Information-statistical analysis of social interaction and communication: An analysis-of-variance approach

Abstract

Using informational statistics, an analysis-of-variance model is developed for separating the exclusive effects of two simultaneously present factors on the variability of the ongoing behaviour of a reference subject. In particular the following factors are considered: the preceding behaviour of the reference subject itself and the preceding behaviour of its partner. Effects due to the latter are usually regarded as representing communication. The model is compared with other information-statistical models for social interaction proposed in ethological research. Three properties are discussed: structural complexity, the rationale for identifying relevant effects and the efficiency in measuring them. With respect to measuring communication it is shown that several existing models confounded inter-individual and intra-individual effects in behaviour sequences. It is pointed out that different analytical frameworks (e.g. Markovian stochastic processes or analysis-of-variance) can use the same information-statistical formalism but give rise to different interpretations. Finally, the relation between the complexity of inter- and intra-individual effects in interaction sequences and the structure of information-statistical models is discussed. In Appendix I computational procedures are specified; in Appendix II a Monte Carlo procedure for testing observed variability measures is presented.