https://core.ac.uk/download/pdf/41236113.pdf"}}
{"status": "OK", "data": {"id": "146503003", "authors": ["Dumay, N"], "citations": [], "contributors": [], "datePublished": "2016-01", "description": "This is the author accepted manuscript. The final version is available from Elsevier via the DOI in this record.There is another ORE record for this publication: http://hdl.handle.net/10871/17864Two published datasets (Dumay & Gaskell, 2007, Psychological Science; Tamminen, Payne,\r\nStickgold, Wamsley, & Gaskell, 2010, Journal of Neuroscience) showing a positive influence of\r\nsleep on declarative memory were re-analyzed, focusing on the \"fate\" of each item at the 0-hr\r\ntest and 12-hr retest. In particular, I looked at which items were retrieved at test, and\r\n\"maintained\" (i.e., not forgotten) at retest, and which items were not retrieved at test, but\r\neventually \"gained\" at retest. This gave me separate estimates of protection against loss and\r\nmemory enhancement, which the classic approach relying on net recall/recognition levels has\r\nremained blind to. In both free recall and recognition, the likelihood of maintaining an item\r\nbetween test and retest, like that of gaining one at retest, was higher when the retention interval\r\nwas filled with nocturnal sleep, as opposed to day-time (active) wakefulness. And, in both cases,\r\nthe effect of sleep was stronger on gained than maintained items. Thus, if sleep indeed protects\r\nagainst retroactive, unspecific interference, it also clearly promotes access to those memories\r\ninitially too weak to be retrieved. These findings call for an integrated approach including both\r\npassive (cell-level) and active (systems-level) consolidation, possibly unfolding in an\r\nopportunistic fashion", "fullText": "In press at Cortex \n1 \n \n  \n \n \nSleep not just protects memories against forgetting, \nit also makes them more accessible \nNicolas Dumay \nDepartment of Psychology, University of Exeter, United Kingdom, and BCBL (Basque Center \non Cognition, Brain and Language), Spain \n \nWord count: \nAbstract \uf02d 199 \nMain text, plus footnotes \uf02d 5,121  \nCorrespondence: \nNicolas Dumay \nDepartment of Psychology \nUniversity of Exeter \nPerry Road \nExeter \nEX4 4QG \nUnited Kingdom \nE-mail: nicolas.dumay@gmail.com \nPhone: +44 1392 724666 \n \nIn press at Cortex \n2 \n \nAbstract \nTwo published datasets (Dumay & Gaskell, 2007, Psychological Science; Tamminen, Payne, \nStickgold, Wamsley, & Gaskell, 2010, Journal of Neuroscience) showing a positive influence of \nsleep on declarative memory were re-analyzed, focusing on the \"fate\" of each item at the 0-hr \ntest and 12-hr retest. In particular, I looked at which items were retrieved at test, and \n\"maintained\" (i.e., not forgotten) at retest, and which items were not retrieved at test, but \neventually \"gained\" at retest. This gave me separate estimates of protection against loss and \nmemory enhancement, which the classic approach relying on net recall/recognition levels has \nremained blind to. In both free recall and recognition, the likelihood of maintaining an item \nbetween test and retest, like that of gaining one at retest, was higher when the retention interval \nwas filled with nocturnal sleep, as opposed to day-time (active) wakefulness. And, in both cases, \nthe effect of sleep was stronger on gained than maintained items. Thus, if sleep indeed protects \nagainst retroactive, unspecific interference, it also clearly promotes access to those memories \ninitially too weak to be retrieved. These findings call for an integrated approach including both \npassive (cell-level) and active (systems-level) consolidation, possibly unfolding in an \nopportunistic fashion. \n   \n \nKeywords: memory consolidation, sleep, forgetting, reminiscence, item fate \n \n \n \n \nIn press at Cortex \n3 \n \n\"The analysis of recall patterns has clearly shown that the \"average\" item is a highly abstract and \nelusive entity having no readily identifiable counterparts in the empirical realm.\" (Tulving, 1967, \np. 183). \n1. Introduction \nIn a study titled \"Obliviscence during sleep and waking\" Jenkins and Dallenbach (1924) had \ntwo of their colleagues learn lists of nonsense syllables either in the morning or late evening. By \ntesting recall after intervals ranging between one and eight hours, these researchers found that \nthe presence of sleep in the retention interval had a protective influence: forgetting curves were \nless steep for intervals filled with sleep than for those filled with active wake. \nNine decades down the forgetting curves, the beneficial impact of sleep on memory is well \nestablished (see Wixted & Cai, 2014, for a review). We know that the sooner the learner sleeps \nafter encoding, the better the memory retention (Ekstrand, 1972; Gais, Lucas, & Born, 2006; \nPayne et al., 2012). We also know that which sleep component is key to the consolidation \nprocess depends on the type of knowledge to be consolidated. For instance, the conscious \nrecollection of factual information or of our past experience (i.e., \"declarative\" memory) almost \nexclusively benefits from slow-wave (slow oscillations) sleep (Plihal & Born, 1997, 1999; \nYaroush, Sullivan, & Ekstrand, 1971). By contrast, the positive effect of sleep on newly learnt \nperceptual or motor skills, i.e., procedural memory, shows a more complex picture, in which \nREM (rapid-eye movement) and non-REM sleep brain correlates have been implicated, \nsometimes in synergy (e.g., Gais, Plihal, Wagner, & Born, 2000; Gaskell et al., 2014; Karni, \nTanne, Rubenstein, Askenasy, & Sagi, 1994; Rasch, Gais, & Born, 2009; Stickgold, James, & \nHobson, 2000). \nIn press at Cortex \n4 \n \nOne of the current debates, however, is whether sleep makes declarative (hippocampus-\ndependent) memories more vivid, and thus more accessible, than they were just after encoding, \nor if instead sleep merely protects them from the deleterious effect of retroactive interference. \nAccording to the \"active consolidation\" account (Diekelmann & Born, 2010), sleep is key to \nmemory consolidation because slow-wave sleep promotes \"neural replay\" (Wilson & \nMcNaughton, 1994). In slow-wave sleep, new memories encoded in the hippocampus are \nrepeatedly re-activated, which drives concurrent reactivation of cortical regions implicated in \ntheir initial capture (Ji & Wilson, 2007). By this feedback from the hippocampus to the cortex, or \n\"systems consolidation\", newly acquired memories are effectively re-experienced, with the result \nthat cortical representations are strengthened. They are also better integrated in pre-existing \ncortical networks, because replay also reactivates similar, long-consolidated material. Hence, \naccording to this account, which emphasizes the role of the hippocampus as the sparring partner \nof the neocortex, sleep has the potential to make memories more accessible. \nIn contrast, theories grounded in the notion of retroactive interference (Wixted, 2004; see also \nMednick, Cai, Shuman, Anagnostaras, & Wixted, 2011) insist that during slow-wave sleep \nsynaptic plasticity in the hippocampus is null. Consequently, hippocampal resources that would \notherwise be allocated to new encoding can now be used to consolidate, at the cell level, \nmemories formed prior to (slow-wave) sleep. Amongst these, those formed earlier during the \nwake should be more eroded. Thus, memory consolidation is fundamentally the antidote to \nretroactive, unspecific interference and thus reduces its product: forgetting. Any factor that \nreduces the encoding activities of the hippocampus ipso facto promotes consolidation. As this \naccount assumes that hippocampal consolidation triggers systems consolidation, anything that is \nforgotten by the hippocampus cannot be recovered via neural replay (Wixted & Cai, 2014, p. \nIn press at Cortex \n5 \n \n30). Thus, according to the anti-forgetting view, sleep\u2014like any other interval of reduced \nencoding\u2014can at best stabilize newly formed declarative memories; it cannot make them more \naccessible. \nSo far, the \"active consolidation\" account has received strong empirical support from \ndemonstrations that declarative memories are better preserved if, while in slow-wave sleep, \nparticipants are cued by an odor or sound also present during encoding (Rash, Buchel, Gais, & \nBorn, 2007), or by translation equivalents, in the case of word lists (Schreiner & Rasch, 2014). \nFrustratingly, however, the data supposedly speaking to the issue of sleep-dependent trace \nenhancement remain ambiguous, providing little support for the idea. \nI suggest that this state of affairs is due to the fact that researchers have been relying \nexclusively on net performance. However, as Tulving (1964, 1967), amongst others, pointed out, \nthis approach is inherently blind to fluctuations at the item-level. Consequently, even though the \ntypically observed pattern is that the sleep group simply shows less forgetting than the wake \ngroup, it may well be that a sleep-dependent trace enhancing mechanism is actually \ncounteracting the effect of a task-specific component that worsens performance at retest. \nConversely, and by the same logic, finding that the sleep group shows more improvement at \nretest (compared to the initial test) than the wake group does not necessarily provide evidence for \nsleep-dependent trace enhancement. A sleep-dependent anti-forgetting mechanism could just \nsupplement a task-specific component which, irrespective of group, helps to maintain/improve \nperformance. Therefore, without information on the trajectory of each item between test and \nretest, it is misguided to use declarative tasks and make inferences about consolidation \nmechanisms supposedly acting on individual representations. \nIn press at Cortex \n6 \n \nConsequently, in the present research I tracked the \"fate\" of individual items from an initial \ntest to a post-sleep retest to better assess the impact of sleep on memory. Specifically, I \ndistinguished between items that were retrieved at test and \"maintained\" (i.e., not forgotten) at \nretest, and items that were not retrieved at test, but were eventually recovered (i.e., \"gained\") at \nretest\uf0bea phenomenon typically referred to as \"reminiscence\". This gave me separate estimates \nof protection against loss and of memory enhancement. These provide a means to determine \nwhether sleep only boosts protection against loss, which would support the anti-forgetting \naccount (Wixted, 2004), or whether it also boosts memory accessibility, which would support the \nactive consolidation account (Diekelmann & Born, 2010). \nItem-fate analysis is routinely performed in cognitive psychology (Ballard, 1913; Brown, \n1923; Erdelyi, 1984; Tulving, 1964, 1967, just to cite the pioneers). However, it is only recently \nthat Fenn and Hambrick (2013) used this approach to look at the effect of sleep, unfortunately \nrather unconvincingly, as I will demonstrate. Fenn and Hambrick had 354 participants learn pairs \nof semantically related items (e.g., table-chair) either in the evening or in the morning. After \ntesting recall of the second member of each pair (e.g., table-?) immediately and after twelve \nhours, and finding improvement for both groups, they classified responses following the above \nmaintained vs. gained distinction. As sleep had its strongest impact on maintained items (+1.35 \nitem compared to wake; versus +.44 for gained), the authors concluded that \"[\u2026] loss prevention \nmay primarily account for the effect of sleep on declarative memory consolidation\" (p. 501). \nAs can be seen on their Fig. 1, however, performance was dangerously close to ceiling, \nespecially at the 12-hr retest (on average, 35 items recalled out of 40). To reassure the reader, \nFenn and Hambrick crosschecked their results, excluding all participants with a retest score four \nitems (i.e., the average improvement across the two groups), or less, away from the ceiling (see \nIn press at Cortex \n7 \n \ntheir Footnote 4 and Supplemental Materials). This is dubious: the average improvement is itself \ndetermined by ceiling height, and therefore, it cannot be used as a safeguard. \nTo assess whether this procedure was at all effective, I computed the correlation between the \n0-hr score and the magnitude of the 12-hr improvement for the 322 remaining participants. As \nshown Fig. 1, there was still a negative correlation: the better a participant was at test, the less \n(s)he improved at retest. In other words, the restricted dataset was still constrained by the ceiling. \nClearly, with so little room for improvement, this experiment hardly gave sleep a chance to \ndemonstrate its potential for enhancing memory\u2014the same criticism applies to a follow-up just \npublished (Fenn & Hambrick, 2015), which used the same parameters, task, crosscheck, and \nanother 435 participants. \nGiven this concern, I re-analyzed the data from two published studies (Dumay & Gaskell, \n2007; Tamminen, Payne, Stickgold, Wamsley, & Gaskell, 2010; henceforth 'DG' and 'TPSWG'), \nthat showed a positive effect of sleep on free recall under conditions in which performance was \nnot constrained by the ceiling (or the floor). These studies used similar methodologies, which \nmakes them easy to combine. Participants were trained on made-up words either in the evening \nor in the morning in a phoneme detection task, and urged to learn. Memory tests were then \nadministered immediately and again after twelve hours, meaning, after nocturnal sleep or a day \nawake, respectively. Crucially, because of the nature of the to-be-learnt stimuli, all competitors \nof existing words (e.g., 'caravoth' for 'caravan'), both studies could demonstrate overnight \nintegration of the new items in lexico-semantic memory (cf. Gaskell & Dumay, 2003; Leach & \nSamuel, 2007). In both studies, emergence of the competitive influence from the novel words on \nperception of their base words was contingent upon the occurrence of sleep during the test-retest \ninterval (see Dumay & Gaskell, 2005). Thus, on the basis of these non-linear changes, we know \nIn press at Cortex \n8 \n \nthat for the sleep group something akin to systems consolidation had been triggered (e.g., \nTakashima et al., 2009). \nWith this appropriate data set, item fate can show whether the net improvement in declarative \nknowledge observed after sleep was driven by protection against loss, by memory enhancement, \nor by both. If protection against forgetting is the driving force, the sleep group should have a \nhigher proportion than the wake group only on maintained items; in contrast, if active trace \nenhancement is operating, then the sleep group should instead differ from the wake group only \non gained items. \n2. Original Methods \nIn DG, training involved 24 made-up words (such as 'frenzilk') created by adding two \nconsonants to vowel-final nouns, each played 36 times (see Dumay & Gaskell, 2012, for the list \nof materials). In TPSWG, the novel words were 30 onset-overlapping competitors (such as \n'caravoth'); each was played 30 times in phoneme detection, plus four times in oral repetition. \nTraining took place between 8:00 and 8:45 a.m./p.m. in DG, and between 9:15 and 10 p.m., and \n8:30 and 10 a.m. in TPSWG. Retest followed exactly 12.5 hr after the immediate test in DG, \nwhereas this was after 9.5 hr for the wake group and 10 hr for the sleep group in TPSWG.\ni\n In \nboth studies, free recall always came first in the series of declarative tasks. However, as the focus \nwas on whether the newly learnt, novel word representation influenced perception of the long-\nconsolidated base word, priority was given to the so-called \"lexical competition\" test, in which \nonly the overlapping base words (e.g., 'caravan'), not the made-up words themselves, were \npresented. In the free recall task, participants had three minutes to report as many novel words as \nthey could remember from the training phase. Responses were digitally recorded and checked for \naccuracy afterwards. This was followed by a two-alternative recognition test only for half the \nIn press at Cortex \n9 \n \nparticipants in DG, and a cued recall and an old/new recognition test in TPSWG. See the original \narticles for further details. \n3. Analysis One\u2014Free Recall \nThe first analysis combined the free recall data of the 42 \"best\" participants out of the 64 \ntested by DG and of the 30 \"best\" participants out of the 59 reported in TPSWG (see below). The \nresulting sleep and wake groups thus comprised 36 participants each, of whom 21 came from \nDG. The use of a restricted sample was motivated by the presence of a floor effect in the wake \ngroup in the total sample, as shown by a significant negative correlation between test scores and \namount of forgetting at retest (r(61) = -.30, p < .02). To eliminate this problem, participants with \nthe poorest test scores were removed until this correlation became non-significant, and close to \nzero (r(36) = -.08, p > .65). The sleep group showed no sign of a ceiling/floor effect in the total \nsample (r(62) = -.09, p > .50), or in the restricted set (r(36) = -.06, p >.73). Using this procedure \nassured that in both groups performance was free to vary either way between the 0-hr test and the \n12-hr retest. Note that the full dataset showed overall the same pattern of results as that reported \nfor the restricted sample. \nI first re-analyzed the data using the classical approach based on percent correct. Included in \nthe analysis of variance (ANOVA) was a dummy factor coding which study the participant came \nfrom. As this factor had no interactive effect, I will not consider it any further. \nAs shown Fig. 2a, the sleep and wake groups did not differ significantly on their performance \nat the 0-hr test (21.4 vs. 20.7%; F < 1), but diverged substantially by the 12-hr retest (Group x \nSession: F(1,68) = 24.67, p < .0001). Although the sleep group simply reproduced the \nimprovement (+5.7%) already seen in the individual datasets (F(1,34) = 13.21, p < .001), the \nwake group, now unbound, statistically confirmed the trend towards forgetting only suggested in \nIn press at Cortex \n10 \n \nthe original studies (-4.8%; F(1,34) = 11.47, p < .002). Similar ANOVAs on arcsined \nproportions showed exactly the same results. \n3.1 \"Item fate\" analysis \nTo examine the effect of sleep on protection against loss and on memory enhancement \nseparately, I first classified each item on its recall trajectory between test and retest for each \nparticipant. Each item was entered in one of the following categories: (1) \"never recalled\"; (2) \n\"lost\" (i.e., recalled only at test); (3) \"maintained\" (i.e., recalled at both test and retest); or (4) \n\"gained\" (i.e., recalled only at retest). As \"never recalled\" items were not useful to my purpose \nand \"lost\" items were just the complement to maintained items, the first two categories were \ndiscarded. \nThe next step was to derive measures that took into account how many items can theoretically \nbe maintained, and how many items can theoretically be gained, between test and retest. Note \nthat by just relying on raw item counts, Fenn and Hambrick (2013, 2015) failed to take these \ndimensions into account.\nii,iii\n Consequently, for each participant I computed proportions, by \ntaking, respectively, \n(1) for the maintained items: the number of items maintained at retest relative to the number \nof items recalled at test (i.e., the maximum number of items s/he could have maintained); \n(2) for the gained items: the number of items gained at retest relative to the total number items \nminus the number of items recalled at test (i.e., the maximum number of items s/he could have \ngained). This is exactly the procedure introduced by Roediger and Challis (1989, p. 179). \nGaining an item (i.e., \"reminiscence\") was much less likely than maintaining one, as shown \nby the proportions average across conditions (gained: 12.1% vs. maintained: 57.0%). Crucially, \nhowever, gained items were the ones that benefitted most from the participants sleeping during \nIn press at Cortex \n11 \n \nthe test-retest interval, as indicated by their sleep-to-wake ratio of 2.30 (from 7.3% for wake, to \n16.9% for sleep) compared to 1.49 for maintained items (from 45.6% to 68.3%, respectively). \nTo back this up and test for the interaction, I first rescaled the proportions of maintained and \ngained to take into account the difference in probability in the two types of recall trajectory. A \nnormalization-by-the-mean was applied, in which the proportion of maintained and the \nproportion of gained were divided by their respective means (by participants), computed over the \nwhole dataset in order to be neutral with respect to the sleep/wake contrast (see Fig. 2b). I then \nran an ANOVA similar to the one performed on percent correct. The effect of sleep was \nsignificant for both gained items (from .61 for wake, to 1.39 for sleep; F(1,68) = 15.79, p < \n.0002) and maintained items (from .80 to 1.20, respectively; F(1,68) = 11.50, p < .002), but the \ninteraction was significant as well (F(1,68) = 4.07, p < .05), which confirmed that sleep had a \nstronger influence on gained items.\niv\n \nIn view of these results, clearly, sleep protects against forgetting, as shown by the higher \nproportion of maintained items after sleep, compared to wake. But its most compelling impact is \non promoting access to memories initially too weak to be recalled, as shown by its effect on \ngained items. To provide a conceptual replication of these findings, in a different declarative \ntask, I applied the same analysis to the old/new speeded recognition data from the 60 participants \ntested by TPSWG. Note that in the original paper only the latency data are reported. \n4. Analysis Two\u2014Old/New Recognition \nIn TPSWG the recognition task included only half of the 30 target made-up words at the 0-hr \ntest (mixed with minimally diverging foils, e.g., \"caravol\"), but all of them at the 12-hr retest. On \neach trial participants had to decide, as quickly as possible, whether they recognized the stimulus \nIn press at Cortex \n12 \n \nas part of the to-be-learnt set. The present analysis is based on these 15 items (which varied \nacross participants). \nAlthough the wake group exhibited no significant variation in recognition speed between the \ntest and the retest (see their Fig. 2c), the accuracy data\uf0beas re-analyzed here\uf0beshowed forgetting \n(-12.2%; F(1,29) = 13.28, p < .002; see this Fig. 2c). In contrast, the sleep group showed the \nreverse: despite a robust improvement in terms of speed (possibly due to consolidated practice, \nwhich the authors did not allude to), these participants showed no reliable change in accuracy (+ \n.4%; F < 1; Group x Session: F(1,58) = 7.37, p < .009). Again, ANOVAs on arcsined \nproportions showed exactly the same results. \nHence, on the surface (i.e., based on net performance), the recognition data seem to support \nthe view that sleep only prevents memory loss. To make this more concrete, just compare Fig. 2a \nwith Fig. 2c. Given this apparent conflict with the free recall results, an analysis of item \ntrajectory/fate is clearly needed. \n4.1 \"Item fate\" analysis \nAs in free recall, the probability of recovering a previously unretrieved item was weaker than \nthat of maintaining one, across test and retest (i.e., gained: 47.3% vs. maintained: 80.7%). But \nonce again, gained items had a higher sleep-to-wake ratio, of 1.60 (from 36.3% for wake, to \n58.2% for sleep) than maintained items, with 1.06 (from 78.3% to 83.0%; see Fig. 2d), indicating \nthat gained items were the ones to benefit most from sleep. \nThis difference was attested by the presence of a significant interaction between group and \nitem trajectory (F(1,58) = 6.99, p < .02) on the normalized proportions. In contrast to free recall, \nhowever, the effect of sleep was significant only for the gained items (F(1,58) = 8.47, p < .006), \nnot for the maintained items (F(1,58) = 1.28, p > .26).\nv\n In short, although the maintained items \nIn press at Cortex \n13 \n \nshowed a weaker sleep effect here, compared to free recall, the same configuration emerged: \nsleep did not just prevent forgetting, it also enhanced memory accessibility. In fact, the memory \nenhancement was stronger than the prevention of forgetting. \n5. Correlational Analysis \nTo determine whether this accessibility boost was driven by the same mechanism as \nprotection against loss, I computed the correlations between participants' ability to maintain \nitems and their ability to gain others during the twelve hours of the test/retest interval. The two \ntasks (free recall and old/new recognition) showed the same dissociation (see Fig. 3 a-d): \nwhereas the proportions of maintained and gained items were positively correlated after wake \n(r(36) = .42, p < .02; r(30) = .38, p < .04), they were not correlated after sleep (r(36) = .19, p > \n.27; r(30) = .10, p > .6, respectively). The positive correlations after wake may indicate that \nparticipants who were good at structuring their recall, not only maintained more items, but also \nhad more time left to look again and gain items. But this is as far as it goes: these differences in \nrecall strategies did not seem determine how well participants maintained and/or gained items in \nthe recognition task, as indicated by the absence of correlations across tasks (maintained: r(29) = \n.16, p > .42; gained: r(29) = .17, p > .37; pooled: r(29) = .05, p > .80)\nvi\n. More strikingly, the \npresence of a non-linear change, from covariation (during wake) to independence (during sleep), \ntells us that during sleep something different was going on, and that protection against loss and \naccessibility enhancement, at that point, no longer reflected the same mechanism. \n6. Discussion \nSleep improves the accessibility of declarative memories more than it protects them against \ninterference. In two studies in which net recall/recognition showed forgetting after 12 hours of \nwake, item fate analyses revealed that the post-sleep advantage was predominantly due to sleep \nIn press at Cortex \n14 \n \npromoting access to memory traces that had initially been too weak to be retrieved. This is at \nodds with the conclusions drawn by Fenn and Hambrick (2013), who argued only for protection \nagainst loss. As I have shown, their dataset was flawed by a ceiling effect, which may have \nmasked the true effect of sleep on memory. This also casts doubt on their recent paper (Fenn & \nHambrick, 2015) on consolidation and IQ, in which in all likelihood a similar flaw was at play. \nThe observed post-sleep boost in memory accessibility is not easily explained by the \"anti-\nforgetting\" account (Wixted, 2004; see also Mednick et al., 2011). It is even less so in view of \nthe fact that sleep seems to wipe out the correlations between the proportion of maintained and \ngained items that can be seen after wake (see above). In this model, the fundamental hypothesis \nis that, during slow-wave sleep, the hippocampus is unable to form new memories. As new \nhippocampal encoding would necessarily erode recently formed traces, by blocking this process \nsleep could only prevent retroactive interference, and thereby memory loss (cf. Wixted, 2004, p. \n261). \nStill, proponents of the anti-forgetting view may argue that the sleep-associated supplement of \ngained items is tied to the endogenous nature of the free recall task. Attempts made at recalling \nitems at the 0-hr test could set the brain in motion to search actively\uf0beeven offline\uf0befor those \nmemories that it failed to access. Under this view, the post-sleep accessibility boost therefore \nwould not reflect a genuine representational enhancement, but rather the fact that this offline \nsearch has been promoted by sleep, as is the case for similar phenomena, such as insight \n(Fischer, Drosopoulos, Tsen, & Born, 2006) and problem solving (Gupta, Jang, Mednick, & \nHuber, 2012; Sio, Monaghan, & Ormerod, 2013). And since participants in the recognition task \nwere all tested in free recall as well, this explanation would hold also for the boost observed in \nthe more exogenous, old/new recognition. \nIn press at Cortex \n15 \n \nTo test this hypothesis, I looked at the trajectory of each item in old/new recognition, taking \ninto account whether or not the item had been successfully recalled at the 0-hr test. Since the \nputative search should not apply to items correctly recalled at test, these  should return a weaker \naccessibility boost (compared to unrecalled items) \uf0beif not also less protection against loss\uf0beat \nthe post-sleep recognition retest. As shown by the pooled proportions of gained and maintained \nitems (taken as an estimate of overall consolidation), the sleep advantage in the recognition data \n(expressed as the sleep-to-wake ratio) was unaffected by the fate of the item at the 0-hr recall \n(recalled: 1.27 (from 1.12 to .88) vs. unrecalled: 1.21 (from 1.10 to .90); F(1,58) = 1.35, p > .25). \nThis clearly does not support the \"offline search\" interpretation. \nThe post-sleep boost in memory accessibility seems to indicate instead that declarative \nmemories can be sharpened overnight, an option clearly envisaged by the active consolidation \naccount (Diekelmann & Born, 2010). If by reactivating cortical regions that originally \ncontributed to its encoding of a new trace, the hippocampus promotes binding amongst those \nregions, changes in memory strength could ensue, resulting in more accessible, more vivid \nrepresentations. From this perspective, the present results nicely complement the cued \nreactivation effects obtained in slow-wave sleep (e.g., Rasch & Born, 2007) in supporting this \naccount. They also predict that the latter effects should be most visible on gained (as opposed to \nmaintained) items. \nThe benefit of sleep on maintained items obtained in the free recall task is, nonetheless, in \nkeeping with the predictions of the anti-forgetting account (Wixted, 2004). If the impossibility \nfor the hippocampus to encode during slow-wave sleep means that memories formed shortly \nbeforehand can be hardened more promptly and with minimal damage, then sleep (as opposed to \nwake) should result in less forgetting. This is exactly what the present data demonstrate at the \nIn press at Cortex \n16 \n \nitem/representational level, at least when the total number of items that could have been \nmaintained is taken into account (cf. Roediger & Challis, 1989). \nIn sum, a statistical approach that focuses on item fate while bearing in mind what is \ntheoretically possible within each category of item-trajectory provides a much more nuanced and \ninformative picture of the role of sleep on memory consolidation. In this picture, we see distinct \nevidence confirming complementary aspects of both the anti-forgetting and the active-\nconsolidation account (e.g., Diekelmann & Born, 2010; Wixted, 2004). However, we also see \nthat, contrary to what net performance has led us to believe, protection against forgetting plays \nonly a minor part. \nAlthough the present data compared nocturnal sleep to active wake, they should not be taken \nto indicate that sleep plays a unique role in consolidating memories. According to the notion of \n\"opportunistic\" consolidation (Mednick et al., 2011, Wixted & Cai, 2014; see also Hasselmo, \n1999), the brain consolidates recently acquired information whenever the hippocampus is not \nbusy encoding new traces. A period of reduced encoding, whether achieved by slow-wave sleep, \nor any other means, including wakeful rest (e.g., Dewar, Alber, Butler, Cowan, & Della Sala, \n2012) would be sufficient to promote synaptic (hippocampal) consolidation and systems \nconsolidation, possibly with some temporal overlap. Therefore, the sleep manipulations which I \nrelied upon may be best seen as manipulations of the amount of trace encoding between \nacquisition and retest. In any case, if one assumes that protection against loss and increased \naccessibility reflect synaptic vs. systems consolidation, the overnight disappearance of their \ncorrelation suggests that although they may be triggered by the same factor, synaptic and \nsystems consolidation quickly become free from each other. \n \nIn press at Cortex \n17 \n \nAcknowledgments \nI thank Kimberly Fenn and Jakke Tamminen, for making their raw data available. I am also \ngrateful to Arty Samuel, Magdalena Abel, Henry Roediger III and two anonymous reviewers, for \ncommenting on earlier versions of this manuscript. \n \n  \nIn press at Cortex \n18 \n \nReferences \nBallard, P. B. (1913). Oblivescence and reminiscence. British Journal of Psychology Monograph \nSupplements, 1, 1-82. \nBrown, W. (1923). To what extent is memory measured by a single recall trial? Journal of \nExperimental Psychology, 6, 377-382. \nDewar, M., Alber, J., Butler, C., Cowan, N., & Della Sala, S. (2012). Brief wakeful resting \nboosts new memories over the long term. Psychological Science, 23, 955-960. \nDiekelmann, S., & Born, J. (2010). The memory function of sleep. Nature Reviews \nNeuroscience, 11, 114-126. \nDumay, N., & Gaskell, M. G. (2005). Do words go to sleep? Exploring consolidation of spoken \nforms through direct and indirect measures. Behavioral and Brain Sciences, 28, 69-70. \nDumay, N., & Gaskell, M. G. (2007). Sleep-associated changes in the mental representation of \nspoken words. Psychological Science, 18, 35-39. \nDumay, N., & Gaskell, M. G. (2012). Overnight lexical consolidation revealed by speech \nsegmentation. Cognition, 123, 119-132. \nEkstrand, B. R. (1972). To sleep, perchance to dream (about why we forget). In C. P. Duncan, L. \nSechrest, and A. W. Melton (Eds.), Human memory: Festschrift for Benton J. Underwood \n(pp. 59-82). New York: Appelton-Century-Crofts. \nErdelyi, M. (1984). The recovery of unconscious (inaccessible) memories: Laboratory studies of \nhypermnesia. In G. H. Bower (Ed.), The psychology of learning and motivation: Advances in \nresearch and theory (Vol. 18, pp. 95-127). New York: Academic Press. \nFenn, K. M., & Hambrick, D. Z. (2013). What drives sleep-dependent memory consolidation: \nGreater gain or less loss? Psychonomic Bulletin & Review, 20, 501-506. \nIn press at Cortex \n19 \n \nFenn, K. M., & Hambrick, D. Z. (2015). General intelligence predicts memory change across \nsleep. Psychonomic Bulletin & Review, 22, 791-799. \nFischer, S., Drosopoulos, S., Tsen, J., & Born, J. (2006). Implicit learning-Explicit knowing: A \nrole for sleep in memory system interaction. Journal of Cognitive Neuroscience, 18, 311-319. \nGais, S., Lucas, B., & Born, J. (2006). Sleep after learning aids memory recall. Learning & \nMemory, 13, 259-262. \nGais, S., Plihal, W., Wagner, U., & Born, J. (2000). Early sleep triggers memory for early visual \ndiscrimination skills. Nature Neuroscience, 3, 1335-1339. \nGaskell, M. G., & Dumay, N. (2003). Lexical competition and the acquisition of novel words. \nCognition, 89, 105-132. \nGaskell, M. G., Warker, J., Lindsay, S., Frost, R., Guest, J., Snowdon, R., & Stackhouse, A. \n(2014). Sleep underpins the plasticity of language production. Psychological Science, 25, \n1457-1465. \nGupta, N. G., Jang, Y., Mednick, S. C., & Huber, D. E. (2012). The road not taken: Creative \nsolutions require avoidance of high-frequency responses. Psychological Science, 23, 288-294. \nHasselmo, M. E. (1999). Neuromodulation: acetylcholine and memory consolidation. Trends in \nCognitive Sciences, 3, 351-359. \nJenkins, J. M., & Dallenbach, K. M. (1924). Obliviscence during sleep and waking. American \nJournal of Psychology, 35, 605-612. \nJi, D., & Wilson, M. A. (2007). Coordinated memory replay in the visual cortex and \nhippocampus during sleep. Nature Neuroscience, 10, 100-107. \nKarni, A., Tanne, D., Rubenstein, B. S., Askenasy, J. J. M., & Sagi, D. (1994). Dependence on \nREM sleep of overnight perceptual skill. Science, 265, 679-682. \nIn press at Cortex \n20 \n \nLeach, L., & Samuel, A. G. (2007). Lexical configuration and lexical engagement: When adults \nlearn new words. Cognitive Psychology, 55, 306-353. \nLoftus, G. R., & Masson, M. E. J. (1994). Using confidence intervals in within-subject designs. \nPsychonomic Bulletin & Review, 1, 476-490. \nMednick S. C., Cai D. J., Shuman T., Anagnostaras S., Wixted J. T. (2011). An opportunistic \ntheory of cellular and systems consolidation. Trends in Neurosciences, 34, 504-514. \nPayne, J. D., Tucker, M. A., Ellenbogen, J. M., Wamsley, E. J.,Walker, M. P., Schacter, D. L., & \nStickgold, R. (2012). Memory for semantically related and unrelated declarative information: \nThe benefit of sleep, the cost of wake. PLoS ONE, 7, e33079. \nPlihal, W., & Born, J. (1997). Effects of early and late nocturnal sleep on declarative and \nprocedural memory. Journal of Cognitive Neuroscience, 9, 534-547. \nPlihal, W., & Born, J. (1999). Effects of early and late nocturnal sleep on priming and spatial \nmemory. Psychophysiology, 36, 571-582. \nRasch, B., Gais, S., & Born, J. (2009). Impaired off-line consolidation of motor memories after \ncombined blockade of cholinergic receptors during REM sleep-rich sleep. \nNeuropsychopharmacology, 34, 1843-1853. \nRasch, B., Buchel, C., Gais, S., & Born, J. (2007). Odor cues during slow-wave sleep prompt \ndeclarative memory consolidation. Science, 315, 1426-1429. \nRoediger, H. L., & Challis, B. H. (1989). Hypermnesia: Improvements in recall with repeated \ntesting. In C. Izawa (Ed.), Current Issues in Cognitive Processes: The Tulane Floweree \nSymposium on Cognition (pp. 