Obtaining convincing evidence for spatial memory of natural food sources in wild animals is hard because the observer rarely knows as much about the available food as does the animal, and the ability of the animal to detect novel food sources is usually not measured. In this study, I took advantage of the scarcity of natural fruit sources in the subtropical winter to present a wild group of brown capuchin monkeys with a large-scale array of 15 feeding platforms spaced ca 200 m apart. With this array, I could control the location, productivity and renewal schedule of the major fruit sources used by these animals. Combining an independent measurement of their detection field for these platforms with the known locations of the platforms, I calculated the expected patterns of movement among platform sites by the group under various models of 'random' foraging. These expected patterns were compared to the actual spatial movements of the group. The capuchin group moved significantly more often toward closer platforms and in straighter lines than expected by any random search model using their observed detection field of 82 m, although their behaviour did agree with such models for unrealistically large search fields of 225-350 m. I infer that the movements of this study group are likely to be guided by spatial memory. However, straight-line movement and a preference for closer platforms are in general not convincing evidence for spatial memory unless the detection field of the forager for the resources is known. Copyright 1998 The Association for the Study of Animal Behaviour. Copyright 1998 The Association for the Study of Animal Behaviour.