The Princeton Guide to Evolution

Units and Levels of Selection

Samir Okasha

OUTLINE

1. The group selection controversy

2. Kin selection, inclusive fitness, and the gene’s-eye view

3. Species selection

4. Major evolutionary transitions

The levels-of-selection question asks at which level or levels of the biological hierarchy the process of natural selection takes place. After a brief historical introduction, this chapter sketches some of the more important positions in the levels-of-selection debate. Topics discussed include group selection and kin selection, the gene’s-eye view of evolution, species selection, and the major evolutionary transitions.

GLOSSARY

Altruism. A behavior that is costly for the individual that performs it but beneficial to others, where the costs and benefits are measured in units of Darwinian fitness: expected number of offspring.

Genic Selection. Sometimes used to describe any selection process resulting in gene frequency change; also used more narrowly to refer to selection between the genes within the genome of a single organism.

Group Selection. The idea that natural selection can operate on whole groups of organisms, favoring some types of group, or group traits, over others.

Kin Selection. The selection of a behavior because of the behavior’s effect on the reproductive success of the relatives of the organism performing it.

Level of Selection. A level in the biological hierarchy at which natural selection takes place; thus, if selection favors some types of individual in a population over others, selection occurs “at the level of the individual,” and so on.

Multilevel Selection. The idea that natural selection can operate simultaneously at more than one level of the biological hierarchy, for example, at the individual level and at the group level.

Species Selection. Selection acting at the level of species, favoring those species best able to avoid extinction and/or to leave daughter species. Sometimes thought to play an important role in macroevolution.

Unit of Selection. Used by some authors as a synonym for level of selection, or the entity on which natural selection acts, but by others to mean a unit transmitted intact across the generations, also called a replicator.

The levels-of-selection question is a fundamental one in evolutionary biology, for it arises directly from the logic of Darwinian theory. In On the Origin of Species, Darwin argued that if the organisms in a population vary, and if some variants are more successful than others in the “struggle for life,” and if offspring tend to resemble their parents, then evolution by natural selection will occur—over time, the fittest variants will eventually supplant the less fit. It is easy to see that in principle, Darwin’s argument could apply to biological entities other than individual organisms, for example, genes, cells, groups, colonies, or species. Any entity exhibiting variation, differential fitness, and heritability (or parent-offspring resemblance) could in theory be subject to evolution by natural selection, no matter the level of the biological hierarchy occupied by that entity. This possibility is what gives rise to the levels-of-selection question.

The question of levels of selection is intimately linked with the paradoxical problem of altruism, both historically and conceptually. Altruism in biology refers to a behavior that is costly to the individual that performs it but beneficial to others, where the costs and benefits are measured in terms of fitness—that is, survival and reproduction. At first glance, the existence of altruism seems hard to square with Darwinian principles: surely natural selection should lead individuals to behave in ways that benefit themselves, not others? Yet altruism is quite common in nature, particularly among social animals (and has even been found among microbes). For example, sterile workers in social insect colonies devote their entire lives to assisting the reproductive efforts of the queen, foregoing personal reproduction. One possible solution to this paradox, first suggested by Darwin himself, is to invoke selection at the group or colony level, rather than at the individual level. Groups containing many altruists, all working for the common good, might have an advantage over other groups, thus leading to evolution of altruism. Darwin suggested that self-sacrificial behaviors in early hominids might have evolved by this mechanism, an idea still discussed.

The phenomenon of altruism suggests that selection may sometimes operate at levels above that of the individual organism. There is also evidence that selection can operate at levels below the individual, for example, on cells and genes. As early as 1903, August Weismann argued that selection could operate on the hereditary particles within the “germplasm” of an individual, a process he called “germinal selection.” Today, it is quite common to think of mammalian cancer as a form of intraindividual selection, in which some somatic cell lines gain a short-term replicative advantage over others, to the detriment of the whole organism. More importantly, abundant evidence has been found that selection can occur among the genes within a single individual’s genome. This arises because, in sexual species, the genes within an individual are not transmitted en masse to its offspring. So, for example, some genes are able to subvert the rules of fair meiosis and gain access to more than half the host organism’s gametes, a phenomenon known as meiotic drive. Selection of this sort, sometimes called genic selection, leads to a conflict of interest between the genes within a single organism, known as intragenomic conflict (see chapter IV.7).

A note on terminology: the expressions level of selection and unit of selection are sometimes used interchangeably. With this usage, if selection operates at the colony level, for example, it follows that the colony is the unit of selection, and vice versa; however, other authors have distinguished levels from units of selection, using the latter to mean, roughly, entities transmitted intact from one generation to another, or replicators, as they are sometimes called. On this usage, the unit of selection is almost always the gene or allele. To avoid confusion over terminology, the expression “unit of selection” is not used in this article.

