From Spelling Test to IQ to g
Ability without essentialism
Our intuitions do pretty well with a spelling test. Few of us are prone to think that a percentage-correct score on a 25-item spelling test is a measure of “innate spelling ability”. On the other side, there aren’t movements to outlaw spelling tests in schools. People recognize spelling tests for what they are: measures of spelling in the ordinary psychometric sense, useful enough if you are trying to find people who are good spellers but not a measure of some essential quality in people’s heads or genes.
But go back to the example of the person who creates a little spelling test to support their business. Turns out that they also do some online bookkeeping, so they also want people who are good at arithmetic. They create a little math test, just like the spelling test, and it turns out the same way: people who get one math item correct also tend to get other math items correct. (A psychometrician would say that the test has a high coefficient alpha, internal reliability; the point is that the correlation among the items is what makes “math ability” a trait. If the items weren’t correlated it wouldn’t make sense to try to measure it. It would just be a random process, like scoring how many sixes you roll in 25 tries.)
But then comes the interesting part. It also turns out the the math score and the spelling score are positively correlated with each other. This is where it is easy to start the slide into essentialism. The fact that spelling scores and math scores are positively correlated violates a certain naive intuition about human ability: some people are good at some things, some at others, but when you average it out we are all about the same in terms of general tendency to get correct answers. You might be good at math questions, me at spelling, but it all comes out in the wash. The first part of this is true: everyone is better at some things than others. But, alas, the second part isn’t. On average, some people are better at getting correct answers across the board.
Another step in the slide into essentialism is giving fancy names to simple phenomena. The fancy name for the correlation between math and spelling is the “positive manifold,” a term that actually has a mathematical meaning in multivariate calculus, but here is just a description of the fact that all tests containing questions with right-wrong answers are positively correlated. That’s an interesting fact about the world, not something to be easily dismissed.
The positive manifold is fodder for factor analysis. Once you have a bunch of tests of different abilities, you can correlate them with each other, and fit a single common factor to the correlations. Charles Spearman did that back in the thirties, and called the resulting factor “g”, for general intelligence. The slide into essentialism continued. To be clear: tests of general ability can be carefully constructed and useful, but they are still just elaborate spelling tests, measures of the tendency to get correct answers to questions on the widest possible domain of right-wrong questions. Psychometrically sound, potentially useful, but not a measure of anything essential inside people’s heads.
Yet for some reason we all— me included— get sucked in to thinking of g or IQ as an essential property that people “have”. Everything about IQ pushes us in this direction. There is no way to describe what it is in simple test-taking terms, like “spelling test.” Thus all my forced constructions about “tendency to get correct answers to right-wrong questions.” IQ (Binet) and g (Spearman), are names for “spelling test score” that over a century have gotten so baked into our brains that we can’t think about test scores without them.
Why some people do better on ability tests than others is a legitimate question, and I don’t dismiss out of hand the possibility that there is something in people’s genes or brains that determine our scores. My point for now is that any evidence for that kind of explanation can’t be found in the ability test itself. All the test produces is a score, even if it is a carefully constructed score using factor analysis or IRT.
What Jon Haidt thinks people are “denying” is not the observation that some people are better at answering questions than other people, or even that the ability to do so is an important trait with consequences for doing well in the modern world. I think that much is perfectly obvious, undeniable. I would actually like to hear from someone who thinks that everyone they meet has the same level of thinking ability, who never thinks of someone as “smart” or, um, the opposite.
What Jon thinks people deny is an essentialist explanation of the obvious fact that people differ in their outward abilities, that there is this thing called “IQ” that explains why some people are better correct-answer-getters than others. In fact, it is Jon and the essentialists who are wrong about this, not the libs. Modern science of intelligence has shown it. More tomorrow.