book reviews The origin of altruism

Unto Others: The Evolution and about? I think the argument is largely seman- degree of relatedness. So we have a single Psychology of Unselfish Behavior tic, and could not be settled by observation. model, but two ways of analysing it. by Elliott Sober and David Sloan Wilson Two examples will make this clearer. More briefly, here is a second example. First, consider Wilson’s ‘trait group’ Evolutionary game theory was first devel- Harvard University Press: 1998. Pp. 394. $29.95, £19.95 model, first published in 1975 but still very oped to explain the ritualistic nature of ani- John Maynard Smith much part of his thinking. A population is mal fights. This had often been explained8 in divided into trait groups, and selection acts terms of the ‘good of the species’: it seemed It is obvious that the parts of organisms have upon them. The members then disperse and desirable to George Price and myself to specific functions. Since Darwin, this has mate randomly, and their offspring come attempt an explanation in terms of individ- been explained by natural selection, but together again in trait groups. There are two ual selection. Sober and Wilson reconsider what is the target of selection? Should indi- kinds of individual: altruists, who benefit (in this model. They come to exactly the same vidual organisms be thought of as the units terms of fitness) each of their fellow group conclusions that we reached, but argue that it of selection, or must one also consider levels members to a degree b, at a cost to themselves is a case of group selection. Why? Essentially, above or below the individual, such as c; and non-altruists, who do not. If costs and because it is a model of the interaction of two groups of individuals, or genes within indi- benefits combine additively, and groups are individuals, and, for them, two individuals viduals? In particular, how can one account formed randomly, then altruism cannot constitute a group. They ask: “is it really fair for altruistic behaviour, by animals and by evolve. But if altruists tend to associate with to call a pair of individuals a group, especially humans, or supra-individual structures such altruists, and non-altruists with non-altruists, if they interact only briefly... never to meet as termite mounds or human institutions, then altruism can evolve. This conclusion is again?” They conclude that it is. In effect, which have features ensuring their survival? agreed. Sober and Wilson interpret this result they say that any situation in which fitness is These are the questions discussed in this as arising from a conflict between within- determined by interactions between individ- book. Readers unfamiliar with evolutionary group selection (favouring non-altruists) and uals is a case of group selection. biology should be warned that these questions between-group selection (favouring altruists, Does it matter what words we use to have been the topic of long and occasionally if assortment is non-random). But it can also describe a model if we agree about its conse- acrimonious debates, in which the authors be seen as an example of kin selection, an idea quences? Perhaps it does. We need formal have been active participants. Readers of this developed by William Hamilton in 1963: the models, but we also need intuition about review should also be warned that I have also values of b, c and degree of assortment why the models give the results they do, and been a participant, often on the other side. required for altruism follow at once from his the words used guide our intuitions and The first half of the book discusses the famous inequality, b > rc, where r is the tell us what to look for. A group selection role of population structure in the evolution of animal behaviour, in particular of altruis- tic behaviour. Although I have been a partici- pant, I find it quite hard to decide what the current debate is about. Is it about what the world is like, or about the best words to use when we describe it? When I first became interested, back in the 1960s, it seemed fairly clear. Although, most of the time, Darwin regarded the individual organism as the tar- get of selection, it was quite common for biologists to speak as if organs or behaviours existed for the good of the species as a whole. This view was expressed with particular clarity in Verne Wynne-Edwards’ 1962 book Animal Dispersion (Oliver and Boyd), which argued that animals limit their breeding to prevent the population exceeding its food supply. His great merit was to see that, for this to be true, selection must be acting on the population as a whole: species whose mem- bers limit their breeding survive, whereas those that do not become extinct. It is not hard to see that this will not work, and that the observed behaviours — such as territorial behaviour — can be explained in other ways. Elliott Sober and David Sloan Wilson would agree, but they would also argue, I think cor- Floral prints rectly, that if a species is divided into partially isolated groups, each established by a few In 1985, a group of botanists from the University plants. Her work, including Nepenthes pervillei founders, this will affect the traits that evolve. of Oxford studied the flora of the Seychelles. (above), can be found in A Fragile Eden: Portraits So, if the ‘old’ group selection no longer Their artist, Rosemary Wise, spent the next ten of the Endemic Flowering Plants of the Granitic has supporters, and we agree that group years painting these unusual and threatened Seychelles (Princeton University Press, $75). structure is important, what are we arguing

