This is the style for the title: if it exceeds the width of the page, let it wrap

Behavior and Philosophy, 30, 21-42 (2002). 2002 Cambridge Center for Behavioral Studies THE RACE CONCEPT: A DEFENSE
The Graduate Center, City University of New York ABSTRACT: It is argued against critics that the concept of race is well-formed. The issueis formulated in terms of the classic sense/reference distinction and shown that “race” has asense specified in terms of geographic ancestry, and thereby a reference. Excessiveconstraints on “race,” for instance that races must by definition have signature genes, arerejected. Empirical validation is considered, although the emphasis here is to placeempirical validation in a philosophical context, not answer the empirical questionsthemselves. At several junctures the familiar divisions of the races are compared to thestellar luminosity types of astronomy, which are still serviceable although representing anearlier state of astrophysical knowledge.
Key words: race, clade, genetics, typology, reference.
This paper concerns the concept of race. Much of it centers (naturally enough) on conceptual issues, particularly the meaning or lack thereof of “race,” butquestions of empirical validity are also addressed. With some qualifications Iconclude that the concept is vindicated.
A prefatory word is in order about the burden of proof. Generally speaking, any word used widely and with apparent comprehension must be presumed toexpress some coherent idea. The presumption is defeasible but absent decisivereasons for rejecting such a word, it should be considered well-formed. Vocabularyis acceptable until shown defective.
The Issue
I begin by setting discussion of “race” in the context of standard semantic According to this theory, every descriptor D possesses both a connotation—a set of defining properties—and a denotation, the set (possibly null or singleton) ofobjects (liberally understood) D is true of. D’s connotative properties are those anobject must satisfy to be denoted by D, and jointly they capture the intuitive idea ofD’s meaning; “gene,” for instance, connotes “bearer of genetic material” andthereby denotes bits of DNA, the actual bearers of genetic material.
The example may suggest that a scientific descriptor is legitimate just in case it denotes, but this condition is too strong. For one thing, science deploysnondenoting idealizations such as “perfect insulator.” A more instructive problem AUTHOR’S NOTE: Michael Levin is Professor of Philosophy at the City College and GraduateCenter of the City University of New York. He may be reached at [email protected] [email protected].
is raised by “atom,” freely used of the smallest chemically active units of mattereven though, contrary to the traditional connotation of indivisibility attached to“atom,” these units are composite. Such ostensible conflicts of connotation withdenotation are surprisingly frequent, as witness “fixed star” and “the mass of anobject;” they occur whenever a term is cued by observational properties widely butmistakenly thought to signal its entire connotative suite. Nineteenth centurychemists were led by the constant relative combining weights of substances toconclude—in error—that the ultimate interactants were simple, hence to think itappropriate to call them “atoms.” This sort of case forces a distinction betweenwhat a term denotes and what it is used to talk about. Denotation must still satisfyconnotation, so that, despite its currency, “atom” denotes neither chemical minimanor anything else in nature, since nothing is both indivisible and a unit of achemical element. At the same time, since everyone knows the scientificcommunity employs “atom” with chemical units in mind, the word directsattention to them. And since the scientific community does successfully talk aboutthese units via “atom,” they are generously allowed to count as atoms.
But mere successful use in speaking of something does not sanction a descriptor, for “phlogiston” is spurious despite having been regularly used ofdeoxygenated air. (Recall that phlogiston was a supposed fluid of negative weightreleased during combustion.) The great advantage of “atom” over “phlogiston” isthat the entities “atom” is used to talk about strongly resemble what atoms must bydefinition be. These entities are almost atoms, in virtue of their minuteness, fixedcombining ratios and within-element uniformity, three of “atom”’s connotativeproperties. (Isotopes introduce some within-element heterogeneity.) Deoxygenatedair by contrast is very unlike phlogiston—in weight, for instance—so that althoughoxidation may have cued “release of phlogiston,” there is little pressure to countone as the other. A well-founded descriptor D, it would seem, must not only beused to talk about something, that something must satisfy most of D’s connotation.
(A near-miss descriptor like “atom” may be said to be used to denote possessors ofmost of its connotative traits.) On an alternative analysis proposed by Kripke (1972) and Putnam (1975), scientific terms are compared to connotationless proper names, and allegedlyconnotative properties are construed as theoretical beliefs, some true, some false,about the objects named. This analysis takes “atom” to denote atoms, which oncewere but are no longer believed to be indivisible and need no-one’s generosity toexist. However, the problem then remains of how denotation is acquired, if not viasatisfaction of connotation. Defenders of the Kripke-Putnam view (see Kitcher1993; also Siderits, 1987) typically suggest that a term denotes the cause, orprimary cause, or sustaining cause of its use, but these approaches all seem toimply that “dephlogisticated air” did (and does) after all denote oxygen—animplication widely perceived as a weakness. And the very fact that this implicationis perceived as a weakness confirms that a word’s denotation must not stray too farfrom what its users think it denotes. Perhaps a two-factor account which combinesnaming and connoting will in the end prove most satisfactory (see Devitt & Sterelny, 1999, secs. 2.6, 9.4; their own view, that sense is mode of presentation, istwo factor.) “Atom” survives because some things approximate what atoms are supposed to be, while “phlogiston” has perished because nothing remotely resembles whatphlogiston was supposed to be. But there are intermediate cases, famously typifiedby “witch.” Notwithstanding a medieval consensus on indicia of witchcraft anddisplay by some women of such relevant connotative traits as haggardness, witchesnever existed. Evidently this is because crones who looked right and were therebysomewhat witch-like lacked too many other connotative traits, among thempossession of magical power. They were not witch-like enough. (The persistenceof “witch” as a metaphor is no doubt explained by possession by some individualsof a few of these connoted properties.) Exact criteria for resemblance have yet to be worked out, leaving the range of cases from “gene” to “atom” to “witch” to “phlogiston” somewhat vague. Even so,it allows a compact statement of skepticism about race: skepticism is the view that“race” belongs with “witch” and “phlogiston” rather than “atom” and “gene.”More explicitly, it is the view that no population satisfies sufficiently many of theconnotative properties of “race” for “race” to denote or be used to denote it. To besure, skeptics disagree among themselves on the extent of the mismatch between“race” and reality. According to Ashley Montagu (1964, p. xii), “Race is thephlogiston of our time,” by which he appears to be claiming that no groupresembles even slightly what a race is supposed to be. More moderate skepticssuch as Appiah and Gutmann (1996) compare “race” to “witch.” They grant thatracial categories have a stable interpersonal use that facilitates noncollusiveagreement in sorting human beings but argue that this reliability rests on the sharedfalse belief that skin color, hair texture, and facial features indicate (fictitious)deeper attributes. Cavalli-Sforza, Menozzi, and Piazza (1994), perhaps the mostrespected scientific critics of race, take a similar position: “racial stereotypes havea consistency that allows even the layman to classify individuals. However, themajor stereotypes, all based on skin color, hair color and form, and facial traits,reflect superficial differences that are not confirmed by a deeper analysis” (p. 19).
(Cavalli-Sforza, 2000, restates his antirace argument more popularly.) Nonskepticsfor their part place “race” on the “atom”/“gene” side of the spectrum, as a properlydefined term that denotes or is used to denote.
Useful as this formulation of the issue is, especially in light of the analogies pressed by skeptics, it must be qualified. “Race,” or more properly “races,” is notmeant to name a substance or class of individuals, as are “phlogiston,” “witch,”and the other terms cited, but to mark differences. Affirming the “reality of race”affirms the reality of races, the distinctness of members of differing races.