175-199). Hillsdale (NJ): Lawrence Erlbaum Associates. \nSchreiner, T., & Rasch, B. (2014). Boosting vocabulary learning by verbal cueing during sleep. \nCerebral Cortex (Epub ahead of print). DOI: 10.1093/cercor/bhu139 \nIn press at Cortex \n21 \n \nStickgold, R., James, L. & Hobson, J. A. (2000). Visual discrimination learning requires sleep \nafter training. Nature Neuroscience, 3, 1237-8. \nSio, U. N., Monaghan, P., & Ormerod, T. (2012). Sleep on it, but only if it is difficult: Effects of \nsleep on problem solving. Memory & Cognition, 41, 159-166. \nTakashima, A., Nieuwenhuis, I. L. C., Jensen, O., Talamini, L. M., Rijpkema, M., & Fern\u00e1ndez. \nG. (2009). Shift from hippocampal to neocortical centered retrieval network with \nconsolidation. Journal of Neuroscience, 29, 10087-10093. \nTamminen, J., Payne, J. D., Stickgold, R., Wamsley, E. J., & Gaskell, M. G. (2010). Sleep \nspindle activity is associated with the integration of new memories and existing knowledge. \nJournal of Neuroscience, 30, 14356-14360. \nTulving, E. (1964). Intratrial and intertrial retention: Notes towards a theory of free recall verbal \nlearning. Psychological Review, 71, 219-237. \nTulving, E. (1967). The effects of presentation and recall of material in free-recall learning. \nJournal of Verbal Learning and Verbal Behavior, 6, 175-184. \nWilson, M. A., & McNaughton, B. L. (1994). Reactivation of hippocampal ensemble memories \nduring sleep. Science, 265, 676-679. \nWixted, J. T. (2004). The psychology and neuroscience of forgetting. Annual Review of \nPsychology, 55, 235-269. \nWixted, J. T., & Cai, D. J. (2014). Memory consolidation. In K. Ochsner & S. Kosslyn (Eds.), \nOxford Handbook of Cognitive Neuroscience (pp. 1-59). New York: Oxford University Press. \nYaroush, R., Sullivan, M. J., & Ekstrand, B. R. (1971). The effect of sleep on memory: II. \nDifferential effect of the first and second half of the night. Journal of Experimental \nPsychology, 88, 361-366. \nIn press at Cortex \n22 \n \nFigure captions \nFig. 1. Participants' performance at the 0-hr test plotted against magnitude of the 12-hr \nimprovement (%) in Fenn and Hambrick's (2012) restricted dataset, plus trend line and \ncorrelation value. \nFig. 2. (a) Free recall percent correct for the sleep and wake groups in the DG and TPSWG \nrestricted dataset (n = 72). (b) Normalized proportions of \"maintained\" and \"gained\" items. (c) \nOld/new recognition percent correct in TPSWG (n = 60). (d) Normalized proportions of \n\"maintained\" and \"gained\" items. Error bars show 95% confidence intervals based on Loftus and \nMasson (1994). \nFig. 3. Proportion of gained items plotted against that of maintained items, plus trend line and \ncorrelation value for: (a) free recall after wake, (b) old/new recognition after wake, (c) free recall \nafter sleep, and (d) old/new recognition after sleep. Note. In (a) the correlation is still substantial \neven after removing the outlier with a normalized proportion gained of 3.75 (r(35) = .39, p < .02; \nsee alternate plot). \n  \nIn press at Cortex \n23 \n \nFig. 1 \n \n \n \n \n  \n-30\n-20\n-10\n0\n10\n20\n30\n40\n50 60 70 80 90 100\n1\n2\n h\nr \nC\nh\nan\nge\n (\n%\n)\nPercent Correct At Test\nr(322) = -.28, p < .0001 \nIn press at Cortex \n24 \n \nFig. 2  \n \n(a) \n \n \n(b) \n \n(c) \n \n(d) \n \n \n  \n0.0\n5.0\n10.0\n15.0\n20.0\n25.0\n30.0\n35.0\n0 hr 12 hr\nP\ne\nrc\ne\nn\nt \nC\no\nrr\ne\nc\nt\nSleep\nWake\n1,20\n1,39\n0,80\n0,61\n0.0\n0.2\n0.4\n0.6\n0.8\n1.0\n1.2\n1.4\n1.6\n1.8\n2.0\nMaintained Gained\nN\no\nrm\na\nliz\ne\nd\n P\nro\np\no\nrt\nio\nn\nSleep\nWake\n35.0\n40.0\n45.0\n50.0\n55.0\n60.0\n65.0\n70.0\n0 hr 12 hr\nP\ne\nrc\ne\nn\nt \nC\no\nrr\ne\nc\nt\nSleep\nWake\n1,04\n1,26\n0,96\n0,70\n0.0\n0.2\n0.4\n0.6\n0.8\n1.0\n1.2\n1.4\n1.6\n1.8\n2.0\nMaintained Gained\nN\no\nrm\na\nliz\ne\nd\n P\nro\np\no\nrt\nio\nn\nSleep\nWake\nIn press at Cortex \n25 \n \nFig. 3 \n \n \n(a) \n \n(b) \n \n \n(c) \n \n(d) \n \n \n \n                                                          \ni\n The data of the second retest at 24 hours or seven days are irrelevant to the present argument. \nii\n Just to illustrate, if participant A maintains 4 items at retest out of 8 recalled at test, and \nparticipant B also maintains 4 items, but out of 6 recalled at test, Fenn and Hambrick would give \nthem both a count of 4 maintained items, while in fact participant B maintains relatively more of \nwhat s/he initially acquired. Similarly, if both participants gain 4 items at retest, participant A \n0\n0.2\n0.4\n0.6\n0.8\n1\n1.2\n1.4\n1.6\n1.8\n0 0.5 1 1.5 2\nG\nai\nn\ned\n (\nN\no\nrm\nal\niz\ned\n %\n)\nMaintained (Normalized %)\nfree recall after wake:\nr (35) = .39, p < .02\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0 0.5 1 1.5 2\nG\nai\nn\ned\n (\nN\no\nrm\nal\niz\ned\n %\n)\nMaintained (Normalized %)\nfree recall after wake:\nr (36) = .42, p < .02\n0\n0.5\n1\n1.5\n2\n2.5\n0 0.2 0.4 0.6 0.8 1 1.2 1.4\nG\nai\nn\ned\n (\nN\no\nrm\nal\niz\ned\n %\n)\nMaintained (Normalized %)\nrecognition after wake:\nr (30) = .38, p < .04\n0\n0.5\n1\n1.5\n2\n2.5\n3\n3.5\n4\n0 0.5 1 1.5 2\nG\nai\nn\ned\n (\nN\no\nrm\nal\niz\ned\n %\n)\nMaintained (Normalized %)\nfree recall after sleep:\nr (36) = .19, p > .27\n0\n0.5\n1\n1.5\n2\n2.5\n0 0.2 0.4 0.6 0.8 1 1.2 1.4\nG\nai\nn\ned\n (\nN\no\nrm\nal\niz\ned\n %\n)\nMaintained (Normalized %)\nrecognition after sleep:\nr (30) = .097, p > .61\nIn press at Cortex \n26 \n \n                                                                                                                                                                                           \ngains more of what s/he did not recall at test (say 12 if the maximum score is 20), but still, both \nwould get the same count of gained items by the Fenn and Hambrick's procedure. \niii\n And the fact that, in both papers (Fenn & Hambrick, 2013, 2015), the sleep and wake \ngroups were (on average) matched for number of items recalled at test does not solve the \nproblem. What counts is how many items can theoretically be maintained vs. gained at the \nparticipant level. Imagine a wake group of two subjects, with respectively 4 and 6 out of 10 as \nnumber of items recalled at test, and a sleep group of also two, with scores of 3 and 7. Although \ntheir average test scores are identical, a subsequent gain of three items for every participant \nactually mean an average proportion gain of .63 for the wake group, but one of .71 for the sleep \ngroup. Clearly, unless the groups are matched pairwise, raw item counts would miss out on an \neffect of sleep. \niv\n The same analysis on raw item counts detected an effect of sleep on gained items (sleep: 3.1 \nvs. wake: 1.3; F(1,68) = 23.33, p < .0001), but missed out on the corresponding effect on \nmaintained items (sleep: 3.9 vs. wake: 2.8; F(1,68) = 1.68, p > .19; Group x Item trajectory: \nF(1,68) = 7.17, p < .01). \nv\n The same analysis on raw item counts showed a similar pattern to normalized proportions, \nwith no effect of sleep on maintained items (sleep: 9.33 vs. wake: 9.60; F < 1), but one on gained \nitems (sleep: 2.13 vs. wake: 1.1; F(1,58) = 9.22, p < .004; Group x Item trajectory: F(1,68) = \n7.52, p < .009). \nvi\n Note that the latter analysis should be taken with caution, as it required inclusion of all \nparticipants tested by TPSWG.  \n \n", "identifiers": ["oai:ore.exeter.ac.uk:10871/31235", "10.1016/j.cortex.2015.06.007"], "journals": [{"title": null, "identifiers": ["issn:0010-9452", "0010-9452"]}], "publisher": "Elsevier", "relations": ["http://hdl.handle.net/10871/17864"], "repositories": [{"id": "504", "openDoarId": 0, "name": "Open Research Exeter", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "baseId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 1, "metadataAdded": 1517985833000, "metadataUpdated": 1591071170000, "depositedDate": 1517270400000}, "subjects": ["Article"], "title": "Sleep not just protects memories against forgetting, it also makes them more accessible", "topics": ["memory consolidation", "sleep", "forgetting", "reminiscence", "item fate"], "types": [], "year": 2016, "fulltextIdentifier": "https://core.ac.uk/download/pdf/146503003.pdf", "doi": "10.1016/j.cortex.2015.06.007", "oai": "oai:ore.exeter.ac.uk:10871/31235", "downloadUrl": "https://core.ac.uk/download/pdf/146503003.pdf"}}
{"status": "OK", "data": {"id": "18509190", "authors": ["Homburg, V.M.F. (Vincent)", "Dijkshoorn, A.J.D. (Andres)", "Thaens, M. (Marcel)"], "citations": [], "contributors": [], "datePublished": "2013-01-01", "description": "In many European countries, municipalities are becoming increasingly important as providers of electronic public services to their citizens. One of the horizons for further expansion is the delivery of personalised electronic services. In this paper, we describe the diffusion of personalised services in the Netherlands over the period 2006-2009 and investigate how and why various municipalities adopted personalised electronic services. In achieving this, we analyse data that were gathered during interviews with key stakeholders in ten selected Dutch municipalities. We synthesise the findings in an explanatory model of personalised electronic service delivery diffusion. The model emphasizes persuasive pressures that are channelled to potential adopters of personalised services. Furthermore, the model shows how persuasive pressure (as perceived by adopters) is followed-up by organisational search activities, and how, in various circumstances, the idea of personalised services is \u2018framed\u2019 by innovation champions, knowledge brokers and new members of staff as to appeal to specific organisational priorities and ambitions. In doing so, this article contributes to an institutional view on adoption and diffusion of innovations, in which (1) horizontal and vertical channels of persuasion and (2) human agency, rather than technological opportunity and rational cost-benefit considerations, account for actual diffusion of innovations", "fullText": " 1 \nDiffusion of personalised services among Dutch municipalities: evolving \nchannels of persuasion  Paper submitted to Local Government Studies  Vincent Homburg (corresponding author) Erasmus University Rotterdam Faculty of Social Sciences, Public Administration, Comparative Public Services Innovation PO Box 1738 NL3000 DR Rotterdam The Netherlands Phone +31 10 4081863 Email: homburg@fsw.eur.nl  Andres Dijkshoorn Erasmus University Rotterdam Faculty of Social Sciences, Public Administration, Comparative Public Services Innovation PO Box 1738 NL3000 DR Rotterdam The Netherlands Phone: +31 10 4082526 Email: dijkshoorn@fsw.eur.nl   Marcel Thaens Erasmus University Rotterdam Faculty of Social Sciences, Public Administration, Comparative Public Services Innovation PO Box 1738 NL3000 DR Rotterdam The Netherlands Phone: +31 10 4082331 Email: thaens@fsw.eur.nl    Main text (including tables and figure, excluding footnotes and references): 6109 words  \n 2 \nDiffusion of personalised services among Dutch municipalities: evolving \nchannels of persuasion  \nABSTRACT  In many European countries, municipalities are becoming increasingly important as providers of electronic public services to their citizens. One of the horizons for further expansion is the delivery of personalised electronic services. In this paper, we describe the diffusion of personalised services in the Netherlands over the period 2006-2009 and investigate how and why various municipalities adopted personalised electronic services. In achieving this, we analyse data that were gathered during interviews with key stakeholders in ten selected Dutch municipalities. We synthesise the findings in an explanatory model of personalised electronic service delivery diffusion. The model emphasizes persuasive pressures that are channelled to potential adopters of personalised services. Furthermore, the model shows how persuasive pressure (as perceived by adopters) is followed-up by organisational search activities, and how, in various circumstances, the idea of personalised services is \u2018framed\u2019 by innovation champions, knowledge brokers and new members of staff as to appeal to specific organisational priorities and ambitions. In doing so, this article contributes to an institutional view on adoption and diffusion of innovations, in which (1) horizontal and vertical channels of persuasion and (2) human agency, rather than technological opportunity and rational cost-benefit considerations, account for actual diffusion of innovations.   \nINTRODUCTION  The use of information and communication technologies (ICTs) is central in many national policies as the means to actually realise government modernisation and transformation agendas. In many western countries, it is especially local governments that are developing one-stop shops that serve as a point-of-entry to the whole range of government (Ling, 2002; Ho, 2002; Beynon-Davies and Martin, 2004; Mulgan, 2005) in such a way that the relationship between public sector organisations and citizens is re-engineered (Beynon-Davies and Martin, 2004). This redesign phenomenon has been termed electronic government (Bekkers and Homburg, 2005).   Internationally, the actual implementation and take-up of the e-government phenomenon by public sector organisations has lagged behind policy ambitions. In trying to explain the actual diffusion of e-government, various empirical studies have identified city size, citizen demand, organisational structure, geographic location and capacity (see Table 1 for details of the literature reviewed) as the most important determinants of e-government adoption by public sector organisations.   --- Table 1 somewhere 1  Although useful in itself, these determinants do not provide insights into questions such as why public sector organisations actually adopt (or fail to adopt) e-government innovations, and how such organisations actually learn to \n 3 \ninnovate. Further, theoretical insights such as Orlikowksi\u2019s (2000) practice lens, Cziarniawska and Sevon\u2019s (2005) travelling of ideas, Bekkers and Homburg\u2019s (2005) information ecology of e-government and Homburg and Georgiadou\u2019s (2009) technological \u2018contact infection\u2019 point towards human agency and institutional pressures as potential explanations for the variance in the prevalence of e-government services among populations of public sector organisations. Zorn et al. (2011) have argued that, especially for non- and not-for-profit organisations, technological innovations are a means for establishing legitimacy in the eyes of key stakeholders as much as they are means for enhancing operational efficiency. Further, and related to this, Frumkin and Galaskiewicz (2004) indicate that institutional pressures to adopt specific innovations are very relevant for public sector organisations.    Our hope is that, by focusing on \u2018agency\u2019 alongside \u2018structure\u2019 (Orlikowksi and Barley, 2001), more light will be shed on the process of technological and organisational change (Pettigrew, 1985) such that, eventually, the diffusion of e-government will be better understood. This paper seeks to examine the diffusion of a specific, fairly mature form of e-government, namely personalised e-government, in a specific setting (municipalities in the Netherlands). The analysis reported in this paper extends the existing literature on e-government to include both \u2018agency\u2019 and \u2018process\u2019 aspects of e-government innovation.  \nPERSONALISATION AND PERSONAL PUBLIC SERVICE DELIVERY  Various authors have argued that various \u2018stages\u2019 or \u2018levels of maturity\u2019 can be discerned in electronic service delivery (Layne and Lee, 2001; Anderson and Henriksen, 2005), ranging from public organisations offering one-way transmission of static information, through organisations offering online transactions, to organisations offering integrated services (which incorporate information from sources external to the organisation and through which the citizen may interact in the first place). A supposed characteristic of the \u2018final stage\u2019 in these maturity models is that there is a seamless integration of information services across administrative boundaries.   Recently, the idea of integration was further pushed by it being given strong support (OECD, 2009) and the related academic discussion of personalised integrated services (King and Cotterill, 2007; Peterson et al., 2007; Homburg and Dijkshoorn, 2011). A characteristic of personalised services is that they generally, when using Customer Relationship Management systems (King and Cotterill, 2007), make use of authorisation, profiling and customisation in such a way that, eventually, one-to-one relationships between public service providers and citizens are established. One-to-one relationships may provide citizens with, for example, pre-filled forms, suggestions for permits or benefits that maybe relevant given past requests, automatically generated reminders, and news updates based on customer preferences. The eventual aim is to provide services that are geared towards the needs of citizens, and less towards the existing supply-oriented organisational routines of service providers.    Examples of personalised services at national or federal levels are the Belgian MyMinFin e-tax initiative, the Danish borger.dk portal, the Estonian eesti.ee \n 4 \ninitiative, the French mon.service-public.fr website, the Norwegian Norway.no portal, the British direct.gov.uk site and the Dutch mijnoverheid.nl site. In the Netherlands, various municipalities offer personalised sections on their websites including http://www.eindhoven.nl/mijn-eindhoven.htm for the city of Eindhoven, and http://www.rotterdam.nl/mijn_loket _digid for Rotterdam.  Implementing personalised electronic government services is far from trivial and the literature reports many obstacles. These include socio-political problems of information sharing across traditional organisational boundaries (Mulgan, 2005; Homburg, 2008), the existence of legacy systems (Pieterson et al., 2007) and concerns related to privacy issues (Lips et al., 2004).  In this paper, we are not arguing that personalised e-government service delivery is, or should be, a necessary next step, nor do we intend to discuss whether personalisation should or should not be adopted by public sector organisations. Rather, we analyse personalised e-government services as a \u2018case\u2019 of the diffusion of a specific innovation and analyse how the diffusion of personalised e-government service delivery takes place. In order to be able to explain the diffusion, we focus on a population of 441 potential adopters in a single national jurisdiction, the Netherlands, which can be seen as a decentralised unitary state (Esping-Andersen, 1990; Pollitt and Bouckaert, 2004). Dutch municipal governments are relatively autonomous vis-\u00e0-vis central government with respect to management issues, including the design, implementation and maintenance of electronic services (van Os, 2011). In practice, however, various initiatives exist that are aimed at coordinating e-government development. For instance, the ICTU Foundation is an outreach initiative, governed jointly by the national Ministry of the Interior and Kingdom Relations and the Association of Dutch Municipalities, which attempts to facilitate and support all kinds of public organisations including municipalities in deploying ICTs. In specific policy sectors, initiatives (such as the BKWI in the social security and welfare domain) exist in which local and central governments cooperate in the development of electronic services. At the same time, municipalities cooperate in the context of outreach initiatives such as DIMPACT to share knowledge and to collectively purchase services, hardware and software. These initiatives and organisations are mentioned here to illustrate that although municipalities are relatively autonomous, the development and management of ICTs, including electronic services, does not take place in a vacuum and that therefore explanations that explicitly include contexts might be relevant in explaining the diffusion of personalised e-government services (King et al., 1994).   \nRELATED RESEARCH: THE THEORETICAL ANTECEDENTS OF DIFFUSION Diffusion of a new idea, product or service is defined as the spread of its use in a population of potential adopters (Rogers, 1995; King et al., 1994). The process of diffusion has been linked to characteristics of the innovation itself, the social system (community of potential adopters), channels of communication, and time (Rogers, 1995; Mahajan and Peterson, 1985). Given the lack of a rigorous theory that acknowledges the embeddedness of e-government diffusion processes  by municipalities in the larger context of governance (including central-local relationships and the activities of outreach programmes), we scanned and \n 5 \nsearched related concepts in the organisational sociology and institutional theory disciplines - two related literature streams that explicitly address how organisations absorb prevailing ideas from their environments into their operations and structures. However, this literature tends to be relatively abstract, and does not address the diffusion of personalised e-government in the specific context of municipalities.   Advancements in the discipline of organisational sociology in recent decades, such as the emergence of \u2018new institutionalism\u2019 (DiMaggio and Powell, 1983; Tolbert and Zucker, 1996), have highlighted the significance of the professional and/or legal rules, cognitive structures, norms and the prevailing values in which innovation takes place. Institutionalism holds that the adoption of innovations does not take place because individuals or (private as well as public) organisations make rational, calculated decisions regarding costs and benefits (see, for instance, Venkatesh et al., 2003), but rather that: (1) decisions whether or not to adopt innovations are influenced by rules, conventions, principles, belief systems and social context criteria,  and   (2) attempt to shape and affect these overarching structures of norms, beliefs and rules that together form the concept of institutions.  DiMaggio and Powell refer to these kinds of decisions as \u2018collectively rational\u2019, as opposed to individually rational, because the rationality involved in the decision-making processes is influenced through information provided by others. Hence, institutionalism emphasises the persuasive control over the practices, beliefs and belief systems of individuals or organisations through an institution\u2019s sway (Kimberley, 1979, in King et al., 1994). Persuasion can be achieved not only through directives, but also through more gentle but nevertheless potentially convincing means such as deploying specific knowledge, subsidising activities deemed \u2018appropriate\u2019 by national government, standard-setting, raising awareness and generally promoting specific technologies (King et al., 1994). Moreover, Venkatraman et al. (1994) have stated that persuasion can occur both through vertical channels of communication (initiated by actors outside the set of potential adopters, such as central government; see also Leonard-Barton and Rogers, 1981; Moon and Bretschneider, 1997; Bobrowski and Bretschneider, 1994) as well as through processes of mimicking and \u2018word-of-mouth\u2019 diffusion (Wang and Doong, 2010) involving communication, interaction and persuasion among potential adopters (DiMaggio and Powell, 1983). Innovation through mimicry is likely to occur when innovations are socially visible (Mahajan and Peterson, 1985; Dos Santos and Peffers, 1998), when causes, conditions and consequences are known (absence of causal ambiguity, Barney, 1991; Loh and Venkatraman, 1992) and when the success of the innovation is unlikely to be determined by path dependencies (Barney, 1991; Loh and Venkatraman, 1992).  The organisational sociology and institutional theory literatures are helpful in identifying an organisation\u2019s environment as a source of \u2018pressure\u2019 for organisations to behave in a particular way. Nevertheless, these literature streams do not explicitly address how, through \u2018pressure\u2019, innovations are diffused among organisations. Additional theoretical guidance related to this issue is provided by the so-called Scandinavian Institutionalism (Czarniawska and Sevon, 2005). This school of thought sees innovations as \u2018ideas\u2019 insofar as \n 6 \nthey can be viewed as artefacts. In order for ideas (such as \u2018personalisation\u2019) to spread (using either horizontal or vertical channels of communication) they must be translated into, or associated with, a success story, image or even a myth. During its travel, the idea itself is likely to change (Czarniawska and Sevon, 2005). As such, the idea of translation is a much more complex concept than the notions of \u2018diffusion, \u2018adoption, \u2018mimicry\u2019 and \u2018direction\u2019 suggest, and involves various phases in which so-called change agents (Caldwell, 1996) are actively engaged in the process of adoption. Agency in the process of translation takes place in horizontal as well as vertical modes of persuasion, and change agents (experts, boundary-spanning agents, consultants, knowledge brokers etc.) may operate on both the supply and the demand sides of the translation (King et al., 1994).   To sum up, we draw on concepts selected from organisational sociology and institutional theories that are relevant to the diffusion of a specific instance of e-government. The general sociological and institutional literatures sensitise us to horizontal and vertical \u2018persuasive\u2019 pressures, in the form of beliefs, cognitive structures, norms and values, that are put on municipalities to adopt specific innovations (including e-government). Scandinavian Institutionalism adds the aspect of \u2018agency\u2019 to this: diffusion is not a deterministic process but an intricate social process that involves translation activities by experts, boundary-spanning agents and knowledge brokers. At this moment, however, there is no complete theory available that fully explains the diffusion process. In the analysis section of this paper, we attempt to provide a diffusion model that takes account of the above notions of change.   \nMETHODS AND DATA In order to explain the diffusion of personalised e-government services, we first define our population of interest, that is, the set of potential adopters of the innovation being scrutinised: personalised e-government. We chose to analyse diffusion within one national jurisdiction (in our case, the Netherlands), and our first step was to describe the diffusion pattern of personalised e-government services over a four-year period (2006-2009) delineated by two consecutive municipal elections. For this purpose, we used an existing dataset that is assembled annually by the Dutch \u2018Government has an answer\u2019 project. This project monitors directly observable characteristics of e-government initiatives by Dutch ministries, provinces, municipalities and water regulatory authorities, and annually reports its findings. This monitor resembles the practice of UK local government league tables; for American examples refer to McClure et al. (2000).  In line with our objective to further extend the e-government body of knowledge (that is, to actually build theory), we first described general trends in a population of 441 municipalities1 and then, as a second step, we selected ten municipalities for an in-depth analysis of adoption processes. As the e-government literature consistently reports city size as being a major determinant of e-government adoption (see Table 1; see also Reddick, 2004),                                                         1 In the interval 2006-2009, the population of municipalities shrunk from 458 to 441 due to planned mergers \n 7 \nfrom the 2008 data we selected both adopters (\u2018early adopters\u2019) as well as non-adopters (\u2018laggards\u2019) from substrata, based on city size, of the population (see Appendix A for a list of the selected municipalities). In each of the selected municipalities, qualitative interviews were held with key stakeholders such as council members, city managers, senior ICT managers and managers of public service provision. The starting point for the interviews was a topic list (see Appendix B) through which it was attempted to identify the antecedents, critical events, ongoing activities and interactions with internal and external stakeholders in relation to e-government and public service delivery. All the interviews were recorded, transcribed and analysed2 using back-and-forth coding techniques3 (Miles and Huberman, 1994; Patton, 2002). Following recommended practices for qualitative research, data analysis was conducted in parallel with data gathering, starting with the constructs that emerged from the theoretical antecedents of diffusion. In an iterative process, construct categories were refined, added or combined, and new categories sometimes led to new constructs being identified in the empirical data. Furthermore, the five \u2018non-adopter\u2019 cases were compared to the \u2018adopter\u2019cases based on the occurrence of specific, coded constructs or themes. As a final step, using both theoretical plausibility and induction, relationships between constructs were made explicit: first through \u2018conjectures\u2019 (statements that, through induction, follow from empirical observations, patterns and theoretical analogy), and second through including concepts and relationships in a conceptual framework for understanding the adoption of personalised e-government.  \nANALYSIS: EXPLAINING THE DIFFUSION OF PERSONALISED E-\nGOVERNMENT \n \nDescription of personalised e-government services in Dutch municipalities Table 2 lists the prevalence of certain attributes of personalised electronic service delivery by Dutch municipalities in the years 2006, 2007, 2008 and 2009. Overall, in the period covered, there is an increase in the offered possibility of using DigiD4 authentication (from 20.7% in 2006 to 88.2% in 2009) and on-line payments (from 15.9% in 2006 to 80% in 2009). The growth of possibilities for receiving personalised newsletters, using pre-completed forms, assessing personalised policy consequences and using personalised counters lagged somewhat.  --- Table 2 somewhere here \nPressure: persuasive influence on adoption decisions We identified various categories of pressure (see Table 3) as sources of \u2018pressure\u2019 on decisions to adopt personalised e-government services. Most prominent was the perceived expectations of citizens. As one councillor phrased                                                         2 Note that the interviews were held in Dutch; the authors have translated the quotations given in the analysis section into English. \n3 Using the MaxQDA qualitative analysis tool. \n4 DigiD stands for Digital Identity. With a DigiD users can access a large number of online services offered by Dutch (central and municipal) government agencies. \n 8 \nit:  \n\u201c\u2026 a clamour for service provision, less bureaucracy, transparency: that is \nexternal pressure, as I perceive it. (\u2026) Simply because society does not \ntolerate other kinds of organisational behaviour \u2026\u201d (Councillor)  Another form of influence that was mentioned quite frequently was the existence of benchmarks against which municipalities can be judged. Several respondents indicated that low performance resulted in questions from, for instance, city council members, especially when neighbouring or otherwise comparable municipalities scored considerably higher (\u2018peer rivalry\u2019). As a manager of service provision explained:  \n\u201cTo score well is felt to be important within municipalities. How often is your \nmunicipality mentioned in professional journals, are you in the Top 3\u2026. that \nis considered to be very important\u201d (Manager of service provision)  The fact that municipalities keep a sharp eye on benchmarks and rankings sometimes results in somewhat perverse incentives to adopt personalised services, such as one respondent noted:   \n\u201cOur decision to implement personalised service delivery was due to our low \nranking \u2026 Our councillor wanted to improve our ranking, and we found out \nthat we could improve our ranking quite easily by implementing a \nPersonalised Internet Page \u2026 and so we did\u201d (Project manager).   To summarise, what can be witnessed is that, in line with institutional theory (DiMaggio and Powell, 1983; Ashworth et al., 2009; Lai et al., 2006), all the municipalities reported perceiving persuasive pressure to adopt personalisation measures, both from outside the set of potential adopters (referring to norms to conform to citizens\u2019 needs, or to be receptive towards national initiatives) as well as from within the set of potential adopters (referring to the norm to excel in relation to one\u2019s peers). Adopters are associated with a higher perceived persuasive pressure than non-adopters. A notable exception here is legislation as a source of pressure. Legislation here refers to regulations that enable the issuance of integrated permits based on a range of acts (such as the WABO: an act on general environmental issues). Although this regulation does not force personalised service delivery, municipalities reported that they expected it to be easier to comply with future audits and assessments if they adopt personalised electronic services. As one council member explained:  \n\u201cThe WABO describes how we should comply and what kind of checks, audits \nand assessments are in place. In the end, there is less space to manoeuvre\u201d (Councillor).   Implementing technology in anticipation of future audits has also been reported by Svensson (2002) in a study on the adoption of legal expert systems in municipal social security systems. In our study, pressure from legislation was also reported, although non-adopters reported this more frequently than \n 9 \nadopters. Legislative pressure therefore cannot explain the adoption of personalised e-government.   --- Table 3 somewhere here  \nOrganisational search: scanning for knowledge, experiences and courses of action Respondents from the municipalities reported that perceptions of persuasive pressure were followed by organisational searching and scanning activities, through which municipalities attempted to seek, identify and choose relevant knowledge, experiences and courses of action (see also Levinthal and March, 1982; Tidd et al., 2009). One respondent clearly illustrated how municipalities react to pressure:  \n\u201cOne member of our support staff made an inventory of associations that staff \nmembers are participating in, and she managed to compile a list of three or \nfour pages\u2026\u201d (Manager of service provision)  Typical targets for search and scan operations are forums and national outreach programmes, companies and financial institutions, other municipalities generally, municipalities with which the municipality has an alliance, and within the organisation itself (see Table 4).   --- Table 4 somewhere here  Respondents reported that pressure did not directly result in new connections with other organisations, but rather that it resulted in more intensive contact with forums and associations (for instance, the Public Service Provision Managers\u2019 Association, the Association of Dutch Municipalities, and also outreach programmes such as GovUnited and DIMPACT) with which one was already involved. As one public manager reported:  \n\u201cWe meet each other at meetings of municipalities with 100 000+ residents. \nOne talks to others, exchanges experiences, we report our practices, and listen \nto other ideas \u2026 in this way, we converge on similar solutions as we all offer \nsimilar services\u201d (Manager Population Affairs and Taxes)   One respondent explained how one\u2019s own organisation could serve as a source of relevant knowledge (see also Isabella, 1990):  \n\u201cIt happened partly due to one of our developers (\u2026) She said she had  \nknowledge of how personalised counters could be implemented but, until \nthen, she had no time to work on them. She said it could be realised given that \nwe already owned the necessary software packages - and she could develop it \nfurther. So we felt we were quite lucky to have such a developer who was able \nto build these facilities\u2026\u201d (Programme manager)  To summarise, in all the selected municipalities we witnessed how pressure was followed by organisational search activities, both in the external environment, as well as in existing alliances and within one\u2019s own organisation for solutions and \n 10 \ninspiration that might already be available. It can also be concluded that both adopters and non-adopters seek and scan for relevant knowledge in national forums, outreach initiatives and companies, but that adopters more frequently report seeking out ideas, inspiration and solutions in alliances, other (often similar) municipalities and their own organisation than non-adopters do.   \nFraming: translating pressure into local priorities and opportunities According to Sahlin and Wedlin (2008; see also Silva and Hirscheim, 2007), and in line with the Scandinavian Institutionalism mentioned in the discussion on theoretical antecedents of diffusion, knowledge and ideas cannot simply be transfused from one organisation to the other: rather, ideas, concepts and knowledge has to be repackaged and re-embedded (Isabella, 1990; MacDonald, 1995; Sahlin and Wedlin, 2008). In our field study, respondents explained that comparable ideas and \u2018chunks\u2019 of knowledge on personalisation were framed completely differently in various adopting organisations. Personalisation was sometimes framed in terms of: \n\u2022 A precursor to an organisation becoming a service champion (enabling genuine citizen-centric service delivery);  \n\u2022  A means for achieving efficiency (\u201cIf the processes are well-organised, I am \nconvinced that in the long run we can do without large numbers of staff\u201d , Councillor); \n\u2022 Boosting reputation (\u201cWe think that we, being part of a high technology \nregion, are obliged to modernise our service delivery\u201d, Head of Customer Relations Department); and \n\u2022  Exerting control (\u201cNow the focus is on the front office \u2026 but in the near \nfuture we intend to reengineer processes in the back office as well, so as to \nsimplify and speed up processes\u2026\u201d, Project Manager Service Delivery).   Table 5 summarises these assessments.  --- Table 5 somewhere here Overall, we conclude that persuasive pressures are actively framed by stakeholders so as to appeal to local priorities and ambitions, which are found to be enhancing citizen orientation, organisational efficiency, reputation and management & control. Adopters differ from non-adopters in terms of the first three framings (being a service champion, striving for efficiency and reputation enhancement), but not in terms of the fourth (management and control).  \nActivation triggers: enabling episodic changes From the transcripts of the various interviews, we distilled concepts that did not directly affect adoption but that, nevertheless, can be interpreted as being important in explaining adoption. Activation triggers of various kinds (coined discrepant events by Tyre and Orlikowksi (1994); see also Kim (1998); Zahra and George (2002); Tidd et al. (2009)) were reported by respondents as moderating the impact of persuasive pressures. Van Waarden and Oosterwijk (2006) see activation triggers as precursors of episodic as opposed to more linear forms of diffusion. In our study, we came across general shocks, staff changes and organisational mergers as activation triggers (see Table 6). \n 11 \n --- Table 6 somewhere here    One clear example of an activation trigger was a firework factory exploding in the municipality of Enschede. This triggered a political crisis and, in the subsequent reorganisation, personalised service delivery was seen as an opportunity to help shape the new organisation. Respondents also mentioned new members of staff. As one respondent explained:  \n\u201cWith new members of staff, new ideas and new energy entered our \norganisation (\u2026). A new city manager, ICT developers, new departmental \nmanagers: they all managed to get up to speed with evolving developments\u201d (Head of Service Delivery)   The merging of municipalities was also reported as a trigger event that helped boost ongoing developments. The idea of service delivery in general, and the more advanced forms of personalised e-government in particular, were embraced by some stakeholders as helping to shape the identity of a new municipal organisation and to get away from the pre-existing local identities of the former smaller municipalities. The above activation triggers were reported more often by respondents from adopting councils than by respondents from non-adopters (see Table 7).   \nSocial integration mechanisms: translating and framing by actors From fieldwork observations, and informed by our theoretical discussion of Scandinavian Institutionalism, we could see that translation, transfusion and repackaging of knowledge and ideas does not take place in a vacuum, but is a social integration process in which specific actors play a role (Czarniawska and Sevon, 2005; see also Pawlowski and Robey, 2004). Szulanski (1996, p. 29) states that \u201cas time passes, a shared history of jointly utilizing the transferred knowledge is built up in the recipient, actions and actors become typified, and types of actions are associated with types of actors\u201d.   The actual framing, transformation and transfusion of ideas and knowledge, that occurs between the processes of organisational searching and scanning and the framing of ideas, involves various social integration processes. An example was given by one respondent who commented:   \u201c\u2026 John Doe, of Consulting Inc5, he is a remarkable character. He has access \nto senior management levels, where normally no-one understands the \npotential of modern ICTs. However, he is able to come up with brilliant \napplications, stories and examples\u2026\u201d (Programme manager)                                                          5 The names of the respondent and the consultancy firm have been changed to maintain anonymity.  \n 12 \nIn our study, we identified three types of social integration: activities by in-house innovation champions, activities involving knowledge brokers and staff exchanges (see Table 7).   --- Table 7 somewehere here   The importance of innovation champions can be illustrated through a quote by a departmental manager who described the activities of a specific stakeholder as follows:   \n\u201c\u2026 I think it all depends on the person. The former manager of Employment, \nWelfare and Income, she really was a citizen- and service-oriented person. \nShe really put quite some energy into the whole idea of service delivery. She \nasked me to team up with her, and to manage a service delivery programme. \nThen, a whole bunch of enthusiastic people joined us\u2026 this group of people, \nwho were all actively participating in the Super Pilots, continued to grow\u2026\u201d (Departmental Service Operations Manager)  Another respondent similarly mentioned a specific innovation champion:   \n\u201c\u2026the idea of personalisation, that was more or less Erik\u2019s business. Our \ncouncillor was never a digital enthusiast, and despite limited resources we \nmanaged to make things happen. And that was due to the persistence of Erik, \nwho kept on fighting, although the councillor said that he\u2019d never succeed\u2026\u201d (ICT policy advisor).   Staff exchange was also mentioned as a moderator of adoption. One respondent explained:   \n\u201c\u2026 We moved on when a member of Enschede\u2019s staff temporarily joined our \norganisation. Due to personal circumstances, he was able to stay for a while \nin the Province of Noord-Brabant and we were able to make use of his \nexperience. That helped quite a lot\u201d (Departmental Manager Service Operations).   It is especially through these kinds of integration mechanisms that ideas and knowledge are actually transfused, translated and framed rather than merely being imitated (Howell and Shea, 2001; see also Bressant and Rush, 1995; Rice and Rogers, 1980; Sahlin and Wedlin, 2008; Silva and Hirscheim, 2007).  Concluding, we infer that genuine framing, transfusion and translation of knowledge, experiences and ideas takes place through social integration mechanisms, activities of internal innovation champions and external knowledge brokers, and through staff exchange. These activities were more frequently mentioned by adopters than by non-adopters.   \nSYNTHESIS: CONJECTURES AND EXPLANATORY MODEL The findings from this study provide a detailed account of how persuasive pressure is channelled within organisations, and of how various processes (searching and seeking, framing) and moderators (external shocks, social \n 13 \nintegration) are related to the adoption of personalised electronic service delivery.   Processes and moderators, as identified in and distilled from the presented empirical data, can be summarised in a number of conjectures (King et al., 1994) and presented graphically in a model that displays the processes of persuasion at work.   Our first conjecture relates to the concept of environmental pressure. In all our interviews, respondents reported perceiving persuasive pressure from their respective environments (perceived expectations of citizens, legislation, \u2018naming and shaming\u2019 pressure from benchmarks, general considerations relating to the reputation of municipalities, effects of national programmes and peer rivalry) in forms that promote electronic service delivery.  However, this perceived pressure was more prevalent in adopters than in non-adopters, leading to the following conjectures:  Conjecture 1A: Municipalities experience environmental pressures that persuade them to adopt personalised electronic service delivery.   Conjecture 1B: Perceptions of persuasive pressure are reported more frequently by adopters than by non-adopters.   In general, persuasive pressure was followed up by organisational searching and scanning activities, through which stakeholders in municipalities attempt to gather knowledge, inspirations and suggestions for courses of action both from the organisational environment (forums and outreach activities, companies, other municipalities in general and allied municipalities in particular) as well as from within their own organisation. In all cases, respondents reported scanning and searching activities, albeit more frequently by adopters than by non-adopters. This leads to our second pair of conjectures:  Conjecture 2A: Municipalities respond to persuasive external pressure by searching and scanning for knowledge, inspiration and suggested courses of action related to personalised electronic service delivery.   Conjecture 2B: Searching and scanning activities are more prevalent in adopters than in non-adopters.   In the various selected municipalities, it could be observed that persuasive pressure and the knowledge and ideas sought were actively \u2018framed\u2019 as compelling arguments for adopting personalised e-government services. Active framing was reported more frequently in adopting organisations than in non-adopting organisations; leading to our third pair of conjectures:   Conjecture 3A: In municipalities, persuasive pressure and sought knowledge are actively framed in order to legitimise the adoption of personalised e-government.   \n 14 \nConjecture 3B: Framing occurs more frequently in adopters than in non-adopters.   With these three pairs of conjectures, the processes underlying the adoption of personalised e-government have been identified. In an inductive fashion, moderating influences could also be observed. The first moderating influence relates to external trigger events. Such trigger events are assumed to moderate the influence of persuasive external pressure on search activities by organisations. This leads to a fourth pair of conjectures:  Conjecture 4A: External trigger events moderate the impact of perceived persuasive pressure on organisational searching and scanning activities.  Conjecture 4B: External trigger events are reported more frequently by adopters than by non-adopters.    A second moderating influence stems from the activities of internal innovation champions, external knowledge brokers and exchanges of staff members:   Conjecture 5A: Knowledge gathered on personalised e-government services is spread and organisationally embedded through the activities of innovation champions, knowledge brokers and temporarily assigned staff members.   Conjecture 5B: Activities aimed at integrating the knowledge gathered within municipalities are more prevalent in adopters than in non-adopters.    All these conjectures can be synthesised into a conceptual explanatory framework (see Figure 1) which visualises how persuasive pressure is followed by specific organisational processes. The format of the framework is similar to Zahra and George\u2019s (2002) absorptive capacity (ACAP) model that generally explains organisational innovativeness. However, much more than in Zahra and George\u2019s ACAP model, the model displayed in Figure 1 acknowledges the \u2018push\u2019-influence of institutions through channels of persuasion, rather than \u2018pull\u2019-absorptive capacities (initiated by adopting organisations) emphasised by Zahra and George. Furthermore, the rendering of processes is based on a perspective that conceives of social structure and human agency as interdependent constructs (Orlikowksi and Barley, 2001) through which processes gradually enable and constrain actors in pursuing specific lines of action.   --- Figure 1 somewhere here  The line of reasoning depicted in Figure 1 combines and extends a number of previous findings in specific ways. First of all it extends and combines King et al\u2019s (1994) work on institutional influences on ICT adoption and Rogers\u2019 concept of channels of communication (see also Venkatraman et al., 1994; Leonard-Barton and Rogers, 1981; Moon and Bretschneider, 1997; Bobrowski and Bretschneider, 1994) by stipulating channels of persuasion that confront municipalities. Second, it provides an understanding of how municipalities react to persuasive pressures (i.e., by organisational searching and framing of ideas and knowledge, see \n 15 \nCzarniawska and Sevon (2005)). Further, by following Orlikowski and Barley (2001), this study engages with both structure as well as agency in explaining the adoption of technologies in organisations.   \nDISCUSSION AND CONCLUSIONS This paper has explored the process through which public organisations \u2013 or rather Dutch municipalities \u2013 adopt personalised e-government services. In so doing, it builds upon an institutional tradition of technology diffusion in which technology diffusion and adoption are associated with an organisation\u2019s attempts to cope with a range of prevailing norms, values, belief systems and rules that are imposed upon them, rather than primarily with individual rational cost/benefit considerations. Furthermore, an objective has been to highlight the role of human agency in the process of innovation, rather than focus on explaining the outcomes as such.   Our analysis has concentrated on persuasive environmental pressure and the ways in which knowledge and ideas regarding innovations are dealt with in municipal organisations. These concepts and their interrelationships have been conceptualised in the form of conjectures and, from these, a conceptual model of institutional influence on the adoption of personalised e-government services has been developed. Overall, the conjectures and our distilled conceptual framework show how \u2013 in this case, municipal \u2013 organisations are confronted with, and then deal with, innovations that have been strongly advocated, for example by the OECD (2009).   One notable conclusion is that municipalities are confronted with channels of persuasion that are both vertical (stemming from beyond the set of potential adopters) and horizontal (related to reputation and rivalry considerations that stem from within the set of potential adopters). The reactions prompted by this persuasive \u2018control\u2019 are primarily horizontally oriented: knowledge, ideas and solutions are sought from contacts within other municipalities in general, existing alliances with other municipalities in particular, and in one\u2019s own organisation.     A second conclusion is that, as Scandinavian Institutionalism  argues, the \u2018idea\u2019 of personalised e-government is indeed actively framed, translated and transfused throughout the process of adoption by the activities of knowledge brokers, internal service champions and temporary members of staff in social integration processes. Clearly, persuasive institutional pressure and the framing of ideas and knowledge are far more important than is often acknowledged in the innovation and e-government literature streams.   These findings raise a number of questions for further research.   First, the direction and source of institutional pressures (horizontal, vertical or mixed) may depend on the differences among centralised, decentralised or decentralised unitary state regimes. Comparative research is needed to reveal differences and similarities in this respect.   \n 16 \nSecond, the resource based view on the firm (Winter, 1987; Zahra and George, 2002) suggests that the so-called appropriability regime (the extent to which organisations are risk averse, or willing to accept ambiguities inherent in framing) moderates the relationship between framing and eventual adoption of innovations. More generally, in an era of administrative change and transformation, it is useful to remember that organisations, and the actors within them, display complex responses in attempting to promote substantive results (service delivery) and legitimacy (reputation).\n 17 \nREFERENCES \n \n Andersen, K. V., Henriksen, H. Z. (2006). E-Government Maturity Models: Extension of the Layne and Lee Model. Government Information Quarterly, 23(2), pp. 236-248. Ashworth, R., Boyne, G., & Delbridge, R. (2009). Escape from the iron cage? organizational change and isomorphic pressures in the public sector. Journal of \nPublic Administration Research and Theory, 19(1), 165-187. doi:10.1093/jopart/mum038  Barney, J. (1991), Firm Resources and Sustained Competitive Advantage, Journal \nof Management, 17, 99-120 Bekkers, V. J. J. M., & Homburg, V. M. F. (Eds.). (2005). The information ecology of \nE-government (E-government as institutional and technological innovation in \npublic administration) (2nd ed.). Amsterdam: IOS Press.  Beynon-Davies, P., & Martin, S. (2004). Electronic local government and the modernization agenda: progress and prospects for public service improvement. \nLocal Government Studies. 30(2). Pp. 41-62.  Bobrowski, P., & Bretschneider, S. (1994). Internal and external interorganizational relationships and their impact on the adoption of new technology: An exploratory study. Technological Forecasting and Social Change, \n46(3), 197-211.  Bessant, J., & Rush, H. (1995). Building bridges for innovation - the role of consultants in technology-transfer. Research Policy, 24(1), 97-114. doi:10.1016/0048-7333(93)00751-E  Caldwell, R. (2003). Models of change agency: A fourfold classification. British \nJournal of Management, 14(2), 131-142.  Czarniawska, B., & Sevon, B. (2005). Global ideas: How ideas, objects and practices \ntravel in the global economy. Copenhagen: Copenhagen Business School Press.  Dos Santos, B., & K. Peffers. (1998). Competitor and vendor influence on the adoption of innovative applications in electronic commerce. Information & \nManagement 34(3). 175-184.  DiMaggio, P. J., & Powell, W. W. (1983). The iron cage revisited: Institutional isomorphism and collective rationality in organizational fields. American \nSociological Review, 48(2), 147-160.  Esping-Andersen, G. (1990). The Three Worlds of Welfare Capitalism. Cambridge: Polity Press & Princeton: Princeton University Press, 1990. Frumkin, P., & Galaskiewicz, J. (2004). Institutional isomorphism and public sector organizations. Journal of Public Administration Research and Theory. 14(3). Pp. 283-307.  \n 18 \nGilbert, D., Balestrini, P., & Littleboy, D. (2004). Barriers and benefits in the adoption of e-government. International Journal of Public Sector Management, 17(4), pp. 286-301.  Ho, A. T. (2002). Reinventing local governments and the E-government initiative. \nPublic Administration Review, 62(4), pp. 434-444.  Holden, S. H., Norris, D. F., & Fletcher, P. D. (2003). Electronic government at the local level. Public Performance & Management Review, 26(4), pp. 325-344.  Homburg, V. M. F. (2008). Understanding E-government: Information systems in \npublic administration. London: Routledge.  Homburg, V. M. F., & A. J. D. Dijkshoorn (2011). Diffusion of Personalised E-Government Services among Dutch Municipalities (An Empirical Investigation and Explanation). International Journal of E-Government Research. 7(3). 21-37.  Homburg, V. M. F., & Georgiadou, Y. (2009). A tale of two trajectories: How spatial data infrastructures travel in time and space. Information Society, 25(5), pp. 303-314.  Howell, J., & Shea, C. (2001). Individual differences, environmental scanning, innovation framing, and champion behavior: Key predictors of project performance. Journal of Product Innovation Management, 18(1), 15-27. doi:10.1016/S0737-6782(00)00067-9  Horst, M., Kuttschreuter, M., & Gutteling, J. M. (2007). Perceived usefulness, personal experiences, risk perception and trust as determinants of adoption of e-government services in the Netherlands. Computers in Human Behavior, 23(4), pp. 1838-1852.  Isabella, L. (1990). Evolving interpretations as a change unfolds - how managers construe key organizational events. Academy of Management Journal, 33(1), 7-41. doi:10.2307/256350  Kim, L. (1998). Crisis construction and organizational learning: Capability building in catching-up at hyundai motor. Organization Science, 9(4), 506-521.  Kimberley, J.R. (1979). Issues in the Creation of Organizations: Initiation, Innocation and Institutionalization. Academy of Management Journal. 22. 437-457.  King, J., Gurbaxani, V., Kraemer, K., McFarlan, F., Raman, K., & Yap, C. (1994). Institutional factors in information technology innovation. Information Systems \nResearch, 5(2), 139-169. King, S., & Cotterill, S. (2007). Transformational government? The Role of Information Technology in Delivering Citizen-Centric Local Public Services. Local \nGovernment Studies.  33(3). Pp. 333-354.  Lai, K., Wong, C. W. Y., & Cheng, T. C. E. (2006). Institutional isomorphism and the adoption of information technology for supply chain management. Computers in \nIndustry, 57(1), 93-98. doi:10.1016/j.compind.2005.05.002  \n 19 \nLayne, K., & Lee, J. (2001). Developing fully functional E-government: A four stage model. Government Information Quarterly, 18(2), pp. 122-136.  Leonard-Barton, E., & E. Rogers. (1981) Horizontal Diffusion of Innovations: \nAlternative Paradigm to the Classical Model. Sloan School of Management Working Paper 1214, MIT, Cambridge. Levinthal, D. A., & March, J. G., A Model of Adaptive Organizational Search.  \nJournal of Economic Behavior and Organization, 2 (1981) 307-333. Ling, T. (2002). Delivering joined-up government in the UK: dimensions, issues and problems. Public Administration. 80(4). Pp. 615-642.  Lips, A. M. B., Hof, v. d., S., Prins, J. E. J., & Schudelaro, A. A. P. (2004). Issues of online personalisation in commercial and public service delivery. Nijmegen: Wolf Legal.   \nLoh, L. & Venkatraman, N. (1992). Diffusion of information technology outsourcing \ninfluence sources and the Kodak effect. Information Systems Research, 334-358. MacDonald, S. (1995). Learning to change - an information perspective on learning in the organization. Organization Science, 6(5), 557-568. doi:10.1287/orsc.6.5.557  Mahajan, V., & Peterson, R. A. (1985). Models for innovation diffusion. Beverley-Hills CA: SAGE.  McClure, C.R., Sprehe, J., & Schenfelder, K. (2000). Performance Measures for \nFederal Agencies: Final Report. Washington DC: Defence Technical Information Center, the Energy Information Architecture, and the US Government Printing Office.   Moon, M. J. (2002). The evolution of E-government among municipalities: Rhetoric or reality? Public Administration Review, 62(4), pp. 424-433.  Moon, M. J., & Bretschneider, S. (1997). Can state government actions affect innovation and its diffusion?: An extended communication model and empirical test. Technological Forecasting and Social Change, 54(1), 57-77.  Moon, M. J., & Norris, D. F. (2005). Does managerial orientation matter? The adoption of reinventing government and e-government at the municipal level. \nInformation Systems Journal, 15(1), pp. 43-60.  Miles, M. B. & Huberman, A. M. (1994). Qualitative data analysis (2nd ed.). Thousand Oaks, CA: Sage.  Mulgan, G. (2005). Joined-up government: past, present and future. In: Bogdanor, V. (ed), Joined Up Government. Oxford: Oxford University Press. Pp. 175-189.  Norris, D. F., & Moon, M. J. (2005). Advancing E-government at the grassroots: Tortoise or hare? Public Administration Review, 65(1), pp. 64-75.  \n 20 \nOrlikowski, W.J. (2000). Using Technology and Constituting Structures: A Practice Lens for Studying Technology in Organizations, Organization Science, 11, 4, 2000: 404-428. Orlikowski, W.J. & Barley, S. (2001). Technology and Institutions: What Can Research on Information Technology and Research on Organizations Learn from Each Other? MIS Quarterly 25(2). pp. 145-165  OECD. (2009). Rethinking e-government services (user-centred approaches). Paris: OECD. Os, G. S. van (2011). The challenge of coordination: Coordinating integrated electronic service delivery in Denmark and the Netherlands. Information Polity, 16(1), pp 51-61. Patton, M. Q. (2002). Qualitative Research and Evaluation Methods. Thousand oaks, CA: Sage Publications. Pawlowski, S. D., & Robey, D. (2004). Bridging User Organizations: Knowledge Brokering and the Work of Information Technology Professionals. MIS Quarterly. 28(4). Pp. 645-672.  Pettigrew, A.M. (1985). The Awakening Giant. Oxford, Blackwell  Pieterson, W., Ebbers, W., & van Dijk, J. (2007). Personalisation in the public sector: An inventory of organizational and user obstacles towards personalisation of electronic services in the public sector. Government \nInformation Quarterly, 24(1), 148-164.  Pollitt, C., & Bouckaert, G. (2004). Public management reform: A comparative \nanalysis. Oxford: Oxford University Press.  Reddick, C. G. (2004). Empirical models of E-government growth in local governments. E - Service Journal, 3(2), pp. 59-84.  Reddick, C. G. (2009). Factors that explain the perceived effectiveness of E-government: A survey of united states city government information technology directors. International Journal of E-Government Research, 5(2), pp. 1-15.  Rice, R. E., & Rogers, E. M. (1980). Reinvention in the innovation process. \nKnowledge-Creation Diffusion Utilization, 1(4), 499-514.  Rogers, E. (1995). Diffusion of innovations (4th ed.). New York: Free Press.  Sahlin, K., & Wedlin, L. (2008). Circulating ideas: Imitation, translation and editing. In R. Greenwood, C. Oliver, K. Sahlin & R. Suddaby (Eds.), Organizational \ninstitutionalism (pp. 218-242). London: Sage.  Silva, L., & Hirschheim, R. (2007). Fighting against windmills: Strategic information systems and organizational deep structures. MIS Quarterly, 31(2), 327-354.  \n 21 \nSvensson, J. (2002). Legal expert systems in social administration: From fearing computers to fearing accountants. Information Polity. 7(2,3). Pp. 143-154.  Szulanski, G. (1996). Exploring internal stickiness: Impediments to the transfer of best practice within the firm. Strategic Management Journal, 17, 27-43. Tidd, J., Bessant, J., & Pavitt, K. (2009). Managing innovation. Chichester: John Wiley & Sons Ltd.  Tolbert, P. S., & Zucker, L.G. (1996). The Institutionalization of Institutional Theory. In S. R. Clegg, C. Hardy y Walter R. Nord (Eds.), Handbook of Organization \nStudies (pp. 175-190). Thousand Oaks, CA: Sage.  Tyre, M. J., & Orlikowski, W. J. (1994). Windows of opportunity - temporal patterns of technological adaptation in organizations. Organization Science, 5(1), 98-118.  Venkatesh, V., Morris, M. G., Gordon B. Davis, & Davis, F. D. (2003). User acceptance of information technology: Toward a unified view. MIS Quarterly, 27(3), pp. 425-478.  Venkatraman, N., Lawrence Loh & Jeongsuk Koh (1994). The Adoption of Corporate Governance Mechanisms: A Test of Competing Diffusion Models. \nManagement Science. pp. 496-507  Van Waarden, F., & Oosterwijk, H. (2006). Turning tracks? Path dependency, technological paradigm shifts, and organizational and institutional change. In J. T. Hage, & M. Meeus (Eds.), Innovation, science, and institutional change. Oxford: Oxford University Press.  Wang, H., & Doong, H. (2010). Does government effort or citizen word-of-mouth determine e-government service diffusion? Behaviour & Information Technology, \n29(4), 1-15.  Winter, S. G. (1987). Knowledge and competence as strategic assets. In D. Teece (Ed.), The competitive challenge - strategies for industrial innovation and renewal. Cambridge, MA: Ballinger.  Zahra, S. A., & George, G. (2002). Absorptive capacity: A review, reconceptualization, and extension. The Academy of Management Review, 27(2), 185-203.  Zorn, T. E., Flanagin, A. J., & Shoham, M. D. (2011). Institutional and Noninstitutional Influences on Information and Communication Technology Adoption and Use Among Nonprofit Organizations. Human Communication \nResearch. 37(1). Pp. 1-33.  \n 22 \nAppendix A: municipalities selected for field study \n \nSelected municipalities Number of inhabitants Personalisation adopted  Personalisation not adopted 200,000+ Eindhoven Tilburg 150,000-200,000 Enschede  Nijmegen 100,000-150,000 Haarlem Amersfoort 50,000-100,000 Capelle aan den IJssel Lelystad 0-50,000 Moerdijk Lemsterland  \n 23 \nAppendix B: topic list for qualitative interviews \n \nIntroduction of research and researcher  \n \nBackground of the respondent - Background (work experience), function, role in organisation - Work environment: involvement in public services, e-government services, organisational structure, project structure, organisational environment \nOrigins of e-government and public services strategies, contacts regarding these \nissues with other municipalities, public service providers and other organisations  - Origin of ideas regarding service delivery, e-government - Relevant internal and external stakeholders \nOrganisational search, learning, development - Sources of knowledge, activities, contacts \nInformation strategy  - Implicit or explicit, rules, policies, key stakeholders   \n 24 \nTables and figures, to be included in the text \n  Table 1  \nDeterminants of e-\ngovernment adoption \nAuthor(s) City size Moon (2002); Reddick (2004); Moon and Norris (2005); Norris and Moon (2005); Homburg and Dijkshoorn  (2011) Citizen demand (perceived usefulness)  Holden et al. (2003); Reddick (2004); Gilbert et al. (2004); Horst et al. (2007) Organisational structure  Moon (2002); Reddick (2004); Holden et al. (2003);  Geographic location Holden et al. (2003); Reddick (2004); Norris and Moon (2005) Managerial, financial and technological capacity  Reddick (2004, 2009); Moon and Norris (2005) \nTable 1: literature review of explanatory e-government studies  Table 2 \n 2006 \n(n=458) \n2007 \n(n=443) \n2008 \n(n=443) \n2009 \n(n=441) DigiD authentication 20.7% 56.7% 76.3% 88.2% Personalised newsletter 16.4% 21.2% 21.2% N/A Tracking and tracing 10.0% 16.0% 28.2% 26.5% Payment 15.9% 42.4% 61.4% 80.0% Pre-completed forms N/A N/A 17.8% 19.1% Personalised counters (MyMunicipality.nl) 5.2% 14.2% 23.7% 28.8% Personalised policy consequences N/A N/A 19.4% 18.7% \nTable 2: prevalence of personalisation attributes in Dutch municipal e-government \nservices   \n 25 \nTable 3 \nPerceived pressures Adopters Non-adopters Citizen expectations 56 36 Legislation 40 50 Benchmarks  57 26 National initiatives 52 24 Reputation 54 17 Other municipalities  16 7 \nTable 3: prevalence of \u2018pressure\u2019 on adoption decisions  Table 4 \n \nOrganisational search Adopters  Non-adopters Forums and outreach programmes 59 48 Companies 49 51 Other municipalities 60 27 Municipal alliances 29 7 Own organisation 22 9 \nTable 4: organisational search    \n 26 \nTable 5  \nFraming in terms of: Adopters  Non-adopters Service delivery, being a \u2018service champion\u2019 84 40 Efficiency 46 21 Reputation  24 16 Management and control 13 18 \nTable 5: framing of personalised e-government services  Table 6  \nActivation triggers Adopters  Non-adopters External shocks (accidents, rehousing) 17 4 Staff changes 8 5 Organisational merger 5 2 \nTable 6: activation triggers   Table 7 \nSocial integration \nmechanisms \nAdopters  Non-adopters \nActivities of innovation champions (internal) 43 11 Activities of knowledge brokers (external experts) 26 14 Staff exchange 13 6 \nTable 7: social integration   \n 27 \nFigure 1  \n \nEnvironmental \npressure\n(through\ncommunication\nand persuasion)\n- citizen expectations\n- benchmarks\n- national programmes\n- reputation\n- peer rivalry\nSearch\n- forums\n- alliances\n- other \nmunicipalities\n- organisation \nitself\nActivation\nTriggers\n- external shocks\n- staff change\n- merger\nSocial \nIntegration\n- innovation\nchampion\n- knowledge brokers\n- staff exchange\nFraming\n- service \nchampion\n- efficiency\n- image\nAdoption\n \n \nFigure 1: model of institutional influence on adoption of personalised e-\ngovernment  \n", "identifiers": ["oai:repub.eur.nl:39183", null], "relations": [], "repositories": [{"id": "554", "openDoarId": 0, "name": "Erasmus University Digital Repository", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "baseId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 1, "metadataAdded": 1387159910000, "metadataUpdated": 1589337608000, "depositedDate": 1520553600000}, "subjects": ["info:eu-repo/semantics/article"], "title": "Diffusion of personalised services among Dutch municipalities: evolving channels of persuasion", "topics": [], "types": [], "year": 2013, "fulltextIdentifier": "https://core.ac.uk/download/pdf/18509190.pdf", "doi": "10.1080/03003930.2013.795892", "oai": "oai:repub.eur.nl:39183", "downloadUrl": "https://core.ac.uk/download/pdf/18509190.pdf"}}
{"status": "Not found", "data": null}
