DOES VIVIPARITY EVOLVE IN COLD CLIMATE REPTILES BECAUSE PREGNANT FEMALES MAINTAIN STABLE (NOT HIGH) BODY TEMPERATURES?

Evolution, 58(8), 2004, pp. 1809 1818 DOES VIVIPARITY EVOLVE IN COLD CLIMATE REPTILES BECAUSE PREGNANT FEMALES MAINTAIN STABLE (NOT HIGH) BODY TEMPERATURES? RICHARD SHINE School of Biological Sciences, A08, University of Sydney, New South Wales 2006, Australia E-mail: rics@bio.usyd.edu.au Abstract. Viviparity (live bearing) has evolved from egg laying (oviparity) in many lineages of lizards and snakes, apparently in response to occupancy of cold climates. Explanations for this pattern have focused on the idea that behaviorally thermoregulating (sun-basking) pregnant female reptiles can maintain higher incubation temperatures for their embryos than would be available in nests under the soil surface. This is certainly true at very high elevations, where only viviparous species occur. However, comparisons of nest and lizard temperatures at sites close to the upper elevational limit for oviparous reptiles (presumably, the selective environment where the transition from oviparity to viviparity actually occurs) suggest that reproductive mode has less effect on mean incubation temperatures than on the diel distribution of those temperatures. Nests of the oviparous scincid lizard Bassiana duperreyi showed smooth diel cycles of heating and cooling. In contrast, body temperatures of the viviparous scincid Eulamprus heatwolei rose abruptly in the morning, were high and stable during daylight hours, and fell abruptly at night. Laboratory incubation experiments mimicking these patterns showed that developmental rates of eggs and phenotypic traits of hatchling B. duperreyi were sensitive to this type of thermal variance as well as to mean temperature. Hence, diel distributions as well as mean incubation temperatures may have played an important role in the selective forces for viviparity. More generally, variances as well as mean values of abiotic factors may constitute significant selective forces on life-history evolution. Key words. Adaptation, developmental plasticity, embryo, lizard, phenotype, thermal. 2004 The Society for the Study of Evolution. All rights reserved. Received February 23, 2004. Accepted May 4, 2004. 1809 Understanding the selective forces that have shaped major life-history transitions is an important aim of ecological research (e.g., Calder 1984; Roff 1992; Charnov 1993). Inevitably, the challenge is a difficult one given that we need to infer processes at work in evolutionary history from present-day phenomena (Harvey and Pagel 1991). Indeed, it is difficult to imagine any way to robustly test among alternative explanations for some transition that has occurred only once, in the distant past. Nonetheless, some transitions have occurred so frequently, and are so clearly linked to specific environmental factors, that they provide opportunities to clarify the operation of microevolutionary forces responsible for important macroevolutionary transitions. The evolution of viviparity (live bearing) from oviparity (egg laying) in squamate reptiles offers unique advantages in this respect. First, at least 100 separate lineages of lizards and snakes have undergone this transition, offering immense potential for comparative study (Blackburn 1982, 1985, 1999; Shine 1985; Lee and Shine 1998; Shine and Lee 1999). Second, phylogenetic analyses indicate that most of the recent transitions (i.e., those for which we can identify closely related oviparous and viviparous taxa) have occurred in cold climates (Blackburn 1982, 1985; Shine 1985; Qualls et al. 1995). Third, present-day geographic distributions of viviparous reptiles show the same pattern: the proportion of viviparous taxa is higher in colder climates, across all continents that contain reptiles (Sergeev 1940; Tinkle and Gibbons 1977; Greer 1989; Qualls et al. 1995). This association between cold climates and viviparity is evident within many reproductively bimodal lineages, albeit not within all such taxa (e.g., Mabuya: Blackburn 2000; Blackburn and Vitt 2001). Fourth, there is a plausible adaptationist hypothesis to explain this association between cold climates and viviparity: that maternal thermoregulation in cold climates exposes the developing offspring to higher temperatures than would be experienced by eggs in a nest, and that this highertemperature incubation accelerates hatching and enhances viability of the offspring (Mell 1929; Weekes 1933; Sergeev 1940). Reflecting these advantages, selective forces for the evolutionary transition to viviparity in reptiles have attracted considerable research. This work has broadly supported the feasibility of many assumptions underpinning the cold-climate hypothesis outlined above (Shine 1983, 2002a). For example, at very high elevations (above those at which oviparous species occur), there are no thermally suitable nest sites for egg-layers (Shine et al. 2003) and eggs translocated to potential nest sites in these areas experience high rates of mortality (Shine 2002a). However, although these data bear upon the present-day distributions of egg-layers and livebearers, they do not address the question of adaptive shifts from one reproductive mode to the other. The shift from oviparity to viviparity presumably occurred under conditions similar to those that are experienced at the upper elevational limit for oviparous reproduction. Thus, we need to examine conditions at these elevations, not simply at much higher (colder) sites where oviparity clearly is not a feasible lifehistory tactic. The results of recent studies at the upper elevational limits to oviparity pose a strong challenge to the central tenet of the cold-climate hypothesis: the assumption that eggs retained in utero will be kept warmer than eggs laid in a nest. Andrews (2000) found only minor disparities in mean temperature between nests and pregnant female lizards in three North American phrynosomatid species living at high elevations. However, thermal variances were higher for female body temperatures than for nests (Andrews 2000). Subsequent studies on Australian scincid lizards produced very similar results. At the upper elevational limit for oviparous

