The /r/linguistics thread on this subject does a pretty good job of pointing out the flaws in the way this data is presented: http://www.reddit.com/r/linguistics/comments/288ed6/12_engli...

The test that they run is pretty cool, but this conclusion is super weak.