The Bell Curve Case for Heredity

PHILOSOPHY OF THE SOCIAL SCIENCES / September 1999Hocutt, Levin / BELL CURVE CASE FOR HEREDITY

The Bell Curve Case for Heredity

MAX HOCUTTUniversity of AlabamaMICHAEL LEVINCity College of New York

The hereditarian theory of race differences in IQ was briefly revived with theappearance of The Bell Curve but then quickly dismissed. The authors attempt adefense of it here, with an eye to conceptual and logical issues of special interestto philosophers, such as alleged infirmities in the heritability concept. At thesame time, some relevant post–Bell Curve empirical data are introduced.

Intelligence potential is distributed among Negro infants in the sameproportion and pattern as among Icelanders or Chinese, or any othergroup. There is absolutely no question of any genetic differential.

U.S. Department of Agriculture, 1965

INTRODUCTION

Some scientific disputes are momentous enough to raise philo-sophical questions about the measurability of key concepts, the causalinterpretation of data, and the relation of facts to social policy. Onesuch dispute, we believe, concerns the 15-point gap between the aver-age IQ of blacks and that of whites.

The existence of this large—one standard deviation—and impor-tant difference can no longer be questioned,1 and it calls for an expla-nation. One hypothesis holds that the difference2 is due entirely topast and present disadvantages imposed on blacks by whites. In Lyn-don Johnson’s vivid metaphor, blacks trail because they have beenmade to run the race while hobbled. This social-environmental (or,following E. O. Wilson [1998], “nurturist”) hypothesis has been con-ventional wisdom for nearly half a century, and those who have chal-lenged it have been made objects of public derision. The assertion

389

Arthur Jensen and Margarita Garcia made helpful comments about earlier drafts.

Philosophy of the Social Sciences, Vol. 29 No. 3, September 1999 389-415© 1999 Sage Publications, Inc.

at UNIV OF CALIFORNIA SANTA CRUZ on October 1, 2015pos.sagepub.comDownloaded from

http://pos.sagepub.com/

quoted at the head of this article was made without supporting evi-dence, indicating that proponents of the social-environment theoryregard it as beyond dispute, even self-evident. If they are right, thegap in IQ can be completely eradicated by a determined and massiveeffort to equalize opportunities.

A second hypothesis, favored by, for instance, Thomas Sowell(1995), holds that lower black IQ is the result not of impediments cur-rently facing blacks but of habits that have taken root in black commu-nities, perhaps not only as the long-term residue of slavery and JimCrow but also perhaps as the result of ancient African traditions. Onthis “cultural” theory, black poverty, crime, illegitimacy, anddependence on welfare are not imposed on blacks by whites but arefreely chosen by blacks themselves. These counterproductive habitsare thus not the consequence but the cause of lower black IQ: the chil-dren of lower IQ blacks are being raised and educated in ways thatensure they too will develop low IQ. On this hypothesis, the gap in IQcan be closed, but only if blacks adopt new forms of behavior.

The third hypothesis, cautiously reintroduced a few years ago byHerrnstein and Murray (1994) in The Bell Curve (henceforth BC) andimmediately dismissed by hostile critics as not merely baseless butmorally and politically unworthy, is that the difference in IQ is partly,though not wholly, hereditary. We will also refer to this theory as BC.3

(It could also be called the hereditarian theory; its critics use less neu-tral names.) Itself taken to be beyond dispute a century ago, the BChypothesis was gradually abandoned under the onslaught of chargesthat belief in hereditary differences is a myth created by a ruling classto justify its power and privileges. If BC is correct, efforts at equalizingopportunity might reasonably be expected to diminish the gap in IQto some extent but not close it. On this hypothesis, moreover, policiesmeant to compensate blacks for what is thought to be harm done themby whites are misconceived since black competitive failure is largelydue to genetic factors for which whites are not responsible.

We wish to revive the brief-lived and unsatisfactory debate overthe merits of these hypotheses,4 especially the third. It is our view thatthere is more to be said for it than is presently allowed. Since, as sug-gested, the debate over BC is to a considerable extent conceptual andmethodological, we believe the participation of philosophers is par-ticularly appropriate.

We claim not that BC is demonstrably correct but that, correctlyconstrued, it is more plausible and its alternatives less so than criticshave acknowledged. Although interest in the issue has been

390 PHILOSOPHY OF THE SOCIAL SCIENCES / September 1999



heightened by its relation to public policy, we emphasize the scien-tific questions. Policy must be posterior to well-founded scientificbeliefs. The facts come first.

WHAT DOES BC MAINTAIN?

It would be well to make sure at the outset that the question is clear.Nobody, certainly not Herrnstein and Murray (1994), maintains thatall 15 points of the racial gap in IQ is hereditary. By contrast, advo-cates of the social-environment hypothesis do appear to believe thatthe entire gap is environmental. They certainly argue as if they believethis, vehemently opposing all suggestions that any part might behereditary.5 The issue, then, is not between those who say that all isgenes and those who say none is. Nor, in strictness, is it over howmuch of the gap is hereditary and how much environmental.Although they clearly think it may be substantial, Herrnstein andMurray venture no estimate of the portion of the gap that is genetic.They do offer 60% as a middling guess at the heritability of IQ, but thisis a within-race, not a between-race, estimate. The proposition theyendorse is “some of the gap is hereditary,” the contradictory of whichis not that some of the gap is environmental but that all of it is. Thequestion, then, is whether we have reason to believe that any part ofthe gap is hereditary, as Herrnstein and Murray claim, or whether weshould presume that it is all environmental, as proponents of theother two hypotheses believe.

It should also be noted that, despite a widespread impression tothe contrary, The Bell Curve is not primarily about race, nor is its cen-tral thesis that the racial gap in IQ is (in good part) genetic. The bookconcerns the statistical distribution of intelligence and how it relatesto social and economic class; race takes up only 120 pages of 555 pagesof text, only 48 of which concern heredity. In accordance with theselarger aims, BC‘s first 12 chapters document the importance of IQamong whites to poverty, schooling, employment, family relations,welfare dependency, parenting, crime, and citizenship—variableschosen for study because they are known to correlate with IQ—andthe authors want to decide in each case which is cause and which iseffect.

They conclude that, for whites, IQ makes a huge systematic differ-ence. Smart people earn more, achieve higher grades in school, per-form better on the job, have more satisfactory personal relationships,

Hocutt, Levin / BELL CURVE CASE FOR HEREDITY 391



depend less on public assistance, commit fewer crimes, and make agreater contribution to their communities. Smart folk are also moredependable, more likable, and healthier—probably better lookingtoo. Although these are things most people already suspect, they arealso frequently disputed, perhaps in part because being more intelli-gent is still often confused with being better educated. Herrnstein andMurray (1994) do not insist that IQ is all that counts. “This thing weknow as IQ is important but not a synonym for human excellence,”they say. Character, personality, health, and luck matter too, but theseare correlated with IQ. Smart people, as the saying goes, have all theluck.

REGRESSION ANALYSIS

To illustrate BC‘s methodology, consider IQ and poverty. Nobodydoubts the two are connected, but many deny that low IQ is a cause ofpoverty. The prevailing view, in fact, reverses the causal arrow: peo-ple are stupid because they are poor. To test this view, Herrnstein andMurray (1994) plot (see Figure 1) both IQ and parental social-economic status (SES) against the probability of being in poverty,using standard scores to allow comparison. (That is, the scale for bothIQ and parental SES is the distance from each variable’s mean in stan-dard deviation [SD] units; thus, the prospects of someone 1 SD belowthe mean in parental SES can be compared to those of someone 1 SDbelow the mean in IQ. Statisticians know this technique as “z scor-ing.”) Results: an increase in IQ from low to high (with SES kept con-stant) predicts a much greater reduction in the likelihood of povertythan a corresponding increase (with IQ kept constant) in parentalSES. A similar graph compares schooling with IQ to similar effect:low IQ is far more likely than minimal schooling to result in poverty.Herrnstein and Murray acknowledge the importance of other fac-tors—for example, the presence of a responsible father in thehome—but leave little doubt that IQ is the variable of greatestsignificance.

Note here the implicit use of two methodological rules to disam-biguate correlational data: (1) causes (at least above the subatomiclevel) precede effects, and (2) variation in the cause must be followedby variation in the effect. Given (1), the nurturist must hold that it ischildhood environment—measured by parental SES—that causesboth adult IQ and adult poverty, with the IQ/poverty correlation an




induced artifact. Hence, by (2), he must expect the probability of pov-erty to vary widely in those cases across which parental SES varieswidely but IQ turns out about average; conversely, where parentalSES is average but IQ varies widely, the chances of later povertyshould vary very little. That the reverse pattern is observed under-mines the nurturist interpretation of the data. This theme—correlationversus causation—will recur.

The principal objections to this phase of the BC argument havebeen, as implied, conceptual rather than empirical. Thus, two econo-mists, Goldberger and Manski (1995, 769), complain that they can“find no substantively meaningful way” to interpret normalizedcomparisons. In place of causal analysis by regression of dichoto-mous life outcome variables, they propose that the effect on a depen-dent variable of spending some fixed resource to change socioeco-nomic status be compared with the effect of spending the sameresource to change IQ.

This criticism proves too much and too little. As noted parentheti-cally, regression on normalizations to compare the effects of two vari-ables, thus obviating the objection that you cannot compare appleswith oranges, is commonplace in statistics. It is the financial


Figure 1: IQ versus Parental SES in Determining PovertySOURCE: Herrnstein and Murray (1994, 134).



experiment described that makes little scientific sense: one might aswell seek to gauge the nutritional value of various foods by measur-ing the vitamins in a dollar’s worth of each. The econometric proposalalso assumes an ability to boost IQ that we do not have.6 To follow thatdesign amounts to abandoning tried-and-true methods for an experi-ment we have no idea how to conduct.

Goldberger and Manski (1995) stress7 that in a normalized regres-sion, the slope of the regression line depends on the regressor’s SD;SES would predict poverty as well as IQ were its SD larger. This is so,but it in no way makes the greater predictiveness of IQ a scaling arti-fact: the actual slope difference is just another way of saying that IQcovaries with poverty more than does SES. It is also true, as Goldbergerand Manski state, that the slopes will equalize if IQ and SES stratify asBC predicts they will. But again, far from being an embarrassment toBC, this just restates BC‘s conclusion that as IQ becomes more impor-tant, it displaces other determinants of status.

OBJECTIONS TO IQ TESTS

BC‘s measure of IQ is the Armed Forces Qualification Test compo-nent of the National Longitudinal Survey of Youth, which suppliesthe data for its analyses. An inevitable complaint is that such tests donot measure the highly varied manifestations of intelligence.8 Herrn-stein and Murray’s (1994) use of “cognitive ability” in place of “intelli-gence” does not avoid verbal issues, since anyone who denies thatIQ measures intelligence will also deny that it measures cognitiveability. The proper answer to these doubts (also given in BC) is theample evidence of the validity of standardized IQ tests—their well-documented ability to predict success in a variety of academic, eco-nomic, and social activities calling on what plain folk call intelligence.9

People who do well on IQ tests also do well in school and go far in theprofessions. Any well-defined claim of the form “test T does not regis-ter the presence of trait P” must cite a task intuitively requiring P thatdoes not correlate with T. We know of no criticism of IQ tests that sat-isfies this constraint.

It will be replied that evidence concerning whites does not showthat IQ tests also measure intelligence for blacks, but between-racevalidity is also well confirmed, as BC also points out.10 High-IQ blacksalso make better students, better employees, better parents, and bet-ter citizens; they are also convicted of fewer crimes, get more years of




schooling, and are more successful financially and socially than theirlower IQ counterparts. An IQ of 120 means the same thing no matterthe race of the person who achieves it.

