Introduction

In mammals, sex allocation theory predicts a sex bias in maternal expenditure on offspring where maternal fitness differs with increased allocation of resources to one sex over the other (Trivers and Willard 1973; Frank 1990; Clutton-Brock 1991; Hewison and Gaillard 1999; Hardy 2002). Thus, maternal allocation of resources should be greater in the sex that most increases the mother’s fitness by having a higher future reproductive potential, implying that maternal expenditure affects the fitness of offspring of each sex differently. This expectation is an extension of the Trivers–Willard hypothesis (TWH; Trivers and Willard 1973), which proposes that if the cost of producing each sex differs, the offspring sex ratio will vary according to the mother’s ability to allocate resources. The potential advantages of sex ratio adjustment predicted by TWH assume that: (1) offspring condition, and therefore, quality, is correlated with the mother’s condition; (2) condition differences at the end of maternal care translate into condition differences at reproduction; and (3) there is variation in reproductive success between the sexes.

For polygynous species, males may have more variable reproductive success than females (Clutton-Brock et al. 1988) because high quality males gain disproportionately high numbers of matings relative to low quality males, whereas female reproductive output is constrained by the number of offspring that can be produced. In such species, inclusive maternal fitness—defined as the total reproductive output of offspring and grand-offspring (Hewison and Gaillard 1999)—is greater for females that produce males when they are in good condition. The increase in fitness occurs because: (1) the offspring of higher quality mothers have higher survival and (2) the reproductive success of sons will increase more than that of daughters where improved condition during early development translates to greater fitness throughout reproductive life. The improvement in offspring fitness through higher maternal care implies that a given additional unit of maternal allocation increases the fitness of sons more than daughters (Trivers and Willard 1973; Hewison and Gaillard 1999). Such differences in fitness benefits to mothers are often associated with sexual size dimorphism and consequent differences in physiology and/or growth rates. TWH effects are expected to be stronger for species with strong sexual selection on males since the inclusive fitness of the mother is expected to vary more as a result of sex-biased allocation under such circumstances (Hewison and Gaillard 1999; Sheldon and West 2004).

An alternative explanation for sex ratio variation, the extrinsic modification hypothesis (EMH), proposes that sex ratios vary because extrinsic environmental forces affect maternal condition in dimorphic vertebrates in variable environments (Post et al. 1999). The theory suggests sex ratios vary stochastically through sex-differential survival related to differences in growth rates or energetic demands (e.g., lower survival associated with higher growth of male offspring in dimorphic vertebrates) and/or sensitivity to environmental conditions that affect the mother. Therefore, EMH is independent of maternal tactics of sex-biased allocation of resources. The broader importance of EMH is that it identifies the direct potential for abiotic factors to induce variation in secondary sex ratios in vertebrates. Such factors are not directly considered by TWH.

Krackow (2002) argues that biased vertebrate sex ratios are not evidence for either maternal control or adaptive variation but are a by-product of physiological and developmental differences between the sexes associated with extrinsic factors. For example, environmental stochasticity affects nutritional conditions, physiological stress, steroid hormone levels, and mating dynamics that may in turn affect sex ratios (Krackow 2002; Cameron 2004; Love et al. 2005). Similarly, any extrinsic factors that affect maternal body condition, including population density (Kruuk et al. 1999) and climate (Post et al. 1999), may also affect sex ratios.

Marsupial reproductive tactics and sex ratio variation

In most mammals, maternal resource allocation to offspring is greater than paternal allocation because of the demands of pregnancy and lactation. The total maternal expenditure per offspring is similar in marsupials and placental mammals (Krockenberger 2006). Nonetheless, marsupial and placental mammals have contrasting reproductive patterns. Marsupial offspring are born as altricial neonates after a short gestation. Lactation occurs throughout development, and the milk produced from a mammary gland varies in quantity and composition with changes in the energetic requirements of offspring as they grow. In contrast, gestation encompasses a much higher proportion of the period of maternal care in most placental mammals. Relative to marsupials, offspring are generally precocial at birth, the time from conception to weaning is markedly shorter, and milk composition is relatively homogenous during lactation (Krockenberger 2006). Also, energy transfer through lactation in marsupials is under maternal control (Isaac and Johnson 2005) and is less responsive to offspring suckling demand (Krockenberger 2006). Whereas in placental mammals, energy transfer is regulated by foetal hormones, and milk production is more responsive to offspring demand (Isaac and Johnson 2005).

