Contemporary evolution of secondary sexual traits in the wild


†Author to whom correspondence should be addressed. E-mail:


  • 1Sexual selection has been studied intensively, and is often strong in natural populations. Theoretical models, comparative studies and laboratory selection experiments all suggest that evolution driven by sexual selection should be rapid and common.
  • 2However, there are relatively few documented cases of the contemporary evolution of secondary sexual traits in natural populations. Moreover, the causes are not always due to sexual selection, but often due to altered natural selection regimes, or an altered balance between natural and sexual selection.
  • 3Here we discuss recent empirical studies that have demonstrated the contemporary evolution of secondary sexual traits in natural populations. Our examples include both continuous traits and discrete polymorphisms. Taxa in which the contemporary (rapid) evolution of secondary sexual traits have been demonstrated include fish, insects, mammals, reptiles and birds. The evolutionary rates of these changes range from 0·005 to 0·570 haldanes (arithmetic mean = 0·14; geometric mean = 0·095; median = 0·12).
  • 4The relative rarity of examples could be explained by different genetic architectures of sexually selected traits compared with naturally selected traits, or by sexual selection regimes not being sustained in the long term. These factors could potentially slow down evolutionary rates of secondary sexual traits and make the detection of the contemporary evolution empirically more difficult.
  • 5Promising and underutilized approaches to sexual selection research include reciprocal transplant experiments and estimation of sexual isolation between conspecific populations.


Sexual selection and related processes (e.g. sexual conflict) have been major topics of interest among behavioural ecologists and evolutionary biologists for several decades. Although the theory of sexual selection by female choice was developed by Darwin (1871), it took more than a century until it became widely accepted by evolutionary biologists. After some pioneering manipulation experiments on long-tailed widow birds, Euplectes progne (Andersson 1982), experimental studies aimed at identifying traits that are targets of sexual selection have exploded in number (Andersson & Simmons 2006). Many studies have clearly demonstrated the existence of both female choice and male–male competition in natural populations, and confirmed that many secondary sexual traits are targets of inter- or intrasexual selection, or both.

The importance of sexual selection in natural populations has also been verified in non-experimental contexts. Studies estimating the strength of selection in natural populations using regression approaches became popular following the publication of a seminal paper by Russell Lande and Stevan J. Arnold (Lande & Arnold 1983; Endler 1986). Recent meta-analyses of such selection studies indicate that sexual selection gradients are stronger in magnitude than are natural selection gradients (Kingsolver et al. 2001), suggesting an important role for sexual selection in natural populations. Although such patterns could potentially have arisen due to different time intervals over which sexual and natural selection were estimated (Hoekstra et al. 2001), the conclusion that sexual selection is often stronger than natural selection appears to be robust, at least in some species (Svensson et al. 2006).

There are other reasons to believe that sexual selection is strong and can fuel rapid evolution. Verbal arguments (West-Eberhard 1983), formal mathematical models (Lande 1981, 1982) and comparative studies (Eberhard 1985, 2004; Arnqvist 1998; Mila et al. 2007) all indicate that secondary sexual traits, and other traits involved in social interactions, should exhibit rapid evolutionary change. Such traits could subsequently be involved in speciation if they also function as incipient species-recognition characters. Similar arguments apply to the closely related field of sexual conflict, in which male and female reproductive traits involved in mating conflicts have been shown to evolve rapidly in laboratory settings (Rice 1996), which could also promote sexual isolation between allopatric populations (Martin & Hosken 2003). Sexual conflict models confirm the predictions from models of classical sexual selection by female choice: evolution driven by sexual conflict is expected to be very rapid, and sexual conflict may be important as an engine of speciation (Gavrilets 2000, 2004).

Given these diverse and independent lines of evidence, one would expect to find numerous, well documented cases of the contemporary evolution of secondary sexual traits in natural populations. But the empirical evidence is very scant, contrasting markedly with several now classical examples of contemporary evolutionary change driven by natural selection, such as beak morphology in Darwin's finches (Grant & Grant 2002), wing length in Drosophila flies (Huey et al. 2000), and life-history traits in guppies (Reznick et al. 1997). These and many other documented cases of contemporary evolution driven primarily by natural selection have been discussed elsewhere (Hendry & Kinnison 1999; Kinnison & Hendry 2001). Our purpose here is to discuss the apparent scarcity of the contemporary evolution of secondary sexual traits in natural populations. We also discuss some promising avenues for the future and exemplify with some data from our own recent empirical work.

