Quantifying the adaptive value of learning in foraging behavior

The value of acquiring environmental information depends on the costs of collecting it and its utility. Foragers that search for patchily distributed resources may use experiences in previous patches to learn the habitat quality and adjust their behavior. We map the ecological landscape for the evolution of learning under a range of conditions, including both spatial and temporal heterogeneity. We compare the learning strategy with genetically fixed patch-leaving rules and with strategies of foragers that have free and perfect information about their environment. The model reveals that the efficiency of learning is highest when low encounter stochasticity results in reliable estimates of patch quality, when there is no or little temporal change, and when there is little spatial variability. This partially contrasts with the value of learning, which is highest when there is temporal change, because flexible strategies may track the environmental trend, and when there is spatial variability, because there is a need to distinguish between good and bad patches. Learning rules with short-term memory are beneficial when patch information is accurate and when there is temporal change, whereas learning rules that update slowly are generally more robust to spatial variability.