Richard Shine, School of Biological Sciences A08, University of Sydney, Sydney, NSW 2006, Australia. Tel.: +61-2-9351-3772; fax: +61-2-9351-5609; e-mail: firstname.lastname@example.org
The underlying genetic basis of life-history traits in free-ranging animals is critical to the effects of selection on such traits, but logistical constraints mean that such data are rarely available. Our long-term ecological studies on free-ranging oviparous snakes (keelbacks, Tropidonophis mairii (Gray, 1841), Colubridae) on an Australian floodplain provide the first such data for any tropical reptile. All size-corrected reproductive traits (egg mass, clutch size, clutch mass and post-partum maternal mass) were moderately repeatable between pairs of clutches produced by 69 female snakes after intervals of 49–1152 days, perhaps because maternal body condition was similar between clutches. Parent–offspring regression of reproductive traits of 59 pairs of mothers and daughters revealed high heritability for egg mass (h2 = 0.73, SE = 0.24), whereas heritability for the other three traits was low (< 0.37). The estimated heritability of egg mass may be inflated by maternal effects such as differential allocation of yolk steroids to different-sized eggs. High heritability of egg size may be maintained (rather than eroded by stabilizing selection) because selection acts on a trait (hatchling size) that is determined by the interaction between egg size and incubation substrate rather than by egg size alone. Variation in clutch size was mainly because of environmental factors (h2 = 0.04), indicating that one component of the trade-off between egg size and clutch size is under much tighter genetic control than the other. Thus, the phenotypic trade-off between egg size and egg number in keelback snakes occurs because each female snake must allocate a finite amount of energy into eggs of a genetically determined size.
Organisms display an immense variety in life-history traits such as reproductive modes, reproductive frequencies, the relative allocation of resources among offspring size vs. number, and so forth (Charnov, 1982, 1993; Clutton-Brock, 1991). In comparisons among individuals within a single population, such variation bears a direct causal link to microevolutionary fitness, which by definition depends upon the numbers of surviving progeny that are produced during an organism's lifetime. This strong link to fitness has stimulated an extensive literature on the adaptive significance of life-history variation, ranging from mathematical models that predict which trait values will maximize fitness, through to field studies that examine such predictions (Roff, 1992, 2002). However, any adaptationist interpretation of life-history variation necessarily makes a set of assumptions about the traits being studied. Chief among these assumptions is the idea that the observed phenotypic variation (in this case, in life-history traits) has a genetic basis. If this is not true – for example, if all variation in a phenotypic trait is because of direct impacts of the local environment (incubation conditions, food supply, weather, parasites, etc.) – then there will be no cumulative change in frequencies of life-history traits across generations (i.e. no evolution) even if trait frequencies are modified by selection (reflecting differential fitness) within each generation.
The functional link between empirical studies of phenotypic selection on the one hand, and evolutionary models on the other, is the underlying genetical structure of the traits in question. Quantifying the genetic bases and correlations among selectively important traits in wild populations is an onerous but vital task for evolutionary biologists (Lynch, 1999; Réale & Festa-Bianchet, 2000; Réale et al., 2003). The genetic basis of life-history traits is of special interest because such traits are directly linked to organismal fitness. For these reasons, we need empirical studies on natural populations (i.e. of free-ranging rather than captive animals) to examine the critical assumption of a genetic basis to within-population variation in life-history traits. Unfortunately, logistical obstacles have meant that field studies of heritability display a strong taxonomic bias. For example, Mousseau & Roff's (1987) review of heritabilities in natural populations did not include any studies on reptiles. Since that time there has only been one (Sinervo & Doughty, 1996).
Two main approaches have been used to assess the heritability of life-history traits. First, it is relatively easy to measure repeatability: for example, the degree of consistency of reproductive traits (egg size, number, etc.) in successive clutches produced by the same female snakes. If successive clutches are no more similar to each other (in the number and size of eggs, etc.) than they are to the clutches of other female snakes, then it is likely that maternal genotype is not an important contributor to overall life-history variation within the population. The second and more robust (but logistically demanding) method to quantify heritability of reproductive traits is to examine inheritance directly (e.g. by comparing reproductive traits of mothers with those of their adult daughters). If genetic factors play an important role, we would expect daughters to produce similar kinds of clutches (e.g. a few large eggs vs. many small eggs) as produced by their mothers.