Theoretical considerations

The vast majority of secondary sexual traits, if not all, are likely to be targets of both sexual and natural selection. The evolutionary dynamics and change in secondary sexual traits in nature will thus, in most cases, be due to an interaction between natural and sexual selection, and evolutionary changes in secondary sexual traits could often have been caused by a changed balance between these selective forces. For instance, if natural selection against exaggerated secondary sexual traits were to become relaxed when a group of animals invaded a novel and favourable environment (e.g. with few predators), natural selection against exaggerated secondary sexual traits would weaken and one would expect to see rapid evolutionary change driven entirely by sexual selection. We will return to this important issue later, but first we discuss briefly why secondary sexual traits are expected to exhibit rapid evolutionary change due to the intrinsic and unstable nature of sexual selection regimes alone.

R. A. Fisher was the first to suggest that the evolution of female choice and male secondary sexual characters that were targets of such female choice would coevolve in a runaway fashion (Fisher 1930). Fisher recognized that even if female choice and male traits were governed by separate sets of loci, non-random female choice of attractive males would create a link in their offspring between male trait alleles and female choice alleles. This genetic correlation between female choice alleles and male trait alleles arises because non-random mating causes linkage disequilibrium (LD) between the separate loci. If this LD becomes sufficiently large, the two traits will coevolve together and reinforce each other. The co-evolutionary dynamics between female mate preference and male trait loci were subsequently modelled by Lande (1981), who basically confirmed Fisher's original theory.

Later theoretical work has shown that sexual selection could also cause rapid evolution through other mechanisms, including female choice of ‘good genes’ for high offspring viability (Houle & Kondrashov 2001), and male–female antagonistic co-evolution driven by sexual conflict (Gavrilets 2000). Although rapid evolution by sexual selection can also arise through good-gene mechanisms (Houle & Kondrashov 2001), evolutionary change driven by good genes is generally expected to be slower compared with Fisherian runaway processes (Iwasa & Pomiankowski 1995; Iwasa & Pomiankowski 1999). Good-gene models are therefore less likely to explain rapid population divergence of secondary sexual traits (Lorch et al. 2003) compared with other sexual selection mechanisms, including runaway processes caused by Fisherian or sexual conflict mechanisms, or direct fitness benefits to females of choosing a good male parent (Schluter & Price 1993; Iwasa & Pomiankowski 1999).

Runaway processes are thus by no means restricted to classical sexual selection by female choice, but are indeed possible under a wide range of conditions as long as there is some form of mating interaction between the sexes that causes a genetic covariance between male and female reproductive traits. An excellent review of quantitative genetic models of sexual selection processes was recently published by Mead & Arnold (2004). Also pure intrasexual selection processes, such as male–male competition, have the potential to cause rapid evolutionary change. Male–male competition and other forms of social interaction could, due to their frequency-dependent nature, be expected to fuel rapid evolution through contest-driven intraspecific arms races (West-Eberhard 1983; Rice & Holland 1997; Hardling 1999).

Another idea with implications for rapid evolutionary change in secondary sexual traits is phenotypic plasticity followed by genetic assimilation (Price et al. 2003; West-Eberhard 2003). For instance, carotenoid colours in bird feathers often function as sexual and social signals, are condition-dependent and sensitive to the availability of carotenoid-rich food in the environment, and are thus highly plastic traits (Price 2006). Bird populations that invade novel environments with lower or higher levels of carotenoid-rich food sources thus could potentially change their secondary sexual characters as a purely plastic response, at least initially (Price 2006). Phenotypic plasticity of this kind can rapidly cause a shift from one adaptive peak to another, and the initially plastic response would subsequently be followed by genetic assimilation and refinement of the novel sexual phenotype (Price et al. 2003; Price 2006).

Empirical evidence

Secondary sexual traits can evolve due to natural selection alone, due to sexual selection alone, or (perhaps most common) due to a change in balance between natural and sexual selection. Models of sexual selection have provided us with a theoretical framework to expect rapid evolution of secondary sexual traits in natural populations. Rapid evolutionary change driven by sexual selection has indeed been confirmed and observed in relatively short time frames such as 34–41 generations in laboratory experiments (Rice 1996; Martin & Hosken 2003). However, empirical data demonstrating the contemporary evolution of secondary sexual traits in natural populations are quite limited, and in most cases sexual selection was not the main selective cause. This contrasts markedly with the data for contemporary evolution by natural selection, which include case studies from dozens of animal species (Hendry & Kinnison 1999; Kinnison & Hendry 2001).

If we apply the criteria for contemporary evolution used by the authors above (evolution on a time scale of 100 generations or fewer), we are aware of 11 well documented studies on 10 different organisms for which the contemporary evolution of secondary sexual traits has been documented in natural populations (Table 1). Most of these examples are secondary sexual traits that have changed due to either natural selection alone or a change in the balance between natural and sexual selection (Table 1). In only a few cases did the secondary sexual traits evolve primarily or only by sexual selection (Table 1).There is thus a clear challenge for future investigators to document additional cases of contemporary evolution driven primarily by sexual selection.

