1Mathematical models suggest that ‘costs of reproduction’ (decrements in an organism's probable future reproductive output due to investing in current reproduction) are major determinants of life-history evolution.
2Pregnancy decreases locomotor performance in many taxa, and could render reproducing females more vulnerable to predators. To evaluate the importance of this ‘cost of reproduction’, however, we need to know the magnitude of performance decrement induced by reproduction relative to that induced by other factors in the animal's biology.
3Studies on adult female skinks (Lampropholis guichenoti) confirm that pregnancy significantly impairs maternal locomotion, but also show that factors such as a moderate decrease in body temperature, a large meal or loss of the tail reduced locomotor speeds even more than did pregnancy. Thus, reproductive state probably causes only a minor proportion of the total temporal variation in a female skink's locomotor ability.
4In such a system, even a large effect of pregnancy on running speeds may not impose a significant selective pressure on reproductive investment.
5Climate and foraging modes may affect the degree to which locomotor speeds are influenced by pregnancy vs. other factors, offering a potential explanation for the lower overall reproductive investment per clutch in tropical vs. temperate-zone reptiles, and in lizards vs. snakes.
Although the idea of ‘costs of reproduction’ is simple and powerful, there is unlikely to be any simple relationship between the magnitude of maternal locomotor impairment and the magnitude of resulting selective forces on reproductive investment. For example, running speeds of the prey may have little impact on their vulnerability to some types of predators (Schwarzkopf & Shine 1992; Downes & Shine 2001); and reproducing animals can reduce their movements and stay close to shelter, thereby increasing their probability of evading predators and thus lowering the survival cost of reproduction (Bauwens & Thoen 1981; Brodie 1989; Cooper et al. 1990; Brana 1993; Weiss 2001). Although reduction of locomotor performance due to ‘pregnancy’ is widespread, such complications make it difficult to predict the resulting intensity of selection (if any) for reduced reproductive investment.
In this paper I address a related issue that may be even more significant for the degree to which maternal locomotor impairment constitutes a selective force on reproductive investment. To relate the magnitude of such performance decrements to the magnitude of ‘costs of reproduction’, we need to place the reproduction-induced decrement in a female's locomotor speed within the context of other factors that modify her speed. If factors other than reproduction exert little effect on locomotor ability, then the influence of pregnancy in this respect may well be significant. That is, if much of the variance in an individual's locomotor performance over some biologically meaningful time period is attributable to its reproductive state, ‘costs of reproduction’ may be significant selective forces on reproductive investment. On the other hand, it is possible to envisage a situation in which reproductive state is only one of several factors that affect a female's speed, and in which the magnitude of reproductive effects is smaller than those due to other factors. In the latter case, reproductive state generates only a small proportion of the observed variation in locomotor performance, and would be unlikely to engender a major selective force.
To clarify this question, I measured running speeds in lizards, and quantified the proportional decrement due to pregnancy compared to that induced by other factors (feeding; tail loss; variation in body temperatures) also likely to influence locomotion in these animals.
Materials and methods
Garden Skinks, Lampropholis guichenoti, are small (to 70 mm total length) oviparous scincid lizards that are abundant in coastal and montane regions of eastern Australia (Cogger 2000), including in highly modified suburban habitats (Swan 1990). Females lay one to three clutches in the warmer months of the year (Shine 1980; Joss & Minard 1985); clutches contain one to five eggs (Qualls 1996; Qualls & Shine 1997). Relative clutch masses (clutch mass divided by maternal body mass) vary spatially and temporally, but generally average around 0·25–0·30 (Pengilley 1972; Qualls 1996).
