2200 Variation in offspring sex ratios in the northern water snake (Nerodia sipedon) Patrick J. Weatherhead, Gregory P. Brown, Melanie R. Prosser, and Kelley J. Kissner Abstract: We used data from 88 litters of northern water snakes (Nerodia sipedon) to test predictions about how mothers would adaptively vary the sex ratios of their offspring. Larger mothers produced significantly more daughters (r 2 = 0.04, P = 0.05), and mothers producing larger offspring produced significantly more daughters (r 2 = 0.06, P = 0.02). Because neonate size did not vary with maternal size, these sex-ratio patterns were independent of each other. These patterns were more pronounced for wild females than for females maintained in captivity while gravid, but rearing conditions did not have a significant effect on sex ratio. Also, because sex ratios were similar between captive and free-living females despite captive females giving birth 16 days earlier, on average, and because sex ratios did not vary with birth date within the two groups of females, gestation appeared not to affect sex ratio. If females vary sex ratios adaptively, only the relationship between sex ratio and neonate size was consistent with our predictions. Limited evidence from other snake species also indicates variation in neonatal sex ratios that is nonrandom but not necessarily adaptive. A better understanding of these patterns will require information on the factors that affect the fitness of male and female neonates differently. An unexpected sex-ratio pattern that we found was that 14 of 19 stillborn young were male. We speculate that this pattern could be a result of male embryonic sensitivity to temperature. Thus, the need for gravid females to maintain a high body temperature so that their young are born with enough time to find hibernation sites may conflict with the need for embryos to develop at a safe temperature. Résumé : Nous avons utilisé des données sur 88 portées de Couleuvres d’eau (Nerodia sipedon) pour vérifier les prédictions sur la façon dont les mères doivent contrôler le rapport mâles:femelles de leur progéniture de façon adaptative. Les mères de plus grande taille produisent significativement plus de femelles (r2 = 0,04, P = 0,05) et les mères qui produisent le plus de rejetons donnent naissance à significativement plus de femelles (r2 = 0,06, P = 0,02). Comme la taille des couleuvres néonates ne dépend pas de la taille de la mère, les patterns de ces rapports mâles : femelles sont indépendants les uns des autres. Ces patterns sont plus accentués chez les femelles sauvages que chez les femelles gardées en captivité pendant leur grossesse, mais les conditions d’élevage n’affectent pas significativement le rapport mâles : femelles. De plus, comme les rapports sont similaires chez les femelles gardées en captivité et chez les femelles libres, même si les femelles en captivité mettent bas 16 jours plus tôt en moyenne, et comme les rapports ne varient pas en fonction de la date de naissance chez l’un ou l’autre de groupe de femelles, la gestation ne semble pas affecter le rapport mâles:femelles. Si les femelles contrôlent les rapports mâles:femelles de leur progéniture de façon adaptative, seule la relation entre le rapport mâles:femelles et la taille des néonates s’accorde avec nos prédictions. De rares données sur d’autres espèces de couleuvres indiquent également une variation non aléatoire du rapport mâles : femelles chez les néonates, mais cette variation n’est pas nécessairement adaptative. Une meilleure compréhension des phénomènes suppose l’acquisition d’informations sur les facteurs qui affectent différemment le fitness des mâles et des femelles. Une observation surprenante : 14 sur 19 néonates morts à la naissance étaient des mâles. Nous soupçonnons que ce phénomène peut être attribuable à une sensibilité plus grande des embryons mâles à la température. Il peut donc se produire un conflit entre, d’une part, la nécessité pour les femelles gravides de maintenir leur température élevée pour mettre bas suffisamment tôt pour que leurs rejetons puissent trouver des sites d’hibernation et, d’autre part, les besoins thermiques des embryons qui doivent se développer à une température adéquate. [Traduit par la Rédaction] Weatherhead et al. Introduction The ratio of sons to daughters among the offspring produced by a female may be a simple consequence of random segregation of sex chromosomes. More interesting from an 2206 evolutionary perspective, however, is the possibility that females are able to manipulate the sex ratio of their brood to enhance their own fitness (Charnov 1982). Various ecological circumstances that favor such adaptive sex allocation have been proposed, and evidence continues to be found Received January 30, 1998. Accepted July 20, 1998. P.J. Weatherhead,1 G.P. Brown, and K.J. Kissner. Department of Biology, Carleton University, 1125 Colonel By Drive, Ottawa, ON K1S 5B6, Canada. M.R. Prosser. Department of Biology, McMaster University, 1280 Main Street West, Hamilton, ON L8S 4K1, Canada. 