Our long-term mark-recapture study on an oviparous snake (the keelback, Tropidonophis mairii, Gray, 1841) in the Australian wet–dry tropics provides an opportunity to utilize both of these methods to investigate the quantitative genetics of female reproductive traits in free-ranging reptiles. Several aspects of the ecology of these snakes make them useful models for studies of quantitative genetics in the wild. First, we are able to collect large numbers of female snakes as they move to nesting sites, so that we can obtain data on their clutches laid within a few days of capture (hence minimizing artefacts because of prolonged captivity). Second, we can incubate the eggs under standardized conditions, thereby reducing environmental sources of variance in phenotypic traits of hatchlings (e.g. nest-site selection: Rhen & Lang, 1995; Deeming, 2004). Third, hatchlings that are individually marked and released in our study area usually reach sexual maturity in less than 1 year (Brown & Shine, 2002) and they in turn can be captured while gravid and their clutches collected. Thus, within a relatively short span of time it is possible to acquire data on the clutches of female snakes and their daughters. Like most oviparous snakes, keelbacks lack parental care and a mother's investment in each egg is completed at ovulation (Somma, 2003). Hence, there is no prolonged period of close contact between the mother and her developing offspring (such as gestation in viviparous animals) during which maternal behaviour or physiology could affect offspring phenotypes. Thus, keelbacks are not subject to many of the potential sources of maternal effects that are present in studies of other taxa.
Materials and methods
Study site and species
The study took place at the Fogg Dam Conservation Reserve (12°34′S, 131°18′E), 60 km SE of Darwin in the wet–dry tropics of Australia's Northern Territory. Fogg Dam is a 200-ha artificial impoundment formed by a 1300-m-long earthen wall approximately 3 m high. The area north of the dam wall is an extensive floodplain of the Adelaide River. The study area has been described and illustrated previously (Madsen & Shine, 1996; Brown & Shine, 2002). Keelbacks (T. mairii) are medium-sized (female snakes grow to 80 cm snout-vent length, SVL) nonvenomous natricine snakes widely distributed through coastal and near-coastal areas of northern and north-eastern Australia (Cogger, 1996). Keelbacks are oviparous and in our study area nesting occurs throughout the drier months of the year (April–November). Some female snakes produce two or more clutches during this period (Brown & Shine, 2002).
Methods of data collection
We patrolled the dam wall by car and on foot looking for snakes during 1804 nights between May 1998 and December 2003. Keelbacks were captured by hand and returned to a field laboratory where they were marked, weighed and measured. If shelled eggs were evident by palpation of a female snake's abdomen, she was retained in captivity. Gravid female snakes were housed indoors in plastic cages containing a water dish and a container of moist vermiculite for oviposition. On average, gravid female snakes oviposited 6.5 days (SD = 3.4) after capture. After oviposition, female snakes were reweighed and then released at their capture location.
Eggs were measured and weighed within 12 h of being laid. Each clutch was placed in a separate plastic bag containing damp vermiculite and incubated in an insulated (Styrofoam) container at room temperature (average 24.0 °C, SD = 3.0, range 16.0–32.0 °C). Most eggs were incubated on a mixture of 1:1 vermiculite to water by mass, but some were incubated on vermiculite-water mixtures ranging in water content from 10 % to 600 % as part of experiments on the effects of incubation moisture on hatchling phenotypes (Brown & Shine, 2004). Within 24 h of hatching, neonates were sexed, measured (SVL, head length, tail length and mass) and individually marked. They were then released at their mother's capture location.
The suite of reproductive traits we investigated comprised: (1) mean egg mass; (2) clutch size; (3) total clutch mass; and (4) maternal post-partum body condition (i.e. mass relative to snout-vent length). Because snakes continue to grow after maturation, a population typically contains adult individuals that span a wide range of body sizes (Andrews, 1982). A female snake's body size strongly affects her reproductive output: larger female snakes typically produce more (and sometimes, larger) offspring (Seigel & Ford, 1987). Thus, analyses of heritability need to be based on size-corrected measures; otherwise, the degree of similarity would depend upon the respective body sizes of the two female snakes being compared. To correct for body size effects, we calculated residual scores from linear regressions of ln-transformed clutch traits on ln-transformed maternal SVL. Residual scores from the regressions of post-partum mass on maternal SVL were also calculated, to provide estimates of post-partum body condition. The residual scores from these regressions were then used in the calculation of repeatabilities and heritabilities.
