Go to Semiotic Review of Books Home Page
Go to SRB Highlights
Go to SRB Archives


SRB Archives

This review appeared in Volume 10(2) of The Semiotic Review of Books.


Ronald deSousa

Unto Others: The Evolution and Psychology of Unselfish Behavior. Elliott Sober and David Sloan Wilson. Harvard University Press 1998. 394pp ISBN 0-674-93046-0 (hardcover)

Such progress as there is in scientific thinking is not linear, but dialectic. A thinker's reputation is made by rejecting what his predecessors took to be obvious, and thereby constitutes itself as the orthodoxy which must in its turn be overthrown. But "overthrowing" the opposing model often involves showing that the two weren't really so sharply opposed as it seemed.

That view of science is now so commonplace that the time must surely have come for it to prescribe its own rejection. But the book under review doesn't afford the right opportunity, since it exemplifies the truism to perfection.

The story in which it figures as the latest swing of the pendulum concerns the vicissitudes of the concept of altruism in biology. Altruism is not a pre-existing problem which Darwinists among others must solve. It is a paradox created by Darwinism itself, as surely as the problem of evil is a paradox created by theology. No God, no Problem of Evil; no Natural Selection, no Problem of Altruism. By contrast, a Darwinian and anyone else can worry about the problem of the mechanism which accounts for the transmission of inherited characteristics. Similarly, an atheist can wonder about the nature and origin of evil, but there is nothing paradoxical about its existence. In the same way, a non-Darwinian can wonder about the psychology of altruism, but has no reason to find it paradoxical.

A convenient way to see how Darwinism generates the paradox is to reflect on the well-travelled Legend of the Lemmings, according to which lemmings gauge the risk of overpopulation in their own group, and commit mass suicide by drowning in order to avert it, thereby benefiting the species. This story obviously requires that some lemmings remain on shore, as beneficiaries of their fellows' sacrifice, since if they all plunged into the ocean the population would not subsist to reap the benefit afforded by the sacrifice. But now if we grant that those who sacrificed themselves carried a heritable disposition to self-sacrifice, we must also conclude that this beneficial disposition will be sacrificed along with them, and that the inherited characteristics of the remaining lemmings will dictate that they and their offspring will, in future, wait in vain for others to throw themselves into the sea on their behalf. For they have been selected precisely for their selfishness.

The fact that altruism is a problem only for Darwinism might foster either of two possible responses. One is to claim that altruism isn't a concept that belongs in biology, or at least that human altruism isn't, since all behaviour is socially conditioned. Humans transcend their biological nature; there are no genes for behaviour. That attitude received its most bizarre expression in one response to Dawkins's ideas by a prominent British philosopher, who "refuted" Dawkins's Selfish Gene by indignantly proclaiming that genes weren't conscious, and so couldn't form selfish intentions any more than altruistic ones. The more sensible response is to address the puzzle from within Darwinian theory. And that is what has brought on the dialectical pendulum of which Sober and Wilson's book signals the latest swing.

The general strategy for addressing the problem illustrated by the Legend of the Lemmings has been to seek the origin of altruism in units of selection other than individuals. These might be either larger or smaller than individual organisms. Some groups might be thought to endure longer, at least so long as they aren't identified with the sum of their individual members: species for example, or populations. At another scale altogether, genes also endure, and indeed, unlike species or populations, might be envisaged as potentially enduring forever. These are the two preferred candidates for the status of "units of selection" that have alternately focused the efforts of evolutionists. First, (and without even going back to the myth of the pelican), some evolutionists and ethologists such as Wynne-Edwards (1962) and Lorenz (1965), suggested that self-sacrifice existed in nature and could be explained as benefiting the species. Then came tough-minded neo-Darwinists (Williams 1966, Dawkins 1976), who argued that there were no plausible mechanisms for these species-beneficial behaviours to evolve. On the contrary, the known mechanisms of Darwinian selection were bound to make it impossible for genes for altruism ever to take root. A couple more swings of the pendulum later, we reach the presentengaging and persuasive book, which is bound to have a marked effect on what passes for orthodoxy.