From a logical point of view, this fact is sufficient to rebut the oft-heard claim that IQ tests are “culturally biased.” To the extent (whichwe suspect is considerable) that this claim is based on the simple factthat blacks do not score as high on IQ tests, it patently begs the ques-tion. It assumes precisely what is at issue—namely, whether blacksare on average as intelligent as whites. IQ tests measure the same traitfor blacks and whites so long as test results continue to correlatehighly with other indices of ability and success when blacks andwhites are included in the same population. And, to repeat, not onecareful study has shown that a standardized IQ test given to bothraces either overpredicts white success or predicts success in a differ-ent pattern.11

What of those items on some IQ tests—for example, questionsallegedly using words more familiar to whites—that appear to penal-ize black testees? Surprising as it may be from the nurturist point ofview, it is not these items that depress black scores. Blacks do evenworse on “culture-neutral” items that measure pure acuity—what thefollowers of Spearman call g. For instance, blacks fall much more thanone SD below whites on the backwards digit span test, which mea-sures the capacity to remember and repeat strings of characters back-wards. Another telling example involves tests of reaction time—theinterval between the presentation of a stimulus and the onset of aresponse to it. (Time of reaction is presumably related to the speedwith which a situation is perceived, a good index of intelligence.)Jensen (1989) has shown that black reaction time is slower, althoughblack motion time is faster. The intraindividual variation in reactiontime is also higher for blacks than whites, and variance in reactiontime correlates quite strongly with IQ within both races. None ofthese data is easy to explain on the hypothesis that the IQ gap is envi-ronmental or cultural.

It may also be objected that while IQ tests are “fair” in the psycho-metric sense, race disparities in performance on them reflect bias inthe larger society: discrimination, lack of stimulating early environ-ments, and so on. But that nurturist hypothesis, whatever its merits, isirrelevant to the issue of test fairness. Since the trait measured by IQtests is intelligence, whatever causes the race difference in this traitcauses a difference in intelligence. That is what those tests faithfullyreveal. To the extent that they do, they are no more biased than an




X-ray of a torture victim that accurately reveals his wrongfully bro-ken bones.

Is it circular to validate a test by its ability to predict education andincome and then cite variation in the trait measured by the test as a(proximate12) cause of variation in educational and vocational suc-cess? No. Scientists often explain phenomena by posits the sole evi-dence for which, at the time, are the phenomena themselves. Whyisn’t Uranus where it is supposed to be? A new planet. How do weknow a new planet is there? Uranus’s deviant orbit. No circle exists solong as the explanans yields testable consequences beyond what it isenlisted to explain, and this is the trait measured by IQ tests. It pre-dicts all manner of intuitively intellectual abilities not involved in theinitial validation of IQ. To put the general point in terms congenial tophilosophers, when explanandum e is introduced to account for phe-nomenon P, “the cause of P” is an accidental designator that fixes thereference of e. “The unknown influence on Uranus” fixes the refer-ence of “Neptune” without defining it, so “Neptune is influencingUranus” remains nontrivial.

HERITABILITY OF INTELLIGENCE

The discussion up to now has concerned phenotypic intelligence,as distinguished from intelligence insofar as it is innate. Many people,if few psychometricians, conflate the two, assuming that if Ann issmarter than Brad, she must have been born so. However (as Herrnsteinand Murray 1994 fully recognize), differences in IQ or any other traitmay be real without being inherited. The claim that a difference isgenetically based needs separate argument.

BC summarizes what is known of the genetic control of IQ, tenta-tively estimating it at 60%. We must now attend to the heritability ofIQ and the role of this difficult concept in BC‘s overall argument, for itis on this concept that virtually all serious criticisms of BC focus.13

Heritability measures how much of the variation in a trait in apopulation is due to variations in genetic endowment, the rest beingassumed due to variation in environment.14 If all the variation is dueto genes, heritability is 1; if none is, heritability is 0; numbers between0 and 1 indicate the proportion of variation attributable to genes.Heritability may be estimated “directly” by comparing blood rela-tives (for example, monozygotic twins) sharing some known propor-tion of genes but reared apart or “indirectly” by comparing related




or—the most interesting case—unrelated persons reared together.Direct studies have yielded estimates of heritability as high as .8 forIQ, while indirect studies have yielded estimates as low as .4.15 Toapply these numbers to individual differences, we must take theirsquare roots.16 A low-end heritability of .4 for IQ suggests that, onaverage, 63% of the difference in IQ between any two individuals isattributable to heredity—9.5 of every 15 points that separate them canbe attributed to genetic variation—while a heritability of .8 raises thatfigure to 89%. The estimate ventured in BC of .6 implies that, on aver-age, 77% of the difference in IQ scores between pairs of whites in theUnited States can be attributed to differences in genes.17 Thus, evenconservative estimates of heritability leave us attributing most varia-tion in IQ to genetic differences.

A more formal definition of heritability is needed to engage fullywith the issues BC raises. In any population, the values of a quantifi-able phenotype P will vary. The variance of the distribution of P‘s val-ues is P‘s phenotypic variance in that population. Consider next that agenotype of P may express itself differently in different environ-ments. The yield of a variety of wheat raised in Iowa will exceed theyield of the self-same variety cultivated in the Mojave, according tovariations in temperature, moisture, and fertility of soil.18 This varia-tion is the genotype’s reaction range.19 Each genotype for P in a popu-lation will have a mean expressed value over its reaction range; thevariance of the distribution of these mean genotypic values is P‘sgenetic variance in the population. Intuitively, the smaller the geneticvariance in a population—the more the reaction ranges of the geno-types for P resemble each other—the more P‘s phenotypic variation isdue to variation in the environments to which the various genotypesare exposed. If, on the other hand, every gene for P reacts dissimilarlyto the same environments, yielding a large genetic variance, the moreindividual differences in P will be due to genetic differences. Hence,the ratio of P‘s genetic to phenotypic variance is defined to be the heri-tability of P, or h2 (P). At one limit—the same environment for every-one—all phenotypic variation is due to genetic variation, and h2 is 1; atthe other limit—the same heredity for everyone—h2 is 0.

It is important to emphasize that even an h2 of 1 does not mean thatan individual must develop the same IQ no matter what, since thereaction range of his genotype may still be quite wide, so that his IQdepends on the environment he finds himself in. The situation, as inFigure 2, is the following.




If genotype g1 for IQ develops in environment e1 while genotype g2

develops in e2, the difference ∆ between their phenotypic expressionsis due mostly to their genetic difference, implying a large h2; were g2

also in e1, the gap would still be ∆g. (So one can think of ∆e = ∆ – ∆g asthe difference made by the environment.) Nonetheless, the reactionranges R1 and R2 for both genotypes are still wide, and g1 would yielda lower phenotypic value in e2—a phenomenon closely related to“gene/environment interaction.”20 By contrast, to say that just one ora few phenotypic expressions is possible for a genotype is to say thatthe genotype’s reaction range is extremely narrow. Eye color, forinstance, is predictable from genotype alone; it does not vary with theenvironment. In such cases, behavioral geneticists speak of ontoge-netic fixity, and what many people seem to have in mind when callinga trait genetically determined is that it is ontogenetically fixed.21

HERITABILITY: WELL OR POORLY DEFINED?

Heritability is thus relative to given groups and ensembles of envi-ronments.22 The hard part is using data on heritability for one groupto compare groups that may live in markedly different environmentsor to predict how a trait will be expressed for the given group in new


Figure 2: Heritability versus Fixity



environments. Stated abstractly, the problem23 is this. While a geno-type’s reaction range should ideally be defined over all possible envi-ronments, it can be known only over environments to which the geno-type has actually been exposed. In particular, a genotype’s responseto observed environments entails nothing about its response to unob-served ones. As one critic of BC has observed with regard to IQ in par-ticular, “no one know[s] how any human genotype may react to envi-ronments that involve new intellectual machinery” (Block 1995, 124).Genotypes for IQ (e.g., those characteristic of blacks and whites) thatdiverge in some environments (e.g., the United States) may convergeor reverse24 in other, as yet unrealized ones, such as serious remedialefforts might create. High heritability across known environmentsdoes not close this possibility since heritability is determined bygenetic variance and not vice versa. Genotypic variance will decreasein new equalizing environments, and should it decrease faster thanphenotypic variance, the mathematical result will simply be a drop inheritability for IQ. Since, by hypothesis, the typical white and blackgenotypes for IQ express themselves identically in these new envi-ronments, the proportion of between-race variance explained bygenetic difference (i.e., the between-race heritability) would also fall.

Stated more concretely, the problem is that since there are nomonozygotic twins, one of whom is wholly white and one whollyblack, heritability is usually estimated within races—whites beingcompared with other whites, blacks with other blacks—in the same ordifferent environments. These estimates, however high, say nothingabout the role of genes in variations between members of differentraces. To conclude that between-race variance is due to heritablevariation, blacks and whites would have to be raised in demonstrablyidentical environments. Even then the inference would be incom-plete, for while whites and blacks might react differently to the sameknown environments, there might be an unknown environment inwhich black IQ would equal or even exceed white IQ. This is whymany authors criticize BC‘s invocation of heritability, and one goes sofar as to advise that “we should ignore heritability . . . and simply tryout improved environments.”25

This criticism is misplaced. Herrnstein and Murray (1994) them-selves do not claim—in fact, they deny—that within-race heritabilityof IQ provides any basis for concluding that the IQ gap between racesis hereditary. It is always possible that the between-race mean differ-ence is due entirely to environmental factors. On the other hand, thesheer possibility that the BC hypothesis is false does not constitute




any sort of evidence that it actually is false. By the same token, thesheer possibility that environmental intervention might equalizeblack and white IQ, the logical consistency of this supposition withwhat we now know, does nothing whatever to establish that there arepresently unknown environments in which black and white IQswould emerge the same.26

The question of substance, which we now address, is the evidencefor BC‘s conclusion that genes are implicated in the race difference.

BETWEEN-RACE COMPARISONS

This question has already been partly answered. As Herrnsteinand Murray (1994; see, e.g., p. 302) observe, no purely environmentalor cultural difference hypothesis yet envisioned explains the racialpatterning of differences in IQ scores—the fact that blacks scoreworse on items heavily laden with g while doing relatively better onother items. Environmentalists need to explain why the black envi-ronment lengthens black reaction time and causes blacks to do worseon the backwards digit span test, as well as why black performance onculture-loaded tests (e.g., of vocabulary) exceed black performanceon more culture-neutral ones (e.g., of spatial visualization).27

The prospects for a purely environmental explanation of the racialIQ gap have been further weakened by evidence appearing since BC28

that the gap is fully in place by age 3, when the dominant socializingagent is still the mother. There is the further striking fact that blackinfants are more advanced in motor and, apparently, mental develop-ment for the first 15 months of life.29 It is extremely difficult to imagineenvironmental influences that would accelerate black developmentahead of white for a year or so after birth and then retard it.30

Of course, these data all involve comparisons of persons reared inradically different environments—the basic problem—but (as BCpoints out, pp. 309-10) not all between-race studies do so. In the well-known transracial adoption by Sandra Scarr and Richard Weinberg(Weinberg, Scarr, and Waldman 1992), the IQ of adopting white par-ents was found to correlate much more closely with that of their natu-ral children than with that of their black adoptees. Indeed, by age 17,the mean IQ of the cohort of black adoptees was more than 1 SD belowthe mean IQ of the birth children of the adoptive family, 1 SD belowthe mean IQ of white children adopted by these families, and .7 SDbelow the white mean. These results have no obvious explanation on




the black culture theory, and to explain them on the social environ-ment hypothesis, we have to suppose that adopting white parentsmake a point of treating their adopted black children in a radicallyinferior way—an assumption that does not comport with the usualmotives for adoption. It is true that the cited transracial study does notquite satisfy the assumptions of the classical model for independenceof variables, but they come as close as adoption policy allows, whichis close indeed. Rejecting it on grounds of imperfection is the scientificequivalent of burying one’s head in the sand.

