Dr Jean Clobert, Laboratoire d’Ecologie, Université Pierre et Marie Curie et Ecole Normale Superieure, Bâtiment A, 7 quai Saint Bernard, Case 237, 75742 Paris cedex 05, France Correspondence author.
1. There is growing evidence that dispersal is highly phenotypically plastic, i.e. that dispersal is condition-dependent. In the common lizard, dispersal has even been shown to be influenced by the maternal environment during pregnancy. Juveniles in good condition or issued from mothers in good condition disperse earlier or in higher numbers.
2. We hypothesized that plasma corticosterone was the proximate mechanism by which condition and dispersal are linked, and tested this by manipulating the level of circulating corticosterone in pregnant females of the common lizard.
3. After parturition, we measured juvenile attractiveness towards the mother and juvenile dispersal of corticosterone (B) and placebo (P) implanted females.
4. Offspring of B females did disperse in lower number than those of P females. B offspring were also more attracted by the mother's odour than P offspring.
5. In quite a few cases, the behavioural response of juveniles was dependent on the interaction between the hormonal treatment and the mother snout–vent length or condition (body weight corrected for snout–vent length).
6. Corticosterone constitutes therefore one of the proximate mechanisms involved in the prenatal control of juvenile dispersal in this species. Along with other results, it is proposed that prenatal control of dispersal has evolved in order to avoid competition between mothers and their offspring.
Environmental stressors including starvation, competition, predation, and parasitism result in an increase in plasma glucocorticoids (Siegel 1980; Harvey et al. 1984). An increase in plasma levels of these hormones in pregnant or lactating females may directly influence offspring phenotype (Politch et al. 1978; Peters 1982). Corticosterone, the predominant stress hormone of many vertebrates, has repeatedly been found to influence various morphological, behavioural, and life-history traits (Hamm et al. 1983; Pollard 1984; Takahashi et al. 1988; Hardy et al. 1990; Sinervo & Denardo 1996). For example, the ability of an individual to settle in a territory (Silverin et al. 1989), to defend this territory (Wingfield & Silverin 1986), and to interact with potential mates (Seitz et al. 1997; Seitz 1998) is influenced by levels of circulating corticosterone and testosterone. Corticosterone in particular has been found to play a major role in alternate behavioural patterns (Wingfield 1994). However, only two experiments have investigated the relationship between natal dispersal and hormones: the first found that testosterone but not corticosterone in female Belding's ground squirrels (Spermophilus beldingi) influenced juvenile dispersal (Holekamp et al. 1984). The second one reported that implanted corticosterone enhances autumn dispersal in juvenile willow tits (Parus montanus) (Silverin 1997). We are unaware of any study that has investigated the influence of maternal corticosterone level on dispersal of offspring.
Of the many factors that influence natal dispersal of common lizards (Lacerta vivipara), the social environment is one of the most important (Clobert et al. 1994). For example, competition for food between individuals has important consequences on the population dynamics of this species (Massot et al. 1992; Lecomte et al. 1994). Many of these factors influence natal dispersal through a prenatal mechanism (Ronce et al. 1998; Sorci et al. 1994). In particular, the level of food delivered to the mother during gestation has been found to increase offspring dispersal (Massot & Clobert 1995), while the morphological phenotype of the offspring was not modified. This finding is counter-intuitive because the offspring dispersed from a productive environment as indicated by a high mother feeding rate. However, a mother living in a ‘good’ environment has a good prospect of survival. Thus, in a ‘good’ environment, natal dispersal may help to decrease the likelihood of mother–offspring competition. Kin competition is an important force driving the evolution of dispersal (Hamilton & May 1977; Ronce et al. 1998), and indeed this type of competition has proved to be important in shaping natal dispersal in this species (Léna et al. 1998; Ronce et al. 1998).