{"status": "OK", "data": {"id": "81867052", "authors": [], "citations": [], "contributors": [], "datePublished": "2009-12-01", "fullText": "REVIEW\nInformation of the chassis and information of the program\nin synthetic cells\nAntoine Danchin\nReceived: 10 May 2009 / Revised: 8 July 2009 / Accepted: 27 July 2009\n\u00a9 The Author(s) 2009. This article is published with open access at Springerlink.com\nAbstract Synthetic biology aims at reconstructing life to\nput to the test the limits of our understanding. It is based on\npremises similar to those which permitted invention of\ncomputers, where a machine, which reproduces over time,\nruns a program, which replicates. The underlying heuristics\nexplored here is that an authentic category of reality,\ninformation, must be coupled with the standard categories,\nmatter, energy, space and time to account for what life is.\nThe use of this still elusive category permits us to interact\nwith reality via construction of self-consistent models\nproducing predictions which can be instantiated into\nexperiments. While the present theory of information has\nmuch to say about the program, with the creative properties\nof recursivity at its heart, we almost entirely lack a theory\nof the information supporting the machine. We suggest that\nthe program of life codes for processes meant to trap\ninformation which comes from the context provided by the\nenvironment of the machine.\nKeywords Algorithmic complexity \u00b7 Logical depth \u00b7\nPhysical entropy \u00b7 Self-organisation \u00b7 Field \u00b7 Context\nAbbreviations\n3D Three-dimensional\nOS Operating system\nReconstructing life is central to synthetic biology\u2019s efforts,\nas a means to try and understand what life is. I explore the\nconsequences of the model of the cell-as-a-computer,\nwhere the \u201cchassis\u201d is explicitly separated from the pro-\ngram, as in a computer (Danchin 2009a). As a heuristics,\ninformation is viewed here as an authentic category of\nreality. I organise in what follows a tentative philosophical\nreflection on constraints met by synthetic biology around\nfour themes which I see as a true revolution of human\nthinking, the shift from a mechanistic view of the world to\nan algorithmic view, with the result that living organisms\ncan be understood as information traps.\nThe modern reflection on information began with\nHilbert\u2019s problems at the beginning of the XXth century.\nOne of his questions was whether arithmetics, the mathe-\nmatics of whole numbers, was simply a tautology, i.e. its\nconclusions could be automatically drawn and reached\nfrom its premises. The response brought about by Go\u00a8del\nand successors after 1931 recognised that arithmetic was\nincomplete, in that it could bring about conclusions which\nwere understandable only when taking a point of view from\nthe outside of arithmetics. Arithmetics incompleteness\nestablishes that whole numbers theory must be separated\nfrom its meaning for the human creator and observer.\nBriefly, arithmetics is associated to two levels of infor-\nmation: the self-sufficient information carried by strings of\nsymbols, and the information carried by the context: lan-\nguage and civilisation, or more generally, by the biological\nentities we name Homo sapiens. The latter provides\ninterpretations of the demonstrations and theorems created\nby the axioms and definitions of number theory but the\ncorresponding information has not yet been theorised.\nBased as is arithmetics on strings of symbols, the\nalphabetic metaphor of the genetic program sits at the\ncentre of several theses which I try to make explicit, via a\nA. Danchin (&)\nCNRS URA2171/Institut Pasteur, 28 rue du Docteur Roux,\nParis Cedex 15 75724, France\ne-mail: antoine.danchin@normalesup.org\nURL: www.normalesup.org/~adanchin/index_en.html\n123\nSyst Synth Biol (2009) 3:125\u2013134\nDOI 10.1007/s11693-009-9036-5\nrapid travel through history with the aim to place the cat-\negory information within synthetic biology: 1. A thousand-\nyear-old metaphysical/ontological thesis, where I discuss\nthe relationships between shape, form and the process of\nin-formation; 2. An epistemological thesis exploring how\ninformation links models of phenomena to reality in a\nsituation identifying two levels of information, the infor-\nmation of the model, and the meaning of the model; 3. The\nexploration of extant theories of information as a pre-\nrequisite to understand the concept of genetic program in\nsynthetic biology; 4. A conjecture proposing the need to\ncreate a new theory, that of information of the chassis\n(machine) or, why the brain is not a computer.\nMetaphysical thesis: shape, form and information\nWhen searching for life outside Earth we look for \u201cunu-\nsual\u201d shapes, not commonly associated with standard\nchemistry and mineralogy. We restrict our identification of\nforms, looking first for the 3D architecture of the \u201cchassis\u201d\nwhich compartmentalises the living entity (see for example\nthe beginning of Monod\u2019s Chance and Necessity (Monod\n1971)). Typically we draw aside crystalline shapes, and\nlook for more complex shapes such as those of spheroidal\nor tubular objects. Yet, we need more to recognise life, as\ndrawing our conclusions from the geometry of shapes can\nbe misleading (this happens when an artefact is interpreted\nas a biological entity\u2019s signature (Clemmer and Beebe\n1991)). Furthermore, the form of living organisms does not\nreduce to their static shapes, it implies dynamic processes.\nFrom the early times of philosophy, life was identified\nas a phenomenon connected to recognisable autonomous\nbut not independent categories. To account for all phe-\nnomena, Aristotle recognised ten categories: \u03bf\u1f50\u03c3\u03af\u03b1,\n\u03c0\u03c1\u03bf\u03c3\u03cc\u03c4\u03b7\u03c2, \u03c0\u03bf\u03b9\u03cc\u03c4\u03b7\u03c2, \u03c0\u03c1\u03cc\u03c2 \u03c4\u03b9, \u03ba\u03b5\u0131\u02dc\u03c3\u03b8\u03b1\u03b9, \u1f14\u03be\u03b9\u03c2, \u03c4\u03cc\u03c0\u03bf\u03c2, \u03c7\u03c1\u03cc\u03bd\u03bf\u03c2,\n\u03c0\u03c1\u03ac\u03c4\u03c4\u03b5\u03b9\u03bd, \u03c0\u03b1\u03b8\u03b5\u0131\u02dc\u03bd. An essential step in understanding reality\nrequired construction of some entanglement of these cate-\ngories, a process which progressively reduced them to four:\nmatter, space, time, and subsequently energy. A remark-\nable achievement was reached when Einstein combined\nthem together in a surprisingly concise equation, E = mc2.\nYet, it was obvious that these universal categories do not\naccount for many phenomena: no one has been able, for\nexample, to derive the crystal lattice of a mineral as simple\nas sodium chloride from the equations of microscopic\nphysics (Grandy 1992). Imagine the challenge for synthetic\nbiology!\nEven understanding what became the modern category\nmatter was never simple: matter displayed itself as an\nimmense variety of entities (shapes, processes, phase\ntransitions and even transmutations\u2026). One could think\nabout substance (\u03bf\u1f50\u03c3\u03af\u03b1, \u1f55\u03bb\u03b7 and the like, with all kinds of\nidiosynchrasies) but something more was required. Ana-\nlysing movement of matter, the Atomists invented a\nprocess permitting matter to take form, to be in-formed, via\nthe interaction of indivisible Parmenidian tiny material\nparticles, the atoms. This process had the excellent prop-\nerty to account for an infinity of forms, but it asked for a\nprocess causing interaction. The core of the Atomists\u2019\nthought was that necessity (\u1f00\u03bd\u03ac\u03bd\u03b3\u03ba\u03b7), associated to some\nconsistency, \u03bb\u03cc\u03b3\u03bf\u03c2, not chance (which is neither a Greek\nnor a scientific concept (Grandy 1992)), was the ultimate\ncause (Danchin 1986). In short, the problem of creation of\nform, in-formation, superimposed on the general problem\nof movement. In addition to phenomenology, understand-\ning in-formation required a process of synthesis. In this\ncontext, synthesis became a direct exploration of the con-\ncept of creation, uncovering a link between in-formation\nand creation. This was understood soon after chemistry was\nborn\u2014chemical synthesis is at the heart of modern chem-\nistry\u2014and it is therefore not unexpected that biology,\nwhere form is apparent everywhere, should develop into\nsynthetic biology.\nAlready asked by the presocratic philosophers the\nquestion of the nature and origin of form was renewed by\nAristotle. After him, the question kept developing with\nGreeks in Alexandria and southern Italy, Arab and scho-\nlastic philosophy from the fall of the roman empire till the\nfourtheenth century. Between Aristotle and the present\ntime, I retain John Scotus Eriugena because of the way he\ntackled the problem of creation (Erige`ne trans. 1995). A\nneo-platonist, Eriugena divided Nature into four species:\n(1) Nature which creates and is not created; (2) Nature\nwhich creates and is created; (3) Nature which does not\ncreate and is created; (4) Nature which neither creates nor is\ncreated. To make a long argument short, asking questions\nthis way leads to fivemodes of opposition, which introduce a\nhierarchy in natural entities, and in particular in living\nbeings. As in the platonistic tradition, the material world of\nour experience is composed of ideas clothed in matter.\nHowever, Eriugena attempted to reconcile Plato with Aris-\ntotle, discussing Aristotle\u2019s ten categories. Time and space\nwere discussed as central for human perception of phe-\nnomena, matter is without form or limit, but it needs an\nexternal agent to take form, it needs to be in-formed. Inter-\nestingly, God, as defined by scriptures, escapes all categories\nexcept one, relatio (\u03c0\u03c1\u03cc\u03c2 \u03c4\u03b9, ad aliquid), which I retain here\nas it lies behindwhat we now name information. This elusive\ncategory remains central today (relationships appears typi-\ncally in the god-like self-organisation (Grandy 1992)).\nThe second name I keep in this sequence is Averroe\u00a8s.\nHis commentaries of the metaphysics of Aristotle had\nimmediate and lasting success. I retain sentences of his\nTahafut al Tahafut: \u00abMatter only becomes in so far as it is\ncombined with form. Everything that comes into being\n126 A. Danchin\n123\ncomes into being from something else, and this must either\ngive rise to an infinite regress and lead directly to infinite\nmatter which is impossible, even if we assume an eternal\nmover, for there is no actual infinite; or the forms must be\ninterchangeable in the ingenerable and incorruptible sub-\nstratum, eternally and in rotation.\u00bb (Averroe\u00a8s trans. 1497).\nThat substratum, substance, was difficult to make explicit.\nFirst split into the four elements, fire, air, water and earth,\nand, subsequently seen as atoms (which, despite their\nname, recently split into further particles), it needed asso-\nciation with something which makes the root of variety in\nthe world, form (\u03b5 \u03b4\u03bf\u03c2).\nHow could form combine with matter? A variety of\nanimas (souls and spirits) were invented to account for the\nbirth, development and conservation of movement, until\nenergy came in. This category permitted some entangle-\nment of matter with space and time, and long took the role\nof the animating principle needed to account for life (see\nmesmerism and its \u201canimal magnetism\u201d and \u201cpositive\nenergy\u201d in small talk or the vocabulary of sects today).\nMany further categories required to account for life were\ndiscussed for centuries in the western world, most often\nbased on the assumption that reality had to fit with explicit\nrevelation by God of its characters, as written in the\nScriptures. Thomas Aquinas used Averroe\u00a8s\u2019 Grand Com-\nmentary of Aristotle\u2019s Metaphysics as his model. In his\nSumma theologica he analysed the questions of Trinity and\nof creation, showing that standard reasoning tells us that\ncreation is related with in-formation, placing relationships\nbetween all kinds of entities (including abstract entities) at\na central position.\nThese analyses may be condensed in the question asked\nby the Pythia in Delphi: \u201cI have a boat made of planks of\nwood. The planks are progressively replaced as they rot\naway. After some time, all have been replaced, none of the\noriginal ones remain: is it the same boat?\u201d (Danchin 2003).\nTo understand what life is, we need to understand the\nrelationships between entities recognised as belonging to\nlife, whether they are material, processes, or abstract, such\nas language. And we need to try and understand the process\nof in-formation (which we would in the modern terms\nname creation of information, noting that in the evolution\nof languages redundancy is a ubiquitous trend (Livingstone\n2003)).\nInformation links models with reality\nBelonging to reality we cannot behave as outsiders con-\ntemplating the world. Understanding information asks us to\ninvestigate the way science is constructed. Presocratic\nGreek philosophers recognised that our limitations in\nunderstanding truth (\u1f00\u03bb\u03ae\u03b8\u03b5\u03b9\u03b1) only allowed us to give our\nviews (\u03b4\u03cc\u03be\u03b1) on reality. \u201cAnd for a certain truth, no man\nhas seen it nor will there ever be a man who knows about\nthe gods and about all the things I mention. For if he\nsucceeds in the end in saying what is completely true, he\nhimself is nevertheless unaware of it; and opinion is fixed\nby fate upon all things\u201d said Xenophanes (Diels and Kranz\n1935). Approaching truth is to place a fragment of reality\nin a particular perspective, where we can understand its\nrelationships with human beings. The central tenet of sci-\nence\u2014often ignored, as many scientists behave as priests\nof a revealed religion when interacting with mass media\u2014\nis that we construct models distinct from reality. We match\nmodels with phenomena, expressing local instances of\nreality in a particular context. Models may display a certain\ndegree of adequacy, if not truth, with reality. The model/\nreality separation is so significant that several concomitant\nmodels may express our knowledge about a particular side\nof reality.\nThe importance of models to understand reality trig-\ngered the creation of axiomatics, mathematics and logics.\nThis effort was well fitted to the Renaissance trend to\nreplace Aristotle by Plato, removing the thought of the\nformer to the \u201cdark centuries\u201d of Middle Ages. At the heart\nof platonistic philosophy, the shadows of mathematical\narchetypes had to be discovered by persons who were\nilluminated by their truth. This attitude placed mathematics\nin the world of idealities, suggesting that mathematical\ncertainties existed separately. The medieval reflection on\nin-formation was soon replaced by a geometrical view of\ncombinatorial creation of forms associated to the general\nstructure of space, initially studied in one, two or three\ndimensions and later generalised to all kinds of dimensions.\nIn parallel, and following a medieval trend of arabic\nmathematics, arithmetics and number theory slowly\nemerged as algebraic equations. Models recognised as of\nthe highest quality were mathematical models, developing\non their own, independently of reality with their in-built\nconsistency (information). Trying to match models with\nreality allowed scientists to progress by producing better\nand better adequation with reality (Danchin 1992; Putnam\n1988). However, the match between models and reality\ncould never be direct (a mathematical model of an aero-\nplane does not fly). It rested on interpretations (processes\nrooted in culture and language, thus associated to a prop-\nerty that we might name context and linked to a research\nprogramme (Lakatos 1976, 1980)).\nIf constructing models while confronting them to reality\ndefines science (Popper 1959), then the effort to establish\nan explicit demarcation between science and non-science is\ndominated by a particular category of reality, information\nagain, using the word with all its fuzzy connotations\n(Popper 1963). Defining what science is emphasises two\ntypes of information, information of the (mathematical)\nInformation of the chassis and information 127\n123\nmodel and information of the context. Both types place the\nold category, relatio, at their heart, but only the former has\nyet been theorised, in the chaining of axioms and defini-\ntions, demonstrations and theorems (Danchin 2003).\nThe way synthetic biology is developing illuminates\nthese points. Starting from preconceived biological views,\nit abstracts specific features into axioms and definitions,\nand builds up models, whether mathematical or experi-\nmental (e.g. engineering models) (Endy 2005). The models\nunfold with their own rules of consistency: a demonstration\nin mathematics, yielding a theorem, a computer output in a\nsimulation, a genetically modified cell in an experiment\u2026\nSubsequently one goes back to reality by proposing a\nconcrete instantiation of the output, predicting a particular\nphenomenon. This prediction is of two major types: Either\nthe prediction of a novel, previously unknown or unrec-\nognised entity (a structure, a process, a metabolite\u2026), or\nthat of a particular behaviour of reality, which should\nmanifest itself along lines predicted by the model (Fig. 1).\nA model is (temporarily!) valid when all its predictions are\nrecognised in actualisations of reality. Typically, in syn-\nthetic biology bacteria have been constructed which\ndisplay, as expected, some type of multistable behaviour or\noscillations (Elowitz and Leibler 2000) or phages with\nartificial regulatory regions have been shown to display the\nability to grow on cells (Chan et al. 2005). In neurosciences\nthe basis of neuromimetic networks rests on a vast number\nof works where selective processes play a central role\n(Changeux et al. 1973; Edelman 1987).\nHowever, because the model is not reality, this ideal\noutcome never develops for a long time. Even when we\nproduce a new entity not recognised before the model\u2019s\nconstruction\u2014a great success, comes a time when a phe-\nnomenon does not fit the model\u2019s predictions. To proceed\nwith our example: in bacteria, bistability is not stable in\ntime (Veening et al. 2008). Initial attempts to solve the\ncontradictions between model predictions and observed\nphenomena do not immediately discard the model. The\ncommon practice we witness in synthetic biology is re-\ninterpretation of the instantiation process that matched the\nmodel to reality. Typically: \u201cexceptions make the rule\u201d, or\n\u201cthis is not exactly what we meant, we need to focus more\non this or that feature\u201d\u2026This polishing step permits the\ncontext of the model and its associated phenomena to be\ndefined as accurately as possible. It marks the moment\nwhen technically arid efforts such as normalisation,\ndefining a proper nomenclature, a database data schema\nhave a central role. We witness this today in synthetic\nbiology in the standardisation effort of the community\n(Endy 2005). Despite all efforts to reconcile predictions\nand phenomena, the inadequacy between the model and\nreality becomes insoluble. This contradiction implies that\nwe need to reconsider the axioms and definitions upon\nwhich the model has been constructed, triggering a spiral\nof further models, making science as we know it. As\nalways with exploration, this exploratory attitude meets\nresistance: most of our contemporaries would be happy to\nbe believers, and forget about the impossible but necessary\nquest of truth (Danchin 1992). This may explain both the\nhype and the reluctance to accept synthetic biology.\nIn the subsequent inflation of models there is a hierar-\nchy. A mathematical demonstration is perceived as the\nultimate proof (Popper 1963). This justifies the huge\nnumber of mathematical models published in systems and\nsynthetic biology. Do they result in non-trivial predictions?\nI am afraid that, more often than not, most models are\n\u201cretrodictions\u201d, finding what is already well known (how\noften do metabolic models \u201cdiscover\u201d the Krebs cycle?),\nrather than predictions. Indeed, assessing the interpretation\nof postulates which have not been expressed in a precise\nway has deep consequences, including in mathematics,\nwhich illustrates the importance of the category informa-\ntion, connecting it with the standard categories of reality\n(time in particular). Deep features of axiomatics were\nunderstood when we discovered that something taken for\ngranted was overlooked. Zermelo\u2019s axiom of choice (given\nany two sets, one set is in one-to-one correspondence with\nsome subset of the other: this looks trivial, but is not) is a\nfamous example of this situation. Similarly, and in line\nwith the Pythagorean/Platonistic tradition, we accept syn-\nchrony in the way we use mathematics, making it\nindependent of time: when reasoning by recurrence, if we\nshow that something is true for n+1, knowing it is true for\nn, this will be valid, whatever the size of n. We take for\ngranted very large numbers eventhough it will be\nCommon ideas Language\nExistential Refutable\ninterpretation\ninstantiation\nabstraction\ninterpretation\nde\nm\non\nst\nra\ntio\nn\nFig. 1 Schematic representation of the dialogue between models and\nreality. Note that the context is essential in isolating postulates. Also,\nmany models can co-exist, and, beside mathematical models\napproaches using analogies and simulations can behave as models.\nContrary to Karl Popper\u2019s wish there is no clearcut link between\nmodels and reality, precluding universal processes to define the exact\ncontours of Science\n128 A. Danchin\n123\nimpossible to reach them in any realistic time frame. This\nimplies that there is no time involved (no computation) to\naccess them, assuming that the nature of mathematics does\nnot change if n is very large. What would happen if we\nmodified this axiom? Non-standard analysis explores our\nlimitations if we accept that the behaviour of mathematics\nchanges for infinitely small or infinitely large numbers (an\neffort that permitted Leibniz to invent differential equa-\ntions) (Robinson 1996). This is mentioned here as another\nexample of the fact, well established by Ruelle (2000), that\nmathematics do not exist outside reality (Desanti 1968), but\nbelongs to it.\nThe present status of information in synthetic biology\nMost developments of synthetic biology consider the\ngenetic program as an algorithm, implicitly assuming that\nthe cell behaves as a computer, a machine manipulating\ninformation. I will not repeat the argument meant to justify\nthe model of the cell as a Turing Machine (Danchin\n2009a). Suffices it to say that this implies the existence of\ntwo entities, associated via a read/write process. A machine\nis moving a device that carries a support with a linear\nstring of symbols written in a finite alphabet; the data of the\nstring of symbols, read by the machine, triggers its future\nactions. The focal point of representing the atom of life, the\ncell, as a Turing Machine, assumes the physical separation\nbetween machine (\u201cchassis\u201d) and data/program, repre-\nsented by one or several linear strings of symbols. The crux\nof the model is that one should be able to isolate the entity\ncarrying the program, put it back in a recipient host, and\nobserve that the program in its new location displays\nphenomena specific of the information it carries. Beside\nexperiments showing that pieces of program can be han-\ndled by cells (viruses and horizontal gene transfer),\nexperiments produce results consistent with the model: 1.\nAnimal cloning (Wilmut et al. 1997) is now commonplace;\n2. The genome of a Mycoplasma species M. mycoides, was\ntransplanted into another species, M. capricolum, and after\nseveral rounds of reproduction (reproduction of the\nmachine and replication of the program, see below) the\nhost species was replaced by a colony of the donor genome\n(Lartigue et al. 2007). This latter experiment is so impor-\ntant conceptually that it is essential for synthetic biology\nthat it is reproduced in many laboratories. Yet, this might\nnot satisfy us that the model is adequate to represent\nreality. Three main lines of reasoning argue against the\ncell-as-a-computer model.\n1. The first counterargument explores the concept of\nOperating System (OS) (von Neumann 1958). Because\nthe machine is separated from the program, a subset of\nthe program must be devoted to the interaction with\nthe machine and its \u201cusers\u201d (in the most general sense)\n(Danchin 2009a). If a particular routine is meant to\nreproduce the machine, then a subset of the program\nmust be somehow linked to the architecture of the\nmachine. Analysis of the genes giving bacteria their\nshape showed that there is indeed an unexpected\ncoincidence between gene clustering in genomes and\nshape of bacteria (Tamames et al. 2001). In multicel-\nlular organisms, the distribution of control genes, the\nhomeogenes, parallels the body plan: changing the\norder of some homeogenes in the chromosomes\nchanged the shape of the organism, putting organs in\nthe place of others (Gaunt 1991). Rather than an\nobjection, the existence of a correlation between the\norganisation of the program and the architecture of the\norganism fits a prediction of the model.\n2. The second counterargument is that the program is\ncarried by some material structure, bringing about\ncontextual information. However, this is true in\ncomputers as well: the material support of the program\nhas its saying in permitting the machine to run\nproperly. Different machines may be driven by the\nsame program on different supports. Thus, even the\ncloning experiment, which does not involve naked\nDNA but a whole nucleus, with its envelope, its\nproteins and its RNAs, is not different from a material\nsupport of a program in a computer. Indeed, nocturnal\nanimals use chromatin in the nuclei of neurons using\nthe retina in an extraordinary way. Their retina can\ndetect one unique photon. Yet, the photon receptors are\nlocated behind neurons, which absorb or diffuse\nphotons rather than preciously conserve them. When\nlight is dimmed, the chromatin changes transcription\nand reorganises in such a way that its material behaves\nas a lens, focusing photons on receptors located behind\nthe neurons (Solovei et al. 2009)! This novel function\nfor DNA, which has nothing to do with its role in\ncarrying the genetic program, shows that another type\nof information has to be taken into account. In the\nsame way, in many computers the support of the OS\nbelongs the casing part of the chassis.\n3. A third counterargument is that many rules prescribe\nthe organisation of the cell soma, reflecting a large\namount of information unrelated to the information in\nthe program. Quite true, but this is true again for\ncomputers as well. The design of the interfaces, the\nmicroprocessors and the energy supply of the machine\nrequire much information.\nIn summary, two types of information (coupling of a\nparticular form\u2014not simply shape\u2014with matter, energy,\nspace and time), information of the chassis (casing +\nInformation of the chassis and information 129\n123\nmetabolism) and information of the program are associated\ntogether in a cell (Tanaka 1984). A synthetic cell needs the\nassociation of a chassis developing metabolism (not a\nsimple 3D casing) and a program similar to that found in\ncomputers. The conclusions of Dyson\u2019s argument on the\ndouble origin of life, with reproducing metabolism pre-\ndating replication are therefore a pre-requisite for synthesis\nof life (Dyson 1985). This dichotomy is visible in present\nsynthetic biology, with a fairly clear separation between\nthose who study the chassis (and are often also interested in\nthe origin of life) (Kuruma et al. 2009; Shenhav and Lancet\n2004) and those who think that life is essentially due to the\ngenetic program, organising their activity around con-\nstruction of program biobricks, or even as complete\ngenomes (Gibson et al. 2008).\nInformation of the program\nThe study of the genetic program as a text, applying\naccepted rules of the theory of information (Shannon and\nWeaver 1949; Cover and Thomas 1991) to its analysis\n(He\u00b4naut et al. 1996) resulted in the emphasis placed on DNA\nin synthetic biology. Schneider created his famous \u201clogo\u201d\nrepresentation of sequences (Schneider and Stephens 1990)\nin a model of molecular machines based on Shannon\u2019s\ninformation (Schneider 1991a, b). His work was based on\nthe intuition that creation of information was consuming\nenergy (Schneider 1991b). Furthermore, it assumed that the\ndata has no meaning (hence no \u201cvalue\u201d), and could be\ncharacterised purely by analysing the probability of pres-\nence of a given symbol in the sequence, generating its logo\n(Schneider and Stephens 1990). A similar trend is visible in\nthe way information is used in the mass media. It is current\nwriting\u2014because all kinds of signals can be digitised\u2014that\neverything has an information coded in sequences of (0,1),\nrestricting the concept of information to that particular view\nof sequences of symbols, and forgetting about in-formation\n(creation and accumulation of information, or a value\nassociated to an information). The common feature of this\nconceptualisation is dematerialisation: the corresponding\ninformation becomes an abstract entity, which can be\nmanipulated using mathematic tools.\nYet pure abstraction is obviously inaccurate in terms of\nwhat we would like to name information. Messages without\nmeaning (random messages) are without value. \u201cO singe\nfort\u201d in German has a meaning totally different from that in\nFrench (Yockey 1992). Can we see, even within the digi-\ntisation (or binarisation) paradigm, whether we should\ngo further? The soviet school of electronics following\nAndronov, Kolmogorov, and the Americans Chaitin and\nSolomonoff constructed formal models of information vs\nchance by considering sequences of symbols as the result of\nan algorithm. Any sequence of symbols has some algo-\nrithmic complexity: the length of the shortest program\ngenerating the sequence. A repeated sequence of 2n bits\n0101010\u2026is coded by a simple program of the type:\nBEGIN DO [1,n] PRINT 01 RETURN END. For n large,\nthe program is much shorter than the sequence. In contrast,\nif the sequence is random (this is proposed as a definition of\nrandomness), the only way to get the sequence is BEGIN\nPRINT \\sequence[ END, i.e. a program with a length\nsimilar to that of the sequence.\nAlgorithmic complexity has been related to Shannon\u2019s\ninformation (Cover and Thomas 1991) and to physical\nentropy: \u201cAlgorithmic randomness provides a rigorous,\nentropy-like measure of disorder of an individual, micro-\nscopic, definite state of a physical system. It is defined by\nthe size (in binary digits) of the shortest message specify-\ning the microstate uniquely up to the assumed resolution.\nEquivalently, algorithmic randomness can be expressed as\nthe number of bits in the smallest program for a universal\ncomputer that can reproduce the state in question (for\ninstance, by plotting it with the assumed accuracy). In\ncontrast to the traditional definitions of entropy, algorith-\nmic randomness can be used to measure disorder without\nany recourse to probabilities\u201d (Zurek 1989). This success\nled many to think that we had a final Theory of Informa-\ntion, which could tell us what information is.\nHowever, we can point out a first difficulty here. We\nknow of an infinite set of transcendent numbers, such a \u03c0,\nwhose digits are generated by fairly short algorithms while\ntheir succession cannot be predicted. They are therefore of\nlimited algorithmic complexity. Yet, they are much more\ninteresting than repeated sequences with the same com-\nplexity. Knowing the exact value of a digit placed in the\ndigits of \u03c0, very far away in the digitisation might be inter-\nesting, but the only way to reach that value is to actualise the\nprocess of computation. Bennett named logical depth the\ntime needed to reach that value and related it to physical\ncomplexity (Bennett 1988b). This is a first indication that we\nare far from having a thorough theory of information.\nHow do we have access to the information of the genetic\nprogram? Practice of computation is fairly old (well before\nal\u2019Khawarizmi algorithms, Erastothenes\u2019 sieve is a familiar\nexample) but we had to wait for Pascal\u2019s computing\nmachine, and for Lovelace and Babbage Analytical Engine\nto reach today\u2019s situation, with the basic concepts proposed\nby Turing, von Neumann and others, coupling the machine\nand the program, via an OS managing the \u201chousekeeping\u201d\nfunctions of the machine. The functions coded by the\ngenetic program are the result of a very long evolution.\nAnd if we keep the algorithmic metaphor, because DNA\ncomes from DNA comes from DNA\u2026 in an endless rep-\nlication process, the nucleotides in the sequence have\nconsiderable logical depth.\n130 A. Danchin\n123\nAs with computer OSs, the housekeeping program is\nabstract and general, yet its concrete implementation,\nresulting from billions of years of evolution, makes that\nseveral OSs may coexist, revealing again two kinds of\ninformation, information of the program and information of\nthe context in which the program is expressed. This has\nconsiderable consequences for synthetic biology: cellular\nfunctions can be general and ubiquitous, whereas there is\nno reason why they should always be performed by\nstructurally related objects. Overall, living cells display\nsimilar abstract features, and the genetic code argues for\nuniversality. Yet, Woese uncovered a significant discrep-\nancy between two unicellular classes, the Archaea and the\nBacteria (Woese et al. 1978). To identify ubiquitous\nfunctions operated by non-ubiquitous structures one had to\ndevise an operational strategy, based on the concept of\ngene persistence (tendency of a given gene to be present in\na quorum of species) (Fang et al. 2005). Different structural\nentities with common functions in different bacterial clades\nwere indeed characterised (Danchin 2009b; Woese 2002).\nA structure is therefore recruited for a particular function,\ndependent on the context in which it operates. The context\ncreates the function.\nA way forward: the information of the machine/chassis\nEmphasis on the idea of information as meaningless strings\nof symbols (Shannon and Weaver 1949), restricted our\nthought to that very limited feature of reality. In the Turing\nMachine, there is a machine. While its actions are explicit,\nnothing is said about its innards, at least when mathema-\nticians analyse its behaviour. This is no longer so when\nengineers build up computers. The same is true for the\nchassis in synthetic biology. Not only does one need to\nmake a machine that performs the actions of the cell/Tur-\ning Machine but this machine needs to be implemented in\nthe real world (Tanaka 1984). It must be made of explicit\nmatter, its actions need to be energised and there must be\nan Environment/Machine interaction with sensors, trans-\nporters, adhesins, safety valves (Danchin 2009c)\u2026\nThe information of the chassis provides the relevant\ncontext which allows it to read the program and interpret it\ninto actions (including modifying the program). Even sys-\ntems which \u201cself-organise\u201d do not organise by themselves,\nbut do so only when placed in proper context, which drives\norganisation (the DNA double helix does not form in\ndimethylformamide) (Grandy 1992). This type of informa-\ntion has a huge variety of properties: shapes, dynamics and\nfluxes. It displays relationships between components of the\nmachine, and between the machine and the environment. It\nexpresses a situation. It has characteristics which are\nsomewhat similar to those of a field but also of a graph.\nTypically, what we name epigenetics carries over chassis-\ntype information. A great many works dealing with the study\nof the brain (Edelman 1987), or of cognition (Clark 1998;\nRyle 1949) has taken into account this type of information. It\nis also at the root of much work on artificial life, learning and\nmemory where reproduction, rather than replication is the\nexplicit goal (see e.g. Bullock et al. 2008). But there is not\nyet, despite many advances, explicit consistent theories of\nthe corresponding information (Tanaka 1984).\nThat we might code this information after digitisation\ndoes not place it automatically within the realm of under-\nstandable or valuable sequence information, as the very\nprocess of digitisation is only efficient knowing which type\nof Turing Machine would read out the corresponding\nsequence. This can be shown as follows. Algorithmic com-\nplexity was meant to define what chance is, because chance\nis the reference that permits definition of physical informa-\ntion: a random sequence displays the highest complexity\n(Cover and Thomas 1991; Zurek 1989). However this defi-\nnition does not hold, as it is context-dependent (Grandy\n1992). Here is an open conjecture (a preliminary version has\nbeen proposed in a different context by Wolfram (1985) and\nit is interesting to follow the analysis of \u03c0 by Simon Plouffe:\nhttp://www.lacim.uqam.ca/~plouffe/). Take once again a\ntranscendent number like \u03c0. Its digits are pseudo-random: for\nany sequence, there is a place in its digital development\nwhere one can find the sequence, whatever it is (this con-\njecture holds for infinite many real numbers and we would\nneed to have a short algorithm to chose the one where the\nposition of the sequence can be readily identified). Now, the\ndigits can be generated by an algorithm of length N. Let us\nchoose a putative algorithmically random sequence of length\nN plus an appropriate constant. The sequence can be gen-\nerated by an algorithm shorter than the sequence, giving the\nalgorithm generating the number and the position of the\nsupposedly random sequence. Hence the sequence is not\nrandom (QED). Note that the value of the information\n(logical depth) is not fixed. It depends on the algorithm, as it\ndiffers in \u03c0 and in any other transcendent algorithmically\ngenerated real number. Said otherwise, the complexity and\ndepth of the sequence depends on the algorithm i.e. on the\ncontext. Provided that we can prove the conjecture, this fits\nexactly the objection raised against the cell-as-a-computer\nmodel: beside information in the program, there must be\ninformation in the machine (providing the context). Hence,\nthe information of the machine is not described by our\npresent theory of information.\nIn the definition of logical depth, we have implicitly used\na property of algorithms, recursivity. In 1931 Go\u00a8del con-\nstructed a recursive algorithm, which, when decoded,\ntranslated into a particular proposition, which, briefly, sta-\nted: \u201cI am impossible to prove\u201d. Moving form one context to\nanother one, recursivity created a novel information, the\nInformation of the chassis and information 131\n123\nstatement of a fact of which the world was previously una-\nware. The genetic code, which enables nucleic acids to be\ntranslated into proteins, which in turn manipulate nucleic\nacids, behaves exactly as Go\u00a8del\u2019s procedure does (Danchin\n2003, 2009a). The consequence of this demonstration is that\na purely deterministic system, with known initial conditions,\nmay have an entirely unpredictable outcome. By contrast\nwith the mechanistic philosophy, even with its more modern\nappendices such as feedback and feedforward loops, recur-\nsivity brings about novelty not because we fail to grasp all\ninitial conditions of a particular phenomenon, but because it\ncan be only understood a posteriori, after it has unfolded in\nspace an time. This implies that synthetic biology, when it\ntakes recursivity into account, develops in a world that is\ntotally irreducible to the world of systems biology, which\nremains an elaborate episode of the study of mechanistic\nautomata. An important consequence is that what we com-\nmonly term the \u201cgenetic program\u201d because it unfolds\nthrough time in a consistent manner is not a programmewith\nan aim (we would not be able to predict any aim) - it is\nmerely there, and functions because it cannot do otherwise.\nWe only perceive a design because the end result is familiar\nto us, and thus seems more \u201cright\u201d than any other possible\nresult (Danchin 2009a).\nIf creation of information depends heavily on the context,\nwe must identify in living organisms functions which permit\nit to accumulate, and relate it to the material world. How are\nparticular structures or processes recruited to this aim? A\nreflection on the coupling between accumulation of infor-\nmation and energy based on work developed by Landauer\nand Bennett, showed that the relationship between energy\nand information is that there exists degradative processes\nwhich \u201cmake room\u201d using energy to prevent degradation of\nwhat is functional ((Bennett 1988a; Landauer 1961),\ndetailed analysis in (Danchin 2009a, d), see also a very\nrecent development (Sagawa and Ueda 2009)). In such a\nsituation it is the context that determines which gene product\nis functional and which is not. The consequence is that, if the\ncontext does not vary too rapidly, then the functions which\nwill be selectively retained are sculpting an image of the\nenvironment within, creating adaptation. This is exactly how\nan information can get a meaning. In terms of synthetic\nbiology, this orients research towards learning and memory,\nrather than towards fixed mechanical engineering.\nIn guise of conclusion: the brain is not a computer,\nyet it manipulates information\nTrying to put vitalism to an end, Claude Bernard placed\nbiology within the realm of physics and chemistry (Bernard\n1865). This led his followers to ask the question: what are\nthe relevant entities (material objects and processes) which\nmake a cell alive? The biochemical inventory stage started\nwell, with the discovery of the ribosomes, of the structure\nof the DNA double helix, of the sequence of the poly-\npeptide chain of insulin, and, rapidly of messenger RNA\n(Judson 1979). Yet, many features of biological entities\nresisted the classical analysis of chemistry and physics.