Portal de Periódicos da CAPES

Changes in behavioural trait integration following rapid ecotype divergence in an aquatic isopod

2011; Oxford University Press; Volume: 24; Issue: 9 Linguagem: Inglês

10.1111/j.1420-9101.2011.02322.x

1420-9101

Shannon L. Harris, Fabrice Eroukhmanoff, K. K. Green, Erik Svensson, Lars Pettersson,

Genetic diversity and population structure

Journal of Evolutionary BiologyVolume 24, Issue 9 p. 1887-1896 Free Access Changes in behavioural trait integration following rapid ecotype divergence in an aquatic isopod S. HARRIS, S. HARRIS Department of Biology, Animal Ecology, Lund University, Lund, SwedenSearch for more papers by this authorF. EROUKHMANOFF, F. EROUKHMANOFF CEES, Department of Biology, University of Oslo, Blindern, Oslo, NorwaySearch for more papers by this authorK. K. GREEN, K. K. GREEN Department of Biology, Animal Ecology, Lund University, Lund, SwedenSearch for more papers by this authorE. I. SVENSSON, E. I. SVENSSON Department of Biology, Animal Ecology, Lund University, Lund, SwedenSearch for more papers by this authorL. B. PETTERSSON, L. B. PETTERSSON Department of Biology, Animal Ecology, Lund University, Lund, SwedenSearch for more papers by this author S. HARRIS, S. HARRIS Department of Biology, Animal Ecology, Lund University, Lund, SwedenSearch for more papers by this authorF. EROUKHMANOFF, F. EROUKHMANOFF CEES, Department of Biology, University of Oslo, Blindern, Oslo, NorwaySearch for more papers by this authorK. K. GREEN, K. K. GREEN Department of Biology, Animal Ecology, Lund University, Lund, SwedenSearch for more papers by this authorE. I. SVENSSON, E. I. SVENSSON Department of Biology, Animal Ecology, Lund University, Lund, SwedenSearch for more papers by this authorL. B. PETTERSSON, L. B. PETTERSSON Department of Biology, Animal Ecology, Lund University, Lund, SwedenSearch for more papers by this author First published: 10 June 2011 https://doi.org/10.1111/j.1420-9101.2011.02322.xCitations: 14 Sanna Harris, Department of Biology, Animal Ecology, Lund University, SE-223 62 Lund, Sweden. Tel.: +46 46 222 17 81; fax: +46 46 222 47 16; e-mail: sanna.harris@biol.lu.se AboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinked InRedditWechat Abstract Colonization of new habitats can relax selection pressures, and traits or trait combinations no longer selected for might become reduced or lost. We investigated behavioural differentiation and behavioural trait integration in the freshwater isopod Asellus aquaticus. This isopod has recently colonized a novel habitat and diverged into two ecotypes which encounter different predator faunas. We investigated sex-specific behavioural differences and phenotypic integration in three behavioural assays: (i) time to emerge (TE) from a shelter, (ii) activity and (iii) escape behaviour. General activity and escape behaviour differed between ecotypes. Furthermore, general activity and TE differed between sexes. Behavioural traits were more frequently correlated in the ancestral habitat, and phenotypic integration tended to be higher in this habitat as well. Our study suggests that different predator types, but also other ecological factors such as habitat matrices and population densities, might explain the differences in behavioural integration in these ecotypes. Introduction In both aquatic and terrestrial systems, predation is a major mortality risk (Lima & Dill, 1990), and predator-mediated selection has been demonstrated to induce changes in prey morphology, life history and behaviour (Pettersson et al., 2000; Reznick et al., 2001; Stoks et al., 2003; Langerhans et al., 2004; Svensson & Friberg, 2007). When a prey population is colonizing a new habitat, predator–prey interactions can rapidly change, resulting in prey divergence (Magurran et al., 1992; Vamosi, 2005; Stoks & McPeek, 2006; Eroukhmanoff & Svensson, 2009). For instance, some predator species may be absent in the novel environment and some new species present, resulting in that the colonizing prey population becomes exposed to a novel predator community. The net effect of such environmental shifts may result in relaxed selection pressures on antipredator traits (for review Lahti et al., 2009). As