Thus, the marsupial reproductive pattern provides an unique perspective from which to assess patterns of sex ratio variation and offspring survival because of: (1) the altricial stage of development of marsupial offspring at birth; (2) the complex and extended period of maternal care; and (3) the capacity for variation in the life history that can be investigated at the birth, pouch emergence, and end of weaning stages of offspring development during maternal care.

Some marsupials living in highly unpredictable environments breed continuously. The effect of extrinsic factors on maternal condition, and consequently on offspring life histories, are expected to be strong in such species. Given differences in the costs of producing each sex, the costs of facultative adjustment of resources during development should be low because the continuous breeding cycle allows mothers to stop allocating resources in uncertain environmental conditions and to immediately produce another offspring. Sex ratios and offspring survival should therefore be related to maternal condition at important life history stages, indicative of differential resource allocation to the sexes predicted under TWH. However, for species where adults are weakly dimorphic, the effect of differential resource allocation between the sexes may be diminished.

In this study, we assess evidence for the effects of maternal factors (i.e., TWH) and environmental variation (i.e., EMH) on the sex ratio and pre-weaning survival of a weakly dimorphic macropodid marsupial at two life history stages based on data collected from a 12-year capture–recapture study of an isolated population in tropical western Queensland, Australia. We used maternal body mass and condition as indicators of the capacity of individual mothers to allocate resources under the TWH. Similarly, we used rainfall and an index of the El Niño-Southern Oscillation as indicators of local environmental and regional climate conditions under the EMH, respectively. We expect that: (1) birth sex ratios will be unrelated to maternal or environmental factors because this species’ reproductive pattern means that maternal condition and environmental circumstances at conception and birth are uncorrelated; (2) higher quality mothers, or those in good condition, will successfully wean relatively more male offspring; and (3) because there is only weak sexual size dimorphism, environmental conditions will affect survival of both sexes equally. Multi-model inference using information criterion to weight models is used to evaluate the relative importance of maternal, environmental, and joint maternal/environmental factors on sex ratio and survival to test our predictions. We show support for our expectation that birth sex ratios do not depend on maternal or environmental factors. Contrary to our expectations, we show that: (1) environmental conditions strongly influenced offspring survival and sex ratios at pouch emergence (in support of EMH), despite no differences in offspring growth during that developmental stage; and (2) maternal quality affects offspring survival and sex ratios at the end of maternal care (in support of TWH), the developmental stage where males grow faster than females.

Materials and methods

Study species

The allied rock-wallaby, Petrogale assimilis Ramsay 1877, is a medium-sized macropodid marsupial that lives in rocky habitats that provide diurnal shelter from high tropical temperatures. It exhibits many features that make it an ideal species to assess the competing sex ratio variation hypotheses outlined above. P. assimilis is a continuously breeding, polygynous marsupial which lives in the climatically variable wet-dry tropics of northern Australia. Although previously described as monomorphic because differences in body size are small (Delaney 1997; see Table S1), P. assimilis is weakly dimorphic with male body mass about 15% greater than females. Also, P. assimilis exhibits sex-biased dispersal; females are philopatric and males disperse (Spencer 1996). Variance of reproductive success in males is higher than females (Spencer et al. 1998; an assumption of TWH), and male success is dependent on an individual’s ability to defend shelter sites and consorts from competing males, as well as the ability to compete with other males for extra-pair mating opportunities (Spencer et al. 1998). Also, the quality and availability of P. assimilis’ diet varies seasonally as a function of environmental conditions (Horsup and Marsh 1992). The body condition of individual P. assimilis is strongly affected by environmental variation, and higher condition during early development leads to higher relative condition in adulthood (Delean 2007; an assumption of TWH).

Study species reproductive pattern

The reproductive pattern of P. assimilis is characterised by a post-partum oestrus and lactation-controlled embryonic diapause (Tyndale-Biscoe and Renfree 1987; Delaney 1997). Within days of birthing, females enter post-partum oestrus and mate, and the conceived blastocyst enters dormancy and remains quiescent until development of the embryo is reactivated (Fig. 1). Single offspring gestate for 30 days (Spencer 1996; Delaney 1997) and newborn neonates climb to the mother’s pouch, attach to one of four teats, and suckle for about 200 days until pouch emergence when the fully developed offspring permanently exits the pouch. Offspring continue to suckle intermittently during the young-at-foot (YAF) stage until the end of weaning (Fig. 1). If the reproductive cycle is interrupted by offspring mortality, development of the quiescent embryo is reactivated and birth follows gestation (Fig. 1). Post-partum oestrus may not result in successful conception, and therefore, the timing of oestrus and conception can be variable (Tyndale-Biscoe and Renfree 1987).

