Rhesus monkeys, Macaca mulatta, on the island of Cayo Santiago, Puerto Rico produce one or more of five acoustically distinctive calls when they find food. Three of these calls ('warbles', 'harmonic arches' and 'chirps') are produced by individuals finding high-quality, rare food items, whereas the other two calls ('coos' and 'grunts') are produced upon encountering lower-quality, common food items, and in non-food contexts as well. To determine how rhesus classify such acoustic variation, I conducted habituation experiments using a subset of the five call types. I designed experiments to reveal whether classification is based primarily on acoustic features or on the basis of a call's functional referent; caller identity was held constant within sessions. Habituation to 'warbles' transferred to 'harmonic arches', and vice versa. Thus, although these two calls are acoustically distinctive, they appeared to be perceptually clustered into one category based on referential similarities. In contrast, habituation to 'grunts' was followed by dishabituation to 'warbles' or 'harmonic arches', and habituation to 'warbles' or 'harmonic arches' was followed by dishabituation to 'grunts'. Dishabituation could be due to acoustic or referential differences. Significantly, the magnitude of the dishabituating response was asymmetric and depended upon the call type used in the habituation series. Thus, when subjects were habituated to 'grunts', they responded significantly more to tests of 'warbles' or 'harmonic arches' than when the sequence was reversed. These results suggest that for rhesus monkey food-associated calls, referential differences carry more weight during perceptual classification than do acoustical differences. Copyright 1998 The Association for the Study of Animal Behaviour. Copyright 1998 The Association for the Study of Animal Behaviour.