Sociobiology of rhesus monkeys. II. Stochastics of social communication.

Abstract The behaviour of an individual depends upon preceding actions within the individual's interacting group. Such antecedent events do not, however, completely determine present behaviour. Rather, probabilistic constraints exist between events within a sequence of actions. Such sequential dependencies between the actions of group members indicate that social communication is taking place. These communicative constraints in sequences of behaviour may be represented by stochastic processes. The sequential constraints in the social communication of rhesus monkeys are given in a series of stochastic models that specify the probabilities of all events and all sequences of events. Estimates of these probabilities were obtained during a two-year field study of rhesus monkeys on Cayo Santiago, an island in the West Indies. Each model considers the constraints from an earlier antecedent than did the last model in the series. Thus, each model forms the null hypothesis for the next one, and is replaced by it because the new model reduces the uncertainty of our predictions of behaviour. This uncertainty is measured, using a method developed in the mathematical theory of communication. This measurement readily leads to several related ones, including an index of behavioural stereotypy. The results are consistent with the theory that rhesus monkeys base their social behaviour upon their memory for preceding events in their social group, and that this memory extends back beyond the immediately antecedent event. The behaviour of the monkeys, considered as isolated events, had an uncertainty of 4·8 bits, out of 6·9 bits maximum for a species with a repertoire of 120 behaviour patterns. Approximately 1·9 bits of information were picked up by the monkeys from the immediately antecedent behaviour in their social group, thus reducing the uncertainty of their behaviour to 2·9 bits. Further reductions resulted from information from still-earlier events in their social group, although the influences of such long-term constraints on behaviour were progressively more difficult to estimate. As a result of these stochastic contingencies, it was possible to measure the degree of stereotypy of the monkeys' behaviour. The index of stereotypy had a value of 0·3 for isolated events, 0·6 for events considering just the immediate antecedent behaviour in the group, and higher values for longer sequences. The development of such probabilistic models involved several sources of error and certain implicit assumptions. They should not be used without cognizance of these limitations. When properly used, however, they should enable one to predict the social behaviour of rhesus monkeys with a far greater accuracy than has previously been possible.