Fig. 1
figure 1

The reproductive pattern of Petrogale assimilis. Births are followed by a post-partum oestrus, and the resulting embryo remains quiescent until lactation-controlled embryonic diapause ends with the mortality or completed development of the offspring. A 30-day gestation period follows the natural death of an offspring or begins when a surviving offspring reaches approximately 170 days of age. Offspring permanently exit the pouch (pouch emergence) at about 200 days and continue to suckle as young-at-foot (YAF) until the end of weaning (approximately 330 days). Births either occur immediately after pouch emergence or after a 30-day gestation following mortality of the previous offspring

Full size image

In our study population, P. assimilis exhibits behavioural pair-bonding (Horsup 1996) and a polygynous mating system (Spencer 1996) with a high proportion of extra-pair matings (33%, Spencer et al. 1998). Heterogeneous mating tactics result in some females always mating with their consort, some never mating with their consort, and other females employing mixed tactics (Spencer et al. 1998). The last group has the highest success in raising offspring to independence. Variability in mating tactics is associated with phenotypic quality; males with longer arms are more likely to father extra-pair offspring (Spencer et al. 1998). Arm length is an index of upper body musculature that likely provides an advantage in male–male aggressive interactions. Thus, the variance of reproductive success in males is higher than females (Spencer et al. 1998).

Data collection

This capture–recapture study was conducted over 12 years at Black Rock (19°04′38″S, 144°27′23″E), a small sandstone escarpment 265 km WNW of Townsville, Australia. Rock-wallabies were captured at night using a grid of 60 baited cage traps set over four to seven consecutive nights on 101 sampling occasions between 1986 and 1997. Individual rock-wallabies were marked on first capture using ear tattoos and passive integrated transponder tags (Animal ID Systems, Sydney).

Female rock-wallabies were examined to determine their reproductive status and record the developmental stage of offspring on first capture on each sampling occasion. Offspring body mass was recorded when individuals had naturally detached from the teat (usually post-100 days of age). The sex was recorded for offspring older than 7–14 days, but newborn offspring were not sexed because: (1) determining sex was difficult and (2) stress from handling increased the risk of capture-related mortality. Offspring of unknown sex were excluded from analyses. The ages of offspring were predicted from body measurements using growth equations (Delean 2007). The fates of most offspring between birth and pouch emergence were known because: (1) the probability of recapturing adult females on the next sampling occasion was high (\( \overline X = 0.92 \), 95% CI = 0.89, 0.93; Delean 2007), and (2) the frequency of sampling allowed females to be captured several times during the development of each offspring that survived to pouch emergence.

Based on the continuous reproductive pattern of P. assimilis, the estimated survival age for individual offspring was determined from the age of the mother’s next offspring on first capture. Offspring age at pouch emergence was calculated as its estimated age at last capture, plus the number of days between the last and penultimate captures of the mother, minus the age of the newborn offspring. Age at mortality for offspring not surviving to pouch emergence was calculated in the same way but also subtracting the length of gestation (i.e., 30.2 days (95% CI = 28.5, 31.9; Spencer 1996, Delaney 1997). The lactation status of mothers’ teats was also used for determining offspring survival to pouch emergence. Offspring that were predicted to have survived to pouch emergence based on the age of the next offspring, but were not subsequently captured as YAF offspring, were assumed to have survived if the teat they suckled from was elongated and producing milk. Teat status was also used in conjunction with captures of YAF offspring to determine the age at mortality for pre-weaning offspring or the offspring age at successful weaning.

A total of 384 offspring (183 females and 201 males) from 77 mothers was used to examine variation in the sex ratio and survival between birth and pouch emergence. Individual mothers had between one and 24 offspring (median = 3; interquartile range (IQR) = 2–9). We identified 27 offspring that possibly died from capture-related causes and were excluded from analysis to avoid bias from including non-natural mortality events. Survival of 205 YAF offspring (101 female and 104 male) between pouch emergence and the end of weaning from 56 mothers was examined. Individual mothers had between one and ten YAF offspring (median = 3; IQR = 1.0–5.3). Body mass and condition index data for 50 pouch offspring (180–200 days old) and 70 YAF offspring near weaning (242–399 days old) were used to assess relationships between offspring and maternal condition.