capture and husbandry
From November 2001 to January 2002, Lampropholis guichenoti were collected by hand from suburban gardens in Sydney. Sex of the skinks was determined by manual eversion of hemipenes; only adult females (>35 mm snout–vent length, SVL) were retained. Captive lizards were maintained in large outdoor enclosures (2·5 × 1 m2) containing ad libitum water, shade and basking sites. Food (live crickets) was provided twice weekly. On 11 January 2002 the lizards were divided into animals that were clearly gravid (based on eggs evident in abdominal palpation; n = 11), those that were clearly non-gravid (no detectable eggs or ova; n = 21), and those of uncertain status (n = 6). Animals in the two former groups were transferred to separate plastic lunchboxes (22 × 13 × 8 cm3) containing a 1-cm substrate of commercial potting mix, a plastic shelter and a water dish. These cages were then arranged on timer-controlled heating racks allowing the skinks to control their own body temperatures (substrates within each cage ranged from 22·65 ± 0·34–32·0 ± 1·14 °C). Heating was provided from 0800 to 1700 hours. Outside these hours, cage temperatures fell to ambient temperature (20 °C). Timer-controlled lighting in the room provided a light : dark photoperiod of 12L : 12D (lights on 0700–1900 hours). Lizards were fed four to eight crickets, depending on cricket size (10–13 mm long × 3–5 mm wide, approx. 0·044–0·095 g each) twice weekly, and water was provided ad libitum.
Three days after the lizards were last fed, the SVL and total length of each lizard (±1 mm) was measured, as was its mass (±0·001 g). Additionally, digital callipers (±0·1 mm) were used to measure axilla–groin length, and the body width and depth at each of three points along the body (at the pectoral girdle, the pelvic girdle and halfway between). These measurements were later used to estimate mean body widths and depths, and thus the total trunk (i.e. chest plus abdomen) volume of each lizard.
Locomotor performance of the gravid and non-gravid females was assessed immediately after they were measured (above), as follows. The lizards were placed in individual plastic containers and left to acclimate undisturbed for at least 30 min at the test temperature prior to the locomotor trials. They were then transferred directly from their containers to the starting area of the 4·5 cm-wide raceway and allowed to run 1 m along the raceway. Animals that refused to run were lightly touched on the tail with an artist's paintbrush. Infrared photocells located at 25-cm intervals along the runway recorded the cumulative time taken for lizards to cross each successive infrared beam. In each of the body temperature trials, lizards were raced three times, with a rest period of at least 10 min between each run. I used these data to calculate a mean running speed over 1 m for each female. Upon completion of these trials, lizards were returned to their original cages and fed.
Three separate experiments were conducted to examine influences on body shapes and locomotor speeds:
1Thermal experiment– effect of body temperature on locomotor speeds. The lizards were tested (as above) at body temperatures of 20, 25 and 30 °C.
2Feeding experiment– effect of a full stomach on body shape and locomotor speeds. Half of the skinks in each group (gravid and non-gravid) were given a large meal of crickets (8–10 crickets per lizard), and left undisturbed to eat their fill. Three hours later, these animals plus the remaining (unfed for 3 days) lizards were scored for morphology and running speed (at 25 °C only) as described above.
3Tail loss experiment– effect of tail autotomy on locomotor speeds. The lizards were again split into two groups, orthogonal to the division into fed and unfed animals in the experiment described above. The first group of lizards had tail loss induced close to the base of the tail, by holding the tail firmly until the lizard voluntarily autotomized it. These animals were then returned to their cages and left undisturbed for 24 h, after which I tested locomotor performance at 25 °C on both of these groups, and the other (similarly disturbed, but not autotomized) group.
