Heritable personality variation is subject to fluctuating selection in many animal taxa; a major unresolved question is why this is the case. A parsimonious explanation must involve a general ecological process: a likely candidate is the omnipresent spatiotemporal variation in conspecific density. We tested whether spatiotemporal variation in density within and among nest box plots of great tits (Parus major) predicted variation in selection acting on exploratory behaviour (n = 48 episodes of selection). We found viability selection favouring faster explorers under lower densities but slower explorers under higher densities. Temporal variation in local density represented the primary factor explaining personality-related variation in viability selection. Importantly, birds did not anticipate changes in selection by means of adaptive density-dependent plasticity. This study thereby provides an unprecedented example of the key importance of the interplay between fluctuating selection and lack of adaptive behavioural plasticity in maintaining animal personality variation in the wild.

Summary1. An increase of competition among adults or nestlings usually negatively affects breeding output. Yet little is known about the differential effects that competition has on the offspring sexes. This could be important because it may influence parental reproductive decisions. 2. In sexual size dimorphic species, two main contradictory mechanisms are proposed regarding sex-specific effects of competition on nestling performance assuming that parents do not feed their chicks differentially: (i) the larger sex requires more resources to grow and is more sensitive to a deterioration of the rearing conditions ('costly sex hypothesis'); (ii) the larger sex has a competitive advantage in intra-brood competition and performs better under adverse conditions ('competitive advantage hypothesis'). 3. In the present study, we manipulated the level of sex-specific sibling competition in a great tit population ( Parus major ) by altering simultaneously the brood size and the brood sex ratio on two levels: the nest (competition for food among nestlings) and the woodlot where the parents breed (competition for food among adults). We investigated whether altered competition during the nestling phase affected nestling growth traits and survival in the nest and whether the effects differed between males, the larger sex, and females. 4. We found a strong negative and sex-specific effect of experimental brood size on all the nestling traits. In enlarged broods, sexual size dimorphism was smaller which may have resulted from biased mortality towards the less competitive individuals i.e. females of low condition. No effect of brood sex ratio on nestling growth traits was found. 5. Negative brood size effects on nestling traits were stronger in natural high-density areas but we could not confirm this experimentally. 6. Our results did not support the 'costly sex hypothesis' because males did not suffer from higher mortality under harsh conditions. The 'competitive advantage hypothesis' was also not fully supported because females did not suffer more in male-biased broods. 7. We conclude that male nestlings are not likely to be more expensive to raise, yet they have a size-related competitive advantage in large broods, leading to higher mortality of their on average lighter female nest mates.

Summary1. Costs and benefits of reproduction are central to life-history theory, and the outcome of reproductive trade-offs may depend greatly on the ecological conditions in which they are estimated. In this study, we propose that costs and benefits of reproduction are modulated by social effects, and consequently that selection on reproductive rates depends on the social environment. 2. We tested this hypothesis in a great tit Parus major population. Over 3 years, we altered parental reproductive effort via brood size manipulations (small, intermediate, large) and manipulated the local social environment via changes in the local fledgling density (decreased, increased) and the local sex ratio (female-biased, control, male-biased). 3. We found that male-biased treatment consistently increased the subsequent local breeding densities over the 3-year study period. We also found that parents rearing small broods in these male-biased plots had increased survival rates compared with the other experimental groups. 4. We conclude that reproductive costs are the product of an interaction between parental phenotypic quality after reproduction and the social environment: raising a small brood had long-lasting effects on some phenotypic traits of the parents and that this increased their survival chances in male-biased environment where habitat quality may have deteriorated (via increased disease ⁄ predation risk or intraspecific competition). 5. Our results provide the first experimental evidence that local sex ratio can affect reproductive costs and thus optimal clutch size.

Individuals of the same species differ consistently in risky actions. Such 'animal personality' variation is intriguing because behavioural flexibility is often assumed to be the norm. Recent theory predicts that between-individual differences in propensity to take risks should evolve if individuals differ in future fitness expectations: individuals with high long-term fitness expectations (i.e. that have much to lose) should behave consistently more cautious than individuals with lower expectations. Consequently, any manipulation of future fitness expectations should result in within-individual changes in risky behaviour in the direction predicted by this adaptive theory. We tested this prediction and confirmed experimentally that individuals indeed adjust their 'exploration behaviour', a proxy for risk-taking behaviour, to their future fitness expectations. We show for wild great tits (Parus major) that individuals with experimentally decreased survival probability become faster explorers (i.e. increase risk-taking behaviour) compared to individuals with increased survival probability. We also show, using quantitative genetics approaches, that non-genetic effects (i.e. permanent environment effects) underpin adaptive personality variation in this species. This study thereby confirms a key prediction of adaptive personality theory based on lifehistory trade-offs, and implies that selection may indeed favour the evolution of personalities in situations where individuals differ in future fitness expectations.

Negative density dependence of clutch size is a ubiquitous characteristic of avian populations and is partly due to within-individual phenotypic plasticity. Yet, very little is known about the extent to which individuals differ in their degree of phenotypic plasticity, whether such variation has a genetic basis and whether level of plasticity can thus evolve in response to selection. Using 18 years of data of a Dutch great tit population (Parus major), we show that females reduced clutch size with increasing population density (slopes of the reaction norms), differed strongly in their average clutch size (elevations of the reaction norms) at the population-mean density and that the latter variation was partly heritable. In contrast, we could not detect individual variation in phenotypic plasticity ('I 9 E'). Level of plasticity is thus not likely to evolve in response to selection in this population. Observed clutch sizes deviated more from the estimated individual reaction norms in certain years and densities, implying that the within-individual between-year variance (so-called residual variance) of clutch size was heterogeneous with respect to these factors. Given the observational nature of this study, experimental manipulation of density is now warranted to confirm the causality of the observed density effects. Our analyses demonstrate that failure to acknowledge this heterogeneity would have inflated the estimate of 'I 9 E' and led to misinterpretation of the data. This paper thereby emphasizes the fact that heterogeneity in residuals can provide biologically insightful information about the ecological processes underlying the data.