1810 RICHARD SHINE reproduction, eggs maintained in utero experienced about the same mean temperature as eggs laid in the nest, but higher thermal variance (Shine et al. 2003). Do these results mean that thermal advantages of uterine retention have not been important selective forces for the evolution of viviparity in squamate reptiles? This inference seems paradoxical, given that comparative analyses strongly link cold climates to the transition from oviparity to viviparity (above). Another possibility is that maternal thermoregulation has indeed played a significant role, but that previous discussions of selective forces have focused on the wrong aspect of thermal regimes: that is, on the mean rather than the variance. Results both from Andrews (2000) and Shine et al. s (2003) work do show a strong effect of uterine retention on the thermal regimes experienced by an embryo, but in diel distributions rather than in mean values. Could a shift in thermal variance, independent of the mean, influence developmental rates or hatchling phenotypes (and thus, potentially, offspring fitness) in oviparous reptiles? In keeping with this possibility, previous studies suggest that thermal variation at a variety of temporal scales can affect developmental rates and/or phenotypic traits of hatchlings. For example, hatchlings of the cold-climate oviparous scincid lizard Bassiana duperreyi are affected not only by mean temperature but also by variance within each day, between days, and across months (Shine and Harlow 1996; Shine et al. 1997; Shine and Elphick 2001; Shine 2002b, 2004). These considerations suggest a novel hypothesis on the selective forces for viviparity: that the fitness advantages to uterine retention of eggs are due not (or not only) to an eggretaining female maintaining high mean temperatures, but to her maintaining relatively stable temperatures by behavioral thermoregulation. Nests display relatively smooth diel cycles of heating and cooling, such that the overall frequency distribution of temperatures experienced will resemble a normal curve (most values near the mean). In contrast, behaviorally thermoregulating female reptiles (at least in heliothermic species) maintain high and relatively invariant temperatures during daylight hours, falling to ambient levels at night. The result will be that in live-bearers the pattern of temperature fluctuation across 24 hours is virtually a square waveform, with sharp transitions, whereas for nests it is a smooth waveform like a sine wave, with gradual transitions. As a result, the overall frequency distributions of body temperature for such an animal will be bimodal, with many low values (at night) and high values (by day), but relatively few records close to the mean. Hence, an egg retained in utero is likely to experience a shift in the way that incubation temperatures change through the course of the day. Can such a shift in diel distributions, independent of any change in mean temperature, modify the phenotypes of hatchlings (and hence, potentially, their fitness)? To clarify these issues I (1) measured temperatures of nest sites and lizards at a site near the upper elevational limit for oviparous reptiles in southeastern Australia, (2) incubated eggs in the laboratory under thermal regimes that mimicked this natural variation, and (3) measured fitness-relevant phenotypic traits of the resultant hatchlings. MATERIALS AND METHODS Study Species Bassiana duperreyi and Eulamprus heatwolei are scincid lizards, abundant and widely sympatric over much of southeastern Australia (Cogger 2000). Although not closely related (Greer 1989), the two species are similar in body size (80 100 mm snout-vent length [SVL]), time of activity (diurnal), food habits (insectivorous), and thermal preferenda (heliothermic, averaging body temperatures of around 30 C during activity; Shine 1983; Greer 1989). Both species mate in spring, with females ovulating a single clutch of about three to six ova in early summer each year (Pengilley 1972). However, they differ in reproductive mode: B. duperreyi produces parchment-shelled eggs that develop for at least two months in the nest before hatching, whereas female E. heatwolei retain their embryos in utero and give birth to live young. Nest temperatures of B. duperreyi are similar to those of most other squamate reptile species in the Brindabella Range; indeed, communal nests sometimes contain eggs of more than one species (e.g., Shine 1983). Body temperature preferenda of the two species also are similar to each other and to those of most other Brindabella diurnal scincid taxa, averaging close to 30 C (Shine 1983; Greer 1989). This interspecific similarity in selected body temperatures means that the body temperatures of Eulamprus should provide a reasonable approximation to the incubation regimes that would be available to embryos inside a hypothetically egg-retaining Bassiana. Study Area The Brindabella Range 40 km west of Canberra, in the Australian Capital Territory, experiences relatively cold winters (July mean temperature 10.4 C) and mild summers (January mean 25.9 C; Australian Bureau of Meteorology). As with other squamate reptiles, oviparous scincid lizards do not extend to as high an elevation as do viviparous species in the Brindabella Range. Bassiana duperreyi extends to higher elevations than do any other egg-laying scincids in this region, with a major oviposition area at the base of a ski run at 1615 m above sea level (ASL) on Mount Ginini (148 46 E, 35 32 S; for details see Shine et al. 2003; Shine 2004). In contrast, E. heatwolei is distributed from 1000 m ASL all the way up to the top of Mount Ginini ( 1800 m ASL; Pengilley 1972). I monitored thermal regimes of natural nests and of viviparous lizards at the base of Mount Ginini during the period immediately after oviposition by the egg-layers. Field Methods Natural nests of B. duperreyi were located by looking under rocks at the base of the Mount Ginini ski run in December 2002. Thermal data-loggers (Dallas Semiconductor, Dallas, Texas, USA; diameter 15 mm, height 6 mm, mass 3.3 g) set to record temperatures every 10 minutes were placed among the eggs in 15 nests under these rocks, and the cover items carefully replaced. The eggs proceeded to develop normally and hatch successfully in almost all of these clutches. For the current analysis I only used data gathered during the time period when body temperatures of lizards in nearby field

