his massive but well-written study by the late Harvard psychologist Richard Herrnstein and public policy analyst Charles Murray is about the role of intelligence in the social structure of American society. Their discussion is organized into four parts. In part 1, they piece together a picture of the emergence of a cognitive elite in American society in the course of this century, through the increasing concentration of individuals with high IQ into the ranks of the educated elites and into high prestige occupations. This trend is driven by the reduced importance of social origin as the major factor of class distinctions and the increasing value of intelligence in the marketplace of modern industrial society, due to the increasing scale of organizations and markets combined with the intellectual demands of complex technologies and intricate regulatory structures. The authors view the cognitive elite as becoming increasingly separated from the rest of the population by a widening economic gap and increasing physical separation in the workplace and in residential communities. Through assortative mating, this process of "cognitive partitioning" is viewed as leading to the creation of a quasi-hereditary cognitive elite. This macrosociological description constitutes an updated and more richly documented version of the earlier work by Herrnstein (1973) on IQ in a meritocracy; sociologists will also find this discussion reminiscent, in a less fictional register, of the work of Michael Young (1958).
In part 2, Herrnstein and Murray present their own empirical analyses of data from the National Longitudinal Survey of Youth (NLSY), based on a longitudinal survey of about 12,000 youths aged 14 to 22 in 1979. Their analytical strategy is to estimate logistic regressions of a wide variety of dichotomous socioeconomic outcomes (dropping out of school, graduating from college, being in poverty, being unemployed, committing a crime, having a child out of wedlock, etc.), showing that in most cases intelligence, measured as the score on the Armed Forces Qualification Test (AFQT), has a stronger effect on the outcome than a composite measure of SES (based on family income and parents' education and occupational prestige). These analyses are carried out with non-Hispanic whites only, so issues of race or ethnicity are not addressed at this stage of the analysis.
Part 3 introduces issues of race and ethnic differences in IQ. The authors review the large literature showing the average IQ of blacks as being about one standard deviation lower that the white average, and the East Asian average (in the U.S. and abroad) as several points higher than the white mean. While maintaining a position of official agnosticism with respect to a possible genetic basis for these group differences in IQ, the authors suggest that the detailed pattern of the differences - such as the systematic difference in intellectual profile between East Asians and whites, with an East Asian advantage concentrated in the nonverbal intelligence scores, and the finding that black and white scores differ most on the most g-loaded tests (i.e., those most related to a general intelligence factor) - strongly points to genetic roots. They go on to estimate the effect of IQ differences on average group differences in a variety of socioeconomic outcomes in the NLSY data. For many socioeconomic outcomes, such as family income and proportion below the poverty line, controlling for IQ substantially reduces the gap between blacks and whites. In some cases, such as wages, the discrepancy vanishes almost entirely when IQ is controlled. For educational attainment and access to professional occupations, however, the gap is reversed so that controlling for IQ, blacks have higher educational and occupational achievement than whites. This part of the book also discusses dysgenic effects of the higher rate of births among people of lower cognitive ability and the apparently contradictory phenomenon of the rise in average IQ that took place in industrial societies during the past few decades (the Flynn effect) and documents from the NLSY data the higher prevalence of low cognitive ability among people visited by a variety of social problems from dropping out of school to ever being interviewed in jail.
Part 4 deals with topics variously related to policy. The authors document the failure of many attempts to raise IQ. While better nutrition may play a role in the historical upward trend in IQ, formal schooling and special programs, notably Head Start, do not seem to raise IQ. Some highly intensive (and therefore costly) programs directed at young children, as well as adoption, might produce an IQ improvement. The authors credit affirmative action in higher education for opening college opportunities for more blacks and Latinos than would have otherwise been there, but spell out the downside of affirmative action programs in producing unfairness, distrust in the larger society of the achievement of minority students, and lowered productivity because of the suboptimal matching of ability and occupation. With respect to affirmative action in the workplace, they call for "scrapping the existing edifice of job discrimination law." In the penultimate chapter, the authors amplify their pessimistic view of a future "custodial state" in which an isolated cognitive elite (now merged with the affluent) protects itself in gated communities from the underclass at the bottom of the cognitive ability distribution, whose leadership has been drained and whose quality of life deteriorates further in the urban ghettos in which it is concentrated. The last chapter of The Bell Curve offers the authors' hopeful vision of an alternative society based on a return to the central role of local communities in which neighbors take care of each other, rules of conduct are made simple to understand and to follow, and everyone finds a "valued place" irrespective of their ability. This vision of vibrant local communities is hardly sinister, but the authors do not say how the devolution to local structures would happen, or even what trend might possibly lead in that direction. This exercise in utopia is largely independent of the rest of the book.
