|Home | About | Journals | Submit | Contact Us | Français|
Trivers and Willard proposed that offspring sex ratio should vary with maternal condition when condition, meant as maternal capacity to care, has different fitness consequences for sons and daughters. In polygynous and dimorphic species, mothers in good condition should preferentially produce sons, whereas mothers in poor condition should produce more daughters. Despite its logical appeal, support for this hypothesis has been inconsistent. Sex-ratio variation may be influenced by additional factors, such as environmental conditions and previous reproduction, which are often ignored in empirical studies. We analysed 39 years of data on bighorn sheep (Ovis canadensis) that fit all the assumptions of the Trivers–Willard hypothesis. Production of sons increased with maternal condition only for mothers that weaned a son the previous year. This relationship likely reflects a mother's ability to bear the higher reproductive costs of sons. The interaction between maternal condition and previous weaning success on the probability of producing a son was independent of the positive effect of paternal reproductive success. Maternal and paternal effects accounted for similar proportions of the variance in offspring sex. Maternal reproductive history should be considered in addition to current condition in studies of sex allocation.
Sex-allocation theory attempts to explain variation in offspring sex ratio at different levels, from individuals and populations to species. Most research on individual sex allocation in vertebrates is inspired by the Trivers–Willard hypothesis (TWH), which proposes that the sex of offspring should vary with maternal condition when it influences the fitness of sons and daughters differently . Trivers–Willard effects are predicted when (i) maternal condition correlates with offspring condition at the end of maternal care, (ii) differences in condition between offspring at the end of care persist to adulthood and (iii) differences in adult condition have a greater impact on reproductive success of one sex. For polygynous and dimorphic species, the theory predicts that females in good condition should preferentially produce sons, whereas females in poor condition should produce more daughters . However, empirical results are mixed [2–6]. Many species that fit all the assumptions of the TWH do not fit its prediction .
Heterogeneous measures of condition may explain inconsistencies in tests of the TWH. Studies measuring maternal condition near conception rather than after birth show stronger support for the hypothesis [3,4]. Furthermore, relative changes in maternal condition seem to better predict offspring sex than absolute values . It is important to note that Trivers and Willard used ‘condition’ to refer to maternal capacity to care. Environmental conditions and previous reproduction may thus confound the relationship between behavioural or morphological measures of condition and sex ratio by influencing maternal ability to care. In red deer (Cervus elaphus), a positive relationship between maternal dominance and production of sons disappeared at high density, but the adaptive significance of this change remains unclear . To our knowledge, no study has explored whether Trivers–Willard effects depend on reproductive history.
We tested whether environmental conditions near conception and previous weaning success (PWS) affected the relationship between relative condition, measured as yearly changes in mass and offspring sex in female bighorn sheep (Ovis canadensis). This species satisfies all the assumptions of the TWH but has repeatedly been found not to support its predictions [9–11].
Data were collected from 1975 to 2014 at Ram Mountain, about 30 km east of the Rockies in Alberta, Canada. From late May to late September, sheep were captured in a trap baited with salt. Females can produce one lamb per year from 2 years of age onwards. Lambs were sexed at first capture, aged from one week to three months. Maternity was accurately determined from field observations of suckling behaviour.
We analysed offspring sex of females aged ≥5 years (see electronic supplementary material, table S1 for female age distribution) to obtain a measure of condition relatively independent of female age. We measured maternal condition as a female's mass adjusted to 15 September each year, about two months before conception, minus her average mid-September mass from age 4 onwards. Offspring sex and maternal condition were measured at age x and age x − 1, respectively. By age 4, females reach about 95% of asymptotic mass . Females can gain mass until 6 years and senescence-related mass loss begins at 11–12 years . To ensure that our analyses were not biased because of late growth and senescence, we re-ran them considering only females aged 7–13 years. Almost all females (95%) were weighed two to seven times each summer from age 4. Mass adjustments to mid-September are described elsewhere . Our measure of condition provides a direct test of the TWH because relative rather than absolute values of mass correlate with reproductive potential .
Because sons are costlier than daughters , we distinguished three levels of PWS: no lamb weaned, weaned a female or weaned a male. We used yearly average mass of yearling females in mid-September as an integrator of environmental conditions before conception . Electronic supplementary material, figure S1 summarizes when each variable was measured.
We ran generalized linear mixed models (GLMMs) with a binomial distribution using the ‘lme4’ library  in R v. 3.1.2 (http://www.r-project.org) to test whether the probability to produce a son was influenced by maternal condition, PWS, environmental conditions and their interactions. Mother identity and year of lamb birth were fitted as random effects.
Both maternal age and paternal reproductive success (the log-transformed percentage of lambs sired by the father that year) influence offspring sex in this population [10,18]. These two variables were included in the final GLMM to check that the results obtained in the above analyses did not change. Paternity is available for most lambs only since 1988. Father identity was entered as an additional random effect in models including paternal effects.
