Adaptation and Constraint: Overview Martin Burd, Monash University, Victoria, Australia

Adaptation Adaptations are evolved features of organisms that enhance fitness in particular environments. As such, they appear to be well designed for their function, although the appearance of design is a result of natural selection, which operates without goals or teleological foresight. Adaptation has been a central idea in evolutionary biology since Darwin, but it has been difficult to characterize adaptations in a single, unified framework, and the term continues to generate controversy. See also: Adaptations: meanings; Adaptation and natural selection: overview Common usage among biologists allows adaptation to describe both a process and the product of that process. There is no disagreement that natural selection is the only process that creates adaptations. Features resulting from chemical or physical necessity are not adaptations. For example, the evaporation of sweat from the skin of some mammals, including humans, cools the body, but the latent heat of vaporization of water, which provides the cooling effect of evaporation, is a physical property that would not be described as adaptive. Most biologists also wish to exclude features that fortuitously provide a benefit, but have not been selected to provide that benefit. Darwin noted, for example, that the unfused sutures of the braincase in fetal mammals are of great benefit during parturition; only well after birth does complete ossification of the skull occur. But unfused sutures also occur in young birds and reptiles that need only escape from an egg. Thus, incomplete ossification in fetal mammals cannot be considered as an adaptation for mammalian birth. See also: Function and teleology; Natural selection: introduction Darwin’s example contains an implicit reference to the evolutionary history of a putative adaptation, and many biologists argue that adaptation must be viewed in a historical context. Recent advances in phylogenetic reconstruction (inference about the evolutionary relationships of lineages) make this task easier. Since any evolutionary change is specific to an ancestral starting

Article Contents . The Nature of Adaptation and Constraint . Adaptive Perfection and Imperfection

Adaptations are features of organisms that have evolved to perform fitness-enhancing functions. Some conceivable adaptations or combinations of adaptations do not evolve because constraints (which may vary from lineage to lineage) limit the potential of adaptive evolution.

The Nature of Adaptation and Constraint

Introductory article

. Examples of Improvement, Perfection and Imperfection . Types of Constraint . Testing for Adaptation and Constraint

doi: 10.1038/npg.els.0004166

condition, adaptations can be identified as derived (i.e. modified) traits that confer superior fitness, relative to the ancestral condition, in the current environment. The evolutionary history of feathers, shown in Figure 1, illustrates the phylogenetic perspective on adaptation. Feathers first appeared in a lineage of theropod dinosaurs. The earliest feathers, as found on the fossil genus Beipiaosaurus, for example, formed tufts of branches with small barbs. We do not know why these feathers evolved, but insulation was a possible function. They could not have aided flight. In lineages that arose later, such as the fossil genus Caudipteryx, a feather appeared with a central rachis, barbs and barbules that form a closed vane. But Caudipteryx could not fly. Only in the lineage including Archeopteryx and living birds, do we find feathers with asymmetrical vanes that could assist in creating lift – the flight feathers.Using the phylogeny of Figure 1, we can identify some adaptations involving feathers. Although the function of the first feathers is not known with certainty, let us assume it was insulation. Insulating feathers could then be considered an adaptation in Beipiaosaurus and all the feathered lineages that arose later, when this trait is compared to the basal lineages of allosauroids that lacked feathers entirely. But the insulating feathers of the house sparrow, Passer domesticus, would not be considered an adaptation when the historical context for comparison includes only other passerines (songbirds), all of which share the trait. Only with reference to the ancient and extinct allosauroids (or some more distantly related vertebrate) we can recognize the insulating feathers of the house sparrow as an adaptation. Similarly, the asymmetrically vaned flight feathers of the house sparrow can be recognized as an adaptation for flight in comparison to the feathers of nonflying theropods like Caudipteryx, but not in comparison to the flight feathers of other passerines. See also: Fitness; Molecular phylogeny reconstruction The tyrannosaurid lineage, which includes the famous fossil Tyrannosaurus rex, may have lost the tufted feathers that had evolved among its ancestors (Figure 1). Palaeontologists may yet discover evidence of feathers on tyrannosaurids, but let us assume that feathers were absent. In that case, the absence is a plausible adaptation

ENCYCLOPEDIA OF LIFE SCIENCES & 2005, John Wiley & Sons, Ltd. www.els.net

1

Adaptation and Constraint: Overview

Feathers absent

Allosaurids

Feathers unknown

Beipiaosaurus

Tyrannosaurids

Caudipteryx

Archaeopteryx

Modern birds

Asymmetrical flight feathers

Loss of feathers?