REPTILE VIVIPARITY 1811 enclosures were also monitored (see below); this period occurred 2 weeks after oviposition by B. duperreyi. To quantify body temperatures of viviparous lizards at the same site at the same time, 19 E. heatwolei were captured and fitted with thermochrons (with external casings removed to reduce mass: Robert and Thompson 2003; glued to the dorsal mid-body). These units weighed 1.5 g each, 10% of lizard mass, and had no overt effect on the lizards behavior. The animals were placed in outdoor enclosures close to the nesting area, in circular open-topped arenas measuring 1.3 m diameter with walls 50 cm high. Food and water were provided ad libitum. The study ended prematurely five days later because of the approach of an intense wildfire. The thermochrons were removed at this time, and all lizards were later released unharmed at their original sites of capture. Laboratory Methods Eggs of Bassiana duperreyi were collected from natural nests at another major laying site in the Brindabella Range (Picadilly Circus, 1040 m ASL) in December 2002, about one week after the peak timing of oviposition. The eggs were placed in moist vermiculite for transfer to Clayson IM550R variable-temperature incubators at the University of Sydney. Because of communal oviposition, I could not determine the clutch to which individual eggs belonged. I thus randomly allocated eggs to one of four thermal regimes, designed to mimic the conditions measured in natural nests and lizards during the previous year (data in Shine et al. 2003). The treatments differed in mean incubation temperatures (18 C vs. 20 C) and in the diel distribution of temperatures (i.e., gradual heating and cooling vs. abrupt heating and cooling), so that the four treatments simulated (1) a warm nest (20 C mean, gradual changes in temperature through the day), (2) a cold nest (18 C, gradual changes), (3) a warm female (20 C, abrupt changes) and (4) a cold female (18 C, abrupt changes). This design was achieved in as realistic a way as possible, by keeping the retained eggs at a temperature close to mean selected body temperatures of E. heatwolei (30 C) during daylight hours (4 h for 20 C treatment, 3 h for 18 C treatment) falling to 13.3 C or 12.9 C overnight respectively. In contrast, the nest eggs followed smooth diel cycles of heating and cooling (up to 30 C by day, down to 13.3 C or 12.9 C overnight: see Fig. 1). Data-loggers within each incubator confirmed that actual mean temperatures deviated from the nominal values by 0.2 C. Because embryonic development in B. duperreyi ceases at temperatures below 16.5 C (Shine and Harlow 1996), the mean effective incubation temperature in terms of embryogenesis may differ from the arithmetic mean value. Andrews (2000) suggested a simple method to correct for this minimum, by replacing all temperatures below the developmental minimum with that developmental minimum value (in the case of B. duperreyi, 16.5 C); this method essentially is equivalent to calculating a degree hours temperature summation above the developmental zero. Calculated in this way from the records of data-loggers inside the incubators, effective mean temperatures for the four experimental treatments were 1.2 2.6 C higher than the arithmetic mean values (19.8 C for the cold nest treatment, 20.6 C for the cold female, 21.2 C for the hot nest, and 21.8 C for the hot female). Eggs were placed in individual 64-ml glass jars for incubation, in vermiculite with 120% water by mass (for details see Shine and Harlow 1996). Plastic food wrap over the top of each jar maintained constant water content within each incubation vial. Incubators were checked daily for hatching, and all hatchlings were immediately removed and measured (SVL, tail, mass). Sex was determined by manual eversion of hemipenes (Harlow 1996). Hatchling B. duperreyi have a red patch of color on the throat when they hatch; this fades with age, and does not reappear until sexual maturity (and then only in males: Greer 1989). When the young lizards were first measured, the intensity of color in this patch was scored on a three-point scale (always by the same observer). The young lizards were maintained individually in boxes (22 13 7 cm) that provided water and a thermal gradient from 20 to 38 C (for details see Shine et al. 1997). At seven days of age, locomotor speeds of the hatchlings were tested on a 1-m raceway in a room kept at 25 C (after 30 min acclimation to this temperature); infrared cells at 25-cm intervals allowed us to measure burst speeds (maxima over 25 cm) as well as mean speeds over 1 m. Each hatchling was run three times, with at least a 10-min rest between successive runs. Statistical analyses were performed on a Macintosh G4 computer using the software programs Statview 5 (SAS Institute 1998) and JMP 5.1 (SAS Institute 2002). Data were checked for conformity to relevant assumptions (normality, variance homogeneity) for statistical tests, and were ln-transformed if necessary to overcome such problems. RESULTS Thermal Regimes of Nests and Lizards Two major patterns were evident, both consistent with the predictions outlined above. (1) Nests were warmer than lizards, although the difference in mean temperature was minor (17.3 C vs. 14.5 C). These arithmetic means may be misleading, because much of the difference involved lower overnight minimum temperatures for lizards than for nests (Fig. 1) and (as noted above) temperatures below 16.5 C are irrelevant for rates of embryogenesis in B. duperreyi. Using Andrews (2000) method to correct for this minimum (i.e., replacing all temperatures 16.5 with a value of 16.5 C), effective mean incubation temperatures were almost identical for nests and lizards (19.41 C vs. 19.51 C). (2) The diel distribution of nest temperatures differed between nests and lizards. Lizards achieved higher temperatures earlier in the morning than did nests, which continued warming throughout the day (Fig. 1). Most lizards maintained relatively stable body temperatures for long periods during the day, but this trend is obscured in Figure 1 by differences among individual animals in the timing of those periods (i.e., more lizards basked early in the morning than in the afternoon, so that mean lizard temperatures declined through the course of the day; Fig. 1). Nonetheless, the end result was that the overall frequency distributions of temperatures (i.e., proportions of records within each category) were unimodal for nests and bimodal for lizards, as predicted (Fig. 2). Inspection of data