In any event, advances in statistics since the appearance ofBC—unnoticed, so far as we know, by all of its critics—do permitinferences from within-group to between-group heritabilities underappropriate conditions (Rowe and Cleveland 1996). The basic tech-nique is to compare the patterns of correlations between test perfor-mance and genetic relatedness within each group; if an environ-mental factor is depressing the performance of one group but not theother, it should manifest itself somewhere in a difference between thewithin-group patterns. Applied to black and white academic perfor-mance, this technique indicates that the causes of between-group dif-ferences resemble the causes of within-group differences (i.e., are sig-nificantly genetic).

SUPPORTING CONSIDERATIONS

Supporting the evidence just surveyed are several other less formalconsiderations. One is the failure of Asians and Jews, who have alsofrequently grown up in disadvantaged environments, to show evi-dence of diminished IQ as a result; in fact, their average IQ is slightlyhigher than that of more privileged whites. Another of some rele-vance is the marked improvement over the past 30 years in the socialand economic environment of blacks. If that environment is not yetequal to the white environment, it is surely less unequal thanbefore—at least in the ways that social environmentalists, if not blackculture theorists, think matter to IQ. Blacks now get better nutritionand better housing, go to the same public schools, see the same mov-ies and television shows, shop in the same stores, and vote in the sameelections as whites. Yet the IQ gap remains constant, and (see n. 21)genetic factors become more salient as environments converge. HeadStart and other early intervention programs are especially designedto provide enriched environments, yet these have failed. Early




reports that they had raised IQ for blacks have been followed bynotices that all gains are lost after a few years.31

It might be argued that since the environment of whites has alsoimproved, there is no reason to expect closure in the IQ gap, but thisreply will have to contend with a striking statistic mentioned byDinesh D’Souza (1998): data from the College Board show that whitesand Asian Americans who come from families earning less than$15,000 a year score higher on both the verbal and math sections of theSAT than African Americans from families earning more than $60,000a year.

The SAT is so highly g-loaded it might be considered a virtual IQtest, so this datum would be predicted by BC. It is, however, not obvi-ously consistent with either the social-environment hypothesis or theblack culture theory.32

The considerations of this and the preceding section are, of course,not conclusive. As we acknowledged, the logical possibility alwaysremains of some unknown environment in which whites and blackswould do equally well on IQ tests or in which blacks would havehigher average IQ than whites. However, given what we know—thatwhite IQ exceeds black IQ in all environments so far examined (andthat the principal reason seems to be genetic differences in reaction tothese environments)—to deny that we may expect this to hold in asyet unspecified environments is not to offer an argument against BC;it is to challenge induction. As Cleanthes reminds Philo in Hume’sDialogues, one cannot withhold assent to a hypothesis to which onehas “nothing particular to object” simply because it might be false.

To be sure, attempts to equalize environments do not exhaust allthe practical possibilities. We might, for instance, also try to achieveequality by enriching the environment for blacks while degrading it forwhites.33 Think of a variation on Lewontin’s horticultural example—two strains of wheat, A and B, such that (1) A grows to 8 feet while Bgrows to 7 when both are planted in Iowa, and (2) A grows to 7 feetand B to 6 when both are planted in the Mojave. We cannot equalizetheir growth in any one environment, but we can equalize it by plant-ing A in the Mojave, B in Iowa. Likewise, we might try to equalizeblack and white IQ by putting blacks in environments that raise theirIQ while putting whites in environments that reduce theirs.

The moral objections to this proposal should be obvious. It contra-dicts the demands for equal treatment of the races made from allpoints on the political spectrum34 and would require totalitarianintrusion into the lives of both whites and blacks. And, empirically, to




repeat one of BC‘s main lessons, no one has any good idea how toboost anyone’s IQ; the proposition that we can is wholly speculative.There is, to take just one example, no evidentiary support for theproposition that formal education raises IQ. Instead, the evidenceprovided by Herrnstein and Murray (1994) suggest that IQ leads toadditional schooling, not the reverse. We do have some idea how todepress IQ or prevent its emergence: cause brain damage or interferewith cerebral development. If we started early enough to starve thebrains of white infants or administer blows to the head, we couldreduce their IQ as much as desired. But these actions, apart from theirmoral repugnance, would be premised on the well-founded beliefthat IQ is related to brain development, which is clearly and indis-putably mediated by genes, a fact that returns us once again to BC.

THE IRRELEVANCE OF THE FLYNN EFFECT

There is an idea that all of this cumulative evidence is refutedsomehow by the Flynn effect, the apparent 15-point rise in average IQfor whites from 1930 to the present (Flynn 1984, 1987). Since thisperiod of time is too short for a diachronic rise to have an obviousgenetic explanation, a natural presumption is that it is environmental.This encourages the inference that the synchronic race difference isalso environmental, but this inference is too hasty, for several reasons.

First, the Flynn data are highly puzzling. The present generationdoes not appear to be markedly more intelligent than its parents andgrandparents. As Flynn (1987, p. 182) observes, there do not appear tobe many more people these days who “find school easy and can suc-ceed in virtually any occupation, [whose] achievements are so clearthat they fill the pages of American Men of Science, [who] resemblethe famous geniuses.” A possibility suggested by these facts is that IQtests, which appear to measure intelligence within generations, doesnot measure it across generations. Deciding whether this is so willrequire further study. The last word on Flynn’s data remains to bespoken.

Second, even taken at face value, the Flynn effect fails to warrant anenvironmentalist explanation of the racial gap in IQ and is in fact con-sistent with a genetic explanation of the continuing difference. Thegeographic scope and uniform rate of the general rise in IQ suggestthat whatever environmental factor is producing it reaches almosteveryone, black and white alike. Yet the racial gap in IQ is as large as




ever before, leaving the Flynn effect irrelevant to whether this gap ishereditary. At best it suggests that unknown improvements in theenvironment can raise IQ, but that is not a proposition BC denies.

Social-environmentalists may reply that the constant IQ differenceis an effect of constant environmental inequalities; everybody’s envi-ronment has improved, but average black IQ continues to be lowerbecause the environmental gap between blacks and whites has stayedconstant. Unfortunately, this reply does not specify the particularimprovements in the environment that both demonstrably alter IQand are enjoyed differentially by whites—so it too is an argumentfrom ignorance or an appeal to sheer possibility. A lockstep increasein black and white IQ over time shows at most that blacks and whitesrespond identically to the Flynn factor, whatever it may turn out be.That IQ may be altered by environment does not imply that IQ differ-ences can be (recall Figure 1). Unless we conflate the malleability of atrait with the malleability of a trait difference, improvements in theFlynn factor for blacks are irrelevant to eradicating the gap in IQbetween blacks and whites.

GENE/ENVIRONMENT CORRELATION

There is a further difficulty with the heritability concept that seem-ingly undermines the significance of even a high between-group heri-tability for IQ.

Genes do not distribute randomly through environments. A genedisposing its carrier to exercise, for instance, may also dispose its car-rier to seek opportunities for exercise, perhaps by moving to a warmclimate or building a tennis court, so it will occur more frequentlywhere there are opportunities for exercise. Genes thus help createenvironments, blurring the line between genetic and environmentaleffects.

A gene may also influence its environment less directly by elicitingreactions from others via its phenotypic expression. The standardexample is a “curiosity gene” that makes a child ask questions,prompting his parents to buy him books, thereby making him morecurious. Such feedback can just as well be negative, however. Were asociety to starve its blond children while giving ample food and ath-letic training to its brunettes, the brunettes would grow far larger andsturdier. Yet, since hair color is highly heritable, between–hair colorphysique variation would be highly heritable as well. More to the




present point, perhaps genetically controlled dark skin triggers white(mis)treatment, which inhibits mental development, creating a spuri-ously high between-race heritability for IQ.35 In this way, one canexplain the failure of Head Start (white supervisors discourage blackchildren) and transracial adoption (black adoptees are treated differ-ently by their adoptive parents and the wider society despite rearingin nominally the same environments as their adoptive siblings).These possibilities show that a high between-race heritability for IQdoes not necessarily indicate that the race difference in IQ is “genetic”in any reasonable sense—any sense that absolves whites from blamefor the black shortfall or shows that environmental intervention isdoomed.

The point underlying this objection is that heritability, both indi-vidual and between-group, is a correlational statistic. It tells us ineffect that certain genes are associated with certain phenotypes butnot why the association holds, hence implying nothing about thedirection of the causal arrow. Nonetheless, here, as previously, com-peting causal interpretations of the correlational data give rise to dif-ferential predictions, which can be and have been tested.

To begin, the supposition that mistreatment is what reduces the IQof black children predicts that the racial gap in IQ will be fairly con-stant over a wide range of mental functions. One would not expectrandom acts of bigotry, for instance, to do more damage to the mental“module” controlling verbal fluency than to that controlling numeri-cal reasoning. In particular, if various IQ subtests exhibit differentialheritabilities among whites, one would expect the white-black differ-ence on these tests to be constant, for there is no obvious mechanismby which bigoted responses can discriminate brain functions on thebasis of heritability. And in fact (although this is not mentioned inBC), it has been found that black-white score differences on IQ sub-tests correlate positively with score heritability within races: the moreheritable an IQ subtest is among whites and among blacks, the widerthe black-white difference, whether within-group subtest heritabilityis determined by standard sibling comparisons or by response toinbreeding depression (Rushton 1989). For our blond-brunette anal-ogy to be sustained, it would have to be supposed that whites some-how calibrate their negative responses to black performance pheno-types by performance heritability, with white discouragement ofblack performances becoming more effective as the performancesbecome less susceptible to environmental influence. The improbabil-ity of such preestablished disharmony leaves it reasonable to suppose




that the black-white IQ difference is functionally related to the genesthat distinguish the two groups.

Second, to use terminology favored by Reichenbach (1938) andmore recently Salmon (1984), we can try to “screen off” a correlationbetween variables B and C by observing the relation between them inthe absence of a purported underlying cause A; if the correlation isspurious, it will then vanish. Thus, if low black IQ is an effect of nega-tive white reactions to genetically black pigmentation, mean black IQsshould approximate whites in mostly or entirely black societies, suchas those of Africa. As BC points out, however (pp. 288-89), the meanIQ of blacks is reported to be if anything lower than that of Americanblacks. Critics (e.g., Kamin 1995) have dismissed these data, but morehave come to light since BC‘s appearance. Zindi (1994), a black Zim-babwean, administered the Revised Wechsler for Children and theRaven’s Progressive Matrices to 202 Zimbabwean secondary schoolchildren and 204 white London children matched for age and, as far aspossible, background. The two groups differed overall by about 2 SD,roughly the gap reported in BC. Also, 15-year-old Ethiopian Falashasin Israel have been found to score 2 SD below Israelis of the same age(Kaniel and Fisherman 1991). The correlation between race and IQsurvives removal of its hypothesized underlying cause.

Finally, recent research (again not mentioned in BC) has begun toclarify the mechanisms that produce intelligence. IQ has been foundto correlate inversely with the rate of cortical glucose metabolism(Haier et al. 1988). A critic might of course reply that any between-race differences in this factor would merely show that racism reducesthe brain’s metabolic efficiency. However, several studies36 havefound within-race correlations in the .4 range between brain size andIQ; since adult brain size is predictable from brain size in infancy andset by the closing of the cranial sutures early in life, it is hard to seehow social practices might affect it. Head size, an approximator ofbrain size, correlates significantly with IQ both within and betweenraces (Jensen and Johnson 1994), implicating brain size in individualdifferences across races. Furthermore, it is now widely conceded that,on average, Caucasoid crania are larger than Negroid crania (Beals,Smith, and Dodd 1984).37 Finally, specific chromosomal sites associ-ated with high intelligence have recently been announced (Plomin et al.1995), although the proportion of variance in overall IQ they explainremains small.

