Effects of sex and mass
We find limited support for the hypothesis that among-individual behavioural differences are explained by sex and/or size (mass). Refitted univariate models provide no statistical support for differences in behavioural means between assigned sexes (Table 3), nor did we detect any significant sex:size interactions. However, we did detect significant main effects of size (mass) on OFT Track Length and OFT Area Covered . Both coefficients are positive meaning heavier individuals on average cover longer distances and a larger area in OFT trials. Estimates of VI and R from these refitted univariate models were very similar to those estimated unconditional on the extra fixed effects (Figure 2), providing further confirmation that neither sex nor size affects considerable among-individual behavioural variation in single traits. This same result also holds for the multivariate phenotype in that the correlation structure in IDsex:mass is very similar to that in ID (Figure 3). Moreover, the first vector ofIDsex:mass captures 65% (95% CI 47.96%-73.26%) of multivariate variance and has trait loadings almost identical to those of idmax (Figure 4). Finally, plotting individual behavioural scores onidmax confirms the absence of clustering by sex (Figure 5). While a net positive association between mean weight andidmax can be seen, it is also clear that size accounts for only a small fraction of the variation present. We present the full table of fixed effects estimates in Table S4 and the fullIDsex:mass . variance-covariance matrix from which the correlations are derived in Table S5.
DISCUSSION
Our repeated measure design provided strong statistical support for consistent among-individual differences in behaviour across the Open field (OFT) and Food and shelter (FST) assays. As expected, multivariate analyses also yielded evidence of strong individual-level correlations among the specific traits assayed. However, in asking whether the structure of multivariate behavioural variation was consistent witha priori expectations of an underlying axis of shy-bold personality variation, our results provided a somewhat mixed picture. Specifically, we did find a dominant leading eigen vector of IDthat explains most (61%) of individual level variation in, and covariation-among, traits. However, several of the behaviours measured load antagonistically on this axis, a pattern that does not match oura priori predictions for a simple shy-bold axis. Further analyses provided little statistical support for major contributions of size- or sex-effects to the personality variation described (though some size effects were found). Below we discuss each of these main results in the context of the existing literature and highlight some important caveats to our conclusions.
Behavioural repeatabilities were statistically significant in seven of the eight behavioural traits tested. This confirms the presence of personality variation in red cherry shrimp and shows it is readily detected using simple testing paradigms adapted from fish models. Although personality has yet to be widely tested in decapods, our results align with the conclusions of ) who demonstrated among-individual variation in startle response duration (a proxy of boldness) in male hermit crabs. Repeatability in traits associated with shy-bold and/or exploratory personality variation has also been demonstrated in the European crayfish (Astacus astacus ; , and the rock pool shrimp (Palaemon elegans ; . In these latter examples, boldness was also negatively correlated with resource holding potential, suggesting a trade-off whereby bolder individuals may be better at finding resources but less able to defend them in competition (. We do not yet know if similar relationships hold in N. heteropoda but investigating the functional significance of personality in relation to competition and other ecological processes ) in this species would be useful. More generally, testing for associations of behavioural profiles with individual life histories would permit scrutiny of whether personality variation is maintained through adaptive processes (e.g., resource allocation trade-offs ), and/or arises through differences in resource acquisition coupled to state-dependent behaviour .
While estimates clearly varied among traits, some repeatabilities were notably high in comparison to the wider animal personality literature. For example, estimates of R=0.63 and R=0.56 for OFT Track Lengthand Area Covered are high compared to a median behavioural repeatability of 0.37 reported by . This may, at least in part, reflect the use of a short inter-observation interval; R is known to decline as the time between observations increases (e.g., ). However, high repeatabilities also have some implications worth noting. First, R sets an upper limit for heritability (and so potential rate of selection response) and facilitates more accurate selection
. Selection on behaviour has been suggested as a way to improve growth and other production traits in commercial shrimp cultivation and in this context, high repeatability of the selection target would be advantageous. In a study of Pacific white shrimp (Litopenaeus vannamei ), found that individuals interacting more with feed had a lower latency of approach and consumed food more rapidly. In our FST assay, latency to enter the zone containing food in the FST was actually not repeatable (and so not heritable). However, if simple behavioural biomarkers of improved feeding could be identified, then selecting on these could be valuable in an industry where food waste is a major source of economic inefficiency .
Although the implications are perhaps less clear at present, personality variation has been widely linked to the likelihood of establishment and/or invasive spreading of species following accidental introduction For instance in the American signal crayfish (Pacifastacus leniusculu s), found repeatable differences in shy-bold type behaviour that predicted individual motivation to disperse by passing through weirs (though not successful at doing so). Invasion processes could therefore be considered a ‘selective filter’ whereby only individuals with appropriate combinations of personality and other traits can invade a novel environment successfully Working on the same species of crayfish, , found invasive populations were bolder and more aggressive on average than populations within their native range. However, this was only true where invasive populations were allopatric to native crayfish, and the extent to which differences were present prior to invasions, as opposed to emerging after as adaptations to a new environment is unclear . Moreover, whether population level repeatability to shy-bold type behaviours (as opposed to individual or population mean) influences invasive potential is unresolved. Low repeatability implies high plasticity- at least relative to ‘fixed’ individual differences- and there is a long-standing hypothesis that plasticity plays an important role in adaptation to novel environments following dispersal . A recent study found differences in behavioural repeatability between native and invasive species of nudibranchs that is consistent with this idea , but more empirical studies are needed to see if low behavioural repeatabilities reliably predict risk of invasive spread.
