VI.14

Species Selection

Emma E. Goldberg

OUTLINE

  1. Concepts and consequences

  2. History and controversy

  3. Empirical tests

The logic of Charles Darwin’s view of evolution by natural selection applies not only to individual organisms within populations but also to other levels of the evolutionary hierarchy. Entire species can differ from one another in traits that interact with the environment to affect speciation and extinction, and when those traits are inherited through lineage divergence, species selection occurs. This process has the potential to drive evolution on a large scale, making some clades more species rich than others and determining how commonly particular traits are possessed across groups of species. The precise scope of the definition of species selection and its feasibility as an evolutionary force have been debated for decades. Research now focuses on the empirical question of its prevalence, strength, and consequences in a variety of study systems. Because natural selection may act simultaneously at multiple levels, a great challenge in this endeavor is to separate the contribution of species selection itself from evolution at other levels, especially adaptation within species. Analyses of character evolution and diversification among fossilized or living species help illuminate the significance of species selection in shaping patterns of biological diversity.

GLOSSARY

Aggregate Trait. A characteristic of a species that summarizes a trait present in its individual organisms. Variation within each species in this trait should be smaller than differences in the aggregate trait among species. Body size is one example.

Clade. A group consisting of all the species, living or extinct, descended from a particular ancestral species.

Emergent Fitness. The heritable ability of a species to survive and reproduce; the expected difference between speciation and extinction rates, also called the net diversification rate.

Emergent Trait. A characteristic of a species that is not defined by the traits of its individuals. Any particular value of an emergent character may result from many combinations of organismal properties. Geographic range size is one example.

Levels of Evolutionary Hierarchy. Nested units of the complex organization of life. Examples include the species level, the level of populations below it, and the level of clades above it. Selection at any one level may have consequences that appear at other levels.

Species Heritability. The fidelity with which a trait of one species is passed along to its two daughter lineages during a speciation event.

1. CONCEPTS AND CONSEQUENCES

Species-Level Traits, Fitness, and Heritability

There are three basic requirements for evolution by natural selection: variation in the values of some traits, interaction of these traits with the environment to affect fitness (mortality and reproduction), and inheritance of the traits and their fitness consequences from generation to generation. In Darwin’s original view, this process played out among individual organisms within populations. Other units of the evolutionary hierarchy, however, may also be viewed as units subject to selection (discussed more generally in chapter III.2), including entire species, with fitness, trait, and inheritance defined at the species level. Selection at the species level may have profound effects on the distributions of species numbers and characteristics across the tree of life.

The fitness of a species is determined by its survival and reproduction—that is, how well it avoids going extinct and how successfully it gives rise to new species. From the standpoint of species selection, speciation and extinction propensities are properties of the species as a whole and together define its emergent fitness. The tremendous variation in species richness and characteristics among different groups of organisms is shaped to various degrees by environmental conditions, ecological interactions, and random occurrences. Among these many factors, the defining question for species selection is, What differences in emergent fitness are determined by the interaction of the environment with heritable traits at the species level?

What sorts of traits do species have? Each species exists over some limited portion of earth, so geographic range size is one example. Range size is not the property of a single organism but rather is determined from the group of organisms that make up the species as a whole—it is therefore an emergent trait at the species level. Other emergent traits include sex ratio, population density, genetic structure across populations, and social institutions or cultures. Another feature of an emergent trait is that its evolutionary consequences do not depend on how it is determined at lower levels. For example, a species may be narrow ranged because its individuals either lack wings for long-distance dispersal or tolerate only a very particular climate. Geographic range size “screens off” these organismal-level traits, however, if it is simply the small spatial area occupied by the species that puts it at risk of extinction.

In contrast, the body size of an animal is an example of a trait defined at the organismal level. If the variation in body size is lower within species than among them, then it also makes sense to treat average body size as a property of a species. This is an example of an aggregate species trait, one defined by a characteristic of individuals within the species. Other aggregate traits include generation time, the degree of ecological specialization, and various modes of reproduction, such as asexuality versus sexuality, monogamy versus polygamy, or wind versus animal pollination.