sequence of experiments and statistical analyses
All experiments were run over an 8-day period (18–25 January 2002). Morphological data and temperature effects were quantified on day 1, feeding effects on day 5 and tail-autotomy effects on day 8. Thus, gravid females did not shift appreciably in their degree of burdening over this period. All gravid females (but no animals classed as ‘non-gravid’) oviposited < 14 days after conclusion of the trials. Data were tested for conformity to statistical assumptions; all variables met these assumptions after ln-transformation. Two-factor anovas were used to analyse the data, with reproductive state as one factor and the other as the manipulated variable (temperature, feeding status or tail loss). Because the same females were used in each of these three sets of experiments, the data are not independent in this sense. However, the experimental design was orthogonal in terms of allocation of individuals – that is, there was no confounding between temperature treatments (all individuals tested), feeding (half fed, randomly chosen) and tail loss (half autotomized, equally split between the previously ‘fed’ and ‘unfed’ groups). I did not apply Bonferroni ‘corrections’ to the resulting analyses because of subjectivity in this procedure, especially in selecting suites of ‘related’ tests (Cabin & Mitchell 2000). For the present study, whether or not specific results fall above or below the conventional level of statistical significance (i.e. P < 0·05) is less important than the relative magnitude of different effects on maternal locomotor speeds. I thus examined the relative degree to which morphological and performance traits were influenced by pregnancy vs. other factors.
The trials commenced with 11 gravid lizards and 21 non-gravid females, but two lizards oviposited partway through the study and thus, sample sizes differed slightly among experiments.
Table 1 provides morphological data on these animals, and shows that the two groups exhibited similar body lengths (SVL and axilla–groin), but that gravid females averaged slightly (8%) heavier, and had significantly wider (13%) and deeper (7%) abdomens. Trunk volumes of gravid lizards averaged 26% higher than those of non-reproductive animals (Table 1). On visual inspection, the gravid animals were noticeably distended by their eggs; in some cases, the outlines of individual eggs were clearly evident.
Table 1. Morphological traits of gravid and non-gravid adult female skinks (Lampropholis guichenoti) measured 3 days postfeeding. The table shows mean values (SD in parentheses) and results (F and P-values) of one-factor anovas (1,30 df) comparing gravid vs. non-gravid lizards. ‘Trunk’= body segment between the insertion points of fore and hind limbs
Snout–vent length (mm)
Axilla–groin length (mm)
Mean body width (mm)
Mean body depth (mm)
Trunk volume (mm3)
These data were analysed using two-factor anova, with body temperature and reproductive state as the factors. At each of the three temperatures, gravid lizards ran about 18% slower than non-gravid conspecifics (Fig. 1a). A decrease in body temperatures from 30 to 20 °C caused a decline in speed of 27% in gravid skinks, and 38% in non-gravid animals (Fig. 1a). Running speeds of the lizards were substantially affected both by body temperature (F1,26 = 8·27, P < 0·001) and by pregnancy (F1,26 = 6·32, P < 0·02), with no significant interaction between the two factors (F1,26 = 0·05, P = 0·96; see Fig. 1a).
tail loss experiment
The tails of experimentally manipulated lizards averaged only 4·9 (SD = 1·2) mm after autotomy, whereas tails of the control lizards averaged 41·3 (SD = 15·7) mm. Tail loss reduced speeds of non-gravid lizards by approximately 31% (0·44 vs. 0·33 m s−1: Fig. 1b). A two-factor anova on running speeds revealed a significant interaction term between the effects of pregnancy and tail loss (F1,25 = 7·11, P < 0·015). This interaction reflected a decrease in speeds following autotomy in non-gravid lizards, vs. an increase in gravid animals (Fig. 1b). Because of the significant interaction, I repeated the analyses separately for gravid and non-gravid animals. The increment in running speeds for gravid lizards was not statistically significant (F1,6 = 1·16, P = 0·32), but tail loss significantly reduced speeds of non-gravid animals (F1,19 = 10·22, P < 0·005).