Unto Others presents itself as a comprehensive rehabilitation of the concept of altruism, both at the biological and at the psychological level. The book's two parts take up the problem at each of the two levels, with a coda bringing the two back together again. In the first part, Sober and Wilson explain how the concept of altruism, though it obviously has its original home in human motivation, can be redefined in purely behavioural and biogical terms. They then develop a perspective on the vexed problem of "units of selection" which yields a clear model for group selection, which they claim is necessary to the development of biological altruism. In the last chapter of this part the authors survey the anthropological evidence for the existence of the sort of tight group control of individual behaviour. This, they argue, makes it likely that groups sometimes exemplify their model, and can thus be successful not only in expanding their own numbers at the expense of other groups, but also in fostering a significant measure of altruism in their individual members' behaviour. In the second part, they turn to motives, considered as "proximate mechanisms" by which evolution secures certain outcomes, and embark on a philosophical examination of hedonism, egoism, and altruism. These are defined, like Russian dolls, as each inclusively broadening the range of motives provided by the last, which allows the authors to claim that a theory that supports the reality of altruism is ipso-facto a "pluralist" theory. The second part includes a fascinating discussion of the "experience machine" problem, in which subjects are offered a lifetime of pleasurable experience if they would only let their brain be connected to a machine designed to manufacture experience, and thus forgo the objective reality of these experiences's apparent cause. There are also discussions of ingenious psychological experiments designed to determine whether we are really motivated merely by hedonistic motives, or less narrowly by egoistic ones, or most inclusively by genuinely altruistic ones. Claiming to find all these experiments inconclusive, the authors return to the evolutionary argument, and infer that it is more plausible, if not demonstrable, that altruist motives actually exist. How much of a comfort to us this conclusion should be, given the particular way in which it is attained, will be one of the questions addressed in this essay.

To carry the story further, we must be more precise in characterizing altruism. As Sober and Wilson rightly note, an evolutionary perspective on human behaviour "requires us to regard the human mind as a proximate mechanism for causing organisms to produce adaptive behaviors." (200) In particular, intrinsic or ultimate desires, whether or not they turn out to be invariably self-regarding, seem to be psychologically remote from the concrete means of their satisfaction, yet they are the proximate means by which selection secures its "end". So while from the psychological point of view altruism can designate a class of motives, from the biological point of view it is defined strictly as behaviour that "increases the fitness of others and decreases the fitness of the actor." (17). Sober and Wilson should have added that such behaviour must have those consequences not merely by accident, but because it is its function to do so, in some sense that I won't attempt to detail here. The latter part of Sober and Wilson's book is devoted to the psychological question, to which I will turn in a moment. But the larger half of their book is devoted to showing that biological altruism could evolve by natural selection through group-selection mechanisms.

In the legend of the lemmings as I've told it, the role of genes contains the germ of a solution to the problem of altruism. Williams (1966) and Dawkins (1976) stressed that while individuals don't survive at all, and groups or species only in a manner of speaking, survival can be literally attributed to genes, considered as types of DNA sequences linked to heritable traits. Genes are therefore able to benefit directly from the consequences of selection. Since it is genes that are literally replicated, and thus perpetuated, we can think of the evanescent organisms they shape as the instruments of their replication. It is organisms, not genes, that get credited with altruism (or blamed for selfishness). The problem of altruism no more exists for genes than does the problem of evil. It would be meaningless to speak of a gene that sacrificed its own fitness for the sake of another gene. Merely leaving fewer replicas than another allele doesn't yet count as altruism (though the definition quoted above, without the amendment I suggested, wouldn't strictly suffice to exclude it.) For it can't possibly be its biological function thus to lose in the"struggle for existence." On the other hand, if the other allele is of its own type, then it has actually survived, in the relevant sense, not sacrificed itself.

From here on, the problem can be taken over by game theorists. The Master Problem of game theory is the prisoners' dilemma (PD): a type of situation that potentially arises in nearly every group of agents whose behaviour can affect one another and themselves. Here's one non-biological illustration of the PD: at dinner at All Souls College, the wine bill is divided equally between all those Fellows who drink wine, regardless of how much they have consumed. Each Fellow, naturally seeking the best deal for himself, reasons that if he drinks more than the average, he will have bought his wine more cheaply. As a result, each is motivated to drink more than his fellow Fellows. All end up poorer, and the scholarly output of the College suffers. All are worse off than they would have been, had they not been concerned to pursue their own interests. Altruism would solve the problem, if it were understood as a willingness to be penalised for drinking fewer glasses and paying more for each one.

