In the late 1980s, J. Philippe Rushton published an article laying out his "Genetic similarity theory". In our view, it contains many conceptual errors, which are then replicated in his more recent work on race, as well as in Kevin MacDonald's recent books on "evolutionary group strategies". In later installments (and as time permits!), we will comment more fully on Rushton's more recent work. In the meantime, our original commentary criticizing Rushton's BBS target article might be of some interest. Here it is:


Kin selection, genic selection, and information-dependent strategies

John Tooby & Leda Cosmides (1989). Behavioral and Brain Sciences 12, 542-544.

Commentary for Behavioral and Brain Sciences on "Genetic similarity, human altruism, and group selection", by J. Philippe Rushton


Although Rushton explores some interesting phenomena in his target article, the theoretical framework he uses to integrate them suffers from a series of defects. These include 1) the failure to fully understand the theory of kin selection (see, e.g., Dawkins 1979); 2) the failure to distinguish the operation of kin selection as a selection pressure from the operation of adaptations that evolved in response to kin selection (e.g., phenotype matching); and 3) the failure to distinguish circumstances reliably present during human evolutionary history that we can have evolved adaptations to (e.g., encounters with near and distant kin) from recently emerged circumstances that we cannot have evolved adaptations to (e.g., encounters with those of other races).

Kin selection theory explores how natural selection shapes genetically inherited traits that simultaneously impact the reproduction of the bearer of the trait, and the reproduction of other individuals who share the gene(s) underlying the trait (e.g., Hamilton 1964; Williams & Williams 1957; Williams 1966). Rushton proposes an extension to kin selection theory, in which the idea of "genetic similarity" between individuals is substituted for relatedness, as the more general and appropriate concept.

Approached at the level of the individual, there is no single standard of fitness, such as inclusive fitness, that definitively characterizes what the evolutionary process maximizes, because the genome contains subsets of genes whose fitnesses cannot all be simultaneously maximized (Cosmides & Tooby 1981; Dawkins 1982); since selection operates at the genic rather than the individual level, the nature of kin selection and inclusive fitness must be addressed at the genic level (Cosmides & Tooby 1981; Dawkins 1982). Moreover, the question of kin selection is a game theoretic one, about which phenotypic strategy of reproductive trade-offs between bearer and recipient will maximally propagate a gene coding for the strategy; the optimal strategy will depend (in part) on the information available to be used by the strategy. Scrutinized in this way, flaws appear in the intuitive notion of "genetic similarity". At the genic level, there is no genetic similarity: there is either identity, nonidentity, or some information reliably indicating the probability that another individual contains and will propagate a replica. In the absence of constraints on information or strategy implementation, a gene would be selected to promote the reproduction of its replicas, regardless of which individuals they are in. However, situations in which such constraints are absent are vanishingly rare; "green beard" selection (Dawkins 1976) in the real world is limited to aposematic coloration, where predators from other species, through foraging, incidentally solve for the "green beard" genes the otherwise insurmountable problems of the reliable identification of replicas, linkage between the genes used for identification and the genes for conferring benefits, mimicry, and the implementation of altruistic consequences on the "green beard" genes in other individuals.

Leaving aside such exceptional and stringent circumstances, any trait with social consequences will typically involve many genes from many loci, so the issue is, What kin selection principles govern the evolution of adaptations that are polygenic and information-limited? In particular, the question that Rushton addresses is the significance of "genetic similarity", measured across loci, as hypothetically distinguished from genetic relationships that arise due to common ancestry. Rushton's discussion of "genetic similarity theory" (GST) in fact raises two distinct questions: 1) does genetic similarity operate as an evolutionary principle independent of common ancestry, and 2) can and does a phenotype matching process that samples heritable phenotypic markers (in order to modulate altruism and/or mating) operate in humans?

