• Email
  • Single Page
  • Print

Survival of the Nicest?

The hinge of Sober and Wilson’s argument is a rejection of the prejudice that natural selection must operate directly solely on individuals. They point out that groups of organisms may also be the units of differential reproduction and they provide examples from natural history that can only be understood if, in addition to the survivorship of individuals, the survivorship of entire collections of organisms is taken into account. In fact, the effect of selection within groups may be the opposite of the effect between groups, with the group effect dominating the entire evolutionary process.

The story of the rabbit in Australia is a clear example. Rabbits were introduced into Australia in the nineteenth century in order to allow colonial country squires to continue the English tradition of huntin’ and shootin’. Unfortunately the squires’ marksmanship could not keep up with the reproductive rate of the rabbits, who quickly became a serious pest and competitor with sheep. An attempt was made to control the rabbit population by introducing an infectious lethal disease, myxomatosis, which spread rapidly and, at first, decimated the rabbit population. But then the rabbits began to increase again. As was to be expected, rabbits resistant to the disease organism had appeared, presumably by random mutations, and genetic resistance spread through the species by natural selection. In addition, however, the disease organisms had also evolved to be less virulent so that even susceptible rabbits were less damaged by the newly evolved pathogens.

This seems odd, because more virulent strains grow more quickly and should eliminate less virulent strains when competing within rabbits. The trouble with being virulent, however, is that, having taken over a rabbit, the virus kills it. Myxomatosis is spread from rabbit to rabbit by a mosquito and mosquitoes do not bite dead rabbits. So, having taken over the subpopulation of viruses by natural selection within a rabbit, virulent viruses have guaranteed their own eventual elimination by failure of that subpopulation to be transferred to new hosts. Ethnic cleansing has been the pathway to national suicide and the more benign eventually survive globally.

In considering groups as units of selection, it is important not to take too impoverished a view of what constitutes a group. It need not be a spatially defined unit, as in the myxomatosis case, where all the virus particles contained physically within one rabbit form a selected group that then may contribute to new groups in new rabbits. Groups may be delimited by any shared property such as food preference, temporal pattern of activity, gender, social class, or e-mail list. Anything that sorts individual organisms will do, including kinship. Thus Sober and Wilson swallow up kin selection as a special case of their more general theory. For group selection to operate, all that is required is that there be collections of individuals that interact with each other in some way separate from the interactions of other individuals in other groups, and that subsequently the groups contribute differentially to the next generation.7

In fact, it is hard to imagine any real biological species that is not broken up into such “trait groups” according to many different traits, so that a mixture of individual-level selection within collections and group-level selection between them ought to be the rule. The exploitation of plants as a food resource for insects is an example. Female moths make a choice of plants on which to lay eggs, depending on various physical and chemical properties of the plants. There is variation in preference among females within a moth species, partly as a consequence of genetic differences between moths, but also as a result of exposure to different plant chemicals during their own development as caterpillars. All the caterpillars that hatch out on a given plant are a “trait group,” having in common the consumption of a particular plant with particular characteristics. During their lives as caterpillars some will be more efficient or voracious than others, and so survive better to adulthood. Their genetic type will therefore increase within a group.

But this fact alone is not sufficient to predict whether their type will increase in the species as a whole. That increase or decrease depends in part on the commonness or rarity of the plant variety on which they survived and on how choosy they will be as adults when they are ready to lay eggs. A type that survives extremely well in competition on a rare plant will decrease in the species as a whole if its members insist on laying eggs on that rare plant, but may increase if they are more catholic in their tastes for plants. The ultimate pattern of plant consumption for the moth species as a whole is, then, a consequence of a mixture of individual and group selection in which the balance depends in part on how widely the individuals born on one sort of plant will disperse to other sorts of plants to produce the next generation.

Using the trait group approach, Sober and Wilson show quite convincingly that species may evolve altruistic behavior provided that the frequency of altruistic types within groups has an effect on the contribution of the group as a whole to the next generation of the species. If some group has, by chance, a higher frequency of altruistic individuals, and if the consequence is a larger number of offspring for the group as a whole, then even though there is some selection against the altruists within each group, altruism may come to characterize the species. Demographic studies of Native American groups have confirmed that, as might be expected, so-called “war chiefs” who were elected to lead their fellows into battle had smaller numbers of children as a consequence of their greater likelihood of death in combat. Success in war presumably enhanced group survival, however, so the altruistic act of the war chief, sacrificing himself for the group, would nevertheless lead to a survival and spread of the altruistic institution.