Measuring repeatability is often suggested as an initial step for quantitative genetics studies. Significant repeatability in a trait within individuals suggests that the trait may have a genetic basis, and values of repeatability are often thought to estimate an upper limit to heritability (but, see Dohm, 2002). Over the study period, we captured a total of 69 individual female snakes that were gravid at more than one capture. Thus, we could calculate within-individual repeatability scores for the clutch and maternal characteristics enumerated above, using intraclass correlations (Lynch & Walsh, 1998).
A total of 59 female neonatal keelbacks that hatched in the laboratory were later recaptured as gravid adults. Thus, we could compare the clutch characteristics of these 59 daughters to those of their mothers using parent–offspring regressions [a single outlying data point for clutch size where a mother that produced a small clutch had a daughter that produced a large clutch (see Fig. 1b) was removed from analyses]. In parent–offspring regression analyses, it is important for variances of traits to be equivalent between parental and offspring generations (Lynch & Walsh, 1998). Thus, we standardized size-corrected traits (such that mean = 0, variance = 1) before regression analyses were carried out. Standardized, size-corrected values of each reproductive trait for each daughter were regressed on those of her mother using ordinary least squares. The slope of the linear regression equation was multiplied by two to yield narrow-sense heritability (h2), and the standard error of the slope was multiplied by two to provide the standard error of h2 (Lynch & Walsh, 1998). h2 measures the proportion of phenotypic variance explained by additive genetic variance; low h2 values indicate a high degree of environmental influence on phenotypes whereas a high h2 indicates a high degree of additive genetic influence. Quantitative genetic parameters were estimated using a bootstrap approach to estimate (with replacement) heritability and genetic correlations and their respective standard errors (program h2boot: Phillips, 1998). We report estimates based on 10 000 bootstrap replicate samples.
The mean duration between successive clutches by the same female snake was 336 days (SE = 19.6 days, range = 49–1152). Both maternal body size and clutch size increased significantly from the first to the second clutch (respective means 62.0 vs. 67.3 cm SVL, paired t68 = 11.51, P < 0.0001; 9.9 vs. 12.3 eggs, paired t68 =6.95, P < 0.0001), but mean egg mass remained relatively constant (2.84 vs. 2.80 g; paired t68 = 0.70, P =0.48). Most reproductive traits were relatively consistent in comparisons between successive clutches produced by the same individual female snakes (Table 1). Repeatabilities of egg mass and clutch size were similar (0.26 vs. 0.28). However, post-partum body condition was even more highly repeatable, suggesting that any environmental effects related to maternal body condition would have tended to generate high repeatability in clutch traits. In keeping with this possibility, post-partum maternal condition after the first clutch was highly correlated with pre-partum condition at the next clutch (N = 69, r = 0.52, P < 0.0001). In other words, the energetic circumstances of a female snake apparently did not vary much between her two successive clutches.
Table 1. Repeatability of reproductive traits from 69 free-ranging female keelback snakes that each produced two clutches of eggs during our study.
All analyses are based on trait values corrected for the influence of maternal body size (residuals of ln-trait regressed on ln-SVL). F, P and repeatability values are derived from anovas using maternal identity number as the independent variable. Pearson correlation coefficients are based on correlations between values for clutch 1 and clutch 2. All traits are more similar between successive clutches from the same female than would have been expected by chance. Boldface font shows P < 0.05.
These significant repeatabilities suggest that reproductive traits may have a genetic component, although we emphasize that the repeatability could also be entirely because of environmental or nongenetic factors. If we found no repeatability it would be reasonable not to continue with heritability calculations (Postma & Van Noordwijk, 2005). The significant repeatability within individual keelbacks indicates that it is worth calculating actual heritabilities.
