The Princeton Guide to Evolution

III.1

Natural Selection, Adaptation, and Fitness: Overview

Stephen C. Stearns

OUTLINE

1. Natural selection explains adaptation

2. Concepts are tools

3. Definitions and complications

4. Fitness and units of selection

5. Connecting selection to fitness in hierarchies

6. Polished adaptations or rough history

7. Adaptationist storytelling

8. How to recognize selection and adaptation

9. Can we do without these concepts? Absolutely not.

This chapter defines natural selection, adaptation, and fitness, the core concepts in the process driving an important part of evolutionary change, discusses how they relate to each other, and comments on their appropriate and inappropriate use.

GLOSSARY

Density-Dependent Selection. Selection that favors different genotypes or phenotypes at different population densities.

Frequency-Dependent Selection. A mode of natural selection in which either rare types (negative frequency-dependent selection) or common types (positive frequency-dependent selection) are favored.

Genotype. The information stored in the genes of one individual; it can refer to anything from one gene to all the genes, depending on context.

Group Selection. Selection generated by variation in the reproductive success of groups.

Individual Selection. Selection generated by variation in the reproductive success of individual organisms.

Interactor. The organism in its ecological role, in which it develops, grows, acquires food and survives, mates, and reproduces.

Kin Selection. Adaptive evolution of genes caused by relatedness; an allele causing an individual to act to benefit relatives will increase in frequency if that allele is also found in the relatives and if the benefit to the relatives more than compensates the fitness cost to the individual.

Life History Traits. Traits directly associated with reproduction and survival, including size at birth, growth rate, age and size at maturity, number of offspring, frequency of reproduction, and life span.

Maladaptation. A state of a trait that leads to demonstrably lower reproductive success than an alternative existing state.

Phenotype. The material organism, or some aspect of it, as contrasted with the information in the genotype providing the blueprint for the organism.

Replicator. The organism in its role as information copier, the mechanism that copies the DNA sequence of the parent and passes it to the offspring.

Trade-off. An evolutionary change that increases fitness in one trait or context but causes a decrease in fitness in another trait or context.

1. NATURAL SELECTION EXPLAINS ADAPTATION

The astonishing precision and elegance with which complex biological structures efficiently function calls out for explanation. The human eye can detect a single photon, allowing it to see a match lit at a distance of 10 miles on a dark night. The nostrils of a migrating salmon can discriminate a difference in concentration of the molecules characteristic of its home stream corresponding to one molecule more in one nostril than the other. The ears and brain of an echo-locating bat can detect a returning echo less than a millionth as loud as the cry it emitted milliseconds earlier and from that information decipher the position, movement, and surface texture of a rapidly and erratically flying insect. Darwin’s greatest idea, natural selection, explains the evolution of the myriad of such astonishing adaptations through the operation of simple mechanisms that can be readily observed. A triumph of human thought, it organizes and explains much of biology, strongly contributes to the impact of biology on other disciplines, and as a major scientific principle not contained in chemistry or physics, elevates biology to their rank in its power to explain the natural world.

2. CONCEPTS ARE TOOLS

Natural selection, adaptation, and fitness are concepts used by biologists to organize their understanding of evolutionary processes and to facilitate their communication; like all scientific concepts, they are tools invented to describe aspects of reality. Here I first define selection, adaptation, and fitness. I then note complications introduced by sex, traits, age structure, relatives, and frequency dependence; the complications call attention to important issues. After comparing definitions of fitness and the units of selection to which they apply, I discuss things on which selection can operate in principle but so inefficiently that it produces rough history rather than polished adaptation. I recount the controversy over adaptationist storytelling; and then discuss how to attenuate that controversy by demonstrating selection and adaptation. I conclude by reemphasizing the stature of these major ideas.

3. DEFINITIONS AND COMPLICATIONS

Natural selection is a process of sorting by reproductive success that occurs in populations of replicating units, whether those units are molecules, cells, organisms, or larger units. Four conditions—all necessary and together sufficient—state when natural selection on a trait will occur and elicit a response:

1. The units expressing the trait must vary in their reproductive success.

2. The trait must vary among the units in the population.

3. The correlation between the trait and reproductive success must be nonzero.

4. For a response to occur, the trait must be heritable.

When these conditions hold, the differential reproductive success of the units expressing a heritable trait correlated with reproductive success will change the frequencies of the states of the trait in the population from one generation to the next. Those positively correlated with reproductive success will increase, and those negatively correlated will decrease in frequency. Natural selection can work in populations of anything that satisfies the conditions, not just populations of organisms. Among other units that have been proposed to experience natural selection are groups, species, words, and ideas.

