Phenotypic plasticity in reproductive traits: geographical variation in plasticity in a viviparous snake



  • 1 Previous experiments showed that Checkered Garter Snakes (Thamnophis marcianus) from south Texas, USA (an environment subject to high seasonal and annual variation in environmental conditions), demonstrated marked phenotypic plasticity in clutch size and clutch mass in response to experimental changes in prey availability.
  • 2 In this study, the extent of phenotypic plasticity in life-history traits in Checkered Garter Snakes from south Texas was experimentally compared with a population of the same species from south-eastern Arizona, where the environment may be more constant.
  • 3 Unlike results from south Texas, Checkered Garter Snakes from Arizona showed no significant phenotypic plasticity in clutch size, clutch mass or any other life-history trait in response to changes in food availability, at least within the boundaries of our experimental conditions.
  • 4 The data indicate that the degree of phenotypic plasticity in life-history traits differs among populations within the same species. However, these differences are subject to both adaptive and non-adaptive explanations.


Phenotypic plasticity has received increasing amounts of attention from evolutionary ecologists, both from an empirical and theoretical perspective (see reviews in Stearns 1989; Travis 1994; Via 1994; De Jong 1995; Gotthard & Nylin 1995; Via et al. 1995). Empirically, there is growing recognition that interpreting spatial or temporal variation in life-history traits in a variety of organisms requires partitioning of the sources of variation into environmentally induced components (plasticity) vs variation caused by non-environmental factors (e.g. Ballinger 1983; Berven & Gill 1983; Kaplan 1987; Niewiarowski & Roosenburg 1993; Qualls & Shine 1996; Sinervo & Doughty 1996). From a theoretical viewpoint, considerable attention has been focused on the evolution of phenotypic plasticity and whether phenotypic plasticity is itself an adaptive trait (e.g. Caswell 1983; Bull 1987; Via 1993, 1994; Gotthard & Nylin 1995; Via et al. 1995).

In a recent series of papers, we showed that changes in food availability had a strong effect on some (but not all) reproductive traits in live-bearing Checkered Garter Snakes (Thamnophis marcianus) and egg-laying Corn Snakes (Elaphe guttata) (Ford & Seigel 1989, 1994; Seigel & Ford 1991, 1992). Specifically, we found that although the number of eggs or offspring per breeding event (clutch size) was strongly affected by changes in food availability in both T. marcianus and E. guttata, offspring size showed no significant response, suggesting that the degree of plasticity with respect to food availability differs among life-history traits. Other authors working with squamate reptiles have found similar results in both laboratory (e.g. Ferguson & Talent 1993; James & Whitford 1994; Gregory & Skebo 1998) and field studies (e.g. Andren & Nilson 1983; Ballinger 1983; Seigel & Fitch 1985; Shine & Madsen 1997; but see Madsen & Shine 1992 and Luiselli et al. 1996). However, most field studies and laboratory experiments have focused on individuals from a single population (but see Bronikowski & Arnold 1999). Several authors have suggested that the degree of phenotypic plasticity may vary among populations (especially if populations are exposed to different environmental conditions), and that geographical differences in plasticity may be a useful test in determining whether plasticity is adaptive (Ferguson & Talent 1993; Gotthard & Nylin 1995; Bronikowski & Arnold 1999).

In this paper, we contrast the degree of phenotypic plasticity shown by our original study population of T. marcianus (where food availability is highly variable temporally), with new data on a population of Checkered Garter Snakes from Arizona (where food availability is more constant). We pose the following questions: (1) Is the degree of phenotypic plasticity in life-history traits variable between populations in Checkered Garter Snakes? (2) Are traits that were canalized in one population (e.g. offspring size) also canalized in a second population under different environmental conditions? (3) When diet is controlled in the laboratory, are observed geographical differences in life-history traits between these populations still present?

Materials and methods


Our original experimental studies on phenotypic plasticity of reproduction (Ford & Seigel 1989) were based on Checkered Garter Snakes collected from southern Texas (La Salle & McMullin counties: 28°25·9′ N, 99°46·5′ W). The environment in this part of southern Texas is marked by extremely wide yearly fluctuations in annual precipitation, especially in the vicinity of our study site (Norwine & Bingham 1985). The activity of garter snakes, the prey available to them and their reproductive characteristics are also all highly variable (Karges 1983; Ford & Karges 1987; N. Ford, unpublished data). For example, both movements of snakes and prey availability (primarily amphibians) are heavily affected by late summer rains resulting from tropical storms (N. Ford, unpublished data). Reproductive phenology is also highly variable; females are not receptive to courting males until at least 28 days posthibernation (Karges 1983; Ford & Cobb 1992), and females may give birth to offspring as early as late May (local late spring) or as late as early October (local early autumn) (Ford & Karges 1987; N. Ford, unpublished data). The natural clutch size of this population averages 9·5 offspring per clutch and mean adult female snout–vent length (SVL) is 52·6 cm (Seigel, Ford & Mahrt 2000).