How could the prenatal environment influence the likelihood of offspring dispersal? We have previously shown that environmental stressors such as a food restriction can raise the level of plasma corticosterone in the mother (Harvey et al. 1984) and that stress hormones can influence the movements of individuals (Wingfield 1994; Silverin 1997). We also showed that female snout–vent length (SVL) or female body condition (body weight divided by SVL) were correlated to their offspring dispersal and sensitivity to their own odour (Léna et al. 1998). Thus, our main goal in this study was to see whether we could increase offspring philopatry by increasing plasma corticosterone levels of pregnant females (mimicking a food-deprived mother), and how this was mediated by their SVL or body condition.
Species, collection, and laboratory conditions
The common lizard (Lacerta vivipara) is a small (adult female snout–vent length [SVL] = 55 mm on average) Lacertidae found in peatbogs and heathland across Europe and Asia. In our populations of the Mont Lozère (Massif Central, south-east of France), males emerge from hibernation in mid-April, followed by yearlings and females in mid-May. Mating takes place as soon as females emerge from hibernation. After 2·5 months of gestation, females lay a clutch of on average five soft-shelled eggs. Offspring hatch within 1 hour. Young (18 mm SVL) have an independent life from their mother right from birth. Finally, individuals enter hibernation again in mid-to late September.
We have been studying this species for 10 years (Clobert et al. 1994; Sorci et al. 1996; Léna & de Fraipont 1998), and have been particularly interested in the importance of maternal effects in determining the phenotype of juveniles (Massot & Clobert 1995; Sorci et al. 1995, 1996; Léna et al. 1998). To accomplish our goals, we have had to ascertain sibling relatedness and to manipulate the maternal environment. Our standard procedure has been to remove pregnant females early in July and house them in individual terraria in the laboratory until parturition early in August. To facilitate thermoregulation, we provided an incandescent lamp as a heat source for 6 h per day, thus creating a gradient of 28–34 °C in each terrarium. We provided water ad libitum, and we offered each female one Pyralis larva each day. Many females do not feed on a daily basis, and we recorded the number of food items taken by each female (see Massot & Clobert 1995).
For the present experiment, we collected 50 pregnant females from a peatbog situated near our long-term study populations. Collections were completed between 25 June and 2 July 1997.
Hormone implantation was done under cold anaesthesia within 3 days after capture. Implants were inserted subcutaneously through a 5-mm lateral incision at the mid-point of the animal (less than 4 min per animal). Twenty-five females received Silastic implants containing crystalline corticosterone (Sigma No. C-2505 St. Louis MO USA, tubing Dow-Corning N°. 602–235, Dow-Corning medical adhesive no. 8916, implant dimension: 6 mm long and 2 mm in diameter), and 25 received empty implants to serve as controls. Implants were punctured 6 times with a 27 gauge hypodermic needle to increase the diffusibility of corticosterone. The locations of cages of corticosterone- vs. placebo-implanted females were randomly assigned in the laboratory. Females were individually marked, and the correspondence between each female's mark and the treatment type was known only to an investigator not participating in the experiment.
To assess the effect of hormone implants on maternal plasma corticosterone levels, we collected 60 μL of blood (taken from the corner of the eye) in a heparined Eppendorf tube, directly after parturition. Blood collection was complete within 30 s. We could not measure the level of corticosterone during pregnancy because female reproduction is strongly reduced when the required quantity of blood is taken during this period. The blood sample was immediately centrifuged, and the plasma was stored at −25 °C. Plasma corticosterone was measured by radioimmunoassay (RIA) following Smith & John-Alder 1999.
Female behaviour during gestation
Following surgical implantation, we observed the behaviour of each female three times per day at 09.30, 13.00 and 16.30 h until parturition. We recorded whether the female was outside or inside her shelter and whether she was active or motionless. We also measured the time needed to seize a prey. Care was taken to ensure that the females’ behaviour was unaffected by the observer. We made no effort to quantify behaviour in any greater detail because captivity itself can deeply affect the behaviour of this species (Avery 1966). The information we recorded provided an overall index of the activity (e.g. active thermoregulation, exploration, activity linked to stress, etc.) of each female. An indication of stress was also given by the number of food items eaten by each female.
Juvenile behaviour at birth
To assess the reactions of juveniles to stressful situations of varying intensity, four types of experiments were conducted just after birth. Some experiments (new ones) were performed twice to assess their repeatability. All terraria were thoroughly rinsed between trials.
