August 1998

IQ Since "The Bell Curve"

Christopher F. Chabris

THIS PAST January, Governor Zell Miller of Georgia asked his legislature for enough money to give a cassette or CD of classical music to every newborn child in the state. The governor cited scientific evidence to support this unusual budget request. "There's even a study," he declared in his State of the State address, "that showed that after college students listened to a Mozart piano sonata for ten minutes, their IQ scores increased by nine points." And he added: "Some argue that it didn't last, but no one doubts that listening to music, especially at a very early age, affects the spatial-temporal reasoning that underlies math, engineering, and chess."

The so-called "Mozart effect" is one of the most publicized recent examples of our ongoing preoccupation with intelligence, a subject that not only refuses to go away but continues to raise whirlwinds of controversy. The largest such controversy, of course, surrounds The Bell Curve (1994), by the late Richard J. Herrnstein and Charles Murray. A mountain of essays and books purporting to refute that work and its conclusions grows and grows to this day. But now we also have the magnum opus of Arthur Jensen,1 a leading figure in IQ research and, like Herrnstein and Murray, a favorite target of academic liberals, as well as a posthumous volume by another leading IQ researcher, Hans Eysenck.2 So it is a good moment to look again at what we know, what we do not know, and what we think we know about this vexed subject.

IN The Bell Curve, Herrnstein and Murray set out to prove that American society was becoming increasingly meritocratic, in the sense that wealth and other positive social outcomes were being distributed more and more according to people's intelligence and less and less according to their social backgrounds. Furthermore, to the extent that intelligence was not subject to easy environmental control, but was instead difficult to modify and even in part inherited, genetic differences among individuals, Herrnstein and Murray posited, would contribute significantly to their futures.

The evidence for this thesis came largely from an analysis of data compiled in the National Longitudinal Study of Youth (NLSY), an ongoing federal project that tested over 10,000 Americans in 1980, with follow-up interviews regularly thereafter. Each participant completed the Armed Forces Qualifying Test (AFQT)--which, like any diverse test of mental ability, can be used as a measure of intelligence--and was then evaluated for subsequent social outcomes (including high-school graduation, level of income, likelihood of being in jail, likelihood of getting divorced, and so forth). As a rule, a person's intelligence turned out to predict such outcomes more strongly than did the socio economic status of his parents. This relationship held for all ethnic groups; indeed, when intelligence was statistically controlled, many "outcome" differences among ethnic groups vanished.

Herrnstein, a professor of psychology at Harvard with an impeccable reputation for scientific integrity, died of cancer just a week before The Bell Curve arrived in bookstores. This in itself may have had something to do with the frenzy of the public response. Had Herrnstein lived to participate in the debate, critics might have found the book harder to malign than it became when Murray, whose training was not in psychology but in sociology, was left to promote and defend it by himself.

Not that Murray, the author of Losing Ground (1984) and a vocal critic of the liberal welfare state, failed to do so energetically. But his lack of credentials as a hard scientist, and his overabundant credentials as a scourge of liberalism, made him a tempting target for an attack that was itself motivated as much by political as by scientific differences, and that was almost entirely focused on a side-issue in the book. That side-issue was differences in intelligence not among individuals but among groups--and specifically between whites and blacks--the degree to which those differences might or might not be explained genetically. So heated, and so partisan, was the furor at its peak that even President Clinton was asked about the book at a press conference. (He had not read it, but disagreed with it nonetheless.)

But the overreaction to what was in essence a moderate and closely reasoned book would also not have surprised Herrnstein in the least. If anything, it was a replay--actually, a more civilized replay--of what had happened to him after he published his first article on intelligence in the Atlantic in 1971. That article, entitled "IQ," besides bringing to public attention several points raised by Arthur Jensen in a 1969 paper in the Harvard Educational Review, offered a more speculative version of the argument that would be fleshed out and documented with NLSY data in The Bell Curve 23 years later.