Speaking abstractly (and for the moment conflating use and mention), since ascheme of categories F1/F2/./Fn is neither a substance nor a class, it cannotliterally be said to exist or fail to exist or to denote or fail to denote. For thisreason, any categorial scheme is almost bound to fail tests for existence that areappropriate for substance terms. However, in an extended sense F1/F2/./Fn can besaid to be real, or more properly well-founded, when (a) there is a criterion that discriminates Fi’s from Fj’s, (b) Fi’s differ from Fj’s in ways beyond that markedby the discriminator, and (c) most of the categories Fi are nonempty. Equivalently,the categories of a scheme “do not exist” if there is no discriminator, or thediscriminator is arbitrary, or most of the categories are empty. In this extendedsense, a scheme’s connotation may be identified with its discriminator and itsdenotation identified with (some function of) the denotations of each category.
Failure to denote on the part of a large minority of categories in a scheme mayleave its denotation indeterminate. (This latter possibility goes beyond the issue ofvagueness, since it can arise even if each Fi is precise.) Note that condition (b)—which is stressed by critics of race in connection with construct validity (taken up below)—may be too demanding. Arguably, adistinction “exists” as long as there is some criterion for assigning objects to itspigeonholes, whatever further significance or utility the pigeonholes may have. Forinstance, there would seem to be a difference between objects now on my left andobjects now on my right, even though there is no other difference between them,and the distinction itself will lapse the moment I move. Most philosophersreluctantly grant Goodman’s (1954) grue/not-grue distinction, where an object isgrue at a time t if t is before Jan. 1, 2010 and the object is green at t, or t is afterJan. 1, 2010, and it is blue at t. Note that all emeralds have so far been grue, andsince the inductive principle (that the future will resemble the past) requires thatemeralds after 2010 will therefore be grue, it seems to imply that they will be blue.
The challenge is to find a way to disqualify induction on grue, but, as I say, the oneway philosophers have not gone is to dismiss grue as unreal. So (a) and (c) alonemay well suffice for the well-foundedness of a distinction, and we will see shortlythat race satisfies them trivially.
The Use of Racial Descriptors to Denote Clades
Neither moderate nor extreme skepticism implies that “race” is devoid of meaning or inherently confused. Both forms of skepticism are consistent with“race” having a perfectly definite connotation, as indeed “witch” and “phlogiston”do. It is consistent with both forms of skepticism that “race” is defective only inbeing true of nothing, as are such innocuous phrases as “natural satellite ofVenus.” “Race” may well be confused, perhaps because of dependence on obscurenotions like biological essence, but this is not implicit in the analogies with“phlogiston” or “witch.” (“Witch” may depend on the confused idea of magic, butmoderate skeptics cite it mainly to illustrate overhasty inference, as the move fromskin color to race supposedly is.) The point may seem tediously obvious, but manypolemicists do in effect treat “race” as if it were a nonsense-syllable on a par withscholastic jargon, and as if those who use it literally have nothing in mind. Thistendentious position should not be imputed to serious critics of “race.” Rather,serious critics who dismiss race as “nonsense” are best viewed as holding that“race” not only fails to denote, but that this failure is an elementary consequence ofmodern biology, just as the failure of “astral influence” to denote is an elementaryconsequence of modern astronomy. On their view, belief in “racial influence” is nonsensical in the way astrology is, and anyone who indulges in racial thinkingdisplays easily remediable (and therefore perhaps willfully culpable) ignorance.
However, critics often reach this conclusion by reading into the race concept unsatisfiable demands that are no part of its actual meaning. The most common ofthese is that races, to deserve the name, must be distinguished by unique geneticfactors. Gould (1984, p. 32) assumes this condition in rejecting race on the groundsthat no “‘race gene’ [is] present in all members of one group and none of another.”Cavalli-Sforza, Menozzi, and Piazza also assume it when they write: The classification into races has proved to be a futile exercise. . . . Allpopulations or population clusters overlap when single genes are considered,and in almost all populations, all alleles are present but in different frequencies.
No single gene is therefore sufficient for classifying human populations intosystematic categories. (1994, p. 19) In the same spirit, Schwartz (2001, p. 1393) argues that “there is only one race—the human race” on the grounds that “there are no alleles that define the blackpeople of North America as a unique population.” (Curiously, while dismissingrace as an “arbitrary social construct,” Schwartz adds that “Racial identificationdoes have importance in the formulation of just and impartial public policies,”2001, p. 1392). A more stringent condition is that there must be genetic signaturesthat in addition fix major psychobiological phenotypes. Thus, Yee (1992, p. 110)objects that the ordinary indicia of race do not “relate inherently to behavior andpotentials.” A third condition is that the boundaries between races must bediscontinuities. For instance, Tooby and Cosmides (1990) identify races with“discrete” human groups. Barbujani, Magagni, Minch, and Cavalli-Sforza (1995,p. 4518) are quite explicit on this point: “a species is divided in races when it canbe regarded as an essentially discontinuous set of individuals.” Evidently, Cavalli-Sforza et al. (1994, p. 19) has the same criterion in mind: Whatever genetic boundaries may have developed [between human groups],given the strong mobility of human individuals, there probably never were anysharp ones, or if there were, they were blurred by later movements. . . . Thesuccessive levels of clustering [of populations by gene frequency] follow eachother in a regular sequence, and there is no discontinuity that might tempt us toconsider a certain level as a reasonable, although arbitrary, threshold for racedistinction.
Finally, it may be required that any two members of one race resemble each othergenetically or phenotypically more than either resemble any member of any otherrace, as when it is objected that within-group variances greatly exceed group meandifferences or constitute a large portion of total variance.
These constraints create an easily dispatched straw man but are no part of the ordinary connotation of “race” or of any serious scientific or normativecontroversy. Not to monger further mystery, “race,” as used by the averageeducated speaker of English, connotes geographic ancestry, by continent or largecontinental subregion. “Negroid” means and is ordinarily taken to mean “of sub- Saharan African descent”; “Mongoloid” means and is ordinarily taken to mean “ofnortheast Asian descent”; and “Caucasoid” means and is ordinarily taken to mean“of European descent.” There are of course other terms with racial overtones, forinstance “Nordic,” “Dravidian,” “Melanesian,” or “native New World.” In fact,(see sec VI) there is some reason to add “southeast Asian” as a fourth largecategory to the conventional Negroid/Caucasoid/Mongoloid trichotomy. At thesame time, the more specific a descriptor, the less pronounced is its racial tone,although many of the objections to the conventional trichotomy to be consideredapply with equal force to narrower groupings. In any event, these other terms tooare ordinarily meant and taken to be meant as referring to descent by geographicalregion.
I suspect that the reader who asks himself what goes through his mind when he thinks of the word “race” will find that it is a geographic criterion. Three lesssubjective lines of evidence also converge on this conclusion. First, it is endorsedby many knowledgeable adversaries of “racism,” among them Andrew Hacker(1992, p. 7), who writes: “In its basic meaning, ‘white’ denotes Europeanantecedents, while ‘black’ stands for Africa.” (E. O. Wilson, 1978, also treats theconstrual of race as ancestry as uncontroversial.) Second, the geographic criterionexplains, as the straw man does not, what opponents of racial discrimination areopposing. It also explains the demand that race-based affirmative actionpreferences and other reparations be directed to descendants of sub-SaharanAfricans brought to America as slaves. Third, the geographic criterion explains, asthe straw man does not, the desire of some American Negroids to be called“African-American.” Since the difference between sub-Saharan African, European, and northeast Asian ancestry is patent, and each category contains numerous individuals, the raceconcept as ordinarily understood satisfies criteria (a) and (c) of the first section.
