||Comments by Robert J. Waldbillig
Department of Psychology, University of Florida, Gainesville, Florida 32611
Offense, defense, submission, and attack: Problems of logic and lexicon. In attempting to bring coherence to the unruly mass of data on the brain mechanisms of aggressive behavior, Adams has taken on a thankless task. I was pleased to read that he was arguing for both a more precise behavioral categorization of responses and a more thorough behavioral analysis. However, his subsequent acknowledgement that the response categories (offense, defense, and submission) were not mutually exclusive was perplexing. It is commonly thought that such categorizations are useful only to the extent that they are based on a logical set of behavioral distinctions. Adams instead suggests that the neural circuitry of behavior can be used as the basis for behavioral categorizations. As a specific example, Adams maintains that "offense consists of those behaviors under the control of an offense motivational system." Because I believe there are problems with this approach. and because these considerations are basic to the neurological model presented, I shall restrict my comments to this area.
Adams's suggestion that a physiological substrate can be used as the basis for behavioral categorization is not new. Such a strategy is used in those investigations that classify stimuli as stressful if they increase corticosterone. Many physiological psychologists hold that while such a categorization is formally logical, it is inappropriately focused. The basis for its logical acceptability Is clear; however, corticosterone can be measured independently of behavior. Unfortunately, a behavior-independent measurement of brain motivational systems is not possible. The interrelationship between the behavior being defined and the construction of the defining mechanism makes the logic associated with a neural categorization of behavior circular. To make this clearer, imagine the task facing a new investigator using the Adams schema to determine which of the so-called aggressive responses should be placed in the offense category. Because Adams maintains that behavioral categorization occurs as the result of manipulating the brain mechanisms, the new investigator's research is logically assured of failure. He cannot begin to localize the brain mechanism, because he has no way of knowing when he has mapped out the appropriate mechanism. Normally this would not be a problem, because manipulations of brain are related to changes in responses already classified according to an independent set of criteria.
Although Adams maintains that his behavioral categorization is "ultimately based on neural circuitry," it is possible that he actually means that neural circuitry data can be used to supplement and strengthen a categorization schema based on behavioral descriptions. In this context, however, the strategy is not helpful, because response topographies appear in more than one category. Possibly, and hopefully, Adams uses undescribed situational variables to specify the appropriate categorization of a response. In the apparent absence of such objective guidelines, responses appear to be shifted between categories merely to support the argument of the moment. Adams provides examples of shifting of responses from one category to another where fleeing is first claimed to be a submissive response, then a defensive behavior. Shifting is again found where the boxing response is first seen as defensive and then as submissive. Another problem with the present categorization is that it cannot constrain the extent of post hoc analyses. At one point, for example, it is suggested that data inconsistent with the model could be made consistent by arbitrarily shifting a response from one category to another. A more widely accepted strategy would be to accept the data as discordant and consider changing the model.
A major difficulty for reviewers of this area is that there is no standardization of terms used to describe behavior. At selected points Adams apparently feels compelled to translate the work of original authors into his own terminology. For example, "attack" is translated by Adams to defense. The legitimacy of this translation is open to question, however, because the term attack is commonly "opposed to defense" (Webster's New Collegiate Dictionary). A rereading of the original work in question (Baxter 1967) makes it fairly clear that the author meant to connote offense with his choice of the term. A similar translation problem arises where "fear" is translated to submission. The experiment Adams refers to here has an interesting interpretive history and typifies the problems of the area. The original author reported that electrical stimulation of the thalamus elicited a low-profile posture with side-to-side head movements and occasional low-profile forms of locomotion. The posture was provisionally termed crouching. At this point the original author created the opportunity for misinterpretation when he noted that the response topography was similar to that seen in what he called "fear producing situations" (Roberts 1962). The use of the term "fear" allowed attention to move from the observable elicited behavior and towards its inferred motivational or emotional basis. Because inference easily becomes acceptance, the term fear became reified, and the original experiment was interpreted as indicating that electrical stimulation of the thalamus produces "fear." Adams takes this sequence further when he translates fear to submission. As a result, the original work is interpreted as indicating that electrical stimulation of the thalamus elicits, if not submissiveness, at least submissive behavior. One begins to wonder how many translations can occur before interpretations are completely divorced from data. It is clear from this example that the area needs to abandon terms referring to inferred mental or emotional states. Instead, clear and simple descriptions of observed responses should be used.
On the matter of behavioral models, I would like to point out that there is an alternative to the "and gate" formulation presented by Adams. The "and gate" model requires that the motivating and releasing stimuli be simultaneously present before a response can occur. Such a model grows out of laboratory experience with solid-state modules and helps to defocus our attempts at understanding behavior. The alternative model is simpler and more focused in its behavioral analysis. It does not distinguish between motivating or releasing stimuli but instead views all behavior as response chains. The length of the chain and the nature of the component responses may, of course, vary between behaviors. The questions for bio-behaviorists is straightforward. What variables (external and internal stimuli) control fhe various components of the chain, and what aspects of brain respond to, or process, these stimuli?
A recent study on mouse-killing behavior has exemplified this approach (Waldbillig 1979). It was found in this work that lesions of the area adjacent to the rat mesencephalic central gray, an area where electrical stimulation elicits mouse-killing in natural nonkillers (Waldbillig 1975), blocked this behavior. Interestingly, however, the lesions blocked only the killing bite. Orienting towards the mouse, tracking the mouse's position, and lunges toward the mouse were not affected by the lesion. Clearly, it would be inappropriate to interpret the lesions' effects as due to the loss of a mesencephalic "motivational mechanism." Instead, a discrete portion of the behavioral chain was altered. Such an interpretation leads naturally toward determining exactly what variables control the killing bite and how this area of the brain is involved in processing these stimuli.
When biopsychology focuses on both the stimuli that naturally control behavior and how the brain responds to these stimuli, it will have taken a major step away from the speculation inherent in intervening variables and virtually unverifiable hypothetical constructs. By avoiding post hoc speculation, and by staying close to the behavioral data, we will ultimately be able to specify the neurological interactions involved in these important behaviors.