The modern ideology of population control predisposes us to think that small families are good for the individual, good for the family, and good for the species as a whole. Large families are harder to support and demand a sharing out of limited domestic resources, and we all know that the world is being ruined by overpopulation. But such an argument confuses the present with the past, owners with peasants, and well-being with numbers. For most of the history of agricultural Europe, land-owning families were well advised to have the largest possible families in order to maximize, by marriage connections, the land and military force on which they could depend. Even today, in rural agricultural India, family prosperity is increased, not decreased, by extra children whose unpaid agricultural labor often means the difference between a net profit or loss to the family.8 Moreover, the question of the increase or decrease of numbers in a population should not be confused with their prosperity. The prosperous few have more than once succumbed to the ill-fed masses.

Obdurate genic selectionists will respond that, irrespective of the mechanical details, all that matters in the end is which genes increase and which decrease. Groups, like individuals, are here today and gone tomorrow. Only the gene remains, so we need to consider only the differential reproduction of different genes, however that may be mediated. By definition, a gene is more fit in evolution if it leaves more copies in the next generation. But this bookkeeping trick confuses causes and effects, or, rather, eliminates material causes by reifying statistical effects. First, it confuses random changes with selective changes. In all countries at all times surnames become more or less common, and even extinct, because of variation in reproductive rates among families that are purely at random with respect to the names themselves. Martin is the most common French surname, whereas Bonaparte is practically nonexistent. This is a consequence of the fact that when surnames were created Martin was a common given name, but also that Martins have left a lot of children and Bonapartes only a few.

We would not want to say, however, that these names in themselves had some causal efficacy in reproduction, whatever the social and biological properties of their carriers may have been. That is, we would not want to ascribe Darwinian fitness to a name. There are many causes for an increase in genes that have nothing to do with the direct physiological consequences of bearing those DNA sequences. Different families leave different numbers of offspring for reasons that are at random with respect to a particular gene, so in a finite population of organisms there will be random changes in the frequency of genes from generation to generation. These random changes eventually cause the complete loss of one form of a gene and its chance replacement by another form.

This process of “neutral evolution” characterizes most changes in DNA during evolution, and its very selective neutrality is the basis for the reconstruction of the past relationships of living species. Sometimes unselected genes are swept rapidly to a high frequency because they happen to be on the same chromosome as a gene that is itself the object of natural selection. To assign fitnesses to genes that increase or decrease by chance or because they hitchhike on the chromosomes of other genes is a tautology that completely obfuscates the actual causal events.

Second, there is a confusion of the level at which causal action is occurring. The effect, at one level, of processes occurring at a different level may give an incorrect picture of what is happening. A famous example was the suspicion during the 1970s that the graduate school at the University of California at Berkeley was discriminating against women, because the overall acceptance rate of male applicants was clearly higher than for females. When each department was looked at separately, however, men and women were being accepted at the same rate at which they applied. The apparent discrepancy arose because women were disproportionately applying to the departments with the lowest overall acceptance rates. That is, they were engaged in riskier behavior than men, so they failed more often when averaged over all groups, although not within any given group.9

A gene with no deleterious effect on its carrier, but which is present in a species that lives only in the rich soil on the sides of slumbering volcanoes, will not decrease in frequency within that species (and may even increase), but its overall prospects are dim. To claim that selection is operating against this gene just because the species within which it occurs lives in a risky environment is an example of what Sober and Wilson call the “Averaging Fallacy.” To call it a “fallacy,” however, misses the point. The issue is not an analytic one as the word “fallacy” implies, but a metaphysical one about causal reality. If one continues to insist that the gene is what “really” matters, or professes a complete lack of interest in material mediation in favor of computing outcomes, then there is no fallacy. Once again the angel of reality seems to have abandoned us.

A large part of Unto Others is taken up with a classic problem in philosophy and psychology that is analogous to the evolutionary question of whether the appearance of altruism at the individual level is really selfishness at the genic level. Is human altruism really egoism, or even pure hedonism, in disguise? Egoism is entirely self-directed. The egoist asks only “Is it good for me?” The answer, of course, may involve whether it is good for others, who may also be egoists, because it is beyond dispute that we may sometimes benefit ourselves by means of benefiting others. The egoist must spend a certain amount of time doing cost-benefit analyses, but with confidence that some bread is worth casting on some waters.