\nThis was apparent in the laws of genetics, where linear\narrangements of the elusive genes was central (Gayon\n2007). Even in biochemistry the shape of molecules posed\nan enigma: La dissyme\u00b4trie, c\u2019est la vie, insisted Pasteur.\nBut the involvement of shape was deeper than usual: the\nvery process of replication placed the concept of form in a\nworld quite different from the simple arrangement of a\nparticular setup in 3D as shape would suggest. Replication\nshifted the idea of a chemical as the substrate of a recog-\nnition process to that, abstract, of a template, in this case\nfor a duplication process doubling the number of the initial\nmolecule. Subsequently, the discovery of transcription,\ntranslation, and associated control and coding processes,\ncontinued to shift emphasis from shape to form in an\nabstract way, commonplace in mathematics.\nInformation\u2014creating and manipulating form\u2014was\nessential to account for life processes. From the world of\nPlato\u2019s archetypes, those who explored the basic concepts\nof life resorted to discussions which began with Aristotle\nand placed form as a central category of reality. For some\ntime, and this is still quite visible in systems biology as\nwell as in synthetic biology, living organisms were seen as\nmechanistic automata, with feedback and feedforward\nloops as paradigmatic entities. The purpose of the present\nreflection was to try and show that investigating the con-\ncept of information shifts eighteenth century\u2019s automata to\nmodern algorithmic machines, capable of authentic crea-\ntion. This implied replacing feedback by recursivity, a\nmuch deeper process. Recursivity, associated to appropri-\nate management of energy (Bennett 1988a; Landauer 1961;\nSagawa and Ueda 2009), creates information (Danchin\n2009a). It does so by identifying two domains where\ninformation must be taken into account: information of a\nprogram and information of a machine. However, while the\ninformation of the program is fairly deeply explored by a\nvast community of investigators, this is not so of the\ninformation of the machine/chassis, which involves some\nkind of measurement of the context (in terms of imple-\nmentation within the four categories, matter, energy, space\nand time) (Tanaka 1984; Sagawa and Ueda 2009).\nIt is perhaps in the functioning of the brain that we can\nmake the latter type of information most prominent.\nIndeed, while von Neumann and others invented computers\nwith mimicking the brain in mind (von Neumann 1958),\nthe brain does not appear to behave as a Turing Machine\n(Edelman 1987). There is no \u201cgost in the machine\u201d (Ryle\n1949). However, nobody would doubt that brain manages\n132 A. Danchin\n123\ninformation, and in a very efficient way (Clark 1998;\nBullock et al. 2008). To my view this is a strong indication\nthat the information we describe when considering mes-\nsages is a tiny part of what information is. Because we use\nlanguage, built on the exchange of sequences of symbols,\nexactly as programs are exchanged in computers, linguists\noften saw the brain as a Turing Machine. But language is\ndeeply associated to meaning: I had in 1974 at a meeting of\nthe Centre Royaumont pour une Science de l\u2019Homme at the\nMIT, a heated argument with Noam Chomsky about other\nfeatures of human languages, such as rhythm (in the west\nafrican language mo\u00a8re\u00b4, a speaker may begin a rythmic\nsentence, which is answered preserving rythmic rules by\nsomebody in the audience, triggering another ejaculation of\nthe speaker, with related rules, etc) suggesting that beside\ngrammatical syntactic structures, there may exist a variety\nof superimposed contexts which transmit information\nmediated by channels that are not those usually considered\n(Danchin 1987, Danchin and Marshall 1987; Marshall et al.\n1987). As in Dyson\u2019s scenario of the origin of life, the\nbasic functioning of the brain would base on reproduction,\nwhile invention of language with its linear sequences of\nphonemes, when spoken, and letters when written, would\nbe, in Man, the transition moment when it would begin to\ndiscover recursivity in linear strings of symbols (pho-\nnemes) which can be propagated from brain to brain, as\nprograms in a Turing Machine. In any event, in the few\ncases where it might do so, it would be an extremely slow\none (Sackur and Dehaene 2009). With this view, Nature\nwould have discovered twice the importance of coding and\nrecursivity, in the emergence of life first, and in the\nemergence of language, quite recently.\nAcknowledgements I wish to thank Philippe Binder for pointing\nout to me Wolfram\u2019s article, Jean-Baptiste Masson for pointing out\nSagawa and Ueda\u2019s article and anonymous reviewers for very con-\nstructive comments. This work developed in the Stanislas Noria\nnetwork is supported by the PROBACTYS and the TARPOL Euro-\npean Union programmes in Synthetic Biology (http://www.normale\nsup.org/~adanchin/causeries/causeries_en.html).\nOpen Access This article is distributed under the terms of the\nCreative Commons Attribution Noncommercial License which per-\nmits any noncommercial use, distribution, and reproduction in any\nmedium, provided the original author(s) and source are credited.\nReferences\nAverroe\u00a8s, Tahafut al Tahafut, EJW Gibb Memorial Trust, Cambridge\n(UK) ca 1175 (trans. latin 1497)\nBennett C (1988a) Notes on the history of reversible computation.\nIBM J Res Dev 44:270\u2013277\nBennett C (1988b) Logical depth and physical complexity. In: Herken\nR (ed) The universal turing machine: a half-century survey.\nOxford University Press, Oxford\nBernard C (1865) Introduction a` l\u2019e\u00b4tude de la me\u00b4decine expe\u00b4rimen-\ntale. Garnier-Flammarion (re\u00b4ed. 1966), Paris\nBullock S, Noble J, Watson R, Bedau MA (ed) (2008) Artificial life\nXI. MIT Press, Cambridge\nChan LY, Kosuri S, Endy D (2005) Refactoring bacteriophage T7.\nMol Syst Biol 1:0018\nChangeux JP, Courrege P, Danchin A (1973) A theory of the\nepigenesis of neuronal networks by selective stabilization of\nsynapses. Proc Natl Acad Sci USA 70:2974\u20132978\nClark A (1998) Being there: putting brain, body, and world together\nagain. MIT Press, Cambridge\nClemmer CR, Beebe TP Jr (1991) Graphite: a mimic for DNA and\nother biomolecules in scanning tunneling microscope studies.\nScience 251:640\u2013642\nCover T, Thomas J (1991) Elements of information theory. Wiley,\nNew York\nDanchin A (1986) Order and necessity. In: Quagliariello E, Bernardi G,\nUllmann A (eds) From enzyme adaptation to natural philo-\nsophy: heritage from Jacques Monod. Symposium J Monod and\nmolecular biology, yesterday and today. Elsevier Sciences\nPublishers, Amsterdam, Trani, Italy: 13\u201315Dec 1986, pp 187\u2013196\nDanchin A (1987) Biological foundations of language: a comment on\nNoam Chomsky\u2019s approach of syntactic structures. In: Mogdil S,\nMogdil C (eds) Noam chomsky, consensus and controversy. The\nFalmer Press, New York, pp 29\u201339\nDanchin A, Marshall JC (1987) Cross replies. In: Mogdil S, Mogdil C\n(eds) Noam chomsky, consensus and controversy. The Falmer\nPress, New York, pp 50\u201355\nDanchin A (1992) Science and technology: a western imbroglio.\nProjections 7/8:39\u201348\nDanchin A (2003) The Delphic boat. What genomes tell us. Harvard\nUniversity Press, Cambridge\nDanchin A (2009a) Bacteria as computers making computers. FEMS\nMicrobiol Rev 33:3\u201326\nDanchin A (2009b) A phylogenetic view of bacterial ribonucleases.\nProg Mol Biol Transl Sci 85:1\u201341\nDanchin A (2009c) Cells need safety valves. Bioessays 31:769\u2013773\nDanchin A (2009d) Natural selection and immortality. Biogerontol-\nogy 10:503\u2013516\nDesanti J-T (1968) Les Ide\u00b4alite\u00b4s Mathe\u00b4matiques. Le Seuil, Paris\nDiels H, Kranz W (1935) Die Fragmente der Vorsokratiker.\nWeidmann, Berlin\nDyson FJ (1985) Origins of life. Cambridge University Press,\nCambridge\nEdelman G (1987) Neural darwinism. The theory of neuronal group\nselection. Basic Books, New York\nElowitz MB, Leibler S (2000) A synthetic oscillatory network of\ntranscriptional regulators. Nature 403:335\u2013338\nEndy D (2005) Foundations for engineering biology. Nature 438:\n449\u2013453\nErige`ne JS De la division de la Nature (Periphyseon), Presses\nUniversitaires de France, Paris ca 850 (trans 1995)\nFang G, Rocha E, Danchin A (2005) How essential are nonessential\ngenes? Mol Biol Evol 22:2147\u20132156\nGaunt SJ (1991) Expression patterns of mouse Hox genes: clues to an\nunderstanding of developmental and evolutionary strategies.\nBioessays 13:505\u2013513\nGayon J (2007) The concept of the gene in contemporary biology:\ncontinuity or dissolution? Logic, Epistemol Unity Sci 6:\n91\u201395\nGibson DG, Benders GA, Andrews-Pfannkoch C, Denisova EA et al\n(2008) Complete chemical synthesis, assembly, and cloning of a\nMycoplasma genitalium genome. Science 319:1215\u20131220\nGrandy WT (1992) On randomness and thermodynamics. Found Phys\n22:853\u2013866\nInformation of the chassis and information 133\n123\nHe\u00b4naut A, Danchin A (1996) In Neidhardt F (ed) Escherichia coli and\nsalmonella, cellular and molecular biology. ASM Press, Wash-\nington, pp 2047\u20132065\nJudson H (1979) The eight day of creation. The makers of the\nrevolution in biology. Simon and Schuster, New York\nKuruma Y, Stano P, Ueda T, Luisi PL (2009) A synthetic biology\napproach to the construction of membrane proteins in semi-\nsynthetic minimal cells. Biochim Biophys Acta 1788:567\u2013574\nLakatos I (1976) Proofs and refutations. Cambridge University Press,\nCambridge\nLakatos I (1980) The methodology of scientific research programmes,\nvol 1: philosophical papers. Cambridge University Press,\nCambridge\nLandauer R (1961) Irreversibility and heat generation in the\ncomputing process. IBM J Res Dev 3:184\u2013191\nLartigue C, Glass JI, Alperovich N, Pieper R et al (2007) Genome\ntransplantation in bacteria: changing one species to another.\nScience 317:632\u2013638\nLivingstone D (2003) School of computing. University of the West of\nScotland, Paisley Campus, Scotland, p 193\nMarshall JC (1987) Language learning, language acquisition, or\nlanguage growth? In: Mogdil S, Mogdil C (eds) Noam chomsky,\nconsensus and controversy. The Falmer Press, New York,\npp 41\u201349\nMonod J (1971) Chance and necessity: an essay on the natural\nphilosophy of modern biology. Alfred A Knopf, New York\nPopper K (1959) The logic of scientific discovery. Hutchinson,\nLondon\nPopper K (1963) Conjectures and refutations: the growth of scientific\nknowledge. Routledge, London\nPutnam H (1988) Representation and reality. MIT Press, Cambridge\nRobinson A (1996) Non-standard analysis. Princeton University\nPress, Princeton\nRuelle D (2000) Mathematical platonism reconsidered. Nieuw\nArchief voor Wiskunde 5:30\u201333\nRyle G (1949) The concept of mind. The University of Chicago Press,\nChicago\nSackur J, Dehaene S (2009) The cognitive architecture for chaining of\ntwo mental operations. Cognition 111:187\u2013211\nSagawa T, Ueda M (2009) Minimal energy cost for thermodynamic\ninformation processing: measurement and information erasure.\nPhys Rev Lett 102:250602/250601\u2013250602/250604\nSchneider TD (1991a) Theory of molecular machines. I. Channel\ncapacity of molecular machines. J Theor Biol 148:83\u2013123\nSchneider TD (1991b) Theory of molecular machines. II. Energy\ndissipation from molecular machines. J Theor Biol 148:125\u2013137\nSchneider TD, Stephens RM (1990) Sequence logos: a new way to\ndisplay consensus sequences. Nucleic Acids Res 18:6097\u20136100\nShannon CE, Weaver W (1949) The mathematical theory of\ncommunication. University of Illinois Press, Urbana\nShenhav B, Lancet D (2004) Prospects of a computational origin of\nlife endeavor. Orig Life Evol Biosph 34:181\u2013194\nSolovei I, Kreysing M, Lancto\u02c6t C, Ko\u00a8sem S et al (2009) Nuclear\narchitecture of rod photoreceptor cells adapts to vision in\nmammalian evolution. Cell 137:356\u2013368\nTamames J, Gonzalez-Moreno M, Mingorance J, Valencia A, Vicente\nM (2001) Bringing gene order into bacterial shape. Trends Genet\n17:124\u2013126\nTanaka M (1984) A physical characterization of biological informa-\ntion and communication system model of ecosystems. J Theor\nBiol 110:619\u2013635\nVeening JW, Smits WK, Kuipers OP (2008) Bistability, epigenetics,\nand bet-hedging in bacteria. Annu Rev Microbiol 62:193\u2013210\nvon Neumann J (1958) The computer and the brain. Yale University\nPress, New Haven (reed 1979)\nWilmut I, Schnieke AE, McWhir J, Kind AJ, Campbell KH (1997)\nViable offspring derived from fetal and adult mammalian cells.\nNature 385:810\u2013813\nWoese CR (2002) On the evolution of cells. Proc Natl Acad Sci USA\n99:8742\u20138747\nWoese CR, Magrum LJ, Fox GE (1978) Archaebacteria. J Mol Evol\n11:245\u2013251\nWolfram S (1985) Origins of randomness in physical systems. Phys\nRev Lett 55:449\u2013452\nYockey HP (1992) Information theory and molecular biology.\nCambridge University Press, Cambridge\nZurek W (1989) Algorithmic randomness and physical entropy. Phys\nRev A 40:4731\u20134751\n134 A. Danchin\n123\n", "identifiers": ["doi:10.1007/s11693-009-9036-5", "10.1007/s11693-009-9036-5"], "publisher": "Springer", "relations": ["http://dx.doi.org/10.1007/s11693-009-9036-5"], "repositories": [{"id": "2612", "openDoarId": 0, "name": "Springer - Publisher Connector", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 1, "metadataAdded": 1493847573000, "metadataUpdated": 1493847573000, "depositedDate": null}, "subjects": [], "title": "Information of the chassis and information of the program in synthetic cells", "topics": [], "types": [], "year": 2009, "fulltextIdentifier": "https://core.ac.uk/download/pdf/81867052.pdf", "doi": "10.1007/s11693-009-9036-5", "downloadUrl": "https://core.ac.uk/download/pdf/81867052.pdf"}}
{"status": "OK", "data": {"id": "82854584", "authors": ["Daniel Collerton"], "citations": [], "contributors": [], "datePublished": "2013", "fullText": "MINI REVIEW ARTICLE\npublished: 06 September 2013\ndoi: 10.3389/fpsyg.2013.00548\nPsychotherapy and brain plasticity\nDaniel Collerton1,2*\n1 Clinical Psychology, Northumberland, Tyne and Wear NHS Foundation Trust, Gateshead, UK\n2 Newcastle upon Tyne, Newcastle University, UK\nEdited by:\nElaine K. Perry, Newcastle\nUniversity, UK\nReviewed by:\nUrsula Voss, Rheinische\nFriedrich-Wilhelms University Bonn,\nGermany\nElaine K. Perry, Newcastle\nUniversity, UK\n*Correspondence:\nDaniel Collerton, Clinical Psychology,\nNorthumberland, Tyne and Wear\nNHS Foundation Trust, Bensham\nHospital, Saltwell Road, Gateshead,\nNE8 4YL, UK\ne-mail: daniel.collerton@ncl.ac.uk\nIn this paper, I will review why psychotherapy is relevant to the question of how\nconsciousness relates to brain plasticity. A great deal of the research and theorizing on\nconsciousness and the brain, including my own on hallucinations for example (Collerton\nand Perry, 2011) has focused upon specific changes in conscious content which can be\nrelated to temporal changes in restricted brain systems. I will argue that psychotherapy,\nin contrast, allows only a focus on holistic aspects of consciousness; an emphasis which\nmay usefully complement what can be learnt from more specific methodologies.\nKeywords: plasticity, psychotherapy, functional magnetic resonance imaging, multivoxel pattern analysis,\nemotion, consciousness\nINTRODUCTION\nFor the last century or so, psychotherapy has aimed to change\nthe mind. And if it has been effective in doing so, it must have\nlead to lasting changes in conscious content, and potentially in\nthe process of consciousness itself. Few other human endeavors\nseek so systematically to produce predictable enduring variation\nin emotion, cognition, behavior and somatic perceptions; changes\nwhich can persist for many years beyond the end of therapy. Over\nthe last few decades, as methods have become available for mea-\nsuring brain structure and function, it has provided a potential\nreal-life method by which meaningful changes in consciousness\ncan be related to measures of brain function. Admittedly, thus\nfar interest has been on understanding what brain function can\ntell us about how psychotherapy works (for example, Jokic\u00b4-Begic\u00b4,\n2010) rather than the aim of this paper\u2014what psychotherapy can\ntell us about how the brain and mind are linked. However, in\ntheory, it should be possible to assess consciousness, or at least\nsome aspects of it, before and after psychotherapy, and to relate\nthese to brain changes. And indeed, as described later, such stud-\nies have been done. The evidence is still sporadic and somewhat\ncontradictory, but there is more potential now than ever before\nto correlate psychotherapy-related changes in mind to changes in\nbrain.\nIn this paper, I will survey the current situation, outlining the\nformidable conceptual and practical difficulties which still need to\nbe overcome, and suggest a potential strength of psychotherapy\nas a tool for understanding consciousness which, in combina-\ntion with advances in functional imaging analysis, may give a way\nforwards.\nCHALLENGES OF DEFINITION AND MEASUREMENT\nThere are a number of intractable definitional and measurement\nissues in this field which have been only partially solved.\nConsciousness itself is a fuzzy concept\u2014and its components\nare no clearer, as illustrated by the continuing discussions on\nwhether it even exists as ameaningful phenomenon. James explic-\nitly posed the question in the first, 1904, volume of the Journal of\nPhilosophy, Psychology, and Scientific Methods and over a cen-\ntury later the answer is still not clear. As exemplified by Newell\nand Shanks (in press), the nature and role of consciousness and\nits relationship to behavior are still under discussion.\nThe term psychotherapy equally lacks definition. It has been\napplied to an exceptionally wide range of approaches with dif-\nfering models, techniques, goals, and outcomes. Feltham and\nHorton\u2019s (2012) Handbook, for example, surveys some 23 major\napproaches from Dialectical Behavior Therapy to Psychodrama.\nFinally, though methods of investigating in vivo brain function\nare incomparably better than even a decade ago, they are still lim-\nited in cognitive, temporal, and spatial resolution, while the large\nand intrusive technology involved limits applications to settings\nwhich bear little relationship to everyday life.\nIn the face of fuzzy concepts and methods, there is a scientific\ntemptation to retreat into investigating relatively specific, easily\nmeasureable, aspects of consciousness and the brain. However,\nlooked at in another way, this very lack of focus may be a strength\nrather than a weakness in that it forces attention to holistic\nchanges in consciousness which can then be related to systemic\nchanges in neural networks.\nTHE EFFECTS OF PSYCHOTHERAPY\nWithin the vast variety of psychotherapies, Cognitive Behavioral\nTherapy (CBT) has become the most widely accepted approach,\nand has amassed a strong body of evidence of effectiveness (Butler\net al., 2006). It leads to long lasting, reproducible changes in emo-\ntion, cognition, behavior, and somatic symptoms across a range\nof mood and other psychological disorders. This strongly sug-\ngests that CBT is a powerful means of changing consciousness.\nCBT has always had a marked emphasis on measurement and it\nis also the therapy whose effects have most often been related to\nbrain changes. For these reasons, I will take it as the exemplar\nwww.frontiersin.org September 2013 | Volume 4 | Article 548 | 1\nCollerton Brain plasticity and psychotherapy\npsychotherapy for the purpose of this paper, while acknowledging\nthat other approaches may be equally valid.\nIf we want to relate the effects of psychotherapy on conscious-\nness to brain changes, it would seem necessary to know the\nchanges it produces and how it does so. Consciousness itself is\nnot directly accessible of course, so, as summarized in Table 1,\nresearchers have taken a number of routes to inferring what is\nin the consciousness of a patient. Self-report, observer report,\nbehavioral measures, and experimental tasks can be used to more\nor less directly infer changes in the content of patient\u2019s con-\nscious thought. Taken together, these methodologies have pro-\nduced good evidence that the complex combinations of emotions,\ncognitions, behaviors and somatic symptoms which characterize\nmood disorders do shift as a result of CBT.\nWHAT CHANGES AS A CONSEQUENCE OF\nPSYCHOTHERAPY?\nHowever, evidence is lacking as to what specifically changes as a\nconsequence of psychotherapy (see, for example, Murphy et al.,\n2009). Despite the range of different ways of measuring the effects\nof psychotherapy noted above, it is striking how closely these are\nrelated. Thus, change in one symptom area, for example cogni-\ntion, is accompanied by changes in other symptom areas such as\nemotion or behavior; at least as averaged over the timescales and\ngroup numbers common in treatment trials.\nSimilarly, though there is evidence that different modalities of\ntherapy may have different levels of effectiveness (see Tolin, 2010\nfor a meta-analytic comparison of CBT with other therapies),\nwhere this does occur this appears to be more a quantitative than\na qualitative difference. The outcomes of psychodynamic, person-\ncentred, and behavioral psychotherapy are broadly equivalent\ndespite their varieties of approaches and targets for therapeu-\ntic change (Stiles et al., 2008; Budd and Hughes, 2009) perhaps\nbecause they work via common final paths (Mansell, 2011).\nEven treatments as different as CBT and pharmacotherapy\nappear to have broadly similar outcomes in, for example, depres-\nsion (DeRubeis et al., 2008) with no reliable evidence of the\ndifferential effects of CBT on negative cognitions or medication\non somatic symptoms as might have been expected from their\nmechanisms of action. Attempts to predict which patients might\nbenefit from any specific intervention have not been successful.\nTaking all of these comparisons together, there is very lit-\ntle evidence that specific areas of consciousness can change in a\nmeaningful way without these effects rippling through the rest of\nconsciousness. This suggests that trying to fractionate the effects\nof psychotherapy on consciousness into components might not\nbe possible. It is the totality of consciousness which changes as a\nresult of CBT rather than one specific aspect of it.\nIMPLICATIONS FOR LINKING THE EFFECTS OF\nPSYCHOTHERAPY TO BRAIN PLASTICITY\nThis has implications for relating the effects of CBT to brain\nplasticity. Traditionally, neuroscience has adopted a reductionist\napproach to the brain; relating specific psychological functions\nto specific brain areas. This has been enormously successful with\nmany psychological functions. Amongst many other pairings,\nneuropsychological localization has linked learning to the hip-\npocampus and other structures in the medial temporal lobe,\nobject perception to the ventral visual stream, and language to\nthe left temporal cortex. However, the localization paradigm has\nleft in its wake the binding problem\u2014how the functions of dis-\nparate brain areas are tied together to produce the usual subjective\nsense of a single coherent consciousness. In this context, perhaps\nthe holistic effects of CBT and other psychotherapies on con-\nsciousness become a means of side stepping the binding problem.\nIf significant changes in bound consciousness can be related to\nrestricted changes in brain function, that might narrow down the\ncandidate brain areas which may underlie consciousness.\nEFFECTS OF CBT ON MEASURES OF BRAIN FUNCTION\nThe plasticity in human brain which underlies individual,\nidiosyncratic and instantaneous elements of consciousness\u2014\nspecific words, thoughts, images, feelings, and memories\u2014comes\nfrom tiny, subtle, and dynamic changes which are embedded\nwithin and across networks of microscopic cells. In comparison,\nour ways of measuring plasticity in the human brain have to trade\noff cognitive (the ability of scans to resolve precise psychological\nstates, particular memories for instance), spatial, and tempo-\nral resolutions with even the best resolution vastly greater than\nthe fundamental mechanisms of plasticity. The spatial resolution\nlimit of a Magnetic Resonance Imaging (MRI) scan, our best cur-\nrent means of directly assessing brain structure and function,\ncontains somewhere around 9 million brain cells (deCharms,\n2008).\nHowever, this has not prevented functional imaging, partic-\nularly functional MRI (fMRI) to identify which areas of the\nbrain change following psychotherapy. (There have been a rather\nsmall number of structural imaging studies of the effects of psy-\nchotherapy, mainly in eating disorders, but many of these have\nbeen confounded by the effects of weight gain and loss on brain\nstructure (Lobera, 2011) making it difficult to interpret their\nresults).\nThe dominant paradigm has been to compare levels of brain\nactivity pre and post CBT to see what changes. This approach\nhas mainly been used in depression and has identified that\nchanges are localized to specific frontal, cingulate, and limbic\nareas. There is decreased activity in the limbic system, especially\nTable 1 | Indicators of consciousness used in CBT outcome studies.\nType of measure Example measures Illustrative study\nSelf-report Structured questionnaires e.g., Beck Depression Inventory (Beck et al., 1988) Elkin et al., 1995\nObserver report Structured observer ratings e.g., Hamilton Depression Rating Scale (Bagby et al., 2004) Teasdale et al., 2000\nBehavioral measures Observable behavioral change e.g., return to employment Della-Posta and Drummond, 2006\nExperimental tasks Measures of perceptual distortion e.g., body morphing (Benson et al., 1999) Cornelissen et al., 2013\nFrontiers in Psychology | Consciousness Research September 2013 | Volume 4 | Article 548 | 2\nCollerton Brain plasticity and psychotherapy\nthe amygdala, with dorsolateral prefrontal cortex becoming rela-\ntively more active and orbitomedial and cingulate cortex less so; a\nmove toward normality from patterns observed before treatment\n(Ochsner et al., 2002; Goldapple et al., 2004; Malhi et al., 2004;\nRitchey et al., 2011; H\u00f6flich et al., 2012) and consistent with what\nis know of the processing of emotional stimuli (Simpson et al.,\n2000; Northoff et al., 2004; Lepp\u00e4nen, 2006; Beck, 2008). Pre-\ntreatment levels of cingulate activity can even predict response\nto CBT with some reliability (Konarski et al., 2009; Ritchey et al.,\n2011; Siegle et al., 2012).\nThere have not been comparisons between the effects of differ-\nent types of psychotherapy on the brain (potentially interesting\nin view of their equivalent effects on consciousness), but there\nare conflicting reports of the effects of pharmacotherapy; similar\nchanges in brain activity to those following psychotherapy were\nnot seen after antidepressant treatment by Goldapple et al. (2004)\nthough they were identified by Furmark et al. (2002); opening up,\nbut not confirming, the possibility that different brain changes\nmight have similar effects on consciousness.\nTaken as a whole, this evidence would suggest that the holistic\nchanges in consciousness seen after psychotherapy for depression\nare associated with changes in a relatively restricted number of\nbrain areas; mainly frontal, cingulate, and limbic cortex, with the\nimplication that plasticity in those areas is particularly associated\nwith persistent variations in consciousness.\nHowever, a one to one correspondence between change in\ndepression and change in specific brain areas may be over stated\n(Linden, 2006; Frewen et al., 2008; Dichter et al., 2012). For\nexample, very similar changes in those brain areas are seen after\nCBT and other psychological treatments for anxiety (Furmark\net al., 2002; Paquette et al., 2003; Straube et al., 2006; Porto\net al., 2009; Freyer et al., 2011), schizophrenia (Wykes et al.,\n2002), eating disorders (Vocks et al., 2011) and Irritable Bowel\nSyndrome (Lackner et al., 2006). This is consistent with a con-\nsciousness network which depends upon these brain areas (and\nno doubt others) but it also suggests that there is a large over-\nlap in the brain changes associated with different holistic states\nof consciousness. At present, fMRI data would suggest that, sim-\nply put, CBT is associated with a decrease in emotionality (less\nlimbic activity) and an increase in thoughtfulness (increased dor-\nsolateral frontal activity), as would be expected from its aims\nand methods (Clark and Beck, 2010). Though we may be able to\nlink consciousness to a subset of anatomical structures using psy-\nchotherapy as an investigative tool, we appear to lack specificity in\nour account of how different states of consciousness could arise. Is\nit that in using psychotherapy to localize holistic changes in con-\nsciousness to a restricted set of brain structures, we have lost the\nability to account for why consciousness is so idiosyncratic and so\nchangeable?\nAPOTENTIAL WAY FORWARD\nAn analogous challenge has arisen in fMRI studies of visual per-\nception. Early attempts to localize specific perceptions to specific\nbrain areas worked only for grossly different stimuli\u2014visualizing\nnavigating a house compared to imagining playing tennis (Owen\net al., 2006)\u2014or for simple stimuli in early, highly specialized,\nvisual areas (Kay et al., 2008). More latterly, however, multivoxel\npattern analysis (MVPA), in which patterns of activity across wide\nareas of the brain are analyzed, has produced a significant increase\nin cognitive resolution. Fairly similar stimuli, for example chairs\nand shoes (Norman et al., 2006; deCharms, 2008; Poldrack, 2011),\nor over-riding categories of images such as living or non-living\n(Naselaris et al., 2012) can now be recognized from pattern infor-\nmation fMRI (Formisano and Kriegeskorte, 2012) data. Not only\nperceptions, but also images and memories (Chadwick et al.,\n2012; Rissman andWagner, 2012) are starting to be distinguished.\nBeginnings are starting to be made in reproducing data across\nas well as within subjects (Accamma and Suma, 2012; Raizada\nand Connolly, 2012). Significantly, cognitive resolution appears\nto increase as the focus of the analysis is widened to include more\nbrain areas.\nMVPA might therefore lead to the ability to map holis-\ntic changes in consciousness to patterns within and across the\nregions that classic fMRI has identified as responsive to psy-\nchotherapy (Siegle et al., 2007). Thus it may provide the mech-\nanism to bridge holistic and specific variations in consciousness\nand brain.\nCONCLUSIONS\nIt is clear that CBT, and probably other psychotherapies, alters\nconsciousness in personally important, lasting, and measurable\nways. Brain function and brain structure are different after CBT.\nLooking at functional changes in the brain suggests that con-\nsciousness changes in response to plasticity in the linked systems\nof the frontal, cingulate, and limbic cortices. However, we do not\nknow how modulations in those areas link to different states of\nconsciousness. Using the most recent imaging analysis to map\nactivity simultaneously across these regions might give the miss-\ning specificity; allowing whole brain changes to be mapped to\nholistic changes in consciousness.\nIn order to do this, our next challenge will be to develop ways\nof capturing the experience of consciousness as a whole rather\nthan, as we have tried to do in the past, the individual thoughts,\nimages, and emotions which are bound together to produce it.\nREFERENCES\nAccamma, I. V., and Suma, H. N.\n(2012). Feature selection for decod-\ning of cognitive states in multiple-\nsubject functional magnetic reso-\nnance imaging data. Adv. Intell. Syst.\nComput. 174 2012, 997\u20131004. doi:\n10.1007/978-81-322-0740-5_121\nBagby, R. M., Ryder, A. G., Schuller,\nD. R., and Marshall, M. B. (2004).\nThe hamilton depression rat-\ning Scale: has the gold standard\nbecome a lead weight? Am. J.\nPsychiatry 161, 2163\u20132177. doi:\n10.1176/appi.ajp.161.12.2163\nBeck, A. (2008). The evolution of\nthe cognitive model of depression\nand its neurobiological correlates.\nAm. J. Psychiatry 165, 969\u2013977. doi:\n10.1176/appi.ajp.2008.08050721\nBeck, A. T., Steer, R. A., and Carbin,\nM. G. (1988). Psychometric prop-\nerties of the Beck depression inven-\ntory: twenty-five years of evaluation.\nClin. Psychol. Rev. 8, 77\u2013100. doi:\n10.1016/0272-7358(88)90050-5\nBenson, P. J., Emery, J. L., Cohen-Tov\u00e9e,\nE. M., and Tov\u00e9e, M. J. (1999).\nA computergraphic technique for\nthe study of body size perception\nand body types. Behav. Res. Methods\nInstrum. Comput. 31, 446\u2013454. doi:\n10.3758/BF03200725\nBudd, R., and Hughes, I. (2009). The\nDodo Bird Verdict\u2014controversial,\ninevitable and important: a\ncommentary on 30 years of\nmeta\u2212analyses. Clin. Psychol.\nPsychother. 16, 510\u2013522. doi:\n10.1002/cpp.648\nwww.frontiersin.org September 2013 | Volume 4 | Article 548 | 3\nCollerton Brain plasticity and psychotherapy\nButler, A. C., Chapman, J. E., Forman,\nE. M., and Beck, A. T. (2006).\nThe empirical status of cognitive-\nbehavioral therapy: a review of\nmeta-analyses. Clin. Psychol. Rev.\n26, 17\u201331. doi: 10.1016/j.cpr.2005.\n07.003\nChadwick, M. J., Bonnici, H. M., and\nMaguire, E. A. (2012). Decoding\ninformation in the human\nhippocampus: a user\u2019s guide.\nNeuropsychologia 50, 3107\u20133121.\ndoi: 10.1016/j.neuropsychologia.\n2012.07.007\ndeCharms, R. C. (2008). Applications\nof real-time fMRI. Nat. Rev.\nNeurosci. 9, 720\u2013729. doi: 10.1038/\nnrn2414\nClark, D. A., and Beck, A. T. (2010).\nCognitive theory and therapy of\nanxiety and depression: conver-\ngence with neurobiological findings.\nTrends Cogn. Sci. 14, 418\u2013424. doi:\n10.1016/j.tics.2010.06.007\nCollerton, D., and Perry, E. (2011).\nDreaming and hallucinations\n\u2013 continuity or discontinuity.\nPerspectives from dementia with\nLewy bodies. Conscious. Cogn. 20,\n1016\u20131020. doi: 10.1016/j.concog.\n2011.03.024\nCornelissen, P. L., Johns, A., and\nMartin, J. T. (2013). Body size over-\nestimation in women with anorexia\nnervosa is not qualitatively dif-\nferent from female controls origi-\nnal research article Body Image 10,\n103\u2013111.\nDella-Posta, C., and Drummond, P.\nD. (2006). Cognitive behavioural\ntherapy increases re-employment\nof job seeking worker\u2019s compensa-\ntion clients. J. Occup. Rehabil. 16,\n217\u2013224. doi: 10.1007/s10926-006-\n9024-5\nDeRubeis, R. J., Siegle, G. J., and\nHollon, S. D. (2008). Cognitive ther-\napy versus medication for depres-\nsion: treatment outcomes and neu-\nral mechanisms. Nat. Rev. Neurosci.\n9, 788\u2013796. doi: 10.1038/nrn2345\nDichter, G. S., Sikich, L., Song, A.,\nVoyvodic, J., and Bodfish, J. W.\n(2012). Functional neuroimaging\nof treatment effects in psychiatry:\nmethodological challenges and rec-\nommendations. Int. J. Neurosci. 122,\n483\u2013493. doi: 10.3109/00207454.\n2012.678446\nElkin, I., Gibbons, R. D., Shea, M.\nT., Sotsky, S. M., Watkins, J. T.,\nPilkonis, P. A., et al. (1995). Initial\nseverity and differential treatment\noutcome in the national institute\nof mental health treatment of\ndepression collaborative research\nprogram. J. Consult. Clin. Psychol.\n63, 841. doi: 10.1037/0022-006X.63.\n5.841\nFeltham, C., and Horton, I. (eds.).\n(2012). The SAGE Handbook of\nCounselling and Psychotherapy.\nLondon: SAGE Publications.\nFormisano, E., and Kriegeskorte, N.\n(2012). Seeing patterns through\nthe hemodynamic veil\u2014the future\nof pattern-information fMRI.\nNeuroimage 62, 1249\u20131265. doi:\n10.1016/j.neuroimage.2012.02.078\nFrewen, P. A., Dozois, D. J., and\nLanius, R. A. (2008). Neuroimaging\nstudies of psychological interven-\ntions for mood and anxiety disor-\nders: empirical and methodologi-\ncal review. Clin. Psychol. Rev. 28,\n228\u2013246. doi: 10.1016/j.cpr.2007.\n05.002\nFreyer, T., Kl\u00f6ppel, S., T\u00fcscher, O.,\nKordon, A., Zurowski, B., Kuelz, A.\nK. et al. (2011). Frontostriatal\nactivation in patients with\nobsessive\u2013compulsive disorder\nbefore and after cognitive behav-\nioral therapy. Psychol. Med. 41,\n207\u2013216. doi: 10.1017/S003329\n1710000309\nFurmark, T., Tillfors, M.,\nMarteinsdottir, I., Fischer, H.,\nPissiota, A., Langstrom, B., et al.\n(2002). Common changes in\ncerebral blood flow in patients\nwith social phobia treated with\ncitalopram or cognitive-behavioral\ntherapy. Arch. Gen. Psychiatry\n59:425. doi: 10.1001/archpsyc.59.\n5.425\nGoldapple, K., Segal, Z., Garson, C.,\nLau, M., Bieling, P., Kennedy,\nS., et al. (2004). Modulation of\ncortical-limbic pathways in major\ndepression: treatment-specific\neffects of cognitive behavior ther-\napy. Arch. Gen. Psychiatry 61, 34.\ndoi: 10.1001/archpsyc.61.1.34\nH\u00f6flich, A., Baldinger, P., Savli, M.,\nLanzenberger, R., and Kasper, S.\n(2012). Imaging treatment effects\nin depression. Rev. Neurosci. 23,\n227\u2013252. doi: 10.1515/revneuro-\n2012-0038\nJames, W. (1904). Does \u2018Consciousness\u2019\nexist. J. Philos. Psychol. Sci. Methods\n1, 477\u2013491. doi: 10.2307/2011942\nJokic\u00b4-Begic\u00b4, N. (2010). Cognitive-\nbehavioral therapy and\nneuroscience: towards closer\nintegration psychological topics. 19,\n235\u2013254.