Table 1.  Contemporary evolution of secondary sexual traits in natural populations. The table contains examples of both quantitative (continuous) traits and discrete traits (polymorphisms). The criterion for contemporary evolution is that the change took place during 100 generations or less (see text). Five criteria had to be fulfilled for inclusion in this list. (1) The trait should have, or be likely to have, some genetic basis (at least partly). (2) It should be either a longitudinal study over several generations (consist of ‘before’ and ‘after’), or come from populations with a common ancestor at a known time in the recent past. (3) Selective causes and selective agents should be known or inferred. (4) The trait should have some clear function in sexual selection. (5) The study should have been performed in a natural population.Estimates of the rates of change are included (in haldanes where applicable). Estimates of evolutionary rates were taken from original publications or estimated after contacting the authors and obtaining complementary data.
OrganismTraitTrait typeGenerations (years)Selective agentRate/directionSource
  1. Abbreviations: P, polymorphic trait ; Q, quantitative (continuous) trait ; S, sexual selection; N, natural selection.

  2. NA, not available or not applicable (in the case of polymorphic traits).

  3. Superscripts: b, breeding values; p, phenotypic values.

  4. –, trait has decreased in average value (or frequency in the case of polymorphic traits); +, trait has increased in average value (or frequency in the case of polymorphic traits).

Barn swallowTail lengthQ10 (20)Climate change (female choice)0·130p/+Möller & Szep (2005)
 Hirundo rustica      
Bighorn sheepMale weightQ 5 (30)Humans (trophy hunting)–0·14b/–Coltman et al. (2003)
 Ovis canadensisMale horn sizeQ 5 (30)Humans (trophy hunting)–0·15b/– 
Blue-tailed damselflyFemale colour morph frequencyP 4 (4)Male mating harassmentNA/fluctuatingSvensson et al. (2005)
 Ischnura elegans      
Chinook salmonJaw length, hump sizeQ26 (90)Migration costs0·101p/–Kinnison et al. (2003)
 Oncorhynchus tshawytscha Q26 (90) 0·045p/– 
Collared flycatcherForehead patch sizeQ11 (22)Climate change–0·053p/–Garant et al. (2004)
 Ficedula albicollis    –0·005b/– 
Collared flycatcherForehead patch sizeQ 8 (15)Climate change–0·093p/–Hegyi et al. (2006)
 F. albicollis      
Dark-eyed juncoWhite on outer tail feathersQ 8 (19)Climate?–0·19p/–Yeh (2004)
 Junco hyemalis      
Field cricketMale courtship (song)P 5 (20)Parasitoid flyNA/–Zuk et al. (2006)
 Teleogryllus oceanicus      
GuppyMale coloration (four traits)Q 3 (2)Predatory fish0·570p/+Endler (1980)
 Poecilia reticulata      
Side-blotched lizardColour morph frequenciesP 7 (7)Intrasexual interactionsNA/cyclicSinervo & Lively (1996)
 Uta stansburiana     Sinervo et al. (2000)
Sockeye salmonMale body depthQ13+Male–male competition0·098–0·157pHendry et al. (2000)
 Oncorhynchus nerka     Hendry 2001

Our list of examples (Table 1) includes 10 different organisms and 12 traits; the number of generations over which the evolutionary change took place varies between three and 90 generations. Of the 12 traits that have been studied, nine were quantitative, continuous traits. We present estimates of evolutionary rates (in haldanes: trait changes expressed standard deviations per generation) where data are available and where applicable (Table 1). Estimates of evolutionary rates range from 0·005 (breeding values for the forehead patch in collared flycatchers, Ficedula albicollis) to 0·570 (mean rate of all coloration traits in guppies, Poecilia reticulata) with an arithmetic mean of 0·14 (SD = 0·14; N = 12 estimates), a geometric mean of 0·095 and a median of 0·12. The average rate estimates for secondary sexual traits are almost three times higher than similar estimates for all kinds of trait during time frames <80 generations (geometric mean 0·032; median 0·035) (Kinnison & Hendry 2001). We note that our examples include one of the fastest evolutionary rates documented so far: the change in male guppy coloration following an experimental introduction (Hendry & Kinnison 1999). However, this rate was estimated over only three generations, and estimated rates scale negatively with the number of generations (Gingerich 1983). It is thus possible that the average evolutionary rate for secondary sexual traits would become lower and approach the level of other traits if they were to be estimated over longer time intervals. Our list also includes three discrete polymorphic traits, including sexually selected colour polymorphisms in lizards (Sinervo & Lively 1996; Sinervo et al. 2000) and damselflies (Svensson et al. 2005), as well as the recent and rapid increase in the frequency of a quiet male cricket type (‘flatwing’) in response to natural selection against singing males from acoustically oriented parasitoid flies (Zuk et al. 2006).