a consequence, antipredator traits that were maintained and expressed in the former selective environment can become reduced or entirely lost in the novel environment, if they are no longer important to prey fitness (Coss, 1999; Blumstein & Daniel, 2005; Cresko, 2008; Lahti et al., 2009). The opposite direction can also be true, i.e. instead of generating relaxed selection, environmental shifts might increase selection pressures in the colonizing population depending on the predation regime in the ancestral habitat. Although many studies have quantified differences in mean trait values between ancestral environments and novel environments (e.g. Reznick et al., 1997), relatively few have investigated how novel selective regimes might favour new trait correlations or decouple old maladaptive trait correlations in prey. This is important, as predators are unlikely to select for single traits in isolation, but are more likely to favour certain combinations of traits (Brodie, 1992; Svensson et al., 2001; Sinervo & Svensson, 2002; Mikolajewski et al., 2006). Most studies of correlational selection have focused on correlations between morphological characters (Sinervo & Svensson, 2002), but there is now evidence that behavioural correlations can be addressed using a similar approach, particularly in the context of so-called behavioural syndromes (Sih et al., 2004a,b; Bell, 2007; Réale et al., 2007). Behavioural syndromes may be defined as a suite of correlated behaviours across different contexts and situations (Sih et al., 2004a,b). A population or species can exhibit a behavioural syndrome, as the positive correlation between aggressiveness and boldness seen in sticklebacks in the presence of predators (Huntingford, 1976; Bell, 2005), or the correlation between exploration and dispersal in great tits (Dingemanse et al., 2003). Two hypotheses have been proposed to explain the evolution of behavioural syndromes (Bell, 2005). The constraint hypothesis suggests that these syndromes are mainly the result of the robustness of the underlying genetic correlations and are thus not necessarily adaptive (Sih et al., 2004b). Moreover, the constancy of the covariance structure of a set of traits might also constrain their individual evolution towards a new optimum (Price & Langen, 1992; Lynch & Walsh, 1998; Sih et al., 2004b). One prediction from this hypothesis is that correlated behaviours should be expressed in the same direction across populations as within populations (Bell, 2005). The alternative, adaptive hypothesis suggests that different selective environments will favour different adaptive combinations of traits, and selection will decouple traits if traits combinations become maladaptive (Bell, 2005). Empirical support for the adaptive hypothesis has been provided by studies of different populations of sticklebacks from different predation environments (Bell, 2005; Dingemanse et al., 2007). In predator-sympatric populations, the correlation between boldness and aggressiveness was significantly tighter than in predator-naïve populations (Bell, 2005; Dingemanse et al., 2007). In an elegant experimental study of these patterns, predator-naive individuals were exposed to real predation, and it became evident that predation, possibly in combination with behavioural plasticity, can induce a phenotypic correlation between boldness and aggressiveness (Bell & Sih, 2007). Hence, predators can strengthen and favour particular behavioural correlations in prey populations (Bell & Sih, 2007; Dingemanse et al., 2007). The aquatic isopod Asellus aquaticus is an example of a species that is strongly affected by predation and which has recently established itself in a novel selective environment. It has rapidly diverged phenotypically and in parallel into two distinct ecotypes in several different south Swedish lakes (Hargeby et al., 2004, 2005b; Eroukhmanoff et al., 2009a). This ecologically and phenotypically variable isopod species provides an excellent opportunity to address questions about behavioural correlations and differentiation in behavioural traits following the colonization of a novel environment. Rapid phenotypic changes in A. aquaticus morphology and behaviours have