Intelligence changes


Jason Collins


November 15, 2011

Scott Haufman has written a post on the variation in IQ over a person’s life. He writes:

In 1932, the entire population of Scottish 11-year-olds (87, 498 children) took an IQ test. Over 60 years later, psychologists Ian Deary and Lawrence Whalley tracked down about 500 of them and gave them the same test to take again.

Turns out, the correlation was strikingly high – .66, to be exact. Those who were at the top of the pack at age 11 also tended to be at the top of the pack at age 80, and those who were at the bottom also tended to stay at the bottom. Equally as interesting, the correlation was far from perfect. A few outliers could be found. One person had an IQ of over 100 at age 11, but scored just over 60 at age 80. There are many possible reasons for this outlier, including dementia. Other folks showed IQ increases as they aged. On average, people’s individual (or absolute) scores on the test taken again at age 80 was much higher (over one standard deviation) than their scores had been at age 11, even though the rank ordering among people stayed roughly the same.

These results are illustrative of what psychologists find over and over again. IQ tends to remain relatively stable over the lifespan. The key word here is relative. IQ researchers are interested in explaining differences. Developmentally speaking, an individual’s intelligence is not fixed at birth. Although rank ordering of scores tends to remain stable (relative change), scores within individuals fluctuate quite a bit (absolute change).

In his analysis, Haufman rightly points out that this is further evidence that IQ is not fixed. However, he hasn’t noted one element of the change in IQ - the heritability of IQ increases through childhood and adolescence. The idiosyncratic environmental influences wash out. Some of the variation is movement reflecting some fixed underlying factors.

If the experiment reported changes in IQ between the ages of 20 and 80, I expect the variation would be smaller.