Predicting flock vigilance from simple passerine interactions: modelling with cellular automata

Vigilance in flocks can be described and modelled as a plausible set of local interactions between neighbouring birds. Each bird in the modelled flock chooses to feed or to scan based solely on whether or not its neighbours are feeding or scanning. This simple model has the ability both to reproduce observations that have not been previously explained and to predict flock behaviours that might be confirmed with future field studies. Examples include simulations showing decreased vigilance with increased flock size (as observed in the field), greater time spent scanning when obstacles such as trees are present (as observed) and a coordinated feed/scan pattern (that is predicted to become increasingly coordinated when the birds look up from feeding more frequently). The numerical model also predicts that hock geometry influences vigilance. If two flocks are the same size, individuals in the one with the larger perimeter will spend more time scanning. This prediction could be tested with field studies and already has been observed empirically for two limiting cases: birds arranged in a line (high perimeters, high scan times) and birds in a circle (lower perimeters, lower scan times). As demonstrated by its multiple successes, cellular models of this type are a powerful new approach to understanding bird flock behaviours. (C) 1999 The Association for the Study of Animal Behaviour.