Whatever further connotative properties of “race” these populations may lack, theyare sufficiently race-like for “race” in its ordinary sense to be used to denote them.
On this analysis, overt traits including skin color and facial bone structure cueracial descriptors because of an inference from these traits to ancestry. Unlike theassociation of haggard appearance with supernatural power, however, thisinference is highly reliable, especially when, as commonly happens, these overttraits coalesce into gestalts. Thus, light-skinned African-Americans are usuallycorrectly identified on the basis of lip eversion and hair texture as ofpredominantly sub-Saharan African ancestry while darker-skinned Hindustanis areseldom misclassified as sub-Saharan African. The widespread consensus in the useof “race” does not rest on systematic error.
“Race,” in a word, is used to denote continental or subcontinental clades.
Cladistic taxonomy (Kitching et al., 1998; also see Andreasen, 1998, for a similarapproach) takes the basic taxon to be the genealogical unit, ancestors–plus–line–(or tree)–of–descent, what according to the present analysis races are. To be sure,the rough-and-ready cladism of customary usage must accommodate threephenomena commonly cited as objections to it: (i) mankind’s origin in sub- Saharan Africa and subsequent radiation into Europe and Asia; (ii) hybridsresulting from contact between previously isolated groups, and (iii) offshootpopulations. The required adjustments create no fundamental difficulty.
(i) Ordinary speakers acquainted with the out-of-Africa scenario are most charitably construed as intending “Negroid” to denote individuals whose ancestors15 to 5000 generations ago (with Harris & Hey, 1999, counting a generation as 20years) were sub-Saharan African and mutatis mutandis for “Caucasoid” and“Mongoloid.” (ii) Hybrid populations with multiple lines of descent are to be characterized in just those terms: as of multiple descent. Thus, American Negroids areindividuals most of whose ancestors from 15 to 5000 generations ago were sub-Saharan African. Specifying “most” more precisely in a way that captures ordinaryusage may not be possible. “> 50%” seems too low a threshold; my sense is thatordinary attributions of race begin to stabilize at 75%. An individual, half of whoseancestors are East Asian and half Caucasian, is to be categorized as just that, ofhalf northeast Asian and half Caucasian ancestry. Nothing in continental cladisticsprecludes mixed ancestry, any more than the concept of a breed of dog excludesmixtures. To be sure, there are populations, for instance in the Caribbean andSouth America, whose ancestry is highly complex from the continentalperspective. Moreover, the proportion of individuals with no dominant continent ofancestry has increased in recent centuries because of greater individual mobilityand contact between previously separated populations. Someday this may be trueof most human beings, and further in the future most human beings may haveroughly similar ancestries. Under those circumstances continental clades would bemostly empty and distinct races would no longer exist, just as, should all dogssomeday be mixtures, breeds will no longer exist (although all dogs and allhumans will still have pedigrees). However, races exist so long as most individualscan trace their ancestry to one of a small number of continental clades, as iscurrently the case.
(iii) Local groups such as Australids can be regarded either as separate groups (although not “races” if they are too small) or branches of larger ones; clades aresufficiently open-textured to permit and sometimes require these questions to besettled by stipulative decision. At the same time, empirical investigation ofdistributions of gene frequencies and heritable phenotypes can motivate somedecisions as the most reasonable (see below).
Marginal hybrids invite further discussion, since their very existence is often said to discredit racial taxa. This negative conclusion follows only if racialboundaries must be discontinuous, an unduly restrictive condition that (as alreadynoted) misdescribes the everyday race concept, and, consistently applied, would 1 A monogenist account of human origins, although not without detractors, is the current consensus:see, for example, S. Horai et al. (1995) and C. Stringer and P. Andrews (1988). For a review stressinggenetic and linguistic evidence, see Cavalli-Sforza, Menozzi, and Piazza (1994). The primaryAfrican/non-African split is usually estimated to have occurred 100,000+ years ago; E. Harris and J.
Hey (1999) push the split back toward 200,000 years ago on the basis of African/non-Africandivergence on an X chromosomal gene; see below.
disqualify many well-established scientific taxonomies. Astronomy, for instance,classifies stars by luminosity class despite borderline cases such as Canopus, andthe anomalous subgroup of white dwarfs whose temperature correlates abnormallywith magnitude. Barbujani, Magagni, Minch, and Cavalli-Sforza (1995) take theobservation that “gene frequencies form smooth clines over all continents” aschallenging “a biological basis for human race classification,” and one supposesthey would reject as arbitrary any fiat boundaries imposed over this variation. Yetscience often imposes sharp boundaries on continua. To recur to astronomy,spectral lines whose strengths vary continuously with temperature are divided intothe 70 slots of the stellar spectral sequence. Since spectral types can overlaysmooth gradients of temperature, it is unclear why population categories cannotoverlay smooth gradients of genetic similarity. The argument from continuityactually impugns not just continental clades but most other classifications byancestry, including families, since kinship diminishes by steps of decreasing sizeas monozygotic twins give way to siblings, who give way to cousins, and so onoutward. It should also be borne in mind that over sufficient (possiblyintercontinental) distance gradual clinal change yields substantial differences.
Nor does the definition of race by ancestry demand genetic signatures, another abatement consistent with post-Darwinian gradualism. Groups differentiategenetically when daughter populations straying from a parental stock accumulatedifferences through natural selection and drift, a process admitting gene backflow,crossflow, and retention of parental alleles. The nontransitivity fostered by thisgradualism stymies even the classical interbreeding test for conspecifics: there areorganisms 1, 2, . . . , n such that each i can breed with i+1 while pairings of 1 withn are sterile. The Aristotelian idea of natural types rigidly demarcated per genus etdifferentia is foreign to population biology, the ordinary conception of race, andcladistics. Kitching et al. (1998) stress that the dyadic relation “taxon x is/is not thesame as taxon y” is ill-defined; the primitive cladistic comparison is the triadic“taxon x is closer to taxon y than to taxon z.” Divergent branches of the sameancestral tree inevitably share a good deal genetically, their similarities attenuatedrather than destroyed by distance over space and time. The various branches ofmankind should be expected to differ at most in the frequencies of shared alleles(see secs III and V). The cladistic definition of race does not positively precludeunique racial genes, and such a “fixed DNA sequence difference” has apparentlybeen found at a region of the X chromosome that controls glucose metabolism(Harris & Hey, 1999). At the same time the definition does not require that suchgenes, should they exist, be expressed as adaptive or socially significantphenotypes, or expressed at all, and indeed the racially differentiatedpolymorphisms just cited are thought to be “synonymous” (Harris & Hey, 1999),that is, not functionally significant or subject to natural selection.
The methodological moral is quite general: given a definition of race by continental clades, the distribution of genes across races becomes a whollyempirical question, so that, in particular, the existence of races is consistent withany discovery about the distribution of genes, just as the existence of Italy isconsistent with any discovery about the distribution of genes over nations. (This point must be distinguished from the possibility of eventual homogeneousancestry.) Because divergent ancestry implies nothing about genetic variety, therepeatedly made observation that all human groups share the vast majority of theirgenes, in excess of 99.9%, is not an objection to the cladistic definition of race.