The first experiment measured the reaction of a juvenile to an acute stress. Each juvenile was placed at the centre of a terrarium (18 × 12 × 12 cm) which contained a shelter in one of its corners. After 10 min of acclimation, an investigator tapped the juvenile's tail with a stick (a modification of the Haffner's test used in the study of stress and analgesia in pharmacology, Bianchi & Franceschini 1954). Three types of behaviour were observed: (i) the juvenile remained motionless; (ii) the juvenile reacted by running within the terrarium; (iii) the juvenile reacted by running into the shelter.
The second experiment was designed to investigate the juvenile's reaction in a new situation during daylight with and without a familiar landmark (mother's odour). We placed one juvenile into an empty terrarium and focused a video camera on the juvenile from above. After 10 min of acclimation, the juvenile was filmed for 10 min. We then introduced a piece of paper impregnated with the odour of its mother, and the juvenile's behaviour was recorded for an additional 10 min. We recorded the time spent active (walking) and the position in the terrarium (centre vs. along the side). We also measured the time spent scratching the wall, a behaviour we interpret as an attempt to escape from the terrarium.
The third experiment investigated the rapidity with which a juvenile accepts a prey after being placed into an unfamiliar environment. A juvenile was placed alone into a terrarium and given 10 min to acclimate. Then, a small cricket was offered for 1 minute, and we recorded if the juvenile seized the prey within this time.
The fourth experiment was designed to investigate the reaction of a juvenile to the odour of its mother, used here as an indication of a familiar environment. In a previous experiment, the affinity of juveniles for their maternal odour was found to depend on the prenatal environment (Léna & de Fraipont 1998). Juveniles that did not exhibit an attraction to their maternal odour were those that were found to disperse in the field (Léna et al. 1998; Léna et al. unpublished). The present experiment was designed to investigate whether corticosterone could be a determining mechanism of this behavioural variation. Two shelters were randomly placed in a terrarium, one with no odour and one containing the maternal odour (Léna & de Fraipont 1998). The juvenile was introduced into the terrarium at 16.00 h on the day of its birth. At 24.00 h, we recorded the position of the juvenile with regard to three possibilities: (i) outside of both shelters; (ii) inside the odourless shelter; (iii) inside the shelter containing maternal odour.
The measure of juvenile dispersal in nature requires large sample sizes incompatible with the present study. Thus, we followed a surrogate procedure to investigate dispersal without releasing juveniles into their natal population. Natal dispersal occurs almost exclusively within 10 days of birth in Lacerta vivipara (Clobert et al. 1994; Massot & Clobert 1995). Thus, we can assess the likelihood that a juvenile will disperse by observing its behaviour under semi-natural conditions during this period of time (Clobert et al. 1994; Léna et al. 1998). The experimental design is as follows (see Léna et al. 1998): each experimental set consists of two enclosures of 4 m2 connected by small holes large enough only to allow passage of juveniles. One day prior to the introduction of juveniles, the mother was placed into one arbitrarily chosen enclosure of each set. Then, the mother's offspring were introduced at 4 days of age (no dispersal has previously been observed prior to this age, Léna et al. 1998). The offspring were allowed to move freely within the experimental set for 6 days. Subsequently, we recorded the number of juveniles found in the enclosure not containing the mother. This experiment is a good surrogate of a natural situation since previous experiments have shown that when juveniles disperse to the non-maternal enclosure, they never return to the enclosure from which they emigrate, and that the timing and rate of dispersal as well as the magnitude of the family effect are the same as under natural conditions (Clobert et al. 1994; Lecomte & Clobert 1996; Léna et al. 1998). Furthermore, dispersers in natural populations had characteristics similar to those of juveniles observed to colonize the non-maternal enclosure (Massot & Clobert 1995; Léna et al. 1998; Ronce et al. 1998). Between experiments, experimental sets were thoroughly rinsed. After the experiment, all juveniles and their mother were released into their population of origin.
