Back to the Syllabus

Levels of Selection

Barry Sinervo©1997


Levels of Selection

Genic Selection

What is the unit of Selection: individuals or genes?

t-allele in Mice

Individual Selection and the "Whole-organism Approach"

Kin Selection and Hamiltons Rule

Social Hymenoptera (Hamilton)

Kin Selection, Inbreeding, and Relatedness in Termites and Naked Mole Rats(***)

Social Selection and Genes for Queen Determination in Ants

Cannabalism in Tiger Salamander Larvae

Genic Selection versus Kin Selection

Group Selection

Reproduction and Density

Back to the t-allele in mice

Artificial Group Selection in Flour Beetles (Wade *** get figures)

Levels Reconsidered

Levels of Selection

The genetic makeup of a population is altered through an interaction with the ecology of the organism. We refer to this interaction as the process of natural and sexual selection. The fundamental premise of Darwinian selection is that natural selection acts on the individual, or more properly, differences in phenotype among individuals within a population. In recent years a number of authors have argued that selection might act at a number of different levels and these levels of selection are loosely structured according to heirarchies of biological organization:

genes -> individuals -> kin -> groups -> species

We can arrange the five levels of selection along a continuum that describes degrees of genetic relationship. Genes are the fundamental unit by which we can define genetic relationships. If a gene is identical by descent to another gene, we refer to a lineal relationship -- at some point in the past, the two genes were derived from a single gene by a duplication event (e.g., meiosis in a sexual organisms). Dawkins has been the most vocal advocate that selection acts not on individuals per se, but on genes. The theory of genic selection is often referred to as the selfish gene theory (see Side Box 4.1 for a discussion of genic selection terminology).

Individual selection is next in the hierarchy and we have given individual selection extensive treatment in Chapter 3. Kin selection is a concept which is closely related to individual selection. Kin are defined as closely related individuals, and it is thought that it might benefit an organism to aid kin in order to promote inclusive fitness of an individual. Related individuals are likely to share many genes in common with a given individual. Asexual clones share all their genes. A sexual individual is comprised of genes from male and female parents. Half of the genes came from one parent and half from the other. Kin tend to share genes. Sibs tend to share 0.5 of alleles at most loci and cousins share 1/8th of their genes as their genetic relationships are further removed.

Members of a group do not necessarily have to share genes, but the group lives or dies as a function of how their genes (and individuals) interact. Group selection a more difficult concept to grasp, is also the subject of an ongoing debate in evolutionary biology. Suffice it to say that selection at the level of groups occurs when a group of individuals produces more groups than other groups. Group selection is a process entirely analogous to individual selection, but acting at the level of groups. Again the traits underlying group selection must be heritable if the species is to evolve by group selection.

Species selection acts at yet a higher level. Members of a species tend to share more genes than two individuals from a different species. Those members of a clade (e.g., a group of closely related species) that produce more species, or perhaps have greater longevity in the earth's history (e.g., lower extinction probability) are going to become overly represented on the planet in much the same way that one allele at a particular genetic locus spreads through a population of individuals and increases in frequency at the expense of alternative alleles.


Genic Selection

Is the unit of selection the gene?

Dawkins and other adherents to the theory of genic selection argue that because selection changes combinations of ofalleles at a genetic locus, the gene or genetic locus must be the fundamental unit of selection. This is the epitome in atomizing concepts of selection. At first glance, this argument seems imminently reasonable. For if natural selection acts to change allele frequencies at a genetic locus, then a purely "reductionist approach" would reduce the problem to the lowest common denominator in all evolutionary theories -- the gene.

The gene is after all the fundamental unit of inheritance. I suppose an even purer reductionist might go further and say that a nucleic acid base pair was even closer to the true unit of inheritance, but that might be going to far. While nucleic acids base pairs comprise the fundamental unit of genetic variation and information, nucleic acids do not form a functional unit. We also want a theory that links selection and inheritance with at least some aspect of the phenotype. Thus since a gene codes for a protein, the protein would be considered the smallest unit of the phenotype that comprises a functional unit. These philosphical issues require a bit of thought beyond this text, and I highly recommend reading the paper by Dawkins (XXXX).