A large meal substantially changed both the body mass and the body shape of lizards. Non-gravid skinks that were allowed to feed ad libitum after 3 days without food, gained an average of 21% in mass, 9% in body width, 17% in body depth and 26% in trunk volume (Fig. 2). Mean proportional increments due to feeding were very similar for gravid animals (17, 8, 15 and 26%, respectively) as for their non-gravid conspecifics. That is, pregnancy caused a greater lateral distension of the abdomen than did feeding, but a large meal added more to a lizard's mass and abdominal depth than did the clutch of eggs. In consequence, the total increase in trunk volume due to pregnancy was similar to that seen after ingestion of a large meal. I analysed these data using two-factor anovas with reproductive condition and feeding status as the independent variables (see Fig. 2). The body masses of female skinks were significantly greater after feeding (F1,26 = 8·00, P < 0·01) but did not differ between gravid and non-gravid animals (F1,26 = 1·50, P = 0·23) nor was there any significant interaction between these two factors (F1,26 = 0·04, P = 0·85). Virtually identical results were obtained for mean body depth (for feeding, F1,26 = 29·53, P < 0·0001; for pregnancy, F1,26 = 2·58, P = 0·12; interaction, F1,26 = 0·19, P = 0·67). Mean body width was greater in gravid lizards (F1,26 = 12·21, P < 0·002) and in recently fed animals (F1,26 = 13·11, P < 0·002), with no significant interaction (F1,26 = 0·001, P= 0·98). Ingestion of a large meal reduced mean speeds by 19% in non-gravid lizards and 36% in gravid lizards (Fig. 2). Thus, locomotor speeds of the lizards were reduced after feeding (F1,26 = 4·76, P < 0·04) but showed no significant effect of pregnancy (F1,26 = 1·83, P = 0·18) nor any significant interaction between these factors (F1,26 = 0·39, P = 0·53).
Pregnancy reduced locomotor speeds of Garden Skinks by about the same amount as did a large meal, a temperature decrease of about 5 °C, or autotomy of part of the tail. Lampropholis guichenoti are often active at relatively low body temperatures (e.g. Shine 1983; Greer 1989), and the lizards are vulnerable to predators even at night within their retreat-sites (nocturnally active snakes are common, e.g. Downes & Shine 2001). Thus, a high proportion of potential encounters with predators must occur when the lizards’ body temperatures are so low as to substantially reduce their locomotor performance. Similarly, tail loss is frequent in this species (67% of adults exhibit regrown tails: Downes & Shine 2001) and the lizards frequently consume large meals relative to their own body mass (Greer 1989). Indeed, when collecting skinks for the present study, many apparently ‘gravid’ lizards proved to be non-gravid animals with their alimentary tracts grossly distended with prey. Thus, all of the factors that contributed to reduced locomotor speeds of lizards in this study are likely to be experienced frequently by Garden Skinks in the wild. In consequence, much of the total variation in locomotor performance of a female Lampropholis over any meaningful time period in the field will be induced by these factors rather than simply by the animal's reproductive condition.
This conclusion has strong implications for the degree to which variations in reproductive investment (e.g. relative clutch mass) in Lampropholis guichenoti are subject to natural selection based on their effects on locomotor performance. Two issues are important here.
First, the intensity of selection via this pathway will be limited by the proportion of locomotor-speed variation engendered by pregnancy vs. that engendered by other traits (Falconer 1981). If factors such as wide fluctuations in body temperature generate most of the observed variation in locomotor speeds of female lizards, then mobility-based selection for reduced RCM will probably be weak. Imagine two lizard populations with identical survival schedules, with all mortality due to predators. In both populations speed determines survival, and pregnancy lowers speed. In one population, all of the variance in speed is due to pregnancy; in the other, pregnancy is only one of several factors (such as temperature, feeding, autotomy) that reduce maternal speeds. Then, a 20% reduction in reproductive investment will reduce the locomotor burden for pregnant females equally in the two populations, but will generate a much higher benefit (overall survival increment) in the former situation than in the latter one. Thus, costs of reproduction will more strongly favour a lower reproductive investment if pregnancy is the only (or the main) aspect that curtails locomotor speeds.
