'The Bell Curve' Turns 20
This month marks two decades since the publication of The Bell Curve, the massive--and massively controversial--ode to intelligence testing from Harvard psychology professor Richard J. Herrnstein (who passed away shortly before the book hit shelves) and American Enterprise Institute fellow Charles Murray. The book was carefully written and packed with original statistical analyses, graphs, and notes and appendices. It was also sure to provoke outrage: One late chapter focused on genes and the black-white IQ gap, and the authors acknowledged that many readers would probably skip right to it. Herrnstein and Murray knew what they were doing, in more ways than one.
But two decades later, the topics here are as important and feared as ever. As technology becomes ever more complicated, and with an age of robotics on the horizon that could upend the labor market, we wonder if there will long be a valued place in society for people with low intelligence. Many already see a college degree as a necessity for a decent life. The continuing rise of cheap travel, opportunities for women, and college attendance has enabled the brightest people to increasingly segregate themselves socially, solidifying the "cognitive elite" Herrnstein and Murray wrote about. Racial and class gaps in test scores haven't changed much and have often gotten worse, No Child Left Behind and Head Start be damned.
With the immediate furor over the book having subsided, and yet with the subject of intelligence retaining its quiet power over modern discourse, it's worth revisiting The Bell Curve. Some of it hasn't aged well. But even in light of new evidence, the book offers a convincing argument that intelligence, as measured by IQ tests, is a significant contributor to the social patterns we see around us.
The key insight behind IQ is that most mental tests have something in common: A person who does well on one test is likely to do well on others too, even if the tests are quite different in content. The correlation, though far from perfect, allows psychologists to statistically identify the "factor" that the tests share, and to measure it in individuals. There's a lot of debate over what exactly this factor--g, or general intelligence--represents, and whether it's better to break intelligence up into smaller categories, such as "crystallized" and "fluid" (terms that refer, roughly, to general knowledge and problem-solving ability).
But whatever IQ is, it's difficult to change even with concerted effort, it relates to the ability to process complex information, and differences between individuals are partly genetic. All this was already well-established by the time of The Bell Curve's publication. Most importantly, studies of twins and adopted children had shown that sharing genes tends to give children similar IQs, even if the kids don't also share a home environment. Herrnstein and Murray reported estimates that IQ is 40 to 80 percent heritable, and assumed 60 percent in their calculations.
Twenty years on, the evidence has gotten better, even if it remains difficult to nail down a precise number. ("About half," wrote three psychologists in Slate last month.) One recent study, rather than focusing on siblings, analyzed the genomes and intelligence-test scores of unrelated individuals, looking to see if individuals who were similar genetically were also similar in intelligence. They were. The researchers established lower-bound estimates that crystallized intelligence is 40 percent heritable while fluid intelligence is 51 percent heritable. "Our results unequivocally confirm that a substantial proportion of individual differences in human intelligence is due to genetic variation," the team wrote.
Intelligence seems to be the product of a large number of genes, each of which has very little influence on the final result. This makes it difficult to tell exactly which genes matter and how they work, but progress is steady. Just last month, a study identified three genetic variants that may increase IQ scores by about 0.3 points apiece by affecting a "neurotransmitter pathway" in the brain.
But even those who admit IQ is a measurable human trait might deny that it matters. After all, one big predictor of where kids end up is how rich their parents are. Can intelligence really make a difference when class is so strong?
To find out, Herrnstein and Murray turned to the National Longitudinal Survey of Youth, a government-funded study that has followed thousands of Americans since they were adolescents and young adults in 1979. (Another wave began in 1997.) Herrnstein and Murray estimated IQ scores based on these folks' performance on the Armed Forces Qualifying Test, a battery of cognitive exams, and measured parental socioeconomic status (SES) by constructing an index that took into account income, occupation, and education. And to remove the effects of race, a topic they avoided until the book's later chapters, Herrnstein and Murray focused only on whites.
From there, it's a statistical clash of the titans: The IQ and SES data are entered into a single equation (along with age) and used to predict life outcomes such as wages, welfare dependency, out-of-wedlock childbearing, and crime. Overwhelmingly, IQ turns out to be a better predictor than SES. In other words, people with high IQs tend to have better outcomes, even after SES has been accounted for--and the results suggest that if you find yourself in a science-fiction movie and get to choose between a high IQ and high-status parents, you should go with the former.
Of course, that people with high IQs have better outcomes doesn't prove that high IQs cause better outcomes. But for many of the topics The Bell Curve covers, skepticism along these lines borders on the absurd. Take wages, for instance: It is obvious from casual observation, and supported by research, that many of the highest-paying occupations (doctor, lawyer, accountant) require an ability to deal with complex information. IQ also has clear benefits when it comes to getting into and graduating from a demanding college--kids with higher IQs do better in school at all levels, and the standardized tests colleges use to screen applicants are highly correlated with IQ--or even performing well in a more typical work environment.
