J. L. Isaac, School of Tropical Biology, James Cook University, Townsville Qld 4811, Australia. Tel:. 61 747815715; Fax: 61 747251570; E-mail: firstname.lastname@example.org
1Maternal control of offspring sex remains a contentious topic in the evolutionary and behavioural ecology of vertebrates. Two models of adaptive sex allocation − the Trivers–Willard (TW) and local resource competition (LRC) hypotheses − have been supported in some studies, but in many other cases the forces shaping offspring sex ratios are unclear. One reason that past studies may not have provided clear support for either model is that the two processes interact, producing complex patterns of sex allocation in relation to the life histories of individual mothers.
2The TW hypothesis predicts that in dimorphic, polygynous mammals, females in good condition should be more likely to produce male offspring. The LRC hypothesis predicts an independent effect of age on the offspring sex ratio; females should be more likely to produce sons early in their reproductive lives and daughters later, in order to minimize the duration and reproductive cost of mother–daughter competition. We tested these two predictions in a study of sex allocation in the common brushtail possum, Trichosurus vulpecula.
3Offspring sex was strongly dependent on maternal age: females breeding for the first time were likely to produce sons, and at subsequent breeding attempts the sex ratio of offspring was slightly biased towards females. After the first breeding season, sex of offspring was influenced more strongly by body condition than by age, and females in good condition were more likely to have sons. Individual females that showed a between-year gain in condition were also more likely to produce a male in the subsequent year.
4Body mass of male offspring at 5 months was positively correlated with maternal condition. No relationship was found between maternal condition and mass of female offspring.
5We conclude that patterns of change in sex allocation in relation to individual life histories of female common brushtail possums can be explained by the interaction of TW and LRC effects.
That natural selection may favour maternal control of offspring sex continues to be a controversial topic in evolutionary biology (Saltz 2001; Cameron & Linklater 2002; Carranza 2002; Hewison et al. 2002; Saltz & Kotler 2003). Fisher (1930) predicted that natural selection should favour an equal sex ratio at birth because if one sex is rare, it gains a frequency dependent mating advantage. However, a critical assumption of Fisher's model − that sons and daughters are equally costly for a mother to produce − is not met in most polygynous mammals where sons are often larger than daughters at the end of lactation (Kojola 1998) and the costs of rearing male offspring are thought to exceed those of rearing a female (Clutton-Brock, Albon & Guinness 1981).
Trivers & Willard (1973) proposed that maternal condition could influence the sex ratio of offspring in dimorphic polygynous mammals. They argued that females in good condition would produce larger offspring that, if male, would have a disproportionately high mating success. The Trivers–Willard (TW) model is dependent upon three key assumptions: 1, offspring condition/phenotype is correlated to the condition/phenotype of its mother; 2, differences in offspring condition endure into adulthood; and 3, in polygynous mating systems with high variance in male reproductive success, adult males benefit more than females from an advantage in condition (see reviews in Clutton-Brock & Iason 1986; Cockburn 1990; Hewison & Gaillard 1999).
The local resource competition (LRC) hypothesis (Clark 1978) proposes that competition for resources between mothers and philopatric offspring imposes a cost of producing offspring of the philopatric sex. This hypothesis predicts that when offspring of one sex remain philopatric and the other disperses, females should bias offspring sex ratio toward the dispersing sex to avoid ecological competition with their offspring. The majority of mammals show female philopatry, while males disperse. Therefore, according to the LRC model, if resources are limiting females should be expected to produce a male-biased primary sex ratio (Johnson 1988; Cockburn 1990; Dittus 1998).
It has been suggested that it is unlikely that a single selective pressure is responsible for generating sex ratio biases and instead several selective pressures may operate simultaneously to produce any bias or pattern of bias (Dittus 1998; Cockburn et al. 2002). Thus, results may reflect the sum of those pressures and, as a consequence, are likely to be inherently difficult to interpret (Cockburn et al. 2002).
The TW hypothesis considers sex ratios at the level of the individual while the LRC model is usually applied at the level of the population or species (Ward 2003). However, LRC can also shape sex allocation through life-history variation in individual females, as the duration of mother–daughter competition is influenced by maternal age (Dittus 1998; Cockburn et al. 2002). It might be predicted therefore that patterns of sex-ratio bias will be influenced by age-related shifts in the importance of TW and LRC effects. LRC effects are expected to be more pronounced in young, primiparous females (Ward 2003), who should show a male bias in order to avoid the potential for lifetime competition with a philopatric daughter. A consequence of this may be that after producing a male bias early in life, females should show a shift or reversal in their offspring sex ratio (Ward 2003). Older mothers might also produce more females, as they are less likely to live to compete with a surviving daughter (McShea & Madison 1986; Cockburn 1994).