Many of these traits might reasonably be expected to affect rates of speciation or extinction, although proving so is a more difficult matter. Species with small geographic ranges are especially extinction prone because habitat quality, climate, or predation need turn unfavorable in only one location. The effect of range size on speciation is less clear intuitively and more variable empirically. Larger ranges present more opportunities for geographic barriers to arise and separate existing populations, but small-ranged species may be more sensitive to such barriers and hence more likely to become reproductively subdivided. Empirical results on the relation between body size and diversification are also mixed. Associations among large body size, small population size, and long generation time may make large-bodied species more prone to extinction and slower to complete the speciation process. In contrast with sexually reproducing species, lineages that can reproduce asexually (for example, tulips producing bulbs or lizards developing from unfertilized eggs) are expected to exhibit higher extinction than speciation rates, owing perhaps to low genetic diversity reducing the ability to adapt during environmental changes. In plants, populations with greater pollinator specificity may more rapidly become reproductively isolated and hence have higher speciation rates.

For species selection to occur, not only must species have traits that affect emergent fitness, but those traits must be inherited during the process of speciation. For many aggregate traits, species-level heritability follows naturally from organismal-level inheritance. Unless rapid evolution of the trait in question drives the divergence, as during ecological speciation, each new daughter species will be composed of a lineage of individual organisms drawn from the same pool of parental variability. Emergent traits can also be inherited across speciation events, though this is more difficult to show. The best examples again come from studies of species geographic distributions, which find more similar ranges among more closely related species of mollusks, birds, mammals, and plants.

Consequences of Species Selection

When all the ingredients are present for species selection to occur—heritable variation in traits that affect fitness at the species level—what are the potential consequences of this process? Only if trait variation already exists among species can it drive differences in speciation and extinction rates, so species selection cannot directly cause adaptation within species. Once a species acquires a new trait, however, if that trait increases extinction risk, it can be removed from circulation through species selection. Alternatively, if the trait increases speciation rate, species selection can result in a large clade in which the trait is common. Even characters that do not affect diversification may become more or less prevalent if they tend to be associated with, or “hitchhike” on, another trait subject to species selection. Its effects can also extend beyond simply altering the relative frequencies of characters. Traits that persist over longer timescales are more available for possible further modification, so species selection shapes the background from which new characters evolve.

One of the most celebrated features of species selection is its potential to drive evolution in a different direction than does selection at the organismal level. In extreme cases of cross-level conflict, traits that are advantageous for individuals within populations also increase extinction risk. Evolutionary change that compromises the amount of sexual reproduction or outcrossing, argued George C. Williams, is especially likely to exhibit a balance between two different levels of selection. Within a species, individuals that gain the ability to reproduce without a mate will have a marked advantage and become increasingly common. Furthermore, once traits like the ability to self-fertilize, propagate vegetatively, or develop from unfertilized eggs catch on, they rarely disappear from within a species. Selection at the species level can counter this trend toward reproductive self-sufficiency. Lower extinction rates or higher speciation rates of sexually reproducing species may help explain why most animals require mates and why many plants have intricate adaptations that encourage outcrossing.

Species selection may be more difficult to detect when it works in the same direction as organismal selection. Selection can reduce heritable variation in a trait, and when this happens at both levels simultaneously, trait variation and hence the possibility of evolution will exist over a shorter window of time. Therefore, even if species selection was strong in the past, its signature may not be apparent in living species. The separate contributions of selection at two levels will also be more difficult to measure when they have similar effects. In particular, methods that infer species selection by showing a balance between selection at different levels or by ruling out selection at other levels will not be applicable.

Finally, species selection can also drive trends in the absence of selection at other levels. This is a prominent component of the theory of punctuated equilibrium, discussed later (see also chapter VI.12).

2. HISTORY AND CONTROVERSY

The term species selection was first applied to evolution by differential proliferation of lineages by Hugo de Vries in 1905, but it did not come into popular use until 70 years later. It has had a stormy history in the scientific literature. Disagreement over the scope of its definition and collateral damage from debates on related topics have at times muddled its interpretation and thrown its utility into question. A clearer picture of species selection has emerged, however, leading to a firmer grasp on designing and conducting tests of its empirical significance.

Natural Selection, Species Sorting, and Effect Macroevolution

Including species selection under the umbrella of natural selection is not universally accepted. Darwin’s original arguments for evolution by natural selection were formulated in terms of individual organisms, but his compelling logic can apply to any level of the biological hierarchy. Illustrious names appear on both sides of the issue, but the prevailing view in multilevel selection theory is that species selection is one form of natural selection.

