David Sloan Wilson

Department of Biological Sciences
State University of New York at Binghamton
Binghamton New York 13902-6000

Elliott Sober
Department of Philosophy
University of Wisconsin
5185 Helen C. White Hall
600 North Park Street
Madison Wisconsin 53706


culture; evolution; group selection; kin selection; inclusive fitness; natural selection; reciprocity; social organization; units of selection.


In both biology and the human sciences, social groups are sometimes treated as adaptive units whose organization cannot be reduced to individual interactions. This group-level view is opposed by a more individualistic view that treats social organization as a byproduct of self-interest. According to biologists, group-level adaptations can evolve only by a process of natural selection at the group level. During the 1960's and 70's most biologists rejected group selection as an important evolutionary force but a positive literature began to grow during the 70's and is rapidly expanding today. We review this recent literature and its implications for human evolutionary biology. We show that the rejection of group selection was based on a misplaced emphasis on genes as "replicators" which is in fact irrelevant to the question of whether groups can be like individuals in their functional organization. The fundamental question is whether social groups and other higher-level entities can be "vehicles" of selection. When this elementary fact is recognized, group selection emerges as an important force in nature and ostensible alternatives, such as kin selection and reciprocity, reappear as special cases of group selection. The result is a unified theory of natural selection that operates on a nested hierarchy of units.

The vehicle-based theory makes it clear that group selection is an important force to consider in human evolution. Humans can facultatively span the full range from self-interested individuals to "organs" of group-level "organisms." Human behavior not only reflects the balance between levels of selection but it can also alter the balance through the construction of social structures that have the effect of reducing fitness differences within groups, concentrating natural selection (and functional organization) at the group level. These social structures and the cognitive abilities that produce them allow group selection to be important even among large groups of unrelated individuals. 

The existence of egoistic forces in animal life has long been recognized. It is not so well known that the idea of group-centered forces in animal life also has a respectable history. (Allee 1943, p 519)

It is a crude oversimplification to conceive of social motives as being capable of direct derivation from a hedonic algebra of self-interest--real or fictitious--based on a few universal human drives, whatever the choice of the drives may be. (Tajfel 1981, p36)

These quotations illustrate a perspective in which social groups have a primacy that cannot be reduced to individual interactions. The group-level perspective can be found in biology and all branches of the human behavioral sciences (e.g., Anthropology, Economics, Psychology, Sociology). It is opposed by another perspective that treats individuals as primary and social groups as mere consequences of individual interactions. Although the conflict between the two perspectives is often dismissed as semantic, it refuses to go away, suggesting that substantive issues are involved.

In biology, the conflict between the perspectives has had a remarkable history. Prior to 1960 it was quite acceptable to think of social groups and even whole ecosystems as highly adapted units, similar to individuals in the harmony and coordination of their parts1. Williams (1966) and others argued that group-level adaptations require a process of natural selection at the group level and that this process, though theoretically possible, was unlikely to be important in nature. Their verdict quickly became the majority view and was celebrated as a major scientific advance, similar to the rejection of Lamarkianism. A generation of graduate students learned about group selection as an example of how not to think and it became almost mandatory for the authors of journal articles to assure their readers that group selection was not being invoked. Nevertheless, a positive literature began to grow in the 1970's and is rapidly expanding today (table 1).2 It is no longer heretical for biologists to think of natural selection as a hierarchical process that often operates at the group level.

The most recent developments in biology have not yet reached the human behavioral sciences, which still know group selection primarily as the bogey man of the 60's and 70's. The purpose of this paper is to re- introduce group selection to the human behavioral sciences. We think that group selection can provide a firm foundation for a group-level perspective in the human sciences, as it has in biology. Before beginning, however, it is important to note a complication. Evolutionary approaches to human behavior have become increasingly common, as readers of Behavioral and Brain Sciences well know. Unfortunately, some of the most prominent evolutionary biologists interested in human behavior have themselves failed to incorporate the recent literature and still present group selection as a bogey man (e.g., Alexander 1979,1987, Daly and Wilson 1988, Trivers 1985; but see Table 1 entries marked 'H' for exceptions). We therefore must re-introduce group selection to human sociobiology as well as to the more traditional branches of the human sciences.


The adaptationist program. In an influential paper, Gould and Lewontin (1979) criticize evolutionists for using adaptation as their only explanatory principle, to the exclusion of other factors such as genetic drift and genetic/developmental constraints. They coined the term "adaptationist program" as a pejorative and their conclusion that it cannot be the only tool in the evolutionist's toolkit is well taken. At the same time, their message should not obscure the fact that the adaptationist program, or "natural selection thinking" (Charnov 1982), is an extremely powerful tool for predicting the properties of organisms.

One of the virtues of the adaptationist program is that it can be employed with minimal knowledge of the physiological,biochemical and genetic processes that make up the organisms under examination. For example, imagine studying the evolutionary effects of predation on snails, seeds and beetles. Suppose you discover that for all three groups, species exposed to heavy predation have harder and thicker exteriors than species not so exposed. The property ohard exterior@ can be predicted from knowledge of the selection pressures operating on the populations. Since the exteriors of snails, beetles, and seeds are made of completely different materials, there is a sense in which these materials are irrelevant to the prediction (Campbell 1974, Wilson 1988). That is why Darwin was able to achieve his fundamental insights in almost total ignorance of the mechanistic processes that make up organisms. Adaptationist explanations have the power to unify phenomena that are physiologically, biochemically and genetically quite different .

The adaptationist program is valuable even if its predictions turn out to be untrue. If we know the traits that organisms will have if natural selection is the only influence on evolutionary trajectories, then deviations from these traits constitute evidence that factors other than natural selection have played a significant role. To discover whether adaptationism is true or false, optimality models are indispensable (Sober 1993, Orzack and Sober in press).

Although the following discussion is, in effect, a view about how the adaptationist program should be pursued, it involves no substantive committment to the success of that program. Regardless of the scopes and limits of adaptationism, the question owhat would organisms be like if they were well adapted@ is of paramount importance in evolutionary biology.

The adaptationist program and the biological hierarchy. The question "What would they be like if they were well adapted?" is more complicated than it sounds. To see this, consider an imaginary population of rabbits inhabiting an island. A mutant arises that grazes more efficiently--so efficiently that a population of such mutants will overexploit their resource and go extinct. The mutation is adaptive in the limited sense of causing its bearer to have more offspring than other rabbits, but maladaptive in the larger sense of driving the population extinct.

This example should sound familiar to human behavioral scientists because it resembles the social dilemmas that abound in human life. It corresponds to the tragedy of the commons popularized by Hardin (1968), the voting problem of economics (Margolis 1982) and the prisoner's dilemma of game theory (Rapoport and Chammah 1965). For humans and nonhumans alike, individual striving can lead to social chaos.

As previously mentioned, many biologists prior to the 1960's uncritically assumed that natural selection evolves adaptations at upper levels of the biological hierarchy. In our imaginary example they would assume that the population of rabbits evolves to manage its resources. The possibility that adaptation at one level of the hierarchy can be maladaptive at another level was either ignored or assumed to be resolved in favor of the higher level. These sentiments, which today are called "naive group selectionism", permeated the textbooks and were espoused by many eminent biologists, including Alfred Emerson (1960), who believed that all of nature was as functionally integrated as a termite colony. As a young post-doctoral associate at the University of Chicago, G.C. Williams attended a lecture by Emerson and left muttering "Something must be done...". The result was a modern classic, Adaptation and Natural Selection (Williams 1966)3.

Williams' argument against higher-level adaptations came in three parts. First, he claimed that adaptation at any level of the biological hierarchy requires a process of natural selection operating at that level. Returning to our population of rabbits, it is easy to see that efficient grazers will evolve because they have more offspring than inefficient grazers. The negative consequences at the population level are irrelevant. However, if we imagine an archipelago of islands, only some of which contain the mutant strain, then populations driven extinct by the mutant can be replaced by other populations without the mutant. The population- level adaptation can now persist, but only because we have added a process of natural selection at that level; fit populations replace unfit populations in the same sense that fit rabbits replace unfit rabbits within populations. This is what evolutionary biologists term group selection.

Second, Williams argued that group selection is unimportant in nature despite the fact that it is theoretically possible:

It is universally conceded by those who have seriously concerned themselves with this problem that such group-related adaptations must be attributed to the natural selection of alternative groups of individuals and that the natural selection of alternative alleles within populations will be opposed to this development. I am in entire agreement with the reasoning behind this conclusion. Only by a theory of between-group selection could we achieve a scientific explanation of group-related adaptations. However, I would question one of the premises on which the reasoning is based. Chapters 5 to 8 will be primarily a defence of the thesis that group-related adaptations do not, in fact exist. (Williams 1966 p 92 ) Part of Williams' skepticism can be illustrated with our rabbit example. If migration occurs between islands, what is to prevent the mutant from "infecting" the other islands before the original population goes extinct? Or perhaps the mutant population doesn't go extinct but merely hobbles along in a malnourished state, in which case the occasional migrant from other islands would be unable to survive. At least for this example, it seems that the parameters of the model must be very finely tuned for group-level selection to prevail against individual-level selection.

Third, Williams developed a concept of the gene as the "fundamental unit of selection" that has become a major theme in evolutionary biology, especially as amplified and extended by Dawkins (1976,1982). Williams claimed that groups and even individuals cannot be units of selection because they are ephemeral and do not replicate with sufficient fidelity. Every sexually reproducing organism is a unique combination of thousands of genes that will never exist again, no matter how successful reproductively. At the individual level, only clonal organisms replicate with sufficient fidelity to qualify as units of selection. For sexually reproducing organisms, the gene is the unit that is transmitted through time with high fidelity and is therefore the fundamental unit of selection (the replicator, in Dawkins' terminology). This is frequently used as an argument against group selection. For example, Alexander (1979 p36) states:

In 1966 Williams published a book criticizing what he called "some current evolutionary thought" and chastised biologists for invoking selection uncritically at whatever level seemed convenient. WilliamsY book was the first truly general argument that selection is hardly ever effective on anything but the heritable genetic units of "genetic replicators" (Dawkins 1977) contained in the genotypes of individuals.

Individuals and groups appear in Williams' scheme, not as units of selection, but as environments of the genes. As the simplest example, consider two alleles (A,a) at a single diploid locus in a randomly mating population, yielding the familiar three genotypes (AA,Aa,aa) in Hardy- Weinberg proportions. Suppose the fitnesses of the three genotypes are WAA=1, WAa=0.75 and Waa=0.5. From the gene's-eye view, the A-allele can be said to inhabit two "genotypic environments", AA and Aa, and its average fitness can be easily calculated:

WA= pWAA + (1-p)WAa (1)

The term p, in addition to being the frequency of the A-allele in the population, is also the proportion of A-alleles that exist in the AA "environment" in a randomly mating population. The fitness of the a-allele can similarly be averaged across its two genotypic environments (Aa,aa) to yield

Wa= pWAa + (1-p)Waa (2)

The A-allele will evolve whenever WA>Wa, which is always the case when WAA>WAa>Waa. Note that A and a have the same fitness within the one genotypic environment that they inhabit together (the heterozygote). It is only by averaging across genotypic environments that differences in the fitness of A and a occur. Biologically informed readers will recognize WA and Wa as the "average effects" of the two alleles used to calculate breeding values and narrow-sense heritability at the individual level (e.g., Falconer 1982, Wilson and Sober 1989).

More complicated examples can be constructed in which the population is divided into social groups that differ in allele frequencies and genotypic fitnesses. In these cases the genes inhabit a more complicated array of "environments" but in principle it is always possible to calculate gene-level fitness by averaging across genotypic and social contexts. In addition, it will always be the case that A replaces a when WA>Wa. This is why Williams (1986,1992) refers to genes as "bookkeeping" devices that automatically record the net effect of multiple selection pressures.

Williams' case against group selection was strengthened by two other theories in evolutionary biology that were developed during the 60's and 70's. The first was inclusive fitness theory (also called kin selection; Hamilton 1964, Maynard Smith 1964), which explained how altruism could evolve among genetic relatives. The second was evolutionary game theory (Axelrod and Hamilton 1981, Maynard Smith 1982, Trivers 1971, Williams 1966), which explained how cooperation could evolve among non-relatives. These theories seemed to account for many of the phenomena that group selection had been invoked to explain. With the problems raised by Williams and two robust alternatives, the theory of group selection, never well articulated to begin with, collapsed.

Not all evolutionary biologists are familiar with the details of Williams' arguments against group selection, but the bottom-line conclusion has been adopted with such conviction that we will call it Williams' first commandment: "Thou shalt not apply the adaptationist program above the level of the individual." All adaptations must be explained in terms of the relative fitness of individuals within populations. Individual-level adaptations may have positive or negative effects at the group level, but in both cases the group-level effects are irrelevant to evolutionary change. Williams' first commandment was repeated like a mantra throughout the 60's and 70's, as every evolutionary biologist knows. Unfortunately, the mantra still echoes through the numerous accounts of evolutionary theory that are written for the human sciences and popular audiences today (e.g., Alexander 1987, Archer 1991, Cronin 1991, Daly and Wilson 1988, Frank 1988, Krebs 1987, MacDonald 1988, Noonan 1987, Sagan and Druyan 1992).

Examining the edifice. Although Williams' and Dawkins' gene-centered view has enjoyed enormous popularity, it has one flaw that should be obvious, at least in retrospect. Naive group selectionists thought that upper levels of the biological hierarchy were like individual organisms in the coordination and harmony of their parts. According to Williams and Dawkins, however, even sexually reproducing organisms do not qualify as units of selection because they, like groups, are too ephemeral. If a creature such as a bird or a butterfly is not a unit of selection, then what endows it with the internal harmony implied by the word "organism"?

To answer this question, an entirely different concept needed to be invoked which Dawkins (1976) called "vehicles of selection" ("interactors" in Hull's 1980 terminology). Employing one of DawkinsY own metaphors, we can say that genes in an individual are like members of a rowing crew competing with other crews in a race. The only way to win the race is to cooperate fully with the other crew members. Similarly, genes are "trapped" in the same individual with other genes and usually can replicate only by causing the entire collective to survive and reproduce. It is this property of shared fate that causes "selfish genes" to coalesce into individual organisms.

So far, so good, but if individuals can be vehicles of selection, what about groups? After all, we are interested in comparing groups with individuals, not with genes. Yet gene-centered theorists have scarcely addressed this question.4 The situation is so extraordinary that historians of science should study it in detail: A giant edifice is built on the foundation of genes as replicators, and therefore the "fundamental" unit of selection, which seems to obliterate the concept of groups as organisms. In truth, however, the replicator concept cannot even account for the organismic properties of individuals. Almost as an afterthought, the vehicle concept is tacked onto the edifice to reflect the harmonious organization of individuals but it is not extended to the level of groups. The entire edifice therefore fails to address the question that it originally seemed to answer so conclusively and that made it seem so important.

This is such a crucial and unappreciated point that we want to reinforce it by quoting from The Ant and the Peacock (Cronin 1991), one of the most recent book-length treatments of evolution for a popular audience.5 Cronin is a philosopher who has a part-time appointment at Oxford University's Zoology Department. Her book was chosen as one of the year's best by the New York Times and has been cited with approval by authorities such as G.C. Williams (1993) and John Maynard Smith (1992) and Daniel Dennett (1992) 6. There is every reason for the reader to think that it represents state-of-the-art evolutionary biology.

Cronin agrees with us that naive group selectionists compared groups to individuals:

Many an ecologist, equipped with no more than a flimsy analogy, marched cheerfully from the familiar Darwinian territory of individual organisms into a world of populations and groups. Populations were treated as individuals that just happened to be a notch or two up in the hierarchy of life...(p278).

Her treatment of Williams is also close to our own: "Williams retaliated with two types of argument. He spelled out why genes are suitable candidates for units of selection whereas organisms, groups and so on are not...(p286)." Here Cronin commits (along with Williams) the fallacy that we outlined above. If individuals and groups are not replicators, then the replicator concept cannot be used to argue that they are different from each other! Faced with this dilemma, Cronin dutifully invokes vehicles to explain the organismal properties of individuals, with a nod to groups:

If organisms are not replicators, what are they? The answer is that they are vehicles of replicators...Groups, too, are vehicles, but far less distinct, less unified...In this weak sense, then, 'group selection' could occur...But even if they [group-level adaptations] did arise--which as we've seen is unlikely--they would in no way undermine the status of genes as the only units of replicator selection. This does not mean that higher level entities are unimportant in evolution. They are important, but in a different way: as vehicles (p289).

But this is all that naive group selectionists ever claimed--that groups are like individuals by virtue of the adaptive coordination of their parts! Finally, Cronin concludes that group selection is unimportant even in the so-called weak sense:

But group selectionism (weak group selectionism) makes claims about adaptations, about characteristics that satisfy the fragmented purposes of all the genes in the group and, what's more, confer an advantage on that group over other groups. Group-level adaptations, then, are a very special case of emergent properties--so special that it would be rash to expect them to have played any significant role in evolution. Of course, the question of what role they have actually played is an empirical, not a conceptual issue. It is a factual matter about which adaptations happen to have arisen at levels higher than organisms, about the extent to which groups and other higher-level vehicles happen to have been roadworthy. [p290]

Cronin is in the unhappy position of a circus artist who stands on the backs of two horses, replicators and vehicles, as they gallop around the ring. The only way that she can perform this dazzling feat is by making the horses gallop in parallel. Thus, groups must fail not only as replicators but as vehicles. What Cronin cannot bring herself to say is that the replicator concept that forms the inspiration for her book is totally irrelevant to the question that is and always was at the heart of the group selection controversy--can groups be like individuals in the harmony and coordination of their parts? To answer this question we must restructure the entire edifice around the concept of vehicles, not replicators. That is exactly what the positive literature on group selection does7.

Taking vehicles seriously. The essence of the vehicle concept is shared fate, exemplified by the adage (and by Dawkins' rowing crew metaphor) "we're all in the same boat." Our restructured edifice must first be able to identify the vehicle(s) of selection in any particular biological or human situation.