The datum properly stated is that typical members of the conventionallydistinguished races have 99.9% of their genes in common. The interpretation ofthis overlap among groups is a further empirical question, namely whether .01% orany other specified mean difference in genome between lineages significantlydifferentiates phenotypes by some independent criterion of significance. That a.01% difference is causally inert is not known a priori. It is, for instance, entirelypossible that the genome of the average virtuoso pianist differs from that of theaverage member of the populace (fixing traits independent of musical aptitude) byless than this. Perhaps the most profitable course at present is to estimate theimpact of cladal disparities from the proportion of between-clade phenotypicvariance explained by genotypic variance, as determined by conventional behaviorgenetic designs (kinship, cross-fostering). This option is exercised from a slightlydifferent angle below.
A related issue concerns between-group differences in gene frequency. With striking consistency, the variance in gene frequencies within single populations,including small communities, has been found to be 85% of the variance infrequencies across the human race (Barbujani, Magagni, Minch, & Cavalli-Sforza,1995; Jorde, Watkins, Bamshad, Dixon, Ricker, Seielstad, & Batzer, 2000), afinding sometimes cited to show that the between-group variance component is toosmall to support the race concept. Cavalli-Sforza, Menozzi, and Piazza (1994, p.
19) write: The same polymorphisms are found in most populations, but at differentfrequencies in each, [but t]here has been too little time for the accumulation of asubstantial divergence. The difference between them is therefore small whencompared with that within the major groups, or even within a single population.
As before, however, the frequency variance is an empirical datum about racesconstrued as continental clades, or as Cavalli-Sforza et al. (1994) prefer to say,“major groups.” Indeed, this finding cannot be formulated unless “major groups”are identified by nongenetic criteria. If the existence of races did require thebetween-race component of variance to exceed (let us say) 50% for at least 20% ofany random sample of loci, it would be undermined by the variance data, but, onceagain, the cladistic concept makes no such quantitative demand. Gene frequenciesmay consistently be supposed to be roughly or exactly the same in all majorgroups, since the group discriminator does not involve genetic polymorphisms.
And the variance data too must be interpreted carefully, for, given reasonableassumptions and standards of importance, a 15% between-group variancecomponent may be highly important. To take a simplified but not whollyunrealistic example, suppose variation among certain genes corresponds in 1-1fashion to a normally distributed phenotype, for instance IQ. In particular, imagineequally numerous populations A and B for which µ(A) = 95 and µ(B) = 105, while σ(A) = σ(B) = 14 (var = 196). Then var for A ∪ B is 225, of which the between-group component is somewhat less than 15%. Let it finally be assumed that an IQof 135 is necessary for scientific innovation. Now, 135 lies 2.14z above µ(B) and2.85z above µ(A), hence is reached by 1.3% of B but only .2% of A. There maytherefore be expected to be about six creative scientists in population B for eachone in A, with corresponding group differences in comprehension of nature,technology, medicine and related outcomes. Mean differences that appear smallrelative to intragroup variation may matter greatly.
The foregoing discussion suggests how to operationalize any hypothesis connecting race to other phenotypic and genotypic variables. Saying thatCaucasoids are on average taller than Mongoloids, for instance, amounts to sayingthat individuals of European descent are on average taller than individuals ofnortheast Asian descent. Saying that Negroids possess on average less body fatthan Caucasoids amounts to saying that the mean fat-to-body-mass ratio amongpeople of European descent exceeds that of sub-Saharan African. A comparison oftwo races with respect to a genetic trait is tantamount to a comparison of the meangenetic value of that trait for individuals of the one ancestry to the mean geneticvalue of that trait for individuals of the other.
Validation
A serious critic of race will, I think, grant the existence of continental clades.
He might add with some justice that this much was never in dispute; his claim,rather, is that classifying mankind by continental clade is pointless. All manner ofclassificatory schemes are conceivable; I am told that Borges playfully imaginesthe canine kingdom divided into dogs that are surprised and dogs that are not.
Pressing criterion (b), the critic maintains that the racial division of mankind is nomore interesting than Borges’.
This critical stance must be distinguished from an overall methodological “splitter” preference for narrow taxa as opposed to the “lumping” of mankind intolarge groups. The a priori objection to taxonomic lumping is its arbitrariness; giventhat, in general, every lineage includes subtrees and is itself a subtree of largerstructures, nothing distinguishes the sub-Saharan African/northeastAsian/European partition or any augmentation thereof: “There are clearly noobjective reasons for stopping at any particular level of taxonomic splitting. In fact,the analysis we carry out . . . shows that the level at which we stop ourclassification is completely arbitrary” (Cavalli-Sforza, Menozzi, & Piazza, 1994).
At the level of abstraction at which this argument is pitched, however, nosubstantive disagreement separates splitters from lumpers, because any set ofobjects, including human beings, can be carved up in many ways, into a largenumber of small bits or a small number of large bits. None is more correct than,and all are consistent with, any other. This is obvious when one partition refines another; there is no anomaly in an individual being both European and westernMediterranean. It holds as well for nonoverlapping partitions; there is no oddity inan individual being both European and 6’ tall. Logicians would note that a set ofpredicates over a domain is neither right nor wrong, notions that enter only whenone of the predicates is ascribed to a member of the domain. It is true but entirelytrivial that classification by continental clade is arbitrary; so are all narrower (andwider) classifications. (Here one recalls Locke’s insistence that nominal essencesare “the workmanship of the understanding,” and that no concrete thing has aunique nominal essence. But Locke did not overestimate the significance of thispoint.) The intended complaint against continental clades must then be made more specific. Thinking again of Borges’ dogs, it must be that the facts of genetic varietypreclude interesting generalizations linking geographically extended lineages toother traits, particularly genetic ones, in explanatorily fruitful ways. “Race” is not ascientifically valid construct, or, in older philosophical parlance, it is unprojectible.
In more recent parlance, predicates that do occur in explanatory generalizations aresaid to pick out “natural kinds,” and in this sense races are not natural kinds.
It is quite true that scientists who employ the race concept do coordinate the geographic criterion with a genetic one. Brues’ (1977, p. 1) text People and Racesinitially characterizes a race as “a division of a species which differs from otherdivisions by the frequency with which certain hereditary traits appear among itsmembers,” only afterwards associating races with “particular geographic areas.”So without systematic heritable differences between ancestries derived fromdifferent continents, races as conventionally understood may be ignored. I hastento add that it would not therefore follow that races do not exist, since races wouldcontinue to satisfy criteria (a) and (c) discussed in the first section.
