Collective Escape Dynamics and Leadership in Group-living Animals

Abstract

Group-living organisms across taxa coordinate their movement to evade threats or predators. However, how information about threats, often available only to a few individuals within the group, efficiently propagates among group members, and how animals use information about predators to coordinate their movement, remains less explored. In this thesis, we aim to study collective escape responses, information propagation, and context-dependent hierarchical leadership in collectively escaping groups, using both data and models. First, to study escape dynamics and emergent leadership, we develop a computational model of collective motion inspired by empirical observations. Classical models of collective motion assume animals interact with all the neighbours in their neighbourhood via local averaging rules. However, empirical studies suggest that animals interact through rules simpler than local averaging, such as stochastically aligning towards only one randomly chosen neighbour. Here, we study the role of stochastic interactions on group properties. We show that group cohesion and polarisation can be achieved in finite groups even when organisms randomly choose only one neighbour to interact with. We then adopt similar principles of stochastic neighbour selection in the models developed in subsequent chapters to study escape dynamics. When animals collectively respond to threats, it is difficult to know whether individuals are directly responding to the threat or to their neighbours’ responses. Next, we study high-resolution data from a controlled experimental setup with fish (tiger barbs), where an individual trained to a threat stimulus via aversive conditioning escapes the stimulus, thus precisely controlling which individuals react to the threat (and thus have information about the threat). We show that in a group of five fish with only one conditioned fish, the escape behaviour of one conditioned fish could trigger collective escape responses in all the fish. We use lagged cross-correlation analysis of the speeds to analyse information propagation and leadership. Under unperturbed conditions, we do not observe any hierarchical leadership. However, when we turn on the green light, and the conditioned fish responds by crossing the barrier, we observe a hierarchical transfer of information from the conditioned fish to the naive ones. Further, by using spatially-explicit agent-based model, we show that the hierarchical transfer of information occurs because, once the green light is turned on, the conditioned fish reduces its interaction strength with all the naive fish until it crosses the barrier, while the naive fish respond to the conditioned fish due to its rapid change in speed and direction. Finally, we investigate the collective responses of a sheep flock (Ovis aries) to a herding dog (border collie). We observed that the sheep flock remained highly cohesive throughout the herding events, consistent with the selfish herd effect, a known mechanism hypothesised to reduce predation risk. Sheep moved faster as the dog increased speed, while being highly polarised but less cohesive. This suggests that cohesion alone may not adequately explain the anti-predatory benefits of group living, especially in groups exhibiting synchronous collective motion, as seen in our sheep flock experiments. Using lagged cross-correlation analysis of the time series of the direction of different individuals, we identified a clear hierarchy among sheep in terms of their directional influence on the flock. We found that the average spatial position of a sheep along the front-back axis of group velocity strongly correlates with its influence on group movement. To explain these results, we developed a computational model in which sheep follow simple interaction rules: repulsion from the dog and a tendency to move towards and align with neighbours. This model reproduced empirically observed patterns. Consistent with experimental findings, the model predicts that the individuals at the front of the flock had greater directional influence on the group. Furthermore, we developed a null model of herding that excludes the dog’s chasing behaviour. Such a model fails to reproduce the hierarchical information flow, suggesting that the observed empirical patterns are characteristic of collective escape response. In summary, this thesis reveals that during the initial attack by predators, the information about the threat propagates via sudden changes in the speed of informed individuals. However, when the predator continuously chases the group, information spreads more strongly through changes in the direction of the individuals at the front. Further, we can use computational models to both explain these patterns and infer the broad nature of interactions among group members as they escape threats. Thus, combining results from all these studies, from highly controlled to natural settings, our study revealed some general principles of collective escape dynamics in group-living organisms.