In figure 1, the biological hierarchy is shown as a nested series of units, each of which is a population of lower level units. An individual can be regarded as a population of genes and a group is obviously a population of individuals. A metapopulation is a population of groups. For example, a single field might contain hundreds of ant colonies. Each colony certainly deserves to be called a group and yet we must also recognize the collection of groups as an important entity. The hierarchy has been left open on both ends because genes are composed of subunits and metapopulations can exist in higher-order metapopulations, a fact that will become important later.

Vehicles of selection can be identified on a trait-by-trait basis by the following simple procedure: Starting at the lowest level of the hierarchy8, ask the question "Do genes within a single individual differ in fitness?" If the answer is "no", then they share the same fate and are part of the same vehicle. Proceeding up the hierarchy, ask the question: "Do individuals within a single group differ in fitness?" If the answer is "no" then once again they share the same fate and we must proceed up the hierarchy until we find the level(s) at which units differ in fitness. This is the level (or levels) at which natural selection actually operates, producing the functional organization implicit in the word "organism@.9 Everything below this level will acquire the status of organs and everything above this level will be vulnerable to social dilemmas.10

Already we can make three fundamental points: First, focusing on vehicles makes it obvious that the concept of "organism" is not invariably linked to the "individual" level of the biological hierarchy. To the extent that genes can differ in fitness within single individuals, the genes will become the organisms and the individual will become a dysfunctional collection of genes. To the extent that individuals in the same group are in the same "boat" with respect to fitness, they will evolve into harmonious organs of group-level organization. In short, the organ-organism- population trichotomy can be frame-shifted both up and down the biological hierarchy. Frame-shifts in both directions have been documented and examples will be provided below.

Second, the status of organ vs. organism vs. population must be assigned on a trait-by-trait basis. It is possible for a single creature such as a wasp to be an organ with respect to some traits, an organism with respect to other traits, and a population of organisms with respect to still other traits. This may sound strange but it follows directly from the fact that fitness is a property of traits, not organisms (Sober 1984). For example, in the parasitic wasp Nasonia vitripennis, some males harbor what has been called the oultimate@ selfish gene, because it destroys all the other genes in the male to facilitate its own transmission (Werren 1991,1992). In this case the gene is the vehicle of selection but most other genes in the same species evolve by standard Darwinian selection, in which case the individual is the vehicle of selection.

Third, fitness differences are not always concentrated at one level of the biological hierarchy. Individuals with trait A can be less fit than individuals with trait B within single groups, while groups of individuals with trait A are more fit than groups of individuals with trait B. In these cases we cannot assign the status of organ, organism or population and must settle for some hybrid designation. As one example, Williams (1966) showed that, given certain assumptions, natural selection within single groups favors an even sex ratio while natural selection between groups favors an extreme female-biased sex ratio. He thought that the absence of female-biased sex ratios in nature provided conclusive evidence against group selection. Since then, moderately female-biased sex ratios have been discovered in literally hundreds of species, which reflect an equilibrium between opposing forces of within- and between-group selection (Charnov 1982, Colwell 1981, Frank 1986, Wilson and Colwell 1981).11 As we will show, altruism is another example of a hybrid trait that is selected against at the individual level but favored at the group level.

Now we will document our claim that the organ-organism-population trichotomy can be frame-shifted both up and down the biological hierarchy.

Individuals as dysfunctional populations of genetic elements. Individuals are traditionally viewed as stable entities that (barring mutation) pass the same genes in the same proportions to their offspring that they received from their parents. However, this is not always the case. For example, a diploid individual can be regarded as a population of N=2 alleles at each locus. The rules of meiosis usually dictate that each allele is equally represented in the gametes. Occasionally a mutation arises that "breaks" the rules of meiosis by appearing in greater than 50% of the gametes, a phenomenon known as "meiotic drive" (Crow 1979). These same alleles often decrease the survival of individuals that possess them and can even be lethal in homozygous form. Let us apply our simple procedure to this example to identify the vehicle(s) of selection. Can genes within a single individual differ in fitness? The answer is "yes" because the driving allele exists at a frequency of p=0.5 in heterozygotes and occurs in the gametes of those heterozygotes with a frequency of p>0.5. Natural selection therefore operates at the gene level, favoring the driving allele. Now proceed up the hierarchy. Do individuals within a single population differ in fitness? The answer is again "yes" because individuals with the driving allele suffer higher mortality than individuals without the driving allele. Natural selection therefore operates against the driving allele at the individual level. Both the gene and the individual are vehicles of selection. If gene-level selection is sufficiently strong, the driving allele can evolve despite its negative effects on individuals.

Many other examples of natural selection within individuals could be cited involving chromosomal genes (Dover 1986), cytoplasmic genes (Cosmides and Tooby 1981), and competing cell lineages (Buss 1987). These examples have been received with great fanfare by gene-centered theorists as some sort of confirmation of their theory. But these examples do not confirm the thesis that genes are replicators--all genes are replicators by definition and no documentation is needed. These examples are remarkable because they show that genes can sometimes be vehicles. They seem bizarre and disorienting because they violate our deeply rooted notion that individuals are organisms. They force us to realize that individuals are at least occasionally nothing more than groups of genes, subject to the same social dilemmas as our imaginary population of rabbits.

Why aren't examples of within-individual selection more common? Several authors have speculated that the rules of meiosis and other mechanisms that suppress evolution within individuals are themselves the product of natural selection acting at the individual level. Genes that profit at the expense of other genes within the same individual are metaphorically referred to as "outlaws" (Alexander and Borgia 1978) and the regulatory machinery that evolves to suppress them is referred to as a "parliament" of genes (Leigh 1977). Ironically, most of the authors who employ these metaphors are reluctant to think of real parliaments as regulatory machines that reduce fitness differences within groups, thereby concentrating adaptation at the group level. Gene-centered theorists frame-shift downward with enthusiasm but they are much more reluctant to frame-shift upward.

Groups as organisms. Social insect colonies have been regarded as "superorganisms" for centuries. Sterile castes with division of labor, colony-level thermoregulation and patterns of information processing that transcend single brains all suggest intuitively that colonies are functionally organized units, built out of individual insects. This interpretation was rejected by gene-centered theorists, however, who claimed to explain the social insects without invoking group selection. Their scorn for the earlier view is illustrated by West-Eberhard (1981 p 12; parenthetical comments are hers): "Despite the logical force of arguments against group (or colony) selection, and the invention of tidy explanations for collaboration in individual terms, the supraorganism (colony-level selection) still haunts evolutionary discussions of insect sociality."

Let us apply our simple procedure to locate the vehicle(s) of selection in the social insects. Can genes differ in fitness within individuals? Yes-- the social insects resemble other species in this regard--but the products of selection at this level are unlikely to enhance colony function. Can individuals differ in fitness within single colonies? Yes; as one example, honey bee queens usually mate with more than one male, leading to multiple patrilines among the workers. Many insects can detect genetic similarity using odor cues and it is plausible to expect workers tending future queens to favor members of their own patriline. As with evolution within individuals, however, this kind of palace intrigue is more likely to disrupt colony function than to enhance it (Ratnieks 1988, Ratnieks and Visscher 1989). We therefore must proceed up the hierarchy and ask "Can groups (=colonies) differ in fitness within a metapopulation?"

Unlike our archipelago of rabbits, in which the metapopulation seemed somewhat contrived, the social insects obviously exist as a population of colonies. Consider a mutation that is expressed in honeybee workers and increases the efficiency of the hive, ultimately causing the queen to produce more reproductive offspring. It is obvious that this mutation will spread, not by increasing in frequency within the hive, but by causing hives possessing the mutation to out-produce other hives. Thus, for the majority of traits that improve colony function, the colony is the vehicle of selection and can legitimately be called an organism. Focusing on vehicles, not replicators, as the central concept makes West-Eberhard's statement sound absurd. Notice also that Williams' first argument, that group-level adaptations require a process of natural selection at the group level, is correct. But his empirical claim that group selection is weak and group-level adaptations don't exist is just plain wrong in the case of the eusocial insects--both the process and the product are manifest. The focus on genes as the "fundamental" unit of replication merely distracts from the more relevant framework based on vehicles. Fortunately, most social insect biologists now realize this and once again regard social insect colonies as "group-level vehicles of gene survival" (Seeley 1989), at least to the degree that they evolve by between-colony selection.

Before leaving the social insects it is worth asking a question that we will pose later for humans: What does it mean for a creature such as an ant or a honeybee, itself an organism in some respects, to also be part of a group-level organism? A partial answer is provided by Seeley (1989), whose elegant experiments reveal the mechanisms of colony-level adaptation. A honeybee hive monitors its floral resources over several square miles and maximizes its energy intake with impressive accuracy. If the quality of a food patch is experimentally lowered, the hive responds within minutes by shifting workers away from that patch and toward ones that are more profitable. Yet individual bees visit only one patch and have no frame of comparison. Instead, individuals contribute one link to a chain of events that allows the comparison to be made at the hive level. Bees returning from the low quality patch dance less and themselves are less likely to revisit. With fewer bees returning from the poor resource, bees from better patches are able to unload their nectar faster, which they use as a cue to dance more. Newly recruited bees are therefore directed to the best patches. Adaptive foraging is accomplished by a decentralized process in which individuals are more like neurons than decision-making agents in their own right (Camazine and Sneyd 1991; see Camazine 1991, Deneubourg and Goss 1989, Franks 1989 and Wilson and Holldobler 1988 for other examples of group-level cognition in social insects). The image of a group-level mind composed of relatively mindless individuals is aptly described in D. HofstadterYs (1979) essay ant fugue. We suggest that some aspects of human mentality can also be understood as a form of group- level cognition (see below).

Finding the vehicles in inclusive fitness theory. How was it possible for West-Eberhard and others to think that the social insects could be explained without invoking group selection? Her "tidy" alternative explanation was inclusive fitness theory, which she and almost everyone else regarded as a robust alternative to group selection. However, inclusive fitness theory is a gene-centered framework that does not identify the vehicle(s) of selection. When we rebuild inclusive fitness theory on the foundation of vehicles we discover that it is not an alternative to the idea of group selection at all (Michod 1982, Queller 1991,1992, Uyenoyama and Feldman 1980, Wade 1985, Wilson 1977, 1980). It would be hard to imagine a more important discovery, yet human behavioral scientists are almost totally unaware of it, in part because their evolutionary informants so assiduously ignore it. Even the most recent accounts of evolution for the human sciences treat inclusive fitness and group selection as separate mechanisms (e.g., Alexander 1987, 1989,1992, Archer 1991, Daly and Wilson 1988, Frank 1988, Krebs 1987, MacDonald 1988, Noonan 1987). We will consider one of these treatments in detail because it allows us to make a number of important points throughout the rest of our paper. Here is Frank's (1988 p37-39) depiction of group selection.

Group-selection models are the favored turf of biologists and others who feel that people are genuinely altruistic. Many biologists are skeptical of these models, which reject the central Darwinian assumption that selection occurs at the individual level. In his recent text, for example, Trivers includes a chapter entitled "The group selection fallacy". With thinly veiled contempt, he defines group selection as "the differential reproduction of groups, often imagined to favor traits that are individually disadvantageous but evolve because they benefit the larger group". Group selectionists have attempted to show that genuine altruism, as conventionally defined, is just such a trait... Could altruism have evolved via group selection? For this to have happened, altruistic groups would have had to prosper at the expense of less altruistic groups in the competition for scarce resources. This requirement, by itself, is not problematic. After all, altruism is efficient at the group level (recall that pairs of cooperators in the prisoner's dilemma do better than pairs of defectors), and we can imagine ways that altruistic groups might avoid being taken advantage of by less altruistic groups...

But even if we suppose that the superior performance of the altruistic group enables it to triumph over all other groups, the group selection story still faces a formidable hurdle. The conventional definition again, is that nonaltruistic behavior is advantageous to the individual . Even in an altruistic group, not every individual will be equally altruistic. When individuals differ, there will be selection pressure in favor of the least altruistic members. And as long as these individuals get higher payoffs, they will comprise an ever-larger share of the altruistic group. So even in the event that a purely altruistic group triumphs over all other groups, the logic of selection at the individual level appears to spell ultimate doom for genuinely altruistic behavior. It can triumph only when the extinction rate of groups is comparable to the mortality rate for individuals within them. As [E.O.] Wilson stresses, this condition is rarely if ever met in practice.

Frank's account of group selection is accurate and similar to our own rabbit example. He also accurately depicts the climate of the group selection debate during the 60's and 70's. Now here is Frank's description of inclusive fitness theory (p25-27):

Biologists have made numerous attempts to explain behavior that, on its face, appears self-sacrificing. Many of these make use of William Hamilton's notion of kin selection. According to Hamilton, an individual will often be able to promote its own genetic future by making sacrifices on behalf of others who carry copies of its genes... The kin-selection model fits comfortably within the Darwinian framework, and has clearly established predictive power... Sacrifices made on behalf of kin are an example of what E.O. Wilson calls "'hard core' altruism, a set of responses relatively unaffected by social reward or punishment beyond childhood." Viewed from one perspective, the behavior accounted for by the kin- selection model is not really self-sacrificing behavior at all. When an individual helps a relative, it is merely helping that part of itself that is embodied in the relative's genes...

FrankYs exposition certainly suggests that group selection and kin selection are alternative theories that invoke separate mechanisms. Frank himself regards them as so different that he calls one non-Darwinian and the other Darwinian!12 Now consider the model in figure 2, which rebuilds inclusive fitness theory on the foundation of vehicles (see Michod 1982, Queller 1991,1992, Sober 1993, Uyenoyama and Feldman 1980, Wade 1985, Wilson 1977, 1980 for more formal treatments). A dominant allele (A) codes for a behavior that is expressed only among full siblings. The behavior decreases the fitness of the actor by an amount c and increases the fitness of a single recipient by an amount b. In figure 2, adults of the three genotypes (AA,Aa,aa) combine randomly to form six types of mating pairs (AAxAA, AAxAa, AAxaa, AaxAa, Aaxaa, aaxaa). Each mating pair produces sibling groups with a characteristic proportion of altruists and non-altruists. Thus, the sibling groups derived from AAxAA matings are entirely altruistic, the groups derived from aaxaa matings are entirely non-altruistic and so on. Since the behavior is expressed only among siblings, the progeny of each mated pair is an isolated group as far as the expression of the behavior is concerned. Thus, any model of sibling interactions invokes a metapopulation of sibgroups.

Now let us employ our simple procedure to locate the vehicles of selection. Beginning at the lowest level of the hierarchy, there is no meiotic drive or other forms of selection within individuals in this example. Moving up the hierarchy, do individuals within single sibgroups differ in fitness? Yes, and natural selection at this level operates against the altruists. In all sibgroups that contain both selfish (aa) and altruistic (Aa,AA) phenotypes, the former are fitter--they benefit from the latter's help without sharing the costs. Sibling groups are similar to other groups in this respect. Continuing up the hierarchy, can sibgroups differ in fitness within the metapopulation? Yes, and it is here that we find the evolutionary force that favors altruism. Since every altruist contributes a net fitness increment of b-c to the sibgroup, the fitness of the collective is directly proportional to the number of altruists in the group. Sibgroups with more altruists outproduce sibgroups with fewer altruists.

The degree of altruism that evolves depends on the balance of opposing forces at the group and individual levels. Figure 3 shows why kin groups are more favorable for the evolution of altruism than groups of unrelated individuals. In the latter case, groups of size N are drawn directly from the global population, forming a binomial distribution of local gene frequencies. In the former case, groups of size two (the parents) are drawn from the global population and groups of size N (the siblings) are drawn from their gametes. This two-step sampling procedure increases genetic variation among groups, intensifying natural selection at the group level. Put another way, altruists are segregated from non-altruists more in kin groups than in randomly composed groups. In both cases there are mixed groups, however, and evolution within mixed groups is the same regardless of whether they are composed of siblings or nonrelatives. Notice that this explanation does not invoke the concept of identity by descent, which seems to be the cornerstone of inclusive fitness theory. There is no physical difference between two altruistic genes that are identical by descent and two altruistic genes that are not. The coefficient of relationship is nothing more than an index of above-random genetic variation among groups (e.g, Falconer 1982 ch 3- 5, Queller 1991,1992).

We invite the reader to go back to Frank's account of group selection to confirm that it exactly describes the process of kin selection that is portrayed in figure 2. Dr. Jekyl and Mr. Hyde are the same person. The only discrepancy between Frank's account and figure 2 involves the concept of extinction. Sibling groups don't last for multiple generations and don't necessarily go extinct, but rather dissolve into the larger population when the individuals become adults and have their own offspring. Thus, sibling groups (and social insect colonies) differ somewhat from our population of rabbits and the groups that Frank and Trivers had in mind. But this does not disqualify sibling groups as vehicles of selection. After all, individuals are transient collections of genes that "dissolve" into the gene pool as gametes. The ephemeral nature of groups in figure 2 makes them more similar to individuals, not less.

Frank's account of kin selection appears so different, not because it invokes a different mechanism for the evolution of altruism, but because it utilizes a different accounting procedure for calculating gene frequency change that does not compare the fitnesses of individuals within single groups. The method correctly predicts the degree of altruism that evolves but obscures the internal dynamics of the process. In fact, when the vehicle-centered approach was first published, many biologists who thought they were familiar with inclusive fitness theory found it hard to believe that altruism is actually selected against within kin-groups and evolves only by a process of between-group selection.

The unification of group selection and kin selection has implications for the distinction between ogenuine@ vs. oapparent@ altruism. This in an important distinction in the human behavioral sciences and evolutionary accounts such as FrankYs seem to provide a tidy answer: The altruism that evolves by group selection is "genuine" because it entails real self- sacrifice, while the altruism that evolves by kin selection is only "apparent" because it is just genes promoting copies of themselves in other individuals. The unified theory reveals that this distinction is an artifact of the way that fitness is calculated. Any trait that is selected at the group level can be made to appear ogenuinely@ altruistic by comparing relative fitness within groups, or only oapparently o altruistic by averaging fitness across groups (Wilson 1992, Wilson and Dugatkin 1992). Thus, evolutionary biologists have so far contributed little but confusion to the distinction between genuine and apparent altruism.13 Finding the vehicles in evolutionary game theory. Evolutionary game theory (also called ESS theory for "evolutionarily stable strategy") is similar to economic game theory except that the strategies compete in Darwinian fashion, as opposed to being adopted by rational choice. It was developed to explore the evolution of cooperation and was universally considered to be an individual-level alternative to group selection. For example, Dawkins (1980 p360) states

There is a common misconception that cooperation within a group at a given level of organization must come about through selection between groups...ESS theory provides a more parsimonious alternative.

