Sexual conflict in Gerris gillettei (Insecta: Hemiptera): intraspecific intersexual correlated morphology and experimental assessment of behaviour and fitness


Julie Turgeon, Département de Biologie, Pavillon Alexandre-Vachon, 1045 av. de la Médecine, Université Laval, Québec, Québec city, Canada G1V 0A6.
Tel.: +418 656 3135; fax: +418 656 2043; e-mail:


The contemporary dynamics of sexually antagonistic coevolution caused by sexual conflicts have seldom been investigated at the intraspecific level. We characterized natural populations of Gerris gillettei and documented significant intersexual correlations for morphological traits previously related to sexual conflict in water striders. These results strongly indicate that sexually antagonistic coevolution contributed to population differentiation and resulted in different balances of armaments between the sexes within natural populations of this species. No-choice mating experiments further revealed that both male and male–female relative arms levels influence copulation duration. However, there were no asymmetries in reproductive behaviour and fitness between sympatric and allopatric mating pairs, suggesting that differentiation by sexual conflict was not sufficient to influence the outcome of mating interactions. Altogether, these results question the relative importance of female connexival spines vs. genitalia traits in mediating pre- and post-copulatory conflict in Gerris.


Sexual conflict arises when the evolutionary interests of the two sexes over reproduction diverge (Parker, 1979, 2006). Conflicts of this kind frequently occur over a variety of traits such as mating rate (Arnqvist & Nilsson, 2000), re-mating behaviour (Stockley, 1997; Arnqvist & Rowe, 2005) and parental care (Westneat & Sargent, 1996; Wedell et al., 2006). Sexual conflict can result in a coevolutionary arms race between the sexes, for example when males develop traits to overcome resistance by females, which, in turn, develop traits to resist males (Arnqvist & Rowe, 2005). Sexually antagonistic selection proceeds rapidly, it can fluctuate both in intensity and in direction, and it is dependent upon population-specific conditions such as density, sex ratio, predation pressure, population size, and genetic and phenotypic variability (Rowe et al., 1994; Gavrilets & Hayashi, 2005). Theoretical models have demonstrated that sexual conflict, often modelled over mating rate, can lead to a variety of outcomes. Among several possibilities, rapid evolutionary changes perpetuating the intersexual arms race can lead to population divergence and allopatric speciation (Rice & Holland, 1997; Parker & Partridge, 1998; Gavrilets et al., 2001; Gavrilets & Waxman, 2002; Gavrilets & Hayashi, 2005). However, when sexually antagonistic traits impose costs on the bearers, the arms race may also come to a halt at some point in the escalation (Gavrilets et al., 2001; Gavrilets & Waxman, 2002; Gavrilets & Hayashi, 2005). This balance of force may be relatively unstable and dynamic along a trajectory of adaptation and counteradaptation in absolute levels of armaments.

The long-term consequences of antagonistic coevolution of traits related to sexual conflict have notably been demonstrated in interspecific comparative studies. Evidence that sexually antagonistic selection can be an engine of speciation is supported by the fact that speciation rate is much more rapid in insect species likely experiencing post-mating conflict (Arnqvist et al., 2000). There is also strong support that sexually antagonistic coevolution has reached a transient or lasting balance in many groups. This is revealed by correlations between male and female traits across related species, exemplifying how different balance of arms level span over an equilibrium coevolutionary trajectory (Arnqvist & Rowe, 2002b; Koene & Schulenburg, 2005; Bergsten & Miller, 2007; Rönn et al., 2007; Anthes et al., 2008).

Evidence that sexual conflict has similar coevolutionary impacts in contemporary populations is less abundant. If sexual conflict fuels intraspecific differentiation among populations, this should eventually translate into differential reproductive success of matings between members of the same versus diverging populations (Parker & Partridge, 1998). Studies testing this prediction have produced mixed results, casting doubts on the widespread importance of sexual conflict as an engine of speciation (Bacigalupe et al., 2007). Sexual conflicts influenced mating interactions between individuals from highly divergent populations (from remote sites or submitted to strong artificial selection) in some studies (Andrés & Arnqvist, 2001; Hosken et al., 2002; Nilsson et al., 2002; Martin & Hosken, 2003) but not in others (Wigby & Chapman, 2006; Bacigalupe et al., 2007). However, when natural and potentially interacting populations were used, mating outcomes differed between members of sympatric versus allopatric populations, suggesting that antagonistic coevolution can fuel reproductive isolation under natural conditions (e.g. Knowles & Markow, 2001; Hebets & Maddison, 2005; Sugano & Akimoto, 2007).

If contemporary sexual conflict results in balanced levels of armaments, evidence for antagonistic coevolution may be inherently difficult to obtain. Evidence for the sole action of sexual conflict can certainly be obtained within single natural population [e.g. sexual conflict over mating in water striders (Rowe et al., 1994; Arnqvist, 1997)]. Also, in cases of frequency-dependent selection, cyclic coevolution may be apparent [e.g. female colour polymorphism in damselflies (Andrés et al., 2002; Svensson et al., 2005)]. However, when sexes are at equilibrium for relatively long periods of time, the balance of force itself may mask significant differences in the outcomes of sexual interactions (Arnqvist & Rowe, 2002a, 2005). Nevertheless, the action of sexually antagonistic selection is population specific, such that different levels of arms balance may be achieved across contemporary natural populations. Thus, in a fashion similar to that observed at the interspecific level, coevolutionary arms race whereby populations reach different balance of armaments could result in the correlated evolution of male and female traits across populations. Without the influence of sexually antagonistic coevolution, such a pattern of correlated morphology is not expected.