Methodologically speaking, while the environmental effects ofgenes should be discounted when gene/environment correlation




leads to an overestimate of the role of genes in phenotypic differences(as in the blond-brunette case), such discounting can also lead to whatis, intuitively, an underestimate. Applying this caution to the racialcase, many of the environmental factors said to retard black intellec-tual development may themselves be further effects of black geno-types. If, for instance, black babies are handicapped by inadequatenutrition, one might argue that the failure of black parents to provideappropriate food is itself an effect of lower mean black IQ, in turnunder genetic control. Similarly, the unsalubriousness of manyghetto neighborhoods may better be viewed as a consequence of thegenetically controlled behavior of their inhabitants than as an exoge-nous imposition.

To put the issue in quantitative terms, suppose the correlation forhealth (by some measure) for identical twins raised apart was foundto be .4. Conventional behavioral genetics would assign health a heri-tability of .4. But suppose that the co-twins in the sample tended to eatsimilar diets and that the concordance of health of the co-twins fell to.3 when diet was controlled for. This might lead us to reduce the esti-mated heritability of health to .3 and chalk up 10% of the total vari-ance to the environmental factor of diet. But then suppose we were todetermine by some independent means that the heritability of diet is.35. Disregarding gene/environment correlation would require us toignore the 35% of the 10% of variance in health explained by diet that,in turn, is explained by genetic variance. In such a case, it would cer-tainly be plausible to add a .35 • .1 correlation term to the adjustedestimate, for a final value of .335.

“Cause” is to some extent context sensitive, and on occasion thepronouncement that genes as opposed to environment (or vice versa)are “the cause” of something does represent a verbal decision ratherthan an empirical claim. Our general point is that, nonetheless, ascrip-tions of causality in behavioral genetics are not hopelessly relative tointerest, purpose, or ideological bent. In particular, a causal interpre-tation of the between-race heritability of IQ is well motivated.

BLACK CULTURE AND HEREDITY

The discussion of the last section should make clear why we havesaid relatively little about the differences between black and whiteculture: the culture in which one develops is an aspect of his environ-ment, so the cultural explanation of individual and group differences




inevitably reduces to the nurturist or BC hypothesis or some combi-nation thereof. Let us see how.

It is safe to say at the outset that culture is itself a function of intelli-gence. This is clear enough when we are talking about individuals.Smart people behave more intelligently; that, in fact, is how we knowthey are smarter. But if this is true of individuals, why not also ofgroups? One would not expect a social group to continue to behaveless intelligently than the mean level of intelligence of its memberspredicts, leaving us with the problem of explaining the cultural dis-parities, mentioned earlier, that were supposed to explain blackbehavior. At this point, it is of course natural to reply that even thoughblacks are not less intelligent, their behavior has been made to look soby slavery and Jim Crow in the United States and colonialism inAfrica. Oppression created environments in which intelligence doesnot thrive because it cannot; the soil and climate won’t let it. But to saythis is to hand over the job of explaining mean black/white differ-ences to the social environment hypothesis. Simply saying that indi-vidual blacks behave less intelligently on average because otherblacks and some whites encourage them to do so—while Jews andAsians, who value intelligence, make a point of encouraging its devel-opment—does not tell us why some groups value intelligence morethan others. How, if blacks are as inherently intelligent as Asians, didthey acquire counterintelligent habits? To answer, “Because theywere punished for acting smart under slavery” is to give up the blackculture hypothesis and return to the social environment one.38

As we have seen, of course, this hypothesis is riddled with difficul-ties. In the present context we may ask why, if African cultures hadnot already been inferior in technology or energy or resourcefulnessto the cultures of their European conquerors, they were so easilydominated. Why would blacks have so quickly internalized a nega-tive self-image? Why has this self-image persisted despite massiveprivate and governmental efforts to enforce equality? An alternativehypothesis, which we think has been shown to be highly plausible, isto construe black culture as an environmental correlate of the geneticpotential that created it. Since this genetic potential has expresseditself as lower (than white) individual phenotypic intelligence inevery environment of which we have any experience, it is natural thatthe cultures created by the interaction of these individual phenotypesshould be less advanced scientifically and technologically. But againthe main point is that the job of explaining cultural variation has been




taken over entirely by the hereditarian hypothesis, and culture hasceased to be a prime mover or even an independent variable.

CONCLUDING REMARKS

Diagnosis of causes is related to questions about what it wouldtake to change effects, so it is unavoidably linked to issues of publicpolicy. We conclude with a few reflections.

We have seen in some detail that the high heritability of IQ, evenwhen distinguished from ontogenetic fixity, provides good reason todoubt that environmental intervention is likely to end the racial dis-parity in IQ. We willingly grant that this conclusion has not beenestablished with certainty but ask in return where the burden of proofnow lies. Given the evidence cited, it would seem to lie with nurturistswho think that the entire gap can be eradicated by equalizing envi-ronments. It is they who must tell us clearly what would be involvedand why they think it would work.

It might be replied that despite the evidence, continued efforts atintervention can be justified as a kind of Pascal’s wager: if the race gapis incompressible for genetic reasons, these efforts will of course fail,but if any of the difference is due to remediable environmental fac-tors, experimenting gives us a chance to find them. One obvious flawin this reasoning is that the envisioned experiments (unlike a Pascalianembrace of theism) carry costs. Remedial programs such as Head Starthave consumed hundreds of billions of dollars in resources that couldhave gone elsewhere. From a purely utilitarian point of view, expectedreturns on such investments must be taken into account. A relatedflaw is that, as mentioned earlier, the rationale for closing the race gapbecomes unclear if the BC hypothesis is in fact correct. Individual andgroup inequalities offend our sense of rightness primarily—somewould say only—when they are believed to have been wrongfullycaused. Given BC, the IQ gap is an outcome of the amoral action ofgenes and, ultimately, natural selection, hence in no obvious wayimproper. Indeed, diverting resources to counteract it threatens towrong those to whom the resources would otherwise have gone.39

The question, as so often in science, is the following: what is themost reasonable hypothesis? We cannot safely assume that none ofthe race difference in biophysical traits, such as hair color or suscepti-bility to high blood pressure, is hereditary. The relevant evidence in




BC and subsequent publications make that assumption about intelli-gence equally hazardous.

NOTES

1. For the data that support this claim, see Herrnstein and Murray (1994, 276ff).2. We speak, of course, of a difference in mean IQ between populations. There is no

intention in this use of ellipsis to suggest that all blacks have lower IQs than all whites.The distribution of black IQ, like that of white, takes the shape of the eponymous bellcurve. Still, the displacement of means is such that most whites have higher IQs thanmost blacks.

3. Its most notable and abused proponent has been Arthur Jensen.4. Our classification follows D’Souza (1998).5. An instance cited by Herrnstein and Murray (1994) is Lewontin, Rose, and

Kamin (1984).6. Programs such as Head Start have encouraged the belief that we do have this

ability, but although such programs raise IQ very slightly at the outset, the gains soonvanish (see below). There is a way to boost IQ in groups—namely, selective breeding.But this prospect is not only politically unacceptable, to say the least, but nurturists—who deny that heredity has anything to do with IQ—cannot consistently endorse it.

7. Goldberger and Manski (1995, 769).8. As Herrnstein and Murray (1994) note, the variety of “intelligences” distin-

guished by Howard Gardner might better be called “talents.”9. It is also noteworthy in this connection that judgments of comparative intelli-

gence by teachers, employers, and others correspond closely to comparative scores onstandard IQ tests.

10. For confirmation, see Neisser et al. (1996).11. There is some evidence that these tests overpredict success for blacks.12. That is, with genetic differences then cited as a factor in trait variance.13. The reader might also consult Sesardic (1993).14. Environmental variance can indeed be defined as the total variation in a trait less

its heritability.15. Devlin, Daniels, and Roeder (1997) argue from a meta-analysis of 212 studies

that 20% of the covariance between monozygotic siblings reared apart, usually attrib-uted in toto to genetic similarity, is due to the identity of the maternal environment.They conclude that the heritability of IQ is .48, unifying direct and indirect estimates.They also urge that its “narrow” heritability—which discounts epistatic interactionsunique to each genotype—is .34. The latter figure bears more closely on intergenera-tional similarity and BC‘s speculation about the emergence of intellectual castes, whichwe do not consider here.

16. The variance of a trait is (SD)2. So a factor explaining n% of its variance explainsn% of individuals’ differences in that trait.

17. We trust that the canard, propagated by Kamin (1974) and Gould (1981, 1995)that Cyril Burt established the nonzero heritability of IQ by faking data, has by nowbeen discredited. The correlations given by Burt for monozygotic twins reared apartdiffer only trivially from those discovered in independent twin studies. For a thoroughairing of the dispute, see Mackintosh (1996).




18. The example is due to Lewontin (1976). Bouchard (1995) questions the accuracyof Lewontin’s botany.

19. Block (1995, 123) attributes to BC the idea that genes “determine the size of themental bucket and then the environment fills the bucket to one level or another,”whereas what genes “truly determine” is the reaction range, “as every populationgeneticist knows.” Apart from this gratuitous implicature of ignorance, Block presentsno evidence that Herrnstein and Murray (1994) adopt the bucket idea or that BC‘s argu-ment depends on it.

20. The tendency of different genotypes to react differently to the same environ-ment, hence for the expressed difference between pairs of genotypes to vary as environ-ment does. Figure 2 also illustrates interaction.

21. Ontogenetic fixity seems to be what Block (1995) has in mind when he says “thekey to part of the fallacy” of The Bell Curve is the distinction between heritability andwhat he calls “genetic determination.” Block nowhere documents the charge thatHerrnstein and Murray (1994) confuse the two or that they ever talk of “genetic deter-mination.” Rather, he says, without supporting evidence, “People who read The BellCurve often suppose that a heritable characteristic is one that is passed down in thegenes” (p. 104). It surely sets up a straw man to criticize writers for a confusion theirincautious readers might make. Having turned BC‘s claim about the heritability of IQinto one about “genetic determination,” Block makes a great show of contrasting thetwo notions—for instance, that genetic determination (p. 104), unlike heritability,depends on the idea of a “normal environment” (p. 104). (Actually, as noted, a trait isontogenetically fixed when it emerges in all environments. Also, behavioral scientiststend to speak of an environment as “normal” for an organism when it resembles one inwhich the organism’s ancestors evolved [see, e.g., Daly and Wilson 1988], rather than,as Block suggests, when it is one that “allows [the organism] to thrive” [p. 105].) Whathe does not do is demonstrate that the BC argument itself depends on conflating thetwo notions. As we will see, the argument needs nothing stronger than heritability. Theevidence Block offers for this conflation is that Murray made it in a TV interview withMichael Kinsley. Murray freely confessed to one of the present authors that he didindeed muddle the distinction on the air but immediately tried to clear it up.

22. Herrnstein and Murray (1994) are perfectly aware of this relativity. They write,“If, one hundred years ago, the variations in exposure to education were greater thanthey are now (as is no doubt the case) and if education is one source of variation in IQ,then, other things [being] equal, the heritability of IQ was lower then than it is now. As ageneral rule, as environments become more uniform, heritability rises” (p. 106).

23. Versions of it may be found in Block (1995), Jencks (1991, 94, 109), Layzer (1976),Lewontin (1976), Block and Dworkin (1976), Hirsch (1970, esp. 93-94), Hirsch andMcGuire (1977, 68-80), and Feldman and Lewontin (1975). Bodmer and Cavalli-Sforza(1970) also link the “interaction” phenomenon to nurturism.

24. See Figures 4 and 5 in Block (1995). Layzer (1976), Hirsch and McGuire (1977),and Block and Dworkin (1976) display similar hypothetical graphs.