1 Author to whom all correspondence should be addressed (e-mail: patrick_weatherhead@carleton.ca). Can. J. Zool. 76: 2200–2206 (1998) © 1998 NRC Canada Weatherhead et al. showing that sex ratios vary as predicted, at least among certain taxa (Godfray and Werren 1996). Reptiles are poorly represented in sex-ratio studies, with the exception of research on the specific issue of temperature-dependent sex determination, particularly in turtles (Bull 1983). Here we use data on offspring sex ratios in a viviparous squamate, the northern water snake (Nerodia sipedon), to test several predictions from sex-allocation theory. Trivers and Willard (1973) proposed a sex-allocation hypothesis that could potentially apply across a broad array of animal taxa. They argued that when the fitness of sons and daughters is affected by maternal factors (e.g., size, condition, social status), a female should produce an excess of the sex that will benefit most (or suffer least) from her current circumstances. In snakes, ways in which a female might influence the fitness of her offspring are limited because parental care is absent. Thus, maternal effects on neonate attributes seem the most likely candidates to be involved in maternal strategies of adaptive sex allocation. For example, if the fitness of neonates of one sex is affected more than the fitness of the other sex by the timing of birth, or by their size or condition at birth, then snake sex ratios might be expected to vary with these attributes. Little is known about the factors that affect neonate fitness in snakes (e.g., Parker and Plummer 1987), let alone how these factors might differ between the sexes. However, because of the widespread occurrence of sexual size dimorphism in snakes (Shine 1978), there is some possibility for adaptive sex allocation. In sexually dimorphic species, large size is presumably more advantageous for one sex than the other. Thus, any attribute that allows neonates to attain adult size more quickly might favor overproduction of the sex that grows larger. Larger size or better condition at birth, or an earlier birth date in species living in seasonal environments, all seem likely to give individuals a head start in growing to adult size. Two studies of snakes have provided evidence that neonatal sex ratios vary with maternal size. Lemen and Voris (1981) found that in the marine snake Enhydrina schistosa, larger females produced larger offspring and proportionately more female offspring. Because sexual size dimorphism is female-biased in E. schistosa, this pattern is consistent with females adaptively varying sex ratios. Dunlap and Lang (1990) found that larger female garter snakes (Thamnophis sirtalis) disproportionately produced sons. Dunlap and Lang (1990) did not measure neonates. However, given that offspring size has been shown to increase with maternal size in T. sirtalis (Gregory 1977), it is likely that larger females produced larger neonates. Given that females grow larger than males in T. sirtalis, according to the reasoning above, larger females should have produced more daughters rather than more sons. Thus, the pattern of sex-ratio variation reported by Dunlap and Lang (1990) was opposite to that predicted. One study of snakes has provided evidence that offspring sex ratios might vary adaptively with gestation period. Dunlap and Lang (1990) reanalyzed Osgood’s (1978) data on neonatal sex ratios in water snakes (Nerodia fasciata) relative to the temperature at which the gravid females had been maintained. Females maintained at higher temperatures produced more daughters relative to females maintained at lower temperatures. Because higher gestation temperatures should accelerate development and parturition (e.g., Brown 2201 and Weatherhead 1997), broods born earlier would have been disproportionately female. Because females of this species grow larger than males, females may benefit more than males from a head start in growth. An alternative, nonadaptive explanation for this sex-ratio pattern in water snakes is that male embryos may be less viable than female embryos when they develop at higher temperatures. Burger and Zappalorti (1988) found differential viability of male and female embryos with incubation temperature in pine snakes (Pituophis melanoleucus), although male embryos were more viable at higher