In several water strider species, sexual conflict occurs over mating. Males frequently harass females, which vigorously resist costly mating attempts (Rowe et al., 1994; Arnqvist, 1997). Both males and females have evolved grasping and antigrasping (G/AG) traits to persist and resist more efficiently during premating struggles (Arnqvist, 1997; Arnqvist & Rowe, 2002b). Grasping traits in males include a flattened distal part of the abdomen, a long genitalia and large fore femurs. These traits allow a better grip and a tighter union to females during premating struggles. Females resist by means of erected connexival spines, a long genitalia bent downward and large fore femurs that allow them to dislodge males more easily during these struggles.

Sexual conflict over mating played an important role during the evolution of Gerris. Indeed, a comparative study of shape variation among Gerris species revealed that the development of G/AG traits is correlated between males and females across species (Arnqvist & Rowe, 2002b). Moreover, the impact of relative male versus female morphologies was linked to the outcome of antagonistic mating behaviour for species with a relatively unbalanced degree of armaments (Arnqvist & Rowe, 2002a). These findings strongly indicate that sexual conflict has lead to antagonistic coevolution, producing a general pattern of balanced arm levels between the sexes across Gerris species (Arnqvist & Rowe, 2002a,b; Rowe & Arnqvist, 2002).

Much less is known about the contemporary role that antagonistic coevolution may play within Gerris species and populations. Two implicit but important conditions have seldom been verified. First, there should be variation within and among populations for male and female G/AG traits under antagonistic selection. It has been shown that there is variation in G. odontogaster males grasping traits among populations of variable local density (Arnqvist, 1992b). However, among populations variation in female antigrasping traits has never been investigated in Gerris. Second, G/AG traits should have an impact on fitness. Certainly, these traits influence mating success in water striders (Arnqvist & Rowe, 1995, 2005; Arnqvist, 1997). However, mating success does not always translate into reproductive success in these insects, as demonstrated in Aquarius remigis where mating success explained only a third of the variance in paternity success (Vermette & Fairbairn, 2002). Also, the dynamics of sexually antagonistic coevolution, if currently at work within species, is undocumented. If relative arm development has reached variable levels of balance in different populations, G/AG traits should be correlated across populations within species. Moreover, if the various levels of armaments are participating in effective population differentiation, the reproductive outcomes of sympatric versus allopatric matings should differ. To our knowledge, these relationships have never been established in Gerris.

In this study, we examine whether sexually antagonistic selection plays a role in the contemporary evolution of G. gillettei (Lethierry & Severin, 1896), a species displaying intermediate level of development of traits related to sexual conflict (Arnqvist & Rowe, 2002b). We document extensive within- and among-population variation in male and female traits involved in sexual conflict. Intersexual correlations for these traits strongly indicate that sexually antagonistic coevolution is currently at work within this species. In addition, no-choice mating experiments revealed that individual male morphology and relative mating pair morphologies influenced behaviour, whereby arms level influenced copulation duration. However, reproductive behaviour and fitness of sympatric versus allopatric mating pairs did not differ, offering no evidence that sexual conflict favours isolation among populations.

Materials and methods


Between 2006 and 2008, nineteen natural populations of G. gillettei (Gagnon & Turgeon, 2010) were sampled near Kuujjuaraapik, northern Quebec, Canada (55°17′N, 77°43′W). Water striders were collected from small tundra ponds where sex ratio was approximately 1 : 1 (M.-C. Gagnon, pers. obs.). In this region, G. gillettei is univoltine and 99% of individuals collected were wingless. In each pond, between 29 and 60 water striders were collected from the water surface (mean = 50.5/pond; N total = 959). Among the nineteen populations sampled, four were used for mating experiments. All individuals collected were preserved in 95% ethanol until morphological characterization.

Morphological characterization

Variation in natural populations of G. gillettei was analysed to assess whether G/AG traits related to sexual conflict (Arnqvist, 1997; Arnqvist & Rowe, 2002b) were variable and whether male and female morphologies for these traits were correlated across populations. Individual pictures were taken with a digital camera (PixeLINK PL-A686C) attached to a dissecting microscope (Olympus B071), with individual animals positioned in lateral view (Arnqvist & Rowe, 2002b). Landmarks were placed with TpsDig version 2 (Rohlf, 2004), and linear measurements were taken with the TMorphGen6 module of the IMP package (Sheets, 2004). Grasping traits measured in males were fore femur width (2), abdomen slope (3), genitalia length (4) and height (5) (Fig. 1a). Antigrasping traits measured in females were fore femur width (2), genitalia length (4) and height (5), connexival spine height (6), width (7) and slope (8) (Fig. 1b). Traits presumably not related to sexual conflict and measured in both sexes were total body length (9), thorax length (10), abdomen length (11) and height (12), and fore femur length (1) (Fig. 1). To estimate measurement variation owing to landmark positioning, landmarks were placed five times on 20 individual pictures (10 individuals of each sex) and measurements were taken anew. Morphological measurements of all traits were highly repeatable, with coefficients of variation between 0.18% and 3.27% for the largest traits (1, 9, 12) and between 1.09% and 7.65% for the smallest traits (2, 4, 5, 6).