Central rachis and vane

Origin of feathers

Figure 1 A partial phylogeny of theropod dinosaurs and birds to illustrate history-dependent definitions of adaptation. The phylogenetic ‘tree’ represents evolutionary relationships, with earlier origins of lineages represented by splits near the base, and later origins by splits near the top. All the groups represented are extinct, except for the modern birds. The absence of feathers in allosauroids is not an adaptation because this trait has not been modified from the ancestral condition of dinosaurs. The origin of feathers in the ancestor of the lineage of Beipiaosaurus–tyrranosaurid–Caudipteryx–Archaeopteryx– modern birds could be an adaptation, provided feathers brought superior fitness in the environment of that lineage. A reversion to the featherless state in tyrannosaurids (fossil evidence on this question is lacking) could also be an adaptation, because it would be a modification of the ancestral condition in this lineage, even though the same featherless state is not an adaptation for allosauroids.

(although how it enhanced fitness may be unknown), because it is a trait that was modified from feather-bearing ancestors. But the same character state – the absence of feathers – would not be an adaptation in allosauroids, because absence is an ancestral rather than derived trait in that group. Even if the lack of feathers provided the same functional advantage for allosauroids and tyrannosaurids, the phylogenetic perspective would allow us to call it an adaptation in only one of these lineages. Other biologists emphasize the role of natural selection in the maintenance, rather than the origin and phylogenetic history, of an adaptation. Adaptive traits are recognized by their fitness consequences relative to an explicit set of alternative phenotypes among which natural selection may choose. This history-free definition allows highly functional traits to be viewed as adaptations if they are maintained by selection 2

against departures from the favoured design. Such departures may occur because a trait is polymorphic, or may arise from time to time by mutations that disrupt the adaptive function. Darwin’s example of the sutures in the mammalian skull could be reinterpreted under this ahistorical definition. Suppose, for example, that mutants with faster rates of ossification during fetal growth had arisen in the mammalian lineage, but had been selected against because of greater difficulties during birth. The nonfused condition of the fetal mammalian skull would be maintained by selection because of superior functionality, and would be an adaptation even if birds and reptiles share the same feature of skull development. However, problems with an historical definition of adaptation may arise if a trait affects two different kinds of functional performance, or two traits affect the same function.

Adaptation and Constraint: Overview

Constraint Many conceivable adaptations or combinations of adaptations have not evolved. Viviparity (‘live birth’), for example, which has appeared independently several times in vertebrate lineages (including sharks, bony fish, amphibians, snakes, lizards and mammals), has never evolved in the birds. Is viviparity simply not adaptive for birds? Or does something prevent a course of evolution that would be adaptive if it could somehow be achieved? Such factors would be constraints on adaptation that restrict the possible outcomes of the evolutionary process. Constraint is of central importance in evolutionary theory, but, like adaptation, it has proven to be a contentious concept. See also: Functional constraint and molecular evolution In an influential essay written in 1979, palaeontologist Stephen Gould and geneticist Richard Lewontin objected to what they saw as an ‘adaptationist programme’ of research in which traits are given facile adaptive interpretations. They criticized the view that adaptations are pervasive in the natural world, pointing instead to pervasive constraints that derive from developmental pathways, and the need for traits to function within the context of other traits in a unified whole organism. The argument of Gould and Lewontin has itself been criticized. While they insisted that a trait must pass stringent criteria before it can be labelled an adaptation, they did not establish specific criteria by which an evolutionary constraint on a trait could be recognized. There seems not to be, at present, any justification for supposing either adaptation or constraint to be the ‘default’ process of evolution. The issue is not whether adaptations and constraints exist, but what relative importance of each has been in evolutionary history. See also: Adaptationist claims – conceptual problems; Evolution: history