NATURE | VOL 393 | 18 JUNE 1998 639 Nature © Macmillan Publishers Ltd 1998 book reviews terminology leads us to look for factors caus- sion the authors think that more than hedonistic and altruistic motives for helping ing a difference between variation within and semantics is at issue. others in the first place. between groups; a kin selection model leads Finally, and perhaps most important, they This book should carry a health warning. us to look for relatedness; and a game theory seem to confuse semantic and empirical Read critically, it will stimulate thought model leads us to look for frequency depen- issues. There are important empirical issues. about important questions. Swallowed dence and non-additive fitness interactions. For example, fascinating experiments by whole, its effects would be disastrous. If this is right, the argument is not about Michael Wade, Charles Goodnight and oth- John Maynard Smith is in the School of Biological what the world is like, or even about how we ers show that selection between groups (and Sciences, University of Sussex, Falmer, Brighton should model it (that is, what simplifying even between two-species communities) can BN1 9QG, UK. assumptions are adequate to explain it), but be more effective than individual selection 8 about what words we should use to explain in producing change. Sober and Wilson New edition our model. Is the trait group model an exam- describe these experiments, although they A second edition of John Maynard Smith’s ple of group selection or kin selection? Is the seem curiously uninterested in the underly- Evolutionary Genetics has recently been game theory model an example of group ing mechanisms. In discussing Wade’s results published by Oxford University Press (£50, selection or individual selection? they say only, in a footnote, “the reasons... are $95, hbk; £19.95, $39.95, pbk). When the For several reasons, I think the authors’ too technical to be treated in this book”. first edition was reviewed in Nature 339, approach to these questions is confusing. There follow two chapters concerned with 107 (1989), R. C. Lewontin wrote: First, they are misleading for historical rea- the evolution of human societies. The empha- “When... a specialized subject is treated by sons. For example, when Price and I pro- sis is on the competition between human an expert... the changes in the intellectual posed our model of animal fighting, we were groups, and the role of social norms in guiding structure, the problems and the methods combating the then prevalent idea, support- the behaviour of individual members of such of the field can be seen. John Maynard ed by no less figures than Julian Huxley and groups. Because social norms can homoge- Smith’s new book is a model example.” Konrad Lorenz, that ritualized fighting nize the behaviour of groups, the result is that behaviour had evolved for the good of the such groups evolve properties ensuring group species. We tried to explain it by individual survival, just as individuals evolve traits selection, not between-species selection. ensuring individual survival. I found these Per ardua ad Essentially, we were arguing about what is an chapters the most rewarding section of the appropriate model of the world. It is there- book. The essential point is that higher-level Stockholm fore confusing to accept our model, but entities (for example, individuals carrying I Wish I’d Made You Angry Earlier: rename it group selection. many genes, or societies comprising many Essays on Science, Scientists, and A second reason why the book is confus- individuals) will evolve characteristics Humanity ing is that, although the authors argue for favouring the success of the group, provided by Max Perutz pluralism, they are not themselves pluralists: there are processes that reduce within-group Cold Spring Harbor Laboratory Press: 1998. for them, the only right way to describe a selection. In human groups, the most impor- Pp. 354. $39 model is in group selection language. Any tant such process is the homogenization of Walter Gratzer attempt to calculate the behaviour of a model behaviour by social norms. by estimating the fitness of individuals is The final part of the book discusses Speaking at a memorial symposium for A. condemned as “fallacious averaging” human altruism and the motives that cause Michelson, Einstein related the following (although Wilson himself carried out just it. Do people help others because they think anecdote: he had, he said, once asked such an averaging process in his original trait they will get pleasure from doing so (hedo- Michelson why, when the velocity of light group model). It is “fallacious” because it nism), or because they have an ultimate was already known with adequate precision, detracts attention from the role of group- desire to help another (true altruism)? Sober he was continuing to measure it with ever level processes (which, for them, include all and Wilson argue that evolutionary biology greater accuracy. “Because,” Michelson had interactions between individuals). The can shed light on this problem. They do not replied, “I get so much fun out of it”. “Das result of this bias is that it gives the impres- say that human traits that evolved by individ- finde ich wunderbar,” was Einstein’s verdict. ual selection are hedonistic, and those that The parallel to Max Perutz’s 60-year Science book prize evolved by group selection are truly altruis- devotion to the red protein, haemoglobin, is tic. Their argument is more subtle than that. not exact of course, for the deeper he has dug Jared Diamond, professor of physiology at They start by reviewing psychological and into its workings, the more rewarding the the University of California Los Angeles School philosophical attacks on the question, and lessons that have emerged, and his work of Medicine, has won the 1998 Rhône-Poulenc decide it has not been decisively answered. stands now as one of the pinnacles of this cen- science book prize for Guns, Germs and Steel: Their approach is to take a specific example, tury’s achievements. The pleasure he has The Fates of Human Societies (Norton/Vintage; parental care, and argue that it increases bio- drawn from his quest shines through this life- for a review see Nature 386, 339; 1997). This logical fitness, and that the most effective affirming selection of his writings, and he makes him the only author to have won the proximate mechanism for generating quotes with approval Noël Coward’s dictum: prize twice (his previous success was The Third parental care is altruism. It is a kind of reverse “Work is fun. There is no fun like work.” Chimpanzee). What’s more, Guns, Germs and engineering argument. Such arguments are The centrepiece of this collection is Steel was also awarded this year’s Pulitzer prize used in biology, but usually as a first step in a Perutz’s account of his experiences during for general non-fiction (see Nature 392, 750; physiological investigation. For example, an the Second World War, when, in the com- 1998). Other shortlisted authors for the £10,000 effective proximate mechanism for bird pany of thousands of desperate refugees Rhône-Poulenc prize, which was announced migration would be a magnetic sense: this from the horrors unfolding in Germany, and last week at the Science Museum in London, has led people to seek direct evidence for of Italian chefs and waiters who had lived were David Deutsch (The Fabric of Reality; such a sense. It is a lot harder to seek direct peaceably in Britain for decades, he was Allen Lane, 1997), Richard Fortey (Life; evidence for mechanisms underlying true incarcerated on the Isle of Man and then Knopf), Ernst Mayr (This Is Biology; Belknap, altruism. I am not really competent to judge deported to Canada. The ageing blimp who 1997) and Simon Singh (Fermat’s Last Theorem; this final part of the book; indeed, I am not commanded the camp was heard to remark Walker, 1997). sure I understood the distinction between that he had had no idea so many of these

640 NATURE | VOL 393 | 18 JUNE 1998 Nature © Macmillan Publishers Ltd 1998