GOOD GENES DRIVE FEMALE CHOICE FOR MATING PARTNERS IN THE LEK-BREEDING EUROPEAN TREEFROG

Authors


Abstract

Investigating the mechanisms underlying female mate choice is important for sexual-selection theory, but also for population-genetic studies, because distinctive breeding strategies affect differently the dynamics of gene diversity within populations. Using field-monitoring, genetic-assignment, and laboratory-rearing methods, we investigated chorus attendance, mating success and offspring fitness in a population of lek-breeding tree-frogs (Hyla arborea) to test whether female choice is driven by good genes or complementary genes. Chorus attendance explained ∼50% of the variance in male mating success, but did not correlate with offspring fitness. By contrast, offspring body mass and growth rate correlated with male attractiveness, measured as the number of matings obtained per night of calling. Genetic similarity between mating partners did not depart from random, and did not affect offspring fitness. We conclude that females are able to choose good partners under natural settings and obtain benefits from the good genes, rather than compatible genes, their offspring inherit. This heritability of fitness is likely to reduce effective population sizes below values previously estimated.

Sexual reproduction imposes asymmetric selective pressures on mating partners. A male's fitness is often limited by his access to females, whereas a female's fitness is mostly limited by her own reproductive investment. Females are thus under selection to choose high-quality partners, which in turn imposes a sexual-selection pressure on males to advertise their quality through honest (i.e., costly) signals (Grafen 1990; Andersson 1994). Benefits for choosy females may be direct, if for example high-quality partners provide more help in rearing the young (Møller and Jennions 2001), or indirect, if offspring inherit either good genes (Andersson 1994; Møller and Alatalo 1999) or compatible genes (Mays and Hill 2004; Neff and Pitcher 2005). Good-genes models predict attractiveness to vary strongly among males, and to correlate with offspring fitness. Compatible-genes models, by contrast, predict offspring fitness to correlate with genetic dissimilarity between mating partners, not with male attractiveness (Mays and Hill 2004).

Most empirical evidence supporting female mate choice for good genes and male quality-display has been gathered from birds (Petrie 1994; Hasselquist et al. 1996; Sheldon et al. 1997). These include lekking species, in which males congregate on display grounds at breeding times, allowing females to compare potential partners before soliciting mating (Höglund and Alatalo 1995). Being nonresource based, these mating systems are expected to provide evidence for good- or compatible-genes effects, because males contribute only genes to their offspring. Empirical evidence is much scarcer for other groups (Welch et al. 1998; Sandvik et al. 2000; Barber et al. 2001; Wedekind et al. 2001; Sheldon et al. 2003; Forsman and Hagman 2006), even though lek-breeding systems are not limited to birds. Furthermore, such evidence has most often been gathered from laboratory experiments rather than field studies (for exceptions see Woodward et al. 1988; Byers and Waits 2006).

Anurans (frogs and toads) are a priori well-suited model organisms for investigating mechanisms of female choice. Males often congregate in leks to display through costly mating calls (Grafe and Thein 2001), and mating systems may be resource-based or not, offering opportunities to isolate genetic effects. Evidence exists from resource-based systems that male calls are honest indicators of paternal quality (Forsman and Hagman 2006). In nonresource-based systems, among-sire variance in offspring fitness has also been repeatedly documented (Travis 1980, 1981; Woodward 1986, 1987; Berven 1987; Travis et al. 1987; Woodward et al. 1988; Mitchell 1990; Sheldon et al. 2003), and male mating success was shown to vary with several male traits, including calling behavior (Lopez and Narins 1991; Passmore et al. 1992; Cherry 1993; Grafe 1997) and body size (Woodward 1986; Mitchell 1990; Howard et al. 1994; Semlitsch 1994). However, correlations of male traits with offspring fitness produced contrasting results, ranging from positive (Woodward 1986) to negative (Semlitsch 1994), absent (Howard et al. 1994), or condition-dependent (Mitchell 1990).