Examples of the contemporary evolution of secondary sexual traits driven primarily by natural selection include the above-mentioned cricket study, and two independent studies on the decline in the sexually selected forehead patch size of male collared flycatchers (Garant et al. 2004; Hegyi et al. 2006). The flycatcher case is particularly interesting as data come from two independent populations (the island of Gotland in Sweden and Hungary in continental Europe). In both these geographically separated flycatcher populations, the male forehead patch declined during the past two decades. In both cases investigators were able to link this decline to changed climatic conditions caused by the North Atlantic Oscillation (Table 1). Both studies also present evidence, based on the strength and sign of selection gradients (male viability), that the observed change was driven by natural selection against large patches, rather than sexual selection. In these flycatchers, climatic factors were attributed to a decline of a male secondary sexual character, in striking contrast to the increase of another secondary sexual character (tail length in swallows) where climate was also suggested to drive evolutionary change (Table 1). The increase in male body depth of introduced beach populations of sockeye salmon, Oncorhynchus nerka (Hendry et al. 2000; Hendry 2001) was probably also a result of local divergent natural selection regimes against deep male body sizes, although the male humps are targets of intrasexual selection mediated by male–male competition (A. P. Hendry, personal communication). There is also one example of a human-induced change in secondary sexual traits: selective harvesting through trophy hunting on horn size and weight in bighorn sheep (Ovis canadensis) caused the estimated breeding values of these two traits to decline over the course of 30 years, corresponding to about five generations (Coltman et al. 2003).

Secondary sexual traits potentially can also change due to an altered balance between sexual and natural selection. For instance, if natural selection changes in strength and/or direction between populations, the trait will evolve even if sexual selection regimes remain constant between populations. One such example could be the changed trade-off between allocation towards secondary sexual traits vs migratory traits among introduced populations of chinook salmon, Oncorhynchus tshawytscha (Kinnison et al. 2003). Lower investment in smaller secondary sexual traits in migratory populations was apparently favoured by natural selection to facilitate migration, although there is no information available as to whether sexual selection regimes between populations remained constant, or changed. Another possible example of an altered trade-off between natural and sexual selection is the study of Yeh (2004) on darkeyed juncos (Junco hyemalis), which have invaded the campus of the University of California, San Diego in southern California from the nearby high-elevation mountain habitats. Over the course of eight generations, both male and female coastal birds reduced the amount of white in their outer tail feathers by 22% (Fig. 1). Yeh provided several lines of evidence indicating that selection, rather than genetic drift or plasticity, was responsible for the observed change in this signalling trait. The selective cause could potentially be due to an altered optimal trade-off between investment in secondary sexual traits vs parental care, given the longer available breeding season in the novel environment (Yeh & Price 2004). In contrast to the results in this junco study, maternal effects have been suggested to play an important role in the rapid morphological population divergence and changed sexual dimorphism that took place when house finches (Carpodacus mexicanus) spread and established new populations across North America (Badyaev et al. 2000, 2002, 2003).

Figure 1.

Contemporary evolution of the amount of white in tail feathers of the darkeyed junco, Junco hyemalis (Yeh 2004). (a) White in the outer tail feathers functions as a sexual and social signal, as when an individual spreads its tail. (b) Geographical locations of putative source populations (▴) in the mountains of California, and the newly established coastal population at the University of California, San Diego, which was founded by immigrants from the mountains (▵). (c) The amount of white in the outer tail feathers differs between mountain and coastal populations, both among wild-caught birds and nestling birds that were hand-reared in a common environment. The graph has been modified with permission from the Society for the Study of Evolution (SSE) (Yeh 2004).

There are only a few examples where sexual selection alone or primarily appears to have been responsible for the contemporary evolution of secondary sexual traits. These cases include the increase in tail length of male barn swallows (Hirundo rustica) in response to climate change (Möller & Szep 2005), and the rapid morph-frequency dynamics across multiple generations in longitudinal population studies of lizards and damselflies. Colour polymorphisms are examples of heritable traits that often have a very simple genetic basis of one or a few loci (Svensson & Abbott 2005). If an investigator tracks several generations of such polymorphic species, changes in morph-frequency composition between generations are (by definition) evolutionary changes, as the genetic composition of the population(s) has changed. The challenge for investigators remains to clarify if such morph-frequency changes are caused by stochastic processes such as genetic drift or by natural or sexual selection, and the selective causes for the fitness differences among morphs.