Is the unit of selection the individual organism?

This would be the fundamental tenet of individual selection. Let's explore the concept of individual selection as a review for previous readings. While we measure selection on individual traits and we tend to atomize the phenotype, this is really only a simplification that is necessary from the point of view of statistical power. The individual trait is not the unit of selection. A researcher indentifies a particularly trait that is under selection and tests to see whether the power of selection is statistically significant. In principle, we should analyze each and every trait in an organism and study the correlations between traits to completely describe the evolution of a species.

Brodie's study on garter snake morphology (striped vs spotty) and behavior (straight slither vs reversals) is a classic example of correlated selection. In the study of behavioral traits we must consider selection on morphological and physiological traits that might be functionally (e.g., selectively) or mechanistically correlated (e.g., by proximate mechanisms) with the behaviors. This approach to studying individual selection focusses on the whole organism and is fleshed out in more detail in the next sub-section Individual selection and the whole-organism approach (below).


t-alleles in Mice: a selfish gene

Some populations of mice possess an interesting mutation that distorts the normal meiotic products of male mice. This mutation is located at the mouse's t-locus (Lewontin). Let us denote the wild type allele at the t-locus +. Thus,we represent the homozygous wild type +/+. Let us denote the mutant allele at the t-locus: t. Thus, we label an individual heterozygous at the t-locus t/+ and a homozygote t/t. The effect of the t-allele on the phenotype is as follows:

+/+ produce the normal wild type sperm.
t/+ produce normal sperm, however the ratio of sperm of the t-allele vs +-allele is distorted from 50:50 that one would expect by normal meiotic processes. These t/+ heterozygotes produce 85-95% t-allele sperm and only 5-15% +-allele sperm.
t/t homozygotes are sterile.

Conditions for the origin of the t-allele and the spread of the t-allele are promoted by the process of genic selection. In a curious way, genic selection promotes high frequencies of the t-allele in the population but a high equilibrium t-allele frequency is counter-balanced by individual selection.

1. The origin and spread of the t-allele

Because t/+ heterozygotes produce an overabundance of t-alleles, the t-allele is refered to as a selfish gene relative to the +-allele. When such a mutation arises in a population it should spread rapidly, as a male that carries one copy will produce 85-90% sons that also carry the t-allele. This is the classic notion of a selfish gene. By outcompeting other meiotic products the t-allele nearly guarantees it's successful transmission to the progeny when it firsts arises or is at very low frequency in the population. Contrast this with the probability of spreading an allele under normal mendelian segregation (see Side Box 2.1). The t-allele greatly enhances the chances that the mutation will spread after the mutant arises.

2. What limits the spread of the t-allele

The case of the t-allele is also useful for illustrating a selfish gene. The t-allele spreads and increases in frequency up to the point where a significant number of males in the population are in fact sterile. Thus, without concern for the fitness of the individual (e.g., sterility in some cases), the t-allele marches to very high frequencies and leads to a considerable depression in the mean fitness of males in the population. Even the females that happen to mate with such t-allele infested homozygotes will produce no offspring. Hold onto this last fact as it will become important in subsequent consideration of the t-allele when I discuss group selection.

3. Equilibrium for the t-allele

The t-allele cannot spread beyond a certain frequency or the whole population will end up sterile. The t-allele needs a few +-alleles around in order to propagate. However, when two t-alleles end up in a single individual, the process of individual selection limits the spread of the allele.


Individual Selection and the Whole-organism Approach

The dichotomy between the whole-organism approach and genic selection appears to have roots in the 'How?' versus 'Why?' questions in evolutionary biology that were first discussed by Ernst Mayr. Recall from the first chapter on History and Method in Animal Behavior: Ultimate versus Proximate Mechanisms that I made a distinction between proximate mechanisms and ultimate mechanism. The concept of the gene as a unit of selection indentifies the gene as the source for all proximate mechanistic processes that lead up to the phenotype. It also identifies the gene as the key element in the selective processes that lead to evolution of the phenotype. In constrast, the whole-organism approach considers the atomization of phenotypic traits to the level of the gene unrealistic.