25. Block (1995, 124-25). He not only repudiates formulating the genetic issue interms of heritability but also chastises Gould and other critics of BC for appearing toaccept the concept.

26. Block (1995, 115-17) notes that as both genes and environment contribute to anygiven phenotypic outcome, there is no reason to denominate genes the cause. All thatfollows, though, is that given an intervention to change environments so that black andwhite IQs converge, it would be arbitrary to attribute race differences (or similarities) to




genes. However, this point has no bearing on “the cause” of race differences if there isno such environment. Once again, a bare logical possibility has no bearing on whatcauses are actually at work.

27. The ad hoc character of recent “cultural” approaches to achievement discrep-ancy is typified by a study by Harber (1998a, 1998b). When he asked two groups of stu-dents to evaluate identical essays—the first group being told that the essays were byfellow white students, the second group being told they were by fellow black stu-dents—the essays thought to be by blacks were graded more leniently. Harber (1988b,9) interpreted this bias as harming blacks because it “deprive[s] minorities of condi-tions [academic challenges] in which they are most likely to excel.”

28. Peoples, Fagan, and Drotar (1995); Brooks-Gunn, Klebanov, and Duncan (1996).29. See Lynn (1998) and references therein.30. A hereditarian explanation would be that blacks have an overall faster life cycle

than whites. Corroborative evidence is the 5-year advantage in longevity enjoyed bywhites over blacks in the United States, which has persisted despite black access tomodern medical technology.

31. See Spitz (1986, 90, 103ff).32. Diamond (1997) attempts to explain the attainments of all human groups wholly

in terms of geography and ecology (e.g., farming developed in southwest Asia but notsub-Saharan Africa because of the presence in the former but not the latter of numerousspecies of heavy-seeded grasses). However, Diamond neglects the selectional effects offarming itself. He also assumes from the outset that “primitive” peoples are if anythingmore intelligent than modern Westerners, leaving it unclear how his analysis canaccommodate the 2 SD difference between white and native African IQs reportedbelow.

33. Goldberger and Manski (1995, 764) hint at such a policy. After observing that“[BC‘s] thought experiment called for equalizing environments,” they antisepticallyinvite us to suppose instead that “we make U [environment] perfectly negatively corre-lated with Z [genotype] by introducing an extreme compensatory policy.”

34. Even compensatory preferences for blacks are justified as restoring to them whatthey would have had their ancestors been treated fairly, so rest on a norm of equal treat-ment. We note, however, that Ronald Dworkin (1998) construes the maxim of equalityas a demand that people receive equal consideration, which can in practice mean quitedisparate treatment—for instance, he says, preference for blacks over whites. SoDworkin might regard the scheme described as constituting equal treatment.

35. See Block (1995), Jencks (1991, 99, 107), and, more informally, Hacker (1992, 27).36. Willerman, Schultz, and Rutledge (1991); Andreasen et al. (1993); Raz et al.

(1993); Wickett, Vernon, and Lee (1994).37. In a widely cited article, Gould (1978) accused the American craniotomist Sam-

uel Morton of underestimating the size of Negroid crania from an unconscious desireto prove white superiority. Michael (1988) has vindicated Morton’s measurements,showing that such errors as he made understate Caucasoid/non-Caucasoid differ-ences. In his later work, Gould ([1981] 1997, 66) admitted that he had unconsciouslyunderestimated the size of the Caucasoid crania in his reanalysis of Morton’s samplesout of a desire to minimize Caucasian skull size.

38. One might note as an aside that enslavement does not necessarily reduce IQ.Jewish slaves often became teachers and scribes in Greek and Roman households.

39. The greater productivity and law-abidingness of more intelligent people are rea-sons to invest in raising everyone’s intelligence, assuming that were possible. The issue




here, however, is the rationale for closing the race gap, which entails raising the intelli-gence of just one group.

REFERENCES

Andreasen, N., H. Flaum, V. Swayze, D. O’Leary, R. Allinger, G. Cohen, J. Ehrhardt, T.Yuh. (1993). Intelligence and brain structure in normal individuals. AmericanJournal Psychiatry 150:130-34.

Beals, K., C. Smith, and S. Dodd. 1984. Brain size, cranial morphology and timemachines. Current Anthropology 25:301-15.

Block, N. 1995. How heritability misleads about race. Cognition 56:99-128.Block, N., and G. Dworkin. 1976. IQ, heritability and inequality. In The IQ controversy,

edited by N. Block and G. Dworkin, 410-542. New York: Pantheon.Bodmer, W., and L. Cavalli-Sforza. 1970. Intelligence and race. Scientific American

223:19-27.Bouchard, T. 1995. IQ similarity in twins reared apart: Finds and responses to critics. In

Intelligence: Heredity and environment, edited by R. Sternberg and C. Grigorenko.New York: Cambridge University Press.

Brooks-Gunn, J., P. Klebanov, and G. Duncan. 1996. Ethnic differences in children’sintelligence test scores: Role of economic deprivation, home environment andmaternal characteristics. Child Development 67:396-408.

Daly, M., and M. Wilson. 1988. Evolutionary social psychology and family homicide.Science 242:519-22.

Devlin, B., M. Daniels, and K. Roeder. 1997. The heritability of IQ. Nature 338:468-71.Diamond, J. 1997. Guns, germs, and steel: The fates of human societies. New York: Norton.D’Souza, D. 1998. A world without racial preferences. The Weekly Standard, 7 December,

37-42.Dworkin, R. 1998. Is affirmative action doomed? New York Review of Books, 6 October,

56-60.Feldman, M., and R. Lewontin. 1975. The heritability “hang-up.” Science 190:1168.Flynn, J. 1984. The mean IQ of Americans: Massive gains, 1932 to 1978. Psychological

Bulletin 95:29-51.. 1987. Massive IQ gains in 14 nations: What IQ tests really measure. Psychological

Bulletin 101:171-91.Goldberger, A., and F. Manski. 1995. Review article: The Bell Curve by Herrnstein and

Murray. Journal of Economic Literature 333:762-76.Gould, S. 1978. Morton’s ranking of races by cranial capacity. Science 200:503-9.. [1981] 1997. The mismeasure of man. New York: Norton.Hacker, A. 1992. Two nations. New York: Scribner’s.Haier, R., B. Siegel, K. Nuechterlein, E. Hazlett, J. Wu, J. Poeh, H. Browning, M.

Buchsbaum. 1988. Cortical glucose metabolic rate correlates of abstract reasoningand attention studied with positron emission tomography. Intelligence 12:199-217.

Harber, K. 1998a. Feedback to minorities: Evidence of a positive bias. Journal of Personal-ity and Social Psychology 74:622-28.

. 1998b. Is feedback to minorities positively biased? Psychological Science Agenda,May/June, 8-9.

Herrnstein, R., and C. Murray. 1994. The bell curve. New York: Free Press.




Hirsch, J. 1970. Behavior-genetic analysis and its biosocial consequences. Seminars inPsychiatry 2:89-105.

Hirsch, J., and T. McGuire. 1977. General intelligence (g) and heritability (H2, h2). In Thestructuring of experience, edited by I. Uzgiris and F. Weizmann. New York: Plenum.

Jencks, C. 1991. Rethinking social policy. Cambridge, MA: Harvard University Press.Jensen, A. 1989. Educability and group differences. New York: Harper & Row.Jensen, A., and F. Johnson. 1994. Race and sex differences in head size and IQ. Intelli-

gence 18:309-33.Kamin, L. 1974. The science and politics of IQ. Hillsdale, NJ: Lawrence Erlbaum.. 1995. Behind The Curve. Scientific American, February, 99-103.Kaniel, S., and S. Fisherman. 1991. Level of performance and distribution of errors in

the progressive matrices test: A comparison of Ethiopian immigrant and nativeIsraeli adolescents. International Journal of Psychology 26:25-33.

Layzer, D. 1976. Science or superstition? A physical scientist looks at the IQ contro-versy. In The IQ controversy, edited by N. Block and G. Dworkin, 194-241. New York:Pantheon.

Lewontin, R. 1976. The analysis of variance and the analysis of causes. In The IQ contro-versy, edited by N. Block and G. Dworkin, 179-93. New York: Pantheon.

Lewontin, R., S. Rose, and L. Kamin. 1984. Not in our genes. New York: Pantheon.Lynn, R. 1998. New data on black infant precocity. Personality and Individual Differences

25:801-4.Mackintosh, N., ed. 1996. Cyril Burt: Fraud or framed? Oxford: Oxford University Press.Michael, J. 1988. A new look at Morton’s craniological research. Current Anthropology

29:349-54.Neisser, U., G. Boodoo, T. Bouchard, A. Boykin, N. Brody, S. Ceci, D. Halpern, J.

Loehlin, R. Perloff, R. Sternberg, and S. Urbina. 1996. Intelligence: Knowns andunkowns. American Psychologist 51:77-101.

Peoples, C., J. Fagan, and D. Drotar. 1995. The influence of race on 3-year-old children’sperformance on the Stanford-Binet: Fourth edition. Intelligence 21:69-82.

Plomin, R., G. McClearn, D. Smith, P. Skuder, S. Vignetti, M. Chorney, K. Chorney,S. Kasarda, L. Thompson, D. Detterman, S. Petrill, J. Daniels, M. Owen, andP. McGuffin. 1995. Allelic associations between 100 DNA markers and high versuslow IQ. Intelligence 21:31-48.

Raz, N., I. Torres, W. Spencer, D. Millman, J. Baertshci, and G. Sarpel. 1993.Neuroanatomical correlates of age-sensitive and age-invariant cognitive abilities:An in vivo MRI investigation. Intelligence 17:407-21.

Reichenback, H. 1938. Experience and Prediction. Chicago: University of Chicago Press.Rowe, D., and H. Cleveland. 1996. Academic achievement in blacks and whites: Are the

developmental processes similar? Intelligence 23:205-28.Rushton, J. 1989. Japanese inbreeding depression scores: Predictors of cognitive differ-

ences between blacks and whites. Intelligence 13:43-51.Salmon, W. 1984. Scientific Explanation and the Casual Structure of the World. Princeton:

Princeton University Press.Sesardic, N. 1993. Heritability and causality. Philosophy of Science 60:396-418.Sowell, T. 1995. Race and culture: A world view. New York: Basic Books.Spitz, H. 1986. The raising of intelligence. Hillsdale, NJ: Lawrence Erlbaum.Weinberg, R., S. Scarr, and I. Waldman. 1992. The Minnesota Transracial Adoption

Study: A follow-up of IQ test performance at adolescence. Intelligence 16:117-35.




Wickett, J., P. Vernon, and D. Lee. 1994. In vivo brain size, head perimeter, and intelli-gence in a sample of healthy adult females. Personality and Individual Differences6:831-37.

Willerman, L., R. Schultz, and J. Rutledge. 1991. In vivo brain size and intelligence.Intelligence 15:223-28.

Wilson, E. O. 1998. Consilience. New York: Knopf.Zindi, F. 1994. Differences in psychometric performance. The Psychologist 7:549-52.

Max Hocutt (Ph.D. Yale), a former editor of Behavior and Philosophy and a specialistin philosophical psychology, has published articles on a variety of topics in Psychologi-cal Review, Behavioral and Brain Sciences, Ethics, Philosophy, Philosophia, Phi-losophy and Phenomenological Research, American Philosophical Quarterly,Review of Metaphysics, and other journals.

Michael Levin (Ph.D. Columbia) is the author ofWhy Race Matters(1997) and theforthcomingSexual Orientation and Human Rights(with Laurence Thomas). Amongrecent publications is “Putnam on Reference and Constructible Sets” in theBritish Jour-nal for The Philosophy of Science(1997).




