The STARFLAG handbook on collective animal behaviour: Part I, empirical methods

The most startling examples of collective animal behaviour are provided by very large and cohesive groups moving in three dimensions. Paradigmatic examples are bird flocks, fish schools and insect swarms. However, because of the sheer technical difficulty of obtaining 3D data, empirical studies conducted to date have only considered loose groups of a few tens of animals. Moreover, these studies were very seldom conducted in the field. Recently the STARFLAG project achieved the 3D reconstruction of individual starlings' positions in cohesive flocks consisting of thousands of birds under field conditions, thus opening the way to a new generation of quantitative studies of collective animal behaviour. Here, we review the main technical problems in 3D data collection of large animal groups and we outline some of the methodological solutions adopted by the STARFLAG project. In particular, we explain how to solve the stereoscopic correspondence-or matching-problem, which was the major bottleneck of all 3D studies in the past. 3 The first speculations about collective animal behaviour date back to the observations of Pliny about flocks of starlings (translated by Rackham 1933). His remarks were necessarily very qualitative, although very reasonable too. Most of the hypotheses formulated almost two thousands years later were based on equally qualitative observations (Selous 1931; Emlen 1952). At times, the study of collective animal behaviour has been complemented by analogies with instances of collective behaviour in other fields of science, prominently physics (Radokov 1973). However, in the absence of any quantitative empirical insight, attempts to address the fundamental issues of collective behaviour rapidly became a matter of pure speculation. It was only in the mid 1960s that the first empirical studies of collective animal behaviour in three dimensions led to some quantitative results. Cullen and co-workers studied small groups of up to 10 fish in a tank using the shadow method for three-dimensional (3D) reconstruction (Cullen et al. 1965). The same technique was later employed to study groups of up to 30 fish (Pitcher & Partridge 1979; Partridge et al. 1980; Partridge 1980). For the first time it was possible to measure interesting quantities, like the average nearest neighbour distance among individuals, and their angular distribution. Several other studies of fish groups have been performed since (see, for example, Van Long et al. 1985), although always with very small groups. Empirical 3D data for birds have been harder to come by, because working in the field imposes serious …