Environmental and maternal covariates

Cumulative rainfall totals and the mean Southern Oscillation Index (SOI; representing regional climate variability associated with the El Niño-Southern Oscillation phenomenon) were calculated over the 3, 6, 9, and 12 months preceding each sampling occasion. These environmental covariates over each lag period were included in models separately, and the best explanatory lag periods were chosen in a preliminary stage of model selection (see “Statistical methods”). The best lag period for each variable was then used in all subsequent analyses. The population size on the preceding sampling occasion was used to investigate density-dependent relationships. The population size variable was retained in all candidate models where a density effect was evident in the most complex model and was otherwise excluded.

The body mass of each individual (at each capture) and the body condition index (estimated as residual deviations from the relationship between body mass and head length (Delean 2007)) were used as measures of size and nutritional status, respectively. Maternal age, determined from body measurements or dental radiography (Delean 2007), was used as a covariate in preliminary analyses, but did not explain significant variation in sex ratios or survival and was therefore excluded from further analyses. The explanatory variables we used were generally uncorrelated, but there were some weak yet significant correlations (see Table S2).

Statistical methods

We examined candidate sets of models for six response variables: (1) sex ratio at birth, pouch emergence, and weaning; (2) survival from birth to pouch emergence and from emergence to weaning; and (3) offspring body mass at weaning. We fitted three sets of candidate models to each response made up of covariates representing: (1) maternal effects of body mass and condition associated with TWH, (2) environmental effects of cumulative rainfall and mean SOI associated with EMH, and (3) combinations of maternal and environmental effects indicating joint TWH/EMH effects. Each candidate set contained models for each covariate separately and for the additive effects of the covariates. Model sets for the analysis of survival and weaning mass also contained the indicator variable offspring sex and interactions between sex and each covariate. Offspring age was included as a covariate in models for weaning mass to account for age differences in the timing of weaning. For each of the response variables, we selected the model with the highest support (see below) from each of the three sets of candidate models representing the best TWH, EMH, and joint TWH/EMH model. The list of all candidate models fitted to each response and summary model selection statistics are provided in Table S3 (A–F). A summary table of the TWH and EMH effects on each response variable, and associated assumptions, is also provided (Table S4).

Model selection was based on Akaike’s information criterion (AIC, Akaike 1973) corrected for finite sample sizes (AIC c , Burnham and Anderson 2002), and the relative support for each model was expressed as Akaike weights (Burnham and Anderson 2002). The model weights among the best TWH, EMH, and joint TWH/EMH model may be inflated relative to their weight among all models assessed if more than one model in each candidate set of models had substantial support. Thus, model weights among all models considered are presented in Table S3 (A–F). The percent deviance explained by each model provided a measure of its predictive power.

Offspring sex ratios were analysed using conditional generalised linear mixed models (Schall 1991) with binomial variance function and logit link. Such models included independent random effects to account for: (1) the correlation between repeated observations on individual mothers and (2) random between-year variation. Model parameters were estimated using maximum likelihood. Effect sizes are presented on the scale of the odds of a male birth, and the partial effects of covariates are shown graphically on the scale of the probability of a male birth. Confidence intervals (95%) are provided as measures of precision (unless otherwise stated) and are presented in parentheses.

Survival analysis models (Therneau and Grambsch 2000) were used to examine the relationships between time to mortality and environmental and maternal factors. Survival analysis accommodates censored data that occurred in this study because of: (1) emigration from the study area, (2) survival beyond the life history stage of interest, or (3) survival beyond the end of the study period. For the semi-parametric Cox proportional hazards survival model (Therneau and Grambsch 2000), the effect of each covariate multiplies the hazard by a constant factor, and change is inversely proportional to the likelihood of survival. Such models assume that hazards associated with covariates are proportional over time.

Time-dependent survival models extend the proportional hazards model to estimate the effects on the hazard of covariates that change over time (Therneau and Grambsch 2000). The time-dependent survival model does not assume a constant hazard over time or that survival between individuals with identical covariates is independent. The time-dependent approach used all observations recorded on each offspring within a life history stage to estimate how changes in time-varying covariates affected survival as offspring aged. Importantly, time-dependent models also allowed individuals to enter the risk set at any time over the risk interval. Frailty terms (Therneau and Grambsch 2000) were used to include mother identity as a random effect in the analyses. Effect sizes are expressed as percent differences in the relative hazard (i.e., likelihood of mortality) between individuals whose covariate values differ by one unit, with all other covariates set to their mean or reference levels. Effects are presented graphically as the probability of survival to the mean age at pouch emergence or the end of weaning.