We will explore the relationship between game theory and group selection with a fanciful example that is based on Dawkins' rowing crew metaphor. A species of cricket has evolved the peculiar habit of scooting about the water on dead leaves in search of its resource (water lily flowers). A leaf can be propelled much better by two crickets than by one so they scoot about in pairs. Initially they were quite awkward but natural selection eventually endowed them with breathtaking morphological and behavioral adaptations for their task. Especially impressive is the coordination of the pair. They take their stations on each side of the leaf and stroke the water with their modified legs in absolute unison, almost as if they are part of a single organism. Coordination is facilitated by one member of the pair, who synchronizes the strokes by chirping at regular intervals. On closer examination it was discovered that the chirps not only coordinate movements but also steer the little craft. A low-pitched chirp causes the chirper to row harder and a high-pitched chirp causes the non- chirper to row harder. The captain (as the chirper came to be called) adjusts its pitch to correct for asymmetries in the shape of the leaf and also to change direction as lily pads hove into view. Either member of the pair can act as captain; the important thing is that there be only one.

The evolution of any particular trait in this example can be examined with a 2-person game theory model. For example, consider two types (A1 and A2) that differ in their ability to synchronize with their partner's movement. If p is the frequency of A1 in the population and if pairing is at random then three types of pairs exist (A1A1, A1A2, A2A2) at frequencies of p2, 2p(1-p) and (1-p)2. Coordination, and therefore fitness, is directly proportional to the number of A1 individuals in the pair, as shown by the payoff matrix in figure 3a. The fitness of the two types, averaged across pairs, is WA1=5p+4(1-p) and WA2=4p+3(1-p).

This is not a very interesting game theory model because it doesn't pose a dilemma. WA1>WA2 for all values of p, making it obvious that A1 will evolve. However, this should not obscure a more fundamental point, that the pair is the vehicle of selection. If we apply our procedure we find no fitness differences between individuals within a pair, in which case A1 can evolve only by causing pairs to succeed relative to other pairs. The fact that the pairs are ephemeral, perhaps lasting only a fraction of an individual's lifetime, is irrelevant. Persistence is a requirement for replicators, not vehicles. Coordination evolves among the individuals for exactly the same reason that it evolves among genes within individuals, because they are "in the same boat" as far as fitness differences are concerned.

More generally, evolutionary game theory deploys a metapopulation model, in which individuals exist within groups that exist within a population of groups. When this elementary fact is recognized, Dawkins' statement quoted above looks just as absurd as West-Eberhard's statement about the social insects. Cooperation evolves by group-level selection in a game theory model as surely as cooperation among genes evolves by individual-level selection in a standard population genetics model.14 In fact the two models are mathematically identical; we can go from one to the other merely by relabelling A1 and A2 as "alleles" rather than as "individuals" and calling the pair a zygote (Hamilton 1971, Holt 1983, Maynard Smith 1987, Wilson 1983, 1989, 1990).

Continuing our example, suppose that a mutant type (A3) arises that rushes onto the lily pad at the moment of arrival, kicking the boat away and setting its hapless partner adrift. If both members of the pair are the A3 type, however, they collide and have a probability of drowning. The pay-off matrix for this situation is shown in figure 3b and the average fitness of the two types is WA1=p(5)+(1-p)(0) and WA3=p(10)+(1-p)(2).

This model is more interesting because it constitutes a social dilemma. A3 evolves despite the fact that it disrupts group-level functional organization. Applying our procedure, we find that the nasty behavior is favored by within-group selection; A3 is more fit than A1 within pairs. Cooperation, as before, is favored by between group-selection; A1A1 and A1A3 pairs are more fit than A3A3 pairs. By renaming the individuals "alleles" and the pairs "zygotes", we have the example of meiotic drive described on page 14.

Continuing our example, suppose that a new mutant (A4) arises that can remember the previous behavior of its partner. It acts honorably toward new partners and thereafter imitates its partner's previous behavior. This is the famous Tit-for-Tat strategy (Axelrod and Hamilton 1981) that can evolve above a threshold frequency, given a sufficient probability of future interactions (fig 3c). Applying our procedure, we find that natural selection still favors A3 over A4 within pairs because A4 loses during the first interaction. A4 reduces but does not eliminate its fitness disadvantage within groups by changing its behavior and it evolves only because groups of A4A4 outperform groups of A4A3 and A3A3.15

Finally, suppose that yet another mutant arises (A5) that grabs hold of its partner with one of its free legs, preventing it from leaping prematurely onto the lily pad. The pay-off matrix for A5 vs. A3 is shown in figure 3d. Applying our procedure, we find that fitness differences within groups have been eliminated while between-group selection still favors A5A5 and A5A3 over A3A3. A5 is like a dominant allele in the sense that A5A5 and A5A3 groups are phenotypically identical. Within- group selection has been eliminated by an evolved trait. Once again the pair has achieved a harmony and coordination that invites comparison with an organism, but with some safe-guards built in, similar to the rules of fair meiosis at the genetic level.

How was it possible for Dawkins and virtually all other evolutionary biologists to regard game theory as an individualistic theory that does not require group selection? The answer is that groups were treated as "environments" inhabited by individuals, in exactly the same sense that Williams regarded individuals as "environments" inhabited by genes. Averaging the fitness of individual types across groups combines selection at all levels into a single measure of "individual fitness" that correctly predicts the outcome of natural selection but loses sight of the vehicles that natural selection actually acts upon. Selection can operate entirely at the group level (as it does in figure 3a and d) and still be represented in terms of individual fitnesses simply because the average A2 (or A5) is more fit than the average A1 (or A3). This definition of what "individual selection" favors is synonymous with "anything that evolves, regardless of the vehicles of selection". Of course, individuals are not replicators and we can make them disappear along with groups by averaging the fitness of genes across all contexts, arriving at a definition of "gene selection" as "anything that evolves, regardless of the vehicles of selection". These bloated definitions of individual and gene selection have misled a generation of biologists into thinking that natural selection almost never occurs at the level of groups.

In this review we have concentrated on showing how the seemingly alternative theories of kin selection, evolutionary game theory and group selection have been united into a single theory of natural selection acting on a nested hierarchy of units. The unified theory does more than redescribe the familiar results of kin selection and game theory, however; it also predicts that natural selection can operate on units that were never anticipated by kin selection and game theory, such as multigenerational groups founded by a few individuals (e.g., Aviles 1993, Wilson 1987), large groups of unrelated individuals (Boyd and Richerson 1985, 1990a,b), and even multispecies communities (Goodnight 1990a,b; Wilson 1976,1980, 1987). For example, accounts of human evolution that are based on nepotism and reciprocity often conclude that prosocial behavior in modern humans is maladaptive because it is not confined to genetic relatives and is often given without expectation of return benefits (e.g., Ruse 1986; but see Alexander 1987). Later we will argue that these prosocial behaviors can be adaptive because group-level vehicles exist that are larger than the kin groups and very small groups modelled by kin selection and evolutionary game theory.

We summarize our review of group selection in biology as follows: Williams' (1966) argument against group selection came in three parts: a) higher-level adaptations require higher levels of selection, b) higher levels of selection are theoretically possible but unlikely to occur in nature, c) the gene is the fundamental unit of selection because it is a replicator. The third part of this argument is irrelevant to the question of whether groups can be like individuals in the harmony and coordination of their parts. As far as we can tell, all gene-centered theorists now concede this point (e.g., Dawkins 1982, 1989, Grafen 1984, Williams 1992). Taking vehicles seriously requires more than acknowledging a few cases of group selection, however; it demands a restructuring of the entire edifice. It is a mistake to think there is one weak group-level theory and two strong individual-level theories to explain the evolution of altruism/cooperation. Rather, there is one theory of natural selection operating on a nested hierarchy of units, of which inclusive fitness and game theory are special cases. When we focus on vehicles of selection, the empirical claim that constitutes the second part of Williams' argument disintegrates but the first part remains intact. Adaptation at any level of the biological hierarchy requires a process of natural selection at that level.

As might be expected from such a radical restructuring, some biologists who previously regarded group selection with contempt have found it difficult to accept this Cinderella-like reversal of fortunes. Thus, a large group of knowledgeable biologists who are perfectly comfortable with the hierarchical approach (see table 1) coexists with another large group whose members adhere to the earlier view. We think that the views of the former group are in the process of replacing the views of the latter. The replacement process is painfully slow, however, partly because the gene- centered view is so thoroughly entrenched and partly because the major gene-centered theorists have been reluctant to acknowledge the consequences of taking vehicles seriously. As one example, Sterelny and Kitcher (1988) manage to defend the selfish gene concept without even considering the question of whether groups can be vehicles of selection.16 We make these bold statements to provoke a response. If gene-centered theorists wish to rebut our account, let them speak in the commentary section that follows this paper. Otherwise, let the replacement process continue at a faster pace. All of the major developments that we have reviewed are over ten years old and it is time for them to be acknowledged generally.


In his description of honey bee colonies as superorganisms, Seeley (1989 p546) wrote that "...larger and more complex vehicles have evidently proved superior to smaller and simpler vehicles in certain ecological settings. By virtue of its greater size and mobility and other traits, a multicellular organism is sometimes a better gene-survival machine than is a single eukaryotic cell...Likewise, the genes inside organisms sometimes fare better when they reside in an integrated society of organisms rather than in a single organism because of superior defensive, feeding, and homeostatic abilities of functionally organized groups."

This statement applies almost as well to humans as to honeybees. Nevertheless, group-level functional organization in humans is usually portrayed as a byproduct of individual self-interest. Even the most recent evolutionary accounts of human behavior are based on Williams' first commandment and the triumph of "individual selection" in biology is often used to justify the individualistic perspective in the human behavioral sciences.

We think that the hierarchical theory of natural selection leads to a very different conclusion. Individualism in biology and in the human sciences both fail for the same reasons. As far as human evolution is concerned, group-level functional organization is not a "byproduct" of self-interest in humans any more than it is in honeybees. The metapopulation structure of human interactions is manifest; individuals live in social groups which themselves comprise a population of social groups. Even a relatively small social unit such as a village is a metapopulation of still smaller groups such as kinship units or coalitions of unrelated individuals. Genetic variation among human groups is not as great as among bee hives, but, as we will attempt to show, human cognitive abilities provide other mechanisms for concentrating natural selection at the group level, even when the groups are composed of large numbers of unrelated individuals (also see Alexander 1987, 1989, Boyd and Richerson 1985,1990, Knauft 1991).

Individualistic accounts of human behavior do not ignore these facts (e.g., Alexander 1979, 1987, 1989, 1992), but they are able to remain individualistic only by ignoring the concept of vehicles. As soon as we make vehicles the center of our analysis, group selection emerges as an important force in human evolution and the functional organization of human groups can be interpreted at face value, as adaptations that evolve because groups expressing the adaptations outcompeted other groups. The same adaptations can be and often are selectively neutral or even disadvantageous within groups. In the following sections we will sketch some of the implications of the hierarchical view for the study of human behavior.

The new group selection is not a return to naive group selection. Some biologists have been reluctant to accept group selection in any form because they fear it will encourage the uncritical thinking of Emerson and others who simply assumed the existence of higher-level adaptations (e.g., Maynard Smith 1987a,b). Behavioral scientists may share this reluctance because every branch of the human sciences seems to have thinkers like Emerson (1960) and Wynne-Edwards (1962, 1986) who treat social groups as the unit of adaptation as if individuals and their strivings scarcely exist. We therefore want to stress, in the strongest possible terms, that these views are not supported by modern group selection theory. Consider the example within biology of the Gaia hypothesis (Lovelock 1979), which portrays the entire planet as a self-regulating organism. Even a passing knowledge of group selection theory exposes Gaia as just another pretty metaphor because planet-level adaptation would require a process of between-planet selection (Wilson and Sober 1989). Grandiose theories of human societies as organisms would be correct only if natural selection operated entirely at the society level, which no one proposes. The hierarchical theory's attention to mechanism makes it easy to discredit such "theories" both in biology and the human sciences.

Groups are real. Having distanced ourselves from naive group selection, we want to stress with equal force that it is legitimate to treat social groups as organisms, to the extent that natural selection operates at the group level. Williams' first commandment ("Thou shalt not apply the adaptationist program above the individual level") is fundamentally wrong. To see this, consider a simplified situation in which natural selection acts entirely at the individual level, in which case genes within individuals become entirely cooperative and individuals within the population frequently face conflicts of interest that lead to social dilemmas. Employing the adaptationist program at the individual level leads to the celebrated insights that we discussed at the beginning of this paper. Employing the adaptationist program at the population level leads to the errors of naive group selection that Williams so effectively exposed. But now suppose that someone misleadingly suggests that we should not employ the adaptationist program at the individual level--that the fitness of individuals is actually irrelevant to the evolutionary process; it is only gene-level fitness that counts. This misleading advice would have us apply the adaptationist program below the level at which natural selection actually operates.

In a sense, this is just what the gene's-eye view of Williams and Dawkins invites us to do. Even they do not take it seriously enough to abandon the individual's-eye view, however, since they assert the equivalence of gene fitness and individual fitness when the latter are vehicles of selection. In practice, most biologists pay passing tribute to the gene as the "fundamental" unit of selection and think about adaptation at the individual level as they always have (e.g., Grafen 1984, quoted in note 4; Maynard Smith 1987a, p 125). We submit that evolutionary biologists would be severely handicapped if they could not ask the simple question "what would a well adapted individual be like?" Yet that is the very question that is prohibited at the group level by Williams' first commandment. If commandments are needed, we suggest the following: "Thou shalt not apply the adaptationist program either above or below the level(s) at which natural selection operates". This statement avoids both the excesses of naive group selection and the excesses of naive individual and gene selection that we have outlined above.

According to Campbell (1993 p1), the human behavioral sciences are dominated by something very similar to Williams' first commandment:

Methodological individualism dominates our neighboring field of economics, much of sociology, and all of psychology's excursions into organizational theory. This is the dogma that all human social group processes are to be explained by laws of individual behavior--that groups and social organizations have no ontological reality--that where used, references to organizations, etc. are but convenient summaries of individual behavior...We must reject methodological individualism as an a priori assumption, make the issue an empirical one, and take the position that groups, human social organizations, might be ontologically real, with laws not derivable from individual psychology...One of my favorite early papers (Campbell 1958) explicitly sides with that strident minority of sociologists who assert that "Groups are real!" even though it finds human organizations "fuzzier" than stones or white rats.

The hierarchical theory of natural selection provides an excellent justification for regarding groups as "real". Groups are "real" to the extent that they become functionally organized by natural selection at the group level. However, for traits that evolve by within-group selection, groups really should be regarded as by-products of individual behavior. Since group selection is seldom the only force operating on a trait, the hierarchical theory explains both the reality of groups that Campbell emphasizes and the genuinely individualistic side of human nature that is also an essential part of his thinking.18

Altruism and organism. Group selection is often studied as a mechanism for the evolution of altruism. We have also seen that groups become organisms to the extent that natural selection operates at the group level. Although the concepts of altruism and organism are closely related, there is also an important difference. Altruism involves a conflict between levels of selection. Groups of altruists beat groups of nonaltruists, but nonaltruists also beat altruists within groups. As natural selection becomes concentrated at the group level, converting the group into an organism, the self-sacrificial component of altruism disappears. In other words, an object can be an organism without tts parts behaving self- sacrificially.

The distinction between altruism and the interactions among parts of an organism is illustrated by our fanciful cricket example. The four pay- off matrices in figure 3 represent a) pure between-group selection, b) strong conflict between levels of selection, c) weak conflict between levels of selection, and d) a return to pure between-group selection. Within-group selection is absent from the first example by virtue of the situation, since coordination has an equal effect on both occupants of the leaf. Within-group selection is absent from the fourth example by virtue of an adaptation, since the "outlaw" A3 type cannot operate in the presence of the "parliament" A5 type.

It might seem that group-level adaptations would be easiest to recognize in group-level organisms. Ironically, the opposite is true, at least from the individualistic perspective. Individualists acknowledge group-level adaptations when they are easily exploited within groups, but when they are protected, or when exploitation is not possible by virtue of the situation, group-level adaptations are seen as examples of individual self-interest, despite the fact that they evolve purely by between-group selection and result in total within-group coordination. Payoff matrices such as 3a and 3d are seldom even considered by game theorists because their outcome is so obvious. In the absence of fitness differences within groups, any amount of genetic variation between groups is sufficient to select for A1 and A5, including the variation that is caused by random pairing. It is only by adding within-group selection that we can generate the social dilemmas that are deemed interesting enough to model. But A1 and A5 should not be viewed as examples of self-interest just because they easily evolve! As we have seen, groups are the vehicles of selection in these examples as surely as individuals are the vehicles in standard Darwinian selection. To call A1 and A5 examples of self-interest is to place them in the same category as A3, which evolves by within-group selection and disrupts group-level organization. Putting it another way, by lumping together the products of within- and between-group selection, the individualistic perspective does not distinguish between the outlaw and the parliament, turning oself-interest@ into a concept that is as empty as it is universal.