In this study, we test these predictions in the common brushtail possum (Trichosurus vulpecula Kerr). T. vulpecula meets assumptions inherent in both the TW and LRC hypotheses. It has a polygynous mating system (Smith & Lee 1984) in which males compete for access to oestrus females (Winter 1976). There is male-biased sexual dimorphism (Isaac & Johnson 2003) and in Australian populations larger males have higher reproductive success (Clinchy et al. 2004). A number of studies have also shown that dispersal in T. vulpecula is male biased, while daughters are philopatric and settle in a home range that often overlaps or is adjacent to that of their mother (Winter 1976; Clout & Efford 1984; Johnson et al. 2001). Females produce only a single offspring at each reproductive event, avoiding the confounding influence of multiple offspring and mixed-sex litters (Cockburn et al. 2002). Young can also be sexed, aged and studied alongside their mother while in the pouch. Previous research on offspring sex ratios among populations of T. vulpecula has shown that facultative adjustment of the sex ratio occurs when resources (den sites) are limiting, consistent with the predictions of the LRC model (Johnson et al. 2001) and that biases in offspring sex ratios are established prior to birth (Johnson & Ritchie 2002).
The study was conducted over three breeding seasons (2001–03) on a ∼10 ha site in mixed open eucalypt woodland on Magnetic Island, North Queensland (19°10′ S, 146°50′ E). A trapping grid was established at the site in May 2001, consisting of 30 trap sites spaced approximately 25 m apart, each with two traps (i.e. 60 traps in total). Trapping sessions were conducted once monthly for the duration of the study; 28 sessions in total. The majority of offspring were sexed while still pink and permanently attached to the teat and the remainder were sexed prior to weaning, while still riding on their mothers’ back. A number of females were known to give birth to two offspring during the same year and second births were not included in the analysis unless specified.
Females first captured as independent juveniles or adults were aged using a tooth-wear index developed specifically for T. vulpecula by Winter (1980). Tooth wear of the first upper molar was determined by visual inspection during trapping sessions; it was unnecessary to anaesthetize animals for this procedure. Exact ages were known for all females born during the study. Animals were also weighed each month and measurements of head length were taken.
A condition score was calculated for each female as the ratio of her observed mass to that predicted from the regression of body mass on head length (see Johnson et al. 2001). Condition scores used are those calculated the month prior to birth of offspring, as this represented most effectively the mother's condition at the time of conception (see Cameron 2004). Age and maternal condition were positively correlated (R2 = 0·32, P 0·0001, n = 65) and therefore, to avoid potentially confounding results (see Saltz 2001; Hewison et al. 2002; Saltz & Kotler 2003), we used the residuals of maternal condition on age in all analyses except when looking at the raw data. Sample sizes vary, as not all information was available for all mother–offspring pairs.
We used a nominal logistic regression model with random effects to test for the effects of maternal age and condition on offspring sex (SAS Institute 1995). Maternal identity and year were entered into the model as random variables to account for repeated observations of females contributing more than one offspring to the data set and any between-year variation in pouch-young sex (see Côté & Festa-Bianchet 2001). Observed sex ratios were compared to an expected random (binomial) distribution with a mean of 0·5 using Fisher's exact test. Sex ratios were expressed as the proportion of males [males/(males + females)] in all cases.
We checked observed relationships between offspring sex ratio and reproductive stage of the mother (primiparous vs. multiparous) and maternal body condition (measured as above) using data from six mainland populations at the same latitude as Magnetic Island studied by Johnson et al. (2001) and Johnson & Ritchie (2002). We do not have detailed data on age for females in these populations so we cannot extend the analysis to age-specific effects, as we do for the Magnetic Island study population.
maternal age and offspring sex ratios
For the Magnetic Island study population a total of 34 individual breeding females were sampled and 53 pouch-young were sexed and used in the analysis. Mean age at primiparity was 1·91 ± 0·66 years (n = 13), ranging from 1 to 3 years. The overall sex ratio, including all females, in 2001–03 did not differ from parity (25 males: 28 females, P = 0·1, n = 53). We found a significant difference in offspring sex ratios for primiparous and multiparous mothers. At Magnetic Island, females breeding for the first time produced more sons than daughters, and at subsequent reproductions females produced more daughters than sons. There was a similar direction of change in offspring sex ratio in the mainland populations, although in the mainland sample the sex ratio of offspring born to multiparous mothers was still male-biased. The birth sex ratio in primiparous females was significantly male-biased on Magnetic Island, the mainland population and a pooled sample (Table 1). The proportion of male and female offspring born to multiparous and primiparous females differed significantly in the Magnetic Island sample and in a pooled sample including Magnetic Island and the mainland populations. This trend was present, but not significant, in the mainland populations alone (Table 1).