Figure 1.

 Morphological traits measured in (a) males and (b) females. Fore femur length (1) and width (2), abdomen slope (3), genitalia length (4) and height (5), connexival spine height (6), width (7) and slope (8), total body length (9), thorax length (10) abdomen length (11) and height (12).

Principal component analyses (PCA) were performed using all morphological traits, but separately for males and females. First, all sampled populations were included in the analyses so as to characterize the overall variation present within and among natural populations. For each population, mean score values of males and females on each principal component (PC1–PC3) were used in correlation analyses. Second, PCA analyses were performed using only the experimental populations to better represent the variation and relative morphologies of this particular set of populations. In both analyses, PC4 and subsequent PCs accounted for approximately 10% of the variation or less and were not analysed any further. All analyses were carried out using SAS 9.1 (SAS Institute Inc., 2002).

Mating experiments

In 2007, no-choice mating experiments were performed to assess whether individual and mating pair morphologies, in particular for G/AG traits, influenced reproductive behaviour and fitness. These experiments also allowed testing for differences in reproductive behaviour and fitness when mating involved male and female from the same (sympatric) or different (allopatric) populations.

Adults were captured from four ponds early in the season. Phenology of sexual maturation is highly variable within and among ponds (M.-C. Gagnon and J. Turgeon, pers. obs.); experimental populations were chosen on the basis of synchronicity in sexual maturity. All individuals were collected before any mating activity could be observed. Although most females were likely virgin so early after ponds thawed in the spring, sexes were kept separately in the laboratory during a female sperm depletion period of approximately 10 days (Arnqvist & Danielsson, 1999; Danielsson & Askenmo, 1999; Danielsson, 2001). Because the number of adults available in natural populations was limited, two sets of two populations were formed. Within each set, males and females were randomly chosen to form sympatric or allopatric mating pairs. Thirty mating trials were performed for each mating type in each population set [30 trials × 2 mating types (allopatric/sympatric) × 2 population sets = 120 trials]. For each trial, a mating pair was put in a plastic container half filled with water (16 × 12 × 7 cm) where reproductive behaviour was observed for one hour. In all, mating trials extended over a period of 5 weeks. Individuals were fed ad libitum with frozen Drosophila and a mix of fresh insects collected outdoors during the entire experiment.

For each mating pair, six reproductive behaviour variables were recorded (Arnqvist, 1992a; Rowe, 1992): (i) number of male mating attempts, (ii) number of female rejects, (iii) time elapsed before copulation (s), (iv) number of female somersaults while in copulation, (v) copulation duration (s) and (vi) guarding duration (s). If copulation did not happen after 1 h of observation, mating pair members were replaced. After copulation, males were immediately put in ethanol whereas pieces of styrofoam were added to the plastic containers to serve as resting and oviposition substrate for the females.

Three fitness variables were estimated for each female: (i) fecundity (number of eggs laid), (ii) fertilization success (% eggs fertilized) and (iii) hatching success (% hatching eggs). Fecundity and fertilization success were monitored daily over a period of 12 days, after which females were put in ethanol for morphological measurements. Eggs were kept for 12 more days to record the proportion of fertilized eggs that hatched during this period.

Principal component analyses revealed that individual PC scores reflect within sex differences in G/AG traits in male and female experimental populations (see Results). Thus, individual PC scores were used to relate individual and mating pair morphologies with behaviour and fitness. First, correlations were performed between individual PCs scores, behaviour and fitness variables, treating sexes separately. Second, the relative morphologies of mating pair members was characterized to investigate whether specific matches influenced behaviour and fitness. For each mating pair, the differences between male and female PC scores (exp. PC1M–PC1F) were computed, resulting in nine morphological distances. Each distance describes the relative development of male versus female traits specifically associated with the PC axes considered. Positive and negative distances indicate that male characteristics are relatively more and less developed relative to those of the other sex, respectively. Distances near zero mean that trait values of both sexes are equally distant from average values.

Finally, the impact of mating type, population and their interaction on behaviour and fitness variables was assessed by means of manovas performed separately for each set of populations and each set of dependent variables (reproductive behaviour and fitness). If manova was significant, anovas were then performed on each dependent variable to determine which variable contributed to the significant effect detected. As the dependent variables did not conform to the assumptions of normality and variance homogeneity, all analyses were performed on ranks. Analyses were carried out using SAS 9.1.


Morphological variation and trait association

Principal component analyses revealed substantial morphological variation within and among all natural populations, as well as among those used in mating experiments (Table 1, Fig. 2). In both analyses and for both sexes, principal components described similar patterns of variation; variables were generally similarly associated, albeit with some differences in strength. PC1 captured variation linked to overall size and explained 33.0% and 33.4% (43.4% and 38.7%) of variation observed in females and males of natural (experimental) populations. All variables except connexival spine slope in females and abdomen slope in males were positively correlated to this axis (Table 1).