temperatures. Collectively, the available evidence from snakes indicates that although some intriguing patterns of sex-ratio variation occur, it is far from established that these patterns reflect adaptive allocation of sex ratios by females. Our objective in this study was to assess offspring sex ratios in N. sipedon relative to aspects of the mothers, the offspring themselves, and gestation conditions. In our study population, female N. sipedon grow 20% longer and 50% heavier than males (Weatherhead et al. 1995). Thus, we predicted that if females modify sex ratios adaptively, neonates that would develop at higher temperatures, would be born earlier, and would be larger at birth should be disproportionately female. In addition to relying on natural variation in these variables, we also maintained some gravid females in captivity under optimal conditions of food and temperature to increase the variance in the offspring traits of interest, particularly the date of birth. Because neonate size does not vary with maternal size in N. sipedon (G.P. Brown and P.J. Weatherhead, unpublished data), we also predicted that, contrary to the patterns found in E. schistosa (Lemen and Voris 1981) and T. sirtalis (Dunlap and Lang (1990), offspring sex ratios would not vary with maternal size. Materials and methods We conducted this study at the Queen’s University Biological Station in southeastern Ontario, Canada (45°37′N, 76°13′W), during 5 years (1990 and 1994–1997). We had two categories of adult females (>55 cm snout–vent length (SVL); Weatherhead et al. 1995) as follows. (1) Captive females: females that were caught in late April or early May of each year (except 1990) immediately after emergence from hibernation and prior to mating (early May to early June). These females were housed with adult males during the mating season, following which the males were released. Captive females were kept in captivity until they gave birth or we were certain that they were not gravid, after which they were returned to the wild. (2) Free-living females: females that were not captured until early August just prior to parturition and were returned to the wild following parturition. When in captivity, all snakes were held in heated rooms in tanks lined with artificial grass carpeting or wood shavings. Each tank contained a water dish large enough to allow the snake to submerge and most had heating rocks or coiled heating cables. Room temperature varied with ambient conditions but was maintained above 22°C. Snakes were fed minnows (live or previously frozen) ad libitum one to three times per week. Additional methodological details are available in Barry et al. (1992) for 1990 and Brown and Weatherhead (1997) for 1994–1997. The snakes used in this study were cared for in accordance with the principles and guidelines of the Canadian Council on Animal Care. We measured the SVL of each adult female within 12 h of parturition and each neonate within 48 h of parturition. We measured the SVL of adults by running a string along the body, following all the contours, and then measuring the string. We used the mean of © 1998 NRC Canada 2202 Can. J. Zool. Vol. 76, 1998 Table 1. Size, sex ratios, and birth dates (Julian date) of northern water snake (Nerodia sipedon) litters from free-living (FL) and captive-reared (CR) females. No. of litters Litter size Mean Range Percent male Mean Range Birth date Mean Range 1990 1994 (FL) FL 1995 CR FL 3 8 1996 CR 9 FL 19 1997 CR 7 FL 14 12 9 23.1 8–37 19.0 15–25 19.0 14–28 21.1 17–29 20.7 13–32 17.0 10–25 18.9 10–24 20.8 12–36 47.5 37–66 50.6 14–75 55.6 43–67 53.5 35–72 52.3 26–75 44.1 13–71 50.1 33–68 253.6 245–262 244.0 236–260 240.0 235–246 240.0 236–247 219.0 214–229 245.5 234–256 229.6 222–237 Overall CR 7 FL CR 62 26 20.7 14–31 19.9 8–37 20.3 10–32 47.2 31–63 44.5 33–56 47.8 13–75 50.0 26–75 245.4 239–252 240.3 231–247 246.3 234–262 230.0 214–247 Note: No females were maintained in captivity in 1990. two replicate measurements for each snake. If the second measurement differed from the first by more than 1 cm, we remeasured the snake and, among those measurements that differed by less than 1 cm, we used the two that were most similar. We measured neonate SVL by placing the neonates directly on a ruler. We recorded the date of birth as Julian date. All neonates (including stillborns) were sexed by probing for the presence of hemipenes. We confirmed the accuracy of this method by everting the hemipenes (or attempting to do so for females) for a subsample of individuals and by measuring tail length. Individuals confirmed to be males had tails that were approximately 5 mm longer than those of females. This combination of approaches resulted in