Results

Sex ratio variation at birth

The birth sex ratio did not differ significantly from parity (probability of male = 0.52 (95% CI = 0.47, 0.57); n = 384), and did not depend on maternal factors or environmental variation. The null model containing only mother identity had highest support (AIC c weight = 0.65; Table 1). Also, the estimated between-mother variance was effectively zero (σ 2ID  = 1.1 × 10−5), indicating heterogeneity in the sex ratio between mothers was negligible and suggesting a constant-only model would provide the most parsimonious explanation of the data.

Table 1 Model selection statistics for the best Trivers–Willard hypothesis (TWH), extrinsic modification hypothesis (EMH), and joint TWH + EMH models from the sets of candidate models (see Table S3 (A–E)) for sex ratios and survival of Petrogale assimilis offspring at two life history stages
Full size table

Survival from birth to pouch emergence

There was strong support for EMH in survival from birth to pouch emergence; survival differed between the sexes depending on rainfall variation (Table 1). Survival increased for males with a doubling of total rainfall over the preceding 6 months but did not change substantially for females (Table 2; Fig. 3a), showing sex-differential survival resulted from lower male survival under poor environmental conditions. There were also additive maternal and environmental effects on survival that were common to both sexes (AIC c weight = 0.81; Table 1). Survival increased under improved climatic conditions (as indicated by SOI) and was lower at higher population sizes (Table 2). Survival was also positively related to changes in maternal mass; the offspring of heavier mothers had higher survival than the offspring of lighter mothers (Table 2). The combined EMH and maternal mass model was about four times more likely than EMH-only but had many times more support than maternal mass alone (Table 1). Environmental factors were therefore relatively more important for survival to pouch emergence.

Table 2 Estimated coefficients from the analysis of environmental and maternal factors on survival and sex ratio of Petrogale assimilis offspring at two life history stages
Full size table

The mean age of offspring at pouch emergence was 199 days (95% CI = 197, 200). On average, there was no significant difference in survival between the sexes (χ 2 = 0.18, df = 1, P = 0.671; Fig. 2a), and the assumption of proportional hazards was satisfied (χ 2 = 0.22, df = 1, P = 0.639). The survival probability to median age at pouch emergence was 0.57 (0.49, 0.67) for females and 0.62 (0.54, 0.71) for males.

Fig. 2
figure 2

Baseline probability of survival from: (a) birth to pouch emergence for all known-sex individuals; and (b) pouch emergence to weaning, conditional on surviving to pouch emergence, for female (thin line) and male (thick line) Petrogale assimilis at Black Rock. Solid lines represent mean survival; dashed lines represent 95% confidence intervals

Full size image

Sex ratio variation at pouch emergence

Additive TWH and EMH effects best explained the sex ratio at pouch emergence (AIC c weight = 0.44; Table 1). The sex ratio was male-biased under good environmental conditions (positive SOI) and female-biased in poor conditions (Table 2). There was some evidence that the sex ratio was male-biased for mothers in good condition but unbiased for mothers in poorer condition; however, the EMH-only model had almost equal support and was more parsimonious (Table 2). The additive TWH + EMH model was 1.3 times more likely than EMH-only but more than four times more likely than TWH-only (Table 1). Therefore, climatic conditions were relatively more important predictors of male-biased sex ratios than maternal condition.

The overall offspring sex ratio at pouch emergence did not differ significantly from parity (probability of male = 0.53 (0.46, 0.60); n = 195). The estimated between-mother variance was again negligible (σ 2ID  = 8 × 10−7), showing individual maternal traits did not influence offspring sex ratio at pouch emergence.

Body mass and condition at pouch emergence

Although EMH was an important predictor of pouch emergence sex ratios, there was no evidence for differences in growth rates because offspring body mass at pouch emergence did not differ between the sexes (t 48 = 0.64, P = 0.263). Offspring that subsequently survived to weaning had slightly higher mass at pouch emergence (t 48 = 1.79, P = 0.040), and this relationship was independent of sex (F 1, 46 = 0.44, P = 0.510). The mean pouch emergence body mass of individuals that survived until weaned was 434 g (396, 473), and for individuals that did not survive was 386 g (348, 425).

Consistent with one of the TWH assumptions, offspring body condition was correlated with maternal body condition (r = 0.33 (0.05, 0.56); t 46 = 2.36, P = 0.011), and this relationship did not differ between the sexes (F 2, 44 = 0.16, P = 0.853). There were no significant sex differences in offspring body condition at pouch emergence (t 47 = 0.34, P = 0.368).