1. THE GROUP SELECTION CONTROVERSY

The issue of group-level selection has long been a source of controversy in biology. Darwin himself discussed selection primarily at the level of the individual organism, though he did countenance the possibility of colony- or group-level selection in a few cases. Similarly, the founders of the neo-Darwinian synthesis—R. A. Fisher, J.B.S. Haldane, and Sewall Wright—were concerned primarily with selection acting on individual organisms (however, Wright’s “shifting balance” theory can be interpreted as involving a limited form of intergroup selection). This focus on individual-level selection is easy to understand, for the bulk of the adaptations we see in nature appear designed to benefit individual organisms, not their groups.

Fisher and Haldane stressed that natural selection acting on individuals would not necessarily lead to adaptations beneficial to entities at higher levels. Thus, Haldane gave examples of adaptations that were “biologically advantageous for the individual, but ultimately disastrous for the species,” while Fisher wrote that it was “entirely open” whether selection between individuals would on aggregate benefit or harm the whole species. Nonetheless, by the mid-twentieth century, many biologists routinely assumed that evolution would produce adaptations for the “good of species,” or the “good of the group,” even though the evolutionary mechanism they had in mind was apparently individual-level selection. Thus, for example, the ethologist Konrad Lorenz argued that the submissive displays of weaker animals were a feature designed to benefit the species, by preventing costly conflicts. From a modern perspective the fallacy in Lorenz’s reasoning is obvious, but this is thanks to the considerable wisdom of hindsight.

The “good of the group” tradition came under severe attack in the 1960s and 1970s by biologists such as George C. Williams, John Maynard Smith, and Richard Dawkins. In Adaptation and Natural Selection, Williams (1966) argued that group selection, while logically possible, was unlikely to have been a potent evolutionary force, and would usually be trumped by individual selection, owing to individuals’ having a faster rate of turnover than groups. He also argued that the hypothesis of group selection was unnecessary—the empirical facts could be explained without it. A similar argument by Maynard Smith, that altruism was better explained by invoking kin selection and inclusive fitness theory (see chapter III.4) than group selection, was bolstered by early mathematical models. They seemed to show that group selection would have significant effects only for a limited range of parameter values; more recent work suggests the situation is somewhat more complicated than this, however.

Williams also made an important distinction between group adaptation and what he called fortuitous group benefit. The former refers to a feature of a group that benefits the group and evolved for that reason, while the latter refers to a feature of a group that happens to benefit it but did not evolve for that reason. For example, it is conceivable that sexual reproduction is beneficial for a species, as it increases the amount of genetic variation in the species (compared with parthenogenesis), thus allowing rapid evolutionary response to environmental change. This may well be true, but it is unlikely the reason sexual reproduction evolved in the first place—more likely, the benefit is an incidental side effect. If so, then sexual reproduction does not count as a group adaptation, by Williams’s light, as it did not evolve because of the benefit it confers on the whole species.

As a result of the work of Williams, Maynard Smith, and others, the concept of group selection fell into widespread disrepute in evolutionary biology, where it remained for decades. More recently, this situation has begun to change somewhat, partly because of theoretical developments suggesting that a “trimmed down” version of group selection, often called multilevel selection, might be effective in certain contexts; partly because of experimental work showing an unexpected potency to group selection in the laboratory; partly because of conceptual developments suggesting that group and kin selection are really two sides of the same coin (see below); partly because of recent work on the “major evolutionary transitions” suggesting that group selection may have played a role (see below); and partly because there is some evidence to suggest that selection between competing hominid groups may have been an important factor in the evolution of our own species. The issue continues to be debated, but all parties agree that the naive “good of the group” tradition of the 1950s and 1960s was a flawed way of understanding evolution.

2. KIN SELECTION, INCLUSIVE FITNESS, AND THE GENE’S-EYE VIEW

Kin selection theory, and the associated concept of inclusive fitness, emerged from W. D. Hamilton’s seminal work on the evolution of altruism in the 1960s and 1970s. The basic idea is simple. As noted above, altruism poses a prima facie challenge to Darwinism, because altruists sacrifice their own fitness to help others; so it seems that altruism, and the genes that cause it, should be disfavored by natural selection. Hamilton realized that the logic of this argument breaks down if altruists can direct their benefits toward relatives, rather than toward unrelated members of the population. Relatives share genes, so there is a certain probability that the beneficiary of the altruistic action will itself be an altruist; if so, then the gene for altruism may spread. This idea is encapsulated in Hamilton’s rule, which says that a gene causing an altruistic action will spread so long as b > c / r, where c is the cost to the donor, b the benefit to the recipient, and r the coefficient of relationship between them (roughly, the probability that donor and recipient both inherited the gene from a common ancestor). In other words, altruism can spread by natural selection, so long as the cost to the donor is offset by a sufficient amount of benefit to sufficiently closely related relatives (see chapters VII.10 and VII.13).