For dependent variables that are continuous (e.g. time spent, body length, body mass), we used the GLM procedure of the statistical package of SAS Institute (1990a), after having verified their normality. In analyses of juveniles, the female effect was nested within the treatment effect because juveniles of the same family cannot be considered as independent units. Using this general design, we performed nested analyses of variance or covariance. Level of significance were assessed through F-tests. However, in this experiment, the female effect was proved either to have not much influence on the results (and therefore could be dropped from the analyses, McCullagh & Nelder 1989) or could be successfully modelled through the inclusion of both the snout–vent length (SVL), and/or the body condition (body weight divided by SVL) of the mother. We did include both variables in the analysis, but they were never found significant together. We also included juvenile sex in all the analyses, but, as before (Massot & Clobert 1995; Léna & de Fraipont 1998; Léna et al. 1998), this variable was never found to have a significant effect, and therefore was not reported.
For dependent variables that are percentage measurements (e.g. dispersal rate, percentage of time spent), we used the GENMOD procedure of SAS Institute (1990b). This procedure performs logistic-linear regression analysis particularly suited to this type of dependent variable (McCullagh & Nelder 1989). Log-likelihood ratio tests (χ2 values) were used to assess significance of the effects. The model design used was the same as that for continuous dependent variable. We verified that the fit of the final model was good (variance inflation factor), and corrected our tests when necessary (noted χ2 corr).
In both cases, we use type III sum of squares (non-sequential decomposition). We started with a general model including all the potential effects and their interactions. We then dropped the non-significant effects (backward selection, McCullagh & Nelder 1989), starting with the most complex interaction terms. Only the results of the final model are reported.
Of the 25 females that received a corticosterone implant (B), seven died within a few days of surgery and six aborted before parturition. Of the 25 females that received placebos (P), five died within a few days of surgery and two aborted. Laying date was not significantly different among groups (Fl,28 = 0·56 P = 0·48). Thus, hormone implantation itself did not increase mortality or the disruption of reproduction (
= 2·83 P = 0·10).
Females were of the same length and body mass in the B and P groups (mean = 53·1 ± 6·4 mm, Fl,28 = 1·66, P = 0·21; mean = 5·9 ± 0·9 g, Fl,28 = 0·06, P = 0·80). Treatments did not affect the female post-parturition condition (mean = 2·6 ± 0·6 g, Fl,28 = 0·24, P = 0·63) or the number of live offspring (mean = 3·04 ± 0·23, Fl,28 = 0·03, P = 0·86). After parturition, females did not differ in their plasma corticosterone (Fl,19 = 1·09 P = 0·347; mean: B = 12·22 ± 2·6 ng mL−1, P = 12·40 ± 3·9 ng mL−1, only 19 measurable samples). Offspring SVL was not different between treatments (mean = 179 ± 19·8 mm, F1,28 = 0·2, P = 0·68). Juveniles of B females did not have significantly lower body condition than those from P females (Fl,28 = 1·0, P = 0·56; mean: [B] = 182·11 ± 15·4 mg, [P] = 176·61 ± 22·5 mg). In total, 90 juveniles participated in the behavioural experiments, 36 born to B females and 54 to P females.
Female behaviour during gestation
During the 3-week period between implantation and parturition, the total number of food items eaten by females did not differ between treatments (mean = 4·5 ± 1·3 items of prey;
= 0·19, P = 0·66), and the time required for pregnant females to seize the prey item did not differ between treatment groups (357 ± 153 s; F1,28 = 0·84, P = 0·36). Placebo females were found significantly more often in their shelters than B females and this was not dependent on the time of the day (treatment
= 9·01, P < 0·01; interaction treatment time of the day:
= l.25, P = 0·32). However, the behaviour of females outside of their shelters did not differ between treatments (
= 0·24, P = 0·71, other terms not significant).