The best evidence for good genes driving mate choice in amphibians comes from lekking hylids. In Hyla crucifer, the offspring of males that obtained a mating in the wild displayed higher growth rates in the laboratory than those of unsuccessful males (Woodward et al. 1988). However, the question remains whether mating success occurred through female choice or intrasexual (male–male) competition. In the gray treefrog Hyla versicolor, long-calling males are preferred by females under laboratory conditions (Klump and Gerhardt 1987; Gerhardt et al. 1996), and also produce offspring with enhanced growth and survival (Welch et al. 1998; Doty and Welch 2001). However, female preference seems less clear under more realistic conditions (Schwartz et al. 2001), and, in natural choruses, male mating success appears unaffected by call rate or call duration (Sullivan and Hinshaw 1992). Further doubts on the potential for a “good genes” choice by female amphibians have been raised by studies of the European treefrog, Hyla arborea, a species in which amplexi are formed strictly on the female initiative (Friedl and Klump 2005). On the one hand, recent laboratory experiments by Gomez et al. (2009) pointed to a role for visual cues (vocal bag coloration and flank stripe contrast) in H. arborea female choice, and male mating success was shown to be affected by calling parameters in the closely related Italian treefrog H. intermedia (Castellano et al. 2009). On the other hand, field studies by Friedl and Klump (2002, 2005) failed to find evidence of any consistent female choice. Although males differed in several call parameters (Friedl and Klump 2002), calling patterns did not seem to be under any directional selection (Friedl 2006), male mating success being only affected by chorus attendance (number of nights spent calling, Friedl and Klump 2005). The mechanisms underlying female choice in H. arborea and related species are therefore still a matter of debate.

The mechanisms underlying female choice in amphibians matter not only in the context of sexual-selection theory, but also regarding the dynamics of gene diversity within populations. Sexual selection in lek-breeding species is expected to drastically reduce male effective population size by increasing the among-males variance in mating success (Emlen and Oring 1977; Wagner and Sullivan 1992; Murphy 1994; Jehle et al. 2001). Lowering effective population size should enhance extinction risk by increasing the rate of accumulation of deleterious mutations (Lande 1995), by impeding the fixation of beneficial alleles (Whitlock and Bürger 2004), or by limiting the evolutionary potential of populations (Frankham 2005). The mechanisms underlying female choice might either exacerbate or dampen these negative effects. If benefits are only direct (stemming e.g., from paternal condition), then female choice should only marginally affect male effective population size. If females choose mates according to compatible genes (genetic-dissimilarity models), then mate choice should help maintaining genetic diversity within populations (e.g., Perez-Gonzalez et al. 2009). Alternatively, if females choose males for their good genes, negative effects are twofold. First, convergence of female preferences should increase the among-male variance in reproductive success, and second, the heritable component of male fitness should further depress effective population size in the long term (Nei and Murata 1966; Kelly 2001; Nomura 2002). Broquet et al. (2009) investigated the impact of mate choice and sexual selection on genetic drift in a population of lek-breeding European treefrog, H. arborea. From their findings, lek breeding seemed to have an unexpectedly low impact on male effective population size, but calculations were made assuming no heritability of fitness, which might be an important caveat.

Here, we build on the field data gathered by Broquet et al. (2009), combining their results on male reproductive success with additional information on offspring fitness, to test whether attractive males also provide fitness genes (i.e., genes conferring improved survival or growth) to their offspring. From our new analyses, offspring growth rate and body mass correlate with male attractiveness (measured as the number of matings per night of calling activity). These results suggest, not only that females are able to choose partners under natural settings and benefit from this choice through the good genes their offspring inherit, but also that male effective population size might be further depressed by the heritability of fitness traits.

Materials and Methods

Detailed information on sampling design, genotyping, and parentage analyses are provided in Broquet et al. (2009), and thus will be only briefly presented in this article.

CHORUS ATTENDANCE

The study site (four neighboring ponds at Les Mossières, western Switzerland) was visited on 22 nights throughout the 2006 breeding season. At each visit, all calling males were counted, localized, captured, and individually identified via a photographic data bank recording the patterns of their black lateral line, which allows unambiguous individual recognition (Pellet et al. 2007; Broquet et al. 2009). All males captured during the breeding season and all females encountered by chance were sampled for buccal cells (sterile cotton swab, Pidancier et al. 2003; Broquet et al. 2007) for subsequent parentage genetic analyses.

OFFSPRING FITNESS

Ponds were further visited by daylight every four days and searched for clutches of eggs (Broquet et al. 2009). All clutches found were collected and development stage at collection was recorded according to three classes (corresponding to Gosner (1960) stages 1–14, 15–18, and 19–20, respectively). All clutches were then reared in the laboratory until hatching (200 mL tanks). Individual hatching success was recorded as a binary variable (0 for a dead larva and 1 for a living one). Newborns were released in their pond of origin, except for 10 individuals per clutch, kept in stable laboratory conditions (22°C, 13 h light per day) for fitness assessment and parental assignment.