taken place over the last two decades (or approximately 40 isopod generations), when isopods from reed stands colonized a newly emerged habitat consisting of submerged macrophytes (Hargeby et al., 1994, 2007; Eroukhmanoff et al., 2009a,b). Parallel phenotypic divergence in pigmentation (Hargeby et al., 2004, 2005b; Eroukhmanoff et al., 2009a), morphology (Eroukhmanoff & Svensson, 2009) and sexual behaviour (Eroukhmanoff et al., 2009a; Karlsson et al., 2010a) has recently been demonstrated. The changes in morphology have an underlying additive genetic basis, suggesting that microevolutionary change was involved and not only phenotypic plasticity (Eroukhmanoff et al., 2009b). In addition, the ecotypes are at least partly reproductively isolated, and gene flow is probably quite reduced (Eroukhmanoff et al., 2011). Furthermore, in the vegetation-free areas between the reed and the stonewort habitats, A. aquaticus is absent (Hargeby et al., 1994). The ecological shifts between habitats involved a shift from a predation regime dominated by mainly invertebrate predators to a novel predation regime involving fish predators, and this has been suggested to be the main selective factor behind these phenotypic changes (Hargeby et al., 2004, 2005b; Eroukhmanoff & Svensson, 2009). Predation from visually hunting fish presumably favours smaller and brighter isopods in the novel habitat compared with larger and darker source populations in the reed where predators relying on tactile cues are the main threat (Hargeby et al., 2004; Eroukhmanoff & Svensson, 2009). However, in addition to the predator community, other ecological factors that differ between the habitats, such as habitat complexity and population density, might also have an influence in this system (Karlsson et al., 2010b). Here, we have investigated behavioural correlation and differentiation between behavioural traits in these isopods, in three behavioural assays: (i) latency to emerge from a shelter, (ii) general activity in a novel environment and (iii) escape behaviour under predation risk. We also quantified the degree of phenotypic integration in behavioural traits and discuss our findings in relation to the selective regimes in the different habitats. Materials and methods Study organism ecology and general field work procedures The isopod A. aquaticus is commonly found in lakes, ponds and slow-flowing streams across Europe and Asia (Hargeby et al., 2004; Verovnik et al., 2005). In lakes, A. aquaticus typically inhabit the reed belts (Phragmites australis) along the shores, feeding on decaying leaves (Adcock, 1982). In two shallow Swedish lakes, Lake Krankesjön (55°42′N, 13°28′E) and Lake Tåkern (58°21′N, 14°50′E), repeated and dramatic shifts between a phytoplankton-dominated state with turbid water towards a macrophyte-dominated state with clear water have occurred during the past 20 years (reviewed in Hargeby et al., 2007). These ecological shifts resulted in colonization of bare sediment areas in the limnetic zone of stonewort (mainly Chara tomentosa,Hargeby et al., 1994, 2007). Shortly after the emergence of this stonewort habitat, isopods from the reed populations colonized the novel stonewort habitat and established large populations (Hargeby, 1990; Hargeby et al., 2004). In the reed, isopods are generally darker and larger in size, compared with the stonewort isopods that are lighter pigmented and smaller in body size (Hargeby et al., 2004, 2005b; Eroukhmanoff et al., 2009a). Aquatic isopods are common prey for both fish (Rask & Hiisivuori, 1985) and invertebrate predators (Thompson, 1978). In the reed belts, invertebrate predators (e.g. dragonfly larvae: Aeshna spp., Cordulia aenea) are more common than predatory fish (Eroukhmanoff & Svensson, 2009). In contrast, fish predators (mainly perch, Perca fluviatilis) are common in the stonewort (Hargeby et al., 2004, 2005a), whereas dragonfly larvae are scarce (Eroukhmanoff & Svensson, 2009). In spring 2008, we collected isopods from both habitats (reed and stonewort) in Lake Krankesjön in the middle of their reproductive season (April–May). All individuals were caught as pairs in precopula (amplexus), a state where the male is holding the female