The remoteness of this example from biology actually illustrates an important point: it's the form of the problem that counts, not the exact nature of the "selfishness" or "altruism" involved. Since fitness is the currency of the game of natural selection, fitness is what "altruists" sacrifice and "egoists" will try to maximize; fitness is also what all will end up losing if they act selfishly in PD situations. For the gene controlling an organism's behaviour, the problem is to make the organism sacrifice its individual fitness in those cases where this could favour the replication of its own gene type.

Prominent among the models generated by this way of thinking about the problem was Hamilton's (1963) notion of kin selection, prefigured by Haldane's remark that it would make sense to sacrifice myself for two siblings or four cousins, since any of my genes has a 50% chance of finding itself in my sibling and a 25 chance of finding itself in my cousin. The standard Darwinian story, then, can account for at least some cases of altruism in organisms. Is this a case of group selection? Sober and Wilson argue that it is, noting that Hamilton himself sees it this way (42). Most gene selectionists have denied this, but for reasons that will be clear below I think they are mistaken. They are misled by a certain common-sense view of what counts as "my interests".

However that may be, fans of Kropotkin, socialist biologists and other sworn enemies of sociobiology have pursued the quest for a richer variety of group-selection models that worked. They have attacked the claims of gene selectionism, sometimes with a puzzling intensity of invective. A specific argument was offered by Sober and Lewontin (1982), based on heterozygote superiority. This phenomenon, they claimed, requires us to consider a pair of alleles rather than any individual gene as the minimum unit of selection, thus invalidating the "gene-selectionist" point of view. The paradigm case of heterozygote superiority is the gene for sickle-cell anemia, which subsists in populations exposed to malaria because despite the fact that homozygotes for sickle-cell die early and don't reproduce, heterozygotes having one sickle-cell allele are protected from malaria, and therefore have an advantage over the non-sickle cell homozygotes in those climates. Sober and Lewontin argued that given that the SC allele was sometimes beneficial and sometimes lethal, one could ascribe it no consistent causal power, and therefore it couldn't be said to be the relevant unit of selection. As critics showed, however (Sterelny and Kitcher 1988, 155-157), this argument is fallacious, and results from not applying the gene's-eye-view account consistently. Doing so would lead one to find, first, that the actual causal power on the biochemistry of the blood was identical in both cases, but that it makes a difference whether there is one gene or two affecting the blood in its specific way; second, that the resulting fitness of each gene is simply frequency-dependent, meaning that it has different consequences in different environments where it is more or less likely to meet (and get paired with) another like itself. But that is quite consistent with the strictest gene-selectionism.

In the present book, Sober and Wilson tacitly concede the point, making no mention of the sickle cell example. But they continue to plead for "pluralism", accusing gene-selectionism of being too narrow, in rejecting cases of selection at the group level, or just trivial, in simply interpreting such cases in terms of the resulting gene frequency. Their strategy is essentially based on devising models that exemplify Simpson's paradox, which shows that in some circumstances, itcan be true that in each of two subgroups, classes A and B have a certain statistical relation, which they fail to exhibit in the union of those two groups. Simpson's paradox is often illustrated in terms of a real case of sexual discrimination brought against the University of California at Berkeley, based on the fact that the proportion of women admitted was lower than the proportion of men. (I make no claim to describe the actual facts here.) Upon investigation, it was found that in each division taken singly, the opposite was actually the case. This was paradoxical, but was due to the fact that women applied disproportionately often to divisions, let's say English or Psychology, where the numbers of applicants were high and the overall rejection rate was low. In divisions such as Engineering or Physics, smaller numbers applied and fewer were rejected. We can imagine the following case, in which a higher proportion of women are admitted in each division, but a lower proportion overall, as represented in the following table:

