while interesting, i would hesitate to call that experiment a scientific study.
there were no controls and far too many variables {sound equipment, test data, sample selection, sample size}.

however, there are studies that also suggest that perceived audio quality is not directly correlated with data rates. most people cant tell the difference between average and high data rates. however, studies {including the one you posted} also suggest that a non-trivial portion of people who care about it can do much better than average.

the moral of the story is to use ogg vorbis at quality 5 and get better perceived quality, a smaller file size, and a codec unencumbered by patents or commercial ill will.