1812 RICHARD SHINE FIG. 1. Patterns of diel variation in potential incubation regimes for embryonic lizards. (a) Mean hourly temperatures (and associated SEs) inside the nests of oviparous lizards (Bassiana duperreyi), and body temperatures of viviparous lizards (Eulamprus heatwolei), measured in midsummer (December 2002) near the base of Mount Ginini in the Brindabella Range. Data were taken at 10-min intervals over a five-day period from 15 nests and 19 female lizards. (b) Thermal regimes used in the laboratory for incubation of B. duperreyi eggs, based on records from thermal data-loggers inside incubators over a five-day period. (c) As for (a), except that temperatures have been corrected to effective incubation temperatures by changing all values below the developmental minimum for embryogenesis in B. duperreyi (16.5 C) to that developmental minimum level, because fluctuations in temperature below this level do not influence the rate of embryogenesis. (d) As for (b), except that values have been corrected to effective incubation temperatures as above. for individual nests and individual lizards revealed strong consistencies in thermal distributions, so data were pooled for Figure 2. Effects of Incubation Temperatures on the Phenotypic Traits of Hatchling Lizards Overall, hatching success was 90% in all four treatments (range 91.1 97.0%) and contingency-table analysis revealed no significant treatment effect on this variable ( 2 3.12, df 3, P 0.37). I analyzed data on phenotypic traits using three-factor MANOVA, with the factors being mean incubation temperature (18 C vs. 20 C), the diel distribution of temperatures (gradual vs. abrupt changes in temperature), and the sex of the offspring. Although sex seemed unlikely to be of particular interest in the current study, it influences many phenotypic traits in B. duperreyi (Elphick and Shine 1999) and thus its inclusion removes this source of otherwise confounding variation. The dependent variables were the traits I measured on hatchling lizards. For variables that were highly correlated with offspring SVL (tail length, ln mass), I included ln SVL as a covariate in ANCOVAs on these data. Because one variable (throat color) was ordinal (light-medium-bright), I used ordinal multiple logistic regression using the same independent variables as for the ANOVAs and ANCOVAs (Table 1). However, I treated these three variables (mass, tail length,

REPTILE VIVIPARITY 1813 FIG. 2. Frequency distributions of temperatures inside the nests of oviparous lizards (Bassiana duperreyi), and body temperatures of viviparous lizards (Eulamprus heatwolei), measured in midsummer (December 2002) near the base of Mount Ginini in the Brindabella Range. Data were taken at 10-min intervals over a five-day period from 15 nests and 19 female lizards. throat color) differently in the MANOVA, which requires all included analyses to be of the same design. For this purpose, I calculated relative measures for body proportions (residual scores from general linear regressions of ln mass and tail length against SVL) to generate size-independent descriptors of shape. Although use of residual scores has attracted recent criticism, statistical problems are negligible if the two variables involved are very highly correlated with each other (Weatherhead and Brown 1996; Garcia-Berthou 2001). I treated the throat-color score as a continuous variable for inclusion in the MANOVA; this procedure yielded identical conclusions to that based on logistic regression (Table 1). The MANOVA on these data for hatchlings showed highly significant main effects of mean incubation temperature (Wilk s lambda 0.11, F 7,347 399.79, P ), whether temperature changed gradually versus abruptly over the course of the day (Wilk s lambda 0.88, F 7,347 6.41, P ) and offspring sex (Wilk s lambda 0.56, F 7,347 38.63, P ; see Fig. 3). None of the interaction TABLE 1. Effects of incubation temperatures and sex on developmental rates and phenotypic traits of hatchling Bassiana duperreyi. The table shows results of three-factor statistical tests (ANOVAs, ANCOVAs, logistic regression) incorporating offspring sex as well as means and variances of thermal regimes as factors. Relative scores are based on ANCOVA with ln (snout-vent length) as a covariate; main effects are shown after deletion of nonsignificant interaction terms. The analysis of throat color is based on an ordinal logistic regression, with 2 values shown in the F column. All F-values have df 1,394. Values in bold indicate P 0.05. Mean temperature distribution sex Distribution sex Mean temperature sex Mean temperature distribution Offspring sex Frequency distribution of temperature Mean incubation temperature Trait 0.40 0.37 0.94 0.11 0.71 0.82 0.01 2.61 0.37 0.99 0.11 0.55 0.80 0.003 2.71 0.35 0.73 0.55 0.87 0.53 0.12 0.35 0.03 0.40 0.69 0.67 0.22 0.12 0.16 0.18 1.51 2.40 0.62 0.25 80.20 87.78 71.35 0.0012 0.15 0.22 0.02 10.61 2.05 1.50 5.22 0.08 0.21 2521.35 3.08 56.95 1.62 Incubation period (days) Snout-vent length (cm) Relative tail length Relative mass 0.88 0.42 0.60 0.43 0.02 0.65 0.27 0.62 0.52 0.26 0.19 0.02 0.42 1.25 1.73 5.12 0.57 0.86 0.99 0.20 0.32 0.03 1.63 0.40 0.98 0.98 0.24 0.71 0.001 0.001 1.41 0.75 0.0002 0.34 0.10 15.50 14.43 0.88 0.72 0.79 0.63 0.13 0.07 0.24 22.47 0.001 0.03 11.04 17.09 23.05 4.98 Mass change in first week after hatching Speed (m/sec) over 1 m Speed (m/sec) over 25 cm Throat color

1814 RICHARD SHINE FIG. 3. Effects of incubation temperatures and offspring sex on incubation periods and offspring phenotypes in the scincid lizard Bassiana duperreyi. Eggs were maintained under conditions simulating natural nests (nest, with unimodal thermal distribution) or uterine retention (female, with bimodal thermal distribution), and with mean temperatures in either case of 18 C or20 C. Because mass and tail length were highly correlated with snout-vent length (SVL), the graphs show residual scores (from linear regressions against SVL) for each of these variables. The points represent mean values and associated standard errors.