In contrast, we have been studying a second population of Checkered Garter Snakes from southeastern Arizona where local environmental conditions are apparently more constant. At this site (Graham County, 32°46·6′ N, 109°40·8′ W; 1320 m a.s.l.), prey availability is both more constant and predictable, because adult snakes feed mainly on Bullfrog tadpoles (Ranacatesbeiana) which are available year round in an artificial pond built in the 1950s (Rayburn 1990; Rossman, Ford & Seigel 1996, p. 204; see Seigel et al. 2000 for data on feeding habits). Although the pond is artificial, the water source for the pond is a natural mountain-fed spring that flows in all but the driest years, and tadpoles are apparently available all year. Reproductive timing in T. marcianus from this site is much less variable than in south Texas; females are receptive to courtship within a day after emerging from hibernation (Ford & Cobb 1992), and virtually all females produce a single brood of offspring in mid to late May or early June (Seigel et al. 2000). The mean natural clutch size for this population is 14·1 offspring/female and mean female SVL is 63·9 cm (Seigel et al. 2000). Offspring size is also larger in Arizona (Texas, neonate SVL = 15·0 cm; Arizona, 17·7 cm). Thus, females in Arizona are much larger and give birth to larger clutches and longer offspring (Seigel et al. 2000).

In this paper preliminary data are also included on Checkered Garter Snakes from northern Texas (Lubbock County), where, as in south Texas, the environment is highly variable in terms of rainfall and prey abundance. These snakes were collected and maintained at the same times as were the snakes from Arizona, but, because of the limited sample sizes available (five snakes per feeding group), full statistical analyses for the north Texas population are not presented here.


This study replicates the procedures reported in Ford & Seigel (1989) in all respects, except this time using snakes from Arizona and north Texas. The independent variable in this experiment was differential food intake during the period of secondary vitellogenesis, which occurs after hibernation in both populations of T.marcianus (personal observation). This type of reproductive pattern, referred to as Type I secondary vitellogenesis (Aldridge 1979), is characteristic of virtually all species of garter snakes (Rossman et al. 1996, p. 63). Hence, our comparisons between localities focus on the same component of the female reproductive cycle.

All snakes were collected in the field between May and October and placed on an intermediate diet (15% body mass/week in Bullfrog, Ranacatesbeiana, tadpoles) for at least 3 months prior to hibernation. This diet was adequate for normal body mass maintenance in non-breeding females. Animals were housed individually and hibernated artificially for 4 months at 12 °C with a 0L:24D cycle to synchronize reproduction in the spring. Immediately following hibernation, female snakes were assigned randomly into either a high-energy or low-energy diet and placed on a photoperiod of 14L:10D at 28 °C. The high-energy group received 30% of their body mass in tadpoles twice per week whereas the low-energy group was given only 10% twice per week. This high diet was approximately the maximal amount of food females would accept consistently and the low diet was the minimum that allowed normal reproduction without health problems occurring. These diets produced snakes that were in approximately the same range of body conditions seen in wild-caught animals, and so appear to be relevant to natural environmental conditions. Females were given multiple opportunities to mate with males, but housed individually at all other times. Females were weighed to the nearest gram and measured (SVL) to the nearest cm every month. At parturition females were weighed and measured again as were all neonates. Clutch size was calculated as the number of viable plus normal-appearing stillborn offspring and clutch mass as the total mass of those neonates. Relative clutch mass (RCM) was calculated as the mass of the offspring divided by the postparturient mass of the female.


Data were analysed using systat (systat Inc. 1992), SAS (SAS Inst. Inc. 1985) or JMP (SAS Inst. Inc. 1996). All data were tested for assumptions of parametric tests (homogeneity of variances and normality). Because of correlations between maternal body size and many reproductive characteristics in these snakes, all reproductive variables except RCM were analysed by analysis of covariance (ancova), using SVL as the covariate. Although RCM is expressed as a ratio for illustrative purposes, differences in RCM between diets and populations were analysed using an ancova with clutch mass as the dependent variable and female mass after parturition as the covariate. Except as noted, differences in groups using ancova represent differences in y-intercepts, not slopes. Means are followed by ±1 SE



The SVL of females in the two diet groups were virtually identical (Table 1). Females on the high-energy diet consistently ate their entire food allotment (although some females occasionally would leave a small amount of food in their feeding tray), and high-energy females were significantly heavier (23·5%, P = 0·028) than low-energy females prior to parturition (Table 1).