Survival from pouch emergence to weaning

There was strong evidence for TWH effects in YAF offspring survival from pouch emergence to weaning (Table 1); for lighter mothers, survival of male YAF offspring was lower than females, but there were no sex differences in survival of YAF offspring of heavier mothers (Table 2; Fig. 3c). There were also additive environmental effects on survival that were common to both sexes (AIC c weight = 0.62; Table 1). A doubling of total rainfall in the previous 6 months increased YAF offspring survival (Table 2). Also, survival to weaning changed nonlinearly with mean SOI (Fig. 3b); survival increased from low to average SOI values and subsequently decreased toward higher SOI values. The decrease in survival under positive SOI conditions was counterintuitive but resulted from one period in 1989 where SOI was very high but survival was very low. This lower survival probably resulted from predation by a feral cat that significantly impacted pre-weaning survival, particularly during that period (Spencer 1991). The combination of TWH and environmental effects was 1.6 times more likely than environment-only but many times more likely than TWH-only (Table 1). Thus, environmental conditions were relatively more important for survival to weaning.

Fig. 3
figure 3

Partial effects plots of: (a) rainfall (log2) on the sex-specific probability of survival to pouch emergence (200 days); (b) 6-month mean Southern Oscillation Index on the probability of survival from pouch emergence to weaning (330 days); and (c) maternal mass (kg) on the sex-specific probability of survival from pouch emergence to weaning (330 days) for Petrogale assimilis at Black Rock. Effects were estimated constraining all other model parameters to their mean values. Thick lines in (a) and (c) show male survival; dashed lines represent 95% confidence intervals

Full size image

On average, survival from pouch emergence to weaning was lower for males than females (χ 2 = 5.6, df = 1, P = 0.018), with male survival 1.5 (1.1, 2.2) times less likely than female survival (Fig. 2b). Also, the hazard was proportional between the sexes (χ 2 = 0.064, df = 1, P = 0.800). The mean age of 69 YAF offspring (33 females and 36 males) known to survive to the end of weaning was 335 days ((328, 342); range = 286–399). The survival probability to mean weaning age was 0.46 (0.35, 0.57) for females and 0.28 (0.19, 0.38) for males.

Sex ratio variation at weaning

TWH effects best explained the sex ratio at weaning (AIC c weight = 0.66; Table 1); sex ratios were female-biased for lighter mothers and male-biased for heavier mothers (Table 2). The TWH model was twice as likely as the joint TWH/EMH model (Table 1). There was substantial between-mother variance (σ 2ID  = 0.7), suggesting that some mothers were more likely than others to successfully wean males. However, the between-mother variance was effectively zero after accounting for maternal body mass (σ 2ID  = 10−6), suggesting that differences in maternal quality associated with body mass influenced the capacity of mothers to successfully wean male or female offspring. The overall offspring sex ratio at the end of weaning did not differ significantly from parity (probability of male = 0.39 (0.25, 0.55); n = 70).

Body mass at weaning

Offspring body mass at weaning differed between the sexes after conditioning on the age at weaning and also depended on maternal body mass (AIC c weight = 0.57; Table S3-F). At the end of the period of maternal care, mothers heavier by 1 kg weaned YAF offspring that were heavier by 0.16 kg (0.03, 0.28). There was no association between the index of body condition of YAF offspring at weaning and either maternal or environmental factors.

The body mass of males at weaning was 0.23 kg (0.09, 0.37) heavier than females, so males increased in mass faster than females during this developmental stage. Mean body mass at weaning was 1.37 kg (1.28, 1.46) for females and 1.60 kg (1.49, 1.71) for males. Thus, males were about 15% heavier than females at the end of maternal care, the same relative difference as for adult body mass. Also, the rate of increase in body mass with age at weaning differed between the sexes; 50-day increases in age at weaning resulted in an average increase in body mass of 0.17 kg (0.09, 0.25) for females and 0.35 kg (0.23, 0.48) for males (Fig. 4).

Fig. 4
figure 4

Partial effects plot of age (days) on young-at-foot (YAF) offspring body mass at weaning for Petrogale assimilis at Black Rock. Solid lines represent estimated body mass for females (thin lines) and males (thick lines); dashed lines represent 95% confidence intervals

Full size image

Discussion

Birth sex ratio variation

The birth sex ratio for P. assimilis was unity and did not depend on maternal factors or environmental variation in the months preceding birth. There are many examples of birth sex ratio varying with resource availability in mammals (Clutton-Brock and Iason 1986; Hewison and Gaillard 1999; West and Sheldon 2002), and the direction of this variation in sexually dimorphic macropodid marsupials is generally consistent with TWH predictions (Stuart-Dick and Higginbottom 1989; Sunnucks and Taylor 1997; Fisher 1999). However, sex ratio adjustment is expected to be greater where environmental conditions are predictable at conception (West and Sheldon 2002) and is therefore more likely evident if maternal condition is measured at conception rather than at birth (Cameron 2004). For example, in the continuously breeding bridled nailtail wallaby, Onychogalea fraenata, females in good condition produce males, while those in poor condition produce females (Fisher 1999). These biases occurred at conception rather than later during the period of maternal care (Fisher 1999), and females enter oestrus and conceive just prior to birth so maternal condition at conception and at birth are correlated.