Table 1.   Principal components analysis of morphological traits in a) females and b) males of all natural populations sampled (NAT, N = 19) and of populations used in mating experiments (EXP, N = 4). Collectively, the three principal components explain 59.2% (66.9%) of the variance in the female data set from natural (experimental) populations and 66.1% (67.2%) of the variance in the male data set from natural (experimental) populations. PC loadings >0.40 are in bold.
MeasurementPC1 (NAT)PC1 (EXP)PC2 (NAT)PC2 (EXP)PC3 (NAT)PC3 (EXP)
(a) Females
 Total length0.9100.9050.1310.0420.1130.205
 Thorax length0.7660.8390.1940.0600.2580.167
 Abdomen length0.8610.8910.057−0.095−0.3320.341
 Abdomen height0.4100.678−0.450−0.0940.073−0.321
 Fore femur length0.6540.7480.0630.1490.027−0.106
 Fore femur width0.3620.465−0.1340.0860.4590.431
 Genitalia length0.4670.4730.170−0.2910.6500.660
 Genitalia height0.4540.665−0.2670.0040.524−0.236
 Spine height0.3010.4140.6020.645−0.055−0.170
 Spine width0.4720.613−0.385−0.250−0.254−0.317
 Spine slope−0.096−0.0800.8640.8820.2700.294
(b) Males
 Total length0.9010.891−0.164−0.213−0.107−0.191
 Thorax length0.6730.772−0.2810.006−0.3100.425
 Abdomen length0.8430.828−0.2510.4460.2600.005
 Abdomen height0.4550.6960.6980.2370.2400.434
 Fore femur length0.6050.6280.1030.257−0.024−0.044
 Fore femur width0.3470.3420.1950.4810.4490.049
 Genitalia length0.4360.322−0.1960.5800.7210.685
 Genitalia height0.3660.5550.6480.472−0.314−0.017
 Abdomen slope−0.108−0.0180.4710.5880.4830.433
Figure 2.

 Variation for antigrasping traits in females (a and b) and grasping traits in males (c and d) from natural (NAT; a and c) and experimental (EXP; b and d) G. gillettei populations. In experimental populations, circles and triangles represent populations from set 1 (1A and 1B) and set 2 (2A and 2B). Dots represent population means and bars standard errors. Refer to Table 1 for a description of PC loadings.

In females, PC2 and PC3 mostly described variation in antigrasping structures and explained 14.9% (12.7%) and 11.3% (10.8%) of total variation in natural (experimental) populations. PC2 clearly described connexival spine variation in both analyses, with positive PC2 values corresponding to relatively tall and erected spines (Table 1a). PC3 described genitalia length and femur width variation in both analyses. Although signs of correlations differed between analyses, traits were similarly associated, indicating that females with wide femurs also had short genitalia. In natural populations, positive values on this axis were also strongly associated with thick genitalia.

In males, PC2 and PC3 mostly described variation in grasping structures and explained 15.4% (16.7%) and 14.3% (11.9%) of total variation in natural (experimental) populations. PC2 described abdomen slope and genitalia height variation in both analyses, with positive values indicating abdomens with a pronounced slope and thick genitalia (Table 1b). In natural populations, this axis was also strongly associated with thick abdomens, whereas in experimental populations, PC2 was more strongly associated with large femurs, short abdomens and short genitalia. In both analyses, PC3 mostly described genitalia length variation, along with abdomen slope. When considering all natural populations, positive PC3 values also indicated narrow femurs. In the experimental populations, PC3 was also associated with small thorax and thick abdomens.

Natural populations spanned over the multivariate PCA space (Fig. 2). Among all natural populations, male mean PC2 and PC3 values were correlated (rP = 0.481, P-value = 0.037), indicating that male populations varied along a gradient ranging from less (negative PC2 and PC3 values) to more (positive PC2 and PC3 values) developed grasping traits (Fig. 2c). In females, there was no obvious association between mean PC2 and PC3 values (rP = −0.100, P-value = 0.685). Tall and erected connexival spines tended to be associated with either short or long genitalia, as well as with wide or narrow femurs (positive PC2 with either positive or negative PC3 values) (Fig. 2a). In experimental populations, males from both population sets were similarly differentiated regarding abdomen slope and genitalia height variation (PC2), but set 2 populations seemed more differentiated in terms of genitalia length and associated morphology (PC3) (Fig. 2d). In females, a similar pattern was observed. Both population sets were similarly differentiated regarding connexival spine variation (PC2), but populations from set 2 seemed more differentiated in terms of genitalia length and femur width variation (PC3) (Fig. 2b).