highly accurate identification of sex. Unless stated otherwise, we do not differentiate stillborn young from live-born young in our analyses because stillbirths accounted for only 1% of all sexed neonates. We excluded from analysis one female that gave birth to only four live young and 16 undeveloped follicles. Slight variations in sample size reflect missing values for several litters. We calculated sex ratio as the proportion of males within a litter. Sex ratios were arcsine transformed when sex ratio was used as a variable in a linear model. To test our hypotheses of how sex ratio should vary with maternal or neonate traits, we used analysis of covariance (ANCOVA), with litter sex ratio as the dependent variable, the maternal or neonate trait as a continuous independent variable, and year as a categorical independent variable. Models were initially run with interactions between the independent variables included. If the interaction term did not contribute to the model with a significance level of P < 0.10, the interaction term was removed and the reduced model run. We only provide details of interaction terms that were retained in analyses. We analyzed data from captive and free-living females separately and combined data where appropriate. These analyses were done using JMP (version 3.1) statistical software (SAS Institute Inc. 1994). Results We obtained sex-ratio data from a total of 88 litters (Table 1) composed of 883 female and 821 male neonates. The overall sex ratio of 0.48 males was not significantly different from equality (t = –1.16, P = 0.25). Maintaining gravid females in captivity resulted in captive females giving birth, on average, 16 days earlier than free-living females (t = 8.35, P < 0.0001). We used two approaches to test the hypothesis that the sex ratio would become female biased when the length of gestation decreased. First, using the shortened gestation among captive females, we compared mean sex ratios of litters from captive and free-living females. Captive females actually tended to produce litters that were relatively more malebiased than those of free-living females (0.50 vs. 0.48), but the difference was not significant (t = –0.78, P = 0.44). A second approach was to examine how sex ratio varied with birth date separately within captive and free-living samples. Among captive females the sex ratio became more malebiased with birth date in 1996 (r = 0.78, N = 7, P = 0.04) but not in 1995 (r = –0.32, N = 7, p = 0.40) or 1997 (r = 0.07, P = 0.89, N = 7), the only years with captive samples adequate for analyzing separately. A reduced ANCOVA model for captive litters (N = 26) revealed no effect of year (F[3] = 0.96, P = 0.43) and no evidence that sex ratio varied with birth date (F[1] = 0.53, P = 0.47). Among free-living females, sex ratios tended to become more male-biased with later birth dates in 1990, 1995, and 1997 and more femalebiased in 1994 and 1996, but the relationship was only significant in 1995 (r = 0.76, P = 0.03, N = 8). The reduced ANCOVA model for free-living females (N = 61) revealed no overall effect of year (F[4] = 0.92, P = 0.46) and no overall relationship between sex ratio and birth date (F[1] = 0.29, P = 0.60). Overall, there was no support for the prediction that mothers would vary the sex ratios of their litters in response to the duration of gestation. Contrary to our prediction that litters with smaller neonates would be disproportionately male, the proportion of males in captive litters did not vary significantly with mean neonate size, either when the data were analyzed separately by year (1995: r = 0.04, P = 0.93, N = 7; 1996: r = 0.30, P = 0.56, N = 7; 1997: r = –0.55, P = 0.20, N = 7) or when they were analysed together (reduced ANCOVA model (N = 25): F[3] = 0.79, P = 0.51, F[1] = 0.03, P = 0.86, for year and neonate SVL, respectively). However, sex ratio varied with mean neonate size as predicted for litters from free-living females. The proportion of male neonates tended to decrease as mean neonate SVL increased in free-living litters each year (r = –0.17 to –0.55), and the relationship was significant in 1990 (r = –0.55, P = 0.04, N = 14). In the reduced ANCOVA model (N = 62), year did not have a significant effect (F[4] = 0.67, P = 0.62), but mean neonate SVL did have a significant effect on sex ratio (F[1] = 5.33, P = 0.02). The negative relationship between sex ratio and neonate size © 1998 NRC Canada Weatherhead et al. 2203 Fig. 1. Variation in neonate sex ratios with neonate snout–vent length (SVL) in northern water snakes (Nerodia sipedon). Fig. 2. Variation in neonate sex ratios with maternal snout–vent length (SVL) in Nerodia sipedon. remained significant when