Just as with The Bell Curve, only a small portion of Herrnstein's 1971 article dealt with differences among groups, and only a portion of that portion dealt with possible genetic influences on those differences; and, just as with The Bell Curve, these were the passages that received the greatest attention. In his article, Herrnstein concluded that "although there are scraps of evidence for a genetic component in the black-white difference, the overwhelming case is for believing that American blacks have been at an environmental disadvantage" (emphasis added). This did not stop one Nathan Hare from writing in response that "one would think that the pseudo-scientific generalizations surrounding race and IQ had long been put to rest. But the ghoulish die hard." Nor did it keep students at Harvard and elsewhere from putting up posters accusing Herrnstein of racism and calling him "pigeon-man" (in reference to his animal-learning research). His lectures were filled with protesters, and his speeches at other universities were canceled, held under police guard, or aborted with last-second, back-door escapes into unmarked vehicles. Death threats were made.

PEOPLE OFTEN react most defensively when challenged not on their firmly held beliefs but on beliefs they wish were true but suspect at some level to be false. This is the psychology behind the controversy that ensued after "IQ" in 1971 and The Bell Curve in 1994.3 On each occasion intemperate articles were written (some by the same people, barely updated), and the most strident positions were taken by those least qualified to comment on the science.4

By now, five major books have been published in direct response to The Bell Curve. Two of them, though critical, are within the bounds of reasonable discourse. Thus, Intelligence, Genes, and Success (1997), edited by four professors from the University of Pittsburgh who seem opposed to the book's public-policy conclusions, offers a fairly balanced range of scholarly views. On the sensitive question of heritability, what is especially notable is that the argument takes place mainly at the margins; although some of the book's contributors contend that the heritability of intelligence falls within a range lower than the 40-80 percent given by Herrnstein and Murray, that range is in every case much greater than zero.

A tougher line is taken in Inequality by Design: Cracking the Bell Curve Myth (1996), written by six Berkeley sociologists. This book addresses Herrnstein and Murray's main argument--that intelligence is an important determiner of social outcomes in America. To their credit, the authors do some old-fashioned hard work, reanalyzing the NLSY data and even making one correction that strengthens The Bell Curve's conclusions. But their main effort is to show, by adding variables other than parental socioeconomic status to the mix of factors predicting outcomes, that intelligence is not as important as The Bell Curve claims. Murray has since responded to this argument in a pamphlet entitled Income Inequality and IQ (published by the American Enterprise Institute); there, by considering only the NLSY data from sibling groups, within which parental background is by definition equal, he is able to show that intelligence still has very strong effects.

The conclusion one may reasonably draw from these two books, and from Murray's response, is that while intelligence may matter more or less than family background, it certainly matters, and that if it is not entirely heritable, it is heritable in some degree. It is useful to bear this in mind when considering the other three books, for one would scarcely know from reading them that such a view has any reputable backing at all. Though a few chapters in Measured Lies (1996), the most vituperative and scientifically irrelevant of the five volumes under consideration, attempt data-based argumentation, most settle for sarcasm, self-righteousness, and name-calling. And then there are The Bell Curve Debate and The Bell Curve Wars (both published in 1995); the former is an anthology of historical documents and reviews, mostly negative, which the editors rightly claim represent the general trend among responses to Herrnstein and Murray's book; the latter is a set of essays, also mostly negative, that originally appeared in a single issue of the New Republic when The Bell Curve was first published, with a few similar pieces added for effect.

According to its back cover, The Bell Curve Wars "dismantles the alleged scientific foundations . . . of this incendiary book." Since, however, the vast majority of those commenting on The Bell Curve in the anthology's pages have little or no scientific authority, whoever wrote those last words probably had in mind the single entry by the Harvard zoology professor Stephen Jay Gould. That essay, entitled "Curveball," was originally published in the New Yorker and appears both in The Bell Curve Wars and The Bell Curve Debate, occupying the first position in each. In it, Gould repeats many of the same accusations of racism and attributions of political motive that he made in his 1981 book, The Mismeasure of Man, written in response to the earlier controversy sparked by Jensen and Herrnstein.

