Significance of memory properties in prey choice decisions

Abstract

To forage efficiently in a spatially and temporally heterogeneous environment requires that an individual’s information from the immediate past is combined with information from the more distant past to track environmental change. We made use of a model involving exponentially devaluating weights for past events to emulate behaviour of the individual’s memory. As the devaluation rate increases, more weight is given to the most recent events. First, performance of individuals with different memory properties was tested in simulations in which two prey types with different profitabilities were available in different proportions. In a structurally stable prey environment a low memory devaluation rate gave better estimation of prey proportions than a high memory devaluation rate. In a highly variable environment, on the contrary, individuals with high devaluation rate could more quickly correct their estimates as prey availability changed, although this was achieved with the cost of high error rate of the estimate. Second, the ability to reliably assess relative abundances of the prey types proved to increase an individual’s success in prey choice (according to the decision rules by the optimal prey choice model). Third, in further simulations individuals were allowed to adjust their memory devaluation rate according to experience from their success in prey choice decisions in previous patches. We found that there was no need to adjust a high devaluation memory in a highly variable environment, but foragers starting with low devaluation value rather rapidly shifted to high devaluation rates. In a relatively stable environment the situation was reversed and finally all foragers used low devaluation rates. These results imply that the variation in estimation efficacy of prey availability may be critical in terms of optimal prey choice and thus memory properties should be included in examinations of prey choice. Including individual variation in foraging performance in individual-based models could increase our understanding of the consequences of these differences at the population and community levels.