In contrast, P. assimilis are continuous breeders with lactation-controlled embryonic diapause and post-partum oestrus (Spencer 1996; Delaney 1997), so conception usually occurred in the days immediately after the birth of offspring (though the exact timings of conception were unknown). As environmental stochasticity can result in low and variable correlations between environmental conditions at conception, birth, and during the period of maternal care (Nager et al. 1999), the environmental conditions and nutritional state of P. assimilis mothers at conception were unlikely to be good predictors of environmental and maternal conditions up to 7 months later at birth. Consequently, selection pressure for maternal control of birth sex ratios would be low.

Survival and sex ratio variation at pouch emergence

Our results showed environmental conditions strongly influenced offspring survival from birth to pouch emergence (in support of EMH); male survival depended on rainfall, but female survival did not. These EMH effects consequently influenced offspring sex ratios at pouch emergence. Also, maternal quality and regional El Niño-Southern Oscillation climate patterns affected the survival of each sex equally, and the cumulative positive effects of these factors produced male-biased sex ratios, though climate effects were relatively more important. Despite our evidence for EMH, offspring body mass at emergence, and hence growth rates during pouch development, did not differ between the sexes.

There is evidence that mammalian sex ratios are related to sex-specific differences in foetal growth that interact with environmental conditions (Forchhammer 2000) or developmental asynchrony resulting from sexual selection (Krackow et al. 2003). Environmental variability can be expected to have a stronger effect on offspring survival in marsupials than in placental mammals because maternal control in the rate of energy transfer to marsupial offspring (Isaac and Johnson 2005) may afford greater flexibility to compensate for fluctuations in resources and maternal condition. We found higher survival for P. assimilis offspring of superior quality mothers and male-biased pouch emergence sex ratios for mothers that maintained condition, suggesting that males may have had higher energy requirements, or were more susceptible to changes in maternal condition affected by environmental conditions, late in their pouch development.

The evidence for sex ratio manipulation in mammal populations is inconsistent (Hewison and Gaillard 1999), but this inconsistency may be due to the fact that multiple factors appear to influence sex ratios (Kruuk et al. 1999; Post et al. 1999; Mysterud et al. 2000; Cockburn et al. 2002; Isaac et al. 2005). The direct and indirect influences of environmental stochasticity (i.e., EMH) on sex ratios can be substantial (Krackow 2002; Cameron 2004; Love et al. 2005), and their importance is evident in mammal populations (Kruuk et al. 1999; Post et al. 1999). In fact, any extrinsic factors that affect maternal condition, including population density and climate, may also affect sex ratios. For example, climate conditions driven by the North Atlantic Oscillation (NAO) influence offspring sex ratios in red deer, Cervus elaphus, through the direct impact of winter severity on maternal condition (Post et al. 1999), but the effect of NAO varies among populations, and population density correlates more strongly with the nutritional stress that predicts sex ratio variation (Mysterud et al. 2000).

The inability to distinguish between the modification of sex ratios at conception and sex-differential mortality during gestation that result from environmental conditions is a major shortcoming of studies of sex ratio variation in ungulates (Hewison and Gaillard 1999). The extrinsic modification of sex ratios in ungulate populations, predicted from relationships with climate and population density (Kruuk et al. 1999; Post et al. 1999; Mysterud et al. 2000), suggests sex allocation may not be adaptive (Mysterud et al. 2000). For example, the mechanisms hypothesised to determine sex ratios in C. elaphus include both higher in utero mortality of males due to poor environmental conditions and modification of the primary sex ratio at conception (Mysterud et al. 2000), though the specific mechanism is unknown. For P. assimilis, the mechanism determining male-biased pouch emergence sex ratios was lower male survival in poor environmental conditions, as no bias was evident at birth.