WITHIN THE social-science community and the academic world in general, Gould's critique has been widely accepted as the canonical demonstration that the concepts of intelligence and its heritability are at best nonscientific and at worst racist and evil. (For instance, all of the contributors to Measured Lies who cite Gould's essay do so approvingly, if we count the one who asserts that it does not go far enough.) Indeed, so well has The Mismeasure of Man endured that in 1996 its publisher reissued it with a new introduction and appendices, including the ubiquitous "Curveball," but left the main text essentially unrevised.

Gould charges that the craniometrists of the 19th century, and later intelligence researchers as well, operated from racist assumptions, and implies that on those grounds their work should be ignored or even suppressed. Insofar as the charge is meant to include figures like Herrnstein and Murray, it is absurd as well as malicious. But even in those cases in the past in which racist assumptions can indeed be demonstrated, the proof of the pudding remains in the eating, not in the beliefs of the chef. Useful science can proceed from all sorts of predispositions; nor--it seems necessary to add--do the predispositions of scientists always point in the same direction, especially where discussions of human nature are concerned.

Before World War II, for example, the anthropologist Margaret Mead, presumably basing herself on her observations of non-Western cultures, wrote: "We are forced to conclude that human nature is almost unbelievably malleable, responding accurately and contrastingly to contrasting cultural conditions." Later, Mead admitted that what forced this conclusion was not the data she had collected but the political goals she espoused: "We knew how politically loaded discussions of inborn differences could become. . . . [I]t seemed clear to us that [their] further study . . . would have to wait upon less troubled times." As the shoot-the-messenger responses of Gould and others show, the times may still be too troubled for the truth.

But what about Gould's main scientific contention--that, as he puts it in his 1996 introduction to The Mismeasure of Man, "the theory of unitary, innate, linearly rankable intelligence" is full of "fallacies"?

The theory that Gould is attacking usually goes under the name of general intelligence. Its advocates, practitioners of the hybrid psychological-statistical discipline known as psychometrics, argue simply that while individuals differ in their abilities in a wide range of intellectual realms, a relationship exists among these variations that can be attributed to a common factor. This common factor is what the psychometricians label general intelligence, or g.

A brief example will illustrate the evidence they adduce for this proposition. Suppose a group of students takes a set of ten, timed mental-ability tests, five based on verbal materials (such as choosing antonyms) and five based on spatial materials (such as drawing paths through mazes). Each student will receive ten scores, and each student will have a unique profile of scores, higher on some tests than others.

Now suppose we correlate mathematically the students' scores on the five verbal tests. We will probably find them positively, though not perfectly, correlated--that is, the score on one will predict reasonably well the scores on the others. With the aid of a statistical procedure known as factor analysis, we can examine the pattern of these positive correlations and infer that they can be explained by the existence of a common factor, the most logical candidate being the "verbal ability" of the students who took the tests. Analogous results would likely occur if we factor-analyzed the set of five spatial tests.

What if we combined all ten tests in a single analysis, looking at all the possible correlations? Most likely we would find separate verbal and spatial factors at work. But those factors themselves will almost always be correlated. A superordinate, or "general," factor--g--can then be extracted to account for the commonalities across all the tests, though this factor will be revealed more by some tests than by others; such tests, known as "highly g-loaded," are taken as especially good measures of general intelligence.

TO THE extent that it is not simply political, the debate that followed The Bell Curve and "IQ," and that lies at the heart of Gould's critique in The Mismeasure of Man, is over the very existence and coherence of general intelligence. Each side has made the same points over and over, and each side believes it has refuted the other side's arguments. The reason this is so is that the two sides proceed according to different definitions of intelligence.

The psychometric camp, which includes Herrnstein and Murray, Jensen, Eysenck, John Carroll (whose 1993 treatise, Human Cognitive Abilities, offers the most extensive factor-analysis of mental tests), and most psychologists who have traditionally studied the topic, hold to a conception of intelligence that closely matches what common sense and the dictionary tell us the term means. The opposing side, which sports a more eclectic set of disciplinary backgrounds and prides itself on a more sophisticated and inclusive perspective, divides human abilities into broad classes--logical, spatial, interpersonal, verbal, etc.--and labels each class an "intelligence." The two sides then proceed to talk past each other.