Table 1.  Summary of the reproductive characteristics of Thamnophismarcianus from Arizona on different dietary regimens. Means are followed by one SE. All reproductive traits are adjusted for female SVL via ancova (see text). F- and P-values test differences between diets
DietNFemale SVL (cm)Mass before parturition (g)Mass after parturition (g)Clutch sizeClutch mass (g)RCMOffspring SVL (cm)Offspring mass (g)
Low1161·8197·4 ± 14·29154·1 ± 11·6317·3 ± 1·5942·3 ± 3·5622·3 ± 1·1317·5 ± 0·292·55 ± 0·405
High1462·9243·3 ± 13·14176·9 ± 10·2014·7 ± 1·4641·2 ± 3·2719·5 ± 1·0018·2 ± 0·272·84 ± 0·104
F  5·572·161·380·052·382·883·73
P  0·0280·1570·2570·810·1380·1050·067

Despite the differences in feeding rates and consequent body mass, no evidence was found for phenotypic plasticity in any reproductive trait within the Arizona population (Table 1). Indeed, clutch size in the low energy group was actually higher than in the high-energy group, but these differences were not close to statistical significance (P = 0·26; see Table 1 for full statistical results). The only trait that showed even a marginally significant difference between the diet groups was mean offspring mass (P = 0·067), where offspring in the high-energy group weighed 0·29 g (11%) more than offspring from the low-energy group (Table 1).

A retrospective power analysis (see Thomas 1997) was used with a hypothesized range of effect sizes (differences in group means) based on our original experiments (Ford & Seigel 1989). The power analysis indicated we had sufficient power (0·80) to detect an effect size as low as 20% between the diet groups for clutch size. By contrast, differences were observed in clutch size between diets of over 45% in the south Texas population (see Table 2). In other words, the lack of significant differences between diets for clutch size and clutch mass in Arizona was not due to low sample size, but to a small difference between diets (17·7% difference for clutch size, 2·7% difference for clutch mass).

Table 2.  Percentage differences in reproductive traits between diets for Checkered Garter Snakes from three localities. Except for RCM, percentage differences were calculated as the difference between the lowest and the highest value, divided by the lowest value. Because RCM is already expressed as a percentage, percentage differences were calculated as the highest minus the lowest value. Data from south Texas are from Ford & Seigel (1989). Data from north Texas are for comparative purposes only (see text)
LocalityClutch sizeClutch massRCMOffspring SVLOffspring mass
South Texas47·7%53·1%3·1%1·3% 0·0%
Arizona17·7% 2·7%2·8%4·0%11·4%
North Texas33·8%32·9%3·6%1·2% 2·0%

Data on plasticity for the snakes from north Texas showed patterns similar to those from south Texas; high degrees of plasticity for clutch size and clutch mass but very little effect of diet on RCM, offspring mass and offspring length (Table 2). Because of the small sample sizes, full statistical analyses were not conducted for this locality.


The contrast in the ‘norms of reaction’ for clutch size and offspring size are shown in Figs 1 and 2. Clutch size increases significantly with increasing diet in Texas, but decreases (non-significantly) in Arizona, resulting in norms of reaction with different directions (Fig. 1). Conversely, offspring size does not vary with diet in either population, resulting in parallel norms of reaction (Fig. 2).

Figure 1.

Graphical comparison of the degree of phenotypic plasticity (means ±2 SE) between two populations of Checkered Garter Snakes for clutch size. The non-parallel lines in the graph for clutch size indicate a significant interaction between population and diet, showing differences in the degree of plasticity between populations. See Table 3 for statistical analysis.

Figure 2.

Graphical comparison of the degree of phenotypic plasticity (means ±2 SE) between two populations of Checkered Garter Snakes for offspring size. The parallel lines in the graph for offspring size indicate no difference in plasticity between the populations for this trait. See Table 3 for statistical analysis.

