Brain Mechanisms

Motives: Metaphors in motion. Robert Frost used to challenge his students, and critics, with the query "What is a meta for?" As a poet who had mastered clean lines of description he was keenly aware of both the power afforded by language that transcends direct observation and the dangers of taking one's symbols too literally. The behavioral and brain sciences face a similar problem with their meta language. Motivation is a case in point.

Adams apparently presumes that we (readers) agree on what motives are and that they are real. I suggest that motivation is a construct of diverse definition designed to help us gain awareness of changing relations among observed fragments of input and output as they operate in the intact organism. This is its power in setting problems, and its limitation in solving them. Brains have cells, not drives, and strict localization of network transfer functions defined for the intact organism may not be possible (e.g Luria 1976). Like the Cheshire cat, we are left with but a faint smile of our original image.

The image in the first instance reflects a distinction between stimuli that motivate behavior and those that merely release it . Adams is perfectly right that this is the classical view of early ethology, but it now appears much too simple. The distinction is relative not absolute, and multidimensional rather than unitary Two illustrations will suffice.

First, the dichotomy between "functional groupings" and "discrete" motor patterning mechanisms is imperfect and a matter of perspective. The single swipe of a cat's paw may be a descriptively discrete component of the overall melee, but no doubt contains a finely tuned, and functional, grouping ot cooperative and antagonist muscles, excitatory and inhibitory circuitry, and so on. Adams's early implication, from which he retreats at the end, that such an act can be generated "without regard to its functional or temporal relationship to other motor patterns" cannot stand for a variety of considerations, ranging from Sherrington's research on reflex integration to modern research on coarticulation in human speech (e.g. Studdert-Kennedy 1976).

A second division is the time course of action, with "releasing" functions being short and "motivating" functions somewhat longer. This is fine if it leads the investigator to examine explicitly different time courses of action, but it fails as a dichotomy. For example, repeated application of "releasing" stimuli can reveal both incremental and decremental consequences upon subsequent readiness to respond that have quite protracted decay functions (e.g. Heiligenberg 1976). Adams gets himself into some related difficulties in logic when, for example, he distinguishes low latency threat from longer latency attack; is the former "released" and the latter "motivated?" Similarly, with examples such as escape in response to footshock I am not certain whether I am supposed to think "motivate" or "release." The author does point out that the same stimulus can have both consequences, but that only serves to weaken his distinction further. Occasionally Adams bypasses the problem by using other terminology, such as "activate." His brief attempt to superimpose modalities upon the division is fortunately abandoned. As a final point, if a stimulus initiates a behavior it is easy to think of a releasing function, whereas if the same stimulus facilitates the same behavior already in progress one may be more inclined to think of it as motivational, even though it might be operating through the same principles in each case. The distinction has difficulties (cf. Fentress 1977).

One is led next to the division among offense, defense, and submission as distinct motivational systems. This is an obvious releaser (or motivator - take your pick) for comment. It does not help much to define, for example, the defense motivational system as that which contains defense motor patterns such as attack and defense. Nor is the distinction helped when the same motor patterns are found to be shared in pairs of systems. Neither does the contextual definition always help, as in the case of "offense" shown by the rat defending its territory, and "defense" by the rat on the offensive (i.e. invading the territory). Maternal aggression, we are told, is a mixture of offense and defense, and data that are apparently contradictory to the model are viewed suspiciously as resulting from an understandable mislabeling by the investigator. Function, evolution, genetics, and learning join forces to clarify the picture, but the result is not always clear. Offense is reportedly seen more often in males (but see Ryon 1979), and is often communicative (undefined) along with submission, where defense is not, since the latter evolved earlier as a response to predators (presumably this goes for wolves and grizzly bears as well). and submission apparently evolved in the lab - and so on. I think some of the metaphors are mixed.

While it is difficult to construct a logically tight package from the present survey, one positive idea that emerges is that "offense" and "submission" represent ends of a continuum, whereas "defense" is something in between (and thus, e.g., shares motor patterns with each of the others) That is a nice notion, albeit contradictory to the idea that one is dealing with three discrete systems Indeed, if one watches a boxing match between human opponents it is often quite impossible to sort out individual actions as offense, defense, and so on, and even football coaches are fond of stating that the best offense is a good defense (and sometimes the reverse) - just to confuse us all Some of our animal subjects may have the same mentality!

Finally, there is the issue of hardware (or software, depending on one's bias). Ideas such as "homogeneous" neurons lend a ring of reality, but by what criteria are they defined? In some instances they seem to mean "same," as in relating motor mechanisms of defense and submission, but in other instances they refer to the separation of these very systems, as in their action through a "consociate modulator." This last construct relabels the fact that animals sometimes behave differently when they recognize one another than when they do not. Without worrying about problems of localization the "modulator" may have logical troubles, such as when a stimulated animal flees ("submits?") upon the absence of its perceived opponent. Modulate normally means change in degree rather than kind, but this raises the interesting question of how one distinguishes between two different states of the "same" system and two "different" systems. Adams does use the construct of intensity on both the input and output side, but does not separate factors such as current, pulse width, and pulse frequency on the one hand, and amplitude, speed, completedness, interruptability, and so on of behavioral acts on the other (cf Fentress 1973). The summary arguments rest upon a complex mixture of (a) stimuli and their consequences (e.g. "defense and submission inputs"), (b) anatomical loci (e.g. "midbrain central gray"), (c) hypothetical constructs (e.g. "submission motivational mechanism"), and (d) behavioral outputs (e.g. "fleeing"). These are very different logical orders that cannot be boxed together in any simple way (e.g. Figure 2).

We (investigators) obviously attempt to patch coherent pictures out of very incomplete glimpses of nature, and as Frost said about poets, a meta language can provide a powerful ally - as long as we do not forget what we are doing. Where I suspect I differ from Adams is that I view behavioral "systems" as tongue in cheek conveniences for clustering some observations together in distinction from other observations; but I do not take them as being really real beyond that. This leads to the heuristic that one continues to look for signs that favorite systems of the moment are neither unitary (e.g. watch out for "homogeneous neurons") nor totally separate (e.g. watch out for "distinct" and "localized"), and also that their defining boundaries may shift in time as a consequence of both "intra" and "inter" system dynamics (e.g. Fentress 1976a). The most we can do at this stage is to look at how things (processes) cluster under different specified circumstances. Adams has given us one framework for doing just this, and, even though I suspect his solution has difficulties, the effort is precisely what a "meta is for."