REPTILE VIVIPARITY 1815 FIG. 4. Effects of the frequency distribution of incubation temperatures (unimodal vs. bimodal) on the intensity of red throat color in hatchlings of the scincid lizard Bassiana duperreyi. terms were significant (all P 0.08). Table 1 shows results from the individual ANOVAs. Mean incubation temperature significantly affected the incubation period, the offspring s relative tail length, and its running speeds over both 1mand 25 cm. The diel distribution of temperature affected incubation periods also, as well as the young lizard s body shape (mass relative to SVL) and its throat color. This highly significant latter effect was due to higher proportions of lizards with bright red throats from the lizard temperature regime than from the nest temperature regime (Fig. 4). Mean nest temperature also exerted a smaller but nonetheless significant effect on this variable (cold incubation resulted in more intense throat color; Table 1). Male and female hatchlings differed significantly for mean values of all morphological traits and both locomotor speeds that were measured (Table 1), confirming the value of incorporating this factor into the analyses. Although throat colors did not differ overall between males and females, the diel distribution of incubation temperatures affected throat color differently in the two sexes (interaction term in Table 1; sons had more intense color from the maternal than the nest treatment, whereas daughters were less affected). All hatchlings lost mass over the first week of life, but the amount lost was greater for animals from the higher mean incubation temperature (Table 1). Log-likelihood ratio tests from a multiple logistic regression with incubation mean temperature and diel distribution as independent variables, and hatchling sex as the dependent variable, shows that the proportion of sons was significantly lower from 20 C incubation than from 18 C incubation (46.3% vs. 58.1%; 2 5.13, df 1, P 0.025) but was not affected by whether temperatures changed gradually or abruptly during the day (53.45% vs. 51.37%; 2 0.17, df 1, P 0.68), nor by any significant interaction between these factors ( 2 1.57, df 1, P 0.21). Effects of Arithmetic versus Effective Mean Incubation Temperatures The above analyses used arithmetic means to characterize incubation temperature treatments. One effect of the differing diel distributions of nest versus female body temperatures (e.g., Fig. 2) is to generate slight differences in the proportions of time spent above developmental zero (16.5 C) and hence, in effective mean temperatures. Could the effects of differing diel distributions on phenotypic traits of hatchling lizards (as shown above) be due entirely to divergences in effective mean temperature among treatments? To answer this question, I added effective mean temperature (which took values 1.2 2.6 C higher than the arithmetic means; see above) as a covariate in a one-factor MANCOVA with diel distribution (nest vs. female) as the factor, and the same dependent variables as above. If the experimental effects on hatchling phenotypes were due entirely to the slight differences in effective mean temperatures, such effects should disappear in this analysis. In fact, the treatment effects remained strong: the effect of diel thermal distribution in the overall MANOVA was highly significant (Wilk s lambda 0.42, F 7,347 69.53, P ), as were treatment effects on several of the component variables (P 0.05 for relative tail length, throat color, incubation period and running speeds over both 25 cm and 1 m). Thus, the differences in hatchling phenotypes among experimental treatments are not attributable simply to differences in effective mean temperatures. Correlations among Hatchling Traits Correlations among traits may clarify their significance or mechanism of action. Inevitably, many of the traits measured in the present study were significantly correlated with each other. Most of these correlations were easily explicable: either between morphological variables, or between morphology and speed. However, the intensity of throat color bears no intuitive relationship to other traits, and thus warrants closer examination. An ordinal logistic regression with the hatchling s throat color as the dependent variable showed that color was related both to the animal s body shape (mass relative to SVL: 2 9.72, df 1, P 0.002) and to its running speed over 1 m ( 2 12.23, df 1, P 0.001) but not to any interaction between these factors ( 2 0.68, df 1, P 0.41). Lizards with more brightly colored throats tended to be thinner-bodied, and to run more slowly than their siblings. DISCUSSION Measurements of thermal regimes inside natural nests, and body temperatures of lizards in nearby field enclosures, provide compelling evidence that the evolution of viviparity will cause a shift in the diel distribution of incubation temperatures experienced by embryos (Figs. 1 and 2). Mean temperatures also may shift, but the magnitude of this latter difference sometimes may be relatively small and will vary with elevation (Fig. 2; see also Andrews 2000). Hence, it is worth investigating the possibility that changes in the diel distribution of incubation temperatures, independent of any change in mean temperature, might have biologically significant effects. If offspring phenotypes were substantially modified by the diel distribution of incubation thermal regimes, then any analysis of the possible consequences of prolonged uterine retention of eggs for offspring fitness would need to consider this effect. Although many previous