Table 2 summarizes the differences in phenotypic plasticity (based on percentage differences between the experimental diets) between south Texas and Arizona. Although the percentage differences between diets are roughly equivalent for RCM, offspring SVL and offspring mass (none of which showed significant diet effects at either locality), the Arizona population showed a much lower degree of plasticity for clutch size and mass than did snakes from south Texas.


The Arizona and south Texas populations showed marked differences in reproductive traits in the wild (see Materials and methods). To determine if these apparent geographical differences were due to diet, the data from the two experiments were combined in a two-way ancova, using SVL as the covariate and diet and locality as main effects. Our data show that some, but not all, of these differences disappear once diet and maternal SVL are controlled (Table 3). Most notably, geographical differences in clutch size are no longer significant once diet and maternal SVL are controlled. Conversely, females in Arizona give birth to significantly larger offspring, even after diet and maternal SVL are controlled. Locality also had a significant effect on clutch mass and RCM, but these differences are more difficult to interpret owing to a significant interaction between diet and locality for these two traits (Table 3).

Table 3.  Results of two-way ancova on the influence of diet and locality on the reproductive characteristics of Arizona and South Texas populations of Thamnophismarcianus on different dietary regimens. A significant interaction term indicates differences in phenotypic plasticity between the populations (see text). All df = 1,71 except for offspring traits which were 1,72
 Locality effectsDiet effectsSVL effectsInteraction
  • Mass after parturition used as the covariate for testing differences in RCM.

Clutch size0·910·340·840·3649·710·000110·580·0018
Clutch mass16·490·00013·530·06556·220·00015·340·0239
Offspring SVL85·770·00013·020·08671·580·21250·590·4435
Offspring mass36·550·00013·760·05671·650·20311·110·296


Clutch size and clutch mass in Arizona garter snakes were not significantly plastic in response to food availability, whereas the same traits were highly plastic in the south Texas population. Thus, geographical variation in the degree of phenotypic plasticity for these traits has been demonstrated within a single species. It is important to note that we cannot determine whether phenotypic plasticity is absent in Arizona, since it is possible that even lower or higher experimental diets may have produced significant plasticity (we are grateful to an anonymous reviewer for this suggestion). However, our finding that at least the degree of plasticity varies among populations (Figs 1 and 2) suggests that plasticity may be an adaptive trait that is subject to selection in different environments (Gotthard & Nylin 1995). However, our results are subject to both adaptive and non-adaptive explanations. These are reviewed below.


Our data support the hypothesis that the higher degree of phenotypic plasticity in Texas occurs because food availability is much more variable in Texas. Checkered Garter Snakes in Texas appear to feed heavily after tropical storms (N. Ford, personal observation), whereas the primary prey for garter snakes at Roper Lake (Bullfrog tadpoles) are available throughout the active season (Seigel et al. 2000). In this hypothesis, females in Texas are under selection to take advantage of sudden bursts of high food availability by producing larger clutch sizes and clutch masses by diverting energy towards reproduction. Conversely, during periods of low food availability, females reduce clutch size and clutch mass in order to minimize energy demands resulting from reproduction. Such an explanation assumes that females that are able to adjust their clutch sizes (either higher or lower) enjoy a higher lifetime reproductive output than do females that have a lower degree of plasticity for clutch size and clutch mass. In Arizona, it is hypothesized that either no such selection exists or is much weaker, since food availability is relatively constant; hence reproductive traits are partially or completely canalized relative to food availability. Thus, this explanation assumes that plasticity is a derived trait in this group of snakes (but see below).


An important criticism of adaptive explanations is the failure to consider non-adaptive or other hypotheses (Gould & Lewontin 1979). One explanation for the difference in plasticity between our snake populations involves the differences in timing of reproduction between the populations. In south Texas, a minimum of 28 days elapses between emergence from hibernation and mating, whereas the Arizona females mated within a day of emergence (Ford & Cobb 1992). Presumably, there is a longer interval between emergence from hibernation and ovulation in Texas than in Arizona, and this is reflected in the more consistent time of parturition in Arizona (mid to late May vs May to October in Texas). Thus, females in south Texas have a much longer ‘window’ in the spring for adjustment of clutch size and clutch mass relative to food availability than do females in Arizona. Consequently, any additional energy available to females in Arizona immediately after emergence is likely to be devoted to growth and/or storage, whereas in Texas, additional energy after emergence is more likely to be used to increase current reproductive potential. Thus, the observed differences in plasticity between these populations may not be due to selection for plasticity in clutch size or clutch mass per se, but to increased time to modify clutch size in Texas. This, in turn, suggests a possible adaptive association between plasticity in reproductive phenology and plasticity in life-history traits, where plasticity in reproductive phenology results in selection for plasticity in clutch size (or vice versa; K. Gotthard, personal communication). This hypothesis is supported by suggestions by Sinervo & Licht (1991) that physiologically, adjustments to clutch size in the lizard Uta stansburiana are more likely to occur during the early stages of vitellogenesis, which, in these garter snakes, occurs after emergence from hibernation (Rossman et al. 1996).

