One of the main challenges in the study of social networks in vertebrates is to close the gap between group patterns and dynamics. Usually scan samples or transect data are recorded to provide information about social patterns of animals, but these techniques themselves do not shed much light on the underlying dynamics of such groups. Here we show an approach which captures the fission-fusion dynamics of a fish population in the wild and demonstrates how the gap between pattern and dynamics may be closed. Our analysis revealed that guppies have complex association patterns that are characterised by close strong connections between individuals of similar behavioural type. Intriguingly, the preference for particular social partners is not expressed in the length of associations but in their frequency. Finally, we show that the observed association preferences could have important consequences for transmission processes in animal social networks, thus moving the emphasis of network research from descriptive mechanistic studies to functional and predictive ones.

Turbidity, caused by suspended particles in the water column, induces light scattering and shifts in the wavelengths of light. These changes may impair the ability of fish to use physical cues and hence may modify social interactions. We experimentally investigated the social interactions of guppies, Poecilia reticulata, in clear and turbid water. Fish were significantly less active, formed smaller shoals and were found to be more often alone in turbid than in clear water. A Markov chain analysis revealed significant differences in the social dynamics when comparing clear and turbid water conditions. The probability of leaving a particular nearest neighbour and the probability of choosing some neighbour after swimming around alone differed between the treatments. Our results indicate that turbidity has a number of different effects on the social interactions of the guppy, and we discuss their potential costs and benefits and wider implications.

Social network analysis (SNA) has become a widespread tool for the study of animal social organisation. However despite this broad applicability, SNA is currently limited by both an overly strong focus on pattern analysis as well as a lack of dynamic interaction models. Here, we use a dynamic modelling approach that can capture the responses of social networks to changing environments. Using the guppy, Poecilia reticulata, we identified the general properties of the social dynamics underlying fish social networks and found that they are highly robust to differences in population density and habitat changes. Movement simulations showed that this robustness could buffer changes in transmission processes over a surprisingly large density range. These simulation results suggest that the ability of social systems to self-stabilise could have important implications for the spread of infectious diseases and information. In contrast to habitat manipulations, social manipulations (e.g. change of sex ratios) produced strong, but short-lived, changes in network dynamics. Lastly, we discuss how the evolution of the observed social dynamics might be linked to predator attack strategies. We argue that guppy social networks are an emergent property of social dynamics resulting from predator-prey co-evolution. Our study highlights the need to develop dynamic models of social networks in connection with an evolutionary framework.