\nKay, K. N., Naselaris, T., Prenger, R. J.,\nand Gallant, J. L. (2008). Identifying\nnatural images from human brain\nactivity. Nature 452, 352\u2013355. doi:\n10.1038/nature06713\nKonarski, J. Z., Kennedy, S. H., Segal,\nZ. V., Lau, M. A., Bieling, P. J.,\nMcIntyre, R. S., et al. (2009).\nPredictors of nonresponse to\ncognitive behavioural therapy\nor venlafaxine using glucose\nmetabolism in major depressive\ndisorder. J. Psychiatry Neurosci. 34,\n175.\nLackner, J. M., Lou Coad, M., Mertz, H.\nR., Wack, D. S., Katz, L. A., Krasner,\nS. S., et al. (2006). Cognitive therapy\nfor irritable bowel syndrome is asso-\nciated with reduced limbic activity,\nGI symptoms, and anxiety. Behav.\nRes. Ther. 44, 621\u2013638. doi: 10.1016/\nj.brat.2005.05.002\nLepp\u00e4nen, J. M. (2006). Emotional\ninformation processing in mood\ndisorders: a review of behav-\nioral and neuroimaging findings.\nCurr. Opin. Psychiatry 19, 34.\ndoi: 10.1097/01.yco.0000191500.\n46411.00\nLinden, D. E. J. (2006). How psy-\nchotherapy changes the brain\u2013the\ncontribution of functional neu-\nroimaging. Mol. Psychiatry 11,\n528\u2013538. doi: 10.1038/sj.mp.\n4001816\nLobera, I. J. (2011). Neuroimaging\nin eating disorders. Neuropsychiatr.\nDis. Treat. 82, 45.\nMalhi, G. S., Lagopoulos, J., Ward, P. B.,\nKumari, V., Mitchell, P. B., Parker,\nG. B. et al. (2004). Cognitive gener-\nation of affect in bipolar depression:\nan fMRI study. Eur. J. Neurosci. 19,\n741\u2013754. doi: 10.1111/j.0953-816X.\n2003.03159.x\nMansell, W. (2011). Core processes\nof psychopathology and recovery:\n\u201cDoes the Dodo bird effect have\nwings?\u201d Clin. Psychol. Rev. 31,\n189\u2013192. doi: 10.1016/j.cpr.2010.\n06.009\nMurphy, R., Cooper, Z., Hollon, S. D.,\nand Fairburn, C. G. (2009). How\ndo psychological treatments work.\ninvestigating mediators of change.\nBehav. Res. Ther. 47, 1. doi: 10.1016/\nj.brat.2008.10.001\nNaselaris, T., Stansbury, D. E., and\nGallant, J. L. (2012). Cortical rep-\nresentation of animate and inan-\nimate objects in complex natu-\nral scenes. J. Physiol. Paris 106,\n239\u2013249. doi: 10.1016/j.jphysparis.\n2012.02.001\nNewell, B. R., and Shanks, D. R.\n(in press). Unconscious influ-\nences on decision making. Behav.\nBrain Sci.\nNorman, K. A., Polyn, S. M., Detre, G.\nJ., and Haxby, J. V. (2006). Beyond\nmind-reading: multi-voxel pattern\nanalysis of fMRI data. Trends Cogn.\nSci. 10, 424\u2013430. doi: 10.1016/j.tics.\n2006.07.005\nNorthoff, G., Heinzel, A., Bermpohl,\nF., Niese, R., Pfennig, A.,\nPascual\u2212Leone, A., et al. (2004).\nReciprocal modulation and atten-\nuation in the prefrontal cortex: an\nfMRI study on emotional\u2013cognitive\ninteraction. Hum. Brain Mapp. 21,\n202\u2013212. doi: 10.1002/hbm.20002\nOchsner, K. N., Bunge, S. A., Gross,\nJ. J., and Gabrieli, J. D. (2002).\nRethinking feelings: an fMRI\nstudy of the cognitive regulation\nof emotion. J. Cogn. Neurosci.\n14, 1215\u20131229. doi: 10.1162/\n089892902760807212\nOwen, A. M., Coleman, M. R., Boly,\nM., Davis, M. H., Laureys, S., and\nPickard, J. D. (2006). Detecting\nawareness in the vegetative state.\nScience 313, 1402\u20131402. doi:\n10.1126/science.1130197\nPaquette, V., L\u00e9vesque, J., Mensour,\nB., Leroux, J. M., Beaudoin,\nG., Bourgouin, P., et al. (2003).\n\u201cChange the mind and you\nchange the brain\u201d: effects of\ncognitive-behavioral therapy\non the neural correlates of spi-\nder phobia. Neuroimage 18,\n401\u2013409. doi: 10.1016/S1053-8119\n(02)00030-7\nPoldrack, R. A. (2011). Inferring\nmental states from neuroimaging\ndata: from reverse inference to\nlarge-scale decoding. Neuron 72,\n692\u2013697. doi: 10.1016/j.neuron.\n2011.11.001\nPorto, P., Oliveira, L., Mari, J.,\nVolchan, E., Figueira, I., and\nVentura, P. (2009). Does cogni-\ntive behavioral therapy change\nthe brain. a systematic review\nof neuroimaging in anxiety\ndisorders. J. Neuropsychiatry.\nClin. Neurosci. 21, 114\u2013125.\ndoi: 10.1176/appi.neuropsych.21.\n2.114\nRaizada, R. D., and Connolly, A. C.\n(2012). What makes different peo-\nple\u2019s representations alike: neural\nsimilarity space solves the prob-\nlem of across-subject fmri decoding.\nJ. Cogn. Neurosci. 24, 868\u2013877. doi:\n10.1162/jocn_a_00189\nRissman, J., and Wagner, A. D. (2012).\nDistributed representations in\nmemory: insights from functional\nbrain imaging. Annu. Rev. Psychol.\n63, 101\u2013128. doi: 10.1146/annurev-\npsych-120710-100344\nRitchey, M., Dolcos, F., Eddington, K.\nM., Strauman, T. J., and Cabeza,\nR. (2011). Neural correlates of\nemotional processing in depression:\nchanges with cognitive behavioral\ntherapy and predictors of treat-\nment response. J. Psychiatr. Res. 45,\n577\u2013587. doi: 10.1016/j.jpsychires.\n2010.09.007\nSiegle, G. J., Ghinassi, F., and Thase,\nM. E. (2007). Neurobehavioral\ntherapies in the 21st century:\nsummary of an emerging field and\nan extended example of cognitive\nFrontiers in Psychology | Consciousness Research September 2013 | Volume 4 | Article 548 | 4\nCollerton Brain plasticity and psychotherapy\ncontrol training for depression.\nCogn. Ther. Res. 31, 235\u2013262. doi:\n10.1007/s10608-006-9118-6\nSiegle, G. J., Thompson, W. K., Collier,\nA., Berman, S. R., Feldmiller, J.,\nThase, M. E., et al. (2012). Toward\nclinically useful neuroimaging in\ndepression treatment prognostic\nutility of subgenual cingulate\nactivity for determining depres-\nsion outcome in cognitive therapy\nacross studies, scanners, and\npatient characteristic ssgacc and\ncognitive therapy outcome. Arch.\nGen. Psychiatry 69, 913\u2013924. doi:\n10.1001/archgenpsychiatry.2012.65\nSimpson, J. R., \u00d6ng\u00fcr, D., Akbudak,\nE., Conturo, T. E., Ollinger, J. M.,\nSnyder, A. Z., et al. (2000). The\nemotional modulation of cognitive\nprocessing: an fMRI study. J. Cogn.\nNeurosci. 12(Suppl. 2), 157\u2013170.\ndoi: 10.1162/089892900564019\nStiles, W. B., Barkham, M.,\nMellor-Clark, J., and Connell,\nJ. (2008). Effectiveness of\ncognitive-behavioural, person-\ncentred, and psychodynamic\ntherapies in UK primary-care\nroutine practice: replication in a\nlarger sample. Psychol. Med. 38,\n677\u2013688. doi: 10.1017/S0033291\n707001511\nStraube, T., Glauer, M., Dilger, S.,\nMentzel, H. J., and Miltner, W.\nH. (2006). Effects of cognitive-\nbehavioral therapy on brain\nactivation in specific phobia.\nNeuroimage 29, 125\u2013135. doi:\n10.1016/j.neuroimage.2005.07.007\nTeasdale, J. D., Segal, Z. V., Williams,\nJ. M. G., Ridgeway, V. A., Soulsby,\nJ. M., and Lau, M. A. (2000).\nPrevention of relapse/recurrence in\nmajor depression by mindfulness-\nbased cognitive therapy. J. Consult.\nClin. Psychol. 68, 615. doi: 10.1037/\n0022-006X.68.4.615\nTolin, D. F. (2010). Is cognitive\u2013\nbehavioral therapy more effective\nthan other therapies?: a meta-\nanalytic review. Clin. Psychol. Rev.\n30, 710\u2013720. doi: 10.1016/j.cpr.\n2010.05.003\nVocks, S., Schulte, D., Busch, M.,\nGr\u00f6nemeyer, D., Herpertz, S., and\nSuchan, B. (2011). Changes in neu-\nronal correlates of body image\nprocessing by means of cognitive-\nbehavioural body image therapy for\neating disorders: a randomized con-\ntrolled fMRI study. Psychol. Med. 41,\n1651\u20131663. doi: 10.1017/S00332917\n10002382\nWykes, T., Brammer, M., Mellers,\nJ., Bray, P., Reeder, C., Williams,\nC., et al. (2002). Effects on the\nbrain of a psychological treatment:\ncognitive remediation therapy\nfunctional magnetic resonance\nimaging in schizophrenia. Br. J.\nPsychiatry 181, 144\u2013152.\nConflict of Interest Statement: The\nauthor declares that the research\nwas conducted in the absence of any\ncommercial or financial relationships\nthat could be construed as a potential\nconflict of interest.\nReceived: 31 January 2013; accepted:\n02 August 2013; published online: 06\nSeptember 2013.\nCitation: Collerton D (2013)\nPsychotherapy and brain plasticity.\nFront. Psychol. 4:548. doi: 10.3389/fpsyg.\n2013.00548\nThis article was submitted to\nConsciousness Research, a section of\nthe journal Frontiers in Psychology.\nCopyright \u00a9 2013 Collerton. This is\nan open-access article distributed under\nthe terms of the Creative Commons\nAttribution License (CC BY). The use,\ndistribution or reproduction in other\nforums is permitted, provided the origi-\nnal author(s) or licensor are credited and\nthat the original publication in this jour-\nnal is cited, in accordance with accepted\nacademic practice. No use, distribution\nor reproduction is permitted which does\nnot comply with these terms.\nwww.frontiersin.org September 2013 | Volume 4 | Article 548 | 5\n", "identifiers": ["10.3389/fpsyg.2013.00548"], "publisher": "Frontiers Media SA", "relations": ["http://dx.doi.org/10.3389/fpsyg.2013.00548"], "repositories": [{"id": "2614", "openDoarId": 0, "name": "Frontiers - Publisher Connector", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 1, "metadataAdded": 1494862869000, "metadataUpdated": 1494862869000, "depositedDate": null}, "subjects": ["journal-article"], "title": "Psychotherapy and brain plasticity", "topics": [], "types": [], "year": 2013, "fulltextIdentifier": "https://core.ac.uk/download/pdf/82854584.pdf", "doi": "10.3389/fpsyg.2013.00548", "downloadUrl": "https://core.ac.uk/download/pdf/82854584.pdf"}}
{"status": "OK", "data": {"id": "55853555", "authors": ["Vanoutrive, Thomas", "Van De Vijver, Elien", "Van Malderen, Laurent", "Jourquin, Bart", "Thomas, Isabelle", "Verhetsel, Ann", "Witlox, Frank"], "citations": [], "contributors": [], "datePublished": "2012", "fullText": " Published as: Vanoutrive, T., E. Van de Vijver, L. Van Malderen, B. \nJourquin, I. Thomas, A. Verhetsel & F. Witlox (2012) \u201cWhat determines \ncarpooling to workplaces in Belgium: location, organisation, or promotion\u201d. \nJournal of Transport Geography. Vol. 22 (1), pp. 77-86. \n \n \nWHAT DETERMINES CARPOOLING TO WORKPLACES IN BELGIUM: LOCATION, \nORGANISATION, OR PROMOTION? \n \n1 Introduction \nCarpooling is one of the many travel alternatives promoted by transport policies to reduce the \namount of vehicles on the road. It was promoted during World War II to deal with oil and rubber \nshortages and during the oil crisis of the 1970s (Ferguson, 1997b; 2000; Gilbert and Perl, 2008). \nMore recently, carpooling was also advocated during the 2008 Olympics in Beijing as a \nresponse to driving restrictions (Wang, 2011). Nowadays, carpooling is promoted by mobility \nmanagement policies to put more emphasis on the issue of sustainable transport. The main \ntargets here are a reduction of transport-related pollution (PM10, NOx and CO2),  noise nuisance \nreduction and a decrease of congestion levels. These sustainable mobility policies are called \nmobility management or travel/transportation demand management (TDM) to stress that the \nfocus is not on infrastructure supply but on managing the demand-side, i.e. using the transport \nsystem in the most optimal way to fulfil our lifestyle needs (Fr\u00e4ndberg and Vilhelmson, 2010; \nLyons and Urry, 2005).  \n \nUsually, commuting traffic is a major topic in mobility management schemes since most people \ncommute to already congested urban areas during the peak travel period. The involvement of \nemployers in transport policy is a logical consequence of this focus on commuting. Employers \nhave a privileged relationship with their employees and are therefore regularly used as \nintermediaries between government and individual travellers (DeHart-Davis and Guensler, \n2005; Ferguson, 1997b; 2007). Furthermore, employers are also more efficiently organized to \nimplement mobility management measures such as parking restrictions, allowances and bicycle \nfacilities. Finally, involving the private sector also reduces the burden of transport policies on the \npublic budget (Cairns et al., 2008; Roby, 2010; Rye, 2002).  \n \n Despite the policy focus on employers, commuting research traditionally focuses on individual \ncommuters or on aggregated spatial units (e.g. municipalities). Standard mode choice research \ntakes the individual or the household as the unit of observation since individual and household \ncharacteristics determine the choice process (Van Acker et al., 2010). The advantage of \naggregating individuals in geographical areas on the other hand, is that the effects at higher \ngeographical scales are understood. Indeed, one commuter does not cause a traffic jam; it is \nthe concentration of commuters in particular places and at particular times of the day that can \nbe considered as the main causes of congestion and pollution. We acknowledge the merits of \nboth the individual and the area-wide perspectives. However, we argue that a workplace \nperspective might enrich transport research for a number of reasons. First, employers are used \nas intermediaries in mobility management strategies and the set of available carpool incentives \ndiffers between workplaces. Second, workplaces are physical locations with specific \ncharacteristics, even within a particular geographical area considerable differences in \naccessibility levels may exist between workplaces. Third, the workplace is more than a physical \nenvironment; it is also a social environment. Accordingly, what people at your workplace think \nand do (the subjective norm, corporate culture) influences your travel behaviour (Bonham and \nKoth, 2010; Heinen et al., 2011; McDonald, 2007; Van Acker et al., 2011). \n \nThe aim of the current study is to explain the differences in shares of carpooling among \nemployees at large workplaces in Belgium. In contrast to studies like Habib et al. (2011) and \nCanning et al. (2010), which analysed respectively one and six employer-led carpooling \nschemes with more detail, we employ a large dataset containing several thousands of \nworkplaces. Although our main focus is on workplaces, a multilevel perspective is used \n(Wegener, 2011). We conceptualise workplaces as meaningful units since they are physical and \nsocial environments, but we take into account that on a lower level, the behaviour and \ncharacteristics of individual actors matter (Whitmeyer, 1994). Therefore, we make use of \nCensus data to provide the socio-demographics of carpooling in Belgium and we refer to \nindividual characteristics while discussing differences between workplaces. By aggregating \nworkplaces using on the one hand, activity sectors, and on the other hand, municipalities, we \nadd levels of analysis above the workplace level. This multilevel approach implies that we can \nput the importance of workplaces into perspective. \n \nThe paper is organised as follows. In the next section, an overview of the characteristics and \ndeterminants of carpooling is given. Section 3 introduces the data employed in this paper. Next, \n an exploratory analysis is carried out (Section 4). Two perspectives were used; the first is a \nspatial one, as we assume that the modal split at workplaces is influenced by the context \n(location). Besides this exploratory spatial data analysis (ESDA), we explore the differences \nbetween activity sectors. In Section 5 the methodological framework is presented in which we \nuse a multilevel regression model to explain the differences in the popularity of carpooling \nbetween workplaces. Section 6 reports the results of this model. Finally, we end with a \ndiscussion (Section 7) and a conclusion (Section 8). \n \n2. Definition and determinants of carpooling \n2.1 Definition \nOne single definition of carpooling does not exist. Furthermore, the terms carpooling, \nridesharing and car-sharing may or may not be used interchangeably. In the broadest sense of \nthe word, \u2018ridesharing exists when two or more trips are executed simultaneously, in a single \nvehicle.\u2019 (Morency, 2007, p. 240). We prefer to use the term carpooling since it stresses the \nformation of a pool, i.e. a relatively stable arrangement. We will not use the term car-sharing \nsince this is regularly understood as a service in which a car can be booked by persons who \nonly occasionally need a \u2018rental\u2019 car for e.g. their weekly trip to the supermarket.   \n \nIn the literature most authors distinguish between household-based and non-household-based \ncarpools which are also called internal and external carpools respectively (Buliung et al., 2010; \nCorreia and Viegas, 2011; Ferguson, 1997a; Morency, 2007; Teal, 1987). This distinction is \nrelevant for two reasons. First, members of the same household have their trip origin in \ncommon; as a result, no time is lost for picking up a passenger. Second, the level of trust is high \nbetween members of the same household and this is considered to be important in the \nformation of carpool clubs. Therefore, Correia and Viegas (2011) classify relations between \ncarpool members on the basis of the level of trust which is assumed to be higher between \nmembers of the same household than between (in decreasing order of trust) friends, colleagues \nand unrelated persons.  \n \nSome authors classify carpool trips on the basis of the types of matching between origins and \ndestinations (Morency, 2007; Rietveld et al., 1999). The most simple carpool structure \nencompasses that both driver and rider(s) have their origin and destination in common. If origins \nand/or destinations are not the same, more complex structures appear. Furthermore, carpool \nmembers can meet at a carpool parking at an intermediate location. All this makes of carpooling \n a complex and hybrid concept. Accordingly, it is no coincidence that the two data sources used \nin this paper employ a different definition of carpooling.  \n \n2.2 Determinants of carpooling \nMost studies do not detect strong correlations between socio-demographics and carpooling \npropensity (Buliung et al., 2010; Canning et al., 2010; Ferguson, 1997a; Teal, 1987). \nNevertheless, some general patterns appear. Lower income classes seem to be associated with \na higher propensity to carpool, for the most part in the form of internal carpools. The income \nvariable is associated with vehicle ownership and auto availability. Furthermore, members of \nmultiple worker households carpool more since match-making is easier within the household \nand car availability is limited. More educated employees carpool less, a factor which also relates \nto income. Unsurprisingly, females with young children carpool less since their commuting trips \nare often more complex as they have to drop of and/or pick up children at school or nursery.  \n \nPsychological barriers, attitudes and perceptions seem to have a larger influence on the \ndecision to carpool than socio-demographics. Drivers dislike a delegation of control and \npositively experience their solitary personal space (Gardner and Abraham, 2007). Privacy \nissues and the fear to ride with strangers thus limit the potential of carpooling, although a \nminority positively values the \u2018more sociable travel\u2019. However, financial motives and \nenvironmental concern are considered more important attitudinal factors than the social aspect \n(Canning et al., 2010). \n \nThe literature states that carpool commuting trips are generally longer than the journeys of \nsingle occupant vehicle (SOV) drivers (Ferguson, 1997a). However, the relation between \ndistance and carpooling is multifaceted. First, a driver often needs to make a detour to pick-up \nor drop-off the passenger. This extra travel is also known as circuity (Shoup, 1997, p. 205), and \nRietveld et al. (1999) estimate a travel time increase of 17% compared with solo driving (based \non a limited sample). This pick-up/drop-off delay and extra travel and waiting time make \ncarpooling less suitable for short distances. Second, the savings made by sharing travel costs \nincrease with distance which makes carpooling more attractive for longer trips. Third, \u2018pool \ngeography\u2019 is related to distance. Finding a carpool partner with the same origin and destination \nzone may be difficult, especially in low-density areas (Tsao and Lin, 1999) and at larger \ndistances from the destination. As a result, Buliung et al. (2010) note that there is a threshold \ndistance above which carpooling is less likely to occur, besides the positive relationship \n between distance and carpool propensity. They also stress the importance of the pool-size \neffect, which is present both at the origin and destination side of the trip. The spatial clustering \nof commuters at the home-end is a crucial factor in the formation of carpools, but also firm size \nmatters since a larger pool of employees within the same work environment increases the \nnumber of potential carpool partners. Similar work schedules and higher levels of trust between \ncolleagues (Correia and Viegas, 2011) further increase the potential of workplaces as matching \nplaces.  \n \nConsidering spatial structure, the more congested downtown areas, associated with a high \ntransit access, less parking availability and higher parking costs, are stronger correlated with a \nhigher use of SOV alternatives (Hwang and Giuliano, 1990). However, when public transport is \nof good quality, carpooling is less attractive since congestion and parking scarcity remain \nobstacles for car commuting to city centres. Therefore, Teal (1987, p.211) concludes that \n\u2018carpooling picks up the slack for transit in environments where the latter is of poor quality\u2019. \n \n2.3 The promotion of carpooling \nA particular set of carpooling determinants are the (dis)incentives present in mobility \nmanagement schemes which aim to increase the popularity of carpooling. The rationale behind \nthe promotion of carpooling is that every carpooling employee implies one car less on the road. \nThe quoted benefits of carpooling are self-evident: driving costs may be shared (drivers may get \ncompensation, because passengers pay a part of the commuting cost and still enjoy the comfort \nof a car), and commuters are not dependent on schedules and/or public transport networks. \nHowever, the majority of workers does not carpool, the advantages of carpooling are most of \nthe times not strong enough to entice commuters to give up the comfort of driving alone \n(Comsis Corporation, 1993; Hwang and Giuliano, 1990; Kingham et al., 2001; Tsao and Lin, \n1999). Therefore, a variety of instruments are used to promote carpooling. \n \nGiven the low cost of most measures that encourage carpooling, employers indicate them as \nacceptable, with the exception of parking restrictions (Rye, 1999a; 1999b). Most research \nclassifies the guaranteed ride home as effective (Correia and Viegas, 2011; Giuliano et al., \n1993; Kingham et al., 2001; Menczer, 2007; Rye, 1999a, 1999b). However, Hwang and Giuliano \n(1990) identified it as less effective, together with other \u2018carrots\u2019 like preferential parking, \nalternative work hours, a matching service, and marketing. In what follows, we put the \neffectiveness of measures into perspective. As Canning et al. (2010) point out, the success of \n preferential parking depends on the actual parking pressure at a site. Regarding work \nschedules, most studies state that a regular work schedule facilitates finding carpool-partners \nwith the same working hours and that flexible work schedules positively influence public \ntransport patronage. Flexitime and the promotion of carpool are then seen as conflicting mobility \nmanagement measures (Buliung et al., 2010; Huang et al., 2000). However, although a flexible \nwork arrangement reduces the probability that carpooling enters the choice set of a commuter, \nonce the commuter perceives carpooling as a viable option, there is a positive effect of flexitime \non carpooling (Habib et al., 2011). Obviously, the development of ICT brings along more \nadvanced (on-line) carpool-matching tools (Buliung et al., 2010; Canning et al., 2010). Some of \nthese tools try to overcome the poor schedule flexibility and offer a dynamic ride-matching \nservice. This involves the creation of a large pool of potential carpoolers which are matched in a \nflexible way without the necessity to carpool every day and each time with the same person(s) \n(Correia and Viegas, 2011). It is yet unclear whether these casual carpooling strategies can \nsubstantially increase the number of carpool trips. A last \u2018soft\u2019 measure to promote carpooling is \nmarketing. This, and information provision in general, is needed to inform the audience about \nthe carpool measures that exist. Although information provision is considered as cheap, a tailor-\nmade approach might require significant budgets, especially if an in-house carpool coordinator \nis present (Buliung et al., 2010). \n \nHwang and Giuliano (1990) indicate that \u2018sticks\u2019 like parking charges and restrictions, and \ntransport allowances as more effective than \u2018carrots\u2019 like marketing, preferential parking and \nsetting up a matching service. Also Kingham et al. (2001) point to financial incentives as \nmeasures with a high potential. The potential attributed to parking charges and restrictions is in \nline with the general finding that \u2018sticks proved to have a generally greater influence on stated \nmode choice than the carrots\u2019 (O'Fallon et al., 2004, p. 28). However, as is clear from the \ndiscussion above, there is no consensus about the effectiveness of measures promoting \ncarpooling.  \n \n3 Data \nThe aim of the present paper is to analyse carpooling in Belgium using a workplace perspective. \nTherefore, we need data that is national in scope and that provides data on workplace-related \nfactors like work schedules and mobility management measures. As main source of data we \nemploy the Belgian database home-to-work-travel (HTWT) 2005. This database is the result of \na mandatory questionnaire about the home to work displacements and the mobility \n management measures at large workplaces in Belgium (Vanoutrive et al., 2010). A large \nworkplace is defined as a site containing at least 30 employees of a company with at least 100 \nemployees. The database HTWT 2005 contains 7460 work sites with at least 30 employees \nwhich employ 1 342 119 employees in total. In addition, we will also use data stemming from \nthe Belgian 2001 Census for illustrative and comparative purposes (Verhetsel et al., 2009). This \ncensus does not contain information on mobility management initiatives at workplaces and other \norganisational characteristics. Nevertheless, it helps to complete the picture of carpooling in \nBelgium. \n \nPlease note that the definition of carpooling differs between the two data sources. The \nquestionnaire HTWT only collects information on external carpools (without family members) \nwhile the 2001 Census includes household-based (internal) carpools. For an employer it can be \nhard to distinguish between household- and non-household-based carpools. For this and other \nreasons the quality of figures reported by employers is likely to be variable. However, the \nobligatory discussion of the questionnaire with the representatives of the employees (unions) in \nthe works council, acts as a kind of quality check, together with some extra quality checks \ncarried out by the Federal Public Service Mobility and Transport.  \n \n4 Exploratory data analysis \nThe 2001 (i.e. most recent) Belgian Census produces some basic facts about carpooling in \nBelgium (Table 1). In Belgium, 6.1% (203 024 commuters) of the workforce commutes as a car \npassenger, 66.1% as a car driver, 6.5% uses the bicycle, 6.0% is a rail user, 6.2% uses regional \npublic transport (bus, tram or metro) while other modes have a share below 5.0%. If we assume \nthat each passenger shares a car with one driver (cf. Buliung et al., 2010), the share of carpool \nis 12.3%. This figure is comparable to levels observed in the US and Canada (Buliung et al., \n2010), but a major difference with the US is the higher share of public transport in Belgium \n(12.2%). This can be explained by the more extensive public transport systems in Belgium and \nother parts of Europe (Buehler, 2011; Vanoutrive et al., 2012). The Belgian Census revealed \nthat women carpool more than men. In contrast with the general belief that carpooling suits \nbetter with longer commutes, (solo) car drivers and rail commuters travel more kilometres to \ntheir workplace. The high average commuting distance in Belgium may be an explanation for \nthis fact. A clear spatial pattern appears when mapping the share of carpooling residents per \nmunicipality (Figure 1, Verhetsel et al., 2009). Commuters who live in the northwest of Belgium \nor in the Brussels Capital Region carpool less than average. These are the areas with lower \n average commuting distances (Boussauw and Witlox, 2009). Carpooling is more popular in the \nsouth of the country, especially on the former industrial east-west axis (which contains cities like \nCharleroi (C) and Li\u00e8ge (L)), and in the east. Note that south of this axis, Belgium is less \ndensely populated. However, our focus is not on the place of residence, but on the work end of \nthe commuting trip. Therefore, we now turn to an analysis of the workplace data in the database \nHTWT. Figure 1 shows a clear spatial pattern of carpooling, measured at the place of residence. \nTherefore, we start the examination of the database HTWT with an exploratory spatial data \nanalysis (ESDA), to investigate whether a spatial structure at the work end of the commute is \npresent or not.  \n \nInsert Table 1 here \nTable 1: Basic figures on carpooling in the 2001 Census \n         \npercentage of commuters that carpool (passenger) 6.1 \npercentage of female carpooling commuters   59 \naverage commuting distance (single trip, in km) \n car as passenger    17.6 \n car driver     20.1 \n rail      46.6 \n overall      19.1 \naverage commuting time (towards workplace, in minutes) \n car as passenger    26 \n car driver     27 \n rail      66 \n overall      29  \nSource: Verhetsel et al., 2009 \n \nInsert Figure 1 here \n \n \n \n \n \n \n \n \n \n \n \n  \nFigure 1: Share of carpooling in commuting per municipality of residence (Data source: Belgian 2001 \nCensus; cartography by the authors; equal group sizes) \nMajor cities: A: Antwerp; B: Brussels; C: Charleroi; G: Ghent; L: Li\u00e8ge \n \n4.1 Exploratory spatial data analysis (ESDA) \nThe accessibility of a workplace is a function of its location. Since accessibility (and thus \nlocation) is a mode choice determinant, we might expect a spatial pattern when mapping the \nshares of carpooling in Belgium. Therefore, we start the analysis with an exploratory spatial data \nanalysis (ESDA). The variable of interest is the percentage of employees at a worksite which \ncarpools to make the daily commute. According to the database HTWT 2005, on average 3.31% \nof the employees on a worksite are carpooling. To analyse the spatial pattern of carpooling \namong Belgian workplaces, the workplaces in the database HTWT are aggregated at the \nmunicipality level. One of the main issues in ESDA is the measurement of spatial \nautocorrelation. Observations are spatially autocorrelated if the values of neighbouring \nmunicipalities are more similar than those of more distant observations. The most common \nstatistic to measure the overall spatial autocorrelation is the Moran\u2019s I, which ranges between -1 \n(negative spatial autocorrelation) and +1 (positive spatial autocorrelation) (Getis, 2007; \nLegendre, 1993). Using the data of the 2001 Census (Figure 1), the Moran\u2019s I statistic confirms \nthe presence of a clear spatial pattern at the place of residence (0.55 taking into account all \nmunicipalities within a range of 20km, and 0.70 when using the four nearest municipalities). \nHowever, using the database HTWT, a low value for the Moran\u2019s I statistic indicates the \nabsence of spatial autocorrelation (0.056, taking all municipalities within a range of 20km as \nneighbours). Note that there are significant differences between the two datasets (definition of \ncarpooling, place of residence versus workplace, all commuters versus only large companies). \nThe map with the average share of carpooling per municipality (Figure 2) confirms the absence \nof a clear spatial pattern. Besides statistics which measure the overall spatial autocorrelation, \nlocal indicators of spatial association (LISA) exist (Anselin, 1995). A LISA indicates for each \nobservation how different its value is from neighbouring observations. Based on this LISA, \nspatial clusters can be defined as shown in Figure 3. On the LISA map (Figure 3), a cluster of \nmunicipalities with low carpool shares is situated in the centre of the country and carpooling \nseems more popular in the east and in some other more peripheral locations. Note that these \nmaps are based on data about the destination of the home to work trip. There is a notable \ncontrast with the map of the share of carpooling per municipality of residence in Figure 1.  \n \n  \n \n \n \n \n \n \n \n \nInsert Figure 2 here \n \n \n \n \n \n \n \n \n \nFigure 2: Map of carpool share per municipality (work location) \nSource: database HTWT 2005 (cartography by the authors; natural breaks) \n \nInsert Figure 3 here \n \n \n \n \n \n \n \n \n \n \nFigure 3: LISA map of carpool share per municipality (work location) \nSource: database HTWT 2005; Software: Geoda (Anselin, 2005) and ArcGIS (ESRI) \nLISA statistic takes all municipalities into account within a range of 30km \n(cartography by the authors) \n \n 4.2 The promotion of carpooling \nTable 2 shows the carpool-related mobility management measures which could be indicated in \nthe questionnaire HTWT. Figures are given for the whole sample (n=7460) as well as for a \nsubsample which only contains workplaces where at least one employee carpools (n=3353). In \nboth samples, the majority of the workplaces do not report carpool-related mobility management \nmeasures. The organisation of carpooling, connecting to a central carpool database and the \ndelivery of information about carpooling, are relatively the most common measures, while \npreferential parking for carpooling employees and the organisation of a guaranteed ride home \nare rare.  \n \nInsert Table 2 here \nTable 2: Percentage of worksites where carpool-promoting measures are taken \n        all sites  sites with CP- \n          employees \nCarpool promoting measure     (n = 7460) (n = 3353)  \n\u2018Organising a car pool on the site\u2019    5.2  6.5  \n\u2018Connecting to a central database\u2019    4.6  5.7  \n\u2018Dispersion of information about carpooling\u2019   4.2  5.0  \n\u2018Reserved parking places for carpooling employees\u2019  1.9  2.4  \n\u2018Guaranteed ride home for carpool passengers   1.6  1.9  \nin case of unpredicted circumstances\u2019       \n\u2018No carpool measures\u2019      86.6  83.9   \nSource: questionnaire HTWT 2005 \n \n4.3 Activity sector \nIn the previous paragraphs we focused on the location-related differences between workplaces \nand on the measures taken to promote carpooling. However, workplaces also differ for \norganisational reasons. Therefore, the workplaces were classified in 19 activity sectors by \nmeans of the Crossroads Bank for Enterprises (CBE) code of every company in the database \nHTWT. With this code we identified the economic sector (Nacebel 2003, based on the European \nNACE Statistical Classification of Economic Activities in the European Community) using the \nBELFirst database. Almost half of the sites (3445) could not be linked to an economic sector on \nthe basis of their CBE code. We explored the data and it turned out that these sites are mostly \nfound in health, education, public transport and other government sectors. We classified these \nworkplaces semi-automatically on the basis of keywords in the names of the institutions. This \ncould be done since names of institutions in the public sector often contain official abbreviations. \nWe inspected the database and could not detect misclassified workplaces. In general, 13% of \nthe variance in carpooling at Belgian worksites can be attributed to the activity sector. We will \n explain the estimation of this variance partitioning in Section 6. Table 3 shows the different \neconomic sectors together with their average number of carpool-oriented mobility management \nmeasures (see Table 2) and the share of carpool in the modal split. The construction sector is \ndefinitely the \u2018number one\u2019 in carpooling. Other carpool-oriented sectors are manufacturing, \nelectricity, gas and water, and transport warehousing and communication. Low levels of \ncarpooling are found for universities, post, public transport companies and in the health sector. \nThe financial sector and universities take most measures to promote carpooling. Police, post \nand public transport companies take the least measures to promote carpooling.  \n \nInsert Table 3 here \n \nTable 3: Average percentage of carpooling employees at a worksite, average number of carpool-\nmeasures and number of worksites per economic sector \n    average % average CP- \neconomic sector  carpoolers measures # worksites # employees \n    n = n = n = n = n = n = n =     n =  \n    7460 3353 7460 3353 7460 3353 7460     3353  \nConstruction   10.34 24.28 0.23 0.28 108 46  13927     7681 \nManufacturing   6.82 9.81 0.23 0.23 1092 759      315246  248855 \nTransport, warehousing  6.53 11.56 0.25 0.23 280 158 58232   36117 \nand communication \nElectricity, gas and water 5.92 9.67 0.16 0.17 116 71  16786   12433  \nOther community, social 4.71 9.09 0.31 0.27 199 103 30048   19734 \nand personal services \nPrimary sector   3.75 6.01 0.29 0.47 24 15    3393     2262  \nPublic administration, defence, 3.18 7.09 0.12 0.13 752 337      142001   87361 \nsocial security insurance \nReal estate, renting  3.00 7.17 0.24 0.26 344 144 56457   31264 \nand producer services \nLocal government  2.88 5.56 0.07 0.08 664 344 93511   56937 \nNon profit   2.77 5.53 0.28 0.36 192 96 31079   19994 \nFinance   2.60 4.28 0.92 0.96 184 112 58933   48165 \nEducation   2.16 6.89 0.08 0.11 917 288 87610   32277 \nHotels and restaurants  2.12 7.93 0.12 0.26 86 23   9124     4163 \nWholesale and retail;  1.93 6.58 0.14 0.19 877 257  97912   43888 \nrepair of motor vehicles  \nand consumer goods \nPolice    1.76 3.80 0.02 0.05 95 44 13777     8433 \nHealth    1.58 3.57 0.13 0.15 653 290      185552 131993 \nPublic Transport Companies 1.57 4.17 0.02 0.02 231 87 42250   27400 \nPost    1.51 5.79 0.03 0.04 306 80 29399   15000 \nUniversities and other  1.05 3.59 0.40 0.62 340 99 56882   26874 \nhigher education institutions \n \nTotal    3.31 7.36 0.18 0.22 7460 3353  1342119  860831 \nSource: questionnaire HTWT 2005 \nNotes. CP-measures: these are the measures shown in Table 2;  \nn = 7460: full sample; n = 3353: only workplaces where at least one employee carpools \n  \n \n5 Multilevel regression analysis \nTo complement the exploratory analysis, we apply multilevel regression analysis to better \nunderstand the differences in carpooling among workplaces in Belgium. Since the activity sector \nto which a workplace belongs appeared to be relevant, we apply multilevel modelling. This type \nof regression models explicitly takes into account that observations are nested in groups, while \nstandard regression analysis assumes the independence of observations. More concretely, this \nindependence assumption is violated if workplaces in the same economic sector or workplaces \nin the same area are not independent. The main aim of multilevel regression models is to deal \nwith this grouping of observations (Goldstein, 1995; Hox, 2002; Luke, 2004). In fact, the \nregression model estimates a separate regression line for every group (i.e. economic sector). If \nthese regression lines run in parallel, the model is called a random intercept model (each line \nhas a different intercept). If the regression lines have different slopes (and different intercepts), it \nis referred to as a random slope model. As in standard regression analysis, the linear multilevel \nregression model can be transformed in a logistic regression to analyse a binary dependent \nvariable using a logit function.  \n \nMultilevel modelling has the advantage of getting a better understanding and more clear \ninterpretation of the effects of higher levels (by estimating and reporting random effects). \nFurthermore, standard regression analysis ignores the grouping of data and this can cause \nunderestimated standard errors of regression coefficients (Goldstein, 1995; Hox, 2002; Maas \nand Hox, 2004; Rasbash et al., 2005; Schwanen et al., 2004). The main disadvantage is that \nmodels become more complex. As a consequence, diagnostics can be more complicated as \nwell. The comparison of different models is often used to evaluate a model. In what follows, we \nwill refer to the so-called empty model (model A), i.e. a model with a multilevel structure (an \nextra error term) but without any independent variables. The empty model is then compared \nwith the full model (model B), i.e. the multilevel model with the full set of independent variables. \n \nFollowing independent variables were included in the multilevel analysis. As measure for the \nsize of a site, the number of employees is used. The percentage of staff with a fixed (regular) \nwork schedule is also considered. To account for site accessibility, three variables are used. \nFirst, the accessibility by car is measured as the potential number of people that can reach a \ngiven municipality and is calculated by Vandenbulcke et al. (2007, 2009). Rail accessibility is \nseen as the inverse of the sum of waiting and walking time between workplace and railway \n station, the calculation method and an example are given in Vanoutrive et al. (2012). Besides \nthese activity-based accessibility indicators, congestion is introduced as an infrastructure-based \naccessibility indicator. The congestion variable is a dummy with a value of 1 if an employer \nindicated in the HTWT survey that the site suffers from road congestion, and 0 if not. Parking \navailability is defined as the number of parking places per employee; the maximum is set to 1 in \norder to avoid the effect of large customer parking spaces of shops and the like.  \n \nAll non dummy variables are manipulated. To reduce the skewness of the variables, the \nlogarithm is taken, except for the parking variable. Next, all independent variables are \nstandardized using z-scores to make results comparable and to centre the variables. This \ncentring is relevant since a random slope model is used (Luke, 2004). Finally, the standardized \nlogarithm of the number of employees is centred around the value which corresponds to 200 \nemployees. This number falls between the average (257) and the median (139) of the dataset \nused in the random slope model (model B). In more than half of the worksites (4107 of 7460) no \nemployees carpool. These worksites are excluded from the main analysis (models A and B) to \navoid biases caused by zero inflated data. Nevertheless, a logistic regression (model C) is \nmade to show the difference between sites without (0) and the sites with (1) carpooling \nemployees.  \n \n6 Results  \nTable 4 lists the results of three multilevel models which used the Belgian database HTWT \n2005. These models are estimated using MLwiN (Rasbash et al., 2009). First, the empty model \n(Model A), i.e. a model with a multilevel structure but without independent variables, indicates \nthat 13% of the variance in carpooling at Belgian worksites can be attributed to the economic \nsector level, while the remaining 87% is variance between worksites, which also covers the \nvariation between individual employees. In the concrete, 13% is the portion of the variance \nattributed to the economic sector (0.027 in model A), relative to the total variance (0.027 + \n0.188). \n \nThe second model (referred to as model B) is the random slope model. The results of the fixed \npart of the model can be interpreted in the same way as results of a standard regression model. \nIn general, small worksites have high shares of carpooling, as have sites with more parking \nspace per employee. More employees with a regular work schedule positively influence the \nsuccess of carpooling. The lower the accessibility by car and by train, the more popular \n carpooling is, while the effect of congestion is less clear. Considering the mobility management \nmeasures, only the guaranteed ride home has a positive, significant relation with carpooling and \nthe result for information delivery even has a negative sign, while the other variables are not \nsignificant at the 95% confidence interval. The random part of the model indicates that there are \nsignificant differences in carpool levels between activity sectors and that these differences are \nlarger between small workplaces than between large workplaces. Figure 4 shows the intercepts \nfor the different economic sectors (the level 2 residuals). The black dots show the differences \nbetween activity sectors after controlling for the variables included in model B while the grey \ndots are the reference (model A). On average, model B explains 11% of the variance between \nworkplaces (0.027-0.024/0.027, comparing model A with B).  \n \nThird, a logistic regression model (model C, Table 4) examines the difference between \nworksites where nobody carpools (0) and sites where at least one employee is carpooling (1). \nThe results of this model are informative with respect to the omitted observations in the other \ntwo models. Model C shows that the probability of at least one carpooling employee is higher if \nmore employees are working on the site, which is quite evident. This result supports the \nliterature which states that more employees imply more possible carpool partners (pool-size \neffect). Nevertheless, at sites where at least one employee carpools, there is a negative size-\neffect (model B). One potential explanation is the fact that our dataset contains only larger sites \n(>30 employees), as a consequence, the majority of workplaces employ more than 100 \nemployees. Presumably, the pool-size effect would be positive if also small workplaces were \nincluded. Furthermore, large concentrations of workers have generally better access to public \ntransport and since the rail accessibility variable does not cover bus services and the like, the \nnegative size-effect can be explained by stronger competition between carpooling and collective \ntransport at larger sites. Finally, the level of trust between employees may be lower at large \nworkplaces consisting of different separate subunits. \n \nInsert Table 4 here \nTable 4: Results of the multilevel regression analysis \n     A    B   C \n     empty model  random slope model logit model  \nrandom part:    est. s.e.   est. s.e.   est. s.e. \nlevel 2 constant/constant  0.027 0.009   0.024 0.009   0.132 0.050 \nlevel 2 employees/const     0.006 0.003   \nlevel 2 employees/employees     0.003 0.001  \nlevel 1 constant/constant  0.188 0.005   0.173 0.004  \n \n fixed part:    \nconstant    0.604 0.039   0.560 0.038   0.259 0.095 \nemployees       -0.087 0.016   0.733 0.029 \nregular        0.041 0.008   0.243 0.028 \nparking        0.010 0.008   0.026 0.027 \ncar accessibility      -0.048 0.008  -0.048 0.028 \nrail accessibility      -0.022 0.008  -0.022 0.028 \ncongestion       0.030 0.016   0.084 0.057 \nmobility management:  \norganisation of carpool     0.033 0.033   0.197 0.126 \n carpool database     0.010 0.037   0.422 0.137 \n preferential parking     0.020 0.051  -0.196 0.192 \n guaranteed ride home     0.122 0.054   0.010 0.208 \n information delivery    -0.083 0.038  -0.024 0.140 \n \n-2 loglikelihood    3971.96   3706.79   \nlevel 2 (activity sector) n =  19    19    19 \nlevel 1 (workplace) n =   3353    3353    7460   \nNotes. Dependent variable Model A and B: log(%carpooling employees at a site); Dependent variable \nModel C: dummy variable: 0 if #employees that carpool=0; 1 if #employees that carpool>0; italics: not \nsignificant \nCharacteristics of raw data: number of employees: min.: 30, max.: 6552 (both samples); mean: 256.7 \n(model B), 179.9 (model C); percentage of staff with a regular (fixed) work schedule: min.: 0, max.: 100 \n(both samples); mean: 38.9 (model B), 36.6 (model C); number of parking places per employee: min.: 0, \nmax.: 1 (both samples); mean: 0.51 (model B), 0.49 (model C); congestion (dummy variable): frequency: \n27.4 (model B), 26.1 (model C); mobility management dummy variables: see Table 4; other variables are \nrelative measures. All non-dummy variables are standardised (mean: 0, standard deviation: 1). \n \n \nInsert Figure 4 here \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n \n  \n \nFigure 4: Level 2 (workplace) residuals of model A (grey) and model B (black) together with their 95% \nconfidence intervals \n \nIn the exploratory spatial data analysis (ESDA), both the computed Moran\u2019s I and Figure 3 \nrevealed a (small) spatial effect in carpooling at the municipality level. To test whether this \nspatial effect is a source of unexplained variance, we estimated a two-level empty model of the \nworkplace (level 1) residuals of model B with the municipality as second level. The variance at \nthe municipality level was not significant (t-value: 1.85) and only 1% of the variance could be \nattributed to this level. Hence, the model accounts for the spatial pattern in the data and as a \nconsequence, the unexplained variance cannot be attributed to differences between \nmunicipalities. The result for the car and rail accessibility variables indicate that carpooling is \nmore abundant in the more peripheral areas of Belgium, as already indicated by the exploratory \nmap in Figure 3. This confirms the literature which suggests that carpooling is particularly \nsuccessful in places with a lack of public transport services.  \n \n7 Discussion \nGiven the potential competition between carpooling and public transport, the promotion of \ncarpooling in public transport-rich areas can be a counterproductive mobility management \nstrategy. Therefore, we follow the line of reasoning developed by Wang (2011) who states that \ngovernments should remove unnecessary barriers to carpooling, but that excessive subsidies to \ncarpooling are detrimental to collective welfare since bicycles and public transport produce less \nemissions and use space more efficiently. Unwanted barriers to carpooling are taxation and \ninsurance issues like the uncertainty of being insured while making a detour to pick up a \npassenger or when using a company car. \u2018Innocent\u2019 carpool incentives are on-line ride-matching \nservices, preferential parking and guaranteed ride home services (preferably also applicable to \npublic transport). In contrast, allowances or free parking for carpoolers in areas with high \nparking costs is oversubsidising since public transport is most of the times a viable alternative in \nthese areas. For the same reason, we do not advise the introduction of carpool lanes in Belgian \ncities. Analogously, carpool parkings and park and ride facilities should be carefully planned \nsince they often generate additional traffic and encourage car-oriented land use development \noutside urban areas (Meek et al., 2008; Parkhurst, 2000). If carpooling is heavily promoted at a \nworkplace due to a lack of public transport, one should always check whether this is caused by \nbad land use planning or poorly organized public transport. Workplaces can be located outside \n agglomerations for safety and environmental reasons and carpooling might be the most efficient \nway to reduce levels of SOV driving in such cases. However, the clustering of businesses in e.g. \nchemical industrial parks (Reniers et al., 2010) creates the opportunity to invest in bus services \nas can be observed in the Antwerp port area where large chemical companies have established \nan extensive network of bus routes. When scale effects are absent, carpooling might be \npreferred over collective transport. Also for companies with particular characteristics, there are \nfew alternatives to carpooling to reduce levels of SOV commuting. A noticeable example is the \nconstruction sector where the changing location of work makes of ridesharing the most rational \nway of travelling. Note that in this case it is hard to distinguish between carpooling/ridesharing \nand transport organized by the employer (Meersman et al., 1998). The high levels of carpooling \nin construction, manufacturing and transport (Table 3, Figure 4) indicate that in these sectors, \ncarpooling has the highest potential to reduce the amount of SOV commuters. As discussed \nearlier, the promotion of carpooling may, however, not result in increased urban sprawl or lower \nlevels of public transport or bicycle use. \n \nAlthough transport professionals should be aware of the possible unintended consequences of \ncarpool-oriented measures, their potential impact is still a relevant issue from a policy \nperspective. We included five dummy variables in our regression model which indicated the \npresence of carpool incentives at a workplace. A regression does not assume a causal \nrelationship, nevertheless, the results might be indicative for the impact of the measures. In the \nmain model, we omitted workplaces without carpooling employees which excludes the \npossibility that the absence of carpooling employees causes the absence of carpool measures. \nFurthermore, we controlled for location, accessibility and economic sector effects and only for \nthe guaranteed ride home a positive significant result appeared. Note that this is a cross-\nsectional analysis and that more advanced research designs can more comprehensively \nevaluate the effectiveness of measures. However, the relative low amount of explained variance \nmakes methods like 2SLS less suitable. Also note that the guaranteed ride home is only present \nat 1.9% of the workplaces which makes the extrapolation of its effect less reliable. Some \ncarpool measures might be successful at some sites, but in general, they do not seem to \ncontribute much to the success of carpooling. Measures are often part of general HRM and \ncorporate sustainability strategies and these general strategies often lack the right mix of \nmeasures to tackle the site-specific accessibility problems. Finally, some authors classify the \ninvestigated mobility management measures as less effective and expect more from financial \nincentives and \u2018hard\u2019 parking restrictions. However, we obtained a non significant estimate for \n the parking pressure variable. This lack of significance might be attributed to the correlation \nbetween the parking and the accessibility variables. The number of parking places per \nemployee (parking index) is remarkably lower inside agglomerations (average 0.46) than \noutside (0.58; Belgian agglomeration as defined by Luyten and Van Hecke, 2007). Excluding \nthe two aforementioned accessibility variables, the model generates a positive significant result \nfor the parking variable (0.019; with s.e. 0.008). In our model, parking seems thus a measure for \nagglomerations and density (Chen et al., 2008).  \n \nFurther research could shed more light on the different spatial patterns which are found at the \norigin and the destination side of the commuting trip. The different definitions employed in the \ntwo questionnaires may be one of the causes. The term carpooling covers a broad range of \ntravel arrangements and a deeper insight can be obtained by analysing the differences across \nregions and activity sectors with regard to the different types of carpooling. Furthermore, the \ninclusion of small workplaces would increase the scope of the research. More advanced \nresearch designs could lead to a better understanding of the effectiveness of mobility \nmanagement measures that promote carpooling. One could focus on the interdependencies \nbetween measures that promote different modes since competition between public transport \nand carpooling might occur. Finally, we did not investigate the time dimension in this paper. The \ndramatic decline in carpooling in the US received considerable attention (Ferguson, 1997a), but \nit is yet unclear what the effect would be of expected higher fuel prices (peak oil) (Gilbert and \nPerl, 2008; Wegener, 2011). \n \n8 Conclusion \nBoth governments and employers promote carpooling as a commuting alternative in order to \nreduce the number of single occupant vehicle (SOV) users. This paper takes the workplace as \nprime research unit. The Belgian questionnaire home-to-work-travel (HTWT) proved to be a \nunique source of data generated at the worksite level. We controlled for three factors which \ncould explain the differences in carpool shares at workplaces. A first relevant characteristic of a \nworkplace is its location, which influences its accessibility. Second, organisational factors such \nas work schedules and the activity sector determine the attractiveness of carpooling. To account \nfor the nesting of workplaces in activity sectors, we chose a two-level multilevel model which \nsimultaneously modelled both the worksite and the economic sector levels. Third, employers \npromote carpooling by mobility management measures such as carpool databases, preferential \nparking and emergency ride home services. However, a model which checked for location, \n organisation and promotion, left the majority of the variance in carpooling unexplained. Despite \nthe relative low explanatory power of the model, the analysis gives insight in the distribution of \ncarpooling among workplaces in Belgium. \n \nThe most carpool-oriented sectors are construction and manufacturing and also in the \nwholesale and retail sectors carpool is popular. Carpooling is rather unpopular at universities, in \nthe health sector and in public transport companies, and seems to be an alternative at locations \nwhere rail is no real alternative. Regular work schedules and a smaller number of employees at \na site are positively correlated with a higher share of carpooling employees. For the \neffectiveness of carpool promoting measures no evidence could be found, except for the \nguaranteed ride home. Measures that discourage car use, like parking charges, seem to be \nmore effective than soft carpool-promoting initiatives. Finally, we argue that the promotion of \ncarpooling must not result in more car-oriented development and lower levels of cycling and \nusing public transport . \n \nAcknowledgements \nThis research is conducted within the ADICCT-project (Assessing and Developing Initiatives of \nCompanies to control and reduce Commuter Traffic) which is financed by Belgian Science Policy in the \nScience for a Sustainable Development research programme. We are also grateful to Gr\u00e9gory \nVandenbulcke (UCL) and the FPS Mobility and Transport for delivering data and information. The authors \nare grateful for advice from the anonymous referees.  \n \nReferences \nAnselin, L., 1995. Local Indicators of Spatial Association-LISA. Geographical Analysis 27, 93-115. \nAnselin, L., 2005. Exploring Spatial Data with GeoDaTM : A Workbook. Center for Spatially Integrated \nSocial Science, Urbana. \nBonham, J. and Koth, B., 2010. Universities and the cycling culture. Transportation Research Part D 15, \n94-102. \nBoussauw, K. and Witlox, F., 2009. Introducing a commute-energy performance index for Flanders. \nTransportation Research Part A-Policy and Practice 43, 580-591. \nBuehler, R., 2011. Determinants of transport mode choice: a comparison of Germany and the USA. \nJournal of Transport Geography 19, 644-657. \nBuliung, R.N., Soltys, K., Bui, R., Habel, C., and Lanyon, R., 2010. Catching a ride on the information \nsuper-highway: toward an understanding of internet-based carpool formation and use. Transportation 37, \n849-873. \nCairns, S., Sloman, L., Newson, C., Anable, J., Kirkbride, A., and Goodwin, P., 2008. Smarter choices: \nAssessing the potential to achieve traffic reduction using 'soft measures'. Transport Reviews 28, 593-618. \n Canning, P.E., Hughes, S.J., Hellawell, E.E., Gatersleben, B.C.M., and Fairhead, C.J., 2010. Reasons for \nparticipating in formal employer-led carpool schemes as perceived by their users. Transportation \nPlanning and Technology 33, 733-745. \nChen, C., Gong, H., and Paaswell, R., 2008. Role of the built environment on mode choice decisions: \nadditional evidence on the impact of density. Transportation 35, 285-299. \nComsis Corporation, 1993. A guidance manual for implementing effective employer-based travel demand \nmanagement programs.  The Institute of Transportation Engineers, Washington \nCorreia, G. and Viegas, J.M., 2011. Carpooling and carpool clubs: Clarifying concepts and assessing \nvalue enhancement possibilities through a Stated Preference web survey in Lisbon, Portugal. \nTransportation Research Part A 45, 81-90. \nDeHart-Davis, L. and Guensler, R., 2005. Employers as Mediating Institutions for Public Policy: The Case \nof Commute Options Programs. Policy Studies Journal 33, 675-697. \nFerguson, E., 1997a. The rise and fall of the American carpool: 1970-1990. Transportation 24, 349-376. \nFerguson, E., 1997b. Privatization as choice probability, policy process and program outcome: The case \nof transportation management associations. Transportation Research Part A: Policy and Practice 31, 353-\n364. \nFerguson, E., 2000. Travel Demand Management and Public Policy. Ashgate, Aldershot, UK. \nFerguson, E., 2007. Transportation Management Associations: A Reappraisal. Journal of Public \nTransportation 10, 1-26. \nFr\u00e4ndberg, L. and Vilhelmson, B., 2010. Structuring Sustainable Mobility: A Critical Issue for Geography. \nGeography Compass 4, 106-117. \nGardner, B. and Abraham, C., 2007. What drives car use? A grounded theory analysis of commuters' \nreasons for driving. Transportation Research Part F 10, 187-200. \nGetis, A., 2007. Reflections on spatial autocorrelation. Regional Science and Urban Economics 37, 491-\n496. \nGilbert, R. and Perl, A., 2008. Transport Revolutions - Moving People and Freight without Oil. Earthscan, \nLondon, UK. \nGiuliano, G., Hwang, K., and Wachs, M., 1993. Employee trip reduction in Southern California: First year \nresults. Transportation Research Part A: Policy and Practice 27, 125-137. \nGoldstein, H., 1995. Multilevel Statistical Models. Arnold, London \nHabib, K.M.N., Tian, Y., and Zaman, H., 2011. Modelling commuting mode choice with explicit \nconsideration of carpool in the choice set formation. Transportation 38, 587-604. \nHeinen, E., Maat, K., and van Wee, B., 2011. The role of attitudes toward characteristics of bicycle \ncommuting on the choice to cycle to work over various distances. Transportation Research Part D 16, \n102-109. \nHox, J.J., 2002. Multilevel Analysis: Techniques and Applications. Erlbaum, Mahwah, NJ. \n Huang, H.-J., Yang, H., and Bell, M.G.H., 2000. The models and economics of carpools. Annals of \nRegional Science 34, 55-68. \nHwang, K. and Giuliano, G., 1990. The Determinants of Ridesharing: Literature Review. UCTC, Berkeley, \nCalifornia. \nKingham, S., Dickinson, J., and Copsey, S., 2001. Travelling to work: will people move out of their cars. \nTransport Policy 8, 151-160. \nLegendre, P., 1993. Spatial Autocorrelation: Trouble or New Paradigm? Ecology 74, 1659-1673. \nLuke, D.A., 2004. Multilevel Modeling. Sage, Thousand Oaks. \nLuyten, S. and Van Hecke, E., 2007. De Belgische Stadsgewesten 2001. Statistics Belgium Working \nPaper, Brussels. \nLyons, G. and Urry, J., 2005. Travel time use in the information age. Transportation Research Part A 39, \n257-276. \nMaas, C.J.M. and Hox, J.J., 2004. Robustness issues in multilevel regression analysis. Statistica \nNeerlandica 58, 127-137. \nMcDonald, N.C., 2007. Travel and the social environment: Evidence from Alameda County, California. \nTransportation Research Part D 12, 53-63. \nMeek, S., Ison, S., and Enoch, M., 2008. Role of Bus-Based Park and Ride in the UK: A Temporal and \nEvaluative Review. Transport Reviews 28, 781-803. \nMeersman, H., Pauwels, T., and Van de Voorde, E., 1998. Bouwen aan een duurzame mobiliteit - Het \nwoon-werkverkeer in de Belgische bouwsector: situering, analyse en strategische alternatieven. \nTijdschrift Vervoerswetenschap 34, 361-377. \nMenczer, W.B., 2007. Guaranteed Ride Home Programs: A Study of Program Characteristics, Utilization, \nand Cost. Journal of Public Transportation 10, 131-149. \nMorency, C., 2007. The ambivalence of ridesharing. Transportation 34, 239-253. \nO'Fallon, C., Sullivan, C., and Hensher, D.A., 2004. Constraints affecting mode choices by morning car \ncommuters. Transport Policy 11, 17-29. \nParkhurst, G., 2000. Influence of bus-based park and ride facilities on users' car traffic. Transport Policy \n7, 159-172. \nRasbash, J., Steele, F., Browne, W., and Prosser, B., 2005. A User's Guide to MLwiN Version 2.0. Centre \nfor Multilevel Modelling University of Bristol, Bristol. \nRasbash, J., Steele, F., Browne, W.J., and Goldstein, H., 2009. A User's Guide to MLwiN - version 2.10. \nCentre for Multilevel Modelling, University of Bristol, Bristol. \nReniers, G., Dullaert, W., and Visser, L., 2010. Empirically based development of a framework for \nadvancing and stimulating collaboration in the chemical industry (ASC): creating sustainable chemical \nindustrial parks. Journal of Cleaner Production 18, 1587-1597. \n Rietveld, P., Zwart, B., van Wee, B., and van den Hoorn, T., 1999. On the relationship between travel \ntime and travel distance of commuters - Reported versus network travel data in the Netherlands. Annals \nof Regional Science 33, 269-287. \nRoby, H., 2010. Workplace travel plans: past, present and future. Journal of Transport Geography 18, 23-\n30. \nRye, T., 1999a. Employer attitudes to employer transport plans: a comparison of UK and Dutch \nexperience. Transport Policy 6, 183-196. \nRye, T., 1999b. Employer transport plans - a case for regulation? Transport Reviews 19, 13-31. \nRye, T., 2002. Travel plans: do they work? Transport Policy 9, 287-298. \nSchwanen, T., Dieleman, F.M., and Dijst, M., 2004. The Impact of Metropolitan Structure on Commute \nBehavior in the Netherlands: A Multilevel Approach. Growth and Change 35, 304-333. \nShoup, D.C., 1997. Evaluating the effects of cashing out employer-paid parking: Eight case studies. \nTransport Policy 4, 201-216. \nTeal, R.F., 1987. Carpooling - Who, How and Why. Transportation Research Part A 21, 203-214. \nTsao, H.-S. and Lin, D.-J., 1999. Spatial and Temporal Factors in Estimating the Potential of Ride-sharing \nfor Demand Reduction. California PATH Research Report, UCB-ITS-PRR-99-2, \nVan Acker, V., van Wee, B. and Witlox, F., 2010. When transport geography meets social psychology: \ntoward a conceptual model of travel behaviour. Transport Reviews 30, 219-240.  \nVan Acker, V., Mokhtarian, P.L. and Witlox, F., 2011. Going soft: On how subjective variables explain \nmodal choices for leisure travel. European Journal of Transport and Infrastructure Research 11, 115-146. \nVandenbulcke, G., Steenberghen, T., and Thomas, I., 2007. Accessibility indicators to places and \ntransports: Final Report. Belgian Science Policy and FPS Mobility and Transports, Brussels. \nVandenbulcke, G., Steenberghen, T., and Thomas, I., 2009. Mapping accessibility in Belgium: a tool for \nland-use and transport planning? Journal of Transport Geography 17, 39-53. \nVanoutrive, T., Van Malderen, L., Jourquin, B., Thomas, I., Verhetsel, A., and Witlox, F., 2010. Mobility \nManagement Measures by Employers: Overview and Exploratory Analysis for Belgium. European Journal \nof Transport and Infrastructure Research 10, 121-141. \nVanoutrive, T., Van Malderen, L., Jourquin, B., Thomas, I., Verhetsel, A., and Witlox, F., 2012. Rail \nCommuting to Workplaces in Belgium: A Multilevel Approach. International Journal of Sustainable \nTransportation 6, 67-87. \nVerhetsel, A., Van Hecke, E., Thomas, I., Beelen, M., Halleux, J.-M., Lambotte, J.-M., Rixhon, G., and \nM\u00e9renne-Schoumaker, B., 2009. Pendel in Belgi\u00eb. FOD Economie,K.M.O.,Middenstand en Energie, \nBrussels. \nWang, R., 2011. Shaping carpool policies under rapid motorization: the case of Chinese cities. Transport \nPolicy 18, 631-635. \nWegener, M., 2011. From Macro to Micro - How Much Micro is too Much? Transport Reviews 31, 161-\n177. \n Whitmeyer, J.M., 1994. Why Actor Models Are Integral to Structural Analysis. Sociological Theory 12, \n153-165. \n \n abstract \nHome to work travel remains the prime focus of mobility management policies, in which the \npromotion of carpooling is one of the main strategies. Besides governments, employers are key \nplayers in this strive for a more sustainable commute. However, commuting research tends to \nfocus on individual commuters and their place of residence, rather than on workplaces and \ncompany-induced measures. Therefore, this paper takes the workplace as research unit to \nanalyse the popularity of carpooling in Belgium. After an exploratory (spatial) data analysis, we \nincorporate three groups of factors in a multilevel regression model which predicts the share of \ncarpooling at large workplaces: location (accessibility), organisation (activity sector), and \npromotion (carpool-oriented mobility management measures). Higher levels of carpooling are \nfound at less accessible locations, and in the activity sectors construction, manufacturing and \ntransport. This analysis gives insight in the determinants of carpooling, and may thus contribute \nto the development of sustainable transport policies. \n \nKeywords: carpool; commuting; Belgium; mobility management; multilevel modelling \n \nResearch Highlights \n \n-We model the share of carpooling in commuting at large workplaces in Belgium. \n \n-Mobility management policies promote carpooling as sustainable commuting alternative.  \n \n-Carpooling is more popular at workplaces with limited accessibility by rail. \n \n-There are more carpooling employees in construction, manufacturing and transport. \n \n-The promotion of carpooling must not result in more car-oriented development and lower levels \nof cycling and using public transport. \n \n \n \n \n Thomas Vanoutrive \nDepartment of Transport and Regional Economics, University of Antwerp and Department of Geography, \nGhent University \nthomas.vanoutrive@ua.ac.be \nCorresponding author \n \nElien Van De Vijver \nDepartment of Transport and Regional Economics, University of Antwerp and Department of Geography, \nGhent University \nelien.vandevijver@ugent.be \n \nLaurent Van Malderen \nFacult\u00e9s Universitaires Catholiques de Mons (FUCaM) and Universit\u00e9 catholique de Louvain, C.O.R.E. \nlaurent.vanmalderen@fucam.ac.be \n \nBart Jourquin \nFacult\u00e9s Universitaires Catholiques de Mons (FUCaM) \nbart.jourquin@fucam.ac.be   \n \nIsabelle Thomas \nUniversit\u00e9 catholique de Louvain, C.O.R.E. \nisabelle.thomas@uclouvain.be   \n \nAnn Verhetsel \nDepartment of Transport and Regional Economics, University of Antwerp \nann.verhetsel@ua.ac.be \n \nFrank Witlox \nDepartment of Geography, Ghent University \nfrank.witlox@ugent.be   \n \n \n", "identifiers": ["oai:archive.ugent.be:2967223", "10.1016/j.jtrangeo.2011.11.006"], "relations": [], "repositories": [{"id": "1493", "openDoarId": 0, "name": "Ghent University Academic Bibliography", "uri": null, "urlHomepage": null, "urlOaipmh": null, "uriJournals": null, "physicalName": "noname", "source": null, "software": null, "metadataFormat": null, "description": null, "journal": null, "roarId": 0, "baseId": 0, "pdfStatus": null, "nrUpdates": 0, "disabled": false, "lastUpdateTime": null, "repositoryLocation": null}], "repositoryDocument": {"pdfStatus": 1, "metadataAdded": 1478954803000, "metadataUpdated": 1590585838000, "depositedDate": 1534114800000}, "subjects": ["journalArticle", "info:eu-repo/semantics/article", "info:eu-repo/semantics/publishedVersion"], "title": "What determines carpooling to workplaces in Belgium: location, organisation, or promotion?", "topics": ["Social Sciences", "REDUCTION", "ENVIRONMENT", "NETHERLANDS", "TRAVEL-TIME", "MODE CHOICE", "CALIFORNIA", "MULTILEVEL APPROACH", "SPATIAL AUTOCORRELATION", "Carpool", "Commuting", "Belgium", "Multilevel modelling", "Mobility management", "EMPLOYERS", "COMMUTERS"], "types": [], "year": 2012, "fulltextIdentifier": "https://core.ac.uk/download/pdf/55853555.pdf", "doi": "10.1016/j.jtrangeo.2011.11.006", "oai": "oai:archive.ugent.be:2967223", "downloadUrl": "https://core.ac.uk/download/pdf/55853555.pdf"}}
{"status": "Not found", "data": null}