Male and female traits were significantly correlated across natural populations. First, male and female mean PC1 values were correlated (Fig. 3a; rP = 0.742, P-value < 0.001), indicating that male and female body sizes are correlated across populations. This correlation remained significant after removing an outlier data point (rP = 0.521, P-value = 0.027). Once overall size was accounted for, correlations revealed associations between male grasping and female antigrasping traits. Male and female mean PC3 values were correlated (Fig. 3b; rP = −0.719, P-value < 0.001), indicating that long genitalia and narrow femurs covary in males and females across populations. Female mean PC3 values were also correlated with male mean PC2 values (Fig. 3c; rP = −0.674, P-value = 0.002), suggesting that a more flattened distal part of the abdomen and a thick genitalia in males is correlated with long genitalia and narrow femurs in females among the populations analysed. By contrast, variation in female connexival spine morphology was not related to variation in male grasping traits (female mean PC2 vs. male mean PC2 and male mean PC3 values: rP = −0.185, P-value = 0.448 and rP = −0.100, P-value = 0.685). When considering single trait values (but size-corrected), female genitalia length (high negative loading on PC3) was correlated with male abdomen slope (high positive loadings on both PC2 and PC3), indicating that in populations where females had long genitalia, male tended to have flattened abdomens (rP = 0.526, P-value = 0.021). This result is highly consistent with PCA results of Fig. 3b and c.

Figure 3.

 Correlation between male and female mean population PC scores representing (a) body size (rP = 0.742, P-value < 0.001) and (b, c) grasping/antigrasping traits (rP = −0.719, P-value < 0.001 and rP = −0.674, P-value = 0.002 for (b) and (c) respectively). Refer to Table 1 for a description of PC loadings.

Mating experiments

Individual and relative male–female morphologies influenced reproductive behaviour, but not the fitness components estimated in this study. In males, PC2 values, representing abdomen slope and genitalia height variation, were negatively correlated to copulation duration (Fig. 4a; rS = −0.265, P-value = 0.004). Thus, males with a more flattened distal part of the abdomen and thick genitalia tended to have shorter copulations. The relative morphology of male and female engaged in mating was variable and influenced copulation duration. This was indicated by the negative correlation between copulation duration and the intersexual morphological distance capturing the relative importance of male grasping versus female antigrasping traits development (PC2M–PC2F; Fig. 4b; rS = −0.304, P-value = 0.001). Thus, relatively ‘armed’ males paired with females with relatively short and bent spines had shorter copulation times than other mating pairs. In females, PC3 values, representing genitalia length and femur width variation, were negatively correlated with guarding duration (Fig. 4c; rS = −0.198, P-value = 0.031). Thus, males achieved shorter guarding duration with females having long genitalia and narrow femurs. By contrast, neither individual nor relative male and female morphologies influenced fitness (Appendix 1).

Figure 4.

 Correlations between (a) individual male morphology (PC2M) and copulation duration (rS = −0.265, P-value = 0.004), (b) the difference between male and female PC2 (PC2M–PC2F) and copulation duration (rS = −0.304, P-value = 0.001), (c) individual female morphology (PC3F) and guarding duration (rS = −0.198, P-value = 0.031). Refer to Table 1 for a description of PC loadings.

Mating experiments revealed no asymmetry in reproductive behaviour or fitness between sympatric and allopatric mating pairs. For reproductive behaviour variables, and for both population sets, there was no evidence of differences associated with populations, mating type or interaction between these factors (Table 2a and Appendix 2). For fitness variables, no differences were associated with mating type or interaction with population (Table 2b and Appendix 2). However, there were significant differences between populations in set 2 (Table 2b). anova analyses further indicated that fecundity and fertilization success differed between these populations (Table 3a, b,Appendix 2).

Table 2.   Results of manovas for (a) reproductive behaviours and (b) fitness variables in the first set of experimental populations (Set 1) and the second set of experimental populations (Set 2).
EffectSet 1Set 2
  1. Significant values (P < 0.05) are given in bold.

(a) Reproductive behaviour variables
 Population6, 510.8271.780.1226, 510.9440.510.801
 Type6, 510.7982.150.0646, 510.8351.680.146
 Population × type6, 510.8551.440.2176, 510.8951.000.436
(b) Fitness variables
 Population3, 540.9041.910.1393, 540.63210.46<0.001
 Type3, 540.9960.070.9753, 540.9840.300.824
 Population × type3, 540.9760.430.7293, 540.9271.420.246
Table 3.   Results of anovas for each of three fitness variables (fecundity, fertilization success and hatching success) in the second set of experimental populations (Set 2).
  1. Significant values (P < 0.05) are given in bold.

 Population × type12464.002.180.146
Fertilization success
 Population × type13139.273.540.065
Hatching success
 Population × type1576.600.490.486


Natural populations of G. gillettei displayed substantial intra- and inter-morphological variation, in particular for traits previously linked to sexual conflict over mating in water striders (Arnqvist, 1997; Arnqvist & Rowe, 2002b). In both sexes, variation in size was important, but once the effect of size was removed, most of the remaining variation was explained by differences in G/AG traits (Table 1). In water striders, theory suggests that suites of traits should ensure better persistence and resistance in male–female conflicts. Long genitalia and wide femurs should be associated with a flattened distal part of the abdomen in males, and with tall and erected connexival spines in females (Arnqvist & Rowe, 2002b). In G. gillettei males, abdomen and genitalia traits did covary and defined variable levels of armaments across populations. In females, however, well-developed connexival spines were associated with variable genitalia characteristics across populations. Moreover, narrow femurs were unexpectedly associated with long genitalia in both males and females, and sometimes with tall and erected spines in females. These results concur with other studies indicating that the entire suite of G/AG traits may not be necessary to persist and resist in male–female water strider conflicts. Indeed, G/AG traits are rarely all significantly related to mating success, and the relationship between these traits and mating success vary between populations and species (Arnqvist, 1989; Preziosi & Fairbairn, 1996; Weigensberg & Fairbairn, 1996; Arnqvist et al., 1997; Ferguson & Fairbairn, 2000). Moreover, despite a significant correlation between male and female femur width among Gerris species, this trait was not significantly correlated with other G/AG traits within sex, suggesting that its evolution was independent from the other traits related to sexual conflict (Arnqvist & Rowe, 2002b).

