Hamilton, who's been called "the greatest mathematical biologist of the last half of the 20th century", believed group selection "may be a really appropriate term for many human situations". See also Hamilton on inclusive fitness and social behavior in humans and Robert Axelrod on the evolution of ethnocentrism.
Some excerpts from "Selection of selfish and altruistic behavior in some extreme models":
RECOGNITION OF RELATIVES
Galton 36 stated that cows that have dropped out of a moving herd in order to calve will attempt to fight off predators, although a solitary condition is one which usually causes great terror. This phenomenon, of course, is well known with many normally timid species and as an aspect of parental care it is easily understood even in the classical models of natural selection. But the case can also be viewed as conditional altruism encouraged by a particular kind of assortation of like genotype s; the mother and offspring have half their genes in common by direct replication.
Other cases of relationship may be viewed in the same way. The relatives of an individual can be considered to carry his genes in a statistically diluted state with the dilution depending in a definite way on the structure of the relationship (see Part I of the paper in Chapter 2). Distant relationships are obviously more dilute. Indeed, the required precise measure of relationship corresponds more or less to the vague popular notion of 'blood' similarity.
These ideas are illustrated in Fig. 6.2. In general, if we know all the paths that connect two relatives through their common ancestors, we can say what fraction of the genes carried by one can be expected to occur as replicas in the other. When we are concerned with selection at a particular locus it may be useful to think of this fraction, which is probabilistic for any particular pair of relative s, as being the expectation applying to a large set of relatives having like pedigree connections. The fraction can be called the coefficient of relatedness. For relationships which do not involve inbreeding it is the same as Wright's coefficient of relationship, which is the correlation between relatives on the basis of wholly additive gene effects.
Consider a relationship with coefficient b between two individuals A and B. If A knew the relationship and could intelligently consider the consequences of a selfish or altruistic act toward B, he would see that a fraction b of the effect he caused to B's fitness consisted, in a statistical sense, of pure loss or gain to genes which he himself carries. The remaining fraction of the effect he would see as a lo ss or gain to the undifferentiated fragment of the gene pool that makes up the rest of B's 'expected genotype'. If A was interested in the natural selection of the genes which he carries he would regard the effect which might be caused to this second part with indifference: it is merely an increment or decrement to the gene pool as it already stands. The effect which he might cause to the 'related' (effectively identical) part of B's genotype he has to weigh against the magnitude of the effect his action has on his own fitness. In the case of a selfish act, for example, he has to consider whether the gain to personal fitness is going to be greater than a fraction b of the amount of fitness which he destroys in his relative. If it is greater, then the concentration of his own genes in the gene pool will be increased. The result will be either a contribution to selective advance or perhaps, if A is a heterozygote, a contribution to the maintenance of equilibrium.
Of course there is no need to suppose that individuals think intelligently about natural selection or about the genetic constitution of their relatives: this notion is brought in merely to aid explanation. The model we are concerned with here supposes that A's genotype causes him to act in a particular way toward B; if OA and OB are additive effects to fitness of A and B due to A's action, then A's genotype is positively selected if (OA + bOB) > O. On this basis I have suggested that it may be useful in contexts of social evolution to replace the classical concept of individual fitness with a concept of 'inclusive' fitness, consisting of A's own fitness plus the sum of all the effects he causes to the related parts of the fitnesses of his relatives. (Note that A's own fitness in this aggregate has to be taken not as the fitness A expects to express but as what he would expect to express were he not subject to the social effects that come from his relatives.)
The evolutionary outcome of such selection might well be that A appeared in his social behaviour to value his relatives' fitness against his own according to weightings given by b AB . He would always value a unit of fitness in a relative less than a unit of his own fitness except in the special case in which the relative is clonal (as in the case of an identical twin): then b AB = 1. In humans, twinning is too rare for any special social adaptations to have arisen upon this relation- ship, but the idea that clonal groups should show almost perfect co-operation helps to explain other contrasts in nature- for example, the contrasting beha- viour in budded clones of sessile marine organisms and in sessile colonies derived from settling larvae; and the similar but weaker contrast between the perfect co-operation of cells within a multicellular organism and the imperfect co-operation among the sexually produced individuals of the colonies of social insects.