We tested successively interaction terms and, if these were not significant, the main effects of variables using likelihood-ratio tests based on bootstrapping of 400 replicates (R script available in electronic supplementary material, S3). Model simplification proceeded until only terms significant at the p < 0.05 level remained in the model. We reported standardized regression coefficients for the final model by first centring and then dividing all continuous variables by 2 s.d., to compare effect sizes . We also calculated total variance explained by random and fixed effects . Generalized variance-inflation factors (GVIFs) were calculated for each predictor and full model with the function ‘vif’ of package ‘car’. Collinearity was not a problem because none of the GVIFs exceeded 4.5 .
Sex ratio did not differ from 0.5 (267 sons, 293 daughters; p = 0.29). A significant interaction between maternal condition and PWS influenced the probability of producing a son, independently of environmental conditions (table 1). After weaning a son, sex ratio varied from a strong daughter bias for mothers in poor condition to a bias towards sons for mothers in good condition (figure 1a). Production of sons appeared to decrease with condition when mothers failed to wean a lamb the previous year, but this pattern was not statistically significant (figure 1c). Neither environmental conditions nor their interaction with maternal condition influenced the probability of producing a son (table 1). Restricting the analysis to females aged 7–13 years yielded similar results (electronic supplementary material, table S2).
Re-running the final model (table 1) with paternal reproductive success or maternal age as additional explanatory variable did not alter the conclusions (electronic supplementary material, tables S3 and S4). The probability of producing a son was positively influenced by paternal success (slope = 0.886 ± 0.281, p = 0.002), but not by maternal age (slope = 0.057 ± 0.190, p = 0.77). The model with paternal success and the interaction between maternal condition and PWS (electronic supplementary material, table S4) explained 11.9% of the variation in offspring sex, of which 9.9% was explained by the fixed effects. Maternal and paternal fixed effects taken separately accounted for 4.6% and 4.4%, respectively, of the model variance.
Tests of the TWH in mammals have produced inconsistent results [2,4]. Understanding the causes of such variation is a challenge for sex-allocation theory. Schindler and colleagues recently suggested that sexual differences in lifetime reproductive value explain the mismatch between TWH predictions and observations . After accounting for sex differences in demography, however, their model still predicted a Trivers–Willard effect in bighorn sheep. We found that production of sons increased with maternal condition only for mothers that weaned a son the previous year.
A correlation between offspring sex and maternal condition may arise simply as a consequence of higher male mortality during early life . We did not know sex ratio at birth, but neonatal mortality was relatively low (17%) compared with subsequent juvenile mortality (45%) with no evidence that it was sex-specific . It is also unlikely that our results were due to sex-specific neonatal mortality since the relationship between sex ratio and maternal condition after weaning a son was symmetrical around 0.5 [3,7]. Furthermore, the hypothesis invoking male-biased early mortality predicts excess daughters following years of poor environmental conditions , which was not the case.
Our findings may instead reflect adaptive manipulation of offspring sex. In polygynous dimorphic ungulates, higher energetic demands of sons over daughters can lead to higher costs in terms of future reproduction and survival for mothers that raised sons . A previous study reported that bighorn females avoided producing sons in consecutive years . Our analyses reveal that this is true only when females are in poor condition, likely because the costs of reproduction are highest. Females in good condition after weaning a son likely acquired a large amount of resources, so it would be advantageous for them to bias care towards sons. A combination of these two effects can explain why only mothers that weaned a son the previous year behave as predicted by Trivers and Willard.
Generally, sex-ratio studies of vertebrates yield small effect sizes . A recent meta-analysis reported that paternal attractiveness explained less than 1% of variation in offspring sex ratio . About 10% of variation in offspring sex of bighorn sheep was explained by an equal contribution of maternal and paternal effects. Our results emphasize the importance of considering maternal reproductive history in addition to condition to predict sex ratio, because the product of these two variables better reflects a mother's capacity to care.
We thank A. Hubbs, C. Feder and J. Jorgenson for their support of the Ram Mountain research programme; all assistants and students who worked on Ram Mountain over decades; D. Coltman for providing data on paternity; G. Pigeon for providing adjusted body mass and three anonymous reviewers for constructive comments on the manuscript.
Animal-handling procedures were approved by the Animal Care Committee of the University of Sherbrooke (MFB2009-01 and FP2012-01), affiliated to the Canadian Council for Animal Care.
M.D. developed the idea and carried out statistical analyses. All authors participated in the study design, wrote the manuscript, agree to be held accountable for the content therein and approved the final version of the manuscript.
We have no competing interests.
This work was funded by the Natural Sciences and Engineering Research Council of Canada (2014-05572, 355492), including grants to F.P. and M.F.-B. and a CREATE Environord postdoctoral fellowship to M.D. and the Alberta Conservation Association. F.P. holds the Canada Research Chair in Evolutionary Demography and Conservation (229221).