Table 2 presents calculated heritabilities of reproductive traits for the 59 cases in which we recorded clutches from female snakes and their adult daughters. The mean interval between the dates that clutches were produced by mothers vs. their daughters was 640 days (SE = 31.3 days, range = 323–1453). On average, the SVLs of reproducing female snakes in the first generation were larger than those of their daughters when they reproduced (mean SVLs 63.4 vs. 59.7 cm; paired t58 = 3.47, P = 0.001) and as a result clutch sizes in the first generation were larger than those in the second generation (means of 10.95 vs. 8.95 eggs; paired t58 = 3.65, P < 0.001). However, the mean mass of eggs produced was similar in the two generations (mean egg masses 2.73 vs. 2.82 g; paired t58 = 1.50, P = 0.14). The clear result from our calculation of heritabilities is that only one of these four traits is strikingly similar between mothers and daughters: the size of eggs that they produce (Table 2). Based on 95 % confidence intervals, the lowest bound for the estimate of true heritability of egg mass is 0.25. Raw measures of egg mass (i.e. not corrected for maternal SVL, or standardized to equal variance) also are significantly heritable between mothers and their daughters (slope = 0.54, SE = 0.20; P < 0.01). Despite the large standard error associated with egg mass heritability, it is the only trait for which the mother–daughter regression is statistically significant. Heritability of the other traits is low, and nowhere close to statistical significance. For example, the estimated heritability of clutch size is 0.04, with an upper 95% confidence limit of 0.58 (Table 2). Thus, it appears that egg mass could respond rapidly to selection within this population whereas clutch size could not.
Table 2. Mother–daughter regression results for reproductive traits from 59 pairs of female keelbacks and their daughters.
All analyses are based on trait values corrected for the influence of maternal body size (residuals of ln-trait regressed on ln-SVL). Trait values were then standardized to zero mean and unit variance. Heritability values and standard errors were derived from 10 000 bootstrap samples. Of the four traits examined, only egg mass appears to be significantly heritable. Boldface font shows P < 0.05.
The heritability of egg mass, estimated by parent–offspring regression, may be inflated because it includes nongenetic maternal effects (Lynch & Walsh, 1998). For example, a tendency to produce large eggs could be transferred from mother to offspring simply because larger eggs produce larger babies that in turn, grow quickly and tend to produce large eggs of their own. That is, some direct fitness consequence of hatchling size (rather than egg size per se) could generate a consistency across generations in egg size. We could test this scenario by experimentally manipulating offspring size relative to egg size, to break apart the usual correlation between these traits. One approach to achieve this aim would be allometric engineering (yolk removal, follicle ablation: Sinervo & Huey, 1990). However, a simpler method is fortuitously available from our data: modification of incubation conditions to increase or reduce hatchling size experimentally relative to egg mass. In keelbacks, incubation on relatively dry substrates severely reduces hatchling size, because much of the yolk mass desiccates and is left behind in the eggshell (Brown & Shine, 2004). We have data on moisture uptake (mass of egg at hatching vs. at oviposition) for 32 of the 59 maternal eggs from the mother–daughter data set analysed above. Thus, we conducted stepwise multiple regression with the daughter's mean egg mass as the dependent variable and traits relating to the egg from which she originally hatched (egg mass, water uptake of egg and incubation period) as well as her traits as a hatchling (SVL and mass) as the independent variables. If the size of eggs produced by a female keelback depend upon her own size at hatching rather than the mass of the egg from which she emerged, such a regression should identify water uptake during incubation or hatchling SVL, as well as original egg mass, as significant influences on subsequent egg sizes. In practice, this regression identified original egg mass as the sole significant variable (F1,30 = 14.97, P < 0.001): that is, the size of egg a female snake produced depended upon the size of the egg from which she originated, and not on the incubation conditions of that egg nor on her size as a hatchling. This has to be either a genetic effect, or a maternal effect of egg size per se (e.g. if hormone or nutrient levels consistently covary with egg size) rather than a maternal effect acting through large size at hatching. Similar analyses on relative fecundity of daughters (clutch size relative to maternal body size) revealed that factors such as the mass of the egg from which a female keelback emerged, or the size of the clutch from which she came, were not significantly related to her subsequent fecundity (all F < 2.26, all P > 0.36).