PHILOSOPHY OF THE SOCIAL SCIENCES / September 2001Keita / THE BELL CURVE AND HEREDITY

Discussions

The Bell Curve and HeredityA Reply to Hocutt and Levin

L. D. KEITAFourah Bay College, Sierra Leone

In “The Bell Curve Case for Heredity,” Hocutt and Levin (1999, 389-415) argued that the average black-white interracial difference of fif-teen points registered on IQ tests is to be attributed maximally togenetic inheritance. Hocutt and Levin began their discussion by not-ing three hypotheses concerning the 15-point differential: (1) the 15-point gap is to be accounted for by social-environmental differentialsgiven the well-known differences in the sociology and history ofblacks in the United States; (2) the 15-point gap derives from culturalpractices specific to the sociological conditions blacks were con-strained to adopt in environments configured for them by the major-ity, racial caste-conscious society; and (3) the 15-point gap is due pri-marily to genetic factors and is impervious to piecemeal egalitariangestures designed to narrow this gap. This is the position taken by theBell Curve (Herrnstein and Murray 1994) and fully supported byHocutt and Levin in their article.

The key point made by Hocutt and Levin was that IQ tests measureintelligence (no matter what causal factors are involved) and thatsuch scores are causally correlated on the average with the so-calledraces. Hocutt and Levin also argued that attempts to improve scoresby way of governmental programs such as Head Start have not beensuccessful. The reason for this, they claimed, is that since IQ scores arereflective of innate intelligence, attempts to change them would beunsuccessful.

In this rejoinder to Hocutt and Levin, I want to argue that thereexists ample empirical evidence to refute the claim that so-called

386

Philosophy of the Social Sciences, Vol. 31 No. 3, September 2001 386-394© 2001 Sage Publications



racial phenotype is necessarily a predictor of IQ and that the explana-tions usually advanced to support such a claim cannot withstandepistemological scrutiny. I argue also that IQ scores, unlike ontic traitssuch as eye color, are a function of tests subjectively designed withinspecific cultural contexts (hence the evident meaningfulness andvalidity of the Flynn effect).

IQ SCORES ARE NOT RACE SPECIFIC

Hocutt and Levin sought to support the Bell Curve hypothesis thatthe black-white IQ difference may be partly if not totally due to hered-ity. This would mean, on Hocutt and Levin’s reckoning, that race asdetermined by phenotype correlates with IQ. But I want to arguecontra Hocutt and Levin that racially designated phenotype does notcorrelate with IQ. For example, the average IQ of Spaniards in Spainhas been calculated at 87; that of Greece, 89; Yugoslavia, 89; Iran, low80s; Iraq (Baghdad), 80 (Lynn, 1978). Lynn also writes,

In India, there is considerable literature on intelligence testing. Fiftyyears ago, the Stanford Binet was given to a sample of 25 postgraduatestudents at the University of Calcutta. The mean IQ was found to be 95(Maity, 1926). A more recent investigation, using a small sample of 25postgraduate students at the University of Calcutta, who took Raven’sTest, produced an incredibly low mean IQ of 75 (Sinha, 1968) . . . .Sinha’s review provides data for 17 groups of children aged between 9and 15 years drawn from a variety of Indian states and numbering inexcess of 5000 cases. All the mean IQ’s lay in the range from 81 to 94, theoverall mean being about 86. (1978, 269)

Obviously, Hocutt and Levin’s claim that “Asians and Jews whohave also frequently grown up in disadvantaged environments”(p. 401) have IQ scores higher than those of more privileged whites isfalse. The population of India constitutes approximately 40% of thepopulation of Asia. And the average IQ of Chinese from mainlandChina (where another 42% of Asians live but where IQ testing is rare,possibly because it is regarded an instrument of class oppression) isno more than that of persons of European ancestry (Flynn 1991, 4).

In the case of Jews, there are reported IQ differentials of such mag-nitude between subgroups that Hocutt and Levin’s claim concerningJews cannot be supported. Jean-Pierre Hèbert (1977) reported on anIQ differential of seventeen points between Jewish youth of Ashke-

Keita / THE BELL CURVE AND HEREDITY 387



nazi and Sephardic ancestry living in the same social environment inNew York City. The only phenotypical difference between bothgroups was that the Ashkenazi environment did not stress the acqui-sition of material goods while the Sephardic households emphasizedwealth acquisition. According to Hèbert, “The Ashkenazi mothersstate that the wealth of their children was not their main concern; theSephardic mothers hoped that their children live in conditions ofgreat wealth” (p. 156). Hèbert reported that the average IQ score ofJews is 103 (p. 162).

HOCUTT AND LEVIN AND THEHERITABILITY OF INTELLIGENCE

The “black box” question concerning the issue of IQ is what por-tion of the variability of IQ scores between individuals and groupscould be attributed to genetic differences. Hereditarians want toargue that “the estimate ventured in BC of .6 implies that, on average,77% of the difference in IQ scores between pairs of whites in theUnited States can be attributed to difference in genes” (Hocutt andLevin, p. 397). The influence of genetics on IQ scores is generallyassumed to range from the highest heritability for monozygotic twinsto the lowest for unrelated individuals of the same racial group.

On account of the assumed genetic equality between monozygotictwins, theorists concerned to determine the influence of the environ-ment on behavioral phenotypes have sought to compare the IQ scoresof monozygotic twins reared apart. Jensen (1970) and Bouchard,Lykken, et al. (1990) offered such comparisons. The approximateheritability factor (h2) for all these tests was put at .75, but there are suf-ficient exceptions to warrant caution. Newman, Freeman, andHolzinger (1937) pointed out in their study of monozygotic twinsreared apart that socially divergent environments could yield h2 fac-tors of less than .45.

But again, even heritability factors as high as .75 cannot be shownto exhaust the environmental range. After all, most of themonozygotic twin studies were conducted in societies where cul-tural, sociological, and economic differentials between individualsare reduced on account of concerns for general social welfare. Thesetests were conducted in Denmark, Britain, and the United States(where the cultural, economic, and sociological differentials between




individuals of European ancestry are less than those concerning othergroups). The deliberate choice of home for monozygotic twins rearedapart is also a factor to be considered.

What is most important, though, is that in the Jensen study (I ex-clude the suspect Burt study), 10% of the twins had IQ divergences ofat least 15 points. The Bouchard et al. study (1990) demonstrated thatthere was a reported IQ divergence of 29 points for one pair of themonozygotic twins. These divergences do cast doubt on the heredi-tarian thesis that the 15-point black-white difference must contain agenetic component.

HOCUTT AND LEVIN ONGENE/ENVIRONMENT CORRELATION

Hocutt and Levin sought to justify their claims about the herit-ability of intelligence by arguing that the IQ scores of Zimbabweanchildren and 15-year-old Ethiopian Falashas are reportedly two stan-dard deviations below their peers in England and Israel. There is aproblem with the results here, given that it is not known how cultur-ally compatible such tests were. Yet Lynn’s study (1978) stated that IQtests on some Africans reported scores ranging from 75 to 88. The cal-culated average of these scores is 84, which is higher than that ofNative Americans, reported at 70 (Garth 1921) and 67 (Hunter andSommermier 1922). Lynn (1978) also claimed that IQ tests on Eskimoswere reported as 70 to 80 in one set of tests. And Kamin (1995) statedthat the average IQ score of Soweto South Africans was higher thanthat of the white mean for comparable subjects.

What is evident from the above is that the IQ scores from industri-alized nations are higher than those of nonindustrialized nations,which are not limited to those of Africa. Asian Indians (86), Eskimos,Native Americans, Mexicans (IQ 83.4, see Hèbert 1977), Iranians (83),Greeks (89), and so on score approximately one standard deviationbelow the norm of industrialized nations. This explanation is borneout by the fact that the reported IQ scores of black army recruits(World War I) from four northern American states were higher thanthose of white recruits from eleven southern states (Shuey 1966, 324-25). These facts would seem to presage aspects of the Flynn effect—given that the northern states of the United States were more industri-alized than those of the South during World War I.




HOCUTT AND LEVIN ON CULTURE AND HEREDITY

Hocutt and Levin argued that since in the case of individuals “cul-ture is itself a function of intelligence,” it should follow that the sameprinciple could hold for groups of individuals. In the case of AfricanAmericans, a reasonable counterargument has been that the oppres-sive conditions under which they were forced to survive led to theirbeing not as equally exposed to the kinds of environmental conditionsthat prepare individuals for the more advanced levels of education.

To support their hypothesis, Hocutt and Levin made the claim thatpersons of African descent in Africa are naturally disposed to“counterintelligent habits.” The authors attempt to substantiate thisquestionable assumption with the following:

In the present context we may ask why, if African cultures had notalready been inferior in technology or energy or resourcefulness to thecultures of their European conquerors, they were so easily dominated.(P. 408)

This argument is specious. First of all, one could make a similar argu-ment in the case of the Native Americans, the Inuit (Eskimos), theAustralian Aborigines, and the Maori of New Zealand, who werecompletely overwhelmed and dispossessed by their European con-querors. Second, would Hocutt and Levin argue that the Roman con-quest of the Vandals, Celts, Saxons, and Visigoths of Europe, and thesubsequent imposition of Roman customs, technology, and languageon the conquered constitute proof that the Romans were more intelli-gent than the people they conquered? Would Hocutt and Levin alsoaccept the spurious argument that the Nazi decimation of the Jewishpeople of Germany and Poland—who allowed themselves to beherded into concentration camps—constitute evidence of the intellec-tual inferiority of the Jewish people?

Hocutt and Levin made the related argument (which they claimed,without any proof, to be highly plausible) that one can construe

black culture as an environmental correlate of the genetic potential thatcreated it [and that] since this genetic potential has expressed itself aslower (than white) individual phenotypic intelligence in every envi-ronment of which we have any experience, it is natural that the culturescreated by the interaction of these individual phenotypes should be lessadvanced scientifically and technologically. (P. 408)




This hypothesis is false. The technological environments of parts ofAfrica were much more advanced than any environment in Europefor at least three thousand years in some cases. The technological envi-ronments of ancient Egypt and Nubia were much in advance of thosein Europe for millennia. Hocutt and Levin might note the fact that Greekwriters such as Herodotus (History, Book 2, 115) and Aristotle (Prob-lems [XIV], 909a, and Physiognomica [VI], 812a) depicted the ancientEgyptians and Nubians as being “black skinned and woolly haired”—obvious markers of the African phenotype. The technological envi-ronments of medieval Africa (Ghana, Mali, and Songhay) were also atleast as technologically advanced as parts of Europe in the tenth cen-tury A.D. and similarly for other parts of the African continent (theenvironments of Nok, Benin and Zimbabwe, are obvious examples).

IS THE HOCUTT-LEVIN HYPOTHESIS SCIENTIFIC?

Hocutt and Levin accepted the very questionable Jensen and John-son hypothesis that cranial capacity “correlates significantly with IQboth within and between races” (as cited in Hocutt and Levin, 406). Toaccept this hypothesis would be tantamount to accepting the veryquestionable hypothesis that male humans are more intelligent thanfemale humans because male crania are on the average larger thanfemale crania. By the same logic, elephants and whales should bemore intelligent than humans. This is not the case. The authors alsosupport the claim that “caucasoid crania are larger than Negroid cra-nia” (p. 406). This claim cannot amount to much, given the diversity ofbody types and head shapes found in Africa’s diverse populations.And C. A. Diop, for example, wrote that the “largest skulls discoveredto this day are those of these Negroids of Southern Africa. ” (1981, 31).Again, the crania of Neanderthal man were larger than those of Homosapiens Africanus who migrated into Europe from Africa. Yet, Homosapiens is assumed to be more cognitively able than the bigger-brainedNeanderthals.