{"status": "OK", "data": {"id": "78076813", "authors": ["Zimmerer, VC", "Varley, RA"], "citations": [], "contributors": [], "datePublished": "2015-08", "description": "Processing of linear word order (linear configuration) is important for virtually all languages and essential to languages such as English which have little functional morphology. Damage to systems underpinning configurational processing may specifically affect word-order reliant sentence structures. We explore order processing in WR, a man with primary progressive aphasia (PPA). In a previous report, we showed how WR showed impaired processing of actives, which rely strongly on word order, but not passives where functional morphology signals thematic roles. Using the artificial grammar learning (AGL) paradigm, we examined WR's ability to process order in non-verbal, visual sequences and compared his profile to that of healthy controls, and aphasic participants with and without severe syntactic disorder. Results suggested that WR, like some other patients with severe syntactic impairment, was unable to detect linear configurational structure. The data are consistent with the notion that disruption of possibly domain-general linearization systems differentially affects processing of active and passive sentence structures. Further research is needed to test this account, and we suggest hypotheses for future studies", "fullText": "1 \n \nTitle \nA case of \u201corder insensitivity\u201d? Natural and artificial language processing in a man with primary \nprogressive aphasia. \nVitor C. Zimmerer & Rosemary A. Varley \nUniversity College London \n \nCorresponding author: \nVitor Zimmerer \nv.zimmerer@ucl.ac.uk \n \nAbstract \nProcessing of linear word order (linear configuration) is important for virtually all languages and \nessential to languages such as English which have little functional morphology. Damage to systems \nunderpinning configurational processing may specifically affect word-order reliant sentence \nstructures. We explore order processing in WR, a man with primary progressive aphasia. In a \nprevious report, we showed how WR showed impaired processing of actives, which rely strongly on \nword order, but not passives where functional morphology signals thematic roles. Using the artificial \ngrammar learning (AGL) paradigm, we examined WR\u2019s ability to process order in non-verbal, visual \nsequences and compared his profile to that of healthy controls, and aphasic participants with and \nwithout severe syntactic disorder. Results suggested that WR, like some other patients with severe \nsyntactic impairment, was unable to detect linear configurational structure. The data are consistent \nwith the notion that disruption of possibly domain-general linearization systems differentially affects \nprocessing of active and passive sentence structures. Further research is needed to test this account, \nand we suggest hypotheses for future studies. \nKeywords: Syntax; word order; configuration; primary progressive aphasia; artificial grammar \nlearning \n \n  \n2 \n \nIntroduction \nIn natural language, words occur one after another in a \u201clinear\u201d fashion, and processing of linear \nconstituent order, as one aspect of configuration, is mandatory for successful production and \ncomprehension. English, because of its limited inflectional system, relies heavily on linear \nconfiguration. The lion kills the man and The man kills the lion refer to very different events despite \nhaving the exact same lexical items. It has been argued that linearization processes are not specific \nto language (Boeckx, Martinez-Alvarez, & Leivada, 2014). The ability to process auditory and visual \nsequences, even when stimuli are meaningless, has been linked to the ability to process language \nunder noise conditions (Conway, Bauernschmidt, Huang, & Pisoni, 2010; Conway, Karpicke, & Pisoni, \n2007; Conway & Pisoni, 2008). Furthermore, sequence learning in a serial reaction time task \ncorrelated with children\u2019s ability to maintain syntactic structure in a priming task (Kidd, 2012). A \nmeta-analysis of eight studies, collectively examining 186 people with specific language impairment \nand 203 controls, found that pathological groups were poorer at statistical sequence learning (Lum, \nConti-Ramsden, Morgan, & Ullman, 2014). \nImpaired linear processing has also been found in acquired syntactic disorders such as aphasia. \nArtificial grammar learning (AGL) experiments have shown that people with impaired sentence \nproduction and comprehension have difficulties processing regularities in sequence order even \nwhen stimuli are non-verbal and/or visual. Dominey, Hoen, Blanc and Lelekov-Boissard (2003) tested \nseven aphasic participants and found that their ability to learn simple and complex artificial \ngrammars respectively predicted their ability to comprehend simple and complex sentences. \nChristiansen, Kelly, Shillcock, & Greenfield (2010) compared seven aphasic participants with seven \ncontrols matched for age and non-verbal intelligence and found that the former performed poorer in \nAGL. Zimmerer, Cowell, & Varley (2014) investigated AGL in four people with severe aphasia and \nsyntactic disorder, five people with aphasia in absence of syntactic disorder, and ten older controls \nand found learning profiles in patients with syntactic disorder which did not occur in the other \nsamples. Together, these studies indicate that differences between syntactically impaired and \nunimpaired participants were not the result of experimental procedures, non-verbal intellectual \ncapacity or general effects of brain damage, but were related to the syntactic impairment itself. \nHowever, reports of AGL performance in relation to processing of specific sentence structures are \nrare. Hoen et al. (2003) explored the effects of a training task involving sequence order manipulation \non sentence comprehension in six aphasic participants. They reported improvement only for \nsentences which were assumed to involve a similar order transformation. \n3 \n \nWe explore the relationship between general linearization and configurational processing in \nlanguage. In a previous report in Cortex, we described the comprehension ability of WR, a man with \nprimary progressive aphasia (PPA) (Zimmerer, D\u0105browska, Romanowski, Blank, & Varley, 2014). He \ndisplayed a striking and rarely reported syntactic profile: In sentence-picture matching tasks, his \nperformance on active transitives (The man pushes the elephant) and truncated actives with an \nauxiliary (The man is pushing) was at chance. His performance on full passives (The elephant is \npushed by the man) as well as on truncated passives (The man is pushed) was near ceiling. WR had \nsevere sentence production problems. His spoken output was markedly impoverished. In rare \ninstances of sentence-like output in spontaneous speech he strung together content words \nconnected by is a (e.g., Mary is a holiday is a Turkey). In spontaneous writing he produced only a \nsmall number of sentences, most of which were in the passive voice (e.g., Can it be used for \ntreatment?; As research was Vitor created). Naming ability on the PALPA54 subtest (Kay, Lesser, & \nColtheart, 1992) indicated residual lexical capacity with scores of 59/60 for spoken (with no penalty \nfor phonemic paraphasias as long as the target was recognizable) and 59/60 for written naming. \nWR\u2019s comprehension profile showed a dissociation which is the reverse of the predominantly \nreported pattern of good performance on actives and poor performance on passives, and poses a \nsubstantial challenge to conventional explanations for syntactic disorder (Druks & Marshall, 1995, \n1996). English passives are \u201charder\u201d with regard to a number of variables (Caplan & Waters, 1999; \nDrai & Grodzinsky, 2006; Druks, 2002; Grodzinsky, 2000; Mauner, Fromkin, & Cornell, 1993) as they \ncontain more words, more functional morphemes, have a non-canonical word order and, in some \ntheories, involve a transformation from canonical order (or \u201cmovement\u201d of constituents) . \nHowever, functional morphemes in passives contain strong cues for interpretation. The verb phrase \nmorphology (be+TNS V+PastP, e.g., is pushed), which in child development first emerges as the state \npassive (e.g., it\u2019s broken), appears to be grounded in stative use and biased towards assigning its \nsubject an inactive role (Brooks & Tomasello, 1999; Israel, Johnson, & Brooks, 2001; Riches, 2013). \nOn the other hand, the prepositional phrase by + NP (e.g., by the elephant) is a strong cue for \nagency. It is used in passives or agentive nominalizations (e.g., performances by the whole team) in \nan estimated 70% of instances (see Zimmerer, D\u0105browska, Romanowski, Blank, & Varley, 2014, for \ncorpus-based results). Zimmerer et al. (2014) demonstrated that WR used both verb and \nprepositional phrase morphology in isolation or together in order to assign agent/patient roles. \nActive sentences, and in particular those used for testing WR\u2019s syntactic comprehension, lack cues of \nthis type. Functional morphology, if present, was not reliable as a cue for determining semantic \nroles. For instance, the NP in NP be+TNS V+ing can have many roles such as agent (The man is \n4 \n \npushing the elephant), patient (The dress is selling), experiencer (The woman is watching the game) \nor instrument (The computer is enabling her to speak). Interpretation of English actives relies more \non word order. All active sentences used in testing WR\u2019s comprehension could have been \ninterpreted correctly using the common bias that the agent appears first (Ferreira, 2003). \nWR\u2019s profile presented a valuable opportunity to investigate linear order processing and its \nrelationship to configurational processing in language, and specifically, the issue of whether the \ndifficulties with comprehension of actives may be associated with a more general impairment of \nlinear structure processing. We hypothesized that WR would display impaired AGL behavior when \nprocessing linear configurational information in non-verbal sequences. \n \nArtificial grammar learning and the grammar AnBn \nAGL (Reber, 1967) is a commonly employed paradigm that tests processing of sequence structure. In \na training phase, participants are exposed to sequences of (nonsense) stimuli. Each sequence is \nunique but all are generated by a common set of rules. In the test phase, new sequences are \npresented. Some are generated by the same grammar, others violate it. The participant accepts or \nrejects each sequence based on its \u201cfit\u201d to the training set. Acceptance/rejection patterns provide \ninsight into which structural properties of the artificial grammar were learned spontaneously and \ngeneralized to the new sequences. AGL tasks engage neural language areas, in particular left inferior \nfrontal regions (Bahlmann, Schubotz, & Friederici, 2008; Bahlmann, Schubotz, Mueller, Koester, & \nFriederici, 2009; Petersson, Folia, & Hagoort, 2012; Petersson, Forkstam, & Ingvar, 2004). \nFor the current study AGL sequences were generated by the grammar AnBn. In AnBn sequences, a \nnumber of stimuli classified as A are followed by the same number of stimuli classified as B (e.g., \nAABB, AAABBB, AAAABBBB). AnBn sequences have linear configurational properties: All stimuli of one \nclass precede the stimuli of the other class, and the first class is A. AnBn sequences also contain non-\nconfigurational properties: The numbers of stimuli from each class match. There is strong evidence, \nincluding from self-reports after the experiment, that participants who learn non-configurational \nproperties do so by counting and comparing the number of stimuli of each class, and participants \nwithout syntactic disorder more consistently attend to configurational (order) than non-\nconfigurational (counting) structure (de Vries, Monaghan, Knecht, & Zwitserlood, 2008; Hochmann, \nAzadpour, & Mehler, 2008; Perruchet & Rey, 2005; Zimmerer, Cowell, & Varley, 2011; Zimmerer, \nCowell, et al., 2014). Some non-impaired individuals learn the configurational structure only, but \n5 \n \nnone learn only non-configurational structures. Performances are similar across auditory and visual \nmodalities (Zimmerer, Cowell et al., 2011).1 \nZimmerer, Cowell et al. (2014) investigated AnBn learning in four people with severe aphasia and \nsyntactic disorder, five aphasic controls without syntactic disorder, and ten older healthy controls. \nTwo syntactically-impaired participants, SA and SO, displayed atypical behavior. They learned the \nnon-configurational properties (counting) without learning the configurational properties (order), \ni.e., they rejected sequences like AABBB, but accepted sequences like BBBAAA and AABABB. We \nconcluded that they were insensitive to linear configuration in the experiment as the rejection of \nnon-configurational violations only requires tracking of the number of item occurrences, but not of \nany structure beyond that. Both SA and SO had large left perisylvian lesions and severely impaired \nprocessing of both active and passive sentences, and therefore results gave little insight into the \nrelationship between linear configurational impairment in AGL and the ability to process specific \nsentence structures. \nExperiment \nParticipant \nWR was 62 years old at the time of the experiment. Four years previously, he was diagnosed with \nlogopenic PPA (Gorno-Tempini et al., 2011). Non-language cognition was unimpaired during the \ncourse of the investigation reported here. His Performance IQ (Wechsler, 1999) was 119. Structural \nMR scanning revealed bilateral atrophy to fronto-temporal perisylvian regions (Fig. 1), which was \ngreatest in the left superior temporal gyrus. With disease progression, he displayed a decline in \nauditory comprehension. In the third year post-diagnosis and consistent with bilateral temporal lobe \natrophy, WR was diagnosed with cortical deafness. Although his performance in auditory \ncomprehension tasks was at or near chance, written word comprehension was only mildly impaired \nand remained stable over time. Written sentence comprehension and production displayed more \ngradual decline (Table 1), consistent with Thompson et al.\u2019s (2013) observation that some people \nwith logopenic PPA become syntactically impaired over time, and an unusual comprehension \npattern emerged (chance level performance on actives, ceiling performance on passives (Table 2)). \n \n \n \n6 \n \nFigure 1. Structural MR images of WR\u2019s brain in coronal and sagittal (left) views, showing fronto-\ntemporal perisylvian atrophy. \n \n \nTable 1. WR\u2019s performance on written and spoken tests of lexical and grammatical processing in \nyears 1, 2 and 3 post-diagnosis (years 2.5, 3.5 and 4.5 post symptom onset). Chance level is 25% for \nword-picture matching tasks and 50% for all other tasks. \nTest Year 1 Year 2 Year 3 \nSpoken minimal pairs 80% 73% 65% \nSpoken lexical decision 74% 70% 60% \nSpoken word-picture matching 100% 98% 69% \nSpoken synonym judgments 90% 84% 58% \nWritten lexical decision 98% 99% 98% \nWritten word-picture matching 100% 100% 100% \nWritten synonym judgment 97% 98% 96% \nSpoken sentence-picture matching 98% 93% 65% \nWritten sentence-picture matching 81% 78% 52% \n \n \n7 \n \nTable 2. Results of three sentence-picture matching tasks reported in Zimmerer, Cowell et al. (2014). \nTests involved selection from a choice of two pictures. Asterisks denote significant (p<.05) deviations \nfrom chance performance (50% accuracy). \nSentence type Example(s) N. of correct responses \n(% accuracy) \nExp. 1 Reversible active transitives The man pushes the elephant. 18/50 (36%) \nReversible full passive The man is pushed by the \nelephant. \n47/50 (94%*) \nExp. 2 Irreversible active transitives \nwith spatial by-phrase \nThe man by the woman shoots \nthe rabbit. \nThe man shoots the rabbit by \nthe woman. \n1/60 (2%*) \n \nIrreversible transitive actives The man shoots the rabbit. 30/30 (100%*) \nIrreversible full passives The rabbit is shot by the man. 30/30 (100%*) \nExp. 3 Truncated actives The man is pushing. 17/36 (47%) \nTruncated passives The man is pushed. 31/36 (86%*) \n \nMethods \nWe used the same methods as Zimmerer, Cowell et al. (2011, 2014) to allow comparability with \ncontrol and aphasic data reported therein (see online supplement for full Methods description). WR \nwas tested in the visual modality, which is consistent with prior AGL research conducted on people \nwith aphasia. WR participated in the AGL experiment two months after the reversible sentence \ncomprehension test took place. He gave informed consent to participation. Ethical approval was \ngranted by the local NHS Research Ethics Committee (08/H1308/32). \nWR was trained on 72 AnBn sequences (e.g., AAABBB). The test phase contained 120 new sequences, \nof which 60 were violations of the target grammar. Violations sequences were either non-\nconfigurational, i.e., they contained an incorrect number of stimuli (e.g., AABBB) or linear \nconfigurational, i.e., they violated the order in which A and B classes occurred (e.g., BBBAAA) or had \ntoo many transitions between classes (e.g., ABBABA). Each symbol (A or B) was mapped to a \nnonsense shape. Stimuli of class A were blue and rounded, and stimuli of class B were red and \nangled (Fig. 2). WR was asked to decide whether stimuli fit or did not fit the training set. He made his \ndecisions using a computer mouse with a green and a red button. \n8 \n \nFigure 2. Visual stimuli used in the experiment. Stimuli of class A were blue and rounded. Stimuli of \nclass B were red and angled. \n \n9 \n \nResults \nAt the start of the test phase, WR expressed uncertainty as to what was required. The experiment \nwas paused and instructions repeated. As a result, he did not provide responses to the first three \ntrials (two violations, one grammatical sequence). After instructions were repeated WR indicated he \nunderstood the task and responded quickly after each sequence. Ignoring the missed trials, WR \naccepted 96.6% (57/59) of the grammatical sequences. He rejected 50% of violations. However, \nclassified by configurational (order) vs. non-configurational (counting) violations, he rejected 17% \n(6/35) of the former and 100% (23/23) of the latter (see online supplement for further details).  \nIn AGL tasks it is possible that participants respond on the basis of familiarity of single stimuli or \nstimulus combinations, with a bias towards combinations that appeared more frequently. Such \nstatistical sensitivity is likely part of configurational processing. To account for such behavior we \ncalculated three familiarity variables for each test sequence (Knowlton & Squire, 1994; Redington & \nChater, 1996): associative chunk strength (the mean frequency in which bigrams and trigrams \nappeared during training), anchor strength (the mean frequency in which initial and final bigrams \nand trigrams appeared at the respective positions during training) and chunk novelty (the number of \nbigrams and trigrams which did not appear during training). We then ran a logistic binary regression \nwith acceptance/rejection of each sequence as a binary outcome variable. The predictors were the \nthree familiarity variables as well as the categorical variable Sequence Type. A test of the full model \nwas statistically significant, indicating that the predictors distinguished between acceptance and \nrejection of sequences, \u03c7\u00b2=75.649, p<.001, df=8. Nagelkerke\u2019s R2 of .707 indicated a moderately \nstrong relationship between prediction and response grouping. Only Sequence Type made a \nsignificant contribution to prediction, Wald=32.194, p<.001. Familiarity variables made no significant \ncontribution. \nZimmerer, Cowell et al. (2011, 2014) applied a template analysis to explore the use of different \nacceptance/rejection criteria by participants. The template analysis describes a number of possible \ngrammaticality criteria (summarized here as configurational only, non-configurational only, or both \nconfigurational and non-configurational). For instance, to test whether participants only rejected \nconfigurational violations, the analysis classifies not only AnBn sequences, but also non-\nconfigurational violations as grammatical, and all other sequences as ungrammatical. For each \ntemplate, a D Score (see also Perruchet & Pacteau, 1990) was calculated by subtracting the \npercentage of grammatical sequences rejected from the percentage of ungrammatical sequences \nrejected. D Scores range from -100 to 100, with 100 representing fully consistent application of \ngrammaticality criteria and zero representing chance. In addition, we used binomial tests to see if \n10 \n \nthe proportions of accepted and rejected sequences deviated significantly from chance. We \ncompared WR\u2019s scores with previously collected data from four severely aphasic participants with \nsyntactic disorder, five aphasic participants without syntactic disorder, ten older and 20 younger \ncontrols. WR\u2019s D Score for the non-configurational template was 91.5, his score for configurational \nonly was -13, and his score for the combined criteria was 46.6. WR, along with two severely aphasic \nparticipants (SA and SO), were the only participants who consistently rejected non-configurational \nviolations and at the same time accepted configurational violations. One other severely aphasic \nparticipant, JB, showed some sensitivity to linear configuration. \n  \n11 \n \nFigure 3. Behavioral patterns across all test phase sequence types for healthy speakers aged below \n50 years (mean age 27), healthy speakers aged 50 or older (mean age 62), aphasic participants \nwithout syntactic impairment, aphasic participants with severe syntactic impairment and WR. Each \ndata point represents one participant (chance performances excluded). Categories indicate analysis \ntemplate with highest D Score and suggest each participant\u2019s acceptability criteria. Higher D Scores \nindicate consistent application of decision criteria. Decision criteria: AnBn (configurational and non-\nconfigurational: A number of As is followed by the same number of Bs), (AnBn)+(BnAn) \n(configurational and non-configurational: All items of one class precede all items of the other class, \nequal number of As and Bs), A+B+ (configurational: All As precede the Bs), (A+B+)+(B+A+) \n(configurational: All items of one class precede all items of the other class) and EQUAL (non-\nconfigurational: Equal number of As and Bs). \n \nDiscussion \nImpaired sentence processing has been associated with poor or atypical AGL performance.  \nHowever, few studies have attempted to explore which aspects of sequence processing in AGL relate \nto which types of constructions. WR\u2019s very selective comprehension impairment provided a unique \n12 \n \nopportunity to explore the cognitive mechanisms involved in processing sentences that imposed \ngreater loading on constituent-order processing, as is the case with English active sentences. These \nsentences are frequently employed in clinical comprehension tests. WR\u2019s AGL behavior shows a clear \ndissociation between his ability to spontaneously learn linear configurational and non-\nconfigurational information. He was very sensitive to quantitative properties of stimulus sequences \nand reliably rejected quantitative violations, but appeared insensitive to properties of stimulus order \nand consistently accepted order violations. In contrast, participants without syntactic disorder \n(healthy and aphasic) are more sensitive to linear configurational structure. In combination with his \nselective impairment of active sentence comprehension WR\u2019s data are consistent with claims that \nthere is a relationship between AGL and sentence processing. His insensitivity to linear order in AGL \nmay be related to his difficulties with processing English actives which rely on configuration rather \nthan functional morphology. \nOne other AGL study has looked at individual aphasic profiles. WR\u2019s AGL behavior to date has only \nbeen observed in SA and SO, two people with severe syntactic disorder (Fig. 3, Zimmerer, Cowell et \nal., 2014) and large lesions across left perisylvian areas. However, SA and SO displayed chance level \ncomprehension across both active and passive sentences. One possibility is that impairments of \nprocessing functional morphology in addition to failure of general configurational processing might \nbe responsible for wholesale sentence processing failure. JB, another severely aphasic participant, \nwas able to identify violations of configuration. However, in addition to a grammatical profile \ncomparable to SA and SO, JB showed markedly lower lexical capacity (e.g., more impaired in naming \nand auditory/written synonym judgment (see Zimmerer, Cowell, & Varley, 2014, for his full profile). \nIt is possible that this severe lexical impairment masked intact configurational processing in natural \nlanguage processing. \nIt is not possible to generalize from a few cases to populations, and further investigations are \nneeded. However, consistent reports of heterogeneity in agrammatism (Berndt & Caramazza, 1999; \nCaramazza, Capasso, Capitani, & Miceli, 2005; Caramazza, Capitani, Rey, & Berndt, 2001), warrant a \nfocus at individual profiles. WR\u2019s profile is consistent with a claim of impaired configurational \nprocessing, accounting for his difficulty processing active, but not passive, sentences. We suggest the \nfollowing hypotheses for further investigations: \n1. Participants who are insensitive to configuration in AGL will have difficulties (at least) with \nsentences which are word order reliant, at least in comprehension. \n13 \n \n2. Participants who have no difficulties with processing word order in language (when functional \nmorphology is absent) will show a typical profile with regard to linear configurational processing in \nAGL. \nThe hypotheses also underline questions which, because of the progressive nature of WR\u2019s \npathology, could not be addressed in the current report. For instance, our explanations predict \ndifficulties with object-clefts as well subject-auxiliary inversions. Similarly, sensitivity to \nungrammatical word order can be tested in the future via grammaticality judgment tasks, and it \nwould be expected that a participant with WR\u2019s profile would fail to recognize such violations. \nHowever, the relationship between configurational processing in AGL and natural language \nprocessing may differ from language to language. In languages with a less rigid word order and/or \nmore functional morphology, impaired configurational processing may result in less severe linguistic \nimpairment. In other languages, such impairment may not affect actives, but other constructions. \n \n1 Early reports suggested that non-configurational properties of AnBn were processed through \nhierarchical phrase embedding (Fitch & Hauser, 2004; Friederici, Bahlmann, Heim, Schubotz, & \nAnwander, 2006). Some variations of AnBn, but not the one used for this experiment, aim to force \nhierarchical phrase processing (Bahlmann et al., 2008, 2009; Lai & Poletiek, 2011). \nReferences: \nBahlmann, J., Schubotz, R. I., & Friederici, A. D. (2008). Hierarchical artificial grammar processing \nengages Broca\u2019s area. NeuroImage, 42, 525\u2013534. \nBahlmann, J., Schubotz, R. I., Mueller, J. L., Koester, D., & Friederici, A. D. (2009). Neural circuits of \nhierarchical visuo-spatial sequence processing. Brain Research, 1298, 161\u2013170. \nBerndt, R. S., & Caramazza, A. (1999). How \u201cRegular\u201d Is Sentence Comprehension in Broca\u2019s Aphasia? \nIt depends on how you select the patients. Brain and Language, 67(3), 242\u2013247. \nBoeckx, C., Martinez-Alvarez, A., & Leivada, E. (2014). The functional neuroanatomy of serial order in \nlanguage. Journal of Neurolinguistics, 32, 1\u201315. \nBrooks, P. J., & Tomasello, M. (1999). Learning the English passive construction. In B. Fox, D. \nJuracsky, & L. Moichaelis (Eds.), Cognition and Function in Language. Stanford, CA: CSLI \nPublication. \nCaplan, D., & Waters, G. S. (1999). Verbal working memory and sentence comprehension. Behavioral \nand Brain Sciences, 22, 77\u2013126. \n14 \n \nCaramazza, A., Capasso, R., Capitani, E., & Miceli, G. (2005). Patterns of comprehension performance \nin agrammatic Broca\u2019s aphasia: A test of the Trace Deletion Hypothesis. Brain and Language, \n94(1), 43\u201353. \nCaramazza, A., Capitani, E., Rey, A., & Berndt, R. S. (2001). Agrammatic Broca\u2019s Aphasia Is Not \nAssociated with a Single Pattern of Comprehension Performance. Brain and Language, 76(2), \n158\u2013184. \nChristiansen, M. H., Louise Kelly, M., Shillcock, R. C., & Greenfield, K. (2010). Impaired artificial \ngrammar learning in agrammatism. Cognition, 116(3), 382\u2013393. \nConway, C. M., Bauernschmidt, A., Huang, S. S., & Pisoni, D. B. (2010). Implicit statistical learning in \nlanguage processing: Word predictability is the key. Cognition, 114(3), 356\u2013371. \nConway, C. M., Karpicke, J., & Pisoni, D. B. (2007). Contribution of implicit sequence learning to \nspoken language processing: some preliminary findings with hearing adults. Journal of Deaf \nStudies and Deaf Education, 12(3), 317\u201334. \nConway, C. M., & Pisoni, D. B. (2008). Neurocognitive basis of implicit learning of sequential \nstructure and its relation to language processing. Annals of the New York Academy of Sciences, \n1145, 113\u201331. \nDe Vries, M. H., Monaghan, P., Knecht, S., & Zwitserlood, P. (2008). Syntactic structure and artificial \ngrammar learning: The learnability of embedded hierarchical structures. Cognition, 107(2), \n763\u2013774. \nDominey, P. F., Hoen, M., Blanc, J.-M., & Lelekov-Boissard, T. (2003). Neurological basis of language \nand sequential cognition: Evidence from simulation, aphasia, and ERP studies. Brain and \nLanguage, 86(2), 207\u2013225. \nDrai, D., & Grodzinsky, Y. (2006). A new empirical angle on the variability debate: Quantitative \nneurosyntactic analyses of a large data set from Broca\u2019s Aphasia. Brain and Language, 96(2), \n117\u2013128. \nDruks, J. (2002). Verbs and nouns: a review of the literature. Journal of Neurolinguistics, 15(3-5), \n289\u2013315. \nDruks, J., & Marshall, J. C. (1995). When passives are easier than actives: two case studies of aphasic \ncomprehension. Cognition, 55, 311\u2013331. \nDruks, J., & Marshall, J. C. (1996). Syntax, strategies, and the single case: a reply to Zurif. Cognition, \n58, 281\u2013287. \nFerreira, F. (2003). The misinterpretation of noncanonical sentences. Cognitive Psychology, 47, 164\u2013\n203. \nFitch, W. T., & Hauser, M. D. (2004). Computational constraints on syntactic processing in a \nnonhuman primate. Science, 303, 377\u2013380. \n15 \n \nFriederici, A. D., Bahlmann, J., Heim, S., Schubotz, R. I., & Anwander, A. (2006). The brain \ndifferentiates human and non-human grammars: Functional localization and structural \nconnectivity. Proceedings of the National Academy of Sciences, 103(7), 2458\u20132463. \nGorno-Tempini, M. L., Hillis, A. E., Weintraub, S., Kertesz, A., Mendez, M., Cappa, S. F., \u2026 Grossman, \nM. (2011). Classification of primary progressive aphasia and its variants. Neurology, 76(11), \n1006\u201314. \nGrodzinsky, Y. (2000). The Neurology of Syntax: Language Use Without Broca\u2019s Area. Behavioral and \nBrain Sciences, 23, 1\u201371. \nHochmann, J.-R., Azadpour, M., & Mehler, J. (2008). Do Humans Really Learn AnBn Artificial \nGrammars From Exemplars? Cognitive Science, 32(6), 1021\u20131036. \nHoen, M., Golembiowski, M., Guyot, E., Deprez, V., Caplan, D., & Dominey, P. F. (2003). Training with \ncognitive sequences improves syntactic comprehension in agrammatic aphasics. NeuroReport, \n14(3), 495\u2013499. \nIsrael, M., Johnson, C., & Brooks, P. J. (2001). From states to events: The acquisition of English \npassive participles. Cognitive Linguistics, 11(1-2), 103\u2013129. \nKay, J., Lesser, R., & Coltheart, M. (1992). Psycholinguistic Assessment of Language Processing in \nAphasia. Hove, U.K.: Psychology Press. \nKidd, E. (2012). Implicit statistical learning is directly associated with the acquisition of syntax. \nDevelopmental Psychology, 48(1), 171\u2013184. \nKnowlton, B. J., & Squire, L. R. (1994). The information acquired during artificial grammar learning. \nJournal of Experimental Psychology. Learning, Memory, and Cognition, 20(1), 79\u201391. \nLai, J., & Poletiek, F. H. (2011). The impact of adjacent-dependencies and staged-input on the \nlearnability of center-embedded hierarchical structures. Cognition, 118(2), 265\u201373. \nLum, J. A. G., Conti-Ramsden, G., Morgan, A. T., & Ullman, M. T. (2014). Procedural learning deficits \nin specific language impairment (SLI): A meta-analysis of serial reaction time task performance. \nCortex; a Journal Devoted to the Study of the Nervous System and Behavior, 51, 1\u201310. \nMauner, G., Fromkin, V. A., & Cornell, T. L. (1993). Comprehension and acceptability judgments in \nagrammatism: Disruptuions in the syntax of referencial dependency. Brain and Language, 45, \n340\u2013370. \nPerruchet, P., & Pacteau, C. (1990). Synthetic Grammar Learning: Implicit Rule Abstraction or Explicit \nFragmentary Knowledge? Journal of Experimental Psychology: General, 119, 264\u2013275. \nPerruchet, P., & Rey, A. (2005). Does the mastery of center-embedded linguistic structures \ndistinguish humans from nonhuman primates? Psychonomic Bulletin & Review, 12(2), 307\u2013313. \nPetersson, K. M., Folia, V., & Hagoort, P. (2012). What artificial grammar learning reveals about the \nneurobiology of syntax. Brain and Language, 120(2), 83\u201395. \n16 \n \nPetersson, K. M., Forkstam, C., & Ingvar, M. (2004). Artificial syntactic violations activate Broca\u2019s \nregion. Cognitive Science, 28(3), 383\u2013407. \nReber, A. S. (1967). Implicit Learning of Artificial Grammars. Journal of Verbal Learning & Verbal \nBehavior, 6(6), 855\u2013863. \nRedington, M., & Chater, N. (1996). Transfer in artificial grammar learning: A reevaluation. Journal of \nExperimental Psychology: General, 125(2), 123\u2013138. \nRiches, N. G. (2013). Treating the passive in children with specific language impairment: A usage-\nbased approach. Child Language Teaching and Therapy, 29(2), 155-169. \nThompson, C. K., Meltzer-Asscher, A., Cho, S., Lee, J., Wieneke, C., Weintraub, S., & Mesulam, M.-M. \n(2013). Syntactic and morphosyntactic processing in stroke-induced and primary progressive \naphasia. Behavioural Neurology, 26(1-2), 35\u201354. \nWechsler, D. (1999). Wechsler Intelligence Scale for of Intelligence (WASI). San Antonio, TX: Harcourt \nAssessment. \nZimmerer, V. C., Cowell, P. E., & Varley, R. A. (2011). Individual behavior in learning of an artificial \ngrammar. Memory & Cognition, 39(3), 491\u2013501 \nZimmerer, V. C., Cowell, P. E., & Varley, R. A. (2014). Artificial grammar learning in individuals with \nsevere aphasia. Neuropsychologia, 53, 25\u201338. \nZimmerer, V. C., D\u0105browska, E., Romanowski, C. A. J., Blank, C., & Varley, R. A. (2014). Preservation \nof passive constructions in a patient with primary progressive aphasia. Cortex, 50, 7\u201318. \n  \n17 \n \nOnline supplement \n \nMethods (extended) \nWe used the same methods as Zimmerer, Cowell and Varley (2011, 2014) to allow comparability \nwith control and aphasic data reported therein (see online supplement for full Methods description). \nWR was tested in the visual modality, which is consistent with prior AGL research conducted on \npeople with aphasia. WR participated in the AGL experiment two months after the reversible \nsentence comprehension test took place. WR gave informed consent to participation. Ethical \napproval was granted by the local NHS Research Ethics Committee (08/H1308/32). \nEach symbol (A or B) was mapped to a nonsense shape. Stimuli of class A were blue and rounded, \nand stimuli of class B were red and angled (Fig. 2). For each class there were eight different shapes. \nStimuli were presented one after another. Stimulus duration was 400 ms. The interstimulus interval \nwas 200 ms within a sequence and 4.5s between sequences. Between sequences, a fixation cross \nappeared on the screen. The average trial length was 8 seconds. \nThe training phase consisted of 72 AnBn-sequences of three different structures: AABB, AAABBB and \nAAAABBBB. Each structure appeared 24 times. Each sequence had a unique combination of stimuli, \nand each stimulus appeared the same number of times. The test phase consisted of 60 new AnBn-\nsequences (grammatical) and 60 violation sequences (ungrammatical) that employed the same \nsymbol sets. To enable identification of different response patterns, there were five types of \nviolation sequences. Each violation type occurred 12 times. Two violated non-configurational, and \nthree violated configurational properties of the grammar: \n- Type 1 (non-configurational) had mismatching numbers of As and Bs. In addition, it had an \nodd total number of stimuli, while grammatical sequences always contain an even number. \nE.g., AABBB. \n- Type 2 (non-configurational) had mismatching numbers of As and Bs but a (legal) even total \nnumber of stimuli, e.g., AAAABB. \n- Type 3 (configurational) had the reverse class order. They started with the wrong class, \nended with the wrong class and contained the ungrammatical combination BA, e.g., \nBBBAAA. \n- Type 4 (configurational) started with the wrong class and contained the ungrammatical \ncombination BA, e.g., BABAAB. \n18 \n \n- Type 5 (configurational) ended with the wrong class and contained the ungrammatical \ncombination BA, e.g., ABBABA. \nZimmerer, Cowell et al. (2011, 2014) employed the same randomized sequence order across all \nparticipants to eliminate order effects across individuals. The only randomization parameter was \nthat not more than four consecutive test sequences were all grammatical or all ungrammatical. The \nsame order was used for WR. The training and the test phase were each divided into three trial \nblocks of equal length, with a self-timed break between blocks. The entire session took ca. 25 \nminutes. \nDecisions in the test phase were made via a computer mouse. The left button was marked with a \ngreen sticker, the right button with a red sticker. The green button registered acceptance, and the \nred button rejection of a sequence. There was a timeout after 4.5 seconds. Before the training \nphase, WR was instructed to attend to the stimuli on the screen. Instructions were simplified and \npresented in writing (Zimmerer, Cowell et al., 2014). Before the training phase, we asked WR to \nwatch the screen. Before the test phase, we presented him instructions on paper in a 26pt font: \n\u201cNow you will see more shapes. Do they follow the pattern of the shapes you just saw? GREEN \nbutton \u2013 YES \u2013 same pattern. RED button \u2013 NO \u2013 different pattern. Decide quickly.\u201d There was a line \nbreak before each new sentence. The word \u201cgreen\u201d was printed in green font, the word \u201cred\u201d in red \nfont. The experimented provided the instructions orally, supported by gestures (\u201cthumbs up\u201d, \n\u201cthumbs down\u201d). WR signaled that he understood the instructions before proceeding. \n \nResults (extended) \nAt the start of the test phase, WR expressed uncertainty as to what was required. The experiment \nwas paused and instructions repeated. As a result, he did not provide responses to the first three \ntrials (type 1 violation, grammatical sequence, type 5 violation, in that order). After instructions \nwere repeated WR indicated he understood the task and responded quickly after each sequence. \nIgnoring the missed trials, WR accepted 96.6% (57/59) of the grammatical sequences. He rejected \n100% (11/11) of the violations of type 1, 100% (12/12) of the violations of type 2, 0% (0/12) of the \nviolations of type 3, 8% (1/12) of the violations of type 4 and 45% (5/11) of the violations of type 5. \nClassified by configurational (order) vs. non-configurational (counting) violations, he rejected 17% of \nthe former and 100% of the latter. \nIn AGL tasks it is possible that participants respond on the basis of familiarity of single stimuli or \nstimulus combinations, with a bias towards combinations that appeared more frequently. Such \n19 \n \nstatistical sensitivity is likely part of configurational processing. To account for such behavior we \ncalculated three familiarity variables for each test sequence (Knowlton & Squire, 1994; Redington & \nChater, 1996): associative chun