I just ran across an interesting article where Nassim Nicholas Taleb writes about the problems with IQ. Like everything he writes it is hard to read because of his tone. I suspect that he would be far easier to have a conversation with while walking with him or eating a meal with him than he is to read in text. That said, he makes a really interesting point about IQ in this article which I haven’t seen before.

(Quick note: my own position on IQ is that it makes a great metaphor for talking about comparative intelligence and, in the real world, I’ve heard it’s among the best predictors there is for how people will do on IQ tests. That is to say, its only use is as a metaphor for certain specialized conversations.)

The point that Taleb makes is that IQ will always produce strong statistical correlations because it does strongly detect severe learning disabilities. Or, more simply, people who are bad at everything are also bad at taking IQ tests.

He gave a striking example which makes it very clear. Suppose that you have a test which you administer to 10,000 people, 2,000 of which are dead. Dead people have a 0 IQ and will also score 0 on any performance metric. If there is no correlation whatever among the living people, standard statistical methods will give the result that this total population has an IQ-to-performance correlation of 37.5%.

To the best of my knowledge, psychologists don’t give IQ tests to corpses, but the same thing holds for less severe impairments than death, which is why the psychologists keep getting strong statistical significance for the correlation of IQ tests with… anything else.