Adaptive Perfection and Imperfection To be favoured by natural selection, a trait only needs to function better than currently available alternatives. Nothing in this process requires nor ensures that adaptive outcomes will be perfect. Darwin expected that natural selection will produce organisms that are only as perfect, or slightly better, than the competitors against which they struggle for existence. Indeed, Darwin noted that the theory of descent with modification makes comprehensible, in a way that special creation cannot, traits that are imperfect or nonfunctional, such as the webbed feet of upland species of geese that never or rarely swim. Whatever the perfection or imperfections of adaptations, one can ask whether adaptive evolution by natural selection produces improvement or progress in any lineage over time. Progress often implies approach towards a goal, and the process of natural selection admits no goal or

foresight. If we abjure this interpretation, then natural selection, although it may not require progress, is nevertheless consistent with continual improvement in adaptations. However, a difficulty in judging the quality of adaptation is the lack of obvious standards of perfection. We might suppose that fitness itself provides a relative criterion of improvement, but often we are interested in whether and how often adaptations improve, or approach perfection, by some objective standards. Geometry and physics are obvious touchstones against which the success of adaptation is measured, because they seem independent of the details of biological context and cannot be changed by natural selection. But in some cases, criteria drawn from biology may be appropriate. Some examples in which these ideas are applied will show how the quality of adaptation might be assessed. See also: Fitness: physiological problems

Examples of Improvement, Perfection and Imperfection Perfection seems rare in biology, but occasional examples occur. The hexagonal pattern of the comb built by honey bees (Apis melifera) is a geometric design that creates chambers needing a minimum quantity of building material for a given chamber volume. All the chambers share walls with adjacent chambers, so that there is no ‘dead space’ in the comb, and in this sense the hexagonal polyhedrons are a geometrically perfect design. Other bee species produce combs with chambers that are circular in cross-section. For chambers of a given volume, the hexagonal honey bee design requires only about 52% as much wax. In some cases, adaptive evolution must occur within a specific context set by some other biological feature, which can then act as the standard of perfection. In Batesian mimicry, for example, a mimetic species evolves an appearance that is difficult for predators to distinguish from that of a distasteful model species. The model provides the benchmark for adaptive perfection in the mimic, and some instances of mimicry in butterflies appear (to human eyes, in any case) to be nearly perfect. Mimicry occurs also in the eggs of cuckoos (Cuculus canorus), which are deposited by cuckoo females in the nests of other species, so that their offspring parasitize the feeding and care supplied by the hosts. Some host species will eject eggs dissimilar to their own, creating a selective pressure on the appearance of cuckoo eggs. Local populations of cuckoos in the UK deposit better matching eggs in the nests of species that are more discriminating. In this case, host eggs dictate the standard of perfection, and the cuckoo adaptation has demonstrably improved under selection. See also: Mimicry Adaptive improvement of the respiratory pigments of vertebrates, the haemoglobins, can be assessed by their biochemical activity. To function well in oxygen transport, 3

Adaptation and Constraint: Overview

haemoglobin must have a high affinity for oxygen when it is in the gills or lungs, but sufficiently low affinity that it gives up oxygen to tissues in need. These opposing requirements are satisfied in mammalian haemoglobin by so-called cooperative binding among the four subunits of the haemoglobin molecule, such that acquisition of an oxygen molecule by one subunit increases the affinity of the others for oxygen, while loss of oxygen by one subunit makes it easier for the others to lose their oxygen. The four subunits are of two types, termed a and b, which diverged following a gene duplication that occurred about 450 Ma. In the basal vertebrate lineage of lampreys and hagfishes the duplication did not occur, and their haemoglobin is composed of just two subunits which weakly show cooperative binding. The chemical and molecular basis of the cooperative binding is well understood, and there seems no reason to deny that the tetrameric mammalian haemoglobin is an improvement over the dimeric form. See also: Adaptation: genetics; Evolutionary developmental biology: gene duplication, divergence and co-option; Haemoglobin: cooperativity in protein–ligand interactions The evolution of internal gestation in mammals required that the fetus be able to extract oxygen from haemoglobin in the mother’s blood. This is accomplished through fetal haemoglobins that are expressed only in early life. These fetal haemoglobins, formed from a and g subunits, have a higher oxygen affinity than maternal a–b haemoglobin due to a single amino acid change in the g subunit that affects the interaction of the haemoglobin with a regulatory chemical, bisphosphoglycerate. As a result, the fetus can draw oxygen across the placenta from the mother’s blood. In the context of mammalian evolution, the maternal haemoglobin sets a standard of biochemical activity against which fetal haemoglobin must operate. In this restricted sense, fetal haemoglobin seems to be an optimal or perfected adaptation. See also: Reproduction in mammals: general overview; Reproductive strategies Despite such examples, obvious deficiencies in function are easy to find. Human bipedal locomotion, for example, creates or exacerbates effects such as varicose veins and hernias, because the quadrupedal body design from which hominids evolved did not place as much pressure on the leg veins or abdominal wall. The size and position of the pharynx and larynx in the human throat allow a variety of sounds to be produced in speech, but also make it possible for swallowed food to become lodged in the larynx, leading to suffocation. Even in the human eye, which Darwin termed an organ of supreme perfection, the lens progressively hardens with age, so that its shape can less easily be changed to accommodate to near and distant objects. In the eyes of cephalopod molluscs, by contrast, accommodation is achieved not by distorting the shape of the lens but by moving it forward or back within the eye, so that focus is unaffected by ageing of the lens. Other species would provide many more examples of imperfections in adaptations. See also: Darwinian medicine 4

Types of Constraint Like adaptation, the various facets of evolutionary constraints have not been integrated in a single framework. Thus, the following categories need not be mutually exclusive.