Failure to recognize group-level adaptation in the absence of altruism extends far beyond game theory. We present an example from Alexander (1987) in detail, in part because he is one of the most influential biologists writing on human evolution. Alexander envisions moral systems as levelers of reproductive opportunities within groups:

The tendency in the development of the largest human groups, although not always consistent, seems to be toward equality of opportunity for every individual to reproduce via its own offspring. Because human social groups are not enormous nuclear families, like social insect colonies, ...competition and conflicts of interest are also diverse and complex to an unparalleled degree. Hence, I believe, derives our topic of moral systems. We can ask legitimately whether or not the trend toward greater leveling of reproductive opportunities in the largest, most stable human groups indicates that such groups (nations) are the most difficult to hold together without the promise or reality of equality of opportunity (p69).19

Alexander explicitly compares human moral systems to the genetic rules of meiosis that eliminate fitness differences within individuals:

A corollary to reproductive opportunity leveling in humans may occur through mitosis and meiosis in sexual organisms. It has generally been overlooked that these very widely studied processes are so designed as usually to give each gene or other genetic subunit of the genome...the same opportunity as any other of appearing in the daughter cells...It is not inappropriate to speculate that the leveling of reproductive opportunity for intragenomic components--regardless of its mechanism--is a prerequisite for the remarkable unity of genomes...[p69]

Since the rules of meiosis concentrate natural selection at the individual level, producing individual-level organisms, moral rules must concentrate natural selection at the group level, producing group-level organisms--right? Wrong. Here is Alexander's verdict on group selection:

Finally, many easily made observations on organisms indicate that selection is most effective below group levels. These include such things as evidence of conflicts among individuals within social groups, failure of semelparous organisms (one-time breeders) to forego reproduction when resources are scarce, and strong resistance to adopting nonrelatives by individuals evidently long evolved in social groups. None of these observations is likely if the individual's interests are consistently the same as those of the group or if, to put it differently, allelic survival typically were most affected by selection at the group level (p37-8).

All of these examples involve altruistic traits that are highly vulnerable to exploitation within groups. The only evidence that Alexander will accept for group selection is extreme self-sacrifice. Somehow, Alexander manages to combine a strong emphasis on between-group competition and opportunity leveling within groups with a belief that group selection can be dismissed and that everything, parliaments and outlaws alike, are products of self interest.20 To make matters worse, Alexander speaks for the majority of biologists interested in human behavior. For example, here is Daly and Wilson's (1988 p254) tidy statement about human morality:

If conscience and empathy were impediments to the advancement of self- interest, then we would have evolved to be amoral sociopaths. Rather than representing the denial of self-interest, our moral sensibilities must be intelligible as means to the end of fitness in the social environment in which we evolved.

We hope the reader recognizes the familiar pattern of treating groups as "environments" inhabited by individuals and defining self-interest as "anything that evolves" without any consideration of vehicles. Alexander, Daly and Wilson join the anti-group selection chorus and then provide dozens of examples of human groups as vehicles of selection without ever acknowledging what gene-centered theorists have already conceded--that group selection is a "vehicle" question.

Alexander's theory of moral systems can be rebuilt on the foundation of vehicles as follows: Human adaptations can evolve along two major pathways; a) by increasing the fitness of individuals relative to others within the same social group, and b) by increasing the fitness of social groups as collectives, relative to other social groups. Both pathways have been important in the evolution of the psychological mechanisms that govern human behavior. Sometimes group selection is important just by virtue of the situation. For example, the only way to defend a village might be to build a stockade, which benefits the collective by its nature. We are not surprised to see villagers building stockades, even when they are genetically unrelated to each other. We are not surprised when they coordinate their efforts in ways that invite comparison to a single organism. Nor do we regard them as especially morally praiseworthy as they feverishly work to save their collective skins. But building the stockade is not selfish just because it is reasonable. Applying our procedure, we find that the village is the vehicle of selection. We expect the stockade to be built for the same reason that we expect A1 and A5 to evolve in the game theory models; because in this particular situation group-level selection is very strong relative to within-group selection. If we define behaviors on the basis of fitness effects (as all evolutionists do), and if we want our terminology to reflect the vehicle(s) that natural selection acts upon, we should call stockade-building groupish, not selfish.

Many other situations in human life provide opportunities for adaptation via the first pathway, by increasing the fitness of individuals relative to others within the same social group. Even with our stockade example we can imagine a temptation to selfishly cultivate one's own garden or romantic possibilities as others build the stockade. The use of the word selfish is fully appropriate here because the individual is the vehicle of selection whose behaviors tend to disrupt group-level functional organization.

The balance between levels of selection is not determined exclusively by the situation, however. Adaptive human behavior not only reflects the balance between levels,but also can alter the balance between levels. Moral sentiments and moral social systems may function as "rules of meiosis" that often concentrate fitness differences, and therefore functional organization, at the group level. This is the core of Alexander's thesis. When stated in terms of vehicles, however, AlexanderYs theory acquires a familiar and conventional ring that is absent from his own account. Moral systems are defined as "social organizations designed to maximize the benefit of the group as a collective." Immoral behaviors are defined as "behaviors that benefit individuals at the expense of other individuals within the same group." These are close to the concepts of moral and immoral behavior in folk psychology.21 The shock value of Alexander's account, in which the gentle reader is made to face the grim reality that all is self-interest, evaporates when we realize that for Alexander, self-interest is everything that evolves, at all levels of the biological hierarchy.22

We will return to moral systems with an empirical example, but first we must consider the important issue of psychological motivation.

Psychological selfishness and its alternatives. Dawkins portrays genes as psychologically selfish entities that manipulate their environment, including the genotypic environment in which they reside, to increase their own fitness. This image is obviously metaphorical, allowing Dawkins to use a familiar human reasoning process to describe the outcome of natural selection. The metaphor is relatively innocuous because there is no danger that it can be taken literally. No one believes that genes are intentional systems of any sort, much less systems motivated by self interest.

Frame-shifting upward, it is possible to portray individuals as psychologically selfish entities that manipulate their environment, including the social environment in which they reside, to increase their own fitness. This image of "selfish individuals" may also be metaphorical but it is more insidious because it can be taken literally. In other words, it is possible to believe that individuals really are intentional systems motivated entirely by self-interest and this is, in fact, the individualistic perspective that pervades the human sciences.

To distinguish mechanisms from metaphors, it is useful to think of a psychological motive as a strategy in the game theoretic sense, which produces a set of outcomes when it interacts with itself and with other strategies. Thus, a psychologically selfish individual (however defined) will be motivated to behave in certain ways with consequences for itself and others. A psychologically altruistic individual (however defined) will be motivated to behave in other ways with a different set of payoffs. Within an evolutionary framework, the empirical claim that individuals are motivated entirely by self-interest must be supported by showing that the psychologically selfish strategy prevails in competition with all other strategies.

Psychological motives have seldom been analyzed in this way (but see Frank 1988, Alexander 1987) and we suggest that it will be a productive line of inquiry in the future. We also predict that two general conclusions will emerge: First, it is extremely unlikely that any single strategy will prevail against all other strategies. Even the famous Tit-for-Tat strategy, which is robust in the narrow context of Axelrod's (1980 a,b) computer tournaments, is vulnerable to a host of other strategies in more complex and realistic environments (e.g., Boyd and Lorberbaum 1987, Dugatkin and Wilson 1991, Feldman and Thomas 1987, Peck and Feldman 1986). Thus, any monolithic theory of proximate motives is destined to fail, including the monolithic theory of psychological selfishness. We should expect a diversity of motives in the human repertoire that is distributed both within and among individuals.

Second, the very opposite of psychological selfishness can be highly successful, especially when natural selection operates at the group level. To see this, consider an individual who identifies so thoroughly with his group that he doesn't even consider the possibility of profiting at the expense of his fellows. This individual will be vulnerable to exploitation by members of his own group who are less civic-minded. But groups of individuals who think in this way will probably be superior in competition with other groups whose members are less civic-minded. It follows that intense between-group competition will favor psychological mechanisms that blur the distinction between group and individual welfare, concentrating functional organization at the group level. Alexander (1988) himself provides a good example in a review of Richards (1987) when he describes his own military experience:

In the army in which I served one was schooled so effectively to serve the welfare of his unit (community?) that not only the contract altruism that Richards says is inferior to his "pure" altruism, but the intent that he requires,both disappear in a kind of automaticity that ceases to involve any deliberateness, either in maintenance of the contract signed when drafted or enlisted, or in explicitly serving the rest of one's unit [p443].

Quibbles about the definition of altruism aside, nothing more is required to convert a social group into an organism. Critics may argue that the selfless attitude of a well-trained soldier is not adopted by individual choice but imposed by an indoctrination process and reinforced by sanctions against disloyalty that make it disadvantageous to cheat. We disagree in two ways. First, individuals are not always drafted into these groups and often rush to join them, enthusiastically embracing the doctrine, refraining from cheating and enforcing the sanctions against others. Their self-interest is not taken from them but willingly abandoned. Second, even when imposed, indoctrination and sanctions are best regarded as group-level rules of meiosis that reduce the potential for fitness differences within groups, concentrating functional organization at the group level. An entity can be an organism without the parts behaving self-sacrificially (for an evolutionary model of psychological altruism per se, see Frank 1988).

Since humans have lived in small groups throughout their history, it is reasonable to expect the evolution of psychological mechanisms that cause them to easily become "team players" in competition with other groups. We do not expect these to be the only motives that guide human behavior, but rather a module that is facultatively employed under appropriate conditions. In fact, there is abundant empirical evidence that humans coalesce into cooperative teams at the merest suggestion of a metapopulation structure in which groups can compete against other groups (e.g., Dawes et al 1988, Hogg and Abrams 1988, Sherif et al 1961, Tajfel 1981 ). Members of the same group often share a feeling of high regard, friendship and trust that is based not on any prior experience but merely by the fact that they are members of the same group. Exploitation within groups is often avoided even when opportunities are experimentally provided without any chance of detection (e.g., Caporeal et al 1989). Group formation is as spontaneous in children as in adults (e.g., Sherif et al 1961). These are the earmarks of an evolved "Darwinian algorithm" (sensu Cosmides and Tooby 1987) that predisposes humans for life in functionally organized groups. The algorithm appears paradoxical only when we consider its vulnerability to more selfish algorithms within groups. The advantages at the group level are manifest.

It is important to stress that we have not merely converged on a view that is already well accepted within the human sciences. Proponents of alternatives to psychological selfishness are better described as an embattled minority who must constantly defend themselves against a monolithic individualistic world view (e.g., Batson 1991, Caporeal et al 1989, Campbell 1993, Mansbridge 1990, Simon 1991). As one example, most economists assume that individuals act in the interest of the company that employs them only because the company pays them enough to make it worthwhile from the standpoint of the individual's personal utility. According to Simon (1991), real people who are satisfied with their jobs do not distinguish between their own and their company's utility, but rather adopt the company's interest as their own interest. Even the lowest level employees make executive decisions that require asking the question "what is best for the company?" and which go far beyond the actual requirements of the job. In fact, one of the most effective forms of protest by dissatisfied employees is "work to rule", in which people perform their jobs to the letter and the company comes to a grinding halt. In modern life as in ancient times, group-level function requires individuals who to a significant degree take the group's goals as their own. This is a radical proposal within economics, however, and not the majority view.

Group-level cognition. Goal-oriented behaviors are typically accomplished by a feedback process that includes the gathering and processing of information. While the entire process can be described as intentional (e.g., the wolf tries to catch the deer), the elements of the process cannot (the neuron does not try to fire; it merely does fire when stimulated at enough synapses; Dennett 1981).

We are accustomed to regarding individuals as intentional systems with their own self-contained feedback processes. Group selection raises another possibility in which the feedback process is distributed among members of the group. We have already provided an example for honeybee colonies in which individuals behave more like neurons than as intentional agents in their own right. Similar examples have scarcely been considered for humans and our main purpose here is to define the question, rather than answer it.

Modern governmental and judicial systems are sometimes designed to produce adaptive outcomes at the level of the whole system but not at the level of the component individuals. Science is sometimes portrayed as a similar process that generates knowledge only at the group level (e.g., Hull 1988, Kitcher 1993). The invisible hand metaphor in economics invokes the image of an adaptive system that organizes itself out of neuron-like components, although the metaphor is more often stated as an ideology than as a testable research program.23

The invisible hand notwithstanding, research on decision making in small groups reveals a complex process that does not always yield adaptive solutions (Hendrick 1987a,b). Groups even make decisions that would be regarded as foolish by every member of the group (Allison and Messick 1987). This research is important because it shows that intelligent individuals do not automatically combine to form intelligent groups. Adaptive decision-making at the small group level may require a highly specified cognitive division of labor. Since decision-making has occurred in small groups throughout human history, it is reasonable to expect "Darwinian algorithms" that cause individuals to relinquish their capacity to act as autonomous intentional agents and adopt a more limited role in a group-level cognitive structure. The architecture of group-level cognition might simply take the form of "leaders" who act as self- contained intentional agents and "followers" who abide by the decisions of others. Alternatively, even so-called "leaders" may be specialists in a feedback process that is distributed throughout the group. These questions can only be asked by recognizing the group as a potentially adaptive entity.

An example of a human group-level oorganism@. We conclude by providing a possible example of extreme group-level functional organization in humans and the background conditions that make it possible. The Hutterites are a fundamentalist religious sect that originated in Europe in the sixteenth century and migrated to North America in the nineteenth century to escape conscription. The Hutterites regard themselves as the human equivalent of a bee colony. They practice community of goods (no private ownership) and also cultivate a psychological attitude of extreme selflessness. The ultimate Hutterite virtue is oGelassenheit@, a word that has no English equivalent, which includes othe grateful acceptance of whatever God gives, even suffering and death, the forsaking of all self- will, all selfishness, all concern for private property@ (Ehrenpreis 1650/1978). Nepotism and reciprocity, the two principles that most evolutionists use to explain prosocial behavior in humans, are scorned by the Hutterites as immoral. Giving must be without regard to relatedness and without any expectation of return. The passion for selflessness is more than just sermonizing and frequently manifests itself in action. For example, Claus FelbingerYs oConfession of faith@ (1560/1978) provides an eloquent statement that a Hutterite blacksmith gave to Bavarian authorities after their efforts to make him recant had failed and before they executed him for his beliefs.

The extreme selflessness of the Hutterites can be explained in at least three ways. First, many authors, both inside and outside of biology, think of culture as a process that frequently causes humans to behave in ways that are biologically maladaptive. By this account, the Hutterites are influenced by (unspecified) cultural forces and their behavior cannot be explained by any biological theory, including the theory of group selection.

Second, some evolutionists have tried to explain widespread altruism in humans as a product of manipulation, in which the putative altruists are essentially duped into behaving against their own interests for the benefit of the manipulator (e.g., Dawkins 1982,1989). If people can be fooled into believing that a life of sacrifice will lead to a pleasant afterlife, for example, then perpetrating that belief in others becomes an example of individual self-interest. By this account, we might expect some Hutterites (such as the leaders) to profit at the expense of their duped brethren.

Third, it is possible that humans have evolved to willingly engage in selfless behavior whenever it is protected by a social organization that constitutes a group-level vehicle of selection. The relatively small group- level vehicles of kinship groups and cooperating diads are already well recognized. The hypothesis we wish to explore asserts that the Hutterites constitute a less familiar case in which the vehicle is a relatively large group of individuals and families that are genetically unrelated to each other. If this interpretation is correct, then group selection theory should be able to predict some major features of Hutterite social organization and ideology, despite the fact that it is stated in purely religious terms. In particular, the prediction is that the bee-like behavior of the Hutterites is promoted by a social organization and ideology that nearly eliminates the potential for individuals to increase their fitness relative to others within groups.24 Note that the other two interpretations do not make the same prediction. If Hutterite society is governed by independent cultural forces, it is unlikely to have the specific design features of a group-level vehicle. And if selflessness is a product of manipulation, we should find fitness differences between the puppets and the puppeteers.

A number of caveats are in order before proceeding. First, we do not claim to rigorously distinguish among the above three explanations in the confines of this review article. The best that we can do is provide a brief sketch, which is nevertheless important because it makes the preceding discussion less abstract and gives an idea of what a group-level vehicle of selection might look like in humans. Second, we do not regard the Hutterite social organization as a direct product of group selection. Rather, we conjecture that group selection has operated throughout human history, endowing the human psyche with the ability to construct and live within group-level vehicles of the sort exhibited by the Hutterites.25 This enables us to make another prediction, that the Hutterite social organization is not unique but represents a fairly common type of social organization in ancestral environments. Otherwise it could not be interpreted as an evolved adaptation. As one of many possible social organizations in the human repetoire, this one is presumably evoked only under appropriate environmental conditions, yielding another set of testable predictions. Third, evolutionary psychologists rely on fitness maximizing arguments to explain the human psyche, but they do not necessarily expect humans to maximize biological fitness in present day environments. This is because, to the extent that humans are oprogrammed@ by natural selection, it is not to maximize biological fitness per se but only to achieve the more proximate goals that led to high fitness in ancestral environments. Thus, we must focus more on the design features and what they would have meant in ancestral environments than on the present day consequences of the design features (Symons 1992). This is a general issue in evolutionary psychology that applies to the Hutterites as well as any other group.

With these caveats in mind, we now will elaborate the idea that Hutterite society is a group-level vehicle of selection. Although their ideology is stated in purely religious terms, it is clearly designed to suppress behaviors that benefit some individuals at the expense of others within groups:

That is what Jesus means by His parable of the great banquet and the wedding of the king's son, when the servants were sent to call all the people together. Why did his anger fall on those who had been invited first? Because they let their private, domestic concerns keep them away. Again and again we see that man with his present nature finds it very hard to practice true community; true community feeds the poor every day at breakfast, dinner, and the common supper table. Men hang on to property like caterpillars to a cabbage leaf. Self-will and selfishness constantly stand in the way! [Ehrenpreis 1650/1978 p 11-12]

Benefitting the group is exalted as highly as selfishness within groups is reviled:

Where there is no community there is no true love. True love means growth for the whole organism, whose members are all interdependent and serve each other. That is the outward form of the inner working of the Spirit, the organism of the Body governed by Christ. We see the same thing among the bees, who all work with equal zeal gathering honey; none of them hold anything back for selfish needs. They fly hither and yon with the greatest zeal and live in community together. Not one of them keeps any property for itself. If only we did not love our property and our own will! If only we loved the life of poverty as Jesus showed it, if only we loved obedience to God as much as we love being rich and respected! If only everybody did not hang on to his own will! Then the truth of Christ's death would not appear as foolishness. Instead, it would be the power of God, which saves us. [Ehrenpreis 1650/1978 p12-13]

Thus the Hutterites are as explicit as they can possibly be that their members should merge themselves into a group-level organism. They are also explicit about how group-level functional organization can be accomplished. In the first place, the Hutterites believe that selfishness is an innate part of human nature that can never be fully irradicated:

The sinner lies in all of us; in fact to sin, to be selfish,is our present inclination. Left to ourselves we shall end up in damnation, but this does not mean that salvation cannot be attained. On the contrary, salvation is possible on three conditions: we live according to the life of Christ; we live in community; we strive very hard to attain salvation and are prepared to suffer for our efforts. Christ appeared to save us from our sinful nature. This nature is not easily abjured but it can if we try hard enough, both in the sense of personal determination and in the sense of collectively living according to the Word (Shenker 1986 p73).