Let's start with the proximate effect of genes. From the discussion on genetic terms in chapter 2, we saw that an allele at a locus does not produce it's effects on the phenotype in isolation. The dominance state on the other chromosome could modify the expression of the allele at the first locus. If the first gene were recessive, and the second gene was dominant, then the first gene would have no effect on the phenotype of the organism. This is an example of a higher order interaction between alternative alleles that is difficult to reconcile if we were to consider the allele the unit of selection. Even the fitness of the t-allele is dependent on the allele on the other chromosome.

Moving up the causal cascade we could also consider the genetic locus as the unit of selection -- or rather the two alleles that comprise a genetic locus. Several problems arise as we move up to this level of selection. First, the expression of the phenotype as a function of this locus may be contingent on the interaction of this locus with other genetic loci in the organisms -- epistasis may influence the outcome. Moreover, the expression of the gene and its effect on phenoytpe is also shaped by an interaction with the environment. Finally, a gene does not just affect one trait, but can affect many organismal traits through pleiotropy. Similarly, complete expression of a polygenic trait may require the additive effect of many genes.

Proximate mechanisms as the source of correlations between traits

All of these genetic effects serve to link the various organismal traits together in a complicated web of genetic and mechanistic interactions. The end result of such interactions are seen as the correlations between traits. For example, the level of testosterone in an organism is correlated with the aggression as well as with muscle mass (just look at a bull and compare it to a steer of similar age). The essence of the whole organism approach is to try to understand the source of such correlations between organismal traits, and how such correlations govern the evolution of the species. For now, realize that the correlations between traits arise from the proximate mechanisms of genetics and the proximate mechanisms of development, physiology and learning. Such proximate mechanisms govern the forces of selection and they might constrain the direction that evolution can take. [For example, is it possible to get an aggressive, testosterone-laden, care-giving male bird? Is aggression incompatible with parental care?]

Natural and sexual selection as the source for the correlations between traits

Another source for the correlation between traits is the process of natural selection. Natural selection might favor specifc alleles at a gene that controls a behavioral trait or at a gene that controls a morphological trait. Such loci are not linked by the mechanisms of genes, development, or physiology. They are linked by the process of selection. The case of stripped versus spotted snakes (Brodie) may be an example where selection bundles up unlinked genes (e.g., genes that control behavior and genes that control morphology). Sexual selection bundles up the alleles for female choice and the allele for the male trait by a similar process. Thus, a correlation between the traits making up the organism also occurs by the process of individual selection.

Neither the interactions of alleles, genes, traits, or the effect of selection can be predicted from the structure of an allele. It is the existence of higher level properties of genes and proximate mechanisms that presents difficulties for the theory of genic selection. We refer to such higher-level features of genes as emergent properties. The effect of genes cannot necessarily be reduced to a description of a single allele or even alleles at a locus, because the emergent properties of the genes arise from higher levels or organismal integration. Have we saturated the kinds of emergent properties of a gene or are there higher-level interactions to be found in kin and group selection?


Kin selection

J. B. S. Haldane clarified the issues underlying the study of kin selection with his clever and illuminating statements regarding its operation (paraphrased):

I would gladly give up my life for two sibs, or 8 first cousins, etc.

J. B. S. Haldane

Embodied in this statement is the notion of a genetic equivalence with an individual's entire genome and an equal number of shared alleles with related individuals. If 1/2 of the genes are shared with sibs, then it would take 2 sibs to make up for the total genetic material in an individual. Certainly, some of the sharing of some alleles would be lacking, because of the probabilistic relationship of relatedness that sibs share (see Side Box 2.4). However, the average across all potential loci that sibs could share would be made up for by duplicate copies of other genes.