In fact, heritable traits do vary systematically among Negroids, East Asians, and Caucasoids. The most familiar involve the facial features that critics of racedeem superficial: the shape of the nose root, ears, lips, jaw, and cheeks, thicknessof beard, and shape and thickness of hair. Less visible somatic differences betweenclades include age of tooth eruption (earlier in sub-Saharan Africans than in Asiansand Europeans; Bailit, 1976), subcutaneous fat (Caucasoids over Negroids),muscle (Negroids over Caucasoids), and elongation of physique (Negroids tend tonarrow hips, short trunk, long arms and legs; Mongoloids tend to long trunks andshort legs, Caucasoids intermediate). Descent predicts litter size: Negroids produceon average roughly 16 fraternal twin pairs per 1,000 births, Caucasoids 8, andMongoloids 4 (Brues, 1977). Continental clades also differ in the frequency ofblood groups in the A/B/O, M/N, and Rh systems and in protein components of theblood. Since blood groups express genes fairly directly, they are prominent inestimates of genetic diversity and relatedness, discussed below. Entine (2000) citesa variety of descent-related anatomical and microphysiological differences 2 A partition is a scheme of mutually exclusive and jointly exhaustive categories; one scheme refinesanother when every category in the second is partitioned by the first. In view of hybrids and vagueboundaries, the racial scheme is more like the topologist’s open covering.
affecting athletic performance. (Some of these involve intracontinentalpopulations; see the discussion of averaging in the section of that name.) There are also apparent heritable differences in psychological functioning, particularly in phenotypic IQ, a valid, unbiased proxy for intelligence (Neisser etal., 1996). (It is unnecessary in the present context to decide whether “intelligence”names a unitary ability or a cluster of more specific ones.) A mean one-standard-deviation difference in measured phenotypic IQ between American Negroids andCaucasoids is not seriously disputed, and a Mongoloid/Caucasoid difference issuspected. Older studies and more recent psychometric experiments consistentlyshow indigenous sub-Saharan Africans underscoring European referencepopulations and controls by one to two standard deviations (Lynn, 1991; Rushton,2001; Zindi, 1994). While systematic review of the question of genetic factors inthese discrepancies is inappropriate here (see Hocutt & Levin, 1999; Levin, 1997),two lines of evidence support genetic involvement firmly enough to warrantretaining broad racial categories pending further inquiry. The first is the correlationof the size of white/African-American mean differences on IQ subtests with(within-group) subtest heritability, a result difficult to explain by means ofenvironmental variables, for these variables would have to produce larger effectsas sensitivity to environment dwindles. The second concerns the results oftransracial adoption, the experimentum crucis in this subject. The IQs at lateadolescence of Negroids and Mongoloids reared in Caucasoid families have beenfound to differ from the Caucasoid mean and approximate Negroid and Mongoloidmeans, respectively; moreover, the IQs of adoptees with one Caucasoid and oneNegroid parent fall about midway between those of adoptees with two Negroidparents and those of the birth children of the adoptive families, a result hard toreconcile with wholly environmental causation (Frydman & Lynn, 1989; Levin,1994; Weinberg, Scarr, & Waldman, 1992).
In virtue of the differences cited, continental clades appear to satisfy condition The reader might protest that I have ignored studies that fail to find differences between clades, but this objection is ill-founded empirically, and moreimportantly, methodologically. As an empirical matter, few psychological andsociological studies in this area confirm null hypotheses. Where intergroupcomparisons are possible, particularly in heterogeneous societies, the usual patternis phenotypic variation with, for example, northeast Asians most successfulacademically and least crime prone, Caucasoids somewhat less successfulacademically and more crime prone, and Negroids least successful academicallyand most crime prone (Jaynes & Williams, 1989). The sense of controversy in theliterature on racial differences is sustained largely by conceptual criticisms of thephenotypic data, often the alleged shortcomings of IQ tests, and the urgency of thesearch for environmental explanations (however forced) of the outcomedifferences. These criticisms are discussed in the publications cited in Levin (1997)and Hocutt and Levin (1999). A third conceptual criticism, the possibility ofdisaggregating data about continental clades into data about subgroups, isconsidered in a later section.
Methodologically, the selectiveness objection overlooks the asymmetry between claims of sameness and claims of difference. To establish that two thingsare so similar that distinguishing them is pointless, they must be shown to be alikein all (or, more weakly, virtually all) respects belonging to a contextually specifiedbut normally wide range. Wine from vineyard A is indiscriminable from wine fromvineyard B, for instance, only if they share the same aroma, flavor, alcohol content,color, and so on for all oenologic variables. To distinguish two things, on the otherhand, they need only be shown to differ in some significant way (or someminimum number of significant ways), whether or not there are also manysimilarities, some possibly significant. The difference in mass of the nuclei ofuranium 238 and uranium 235 atoms shows that uranium comes in two kinds, aphysical difference not offset by the identical chemical behavior of U235 andU238. This asymmetry also governs the differentiation of groups. The cladisticconstruct is invalid only if there are no significant differences in heritablephenotypes between individuals of differing conventional races (clades), valid ifthere are some significant differences, however many exact resemblances theremay also be. One must not be gulled by the defense attorney who argues “Theprosecution has only a few witnesses who saw my client rob the bank; I havehundreds who didn’t.” A pair of studies (Exner et al., 2001; Yancy et al., 2001) of the interaction of heart medicine with self-identified race clarifies the point. One found that the drugenalapril significantly reduced the risk of hospitalization for heart failure in whitepatients but did not reduce this risk among similar black patients (Exner et al.,2001; the authors speculate that the difference is due to race differences inresponse to certain enzyme inhibiting agents); the other study found no relationbetween race and response to the drug Carvedilol. It would clearly be fallacious toconclude that the similarity of response to Carvedilol cancels the difference inresponse to enalapril so that the studies taken jointly add nothing to our knowledgeof the biochemical role of race. By itself, the enalapril result shows race xmedication interaction and the need for the associated interaction term inmultivariate regression equations seeking to predict heart failure. One editorialaccompanying these studies (Wood, 2001) references further differences among“whites, blacks and Asians” in polymorphisms for receptors for certain types ofdrugs, and remarks: Genetic differences among racial and ethnic groups usually reflect differences inthe distribution of polymorphic traits, which occur at different frequencies indifferent populations, rather than a trait unique to a particular racial or ethnicgroup. . . . Polymorphisms of the enzymes responsible for drug metabolism[and, as noted, reception] are distributed differently among different racial andethnic groups. (p. 1394) This repudiation of the need for unique racial traits is essentially the view takenhere.
Ironically perhaps, the self-identification of race by the participants in the studies enhances rather than reduces their biological significance. Enalapril (like Carvedilol) can hardly be supposed to differentiate individuals on the basis of an“arbitrary social construct” (Schwartz, 2001). It is hard to see how enalapril, amere chemical, knows not to affect blacks, as it would have to if being black orwhite is exhausted by a tendency to self-ascription. How does it know whatpatients call themselves? Self-ascriptions of race evidently correlate with abiological substrate.
There is a curious catch-22 that threatens to stop discussion of racial dimorphism before it starts. One objection to hypotheses of phenotypic andgenotypic race differences—an increasingly common one, as the accumulatingempirical data makes straightforward denial of these hypotheses very difficult—isthat there is no such thing as race. To meet this objection, the existence of racemust be established before the question of dimorphism is reached; as should beclear, the present paper is concerned mainly with this prior issue. But then it isobjected that for races to be said to exist, significant connections between race andother traits must be shown! The present section is partly a response to that secondobjection, but adequately meeting both at once is as impossible as bringing themboth is illicit. Suitably reworded, this double-barrel dilemma can be aimed at anysubject. Astronomy is a pseudoscience; for to state laws about planets you mustfirst show that planets exist, and to show that planets exist you need laws aboutthem. This circle is best broken by operationalizing the disputed term, as “race” isoperationalized in terms of continental clades. The quest for laws involving theterm can then proceed without worry about the clarity of the term itself. Validityissues can still be raised, but the concept whose validity is under scrutiny is clear.