If we were to translate this sentiment into evolutionary language, we would arrive at the claim that within-group selection has been a powerful (but not the only) force in human evolution and has stamped itself upon the human psyche. To the extent that humans are the products of natural selection, they are inclined to benefit themselves at the expense of others within their group whenever it is evolutionarily advantageous to do so (at least in ancestral environments). To create a group-level organism, the part of human nature that has evolved by within-group selection must be constrained by a social organization that plays the same functional role as the genetic rules of meiosis.

The most important ingredient of this social organization is evidently a sense of oshared fate@:

The community can "hang together" only through the members having an identity of fate. In practice this means two things. Members must identify with the past and (more important) with the future of the community, such that their own future and the community's future are one and the same. We rise and fall together. This is another way of saying we have unconditional commitment to our community. We do not say "if the community does or achieves such and such, then I will stay, otherwise I won't", since this implies that there is an individual identity ontologically and morally distinct from the community's. No true community could operate successfully or manifest its raison d'etre with such limiting conditions or separate identities. Identity of fate also means that members relate to each other in an atmosphere of mutual trust, i.e. they consider their presence to stem from a common desire to express their humanity and recognize that this can only be achieved through mutual effort. Should one person claim that he has an inherent right to gain for himself at the expense of others, the entire fabric collapses. Life in the community presupposes that each will work for the benefit of others as much as for himself, that no-one will be egoistic. The moment this assumption is undermined, mutual suspicion, jealousy and mistrust arise. Not only will people probably consider themselves silly for being self- righteous while others are feathering their nest,but operationally the community will have to take a different character (primarily through the use of coercion) and the entire moral nature of the community disappears. (Shenker 1986 p93)

We could not ask for a stronger correspondence between the sentiment expressed in this passage and the concept of ovehicles@ in a group selection model.

One way to establish a sense of shared fate is via egalitarian social conventions that make it difficult to benefit oneself at the expense of others. Hutterite society is elaborately organized along these lines. In addition to practicing community of goods, they discourage individuality of any sort, for example, in the context of personal appearance and home furnishings. Leaders are elected democratically and are subject to long probationary periods before they are given their full authority. The HutteriteYs passion for fairness is perhaps best illustrated by the rules that surround the fissioning process. Like a honey bee colony, Hutterite brotherhoods split when they attain a large size, with one half remaining at the original site and the other half moving to a new site that has been pre-selected and prepared. In preparation for the split, the colony is divided into two groups that are equal with respect to number, age sex, skills and personal compatibility. The entire colony packs its belongings and one of the lists is drawn by lottery on the day of the split. The similarity to the genetic rules of meiosis could hardly be more complete.

In principle, we might imagine that a psychological egoist, who thinks only in terms of personal gain, could decide to become a Hutterite if he became convinced that the group-level benefits (which he shares) are sufficiently great and the social conventions are sufficiently strong that neither he nor anyone else in the group can act as a freeloader. The Shenker passage quoted above suggests, however, that an effective group- oriented society cannot be composed of individuals who are motivated solely by a calculus of self-interest.26 The external social conventions that make freeloading difficult are evidently necessary but not sufficient and must be supplemented by a psychological attitude of genuine concern for others; a direct calculus of group interest rather than self interest is essential. Recall that Simon (1991; discussed on p38) makes a similar point about the behavior of individuals in business organizations. Thus, although we are focusing on the Hutterites, our discussion is not limited to esoteric communal societies, a point that we will return to below.

Even with these attitudes and social conventions, however, selfishness in thought and action cannot be entirely eliminated. The Hutterites therefore have a well specified procedure for dealing with members who benefit themselves at the expense of others.

The bond of love is kept pure and intact by the correction of the Holy Spirit. People who are burdened with vices that spread and corrupt can have no part in it. This harmonious fellowship excludes any who are not part of the unanimous spirit... If a man hardens himself in rebellion, the extreme step of separation is unavoidable. Otherwise the whole community would be dragged into his sin and become party to it...The Apostle Paul therefore says, "Drive out the wicked person from among you."

... In the case of minor transgressions, this discipline consists of simple brotherly admonition. If anyone has acted wrongly toward another but has not committed a gross sin, a rebuke and warning is enough. But if a brother or a sister obstinately resists brotherly correction and helpful advice, then even these relatively small things have to be brought openly before the Church. If that brother is ready to listen to the Church and allow himself to be set straight, the right way to deal with the situation will be shown. Everything will be cleared up. But if he persists in his stubbornness and refuses to listen even to the Church, then there is only one answer in this situation, and that is to cut him off and exclude him. It is better for someone with a heart full of poison to be cut off than for the entire Church to be brought into confusion or blemished.

The whole aim of this order of discipline, however, is not exclusion but a change of heart. It is not applied for a brother's ruin, even when he has fallen into flagrant sin, into besmirching sins of impurity, which make him deeply guilty before God. For the sake of example and warning, the truth must in this case be declared openly and brought to light before the Church. Even then such a brother should hold on to his hope and his faith. He should not go away and leave everything but should accept and bear what is put upon him by the Church. He should earnestly repent, no matter how many tears it may cost him or how much suffering it may involve. At the right time, when he is repentant, those who are united in the Church pray for him, and all of Heaven rejoices with them. After he hasshown genuine repentance, he is received back with great joy in a meeting of the whole Church. They unanimously intercede for him that his sins need never be thought of again but are forgiven and removed forever. [Ehrenpreis 1650/1978 p66-9]

We could not ask for a more explicit awareness of the freeloader problem and what to do about it, including the elements of retaliation and forgiveness that are also part of the tit-for-tat strategy in diadic interactions. If we were to translate this passage into evolutionary language, it would be as follows: Altruism can succeed only by segregating itself from selfishness. Not only does the selfish individual have the highest fitness within groups, but his mere presence signifies a population structure that favors within-group selection, causing others to quickly abandon their own altruistic strategy. Fortunately, in face-to-face groups whose members are intimately familiar with each other, it is easy to detect overt forms of selfishness and exclude the offender. When osubversion from within@ can be prevented to this extent, extreme altruism, in both thought and action, becomes evolutionarily advantageous.

It is remarkable, and crucial for the hypothesis under consideration, that the willingness of the Hutterites to sacrifice for others is accompanied by such an elaborate set of rules that protect self-sacrifice from exploitation within groups. We suggest that there is a causal relationship here, that humans are inclined to adopt selfless behavior in social organizations that provide the functional equivalent of the genetic rules of meiosis. Not only do these social organizations promote selflessness at the behavioral level, but they also promote forms of thinking and feeling that would be classified as non-egoistic in a psychological sense. After all, what is the advantage of psychological selfishness if the most successful way to behave is by contributing to group-level functional organization?

It is also crucial for our hypothesis that group-level functional organization is, in some sense, superior to what can be accomplished by individuals when they are free to pursue their own self interest (recall the Seeley passage quoted on pg 27). This certainly appears to be the case for the Hutterites, who do not have to wait for the hereafter to get their reward. By fostering a selfless attitude towards others and minimizing the potential for exploitation within groups, they are spectacularly successful at the group level. In sixteenth century Europe they were alternately tolerated and persecuted for their economic superiority, much like the Jews, another society that, in its traditional form, is well- organized at the group level (MacDonald 1994). In present-day Canada, Hutterites thrive in marginal farming habitat without the benefit of modern technology and almost certainly would displace the non-Hutterite population in the absence of laws that restrict their expansion. The HutteritesY success can also be measured in reproductive terms, since they have the highest birth rate of any known human society (Cook 1954).27 Finally, Hutterite society is internally stable, with the majority of young people electing to remain when given a choice. Were it not for persecution and legal restrictions imposed by their host nations, Hutterite colonies would be far more common than they are now.

Part of our hypothesis is that the Hutterite social organization is not a unique product of the sixteenth century but reflects an evolved human potential to construct and live within such group-level vehicles. It might seem that the Hutterites are such an esoteric society that our prediction could not possibly be confirmed. On closer reflection, however, it appears that the functional elements of Hutterite society that act as group-level rules of meiosis are repeated in a great many social groups that place a premium on group-level performance, even though the ideologies are superficially different and the purpose of the group can be diametrically opposed to the goals of the Hutterites (e.g., an elite military group). Furthermore, according to Knauft (1991), this kind of egalitarianism characterizes hunter-gatherer groups whenever resources are too widely dispersed to allow the development of status-based societies (i.e., most human groups throughout human evolutionary history). The ethic of ogood company@ (which is extended to non-kin as well as kin; e.g. Knauft 1985) and the de-emphasized sense of self-interest that pervades many tribal societies does indeed resemble the HutteriteYs ocommunity@ and their denigration of oself-will@.

Another part of our hypothesis is that the human potential to build and live within group-level vehicles is facultative and evoked more strongly in some situations than in others. Group-level vehicles should be most commonly observed in situations that place a premium on group-level functional organization, such as extreme physical environments, extreme persecution, or extreme intergroup competition. In more benign situations, the consequences of social dilemmas are not so dysfunctional and the effort that goes into the maintenance of group-level vehicles may be correspondingly relaxed.28

Obviously, we have only skimmed the surface of an enormously complex and poorly understood subject. We hope we have demonstrated the likelihood, however, that group selection in humans extends far beyond nepotism and narrow reciprocity. These two principles cannot account for the full range of prosocial behaviors in humans and evolutionists who rely on them have been forced to invoke other factors; that prosocial behavior evolved in ancestral groups of closely related individuals and is maladaptively expressed in modern groups of unrelated individuals (Ruse 1986); that prosocial behavior is a form of manipulation whereby some individuals profit at the expense of others (Dawkins 1982, 1988); or that prosocial behavior results from cultural forces that promote biologically maladaptive behavior (Campbell 1983). Group selection theory provides a robust alternative: Even large groups of unrelated individuals can be organized in a way that makes genuinely prosocial behavior advantageous.

We have emphasized group-level functional organization in humans as an antidote to the rampant individualism we see in the human behavioral sciences. But it is not our goal to replace one caricature with another. Many human groups are clearly not the oorganisms@ that we have described above and must be explained as the product of conflicting individual interests within the group. Evolutionary theory has the resources to understand both conflict and cooperation. Only by pursuing both problems- -with the group as well as the individual as possible units of functional integration--can the human sciences come to terms with our evolutionary heritage.


Maynard Smith's most recent comment on group selection includes the following passage:

It is ...perfectly justified to study eyes (or, for that matter, ribosomes, or foraging behaviors) on the assumption that these organs adapt organisms for survival and reproduction. But it would not be justified to study the fighting behavior of spiders on the assumption that this behavior evolved because it ensures the survival of the species, or to study the behavior of earthworms on the assumption that it evolved because it improves the efficacy of the ecosystem. (Maynard Smith 1987b p147)

Maynard Smith still resists what we think is the most fundamental implication of natural selection as a hierarchical process: Higher units of the biological hierarchy can be organisms, in exactly the same sense that individuals are organisms, to the extent that they are vehicles of selection. Group organisms may be less common than individual organisms and they may be more vulnerable than individuals to subversion from within, but this must not prevent us from recognizing group-level functional organization where it exists.

As the most facultative species on earth, humans have the behavioral potential to span the full continuum from organ to organism, depending on the situations we encounter and the social organizations that we build for ourselves. We often see ourselves as "organs". We sometimes identify ourselves primarily as members of a group and willingly make sacrifices for the welfare of our group. We long to be part of something larger than ourselves. We have a passion for building, maintaining and abiding by fair social organizations. The individualistic perspective seems to make all of this invisible. Because group-level functional organization can be successful, it is labelled selfish, therefore no different from the kinds of behaviors that succeed by disrupting group-level functional organization. But this is just a conjurer's trick. There are compelling intellectual and practical reasons to distinguish between behaviors that succeed by contributing to group-level organization and behaviors that succeed by disrupting group-level organization. That is what the words "selfish" and "unselfish", "moral" and "immoral" are all about in everyday language. Human behavioral scientists need to focus on these ancient concerns, rather than obscuring them with bloated definitions of "self-interest". A concern for within-group versus between-group processes characterizes the human mind and it should characterize the study of the human mind as well.


Supported by NSF SBE-9212294. DSW thanks A.B. Clark, Lee Dugatkin, Eric Dietrich, Greg Pollock and Binghamton's EEB group.


1) In this article we use the word oindividual@ to refer to single flesh- and-blood creatures such as a bird or a butterfly. We use the term oorganism@ to refer to any biological entity whose parts have evolved to function in a harmonious and coordinated fashion.

2) The purpose of this table is to provide a reasonably complete guide to the modern literature on group selection. A number of controversies exist within this literature that are beyond the scope of the present paper. For completeness we provide references for all sides of these controversies, including those with which we disagree. The philosophical literature on levels of selection has recently been reviewed by Sober and Wilson (1993).

3) Williams was only one of many biologists who reacted against group selection during the 1960's, especially in response to Wynne-Edward's (1962) Animal Dispersion in Relation to Social Behavior. We do not mean to imply that Williams was the only articulate critic, but he has become the icon for the individualistic perspective in biology.

4) Dawkins (1982, 1989) acknowledges that the group selection controversy is a "vehicle" question but asserts that groups are almost never vehicles of selection, with the possible exception of the eusocial insects. Dawkins (1989 p 297-8) and Cronin (1991 p 290) cite Grafen (1984) as the authoritative critique of group selection but Grafen's treatment of groups as vehicles consists of a single parenthetical statement (p76): "(The organismal approach suggested here is not in conflict with the the 'gene selectionism' of Dawkins (1982a,b). In his language, we are saying that the individual is usually a well-adapted vehicle for gene replication, while groups are usually not)". Williams (1986 p8) states that "selection at any level above the family (group selection in a broad sense) is unimportant for the origin and maintenance of adaptation. I reach this conclusion by simple inspection." More recently, Williams (1992) acknowledges that groups are vehicles in the specific cases of eusocial insects and female-biased sex ratio but does not generalize to other cases.

5) Cronin's (1991) The Ant and the Peacock belongs to the same genre as Dawkins' (1976) The Selfish Gene and Gould's (1989) Wonderful life, in which the author attempts to make the subject accessible to a popular audience without sacrificing scholarship. As Gould (1989 p16) put it, "...we can still have a genre of scientific books suitable for and accessible alike to professionals and interested laypeople". Because these books are so accessible they tend to be influential even among academic audiences, which is why Cronin (1991) merits criticism despite its status as a "popular" book. Similar views can be found in the more technical gene- centered literature (references in note 4)

6) Gould (1992) criticizes Cronin's gene-centered approach and advocates a hierarchical view of evolution. However, he accepts the gene-centered framework for the evolution of altruism and does not invoke the concept of vehicles in the same sense that we do. More generally, the concept of "species selection" that Gould emphasizes is somewhat different from the concept of group selection that we review here (for a discussion of the difference, see Sober 1984). This constitutes one of the controversies within the group selection literature mentioned in note 1.

7) The term "unit of selection" has become ambiguous because it refers to both replicators and vehicles, depending on the author. Within the group selection literature, "unit" equals "vehicle" and no word is required for "replicator" because it is (and always was) assumed that natural selection at all levels results in gene-frequency change. We prefer the word "unit" but use the word "vehicle" in this paper to distinguish it from replicators and also to force gene-centered theorists to acknowledge the implications of their own framework.

8) We start at the lowest level and work up the hierarchy for convenience, not because it is required for the procedure. Also, unless there is uncertainty as to where fitness differences are occuring, it is not necessary to invoke WilliamsY (1966) concept of parsimony in this procedure.

9) Even though organisms are defined on the basis of functional coordination among their parts, functional coordination per se does not enter into our definition of vehicles, which is based purely on shared fate. This is because shared fate is the crucial property of the process of natural selection; functional coordination among the parts is a product of the process.

10) The procedure for identifying vehicles requires some precautions that can be illustrated by the following examples. First, imagine that tall individuals are more fit than short individuals regardless of how they are structured into groups. The procedure will (correctly) identify the individual as the vehicle of selection. Nevertheless, groups that contain more tall individuals than other groups will be more productive, suggesting (incorrectly) that groups are also vehicles of selection. To resolve this difficulty we must imagine placing all individuals in a single group. Tall individuals are still most fit, demonstrating that the metapopulation structure is irrelevant . As a second example, imagine that the fitness of everyone in a group is directly proportional to the average hight of the group. Our procedure (correctly) identifies the group as the vehicle of selection because there are no fitness differences between individuals within groups. To confirm this result, imagine placing all individuals in the same group. The fitness of tall and short individuals are identical, demonstrating that the metapopulation structure is necessary for tallness to evolve (Sober 1984; see also Goodnight et al 1992, Heisler and Damuth 1987, Walton 1991). Another problem arises when a trait has already evolved to fixation. To apply the procedure we must conduct a thought experiment (or a real experiment) in which alternative types are present. Although other refinements in our procedure may be needed, we believe that they don't require discussion in the present context.

11) Although female biased sex ratios evolve by group selection, they cannot be used to assess the importance of group selection in the evolution of other traits. In other words, it does not follow that group selection can be ignored for species that have an even sex ratio. This is because the metapopulation structure must be defined separately for each trait (hence the term otrait group@; Wilson 1975, 1977, 1980). The trait group for sex ratio must persist long enough for f1 progeny to mate within the group before dispersing, a constraint that does not necessarily apply to other traits.

12) For most evolutionists, the ultimate rejection is to be labelled "non- Darwinian". In fact, Darwin's (1871) theory for the evolution of human moral sentiments is remarkably similar to the vehicle-based framework that we develop here (Richards 1987).