In side-blotched lizards (Uta stansburiana), the frequencies of three male heritable throat colour morphs change rapidly between generations due to negative frequency-dependent sexual selection in favour of rare male phenotypes (Sinervo & Lively 1996). Each morph has its strengths and weaknesses in pairwise interactions with the two other morphs, which results in cycles of morph frequencies across generations in ‘rock–paper–scissor’ dynamics (Sinervo & Lively 1996). This lizard system thus exemplifies how rapid evolutionary dynamics by sexual selection could be driven by male–male competition, with little or no role for female choice. Similarly, in the polymorphic blue-tailed damselfly (Ischnura elegans) there are three heritable colour morphs in females, one being a male mimic (Cordero 1990; Svensson & Abbott 2005). The fecundities of these three female morphs are negatively frequency-dependent, and this maintains all three female phenotypes in the populations studied to date (Svensson & Abbott 2005). Although there is no strict evidence for regular cycles in this species, unlike in the side-blotched lizards, the system is evolutionarily dynamic, with rapid changes in morph frequencies across generations (Fig. 2). Experiments are currently under way to investigate the ecological causes behind the frequency-dependent fitnesses of the three female morphs (E.S. and T.P.G., unpublished data). Several lines of evidence indicate that male mating harassment of common female morphs leads to a form of apostatic selection that causes frequency-dependent sexual conflict over mating rate (Fincke 2004; Svensson et al. 2005).

Figure 2.

Rapid evolutionary change in morph frequencies in the wild and during four generations of the blue-tailed damselfly, Ischnura elegans (Svensson et al. 2005). Ternary plots show the annual frequencies of three female colour morphs in two different populations during four generations (one generation per year). Each apex shows a situation when the population is monomorphic (100% of a particular morph); interior points are polymorphic situations. The morph frequency of each particular morph is proportional to the distance from each vertex, and the mid-point of each ternary plot represents the situation with a frequency of 1/3 for each morph. A, androchrome females (male mimics); I, infuscans females; Io, infuscans–obsoleta females. Morph frequencies, and hence the genetic composition of both populations (a) and (b), changed significantly between the four generations. Morph-frequency dynamics result from frequency-dependent female fecundities thought to arise through frequency-dependent male-mating harassment of female morphs. Reproduced with permission of the American Society of Naturalists (Svensson et al. 2005).

These two examples of contemporary evolution driven by intrasexual selection (side-blotched lizards) or sexual conflict (damselflies) are thus not directional changes, but were either cyclic or moving towards an equilibrium morph frequency. Sexual selection, sexual conflict and many other forms of social selection are often frequency-dependent processes (West-Eberhard 1983; Gavrilets 2004). Sexually selected signalling traits (e.g. coloration traits) could, in many cases, also be governed by one or a few loci (Price 2002). The combination of frequency-dependent social selection and relatively simple genetic architectures of some secondary sexual traits could make their evolutionary dynamics quite different from naturally selected, quantitative traits subject to unidirectional change, such as those reviewed previously (Hendry & Kinnison 1999).


The empirical data on the contemporary evolution of secondary sexual traits is thus not overwhelming, at least not from natural populations. Here we discuss briefly some explanations for the relative paucity of such data. Although our list of documented changes in secondary sexual traits may not be complete (Table 1), it is obvious that most studies are relatively recent. Of the examples we list, only two (guppies and male side-blotched lizards) were published prior to 2000. Although there may be some additional examples in the past that we have overlooked, we nevertheless believe our list indicates that the study of contemporary evolution of secondary sexual traits has been selected, at least until quite recently.

Although the few documented changes of secondary sexual traits indicate that such changes may have been overlooked, we would not a priori exclude the possibility that there may be some interesting biological explanations behind the relative rarity of good examples. Explanations for lack of evolutionary change when we expect to see it could be based on either ‘internal’ factors (genetic and/or developmental constraints) or ‘external’ factors (strength and mode of selection). Both internal and external explanations echo the classical debate focused on naturally selected traits, and the causes of long-term stasis (absence of evolutionary change) in phenotypic evolution (Eldredge & Gould 1972; Charlesworth et al. 1982; Meriläet al. 2001; Hansen & Houle 2004). We will not dwell on this debate here, but just make a few points that are particularly relevant to the discussion about sexually selected traits.

Explanations for evolutionary stasis based on lack of additive genetic variance (Meriläet al. 2001), or because of strong genetic covariance between traits (Hansen & Houle 2004), have been the most popular and discussed explanations in the past. However, there are several good reasons to believe that these genetic explanations are, in themselves, unlikely to explain stasis (Meriläet al. 2001; Hansen & Houle 2004). Although many secondary sexual characters could have low levels of additive genetic variance (Hall et al. 2004), it is unlikely to be totally absent. Some empirical data and theory indicate that life-history traits and secondary sexual traits often have high levels of additive genetic variance, compared with morphological or physiological traits (Houle 1992; Pomiankowski & Möller 1995; Rowe & Houle 1996).