Table 1. Comparison of proportion of male and female offspring produced by primiparous and multiparous females in mainland populations on Magnetic Island, and a combined sample. Binomial probability values show the probability that the observed proportion differs from that predicted by an expected random (binomial) distribution. χ2 results test for a difference between the number of male and female offspring produced by multiparous and primiparous mothers
Binomal probability P
maternal condition and offspring sex ratios
The raw data (Fig. 1) show a positive relationship between maternal condition and the proportion of male offspring in females aged ≥ 3 years of age. However, this relationship was not evident for primiparous females (aged ≤ 2 years; Fig. 1). There was no effect of either maternal identity or year on offspring sex and these factors were excluded from all further analyses.
On Magnetic Island, condition was found to be the best predictor of offspring sex (maternal condition: Wald statistic = 3·71, P = 0·04; maternal age: Wald statistic = 2·87, P = 0·09); females in better condition were more likely to have sons (Fig. 2). The effect of maternal age was not significant in multiparous females after the effect of condition was removed. The proportion of male offspring produced by females in the mainland populations was also positively related to maternal condition (logistic regression: χ2 = 4·45, P = 0·03, n = 153).
Offspring sex in individual, multiparous females on Magnetic Island was also related to between-year shifts in maternal condition; females who gained in condition from one year to the next were more likely to give birth to a male offspring (Fig. 3: logistic regression: χ2 = 4·53, P = 0·03, n = 17). Between-years variation in maternal condition was not influenced by the sex of offspring born in the preceding year (t = −1·67, P > 0·05, d.f. = 13).
For those females that gave birth to males, offspring mass at 5 months was correlated with maternal condition (Fig. 4, R2 = 0·54, P < 0·05, d.f. = 10); however, no relationship was found between maternal condition and mass of female offspring (Fig. 4, R2 = 0·11, P > 0·05, d.f. = 17).
Females that produced two offspring within a single breeding season were shown to produce a female-biased sex ratio in both cohorts, although the bias did not differ significantly from parity in either cohort (Table 2). There was no evidence that these double breeders produced a different sex-ratio bias between cohorts (Table 2).
Table 2. Comparison of the proportion of male and female offspring produced in the first and second cohort by females that double-bred within an annual breeding season
Binomial probability P
χ2 = 0·43, P = 0·62.
Our results show evidence that TW effects are operating in multiparous females in T. vulpecula, as females in relatively good condition were found to produce more sons. However, given that the TW model is based on individual life-history strategy (Hewison et al. 2002), a more important result is perhaps that individual females were found to adjust their sex ratio adaptively between years according to shifts in maternal condition. Similar results have been found in another marsupial, Onychogalea fraenata (bridled nailtail wallaby: Fisher 1999), where females that lost weight following drought were more likely to produce daughters.
Our data also meet, as far as they are able, assumptions 1 and 3 of the TW model. However, because all male offspring except two dispersed from the site we cannot verify assumption 2, predicting that differences in offspring condition endure into adulthood. In accordance with assumption 1, we found that mass of male offspring at 5 months, just prior to independence, was positively correlated with maternal condition. In T. vulpecula, sons begin to disperse away from their natal range at approximately 6 months of age (Cowan & Clout 2000; Johnson et al. 2001) and it is probable that selection will favour phenotypic adaptations, such as an increased early growth rate, that result in improved first-year survival and dispersal success. Females remain close to their mothers for a much longer period and settle characteristically in a home range that overlaps or is adjacent to that of their mothers and early growth rate is likely to be less important for female offspring. This result also implies indirect evidence for assumption 3 of the TW model: that sons will benefit more than daughters in lifetime reproductive success by an advantage in condition. Clutton-Brock (1984) points out that when male lifetime reproductive success increases with body size, early growth rates are likely to exert an important effect on breeding success as an adult. Hewison & Gaillard (1999) also suggest that this assumption is probably valid in most species where male mating success increases with body size. Paternity analysis in the Magnetic Island population is currently ongoing, but recent molecular work by Clinchy et al. (2004) indicates that, in Australian populations, larger heavier males do have greater reproductive success. Because we are unable to show conclusive evidence that maternal investment has a differential effect on the reproductive value of sons and daughters, females could still be biasing sex ratios in line with the cost of reproduction hypothesis (Myers 1978; Cockburn et al. 2002), rather than demonstrating TW effects. The cost of reproduction hypothesis proposes that females in poor condition may be unable to produce a given sex because it is more costly and may impact on their future reproductive potential. There are many examples that indicate that rearing a male offspring can have a detrimental effect on subsequent maternal fitness in mammals (i.e. Cervus elaphus: Clutton-Brock et al. 1981; Equus caballus: Monard et al. 1997). However, this study found no effect of offspring sex on maternal condition the following year, suggesting that the cost of reproduction hypothesis does not explain sex ratio bias in this population.