13) We think that an evolutionary theory of genuine vs. apparant psychological altruism is possible, but it must be based on the proximate motivations of the actor, which evolutionary accounts ignore by defining altruism and selfishness solely in terms of fitness effects. In other words, we must ask questions such as: oWhen are the behaviors motivated by a ZgenuinelyY psychologically altruistic individual more fit than the behaviors motivated by an ZapparantlyY altruistic individual?@ Frank is actually one of the few authors who are asking these questions, so we are not criticizing his specific proposals about emotions as commitment devices, which make more sense within a vehicle-based framework than within a replicator-based framework. For further discussion, see Wilson (1992) and Sober (in press).

14) Frank (1988) anticipates this conclusion in the passage that we quote above, but does not pursue it further.

15) Anatol Rapoport, who submitted the Tit-for-Tat strategy to Axelrod's (1980a,b) computer tournaments, always appreciated its group-level benefit and individual-level disadvantage (e.g., Rapaport 1991). In contrast, Axelrod and the majority of evolutionary game theorists regard tit-for-tat as a strategy that succeeds "at the individual level."

16) Sterelny and Kitcher recognize that Dawkins' position cannot simply be the empty truism that evolution occurs when the genetic composition of a population changes. They claim (p 340) that the nontrivial thesis that Dawkins advances is that "evolution under natural selection is thus a process in which, barring complications, the average ability of the genes in the gene pool to leave copies of themselves increases with time."

Although this is a nontrivial claim, it is not something we find in Dawkins' writings, and in any case it isn't true as a generality. The average fitness of the alleles at a locus increases under frequency- independent selection. But when a truly selfish gene replaces an altruistic allele, the effect is to reduce average fitness. Dawkins frequently remarks that there is nothing to prevent natural selection (meaning within-group selection) from driving a population straight to extinction. It is also worth noting that group selection can lead the average fitness of the selected alleles to increase. A gene's ability to leave copies of itself can decline under selection as well as increase. And which turns out to occur is a separate issue from whether group selection is present or absent.

17) While group selection has been a controversial topic within biology, the entire subject of evolution has been a controversial topic when applied to human behavior. There are at least three ways that evolution in general (and group selection in particular) can influence human behavior. First, the psychological mechanisms that govern human behavior can be the product of natural selection. In its weak form this statement is uncontroversial, since everyone agrees that basic drives such as hunger, sex and pain exist because they are biologically adaptive. Some psychologists believe that the adaptationist program can be used to explain the architecture of human cognition in much greater detail, however, and this position is more controversial (e.g., Barkow et al 1992). Second, cultural change can itself be described as an evolutionary process with between- and within-group components (e.g. Boyd and Richerson 1985, Findlay 1992). Third, genetic evolution is an ongoing process that can partially explain differences between individuals and populations. Our own thinking is based primarily on the first and second influences. In other words, we think it is imperative to explore the hypothesis that group selection was a strong force during human evolution, resulting in proximate psychological mechanisms that today are universally shared and allow humans to facultatively adopt group-level adaptations under appropriate conditions. The specific nature and precision of these psychological mechanisms are empirical issues. We also propose, along with Boyd and Richerson (1985) and Findlay (1992), that group selection can be a strong force in cultural evolution. Thus, our position is compatable with but does not require a strong form of human sociobiology. Our point is not to prejudge the correctness of adaptationist explanations, but to urge the importance of asking adaptationist questions. Only by doing so can we find out whether and to what degree organisms are well adapted to their environments (Orzack and Sober, in press).

18) Sober (1981) discusses the relationship between methodological individualism and the units of selection controversy in more detail.

19) Opportunity levelling is not restricted to the largest human groups. According to Knauft (1992), the simplest human societies are highly egalitarian and overtly status-oriented societies require a concentrated and stable resource, such as crops or livestock. This improves Alexander's general thesis, especially if the simplest existing human societies represent ancestral conditions.

20) AlexanderYs views on group selection, presented in articles and books from 1974 to 1993, are difficult to represent in a single passage. When evaluating group selection in non-human species, Alexander identifies strongly with the views of Williams and Dawkins, as the passage quoted on p 7 shows. Alexander does speculate that humans may be an exception to the rule because of extreme between-group competition and regulation of fitness differences within groups. The following passage illustrates his pro-group selection side, which is consistent with our own interpretation: oIn sexually reproducing organisms, such as humans, confluences of interest within groups are likely to occur when different groups are in more or less direct competition. As a result, the kind of selection alluded to here [group selection] would be expected to produce individuals that would cooperate intensively and complexly within groups but show strong and even extreme aggressiveness between groups (Alexander 1989 p 463).@ However, in other passages, Alexander clearly minimizes the importance of group selection and attributes the evolution of moral behavior in humans to within-group processes. We provide his most recent statement to this effect: oBecause selection is primarily effective at and below the individual level, it is reasonable to expect concepts and practices pertaining to morality--as with all other aspects of the phenotypes of living forms--to be designed so as to yield reproductive (genetic) gains to the individuals exhibiting them, at least in the environments of history (Alexander 1993 p178).@

At a more technical level, Alexander occasionally seems to appreciate the vehicle concept when evaluating levels of selection (e.g., the 1989 passage quoted above), but more often he implicitly defines anything that evolves as oindividually@ advantageous, even when the group is the vehicle of selection (e.g.,the discussion of Frank, 1988, in Alexander, 1993). We think that a consistent application of our procedure will reveal that Alexander is invoking groups as vehicles of selection much more than he acknowledges in his own writing. We also want to stress, however, that AlexanderYs views on indirect reciprocity, opportunity-leveling within groups and competition between groups remain important within a vehicle-based framework.

21) AlexanderYs theory is conventional in the sense of equating morality with the notion of a common good. However, calling it familiar and conventional does not belittle its importance. Evolutionary theories of human behavior frequently make predictions that correspond closely to folk psychology (e.g., that men tend to value youth in women more than women value youth in men). Since the intuitions of folk psychology are unlikely to be completely wrong, it would be disturbing if an evolutionary theory of human behavior was not familiar and conventional in some sense. Of course, the theory must also go beyond folk psychology by making counter-intuitive predictions, revealing aspects of folk psychology that are false, refining familiar predictions, subjecting predictions to empirical test and so on.

An evolutionary account of morality (including AlexanderYs) does depart from folk psychology in some important respects. The organ-organism- population trichotomy implies that there will always be a level of the biological hierarchy at which social dilemmas will prevail. In other words, moral behavior within groups will frequently be used to generate immoral behavior between groups. This fits well with observed behavior but contrasts with the concept of universal morality that is common in folk psychology and some branches of the human behavioral sciences (e.g., the higher stages of KohlbergYs (1984) theory of moral development; MacDonald 1988). In addition, if moral systems function as group-level rules of meiosis, it becomes difficult to explain the concept of individual rights, which are moral rules that protect individuals from groups. We think that an evolutionary account of morality may ultimately shed light on these topics but it will need to be more sophisticated than current accounts.

22) We provide AlexanderYs most recent statement that humans are motivated entirely by self interest: oIt is not easy for anyone to believe, from his own thoughts about his personal motivation and that of other humans, that humans are designed by natural selection to seek their own interests, let alone maximize their own genetic reproduction (Alexander 1993 p 191-2).@

23) The invisible hand metaphor is the economic equivalent of the Gaia hypothesis. More generally, despite its emphasis on individual self- interest, economic theory is like naive group selection in its axiomatic belief that multi-individual firms maximize a collective utility. A more sophisticated hierarchical approach to economics, along the lines of Campbell 1993, Leibenstein 1976, Margolis 1982, and Simon 1991 will be highly interesting.

24) The sense in which we expect an absence of fitness differences within groups needs to be clarified. In honey bee colonies, there is a set of adaptations that is favored by within-colony selection and has the potential of disrupting colony function. This includes workers laying unfertilized eggs to produce sons and workers favoring their own patriline while tending future queens. These behaviors are seldom observed because of evolved adaptations that prevent them, which qualify as group-level rules of meiosis (Ratnieks 1988). Another set of adaptations is favored by within-colony selection but does not disrupt colony function. For example, a beneficial mutation that increases viability will cause patrilines that have this mutation to be more fit than patrilines that donYt, but there is no reason to expect these kinds of fitness differences to be suppressed by group-level rules of meiosis. Similarly, we expect the Hutterite social organization to suppress fitness differences that correspond to the first set but not those that correspond to the second.

25) Here we are following Tooby and Cosmides (1992) concept of modularity, according to which natural selection has evolved a number of cognitive subsystems that are evoked by appropriate environmental conditions. We do not mean to exclude the possibility of open-ended learning and cultural evolution, however, as envisioned by other evolutionary psychologists (e.g., Boyd and Richerson 1985, Durham 1991, MacDonald 1991).

26) Two caveats are in order here. First, people do not necessarily think the way an ideology exhorts them to think. We think it plausible that Shenker (who was himself an Israeli Kibbutznik) is not simply espousing an ideology but is accurately describing the attitudes and beliefs that exist among members of communal societies. Second, psychological egoism can be defined in many ways and some of the broadest definitions would include the attitudes and beliefs expressed in the Shenker passage. For example, if a Hutterite takes genuine pleasure in helping his group, he might be classified by some as a psychological egoist who is attempting to maximize his pleasure. For the purposes of this discussion, we define a psychological egoist as a person who has a category of oself@ that is separate from the category of oothers@, who acts to maximize perceived self-interest without regard to effects on others, and who does not axiomatically find pleasure in helping others. See Batson 1991 for more detailed discussions of psychological egoism.

27) Although the evaluation of psychological adaptations should be based on design features and their reproductive consequences in ancestral environments, it is still interesting to examine the reproductive consequences in modern environments. The Hutterites have been quite well studied demographically and it should be possible to measure actual fitness differentials within groups.

28) In addition to the environmental situations that we have listed, unstable equilibria leading to majority effects are likely to be important in the evolution and maintenance of group-level adaptations. In other words, group-level adaptations may have difficulty evolving from a low frequency even when they are favored by environmental conditions. Conversely, after they have become established, group-level adaptations may persist even after the environmental conditions that favored them are relaxed (Boyd and Richerson 1990).


Alexander, R. D. (1979). Darwinism and Human Affairs . Seattle: University of Washington Press.

Alexander, R. D. (1987). The biology of moral systems . New York: Aldine de Gruyter.

Alexander, R. D. (1988). Knowledge, intent and morality in Darwin's world. Quarterly Review of Biology, 63, 441-443.

Alexander, R. D. (1989). The evolution of the human psyche. In P. Mellars, & C. Stringer (Ed.), The human revolution (pp. 455-513). Edinburgh: University of Edinburgh Press.

Alexander, R. D. (1993). Biological considerations in the analysis of morality. In M. H. Nitecki, & D. V. Nitecki (Ed.), Evolutionary ethics (pp. 163-196). Albany, N.Y.: State University of New York Press.

Alexander, R., & Borgia, G. (1978). Group selection, altruism and the levels of organization of life. Annual Review of Ecology and Systematics, 9, 449- 475.

Allee, W. C. (1943). Where angels fear to tread: A contribution from general sociology to human ethics. Science, 97, 517-525.

Allison, S. T., & Messick, D. M. (1987). From individual inputs to group outputs, and back again: group presses and inferences about members. In C. Hendrick (Ed.), Group processes (pp. 111-143). Newbury Park: Sage.

Aoki, K. (1982). A condition for group selection prevail over individual selection. Evolution, 36, 832-842.

Aoki, K. (1983). A quantitative genetic model of reciprocal altruism: A condition for kin or group selection to prevail. Proceedings of the National Academy of Sciences, 80, 4065-4068.

Archer, J. (1991). Human sociobiology: Basic concepts and limitations. Journal of Social Issues, 47, 11-26.

Aviles, L. (1986). Sex-ratio bias and possible group selection in the social spider Anelosimus eximius. American Naturalist, 128, 1-12.

Aviles, L. (1993). Interdemic selection and the sex ratio: a social spider perspective. American Naturalist, 142, 320-345.

Axelrod, R. (1980a). Effective choices in the prisoner's dilemma. Journal of Conflict Resolution, 24, 3-25.

Axelrod, R. (1980b). More effective choices in the prisoner's dilemma. Journal of Conflict Resolution, 24, 379-403.

Axelrod, R., & Hamilton, W. D. (1981). The evolution of cooperation. Science, 211, 1390-1396.

Barkow, J. H., Cosmides, L., & Tooby, J. (1992). The adapted mind: evolutionary psychology and the generation of culture. Oxford: Oxford University Press,

Batson, C. D. (1991). The altruism question: Toward a social-psychological answer . Hillsdale, N.J.: Erlbaum.

Bell, G. (1978). Group selection in structured populations. American Naturalist, 112, 389-399.

Boehm, C. (1981). Parasitic selection and group selection: a study of conflict interference in rhesus and Japanese macaque monkeys. In A. B. Chiarelli, & R. S. Corruccini (Ed.), Primate behavior and sociobiology (pp. 161-82). Berlin: Springer-Verlag.

Boorman, S. A., & Levitt, P. R. (1973). Group selection on the boundary of a stable population. Theoretical Population Biology, 4, 85-128.

Boorman, S. A., & Levitt, P. R. (1980). The genetics of altruism . New York: Academic Press.

Boyd, R., & Lorberbaum, J. (1987). No pure strategy is evolutionarily stable in the repeated Prisoner's dilemma game. Nature, 327, 58-9.

Boyd, R., & Richerson, P. J. (1980). Effect of phenotypic variation on kin selection. Proceedings of the national academy of sciences, 77, 7506- 7509.

Boyd, R., & Richerson, P. J. (1982). Cultural transmission and the evolution of cooperative behavior. Human Ecology, 10, 325-351.

Boyd, R., & Richerson, P. J. (1985). Culture and the evolutionary process . Chicago: University of Chicago Press.

Boyd, R., & Richerson, P. J. (1988). The evolution of reciprocity in sizable groups. Journal of Theoretical Biology, 132, 337-356.

Boyd, R., & Richerson, P. J. (1989). The evolution of indirect reciprocity. Social Networks, 11, 213-236.

Boyd, R., & Richerson, P. J. (1990). Culture and cooperation. In J. J. Mansbridge (Ed.), Beyond self-interest (pp. 111-132). Chicago: University of Chicago Press.

Boyd, R., & Richerson, P. (1990). Group selection among alternative evolutionarily stable strategies. Journal of Theoretical Biology, 145, 331- 342.

Brandon, R. (1990). Organism and environment . Princeton: Princeton University Press.

Breden, F. J., & Wade, M. J. (1989). Selection within and between kin groups of the imported willow leaf beetle. American Naturalist, 134, 35-50.

Buss, L. (1987). The evolution of individuality . Princeton, NJ: Princeton University Press.

Camazine, S. (1991). Self-organizing pattern formation on the combs of honey bee colonies. Behavioral Ecology and Sociobiology, 28, 61-76.

Camazine, S., & Sneyd, J. (1991). A model of collective nectar source selection by honey bees: self organization through simple rules. Journal of Theoretical biology, 149, 547-571.

Campbell, D. T. (1958). Common fate, similarity, and other indices of the status of aggregates of persons as social entities. Behavioral Science, 3, 14-25.

Campbell, D. T. (1974). 'Downward causation' in hierarchically organized biological systems. In F. J. Ayala, & T. Dobzhansky (Ed.), Studies in the philosophy of biology (pp. 179-186). New York: MacMillan Press Ltd.

Campbell, D. T. (1979). Comments on the sociobiology of ethics and moralizing. Behavioral Science, 24, 37-45.

Campbell, D. T. (1983). The two distinct routes beyond kin selection to ultra-sociality: Implications for the humanities and social sciences. In D. L. Bridgeman (Ed.), The nature of prosocial development: Interdisciplinary theories and strategies (pp. 11-41). New York: Academic Press.

Campbell, D. T. (1991). A naturalistic theory of archaic moral orders. Zygon, 26, 91-114.

Campbell, D. T. (1993). How individual and face-to-face-group selection undermine firm selection in organizational evolution. In J. A. C. Baum, & J. V. Singh (Ed.), Evolutionary dynamics of organizations New York: Oxford University Press.

Caporeal, L. R., Dawes, R. M., Orbell, J. M., & Van de Kragt, A. J. C. (1989). Selfishness examined: Cooperation in the absence of egoistic incentives. Behavioral and Brain Sciences, 12, 683-739.

Cassidy, J. (1978). Philosophical aspects of the group selection controversy. Philosophy of Science, 45, 574-94.

Cavalli-Sforza, L., & Feldman, M. (1978). Darwinian selection and altruism. Theoretical Population Biology, 14, 268-280.

Chao, L., & Levin, B. (1981). Structured habitats and the evolution of anti- competitor toxins in bacteria. Proceedings of the National Academy of Sciences, USA, 78, 6324-6328.

Charlesworth, B. (1979). A note on the evolution of altruism in structured demes. American Naturalist, 113, 601-605.

Charlesworth, B., & Toro, M. A. (1982). Female-biased sex ratios. Nature, 298, 494.

Charnov, E. L. (1982). The theory of sex allocation . Princeton: Princeton University Press.

Chepko-Sade, B. D., Dow, M. M., & Cheverud, J. M. (1988). Group selection models with population substructure based on social interaction networks. American Journal of Physical Anthropology, 77, 427-33.

Cohen, D., & Eshel, I. (1976). On the founder effect and the evolution of altruistic traits. Theoretical population biology, 10, 276-302.

Colwell, R. K. (1981). Group selection is implicated in the evolution of female-baised sex ratios. Nature, 290, 401-404.

Cook, R. C. (1954). The North American Hutterites: a study in human multiplication. Population bulletin, 10, 97-107.

Cosmides, L. M., & Tooby, J. (1981). Cytoplasmic inheritance and intragenomic conflict. Journal of Theoretical Biology, 89, 83-129.

Cosmides, L., & Tooby, J. (1987). From evolution to behavior: evolutionary psychology as the missing link. In J. Dupre (Ed.), The latest on the best: essays on evolution and optimality (pp. 277-307). Cambridge, Mass: Bradford (MIT Press).

Craig, D. M. (1982). Group selection versus individual selection: an experimental analysis. Evolution, 36, 271-282.

Crespi, B. J., & Taylor, P. D. (1990). Dispersal rates under variable patch selection. American Naturalist, 135, 48-62.

