The spatial memory and foraging strategies of four adult captive-born American black bears (Ursus americanus) were explored in four experiments using a simulated foraging task. In the first three experiments, each session consisted of two phases separated by a delay: During the exploration phase, subjects foraged among a set of baited and unbaited sites. During the delay, the same locations were rebaited and subjects were released again and allowed to search the sites (search phase). In Experiments 1a and 1b, different sites were baited each day and the interval between exploration and search was short (4 hr or 15 min). Subjects were not accurate at recovering the food items in either experiment. In Experiment 2, an "informed forager" paradigm was used in which one subject was given privileged knowledge about the location of the food during the exploration phase and was later released with an "uninformed" competitor during the search phase. The bears did not achieve above-chance recovery accuracy even in the presence of a competitor. In Experiment 3, the same two of four sites were continually baited and the bears were released simultaneously over a period of 20 days, with each baiting separated by 2 or 3 days. As a group, the bears' foraging accuracy with repeated baiting and longer intervals approached greater than chance accuracy. Results suggest some limitations on bears' use of spatial memory in captive environments, but reveal the potential for use of spatial memory over longer delays.