until she moults and is ready to be fertilized (Hargeby et al., 2004). We used only pairs because then we could easily distinguish between the sexes, and by doing so, we avoided the process of sex determination of individuals in the laboratory which sometimes can lower the condition of the animal. In the laboratory, pairs were carefully separated, and males and females were placed singly in containers filled with lake water. Wild-caught isopods were allowed to adjust to laboratory conditions for 24 h prior to behavioural testing. Before the behavioural experiments started, each individual was photographed in a Petri dish with water against a metric reference using a Nikon Coolpix 4500 digital camera. Total lengths of all individuals were quantified using the image analysis software ImageJ (Rasband, 2005), and length measurements were taken dorsally from end to end. The various behavioural assays were carried out in a fixed order with first testing time to emerge (TE), then activity and finally escape behaviour. The underlying rationale was that TE and activity were assumed to have least carry-over effects on subsequent trials, whereas the escape behaviour trial could influence subsequent trials. An hour elapsed between each behavioural observation. Latency to emerge from a shelter The experimental arena consisted of a white plastic container [35 × 26 × 13 cm (length × width × height)] filled up with lake water to the level of 0.5 cm. The water temperature was maintained at room temperature (+20 °C). In the middle of the arena, we placed a refuge (9 × 9 × 2.5 cm) made of grey PVC plastic which was held up by small pins in each corner. The refuge was perforated in the centre where we placed a black plastic cylinder of 3.4 cm diameter, which served as an acclimation compartment. Small weights were glued on top of the refuge so that the acclimation compartment could easily be pulled up and down without moving the whole shelter. Each trial was started by gently placing the isopod in the acclimation compartment, covering the top by a removable lid and then leaving the isopod to acclimate for 1 min. After acclimation, we slowly lifted the cylinder clear of the water (the cylinder still had the lid on and was attached to the refuge), and the isopod was allowed to leave the shelter. TE from the shelter was defined as the time until the isopod's head passed any of the borders of the refuge (Harris et al., 2010). Individuals that failed to emerge within 5 min were given the maximum score of 300 s. (63 of 375 individuals were given the maximum score.) The water was replaced between trials for all behavioural assays to remove any possible chemical cues that could affect behaviour. Latency to exit a refuge is a measure commonly used to estimate an individual's boldness, i.e. its risk-taking behaviour (cf. Herczeg et al., 2009; Harris et al., 2010; Wilson & Godin, 2010). General activity Each individual was placed in a white plastic bucket (18 cm diameter) that was filled up with lake water to the level of approximately 0.5 cm. The isopod was left to settle for 1 min, after which we recorded the total time moving (TM) during 2 min. We also counted the number of stops (NS), i.e. the times the isopod stopped and maintained a still position (cf. Bell, 2005; Wilson & Godin, 2010). Escape behaviour under predation risk To measure escape behaviours, each individual was placed in a white plastic bucket (18 cm diameter) filled up with lake water to the level of 0.5 cm. Individuals were then constantly poked by a small metal rod [13 cm (length), 0.4 cm (diameter)] for 15 s to simulate a predator attack (cf. Eroukhmanoff & Svensson, 2009). When the poking stopped, we recorded (i) the time the isopod continued running [Time Running (TR)] to escape from the 'threat' and (ii) how long it maintained a still position [Time Freezing (TF)] after the run, until it moved again. The maximum behavioural observation time was set to 5 min, but none of the individuals passed this limit. Statistical analysis To estimate behavioural differences between ecotypes in relation to habitat and sex in all three measured behavioural assays, we used a general linear model (GLM). TE, activity (TM; NS) and