Audio Myth Workshop from the 127th AES (Audio Engineer Society) Convention, NY 2009
Very interesting, although when he presents his tests towards the end of the talk (e.g. can you tell the difference between a 16-bit file truncated from 24-bit versus dithered from 24-bit, and so on), I guessed correctly in each instance. Although the difference was not huge, it seemed to me that I could hear the difference, and that one file did sound slightly better than the other.
Now, I could have been imagining this (quite possible) and simply have made lucky choices, but the odds would seem to suggest otherwise. It's possible, though.
I do agree that the role of expectation bias, placebo effect and so forth is so strong that an intelligent double-blind protocol is the only way to reliably determine if differences are truly audible or not. Doesn't mean that subjective evaluations are not fun, interesting and possibly informative, however.