Trade-off between egg size and clutch size
Our data set allows us to examine both phenotypic correlations among reproductive traits, and the underlying genetic covariance. First, we consider phenotypic correlations. Table 3 (above-diagonal values) reveals a significant negative phenotypic correlation (−0.24) between relative egg mass and relative clutch size among the 118 clutches produced in our mother–daughter data set. A similar correlation coefficient is derived from the 138 clutches produced by the 69 ‘repeat’ female snakes (Pearson r = −0.44, P < 0.0001) analysed earlier in this paper. Interpretation of most of the phenotypic correlations in Table 3 is straightforward. Female keelbacks that produce large eggs tend to produce fewer eggs; female snakes that invest more in their clutches tend to have more and larger eggs; and female snakes that invest more energy into their clutches also tend to be in a better body condition after laying.
Table 3. Phenotypic correlations (above diagonal) and genetic correlations (below diagonal) between reproductive traits of keelbacks.
Relative clutch mass
Genetic correlation values are followed by standard errors in parentheses and the values beneath represent the percentage of undefined estimates among 10 000 bootstrap samples. Phenotypic correlations are derived from pooled data on the mother keelbacks and their daughters (hence n = 118). Boldface font shows P < 0.05.
−0.60 (1.35) (45)
Relative clutch mass
1.63 (2.18) (50)
−0.90 (3.82) (73)
0.89 (0.95) (6)
−0.32 (2.14) (50)
0.44 (2.30) (53)
The consistent negative correlation between clutch size and egg mass (after both variables are corrected for maternal body size) suggests a trade-off between these two variables. Similar trade-offs have been documented in many reptile species (Seigel & Ford, 1987; Olsson & Shine, 1997) but only one previous study has examined whether the trade-off has a genetic component, or is simply forced by physical constraints (Sinervo & Doughty, 1996). Unfortunately, calculating genetic correlations from natural populations with relatively small samples involves significant problems (Lynch, 1999; Réale & Roff, 2001). Standard errors were higher than correlation estimates for all of the genetic correlations that we calculated, and negative variance estimates resulted in undefined values in up to 73% of bootstrap samples (Table 3), thus the estimates may be specious. Nonetheless, the signs of the genetic correlations match those of the phenotypic correlations in five of the six cases. The genetic correlation between egg mass and clutch size (−0.60) has two components: one derived from the relationship between a mother's clutch size and the daughter's egg size, and a second relating the mother's egg size to the daughter's clutch size. Looking at these components separately reveals no tendency for mothers that produce large clutches to have daughters that produce small eggs (Fig. 2a; egg mass = 0.00–0.09 × residual clutch size; r2 = 0.008, d.f. = 1,57, P = 0.51), or for mothers that produce small eggs to have daughters that produce large clutches (Fig. 2b; residual clutch size = 0.00 + 0.05 × residual egg mass; r2 = 0.002, d.f. = 1,57, P = 0.74). Thus, the genetic relationship between the two traits remains ambiguous.
For wild populations, often it is only feasible to investigate the heritability of hatchling phenotypes as it is relatively easy to obtain litters of newborns. Investigating the heritability of adult phenotypes (e.g. reproductive traits) is more difficult because this requires waiting for hatchlings of known relatedness to mature and express the phenotype. These logistical constraints have discouraged attempts to estimate heritabilities for life-history traits in free-living vertebrates (Roff & Mousseau, 1987) and data are particularly scarce (virtually unavailable) for nontraditional study organisms.
Lacking data on genetic relationships among free-living animals, many investigators have adopted the more logistically feasible technique of estimating repeatabilities of life-history traits for individuals recorded to reproduce on more than one occasion. Consistency in a trait within an individual suggests, but does not prove, a genetic basis (Postma & Van Noordwijk, 2005). Such analyses on snakes have been conducted for two viviparous European species, the adder Vipera berus, Linnaeus, 1766 (Madsen & Shine, 1992) and the smooth snake Coronella austriaca, Laurenti, 1768 (Luiselli et al., 1996). Data were available for ‘repeat’ litters for only six female adders, but suggested high levels of repeatability for relative clutch size and offspring size, relative clutch mass and post-partum body condition (Madsen & Shine, 1992). In a sample of seven female smooth snakes, Luiselli et al. (1996) reported significant repeatability for several traits (offspring length and mass, RCM, %stillbirths) but most of these effects disappeared after correction for female body size. In a larger oviparous snake species (water python, Liasis fuscus, Peters, 1873) from the same study area as our keelbacks, successive clutches from 13 female snakes (1–2 years apart) revealed significant repeatability in offspring size and hatching success, but not in fecundity or maternal body condition (Madsen & Shine, 1996). In a laboratory study on an oviparous African colubrid snake (Lamprophis fuliginosus, Boie, 1827), Ford & Seigel (2006) reported repeatabilities of egg size and litter size among four successive litters of 0.67 and 0.65 respectively.