HOW TO REFUTE THE HOCUTT-LEVIN HYPOTHESISBY A CRUCIAL THOUGHT EXPERIMENT

Although not stated in Hocutt and Levin’s essay, the explanatoryhypothesis offered by those who support the hereditarian hypothesis




is what might be called the “environmental challenge hypothesis”(Jensen 1998; Lynn 1991). The claim is made that the arctic and tem-perate zone environments in which Europeans and East Asiansevolved were more cognitively challenging than the tropical and sub-tropical areas where Africans evolved. The ice-age conditions of Eur-asia not only produced the lightly pigmented skin color of Europeansand East Asians and the epicanthic eye fold of East Asians, as theargument goes, but also more developed cognitive abilities. Theexplanation is that the cold climate of the last glacial age and thechanging seasons of the Eurasian landmass made survival more diffi-cult and therefore more challenging. According to this theory, theforces of natural selection operating in these more challenging envi-ronments produced human types that were more cognitively ablethan those of the tropical regions of the world.

This theory is purely speculative and can be refuted with a crucialthought experiment. Neanderthals lived much longer (they livedand adapted to the environmental conditions of Eurasia for at least300,000 years) under severe glacial conditions than did Homo sapiensAfricanus migrating from Africa, yet were not cognitively more ablethan the latter. In fact, the ice-age-adapted Neanderthal eventuallybecame extinct while Homo sapiens survived. Furthermore, accordingto this hypothesis, the Inuit (Eskimos), having survived the longest ofall human groups in the coldest environments known, ought to havethe highest IQs. This is not the case. In fact, their reported IQs approxi-mate those of Africans, Asian Indians, and others.

Finally, Hocutt and Levin would have to admit that the most tech-nologically advanced areas of the globe five thousand years ago werefound in tropical and subtropical areas. It is also evident from thearcheological evidence that the originators of the civilizations ofancient Egypt, Nubia, Mesopotamia, and Harappan were indigenousto the tropical and subtropical areas where those technological cul-tures developed.

SCIENCE AND THE HOCUTT-LEVIN HYPOTHESIS

Those who reject the thesis that interracial IQ differences aregenetic explain this difference by appeal to environmental consider-ations. Yet the only explanation offered by Hocutt and Levin is onebased on a logical fallacy: the fact that blacks and whites may differ oncertain ontogenetic traits such as hair color, pigmentation, and so on




does not imply that there are no traits (cognitive abilities being one ofthem) on which both groups do not differ.

One might consider in this regard the fact that the genic differencesbetween blacks and whites is no more than .005 (genic intraracial dif-ference is .0857 while genic interracial difference is .0852) (Nei andRoychoudhury 1982). The same authors wrote concerning the geneticdifferences between European, African, and Japanese populations:“The gene differences between ethnic groups are of the same order ofmagnitude as those between local populations of the house mouseand Drosophila pseudoobscura” (Nei and Roychoudhury 1972, 434-36).

Matters are compounded by the fact that the genetic divergencesbetween the diverse peoples of Africa are at least as great as thegenetic divergences between these groups and individual groupsfrom other continents.

Given the diverse ways in which the Hocutt-Levin hypothesiscould be falsified, the “most reasonable hypothesis” is that humanpopulations are equal in cognitive abilities everywhere in the world,given that the hereditarians have not been able to explain plausiblywhy there should be such natural differences for large populationgroups from any geographical environment. I have pointed out abovethat the hereditarian hypothesis founded on the principle of “climatechallenge” is refutable, given the example of the Neanderthals of theEurasian landmass. The nurturists, on the other hand, do offer con-firmable explanations.

To determine the plausibility of their hypothesis decisively, Hocuttand Levin might argue instead for social transformations that wouldleaven environmental differences. This has already been attempted inthe Scandinavian countries—well-educated populations with envi-able social tranquility—with admirable results. The investments inhuman capital in this instance have been very cost-beneficial, con-trary to what Hocutt and Levin might assume. We are well aware thatapproximately 50% of Scandinavia’s populations would score lessthan 100 on IQ tests. Yet, no one argues that expenditures on the edu-cation and domestic well-being of the members of this sector of thepopulation uses up resources that could have been spent elsewhere(Hocutt and Levin, p. 409).

The challenge to Hocutt and Levin is that if they wish truly to provetheir hypothesis, they should begin to argue vigorously for socialengineering policies that would reduce social environmental differ-ences to a minimum.




REFERENCES

Bouchard, T. J., D. T. Lykken, M. McGue, N. L. Segal, and A. Tellegen. 1990. Sources ofhuman psychological differences: The Minnesota study of twins reared apart. Sci-ence 250:223-28.

Diop, C. A. 1981. Civilization or barbarism. New York: Lawrence Hill.Flynn, J. 1991. Asian Americans: Achievement beyond IQ. Hillsdale, NJ: Lawrence

Erlbaum.Garth, T. 1921. The results of some tests on full and mixed blood Indians. Journal of

Applied Psychology 5:359-72.Hèbert, J -P. 1977. Race et intelligence. Paris: Copernic.Herrnstein, R., and C. Murray. 1994. The bell curve. New York: Free Press.Hocutt, M., and M. Levin. 1999. The Bell curve case for heredity. Philosophy of the Social

Sciences 29:389-415.Hunter, W., and E. Sommermier. 1922. The relation of degree of Indian blood to score on

the Otis intelligence test. Journal of Comparative Psychology 2:257-77.Jensen, A. 1970. IQs of identical twins reared apart. Behavior Genetics 1:133-48.��. 1998. The g factor—The science of mental ability. Westport, CT: Praeger.Kamin, L. 1995. Behind the curve. Scientific American, 272 (February): 99-103.Lynn, R. 1978. Ethnic and racial differences in intelligence: International comparisons.

In Human variation, the biopsychology of age, race, and sex, edited by R. T. Osborne,C. E. Noble, N. Weyl, and C. D. Darlington. New York: Academic Press.

��. 1991. The evolution of racial differences in intelligence. Mankind Quarterly32:99-121.

Nei, M., and A. Roychoudhury. 1972. Gene differences between Caucasian, Negro, andJapanese populations. Science 4047:177.

��. 1982. Genetic relationship and evolution of human races. Evolutionary Biology14:1-59.

Newman, N., F. Freeman, and K. Holzinger. (1937). Twins: Astudy of heredity and environ-ment. Chicago: University of Chicago Press.

Shuey, A. 1966. The testing of Negro intelligence. New York: Social Science Press.




PHILOSOPHY OF THE SOCIAL SCIENCES / September 2001Levin, Hocutt / REPLY TO KEITA

Reply to Keita

MICHAEL LEVINCity College of New YorkMAX HOCUTTUniversity of Alabama

There is much wrong in Keita (2001 [this issue]; all referencesthereto unless otherwise stated).

1. It is a rule of logic that stronger evidence is needed to support aweightier conclusion. By adding weight to our conclusions, Keitamakes our argument appear weaker than it is. This, in fact, is his maintactic. For instance, according to Keita we say that “the 15-point gap isdue primarily to genetic factors and is impervious to piecemeal egali-tarian gestures designed to narrow this gap” (p. 386, emphasis added).In fact, we say no such thing. We limit our skepticism to belief that thewhole gap can be closed by environmental changes. Our acknowl-edgement that only the major part of the gap is hereditary implies thatthere is a part due to environment, a part that (by definition) can beeliminated by environmental means. By page 392, Keita is referring to“the thesis that interracial IQ differences are genetic,” not, notice, tothe hypothesis that they may be partially genetic. Thus, despite ourclear protestations to the contrary, we are represented as holding thatthe entire difference is hereditary. (We repeat that the underlying nor-mative issue is not the forward-looking one of ways to close the racialgap but the backward-looking one of the gap’s causes—whether it iswrongdoing or factors for which no one is to blame.)

Of a piece with this overstatement is a construal of our position as“the claim that so-called racial phenotype is necessarily a predictor ofIQ” (Keita, p. 387, emphasis added). What is the word necessarilydoing here? It is certainly not ours; we speak only of probability andpossibility. The same misrepresentation occurs again when Keitaattributes to us the claim “that the 15-point black-white differencemust contain a genetic component” (p. 389, emphasis added). Wheredid this must come from? We claim only that the genetic hypothesishas great explanatory merit. Thus, again despite our protestations to

395

Philosophy of the Social Sciences, Vol. 31 No. 3, September 2001 395-403© 2001 Sage Publications



the contrary, we are represented as having offered not a hypothesisbut a demonstrated conclusion.

In this connection, note Keita’s argument that we cannot explainthe conquest of Africans by Europeans unless we are prepared toattribute the conquest of the Celts by Rome as proof of inferior Celticintelligence. May has again morphed into must to short-circuit ourpoint, namely, that conquest is an indicator of superior culture, whichmay be due to superior intelligence. (Is Keita sure that British Celticswere as smart as imperial Romans?) Sometimes other contingenciesare plainly at work. For instance, one obvious German advantageover Jews was vastly superior numbers, a factor not at work in the col-onization of Africa. (We also call attention to Jewish overrepre-sentation in pre-Nazi German banking, journalism, music, law, math-ematics, and medicine, without counterpart in black/white relations.)Incidentally, Keita’s counterexample to the superiority of European toAfrican cultures, the Muslim parts of northern Africa during the mid-dle ages, is an irrelevancy. What is at issue is the cultures of the Negroidpeoples of the African subcontinent, not the semitic cultures of Jewsand Arabs.

Finally, Keita objects that we claim plausibility “without anyproof” that (black) culture can be construed as an environmentaleffect of genes. But you need proofs only when claiming necessity. Ahypothesis is shown to be plausible when it is shown to explain thedata in question better than its rivals, and our section “Gene/Envi-ronment Correlation” makes clear the virtues of pointing the causalarrow from genes to environment. One—which seems to have eludedKeita’s radar altogether—is that black/white differences with respectto a cognitive trait vary directly with the trait’s heritability. Asecond isthat placing the origin of the arrow at genes correctly predicts (asreversing it does not) the result of screening off alleged underlyingcausal factors, as in cross-fostering designs.

Contrast the view of culture as an extended phenotypic expressionof a population’s genes with Keita’s own rather diffuse proposal that“industrialization” explains group differences in IQ. To repeat a pointmade in our article, Keita’s proposal leaves it a mystery why somegroups rather than others industrialized. If the English have high IQsbecause of the industrial revolution, where did the industrial revolu-tion come from? Did it just alight in England out of the exogenousblue? Our proposal explains industrial technology. This is not“proof,” but it enhances plausibility.




2. Keita betrays further confusion about our view by attacking theclimatic theory of race differences. He describes the latter as “theexplanatory hypothesis offered by those who support thehereditarian hypothesis” (p. 391), as if the two stood or fell together.Surely the fact that (as he admits) we never mention climate shouldhave told Keita that is not so. As a general matter, since an explanationH entails its explanandum E but not conversely (lest H just repeat E inother words), E is bound to be consistent with ~H. This is why, in gen-eral, the evidence for an explanandum is distinct from that for anyproposed explanations. In particular, while an explanation of geneticrace differences would be welcome, assuming the races do in fact dif-fer, whether they do is to be resolved by twin and adoption studies,genomic analysis, and other data to which evolutionary history isirrelevant.

Alleged facts may reasonably be discounted only when by currentlights they could not be explained. Clairvoyance and extragalactic vis-itations come to mind. Keita’s challenge to hereditarians to explain“plausibly” why there should be intelligence differences for “popula-tions from any geographical environment” lowers the stakesconsiderably.