The Bell Curve has acquired its aura of controversy by challenging deeply held beliefs that make up the ethical-political credo of a large segment of the media and the public, including a large proportion of the sociological profession. While many people accept intuitively, at least in private, the notion that there is such a thing as intelligence, and that people differ in how intelligent or "smart" they are, the notion that part of the difference in intelligence among individuals might have a genetic basis is harder to accept, as it seems to imply a greater inevitability to the inequality in the distribution of ability than if cognitive ability were determined by purely environmental factors that can be changed (it is hoped) through deliberate policy intervention. The idea that average intelligence varies among racial and ethnic groups, and that part of this variation might be genetic is the most scandalous of all. Although the authors do not claim to be certain about a genetic basis of group differences, they review the compelling circumstantial case for the existence of such a genetic basis, thereby committing the ultimate blasphemy against the dominant egalitarian social philosophy of our times. Many sociologists who already find such ideas disturbing as ordinary citizens may be additionally predisposed against attributing a central role to intelligence in explaining social organization by the preposterous notion that social facts should only be explained in terms of other social facts, so that intelligence - being a psychological, not a social, concept - does not constitute a legitimate sociological explanation of social reality.
Amidst the buzz and gossip that greeted the publication of the book in the fall of 1994, there was an almost palpable expectation of a critique that would deal the book a decisive blow by uncovering the "smoking gun," the irreparable defect that would invalidate the book as a whole and thereby save the conventional orthodoxy. Even highly competent reviewers in the social sciences have not discovered this fatal weakness, some contributing valid criticisms (e.g., Goldberger & Manski n.d.; Heckman 1995), others expressing their disagreement with little more than invectives (Hauser 1995). The task of hostile critics is made difficult by the fact that much of The Bell Curve consists of reviews of existing literatures, including the massive literatures on behavioral genetics and the psychometric properties of IQ tests. The authors' interpretation of these bodies of work is by no means outlandish: their summation corresponds by and large to a cautious and even conservative interpretation of the main thrust of the literature, for example in their working estimate of the heritability of IQ in the white population at .6, which is on the low side relative to current estimates in behavioral genetics (Scarr 1994-95). Many of the authors' conclusions such as those concerning the unbiasedness of IQ tests, the heritability of IQ, and the meaning of race and ethnic differences are widely accepted findings within the specialized literatures in which they originated. They appear novel and controversial only because the bulk of this research has been carried out away from the public eye and from the attention of many social scientists. The results of this research have been hidden in specialized journals, to emerge to collective consciousness only during brief episodes of intense public debate of the kind that greeted the work of Jensen (1969) or Herrnstein (1973). With scattered exceptions (e.g., Eckland 1967; Gordon & Rudert 1979; Scarr & Weinberg 1978), these findings have not been well disseminated in the sociological literature. It has been hard for critics of The Bell Curve to challenge convincingly the parts of the book that report well-established findings of sophisticated bodies of research.
The other main source of evidence used by the authors is their own analysis of the NLSY data. In a typical analysis, they estimate the coefficients of the logistic regression of a dichotomous outcome, such as dropping out of high school, on IQ., SES, and age (as a control for the dependence of the AFQT measure of IQ on age). They present the results of each regression graphically, by contrasting on the same graph the curve of the estimated probability of the outcome as IQ ranges from -2 to +2 standard deviations, with SES and age kept constant at their mean sample values, with the curve of the estimated probability of the outcome as SES ranges from -2 to +2 standard deviations, with IQ and age kept constant at their mean sample values. They show that, for many outcomes, the curve representing the probability of the outcome as a function of IQ is steeper than the corresponding curve for SES, and conclude that IQ is a more important factor than SES in determining the outcome. As noted by some reviewers (e.g., G&oldberger & Manski n.d.), this effective graphical presentation is equivalent to the use of standardized coefficients to compare the relative effects of independent variables measured in different metrics. While many sociologists would go on from here to estimate more complex models including larger sets of independent variables, the simple analysis serves well to reinforce the authors' central contention that IQ is a vastly better predictor of important aspects of social performance than SES, a variable that is the centerpiece of so many sociological explanations of such outcomes.