Scientists make bad dictionary writers and worse philosophers. Their main skills are in constructing experiments and generating explanations for what they observe. Neither of these endeavors requires agreement on what the words involved "mean" in any deep or absolute sense, only on ways of converting the elements of the theory at issue into operations that can be carried out in an experiment and repeated later if necessary. Measurement is the most important such operation; as Kelvin pointed out long ago, without a way to measure something it cannot be studied scientifically.

This is why the oft-repeated phrase, "intelligence is nothing more than what intelligence tests measure," is, as an objection, merely a tautology. The truth is that as long as intelligence can be reliably measured--it can be, with a variety of tests--and validly applied--it can be, to predict a variety of outcomes--it is intelligence. If we suddenly started calling it "cognitive ability," "cognitive efficiency," or even "the tendency to perform well on mental tests," it would still have the same scientific properties. Nothing about the natural world would change.

One way to test the schemes of the proponents of "multiple intelligences" would be to apply research techniques that might (or might not) suggest a role in them for general intelligence. But this is an exercise the advocates of multiple intelligences tend to rule out of consideration a priori. Thus, as Howard Gardner correctly notes in Frames of Mind (1983), there is good evidence that different parts of the brain are responsible for different abilities. However, when at a recent seminar a member of Gardner's research group was asked how abilities in the various intelligences are measured, the swift response was, "We don't measure them."

The reason is obvious: any reasonable system of measurement would produce a set of scores whose correlations could be calculated, and the pattern of those correlations would likely reveal a common factor--in other words, g--accounting for some fraction of the total variation. Gardner's theory is very popular in educational circles these days, as is the idea, espoused by Daniel Goleman in Emotional Intelligence (1995), that skill at managing one's own emotions and interpreting those of others is very valuable to social interaction and success in life. Surely both of these ideas are correct, as far as they go. But neither one of them addresses intelligence in a complete way.

ANOTHER CRITICISM of the notion of general intelligence is that it is based on factor analysis, an indirect procedure that deals with the structure of tests rather than the nature of the mind and brain. This is a point raised with special vehemence by Gould. In The g Factor: The Science of Mental Ability, Arthur Jensen shows that the objection is without foundation.5

The g Factor is a deep, scholarly work laden with hundreds of tables, graphs, and endnotes, some of them with tables and graphs of their own. It is balanced and comprehensive, summarizing virtually all the relevant studies on the nature of intelligence and demolishing most of the challenges and alternative explanations of the major findings. (It is not, however, an easy book for nonspecialists to read, which is why we are also fortunate to have Hans Eysenck's much more accessible and even entertaining Intelligence: The New Look.)

In refuting Gould's point, Jensen demonstrates that mental-test scores correlate not just with one another but with many measures of information-processing efficiency, including reaction time (how quickly you can press a button after a light flashes), inspection time (how long two separate line-segments must be displayed for you to judge accurately which is longer), and working-memory capacity (how many random items of information you can remember while doing something else). Jensen also reviews the many direct biological correlates of IQ, such as myopia (a very heritable condition), brain electrical activity, estimates of nerve-conduction velocity (the speed at which brain cells communicate with one another), and the brain's metabolism of glucose. Even brain size, the study of which is richly derided by Gould, has been found with modern imaging technology to correlate with IQ.

These chapters, among the most impressive in Jensen's book, put general intelligence as a psychological trait on a more solid foundation than is enjoyed by any other aspect of personality or behavior. They also speak persuasively to the issue of its heritability, the argument for which becomes more plausible to the extent that intelligence can be associated with biological correlates.

One can go farther. To Stephen Jay Gould and other critics, belief in the heritability of intelligence is inextricably--and fatally--linked to belief in g; destroy the arguments for one and you have destroyed the arguments for the other. But as Kevin Korb pointed out in a reply to Gould in 1994, and as Jensen affirms here, the g factor and the heritability of intelligence are independent concepts: either hypothesis could be true with the other being false. In some alternate reality, intelligence could be determined by wholly uncorrelated factors, or for that matter by wholly environmental (i.e., nonheritable) factors. It is simply less of a stretch to imagine that a general factor both exists and is somewhat heritable, since, as Jensen shows, this combination describes our own reality.

