If blind tests routinely dismiss the top and bottom performers, then those conclusions might apply to the general population, but not to us. Because we probably hear better than the general population.
This is a real problem when interpreting scientific results - unless a reviewer understands the statistics and the methodology, he might not understand the study.
And, bdp, you are absolutely right when you say that someone who identifies 10 out of 10 is unlikely to be doing it by random selection. In fact, it is highly significant - at the 0.001 level (it’s a binomial distribution with 1 way to be dead right, and 1023 ways to be right less often).