data from free-living snakes were pooled across years in a simple linear regression (r = –0.31, P = 0.01). To assess whether the relationship between sex ratio and neonate SVL differed between free-living and captive females, we used ANCOVA with sex ratio as the response variable and neonate SVL and location (captive or free-living) as predictor variables. The interaction between location and neonate SVL was not significant (F[1,83] = 1.59, P = 0.21), indicating that rearing conditions did not influence the relationship between neonate size and sex ratio. The proportion of males in litters declined significantly as neonate SVL increased after data for captive and free-living snakes were pooled in a linear regression (r = –0.25, P = 0.02) (Fig. 1). Sex ratios from captive litters supported the prediction that sex ratio would not vary with maternal size, both when data were analyzed separately by year (1995: r = –0.13, P = 0.75, N = 9; 1996: r = –0.54, P = 0.21, N = 7; 1997: r = – 0.21, P = 0.65, N = 7) and when they were analysed together (reduced ANCOVA model (N = 26): F[3,1] = 0.78 and 1.89, P = 0.52 and 0.18 for year and maternal SVL, respectively). However, data from free-living females indicated that sex ratio varied with maternal size. Sex ratios were consistently negatively correlated with maternal SVL each year (r = – 0.24 to –0.63), with a significant relationship in 1990 (r = – 0.63, P = 0.02, N = 14). The reduced ANCOVA model for free-living snakes showed that year did not have a significant effect overall (F[4] = 1.72, P = 0.16) but that the effect of maternal size was significant (F[1] = 5.67, P = 0.02). Although the effect of year was not significant, when the data for free-living females were pooled across years in a simple linear regression, the relationship between sex ratio and maternal size was no longer significant (r = –0.20, P = 0.12). The direction of the relationship between sex ratio and maternal size was consistent among captive and free-living snakes. In an ANCOVA with sex ratio as the response variable and maternal SVL and location (captive or free-living) as predictor variables, the interaction between location and maternal SVL was not significant (F[1,84] = 0.05, P = 0.82). Pooling the data from captive and free-living females revealed that the proportion of males in litters declined significantly with maternal size (r = –0.21, P = 0.05) (Fig. 2). There are at least two ways in which the relationships between sex ratio and maternal size and neonate size might be misleading. First, if maternal size and neonate size are highly correlated, both variables could be explaining the same variation in sex ratio. However, maternal size and neonate size were not highly correlated (r = 0.13, N = 87, P = 0.24). Second, the negative relationship between sex ratio and mean neonate size could be spurious if neonates are sexually dimorphic, with males being smaller (litters with more males would tend to have a smaller mean neonate size just because of sexual dimorphism). Neonate water snakes are sexually dimorphic, and males are smaller than females within litters (mean SVL = 16.68 and 16.76 cm for males and females, respectively; paired t = 3.07, df = 86, P = 0.003). However, because of the very strong correlation between mean male and female neonate sizes within litters (r = 0.96, N = 87, P < 0.0001), the relationship between neonate size and sex ratio remains virtually identical whether one uses the mean size of males, females, or all neonates within litters as the dependent variable in that analysis. Deviations of sex ratios from random that we have reported could not have been a consequence of sex-biased embryo mortality varying with maternal size or neonate size because there were very few stillborn young and we sexed those individuals and included them in the analyses. It is possible that females biased the sex ratio of their litters by means of selective mortality and resorption of embryos of the unwanted sex. If so, the sex ratio of litters smaller than expected on the basis of maternal size should have been the most biased. We assessed this possibility by examining sex ratios relative to the residuals from a regression of litter size on maternal size and found no relationship (r = 0.13, P = 0.24). Although sex-biased mortality of embryos does not appear to be responsible for the patterns we observed in litter sex ratios, we did note an interesting pattern in the sex ratios of stillborn young. Of 19 stillborn young, 14 were male (χ2 = 4.26, P < 0.05). These 19 young were scattered across 14 different litters, including those from 4 of the 5 years and from both captive and free-living snakes, so this deviation from