Although repeatability is often considered to represent an upper limit to heritability, Dohm (2002) identified situations in which repeatability could underestimate heritability. In the present study, heritability of egg mass (0.73) was substantially higher than repeatability (0.26). Maternal effects may explain the discrepancy between these estimates (Dohm, 2002). If the repeatability of 0.26 represents the true upper limit of heritability of egg mass, then a large maternal effect must be conflating our heritability estimate of 0.73. We demonstrated that this maternal effect is unrelated to hatchling size, but we cannot rule out more subtle maternal effects on hatchling composition. For example, larger eggs may contain higher levels of specific hormones, nutrients or minerals that affect some aspect of physiology of the resultant hatchling, which in turn allows them to produce larger eggs when they attain reproductive age. Testing this hypothesis would require a detailed analysis of yolk composition in eggs of different sizes, as well as manipulations of yolk composition to detect putative cross-generation effects. Such work would be logistically challenging, but it is encouraging to note that in preliminary analyses of keelback eggs we have found a significant correlation between egg size and the concentration of the sex hormone dihydrotestosterone (R. Radder, G.P. Brown and R. Shine, unpublished data).
In contrast to egg mass analyses, our estimates of repeatability of the clutch size and relative clutch mass were substantially higher than our estimates of heritability for these variables. The high repeatability, despite a low heritability, may be attributable to the ‘repeat’ female snakes tending to be in similar energetic condition when they produced their two litters (i.e. repeatability of post-partum maternal condition = 0.56). Repeatability is a much easier parameter to assess in natural conditions than is heritability, but on its own may provide little insight into the genetic framework underlying the trait. Nonetheless, if heritability data are available also (as in the present study), estimates of repeatability may prove useful in interpreting the overall role of genetic, environmental and maternal effects in engendering trait variation.
How heritable is egg mass in free-living keelbacks? Our estimate of heritability for this trait (h2 = 0.73) may be inflated by two effects. First, we removed a potential source of environmental variation by incubating eggs under controlled conditions. Second, potential maternal effects on egg yolk constitution (see above) may have exaggerated mother–daughter similarity in egg mass. Regardless, it is clear that in keelback snakes, egg mass is more heritable than is clutch size (h2 = 0.04). An extensive literature predicts that traits with a large effect on fitness should have low heritability, because selection will erode variance (i.e. any alleles coding for nonoptimal trait values will be eliminated: Roff, 1992, 2002). Thus, additive genetic variance (and hence, heritability) for fitness-relevant traits will be low. In this simplistic form, the hypothesis predicts that in our keelback study population, the trait with highest heritability (egg size) should be relatively unimportant for fitness. Our mark-recapture studies strongly falsify this prediction. Larger eggs tend to produce larger hatchlings (Shine & Brown, 2002; Brown & Shine, 2004), and hatchling size is a primary determinant of a snake's probability of surviving through its first year of life (Brown & Shine, 2004). In these short-lived animals, survival for the first year (i.e. through to maturation: Brown & Shine, 2002) must be responsible for a significant proportion of the total variation in lifetime fitness among individuals within the population. How, then, can heritability be so high for a trait that is so strongly associated with fitness?
The answer to this paradox may lie in the fact that egg mass is not the only determinant of hatchling size. As noted above, the size of a hatchling keelback depends primarily upon two factors: the mass of the egg from which it emerges, and the moisture content of the substrate on which that egg has incubated. At any given egg mass, offspring size is enhanced by incubation in a moister rather than drier nest site (Shine & Brown, 2002; Brown & Shine, 2005a,b). Selection may thus work intensely on genes for maternal nest-site selection, eroding underlying genetic variance in that trait, and coincidentally preserving significant genetic variation for egg mass per se (Brown & Shine, 2005b).
