Abstract: Frames of reference (i.e. sets of loci defining spatial locations) determine animals' performances in object search tasks. Reference frames are used at different scales. Although much behavioural research has been conducted on search strategies in many animal species, relatively little has been done on nonhuman primates. The two experiments reported here focused on the relative strength and the level of functioning of different reference frames at the small-scale level in four capuchins (Cebus apella). Two identical boxes and a landmark were placed on a round platform that could be rotated. A reward was hidden in subject's view under one box, and then a sash-screen was lowered to hide the rotation of the platform; the sash-screen was then lifted and the subject allowed to search for the reward. In experiment 1 the rewarded box was always the closer to the landmark, in experiment 2 it could be either the box closer to or the box farther from the landmark. Capuchins were successful after invisible rotations in experiment 1, but they failed after invisible rotations in experiment 2. Two possible explanations are proposed: (1) capuchins relied heavily on the left-right body-axis as a frame, and they could only substitute it with a simple association between the rewarded position and the landmark; or (2) capuchins failed because they chose external cues in the room, therefore on a inappropriate scale. The latter explanation allows two further inferences: (a) the capuchins' choice was indirectly related to their body-axes; and (b) the capuchins revealed a cognitive asymmetry between small-scale and large-scale spaces, thus differing from humans.