People have a great interest in their blood kin. Popular terminology reflects that interest. Human knowledge of human pedigrees is sometimes amazingly extensive, and it is so especially, considering the dependence on oral tradition, in primitive societies. There is no doubt that an appeal to kinship in general does tend to moderate selfishness and encourage generosity in human social interactions; but, at the same time, human sociability is certainly far from being regulated wholly in accordance with the above principle. Human generosity seems also to depend partly on idealism and partly on correspondences of personality that encourage friendship. Moreover, it is well known that enmities between relatives can be exceptionally bitter and destructive. Progressive human cultures seem to have been rather inclined to reject nepotism. The growing importance in human social life of what might be described as a symbiosis of aptitudes, based in part on training but also in part on genetical endowment; in other words, the growing importance of the civilized socioeco- nomic system, may play some part in this reaction. On the whole, if not in all details, the trend seems to be as one should wish.
It is certain that no other animal has any abstract conception of relation- ship. But many higher animals can recognize individuals by personal attributes, and this permits some discrimination towards very close relatives, possibly to the distance of nephews and grandchildren. In the Japanese macaque a male's position in the rather elaborate social hierarchy seems to depend partly on nepotism and partly on ability as a leader, much as with humans. This and other cases are considered in Crook.40 Less-intelligent animals, which would be unable to remember their siblings or offspring individually, may still be able to differentiate between relatives on the basis of physical similarity and nearness to the place of birth.
There is some evidence, but no proof, that some social insects are hostile to aliens on the basis of physically inherited traits. Such a system would have the advantage of making enslavement and usurpation by parasite queens very difficult, but presumably would have disadvantages through the occasional occurrence of segregating colonies which would fight within themselves. An acquired colony odour homogenized by food exchanges seems to be a more common method by which social insects distinguish the closely related mem- bers of their own colony from aliens. Odour is evidently important as a means of group identification in many species of mammals, especially in rodents.5.42.43 The odours discriminated appear to be acquired rather than inherited.
The phenomenon of visual imprinting suggests how an individual could discriminate on a basis which would be largely genetic. Auditory imprinting could work the same way, but aptitude for vocal mimicry, such as exists in birds, would lessen the reliability of this method.
The simplest case where nearness to place of birth implies relationship is that of nestlings and litter mates. I think the evidence in general shows that agonistic behaviour in groups of siblings is mild compared to what takes place after they leave the nest. There are various seemingly contrary cases where extensive cannibalism takes place among sibling groups- for example, in the embryonic stages of some molluscs. I have suggested reasons for such excep- tions elsewhere (see Chapter 2). A point worth noting here is that structural adaptation for attack and defence in these situations has not evolved, whereas it has done so, for attack at least, in cases where competing young are un- related. 44
If at maturity panmictic dispersal takes place, there is no further distinction to be made; the individuals can be reckoned never to encounter their more distant relatives. If, however, dispersal distances tend to be limited, then neighbouring individuals will tend also to be relatives of varying degree. So also will mating pairs; some degree of inbreeding follows almost inevitably from limited disper- sion. (Panmictic dispersal by only one sex (or type of gamete) is sufficient to ensure random mating. Thus in theory it is possible to have a population structure in which neighbouring individuals have non-inbred relationships through single common ancestors-half sibs, half cousins, and so on. Wright refers to models of this type as involving 'isolation by distance'; more briefly, they can be referred to as 'viscous'. The population is a continuum but in terms of the gene flow of a few generations, distant parts are isolated from one another. Most generally realistic are the cases in which dispersal has something like a normal form with the mode at the place of birth. [. . .]
In all models which lack panmictic dispersal it appears that genetic drift reaches no equilibrium except for that holding in local patches of gene fixation, with the mean size of patches ever increasing. This is so, at least, so long as the factor of mutation is not considered. In the absence of mutation, ultimately each individual should become related with almost unit coefficient to all its neighbours. But during progress to the final state, after a population has been sown from a random mixture, and also subject to the occurrence of mutation (which, of course, an evolutionary model must allow), relationship can be expected to fall off with increasing distance, gradually in the case of the viscous models and stepwise in the case of the stepping-stone models.
