Fibonacci in 52 bytes.

annotated.js

function(n){ // calculate the nth fibonacci number
    for(
        var a = 0, b = 1; // start with a = 0 and b = 1
        n--;              // run the loop n times
    ) b = a + (a = b);    // assign a = b and b = a + b 'in parallel'
    return a              // return the result
}

index.js

function(n){for(var a=0,b=1;n--;)b=a+(a=b);return a}

LICENSE.txt

            DO WHAT THE FUCK YOU WANT TO PUBLIC LICENSE
                    Version 2, December 2004

 Copyright (C) 2011 Jan Kassens <jan@kassens.net>

 Everyone is permitted to copy and distribute verbatim or modified
 copies of this license document, and changing it is allowed as long
 as the name is changed.

            DO WHAT THE FUCK YOU WANT TO PUBLIC LICENSE
   TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION

  0. You just DO WHAT THE FUCK YOU WANT TO.

package.json

{
  "name": "fibonacci",
  "description": "Fibonacci in 52 bytes.",
  "keywords": [
    "140bytes",
    "fibonacci",
    "math"
  ]
}

The last semicolon is not necessary, I think. A recursive version would probably be shorter, but I'm sure you didn't choose it for speed/memory reasons.

@Pat3ris: good catch with the semicolon. Since the outer function should be anonymous, defining an inner function is most likely longer.

Very nice job with b=a+(a=b)
A recursive approach respecting the rules of 140byt.es is around 62 bytes

save 2 bytes:
function(n,a,b){for(b=1;n--;)b=~~a+(a=b);return a}

facepalm How did we miss that ? Well done!

function(n,a,b){for(a=0,b=1;n>1;n-=2,b+=a+=b);return n?b:a;};

longer but even faster (does two steps per loop so two assignments for two steps rather than two for one