Second, a population in which all individuals exhibit very wide temporal variation in locomotor speeds (regardless of causal factors) will be under strong selection to minimize the ‘costs’ (in vulnerability to predation, etc.) of that locomotor variation (cf. Pough 1980). That is, other aspects of their biology will become adapted to the fact that these animals often cannot run very fast. For example, many reptiles shift their antipredator tactics with body temperature (rely on display or crypsis when cold but flee if they are warm enough for rapid locomotion: Greene 1988; Passek & Gillingham 1997; Shine et al. 2000), or remain close to shelter if locomotor capacity is reduced (Bauwens & Thoen 1981). Such behavioural adaptations will further reduce the degree to which decrements in locomotor performance translate into increments in mortality – and in turn, will reduce the intensity of selection on reproductive investment mediated via maternal mobility.
These ideas suggest that although pregnancy substantially reduces locomotor speeds of Lampropholis guichenoti, that performance decrement is not likely to generate strong selection for reduced RCM because it is only one among several such influences. How general is this result? The answer to this question will depend upon interspecific variation in (1) the magnitude of locomotor decrements due to pregnancy vs. other factors, and (2) the frequency with which locomotor performance is affected by factors other than pregnancy.
magnitude of locomotor decrements
The above conclusions will not apply to a species in which locomotor speeds are affected much more by pregnancy than by events such as tail loss, feeding and thermal fluctuations. However, a review of published literature on lizard locomotion suggests that L. guichenoti are fairly typical in these respects:
2Effects of feeding. I found decreases of 19–36%; Huey et al. (1984) reported that recently fed lizards were 33% slower than unfed conspecifics. Endurance capacity may also decline significantly after feeding (Huey et al. 1984; Garland 1983). Long-term starvation did not affect sprint speeds in another lizard taxon (Speedy & Mumme 1994).
3Effects of tail loss. The influence of autotomy on speed presumably depends upon the amount of tail lost, the time since autotomy and interspecific differences in the role of the tail during locomotion (Zani 1996). Tail loss in L. guichenoti caused a mean speed decrease of 12–15% averaged over the first 40 days postautotomy (Downes & Shine 2001), compared with a 31% decrease immediately postautotomy in the current study. Speed decrements after autotomy can be even higher than my own results (36% in Ballinger et al. 1979; 32% and 42% in Punzo 1982; 48% in Martin & Harvey 1998). Sometimes, however, tail loss does not affect running speeds (Huey et al. 1990), and some species actually run faster after autotomy than before (Daniels 1983; Brown et al. 1995).
4Thermal shifts. A decrease in body temperature from 30 to 20 °C decreased L. guichenoti locomotor speeds by 32%. Previous studies have revealed even larger decrements over the same temperature range (mean of 12 species = 68·8%[range 26–102%] in Huey & Bennett 1987; mean of 5 species = 60%[range 47–71%] in Huey et al. 1989; 66 and 72% in Van Damme et al. 1989a).
In summary, the proportional decrements in locomotor speeds due to pregnancy, tail loss and lowered body temperature are broadly similar in L. guichenoti to those reported in other lizard species. Further data are needed on effects of a full stomach on running speeds before we can make a robust comparison for this factor.
frequency of other influences on running speeds
Even if pregnancy causes less reduction in locomotor performance than do other factors, it might still impose effective selection on reproductive investment if the animals only rarely encounter situations where the other factors come into play. The concept is best illustrated with a contrast between two examples. First, imagine a lizard in an equable tropical environment. The benign thermal environment results in very stable body temperatures, and feeding opportunities are constant rather than on a ‘boom and bust’ schedule as can occur due to weather fluctuations in less stable climates. Under such circumstances, reproductive state will generate much of the total temporal variance in locomotor speeds for an individual. Thus, animals with higher RCMs are significantly more likely to be taken by predators, and selection favours a reduction in RCM via this selective pressure. As a contrast, imagine the same lizard in a cooler and more variable climate. Most of the variation in its locomotor performance will be generated by diel shifts in body temperature, and by occasional binge-feeding events when weather conditions facilitate foraging activity. For such an animal, pregnancy may engender only a trivial component of the overall temporal variation in locomotor speed. Thus, we would not expect selection on maternal mobility to reduce RCMs in such a population.