The main empirical prediction of kin selection theory is that individuals should behave more altruistically toward relatives than nonrelatives. This broad qualitative prediction has been amply confirmed in diverse species, and in some cases kin selection models enjoy a close quantitative fit with the data. It seems likely that kin selection played a major role in the evolution of the highly cooperative social insect colonies, like honey bees, as the relatedness of the insects in such colonies to each other and to the queen is typically quite high (see chapters VII.10 and VII.13). Though kin selection theory has its detractors, most evolutionists regard it as a major part of the explanation of mechanisms by which social behavior evolved.

How does kin selection theory relate to the traditional levels-of-selection issue? According to one view, kin selection provides an explanation for a way in which altruism can evolve without appealing to the group selection concept, thus undermining the main motivation for the latter. This was the dominant view for many years; however, many recent theorists regard kin selection and modern versions of group selection as essentially equivalent. This view is underpinned by mathematical results showing that it is often possible to “translate” between kin selection and (at least some) group selection models. Thus, in social insect colonies, for example, one can interpret the foraging behavior of sterile workers as an adaptation for helping relatives, or as an adaptation designed to boost the whole colony’s fitness. These explanations may sound different, but mathematically they amount to essentially the same thing. This is intuitive, because a colony is composed mostly of relatives.

Kin selection is often associated with the gene’s-eye view of evolution, which sees phenotypic adaptations as “strategies” designed by genes to help gain an advantage over their alleles in the competition for increased representation in the gene pool (Dawkins 1976, 1982). The gene’s-eye view is a useful way to understand kin selection, and was employed by Hamilton himself in his early papers. Altruism seems anomalous from the individual organism’s viewpoint, but from the gene’s own viewpoint, it makes good sense. A gene is under selection to maximize the number of copies of itself found in the next generation; one way of doing this is to cause its host organism to behave altruistically toward other bearers of the gene. But interestingly, Hamilton showed that kin selection can also be understood from the organism’s point of view. Though altruistic behavior reduces an organism’s personal fitness, it may increase its inclusive fitness (see chapter III.4), defined as an organism’s personal fitness plus the sum of its weighted effects on the fitness of every other organism in the population, the weights determined by the coefficient of relationship, r. Given this definition, natural selection will act to maximize the inclusive fitness of individuals in the population.

In The Selfish Gene, Dawkins argues that all adaptations, not just social behaviors, should be regarded as “for the good of the gene,” since genes are the ultimate beneficiaries of the evolutionary process. This is sometimes summarized in the slogan “the gene is the unit of selection.” Though heuristically valuable, it is important to see that viewing selection in this “genic” way does not resolve the traditional levels of selection question. Whether natural selection occurs at the individual level, the group level, or some other level, the net result will be the spread of one gene at the expense of its alleles—so it is always possible to take a gene’s-eye view of the selection process. Therefore, it does not make sense to oppose the gene’s-eye view to group selection or to individual selection, as Dawkins himself recognized in his later work.

However, there is a quite different sense in which selection sometimes occurs at the genic level. In cases of intragenomic conflict, there may be selection between the genes within the genome of a single individual (see chapter IV.7). The phenomenon of meiotic drive, discussed earlier, illustrates this. In meiotic drive, selection takes place between the two alleles at a single locus in a heterozygote, leading the organism’s gametes to contain one of the alleles in greater proportion than the other. Genes that bias fair meiosis in their favor, and more generally genes that profit at the expense of their host organism, are known as selfish genetic elements. These genes spread by a selection process that can be called genic or gene-level selection; in this sense, the genic level is a distinct level of selection that can be contrasted with individual- and group-level selection. The important point to note is that gene-level selection in this sense is relatively rare; whether it occurs in any particular case is a matter of empirical fact, not perspective.

3. SPECIES SELECTION

It is sometimes suggested by macroevolutionary theorists that natural selection can occur at the species level (see chapter VI.14). Since species leave daughter species, the Darwinian notion of fitness, the expected number of offspring, is applicable to whole species. Conceivably, natural selection might favor some species over others, depending on their species-level characteristics. For example, it has been argued that mollusk species with a large geographic range have a survival advantage over those with smaller ranges, and that geographic range is a heritable trait, passed on from parent species to their offspring. This may explain why the fossil record appears to indicate that average geographic range, in certain mollusk clades, has increased over time. Similarly, it has been proposed that species selection sometimes favors ecological generalists over ecological specialists within the same clade, as the former are less prone to extinction (see chapter VI.14).