Fitness was assessed for a subset of clutches, limited to 100 due to space and time constraints. All 10 individuals from these clutches were maintained in batches, in 300 mL tanks for the first nine days, then in 750 mL tanks, and fed daily with TetraMin fish flakes ad libitum. Survival was recorded daily. Water was changed at days 9, 13, 17, 20, 24, 28 plus every time one individual was found dead. All tadpoles were weighed individually (Mettler PC220, at ± 0.001 g) at days 13 and 20 after hatching. Growth rate was estimated for each clutch as the difference between average tadpole weight at days 13 and 20. At day 28, individual survival was recorded (binary variable) and tadpoles released in their pond of origin.

When series of clutches were laid on the same day and in the same pond, some of them were not assessed for fitness (due to time and space constraints), being likely laid by the same pair. Tadpoles from these clutches were nevertheless reared to a size large enough to be genotyped without damage, before being released in their pond of origin. These tadpoles were maintained in two containers of 300 mL each (five tadpoles per container) and fed ad libitum, with water changed every three days. On day 28, a tiny portion of the velum (ca. 2 mm2) was removed from the extremity of the tail of all surviving offspring (assessed or not for fitness), and deep-frozen (−80°C) for subsequent genotyping.

PARENTAL ASSIGNMENTS

All adults and about four tadpoles per clutch were genotyped for one mitochondrial sequence and seven to nine nuclear microsatellite markers to carry out parentage analyses (Broquet et al. 2009). Genetic relatedness among partners of all possible mating pairs, realized or not, were calculated with KINSHIP 1.3 (Goodnight and Queller 1999).

MALE TRAITS

The mating success (mi) of male i was defined as the total number of females he mated with over the whole breeding season (as inferred from parental assignments). His attendance (ni) was defined as the total number of nights he was recorded as calling, and his attractiveness inline image as the number of matings per night of calling activity.

STATISTICAL ANALYSES

The effect of male trait x (respectively mating success, chorus attendance, estimated genetic relatedness to female, and attractiveness) on tadpole fitness trait y (respectively hatching rate, mass at days 13 and 20, growth rate and survival at day 28) was assessed in R (R Development Core Team 2007) through linear mixed effect models (R package Lme4, Bates and Sarkar 2007):

image

where Father, mother, and clutch were set as random factors, accounting for the structure of the data (the tadpoles sired by the same male are not independent, as well as tadpoles sharing the same mother or reared in the same container). Clutch effect was not included in the model for growth rate, because our data provided one average value per clutch. Egg stage, attractiveness, chorus attendance, mating success, and mating-pair-estimated relatedness were set as fixed effects. We used mixed effects models fitted by restricted maximum likelihood estimation (REML) for continuous traits (tadpole mass and growth rate) and generalized mixed effects models fitted using the Laplacian approximation with a binomial error family for binary traits (hatching and survival). The effect of each variable (namely male trait x, egg-stage, clutch, mother and father) was tested by likelihood-ratio tests based on the drop in Akaike's information criterion (AIC; a measure of the goodness of fit of an estimated statistical model) following the inclusion of the focal variable into the partial model comprising all other variables. The amount of variance in offspring fitness explained by male traits was calculated as the coefficient of determination (R2) of linear regressions of male breeding values (calculated for each offspring-fitness trait y from the residuals of the partial model y ∼ Egg stage + Clutch+ Mother + Father) on male traits. These breeding values represent the expected fitness value of a male's progeny, once other independent effects are accounted for (Lynch and Walsh 1997).

Heritability for quantitative traits with individual records (i.e., weight at 13 and 20 days) was estimated with two different methods. First, heritability was estimated as h2= 4VSIRE/VTOT (Falconner and McKay 1996), with variance components estimated from the mixed effect model: y ∼ Egg stage + Clutch + Mother + Father (where y is weight at 13 or 20 days). This method applies to half-sib crossing designs with balanced sample size. As our dataset slightly differed from these requirements (the number of partners per male ranged from one to four, and two females mated with more than one male), we also estimated heritability using the Bayesian animal model described in Waldmann (2009). Egg stage was set as fixed effect and the upper bounds for the priors of the additive and error standard deviations (sd.u.add and sd.u.err, Waldmann 2009) were set to 200 each. Three independent Markov chain Monte Carlo (MCMC) runs were performed in Winbugs 1.4.3 (Lunn et al. 2000), each with a burn-in period of 100,000 iterations. Heritability was recorded for a further 10,000 iterations (with a thin of 100, resulting in a total chain length of 1,000,000 after burn-in). The Gelman and Rubin convergence diagnostic (Gelman and Rubin 1992) and trace plots of the variables were used to check for obvious chain convergence problems.