random is not a consequence of a few atypical females or of year or rearing effects. Comparisons of © 1998 NRC Canada 2204 litters with and without stillborn young for all the variables used here revealed no differences that approached significance. Discussion The overall sex ratios of neonatal northern water snakes in this study did not deviate significantly from equality, as would be expected if sex is determined by random segregation of the sex chromosomes. However, contrary to expectation if sex ratios are determined randomly, sex ratios varied with both maternal size and neonate size. Larger mothers produced fewer sons, and mothers that produced larger neonates also produced fewer sons. Although both patterns were more apparent among litters from free-living females than among those from captive females, analyses confirmed that the patterns were not a consequence of rearing conditions. Furthermore, the declines in the proportion of sons produced with increasing maternal size and neonate size were both significant when data from captive and free-living females were combined. We found no evidence that sex ratio varied consistently with the length of the gestation period. These results indicate that sex ratios vary nonrandomly among litters. Bias toward more daughters when neonates were larger is consistent with our prediction of how females should allocate offspring sex adaptively (Trivers and Willard 1973). However, the significant relationship between sex ratio and maternal size and the lack of a relationship between sex ratio and gestation period were both contrary to what we predicted. Interpretation of our results depends to some extent on what we assume regarding the mechanism that biases offspring sex ratios in water snakes. The available evidence indicates that sex is determined chromosomally in snakes and that females are the heterogametic sex (Shine and Bull 1977). Because this evidence is based on relatively few species of snakes, it is possible that in some species not yet examined (which include N. sipedon), gestation or incubation temperature could determine offspring sex, particularly given that this phenomenon occurs in other reptiles (Bull 1983). However, we found no difference in offspring sex ratios between litters from captive and free-living females, even though captive females gave birth substantially earlier because of higher gestation temperatures. We also found no effect of birth date on sex ratio, and at least some of the variation in birth date seems likely to have been a consequence of individual differences in gestation temperature. Thus, it seems most parsimonious to assume that offspring sex in N. sipedon is determined chromosomally. This being the case, and given the evidence that we have presented that litter sex ratios did not appear to be biased via selective embryo mortality, the most likely mechanism for biasing offspring sex ratios is nonrandom segregation of the sex chromosomes during meiosis. We first consider those results that show no evidence of nonrandom variation in sex ratio, adaptive or otherwise. We predicted that females would vary offspring sex ratio according to the length of gestation, based on the assumption that an earlier birth would provide more benefit to female offspring by giving them a head start in achieving a larger size. We found no support for this hypothesis, so it is possible Can. J. Zool. Vol. 76, 1998 that the timing of birth does not differentially affect the fitness of sons and daughters. We also assumed that variation in birth date was a function of variation in gestation length, as opposed to a function of variation in the date of ovulation. We do not know the extent to which each factor affected variation in birth date, although the earlier birth of captive litters suggests that variation in gestation length probably contributed substantially to variation in birth date. Nevertheless, the prediction is unaffected. Whatever the reason for the variation in birth date, we would still have predicted that females would benefit more than males from being born early. An explanation for the negative result may be that females are unable to predict accurately when their young will be born. Given chromosomal sex determination, with females being the heterogametic sex, offspring sex must be determined prior to fertilization. Thus, for females to adjust sex ratios adaptively in response to variation in birth date, they must be able to predict how long gestation will last before it begins. Our captive females had no way of predicting that thermally benign conditions would persist throughout gestation. In the wild, the mean date of birth varied by 2 weeks over the 5 years