The answer to the first question is straightforward: "genetic similarity" does not arise independently from relatedness in the real world, because of the size of the genome (e.g., Bachmann 1972), and the free recombination it displays when genotypes are compared between non-related (i.e. genetically distant) individuals. Although one might, as a thought experiment, imagine random assortment creating by chance individuals who are very similar genetically, given the estimated 100,000 to 200,000 freely recombining genes present in the human genome, the probability that a Pleistocene human during his lifetime would encounter a non-relative who was substantially more "genetically similar" than the local population average was negligible. Nor would it matter if he did. No plausible mechanism can assay genetic "similarity" across all loci in the genome: the most that can be imagined is a mechanism that monitors a restricted subset of the genotype, comparing a limited number of heritable phenotypic markers between individuals. Assuming such a mechanism detected "genetic similarity" in the sense of such shared markers between two non-relatives, this would still provide no basis for the evolution of altruism between them, because, in the absence of common ancestry, the existence of "genetic similarity" at some loci predicts nothing about the identity of alleles at other loci. Because tracking genetic markers provides no information relevant to whether an unlinked gene is present in a non-relative, an independently assorting gene cannot use such information to pursue an altruistic strategy towards non-relatives. Rushton's invocation of hypothesized linked genes and supergenes cannot save "genetic similarity theory" as an evolutionary principle, because sex and recombination interpose so many recombination events between individuals who are genetically distant enough to qualify as "non-relatives", that few or no linked genes are likely to remain (in fact, the dissociation of linked genes throughout the genome is probably the function of sex; see, e.g., Tooby 1982; Seger & Hamilton 1988).

In contrast, kinship (common ancestry) does create what amounts to linkage -- probabilistic associations between alleles across loci. In the presence of common ancestry, sampling genetic similarity (i.e. recognizable heritable phenotypic markers) at distributed loci becomes a useful predictor of the presence or absence of genetic identity at other loci, and hence provides information on which to base a strategy for the regulation of altruistic acts. Because kinship creates these probabilistic associations across loci, it creates the circumstances in which polygenic adaptations regulating altruistic acts towards kin can evolve. Thus, although the answer to question 1 is "no", genetic similarity theory is not sustainable as an extension of kin selection theory, the answer to question 2 is "yes", the monitoring of "genetic similarity" (i.e., phenotype matching) can potentially have evolved via traditional kin selection in humans, as an adaptation for assessing relatedness between kin, in order to regulate kin-relevant behavioral strategies, such as altruism and mating. Kinship in this sense refers to genetic similarity that has arisen because of shared ancestry, however recent or far back, and however aggregated from many small components, such as it commonly is in a local population (particularly in species' with a rich population structure).

Therefore, all that remains of genetic similarity theory are those parts that are consistent with the standard concept of phenotype matching as a kin-recognition mechanism (e.g., Waldman 1982). Given that kin selection creates the selection pressures involved, what can be made of the phenomena that Rushton weaves together under the rubric of "genetic similarity theory"? It is certainly possible that phenotype matching systems supplement other kin-recognition systems, influencing mating, friendship, and altruism in humans, and the data on assortative mating and affiliation based on quantitative characters is interesting and suggestive. (The functions of assortative mating and "assortative affiliation", however, are not entirely clear, and are certainly not explained by GST as a selective principle.) Given paternity uncertainty and the imperfect reliability of other cues (location, identification of sexual contacts, association with mother, etc.) available under Pleistocene circumstances, using information supplied by heritable phenotypic markers could help in reconstructing the local pattern of kinship, and it would be an important advance in our knowledge to trace out the properties of such a mechanism.

However, Rushton's blood group data only bear tangentially on these issues, and other explanations seem sounder. For example, similarity of blood group antigens, after excluding close relatives, predicts with modest reliability the more diffusely aggregated common ancestry arising out of common derivation from the same ancestral population (see e.g., Mourant, Kopec, & Sobczak 1976). Even after migration to the New World, immigrants tended to live near others from their ancestral locality (i.e., those living on the same street in North America were often from the same small village in Europe; Sowell 1981; Whyte 1955). This practice was so pronounced and widespread, that 50 years after such mass immigration ended, 50% of southern Europeans, for example, would have to be relocated to achieve a random distribution (Glazer 1975). Thus, similarity of blood group antigens is likely to reflect common ethnicity and more specifically similarity of ancestral population derivation, which is associated with present residential clustering and cultural background. This could explain Rushton's data: It is not surprising to find that people befriend more often or have more reproductively successful marriages with those of similar cultural and residential backgrounds, although phenotype matching (on quantitative characters) may reinforce such tendencies. On this view, similarity of blood group antigens is a consequence, and not a cause, of the affiliative patterns he reports.

