Classification of Social Signals in Squirrel Monkeys By Means of Cluster Analysis

Abstract

Abstract It is the chief aim of this project to reduce the extremely large number of observable actions of squirrel monkeys to a smaller number of classes of actions, which can be distinguished from each other in respect to their function in the communication processes of the animals. By means of such a classification, we want to increase the predictability of successive social signals. Using an agglomerative cluster method - in which the quantity trans information taken from information theory is responsible for the formation of the classes - such a classification based on two kinds of frequency matrices is found. That we actually obtain classes of actions having the same or at least similar meanings, is proven not only by way of the results derived from cluster-analysis, but also from an explicit study of the frequency matrices on which they are based. The biological relevance of the classifications we have found as well as the mathematical methods we have used is discussed.