Genetic covariance with other traits that have antagonistic effects on fitness could slow down the rate of evolution by sexual selection, but would not entirely prevent change. This is because it is quite unlikely that the genetic correlation(s) would =1, and there would thus still be an amount of ‘free genetic variation’ or ‘conditional heritability’ available to fuel evolutionary change (Hansen et al. 2003; Hansen & Houle 2004). Although low additive genetic variance and genetic covariance between traits are unlikely to explain stasis, they could potentially slow down the rate of evolutionary change and thus make detection of the contemporary evolution of secondary sexual traits a more difficult and challenging empirical task.

Another genetic factor that should slow down evolutionary changes in secondary sexual traits may be the type of genetic variation. If sexually selected traits are subject to continuous and chronic directional selection, quantitative genetic theory predicts that most of the standing genetic variance of these traits will be replenished by new deleterious mutations every generation (Houle 1991, 1992). This mutational input every generation creates a long tail of maladapted genotypes, and the role of sexual selection becomes purifying, and restricted to weeding out these maladapted genotypes or low-quality males (Rowe & Houle 1996). This is a model of sexual selection that is condition-dependent: males in low genetic condition are those males that have many deleterious mutations, which in turn prevents them from fully developing their sexual ornaments (Rowe & Houle 1996). Whether this general explanation for sexual selection is correct or not, we note that sexual selection is not expected to drive the population towards novel trait values in this scenario, as most mutations are deleterious and sexual selection's role is primarily to remove them, leading to a mutation–selection balance at equilibrium (Lorch et al. 2003). The secondary sexual traits would then evolve only if the underlying genetic variation in condition also evolved, permitting changed expression of these condition-dependent traits.

The most frequently invoked external explanation for stasis has been long-term stabilizing selection on the traits around a more-or-less stable optimum (Charlesworth et al. 1982; Hansen 1997). Although sexual selection is often assumed to be directional, as in the runaway models, there are some examples of stabilizing multivariate sexual selection (Brooks et al. 2005). Could such stabilizing selection potentially lead to stasis in secondary sexual traits? We believe this is unlikely, and there are some general objections against invoking stabilizing selection as a general explanation for stasis (Hansen & Houle 2004). It is highly unlikely that the selective factors operating on the trait(s) that are subject to stabilizing (or conflicting) selection would remain entirely constant over a large number of generations, leading instead to a slowly moving optimum, perhaps constrained within certain phenotypic boundaries (Arnold et al. 2001; Hansen & Houle 2004; Estes & Arnold 2007). Such slow-moving optima would lead to slower rates of evolution, but they would not lead to complete stasis and an entire absence of evolutionary change. Stabilizing selection is thus unlikely to be a general explanation for stasis, and a similar argument could be invoked against spatially varying selection or other forms of conflicting selection (Svensson & Sinervo 2004). However, even if they did not lead to stasis in the long term, slow-moving optima would make the empirical detection of contemporary evolution more difficult for empiricists studying extant natural populations.

Finally, gene flow between conspecific natural populations could certainly constrain evolutionary divergence, and has even been put forward to explain stasis (Eldredge et al. 2005). Gene flow will constrain population divergence between populations that are subject to divergent selection due to differences in ecology and associated selection regimes (Hendry et al. 2002; Hendry & Taylor 2004; Svensson et al. 2006; Garant et al. 2007). Although gene flow is certainly not an explanation that is likely to be valid in all cases, and is by no means limited to secondary sexual traits, its role in constraining population divergence in secondary sexual traits is largely unexplored. If gene flow is important in constraining evolutionary divergence by sexual selection in natural populations, we would expect to see pronounced evolutionary changes of both secondary sexual traits and other phenotypic characters following the development of reproductive isolation (Futuyma 1987).

Toward a population biology of sexual selection: future directions

There is a discrepancy between the well recognized role of sexual selection in theoretical models, and the relatively limited empirical evidence for its effect in transforming trait values in natural populations. We have discussed some possible explanations for this gap, and touched briefly on both genetic and selective explanations. We have also suggested that such changes may have been overlooked until recently. It is also possible that the relatively few examples in this area may be due to the specific research traditions in sexual selection. Promising avenues for the future will be to move away from the current focus of identifying the trait(s) that are targets of sexual selection, to try instead to understand spatial variation in sexual selection and the covariance between patterns of sexual selection and population variation in secondary sexual traits (Houde & Endler 1990; Endler & Houde 1995; Svensson et al. 2006).

