Collective behaviour

This work began with the 5-year interdisciplinary Wallenberg-funded project “Linking social behaviour to the brain”, a collaboration between the Maths and IT departments at Uppsala University, and the Zoology department at Stockholm University. This studied whether causal relationships existed between collective behaviour and brain size in guppies (Poecilia reticulata), a model species for the study of shoaling.

Three experiments were run: a comparison of natural populations exposed to high or low predation, resulting in differing levels of aggregation; artificial selection for brain size, and artificial selection for social behaviour.

New computational methods enabled a fine-grained comparison of rules of interaction in low and high predation guppies (Herbert-Read et al., 2017). Dimensionality reduction, however, was able to map these group-based measures onto well-known behavioural axes (Sumpter et al., 2018).

A problem arises when considering social traits: by definition they cannot be measured in isolation. We therefore formulated an experimental method for both sorting for social traits and estimating the trait’s repeatability (Szorkovszky et al., 2017). Three generations of selection on alignment was then enough to show significant differences (Kotrschal et al., 2020).

Subsequent work has focused on modelling of burst-and-glide swimming with social influence. A simple deterministic model of pairs, based on single spiking neurons, can exhibit chaotic intermittency (Gyllingberg et al., 2023). Ongoing work considers a probabilistic alternative, and the question of where apparent randomness in behaviour originates.