Comment Odd methodology, tiny sample size (Score 1, Informative) 101
Typically in sound quality tests, you tell subjects which file is the original, then have them rate how close to the original the other samples are. In this he just gave them four samples, and had them guess which was which, turning it into a more subjective test of guessing what they think the track should like. In addition, based on the table he got a total of 1-4 responses per track, which is far too low to have any statistical significance.
This was a funny joke, but not the gotcha the article played it up to be.