A friend of mine posted a blog entry about a tone-deafness test. Not being a musician, I tried it with the standard speakers on my PC, which are the usual junk. I scored 80.6%, which pleasantly surprised me. For me I think it was more pattern recognition than tone detection. I’m also guessing that since there were only 36 questions, for most people the statistical error for a single test is over +/- 5%. So I’d guess I’m really between 76-86%.
Another thing to note is that the web-test results are biased high by people cheating a little. They say in controlled laboratory testing the highest they got was 88%. So my friend, who is impeccably honest, did really well with 86.1%.
I just did it (with computer speakers and no cheating – how or why would you cheat on something like that anyway?) and got an 86.1%
I might have done better on a second test, as I suspect that a few of my errors were on the first note or two in the first several phrases, but I figured that a second test would defeat the purpose.
Interestingly, I found that as the test progressed, I increasingly used the same skills for recalling musical phrases as I do when drawing from memory and recalling visual details. Says something about the way I remember, I guess…. Interesting.
I think they were talking about biases that would be introduced by re-taking the test, which skews the scores for their database, and makes some people think that they can beat 90%. Those who get a second, higher score will probably cite that as their score.
Basically, if you repeat the test, the questions are in the same order and you are more likely to remember the tones, and also to notice changes in the tones. So you will likely get a higher score, and you will also bias the overall statistical results.
For this to be an honest test, you need to do it only once with a given sequence of tones. To honestly test again you need a different sequence of tones. I just skimmed through the test again and recieved about 88%, but I think the 80.6% result was closer to the truth.
When it was done under controlled lab conditions, the highest anyone got was 88%. They discuss it a bit here.