The broad term species sorting subsumes any process in which speciation and extinction rates differ among lineages, without regard to cause. Species selection is one example, with differences in emergent fitness produced by the interactions of species’ traits with the environment. Species drift, in which speciation and extinction differences are determined by random factors, is another type of species sorting. Sorting on geographic location is a third possibility. For example, if clades diversify rapidly on islands but slowly on the mainland, sorting is driven more by geography than by intrinsic traits.

The discussion so far has considered species selection in the broad sense, allowing a fairly liberal definition of species-level traits. Some prefer to apply the term species selection in a stricter sense, limiting it to cases in which the traits that affect emergent fitness are themselves emergent. When species sorting is instead driven by aggregate traits, the term effect macroevolution is applied. The reasoning here is that when a trait is expressed at the organismal level, selection on it must ultimately be reducible to processes at that level, making apparent species-level effects an artifact of causation cascading upward.

The more common view, however, is that the defining feature of species selection is the level at which selection occurs. Emergent traits may indeed provide the most compelling examples of species selection, especially because they may be harder to explain by organismal-level evolution. Selection at the species level can also act on aggregate traits, however. The interactions by which a trait affects organismal survival and reproduction may in general be quite different from those by which it affects extinction and speciation. Regardless of terminology, the ultimate goal for understanding multilevel selection is to identify the hierarchical level of the unit that is undergoing selection and to establish how its properties interact with the environment to determine its fitness.

Group Selection

The 1960s’ debate over group selection did not touch specifically on species selection, but it affected the general perception of the multilevel framework. Early group selection theories, notably those of V. C. Wynne-Edwards, argued that when an organismal trait is detrimental to the individual but provides an advantage to the group within which it lives, this necessarily implies the action of selection at the group level. Behaviors like giving birth to fewer offspring than is physiologically possible, or deteriorating in health when old, were speculated to evolve “for the good of the group,” to regulate population size. This logic was attacked especially by Williams, who argued that such regulatory adaptations often do not really exist. When there are group benefits, he reasoned further, they are better explained by more careful consideration of organismal-level selection, such as accounting for fitness across the whole life span of an individual.

Williams’s rebuttal very effectively forced more careful treatments of fitness and its consequences for adaptation. Unfortunately, his view that group-level explanations for traits should be called on only when lower-level explanations fall short caused many to unjustly discard selection at higher levels, including species selection, as even a potentially viable force. Williams did not dispute the basic logic of group-level selection when properly applied, however, and he clearly saw the potential power of selection at the level of species or higher taxa to shape the diversity of earth’s biota.

Punctuated Equilibrium

In the wake of the recoil from group selection, Stephen J. Gould was a strong advocate for the importance of multilevel selection, especially in macroevolution. Species selection plays a prominent role in his and Niles Eldredge’s original theory of punctuated equilibrium. In this conceptual model, evolutionary change does not accumulate significantly within species, a situation termed stasis. Instead, trait variation develops rapidly and in any direction during speciation, when peripheral populations become isolated and diverge. It is then the process of species selection that drives trends by preferentially eliminating much of the new variation while allowing some of it to survive and proliferate.

The punctuated equilibrium theory has been controversial since its presentation in the 1970s. One of its hotly contested claims is that natural selection within populations is primarily stabilizing or constrained, yielding long periods of stasis and thus requiring higher-level selection to produce large-scale evolutionary patterns. Although stasis may make the action of species selection more obvious, by clearly defining species as entities and removing a competing explanation for trends, it is not a prerequisite. Species selection operates equally well on variation among species regardless of whether that variation originated through punctuated bursts or gradual accumulation. Therefore, although punctuated equilibrium brought it into the spotlight, species selection should be judged independently, on its own assumptions and evidence.

3. EMPIRICAL TESTS

Assembling and analyzing data from natural systems to test the action of species selection is not straightforward. Nevertheless, several strong cases for species selection have been built, using a variety of data sources and mathematical tools.

Fossil-Based Tests

Clades and traits that are well preserved in the fossil record provide excellent opportunities for tests of species selection. Speciation and extinction rates can be estimated directly from the dated deposits in which fossils are found. Tying those rates to particular traits is more difficult, however.

