After reading about last year's hoax report 'Intelligence Quotient (IQ) and Browser Usage' I realized I was in fact already running a real live experiment measuring number skills: a site were you can solve Calcudoku number puzzles. I analyzed two years' worth of data, consisting of over 1 million solved puzzles. This included puzzles solved 'against the clock,' of three different sizes. For each size, Chrome users were the fastest solvers, Firefox users came second, and IE users were the slowest. The number of abandoned puzzles (started but never finished) was also significantly higher for IE users. Analysis shows that the differences are statistically significant: in other words, they did not happen by chance. I put up more details and some graphs, and also wrote a paper about it(PDF).

http://science.slashdot.org/story/12...ie-users-worst