Abstract: 1. 1. The communication performance of 12 rhesus monkeys was investigated in a situation in which the rewards of both members of a pair of monkeys could not exceed chance levels unless the operator monkey responded to cues provided by the informant monkey which indicated the location of food. Each member of the pair was trained in both operator and informant roles in different phases of the experiment. Communication performance improved progressively to levels consistently above chance. However, communication learning appeared to be specific to the role in which the individual was trained, and when roles were reversed no evidence of transfer was obtained. Tests of foodsharing behaviour showed a substantial increase in the tendency to share food with the partner following communication training. This occurred however, only when the partner was the only social stimulus present; if another monkey was also present there was no evidence of preferential responses to the partner. In all phases of communication training, monkeys which were housed together performed more efficiently than did monkeys housed individually.2. 2. The acquisition of stimulus-producing responses was investigated by causing an opaque screen to remain in front of the informant unless the operator monkey pulled a vertical lever at the front of its restraining cage. Initially, operators responded immediately to the foodcarts, but with further testing there was a steady increase in the tendency to defer the response to the food-carts until the lever had been pulled, revealing the informant monkey.3. 3. Transfer of communication training was tested with new monkey informants, and with two inanimate stimuli, a mechanical puppet, and a stationary plaque. The latter two objects were placed behind the rewarded food-carts before each trial. There was clear evidence of positive transfer to each of these conditions, but marked differences among conditions were obtained. Performance with the monkeys averaged 76 per cent. correct, as compared with 62 and 40 per cent., with the puppet and the plaque, respectively.4. 4. To test the ability of trained operator monkeys to select the appropriate informant on the basis of behavioural cues, the communication situation was arranged so that two informant monkeys were present on all trials. However, on any trial only one of these informants could be rewarded, and the operator's rewards were contingent upon delivering food to this informant. Efficiency of discrimination began at approximately 45 per cent, (chance = 25 per cent. and improved progressively to levels above 75 per cent.