Although we found strong support for personality in red cherry shrimp, multivariate analyses show the structure of behavioural variation differs somewhat from our initial predictions. Our eigen decomposition of the ID matrix is consistent with the presence of a latent personality axis that describes the majority of among-individual (co)variation in the measured behavioural traits . This was expected: all traits analysed were chosen precisely because they are putatively measures of the same underlying shy-bold personality axis. Thus, the important result here is not the presence of correlation structure inID , but the overall ‘shape’ of that structure. Specifically, because traits were scaled such that higher numbers denoted putatively bolder phenotypes, we had also predicted that covariances inID , would be uniformly positive, and all traits would load on the main eigenvector with concordant signs. Some, but not all, relationships were as expected. For example, individuals that travelled further than average in the OFT, also have higher track length in the FST, cover more area (OFT), and freeze less (both assays). These behavioural characteristics meet a priori expectations for the bolder end of the shy-bold continuum indicating that exploratory individuals are associated with the propensity to take greater risks . However, the same individuals also tend to swim closer to the tank walls in the OFT (i.e., be more thigmotaxic) and take longer to visit the food zone in the FST, which are characteristics typically associated with shy personality types. Neither of these traits load significantly onidmax (based on 95% CI not overlapping zero) and among-individual variance for FST -(Food Latency) was not statistically supported in the univariate model. However, FST Time in Open loads significantly on idmax with a negative sign and is also moderately repeatable. This actually means individuals considered bolder and/or more exploratory in the OFT spend longer durations than average in the shelter during FST, a result that is counterintuitive. Very speculatively, it is possible that the shelter provided may have been perceived as a risky environment (rather than a safe one as intended). This could arise if, for instance, our shelter mimicked the type of structure used by drift-feeding and opportunistic foraging predatory fish . With this considered, it would be interesting to investigate whether a refuge with greater structural complexity and/or smaller open spaces may be preferred (as demonstrated in mud crabs Scylla serrata ; .
We find limited evidence for sex and size effects on behaviour and conclude that these aspects of state do not make a major contribution to personality (co)variation in our population. The absence of sex effects is perhaps somewhat surprising given the extensive evidence of behavioural sexual dimorphism in decapods. For example, male rock pool prawns (Palaemon elegans ) are bolder and more active than females , while aggression is sexually dimorphic in white shrimp (Penaeus vannamei ; , rock shrimp (Rhynchocinetes typus ;) and American lobsters (Homarus americanus ;). In these species, males tend to be more aggressive and able to monopolise food for longer durations relative to females. In finding an absence of sexual dimorphism here we acknowledge that uncertainty in sexing the shrimp reduces statistical power. Of the 53 shrimp tested, we ultimately classified 22 as being of unknown sex. These were, on average, smaller individuals than those assigned to male or female categories and likely to be younger (and potentially sexually immature). We therefore cannot exclude the possibility that our study partially conflates sex, age and maturation status in ways that mask any dimorphism. Nevertheless, we also note that several other decapod studies have reported an absence of sex effects on shy-bold type traits. For example, found no difference in mean boldness or activity between male and female Chinese mitten crabs Eriocheir sinensis, while sex did not predict startle response duration in hermit crabs Pagurus bernhardus .
We did find some evidence for size-dependent behaviour. Statistical support was limited to two traits in the OFT, with larger (heavier) individuals traveling slightly further and covering more area on average. These trait-specific effects drive a trend towards larger individuals having higher behaviour scores onidmax . This could potentially be explained by links between behavioural type and life-history strategy as proposed under heuristic frameworks such as the ‘Pace of Life Syndrome’ . For instance, high metabolic rate may be associated with bolder behaviour, increased resource acquisition, and faster growth leading to increased size , albeit at the likely cost of higher mortality risk (e.g. from predation; Wolf et al., 2007). At present we lack individual level data on life histories to test these hypothesised relationships. We also lack sufficient ecological data to explore (arguably) simpler explanations for size-dependence. For instance, argued size-dependent behaviour of mud crabs (Panopeus herbstii ) was linked to size-dependent predation risk. Small crabs used refuges more than large crabs and also increased use in the presence of predators. Other mechanisms proposed for generating size-behaviour relationships in decapods are related to mating traits. For instance, male-male competition can drive size-dependent mating tactics with larger males engaging more in mate guarding and contest behaviour while smaller rivals adopt exploratory mate searching tactics . Although we cannot yet rule out similar processes in cherry shrimp, it is notable we found no sex by size interactions on the behaviours assayed here (i.e., both larger males and females have higher Track Length and Area Covered in the OFT).
In conclusion, this is the first study to our knowledge that describes patterns of among-individual variation in putative measures of shy-bold variation in the cherry shrimp, N. heteropoda. Using simple assays widely applied to small fish models, our results are consistent with the verbal model of a single major personality axis underlying among-individual differences observed. However, the structure of this axis does not fully match a priori expectations of shy-bold. This could be because our initial assumptions of what constitutes ‘riskier’ behaviour in this species are incorrect. We also found no evidence for sexual dimorphism in behaviour and only limited support for size-behaviour relationships. Our description of personality variation in red cherry shrimp adds to the growing picture of this phenomenon in invertebrates generally and decapod crustaceans specifically . We therefore hope this study will set the foundations for future investigations of mechanism and functional significance in this emerging model system, as well as providing baseline data for more applied research across ecotoxicology, aquaculture, and invasion biology.
LITERATURE CITED