of this study, so natural gestation conditions are likely to be difficult to predict. Reanalysis of Osgood’s (1978) sex ratio data for N. fasciata by Dunlap and Lang (1990) revealed that gravid females maintained at higher temperatures produced more daughters, while sons were more abundant at lower gestation temperatures. This pattern would have resulted in femalebiased litters being born earlier, as we had predicted for N. sipedon in our study. However, exactly the same problem that applied to our data, i.e., females must be able to predict thermal conditions during gestation, also applies to Osgood’s (1978) data. Given that females cannot anticipate the nature of captive-rearing conditions, alternative explanations for these data seem to be required. As mentioned earlier, temperature-dependent sex-biased mortality of embryos (Burger and Zappalorti 1988) is one possibility. This would require not only that male embryos were more sensitive to high gestation temperatures, but also that female embryos were more sensitive to low gestation temperatures. If this explanation is correct, then the pattern of sex-ratio variation in N. fasciata clearly is not adaptive, given the high cost of achieving skewed sex ratios. We had predicted that sex ratios would vary with neonate size, based on the assumption that females would benefit more than males from being born larger because they have to grow larger. Despite our evidence supporting this prediction, it would be premature to conclude that our hypothesis is correct. In the two other studies presenting similar data, sex ratio appeared to vary with neonate size, as predicted in one case (Lemen and Voris 1981) and contrary to prediction in the other (Dunlap and Lang 1990). Furthermore, in both earlier studies the relationships between sex ratio and neonate size were inferred from two other relationships: between sex ratio and maternal size and between maternal size and neonate size. In our study, however, sex ratio varied independently with both maternal size and neonate size. Thus, the pattern of sex-ratio variation may actually be different in each of the three species for which data are currently available. To determine if a single hypothesis explains this varia© 1998 NRC Canada Weatherhead et al. tion among species, we need to determine the validity of our assumptions. Specifically, we need to know what factors affect neonate survival and how these factors interact with the size and sex of neonates. Accumulation of additional data on neonatal sex ratio variation in snakes would also be helpful. We had predicted no relationship between sex ratio and maternal size because neonate size does not vary with maternal size. However, we found that larger mothers produced more daughters. One mechanism that could account for this pattern is local mate competition, whereby females produce fewer sons when low natal dispersal results in brothers competing with each other for mates (Hamilton 1967). Local mate competition appears to explain a female bias in the neonatal sex ratio in adders (Vipera berus) (Madsen and Shine 1992). Our study populations are relatively small (about 100 individuals; G.P. Brown and P.J. Weatherhead, unpublished data), and an analysis of variation in microsatellite DNA loci provides some evidence that these local populations are genetically isolated (Prosser et al. 1999). Furthermore, males compete extensively for access to females (Barry et al. 1992). Thus, the prerequisites for local mate competition among brothers might be satisfied by these populations. Arguing against this explanation, however, is the fact that there was no overall female bias in neonate sex ratios. Larger females do produce larger litters (Weatherhead et al. 1995), so their sons might be more likely to compete with each other, but this leaves unresolved the problem of why smaller females produce an excess of sons. An alternative explanation is that maternal size affects offspring quality in some way not reflected by offspring size. Recent models suggest that if maternal quality affects offspring reproductive value, then higher quality females should produce an excess of daughters and lower quality females an excess of sons (Leimar 1996). We currently have no evidence with which to address this hypothesis, but it clearly warrants further exploration. Finally, the male bias among stillborn neonates was unexpected but potentially important. Because stillborn young are detrimental to a female’s fitness, this phenomenon is obviously outside the realm of adaptive sex allocation. However, it may be relevant to another aspect of reproductive strategies in female snakes. Burger and Zappalorti (1988) demonstrated that