Results

CHORUS ATTENDANCE AND MATING SUCCESS

A total of 15 males were monitored as calling over the breeding season, all of which were captured and genotyped, together with eight females. Chorus size varied from 1 to 10 simultaneously calling males (average: 5.2, SD: 3.4). Chorus attendance varied strongly among males, from 1 to 11 nights (average: 5.5, SD: 3.8, Fig. 1, see also Fig. S1 in Broquet et al. 2009). One male exhibited an unusual behavior: it started calling 10 days after the last other males stopped calling, and 18 days after the last clutch was laid (thus it did not reproduce). Assigning this male a chorus attendance of either four (observed number of calling nights) or zero (“effective” chorus attendance) had no qualitative impact on the regression of mating success on chorus attendance.

Figure 1.

Male mating success increases with chorus attendance (number of nights monitored as calling), with 49% of the variance explained by a linear regression (n= 15, P < 0.01). One male was assigned a chorus attendance of zero because he started calling long after the end of the reproductive period.

Of 138 clutches found, 131 were genotyped, and could be assigned without ambiguity to one of the 15 calling males. Five males of the 15 identified failed to sire any clutch (m= 0). By contrast, only 55 clutches (42%) were assigned to one of the eight females captured, the remaining being assigned to 11 unsampled females (see Broquet et al. 2009 for details). Altogether, 22 mating pairs involving 19 females and ten males formed over the eight nights during which clutches were laid. Estimated genetic relatedness between mating partners did not differ from random (t-test between realized [mean relatedness ± SD: −0.03 ± 0.26] and potential mating pairs [−0.02 ± 0.24], P= 0.84). Operational sex ratio (defined as the number of calling males per mating female) averaged 3.3 over these eight nights (range: 1.6–9).

The distribution of male mating success m (average: 1.47 partners, SD: 1.31, range: 0–4; Fig. 1) did not depart from Poisson (Kolmogorov–Smirnov test, P= 0.17). A large part of the variance (49%, P= 0.004) was explained by chorus attendance n (Fig. 1), the residual variance stemming from among-male variance in mating success per night of attendance.

ATTRACTIVENESS AND OFFSPRING FITNESS

Offspring fitness could be assessed for nine of the 10 successful males (one male sired a single clutch, which turned out to be one of the 38 clutches not assessed for fitness). Attractiveness (a) explained a large and significant part of the variance in male breeding values for offspring growth rate and body mass at day 20 (54% and 52% respectively; Table 1, Fig. 2). Effects were positive in all cases (Table 1). After controlling for attractiveness, fathers did not differ significantly in offspring traits, whereas mothers had a highly significant effect on tadpole survival (Table 2). Egg stage affected hatching rate and larval survival, and a clutch effect was significant in all offspring traits measured (Table 2).

Table 1.  Variance in offspring fitness explained by male traits. Given is the proportion of variance (R2) in male breeding values for offspring-fitness traits (hatching rate, survival, weight at 13 and 20 days, and growth rate) explained by male traits (chorus attendance, mating success, and attractiveness). Also given are the sign of the relationship (brackets) and the significance level estimated with linear mixed effect models. Significant values (P<0.01) in bold. NS, not significant; **P < 0.01.
 Chorus attendance, nMating success, mAttractiveness, a
Hatching rate<0.01 (−) ns0.04 (+) ns0.16 (−) ns
Tadpole survival0.08 (−) ns0.03 (−) ns<0.01 (+) ns
Weight at 13 days0.04 (−) ns0.08 (+) ns0.21 (+) ns
Weight at 20 days0.11 (−) ns0.24 (+) ns0.52 (+) **
Growth rate0.12 (−) ns0.23 (+) ns0.54 (+) **
Figure 2.

Tadpole growth rate increases with male attractiveness a (number of matings obtained per night of chorus attendance), with 54% of the variance explained by a linear regression (n= 9, P < 0.01).

Table 2.  Factors influencing offspring fitness traits. Given are significance levels of variables in the linear mixed effect model for attractiveness. Significant values (P<0.05) are in bold. Clutch effect on growth rate could not be assessed, because our data provided only an average value per clutch.
 Egg stage1Attractiveness1Father2Mother2 Clutch2
  1. 1Fixed effect.

  2. 2Random effect.

  3. 3Binomial distribution.

