Social class influences degree of variance sensitivity in wild Siberian jays

Behavioral Ecology doi:10.1093/beheco/arq106 Advance Access publication 16 July 2010 Social class influences degree of variance sensitivity in wild Siberian jays Irja I. Ratikainen, Jonathan Wright, and Anahita J.N. Kazem Centre for Conservation Biology, Department of Biology, Realfagbygget, Norwegian University of Science and Technology, N-7491 Trondheim, Norway he world is a stochastic place due to variation in factors such as weather and the spatiotemporal distributions of food and predators. An optimal choice in foraging therefore includes consideration of not only the average or expected payoff but also the variation in those payoffs. Stochastic variance in foraging rewards may allow individuals to ‘‘gamble’’ for larger fitness returns than might be expected based solely on the average reward. The obvious risk is that more variable rewards are equally likely to provide much poorer fitness returns than the average. Theory on variance sensitivity, also termed risk sensitivity (see discussion of terminology in Ydenberg 2008) considers when such gambling or ‘‘variance-prone’’ behavior—as opposed to more conservative ‘‘variance-averse’’ behavior—might be adaptive (Real and Caraco 1986; McNamara and Houston 1992). More precisely, theory predicts that if fitness is not linearly dependent on energetic state, then the expected fitness increases from high versus low variance in intake rate are not the same, even if the average intake is the same (McNamara et al. 1991; Kacelnik and Bateson 1996). The energy budget rule states that when animals are in poor energetic state relative to the state required for survival, variance-prone foraging will be favored and, conversely, positive energy budgets should generally favor variance-averse foraging (Stephens 1981). The reason for this is that when an animal is in an energetic state with high starvation probabilities each successive intake of energy provides accelerating probabilities of survival. Therefore, any stochastic (symmetrical) variation in foraging success will, on average, increase mean fitness outcomes. Animals in a relatively high energetic state, on the other hand, will obtain decelerating fitness returns from each T Address correspondence to I.I. Ratikainen. E-mail: irja.ratikainen @bio.ntnu.no. A.J.N.K. Coauthor is now at Max Planck Institute for Evolutionary Anthropology, Deutscher Platz 6, Leipzig D-04103, Germany Received 7 January 2010; revised 27 April 2010; accepted 31 May 2010.
The Author 2010. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: unit of energy intake, and so variance in foraging success will on average decrease mean fitness. There is some evidence for the energy budget rule from laboratory studies showing statedependent switches between variance-averse and varianceprone behavior, but the evidence is far from consistent (see Kacelnik and Bateson 1996, 1997). It is, of course, not only the risk of starvation that can select for variance-prone behavior. McNamara et al. (1991) were the first to model how reproduction can affect variance sensitivity. A simple model by Bednekoff (1996) illustrates how varianceprone foraging is predicted below a threshold amount of energy required for individuals to reproduce, whereas variance-averse foraging is predicted once the state threshold has been reached. Essentially, any nonlinear relationship between an animal’s state and its fitness should result in variance sensitivity (McNamara et al. 1991). We might therefore also expect variance-sensitive responses to other thresholds in (energetic) state, such as those associated with achieving and maintaining social rank (see Kuznar and Frederick 2003). However, there have been few experimental tests to confirm this generalized logic of variance sensitivity (but see Hurly 2003). Because variance-sensitivity is expected to be highly state dependent, it is important that animals used in any experimental test are able to assess their own state in a context relevant to the situation in which the behavioral strategy evolved. This may be a problem for studies of variance sensitivity in the laboratory because it is hard to know how captive animals perceive and assess their own state. We cannot necessarily expect adaptive choices to be made by animals that do not have access to those features of the internal and external environment that they have been selected to use in assessments of their own state. It is therefore more plausible to assume that wild animals are able to assess their state in a way that is relevant to any experimental choices presented. However, there have been surprisingly few studies of variance sensitivity in the wild (but see Barkan 1990; Guillemette et al. 1992; Hurly and Oseen 1999; Hurly 2003). Variance sensitivity theory predicts that optimal foragers should take into account not only the mean but also the variance in rewards offered by alternative foraging options. Whereas a positive energy budget should favor variance aversion, a negative one should favor variance-prone choices. The risk of starvation is the most obvious fitness threshold that can select for variance-prone behavior, but additional thresholds may exist such as the energy required for reproduction. Previous studies of variance sensitivity have often been performed in captivity, and few have demonstrated the predicted state-dependent changes in individual variance preferences. We trained groups of wild Siberian jays (Perisoreus infaustus) to forage from one-shot feeders containing 3 colorcoded options differing only in the variance of reward sizes. Not only did we find variance-sensitive behavior under natural conditions but also, for the first time, demonstrate the presence of significant differences in variance sensitivity within groups. Breeders exhibited a preference for high-variance rewards, whereas unrelated subordinates within the same groups preferred low-variance options. These results did not reflect risk of starvation (indexed by body mass) but might be explained by the additional energy needs of breeders compared with subordinates prior to the breeding season. Key words: food hoarding, food storing, optimal foraging, risk sensitivity, stochastic environment. [Behav Ecol 21:1067–1072 (2010)] 1068 This study presented a wild population of Siberian jays (Perisoreus infaustus) with an experimental setup testing for variance sensitivity. We compare the foraging strategies employed by different classes of jays within social groups, and thereby test for differences in state thresholds between breeders versus subordinates as expected from theoretical models of variance sensitivity (McNamara et al. 1991; Bednekoff 1996; Hurly 2003). This design gave us a unique opportunity to explore the foraging decisions of animals in a natural environmental and social setting. MATERIALS AND METHODS Study species and population (...truncated)