In this paper we explore the way foraging animals integrate experience
over time. The marginal value theorem shows that to maximize long-ter
m gain rate, foragers should adjust patch exploitation to the average
travel time for the habitat, and many experiments do find a positive r
elationship between average patch exploitation and average interpatch
travel time. This relationship implies that animals use experience to
determine foraging tactics but, by itself, does not imply that anythin
g but the most recent experience (say, the time taken to find the curr
ent patch) has an effect on behavior. We directly tested the influence
of events before the most recently experienced travel by examining ad
justments in foraging behavior after stepwise changes between two homo
geneous environments, each with a single travel distance. Using starli
ngs (Sturnus vulgaris) in a closed-economy laboratory simulation of a
patchy environment, we found that during periods of active foraging, t
he average number of prey per patch visit is in close agreement with t
hat predicted for rate maximization. After changes in travel time, bir
ds took approximately six full cycles of travel and patch use before r
eaching a new asymptotic behavior. The pattern of adjustment did not v
ary with successive presentations of the environmental change. These r
esults demonstrate that memory for more than one travel episode is inv
olved in the foraging decisions of starlings. We relate our results to
apparently conflicting data from previous experiments and to models o
f memory and information processing.