Averaging
The correlations cited in III might seem to fall short of validating continental clades because they are aggregative. For instance, the cited “Mongoloid” twinningrate is an average over numerous lines of descent, including the very low Japaneserate and higher ones elsewhere in Asia, leaving the overall datum unclear. Similarquestions arise for other “racial” traits. Generalizing about “Africans” on the basisof African-American data might be thought particularly suspect, since mostAmerican blacks are of west African ancestry. (Sober & Wilson, 1998, criticize the“averaging fallacy” in another context.) Averaging is by no means a contrivance for establishing large groups, for it can cloak differences as readily as display them. For instance, should the mean ofgroup A on a trait T be +z and that of equinumerous group B on T be -z, the 0 meanfor A ∪ B obscures the T/A relation. A more fundamental problem with thisobjection is that it implicitly calls into question all statistical description. For whiledata about descendants of northeast Asians, southeast Asians, Europeans, and sub-Saharan Africans all represent averages over diverse subpopulations, so do all non-trivial statistical data. Despite the many forms fires take, fires citywide—acomposite of arson, accidents, and lightning—predicts municipal population,stringency of building codes, local rainfall, and other variables. National farmproductivity for most countries covers disparate climates and soils and could in principle be replaced by individual data points for each acre, or square foot. Dataso atomized are plainly useless, and failure to look beyond them risks blindness toreal trends. Furthermore, once skepticism about averaging becomes a principle(“Don’t generalize over heterogeneous domains”), there is no way to confine itbefore it grinds everything to bits. Doubts about the European-descent categorybased on its subsumption of Celtic, Mediterranean and other lineages must inconsistency be extended to each of these lineages, which are collections of clans,and each of these clans, which are collections of families. There is no stoppingshort of individuals and their genotypes.
Averaging is justified when it reveals real, persistent patterns; nothing further is needed. The within-nation heterogeneity of data sources does nothing toundercut ongoing international discrepancies in farm productivity. If France keepsoutproducing China in wheat, milk, meat, soybeans . . . (more arable land?something in the soil? government policies?) France and China may be treated asunits. The variety of within-continent lineages is likewise irrelevant, if thediscrepancies among humans sorted by continent of ancestry are robust. At the riskof flogging the point to death, there is an obvious sense in which all groupings areman-made. Things do not come with labels to indicate their unity. It is no more orless correct to group (and then average) by continent of ancestry than by any othervariable. One may perfectly properly limit oneself to noting the worldwidedominance in sprinting of descendants of west Africans and the worldwidedominance in long distance track events of descendants of inhabitants of present-day Kenya. But the possible confinement of generalizations to subregions of Africain no way conflicts with the equally accurate generalization that Africans and theirdescendants dominate short and long-distance track.
Ancestry by region is a distinction that is drawn, not an invention of renegade biologists. In Wittgenstein’s words, this language-game is played. All reasonableversions of this distinction, including the one in which the regions are continentalin scale, satisfies conditions (a) and (c) of the first section rather trivially, and, invirtue of observable characteristics reliably correlated with ancestry, the lessexigent condition (b). In virtue of the correlations between heritable functionalphenotypes and ancestry, the conventional division of ancestries by regionapparently satisfies (b) at a deeper level as well. The indeterminacy of aggregationraises secondary issues at best.
Genetic Distance
Systematic differences in heritable traits do not demonstrate variation in allelic frequencies. It is conceivable (I do not speak of probability) that globalvariation in within-group heritable phenotypes such as IQ are wholly due toenvironment factors. The most exhaustive direct study of allelic variation isCavalli-Sforza, Menozzi, and Piazza (1994), which focuses on 120 genetic markersat 49 loci, most controlling blood components, among 42 populations. Since itsauthors repudiate race on the basis of their results, they merit close attention.
Of primary interest is “FST distance,” the metric by which Cavalli-Sforza, Menozzi, and Piazza (1994) encode variation in allelic frequency. FST(X, Y) fordiploid populations X, Y is, intuitively, the sum of the distances of X and Y from ahypothetical mean population. In the simplest nontrivial case, one biallelic gene atone locus, let the alleles be M and N, and let the frequency of M in X be .9 (hencethat of N = .1) and that of M in Y be .2 (hence N = .8). By Hardy-Weinberg thehomozygosity of a population is the sum of the squares of the frequencies of itsalleles, so the homozygosity of X is .92 + .12 = .82 and that of Y is .68.
Heterozygosity is unity less homozygosity, so the heterozygosities for X and Y are.18 and .32, and the mean heterozygosity for {X, Y} is (.18 + .32)/2 = .25. Totalheterozygosity for {X, Y} is derived from the mean frequencies of each allele, here(.9 + .2)/2 = .55 for M and .45 for N. The squares of the mean frequencies yield theaverage homozygosities of each allele, .3025 and .2025, the sum of the squares isthe total homozygosity of the population, and unity less this sum is the totalheterozygosity, here .495. As the separation of X from Y varies directly with thedifference between total and mean heterozygosity, in the example .495 – .25 =.245, and inversely with total heterozygosity, the desired measure, FST, is the ratioof the difference to total heterozygosity. In the example it is .245/.495 = .49.
Reversing the frequencies for Y, so that M = .8, would keep mean heterozygosityat .25 but put the allelic frequencies in the “midpopulation” near those in both Xand Y, reducing total heterozygosity to .255 and FST to .02. Cavalli-Sforza et al.
(1994) measure the separation of groups in their ensemble by averaging FSTs overthe 120 alleles, with correction factors. Distance between two populationsdecreases as genetic variations within them parallel each other.
It is obvious from the construction that, as Cavalli-Sforza et al. (1994) suggest, nothing in this metric corresponds to absolute closeness, apart perhapsfrom the triviality that FST(X, X) = 0. There is no natural value f such that X and Ybelong together just when FST(X, Y) ≤ f. Nothing in the metric permits thejudgment that populations X and Y are “near to” or “far from” each othergenetically. So far this gradualism meshes smoothly with cladistics, which alsorejects absolute closeness (as discussed earlier), and indeed the primitive cladisticcomparison “X is closer to Y than to Z” is operationalizable as FST(X, Y) < FST(X,Z). However, the FST metric, and cladistics, do admit clustering: given an ensembleE of populations, populations X1, . . ., Xn cluster when for all i, j ≤ n and X ∉ {X1, .
. ., Xn}, FST(Xi, X) > max[FST(Xi, Xj)]. To be sure, a cluster in E may dissolvewhen embedded in E'. If in our example neither X nor Y is within .5 of any otherpopulation in an ensemble, they cluster; if other populations lie within .48 ofeither, they do not. Obviously, clustering is undefined for a pair {X, Y}. But withina given ensemble clustering is absolute, invariant under all reasonabletransformations of FST.
3 Cavalli-Sforza, Menozzi, and Piazza (1994) actually reserve the term “genetic distance” for aslightly different statistic. They conceive FST as growing over time t for populations of very similarsize in accordance with a normalized logistic model, so that FST(t) is of the form 1-exp(kt). Then“genetic distance” is -kt.
There prove to be four clusters in the Cavalli-Sforza ensemble: the sub- Saharan African, Caucasian, East Asian, and South Asian populations.
Additionally, each African population is FST-farther from any non-Africanpopulation than virtually any two non-African populations are from each other(1994, pp. 79-83, esp. figs. 2.3.2.A , 2.3.2.B, and 2.3.3., and Table 2.3.2; thedistance from Amerindians to Australids very slightly exceeds that from Africansto some Caucasoids). Cavalli-Sforza, Menozzi, and Piazza (1994) report a factoranalysis of the correlations between allelic frequencies and populations that findsAfrican populations loading heavily and almost exclusively on the first principalcomponent, east Asians loading most heavily on the second principal component,Caucasoids loading about equally on each, and south Asians also loading moreheavily on the second component but nearer the origin. The first three clusteringsreflect the out-of-Africa scenario, in which the hypothesized European/Asian splitoccurred less than half as long ago as the African/European. Cavalli-Sforza,Menozzi, and Piazza (1994) note that “the greatest difference within the humanspecies is between Africans and non-Africans” (p. 83). (They report similar resultsfrom an analysis of mitochondrial DNA mutations from five geographic regions.