W. D. Hamilton captured the essence of kin selection in a simple equation that describes the concept of inclusive fitness. From the point of view of individual selection, what is important is not only the individuals genes, but also the success of closely related individuals. A behavior that is deemed altruistic is one that helps another, but in some way costs the altruist. If the benefits of the the altruistic act outweigh the costs, then selection will favor the behavior. The fitness of a behavior is determined by the direct effect of the behavior on the fitness of the individual, as well as the sumation of indirect fitness effects across all kin that are affected, adjusting for the degree of relationship of the kin. Let's work through Haldane's impeccable logic for a "sib-sacrificial" behavioral act using Hamilton's equation. This argument assumes that the kin were goners unless he intervened.

If Haldane were to sacrifice his own life, the effect of the self-sacrificial behavior on his fitness is W = -1. On the other hand, he saves two sibs, which yields W = +1 for each sib. Each sib has a coefficient of relatedness, r = 0.5. Thus, the indirect effect of the the behavior on his kin = 2 * 0.5 = 1.0, and the net fitness of the behavior of saving two sibs at the cost of his own life is W = -1 + 2 * 0.5 = 0. Hamilton would just break even. If you were to throw in a first cousin (r = 0.125), he comes up positive in the fitness tally. Lets look at a more formal equation that Hamilton devised for inclusive fitness which is expressed in terms of:

  1. W = fitness of a given behavior
  2. d = direct effect of the behavior on the individual
  3. i = indirect effect of the behavior on the kin of the individual
  4. r = coefficient of relatedness.

Inclusive fitness is given by

 W = d + sum(i * r).

 Eqn. 4.1

If d, the direct effect, is deleterious to the invidual, then it is a fitness cost and i should be a fitness benefit to kin for the system to evolve the behavior. The break even point would occur when we set W = 0, in the above equation or and the net fitness d+sum(i*r) should be greater than zero:

0 < d + sum(i * r), 

 Eqn. 4.2

or upon rearranging

 d < sum(i * r),

 Eqn. 4.3

or as long as the costs to indvidual < benefits to kin, the behavior will be favored by kin selection.


Kin Selection and Haplodiploidy in Social Hymenoptera

Hamilton's statements regarding kin selection have a distinct genic selection flavor to them. A behavior can increase in frequency because it favors the fitness of genes per se. Regardless of whether the genes are in the individual or in a closely related kin is irrelevant from the viewpoint of genes. The most striking example of kin selection in the animal kingdom includes those social insects that develop into a sterile worker and allow their sister (the queen) to carry on all reproduction. Such is the case for social hymenoptera: ants, wasps and bees. E. O. Wilson has suggested that such sociality has evolved on at least 11 separate occaissions with this one order of insects. In contrast, only a few examples are found in all the other orders of insects. An unusual aspect of hymenopteran reproductive biology is undoubtedly related to the propensity with which this group of insects evolves kin altruism. Sex determination in the hymenoptera is governed by ploidy of the egg (n or 2n):

  1. eggs that are fertilized develops into a female (2n),
  2. eggs that are unfertilized develops into a male (n),

This form of sex determination, haplodiploidy, leads to some unusual coefficients of relatedness for females and males. For example, a male receives 100% of his genetic material from his mother. Males also produce genetically identical sperm. If a foundress or founding queen of a hive is fertilized by a single male, the daughters from such a union share 0.75 of their genes on average which is considerably higher than the typical 0.50 shared by sibs (see Side Box 4.2). Hamilton has argued that such a high coefficient of relatedness predisposes the hymenoptera to the evolution of kin based altruism.

Consider the actual measures of relatedness in bees relative to the theoretical maximum of 0.75. Many species show estimates of relatedness far less than 0.75 and in some cases it is less than 0.50!

  1. This implies that a single queen mates multiply, and thus the sisters of a colony are really half sibs, or
  2. the queen shares her reproductive roles with other females in the colony.

Thus, haplodiploidy does not invariably lead to high degrees of relatedness.

