Main

Under the postmating sexual selection hypothesis, selection on male genitalia is caused by mechanisms that generate variation in postinsemination paternity success among males. Such mechanisms include: first, any of several female processes that affect male paternity success (that is, cryptic female choice3,4,5); second, competition between male gametes for fertilization (that is, sperm competition6,7); and third, evolutionary arms races between males and females over the control of fertilization (that is, sexual conflict4,8,9,10). The key prediction of this hypothesis concerns the relationship between mating system and the rate of genital evolution1,3. In taxa in which females typically mate with only one male (monandry), there can be little variation in male postinsemination paternity success and postmating sexual selection on genitalia will thus be weak or absent. If females mate with many males (polyandry), on the other hand, there will be ample opportunity forvariation in male postinsemination paternity success and therefore for postmating sexual selection also. Under the lock-and-key hypothesis, selection for hybridization avoidance is suggested to impel the evolution of male genitalia with a proper mechanical fit. In contrast to postmating sexual selection, such selection for pre-insemination reproductive isolation would be expected to be more intense in monandrous species than in polyandrous species. A given occurrence of interspecific matings will generally be more evenly distributed among polyandrous females than among monandrous females, leading to lower variation in female fitness in polyandrous species and therefore to a weaker selection for pre-insemination reproductive isolation. Here I analyse a series of phylogenetic contrast, comparing morphological divergence in pairs of related clades of insects with differing mating systems (Fig. 1a). This is the first general quantitative assessment of the rate of genital evolution under polyandry relative to that under monandry.

Figure 1: Comparison of the rate of genitalic evolution in polyandrous and monandrous clades.
figure 1

a, This study is based on a number of phylogenetic contrasts, where pairs of clades that share a common ancestry are compared. In each contrast, a measure of interspecific morphological dissimilarity within a clade that exhibits a polyandrous mating system (P1–4) is divided by the same measure in a related monandrous clade (M1–4) to form a morphometric distance ratio. Thus, dissimilarities between the two clades are ignored. A ratio higher than unity implies that morphological evolution has been more divergent in the polyandrous clade. b, Morphological dissimilarity within a clade was measured as the average Euclidean distance from the species to the mean of the clade in a common multidimensional shape space. The procedure is illustrated here, in two dimensions only, for male genitalia of several species in the two Dipteran genera Dryomyza (polyandrous, P) and Lucilia (monandrous, M). Dashed lines represent the distances from each species to the mean (filled symbols) of the clade. In this case, species in the polyandrous clade are about four times as different from one another as are the species in the monandrous clade (see Table 1). PC, principal component.

Full size image

Comparisons of the rate of evolutionary divergence of complex morphological traits in a set of related species have been hampered by problems with identifying homologous structures, as well as by alack of appropriate methods for quantifying shape variation. Previous comparative studies have often resorted to various subjective ratings of morphological complexity11,12,13. Here I use one of the new tools of geometric morphometrics14, which not only provides objective and quantitative descriptors of shape but also avoids the problem of defining homologous landmarks (that is, structural points with correspondence resulting from descent from the same point in a common ancestor) across species14,15. By describing the outlines of the genitalia of each species with a nonlinear function (see Methods), and by subsequently analysing morphological shape variation among species as variance in the parameters of the fitted functions, this method allows the ordination of all the species in each contrast in a common multivariate morphological shape space (Fig. 1b).

The results of this analysis show that male genitalia evolve much more divergently in taxa in which females mate many times. The shape of male genitalia of polyandrous species were more dissimilar than were those of monandrous species in 18 out of 19 contrasts, and the average morphological distance between the genitalia of polyandrous species was more than twice that of monandrous species (see Table 1 for tests). This pattern did not differ between orders (Kruskal–Wallis analysis of variance, P = 0.84), and the taxonomic distance between the two clades in each contrast did not significantly affect the relative degree of genital divergence within clades (within versus between-family contrasts; Mann–Whitney U-test, P = 0.80). There was no association between the distance ratios of genitalia and the distance ratios of other traits across contrasts (Spearman rank correlation, P > 0.9). The analysis did not reveal any influence of mating system on evolutionary divergence for morphological traits other than genital traits (Table 1), and the distance ratios of genital traits were indeed significantly larger than those of other traits (paired Wilcoxon signed rank test, P = 0.023; Kolmogorov–Smirnov two-sample test, P = 0.003).

