Personality

The idea of a personality test has so many attractions. The results contain the vagueness and flattery of a horoscope without the obvious pseudoscience of astrology. You get to find out that you are a completely excellent kind of person and that other people are freaks. The Myers-Briggs test is a classic of that kind of personality test.

Even so, it would be very interesting if there were statistically significant personality combinations. It certainly conforms to everyday experience that there would be a limited range of broad personality groups. That sort of impression is frequently wrong, so it's important that large-scale personality characteristics, like a tendency to religiosity, are about as heritable as IQ, with r=.5 if I remember correctly. If personality is heritable and, so to speak, a mental phenotype, then it's not a priori ridiculous to look for phenotypic groupings analogous to the phenotypic groupings of outward appearance.

According to John Derbyshire (and wikipedia) psychological consensus is that the Big 5 or some similar model gets close to capturing personality variation and grouping parsimoniously. It makes no assertions in itself about whether the five axes correspond to five, or any other number, of physical brain arrangements, though that would be the most interesting discovery of all. Whether there is a physical correspondence or not, eHarmony tries to cash in on the model by matching personality types. I don't remember whether they mention the basis for their testing at any point, but the model is clearly Big 5.

Anyway, Derbyshire mentioned the Big 5 again today so I took the most official looking version of the test, which has the advantage of being part of a scientific survey. Reading the results gives a somewhat unreal sensation, like hearing myself on the phone. On the other hand, it is a short test, and as a friend who took it found, coming at a small subset of the questions from an unexpected angle can produce surprising results. I would like to see how the reliability of this short version in subsequent re-testing compares to the full version.

Update: so I took it again, just to see. I didn't try to do it at all differently, but got quite a bit of variation. Results: E 7 vs 10; A 90 vs 76; C 93 vs 66 (!); ES 31 vs 38; O 99 vs 99. I think those percentiles change as more people take the test, but that's what they were when I got the results. That change in conscientiousness is startling. When my friend re-took the test, his C score changed even more sharply, from 2 to 83. He made a good point, doubting whether the questions adequately capture the traits described by C, since there are no questions about civic or religious conscientiousness and questions about work predominate. On his second test he admittedly was trying a different way of handling close calls, but the original 2 does not describe him at all. He did find that his E, ES, and O scores were pretty stable, as mine were. Even so, the instability of the results is disappointing. Oh well, back to oneiromancy.