	Applicants 			  Admitted		 Comparison
				      Males	   Females 

		M	F	    #	 %	    #	 %

English		10	100	    1	10%	   11	11%	     F > M

Engineering	40	10	    6	15%	    2	20%	     F > M

TOTAL		50	110	    7	14%	   13	12%	     M > F

How does this apply to the biological case? One of Sober and Wilson's examples, for which they credit Richard Lewontin, pertains to the degree of virulence of viruses or parasites. Within each host, individual selection will favour the most virulent. But excessively virulent strains may kill their hosts too soon and themselves into the bargain. Less virulent strains, which may be viewed as altruistic since they restrain their replication and thus benefit others in their group, may therefore be favoured at the level of group selection, providing that "altruistic and selfish types ... become concentrated in different groups." (29) Thus individual and group selection act as conflicting forces, the first favouring virulent strains, the second moderate ones. In some cases one should be able to observe the effects of group selection gaining the upper hand. The overall population would then cease to reflect the selective advantage in each host of the more virulent strain, just as the Berkeley admissions pattern on the whole failed to reflect the advantage women enjoyed in each of its divisions. A favoured example of this mechanism is myxomatosis, a rabbit virus introduced into Australia to control the rabbit population. What apparently needed to be explained, is that the virus evolved to a moderate rather than a high degree of virulence. Of course, it's possible that this resulted from an adaptation of the rabbit's level of resistance. But the following group-selectionist explanation is tempting, and something like it must surely happen sometimes. As each particular rabbit is invaded by the virus, it plays the role of a temporarily isolated deme, within which various mutants of the virus compete. Within each of these demes, the most virulent, i.e. those whose reproduction rate is fastest, will be more fit by definition. But on the other hand their host may die too soon, leaving the entire deme stranded. They will therefore not survive to press their advantage in the next generation. Thus they may, as demes, come to be outnumbered in the population as a whole by demes that contain no highly virulent mutants. For those alone will survive to colonize the next rabbit.

Does the existence of such models of group selection invalidate the gene's claim to being the primary unit of selection? The answer is not as clear as one might have hoped. Sober and Wilson seem to give two incompatible answers. They argue that as a matter of sheer logic, "altruism requires a process of group selection to evolve"; for altruism is defined as "behaviors that decrease relative fitness within groups but increase the fitness of groups" (p.99). (The less virulent strains of myxoma give up the chance of producing as many clones as the more virulent; but that serves them well in the end because the group to which they belong does better.) At the same time, Sober and Wilson claim that "the concept of genes as replicators, widely regarded as a decisive argument against group selection, is in fact totally irrelevant to the subject... Selfish gene theory calls the genes responsible for [altruistic] behaviors 'selfish' for the simple reason that they evolved and therefore replicated more successfully than other genes. Multilevel selection theory, on the other hand, is devoted to showing how these behaviors evolved." (p.88) This statement, however, could perfectly well be heartily endorsed by Dawkins, as far as I can see. This nicely illustrates both aspects of the dialectic sketched above: you reject theopponent's view, but also show that your own view entails it.

What of the psychological side of altruism?

In the last part of the book, Sober and Wilson turn to the traditional philosophical problem of the nature of altruism. They are at pains to show that while crude versions of hedonism are easily refuted, it seems always possible to construct a version of the theory of psychological hedonism which even the most subtle experiments can't conclusively disprove. This makes way for their own ingenious argument about the likelihood that altruism was actually promoted, by group selection, as the most direct psychologically proximate means of securing fitness. But they seem not to notice that the crucial move in this strategy has already been accomplished by the view they sought to refute in the first part of the book -- the theory of strict gene selectionism. To see why, we need to return to the crucial property that is being sacrificed or promoted by altruistic behaviour. What sort of property is fitness?

Fitness is commonly attributed to individuals, but it is measured in terms of an individual's offspring. We might question, therefore, whether it is ever truly an intrinsic property of an individual. This question itself is ambiguous. One way to deny that it is an intrinsic property is to assimilate fitness to a certain sort of probability of leaving descendants, and to claim that probability is itself not a property of an individual, but a property of an entire system in which, in the "long run", whatever that is, some sort of notional accounting will determine how fertile each individual has been. That problem can be avoided, however, by decreeing that the sort of probability involved is a propensity, a genuine intrinsic property that just happens to be stochastic (cf. Mellor 1991): a property, in fact, which is only contingently correlated to any actual count of future offspring. (Getting accidentally eaten by a bear cuts off an individual's actual lineage, but it needn't, on this account, be allowed to affect its intrinsic fitness.)