Hatching rate30.00020.3111<0.0001
Tadpole survival30.0360.8810.00030.007
Body mass at 13 days0.0830.130.220.66<0.0001
Body mass at 20 days0.0740.010.210.67<0.0001
Growth rate0.200.0080.330.80

By contrast, offspring traits were not significantly affected by estimated genetic relatedness between mating partners (P-values ranging 0.3 to 1.0). Similarly, breeding values for offspring traits were not significantly affected by male attendance (Table 1), the trend actually being slightly negative. As a result, the effect of mating success on offspring growth rate and body mass (respectively 23% and 24% of variance in male breeding values explained) was not significant (Table 1).

Heritability for weight at 13 and 20 days, approximated as h2= 4VSIRE/VTOT, reached 0.38 and 0.37, respectively. We obtained similar heritability values from the Bayesian animal model (weight at 13: h2= 0.31, 95% highest posterior density interval (HPDI) [0.18; 0.48]; weight at 20: h2= 0.33, 95% HPDI [0.18; 0.53].

Discussion

Investigating the mechanisms underlying female choice for mating partners is important, not only for sexual-selection theory (given the scarcity of field studies providing evidence for good-genes effects in the wild), but also in the broader context of population genetics, because the variance in male fitness and its heritability have long-lasting consequences on the dynamics of genes in natural populations.

In the treefrog population under study, half of the variance in male mating success was determined by chorus attendance. This large amount appears to be a general feature of many frogs and toads with prolonged breeding season (table 5 in Friedl and Klump 2005 for lek-breeding hylids; Castellano et al. 2009). This, however, still leaves half of the variance unexplained by chorus attendance, and to be assigned to among-male variance in mating success per night. Although some of this residual variance was certainly stochastic, our results show a large part to be linked to differences in male quality. Indeed, the offspring of the most attractive males displayed faster growth rates and body masses, generally considered a good indicator of future fitness in amphibians (Altwegg and Reyer 2003 and references therein). How tadpole fitness in the laboratory reflects fitness in the wild, however, remains to be investigated.

By contrast, estimated genetic relatedness between mating partners did not differ from random, and did not affect offspring fitness, hence providing no support for a female preference for compatible genes (Mays and Hill 2004). Even though proper testing of compatible-genes models would require more accurate estimates of relatedness (based on higher numbers of loci) and full-factorial crossing experiments (e.g., Wedekind et al. 2008), the most parsimonious interpretation to our results is that females choose males according to traits displaying good genes.

We are confident that all attractive males were included in our analysis, because all clutches collected and genotyped could be assigned to one of the 15 males identified. We may have missed some unattractive males that did not call during the sampled nights. However, these males would not have been included in the tadpole-fitness analyses, because they did not reproduce, and thus would not have affected our good-genes interpretation of female choice.

The possibility that male “attractiveness” actually resulted from male–male competition rather than female choice can be discarded here, because in H. arborea amplexi are formed strictly on the female initiative, and are not disturbed by unpaired males (Friedl and Klump 2005). An effect of territory quality can also be excluded. First, we collected clutches before hatching, so that larval growth in the laboratory was unlikely to be affected by oviposition site. Second, mating pairs leave the chorus immediately after forming the amplexus, so that oviposition takes place away from the calling site (Friedl and Klump 2005). From our study, the several clutches of a pair were spatially widespread over the laying pond and we found no evidence for a pond effect on offspring mass and growth (data not shown).

A role for differential female investment (Sheldon 2000; Harris and Uller 2009) cannot be ruled out. The best way to properly control for it would involve artificial fertilization with male sperm, using a crossed design (e.g., Welch et al. 1998), but sacrificing adult males is clearly incompatible with the threatened status of H. arborea in Switzerland. However, although differential female allocation has been documented in birds (Sheldon 2000), we consider it unlikely in the present instance. Female anurans finalize egg provisioning before attending choruses (Lofts 1974), and female treefrogs start egg laying shortly after the amplexus is formed (Friedl and Klump 2005, pers. obs.), leaving little time for energy reallocation. We did find a highly significant maternal effect on tadpole survival (presumably linked to differences in egg provisioning), but this was independent of male attractiveness (Table 2). Furthermore, it would make little evolutionary sense to spare energy when mated with a low-quality male, given that, from our results, females usually mate with only one male. Annual survival is 20% in adult females (Friedl and Klump 1997), so that fitness returns would have to be multiplied by five or more for a delay in investment to be selected for. The point should also be made that maternal effects would not negate a good-gene interpretation, because any correlation between female investment and male attractiveness must ultimately rely on selection for genetic differences in male quality.