Constraints of physics and chemistry Organisms must function under laws of physics and chemistry that cannot be changed by natural selection. For example, it is often pointed out that gravity alone prevents the evolution of land mammals as large as the largest whales, because organisms of such size would crush their own lungs on land. The viscous properties of water allow cilia or flagella to be used effectively for locomotion at the scale of microscopic, single-celled organisms, but these adaptations could not provide propulsion at the macroscopic scale of fishes or whales. Conversely, the reciprocating motion of flattened surfaces like fins or flukes would be useless for locomotion by bacteria or protozoa, so that fins could not be adaptations for movement in these tiny organisms. See also: Ecological implications of body size The constraints on adaptation imposed by physics and chemistry may be quite subtle. For example, it has been argued that wheels might be highly functional for some terrestrial animals, but they cannot evolve because of difficulties in arranging a supply of nutrients, blood and nerve impulses to a structure that must rotate freely about an axis. However, although wheels are extremely efficient for locomotion on hard, flat, unrestricted surfaces, they become progressively less useful at small size and on compliant or irregular terrain. Wheeled structures cannot turn sharply. The constraint, it may be argued, is therefore the physical inefficiency of wheels in natural environments rather than the physiological impossibility of constructing a wheel from organic tissues.

Genetic constraints The genetic variation and covariation for an array of traits set an immediate limit to evolutionary potential. Traits for which limited genetic variation exists, so that heritability is at or near zero, can change only weakly or not at all in response to selection. For example, some individuals of the flowering plant Raphanus raphanistrum seem to produce pollen that fertilizes more ovules than rival pollen when the two types compete in the same flower. The traits that confer superiority, such as faster pollen tube growth, would be favoured by selection, but the competitive ability of superior pollen seems not to be heritable, and therefore adaptive evolution cannot proceed. See also: Variation, within species: introduction Adaptive combinations of traits may not evolve due to genetic correlations between traits caused by pleiotropy or

Adaptation and Constraint: Overview

Trade-offs, or functional constraints, derive from a principle of allocation, which supposes that an organism must distribute some finite resource, such as time, metabolic energy, or protein, among competing functions. Adaptive evolution is constrained by trade-offs between functions that demand the same resource. For example, a sunbird might obtain exclusive access to the nectar of flowering plants if it spends time defending a territory against intruders. But if a sunbird cannot simultaneously engage in territorial defence and other activities such as grooming, pursuit of mates, or even feeding itself, then the most adaptive (in the sense of fitness-promoting) effort at territorial defence must be a compromise with these other time-consuming activities. Trade-offs are usually based on an interpretation of physiology or behaviour that suggests a finite resource and competing demands, rather than on an explicit underlying genetic mechanism. Functional trade-offs have been particularly useful in the analysis of behavioural adaptations, whose genetic basis may be difficult to uncover. Of course, the same constraint might be cast in both functional and genetic terms, as in the trade-off of flowering time and growth, discussed above. Few constraints inferred on functional grounds have also been examined genetically, however. Functional constraints might be difficult to detect if there is wide variation among individuals in acquisition of resources. An analogy will illustrate this problem: a tradeoff exists between spending on clothes and on cars, because money allocated to one cannot be used to purchase the other. Nonetheless, a survey might show that wealthy individuals have both fine clothes and fancy cars, while the poor have neither. In this instance, a difference in resource levels masks the underlying trade-off.