No detailed study of social selection in such models has yet been made. There are various difficulties: the analytical complexity of the models, the complexity of the coefficient of inclusive relationship under inbreeding, and even doubt as to the validity of coefficients based purely on pedigrees when there is selection in the system. It seems worthwhile, however, to make some tentative remarks on the basis of a coefficient which is easily obtained from coefficients discussed in the pioneer accounts of the models, and which should certainly serve to point out the main social implications. This coefficient is bAB = 2rAB / (1 + FA), (1) where r AB is the probability of 'identity by descent', or correlation, of random gametes drawn from the two individuals A and B, and FA is the correlation of the homologous genes of A, whose social trait is being considered. To see a rough justification for this formula, it may help to imagine oneself standing in the place of one of the genes of A in the role of matchmaker and considering arranging a successful mating for either A or B; the chances that genes passed on through such matings replicas of the 'matchmaker' gene are! (1 + FA) and r AB, respectively. Alternatively, b AB can be considered as the regression coeffi- cient of the additive genotypic value of B on that of A.
Unless A is specially inbred, FA will be the mean coefficient of inbreeding (F). Wright's work has given much attention to coefficients like F and r. His papers of 1951 and 1965 46 ,47 review a form of analysis of population structure in terms of 'F-statistics' which is particularly useful from the present point of view. This touches, however, a rather technical field and for further informa- tion the reader is referred to Appendix B. The most useful predictions to which the analyses give rise can be briefly summarized as follows (F symbols used in this section are explained in Appendix B):
1. Highly dispersive species (FIT low) will show little positive sociability, although they may be gregarious. They are more likely than indispersive species to be polygamous.
2. Within species, fights will be most damaging when combatants: (a) differ most in heritable characteristics perceptible to the opponent, and (b) derive from distant regions or from different subpopulations.
3. Among species, fighting will be least damaging in species with the most viscous or the most subdivided populations.
4. Co-operative relations and weak social hierarchy within groups will corre- late with the following interrelated factors: (a) small group size, (b) hostility to strangers, (c) endogamy, and (d) high F ST '
5. Conservation of local resources by populations will correlate with the fac- tors cited in prediction 4, subject to the requirement that a population remains in and defends a group territory for many generations.
6. The general level of aggressiveness in a population will rise during periods of increased dispersal and will remain raised until a higher average relation- ship between neighbours has been re-established. The persistence of aggres- siveness after a population crash may be relevant here, but study of FIS and FST over the whole period of a population cycle would be necessary to confirm this.
INTERGROUP HOSTILITY, WARS, AND CRUELTY
In a recent study of biochemical polymorphism in farm populations of the house mouse, Petras 48 found no evidence of other than random mating within local groups, but on the basis of gene-frequency differences between groups he estimated FST = 0.18. Substituting in the formula explained in Appendix B (page 221) this gives b = 0.305, which shows that within groups mice should treat the average individual encountered as a relative closer than a grandchild (or half sib) but more distant than an offspring (or full sib), referring to an outbred population.
It is well known that mice and rats in the wild have unusually united social groups in which dominance behaviour is very restrained and polygamy mod- erate. This mutualism is in strong contrast to the treatment of strangers which is rather vicious; in rats at least, aliens are often killed if they cannot escape. 5 The variability of aggressiveness in captive colonies of mice 49 perhaps depends partly on the extent to which the founding animals came from different wild groups.
Some birds seem to have gone further than rodents in evolving small co- operative groups.35, 50 In some cases several females may contribute eggs to a common nest and take turns in brooding them, or offspring may help their parents to rear subsequent broods, so approaching the situation of some pri- mitively social bees and wasps.51,52 But as with these bees and wasps the stage has not been reached without the appearance of signs of parasitic tendencies (see Part II of the paper in Chapter 2).
In all these cases group territories are defended. From what has been said it would not be surprising to find the demographically stronger groups pushing into the territories of their weaker neighbours. Fission could either precede or follow the taking over of territory. No doubt strong groups would also have higher rates of emission of long-range migrants looking for fresh land to colo- nize or seeking assimilation into other groups, but for the reasons already given I suspect that production of new colonies by budding is relatively more impor- tant in the most mutualistic species.