In agreement with Friedl and Klump (2005), we found male mating success to be Poisson distributed, and to increase significantly with chorus attendance. Contrasting with these authors’ interpretation, however, we note that a Poisson distribution of male mating success (as documented in the present study and in most studies of lek-breeding hylids, Friedl and Klump 2005) is compatible with the existence of female choice. Random components certainly affect all processes involved in female choice, from the distribution of genetic factors underlying individual differences in male quality, to local availability of males (owing to the duration of mating- and egg-laying process, a male is usually unavailable for more than one female per night). It is also worth noting that half of the variance in mating success was linked to attendance, and independent of attractiveness.

Additionally, we found that attendance did not correlate with offspring fitness, opposing the suggestion (Friedl and Klump 2005) that random-mating females might obtain good matings anyway, provided attendance correlates with male quality. Our results suggest instead that females do choose good-quality partners on some phenotypic trait(s), but that low-quality males still have a chance to reproduce, provided they attend the chorus frequently enough (hence the effect of attendance on mating success). A suboptimal choice by females may result from difficulties in assessing males, but also from temporal unavailability of high-quality partners. Low-quality males are certainly in a better situation to obtain matings when the operational sex ratio (number of males per female) is low. In reed frogs (Hyperolius marmoratus) males who mate on one night have a mating advantage on the following night, probably because they saved energy and not because of better genes (Dyson et al. 1998). Such a mechanism might explain why mating success may increase with chorus attendance independently of genetic quality. However, to account for the higher fitness of offspring from attractive males, we have to rely on good-genes arguments.

Empirical support for good-gene models has already been gathered, mostly from birds (Petrie 1994; Hasselquist et al. 1996; Sheldon et al. 1997), fish (Sandvik et al. 2000; Barber et al. 2001; Wedekind et al. 2001), insects (Partridge 1980), and anurans (Welch et al. 1998; Sheldon et al. 2003; Forsman and Hagman 2006). From a meta-analysis by Møller and Alatalo (1999), male traits explained 1.7% of the variance in offspring viability on average, with a considerable heterogeneity in effect size among studies. The effect documented in the present study may seem particularly strong, but the point must be made that our sample size was small, which should generate a large sampling variance (Møller and Alatalo 1999).

Our data provide support for an effective female choice under natural conditions, for which only scarce evidence has been gathered so far from mammals (Byers and Waits 2006), birds (Petrie 1994; Hasselquist et al. 1996; Sheldon et al. 1997), or fish (Barber et al. 2001). This constitutes an important finding, given the doubts recently raised on the very existence of choice in lek-breeding treefrogs (Friedl and Klump 2005). It implies, not only that opportunity for choice exists (as supported by the high operational sex ratio documented here and in other lek-breeding frogs, Friedl and Klump 1997), but also that the constraints imposed by time limitation, predation risks, and background noise are insufficient to prevent effective choice by females. Which male traits attract females remain to be investigated. These traits are likely to be found among calling features (e.g., Castellano et al. 2009), which presumably reflect honestly male quality due to extremely high costs (Grafe and Thein 2001). Recent findings also suggest a potential role for visual cues (Gomez et al. 2009).

However, good genes might also bring bad news for the dynamics of genetic diversity within treefrog populations. The significant heritability for tadpoles mass (h2 > 0.30) should further reduce male effective population sizes. Properly quantifying this reduction (following Nei and Murata 1966; Nomura 2002) would require translating the heritability of weight into a heritability of fitness, which was not feasible here. Nevertheless, larval growth is a good predictor for future fitness in amphibians (Altwegg and Reyer 2003). Whether the benefits of good genes counterbalance indirect costs induced by lowered effective population size remains an open empirical question.


Associate Editor: P. Stockley

ACKNOWLEDGMENTS

E. Ratthey, G. Emaresi, and J. Guélat provided much-welcome help for the field work and the rearing of tadpoles. We thank G. Emaresi, L. Berset-Brändli, and N. Duvoisin for their help and advices in the genetic analyses. This study benefited from discussions with O. Blaser, J. Goudet, C. Grossen, G. Kerth, M. Stöck, and J. Yearsley. We also thank P. Stockley and T. Halliday for valuable comments on a previous draft. The Swiss National Science Foundation provided financial support (grants 3100A0–108100 to NP and PBLAA-122658 to JJ). This experiment was approved by the Veterinary Office of the Canton de Vaud (authorization 1798) and conforms to its regulatory standards.

Ancillary