Use of position tracking to infer social structure in rhesus macaques

Abstract

Rhesus Macaques have clearly defined social dominance hierarchies that can be identified from human observations of social interactions, but determining these social structures through real-time observation is expensive and time consuming. Automating both the observation of animal behavior and inference of social structure reduces the requirement for human observation while providing a greater breadth of data. Automatic tracking of individuals allows researchers to continuously collect data on research animals without requiring expensive real-time observation by trained coders, and eliminates missed observations due to limits in coder attention and reaction time. Our approach works by detecting key events in the tracking data which roughly correspond to behavioral events coded by trained observers. In this work we compare the results obtained by our tracking and analysis methods with a trained human observer, both in terms of events recorded and in the determined dominance hierarchy.