Nonrandom patterns of vigilance in flocks of the greater flamingo, Phoenicopterus ruber ruber

Models of vigilance in groups usually assume instantaneous randomness in scan initiation and sequential randomness in the duration of successive interscans, thus predicting a negative exponential distribution of interscan durations and no predictability in the duration of successive interscans. However, recent models suggest that scanning should instead occur regularly rather than randomly when foragers are threatened by nonobservant predators, thus predicting interscans of constant duration and predictable interscan sequences. I examined whether vigilance patterns departed from instantaneous and sequential randomness in greater flamingos wintering in a tropical lagoon complex in Venezuela. Predation risk in this large species is almost nonexistent, and it is argued that disturbance by people, the most probable threat, should best be detected using a regular pattern of scanning. As predicted, the distribution of interscan durations showed a strong central tendency and differed significantly from the expected negative exponential distribution. Interscan intervals of similar duration occurred in succession more often than predicted by chance. Despite obvious departures from instantaneous and sequential randomness, substantial variability remained in the distribution of interscan durations, and several interscan sequences were quite unpredictable. I relate these findings to potential variation in food supply within the confine of a flock and preemptive scans to detect potential threats. (c) 2005 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All tights reserved.