The larger scientific context in which the authors present their empirical findings is unusual, compared to many previous debates in the social sciences, in that the data they use are so publicly accessible. A CD ROM with the NLSY data can be obtained for a small sum, and many readers of this review can easily replicate the authors' analysis on their own personal computers. Social science research with controversial implications has rarely been as open and aboveboard as this, and one expects that many replications and extensions of the authors' analysis will appear in the future. The authors' analysis has been criticized on the grounds that the composite SES measure is unlikely to capture the entire childhood experience of subjects (which is a criticism of the concept of SES in general, not only the authors' use of it), and that their regression models are underspecified, in particular by not including education as an independent variable to predict other outcomes, since education could affect these outcomes independently of IQ (Heckman 1995). Some critics have also noted that the pseudo-[R.sup.2]s that measure the fit of the logistic models are low for some outcomes. (The measure of fit is substantially higher for educational outcomes, such as dropping out of high school or graduating from college, and for some socioeconomic outcomes such as welfare dependency, than for the other outcomes analyzed by the authors.) Here also, critics are unlikely to find a smoking gun, because the strong effect of IQ that the authors find in their simple models is quite robust, as researchers familiar with the NLSY data know. No amount of nitpicking at the details of the authors' analysis is likely to obscure the remarkable fact that IQ, a measure of a psychological construct, is the best predictor of many aspects of the social performance of individuals many years later, with an effect stronger than any other of the traditional "background" variables including SES, certainly one of the sacred cows of sociological research.
While the major empirical findings of the authors are unlikely to be overturned, the book is not without weaknesses. As pointed out by Heckman (1995), the 875 pages of the book do not constitute a single coherent thread of argumentation, as some issues that are discussed at length by the authors, such as the question of the heritability of intelligence, are arguably separable from the rest of the issues; and the policy vision, particularly the rather romantic account of life in strong local communities, does not follow in any direct fashion from the evidence of the previous chapters. Occasional interpretations of current theories of behavioral genetics are incorrect, for example the discussion of the consequences of reducing the variability of the environment (p. 106), and other passages in which the authors exaggerate the intergenerational rigidity of cognitive classes based on genetic transmission of intelligence. Curiously, the same tendency to exaggerate the rigidity of genetic mechanisms - by downplaying the importance of regression to the mean - had already been criticized in the earlier work of Herrnstein (1973) by Eysenck (1973), who showed that genetic transmission entails much greater intergenerational mobility among IQ categories than Herrnstein had acknowledged.
A major contribution of The Bell Curve to sociological research is its polished presentation of the findings of literatures that have been by and large hidden from social scientists, including the behavioral genetics literature on the strong heritability of IQ, the conclusions of psychometric research on the absence of race (but not linguistic) bias in major IQ tests (Jensen 1980), the studies in the field of human resources of the predictive validity of IQ with respect to job performance, and the program evaluation literature revealing the failure of many intervention programs aimed at raising the cognitive ability of children. This skillful review, together with the wide diffusion of the book and the public attention that it has generated may help unravel the veil of collective denial that has shrouded knowledge of these issues in much of our field. With respect to the genetic basis of individual differences in cognitive ability, the book must also be viewed as part of the larger movement of ideas toward the rehabilitation of biological explanations and evolutionary approaches in the social sciences that has been discussed by Degler (1991). This emerging movement is associated with the discovery of a biological basis for many kinds of behaviors (including gender roles, Udry 1994), as well as the ascent of sociobiology (Nielsen 1994).
Sociology has a natural place to occupy within the scientific division of labor that would tackle the issues raised by Herrnstein and Murray, which is to focus on the aggregate consequences of the central role of IQ for individual outcomes demonstrated by them. Their analysis of processes of cognitive partitioning in modern industrial society in part 1 of the book is evocative but lacks formal precision, as it is pieced together from partial data and case studies with much extrapolation. There is a clear need to better document the existence of the mechanisms of social selection on the basis of cognitive ability postulated by the authors and integrate them with theories of social stratification. A special study under such a heading would include the possible role of cognitive partitioning on the upswing in income inequality that has taken place over the last 25 years. The possible dysgenic consequences of the negative correlation between fertility and IQ are another natural topic of study that is already being investigated by sociologists (Coleman 1993; Preston & Campbell 1993). Finally, in much sociological research on educational, occupational, economic, and behavioral outcomes of individuals, it will be harder than ever for researchers to ignore the importance of IQ as a potential explanatory factor with lame excuses about intelligence being a "psychological variable" and therefore off limits for sociologists.