Simplifications of the Laws of Exercise and Effect

There has been one notable attempt to explain the facts of learning by an even simpler theory than that represented in the laws of exercise and effect. Jennings has formulated as an adequate account of learning the law that: “When a certain physiological state has been resolved, through the continued action of an external agent, or otherwise, into a second physiological state, this resolution becomes easier, so that in course of time it takes place quickly and spontaneously” (‘Behavior of the Lower Organisms,’ p. 289). “The law may be expressed briefly as follows:—The resolution of one physiological state into another becomes easier and more rapid after it has taken place a number of times. Hence the behavior primarily characteristic for the second state comes to follow immediately upon the first state. The operations of this law are, of course, seen on a vast scale in higher organisms in the phenomena which we commonly call memory, association, habit formation and learning” (ibid., p. 291). This law may be expressed conveniently as a tendency of a series of states

A -> B -> C -> D

to become

A -> D

or

A -> B¹ -> C¹ -> D

B¹ and C¹ being states B and C passed rapidly and in a modified way so that they do not result in a reaction but are resolved directly into D.

If Professor Jennings had applied to this law the same rigorous analysis which he has so successfully employed elsewhere, he would have found that it could be potent to cause learning only if supplemented by the law of effect and then only for a fraction of learning.

For, the situations being the same, the state A cannot produce, at one time, now B and, at another time, abbreviated, rudimentary B¹ instead of B. If A with S produces B once, it must always. If D or a rudimentary B¹ is produced, there must be something other than A; A must itself have changed. Something must have been added to or subtracted from it. In Professor Jennings’ own words, “Since the external conditions have not changed, the animal itself must have changed” (ibid., p. 286). And in adaptive learning something related to the results of the S A connection must have changed it.

The series A—B—C—D does not become the series A—D or A—B¹—C¹—D by magic. If B and C are weakened and D is strengthened as sequents of A in response to S, it is because something other than repetition acts upon them. Repetition alone could not blow hot for D and cold for B.

Moreover, as a mere matter of fact, “the resolution of one physiological state into another” through intermediate states does not with enough repetition “become easier so that in course of time it takes place quickly and spontaneously.”

Paramecium does not change its response to, say, an obstacle in the water, from swimming backward, turning to one side and swimming forward by abbreviating and eventually omitting the turn and the backward movement. The schoolboy does not tend to count 1, 2, 10 or to say a, b, z, or give ablative plurals after nominative singulars.

Repetition of a series of physiological states in and of itself on the contrary makes an animal increasingly more likely to maintain the series in toto. It is hard to give the first and then the last word of an oft repeated passage like Hamlet’s soliloquy or the Lord’s Prayer, or to make readily the first and then the last movement of writing a name or address. Repetition never eliminates absolutely and eliminates relatively the less often or less emphatically connected.

Even if supplemented by the law of effect, so that some force is at hand to change the effect of S upon the animal to A D instead of the original A B C D, the law of the resolution of physiological states would be relevant to only a fraction of learning. For example, let a cat or dog be given an ordinary discrimination experiment, but so modified that whether the animal responds by the ‘right’ or the ‘wrong’ act he is removed immediately after the reward or punishment. That is, the event is either S R1 or S R2, never S R1 R2. Let the experiment be repeated at intervals so long that the physiological state, St. R1, or St. R2, leading to the response R1 or R2 in the last trial, has ceased before the next. The animal will come to respond to S by R2 only, though R2 has never been reached by the ‘resolution’ of S R1 R2.

Cats in jumping for birds or mice, men in playing billiards, tennis or golf, and many other animals in many other kinds of behavior, often learn as the dog must in this experiment. The situation on different occasions is followed by different responses, but by only one per occasion. Professor Jennings was misled by treating as general the special case where the situation itself includes a condition of discomfort terminable only by a ‘successful’ response or by the animal’s exhaustion or death.

Assuming as typical this same limited case of response to an annoying situation, so that success consists simply in replacing the situation by another, Stevenson Smith reduces the learning-process to the law of exercise alone. He argues that,—

“For instance, let an organism at birth be capable of giving N reactions (a, b, c, ... N) to a definite stimulus S and let only one of these reactions be appropriate. If only one reaction can be given at a time and if the one given is determined by the state of the organism at the time S is received, there is one chance in N that it is the appropriate reaction. When the appropriate reaction is finally given, the other reactions are not called into play, S may cease to act, but until the appropriate reaction is given let the organism be such that it runs through the gamut of the others until the appropriate reaction is brought about. As there are N possible reactions, the chances are that the appropriate reaction will be given before all N are performed. At the next appearance of the stimulus, which we may call S₂, those reactions which were in the last case performed, are, through habit, more likely to be again brought about than those which were not performed. Let u stand for the unperformed reactions. Then we have N - u probable reactions to S₂. Habit rendering the previously most performed reactions the most probable throughout we should expect to find the appropriate reaction in response to

Thus the appropriate reaction would be fixed through the laws of chance and habit. This law of habit is that when any action is performed a number of times under certain conditions, it becomes under those conditions more and more easily performed” (Journal of Comparative Neurology and Psychology, 1908, Vol. XVIII, pp. 503-504).

This hypothesis is, like Professor Jennings’, adequate to account for only the one special case, and is adequate to account for that only upon a further limitation of the number of times that the animal may repeat any one of his varied responses to the situation before he has gone through them all once, or reached the one that puts an end to the situation.

The second limitation may be illustrated in the simple hypothetical case of three responses, 1, 2 and 3, of which No. 2 is successful. Suppose the animal always to go through his repertory with no repetitions until he reaches 2 and so closes the series.

Only the following can happen:—

and, in the long run, 2 will happen twice as often as 1 or 3 happens.

Suppose the animal to repeat each response of his repertory six times before changing to another, the remaining conditions being as above. Then only the following can happen:—

and in the long run 2 will happen one third as often as 1 or 3 and, though always successful, must, by Smith’s theory, appear later and later, so that if the animal meets the situation often enough, he will eventually fail utterly in it!

Animals do, as a matter of fact, commonly repeat responses many times before changing them,[44] so that if only the law of exercise operated, learning would not be adaptive. It is the effect of 2 that gives it the advantage over 1 and 3. Of two responses to the same annoying situation, one continuing and the other relieving it, an animal could never learn to adopt the latter as a result of the law of exercise alone, if the former was, originally, twice as likely to occur. 1 1 2 would occur as often as 2 and exercise would be equal for both. The convincing cases are, of course, those where learning equals the strengthening to supremacy of an originally very weak connection and the weakening of originally strong bonds. An animal’s original nature may lead it to behave as shown below:—

and yet the animal’s eventual behavior may be to react to the situation always by 2. The law of effect is primary, irreducible to the law of exercise.