Here the major division separates some African populations from the rest ofmankind [including the other Africans]). This tracking of gene frequencies bycontinental clades would appear to reinforce the validation of race by linkage toheritable phenotypes, and it will be noted that three of the four clusters correspondclosely to the “black/white/Asian” categories of the everyday race concept. Thesoutheast Asian cluster, although absent from the everyday concept, may beregarded as an addition to it.
Cavalli-Sforza, Menozzi, and Piazza (1994) dissent from this interpretation on the by-now familiar grounds of gradualism and arbitrariness: By means of painstaking multivariate analysis, we can identify “clusters” ofpopulations and order them in a hierarchy that we believe represents the historyof fissions in the expansion to the whole world of anatomically modern humans.
At no level can clusters be identified with races, since every level of clusteringwould determine a different partition and there is no biological reason to prefer aparticular one. The successive levels of clustering follow each other in a regularsequence, and there is no discontinuity that might tempt us to consider a certainlevel as a reasonable, though arbitrary, threshold for race distinction. (p. 19) Certainly, the populations in the Cavalli-Sforza ensemble can be fused into larger units or fissioned into smaller ones. The next level up consists of unions ofFST-closest groups, with for example the union Eskimo ∪ Chukchi replacing thepair {Eskimo, Chukchi}, and the level above that one fuses the FST-closest of thoseunions. Equally certainly, nature plays no favorites among these levels. Nounification, and that of course includes unification over continents, is the right one.
But while these points may reignite philosophical debate about splitting andaveraging, they do not affect the validation of (essentially continental) clades byclustering. At all levels the FST-neighbor of an African (east Asian, European,south Asian) unit is another African (east Asian, European, south Asian) unit. One level up from the Cavalli-Sforza ensemble, Eskimo ∪ Chukchi is FST-closest toNorth Turkic; one level up from that, Eskimo ∪ Chukchi ∪ North Turkic is closestto Samoyed ∪ Mongol ∪ Tibetan ∪ Korean ∪ Japanese ∪ Ainu, and similarly forother populations. FST neighbors remain spatially contiguous. And at any level theassociated African units remain further from all non-African units than any non-African unit is from any other. African/European/Asian clusters emerge for eachpartition, which show similar factorial structures. Lumping preserves thecorrelation of genetic distance with geographic origin. In practice, Cavalli-Sforza,Menozzi, and Piazza (1994) recognize continental clustering. They describe“European . . . East Asians [and] Africans” as “the three major ethnic groups” and“the major branches of the genetic tree” (p. 100), and remark that “The first split ofthe [human genealogical] tree” is that between Africans and others (p. 135).
The primary aim of this paper has been to explicate “race” as “geographically specified clade.” This explication is not committed to any particular choice ofcontinental regions as the right one, although I have taken pains to explain whythis familiar choice is no less legitimate than its rivals and why the sense of rivalryis largely illusory. With regard to the colloquial black/white/yellow distinction(with southeast Asians added), introduced when demographics was in its infancyand next to nothing was known of genetics, I have already noted its congruencewith the Cavalli-Sforza clusters. In addition to supporting the colloquial distinctionas a reasonable first step toward understanding mankind, this match also suggests afinal remark on taxonomy.
Critics have observed (as I just did) that the race concept was a product of the 18th and 19th centuries, but from this it does not follow that race is ananachronism. The stellar spectral typology mentioned earlier was also introducedin the 19th century, when the Balmer series, the variation in chemical compositionof stars, even the reason they shine at all, were mysteries. Yet spectral typeremains part of astronomy as physicists have discovered the nucleosyntheticprocesses producing the phenomena to which this scheme applies, as well asphenomena like black holes to which it does not. Spectral type is as informative asever about a star’s light, just as conventional race is as informative as ever about anindividual’s ancestry. We know so much more now (about radiation and thetransmission of traits) that the relative informativeness of these descriptors hasdecreased, but in both cases the information can be integrated into morecomprehensive theories. Max Hocutt has argued in a paper that racial distinctionsparallel the earth/air/fire/water distinction of ancient Greek cosmology (2002). Forall the woeful inadequacy of that fourfold classification, we recognize that it wasnot illusory, that it did capture realities the Greeks observed. The impressiveadvances in physics since that time have still not shown the Greeks were wrong tothink that there is such stuff as water or that it differs from fire.
The “phenomenological” branches of physics deal with molar properties of matter without regard to their possible reducibility to more basic microprocesses.
Phenomenological thermodynamics, for instance, states autonomous laws aboutheat, defined merely in terms of observed temperature, as if heat were not knownto be the mean kinetic energy of ensembles of molecules. To some extent these laws reflect the historically accidental order in which physical properties werenoticed, for had some pleistocene Rumford realized that things feel warmer as theirconstituent particles pick up speed, heat phenomena might at once have beenabsorbed into mechanics without first acquiring a special name. Yet thiscontingency neither undercuts thermodynamics, still a staple of first-year physics,nor implies the unreality of heat. Phenotype/race correlations may be regarded asthe phenomenological laws of human genetics, remnants of the historicallyaccidental “chunking” of mankind into easily discernible kinship units before themechanisms of heredity were understood. This contingency neither undercutsgeneralizations about continental clades nor suggests that they are unreal.
Summary and Conclusions
All sides in the race debate agree that humans can be grouped by descent in many ways, that relative to some groupings they vary genotypically, and thatgenetic variance plays some causal role in phenotypic variance. Given this muchagreement, which groups if any to call “races” looks like a purely verbal question,on par with which ground elevations are high enough to call “hills.” Ordinaryusage (worth more, perhaps, than its critics allow) favors black/white/Asian, but inany case the emotional freight carried by “race” demands that something further besaid. The following position strikes me as the most reasonable.
On one hand the coarse division of mankind into sub-Saharan African/European/northeast Asian/southeast Asian/mixtures/ (others classified adlib) is too crude for many purposes. Reliable generalizations about specificpopulation units should be stated as such, not diluted by reference to larger units towhich those populations belong. Better understanding of the genome promisesmore refined knowledge of this sort, which will also “localize” epigenetic accountsof society and societal differences. As the resolving power of human geneticsincreases, uses for the coarse division will decrease commensurately.
On the other hand, they are unlikely to vanish altogether. The coarse division matches the clustering of mankind as presently constituted, suggesting that thereare some trends in human development that can be captured only by generalizingon a continental scale. Moreover, the narrower populations about which specificinformation is coming online will still be individuated by geographical ancestry.
The taxonomic principle which identifies “Caucasoids” as individuals whosedescent can be traced to the European peninsula of Eurasia, is retained when“Lapps” are identified as individuals whose descent can be traced to a specific areaof Europe. Once good-sized clades are admitted, there is no principled basis forbalking at larger ones.
Retention of taxonomy by descent especially is why avoiding “race” or replacing it with circumlocutions like “major group” is apt to accomplish little.