Similar rather closed and highly co-operative societies appear in primates. As would be expected, with increased intelligence, they are much more com- plex, but intergroup hostility in varying degrees still occurs. 53
With still further increase in intelligence, with increase in ability to commu- nicate (and hence also to organize), with invention of new weapons (primarily for hunting) and ability to transmit culturally the techniques acquired, and with increase in possessions that could be carried off or usurped and used in situ, I find no difficulty in imagining that it could become advantageous for groups to make organized forcible incursions into the territory of their weaker neighbours. In other words, I suggest that warfare was a natural development from the evolutionary trends taking place in the hominid stock. (The origin of the sting in Hymenoptera provides an evolutionary precedent for Ardrey's suggestion 54 that in hominids the use of tools for hunting was quickly followed by their use as weapons of combat. The sting was originally the ovipositor of a vegetarian sawfly.55 It then became a paralyser of insect prey. Finally, in many species it has become a lethal weapon against conspecifics, used both in aggres- sion and defence. The structure and use of the sting depends almost wholly on genotype. The human use of tools depends largely on intelligence; thus the parallel conversions of hand-plus-tool in hominids could be expected to occur much more rapidly. There are other examples of appendages converted as murderous weapons in arthropods. 56 Among arthropods, however, mortal combats are more usual than among vertebrates.)
As mentioned by Wilson,5? something very like warfare occurs within many ant species. The human phenomenon of slavery also has a parallel in ants; and robbery, in various forms, occurs in bees, wasps, and ants, with examples both within and between species. Social, insect colonies with multiple queens also parallel humans in having a 'tribalistic' breeding structure, and whether or not they have multiple queens they are certainly highly intra related and show the expected marked contrast between inward co-operation and outward hostility.
The social insects also parallel humans in two out of four of the extrinsic trends mentioned above as occurring in hominids (communication and posses- sions). If we accept that the elaborate instinctive patterns involved in the 'war', 'slavery', and 'robbery' of the social insects are evolved by natural selection, can we consider it unlikely that in humans also the corresponding phenomena have a natural basis? In humans certainly we are concerned with amorphous and variable inclinations rather than instincts; but, of course, considering what a newcomer our species is to our present ecological situation, compared with any of the highly social ants, bees, or wasps, and how differently the ontogeny of human behaviour is planned, this is what we expect.
I hope that by now the relevance to my theme of the quotation from Nietzsche will have become apparent. Populations are usually viscous and subdivided. On considerations of inclusive fitness we do not expect to find everywhere a Hobbesian war of all against all for the necessities of life- the 'Malthus' of Nietzsche's phrase. There should be restraint in the struggle within groups and within local areas in the interest of maintaining strength for the intergroup struggle or for the united repression of outsiders. We now see at least one reason why the 'struggle for existence' should often become restrained and conventional. I suspect that what Nietzsche meant by a 'struggle for power' is what we would now refer to as striving for social dominance, for territory, or for a place in an 'establishment'. From his all-doubting eminence, relying on common observation, Nietzsche seems to have noticed an important discrepancy in Darwin's theory. At the same time, as appears from the rather muddled paragraph following the one quoted, he was probably far from being able himself to explain what he saw, and he offered no amendment to Darwin's argument. Neither for a biological nor for a human context did he indicate how Malthus is to be answered.