Spatial studies of sexual selection cannot, in themselves, reveal the contemporary evolution of secondary sexual traits, in contrast to longitudinal population studies (Table 1). However, spatial studies can complement longitudinal studies by providing more detailed knowledge about the ecological causes of fine-scale variation in sexual selection pressures. Spatial studies may also generate new hypotheses that can subsequently be tested in longitudinal studies. Although spatial studies have been performed previously in natural populations, they have been limited to a few well studied model systems and to comparisons between geographically distant populations. We would especially call for studies performed on small spatial scales, where the antagonism between gene flow and selection will be especially pronounced (Smith et al. 1997; Svensson & Sinervo 2004; Svensson et al. 2004, 2006; Garant et al. 2005). Comparisons between closely related populations will thus be more informative in this context, compared with comparative studies of populations that are widely geographically separated (Badyaev et al. 2000), and where selection can operate more-or-less unopposed by gene flow (Garant et al. 2007). Comparisons of selective regimes between closely located populations experiencing gene flow should reveal both the power and the limits of selection in transforming trait values (Lenormand 2002).

Spatially replicated studies can provide useful information about population variation in preference functions (Mead & Arnold 2004) and estimates of directional or curvilinear selection (Kingsolver et al. 2001). We have recently initiated such spatially explicit studies of sexual selection in two species of damselfly (Odonata), which we discuss briefly here. We do not claim that these studies are complete, or that they provide strong evidence for contemporary evolution driven by sexual selection. Rather, we want to highlight two promising and hitherto underutilized experimental approaches to population variation in sexual selection: reciprocal transplant experiments and visualization of spatial variation in selective regimes using geographic information systems (GIS).

Our first example concerns sexual selection in the blue-banded demoiselle (Odonata: Calopteryx splendens). Both European and North American species of Calopteryx have been classical model organisms in the past, mainly because they are relatively easy to study in the field using mark–resight methods (Svensson et al. 2004). In several species of Calopteryx the males have dark, melanized wing patches, and wing coloration has important functions in intrasexual selection (male–male competition) and in intersexual selection as the target of female choice (Siva-Jothy 1999; Tynkkynen et al. 2004). The amount of melanistic wing coloration is correlated with individual disease resistance (Siva-Jothy 2000), is important in male interspecific territorial interactions (Tynkkynen et al. 2005), and functions in reproductive isolation when species coexist in sympatry (Waage 1979). The multiple functions of melanistic wing coloration contribute to making this an ecologically highly interesting trait that is the target of both natural and sexual selection.

We recently estimated sexual selection on wing coloration and other male traits in several populations of C. splendens, with the goal of understanding how intrapopulation sexual selection affects population divergence and sexual isolation between a series of parapatric populations in southern Sweden (Svensson et al. 2006). Sexual selection on male wing-patch length differs in both magnitude and direction between populations, and also fluctuates between years within populations (Svensson et al. 2004). There is some degree of gene flow between our 12 study populations in southern Sweden, and the degree of neutral population genetic differentiation (measured through amplified fragment length polymorphism analysis) amounts to FST = 0·05 (Svensson et al. 2004). In six out of seven populations, local females prefer local males as their mates, rather than immigrant males, which was revealed by experimental tethering experiments in the wild (Fig. 3a). These results indicate incipient sexual isolation between populations, and we have suggested that sexual isolation may have arisen as a correlated response to divergent sexual selection rather than to divergent natural selection, which is considerably lower in magnitude in this system (Svensson et al. 2006). These data are consistent with so-called ‘divergence-with-gene-flow’ models, in which populations can evolve sexual isolation even in the presence of gene flow (Rice & Hostert 1993; Smith et al. 1997).

Figure 3.

Sexual isolation in the face of gene flow: courtship success of resident and experimental ‘immigrant’ males of the banded demoiselle, Calopteryx splendens (Svensson et al. 2006). Male courtship success of residents and immigrants was tested in tethering experiments, in which female responses were recorded on a nominal scale (average values for each population and 95% CI for each male category). (a) In six of seven populations, resident males had higher courtship success than immigrant males, and this difference is significant. (b) Results from a reciprocal transplant experiment in which immigrants were tested in two directions for two intensively studied populations (Klingavälsåns and Höje Å). There is evidence for a significant interaction between male origin and male category (resident or immigrant), revealing that males are locally adapted to the preferences of their local females. Graph reproduced with permission of the Society for the Study of Evolution (Svensson et al. 2006).

