SOCIAL GROUPS ALL THE WAY DOWN Review of Darwinian Dynamics, by Richard E. Michod Princeton University Press 1999 262pp $45.00

Natural selection increases the fitness of individuals--what could be simpler? Alas, individuality and fitness, two of the most fundamental concepts in evolutionary theory, are anything but simple. Theoretical biologist Richard Michod reviews and extends the state of the art for these subjects in Darwinian Dynamics. His inquiry also includes other large and fundamental subjects, such as multilevel selection, the origin of life, the major transitions of life, and the evolution of sexual reproduction, which turn out to be inseparable from the topics of individuality and fitness.

It is now an established biological fact that the individuals of today--you and me, for example--are communities of subelements that originally led a more autonomous existence. Individuals are social groups whose members have become so functionally integrated that the whole becomes more conspicuous than the parts. It is likely that life has been social from the very beginning, starting with the groups of molecular reactions that constituted the origin of life.

Social life has a problem. It derives its benefits from coordinated action and shared benefits, which are vulnerable to cheating. This is a famous problem in evolutionary biology that has been studied mostly with respect to individuals interacting in social groups. If individuals are themselves social groups, however, the theoretical tools that have been developed to solve the problem become more widely applicable. As Robert Trivers once remarked (personal communication), sociobiologists have always known that they needed to learn genetics, but who would have thought twenty years ago that geneticists would need to learn sociobiology? Or, as Michod puts it (p30), "cooperation (and altruism) and defection are not just special problems in the study of animal behavior, but rather a central issue in transitions to increased complexity in evolution."

The evolution of individuals requires a solution to the cheating problem. Prior to the solution, groups are not individuals but dysfunctional collections of lower-level individuals, who do not provide public goods and succeed primarily at the expense of each other. Or, groups are mixtures of cooperators who supply public goods and cheaters who exploit them, in which case it is difficult to assign individuality to either level of the hierarchy. A transition occurs when mechanisms evolve that suppress fitness differences within groups, causing the group to survive and reproduce as a unit.

This view of life has a cascade of implications that extends to the far corners of biology and even to some nonbiological physical systems. Features of genetics and development, so basic that they are seldom questioned, can be interpreted as mechanisms that prevent individuals from becoming evolving populations of their elements. Examples include the existence of chromosomes, the rules of meiosis, a single-cell stage of the life cycle, the separation of cell lineages into germ and somatic lines, the total number of cell divisions in a life cycle and programmed cell death (apoptosis). One major implication is that the transition is never complete. Even the best integrated individuals are in danger of rogue elements that succeed at the expense of their group and which we classify as certain kinds of diseases.

Viewing individuals as social groups also has an important implication for the traditional study of individuals in social groups. Prior to the 1960's, biologists often compared groups such as fish schools, bird flocks, social insect colonies, human societies and even multispecies communities to single organisms in the harmony and coordination of their parts. The organismic view of social groups was largely rejected in the 1960's because it required a process of group-level selection, which was regarded as too weak to alter the outcome of natural selection within groups (Williams 1966). It is therefore ironic that individuals themselves provide an example of higher-level selection that has prevailed over lower-level selection. Not only is multilevel selection the natural theoretical framework for thinking about the evolution of individuality, but it is almost inconceivable that higher-level selection stops at the level that we currently perceive as individual organisms. If individuals are social groups, then the possibility of social groups as individuals, or at least part way in the direction of individuals, requires a second look. Darwinian Dynamics therefore signals the return to respectability of a theory that was once rejected almost as completely as Lamarkism (see also Bourke and Frank 1995, Dugatkin 1997, Frank 1998, Seeley 1995, Sober and Wilson 1998).

Since individuality is such a fluid concept, it is not surprising that fitness itself becomes difficult to define. In addition, very special assumptions are required to produce a dynamically sufficient mathematical model in which natural selection straightforwardly increases the fitness of individuals, even when they can be precisely defined. Frequency- and density-dependent effects conspire to prevent well-designed organisms from evolving or to maintain a diversity of phenotypes in the population. A valuable feature of Darwinian Dynamics is its thorough review of the evolutionary and philosophical literatures on fitness.

Michod is at the forefront of the material that he covers and does not spare the details. Thus, Darwinian Dynamics is not an easy-to-read book about science for the general reader, but a book of science for the serious reader. Nevertheless, there is an eloquent and philosophical side to Michod that can be appreciated apart from the formal theory. One of his basic messages is that there are no simple answers to some of biology's deepest questions. This message can be better appreciated by reading (however selectively) a book of science that provides the details than a book about science from which the details have been removed.