We caution that our experiments have only tested the effect of differential food levels after emergence from hibernation. It is possible that variable food resources in the autumn preceding hibernation may result in changes in clutch size the following year (see Ford & Seigel 1994; Seigel & Fitch 1985). Thus, plasticity in reproductive traits may, in fact, be present in Arizona as a consequence of the previous year’s environment. Such a possibility has been suggested to occur in the field (Seigel & Fitch 1985; Shine & Madsen 1997) in snakes, but requires further experimental testing.

A second explanation for the differences in phenotypic plasticity between these populations involves phylogenetic effects. Adaptive explanations for phenotypic plasticity inherently assume that plasticity is a derived trait, with plasticity limited or absent in ancestral groups (see Discussions in Coddington 1988; Doughty 1995; Gotthard & Nylin 1995). However, we are unaware of any tests of this assumption for squamate reptiles. For Checkered Garter Snakes it may be as logical to assume that the presence of plasticity is ancestral and the absence of plasticity is derived and adaptive. In order to test this hypothesis, additional data from a variety of squamate reptiles are needed (see Doughty 1995 for specific tests). A recent review of annual variation in reproductive traits in snakes (Seigel et al. 1995) indicated considerable interspecific variation in the presence of plasticity. Indeed, at the same time as these experiments were conducted, limited data were collected on a third population of Checkered Garter Snakes from northern Texas, where the environment is highly variable (as in south Texas). A brief summary of these preliminary data are presented in Table 2; in contrast to the results from Arizona, Checkered Garter Snakes from northern Texas showed degrees of plasticity very similar to those from south Texas. Thus, with additional data, the presence or absence of plasticity could be plotted on a cladogram using the methods of Pagel & Harvey (1989) to determine whether plasticity is derived or ancestral in Checkered Garter Snakes, or, more broadly, in squamate reptiles (Doughty 1995; Qualls & Shine 1996).


Considerable time and effort has been invested in examining geographical differences in life-history traits in reptiles (see review in Fitch 1985). However, it has been argued elsewhere that much of the observed geographical differences in clutch size among populations may be the result of phenotypic plasticity, not selection to the local environment (Ford & Seigel 1989; Seigel & Ford 1991). Our data show that observed differences in clutch size between south Texas and Arizona are a function mainly of diet and maternal SVL; when these factors are controlled, there is no significant difference in clutch size between the sites. However, the two sites differ in their response to diet, as shown by the significant interaction term (Fig. 1; Table 3). Although this may be an artefact of running the experiments in different years, we think this is unlikely, since snakes from north Texas showed essentially the same response to diet as did snakes from south Texas. Thus, our data imply that adaptive explanations for geographical differences in clutch size per se for this species (and other species showing plasticity in clutch size) may be misdirected. Instead, more attention should be given to understanding geographical differences in phenotypic plasticity rather than differences in mean values of traits such as clutch size. Conversely, offspring size was not significantly plastic in either population, yet still showed strong geographical differences among sites, even after correction for diet and female size. Why offspring size varies among populations and why some traits (clutch size) are plastic whereas others (offspring size) are canalized needs further attention.


For reviews of the manuscript we thank Rick Shine, Karl Gotthard, Peter Niewiarowski, David Reznick, Nadia Seigel and an anonymous reviewer. We are grateful to the staff of Roper Lake State Park for logistical support, and to the Arizona Game and Fish Department for permits. We also thank Laura Rayburn Mahrt, Nadia Seigel and Ben Seigel for help with the field work, and the many students who helped collect these data, especially Cliff (Sonny) Fontenot and J. Sean Doody. This research was supported by a grant from the Louisiana Education Quality Support Fund, and by Faculty Development Grants from Southeastern Louisiana University to R.A.S. and University of Texas-Tyler to N.B.F. This research was also supported by DOE contract number DE-FC09–96SR18546 with the University of Georgia Research Foundation and by the Faculty Research Participation Program of the Oak Ridge Institute for Science and Education (ORISE).

Received 26 October 1999; revised 26 April 2000; accepted 7 July 2000