In the years 1979 to 1983, I wrote four reviews on agonistic behavior based on a comparative study of muroid rodents (Adams, 1980) , brain mechanisms (Adams, 1979a), learning effects (Adams, 1979b) and hormone effects (Adams, 1983) and developed a model based on these data. The present paper reviews that model in the light of research carried out during the intervening years.
The model, shown in figure 1, was organized along the lines of the classical ethological distinction between motivating and releasing functions of stimuli, with the motivating function operating through a few motivational mechanisms and the releasing function operating directly on the motor patterning mechanisms of behavior. The particular stimuli and motor patterns were drawn from the analysis of the agonistic behavior of over 20 species of muroid rodents (Adams, 1980) and a comparison to the behavior of primates (Adams, 1981). Three motivational systems are considered, offense, defense and patrol/marking. Two of these, offense and defense, may be considered as forms of aggressive behavior, while all three, patrol/marking as well as offense and defense, may be considered as forms of agonistic behavior according to Scott's classical definition (Scott, 1966).
Although the model is designed for muroid rodents, comparative studies suggest that the innermost neural structure of motivational systems changes very little in the course of evolution. Thus, in the comparison of rodents and primates, only one major difference could be identified, the substitution of a display motivational system in primates for a patrol/marking motivational system in rodents. The results of the primate study on which this was based (Adams and Schoel, 1982) have recently been confirmed under more naturalistic conditions (Partan, 2002).
The model can be applied to other mammalian orders by changing the modality of sensory inputs and motor outputs, while leaving the inner structure intact. Thus, while motivating stimuli tend to be olfactory and releasing/directing stimuli tend to be tactile, including vibrissal in rodents, there is a greater reliance on visual stimuli in carnivores and primates. On the motor side, there is more use of the mouth in rodents and by the hands in primates (Adams, 1981). So-called species-specific aspects of behavior are limited, at least in the case of the muroid rodents, to intraspecific pheromone signaling, as well as some variation in submission and warning signals (Adams, 1980).
The enclosed boxes, ovals and circles in the model are assumed to be neural assemblies and it is assumed that hormones and learning have their effects on them (Adams, 1979b, 1983). Points of hormonal action are shown in the model by letters; points where learning takes place are shown by numbers. The model is deliberately simple, even oversimplified, with about 56 neural assemblies and 70 interconnections. This follows from an assumption that the neural circuitry of aggressive behavior changes very little during the course of evolution and that, once understood, it will turn out to be elegantly simple, just as the base-pair composition of the genetic code turned out to be elegantly simple. It is assumed that the basic characteristics of these neural circuits are determined by a small number of gene-enzyme factors, perhaps several hundred in number (cf Strumwasser, 1967).
Following an initial section describing the behaviors concerned, the following review of recent research is organized in terms of the four components of the hypothetical motivational systems: 1) description of behaviors; 2) motivational mechanisms, 3) sensory analyzers and synthesizers for motivating stimuli, 4) motor patterning mechanisms, and 5) sensory analyzers and synthesizers for releasing and directing stimuli. A concluding section briefly addresses the question of how the model can be tested.