Challenges in identifying the target of species selection are well illustrated by three decades of study of geographic range size and larval dispersal mode in marine mollusks. In this system, species are classified by whether they possess a larval stage that swims and feeds on plankton. Such planktotrophs are carried by ocean currents for weeks or months before settling hundreds of miles or more from their parents; they thus disperse much farther than do nonplanktotrophs. The large yolk required by nonplanktotrophs affects shell shape, so larval mode can be inferred for extinct species. Species’ geographic ranges are measured from the deposits in which their fossils have been found.

Dispersal ability is expected to affect geographic range size and genetic population structure, and consequently perhaps extinction and speciation rates. Work by Thor A. Hansen, David Jablonski, and colleagues uncovered species selection in several groups of gastropods from the Gulf Coast of North America. They found that planktotrophy was associated with larger geographic range size, longer species durations (lower extinction rate), and a lower rate of speciation. Larval mode is not the sole force behind selection at the species level, however. Within each larval mode, there is still substantial variation in diversification that must be attributed to other factors. Ecological specialization and trophic level are not found to be sufficiently explanatory. Contrast with another group is more illuminating: marine bivalves show a similar correlation between geographic range and species duration, but little association between larval mode and geographic range or extinction. From generalized linear models identifying the factors that best predict survivorship, geographic range indeed emerges as the dominant trait, with little additional predictive power provided by larval mode. This last analysis is a particularly important step in choosing from among correlated traits, even across hierarchical levels, those that best account for emergent fitness. The possibility remains that population genetic structure affects speciation more directly than does range size, but it cannot be determined for extinct species. Finally, the heritability of geographic range is established with regressions and randomization tests that show closely related species to have especially similar range sizes.

One complication sidestepped by this case study is trait evolution within lineages. Leigh Van Valen used the mammal fossil record to present genus-level selection as a force opposing previously documented trends of size increase within lineages. (Sufficient data at the level of species were not available, but genus selection is analogous to species selection, with the defining processes being extinction of all species within a genus and the origination of new genera.) Using a method adapted from the balance between mutation and selection in population genetics, he found a selective disadvantage to large body size in mammals. Because their lineage durations were longer, Van Valen concluded that large-bodied genera have lower origination rates.

A different framework for incorporating within-species trends into species selection analyses is provided by the equation named for George R. Price. Overall changes in traits are separated into two components: the correlation between trait and fitness (attributed to species selection) and other changes across generations (attributed to within-lineage evolution and biased trait inheritance). Carl Simpson used this approach in his analysis of complexity of the calyx, the cuplike portion in crinoids that contains reproductive and digestive organs. He estimated the first component by computing origination and extinction rates over time for genera possessing different values of the calyx complexity trait and then regressing net diversification on the trait. A separate set of calculations based on subclade comparisons estimated the second component. Simpson found that calyx complexity decreased over time both because lineages with simpler calyxes diversify more rapidly and because genera tend to be simpler than their ancestors. The reasons for these tendencies are not known, however. The conceptual application of the Price equation allows both within- and among-lineage selection to be treated on equal footing, rather than attributing to higher levels only what cannot be explained at lower levels.

Phylogeny-Based Tests

Studies based on living species are not limited to organisms and traits that fossilize well. The trade-off is that contemporary data will not directly provide a historical record of trait values, speciation events, and extinctions. Using molecular sequence data to quantify the relationships among species is an alternative means of gaining insights into the past. Such phylogenetic trees are rapidly increasing in scope and precision, and the mathematical and computational tools for inferring evolutionary processes from them are likewise advancing.

One popular means of testing whether a trait affects diversification is to compare sister clades, which share a common ancestor and hence have the same age and evolved from the same background (see chapter VI.15). Differences between sister clades in a trait and in their numbers of species, and a consistent association between trait and net diversification differences across many pairs of sister clades, can indicate species selection. From numerous applications of this method, characters related to sexual selection have emerged as one class that may influence diversification. These characters include traits associated with female mating preferences, such as showy male colors and elongated fins or feathers, and also reproductive factors that are antagonistic between the sexes, such as the evolution of seminal fluid chemistry to reduce female remating. Because these traits can evolve quickly and in somewhat arbitrary directions, they can drive rapid reproductive isolation between populations and hence increase speciation rates. Sister clade analyses of sexual dichromatism versus monochromatism in birds and fish, and of polyandry versus monandry in insects, do indeed indicate that traits related to sexual selection increase diversification. These analyses cannot distinguish the separate contributions of speciation and extinction, however, and extinction may play a role here if such traits have detrimental effects—for example, by attracting the attention of predators or reducing total fecundity.