We found no evidence that females that produced two offspring in a single breeding season altered their sex-ratio bias between cohorts. This result contradicts the predictions of the first-cohort advantage hypothesis (FCAH: Wright, Ryser & Kiltie 1995), which proposes that females should increase their fitness by producing more males than females in their first litter. This is proposed because first-cohort males are expected to have an advantage over second-cohort males by being larger in their first year of reproduction (Wright et al. 1995; Ward 2003). Most studies that have found support for the FCAH have been carried out on species that have a short life-span and often reproduce for 1 year only, such as Didelphis virginiana (Wright et al. 1995) and Microtus canicaudus (Wolff, Bond & Krackow 2003). However, Hardy (1997) found no cohort effect in D. marsupialis. We propose that the FCAH may be less applicable to species where males have multi-year opportunities to reproduce. Male possums continue to grow, both in skeletal size and body mass, up to the age of 4 years (J. Isaac, unpublished data) and thus second cohort males may be able to ‘catch-up’, in terms of mass and condition, to some degree at least. However, our data on FCAH suffer from a small sample size as double breeding occurred in only 2 years of our study. Further long-term data will be required in order to understand fully the fitness benefits and costs of double breeding in T. vulpecula.
According to the LRC hypothesis of sex-ratio bias, biases should be related to the potential for resource competition between related females. Although LRC is predicted to affect sex ratio at the level of the population, Cockburn et al. (2002) highlight a number of circumstances where predictions of population-wide sex ratio bias need not be fulfilled in order to meet the requirements of the LRC model. For example, the potential for competition between related individuals is predicted to be highest under conditions of resource scarcity and Johnson et al. (2001) found that the offspring sex ratio in eight populations of T. vulpecula was related to den availability; in populations where den sites were limiting, females showed a male-biased sex ratio. This result is supported by our data as multiparous mothers showed a male bias in the mainland sample, where den availability is a limiting factor. On Magnetic Island, population density is high in comparison to mainland populations (Isaac & Johnson 2003), and individuals given food supplements have been shown to increase in mass considerably (Isaac et al. 2004). It is therefore likely that resources, particularly food, will limit individuals in this population.
The potential for LRC effects is also predicted to be high early in the reproductive life of females and this was supported in our results, as young primiparous mothers were found to bias their offspring sex ratio toward males, the dispersing sex. Cockburn, Scott & Dickman (1985) found similar results in populations of Antechinus spp. In iteroparous species of Antechinus (i.e. A. swainsonii), females were likely to survive to breed twice and mothers were found to bias their first litter toward sons. However, in more semelparous populations (i.e. A. agilis) females rarely breed more than once and produced female-biased litters. A consequence of this may be that after producing a male bias early in life, females should show a shift or reversal in their offspring sex ratio (Ward 2003). This prediction was also upheld, at least for the Magnetic Island population, as the sex ratio of multiparous mothers was female biased. Similar results have been shown in Dasyurus hallucatus (northern quoll: Oakwood 2000) and iteroparous populations of A. swainsonii (Cockburn et al. 1985). However, females of both these species breed only twice in their lifetime and the adaptive significance behind this shift in T. vulpecula is likely to have a more complex explanation. Primiparous females with intrinsically high residual reproductive value may essentially take a bet-hedging strategy and produce a low-quality male without incurring the costs associated with producing a daughter early on in life through LRC. However, the prediction of the LRC model that older females close to the end of their breeding life should overproduce daughters was not met as the female bias found in multiparous females was independent of maternal age.
In conclusion, our data support the prediction that both TW and LRC effects can operate simultaneously within a population to produce differing patterns of bias among mothers according to variation in maternal condition and reproductive status. Further long-term monitoring of this population and ongoing paternity analysis will be imperative in elucidating the influence of maternal condition on the LRS of male offspring.
We would like to thank Brett Goodman, Marco Festa-Bianchet, Fred Ford and Matt Symonds for comments and discussion. Steve Delean and Will Edwards provided help with statistical analysis. JLI would also like to thank the numerous people who helped collect field data on Magnetic Island, particularly Tanya Cornish, Euan Ritchie, Bryan Leighton and Nick Mann. The Australian Research Council and James Cook University provided financial support for this project. Queensland Parks and Wildlife Service provided permits.