Juvenile behaviour at birth
In the presence of an acute stressor
When faced with an acute stressor, juveniles had three options: to remain motionless, to escape into a shelter, or to run within the terrarium. This experiment was done twice, and juvenile behaviour was highly repeatable between trials (
= 66·31 P < 0·001). Only six juveniles out of 94 remained motionless in response to tail tapping. Thus, only the types of escape strategy (escape to a shelter, running inside the terrarium) were included in the statistical analysis. Maternal condition did influence the response of juveniles (maternal SVL: SVL
= 3·38 P = 0·07, SVL × treatment
= 2·52 P = 0·11; maternal body condition: body condition
= 3·02 P = 0·08, condition × treatment
= 4·61 P = 0·01). Although particularly evident for offspring of lean B-implanted females (see condition by treatment interaction), the offspring of B-implanted females fled toward the refuge significantly more often than those of placebo-implanted females (65% vs. 32%; with mother's condition as a covariate:
= 5·38 P = 0·02; without:
= 4·70 P = 0·03).
Unfamiliar environment during the day
When placed for 10 min alone in a terrarium, almost no juvenile remained in the centre of the terrarium. We therefore did not analyse this variable. In the absence of maternal odour, juvenile lizards spent about 25% of their time scratching the sides of the terrarium. The remaining time was divided between periods of walking and inactivity.
While the treatment did not affect the time spent walking (all juveniles mean = 213 ± 22 s, no significant effects; see Table 1), juveniles of B females spent significantly less time inactive than juveniles of P females, mainly when they were born from a relatively small female (treatment means: 218 ± 14 s vs. 249 ± 15 s, interaction SVL × treatment F1,86 = 5·62 P = 0·02; Table 1). More importantly, juveniles of B females spent significantly more time scratching the sides of the terrarium in the absence of maternal odour (treatment means: 168 ± 13 s vs. 141 ± 11 s; see Table 1 and Fig. 1), once more mainly for those born from small females. Conversely, in the presence of maternal odour, juveniles of B females spent significantly less time scratching the sides of the terrarium than juveniles of P females (122 ± 12 s vs. 150·5 ± 19 s, Fig. 2).
Table 1. . Juvenile behaviour when in an unknown environment during the day in absence of the mother's odour (see Juvenile behaviour, experiment 2). Juvenile behaviour was divided into three classes (walking, inactivity, and scratching the side of the terrarium) and was measured by the number of seconds devoted to a given behaviour during 10 min of observation. We performed covariance analyses with either female snout–vent length or corpulence after parturition as the covariable and the treatment (hormone vs. placebo) as factor effect (all F-tests with 1 and 86 degrees of freedom). Here we report the level of significance for the treatment and the covariable for the complete model. The results are no different after having dropped the non-significant effects
Juveniles were placed alone in terraria and were offered one cricket. We recorded whether or not the cricket had been eaten after 1 minute. The repeatability of this experiment was high (
= 15·02, P < 0·01). Overall, the maternal hormone treatment had no effect on the likelihood that a juvenile had eaten the cricket within 1 minute (
= 0·78, P = 0·38). This result was not influenced by female body length (
= 1·47, P = 0·22). However, maternal condition was positively correlated with the probability that a juvenile would eat its cricket within 1 minute (
= 6·64, P < 0·01), although this effect was not affected by the mother's treatment (condition × treatment interaction:
= 1·16, P = 0·28).
Attraction to mother's odour
Each juvenile was in an individual terrarium at the beginning of the night. The juvenile had then the choice to spend the night in a shelter containing the odour of its mother, in a shelter containing no odour or to stay outside the shelters. The treatment had no effect on whether the juvenile decided to stay outside or to enter a shelter even when the condition of the mother (condition
= 0·05, P = 0·82; treatment
= 1·79, P = 0·18; condition × treatment
= 1·18, P = 0·28) or the mother SVL (SVL
= 0·29, P = 0·59; treatment
= 1·79 P = 0·18; SVL × treatment
= 0·30, P = 0·58) was entered into the analysis as a covariable.
There was no overall effect of the treatment on whether the juvenile choose the shelter containing its mother odour or not (
= 0·18, P = 0·67). However, when female SVL or female condition was entered into the analysis, the treatment effect became highly significant (with female SVL
= 5·44, P < 0·01; with female condition
= 6·11, P < 0·01). Juveniles born from small-sized females or from females with low body condition after parturition (SVL or body condition is not significant when body condition or SVL are also entered in the analysis) chose significantly more often the shelter containing their mother's odour when the females were B-implanted than when they were P–implanted (interaction condition × treatment
= 6·28, P < 0·01; SVL × treatment
= 5·56, P = 0·02).
