Behavioural responses of Lampropholis delicata across four groups sizes

Behaviour is a highly labile trait that can be rapidly modified to mitigate the effects of changing environmental conditions. Among the biotic and abiotic factors acting to prompt plastic responses, the social environment has been proposed as being one of the primary modulating forces on behaviour. Being part of a group has particular influence on the expression of risky behaviour in that added eyes and ears serve to decrease a group member's vulnerability to predation resulting in the mean behavioural expression of behaviours such as activity and/or exploration increasing with group size. A large body of work has documented such group size effects. However, as this process may operate at the individual level it is unclear how the social environment affects the consistent expression of personality. Thus, we examined the interactive effects of behavioural phenotype and social context on the stability of individual behaviour in the delicate skink, Lampropholis delicata. Lizards were exposed to a series of assays measuring activity, exploration and boldness in groups of one, two, four and eight. Repeatability was determined across group size treatments and the combined effects of social context and behavioural type on behavioural plasticity were assessed. We found that the predicted patterns of group size effects were only observed when each lizard's behavioural phenotype was considered. Inactive and shy lizards increased their behaviour with increasing group size whereas active and bold lizards exhibited the opposite pattern. Additionally, the degree to which an individual adjusted its behavioural response was contingent upon its behavioural phenotype, with slow lizards showing higher levels of responsiveness than fast lizards. Despite this plasticity, between-individual differences in the expression of activity, exploration and boldness persisted. Thus, our study provides strong evidence that the effects of an individual's personality are stronger than those of group size.Funding provided by: Australian Research CouncilCrossref Funder Registry ID: http://dx.doi.org/10.13039/501100000923Award Number: DP170100684Experiments were conducted from October 2014 to January 2015. All lizards were in a postabsorptive state (fasted for 48 h) during each of the trials as digestion can affect behaviour (Shine, 2003). Behaviour was evaluated in the context of activity, exploration and boldness across four group sizes (one, two, four and eight individuals) following standard methodology (Goulet et al., 2018; Michelangeli et al., 2018). Group sizes were based on natural grouping tendencies observed in this species in the wild as well as those used in a previous study on this species investigating group size effects (Downes & Hoefer, 2004; unpublished data). Lizards were randomly assigned as either focal (N = 20) or stimulus (N = 20) prior to the experiments. Focal lizards were housed separately from stimulus lizards to prevent the effects of social recognition. Each group consisted of a single focal lizard and the required number of stimulus lizards. All focal lizards completed each behavioural assay four times, once for each group size, with at least 2 days between assays. To minimize the risk of carryover effects, lizards went through the behavioural assays in a fixed order, outlined below, where assays that would have the greatest impact upon behaviour were at the end of the experimental schedule. However, the order in which the focal lizards experienced each group size was randomized to control for order effects and habituation. This sequence was predetermined for each focal group and maintained throughout the study. Assays were conducted within opaque experimental arenas (550 x 320 mm and 240 mm high) located in a temperature-controlled room (20 °C). Each trial consisted of a 10 min acclimation period where lizards were placed within a clear plastic container followed by a 30 min test period. Lizard behaviour was recorded using Panasonic HCV130 video recorders suspended above the test arena and data were analysed using Jwatcher (Blumstein 2006). All equipment used was washed with soapy water between trials to remove chemical cues. Activity Activity was measured by placing lizards into a test arena marked with 20 grid squares (80 x 110 mm; Fig. A1a). The number of benign transitions (> half of body over grid line) between squares was recorded. The term benign transitions is used to distinguish between activity within a nonthreatening context and activity within a risky situation (see methods describing the boldness assays below). Exploration An opaque Perspex partition was placed in the centre of the arena, dividing it in half (Fig. A1b). The partition was a trapezium (flush with the base of the experimental arena, tapering to a 15 mm gap at 100 mm in height). This required the lizards to climb and squeeze into the gaps to reach the other compartment. The time it took the focal lizard to manoeuvre around the obstacle and reach the other compartment was recorded. Individuals that did not reach the goal by the end of the trial were assigned 30 min. Boldness Boldness was assessed in two contexts: basking site selection and simulated predator attack. For the basking site selection assays, the arena was divided into three equal zones: open basking site, an intermediate no-preference zone and sheltered basking site (Fig. A1c). Both basking site zones includeda heat lamp (40 W) suspended above a ceramic tile (100 x 100 mm) to encourage natural thermoregulatory behaviour. However, the sheltered basking site also had an opaque plastic shelter placed over the basking tile to provide a safe refuge. The intermediate zone