The important point of the claim for egoism is that the welfare of others enters the calculation only instrumentally. Hedonism, on the other hand, is a particular psychologistic and somewhat unreliable variety of egoism in which the actor asks only “Does it feel good?” without making a calculation. If what I was told about cod-liver oil when I was a child is true, then the senses are an unreliable guide to objective benefit, and assuredly villains smile and smile even outside of Denmark, so how we feel about situations may fool us. Still, risky as hedonism may be, our apparent altruism may be only doing what makes us feel warm inside. That is, we always act only to achieve gratification, but we have been brainwashed by our upbringing and social circumstances into being gratified through being nice. Others then reap the benefit of our ultimate selfishness. A refusal to be tempted by the devil may, after all, only be giving in to the ultimate ego satisfaction of achieving sainthood.

Sober and Wilson are too well versed in the history of this ancient problem to claim they can know the truth of the matter. They know they cannot rule out either egoism or hedonism as the “real” source of human altruism, and they recognize these theories as being ideological predispositions that are not capable of unambiguous empirical tests. They are clearly tempted by the psychological experiments that are most easily interpreted as showing the existence of real empathy for others’ circumstances. For example, a group of test subjects is told that a group of their fellow students needs help in going over their schoolwork. Some of these needy students are described in such a way as to make it easier for the subjects to empathize with them while others are described in a more alienating way. In addition half the subjects are offered a mood-enhancing experience (such as pleasant music) whether or not they helped a student, while the other half were not offered such a reward. In the absence of the reward more subjects helped those with whom they were empathic. The other subjects, despite the promise of an unconditional mood-enhancing experience, still went to the trouble of helping those with whom they felt more empathy. So the claim of the experimenters is that real empathy counts. But Sober and Wilson admit that this sort of experiment is rather unconvincing, and they can offer no suggestion for a better one. They confess that their temptation to believe in a real effect of empathy, in addition to any hedonistic motivation, arises out of their own a priori predisposition to believe in irreducible altruism.

In the end, Sober and Wilson are entirely forthright in saying that they have consciously adopted a pluralistic perspective. In their view evolution occurs at many levels of causation, from the gene to the population. There is natural selection of genes, of individuals, and of whole groups, and all of these are going on in the evolution of altruism. In psychology they accept the existence of egoistic, hedonistic, and irreducibly altruistic motivations for apparently altruistic behavior. To the extent that they support attempts to explain phenomena at lower levels of causation, at the genic rather than the organismal, or the organismal rather than the population level, they do so only for strategic reasons. They are methodological reductionists, because to ascribe actions at higher levels without an attempt to explain them at lower levels invites an indiscriminate obscurantist holism that is the enemy of understanding. I share their pluralism, but, of course, that may really be because it makes me feel good.


Higher Superstition’: An Exchange December 3, 1998

  1. 7

    Even e-mail groups enlarge or diminish over time as a consequence of the attractiveness of their subject matter and the stimulus offered outsiders by the content of their interchanges. This occurs even though the least thoughtful and interesting participants often produce the longest and most frequent messages.

  2. 8

    For a detailed economic analysis of why family planning programs are rejected by Indian small farmers, see Mahmoud Mamdani, The Myth of Population Control (Monthly Review Press, 1973).

  3. 9

    Sober and Wilson’s numerical example is as follows. Suppose that 90 women and 10 men apply to a department with only a 30 percent acceptance rate (Philosophy?). In an unbiased process 27 women and 3 men succeed. In a second department 10 women and 90 men apply but this department accepts 60 percent of its applicants (English?) without bias so 6 women and 54 men succeed. On the average over the whole collection, 100 men and 100 women have applied, they have been accepted without bias, yet only 33 women as compared to 57 men were accepted. Indeed the effect could be seen even if there were a bias in favor of women within departments. Suppose each department accepted 2 more women and 2 fewer men. There would then still be only 37 women and 53 men accepted. This is an example of what is known as “Simpson’s Paradox,” arising from the fact that the probability of an average is not the same as the average of the probabilities.

  • Email
  • Single Page
  • Print