Note that the word selection is misleading, for nothing actively selects. The action of “selection” is caused by whatever contributes to the correlation of a trait with reproductive success, a set of many things that have inspired much research.

These four conditions can be more simply expressed as heritable variation + differential reproductive success = natural selection. While its simplicity is attractive, that shorter definition misses two important points. First, it neglects the key role played by the correlation between the trait and reproductive success. If that correlation is zero, the trait (or gene) drifts aimlessly; it does not change systematically. Thus, the simpler definition misses the point that the conditions for selection and random drift are quite similar, differing only in the value of the correlation with reproductive success. Second, the simpler definition ignores the important distinction between selection and the response to selection. Drawing that distinction calls attention to the two main steps in the process, the two things that need to be measured to establish that selection is occurring and eliciting a response: the correlation of the trait with reproductive success, and its heritability.

Adaptation is both a process and a state. As a process, adaptation describes the portion of evolutionary change in a trait that is driven by natural selection. (There are other reasons for inherited evolutionary change, genetic drift being an important one.) As a state, adaptation describes that aspect of the current condition of a trait that can be reliably ascribed to the past action of natural selection. Although traits are often loosely referred to as “adaptations,” what is meant more precisely is not the entire trait, which usually has a long complex history and is a mosaic product of several processes, but that aspect of the trait that has been produced by natural selection. Identifying that aspect is one aim of the adaptationist research program.

Fitness is a word meant to capture a measure of relative reproductive performance that can be used to predict long-term dynamics. If heritable, traits that confer greater fitness increase in a population subject to natural selection; those that confer less fitness decrease. The meaning of fitness is context specific in ways that illuminate several major issues in conceptualizing evolutionary change. Among the factors defining those contexts are sex, traits, age structure, spatial structure, relationships, and hierarchies.

First Complication: Sex

In an idealized population of asexual organisms reproducing by binary fission, fitness is relatively easy to define. Because every reproductive act results in two daughter cells, there is no variation in the population in number of offspring per reproductive cycle. What can vary are two components of fitness—the time it takes a cell to divide, and the probability that a daughter cell will survive—and the genetic constitutions of the daughter cells, which differ only through mutations. If survival is fixed, then fitness reduces to a measure of cell cycle time; if cell cycle time is fixed, then fitness reduces to probability of survival per cell cycle; a combined composite measure allows projection of the frequencies of performance variants. Here the unit of replication is the entire asexual genome; the unit of interaction is the entire asexual cell; and any changes in performance introduced by a mutation are linked to the entire genome and expressed in the entire cell.

Sex introduces the complication of recombination, which in eukaryotes is caused both by the independent segregation of chromosomes and by crossing-over within chromosomes (see chapter III.10). Sex has two important consequences for fitness: mutations in one gene are no longer linked to the entire genome but to a local region within a chromosome whose size diminishes through generations of crossing-over, and effects on performance are not expressed in one genetic background but in the many genetic backgrounds of the recombinant descendants. Sex thus uncouples the unit of replication—the gene—from the unit of interaction—the organism. It also associates the consequences of a genetic change—a mutation—with average effects on recombinant descendants rather than with cloned daughters that inherit the rest of the genome with its gene-gene interactions intact. This is one reason population geneticists chose to conceptualize evolution as changes in gene frequencies and to associate selection coefficients with alleles. Population genetics is in part an analytic reaction to the existence of sex.

Second Complication: Traits

Organisms can be analyzed into traits produced by a developmental map linking genotypes to phenotypes. Focusing on traits rather than whole organisms unveils a rich structure demanding special analysis. From the perspective of the gene, one gene often affects two or more traits, a pattern called pleiotropy; from the perspective of the trait, one trait is often influenced by more than one gene, a pattern called epistasis; each captures a different aspect of the web of connections between genes and traits. From the perspective of the whole organism, traits are connected in relationships that result in trade-offs, a term that includes the genetic effects of epistasis, pleiotropy, and linkage and adds to them physiological effects mediated by hormones, energy allocation, and signaling conflicts. Trade-offs represent the functional relationships among traits that result in a particular class of correlated responses to selection; a trade-off exists when an evolutionary change that increases fitness in one trait or context causes a decrease in fitness in another trait or context. In one widely used analytical framework that neatly expresses the complications introduced by traits, the linkages among traits are represented as the off-diagonal elements of genetic and phenotypic variance-covariance matrices, and the correlations of the traits with fitness are represented as a vector, the selection gradient (see chapter III.5).