dynamics of development may serve as a constraint on adaptation. For example, vertebrate eyes form in the embryo from outgrowths of the surface of the brain. Photoreceptor cells in the retina differentiate behind a layer of nerve cells, so that light must pass through this diffusing screen before reaching the receptors. This arrangement degrades the quality of the image, but seems to be an unalterable consequence of the pattern of development of the eye in vertebrates. The retina in the eyes of cephalopod molluscs, with a different embryological origin, does not have this compromised structure. See also: Evolutionary developmental biology: developmental and genetic mechanisms of evolutionary change; Evolutionary developmental biology: developmental constraints One body part or feature of an organism often grows in proportion to another part during development, a phenomenon known as ‘scaling’ or ‘allometry’. Allometric relationships between body parts are sometimes thought to be an expression of developmental constraint. For example, among papionid primates (baboons, mandrills and geladas), the relative proportions of the cranium change allometrically as total cranium size varies. The similarity of the allometric pattern among papionid species suggests that they may share a similar developmental pathway that determines cranial shape. The cranium includes the palate and teeth, and any constraint on their size and shape would have especially large consequences for geladas, because their diet, unusually for a primate, is almost exclusively grass. Geladas have large molars with high ridges for grinding tough grass tissue, but these teeth are not so large nor the ridges so pronounced as would be expected by comparison with other grazing animals of similar size. The underlying growth allometry of cranium and teeth in papionids may constrain the evolution of typical grazing adaptations in the dentition of geladas. Perhaps in compensation, geladas have greater manual dexterity than most other primates, and spend much of their foraging time carefully by pinching off the tender young tips of growing grasses. Although allometric patterns may constrain adaptation, such relationships may also preserve functional equivalence of adaptive traits as overall body size undergoes evolutionary change. The difficulty of distinguishing allometry as a source of constraint and adaptation reinforces the idea that neither can be assumed to be the default condition. The absence of a developmental basis for a trait may account for the failure of some adaptations to appear in certain lineages. But the development might also bias the kinds of variation likely to arise from random mutation, leading to directional trends of evolution within lineages, or to parallel adaptive change in lineages that share similar developmental constraints. See also: Life history theory

Developmental constraints

Phylogenetic constraints

Because traits of reproductively mature organisms arise through a developmental pathway, the fundamental

The distribution of traits among related taxa often suggests that evolutionary possibilities are circumscribed in certain

linkage disequilibrium. In the annual flowering plant Brassica campestris there is a strong association between early flowering time and small vegetative size, or between late flowering and large size. Artificial selection experiments demonstrate that this correlation is genetically based, with little possibility for the evolution of the alternative combinations, early flowering and large vegetative size, or late flowering at small size. This genetic constraint is congruent with the trade-off expected from functional constraints, discussed below. Genetic limitations of these sorts may erode over time as mutation and recombination increase the available genetic diversity in a population. See also: Quantitative genetics; Mutations and new variation: overview

Trade-offs

5

Adaptation and Constraint: Overview

lineages. The absence of viviparity in birds, noted above, is one example. Another is the number of ovules per flower in angiosperms, which is uniform among all species in some families, such as the Boraginaceae and Verbenaceae, but differs widely among species in other families, like the Liliaceae. Such patterns suggest the existence of factors that restrict evolutionary potential due to the previous history of evolution in a particular lineage. Such factors are called phylogenetic constraints. The constraints per se (rather than the effects they produce on the taxonomic distribution of a trait) are features acquired by ancestors during earlier episodes of evolution that commit descendent taxa to a limited portion of ‘adaptive space’ in which natural selection can move the phenotype. Much of the conventional wisdom about large-scale patterns in evolution can be viewed as expressions of phylogenetic constraint. For example, Dollo’s law, which states that complex adaptations, once lost, are not regained, is a type of phylogenetic constraint that might, for example, prevent the evolution of snakes with legs. Another constraint, often assumed without much empirical support, is that evolution of specialization in form or function is seldom reversed towards greater generalization. This pattern, if true, would yield lineage-specific limits to the kinds of variation that can be generated by natural selection. A distinction is sometimes made between phylogenetic constraint, in which history actively restricts what selection can do, and phylogenetic inertia, in which similarity due to common descent is passively maintained because there is no need to alter a functional design. It can be difficult to distinguish between these options in the case of highly conserved features of lineages, such as the basic plan of vertebrate limbs.

