||Comments by J.M. Koolhaas
Department of Zoology, University of Groningen, The Netherlands
The risks of using descriptive ethological models in brain research. The ultimate goal of many biological and psychological studies is to understand the way in which an animal is organized to survive in its natural habitat. In this article Adams focuses on the central nervous organization of agonistic behaviour - i.e, behaviour that can be observed in encounters between two male animals of the same species. On the basis ot ethological studies, three subsystems of this agonistic behaviour can be distinguished, namely offense, defense, and submission. It is to Adams's credit that he reviews, with this classification in mind, a large number ot studies in order to construct a general outline of the neural circuitries involved .
Alhough one may express some doubts on ethological grounds about the value of this classification for the purpose of analysing brain mechanisms (see also commentary by Wiepkema), offense, defense, and submission are nevertheless recognisable behavioral strategies in the rather extreme situations in which the animals are usually tested.
The problem now is to analyse how the brain is able to process these strategies. For this, Adams uses descriptive hierarchical models from ethology. In such models a critical role is played by a common causal factor or the motivational mechanism. In Adams's terminology, this motivational mechanism is a hypothetical set of homogeneous neurons whose activity is held to be responsible for the motivational state of the animal. In my opinion, Adams makes a fundamental mistake in trying to give a unitary interpretation of such a common causal factor. Of course, the ethologically observable organisation of behaviour is embodied in the structure of the brain. Its units are groups of neurons, so organized as to produce the various behaviours at the right times. However, such an organization need not necessarily be identical with the boxes posited in a hierarchical model. One box in the model - for example, the motivational mechanism - may be represented in the brain by many structures, possibly the whole limbic system, or even the whole brain.
In his attempt to find neural structures related to the boxes in the model, the author uses somewhat forced arguments. He states, for example, that the midbrain central gray meets the criteria for the motivational mechanism for defense, without explicitly mentioning these criteria. It is argued that neurons in the central gray are specifically active during shock-elicited defense. The original article (Pond, Sinnamon, and Adams 1977), however, only shows that these neurons are most active during shock-induced upright posture. If these units represent the motivational mechanism for defense, one might expect them to be active during other defensive behaviours as well - i.e. during lunge-and-bite attack, squealing, freezing, fleeing, and so forth. Also, if these units are specifically for defense, one might expect them to fail to be active during offense or submission. The original study does not give any answer to these questions, and therefore one cannot argue that these central gray neurons are specifically active during shock-induced defense. Moreover, these units are also activated by vibrissal stimulation, which is thought to be a releasing stimulus for defense. According to the model presented by Adams, such stimuli may not affect the motivational mechanism but the motor-patterning mechanism.
Some of the evidence for the role of the central gray as the motivational mechanism for defense is based upon fear or escape measured under nonsocial conditions (shock-food conflict task, one-way two-way avoidance). The relation of these measures to defensive behaviour in a social situation is far from clear.
Reading the target article, I realized more and more that it is too early to ascribe functions to the various neural structures involved in agonistic behaviour. In the first place, the flexibility of an animal in a social situation has been insufficiently explored ethologically. This means that it is difficult to express any expectancies about the internal organisation of behaviour (see also Wiepkema). Secondly, the effects of brain manipulations have been tested in a wide variety of test situations but rarely in a social setting. If our ultimate goal is to understand how an organism is adapted to its natural environment, knowledge about the relationship between the behaviour in our test situation and that under more natural conditions is a condition sine qua non.
At most we know, for some types of agonistic behaviour, which brain manipulations alter the probability of occurrence of that behaviour. However, questions like why, under what circumstances, and how specifically that probability is changed are often unanswered. In order to answer this, we have to test a wide variety of input variables known to affect the agonistic behaviour of an intact animal These are variables like previous experience, hormones, day-night rhythms, stimuli from the opponent, priorities for other behaviours, and so forth. Manipulation of each of these variables might alter the probability of occurrence, latency, sequential structure, and other mechanisms of agonistic behaviour.
The brain should accordingly contain mechanisms for processing information about these variables. My approach to the function of brain structures in agonistic behaviour is to test, with each brain manipulation, whether it affects one or more of these mechanisms. Not until a number of brain structures have been tested in this way can we speculate about functional organization.
In fact, the main part of this discussion concerns the problem of tuning the size of brain manipulation to the level of integration of brain functioning for agonistic behaviour. Adams has started this discussion, and I do hope that it will lead to a number of fruitful studies resulting in better understanding of the brain mechanisms in agonistic behaviour.