STILL ANOTHER line of attack used by the detractors of g is to point to studies allegedly showing that intelligence is easy to change (and, therefore, a meaningless concept). Arthur Jensen raised a firestorm three decades ago when he asked, "How much can we raise IQ and scholastic achievement?" and answered: not much. This brings us back to the Mozart effect, which purports to do in ten minutes what years of intensive educational interventions often fail to accomplish.

The Mozart effect was first shown in a study by Frances Rauscher, Gordon Shaw, and Katherine Ky that was reported in the British journal Nature in 1993. It is difficult to determine their experimental procedure with precision--their article was less than a page in length--but the essentials appear to be as follows. Thirty-six college students performed three spatial-ability subtests from the most recent version of the Stanford-Binet intelligence test. Before one of the tests, the students spent ten minutes in silence; before another, they listened to ten minutes of "progressive-relaxation" instructions; and before still another, they listened to ten minutes of Mozart's Sonata for Two Pianos in D Major (K. 448). The subjects performed the tests in different orders, and each test was paired with equal frequency against each listening option. The results, when converted to the scale of IQ scores: 110 for silence, 111 for relaxation, and 119 for Mozart.

"Mozart makes you smarter!" said the press releases as new classical CD's were rushed to market. A self-help entrepreneur named Don Campbell trademarked the phrase "The Mozart Effect," published a book by the same name, and began selling cassettes and CD's of his own, including versions designed specially for children. Frances Rauscher testified before a congressional committee and gave many press interviews.

What was wrong with this picture? The article in Nature did not give separate scores for each of the three Stanford-Binet tasks (necessary for comparative purposes), and it used dubious statistical procedures in suggesting that listening to Mozart enhanced overall "spatial IQ" or "abstract reasoning." Nor did the researchers analyze separately the first task done by each subject, to rule out the possibility that prior conditions may have influenced the Mozart score. Finally, they claimed that the effect lasted for only ten to fifteen minutes, but gave no direct evidence; since the subjects were apparently tested only immediately after each listening episode, there was no way to see how this interval was calculated.

IN AN attempt to reproduce the finding that classical music enhances "abstract reasoning," Joan Newman and her colleagues performed a simple experiment: each of three separate groups comprising at least 36 subjects completed two separate subsets of Raven's Matrices Test (a good measure of g) before and after listening to either silence, relaxation instructions, or the Mozart-effect sonata. All three groups improved from the first test to the second, but by the same amount; in other words, Mozart was of no particular help. In another experiment along the same lines, a group led by Kenneth Steele asked subjects to listen to ever-longer strings of digits and repeat them backward; it, too, found no benefit from prior exposure to Mozart. Other independent tests reported similar failures or equivocal results.

In response to these experiments, Rauscher and Shaw have considerably narrowed the scope of their original findings. They now concede that the post-Mozart increase in spatial performance occurred on just one of the three Stanford-Binet tasks, while on the others, varying the listening condition made no difference. According to their revised estimate, only "spatiotemporal" tasks, which require the transformation of visualized images over time, are affected by complex music, not spatial ability or reasoning in general.

Unfortunately, however, neither Nature nor any journal of similar stature has given space to the follow-up experiments, most of which have been reported in Perceptual and Motor Skills or other low-prestige journals that many psychologists never read. And the media have of course moved on, leaving the babies of Georgia with state-sponsored gifts and the public with the vague idea that if ten minutes of music can "make you smarter," then IQ cannot signify very much.

Similarly feeding this mistaken impression are other recent examples of brief treatments affecting performance either negatively or positively on IQ-type tests. Thus, the researchers Claude Steele and Joshua Aronson told one group of black undergraduates at Stanford that a difficult verbal test would diagnose their abilities and limitations, and another group that their answers would be used only for research on verbal processing. In four separate experiments, students did worse under the former conditions than under the latter. Analogous results have been obtained with Asian female students in math tests: stressing to them that the test measures the abilities of their sex reduces their scores (women typically do worse than men in math), but stressing that it measures the abilities of their ethnic group increases their scores (Asians typically do better than other groups). But as Jensen points out, in such cases we are dealing with a stereotype about group differences that serves to increase or decrease test anxiety. That performance goes down when anxiety gets too high is a common enough finding in testing research, and says nothing about g.