The second problem is more interesting. It arises if one questions the very basis of the calculus; but our disposition to take our own interest in our children for granted is so strong that we are inclined not even to notice it at all. For myself considered strictly as an individual, my survival is just the survival of myself. Unfortunately, I as an individual won't survive my body. The common phrase, "survival of the fittest" (a phrase that Darwin didn't use) is actually nonsensical. For an organism may be more or less fit in the sense just described, but its survival is quite another matter, since no individual ever actually survives for more than a puny lifetime. This fact, obvious as it is, is obscured by our fondness for the term "reproduction", which strictly speaking is not something any vertebrate was capable of before Dolly the cloned sheep. But since having children is called "reproducing", we are no doubt encouraged in the assumption that our children are part of ourselves. It's undeniable that people regard themselves as having a direct interest in the welfare of their children: perhaps we are genetically programmed to act as if we thought that, but just as likely we are programmed actually to believe it. "Charity begins at home" rightly passes for a counsel of selfishness, insofar as I can hardly claim any merit for generosity if I regard the beneficiary as just part of myself. Darwin declared that he used the term 'struggle for existence' "in a large and metaphorical sense, including not only the life of the individual but success in leaving progeny." (Darwin 1958, 75). But what is the cash value of the metaphorical idea that "success in leaving progeny" is part of my individual success? As Woody Allen once observed, insofar as I want to achieve immortality, I don't want to achieve it through my works, or through my children, I want to achieve it through not dying. And that method is simply unavailable to individuals. The fact that individuals regard the interests of their children as part and parcel of their own doesn't entail that they are right to do so. It means, instead, that they have been programmed by their genes to endorse an agenda that may not be strictly speaking in their own interest.

And why should they not? From a gene-selectionist perspective, any trick is good so long as it brings the right result in gene fitness. Grouping, merging and regrouping around genes for more or less altruistic dispositions is the trick Sober and Wilson particularly favour. Very likely some cases of Simpson-paradox-type selection do exist, where the altruistic gene is favoured overall despite being swamped by selfish alleles in groups where it must compete. The paradox entirely disappears; there remains only to look at the detailed mechanisms involved.

When we do that, Sober and Wilson make a compelling case that in most cases we simply can'ttell whether the psychological mechanism involved is ultimately selfish or not. The best they can claim is that an altruistic motive to help one's offspring would be one way, but then so would a mechanism that merely linked helping one's offspring to a pleasure centre. They argue, plausibly enough, that the former is slightly more likely than the latter, being more direct and therefore possibly more efficient in a pinch. But this is hardly revolutionary: just one more twist in the rich web of possibilities allowed by Neo-Darwinism.

I conclude with one last irony in the context of the great struggle over Human Nature of the last few decades, between the biophiles who favour Sociobiology and gene selection, and the biophobic social scientists. The social science lobby, who typically have defended the role of group selection, turns up its nose at the slightest whiff of genetic determinism, which they wrongly insist is inseparable from sociobiology; they proclaim the essential dignity and equality of all culture, its power to mould human nature to an indefinite extent into a thousand different forms. But if we are to believe the preliminary conclusions that Sober and Wilson derive from a brief sampling of Murdock's (1967) Ethnographic Atlas, the extent to which cultures can do this on the basis of group selection depends not just on a high degree of isolation, but on a rather nasty form of cultural determinism (Chapter 5). An implacable system of "secondary behaviors" ensures conformity to the most arbitrary and cruel rules, and can do so with impunity from the sanctions of natural selection. The reason is quite simply that bullying your neighbours into submission can be achieved "at trivial cost". (166). The theoretical possibility that altruism can evolve by group selection is just part of a larger truth, which is that "group selection can favor any behavior, depending on the social norm of the group. Between-group selection favors only social norms that lead to functionally adaptive groups." (152). But in this context "functionally adaptive" might just mean "able to beat the others into submission." Group selection, we may infer, is a recipe for Sparta's victory over Athens, or more generally for the triumph of tyrannical militarist cultures over easy-going liberal ones. The tyranny of the group reigns everywhere but in those Western-influenced places that the champions of culture deplore for their "ethnocentric" beliefs in the values of freedom and democracy. The wonder is that any of the latter kind subsist at all; but Sober and Wilson encourage the surmise that this may be a function of size, since "norms may become increasingly difficult to enforce as societies become larger" (180). In which case, we can probably expect that cultural globalization will stop group selection dead in its tracks forever. If Sober and Wilson are right in their claim quoted above, that group selection is required for altruism to evolve, this may mean that it will be downhill all the way from now on for our altruistic genes.

