Predictability and chaos in bird vigilant behaviour

While foraging, many animals alternate between feeding and scanning. Spectral analyses of sequences of successive scan and inter-scan durations suggest the existence of a significant non-random, periodic component in the scanning dynamics of some birds. Evidence for cyclic vigilance remains controversial, however. Here data obtained from a purple sandpiper, Calidris maritima, two Barbary doves, Streptopelia risoria, and two choughs, Pyrrhocorax pyrrhocorax were re-analysed by making use of statistical methods from non-linear dynamics. Predictability portraits obtained by subjecting the data to non-linear forecasting support the view that the vigilant behaviour of the two choughs is periodic with superimposed noise. By contrast, the hypothesis of periodic scanning is rejected for the sandpiper and doves, as well as that of pure randomness. Instead, the vigilant behaviour of the sandpiper and doves bears the signature of deterministic chaos: high, short-term, decaying predictability of scan and inter-scan durations. The sequential structure of the data sets makes them well suited for reliable computation of the rate at which predictability declines; results support the conclusion of chaotic patterns. Finally, a mathematical model is developed to investigate some possible functional benefits of periodicity and chaos in vigilant behaviour, compared with random scanning, in terms of the optimization of corporate vigilance of birds foraging in flocks. If individual vigilance is chaotic, then even loose coordination, based on predictions only one foraging bout ahead, can dramatically reduce individual predictability and enhance the level of group surveillance. (C) 1996 The Association for the Study of Animal