The turmoil surrounding “race” represents a conflict, between an evidently stronghuman interest in ancestry and the conviction that ancestry should bedownplayed—a conviction based in part on empirical doubts about theinformativeness of ancestry but also on the moral precept that people should be judged as individuals. The problem is not racial language but consciousness of thedistinctions it marks. Cavalli-Sforza, Menozzi, and Piazza (1994) consider theword “race” misleading and possibly harmful because it “is coupled in many partsof the world and strata of society with considerable prejudice, misunderstanding,and social problems. Xenophobia, political convenience, and a variety of motivestotally unconnected with science are the basis of racism, the belief that some racesare biologically superior. . . . There is no scientific basis to the belief of geneticallydetermined ‘superiority of one population over another’” (p. 19). It is far fromclear, however, that elimination of the race concept would significantly allay thesemotives. Loss of current racial descriptors might more reasonably be expected toprompt alternatives to express these distinctions, and these euphemisms in theirturn to draw the hostility now directed at “race.” The hypothesis that “major ethnicgroups” differ in intelligence for genetic reasons would surely be execrated asrancorously as is the same hypothesis about race, once the equivalence wasnoticed. Nor is disaggregation apt to restore calm. The hypotheses of an AshkenaziJew/west African or Japanese/Basque difference in genotypic intelligence wouldjust as surely be condemned as baseless and provocative. It is intriguing tospeculate whether in the latter case (assuming psychometric and genetic evidencewere decisive) the existence of these smaller groups would be challenged, perhapswith comparisons to witches or phlogiston. The emotional core of opposition toracial concepts, namely anger at genetic explanations of human behavior, will notbe appeased by banishing a word.
Conventional racial categories will play a diminishing role in biological anthropology, as spectral types do in cosmology. But given mankind’s presentconstitution they will not be completely idle. The more specific categories ofdescent that will play an ever-increasing role are as objectionable as race to thosewho find race objectionable. In light of the probable futility of Newspeak reforms,there is no compelling reason to abandon “race” as customarily employed.
References
Andreasen, R. (1998). A new perspective on the race debate. British Journal for Philosophy of Science, 49, 199-225.
Appiah, A., & Guttman, A. (1996). Color conscious: The political morality of race.
Princeton, NJ: Princeton University Press.
Bailit, H. (1976). Variation in tooth eruption: A field guide. In E. Giles & J. S. Friedman (Eds.), The measures of man (pp. 311-341). Cambridge, MA: Peabody MuseumPress.
Barbujani, G., Magagni, A., Minch, E., & Cavalli-Sforza, L. (1995). An apportionment of human DNA diversity. Proceedings of the National Academy of Sciences of the USA,94, 4516-4519.
Brues, A. (1977). People and races. New York: MacMillan.
Cavalli-Sforza, L. (2000). Genes, peoples and languages (M. Seilenstad, Trans.). Berkeley, Cavalli-Sforza, L., Menozzi, P., & Piazza, A. (1994). The history and geography of human genes. Princeton, NJ: Princeton University Press.
Devitt, M., & Sterelny, K. (1999). Philosophy of language (2nd ed.). Lanham, MD: Entine, J. (2000). Taboo: Why Black athletes dominate sports and why we are afraid to talk Exner, D., Dries, D. L., Domanski, M. J., & Cohn, Jn. (2001). Lesser response to angiotensin-converting-enzyme inhibitor therapy in Black as compared with Whitepatients with left ventricular dysfunction. New England Journal of Medicine, 344,1351-1357.
Frydman, M., & Lynn, R. (1989). The intelligence of Korean children adopted in Belgium.
Personality and Individual Differences, 10, 1323-1326.
Goodman, N. (1954). Fact fiction and forecast. Indianapolis, Indiana: Hackett.
Gould, S. (1984). Human equality is a contingent fact of history. Natural History, Hacker, A. (1992). Two nations. New York: Scribners.
Harris, E., & Hey, J. (1999). X Chromosome evidence for ancient human histories.
Proceedings of the National Academy of Science, 96, 3320-3324.
Hocutt, M. (2002, Summer). Is the concept of race illegitimate? The Independent Review, Hocutt, M., & Levin, M. (1999). The bell curve case for heredity. Philosophy of the Social Sciences, 29, 389-415.
Horai, S., Hayasaka, K., Kondo, R., Tsugane, K., & Takahata, N. (1995). African origin of modern humans revealed by complete sequences of hominoid mitochondrial DNAs.
Proceedings of the National Academy of Sciences, 92, 532-536.
Jaynes, D., & Williams, R. (1989). A common destiny: Blacks and American society. Washington, DC: National Academy Press.
Jorde, L., Watkins, W., Bamshad, J., Dixon, M., Ricker, D., Seielstad, M., & Batzer, M.
(2000). The distribution of human genetic diversity: A comparison of mitochondrial,autosomal and Y-Chromosome data. American Journal of Human Genetics, 66, 979-988.
Kitcher, P. (1993). The advancement of science. New York: Oxford University Press.
Kitching, I., Forey, P., Humphries, C., & Williams, D. (1998). Cladistics (2nd ed.). Oxford: Kripke, S. (1972). Naming and necessity. Cambridge, MA: Harvard University Press.
Levin, M. (1994). Comment on the Minnesota transracial adoption study. Intelligence, 19, Levin, M. (1997). Why race matters. Westport, CT: Praeger.
Lynn, R. (1991). Race differences in intelligence: An international perspective. Mankind Quarterly, 31, 254-296.
Montagu, A. (Ed.). (1964). The concept of race. London: Free Press.
Neisser, U., Boodoo, G., Bouchard, T., Boykin, A., Brody, N., Ceci, S., Halpern, D., Loehlin, J., Perloff, R., Sternberg, R., & Urbina, S. (1996). Intelligence: Knowns andunknowns. American Psychologist, 51, 77-101.
Putnam, H. (1975). Language, mind and reality. Cambridge: Cambridge University Press.
Rushton, J. P. (2001). Black-White differences on the g factor in South Africa: A “Jensen Effect” on the Wechsler Intelligence Scale for Children—Revised. Personality andIndividual Differences, 31, 1227-1232.
Schwartz, R. (2001). Racial profiling in medicine. New England Journal of Medicine, 344, Siderits, M. (1987). Beyond compatibilism. American Philosophical Quarterly, 24, 149- Sober, E., & Wilson, M. (1998). Unto others. New York: Oxford University Press.
Stringer, C., & Andrews, P. (1988). Genetic and fossil evidence for the origin of modern humans. Science, 239, 1263-1268.
Tooby, J., & Cosmides, L. (1990). On the universality of human nature and the uniqueness of the individual: The role of genetics and adaptation. Journal of Personality, 58, 17-67.
Weinberg, R., Scarr, S., & Waldman, I. (1992). The Minnesota transracial adoption study: A follow-up of IQ test performance at adolescence. Intelligence, 16, 117-135.
Wilson, E. (1978). On human nature. Cambridge, MA: Harvard University Press.
Wood, A., (2001). Racial differences in the response to drugs—Pointers to genetic differences. New England Journal of Medicine, 344, 1393-1396.
Yancy, C., Fowler, M., Colucci, W., Gilbert, E., Bristow, M., & Cohn, J. (2001). Race and the response to adrengergic blockade with carvedilol in patients with chronic heartfailure. New England Journal of Medicine, 344, 1358-1365.
Yee, A. (1992). Asians as stereotypes and students: Misperceptions that persist.
Educational Psychology Review, 4, 95-132.
Zindi, F. (1994). Differences in psychometric performance. The Psychologist, 7, 549-552.

Source: https://d1e0u2actw4eb3.cloudfront.net/edu/src/The%20Race%20Concept%20-%20A%20Defense%20(2002)%20by%20Michael%20Levin.pdf