And even in the absence of causation, correlations can matter. We might be less than positive that low IQ causes welfare dependency or crime, for instance--but someone running a program to help the poor might want to know (from the National Longitudinal Survey) that 45 percent of women who've received welfare and 62 percent of men interviewed behind bars are in the bottom 20 percent of the IQ distribution.
The Bell Curve's findings on life outcomes were subject to some criticism, but none of it invalidated the basic point that IQ matters. Here, for example, is a paper that used a more sophisticated measure of SES (including family structure). The result? "Parental family background is at least as important, and may be more important than IQ in determining socioeconomic success in adulthood." So, the clashing titans might be more evenly matched than Herrnstein and Murray claimed, and if you find yourself in that science-fiction movie maybe you should just flip a coin.
Others noted that, statistically, IQ explains a fairly low percentage of the variation in outcomes. This isn't a criticism of the book so much as a quote from it. ("For virtually all of the topics we will be discussing, cognitive ability accounts for only small to middling proportions of the variation among people. It almost always explains less than 20 percent of the variance, to use the statistician's term, usually less than 10 percent and often less than 5 percent.") But more to the point, if we're supposed to ignore IQ because IQ isn't statistically powerful enough, we should probably ignore SES as meaningless too.
Of course, few remember The Bell Curve as the book that made the case for IQ by analyzing an all-white sample: They remember it as the book that argued it was "highly likely" that the sizable black-white gap in IQ scores is partly genetic. On this, The Bell Curve doesn't stand the test of time--in 1994 it was very difficult to argue one way or the other, because our knowledge of the human genome was weak and the other ways of approaching this question are unsatisfactory.
Herrnstein and Murray note, for example, that controlling for parental socioeconomic status eliminates only about a third of the black-white gap--but this is relevant only if blacks and whites with similarly situated parents tend to have similar environments otherwise. They don't, for any number of reasons. (Conservatives and even some liberals note the cultural legacy of slavery and Jim Crow, and liberals have pointed out that blacks with middle-class incomes are far more likely than comparable whites to live in poor neighborhoods. Also: racism.) Herrnstein and Murray further note that the gap is biggest on the tests that most closely measure g, but as they concede, this doesn't prove that g has been suppressed by genes specifically.
But if the arguments in the book aren't looking so hot these days, many arguments against the book are doing no better. A common refrain in the wake of The Bell Curve's publication--it was still very much in vogue when I attended college in the mid-2000s--was that the racial IQ gap couldn't be genetic because race itself is a mere "social construct" with no genetic basis. This theory has been thoroughly discredited (to borrow a word from The Bell Curve's least thoughtful critics) in the years since. We now know that some genetic variants are more common in some racial groups than in others; what we don't know is whether or how those variants influence intelligence. Even those who cling to the idea of race as a "social construct" often admit this, making arguments pertaining to the semantics of the word "race," not to the science of human differences.
Two decades ago we weren't close to knowing how the brain has evolved in the tens of thousands of years since a small groups of humans left Africa to populate the rest of the globe. We're a lot closer now, but there's still a long way to go. Rather than speculating, we should stay calm, wait for the science to sort itself out, and remember that statistical group differences can't justify the poor treatment of individuals.
Another controversial area Herrnstein and Murray wade into is the question of whether human beings are getting smarter or duller over time. Certainly, IQ scores are rising, at a rate of a few points per generation--a phenomenon the authors christened "the Flynn Effect" in honor of James Flynn, an intelligence researcher who has studied it in depth. This rate of change is so high that it's certainly due to environmental changes, not genetics--and it's possible that humans are actually getting duller on the genetic level. Herrnstein and Murray argue this is indeed the case, pointing out that lower-IQ women tend to have more children than their higher-IQ counterparts. Genetically, they suggest, we're losing at least an IQ point per generation.
Unfortunately, they provide policy recommendations to go along with this analysis. One is that we should cut poverty relief for poor mothers to avoid encouraging them to have children. You don't have to be chief of the PC Police--or lack concern about the welfare state's tendency to enable irresponsible decisions--to find this horrifying.
Nonetheless, research and public policy proceed apace. Flynn himself, in his 2007 book What Is Intelligence?, cataloged a variety of odd patterns in the IQ data (such as that scores are rising on some subtests but not others) and explained in detail how the modern environment may have caused the effect that bears his name. A paper last year attributed the IQ gains to improvements in the way people apply rules and heuristics to problems. As for genetics, recent research has speculated that human brain size is falling because IQ is no longer the boon to survival and reproduction that it once was, and even that modern people have lower IQs in some ways than the Victorians did. On the policy level, another of Herrnstein and Murray's suggestions for improving societal IQ--making birth control easily and cheaply available--has become official government policy and remains a cause for many, albeit for reasons (hopefully) having nothing to do with intelligence.
There is much more of interest in The Bell Curve: Extensive data on affirmative action, foreshadows of the arguments Murray would make in Real Education and Coming Apart, a libertarian vision of how to create communities where people with low IQs have a place. It is worth reading or re-reading today. With 20 years' hindsight, the book's flaws are as apparent as ever. But so are its merits.