The elegance of Hamilton's argument is only rivaled by its simplicity. Hamilton illustrated how single proximate factor, haplodiploidy, was ultimately responsibility for an evolutionary predisposition to evolve the most advanced forms of social systems found in the animal kingdom, eusociality. The evolution of eusociality as typified by hymenoptera is the most extreme form of animal society in which members of the colony sacrifice reproductive opportunities for the "common good of the colony". Haplodiploidy and the evolution of eusociality in hymenoptera exemplifies the kind of explanation that we find satisfying in science because it spans a continuum from proximate to ultimate explanations.

While Hamilton's explanation appears to be an adequate explanation of the propensity for hymenoptera to evolve eusocial behaviors, it is not a comprehensive theory as it does not explain how eusociality evolves in other groups that do not have this peculiar form of sex determination. Eusociality is found in other insects that do not have haplodiploidy such as termites, and even in some vertebrates such as the naked mole rat. While eusocial behavior is more rare without haplodiploid sex determination, how is that these groups have evolved eusociality? High levels of genetic relatedness are readily achieved by the haplodiploid mating system and such relatedness appears to predispose the hymenoptera to eusociality. What other mechanisms might predispose a group of kin to high levels of relatedness. It has been hypothesized that genetic relatedness may contribute to the evolution of sociality in these other groups. High genetic relatedness cannot arise by the peculiarities of haplodiploidy, but may arise from inbreeding.

Kin Selection, Inbreeding, and Genetic Relatedness in Termites and Naked Mole Rats

Consider termites which form colonies very similar to ants in their extreme worker specializations. Termites have warrior castes with spray nozzles similar to ants, and there is a single bloated, egg-laying queen in the termite colony that produces all the eggs.

The eusocial system of termites might have been derived via kin selection if individuals breeding in the colony practised brother-sister mating or mother-son matings. Genes become identical by descent across the colony by the process of consanguious matings (see Side Box 4.3: Inbreeding). Consider sib-sib mating versus haplodiploidy. Without inbreeding sibs normally share 1/2 of their genes, but because of the probability of identity by descent, which is 0.25, the probability of shared genetic material is 0.75. This is only after a single generation. Repeated inbreeding events can lead to values for relatedness that rival or even exceed the 0.75 of hymenoptera. If inbreeding is not a problem then inbreeding can readily promote the evolution of eusociality. In a species that has been inbred for a long time, delerious recessive mutations would have been purged from the population long ago, so inbred species may be prone to evolve sociality. All members of a colony have nearly indentical genetic backgrounds.

In addition to consanguinous matings which promote inbreeding, the process of genetic drift can also produce high levels of inbreeding. Genetic drift is the random loss of alleles from a population owing to small population size. As more and more alleles are lost over time, a single allele can increase in frequency and come to predominate the population. In the extreme case, all copies of the allele are idendtical by descent. When colony size is small and colonies do not intermix the circumstances are primed for kin selection.

Naked mole rats have a high degree of relatedness within a colony, and great differentiation among colonies. Such inbreeding might contribute to eusociality.


Social Selection and Genes for Queen Determination in Ants

Perhaps one of the most interesting of all behaviors in the animal kingdom involves the evolution of sociality in which a colony divides labor between sterile workers that maintain the colony and fertile queens that produce all the individuals in a colony. The social hymenoptera form some of the largest and most complex societies in the animal kingdom. Ants and bees can have a number of fertile queens in a colony (polygyne) or a single fertile queen (monogyne). Will the proximate mechanisms of queen determination are varied and can involve a suite of environmental factors (see chapter 14 on development and environment), and at least one case in which a genetic cause has been identified by Kenneth Ross and his colleagues.