1816 RICHARD SHINE studies have mentioned maternal thermoregulation at stable as well as high temperatures as a reason for viviparously retained embryos to be kept warmer than eggs laid in a nest, the possibility that offspring fitness may be directly affected by shifts in diel distribution (rather than mean temperature) seems never to have been addressed. My results on the thermal profiles of nests compared to pregnant female lizards closely accord with those of previous studies by Andrews (2000; based on her studies of phrynosomatid lizards in the western USA) and myself (Shine et al. 2003; based on the same study system as used for the present paper). All three datasets suggest that uterine retention of eggs is likely to have less effect on mean incubation temperatures than on diel variance. Nonetheless, it is impossible to judge the generality of this pattern without data on much longer timescales and in many more systems. For example, mean incubation temperatures at Mount Ginini were about 5 C higher in December 2001 than in December 2002 (compare Figs. 1 and 2 to data in Shine et al. 2003). In the Brindabella Range, the disparity between nest and maternal body temperatures also shifts with elevation (Shine et al. 2003). Thus, there may well be many situations in which uterine retention of eggs does increase mean incubation temperature, as proposed under the usual version of the cold-climate hypothesis. Importantly, however, the diel distribution of temperatures will also shift in such situations. Indeed, such shifts seem inevitable in any heliothermic species with active behavioral thermoregulation, because pregnant females will remain at high and relatively constant temperatures during daylight hours whereas nests will heat and cool on a slower, smoother cycle. My data show this effect clearly, with a peak in lizard (but not nest) temperatures at around 30 C, close to the mean body temperatures selected by active Eulamprus in previous studies (Fig. 2; Shine 1983; Greer 1989). However, the thermal data also reveal a less intuitively obvious effect: lizards exhibited lower temperatures at night than did nests (Fig. 1). There are several possible reasons for this effect (e.g., limited access to warm overnight retreats in the outdoor enclosures), but one possibility is that lizards actively seek cold sites at night. Low nocturnal temperatures would reduce metabolic rates and thus energy expenditure, but (because all potential retreat sites remain well below the thermal minimum for embryogenesis) would have no effect on rates of embryonic development. Thus, females may maximize embryonic development but minimize energy costs by selecting hot sites by day and cold sites by night. The present study was designed to evaluate the relative sensitivity of hatchling phenotypes to mean incubation temperature versus diel thermal variance. The clear result from Table 1 is that hatchling phenotypes are sensitive both to the mean and variance of incubation temperature. Some traits (e.g., incubation period, throat color) were affected by both aspects of the thermal regime, whereas others were affected by either one or the other. For example, mean incubation temperature affected relative tail length, post-hatching growth, and locomotor speeds, whereas diel thermal distributions modified body mass relative to SVL (Table 1). The sex ratios of hatchling B. duperreyi also shifted significantly with mean incubation temperature, with a higher proportion of sons from lower-temperature incubation. The results for effects of mean incubation temperature support and extend those of previous studies on the same species. For several traits, this previous work relied upon comparisons between hatchlings from incubation at more extreme temperatures (Shine 1995, 2002b, 2004; Shine and Harlow 1996; Shine et al. 1997; Shine and Elphick 2001). The present work reveals similar shifts in hatchling phenotypes as a result of relatively small thermal differences in incubation conditions, of a magnitude likely to be encountered between sympatric oviparous and viviparous reptiles in nature. One of the most interesting traits to shift in response to thermal conditions was the sex ratio of hatchlings, with a preponderance of males hatching from lower-temperature incubation (i.e., mean of 18 C vs. 20 C). This effect is unlikely to reflect sex differences in hatching success (Burger and Zappalorti 1988), because 90% of eggs hatched in all treatments. Instead, it seems that incubation temperature can override the heteromorphic sex chromosomes of this species to decouple phenotypic and genetic sex in Bassiana duperreyi (Shine et al. 2002). Effects of incubation temperature were generally similar for males and females (Fig. 3), but the two sexes differ for many traits and hence the sex-ratio shift with incubation temperature means that it is important to include sex as a factor in any statistical analysis of the effects of incubation regimes on phenotypic traits of hatchlings (Table 1). My data do not allow any definitive conclusions about the adaptive significance of shifts in mean temperature versus diel distributions; this will depend on factors such as the duration of effects and