Male and female morphologies were strongly correlated across populations (Fig. 3b and c), providing strong evidence for contemporary antagonistic coevolution owing to sexual conflict in G. gillettei. These correlations are consistent with predictions relating to the dynamics of antagonistic coevolution whereby persistence and resistance traits are expected to be correlated, such that when males traits are more developed, females traits should also be more developed (Arnqvist & Rowe, 2002b). Until now, these predictions have been verified for many organisms but almost exclusively by means of interspecific comparisons (Arnqvist & Rowe, 2002b; Koene & Schulenburg, 2005; Bergsten & Miller, 2007; Rönn et al., 2007; Anthes et al., 2008). Intraspecific intersexual correlated morphologies have been documented among natural populations of diving beetles (Bergsten et al., 2001), but the role of correlated traits in sexual conflict was not firmly established. Here, we show that traits previously linked to sexual conflict (Arnqvist, 1997; Arnqvist & Rowe, 2002b) are clearly correlated between sexes across populations, offering strong evidence for contemporary antagonistic coevolution. The nature of correlated G/AG traits revealed that, as expected, genitalia and femur characteristics were important in both males and females, along with abdomen shape in males. Surprisingly, however, connexival spine morphology, a key antigrasping trait in female water striders (Arnqvist & Rowe, 1995; Ronkainen et al., 2005), was not significantly correlated with any male grasping traits. This result is unexpected under a scenario of differentiation solely driven by antagonistic selection over mating rate. On the one hand, it suggests that female genitalia could be more important than connexival spines for persistence and resistance in G. gillettei male–female conflicts. If so, the observed intersexual correlation can be interpreted as resulting from antagonistic coevolution fuelled by sexual conflict mediated mainly by genitalia traits instead of spine development. On the other hand, other processes, possibly post-copulatory, could also be at work and mask the importance of connexival spines in premating conflicts. In this case, correlated morphologies would result from the combined effects of antagonistic coevolution mediated by genitalia and connexival spines, and these other processes. This interpretation is supported by the role female spine morphology plays in affecting male copulation duration.

It is worth noting that male and female body sizes were also correlated across populations (Fig. 3a), a pattern that can be expected under antagonistic coevolution and that has been documented across Gerris species (Arnqvist & Rowe, 2002b). Indeed, the relative sizes of mating individuals partly determine the outcome of premating struggles in some water strider species (Sih & Krupa, 1992; Arnqvist et al., 1996; Rowe & Arnqvist, 1996). However, correlation between male and female body sizes across populations could also result from similar responses of both sexes to a common selective regime, or more simply, to variable environmental factors. Also, there was no significant link between body size of male and/or female and reproductive behaviour and fitness in this study. Therefore, it is not possible to attribute the pattern observed in this study as resulting without doubt from antagonistic coevolution between the sexes (Arnqvist & Rowe, 2002b).

Correlated intersexual morphologies across populations also suggest that antagonistic coevolution has played a role in the morphological differentiation of G. gillettei natural populations. However, no asymmetry was observed in reproductive behaviour or fitness between sympatric and allopatric crosses (Tables 2 and 3,Appendix 2). Despite morphological differentiation of G/AG traits between experimental populations (Fig. 2c and d), divergence was apparently not sufficient to significantly influence the outcome of mating interactions. Note, however, that differences in behaviour between allopatric and sympatric mating types were nearly significant for one set of populations (Table 2a), suggesting that further experimentations with more trials and populations might provide the statistical power necessary to reveal such influence. In any case, our current results contrast with other studies that found asymmetry in matings between natural populations potentially experiencing sexual conflict. However, these involved populations from relatively remote sites (Knowles & Markow, 2001; Sugano & Akimoto, 2007) or discrete alternative male phenotypes (Hebets & Maddison, 2005). Here, both the close proximity of sites and the subtle morphological differences between populations probably contribute to truly limit reproductive isolation in this species.