The fire ant has become the scourge of the North America since its "invasion" from South America. The particular species of fire ant, Solenopsis invicta, forms polygyne colonies with multiple queens. Egg-laying queens in multiple-queen colonies never possess a particular homozygous genotype (a|a) at the Pgm-3, phosophoglucomutase-encoding gene, reproductive queens are always a|b or b|b genotypes. In contrast, sterile workers and non-reproductive queens can have all three genotypes. What happens to pre-reproductive queens with the a|a genotype which does not permit their reproductive maturity? Pgm-3-a|a queens do not make the transition from non-reproductive to reproductive status because workers kill these queens as the begin to mature and produce oocytes. The workers leave nonreproductive queens with the other two genotypes to mature into reproductive queens. While the proximate cause of the developing a|a queens demise is genetic, selective destruction by the workers forms a more functional cause in the schema of causation developed by Tinbergen (see chapter 1).

Figure 3.X. Genotype proportions at the Pgm-3 locus in introduced (Georgia) and native populations (Argentina) of polygyne or multiple-queen fire ant colonies.

Ross and his colleagues have not ruled out the possibility that another gene which is tightly linked to the Pgm-3 determines the fate of the queen in multiple queen colonies. Not a single reproductive queen from polygyne colonies has been found to be homozygous for the Pgm-3-a allele despite a total of 4600 females that have been genotyped.

Further corroborative evidence that Pgm-3 is the actual gene responsible for selection is provided by patterns seen in monogyne colonies that only have a single queen. Reproductive queens in such single-queen colonies show a pattern of genotypes that is the same as workers and nonreproductive queens. The monogyne form of fire ant does not experience the same kind of worker selection that the polygyne form of fire ant experiences.

Figure 3.X. Genotype proportions at the Pgm-3 locus in introduced (Georgia) and native populations (Argentina) of monogyne or multiple-queen fire ant colonies.


Finally, there is strong directional selection against the Pgm-3-a allele in polygyne colonies, yet it persists at high frequency in the species as whole. Selection against the Pgm-3-a allele in polygyne forms is opposed by gene flow from the monogyne form, in which the allele is common.


Kin Selection and Cannabilism in Tiger Salamander Larvae

Kin selection does not just operate on organisms with eusocial lifestyles. It may operate whenever kin end up in close proximity. Cannabalism is a widespread trait in the animal kingdom. Where you find large monotypic stands of a species and when food becomes limiting, the best item on the menu might just be a conspecific. David Pfennig reared the larvae of the cannabalistic Tiger Salamander in various kin groups.


  1. As a control, he reared unrelated individuals in the same tub
  2. He reared half-sibs in the same tub (0.25 relatedness)
  3. He also reared full sibs in the same tub (0.50 relatedness)

He then scored the probability of cannabalism and compared this among his experimental treatments. If the salamanders had sibs in their tubs the probability of cannabilism was less than that for half sibs and the probability of cannabilism for half sibs was dramaatically lower than tubs with unrelated individuals. These results suggest that: 1) the larvae have kin recognition, 2) kin recognition alters the behavior of cannabalism, and 3) presumably a tadpole that does not eat kin will be favored by kin selection by inclusive fitness.

Pfennig also suggested that recognition of self or more properly kin, may arise from the distinct food ingested by aggregations of tadpoles. The tadpoles aggregate after hatching and a group of related tadpoles might stick around long enough to pick up a common set of odors that they share with related kin.


Genic, Individual, and Kin Selection

In the example of t-alleles in mice we saw that genic selection which favored the production of t-sperm over wild-type sperm was counterbalanced by individual selection that favored homozygous wild type or heterozygous t-allele individuals over homozygous t-allele individuals. Is it also possible for genic selection to be counterbalanced by the action of kin selection? While we do not have concrete examples of such a process it is certainly easy enough to envision with a simple hypothetical example. Consider two parental birds that are rearing a brood of young. Let's assume that a novel "greedy" arises in the nest that causes a chick to eat all the resources at the expense of the other nest mates. While the individual that houses the greedy allele fledges successfully, the other nest mates are left wasting away and die. The greedy gene has a tremendous genic selection advantage. One copy of this hypothetical allele is enough to ensure its short-term spread in the nest (albeit at the expense of its sib mates). However, nests with the greedy gene produces far fewer fledglings than nests which are comprised of pure wild type individuals. Those kin groups which have no greedy gene have a fecundity advantage relative to the kin groups where the greedy gene is present.