Despite these gloomy speculations, I see no theoretical objections to any of the mechanisms described by Sober and Wilson's book. Indeed, I strongly doubt if the most ardent gene selectionist would object to Sober and Wilson's version of pluralism. Still, I'm ready for one more swing of the pendulum, in which someone will explain to us the moral superiority of those groups where gene-selection reigns supreme.


Thanks are due to Paul Thompson, and to the members of the seminar which he and I jointly taught in the Winter of 1999, for many hours of fruitful discussion of Sober and Wilson's book.

This thought was expressed by Freud with startling clarity:

"Biology teaches that ... two views, seemingly equally well-founded, may be taken of the relation between the ego and sexuality. On one view, the individual is the principal thing, sexuality is one of its activities, and satisfaction one of its needs; while on the other view the individual is a temporary and transient appendage to the quasi-immortal germ plasm, which is entrusted to him by the process of generation." (Freud 1915)
Back to where you left off.

The questionof the nature of function or teleology in biology has recently sprung a vast literature. An excellent sampling can be found in Allen, Bekoff and Lauder (1998).
Back to where you left off.

For the pinnacle of the genre,see various contributions to the New York Review of Books by Stephen J. Gould, especially Gould (1992).
Back to where you left off.

Note that the meaningof the phrase "fitness of the group" is not self-explanatory. Is a fit group one that reproduces more groups like itself than a less fit group? Or it is a group the individuals in which leave more offspring on the average? The notion of fitness seems to workdifferently at different levels:

(1) A fit gene is one that leaves more replicas of itself than a less fit.

(2) A fit individual can't leave more replicas of itself, since there are none; a fit individual, rather, is one that leaves more offspring, but that translates into leaving more replicas of his genes, on average, than other individuals.

(3) A fit group can't be a group that leaves more replicas of itself. It must be one in which more individuals descend from the individuals that are members of the group than from members of other groups. So it's a group, among which the average fitness of the individuals is greater than the population average. But by (2) that in turn is defined in terms of genes.

Back to where you left off.


Allen, C., M. Bekoff, and G. Lauder, eds. 1998. Nature's Purposes: Analyses of Function and Design in Biology. Cambridge, MA: MIT Press .

Darwin, C. 1958. The Origin of Species: By Means of Natural Selection of the Preservation of Favoured Races in the Struggle for Life. Introd. by S. J. Huxley. New York: New American Library .

Dawkins, R. 1976. The Selfish Gene. Oxford : Oxford University Press.

Freud, S. 1915. "Instincts and their vicissitudes." In Standard Edition of the Complete Psychological Works, vol. 14, ed. J. Strachey, 117-40. London: Hogarth Press.

Gould, S. J. 1992. New York Review of Books 39 (19 November)"The Confusion Over Evolution. Review of Helena Cronin, The ant and the peacock".

Hamilton, W. 1963. "The evolution of altruistic behavior." American Naturalist 97:354-56.

Lorenz, K. 1965. Evolution and the Modification of Behavior. Chicago: Chicago University Press.

Mellor, D. 1991. "Laws, chances, and properties." D. Mellor. In Matters of Metaphysics, 154-69. Cambridge; New York: Cambridge University Press.

Murdock, G. 1967. Ethnographic Atlas. Pittsburgh: University of Pittsburgh Press.

Sober, E., and R. C. Lewontin. 1982. "Artifact, cause, and genic selection." Philosophy of Science 49:157-80.

Sterelny, K., and P. Kitcher. 1988. "The return of the gene." Journal of Philosophy 85:339-61. Reprinted in David Hull and Michael Ruse, eds., The Philosophy of Biology. Oxford and New York: Oxford University Press, 1998. Citations are to this reprint.

Williams, G. C. 1966. Adaptation and Natural Selection. Princeton, NJ: Princeton University Press.

Wynne-Edwards, V. 1962. Animal Dispersion in Relation to Social Behavior. Edinburgh: Oliver and Boyd.

Ronald de Sousa is Professor of Philosophy at the University of Toronto. His book, The Rationality of Emotion (MIT 1987) was also published in German as Die Rationalitält des Gefülhls. His current research project is on various aspects of Human Individuality.

Go to
Semiotic Review of Books Home Page
Go to SRB Highlights
Go to SRB Archives