DICKEY, B. F.*: Quantification of ‘Consistency’ in Social Interactions

Studies of social behavior use a consensus definition of social dominance: a dominant animal consistently wins interactions against another. A sequence of interaction outcomes is consistent if 1) an animal wins a great majority of the interactions, or 2) neither animal wins a great majority but each animal wins series of consecutive interactions. I therefore suggest two statistical tests for consistency: binomial probability to test for one animal winning more interactions by chance, and runs analysis to test for clumping of wins in time. Using data from crayfish dyads I demonstrate these techniques and show that novel conclusions can be made based on these tests. To further evaluate these techniques, I analyzed 6388 dyads from 103 published dominance matrices. The dominant won significantly more interactions in 38% of dyads studied. A further 30% involved the dominant animal winning all the interactions, but fewer than the five required to determine dominance. Only 5% of dyads were candidates for runs analysis, but these were often dyads with large numbers of interactions where dominance was not established using the binomial test. Runs analysis may therefore be useful for the identification of instability in dyads. The implications for metrics of linearity and the possibilities of group-level measures of dyadic stability are discussed.