What all these experiments do illustrate is that the human brain is a dynamic system whose functioning can change quite quickly. But this is not the same thing as changing intelligence itself. A few weeks of Prozac or another modern antidepressant can radically alter a person's behavior, but we still accept that his basic identity has not changed--he is still the man we knew before. Intelligence, too, is a stable characteristic of a person's behavior across a wide range of situations. He will be able to perform a task much better in one context than in another, with special training than without; but he is still the same person, and his intelligence is also still the same. Although the Mozart effect was promoted as though it were bad news for The Bell Curve and IQ, it is not.

AND NEITHER, finally, is the much-talked-about "Flynn effect." Over the past 50 years, the average score on intelligence tests has risen about three points per decade. This means that we are all getting smarter--indeed, the average adult of 1998 is, psychometrically at least, smarter than 84 percent of the population was in 1948.

The Flynn effect is named after James Flynn, who has been studying it for over fifteen years (although the phenomenon had been noted as early as the 1930's). In a chapter of a new book, The Rising Curve: Long-term Gains in IQ and Related Measures,6 Flynn notes that gains are occurring steadily in every country sampled, mainly in the West and the industrialized parts of Asia, though their size and specific nature varies in different cases. He believes the start of the increases coincided with industrialization in the 19th century, though the data are of course less reliable the farther back one goes. What he does not know is why the gains have been occurring, and the other contributors to The Rising Curve can offer only tentative theories at best.

Psychologists, like all scientists, prefer to test their theories with controlled experiments, but such experiments cannot be performed when the phenomenon to be explained is occurring throughout the world continuously over time. The difficulty in comparing times is that many things have changed with the times: the items on IQ tests are different; education is different; nutrition is better; airplanes, cars, radio, television, movies, computers, and the Internet have been invented; society has become more permissive and also more rewarding of risk-taking; testing is more widespread and people are more accustomed to being tested; birth rates are lower; and so on. Encompassing all of these time-correlated variables, the change in what might be called our cognitive environment has been simply tremendous over the past 150 years. The most relevant factors here are probably better nutrition--a topic Eysenck studied at the end of his career--plus greater, more diverse, and more complex stimulation of the brain by our everyday experiences.

Evidence of such a dramatic environmental effect on IQ scores should reinforce skepticism concerning a genetic basis for group differences. But, in any case, psychometric theory makes no claims about average or absolute levels of intelligence within or between populations, and behavioral genetics allows for complex environmental influences on traits that are still significantly heritable. And so, again contrary to popular belief, the concept of general intelligence remains as sound and as meaningful as ever, Flynn effect or no.

HAVING WITHSTOOD so many attacks, will the psychometric study of intelligence survive? Alas, not necessarily. In a pattern reminiscent of an earlier episode in the annals of modern psychology, the impact of Stephen Jay Gould's critique has been reinforced by the lack of a forceful response to it by psychometricians themselves, leaving the impression even within psychology at large that general intelligence has been routed.

Just as Gould, a paleontologist, has chided psychologists for misunderstanding genetics, so, in a review of B.F. Skinner's Verbal Behavior in 1959, the linguist Noam Chomsky chided behavioral psychologists for misunderstanding language. Like Gould, who has caricatured and ridiculed the notion of general intelligence and the factor analysis used to document it, Chomsky caricatured the tenets and methods of behaviorism, which argued that the task of psychology is to measure only behavior and to explain it only in terms of environmental and genetic causes, without referring to what goes on inside the head.

It took eleven years before a leading behaviorist, Kenneth MacCorquodale, answered Chomsky; the reason none of his colleagues had bothered to reply earlier, he explained, was that they found Chomsky's arguments simply uninformed and irrelevant to the work they did. In the meantime, however, Chomsky's review was widely read and subscribed to by the new wave of cognitive psychologists who were building a framework for psychology that remains dominant today.