While this example involves genic selection and kin groups, it is a simple matter to extend the analogy to groups of individuals who are not genetically related. Can an altruistic gene spread in such groups or will greedy non-altruistic genes spread like wild fire? Answers to this question will require additional theoretical development, which is reserved for the chapters on Social Evolution. However, it is clear that the action of genic selection can be counterbalanced by the action of individual or kin selection, and the action of individual selection can be counterbalanced by the action of kin selection. The additional possibility of social selection as seen in fire ants in which nonreproductive castes alter the fitness outcome of reproductive castes poses additional challenges for the study of levels of selection. Even though workers are nonreproductive their impact on the fitness of the a-allele within a colony is tremendous. Are there additional emergent properties of genes that must considered at the level of the group selection?


Group Selection and Self-limiting Behaviors

"Classic" group selection stems from arguments that Wynne-Edwards made in the 1960's. Wynne-Edwards was arguing for the evolution of a form of population regulation by the mechanism of group selection in which behaviors evolve because they benefit the groups existence. Wynne-Edwards argued that animals might evolve mechanisms that limit reproductive effort for good of species and the arguments are outlined as follows:

  1. avoid stripping resource base and being exterminated by evolving some kind of self-limiting behavior
  2. difference in groups fitness, not individual fitness are what lead to the evolution of such self-limiting behaviors
  3. group benefit takes precedence over individual benefit
  4. the driving force underlying such group selection pressure is variation in rate of increase or extinction among groups
  5. this behavior has a genetic basis (e.g., in the individuals)
  6. in the case of self-limiting behaviors, those groups that evolve individual behaviors that are self-limiting from the groups perspective are the groups that will have a lower probability of extinction.

Charles Darwin was not immune to such conjectures as he wrote:


The group selection advocated by Wynne-Edwards is a form of altruistic behavior in which individuals limit their own reproductive success to benefit the group. The key distinction in this case is that the members of the group are not kin, or the argument might work as kin selection rather than group selection. The group selection argument requires a special kind of altruism towards non-kin to evolve -- true altruism not kin altruism.

The simple problem with group selection is that it can always be invaded by an individual "cheater" strategy. An individual that does not live by the "group rules" and does not limit their own behavior will gain resources relative to other members of the group, get higher fecundity, and thus spread by virtue of enhanced fitness.

Thus, true altruism not an evolutionary stable strategy (ESS) because it can be invaded by a selfish morph. What is an evolutionary stable strategy or ESS? An ESS is a strategy that is so good that it is uninvadable by any (and all) mutant strategies that arise in the population. Because altruism can be invaded by cheaters, altruism is not stable over the long haul. The idea of an ESS was originated by Maynard Smith and Price, and we will use the concept over and over again in subsequent chapters. It is particularly useful in understanding the evolution of behavioral strategies that involve an interaction between two players -- an evolutionary game. When the the fitness of an individual depends on the strategy or frequency with which alternative strategies occur in the population, ESS is quite useful to analyze the long-term evolutionary outcome or equilibrium state.

In the case of group selection, those individuals that are playing the group strategy (e.g., self-limiting) will always do worse than the selfish individual that does not play by the rules. These cheater individuals will produce more progeny at the expense of the other members in the group that are self-limiting. Once the mutant arises, it spreads through the group rapidly.

Another problem with group selection is the rapidity with individuals reproduce compared to the turnover time involved in group selection. It takes far longer for groups to go extinct or for one group to overwhelm the selective advantage of individual selection.

Evolutionary biologists and behavioral ecologists do not say that group selection is impossible, but rather that the conditions required for group selection would rarely be met in natural populations. Groups would have to be very small, and the selective advantage among groups enormous to overwhelm the ever present force of individual selection. Interestingly, the t-allele in mice illustrates how group selection would work. It is important to realize that group selection is working against genic selection in the following example.