That species selection (and lineage-level selection more generally) is a logical possibility is clear, but opinions differ over its empirical importance. The original proponents of species selection, such as Stephen Jay Gould and Niles Eldredge, tried to establish the “autonomy” of macroevolution from microevolution; large-scale evolutionary patterns and trends cannot be understood as the long-term consequences of the within-population evolutionary changes studied by neo-Darwinians, they argued. Rather, macroevolution was governed by irreducible dynamic processes of its own, such as species selection; however, opponents of this “autonomy” thesis have argued that in the most commonly cited examples of species selection, lower-level causal processes, for example, individual selection, are in fact responsible; the differential survival/reproduction of the species is simply a side effect, so selection is not in fact acting at the species level at all. (This point is closely linked to G. C. Williams’s distinction between “group adaptation” and “fortuitous group benefit,” discussed above.) This debate continues today.

Species selection is not simply a higher-level analogue of group selection. In most group selection models, the fitness of a group is defined as the average or total fitness of the individuals in the group. This is because such models have been concerned with explaining the spread of an individual trait, often a prosocial behavior, in a population subdivided into groups. (So it is not assumed that groups must “beget” other groups.) In species selection theory, by contrast, the fitness of a species is defined differently, as the expected number of offspring species that it leaves, a quantity that bears no necessary relation to the average fitness of its constituent organisms. This is because the point of species selection is to explain the changing proportions of different types of species in a clade, not different types of individuals. Hence species selection and group selection, as usually understood, are of different logical types.

4. MAJOR EVOLUTIONARY TRANSITIONS

In the last twenty years, many biologists have become interested in major transitions in evolution, or “evolutionary transitions in individuality” (Buss 1987; Maynard Smith and Szathmáry 1995; Michod 1999). Such transitions occur when a number of free-living biological entities, originally capable of surviving and reproducing alone, become integrated into a cohesive whole, giving rise to a new higher-level entity, and thus an increase in hierarchical complexity. Evolutionary transitions of this sort have occurred numerous times in the history of life; they include the following: individual replicators → networks of replicators; genes → chromosomes; prokaryotic cells → eukaryotic cells with organelles; single-celled organisms → multicelled organisms; solitary organisms → integrated colonies. The challenge is to understand such transitions in Darwinian terms. Why was it advantageous for the lower-level units to come together, sacrifice their individuality, and form themselves into a corporate body? What prevented “cheaters” from selfishly pursuing their own interests and undermining the integrity of the whole?

During an evolutionary transition, the potential exists for selection to act at more than one hierarchical level. Thus, for example, in the transition to multicellularity, the potential exists for selection to act at two levels: between cells within the emerging proto-organisms or cell groups, and between the cell groups themselves. Moreover, in order for cell groups to emerge as genuine organisms, or “evolutionary individuals,” it is necessary for the higher level of selection to dominate; otherwise, selfish cells trying to replicate as fast as possible will undermine the functionality of the cell group. Thus, according to one theory, successful evolutionary transitions require “conflict suppression” or “policing” mechanisms, to ensure the good behavior of the lower-level units, and to align their reproductive interests with those of the whole. In essence, what is required is that the lower-level entities cease to behave as individuals in their own right, and become parts of a larger, integrated whole.

The literature on evolutionary transitions has subtly transformed the original levels-of-selection debate. In the original debate, the existence of the biological hierarchy was taken for granted; the question was about selection and adaptation at preexisting hierarchical levels. By contrast, theorists of the evolutionary transitions are aiming to explain the origin of hierarchical organization itself; that is, why it is that the biological entities we see today form a nested hierarchy. Despite this difference, many themes in the original discussion have proven relevant for understanding evolutionary transitions, including the tension between cooperation and conflict; the importance of kinship in permitting the spread of altruism; the pulling of individual and group selection in different directions; and the suppression of cheaters as a means to promote group welfare. Each of these principles plays a role in recent work on the evolution of multicellularity, for example. The cells within a typical multicelled organism are clonally related, and thus have identical evolutionary interests; sophisticated policing mechanisms exist to prevent rogue cells from undermining the organism’s integrity. This illustrates how ideas originally formulated to explain facets of animal social behavior are applicable much more generally, at many levels of the biological hierarchy.

Finally, the literature on evolutionary transitions shows that the traditional levels-of-selection debate raised issues of real importance. Many biologists regarded the group selection debates of the 1960s and 1970s as rather overblown, arguing that, in practice, selection on individual organisms is the preeminent evolutionary mechanism, whether about other theoretical possibilities. But in light of the evolutionary transitions, this attitude is hard to defend. What we call an “individual organism” is itself a highly cooperative group of cells, each specialized in a different task. Moreover, a eukaryotic cell is itself a multispecies assemblage, as it was formed by the union of two prokaryotic cells, and in addition contains numerous organelles with their own genes, whose evolutionary interests are not always fully aligned with those of their host. Today’s “individual organisms” did not always exist, and were not always the cohesive and integrated entities that they (mostly) are today. Thus “individuality” is a derived trait, something whose evolution we need to explain, not something we can take for granted. Most likely, selection acting at multiple hierarchical levels will constitute an important part of the explanation.