For two of our study populations (Klingavälsåns Naturreservat and Höje Å), we performed reciprocal transplant experiments in which we assessed courtship success of resident and immigrant populations at each site and in both directions (Fig. 3b). Such reciprocal transplant experiments have, to our knowledge, seldom been used in sexual selection studies, but have been a successful research approach in studies of local adaptation and divergent natural selection (Schluter 2000; Kawecki & Ebert 2004). Our reciprocal transplant experiments (Fig. 3b) revealed that this may also be a fruitful experimental approach to estimating the degree of divergent sexual selection (Svensson et al. 2006). In both populations, resident male phenotypes had the highest courtship success (Fig. 3b), and this resulted in a significant interaction between phenotype of origin (population) and status (resident or immigrant). This significant interaction term is exactly analogous to the interaction terms that have been used in the past to infer local adaptation and divergent natural selection (Schluter 2000; Kawecki & Ebert 2004). Potentially, premating isolation of the kind we have documented should facilitate contemporary evolution and population divergence of secondary sexual traits by reducing gene flow (Futuyma 1987). However, the relative importance of female choice in this system is unclear, and its effects in reducing gene flow could possibly be counteracted by forced copulations from males (E.S., unpublished observations). Evolutionary change and population divergence of secondary sexual characters have been argued to reflect a balance between the diversifying effects of female choice and premating isolation, and the constraining effects of male-enforced matings (Magurran 1998; Parker & Partridge 1998).

Our second example concerns spatial variation in sexual selection on male body size in the blue-tailed damselfly (T.P.G. and E.S., unpublished data). We studied sexual selection on male body size along a coastal–inland cline in southern Sweden (Fig. 4). Male and female body size, revealed as the first principal component (PC1) from five morphological traits, increases from coastal to inland populations, presumably because of longer development times as one moves inland However, the strength and magnitude of sexual selection on male body size show no corresponding clear clinal pattern, but instead vary erratically and over a short spatial distance in a mosaic fashion, with both ‘hot spots’ (areas of selection towards large male body size) and ‘cold spots’ (areas of selection towards small male body size) (Fig. 4). This geographical sexual selection mosaic has been observed in all four seasons we have studied to date (2003–06), and appears in part to be driven in part by interactions between local population density and local male body size (T.P.G. and E.S., unpublished data). Given the unpredictable and temporally variable sexual selection regimes in this system, in combination with wind-mediated gene flow between populations (Abbott 2006), evolutionary population divergence of male body size is likely to interfere with these and other stochastic factors. Instability of sexual selection regimes, like those documented in this study, thus could potentially obscure the detection of contemporary evolution driven by sexual selection. The question as to whether sexual selection varies between populations in a clinal fashion (which is the basis of some theoretical models; Lande 1982) or in a more mosaic pattern (as our results indicate) should be a fruitful avenue of future research. Application of new and powerful graphical and statistical tools such as Geographical Information Systems (GIS) will cast new light on this and other issues in the population biology of sexual selection.

Figure 4.

Geographical sexual selection mosaic in the blue-tailed damselfly (Ischnura elegans) along a coastal–inland gradient in southern Sweden (strait of Öresund on left of map). The map illustrates spatial variation in selection differentials on male body size (PC1) with selection for large male size indicated in red and selection for smaller male body size in green. Note the presence of ‘hot spots’, areas of strong selection on male body size (either positive or negative). Map created from estimates of selection differentials across 12 populations (using ArcGIS spatial analysis software). The geographical sexual selection mosaic also varies between different years (2003–06) and does not follow the clinal pattern for male body size (PC1), which increases in a linear fashion from coastal to inland populations across all years (r2 = 0·47, P = 0·002). Fitness functions (cubic splines) and selection coefficients (s and s2, respectively) of male copulation probability vs PC1 are shown for two study populations. Letter combinations are abbreviations for names of study populations.

Summary and conclusions

Although the theory of sexual selection was first formulated over a century ago (Darwin 1871), much work remains to be done, especially in natural populations. We have suggested that further research focusing on the population biology of sexual selection would be a promising avenue for future studies. Our call for spatially and temporally replicated field studies of sexual selection echoes the message in a recent and influential meta-analysis of selection studies in natural populations (Kingsolver et al. 2001). Descriptive estimates of population variation in sexual selection ideally should be complemented with reciprocal transplant experiments and application of powerful new statistical and graphical methods such as GIS and spatial autocorrelation statistics. Future studies in the population biology of sexual selection should integrate such spatially explicit sexual selection work with estimates of gene flow and the degree of sexual isolation to connect to issues explored in current speciation research (Garant et al. 2007; Hendry et al. 2007).


We are grateful to Andrew Hendry for encouraging us to write this review, which forced us to reconsider some of our previous thoughts and assumptions on the topic. We would also like to thank J. Abbott, T.D. Price and J. Wilkinson for criticisms on the first version of this manuscript, and G. Hegyi, D. Garant and A.P. Hendry for providing access to unpublished data. This study was supported by grants from the Swedish Research Council (VR) and the Swedish Council for Environment, Agriculture and Spatial Planning (FORMAS) to E. I. S.