When should a territory resident attack?

Models of territorial defence tend to omit two characteristics of many territorial systems: repeated intrusions by the same individual and the learning processes of residents and intruders. Here we present state-dependent, dynamic models of feeding territories, designed to investigate temporal patterns of resident aggression towards intruders that are capable of spatial learning. We compare two types of models: (1) a nomadic intruder model, in which intruders never visit the same territory twice, and (2) a single, repeat intruder model, in which an intruder may repeatedly intrude into a given territory but is less likely to do so after being attacked. These two models produce qualitatively and quantitatively different patterns of aggression by residents. For instance, residents with intruders that may repeatedly intrude have high initial attack rates, regardless of initial probability of intrusion, but their attack rates decline over time. In contrast, residents in the nomadic intruder model do not attack intruders if intrusion rates are moderately high, and their attack rates are constant and high for most of the period of territory tenure. In addition, residents of both nomadic and repeat intruder scenarios stopped attacking intruders for a short period before voluntarily abandoning their feeding territories. The results of our models suggest that repeated intrusions and learning processes have a dramatic effect on territorial defence.