http://en.wikipedia.org/wiki/Working_dog

The only question is why they replaced "dog" with the snootier "canine," but the sentences parses just the same. /Buzz Killington

Not only is the sample size small, but the conclusion is flawed based on the experiment. All we know is that 1/3 of the 16 picked the "inferior" sample as sounding better. The other 2/3 could have just gotten lucky. After all, if every single one of the 16 people were stone cold deaf and just had to pick randomly, the headline (based on this logic) would still read "1/2 of People Can't Tell..."

A better methodology would involve making the 16 people listen to a couple dozen different samples so the testers can decide if they do substantially better than 50% at picking the "superior" recordings.

Of course, this analysis doesn't even take into account how "better" is entirely subjective...