Tests of the sequential randomness of vigilant behaviour using spectral analysis

Abstract

Many animals interrupt feeding bouts in order to scan for predators; the initiation of a vigilant scan has been interpreted as a random event because the inter-scan interval (ISI) has a frequency distribution that does not differ from that produced by random models (e.g. the negative exponential distribution). This in itself, however, is a test only of instantaneous (or ‘local’) randomness and not of sequential (or ‘complete’) randomness, since it does not consider the temporal sequence of ISIs. Spectral analysis of data sets for a purple sandpiper, Calidris maritima, and Barbary doves, Streptopelia risoria, showed that, while the frequency distribution of ISIs may indicate instantaneous randomness in scanning behaviour, there are significant periodicities such that long and short ISIs cycle with a degree of predictability. Some functional costs of instantaneous randomness and benefits of sequential predictability in scanning behaviour are suggested.