Table 1 Morphometric distance ratios for genital and general traits for 19 different phylogenetic contrasts
Full size table

Many factors other than selection could potentially influence measures of interspecific evolutionary divergence within a given clade (such as age of clade, biogeographic characteristics, taxonomic resolution, genetic architecture and mating-system characteristics other than female mating frequency). Given the confounding role of such factors, it is remarkable that monandrous and polyandrous clades differed so consistently in the relative rates of divergent evolution of male genitalia. This study offers three important and consequential insights. First, it provides strong evidence in favour of the postmating sexual selection mechanism of genital evolution, and thus enhances our knowledge of the processes behind this general evolutionary trend1,2,3,16. Future research should attempt to determine which forms of postmating sexual selection are responsible for genital evolution3,10,16. Second, the results indicate that thesame process (that is, sexual selection) may be responsible for the evolutionary elaboration of both primary and secondary sexual traits, suggesting that this old dichotomy, which Darwin17 realized was problematic but nevertheless adopted, should be reconsidered16,18. Third, as traits evolving by sexual selection tend to be more phenotypically and genetically variable than other traits19,20, this study calls for quantifications of the degree of intraspecific variability in genital traits. The prevailing typological view of intromittent genitalia in taxonomy, especially in species definitions, may need to be reconsidered.

Methods

Case selection and data acquisition. I searched for clades suitable for the phylogenetic contrasts21,22 in previous reviews23 and comparative studies24,25,26, as well as in reference databases and on the internet. Three criteria had to be met for inclusion of a clade in the analysis. First, reliable data on female mating frequencies had to be at hand, typically in the form of female spermatophore/ejaculate counts in natural populations or detailed field and/or laboratory studies of mating behaviour. Second, the phylogeny of the species included in a given contrast had to be well established. Third, taxonomic revisions containing high-quality illustrations of male genitalia had to be existent, as such illustrations were used to characterize each species.

I located 19 phylogenetic contrasts, representing four different orders, that met these criteria; all of these contrasts were independent in the sense that no clade was represented in more than one contrast (Table 1). This selection was based on a large number of published articles, as well as on personal contacts with a large number of colleagues. A complete list of these sources can be found as Supplementary information or obtained from the author on request. For each species, I captured the outlines of two trait types (male genitalia and, when available, a general trait) with a digitizing tablet (Summasketch III), using illustrations presented in published taxonomic revisions. For consistency, only one such source was used for each clade to avoid artifactual intraclade variation in morphology. The general traits were wings (nine contrasts), body parts (two contrasts) or legs (one contrast). When more than one general trait was at hand, the trait that exhibited most interspecific divergence between species in the contrast was included in the analysis.

Elliptic Fourier analysis. For each contrast and trait type, the outlines of all species were included in a common elliptic Fourier analysis14,15,27, using the software EFA-Win (http://life.bio.sunysb.edu/morph/soft-out.html). The Fourier analyses were made invariant of size, position and rotation, and all used 30 harmonics (yielding 120 Fourier coefficients). These functions provided a near perfect fit even to the most complex outlines.

Multivariate ordination. For each contrast and trait type, the 120 Fourier coefficients for each species were treated as variables in a principal component analysis, performed on the covariance matrix28,29. The first seven principal components, collectively describing >98% of the shape variation in all cases, were retained for ordination. These principal components form orthogonal dimensions in a multidimensional shape space, where all species occupy a given location (Fig. 1b). To quantify the morphological dissimilarity between the polyandrous species relative to that of the monandrous species, I calculated a morphometric distance ratio, representing the average Euclidean distance in this multidimensional space of the species in the polyandrous clade to their mean (centroid) divided by the corresponding average distance for the monandrous clade in the contrast (Fig. 1b). This ratio measures the amount of morphological variance within the polyandrous clade relative to that within the monandrous clade, ignoring variance due to differences between the clades. In contrasts in which the number of species in the two clades were skewed (difference > 1), the elliptic Fourier analysis and the subsequent multivariate ordination procedure were repeated many times (>10) using a random subsample from the more speciose clade to match the number of species in the less speciose clade, and the average morphometric distance ratio from these repeated measures was used for analysis.