In any case, this particular shoe belongs on Keita’s foot. The currentconsensus holds that ancestral Caucasoids left Africa 100,000 years agoor more, and that ancestral north Asians left Europe at least 40,000 yearsago. Evolutionary theory virtually requires that populations experi-encing different environments and hence disparate selectional pres-sures over thousands of generations will differentiate—just how, torepeat, being a matter for independent empirical determination. Ifanything, identification of genetic race differences can be expected tothrow light on evolutionary processes. The races have obviouslydiverged in overt morphology and only ultra-Cartesian body/minddualists would doubt that the mind as well adapts. As usual, Keitasets up a straw man when he claims to diagnose the central “logicalfallacy” of such reasoning: “the fact that blacks and whites may differon certain ontogenetic traits such as hair color, pigmentation, and soon does not imply that there are no traits (cognitive abilities being oneof them) on which both groups do not differ” (p. 393). Rectifying thetriple negation, he has imputed to us the absurd tenet that if the racesdiffer genetically on one trait, they (must) differ on all. Such hyper-bole plays no role in our argument.

We note in passing that neither Eskimos nor Neanderthals dis-credit the climatic theory. Eskimos do about as well as European refer-

Levin, Hocutt / REPLY TO KEITA 397



ence populations on the Ravens Matrices (Berry 1966), as the climatictheory predicts. In any case, the theory can only be tested againstspecifications of the environments to which the ancestors of variousmodern populations were subject. Perhaps the major stressors oper-ated on ancestral Europeans and northern Asians after the foundingAmerinds left for the New World. As for Neanderthals, the directionof their evolution is currently conjectural: we certainly cannot say thatearly Neanderthals were as smart as later ones. There is in fact someevidence of an increase in the size of Neanderthal braincases fromabout 100 to 1,400 cubic centimeters during their tenure on earth (InThe Stone Ages 2000). They seem to have been replaced by the moreintelligent Cro-Magnons.

3. Like many Bell Curve critics, Keita calls attention to within-racedifferences, mentioning that Spaniards in Spain have an average IQof 87, Greeks 89, and so on, varying cultures in order to show that“racially designated phenotype does not correlate with IQ” (p. 387).This is poor reasoning. The variation in genetic height within the malesex does not, for instance, negate the genetically based between-sexheight difference. Why could IQ not vary both within and betweenracial groups? What reason has Keita for thinking the cultures hementions are genotypically identical? The standard work on the sub-ject, Cavalli-Sforza, Menozzi, and Piazza (1994), documents geneticvariation within European, north Asian, and sub-Saharan popula-tions (while dividing the human race as a whole into the standardCaucasoid/Mongoloid/Negroid clusters). However wide, within-Caucasoid variance—Lynn (1978, cited by Keita) conjectures that somemay be artifactual—is irrelevant to variance between group means.

This variance emerges most clearly when the IQs of the variousEuropean, north Asian, and African populations are aggregated. Themost pertinent study here is Lynn’s 1991 study, on which Bell Curverelies, not the earlier Lynn 1978 study cited by Keita. Where 100 is themean for a reference population of white Americans, Caucasian IQstend toward the 90s, with northern Europeans at just about exactly100; those of northern Asians tend to be a bit higher than 100; andthose of Negroids fall in the 70s and 80s. Bear in mind that the highestNegroid scores are achieved by American blacks, hybrids with signif-icant (10 to 20 percent) white ancestry. Keita misleadingly cites Kamin(1995), who in turn cited one study in which Sowetan adolescents didas well as whites on the Ravens Matrices, which is a decided singular-




ity in the data. Remember, we are not talking of necessity, but of whatis true by and large and for the most part.

Keita denies that attributing the whole 15-point IQ difference toheredity is consistent with the “small” known genetic differencebetween blacks and whites. (We have been unable to find the precisefigure in Nei and Roychoudhury [1982] that Keita attributes to them.)But he nowhere offers an argument to support the claim that the twoare inconsistent. To put the issue in perspective, recall the old farmerWilliam James quoted as saying, “There is very little differencebetween one man and another, but what there is can be mighty impor-tant.” “Genic differences” that are small by Keita’s measure, namely,variation in heterozygosity, might suffice to produce the seeminglylarge difference in IQ. Keita’s contention, common as it in fact is,involves highly questionable shifts in comparison class. Bear in mindthat chimpanzees are also very “close” to humans in terms of gene fre-quency, simply because most mammalian genes are dedicated tobuilding hearts, lungs, and other structures common to all mammals.Bear in mind too that even if on average whites and blacks differ onlyin one of every two hundred genes, there are thought to be more than3 billion genes in the human genome, meaning the races differ onaverage in 15 million genes. By Cavalli-Sforza, Menozzi, and Piazza’smeasure of “FST distance” (1994, 79-83, especially Figures 2.3.2.A,2.3.2.B, and 2.3.3. and Table 2.3.2), sub-Saharan Negroids are aboutthree times farther from Europeans as they are from each other, andany African population is FST farther from any non-African popula-tion than any two non-African populations are from each other.

4. Keita also (mis)uses outliers when noting the large IQ differencesbetween some monozygotic (MZ) twins.1 Nothing we say rules outthis possibility, which might be caused by, for example, brain damageto one twin. Heritability is a group statistic rather than a record ofindividual cases. A divergence of 29 IQ points for a single pair of MZtwins reared apart is not excessively improbable given the samplesize, even for a heritability of .7.

Highly divergent MZ pairs in the literature Keita cites are young,whereas more recent studies (e.g., Pedersen et al. 1992) indicate thatthe IQs of MZ twins converge with age. The most popular behavioralgenetic explanation for this invokes our old friend, the genetic shap-ing of environments. Preadolescents are situated where others putthem, but as individuals age they choose environments more congru-




ent with their own genetic predispositions. So people with like geneseventually select like environments. Treating environment as anextended phenotype thus predicts the otherwise surprising increaseof h2 with age; nurturism wrongly predicts its decrease under thecumulative impact of experience.

5. A crucial question is what nurturism actually maintains. Citingour argument that a difference in culture is as well explained by a dif-ference in intelligence as conversely, Keita replies that “in the case ofAfrican Americans, a reasonable counterargument has been that theoppressive conditions under which they were forced to survive led totheir being not as equally exposed to the kinds of environmental con-ditions that prepare individuals for the more advanced levels of edu-cation” (p. 390). Notice how extremely vague this claim is. What, spe-cifically, are these environmental conditions supposed to be? Whyhave they been so hard to identify? In particular, why does the racedifference persist even where blacks seem to have the cultural advan-tage? A point we made to which Keita does not respond is that thechildren of middle-class blacks do worse than economically less for-tunate groups. Another point he ignores is that the race differencesappear in early childhood, which rules out most standard environ-mental suspects such as teacher expectation. Another unaddressedpoint is the results of transracial adoption described in our article.Keita seems not to have understood that by “Asians and Jews whohave frequently grown up in disadvantaged environments,” wemeant American Asians and Jews who have suffered poverty and dis-crimination comparable to that of blacks but have managed to over-come these disadvantages. So this point too gets passed over.

Without wishing to seem to endorse positivism, a view we are toldis passé, we wonder what if any empirical content remains in a notionas untestable as “environmental condition.”

6. Keita muddles brain size and IQ, the correlation between whichis not a “hypothesis” but by now a well-established datum (see refer-ences in Hocutt and Levin 1999, note 36; we expect replications toappear by the time the present exchange is published). It does not fol-low that “by the same logic, elephants and whales should be moreintelligent than humans” (p. 391), since IQ rises with brain size whenbody size is controlled for. There is indeed a puzzle about the sexes,because they appear identical in mean IQ although male brains out-




weigh female brains slightly when body mass is fixed (Ankney 1992).It is natural to connect this discrepancy to the specific male advantagein spatial processing, but this interesting issue is not the one Keitaraises.

7. Recurring to a topic mentioned in (2), Keita opposes IQ to whathe rather obscurely calls “ontogenetic” or “ontic” traits like eye color.He says “IQ scores, unlike ontic traits such as eye color, are a functionof tests subjectively designed within specific cultural contexts” (p. 387).Granted, interpreting a verbal analogy test for Australian aboriginesis less than straightforward, but Keita ignores all of the validating cor-relations we presented—the vast number of other measures of apti-tude and achievement that IQ tests predict. These correlations make itreasonable to think IQ tests measure the real, “ontic” comparativeintelligence of people within the same or similar societies. Keita maywell reply that the culture is always different if the mean IQ is, andthere is no a priori way to refute this. But it begs the question unlessKeita can specify the precise environmental variables that, accordingto him, produce the difference, and this, unfortunately, neither he noranyone else has been able to do.

8. Keita concludes, “To determine the plausibility of their hypothe-sis decisively, Hocutt and Levin might argue instead for social trans-formations that would leaven environmental differences” (p. 393). Weare far from clear on what “determine the plausibility of a hypothesisdecisively” means. Is Keita asking us to show that what we claim isevidence really is evidence? Or is he again asking us to prove the BellCurve position beyond question? The positive cash value of his pro-posal is that, since we have not conclusively proved that all of theracial gap in IQ is hereditary, we should get on the side of those whoare trying to change it by altering the environment. This is Block’sidea, discussed in our original article: forget heredity; try improvedenvironments.

We put aside the fact that this proposal would be more reasonableif we knew precisely what environmental improvements we shouldmake and what differences in IQ we could expect them to effect—which we do not. What is more fundamentally wrong with it is thatthe converse argument would be just as good. If we wanted to do so,we could just as well argue that, since the nurturists have not provedit is all environment, they ought to forget environment. They ought to




get on the side of the hereditarians and take up a program of eugenics,or at least forget all intervention programs. We do not endorse thatargument, but it would be just as good.2

We might summarize the dispute between Keita and ourselves inthis way: our hypothesis is that part of the racial gap is hereditary, andKeita replies, in effect, “Since you have not conclusively proven thatall of the gap is hereditary, you ought to conclude that none is.” Thiscombines misrepresentation with fallacy.

NOTES

1. Keita like others worries about environmental correlations due to selective place-ment. Bouchard (1990) maintains that, empirically, such correlations run about .1, andthat overall they distort heritability estimates by about 1 percent.

2. It is often instructive to invert nurturist arguments, for many of them, if made bynaturists, provoke ridicule. Suppose a hereditarian claimed that, since environmentalcauses of within-group clothing differences do not imply environmental causation ofbetween-group clothing differences, the Scots might have a kilt-wearing gene.

REFERENCES

Ankney, C. 1992. Sex differences in brain size: The mismeasure of woman? Intelligence16:329-36.

Berry, J. 1966. Temne and Eskimo perceptual skills. International Journal of Psychology1:207-22.

Bouchard, T. 1990. The genetic architecture of intelligence. In Biological approaches to thestudy of human intelligence, edited by P. Vernon. Norwood, NJ: Ablex.

Cavalli-Sforza, L., P. Menozzi, and A. Piazza. 1994. The history and geography of humangenes. Princeton, NJ: Princeton University Press.

Hocutt, M., and M. Levin. 1999. The Bell curve case for heredity. Philosophy of the SocialSciences 29:389-415.

In the stone ages: Back to prehistoric roots. 2000, 7 June. In Human origins. London: BBCTV/The Learning Channel.

Kamin, L. 1995. Behind the curve. Scientific American 272 (February): 99-103.Keita, L. 2001. The Bell curve and heredity: A reply to Hocutt and Levin. Philosophy of the

Social Sciences 31: 386-394.Lynn, R. 1978. Ethnic and racial differences in intelligence: International comparisons.

In Human variation, edited by R. Osborne, C. Noble, and N. Weyl. New York: Aca-demic Press.

. 1991. Race differences in intelligence: A global perspective. Mankind Quarterly31:254-96.




Nei, M., and A. Roychoudhury. 1982. Genetic relationship and evolution of humanraces. Evolutionary Biology 14:1-59.

Pedersen, N. L., R. Plomin, J. R. Nesselroade, and G. E. McClearn. 1992. A quantitativeanalysis of cognitive abilities during the second half of the life span. PsychologicalScience 3:346-53.




Documents

The Bell Curve Case for Heredity