We examined the ability of capuchin monkeys to use video without immediate visual-kinaesthetic feedback as a source of information to guide their action in the 3-dimensional world. In experiment 1, 2 capuchins learned to retrieve food under 1 of 2 different objects in 1 cage after watching the experimenter hiding food under 1 of 2 replica objects while in another cage. Information space and retrieval space were thus separate. The performance criterion was 71% first correct choices in blocks of 24 trials. However, when the subjects watched prerecorded videos of the hiding events, they chose randomly. In experiment 2, we gave the capuchins further trials with video and we enhanced the object shapes by line drawings. One capuchin eventually learned to use the video clips to locate food and he generalized this learning to 2 new objects.