Survival and sex ratio variation at weaning

We found strong evidence for TWH effects in offspring survival to the end of maternal care; male offspring of lighter mothers were less likely to survive than those of heavier mothers, whereas female offspring survival was independent of maternal mass. Environmental conditions were also important predictors of male and female survival, but sex ratios at the end of maternal care were best explained by differences in maternal mass. Male-biased allocation among higher quality mothers is consistent with TWH in polygynous, sexually dimorphic mammals (Hewison and Gaillard 1999; Cameron 2004; Sheldon and West 2004). However, there is also evidence for TWH for weakly dimorphic ungulate species (Gaillard et al. 1997; Hewison et al. 2005) and in some birds (Frank 1990; Krackow 1995). TWH effects are expected to be stronger in species with more marked sexual size dimorphism (Sheldon and West 2004). However, sexual size dimorphism is not sufficient to expect the hypothesis to hold (Hewison and Gaillard 1999) and in fact can be viewed as a confounding factor in the interpretation of TWH because of the greater cost of males in such species (Cameron and Linklater 2002). Higher quality mothers tend to produce more sons or allocate more resources to sons in sexually dimorphic species, but studies have not generally shown higher allocation in daughters by lower quality mothers (Cameron and Linklater 2000). Differences in allocation by lower quality mothers may be less evident because of the high requirements of all males in dimorphic species associated with their larger size and higher growth rates (Cameron and Linklater 2000).

Yet TWH explicitly predicts that mothers in poorer condition are favoured by producing daughters in the same way that mothers in better condition gain more by producing sons (Hewison and Gaillard 1999; Cameron and Linklater 2000). The former prediction generally receives little attention in empirical sex ratio studies. For the polygynous, weakly dimorphic P. assimilis, lower quality mothers were much more likely to successfully wean daughters than higher quality mothers that were more likely to successfully produce sons, providing support for TWH. This finding is important because sexual size dimorphism in P. assimilis is weak, and the evidence here supports the predictions of TWH in both directions.

However, sons had higher body mass than daughters at the end of maternal care as a result of higher male growth rates between pouch emergence and weaning. The sex difference in body mass (approximately 15%) was of the same magnitude as adult body mass differences due to sexual selection for higher male body mass. Thus, only the highest quality mothers were able to afford the higher cost of successfully raising male P. assimilis offspring to weaning before nutritional stress had a substantial effect on mothers’ condition, which in turn may affect future survival and reproductive potential (consistent with the ‘costs of reproduction hypothesis’ (Cockburn et al. 2002)). In many ungulate herbivores, heavier mothers tend to produce heavier offspring than lighter mothers (Côté and Festa-Bianchet 2001; Weladji et al. 2003; Loison et al. 2004), and heavier offspring tend to have higher survival to weaning and sexual maturity than lighter offspring (Clutton-Brock et al. 1992; Loison et al. 1999).

Thus, in the absence of direct evidence of sex-biased maternal discrimination, sex-biased survival is more likely a function of differences in energetic demands of offspring rather than adaptive adjustment (Moses et al. 1998). The pervasive effects of environmental conditions on survival of P. assimilis offspring throughout maternal care suggest offspring sex ratios at weaning were more likely determined by extrinsic factors affecting mothers’ capacity to meet sex-differing resource requirements. Previous research has shown that maternal condition in P. assimilis correlates with lagged environmental variation, and that survival of prime-aged adult females is high and buffered against environmental variation (Delean 2007). Our results here suggest that the synergistic effects of environmental and maternal factors mean the benefits of producing daughters for lower quality mothers were substantially higher than producing sons that would be in poorer condition during early development and therefore have a lower lifetime reproductive success.

Conclusions

Our results show that extrinsic factors (i.e., EMH) strongly influenced offspring survival and sex ratios at pouch emergence, despite no substantive differences in offspring growth during that developmental stage. Thereafter, maternal quality (i.e., TWH) affected male offspring survival and, consequently, sex ratios at the end of maternal care. However, males grow faster than females during weaning, and environmental effects on survival and sex ratios were more important than maternal factors throughout the period of maternal care. We therefore conclude that offspring sex ratios at the end of maternal care were more likely determined by extrinsic factors affecting the capacity of mothers of different quality to meet sex-differing energy requirements, rather than adaptive allocation of resources. The timing of assessment of sex ratio variation was critical because different proximate factors operated at different life history stages to influence sex-differential survival that contributed to biased secondary sex ratios. These results add to increasing evidence that multiple mechanisms, including environmental stochasticity associated with El Niño-Southern Oscillation-driven climate variation, simultaneously affect sex ratio variation in mammal populations (Kruuk et al. 1999; Cockburn et al. 2002; Isaac et al. 2005).