The hippocampus is critical for remembering locations in a wide variety of species, including humans. However, recent findings from monkeys following selective hippocampal lesions have been equivocal. To approximate more closely the situations in which rodents and birds are tested, we used a spatial memory task in which rhesus monkeys (Macaca mulatta) moved about freely in a large room, on a tether. We used MRI‐guided stereotaxic surgery to produce selective hippocampal lesions in five monkeys, and retained five unoperated control monkeys. In the study phase of each trial of the matching‐to‐location task, monkeys found food in one site in an array of identical foraging sites. During the test, which occurred after a delay, monkeys could return to the site where the food had been found during study to obtain more food. In Experiment 1, normal monkeys showed a small significant tendency to return directly to a site where they had previously found food that day. Operated monkeys showed no such matching tendency. In Experiment 2, further training produced reliable matching‐to‐location performance in both groups at short delays, but monkeys with selective hippocampal lesions rapidly forgot the location of the food. In Experiment 3, we tested whether monkeys used a “cognitive map” to encode the location of the hidden food, by requiring them to relocate the food from a starting location different from that used during study. As a group, monkeys were more accurate than expected by chance, indicating that they did encode the rewarded location with respect to allocentric landmarks; however, both groups of monkeys were significantly worse at relocating the food when required to approach from a different location. In Experiment 4, probe trials using symmetrical test arrays found no evidence for egocentric coding of the rewarded location. © 2004 Wiley‐Liss, Inc.