The objective existence of traits is supported by the genetic observation that pleiotropy is not homogeneously distributed over the genome but is organized into sets of traits, that is, modules within which genes have strong pleiotropic effects and among which they have weaker or even no such effects.

Traits whose combination yields a fitness measure are called components of fitness or life history traits; the most important and frequently used components of fitness are age-specific birth and survival rates. All traits are related to the components of fitness, both mechanistically through the functional connections of genetics, development, morphology, and physiology, and statistically through their correlational impact on survival and reproduction, a level at which ecology and behavior also play a role. The key trade-off between reproduction and survival that shapes the evolution of life span and aging is also called the cost of reproduction (see chapter III.11).

Thus a focus on traits introduces three essential concepts with many consequences: first, the idea that all traits are involved in trade-offs, implying that evolutionary improvements in one trait are linked to evolutionary costs in others; second, the idea that fitness is a composite measure that summarizes the contributions of many traits to reproductive performance; and third, the idea that combinations of traits can yield synergistic fitness effects; for example, light bones and wings are of much less use by themselves than they are in combination.

Third Complication: Ages, Stages, and Sites

There is good reason to introduce the complications of age-, stage-, and site-specific birth and death rates to the analysis of fitness and adaptation: organisms reproducing or surviving at various ages and stages and at different sites can differ in their contributions to future generations, with two very important consequences: aging and maladaptation, which are, in this view, analogous (see chapter III.11).

In any population that is growing or stable, reproduction earlier in life contributes more to future generations than does reproduction later in life, for there is always a risk of dying between the earlier and the later reproductive events, and surviving offspring produced earlier in life will start contributing to future generations through their own reproduction before those produced later in life can do so. That idea, combined with the recognition of trade-offs between reproduction, growth, and survival, yields the theory of the evolution of life histories and aging and with it our understanding of why all living things must grow old and die, why some organisms are large and others small, why some mature earlier and others later, and why some have many offspring and others only a few (see chapter III.11).

Similar insights come from the recognition of spatial heterogeneity in populations. In any population that is distributed among geographical sites, some sites are usually sources and others, sinks. Conditions in sources support excess reproduction that generates emigrants, whereas conditions in sinks do not permit reproduction adequate to maintain the population within the sink, which continues to exist only because of immigration from sources. Organisms will be adapted to sources and maladapted to sinks; this has important implications for both basic science and conservation biology (see chapter III.14).

Just as aging is a by-product of selection for reproductive success early in life, so maladaptation of organisms in sinks is a by-product of selection for reproductive success in sources.

Fourth Complication: Relatives (Kin Selection and Inclusive Fitness)

Because both sexual and asexual organisms have relatives, both genetic systems share a basic property: they encounter situations where actions taken by the focal organism have consequences for the probability that a relative will contribute genes to future generations. Since the same gene can be transmitted to the next generation either through the reproduction of the focal organism or through the reproduction of its relatives, how relatives behave toward each other is one path to reproductive success. Recognition of this fact, and development of its consequences, led to the ideas of kin selection and inclusive fitness, which have yielded many insights into altruism, cooperation, and conflict (see chapter VII.9).

Consider a focal organism, one of its relatives, and an act with fitness consequences for both. A new mutation that affects the probability of that act will invade the population if the benefits gained by the relative (b: its increase in reproductive success as a consequence of the act), multiplied by the coefficient of relationship between the focal organism and the relative (r: the probability that the relative has a copy of the gene that is identical by descent from a shared ancestor), is greater than the cost to the focal organism (c: its decrease in reproductive success as a consequence of the act): that is, if b×r > c. This can also be written as b×r > c×1; that is, the benefit to the recipient, weighted by the coefficient of relationship to the donor, must be greater than the cost to the donor, weighted by its coefficient of relationship to itself, which is 1.

Fifth Complication: Frequency and Density Dependence

Often the fitness of genetic or phenotypic variants depends on the frequencies of the other variants present in the population. Here the tool of choice is game theory, and the questions to ask are, under what conditions will a variant increase when rare, thus invading the population, and under what conditions will a common variant resist invasion by all other variants, persisting in a state of evolutionary stability? (See chapters III.3 and III.5.)

Two examples suggest the importance and scope of frequency dependence. First, what is the stable ratio of males and females in a randomly mating monogamous population? If the population consists mostly of females, males are favored, for each of the rare males will have several mates, whereas each female will have at most one. If the population consists mostly of males, females are favored, for each of the rare females will find a mate, but some males will not mate at all. If offspring sex ratio is heritable, the evolutionary equilibrium will be equal frequencies of males and females: a 50:50 sex ratio. Second, what is the stable frequency of a host genetic variant that confers resistance to a specific pathogen genotype? When that variant is rare, it will increase in frequency, for the commoner variants suffer from infection by the pathogen. As the host variant becomes common, pathogen variants are selected that improve the pathogen’s ability to infect that host genotype, and its spread is halted or reversed. Here being rare is advantageous, and being common is costly. The prediction is that the number of genetic variants conferring resistance will increase until there are so many of them that even the commonest is effectively rare.