Finally, it is important to bear in mind that our complex innate psychological mechanisms evolved during the Pleistocene, and were created by histories of selection (see, e.g., Daly & Wilson 1988): modern phenomena such as friction between people of different "races", war between nation-states, and so on, cannot be adaptations to modern circumstances, but rather reflect the operation of Pleistocene adaptations misfiring under modern circumstances. In fact, non-relatives from one's own "race" are only slightly more genetically similar than non-relatives from a different "race" (Lewontin 1982), and this modest difference could not have led to any behavioral adaptations, because in the Pleistocene, humans would not commonly have encountered people from different "races". Instead, competition could only have been between neighboring groups; typically, intergroup conflict would have reflected cooperation with nearer kin against more distant kin. Although in such small-group conflicts, the relatedness of many of the participants in the same coalition must have been very low, the impact of an individual's decisions on coalition formation, coalition fissioning or exclusion, and coalitional aggression, summed over the members of the two groups, would often have aggregated into substantial inclusive fitness effects. This would have promoted the evolution of specialized mechanisms governing human coalitional psychology (Tooby & Cosmides 1988), without recourse to the group selection that Rushton favors.

It is certainly possible that phenotype matching processes play some role in human coalitional psychology, but this role should be limited by how useful such markers would have been as providers of information about the best inclusive fitness strategy for making coalitional decisions during the Pleistocene. Markers do not seem particularly well suited to this task: although they are useful in tracing close kinship links (e.g., who is the father?), the more distant the relationship tracked, the more likely it is that noisy fluctations in background levels would render the markers erroneous sources of information, particularly in the small local populations characteristic of Pleistocene life (e.g., a Swiss may, by chance, look more like the residents of another Swiss village than he does like his own second cousins; he is, however, still likely to resemble his parents and siblings to a recognizable degree). Non-genetic phenotypic traits that are passed from parents to offspring (such as linguistic patterns or cultural practices), but that decay substantially across several generations, may prove to be better trackers and predictors of relatedness among (say) sets of third or fourth degree kin than is the distribution of genetic markers in relatively homogeneous local populations. Irwin's work (in press) on accent as a badge of group membership adds weight to such a view. Although the mechanism of phenotype matching, misfiring maladaptively under modern circumstances, may contribute to tendencies toward inter-ethnic hostility, it certainly does not swamp other factors. For example, immigrants originally from neighboring villages in Italy were prevented from working together because of the serious violence that would erupt; yet these same individuals lived peacefully among Chinese immigrants (Sowell 1981). In sum, we believe Rushton's interesting empirical results could be pursued more productively and framed more illuminatingly if freed from the distorting influence of genetic similarity theory.



Bachmann, K. 1972. Genome size in mammals. Chromosoma (Berl.) 37: 85-93.

Cosmides, L. & Tooby, J. 1981. Cytoplasmic inheritance and intragenomic conflict. J. theor. Biol., 89: 83-129.

Daly, M. & Wilson, M. 1988. Homicide. New York: Aldine.

Dawkins, R. 1979. Twelve misunderstandings of kin selection. Z. Tierpsychol. 51, 184-200.

Dawkins, R. 1982. The extended phenotype: The gene as the unit of selection. Oxford: W.H. Freeman.

Glazer, N. 1975. Affirmative discrimination. New York: Basic Books.

Hamilton, W. D. 1964. The genetical evolution of social behaviour. Journal of Theoretical Biology, 7: 1-52.

Irwin, C. In preparation. Some observations on the nature and manipulation of "Badged" Group Identity. In: Barkow, J., Cosmides, L. & Tooby, J. (Ed.) The adapted mind: Evolutionary psychology and the generation of culture. Oxford: Oxford University Press.

Lewontin, R. C. 1982. Human diversity. New York: Scientific American Library.

Mourant, A.E., Kopec, A.C. & Sobczak, K. 1976. The distribution of the human blood groups. Second Edition. Oxford: Oxford University Press.

Seger, J. & Hamilton, W. D. 1988. Parasites and sex. In Michod, R. E., & Levin, B. R. (Eds), The evolution of sex: An examination of current ideas. Sunderland: Sinauer.

Sowell, T. 1981. Ethnic America. New York: Basic Books.

Tooby, J. 1982. Pathogens, polymorphism, and the evolution of sex. J. theor. Biology, 97: 557-576.

Tooby, J. & Cosmides, L. 1988. The evolution of war and its cognitive foundations. Proc. Inst. Evol. Studies. 88:1-15.

Waldman, B. 1982. Sibling association among schooling tadpoles: field evidence and implications. Anim. Behav. 30:700 - 713.

Whyte, W. F. 1955. Street corner society. Chicago: University of Chicago Press.

Williams, G. C. 1966. Adaptation and natural selection: A critique of some current evolutionary thought. Princeton: Princeton University Press.

Williams, G. C. & D. C. Williams, 1957. Natural selection of individually harmful social adaptations among sibs with special reference to social insects. Evolution. 17:249-253.