Juvenile dispersal rate
Overall, the percentage of dispersing offspring within the clutch was lower for B-implanted than P-implanted females (33 ± 7% vs. 58 ± 9%,
= 5·91, P = 0·01). Juvenile dispersal was not significantly affected by maternal condition or body length in the presence (
= 0·21, P = 0·81;
= 1·79, P = 0·18) or absence (
= 0·001, P = 0·98;
= 1·12, P = 0·41) of the main treatment effect.
When we removed the hormonal treatment effect from the analysis and replaced it by the outcome of the juvenile shelter selection (SEL) experiment (see above), juvenile dispersal rate was significantly dependent on its sensitivity to mother's odour (
= 4·28, P = 0·04). When the hormonal treatment effect was reincluded in the analysis, the effect of SEL vanished (hormone treatment
= 6·11, P = 0·01; SEL
= 0·32, P = 0·571). The two variables seemed to carry the same information with respect to juvenile dispersal.
Implantation of corticosterone modified the behaviour of pregnant mothers and the behaviour of their offspring in the first 10 days after birth. Corticosterone-implanted females were more active (measured as time outside the shelter), than their placebo-implanted counterparts, but food consumption was not affected by hormone treatment. When placed alone in unfamiliar terraria, the time spent scratching the wall of the terrarium was higher in the absence and lower in the presence of maternal odour in the offspring of B-implanted females compared to P-implanted ones. Similarly, the number of offspring who chose to spend the night inside a shelter marked by maternal odour was higher for the B- than the P-implanted treatment, and dispersal rates were lower in offspring of B- than P-implanted mothers.
Effect of implantation
Surgical implantation of Silastic implants increased mortality of pregnant females. This increase in mortality was due to the surgery itself rather than the specific treatment: nearly all of the mortality occurred within a few days of surgical implantation, and the number of females that died did not differ significantly between treatments. Surgical implantation also increased clutch abortion in comparison to non-implanted females, but the rate of abortion did not differ significantly between B- and P-implanted females as found in other species (Pollard 1984, 1986; Wilson & Wingfield 1992). Overall, surgical implantation approximately doubled mortality and the incidence of abortion in comparison to natural conditions. Of the clutches that were carried to term, however, hormone treatment did not affect clutch size (P = 0·31), maternal mass, or juvenile size at birth. Thus, the differences in behaviour we observed between offspring of B- vs. P-implanted females are the direct result of hormone implantation itself rather than an indirect result of a hormone-induced change in morphology.
We did not detect a significant difference in total plasma corticosterone concentrations in females after parturition. Thus, the behavioural responses did not result from pathological increases in plasma corticosterone, and the mechanism through which the implantation of Silastic tubules containing corticosterone led to behavioural responses is subject to conjecture. One possibility is that the B implants led to small but physiologically significant increases in plasma corticosterone, resulting in a difference between groups too small to be resolved by our assay. An alternative hypothesis is that the B implants were sufficient to replace but not supplement endogenous corticosterone. If true, then endogenous production of corticotropin-releasing factor (CRF) from the hypothalamus and adrenocorticotropic hormone (ACTH) from the pituitary could have been greatly suppressed, and the absence of one or both of these factors rather than an increase in corticosterone itself may have led to the observed behavioural responses. A third, more likely explanation, is that the important production of corticosterone just before parturition (Dauphin-Villemant & Xavier 1986, 1987; Dauphin-Villemant et al. 1990) has overridden the difference induced by implantation. The important point is that the behavioural responses observed in females and their offspring in our properly controlled experiment must be attributed to the presence of corticosterone in the Silastic implants (a similar experiment done with testosterone has induced a quite different suite of consequences, J. Clobert et al. unpublished), even though the mechanism of action of these implants is open to conjecture.