the intensity of selection on those specific traits. Based simply on the numbers of traits affected and the magnitude of those effects (see F and P values in Table 1), a shift in mean temperature of 2 C had more effect than did a shift from gradual to abrupt diel changes in incubation temperature. Importantly, however, both aspects of the incubation regime (i.e., diel distribution as well as mean temperature) modified hatchling phenotypes. Although the microevolutionary consequences of such shifts remain speculative, they may well influence hatchling fitness. For example, the incubation period (and thus, time of hatching) was influenced by both the mean and the variance of incubation temperature, and small differences in the date of hatching may have major fitness consequences in such a cold-climate environment (Ferguson and Fox 1984; Verhulst and Tinbergen 1991; Sinervo and Doughty 1996; Olsson and Shine 1997). The shift from nest to maternal regimes (i.e., to more abrupt changes in incubation temperature) expedited hatching and increased offspring size (Fig. 3), consistent with adaptationist hypotheses. The proximate mechanisms that generate phenotypic shifts with incubation regimes remain unknown. One of the most intriguing results from the current study involves the correlation between hatchling throat color and other traits (running speed, body shape). Incubation effects on color patterns have been documented in other reptile species (Deeming 2004) and color polymorphisms occur in many lizard populations (Sinervo and Lively 1996; Sinervo et al. 2000). Nonetheless, we need more information before we can interpret the phenomenon of throat-color variation in hatchling

REPTILE VIVIPARITY 1817 Bassiana. It may simply be a nonadaptive indication of underlying hormonal variation, or may possess adaptive significance in its own right. Either explanation is potentially consistent with the link between throat color and other phenotypic traits (body shape, running speeds) revealed by the present study. In summary, my data support hypotheses that (1) the evolution of viviparity modifies thermal regimes for egg incubation in reptile species in cold climates, (2) such changes influence phenotypic traits of hatchlings, and thus (3) potentially, this sensitivity might exert selection for prolongation of uterine retention of the developing eggs. However, contrary to conventional wisdom on this topic, (4) the aspect of incubation regimes that will differ most consistently between eggs in the nest versus in utero (and that, in turn, can influence hatchling phenotypes) will be the diel distribution of temperatures rather than mean incubation temperature. Hence, the variance rather than the mean of incubation temperature may constitute a significant selective force for prolongation of uterine retention of developing eggs in cold-climate squamate reptiles. Shifts in diel distributions of incubation temperatures with uterine retention would be expected a priori, and such differences have been shown by previous studies as well as the current work (Andrews 2000; Shine et al. 2003). Why, then, has the extensive published literature on selective forces for the evolution of reptilian viviparity ignored this aspect? The answer probably lies in a more general phenomenon, whereby biologists focus on mean values for traits rather than associated variances (Georges et al. 1994; Shine and Seigel 1996). The other general issue that arises is that tests of adaptationist hypotheses may need to incorporate substantial natural-history detail; simplifying assumptions (such as the primacy of mean temperature over variance) may lead to significant errors. ACKNOWLEDGMENTS I thank M. Elphick and E. Barrott for assistance in the field and the laboratory. The work was funded by the Australian Research Council. LITERATURE CITED Andrews, R. M. 2000. Evolution of viviparity in squamate reptiles (Sceloporus spp.): a variant of the cold-climate model. J. Zool. 250:243 253. Blackburn, D. G. 1982. Evolutionary origins of viviparity in the Reptilia. I. Sauria. Amphib.-Reptilia 3:185 205.. 1985. Evolutionary origins of viviparity in the Reptilia. II. Serpentes, Amphisbaenia, and Icthyosauria. Amphib.-Reptilia 6: 259 291.. 1999. Are viviparity and egg-guarding evolutionarily labile in squamates? Herpetologica 55:556 573.. 2000. Reptilian viviparity: past research future directions, and appropriate models. Comp. Biochem. Physiol. A: Mol. Integ. Physiol. 127:391 409. Blackburn, D. G., and L. J. Vitt. 2002. Specializations of the chorioallantoic placenta in the Brazilian scincid lizard, Mabuya heathi: a new placental morphotype for reptiles. J. Morphol. 254: 121 131. Burger, J., and R. T. Zappalorti. 1988. Effects of incubation temperatures on sex ratios in pine snakes: differential vulnerability of males and females. Am. Nat. 132:492 505. Calder, W. A. 1984. Size, function and life history. Harvard Univ. Press, Boston. Charnov, E. L. 1993. Life history invariants. Oxford Univ. Press, Oxford, U.K. Cogger, H. G. 2000. Reptiles and amphibians of Australia. 6th ed. Reed New Holland, Sydney. Deeming, D. C. 2004. Post-hatching phenotypic effects of incubation in reptiles. Pp. 229 251 in D. C. Deeming, ed. Reptile incubation. Nottingham Univ. Press, Nottingham, U.K. Elphick, M. J., and R. Shine. 