Doctoral defense by Pranav Minasandra

All animals behave. Behaviour allows animals flexibility in dealing with heterogeneous, dynamic environments, and a key goal of the field of animal behaviour is to understand how, when, and why animals do what they do. To better understand behaviour, we can view it as a sequence of discrete behavioural states driven by a behavioural algorithm, a set of principles based on which an animal performs behavioural decision-making. I adopt a multi-time-scale perspective to explore behavioural algorithms from three species of mammals in the wild: meerkats (Suricata suricatta), white-nosed coatis (Nasua narica), and spotted hyenas (Crocuta crocuta), whose behaviours I inferred using accelerometer data. In this talk, I will demonstrate long time-scale structure in the behavioural sequences of all collared individuals of all three study species, showing that behaviour depends on past states of the animal much more than expected from any simple (i.e., Markovian) model of behaviour. This work is likely the most detailed description of long-time-range behavioural structure in wild-living animals reported in the literature. I will then highlight specific behaviours at a slow and a fast temporal scale, considering the 24 h activity patterns of the hyenas and the moment-to-moment vigilance dynamics of the meerkats, and explore various factors, especially social ones, that influence behavioural decisions at these scales.