Corticosterone, stress, and maternal effects
The level of plasma corticosterone is viewed as an integral component of the stress response (reviewed in Holst, in press). Therefore, our use of Silastic implants to experimentally supplement or replace endogenous corticosterone secretion mimics an animal's response to chronic, unpredictable stress. The behavioural responses observed in the present experiment suggest that exogenous corticosterone may have mimicked pregnant females subjected to a recurrent problem, such as food deprivation (Massot & Clobert 1995) or a high parasite load (Sorci et al. 1994). Indeed, food deprivation (Gray et al. 1990; Rogers et al. 1993), parasitism (Oppliger et al. 1998), and social dominance (van Holst 1986) have all been shown to act as chronic stressors.
Prenatal stress is known to affect offspring behaviour in rats, particularly the response of offspring to acute stress (Pollard 1984; Takahashi et al. 1990). In the present experiment, maternal B implantation influenced the behaviour of juveniles placed into an unfamiliar (stressful) environment. The offspring of B-implanted (stressed) females exhibited a higher rate of escape attempts than the offspring of P-implanted (unstressed) females, and this effect was reversed in the presence of maternal odour. The familiarity of maternal odour apparently counteracted the stress of an unfamiliar environment. In the face of acute stress, offspring of B-implanted females sought refuge in higher numbers than those of P-implanted group. Furthermore, when placed into an unfamiliar environment at night, the prenatally stressed offspring selected the shelter with maternal odour more frequently than the non-stressed offspring, in particular when they were born from small or lean B-implanted females. Juveniles which were characterized by this behaviour were also found, like in natural populations (Léna et al., unpublished), to stay philopatric. This further confirms that corticosterone, sensitivity to mother's odour and juvenile dispersal are intimately linked. The offspring of B-implanted females always seemed to adopt a risk-adverse strategy, seeking the situation that would minimize the stress. Perhaps offspring exposed to prenatal stress have a diminished endocrine capacity to respond to postnatal stress as demonstrated in prenatally stressed rats (Pollard 1984; Takahashi et al. 1988, 1990). This will have to be confirmed by examining offspring behaviour in more natural conditions.
The extent to which this behavioural response is adaptive will depend on its implication with regard to exploratory movement (Silverin et al. 1989; Silverin 1997), habitat or mate selection in the wild (DeNardo & Sinervo 1994a,b; Wingfield & Silverin 1986; Moore & Marler 1987; Wingfield et al. 1987), and more generally dispersal, as already suggested by Holekamp et al. (1984), and Belthoff & Dufty (1998). In Lacerta vivipara, kin (mother/offspring) competition seems to be an important force driving the evolution of dispersal (Massot & Clobert 1995; Léna et al. 1998; Ronce et al. 1998). It follows that poor maternal internal condition, indicated by chronic stress, would foretell a poor prospect of maternal survival and would thus promote philopatry in the offspring. This may explain that only juveniles from lean B-implanted females were attracted by the shelter containing their mother's odour, as indeed expected if kin discrimination has evolved to optimize individual spacing behaviour (Léna et al., unpublished). However, if female condition is positively correlated with environmental quality and is more indicative of the level of intraspecific competition, dispersal will be promoted when the condition of the female deteriorates (Ronce et al. 1998) because it will be an indication of a poor environment rather than an indication of a poor health of the mother per se. This may explain why we found results opposite to those of Silverin (1997) and Belthoff & Dufty (1998). Indeed, when applied on the offspring itself, corticosterone will modify juvenile behaviour directly rather than indirectly (through embryonic development) when applied at a prenatal stage. The direct and indirect effect of corticosterone can then elicit different behavioural responses. It may also well be that prenatal elevation of corticosterone induces a general decrease of the activity potential of the offspring (developmental constraint) and, as a by-product, a decrease in dispersal early in life. In this case, it may reflect more a constraint than an adaptation.
To determine if offspring sensitivity to maternal condition (potentially linked to the avoidance of kin competition) through hormone influences is really adaptive (Ketterson & Nolan., 1992), the next step will be therefore to compare the fitness of offspring produced by mothers of different conditions.
We thank the National Park of Cévennes, for their help during our field season, and Manuel Massot, Gabriele Sorci and one anonymous referee for critically reviewing previous drafts of this manuscript.
Received 4 January 1999;revisionreceived 30 September 1999