1999. Sex differences in optimal incubation temperatures in a scincid lizard species. Oecologia 118:431 437. Ferguson, G. W., and S. F. Fox. 1984. Annual variation of survival advantage of large juvenile side-blotched lizards, Uta stansburiana: its causes and evolutionary significance. Evolution 38: 342 349. Garcia-Berthou, E. 2001. On the misuse of residuals in ecology: testing regression residuals vs. the analysis of covariance. J. Anim. Ecol. 70:708 711. Georges, A., C. J. Limpus, and R. Stoutjesdijk. 1994. Hatchling sex in the marine turtle Caretta caretta is determined by proportion of development at a temperature, not daily duration of exposure. J. Exp. Zool. 270:432 444. Greer, A. E. 1989. The biology and evolution of Australian lizards. Surrey Beatty and Sons, Chipping Norton, Australia. Harlow, P. S. 1996. A harmless technique for sexing hatchling lizards. Herpetol. Rev. 27:71 72. Harvey, P. H., and M. D. Pagel. 1991. The comparative method in evolutionary biology: Oxford studies in ecology and evolution. Oxford Univ. Press, Oxford, U.K. Lee, M. S. Y., and R. Shine. 1998. Reptilian viviparity and Dollo s Law. Evolution 52:1441 1450. Mell, R. 1929. Preliminary contributions to an ecology of East Asiatic reptiles, especially snakes. Lingnan Sci. J. 8:187 197. Olsson, M., and R. Shine. 1997. The seasonal timing of oviposition in sand lizards (Lacerta agilis): why earlier clutches are better. J. Evol. Biol. 10:369 381. Pengilley, R. 1972. Systematic relationships and ecology of some lygosomine lizards from southeastern Australia. Ph.D. diss. Australian National University, Canberra. Qualls, C. P., R. Shine, S. Donnellan, and M. Hutchinson. 1995. The evolution of viviparity within the Australian scincid lizard Lerista bougainvillii. J. Zool. (Lond.) 237:13 26. Robert, K. A., and M. B. Thompson. 2003. Reconstructing Thermochron ibuttons to reduce size and weight as a new technique in the study of small animal thermal biology. Herpetol. Rev. 34: 130 132. Roff, D. A. 1992. The evolution of life histories. Chapman and Hall, New York. SAS Institute. 1998. Statview 5. SAS Institute, Cary, NC.. 2002. JMP 5.1. SAS Institute, Cary, NC. Sergeev, A. M. 1940. Researches in the viviparity of reptiles. Moscow Soc. Nat.(Jubilee Issue):1 34. Shine, R. 1983. Reptilian viviparity in cold climates: testing the assumptions of an evolutionary hypothesis. Oecologia 57: 397 405.. 1985. The evolution of viviparity in reptiles: an ecological analysis. Pp. 605 694 in C. Gans and F. Billett, eds. Biology of the Reptilia. Vol. 15. John Wiley and Sons, New York.. 1995. A new hypothesis for the evolution of viviparity in reptiles. Am. Nat. 145:809 823.. 2002a. Reconstructing an adaptationist scenario: What selective forces favor the evolution of viviparity in montane reptiles? Am. Nat. 160:582 593.. 2002b. Eggs in autumn: responses to declining incubation temperatures by the eggs of montane lizards. Biol. J. Linn. Soc. 76:71 77.. 2004. Seasonal shifts in nest temperature can modify the phenotypes of hatchling lizards, regardless of overall mean incubation temperature. Funct. Ecol. 18:43 49. Shine, R., and M. J. Elphick. 2001. The effect of short-term weather fluctuations on temperatures inside lizard nests, and on the phenotypic traits of hatchling lizards. Biol. J. Linn. Soc. 72: 555 565. Shine, R., and P. S. Harlow. 1996. Maternal manipulation of off-

1818 RICHARD SHINE spring phenotypes via nest-site selection in an oviparous lizard. Ecology 77:1808 1817. Shine, R., and M. S. Y. Lee. 1999. A reanalysis of the evolution of viviparity and egg-guarding in squamate reptiles. Herpetologica 55:538 549. Shine, R., and R. A. Seigel. 1996. A neglected life-history trait: clutch-size variance in snakes. J. Zool. 239:209 223. Shine, R., M. J. Elphick, and P. S. Harlow. 1997. The influence of natural incubation environments on the phenotypic traits of hatchling lizards. Ecology 78:2559 2568. Shine, R., M. J. Elphick, and S. Donnellan. 2002. Co-occurrence of multiple, supposedly incompatible modes of sex determination in a lizard population. Ecol. Lett. 5:486 489. Shine, R., M. Elphick, and E. G. Barrott. 2003. Sunny side up: lethally high, not low, temperatures may prevent oviparous reptiles from reproducing at high elevations. Biol. J. Linn. Soc. 78: 325 334. Sinervo, B., and P. Doughty. 1996. Interactive effects of offspring size and timing of reproduction on offspring reproduction: experimental, maternal, and quantitative genetic aspects. Evolution 50:1314 1327. Sinervo, B., and C. M. Lively. 1996. The rock-paper-scissors game and the evolution of alternative male strategies. Nature 380: 240 243. Sinervo, B., E. Svensson, and T. Comendant. 2000. Density cycles and an offspring quantity and quality game driven by natural selection. Nature 406:985 988. Tinkle, D. W., and J. W. Gibbons. 1977. The distribution and evolution of viviparity in reptiles. Misc. Publ. Univ. Mich. Mus. Zool. 154:1 55. Verhulst, S., and J. M. Tinbergen. 1991. Experimental evidence for a causal relationship between timing and success of reproduction in the great tit Parus m. major. J. Anim. Ecol. 60:269 282. Weatherhead, P. J., and G. P. Brown. 1996. Measurement versus estimation of condition in snakes. Can. J. Zool. 74:1617 1621. Weekes, H. C. 1933. On the distribution, habitat and reproductive habits of certain European and Australian snakes and lizards, with particular regard to their adoption of viviparity. Proc. Linn. Soc. NSW 58:270 274. Corresponding Editor: K. Schwenk