Gould's book seems to have had a similar effect on young intelligence researchers. Although Jensen and several others did review The Mismeasure of Man very negatively at the time it appeared, like MacCorquodale they replied in obscure journals read mainly by their own supporters. Thanks in part to Gould's influence (and, of course, to the outrage directed against Jensen and Herrnstein in the 70's), the most popular new theories in the 1980's came to minimize the role of general intellectual ability in favor of other factors, to posit multiple "intelligences," and to give little attention to heritability. Now Eysenck, one of the heroes of psychometrics, and Herrnstein, one of its leading supporters, have died, Jensen and Carroll are approaching the end of their careers, and the psychometricians risk going into the same sort of extended bankruptcy proceedings as the behaviorists before them.

The great irony is that this is occurring just as the field of behavioral genetics has begun to thrive as never before. One of its most striking successes has been to document, through the convergence of numerous family and twin studies, the heritability of intelligence. Now researchers have been able to identify a specific gene whose variations are associated with differences in intelligence. This is a crucial step in building a complete theory of intelligence that can explain individual differences in biological as well as psychological terms. But the new generation of cognitive scientists, who focus on characteristics of the mind and brain that are common to everyone, are not too interested in differences among people, while the psychometricians, who stand to be vindicated, have been sidelined on their own playing field.

THE MOST basic claim put forth by Herrnstein and Murray was that smart people do better than dumb people. What is so troubling about that? We rarely encounter an argument over the fact that beautiful people do better than ugly people, or tall people better than short ones, though each of these propositions is also true. Is an intellectual meritocracy less just or moral than a physical one?

The answer, unfortunately, is that whenever intelligence is said, "race" is heard; whenever race is said, "genetics" is heard; and whenever genetics is said, "inferiority" is heard--even though these issues are not necessarily connected in any way. When I mentioned to friends that I was writing an article on intelligence, many were surprised, and some wanted to know why. I can only imagine how Herrnstein was treated by his colleagues during the last 25 years of his life. The public protests may have bothered him less than the fact that people in his own community never thought of him in the same way again: he had disturbed a pleasant conversation by bringing up unpleasant facts.

Since The Bell Curve, intelligence is stronger than ever as a scientific concept, but as unwelcome as ever as an issue in polite society. It would be reassuring to think that the next twenty years, which promise to be the heyday of behavioral genetics, will change this state of affairs. But if the past is any guide, many more phony controversies lie ahead.

CHRISTOPHER F. CHABRIS, here making his first appearance in COMMENTARY, is a Ph.D. candidate at Harvard specializing in cognitive neuroscience. He is at work on a book about the chess-playing machine Deep Blue and artificial intelligence.

1 The g Factor: The Science of Mental Ability. Praeger, 672 pp., $39.95.

2 Intelligence: The New Look. Transaction, 232 pp., $29.95.

3 For Richard J. Herrnstein's account of what happened after he published his Atlantic article, see his "On Challenging an Orthodoxy," COMMENTARY, April 1973. For Charles Murray's account of the later controversy, see his "The Bell Curve and Its Critics," COMMENTARY, May 1995 and subsequent letters, August 1995.--Ed.

4 In the case of The Bell Curve, a special committee set up by the American Psychological Association to report on the basic science eventually backed all of the book's main claims, as did an open letter signed by several dozen of the nation's most qualified intelligence researchers.

5 Jensen's work should not be confused with another of almost the same title, The g Factor: General Intelligence and its Implications, by Christopher Brand. This provocative but worthy book was published in the United Kingdom early in 1996 and was withdrawn within two months, after negative media coverage and a frenzy reminiscent of the early 1970's. The publisher, Wiley, also canceled the book's distribution in the United States before any copies went on sale. Brand has since been fired from his teaching position at Edinburgh University, and has yet to find a new publisher.

6 Edited by Ulric Neisser. American Psychological Association, 400 pp., $39.95.