The t-allele Revisited

Lewontin calculated that the t-allele should have a phenomenal advantage over the +-allele (wild type) but it would not completely eliminate it. The equilibrium or expected frequency based on the balance between genic selection which favors the t-allele and individual selection that removes homozygous sterile individuals from the breeding pool is t = 0.70. However, the observed frequency in populations was t = 0.36. Why was there such a large discrepancy between expected and observed frequencies?

The spread of the t-allele is in fact limited by the group structure and behavior of mice. Mice live in small demes with usually 2 males, and a few females. The t-allele should increase in frequency to well beyond 70% in the population. With only two males in each group there is a really good possibility that some groups will have no fertile males and other groups will have fertile males.

Table 4.1. Possible group structure for the t-allele.

 1. these groups breed true for +/+, or +/t:
  • +/+ and +/+
  • t/t and +/+
 2. these groups have a t-allele advantage:
  •  +/t and +/+
  • +/t and +/t
  • +/t and t/t
 3. this group is sterile and goes extinct
  •  t/t and t/t

Groups that breed true for +/+ spawn off more groups that breed true and are more resistant to extinction relative to t-allele "infected" groups. The groups that have a t-allele will produce progeny with a greater incidence of the t-allele. Genetic drift, the random sampling of alleles might actually push some of the spawn of these groups over the top and they go to the "t/t and t/t" sterile state. Any sterile group goes extinct. Thus, t=alleles are being continually removed by a group extinction process. Groups that breed true for the wild type +-allele, have a strong group advantage, even over the groups in which the t-allele has an advantage. Eventually the t-allele will increase in frequency in such t-allele infested groups to the point where the spawn of such groups fall into the sterile category. In the case of the t-allele, group selection balances out the effects of genic selection and it leads to a lower frequency of the t-allele than one would expect if genic selection was the only force acting on the population.


Artificial Selection for "Group Traits"

A final example will serve to illustrate how strong group selection must be to overwhelm the force of individual selection. Michael Wade raised lines of Tribolium, flour beetles, under conditions conducive to group. Rather than propogate the lines by choosing the individuals which fared best under high and low density, he propogated lines from those collective groups that produced the highest density colonies. This differs from a classic artificial selection experiment in that one would choose those individuals in each group that performed the best in artificial selection. By choosing entire groups Wade was selecting a group that had the ensemble of genetic attributes that were conducive to self-limiting behaviors. He also raised other lines in which individual selection would be expected to predominate.

Over the long haul, he was successful in altering the "self-limiting capabilities" of the beetles and his lab experiments proved that group selection could alter the evolution of the species. But he was also careful to point out that there are very stringent conditions on the strength and efficacy of group selection. It helps to have small groups, rapid group generation time, and very strong group selection to overwhelm the force of natural selection. Wade also described how such group selection is related to kin selection and cannabilism.


Levels Reconsidered

Convincing cases of genic selection for some traits appear to be quite plausible. Should we consider behavioral traits subject to replicator selection? Tallying up the proportion of shared alleles, without regard to specific traits coded for by those alleles, as is the case for kin selection has a genic selection flavor to the arguments. Concepts of kin selection are central to understanding issues of cooperation which we will treat in greater detail in chapters on social evolution. On the other hand, individual selection envisioned by Darwin has great explanatory power when considering suites of whole-organism traits. In such cases, the emergent properties of genic interactions argue for phenotypic selection as the appropriate level to use when considering the interaction between behavioral traits and the rest of the phenotype.

Whether or not we should reduce all the levels of selection to the replicator is indeed debatable or perhaps even a matter of personal taste. Perhaps an eclectic approach to the issue of level of selection might be the most appropriate. Use theories of individual selection as a default hypothesis, and only consider genic, kin, or group selection as alternative hypotheses when arguments based on natural or sexual selection are not a satisfactory explanation of behaviors.


Back to the Syllabus