escape behaviour (TR; TF) were the dependent variables. Sex and habitat and an interaction term between sex and habitat were included as fixed factors, and total length of each individual was included as a covariate. The covariate and covariate × main factor interactions turned out to be nonsignificant, and the full models were therefore reduced by omitting the covariate and covariate × main factor interactions. Scores for TE, TR and TF were log-transformed prior to the analyses to achieve normality of the residuals. We estimated Pearson correlations between the five behavioural traits for each habitat and sex separately (see Fig. 2). For this part, our alpha level was corrected according to the sequential Bonferroni adjustment (Rice, 1989). To assess whether ecotypes differed in the relationship between the behaviours, we compared the correlation coefficients statistically using χ2-tests (Zar, 1996, p. 386). We also estimated the overall level of phenotypic integration between the behavioural traits of the ecotypes for the five behavioural traits (Proulx et al., 2005; see also Eroukhmanoff & Svensson, 2008, 2009). Phenotypic integration refers to genetic, developmental or functional relationships between traits resulting from the interplay between constraints and selection (Pigliucci, 2003). We estimated the integration levels in each sex and ecotype using two types of phenotypic integration indices: the variance of the eigenvalues of the correlation matrix (Wagner, 1984) and the average of the absolute phenotypic correlations (Cane, 1993). The average connectivity was calculated per trait for each ecotype and sex. To quantify differences in connectivity, we compared the sexes within ecotypes using a GLM with connectivity as the dependent variable, habitat and sex and an interaction term between sex and habitat as fixed factors. Figure 2Open in figure viewerPowerPoint Phenotypic correlation graphs for behavioural traits of both sexes and ecotypes. All the correlations at (P < 0.05) are shown (for Bonferroni correction see Table 1). Abbreviations for the behavioural traits: TE, Time to Emerge; TM, Time Moving; NS, Number of Stops; TF, Time Freezing, TR, Time Running. It should be noted that the experimental design generated dependencies in the data set. In two of the trials, we measured more than a single behaviour (TM and the NS; TR and TF, respectively); hence, these pairs were dependent. Often, such dependencies are dealt with by reducing the data set using, e.g. a principal component analysis (cf. Bell, 2005), or by selecting a subset of the data set. Here, we explicitly included all measures in the design to explore potential and realized dependencies. Using correlation graphs, we compared the expressed phenotypes with the pattern expected from an a priori null model where dependencies were purely a consequence of the experimental design. Therefore, we predicted correlations between behaviours measured in the same behavioural assay such as for activity as well as escape behaviour because they are recorded together in time. In total, our complete data sets contained observations of individual behaviour in all three assays for 375 individuals (reed: 109 females (f); 82 males (m) and stonewort: 91 f; 93 m). The average length (mean ± SE) of reed females was 7.6 ± 0.1 mm, whereas the value for males was 9.9 ± 0.2 mm. Stonewort females were 7.5 ± 0.1 mm, and stonewort males were 10.5 ± 0.1 mm. All statistical analyses were performed in spss 15.0 for Windows (SPSS Inc., Chicago, USA). Results Overall, the results from the multivariate GLM revealed significant differences in behaviour between the two habitats (Wilk's λ = 0.926, F5,367 = 5.91, P < 0.001) and the sexes (Wilk's λ = 0.942, F5,367 = 4.49, P = 0.001), but no significant interaction between habitat and sex (Wilk's λ = 0.979, F5,367 = 1.58, P = 0.166). As the fixed factors were significant in the multivariate analysis, we followed up these multivariate analyses by evaluating all univariate effects of sex and habitat on each behavioural variable. Sex differences in latency to emerge Time to emerge from the shelter differed significantly between the sexes (F1,371 = 11.82, P = 0.001; Fig. 1a). Males