The rapid response of animals to changes in predation risk has allowed behavioral ecologists to learn much about antipredator decision making. A largely unappreciated aspect of such decision making, however, is that it may be fundamentally driven by the very thing that allows it to be so readily studied: temporal variation in risk. We show theoretically that temporal variability in risk leaves animals with the problem of allocating feeding and antipredator efforts across different risk situations. Our analysis suggests that an animal should exhibit its greatest antipredator behavior in high-risk situations that are brief and infrequent. An animal should also allocate more antipredator effort to high-risk situations and more feeding to low-risk situations, with an increase in the relative degree of risk in high-risk situations. However, the need to feed leaves an animal with little choice but to decrease its allocation of antipredator effort to high-risk situations as they become more frequent or lengthy; here, antipredator effort in low-risk situations may drop to low levels as an animal allocates as much feeding as possible to brief periods of low risk. These conclusions hold under various scenarios of interrupted feeding, state-dependent behavior, and stochastic variation in risk situations. Our analysis also suggests that a common experimental protocol, in which prey animals are maintained under low risk and then exposed to a brief "pulse" of high risk, is likely to overestimate the intensity of antipredator behavior expected under field situations or chronic exposure to high risk.

No abstract

No abstract

Sentinels are group members that watch from prominent positions. Sentinel interchanges often appear orderly, and groups with sentinels rarely have zero or many sentinels. A dynamic game was constructed to examine if these observations about sentinels could be based on selfish actions by individual group members. In this game, each group member chose to forage or be a sentinel based on its own energetic state and the actions of others. Sentinels received a selfish antipredator benefit if their ability to detect approaching predators more than compensated for their increased exposure to undetected predators. Provided sentinels were relatively safe and that detection information spread to other group members when sentinels detected predators, sentinels appeared highly coordinated for all combinations of parameters. This apparent coordination was based on mutualism because each individual gained by being a sentinel when other group members were not (and foraging when other group members were being sentinels). The model was very robust, but exact level of sentinel behavior varied somewhat with changes in foraging and predation parameters. This model could best be tested by testing its assumptions about sentinel safety, foraging-predation trade-offs, and information transfer in groups.

Many studies document that individuals visually scan for predators less frequently when in the safety of larger groups. This widely replicated e¡ect has generally been explained in terms of distinct predator detection and risk-dilution e¡ects. We show that a strict distinction between detection and dilution disappears when information about attacks is imperfectly shared (as it is in reality). Furthermore, dilution and detection e¡ects change depending on when during an attack the predator targets a particular prey individual for pursuit. Realistic detection and dilution e¡ects probably interact with each other and also with the targeting behaviour of predators. Instead of considering detection and dilution e¡ects on vigilance, it may be more pro¢table to consider each prey's probability of being targeted during an attack and its probability of escaping if attacked. This perspective emphasizes that a full understanding of safety in numbers requires an understanding of predator targeting strategies.