bindings.js

var SPEECH = 'speech';
var bindings = {
  /* ... */
  
  fullSpeed: {
    device: SPEECH,
    inputId: 'full speed'
  },
  
  stop: {
    device: SPEECH,
    inputId: 'stop'
  }
}

resultHandler.js

var resultHandler = function (evt) {
  var requiredConfidence = 0.75;
  for (var i = evt.resultIndex; i < evt.results.length; ++i) { 
    var result = evt.results[i], 
        primary = result[0]; 
    if (result.isFinal || primary.confidence > requiredConfidence) {
      var transcript = primary.transcript;
      console.log('We have a match:', transcript);
    }
  }
};

setup_1.js

var recognition = new webkitSpeechRecognition();

setup_2.js

var recognition = new webkitSpeechRecognition();
recognition.continuous = true;
recognition.interimResults = true;
recognition.lang = "en_US";

speechHandler.js

inputController = new InputController(bindings);
inputController.registerDeviceHandler(SpeechHandler, 'speech');

speechHandler = inputController.deviceHandlers.speech;
speechHandler.requiredConfidence = 0.8;
speechHandler.onRecognitionEnded = function(){
  // do stuff, e.g. notify user that recognition is inactive
  speechActive = false;
};

startRecognition.js

if(input.toggleSpeechInput){
  if(speechActive){
    speechHandler.stop();
  } else {
    speechHandler.start();
  }
  speechActive = !speechActive;
  input.toggleSpeechInput = false;
}

Please use the unprefixed SpeechRecognition if available by adding something like:

var SpeechRecognition = window.SpeechRecognition || window.webkitSpeechRecognition;
var recognition = new SpeechRecognition();

to setup_1.js and setup_2.js. That way your code won't break when the prefix goes away.