Cronin, H. (1991). The ant and the peacock: Altruism and sexual selection from Darwin to today . Cambridge: Cambridge University Press.

Crow, J. F. (1979). Genes that violate Mendel's Rules. Scientific American, 240, 104-113.

Crow, J., & Aoki, K. (1982). Group selection for a polygenic behavioral trait:a differential proliferation model. Proceedings of the National Academy of Sciences, 79, 2628-2631.

Crow, J., & Aoki, K. (1984). Group selection for a polygenic behavioral triat: estimating the degree of population subdivision. Proceedings of the National Academy of Sciences, 81, 6073-6077.

Crozier, R. H. (1987). Selection, Adaption and Evolution. Journal and Proceedings, Royal Society of New South Wales, 120, 21-37.

Crozier, R. H., & Consul, P. C. (1976). Conditions for genetic polymorphism in social hymenoptera under selection at the colony level. Theoretical Population Biology, 10, 1-9.

Daly, M., & Wilson, M. (1988). Homicide . New York: Aldine de Gruyter.

Damuth, J. (1985). Selection among "species": a formulation in terms of natural functional units. Evolution, 39, 1132-46.

Damuth, J., & Heisler, I. L. (1988). Alternative formulations of multilevel selection. Biology and Philosophy, 3, 407-30.

Darwin, C. (1871). The descent of man, and selection in relation to sex . London: Murray.

Dawes, R. M., Van de Kragt, A. J. C., & Orbell, J. M. (1988). Not me or thee but we: The importance of group identity in eliciting cooperation in dilemma situations: experimental manipulations. Acta Psychologica, 68, 83-97.

Dawkins, R. (1976). The Selfish gene (1st ed.). Oxford: Oxford University Press.

Dawkins, R. (1978). Replicator selection and the extended phenotype. Zeitschrift fur Tierpsychologie, 47, 61-76.

Dawkins, R. (1980). Good strategy or evolutionary stable strategy? In G. W. Barlow, & J. Silverberg (Ed.), Sociobiology: beyond nature/nurture? (pp. 331-367). Boulder, CO: Westview Press.

Dawkins, R. (1982). The Extended Phenotype . Oxford: Oxford University Press.

Dawkins, R. (1989). The Selfish gene (2nd ed.). Oxford: Oxford University Press.

Deneubourg, J. L., & Goss, S. (1989). Collective patterns and decision- making. Ethological and Ecological Evolution, 1, 295-311.

Dennett, D. C. (1981). Brainstorms . Cambridge, Mass: Bradford (MIT press).

Dennett, D.C. (1992) Confusion over evolution: an exchange. New York Review of Books, 40, 44

Dover, G. A. (1986). Molecular drive in multigene families: how biological novelties arise, spread, and are assimilated. Trends in genetics, 2, 159- 165.

Dugatkin, L. A. (1990). N-person games and the evolution of cooperation: a model based on predator inspection behavior in fish. Journal of Theoretical Biology, 142, 123-135.

Dugatkin, L. A., Mesterton-Gibbons, M., & Houston, A. I. (1992). Beyond the Prisoner's Dilemma: towards models to discriminate among mechanisms of cooperation in nature. Trends in ecology and evolution, 7, 202-205.

Dugatkin, L. A., & Reeve, H. K. (in press). Behavioral Ecology and levels of selection: Dissolving the group selection controversy. ,

Dugatkin, L. A., & Wilson, D. S. (1991). Rover: a strategy for exploiting cooperators in a patchy environment. American Naturalist, 138, 687-701.

Durham, W. H. (1991). Coevolution: Genes, culture and human diversity . Stanford: Stanford University Press.

Eberhard, W. G. (1990). Evolution of bacterial plasmids and levels of selection. Quarterly Review of Biology, 65, 3-22.

Ehrenpreis, A. (1650/1978). An Epistle on brotherly community as the highest command of love. In Friedmann (Ed.), Brotherly community: the highest command of love (pp. 9-77). Rifton, N.Y.: Plough Publishing Co.

Eibl-Eibesfeldt, I. (1982). Warfare, Man's indoctinability and group selection. Zeitschrift fur Tierpsychologie, 60, 177-198.

Emerson, A. E. (1960). The evolution of adaptation in population systems. In S.Tax (Ed.), Evolution after Darwin (pp. 307-348). Chicago: Chicago University Press.

Eshel, I. (1972). On the neighbor effect and the evolution of altruistic traits. Theoretical Population Biology, 3, 258-277.

Eshel, I. (1977). On the founder effect and the evolution of altruistic traits: an ecogenetical approach. Theoretical population biology, 11, 410- 424.

Eshel, I., & Montro, U. (1988). The three brothers' problem: kin selection with more than one potential helper: the case of delayed help. American Naturalist, 132, 567-75.

Ewald, P. W. (1993). Adaptation and disease . Oxford: Oxford University Press.

Fagen, R. M. (1980). When doves conspire: evolution on nondamaging fighting tactics in a nonrandom-encounter animal conflict model. American Naturalist, 115, 858-869.

Falconer, D. S. (1981). Introduction to Quantitative Genetics (2nd ed.). London: Longman.

Felbinger, C. (1560/1978). Confession of faith. In R. Friedmann (Ed.), Brotherly community: The highest command of love (pp. 91-133). Rifton, N.Y.: Plough Publishing Co.

Feldman, M., & Thomas, E. (1987). Behavior-dependent contexts for repeated plays of the Prisoner's dilemma. Journal of Theoretical Biology, 128, 297-315.

Findlay, C. S. (1992). Phenotypic evolution under gene-culture transmission in structured populations. Journal of Theoretical Biology, 156, 387-400.

Fix, A. G. (1985). Evolution of altruism inkin-structured and random subdivided populations. Evolution, 39, 928-939.

Frank, R. H. (1988). Passions within reason . New York: W.W. Norton.

Frank, S. A. (1986). Dispersal polymorphisms in subdivided populations. Journal of Theoretical Biology, 122, 303-309.

Frank, S. A. (1986). Hierarchical selection theory and sex ratios. I. General solutions for structured populations. Theoretical Population Biology, 29, 312-342.

Frank, S. A. (1987). Demography and sex ratio in social spiders. Evolution, 41, 1267-1281.

Franks, N. R. (1989). Army ants: a collective intelligence. American Scientist, 77, 139-145.

Gadgil, M. (1975). Evolution of social behavior through interpopulational selection. Proceeding of the national academy of sciences, 72, 1199-1201.

Garcia, C., & Toro, M. A. (1990). Individual and group selection for productivity in Tribolium castaneum. Theoretical and Applied Genetics, 79, 256-260.

Gilinsky, N. L., & Mayo, D. G. (1987). Models of group selection. Philosophy of Science, 54, 515-38.

Gilpin, M. E. (1975). Group selection in predator-prey communities . Princeton: Princeton University Press.

Gilpin, M. E., & Taylor, B. L. (1988). Comment on Harpending and Roger's model of intergroup selection. Journal of theoretical biology, 135, 131- 135.

Goodnight, C. J. (1985). The influence of environmental variation on group and individual selection in a cress. Evolution, 39, 545-558.

Goodnight, C. J. (1990). Experimental studies of community evolution I: The response to selection at the community level. Evolution, 44, 1614-1624.

Goodnight, C. J. (1990). Experimental studies of community evolution II: The ecological basis of the response to community selection. Evolution, 44, 1625-1636.

Goodnight, C. J. (1991). Intermixing ability in two-species communities of Tribolium flour beetles. American Naturalist, 138, 342-354.

Goodnight, C. J., Schwartz, J. M., & Stevens, L. (1992). Contextual analysis of models of group selection, soft selection, hard selection, and the evolution of altruism. American Naturalist, 140, 743-761.

Goodnight, K. (1992). Kin selection in a structured population. American Naturalist, in press,

Gould, S. J. (1980). Is a new and general theory of evolution emerging? Paleobiology, 6, 119-130.

Gould, S. J. (1989). Wonderful life: The Burgess shale and the nature of history . New York: Norton.

Gould, S. J. (1992). The confusion over evolution. New York Review of Books, 39, 47-53.

Gould, S. J., & Lewontin, R. C. (1979). The spandrels of San Marco and the panglossian paradigm: A critique of the adaptationist program. Proceedings of the Royal Society of London, B205, 581-98.

Govindaraju, D. R. (1988). Mating systems and the opportunity for group selection in plants. Evolutionary trends in plants, 2, 99-106.

Grafen, A. (1984). Natural selection, kin selection and group selection. In J. Krebs, & N. Davies (Ed.), Behavioural Ecology: An evolutionary approach (pp. 62-84). Oxford: Blackwell Scientific Publications.

Griesmer, J., & Wade, M. (1988). Laboratory models, causal explanations and group selection. Biology and Philosophy, 3, 67-96.

Griffing, B. (1977). Selection for populations of interacting genotypes. In E. Pollak, O. Kempthorne, & T. B. Bailey (Ed.), Proceedings of the International Congress on Quantitative Genetics, August 16-21, 1976 (pp. 413-434). Ames, Iowa: Iowa State University Press.

Hamilton, W. D. (1964). The genetical evolution of social behavior, I and II. Journal of theoretical biology, 7, 1-52.

Hamilton, W. D. (1991). Selection of selfish and altruistic behavior in some extreme models. In J. S. Eisenberg, & W. S. Dillon (Ed.), Man and beast: comparative social behavior (pp. 57-92). Washington D.C.: Smithsonian Institution Press.

Hardin, G. (1968). The tragedy of the commons. Science, 162, 1243-48.

Harpending, H. C., & Rogers, A. R. (1987). On Wright's mechanism for intergroup selection. Journal of Theoretical Biology, 127, 51-61.

Hausfater, G., & Breden, F. (1990). Selection within and between social groups for infanticide. American Naturalist, 136, 673-88.

Heisler, I. L., & Damuth, J. (1987). A method of analyzing selection in hierarchically structured populations. American Naturalist, 130, 582-602.

Hendrick, C. (1987a). Group processes. Newbury Park: Sage,

Hendrick, C. (1987b). Group processes and intergroup relations. Newbury Park: Sage,

Hofstadter, D. R. (1979). Godel, Escher, Bach: an eternal golden braid . New York: Vintage.

Hogg, M., & Abrams, D. (1988). Social identifications: A social psychology of intergroup relations and group processes . London: Routledge.

Holt, R. D. (1983). Evolution in structured demes: the components of selection. unpublished manuscript, ,

Hull, D. (1980). Individuality and selection. Annual review of ecology and systematics, 11, 311-32.

Hull, D. (1981). The units of evolution--a metaphysical essay. In U. Jensen, & R. Harre (Ed.), The philosophy of evolution (pp. 23-44). Sussex: Harvester Press.

Hull, D. (1988). Science as a process: an evolutionary account of the social and conceptual development of science . Chicago: University of Chicago Press.

Hurst, L. D. (1991). The evolution of cytoplasmic incompatibility or when spite can be successful. Journal of theoretical biology, 148, 269-277.

Jimenez, J., & Casadesus, J. (1989). An altruistic model of Rhizobium- Leguem association. Journal of Heredity, 80, 335-337.

Johnson, M. S., & Brown, J. L. (1980). Genetic variation among trait groups and apparent absence of close inbreeding in Grey-crowned babblers. Behavioral Ecology and Sociobiology, 7, 93-98.

Kelly, J. K. (1992a). The evolution of altruism in density regulated populations. in press,

Kelly, J. K. (1992b). Restricted migration and the evolution of altruism. Evolution, 46, 1492-5.

King, D. A. (1990). The adaptive significance of tree height. American Naturalist, 135, 809-828.

Kitcher, P. (1993). The Advancement of Learning . Oxford: Oxford University Press.

Kitcher, P., Sterelny, K., & Waters, K. (1990). The illusory riches of Sober's monism. Journal of Philosophy, 87, 158-60.

Knauft, B. M. (1985). Good company and violence: Sorcery and social action in a lowland New Guinea society . Berkeley, CA: University of California Press.

Knauft, B. M. (1991). Violence and sociality in human evolution. Current Anthropology, 32, 391-428.

Kohlberg, L. (1984). Essagys on moral development: Vol 2, The psychology of moral development . San Francisco: Harper and Row.

Krebs, D. (1987). The challenge of altruism in biology and psychology. In C. Crawford, M. Smith, & D. Krebs (Ed.), Sociobiology and psychology: ideas, issues and applications (pp. 81-119). Hillsdale, New Jersey: Erlbaum.

Leibenstein, H. (1976). Beyond economic man: a new foundation for microeconomics . Cambridge, Mass: Harvard University Press.

Leigh, E. G. J. (1977). How does selection reconcile individual advantage with the good of the group? Proceeding of the National Academy of Sciences, 74, 4542-4546.

Leigh, E. G. J. (1991). Genes, bees and ecosystems: the evolution of common interest among individuals. Trends in Ecology and Evolution, 6, 257-262.

Levin, B. R., & Kilmer, W. L. (1974). Interdemic selection and the evolution of altruism: A computer simulation. Evolution, 28, 527-545.

Lewontin, R. C. (1970). The units of selection. Annual Review of Ecology and Systematics, 1, 1-18.

Lloyd, E. (1988). The structure and confirmation of evolutionary theory . New York: Greenwood.

Lovelock, J. E. (1979). Gaia: a new look at life on earth . Oxford: Oxford University Press.

MacDonald, K. B. (1988). Sociobiology and the Cognitive-Developmental tradition in moral development research. In K. B. MacDonald (Ed.), Sociobiological perspectives on human development (pp. 140-167). New York: Springer-Verlag.

MacDonald, K. (1991). A perspective on Darwinian psychology: the importance of domain-general mechanisms, plasticity, and individual differences. Ethology and Sociobiology, 12, 449-480.

MacDonald, K. (in prep). Judaism as a group evolutionary strategy .

Mansbridge, J. J. (1990). Beyond self interest. Chicago: University of Chicago Press,

Margolis, H. (1982). Selfishness, Altruism and rationality: a theory of social choice . Chicago: University of Chicago Press.

Matessi, C., & Jayakar, S. D. (1976). Conditions for the evolution of altruism under Darwinian selection. Theoretical Population Biology, 9, 360-387.

Matessi, C., Karlin, S., & Morris, M. (1987). Models of intergenerational kin altruism. American Naturalist, 130, 544-69.

Maynard Smith, J. (1964). Group selection and kin selection. Nature, , 1145-1146.

Maynard Smith, J. (1976). Group selection. Quarterly review of biology, 51, 277-283.

Maynard Smith, J. (1982). Evolution and the theory of games . Cambridge: Cambridge University Press.

Maynard Smith, J. (1982). The evolution of social behavior- a classification of models. In K. C. S. Group (Ed.), Current problems in sociobiology (pp. 29-44). Cambridge: Cambridge University Press.

Maynard Smith, J. (1987). How to model evolution. In J. Dupre (Ed.), The latest on the best: essays on evolution and optimality (pp. 119-131). Cambridge: MIT press.

Maynard Smith, J. (1987). Reply to Sober. In J. Dupre (Ed.), The latest on the best: essays on evolution and optimality (pp. 147-150). Boston: MIT press.

Maynard Smith, J. (1992). Confusion over evolution: an exchange. New York Review of Books, 40, 43.

Mayr, E. (1990). Myxoma and group selection. Biologisches zentralblatt, 109, 453-457.

McCauley, D. E. (1989). Extinction, colonization and population structure: a study of a milkweed beetle. American Naturalist, 134, 365-76.

McCauley, D. E., & Wade, M. J. (1980). Group selection: the genotypic and demographic basis for the phenotypic differentiation of small populations of Tribolium castaneum. Evolution, 34, 813-821.

McCauley, D. E., & Wade, M. J. (1988). Extinction and recolonization: their effects on the genetic differentiation of local populations. Evolution, 42, 995-1005.

McCauley, D. E., Wade, M. J., Breden, F. J., & Wohltman, M. (1988). Spatial and temporal variaition in group relatedness: Evidence from the imported willow leaf beetle. Evolution, 42(1), 184-192.

Mesterton-Gibbons, M., & Dugatkin, L. A. (1992). Cooperation among unrelated individuals: evolutionary factors. Quarterly Review of Biology, 67, 267-281.

Michod, R. (1982). The theory of kin selection. Annual Review of Ecology and Systematics, 13, 23-55.

Michod, R., & Sanderson, M. (1985). Behavioral structure and the evolution of cooperation. In J. Greenwood, & M. Slatkin (Ed.), Evolution - Essays in honor of John Maynard Smith (pp. 95-104). Cambridge: Cambridge University Press.

Mitchell, S. D. (1987). Competing units of selection?: a case of symbiosis. Philosophy of Science, 54, 351-367.

Mitchell, S. D. (1993). Superorganisms: then and now. Yearbook in the sociology of science, ,

Montro, U., & Eshel, I. (1988). The three brothers' problem: kin selection with more than one potential helper: the case of immediate help. American Naturalist, 132, 550-66.

Myerson, R. B., Pollock, G. B., & Swinkels, J. M. (1991). Viscous population equilibria. Games and Economic Behavior, 3, 101-109.

Nagel, E. (1961). The structure of science: problems in the logic of scientific explanation . Indianapolis: Hackett Publishing co.

Noonan, K. M. (1987). Evolution: A primer for psychologists. In C. Crawford, M. Smith, & D. Krebs (Ed.), Sociobiology and Psychology: ideas, issues and applications (pp. 31-60). Hillsdale, New Jersey: Erlbaum.

Nunney, L. (1985). Female-biased sex ratios: invidual or group selection. Evolution, 39(2), 349-361.

Nunney, L. (1985). Group selection, altruism, and structured-deme models. American Naturalist, 126, 212-230.

Nunney, L. (1989). The maintenance of sex by group selection. Evolution, 43(2), 245-257.

Orzack, S., & Sober, E. (in press). Optimality models and the test of adaptationism. American Naturalist

Owen, R. E. (1986). Colony-level selection in the social insects: Single locus additive and non-additive models. Theoretical population biology, 29, 198-234.

Peck, J. R. (1990). The evolution of outsider exclusion. Journal of Theoretical Biology, 142, 565-571.