exhibited shorter latencies to exit the refuge compared with females (mean TE ± SE: males = 95 s ± 7.4, females = 125 s ± 7.6; Fig. 1a). There was no significant habitat effect (F1,371 = 0.013, P = 0.911) or habitat × sex effect (F1,371 = 0.010, P = 0.919). Figure 1Open in figure viewerPowerPoint The Village Bay study population on Hirta shows dramatic fluctuations in population size. Every few years, a combination of high population density, poor weather, low food availability and high parasite burdens cause a sharp decline in population size, known as 'crashes'. Points show population size estimated in October of the year indicated, with filled symbols indicating that the previous winter was not a crash, and open symbols showing years where the preceding winter was a crash. Habitat differences and sex differences in activity The number of times the isopod stopped and maintained a still position differed significantly between isopods from the two habitats (F1,371 = 19.68, P < 0.001; Fig. 1b). Reed isopods stopped more often (mean NS ± SE: reed = 6.8 ± 0.31, stonewort = 5.1 ± 0.28). The NS also differed between the sexes (F1,371 = 5.82, P = 0.016; Fig. 1b), where males stopped more frequently than females (mean NS ± SE: males = 6.4 ± 0.3, females = 5.5 ± 0.3). There was a significant interaction between habitat × sex (F1,371 = 3.76, P = 0.05), which was mainly driven by the males in the reed (Fig. 1b). The total TM did not differ between either habitat or sex (F1,371 = 0.01, P = 0.913; F1,371 = 0.968, P = 0.326, respectively). Escape behaviour under predation risk The time isopods spent on running away after the simulated predatory attacks differed significantly between habitats (F1,371 = 4.52, P = 0.034; Fig. 1c). Stonewort isopods escaped by running for almost twice as long time as individuals from the reed (mean TR ± SE: stonewort = 3.31 s ± 0.43, reed = 1.74 s ± 0.11). Moreover, males tended to flee for a longer time than females (mean TR 3.0 s vs. 2.1 s, respectively; F1,371 = 3.52, P = 0.06). No significant habitat × sex interaction was found (F1,371 = 2.79, P = 0.095), and there were no significant differences in freezing time (habitat: F1,371 = 1.90, P = 0.169; sex: F1,371 = 1.68, P = 0.195). Behavioural correlations and phenotypic integration In general, phenotypic correlations between the behavioural traits were stronger and more significant in the reed population compared with the stonewort population, particularly for the reed females (Fig. 2, Table 1). This was further confirmed from the phenotypic integration indices of the behavioural traits, which revealed higher integration indices in the reed isopods compared with the stonewort isopods, with no obvious sex differences (Fig. 2, Table 2). The mean connectivity per trait was also higher in the reed isopods (Table 2), and we found a significant habitat effect (F1,16 = 8.05, P = 0.012), but no effect of sex (F1,16 = 0.05, P = 0.83) or sex × habitat (F1,16 = 1.19, P = 0.29). Table 1. Pearson correlations (r) between behavioural traits for each habitat and sex. Abbreviations for the behavioural traits: TE, Time to Emerge; NS, Number of Stops; TM, Time Moving; TR, Time Running; TF, Time Freezing. TE NS TM TR TF Females (Reed n = 109; Stonewort n = 91) TE – −0.238* −0.272**† −0.096 0.233* NS −0.060 – 0.485**† −0.144 −0.192* TM −0.347**† 0.416**† – −0.003 −0.203* TR −0.124 0.060 0.159 – −0.254** TF 0.046 0.166 0.021 0.010 – Males (Reed n = 82; Stonewort n = 93) TE – −0.266* −0.478**† −0.082 0.115 NS −0.361**† – 0.537**† −0.162 −0.299**† TM −0.267** 0.336**† – −0.136 −0.223* TR −0.094 −0.080 −0.030 – −0.200 TF 0.060 −0.174 0.015 −0.069 – Values for Reed correlations are above the diagonal, values for Stonewort correlations are below the diagonal. Significant correlations are marked in bold (*P < 0.05, **P < 0.01). †Correlations between traits that are significant after the sequential Bonferroni procedure. Table 2. Phenotypic integration indices and connectivity for behavioural traits in each ecotype. The reed ecotype shows a net tendency for higher integration patterns than the stonewort ecotype, but limited differences between sexes. Connectivity was higher in the reed, but there were no sex differences. I1 