Negative frequency dependence—the advantage of being rare and the cost of being common—efficiently maintains variation in populations. Here there is no single best solution: many rare variants persist stably.

Often the fitness of genetic or phenotypic variants depends on population density. Many things change with population density, among them the importance of traits mediating intraspecific competition, of traits conferring resistance to predators and diseases, and of traits that allow communication with potential mates over short versus long distances. When population densities fluctuate regularly from low to high, traits can experience alternating and sometimes conflicting selection pressures. At low densities many organisms tend to grow rapidly to large size, maturing early and having many offspring. At high densities many organisms tend to grow slowly to smaller size, maturing later and having fewer offspring.

Frequency and density dependence are widespread. Their consequences for population dynamics and community ecology are being analyzed by the growing field of adaptive dynamics, which explicitly considers the effects of population dynamics on fitness and the feedbacks between ecological and evolutionary processes.

4. FITNESS AND UNITS OF SELECTION

If life had remained completely asexual, genes—the replicators—would have remained consistently associated with organisms—the interactors—and there would be little occasion to wonder whether genes or organisms are the units on which selection operates. The evolution of sexual reproduction, which enabled genes to move horizontally into many lines of descent, produced the conditions under which kin selection can be most readily detected. While kin selection does operate in asexual lines of descent, for example, selecting for programmed cell death in the development of multicellular organisms, this was seen only retrospectively after its effects had been recognized in situations where degrees of relationship differed more dramatically.

The realization that it was useful to conceptualize the evolutionary process as driven by differences in the fitness of genes led to ideas like the selfish gene, with support coming from successful analyses of altruism and cooperation in terms of kin selection. Two things need to be said about that. First, that selection acts to increase the copy number of genes in populations rather than promoting the survival of organisms is strongly supported by the theory of the evolution of aging and by the many experiments that have confirmed it. The soma is disposable; it is the genes that persist. This body of evidence is just as strong as or stronger than the behavioral evidence supporting the kin selection argument. Second, it is not necessary to decide whether selection acts on genes or on organisms because both levels are involved: selection acts through the differential reproductive success of organisms—the interactors—to generate differences in populations in the number of copies of genes—the replicators.

5. CONNECTING SELECTION TO FITNESS IN HIERARCHIES

The controversy over whether cooperation and altruism are produced by selection acting on groups, on kin, or on individuals has now continued for nearly fifty years and occupied thousands of pages. Many feel it has been settled in favor of kin selection; others continue to argue for group selection. The space available here will not do the controversy justice, but two brief remarks are in order (see chapter III.2).

First, it helps to distinguish between selection acting on groups and selection acting on genes to increase the fitness of individuals as mediated by their social interactions within groups. Attention to this point reveals that some, but not all, of the evidence produced in favor of group selection is simply evidence that selection has acted to improve the fitness of individuals or genes in the social context of the group.

Second, the conditions that determine the potential strength of selection at any level in a hierarchy are easy to state (although they may be hard to measure). Consider a hierarchy with just two levels—individuals and groups—and an extreme case used here to make a point clearly. If a population of individuals is organized into a set of groups, and the genetic variation in the population is distributed in such a manner that there is no variation within the groups, while each group differs genetically from every other group, then all the potential for a response to selection consists of the differences among groups. And if, in such a group-structured population, there are no differences in the reproductive success of any of the individuals within groups, and large differences in the reproductive success of the groups, then the strength of selection acting on individuals will be zero, and the strength of selection acting on groups will be strong. Any population that fulfills these two conditions will unquestionably experience group selection. Relaxation of these extreme conditions leads to intermediate cases in which selection is acting at both levels of the hierarchy.

The question is how often conditions favoring group selection are fulfilled in reality. There are good reasons to think they are infrequent. In most populations, there is more genetic variation among individuals within groups than there is among groups. That distribution has been well measured, for example, in humans, where about 85 percent of genetic variation is among individuals within groups. Furthermore, the births and deaths of individuals that create variation in individual reproductive success are much more numerous and occur on a much shorter interval than do the splitting and local extinction of groups. During the time it takes for one episode of group selection to happen, there has usually been opportunity for millions of events of individual selection to take place. Under such circumstances, group selection would have to be very strong indeed to overcome modest but consistent differences in individual reproductive success.