Peck, J. R. (1992). Group selection, individual selection, and the evolution of genetic drift. Journal of Theoretical Biology, 159, 163-187.

Peck, J. R. (1993). Friendship and the evolution of cooperation. Journal of Theoretical Biology, 162, 195-228.

Peck, J., & Feldman, M. (1986). The evolution of helping behavior in large, randomly mixed populations. American Naturalist, 127, 209-221.

Pollock. (1988). Suspending disbelief--of Wynne-Edwards and his critics. Journal of Evolutionary Biology, 2, 000-000.

Pollock, G. B. (1983). Population viscosity and kin selection. American Naturalist, 122, 817-29.

Pollock, G. (1989). Population structure, spite and the iterated prisoner's dilemma. American Journal of Physical Anthropology, 77, 459-69.

Pollock, G. B. (1991). Crossing malthusian boundaries: Evolutionary stability in the finitely repeated Prisoner's dilemma. Journal of Quantitative Anthropology, 3, 159-180.

Pollock, G. B. (in press). Personal fitness, altruism and the ontology of game theory. Journal of Quantitative Anthropology, ,

Price, G. R. (1972). Extension of covariance selection mathematics. Annals of Human Genetics, 35, 485-490.

Proctor, H. C. (1989). Occurrence of protandry and a female-biased sex- ratio in a sponge-associated water mite (Acari: Unionicolidae). Experimental and applied acarology, 7, 289-298.

Queller, D. C. (1991). Group selection and kin selection. Trends in Ecology and evolution, 6(2), 64.

Queller, D. C. (1992). Quantitative genetics, inclusive fitness and group selection. American Naturalist, 139, 540-558.

Rapoport, A. (1991). Ideological commitments and evolutionary theory. Journal of Social Issues, 47, 83-100.

Rapoport, A., & Chammah, A. (1965). Prisoner's dilemma . Ann Arbor: University of Michigan Press.

Ratnieks, F. L. (1988). Reproductive harmony via mutual policing by workers in eusocial Hymenoptera. American Naturalist, 132, 217-236.

Ratnieks, F. L., & Visscher, P. K. (1989). Worker policing in the honeybee. Nature, 342, 796-797.

Reed, E. (1978). Group selection and methodological individualism--a Critique of Watkins. British journal for the philosophy of science, 29, 256-62.

Richards, R. J. (1987). Darwin and the emergence of evolutionary theories of mind and behavior . Chicago: University of Chicago.

Richardson, R. (1983). Grades of organization and the units of selection controversy. In P. Asquith, & T. Nickles (Ed.), PSA 1982, v2 (pp. 324-340). E. Lansin: Philosophy of Science Association.

Rissing, S., & Pollock, G. (1991). An experimental analysis of pleometric advantage in Messor pergandei. Insect societies, 63, 205-211.

Rogers, A. R. (1990). Group selection by selective emmigration: the effects of migration and kin structure. American Naturalist, 135, 398-413.

Rosenberg, A. (1983). Coefficients, effects and genic selection. Philosophy of Science, 50, 332-38.

Rosenberg, A. (1985). The structure of biological science . Cambridge: Cambridge University Press.

Ruse, M. (1986). Taking Darwin Seriously . New York: Basil Blackwell.

Rushton, J. P. (1989). Genetic similarity, human altruism and group selection. Behavioral and Brain sciences, 12, 503-559.

Sagan, C., & Druyan, A. (1992). Shadows of forgotten ancestors . New York: Random House.

Seeley, T. (1989). The honey bee colony as a superorganism. American Scientist, 77, 546-553.

Seger, J. (1989). All for one, one for all, that is our device. Nature, 338, 374-5.

Shanahan, T. (1990). Group selection and the evolution of myxomatosis. Evolutionary Theory, 9, 239-254.

Shenker, B. (1986). Intentional communities: ideology and alienation in communal societies . London: Routlege.

Sherif, M., Harvey, O. J., White, B. J., Hood, W. R., & Sherif, C. W. (1961). Intergroup conflict and cooperation: The robber's cave experiment . Norman, OK: The University Book Exchange.

Simon, H. A. (1991). Organizations and markets. Journal of Economic Perspectives, 5, 25-44.

Slatkin, M. (1981). Populational heritability. Evolution, 35, 859-871.

Slatkin, M., & Wade, M. J. (1978). Group selection on a quantitative character. Proceedings of the national academy of sciences, 75, 3531-34.

Slatkin, M., & Wilson, D. s. (1979). Coevolution in structured demes. Proceedings of the National Academy of Sciences, 76, 2084-87.

Smith, D. C. (1990). Population structure and competition among kin in the chorus frog (Pseudacris triseriata). Evolution, 44, 1529-1541.

Smith, R. J. F. (1986). Evolution of alarm signals: role of benefits of retaining memebers or territorial neighbors. American naturalist, 128, 604-610.

Sober, E. (1981). Holism, Individualism and the units of selection. In P. Asquith, & R. Giere (Ed.), PSA 1980 v2 (pp. 93-101). East Lansing: Philosophy of Science Association.

Sober, E. (1984). The nature of selection: evolutionary theory in philosophical focus . Cambridge: Bradford/MIT.

Sober, E. (1987). Comments on Maynard Smith's "How to model evolution". In J. Dupre (Ed.), The latest on the best: essays on evolution and optimality (pp. 133-146). Cambridge, Mass: MIT press.

Sober, E. (1990). The poverty of pluralism. Journal of Philosophy, 87, 151- 57.

Sober, E. (1992). The evolution of altruism: correlation , cost and benefit. Biology and Philosophy, 7, 177-188.

Sober, E. (1992). Screening-off and the units of selection. Philosophy of science, 59, 142-152.

Sober, E. (1993a). Evolutionary altruism, psychological egoism and morality--disentangling the phenotypes. In M. H. Nitecki, & D. V. Nitecki (Ed.), Evolutionary ethics Albany: SUNY Press. p 199-216

Sober, E. (1993b). Philosophy of Biology . Boulder, Co.: Westview Press.

Sober, E. (in press). Did evolution make us psychological altruists? in J. Lennox (ed.). Pittsburgh studies in the philosophy of science. University of Pittsburgh Press.

Sober, E., & Lewontin, R. (1982). Artifact, cause and genic selection. Philosophy of science, 47, 157-80.

Sober, E., & Wilson, D. S. (1993). A critical review of philosophical work on the units of selection problem. submitted, ,

Stanley, S. (1975). A theory of evolution above the species level. Proceedings of the National Academy of Sciences, 72, 646-650.

Stanley, S. (1979). Macroevolution: pattern and process . San Francisco: W.H. Freeman.

Sterelny, K., & Kitcher, P. (1988). The return of the gene. Journal of Philosophy, 85, 339-61.

Symons, D. (1992). On the use and misuse of Darwinism in the study of human behavior. In J. H. Barkow, L. Cosmides, & J. Tooby (Ed.), The adapted mind: evolutionary psychology and the generation of culture (pp. 137-162). Oxford: Oxford University Press.

Tajfel, H. (1981). Human groups and social categories . Cambridge: Cambridge University Press.

Tanaka, Y. (1991). The evolution of social communication systems in a subdivided population. Journal of Theoretical Biology, 149, 145-164.

Tooby, J., & Cosmides, L. (1992). The psychological foundations of culture. In J. H. Barkow, L. Cosmides, & J. Tooby (Ed.), The adapted mind: evolutionary psychology and the generation of culture (pp. 19-136). Oxford: Oxford University Press.

Toro, M., & Silio, L. (1986). Assortment of encounters in the two-strategy game. Journal of theoretical biology, 123, 193-204.

Treisman, M. (1983). Errors in the theory of the structured deme. Journal of theoretical biology, 102, 339-346.

Trivers, R. L. (1971). The evolution of reciprocal altruism. Quarterly Review of Biology, 46, 35-57.

Trivers, R. L. (1985). Social evolution . Menlo Park, CA: Benjamin/Cummins.

Uyenoyama, M., & Feldman, M. W. (1980). Evolution of altruism under group selection in large and small populations in fluctuating environments. Theoretical population biology, 15, 58-85.

Uyenoyama, M. K., & Feldman, M. W. (1980). Theories of kin and group selection: a population genetics perspective. Theoretical population biology, 17, 380-414.

Von Schilcher, F., & Tennant, N. (1984). Philosophy, evolution and human nature . London: Routledge and Kegan Paul.

Voorzanger, B. (1984). Altruism in sociobiology: a conceptual analysis. Journal of human evolution, 13, 33-39.

Vrba, E. (1989). Levels of selection and sorting. Oxford surveys in evolutionary biology, 6,

Wade, M. J. (1976). Group selection among laboratory populations of Tribolium. Proceedings of the National academy of sciences, 73, 4604-7.

Wade, M. J. (1977). An experimental study of group selection. Evolution, 31, 134-153.

Wade, M. J. (1978). A critical review of the models of group selection. Quarterly Review of Biology, 53, 101-114.

Wade, M. J. (1979). The primary characteristocs of Tribolium populations group selected for increased and decreased population size. Evolution, 33(2), 749-764.

Wade, M. J. (1982). The evolution of interference competition by individual, family and group selection. Proceeding of the national academy of sciences, 79, 3575-3578.

Wade, M. J. (1982). Group selection: migration and the differentiation of small populations. Evolution, 36, 949-62.

Wade, M. J. (1985). Soft selection, hard selection, kin selection and group selection. American naturalist, 125, 61-73.

Wade, M. J. (1991). Genetic variance for rate of population increase in natural populations of flour beetles, Tribolium spp. Evolution, 45, 1574- 84.

Wade, M. J., & Breden, F. (1980). The evolution of cheating and selfish behavior. Behavioral ecology and sociobiology, 7(167-72),

Wade, M. J., Breden, F. J., & McCauley, D. E. (1988). Spatial and temporal variation in group relatedness: evidence from the imported willow leaf beetle. Evolution, 42, 184-92.

Wade, M. J., & McCauley, D. E. (1980). Group selection: the phenotypic and genotypic differentiation of small populations. Evolution, 34, 799-812.

Walton, D. (1991). The units of selection and the bases of selection. Philosophy of science, 58, 417-35.

Waters, K. (1991). Tempered realism about the forces of selection. Philosophy of science, 58, 553-73.

Werren, J. H. (1991). The paternal sex-ratio chromosome of Nasonia. American Naturalist, 137, 392-402.

Werren, J. H., & Beukeboom, L. W. (1992). Population genetics of a parasitic chromosome: experimental analysis of PSR in subdivided populations. Evolution, 46, 1257-68.

Werren, J. H., & Beukeboom, L. W. (1993). Population genetics of a parasitic chromosome: theoretical analysis of PSR in subdivided populations. American Naturalist, 142, 224-241.

West-Eberhard, M. J. (1981). Intragroup selection and the evolution of insect societies. In R. D. Alexander, & D. W. Tinkle (Ed.), Natural selection and social behavior (pp. 3-17). NY: Chiron Press.

Whitlock, M. C., & McCauley, D. E. (1990). Some population genetic consequences of colony formation and extinction: genetic correlations within founding groups. Evolution, 44, 1717-24.

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

Williams, G. C. (1971). Group selection. Chicago: Aldine,

Williams, G. C. (1986). A defence of reductionism in evolutionary biology. In R. a. M. R. Dawkins (Ed.), Oxforf surveys in evolutionary biology (pp. 1- 27). Oxford: Oxford University Press.

Williams, G. C. (1992). Natural selection: domains, levels and challenges . Oxford: Oxford University Press.

Williams, G. C. (1993). Hard-core Darwinism since 1859. Quarterly Review of Biology, 68, 409-412.

Wills, C. (1991). Maintenance of multiallelic polymorphism at the MHC region. Immunological reviews, 124, 165-220.

Wilson, D. S. (1975). A general theory of group selection. Proceedings of the national academy of sciences, 72, 143-146.

Wilson, D. (1976). Evolution on the level of communities. Science, 192, 1358-1360.

Wilson, D. S. (1977). How nepotistic is the brain worm. Behavioral Ecology and Sociobiology, 2, 421-25.

Wilson, D. (1977). Structured demes and the evolution of group- advantageous traits. American Naturalist, 111, 157-185.

Wilson, D. S. (1978). Structured demes and trait-group variation. American naturalist, 113, 606-610.

Wilson, D. S. (1980). The natural selection of populations and communities . Menlo Park: Benjamin Cummings.

Wilson, D. S. (1983). The group selection controversy: History and current status. Annual review of ecology and systematics, 14, 159-187.

Wilson, D. S. (1983). Reply to Tresiman. Journal of Theoretical Biology, 102, 459-462.

Wilson, D. S. (1987). Altruism in mendelian populations derived from sibgroups: the haystack model revisited. Evolution, 41, 1059-1070.

Wilson, D. S. (1988). Holism and reductionism in evolutionary ecology. Oikos, 53, 269-273.

Wilson, D. S. (1989). Levels of selection: an alternative to individualism in biolog yand the social sciences. Social Networks, 11, 257-272.

Wilson, D. S. (1990). Weak altruism, strong group selection. Oikos, 59, 135-140.

Wilson, D. S. (1992). Complex interactions in metacommunities, with implications for biodiversity and higher levels of selection. Ecology, 73, 1984-2000.

Wilson, D. S. (1992). On the relationship between evolutionary and psychological definitions of altruism and selfishness. Biology and Philosophy, 7, 61-68.

Wilson, D. S. (1993). Group Selection. Key words in evolutionary biology Cambridge: Harvard.

Wilson, D. S., & Colwell, R. K. (1981). Evolution of sex ratio in structured demes. Evolution, 35(5), 882-897.

Wilson, D. S., & Dugatkin, L. A. (1992). Altruism. In E. F. Keller, & E. A. Lloy d (Ed.), Key words in Evolutionary Biology (pp. 29-33). Cambridge Mass: Harvard University Press.

Wilson, D. S., & Knollenberg, W. G. (1987). Adaptive indirect effects: the fitness of burying beetles with and without their phoretic mites. Evolutionary Ecology, 1, 139-159.

Wilson, D. S., Pollock, G. B., & Dugatkin, L. A. (1992). Can altruism evolve in purely viscous populations? Evolutionary Ecology, 6, 331-341.

Wilson, D. S., & Sober, E. (1989). Reviving the superorganism. Journal of Theoretical Biology, 136, 337-356.

Wilson, E. O. (1973). Group selection and its significance for ecology. Bioscience, 23, 631-38.

Wilson, E. O., & Holldobler, B. (1988). Dense heterarchies and mass communication as the basis of organization in ant colonies. Trends in Ecology and Evolution, 3, 65-67.

Wilson, J. B. (1987). Group selection in plant populations. Theoretical and Applied Genetics, 1987, 493-502.

Wright, S. (1980). Genic and organismic selection. Evolution, 34, 825-843.

Wynne-Edwards, V. C. (1962). Animal disperson in relation to social behavior . Edinburgh: Oliver & Boyd.

Wynne-Edwards, V. C. (1986). Evolution through group selection . Oxford: Blackwell Scientific Publications.

Zeigler, B. P. (1978). On necessary and sufficient conditions for group selection efficacy. Theoretical Population Biology, 13, 356-64.

TABLE 1. A guide to the biological literature on group selection since 1970. "T"= theoretical models (including both mathematical and verbal models), "E"= possible empirical examples (including examples that have not been verified by experiments), "F"=field experiments, "L"=laboratory experiments, "R"=literature reviews, "P"=philosophical treatments, "C"=criticisms of group selection interpretations and "H"=papers that are especially relevant from the standpoint of human evolutionary biology.


Figure 1) A nested hierarchy in which every unit is a population of lower- level units. The hierarchy is left open on both ends because genes are composed of subunits and metapopulations can exist in higher-order metapopulations. For example, a valley can be a metapopulation of villages, which in turn are metapopulations of kinship groups.

Figure 2) A vehicle-centered version of kin selection theory. The dominant A-allele codes for an altruistic behavior. The fitness of altruists (WA) and nonaltruists (WS) in a given group is WA=1-c+b(Np-1)/(N-1) and WS=1+bNp/(N-1), where p=the frequency of altruists in the group, N=group size, c=the cost to the altruist, and b=the benefit to the recipient. Both phenotypes have a baseline fitness of 1. Each altruist can be a recipient for (Np-1) other altruists in the group (excluding itself) who are distributing their benefits among (N-1) members of the group (excluding themselves). Each non-altruist can be a recipient for all Np altruists in the group. For this example N=10, c=0.3 and b=1.0. Random mating among the three genotypes (first line) produces six types of mated pairs (second line), which in turn produce groups of siblings (third line). The third line shows only the average sibling group for each type of mated pair. Random sampling of the gametes will produce variation around the averages. Sibling groups vary in the frequency of altruists (fourth line). Altruism is selected against at the individual level because non-altruists have the highest fitness within all mixed groups. Altruism is favored at the group level, however, because group fitness is directly proportional to the frequency of altruists in the group.

Figure 3. Variation among groups in the frequency of altruists. Altruism is coded by a dominant A allele at a frequency of p=0.25 in the metapopulation, yielding a frequency of 0.438 altruists (AA and Aa) and 0.562 nonaltruists (aa). When groups of N=10 are composed of unrelated individuals, the variation in the frequency of altruists between groups has a binomial distribution, as shown by the black curve. Sibling groups are created by a two-step sampling process in which groups of size N=2 (the parents) are drawn from the global population and groups of size N=10 (the siblings) are drawn from their gametes. This two-step sampling procedure increases genetic variation between groups, as shown by the stipled curve, intensifying natural selection at the group level. Evolution within groups always favors the non-altruist, regardless of whether the groups are composed of siblings or unrelated individuals.

Figure 4. Four pay-off matrices that represent A) pure between-group selection, B) a strong conflict between levels of selection, C) a weak conflict between levels of selection (X is the average number of interactions between members of each pair), and D) a return to pure between-group selection. Within-group selection is absent from the first example by virtue of the situation, since coordination has an equal effect on both occupants of the leaf. Within-group selection is absent from the fourth example by virtue of an adaptation, since the "outlaw" A3 type cannot operate in the presence of the "parliament" A5 type.

. . .


. . .