was empty. The time spent in each zone was recorded. For the simulated predatory attack assays, a test arena marked with 20 grid squares (80 x 110 mm) was divided into three zones: basking zone, an intermediate neutral zone and shelter zone (Fig. A1d). The basking and intermediate zones were set up as previously described while the shelter zone consisted of a plastic refuge. Unlike the previous assays, the lizards were given 10 min to freely acclimate within the arena. After the acclimation period a plastic bird model (700 mm wing span; 50 mm beak to tail length) was flown low over the test arena three times at a predetermined pace (Michelangeli, Chapple, & Wong, 2016; Michelangeli, Chapple, & Wong, 2016). The time spent basking in each zone and the number of risky transitions after the simulated predator attack were recorded. All statistical analysis was conducted using R version 3.3.2 (R Development Core Team 2016). Data were checked for normality (Shapiro–Wilk test) and homogeneity of variance (Fligner–Killeen test) where appropriate. Boldness metrics were found to be correlated; thus, a single measure between pairs of correlated metrics considered to be most representative of an individual's level of boldness (i.e. time in basking zone following simulated predator attack versus time in refuge) was used in the analyses. We, therefore, included time spent in the basking zone under benign conditions (before simulated predator attack), time spent basking under risky conditions (after predator attack) and the number of risky transitions. Each lizard's behavioural phenotype was assigned as being either active or inactive, exploratory or nonexploratory and bold or shy according to whether they scored above or below the median score for each behaviour when tested alone. Median scores were calculated separately for each trial number block (four trials per block for each behaviour) to account for order effects as a function of group size. A generalized linear mixed-effect model (GLMM) with a Poisson distribution was then used to investigate the influence of group size on activity and to test whether individuals differing in behavioural phenotype were affected similarly by social context. The number of benign transitions was the response factor while group size, behavioural type and their interaction as well as trial number were included as fixed factors. Individual identity (ID) and group size were assigned as random factors. By including both ID and group size as random effects, individual intercepts estimated personality while plasticity estimates were represented by the slope of the interaction between ID and group size (Betini & Norris, 2012). Post hoc analyses (package: 'lsmeans') were performed to identify which paired comparisons were significant. Markov chain Monte Carlo linear mixed-effects models (MCMCglmm R package: Hadfield, 2010) were used to evaluate the influence of group size on exploratory and boldness behaviour. Time to cross the barrier was modelled with a log-normal distribution, times spent basking under benign and risky conditions were modelled with a Gaussian distribution and the number of risky transitions was modelled using a Poisson distribution. Group size, behavioural type and their interaction as well as trial number were included as fixed factors and ID and group size as random factors. Prior to running the models, variation in Markov chain lengths, prior specification, thinning intervals and burn-in lengths were explored to obtain models that had adequate sampling of the posterior distribution and showed limited autocorrelation among samples. For the final models, default diffuse uniform priors were used for fixed effects and a random effect variance–covariance matrix prior specification V = 0110 and nu = 0.002 for random effects (Hadfield, 2010). Models were run for 10 million iterations with the first 200 000 discarded (burn-in) that were sampled every 5000 iterations (thinning interval), which resulted in an effective sample size of < 1000. Trace plots were visually inspected to ensure chains had good mixing. Autocorrelation among samples was assessed to ensure levels were low (lag < 0.1) using the autocorr function in the R package coda (Plummer, Best, Cowles, & Vines, 2010). Parameter estimates were considered significant when the credible intervals did not include zero (Hadfield, 2010). Adjusted repeatability of behaviour across group sizes was then calculated using the variance components from the GLMMs described above (ratio of among-individual variation to total phenotypic variation) implemented using the 'rptR' package. Parametric bootstrapping provided 95% confidence intervals (CIs) and statistical significance was evaluated with likelihood ratio tests. Adjusted repeatability was also performed for each behavioural phenotype separately. Nonoverlapping CIs indicated that repeatabilities varied between behavioural phenotypes. Finally, best linear unbiased predictors (BLUPS) of the random effects were extracted from each of the GLMM models to test for correlation between personality and plasticity (within-individual variation) as a way of assessing the level of responsiveness among behavioural phenotypes. Negative correlations would be indicative of a high level of plasticity for individuals expressing a low behavioural phenotype (i.e. inactive, nonexploratory or shy) while a positive relationship would indicate that high behavioural phenotypes were more responsive to changes in group size. ANOVAs were then used to evaluate the difference in levels of plastic responses between behaviours. Where significant differences were identified, post hoc analyses (package: 'lsmeans') were performed to identify which paired comparisons were significant