Tuesday, March 04, 2008

James Flynn: What is Intelligence?

In his new book, What is Intelligence?, James Flynn tries to explain a few things. First he wants to explain what intelligence is and what its components are, second he wants to explain his new understanding of the Flynn effect, and what it implies about genes and intelligence, and thirdly, he wants to convince us that what The Bell Curve said about intelligence and race isn't supported by these new understandings. He succeeds admirably at the latter two; his success in explaining the nature of intelligence is limited.

For anyone who isn't familiar with the Flynn Effect, I'll repeat the findings briefly. More than two decades ago, Flynn noticed that IQs have been going up over time. About 3-5 points per decade, independent of culture, location, sex and race. The people who write intelligence tests have known for quite a while; they reissue their tests every decade or so, and "re-norm" the results. Since the definition of IQ is that 100 is the average score across the population (which population? That's a separate question; read the book if you want the details) they have to measure the results for a standardized group, and set the scoring so the current test will give the right results.

One of the consequences that matters to Flynn is the implication for death penalty cases, which he has been brought into recently. The implication is that if you give someone a test that is 10 to 15 years out of date, their score will be artificially inflated, since they are being measured by the norms of an earlier period. The obvious argument among defenders of capital cases is that death row inmates should be tested by up-to-date standards so as not to inflate their scores and accidentally rate them as competent to stand trial when they are in fact borderline or below it. Flynn points out that it's common for schools in disadvantaged areas and for prisons to not replace their existing stock of test booklets when a revision is issued, so they can be significantly out-of-date, which artificially inflates the scores and negates one escape route.

Like all good scientific revolutions, Flynn starts with four paradoxes arising from the combined data about rising scores.

  • Different sub-tests (e.g. vocabulary, spatial reasoning, abstract analogies, pattern matching) have shown different increases. What's different about the areas in which intelligence is growing the fastest?
  • Given the size of the increase, why doesn't it seem clear in everyday interactions that each generation is significantly smarter? 20 years is almost 10 IQ points, so two generations is nearly 20.
  • How did our ancestors get by if only 100 years ago, everyone was mentally retarded by current standards?
  • The changes are so rapid that they can't be genetic, so they must be due to environmental changes, yet studies comparing twins raised together and apart show that environment makes little difference to adult intelligence. Why does the environment make so much difference in some cases and so little in others?

Flynn's resolution is that the environmental differences that matter are large-scale and societal. He also argues that the societal differences compound, so even though small changes are scattered throughout our schooling, entertainment, child-rearing practices, employment expectations, and hobbies, the effects can be pervasive. The area of change that Flynn pinpoints is reliance on abstraction. This turns out to be a common thread among the sub-tests with the highest increases. Our ancestors dealt with the world much more concretely, and modern child rearing, education and entertainment all exercise our growing competence at abstraction. The results from twin studies are dominated by society-wide practices, and show that which family one is raised in, or which schools one goes to don't matter nearly as much as which era, and which society. An agrarian society that doesn't expect its children to grow up and leave the farm raises them to focus on the here and now. Expectations change when horizons open up, and we should expect every society to undergo a Flynn effect as people expect the next generation to live in cities, work in information-intensive jobs, and socialize with people who aren't all doing the same work their ancestors have done since time immemorial.

Ultimately, Flynn's book provides a satisfying resolution to the problems raised by IQ differences. The implications of Murray and Herrnstein's The Bell Curve for racial differences have been neutralized. Not by ethical arguments or posturing, but by a careful analysis of the data. Murray and Herrnstein were led astray by the surface implications of the data when analyzed within generations. If they hadn't written up their analysis carefully and thoroughly, Flynn wouldn't have been impelled to revisit the data and produce a sounder conclusion. Murray and Herrnstein's conclusions weren't original with them; their contribution was their willingness to explain an unpopular idea carefully enough that its limitations would become visible when the right context became apparent.

Murray and Herrnstein's other conclusions still stand: modern societies do an extremely good job of separating out the (relatively small, we now know) within-generation differences in intelligence, and directing people to different pursuits and occupations. The consequences is a shortage of general problem solvers in areas where intelligence is less valuable, while our institutions evolved in circumstances where general problem solvers were widely distributed.