Beyond correlated morphologies among populations, there was obviously a lot of variation within each population and for each sex (Fig. 2). The maintenance of ample within-population variation is surprising, given that antagonistic selection is constraining male and female morphologies and should deplete variation. This suggests that selection may not be very efficient. This may be attributed to low heritability of G/AG traits, as demonstrated for length and slope of connexival spines in females of G. incognitus, a close relative of G. gillettei (Arnqvist & Thornhill, 1998). Alternatively, but not exclusively, within-population variation could be maintained by temporal changes in the strength and/or direction of selection, a common feature in natural populations (Siepielski et al., 2009). For G. gillettei, local conditions are certainly spatially variable among sampled ponds, which include low-productivity rock pools and highly productive thaw ponds (Breton et al., 2009). Densities also certainly vary from year to year in the highly variable subarctic climates of northern Quebec. In water striders, the intensity of premating sexual conflict is supposed to vary as a function of ecological variables such as density, sex ratio, predation and food availability (Rowe et al., 1994; Arnqvist, 1997). This trend has already been documented in G. odontogaster, where male abdominal processes were shorter at high local density (Arnqvist, 1992b). Overall, high within-population phenotypic variation coupled with correlated male and female morphologies across populations suggest that the balance of armaments levels may be labile within each population, while nevertheless reflecting the intensity of sexual conflict. As such, the overall correlation between male and female G/AG traits despite high within-population variation is rather remarkable.

Experimental mating trials indicated that G/AG traits influenced mating behaviour. First, copulation duration was influenced by individual male morphology as well as by the relative morphologies of mating pair members (Fig. 4a and b). Males with a flattened distal part of the abdomen and thick genitalia had shorter copulations, and this behaviour was also influenced by the relative morphology of their female partner. More specifically, shorter copulations were associated with relatively ‘armed’ males paired with females with relatively short and bent spines. This result may indicate that males with a high mating potential try to maximize encounters by spending a minimum amount of time with each female, resulting in shorter copulations. By contrast, unarmed males, when achieving copulations with females having a better potential to resist, may capitalize on fewer long copulations to insure paternity. Although it has been demonstrated that shorter copulations can indeed be associated with a significant mating advantage in another water strider species (Danielsson, 2001), these tentative explanations remain speculative and must await further investigations. Second, female morphology impacted on guarding duration (Fig. 4c), and females with long genitalia and narrow femurs were guarded for shorter periods of time than other females. This result is surprising, as we would have expected females with wide femurs to have shorter guarding times. Indeed, in water striders, the guarding phase is terminated by post-mating struggles much similar to the premating ones (Rowe, 1992). During the latter, females use their forelegs to dislodge males (Rowe et al., 1994; Arnqvist, 1997). Thus, our observations are difficult to explain in the light of the current knowledge about the role of female femur morphology in pre- and post-mating struggles. One possible explanation could be that cryptic female choice, mediated inside female genitalia, also plays a role in the decision to terminate guarding. For example, females could end guarding more rapidly after mating with a ‘non-preferred’ male, and thus be immediately ready for a new mating. This explanation is speculative, but the pattern of correlated intersexual morphologies indicates that female genitalia characteristics are important in antagonistic interaction with males in natural populations. Investigations focussing on this trait are warranted.

The advantage of our no-choice mating trials was that reproductive behaviour could easily be observed and directly linked to fertilization and hatching success. This experimental approach has successfully been used in many studies investigating sexual selection in general and sexual conflicts in particular (Andrés & Arnqvist, 2001; Knowles & Markow, 2001; Martin & Hosken, 2003; Hebets & Maddison, 2005; Wigby & Chapman, 2006; Bacigalupe et al., 2007; Sugano & Akimoto, 2007). However, by performing a single mating trial, the influence of mating rate per se could not be addressed. Mating rate is part of the basic rationale of sexual conflict in water striders, whereby male fitness should increase with mating rate whereas female are expected to benefit from intermediate mating rate insuring sperm availability while limiting harassment costs (Arnqvist & Nilsson, 2000). Moreover, to assess fitness, single mating trials deliberately involved sperm-depleted females that may have been less reluctant to mate. If mating rate is important, initial mating events may not be those where females express the strongest resistance. Sperm-depleted females have been used previously in water strider mating experiments (Arnqvist & Thornhill, 1998; Arnqvist & Danielsson, 1999; Danielsson & Askenmo, 1999; Danielsson, 2001; Ronkainen et al., 2010), so this bias is unlikely to be specific to our study. Nevertheless, behavioural (and possibly fitness) differences may only be expressed and detected when multiple matings occur. Our results linking morphology and copulation duration suggest that indeed mating rate is important in G. gillettei. Thus, assessment of fitness in a context allowing multiple matings may be more amenable to reveal links between morphology and fitness mediated by behaviour.


We present evidence for intraspecific, intersexual correlated morphologies for traits related to sexual conflict in water striders. This pattern is likely the footprint of contemporary sexually antagonistic selection resulting in different balances of armaments between the sexes across natural populations of G. gillettei. Our analyses reveal unexpected associations between some G/AG traits and suggest that the relative importance of traits in mediating male–female conflicts may be different than previously established. In particular, female genitalia may be more effective than connexival spines in mediating intersexual interactions affecting fitness. Male and mating pair morphologies, including connexival spine variation, influence reproductive behaviour, but not fitness, suggesting that opportunities for multiple matings may be necessary to reveal the real fitness impact of persistence and resistance traits in this, and possibly other, water strider species.


We thank S. Bélanger, A. Bourret, M.-P. Emond and S. Tremblay-Bourgeois for field and laboratory work; C. Cloutier, J.-F. Guay and S. Boudreault for photography set-up and technical advices; J.-F. Simard for statistical advices; G. Colbeck and P. Duchesne for helpful comments. Comments by W.U. Blanckenhorn were also very useful in improving the final manuscript. This work was supported by a NSERC scholarship to M.-C. Gagnon and NSERC research grants to J. Turgeon.