Testing for Adaptation and Constraint A necessary requirement of adaptive status is that a trait has to be functional for the organism in its habitat. Assessment of functionality is one of the normal projects of biological research – the discovery of how something works in a biochemical, anatomical, mechanical or ecological sense. Ideally, it would also be demonstrated that natural selection favours a putative adaptation. This can be accomplished by comparing the fitness of different forms of the trait in question, if there is natural polymorphism or if the trait can be experimentally manipulated to obtain a variety of forms. Functionality and selection are minimal requirements, but would not be sufficient for adaptation under some definitions. Comparisons among taxa, usually employing a phylogeny for the organisms in question, can be a powerful technique to identify adaptations when functionality and selection are difficult or impossible to assess experimen6

tally. An adaptive interpretation is strengthened if the same or similar traits arose independently in several lineages. That is, convergent evolution in several lineages suggests adaptation. A stronger case for adaptation can be made if a trait arose in several lineages following the earlier appearance of conditions that make the putative adaptation functional. This approach has been used to test the suggestion that dioecy (male and female function on separate individuals, as opposed to hermaphroditism) is frequently associated with fleshy propagules among gymnosperms because dioecy allows female plants to devote all their reproductive resources to making a large fruit crop. The abundant fruit would be more attractive to animal frugivores that disperse the seeds, giving greater reproductive success to such females. To be adaptive in this way, dioecy must evolve after the evolution of fleshy propagules. A phylogeny of gymnosperms analysed by botanist Michael Donoghue did not unambiguously support this order of evolutionary events in all lineages, casting doubt on the adaptive interpretation. Constraint, as a counterpart to adaptation, is often inferred by the same techniques from which adaptation is inferred. A consideration of biochemical, physiological or mechanical function may convincingly point to constraints. Phylogenetic comparative approaches may also provide support for constraint arguments. For many groups of organisms, accurate phylogenies are not available, and we must rely on traditional taxonomic classifications. Statistical techniques exist that identify the proportion of variance in a trait associated with each taxonomic rank (genus, family, order and so on) of the organisms in a data set. Such analyses are not definitive, but can help to identify how much of the variation in a trait across many taxa is likely to be adaptive variation or lineage-based constraint. See also: Variation: measures

Further Reading Brooke ML and Davies NB (1988) Egg mimicry by cuckoos Cuculus canorus in relation to discrimination by hosts. Nature 335: 630–632. Darwin C (1859) The Origin of Species. London: John Murray. Donoghue MJ (1989) Phylogenies and the analysis of evolutionary sequences, with examples from seed plants. Evolution 43: 1137–1156. Dorn LA and Mitchell-Olds T (1991) Genetics of Brassica campestris. 1. Genetic constraints on evolution of life-history characters. Evolution 45: 371–379. Dunbar RIM (1991) Adaptation to grass-eating in gelada baboons. Primates 33: 69–83. Goldsmith TH (1990) Optimization, constraint, and history in the evolution of eyes. Quarterly Review of Biology 65: 281–321. Gould SJ (1981) Kingdoms without wheels. Natural History 90(4): 42–48. Gould SJ and Lewontin RC (1979) The spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist programme. Proceedings of the Royal Society of London B 205: 531–598. Harvey PH and Pagel MD (1991) The Comparative Method in Evolutionary Biology. Oxford: Oxford University Press.

Adaptation and Constraint: Overview

LaBarbera M (1983) Why the wheels won’t go. American Naturalist 121: 395–408. Lauder GV, Leroi AM and Rose MR (1993) Adaptations and history. Trends in Ecology and Evolution 8: 294–297. Lord J, Westoby M and Leishman M (1995) Seed size and phylogeny in six temperate floras: constraints, niche conservatism, and adaptation. American Naturalist 146: 349–364. McKitrick MC (1993) Phylogenetic constraint in evolutionary theory: has it any explanatory power? Annual Review of Ecology and Systematics 24: 307–330. Prum RO and Brush AH (2002) The evolutionary origin and diversification of feathers. Quarterly Review of Biology 77: 261–295. Reeve HK and Sherman PW (1993) Adaptation and the goals of evolutionary research. Quarterly Review of Biology 68: 1–32.

Ruse M (1993) Evolution and progress. Trends in Ecology and Evolution 8: 55–59. Seeley TD (1985) Honeybee Ecology: A Study of Adaptation in Social Life. Princeton, NJ: Princeton University Press. Singleton M (2002) Patterns of cranial shape variation in the Papioni (Primates: Cercopithecinae). Journal of Human Evolution 42: 547–578. Snow AA and Mazer SJ (1988) Gametophytic selection in Raphanus raphanistrum: a test for heritable variation in pollen competitive ability. Evolution 42: 1065–1075. Sober E (1984) The Nature of Selection: Evolutionary Theory in Philosophical Focus. Cambridge, MA: MIT Press. Vogel S (1981) Life in Moving Fluids. Princeton, NJ: Princeton University Press.

7