I agree with most of the material in Darwinian Dynamics with some exceptions. Michod assumes (along with most of his colleagues) that life begins with molecules such as RNA, that replicate with relatively high fidelity and form into cooperative groups. Alternatively, it is possible that high-fidelity replication was a product of natural selection rather than an initial condition. Consider a complex mix of chemicals that comes to an equilibrium with each chemical at a characteristic frequency. The equilibrium is dynamic in the sense that every chemical is dissociated into other forms at a certain rate but is reconstituted from other forms at the same rate. Thus, each chemical is replicated in a diffuse sense but this kind of replication occurs, by definition, whenever a mix of chemicals exists at equilibrium. Now imagine that the mix of chemicals is subdivided into a large number of subunits, such as around clay particles or in droplets of water (discussed by Michod on p 39). If the chemical interactions are sufficiently complex, variation in the initial composition of the groups could lead to different equilibria with different properties. A distribution of equilibria would exist for the population of subunits that would be relatively stable across time. My point is that a population of units with heritable variation in their phenotypic properties can exist without molecules that replicate with high fidelity. The replication implicit in any chemical equilibrium may be sufficient.

Michod may also have underestimated the importance of random variation in the evolution of individuality. In standard population genetics models, the random distribution of genes into individuals provides sufficient variation for natural selection to occur. Above-random genetic variation, such as by inbreeding, can increase the rate of natural selection but by no means is thought to be required. In contrast, standard animal behavior models of altruism have been dominated by the theory of kin selection, in which helping others is directly proportional to genetic relatedness. Michod uncritically assumes that individuality is a form of strong altruism that requires genetic relatedness among the members of the primordial groups, but random variation may have been sufficient from the beginning. For example, theories of social behavior are increasingly relying on mechanisms of social control that prevent cheating without themselves being strongly altruistic. The human convention of drawing straws provides an example, which enables groups to perform actions that would be regarded as altruistic if they were performed on a voluntary basis, by randomizing the probability of becoming the "altruist" and enforcing the behavior of the "altruist" by threat of punishment (enforcing the convention is a kind of second-order altruism but one that can require only random variation to evolve; see ch.4 of Sober and Wilson 1998). Genetic relatedness is not required for human groups to draw straws and it may not have been required for primordial groups to evolve similar lottery-like processes. Of course, adding relatedness tips the balance even more in favor of higher-level selection.

Standard animal behavior models of altruism also assume a direct relationship between genes and behavior. Individuals are altruistic because they have altruistic genes, and groups are altruistic in direct proportion to their frequency of altruistic individuals. It follows that phenotypic and genetic variation among groups become tightly linked--the only way to get substantial phenotypic variation among groups is to have substantial genetic variation. In reality, however, genes code for products that interact with each other in a complex fashion to produce individuals, which themselves interact in a complex fashion to produce groups. It is well known that a small genetic change, even an allele substitution at a single locus, can have a large phenotypic effect at the individual level. Experimental studies of group selection (reviewed by Goodnight and Stevens 1997) have shown that minor genetic variation among groups can produce substantial phenotypic variation, precisely because of complex interactions among group members. This is one form of complexity that Michod has yet to incorporate into his models, which again tends to improve the efficacy of higher-level selection (e.g., Boyd and Richerson 1990).

These comments fall well within the paradigm that Michod reviews and extends in Darwinian Dynamics, which reports some of the most exciting advances in modern biology from one of its most able practitioners.

DAVID SLOAN WILSON Department of Biology Binghamton University Binghamton. New York 13902-6000

Reviewed for Human Biology

LITERATURE CITED

Bourke, A., & Franks, N. (1995). Social evolution in ants. Princeton: Princeton University Press.

Boyd, R., & Richerson, P. (1990). Group selection among alternative evolutionarily stable strategies. Journal of Theoretical Biology, 145, 331-342. Dugatkin, L. A. (1997). Cooperation among animals: An evolutionary perspective. Oxford, UK: Oxord University Press.

Frank, S. A. (1998). Foundations of Social Evolution. Princeton, NJ: Princeton University Press.

Goodnight, C. J., & Stevens, L. (1997). Experimental studies of group selection: what they tell us about group selection in nature. American Naturalist 150:S59-S79. Seeley, T. (1995). The wisdom of the hive. Cambridge, Mass.: Harvard University Press.

Sober, E., & Wilson, D. S. (1998). Unto Others: the evolution and psychology of unselfish behavior. Cambridge, MA: Harvard University Press.

Williams, G. C. (1966). Adaptation and Natural Selection: a critique of some current evolutionary thought. Princeton: Princeton University Press.