Appendix 1 Correlations between three fitness variables (fecundity (F), fertilization success (FS) and hatching success (HS)), individual female morphology (PC1F to PC3F), individual male morphology (PC1M to PC3M) and the difference between male and female PC scores (exp: PC1M–PC1F). Refer to Table 1 for a description of PC loadings.

PC1FrS: 0.088rS: 0.125rS: 0.101
P-value: 0.341P-value: 0.179P-value: 0.277
PC2FrS: 0.009rS: 0.046rS: −0.041
P-value: 0.925P-value: 0.623P-value: 0.660
PC3FrS: 0.084rS: 0.159rS: 0.118
P-value: 0.365P-value: 0.086P-value: 0.204
PC1MrS: −0.028rS: −0.118rS: −0.070
P-value: 0.765P-value: 0.205P-value: 0.456
PC2MrS: 0.155rS: 0.151rS: 0.108
P-value: 0.094P-value: 0.104P-value: 0.246
PC3MrS: −0.017rS: −0.030rS: −0.034
P-value: 0.854P-value: 0.752P-value: 0.714
PC1M– PC1FrS: −0.075rS: −0.155rS: −0.125
P-value: 0.424P-value: 0.099P-value: 0.183
PC1M– PC2FrS: 0.001rS: −0.074rS: −0.011
P-value: 0.999P-value: 0.435P-value: 0.905
PC1M– PC3FrS: −0.072rS: −0.184rS: −0.098
P-value: 0.442P-value: 0.050P-value: 0.298
PC2M– PC1FrS: 0.020rS: 0.002rS: −0.029
P-value: 0.831P-value: 0.981P-value: 0.755
PC2M– PC2FrS: 0.155rS: 0.160rS: 0.107
P-value: 0.099P-value: 0.088P-value: 0.255
PC2M– PC3FrS: 0.096rS: 0.039rS: 0.046
P-value: 0.306P-value: 0.679P-value: 0.625
PC3M– PC1FrS: −0.062rS: − 0.085rS: −0.073
P-value: 0.510P-value: 0.367P-value: 0.438
PC3M– PC2FrS: 0.007rS: −0.050rS: 0.013
P-value: 0.937P-value: 0.597P-value: 0.892
PC3M– PC3FrS: −0.068rS: −0.151rS: −0.115
P-value: 0.468P-value: 0.108P-value: 0.222

Appendix 2 Means and standard deviations of reproductive behaviour and fitness variables measured in no-choice experiments. Sets and populations (Set Pop), type of mating (Type) number of male attempts (NMA), number of female rejects (NFR), number of somersaults (NS), time before copulation (TBC), copulation duration (CD), guarding duration (GD), fecundity (F, number of eggs), fertilization success (FS) and hatching success (HS).

1AAllo1.2 (0.8)0.2 (0.8)0.1 (0.4)407 (761)274 (181)129 (127)24.3 (16.0)0.27 (0.22)0.43 (0.42)
Sym2.5 (2.9)1.5 (2.9)2.5 (3.4)882 (1004)212 (188)161 (273)29.5 (18.5)0.31 (0.23)0.51 (0.35)
Overall1.9 (2.2)0.9 (2.2)1.3 (2.7)644 (908)243 (184)145 (210)26.9 (17.2)0.29 (0.22)0.47 (0.38)
1BAllo1.1 (0.4)0.1 (0.4)0.6 (1.7)1255 (1266)356 (315)193 (200)23.5 (14.0)0.28 (0.21)0.63 (0.42)
Sym1.4 (0.9)0.1 (0.5)0.4 (0.9)705 (757)243 (209)177 (199)20.9 (16.7)0.21 (0.17)0.50 (0.44)
Overall1.3 (0.7)0.1 (0.4)0.5 (1.3)980 (1062)299 (269)185 (196)22.2 (15.2)0.25 (0.19)0.56 (0.43)
2AAllo3.7 (10.1)2.7 (10.1)1.7 (4.3)321 (461)311 (275)76 (85)34.0 (18.3)0.40 (0.16)0.63 (0.29)
Sym1.1 (0.3)0.1 (0.3)0.8 (1.5)387 (644)582 (438)108 (134)25.9 (22.8)0.29 (0.26)0.53 (0.37)
Overall2.4 (7.1)1.4 (7.1)1.2 (3.2)354 (551)447 (385)92 (111)29.9 (20.7)0.35 (0.22)0.58 (0.33)
2BAllo1.1 (0.3)0.1 (0.3)0.6 (1.2)644 (1133)391 (345)421 (458)11.1 (13.6)0.09 (0.18)0.32 (0.47)
Sym1.3 (0.8)0.3 (0.8)1 (1.6)1064 (1206)485 (278)127 (209)12.3 (9.6)0.15 (0.22)0.40 (0.46)
Overall1.1 (0.6)0.2 (0.6)0.8 (1.4)854 (1170)438 (311)274 (380)11.7 (11.6)0.12 (0.20)0.36 (0.46)