is based on the variance of the eigenvalues (Wagner, 1984) and I2 on the average of the absolute values of all the correlations (Cane, 1993). Ecotype Sex I 1 I 2 Connectivity (± SE) Reed Male 0.415 0.250 2.0 (± 0.40) Reed Female 0.396 0.212 2.6 (± 0.55) Stonewort Male 0.182 0.173 1.2 (± 0.37) Stonewort Female 0.184 0.141 0.8 (± 0.49) After the sequential Bonferroni correction, some of the correlations in the reed males and females became nonsignificant (Table 1). However, when pairwise correlations were analysed for each ecotype (not separated per sex), several of the correlations in the reed remained significant even after the Bonferroni correction (Table S1). Thus, the overall correlation patterns differed within the habitats. Across all groups, TE and activity (TM) were negatively correlated. This means that individuals with shorter latencies to emerge also were more active in a novel environment (i.e. latency to emerge and activity were positively correlated into a boldness syndrome). In the reed, we found significant negative associations between antipredator behaviour (TF) and activity (NS; TM) (Fig. 2, Table 1). Individuals that froze for a longer period of time after a simulated predator attack were also less active in a novel environment. When compared with the a priori, null model expectations, one of the predicted correlations was detected (TM and NS, Fig. 2). The other a priori expectation, a correlation between TF and TR, was found in the reed females but not in the other three groups. We used the estimated correlation coefficients to compare if the four groups differed in the overall relationship between behaviours. The relationship between activity (NS) and antipredator behaviour (time freeze) was significantly different between the four groups (χ2 = 11.09, d.f. = 3, P = 0.004). A Tukey-type multiple comparison (Zar, 1996, p. 387) revealed that the reed males differed significantly from stonewort females (q = 4.37, k = 4, P = 0.011), and the comparison between the reed females and the stonewort females was close to significant (q = 3.58, k = 4, P = 0.055). The remaining correlation coefficients did not differ between the four groups (TE–NS, χ2 = 4.65, d.f. = 3, P = 0.10; TE–moving, χ2 = 3.37, d.f. = 3, P = 0.19; TE–time freeze, χ2 = 2.27, d.f. = 3, P = 0.32; moving–time freeze, χ2 = 4.94, d.f. = 3, P = 0.08). Discussion Behavioural differences between ecotypes and sexes Here, we have demonstrated sex-specific differences in the latency to leave a shelter across two different isopod ecotypes. Males exhibited a more risk-prone behaviour and left the refuge sooner than females (Fig. 1a). To hide in a refuge prevents detection from predators, but at the same time, there are costs of staying because of lost feeding and mating opportunities (Sih, 1992; Cooper & Frederick, 2007). In the marine isopod, Idotea baltica, there is a sex difference in microhabitat choice with males occurring more often than females on the exposed, apical and light-coloured parts of the macroalga, Fucus vesiculosus, which is suggested to decrease protection from visual predators (Merilaita & Jormalainen, 2000). Thus, males are more risk-prone and trade-off food and shelter differently than females. Growth seems to be an important factor in males, because it increases mating success (Jormalainen & Merilaita, 1995; Merilaita & Jormalainen, 2000; Vesakoski et al., 2008). Sex-specific differences in risk-taking behaviour have also been shown in other organisms such as fish (Magurran et al., 1992; Harris et al., 2010). In guppies, cautious females exhibited antipredator responses that were four times higher than the responses by males, and this pattern is in line with the sex differences in terms of optimal life-history strategies, with males being prone to take higher risks to obtain matings (Magurran & Seghers, 1994). Thus, this might also apply in the A. aquaticus system, and similar sex-specific life-history strategies may be driving behavioural patterns. Individual differences in general activity are often closely associated with tendencies to take risks (e.g. Wilson & Godin, 2010). We found that male activity patterns differed from female ones and that males from t