Those who favor kin and individual selection point out that the conditions under which group selection works—high degrees of relationship within groups, low degrees of relationship among groups—are also the conditions under which kin selection most efficiently selects for individual traits that benefit group cohesion. Those who favor group selection express the opposite perspective on the same situation: namely, that conditions that enable kin selection also favor group selection. A compromise view is that selection can act at any level of a hierarchy, that its efficiency in eliciting a response is determined by the distribution of genetic variation and the variation in reproductive success among units at that level, and that in principle it can act simultaneously at all levels. It usually does so, however, much more efficiently at the levels of genes and individuals than at that of groups.

6. POLISHED ADAPTATIONS OR ROUGH HISTORY

The realization that selection acts simultaneously at many levels, but usually much more frequently, and much faster, at lower levels than at higher ones, suggests two points. First, we will see the appearance of polish and design associated with adaptations only when those traits have experienced a vast number of selective events, for the thoroughness and efficiency with which the space of phenotypic alternatives is explored depend on the number of selective events and the rate at which they occur. For example, the ability of a bat to fly rapidly through a network of closely spaced branches in complete darkness can be explained only by a vast number of selective events. Second, selective events at higher levels that occur infrequently, such as differences in the splitting and extinction rates of species and higher clades, cannot create polished adaptations, for they occur too infrequently to do so, but they can generate the rough and arbitrary look that we attribute to history. For example, the current dominance of mammals and birds in the world fauna has as much to do with the disappearance of the many groups that were eliminated in the end-Cretaceous extinction as with any adaptations characteristic of mammals and birds, both of which originated millions of years before that extinction event (see chapter VI.14). Occasionally selective events at higher levels can produce broad patterns, such as the distribution of asexual reproduction up at the tips of phylogenetic trees regularly associated with sexual ancestors, a pattern that suggests that asexual lineages go extinct more rapidly than do sexual lineages.

7. ADAPTATIONIST STORYTELLING

The power of natural selection to generate complex adaptations from mutations whose effects are random with respect to the needs of the organism strikes some with the force of an epiphany. Such converts are then inclined to explain most of what they see as the product of selection. They usually find it easy to posit a scenario in which selection could have produced the trait in question by invoking some set of conditions tailor-made for the purpose, and they can get carried away and claim that the states of certain traits are adaptations without producing evidence sufficient to exclude alternatives, despite the fact that there are always at least the alternatives that the trait attained that state through genetic drift, or that it emerged as a by-product of selection on other traits, or that it exists in that state because it is constrained and cannot be otherwise. That is why loose thinking within the adaptationist tradition elicited a strong critique, a critique so strong that it produced a temporary overreaction during which any claim of adaptation was suspect. The pendulum has recently swung back to a balanced position that admits that adaptations are frequent but demands evidence to support the claim.

8. HOW TO RECOGNIZE SELECTION AND ADAPTATION

The problem with a claim of adaptation arises when one has not observed the evolutionary process that produced the state of the trait in question. There are at least three ways to support the claim of adaptation. If one has observed heritable changes in the trait that resulted from the correlation of trait state with reproductive success, then the change in the trait is an adaptation. If one can perturb the trait and demonstrate, with credible controls, that the original state has higher fitness than the perturbed states, then the original state was an adaptation relative to the perturbed states. If a developmental change in a phenotype that improves reproductive success relative to the unchanged state occurs only in response to a specific environmental signal, and the change is not expressed without the signal under circumstances where it would decrease fitness if it were expressed, then the change in the trait is an adaptation.

While it is true that adaptation is asserted more frequently than such tests are applied, such tests have nevertheless often been done. For example, altering widowbird tail lengths by cutting and pasting showed that longer tails enhance male reproductive success. And raising mosquitofish in fresh and brackish water demonstrated that those living in freshwater were maladapted to that environment, genetic analysis revealing that the freshwater population was swamped by gene flow from the nearby, much larger population living in brackish water. Both outcomes are possible, and both are accepted.

9. CAN WE DO WITHOUT THESE CONCEPTS? ABSOLUTELY NOT

Selection, adaptation, and fitness are well-tested ideas that have proven their usefulness in organizing information, communicating it efficiently, and motivating research for 150 years (selection) or more (adaptation). They combine to explain the millions of cases in biology of precision and complexity—the pervasive matches of structure to function—that cannot be explained by any other set of concepts. Because of their power and scope in making sense of universal features of biological systems, they belong to the very small set of elite concepts that explain all of biology.