male and female pine snake embryos differed in how temperature affected their viability. All female water snakes in our study were allowed to control their body temperatures behaviorally during gestation. Elsewhere, we show that gravid female water snakes maintain higher body temperatures and thermoregulate more precisely than nongravid females (G.P. Brown and P.J. Weatherhead, unpublished data). If male embryos are more sensitive than female embryos to high temperatures, it is possible that the male bias among stillborns is a consequence of gravid females occasionally having exceeded safe temperatures while thermoregulating. Gravid female water snakes in our study population may be forced by the brevity of the activity season to maintain body temperatures that are close to lethal for their embryos just to ensure that their young are born with adequate time to find hibernation sites before winter. Alternatively, gravid females may have to risk some embryo mortality to achieve other life-history advantages (e.g., growth) if optimal temperatures for mothers differ from those for embryos (Beuchat and Ellner 1987). Collectively, our results 2205 point to the need for more information regarding the factors that affect the fitness of newborn snakes and the extent to which those factors influence female reproductive strategies. Acknowledgements We thank Frances Barry, Perry Comm, and Ian Robertson for assistance in capturing snakes and caring for them in captivity, and Kevin Dufour for statistical advice. The Queen’s University Biological Station provided facilities and logistical support. This study was funded by a Natural Sciences and Engineering Research Council of Canada grant to P.J.W. References Barry, F.E., Weatherhead, P.J., and Philipp, D.P. 1992. Multiple paternity in a wild population of northern water snakes (Nerodia sipedon). Behav. Ecol. Sociobiol. 30: 193–199. Beuchat, C.A., and Ellner, S. 1987. A quantitative test of life history theory: thermoregulation by a viviparous lizard. Ecol. Monogr. 57: 45–60. Brown, G.P., and Weatherhead, P.J. 1997. Effects of reproduction on survival and growth of female northern water snakes, Nerodia sipedon. Can. J. Zool. 75: 424–432. Bull, J.J. 1983. Evolution of sex determining mechanisms. Benjamin Cummings, Menlo Park, Calif. Burger, J., and Zappalorti, R.T. 1988. Effects of incubation temperature on sex ratios in pine snakes: differential vulnerability of males and females. Am. Nat. 132: 492–505. Charnov, E.L. 1982. The theory of sex allocation. Princeton University Press, Princeton, N.J. Dunlap, K.D., and Lang, J.W. 1990. Offspring sex ratio varies with maternal size in the common garter snake. Copeia, 1990: 568– 570. Godfray, H.C.J., and Werren, J.H. 1996. Recent developments in sex ratio studies. Trends Ecol. Evol. 11: 59–63. Gregory, P.T. 1977. Life history patterns of the red-sided garter snake, Thamnophis sirtalis parietalis, in an extreme environment, the interlake region of Manitoba. Natl. Mus. Can. Publ. Zool. 13: 1–44. Hamilton, W.D. 1967. Extraordinary sex ratios. Science (Washington, D.C.), 156: 477–488. Leimar, O. 1996. Life-history analysis of the Trivers and Willard sex-ratio problem. Behav. Ecol. 7: 316–325. Lemen, C.A., and Voris, H.K. 1981. A comparison of reproductive strategies among marine snakes. J. Anim. Ecol. 50: 89–102. Madsen, T., and Shine, R. 1992. Sexual competition among brothers may influence offspring sex ratio in snakes. Evolution, 46: 1549–1552. Osgood, D.W. 1978. Effects of temperature on the development of meristic characters in Natrix fasciata. Copeia, 1978: 33–47. Parker, W.S., and Plummer, M.V. 1987. Population ecology. In Snakes: ecology and evolutionary biology. Edited by R.A. Seigel, J.T. Collins, and S.S. Novak. Macmillan Publishing Co., New York. pp. 253–301. Prosser, M.R., Gibbs, H.L., and Weatherhead, P.J. 1999. Microgeographic population structure in the northern water snake, Nerodia sipedon sipedon, detected using microsatellite DNA markers. Mol. Ecol. 8: 329–333. SAS Institute Inc. 1994. JMP statistics and graphics guide. SAS Institute Inc., Cary, N.C. © 1998 NRC Canada 2206 Shine, R. 1978. Sexual size dimorphism and male combat in snakes. Oecologia, 33: 269–277. Shine, R., and Bull, J.J. 1977. Skewed sex ratios in snakes. Copeia, 1977: 228–234. Trivers, R., and Willard, D.E. 1973. Natural selection of parental ability to vary the sex ratio of offspring. Science (Washington, D.C.), 179: 90–91. Can. J. Zool. Vol. 76, 1998 Weatherhead, P.J., Barry, F.E., Brown, G.P., and Forbes, M.R. 1995. Sex ratios, mating behavior and sexual size dimorphism of the northern water snake, Nerodia sipedon. Behav. Ecol. Sociobiol. 36: 301–311. © 1998 NRC Canada