Characters that evolve multiple times within a clade are especially valuable for tests of species selection. Correlations with other traits may be broken, and a repeated association with changes in speciation or extinction provides stronger support for a causal connection. The sister clade approach does not deal well with traits interdigitated on the tree, but the last decade has seen significant advances in phylogenetic methods that more powerfully integrate trait changes with diversification. A powerful approach is to fit mathematical functions of rates for trait evolution, speciation, and extinction to a phylogeny. Simultaneously accounting for all these processes is difficult, but recent work by Wayne P. Maddison and colleagues has made it possible under some circumstances.

This procedure was used to study the evolution of self-incompatibility, a genetic mechanism that prevents self-fertilization by causing a plant to reject its own pollen. The ability of individuals to reproduce without a mate is expected to be favored within a species and rarely to disappear from it once it takes hold. Analysis by Boris Igić and colleagues of the alleles involved in self-incompatibility indeed shows that evolutionary transitions to self-compatibility are frequent but that the reverse process has not occurred within the nightshade family Solanaceae. Fitting a model of trait evolution and diversification to a large phylogeny from this family provides estimates of the rate of loss of self-incompatibility within species and the rates of speciation and extinction associated with each breeding system. The results match well with Williams’s expectation that species selection can balance organismal selection favoring self-fertilization: the net diversification rate for self-incompatible species was much higher than that for self-compatible species, offsetting the loss of self-incompatibility and causing both states to coexist within the family.

Plants’ interactions with pollinators have been hypothesized to influence speciation and extinction, because relying on specialized pollinators may increase both the ease of reproductive isolation and the risk of insufficient reproduction. To look for such effects, phylogenetic models have also been applied to the evolution of floral traits. In a group of tropical vines, presence of a resin reward, which attracts bee pollinators, was not found to affect diversification. An analysis of morning glories, however, showed higher speciation rates for species with pigmented flowers (typically pollinated by bees, butterflies, or hummingbirds) than those with white flowers (typically pollinated by bats or moths).

Although potentially powerful, this framework has not yet been used to separate the effects of correlated characters. For example, self-compatible species tend to be annual rather than perennial, herbaceous rather than woody, rapidly flowering, and found in temperate climates and on islands; any of these traits could also influence diversification rates. Identifying the true targets of species selection is an ongoing challenge that will continue to be attacked with a wide array of data and techniques.

See also chapter III.2, and chapter VI.11.

FURTHER READING

Arnold, A. J., and K. Fristrup. 1982. The theory of evolution by natural selection: A hierarchical expansion. Paleobiology 8: 113–129. A discussion of natural selection operating at different levels, the Price equation, larval mode, group selection, punctuated equilibrium, and evolutionary constraints.

Jablonski, D. 2008. Species selection: Theory and data. Annual Review of Ecology and Systematics 39: 501–524. A lucid summary of conceptual issues, plus extensive tables of traits proposed to affect speciation and extinction.

Lloyd, E. A., and S. J. Gould. 1993. Species selection on variability. Proceedings of the National Academy of Sciences USA 90: 595–599. Discussion of broad- and strict-sense species selection and effect macroevolution. The amount of variability within a species may itself be subject to species selection.

Rabosky, D. L., and A. R. McCune. 2010. Reinventing species selection with molecular phylogenies. Trends in Ecology & Evolution 25: 68–74. Argues for the widespread importance of species selection, based on modern phylogenetic data and analyses.

Stanley, S. M. 1975. A theory of evolution above the species level. Proceedings of the National Academy of Sciences USA 72: 646–650. Species selection is introduced as an analogue to natural selection within populations. The article argues for its importance via punctuated equilibrium. Fossil-based estimates of speciation and extinction rates for mollusks and mammals are given.

Williams, G. C. 1992. Natural Selection: Domains, Levels, and Challenges. Oxford: Oxford University Press. Gives requirements for species selection (Chapter 3; the term clade selection is used instead of species selection to allow for selection at levels higher than the species) and characters it is most likely to affect. Discussion of group selection and punctuated equilibrium, plus a fascinating range of philosophical and empirical topics.