The evolutionary process certainly has no regard for humanitarian princi- ples. Pain is itself evolved to teach the animal to avoid harmful stimuli. In the immediate situation pain warns of a threat to fitness; when a mortally threa- tened animal cannot extricate itself, it may be expected to experience pain until, approximately, its situation is such that further efforts could not possible avail. Thus whatever the major agencies of natural selection happen to be- whatever kind of 'struggle' occurs- pain must be a usual prelude to elimination, and, as determined by the birthrate, elimination has to go on. Darwin, of course, was aware of this and recently the point has been re-emphasized by Williams 24 and Lack.58
Unfortunately with human beings there may be additions to the basic toll of suffering which the nature of life exacts. We are certainly able to learn to exploit the pain of others selfishly and seem even to have instincts acting in this direction. In the light of the evidence for a cruel streak in human nature, 'natural' systems of ethics, including Nietzsche's own, look less attractive. If as I believe, and as others have thought likely,59-61 vicious and warlike tendencies are natural in humans and were formerly (at least) adaptive, Nietzsche's exhor- tation to follow one's instincts makes it not surprising that some of his admirers developed a cruel warlike ethos. Along with a seeming gloss over the Malthusian issue, Nietzsche's attraction to early Hellenic culture perhaps also helped him to overlook the worst aspects of human 'nature'. The Hellenes in Homer were readily warlike but not wantonly cruel. In the Iliad, for exam- ple, the pictures of two cities which Hephaestus the smith god wrought on Achilles's shield well represent the two aspects of tribalistic society that I have described: the inward harmony and the outward snarl. Yet as described by Homer, the warfare around the besieged city is clean: no captives are being tortured for information, no peace emissaries are being sent back blinded and mutilated into the town. In the siege of Troy likewise, revenge is ruthless but the killing of captives is mentioned with censure and there is no torture. There are none of the atrocities of conquest which a little later the Assyrians proudly perpetuated in their bas-reliefs. Hopefully, these traces of relatively humane conduct in the bloody dark-age tribes who fought against Troy may be con- nected with the marvellous contributions shortly to be made by their descen- dants to human culture; but less hopefully, the restraint may have been due to the fact that they were fighting against their not-very-distant kinsmen. In gen- eral, as Freeman has emphasized,59 the atrocities recorded in human history appear to bear witness to the existence of a spontaneous drive to imagine and carry out cruelty which, at least in some people, in some circumstances, has been given free play and public approval.
Bringing together evidence from social anthropology, from early historical literature, from the fossil record of hominid violence and cannibalism,54 from studies of primate behaviour, and from theoretical considerations of social evolution, I find it only too easy to imagine that the genes that reared cruelty out of the primate's aggressive drive have been favoured by natural selection in the hominid line. For the selection of cruelty, indeed, it is unnecessary even to consider inclusive fitness, except insofar as this may have been involved in the speeding-up of progress in mental and linguistic ability. With animals able to communicate and intelligent enough to see distant objectives, there would be nothing surprising in one refusing to communicate some information useful to another; nor would it be unnatural for the other to try to obtain the information by force. With pleasure in cruelty as the motivation to punishment,62 and with cruelty to create terror, an individualistic argument sounds less plausible; but considering inclusive fitness, or group selection (which may be a really appro- priate term for many human situations), there is no theoretical difficulty. For example, a priori, terror seems as good a weapon as false promises in bringing about the early submission or removal of a threatened tribe.
Against the last-mentioned mode of selection, terror tends to unite other- wise dissident enemies. Also, against all modes of selection, the intrinsic resis- tance to cruelty must continually increase. As was pointed out in connection with simpler cases at the beginning of this presentation, evolution produces shields to its own weapons. Perhaps in a slow natural course of events we may become as unconcerned about cruelty and terror as we are now about lies, given and experienced- but this will be a long time to wait and in any case 'natural' Homo sapiens will have invented fresh horrors by that time. [. . .]
Viscous and subdivided population models
An outline of this analysis can be introduced by quoting an identity, which Wright proves:
(1 - FIT) = (1 - FST)(1 - FIS)
In this identity, FIT is the correlation of uniting gametes relative to the array of gametes of the whole population, FIS is the correlation of uniting gametes relative to the array of gametes of their own subdivision, and FST is the corre- lation of gametes drawn randomly from a subdivision relative to the array of gametes from the whole population.
For evaluation of b from equation (1) (page 213), FIT must be used in the calculation of both rAB and FA. If instead FIS is used, an inclusive coefficient is obtained relevant only to selection within the subdivision. It could easily happen that an altruistic trait that was 'too generous' to increase in frequency within its subdivision (except by drift), increases, nevertheless, in the population as a whole due to the more rapid expansion of those subdivisions which contain altruists in higher frequencies. [WD Hamilton. Selection of selfish and altruistic behavior in some extreme models. Man and beast: comparative social behavior, 1971.]