Capuchin monkeys (Cebus apella) share food even if separated by a mesh restraint. Pairs of capuchins were moved into a test chamber in which one of them received apple pieces for 20 min, and the other received carrot pieces for the next 20 min. Previous research had shown a correlation between the rate of food transfer in both directions across female-female dyads. The present study confirmed this result. Reciprocity across dyads can be explained, however, by symmetry in affiliative and tolerant tendencies between two individuals, provided these tendencies determine food sharing. The present study was designed to exclude this symmetry-based explanation by testing each pair (N=16) of adult females on six separate occasions. There existed a significant covariation across tests of sharing in both dyadic directions, a result unexplained by relationship symmetry. Moreover, control procedures (i.e. testing of a food possessor without a partner, or testing of two individuals with the same food or two different foods at the same time) indicated that behaviour during food trials is not fully explained by mutual attraction or aversion. The monkeys take the quality of their own and the partner's food into account, and possessors limit transfers of high-quality foods. Instead of a symmetry-based reciprocity explanation, a mediating role of memory is suggested, and a mirroring of social attitude between partners. (C) 2000 The Association for the Study of Animal Behaviour.