Although trade-offs between egg size and clutch size play a central role in life-history models (e.g. Smith & Fretwell, 1974; Kaplan & Cooper, 1984; Olsson & Shine, 1997) and have been demonstrated at the phenotypic level in many organisms (including snakes: Seigel & Ford, 1987), we know surprisingly little about the genetic basis of this trade-off. Roff (2002) review of this topic is based on only three empirical examples, involving lizards (Sinervo & Doughty, 1996), fishes (Snyder, 1991) and fruit flies (Schwarzkopf et al., 1999). All of these studies indicate (or at least, hint at) the existence of a negative genetic correlation between egg size and clutch size. However, in a long-term field study on lesser snow geese (Anser caerulescens, Linnaeus, 1758), Lessells et al. (1989) found patterns similar to those seen in our keelback snakes. They reported high heritability of egg size (0.60) whereas heritability of clutch size was low (0.15) and the two traits were correlated phenotypically but not genetically. Furthermore, recent selection studies on insects suggest that the quality of the environment plays a major role in determining whether or not genetic trade-offs between offspring size and number are detected (Czesak & Fox, 2003; Blanckenhorn & Heyland, 2004). Similarly, food supply affects whether or not a phenotypic trade-off is apparent between offspring size and number in the sand lizard, Lacerta vivipara, Jacquin, 1787 (Uller & Olsson, 2005). Such studies demonstrate the complexities of detecting genetic trade-offs, especially in natural populations. Nonetheless, free-living animals are the vehicles through which the quantitative genetics processes that are of interest to evolutionary ecologists occur. Therefore, documenting heritabilities and genetic covariances in nonartificial settings provides necessary benchmarks for interpreting the results of laboratory-based studies.
Our analyses suggest that variation in egg size within the keelback population has a strong genetic underpinning, whereas variation in clutch size is mainly determined by external factors. The contrast is even stronger if one looks at the raw data, without correcting for variation in maternal body size. Mean egg mass shifts only slightly with maternal SVL (2.5–3.2 g, < 30 % difference between smallest and largest female snakes, coefficient of variation = 0.14) whereas clutch size increases by about 500 % from small to large female snakes (4–20 eggs, coefficient of variation = 0.28; Brown & Shine, 2002). Thus, much of the phenotypic variance in clutch size is generated by variation in female body size, in turn reflecting growth rates and age structures. In contrast, egg size may be largely determined by additive genetic variance.
Lastly, we consider the mechanisms responsible for the observed phenotypic trade-off between egg size and clutch size in keelbacks. For simplicity, we can envisage three types of links between egg size and fecundity from a life-history perspective (actually, these are end points on a continua of possibilities). First, both traits might be under tight genetic control, so that their covariation is caused by genetic linkage. This scenario is supported by the studies reviewed by Roff (2002), including the only previous analysis on reptiles (Sinervo & Doughty, 1996). Second, variation in both traits could be generated by proximate environmental factors, so that their covariation results from some kind of constraint (e.g. finite energy or abdominal space) combined with stochastic variation in both egg size and fecundity. Third, one of the traits might be under strong genetic control whereas the other is not, in which case the trade-off arises because one of the traits is determined genetically and the other takes whatever value is possible under space or energy constraints (Olsson & Shine, 1997; Du et al., 2005). Our data on keelbacks clearly support the latter possibility. When a female keelback reproduces, her egg size is largely determined by her genetic inheritance whereas her clutch size is set by whatever proximate cues determine her total energy allocation to reproduction. This conclusion accords well with the observation that despite significant annual variation in traits such as clutch sizes and Relative Clutch Masses, mean egg sizes remain the same among years in the keelback population (G.P. Brown & R. Shine, unpublished data).
We thank C. Shilton and the staff of Beatrice Hill Farm for assistance in a multitude of ways, and the Australian Research Council for funding. The comments of B. Phillips and two anonymous reviewers substantially improved the manuscript.