This discussion predicts that we should see selection for reduced RCMs in reptiles that inhabit tropical rather than temperate-zone habitats. An extensive literature documents exactly this pattern: reproductive investment per clutch is lower for tropical than for temperate-zone reptiles, across an array of continents and phylogenetic lineages (Tinkle et al. 1970; Rand 1982; James & Shine 1988; Shine & Keogh 1996). Previous attempts to explain this pattern in terms of the intensity of competition (Tinkle et al. 1970) or predation (Andrews & Rand 1974; Barbault 1975) have little empirical support. Similarly, prolonged retention of eggs leading to viviparity may be less of a ‘cost’ to maternal locomotor speeds, relative to other factors, for a female reptile in a relatively cool climate than in a warmer climate. The trend for viviparity to evolve in cool rather than warm climates is well known, but has always been interpreted in terms of higher benefits rather than lower costs to retention of eggs (Shine 1983, 1985). The hypothesis also predicts higher RCMs in snakes (many species of which frequently consume prey items very large relative to the size of the predator: Greene 1997) than in lizards (which typically take smaller prey: Webb et al. 2000a,b). Locomotion can be very severely compromised in snakes that have ingested large prey (Jago 1994), and presumably, snakes that take large prey have evolved behavioural adaptations that enable them to survive frequent periods of substantial locomotor impairment. In keeping with the above prediction, RCMs of snakes are typically much greater than are those of lizards (Seigel & Fitch 1984; Shine 1992).
One clear result from this analysis is that locomotor impairment due to reproduction will not necessarily translate into a survival cost (Vitt & Price 1982). Instead, any decrement in running speeds due to pregnancy needs to be seen within the context of other factors. Adding further complexity is the potential for interactions among these factors. For example, pregnant lizards may maintain higher more stable temperatures than do non-pregnant animals (Schwarzkopf 1994), thus erasing the locomotor decrement due to pregnancy (see Fig. 1a); and tail loss might reduce running speeds much more in non-pregnant than in pregnant animals (Fig. 1b). Although the complexity of such interactive effects is a daunting prospect for attempts to measure costs of reproduction, it also offers exciting opportunities to further understand the links between life-history traits, performance measures and selective forces.
The more general message from these analyses is that costs of reproduction, even if empirically demonstrable in terms of variables such as locomotor speeds, feeding rates or thermal biology, do not necessarily impose strong selection on optimal levels of reproductive investment. Simply showing that reproduction influences an organism's behaviour or performance does not mean that such costs are (or have been) significant evolutionary forces. For example, even high mortality or energy costs may be irrelevant to selection on reproductive investment in very short-lived animals (Williams 1966a,b; Shine & Schwarzkopf 1992); performance decrements can be overcome (or translated into different currencies of ‘cost’) by behavioural modifications (Bauwens & Thoen 1981; Brodie 1989); and the magnitude of effects due to reproduction must be viewed within the context of other aspects that generate variance in the same attributes (current study). This cautionary note applies as strongly to ‘cost’ measures based on energy intake as to those based on organismal performance; for example, the reduction in feeding rate due to pregnancy must be seen within the context of other environmentally imposed sources of variation in feeding rates (most obviously, in food supply). Thus, even if reproductive activities exert consistent and easily measured effects on traits such as energy expenditure or locomotor ability, spatial and temporal variation in factors such as predator densities, shelter availability and resource levels may generate massive variation in ‘costs of reproduction’.
I thank Melanie Elphick and George Barrott for their Herculean efforts in catching, measuring, maintaining and running lizards. The study was funded by the Australian Research Council.