A Simple Model for Competition Between Behaviour Patterns

Abstract

This paper explores the properties of a simple model for the interactions between a number of behaviour patterns competing for a single output pathway. In the model, the level of causal factors for each of five competing acts oscillates between start and stop thresholds, rising during non-performance and falling during performance. A sixth, residual, category is shown when none of the other behaviour patterns has been initiated by its level reaching the start threshold. The characteristics of each of the competing acts were chosen, as far as possible, to be similar to those of behaviour patterns previously studied in zebra finches. Were it not for competition, each of the competing acts would have shown cycles in its occurrence with bouts of fixed length taking place at fixed intervals. Competition led several of them to show no cycle in occurrence and to have bout lengths which followed a random, negative exponential, distribution. Only acts with low intrinsic bout lengths exhibited their natural patterns under conditions of competition, the extent to which they did so depending on the degree of competition. As competition became stronger, bout and gap lengths became shorter and, with intense competition, the output was non-stationary. The influence of competition was, however, more than a disruptive one: under some circumstances behaviour patterns showed clear cycles of a different length from those with which they had been endowed, and the matrix of first-order transitions between acts also diverged from randomness. The output of the simulations is discussed in relation to the behaviour of zebra finches and several points of similarity are noted. For example, feeding and ruffling in the finches, like the two simulated acts based on their characteristics, are the behaviours most likely to show a regular pattern of occurrence. Failure to find this in the other patterns of zebra finch behaviour may be because competition hides their intrinsic pattern rather than because they are not scheduled to occur regularly. While the model used is a very simple one, and there are many ways in which it is unrealistic, it demonstrates the marked influence which competition can have on the output of individual systems, so making it hard to discover any pattern intrinsic to those systems. Simple as it is, the model leads to a complex output and a number of testable predictions : it highlights the point that complexity is bound to arise when several acts compete for expression regardless of how simple are the mechanisms underlying each individual behaviour pattern.