The Conditioned Reflex Theory of Learning

From the standpoint of its simplicity, as well as its fruitfulness in suggesting experimental work, the theory of learning as conditioning commands special attention. Stated briefly, the theory asserts that all behavior either is due to innate reflex connections or is acquired by means of conditioning these reflexes. The inherited reflexes need only the lower nervous centers (spinal cord and brain stem) in order to function, but the acquired or conditioned reflexes require an intact cerebral cortex. The nature of interaction between the cortex and the lower levels is largely unknown; but it is generally assumed that even the unconditioned reflex has cortical representation, and that contiguity in cortical fields between this afferent pattern and one associated with an indifferent or incidental stimulus results in a new connection.

In evaluating the vast literature on conditioning, it is necessary to separate the usefulness of the method for attacking experimental problems, such as facilitation and inhibition, from the implications of the theory itself. Such separation makes it clear that while the fact of conditioning exists, it is by no means an answer to the problem of learning. No more is gained by saying that a reaction is conditioned to a new stimulus than is gained by saying that it is associated with it. We can dispense, therefore, with any theoretical discussions which assume that the term explains itself and proceed to two formulations which seek to see behind the phenomenon in terms of neural action.

As we have already mentioned, Holt believes that all learning reduces to the conditioning of originally random movements to concomitant external stimuli by the principle of the reflex circle. He thinks of random movements as being due to the only truly unconditioned impulses and supposes that even the simplest reflexes (unconditioned for Pavlov) are acquired, in the sense that the motor pathways must establish through use a growth connection with the appropriate afferent centers. The fact that the dendrites of growing neuroblasts follow electrical gradients and reach out toward active neurones or nerve bundles. The fact that this process of growth may be continued after birth by means of collateral connections is taken by Holt as the basis of conditioning. At birth few if any patterns exist, since ordered synaptic connections have yet to be made. Afferent impulses, denied specific conduction, spread in all directions toward the muscles. On this background of random spread, definite neural engrams are gradually established by a narrowing down of the motor outlet. This is accomplished by the tendency of every motoneurone to spread its receptive dendrites toward afferent neurones already conducting impulses. Repetition increases the permeability of these new paths, so that conditioned impulses (that is, in specific afferent neurones) pass the synapse of least resistance to excite motor units which have formerly been activated only by unconditioned impulses. Conditioning is thus a transfer of parts of the motor apparatus to specific rather than general patterns of excitation through the growth of dendrites.

Instead of tracing the phenomenon back to neurobiotactic growth, he bases his notions upon Ukhtomski's principle of the dominant, and a drainagequantum theory of nervous action. Each pattern of afferent excitation sets up or tends to set up its own dominant in the central nervous system, and that when "two or more units [or dominants] of organismic action are activated simultaneously or in close succession, the action in the more dominant units will thereby become heightened, while that in the less dominant will thereby become lowered." It is not entirely clear why one pattern of excitation is rendered more dominant than another, the answer will ultimately be reduced to a matter of relative intensity. Without further consideration of this matter he goes on to regard it as proved that, once a dominant is set up, that reaction unit drains off into its own channel the excitation associated with the less intense pattern of afferent excitation. The usual response (movement) to electrical stimulation of a limb was absent, and the dominant swallowing or defecatory act was reinforced. As extended to conditioning this theory assumes that

There is some sort of connection, let it be subliminal, between any stimulus and any response; the activation together of two patterns produces some specific lowering of the threshold between the stimulus of one and the response of the other pattern. When a certain stimulus, let us say an electric shock, is applied to the skin, its excitation is propagated mainly to its own dominant adjustors and effectors, but also in various subliminal--without overt response--degrees to all other centers, the food center included. When now the food pattern is activated, the preceding subliminal irradiation from the shock stimulus to the food response is, by the new excitation, strengthened, and more and more so by repetition, until its path becomes dominant over the original path of the shock stimulus, the application of which will henceforth--before further modifications--be propagated predominantly to the food effector. The further observation that it is always the stimulus of the less dominant, or weaker pattern, that becomes attached to the response of the more dominant, or stronger pattern, may be accounted for by the fact that subliminal irradiation from a stimulus of a weaker pattern to a response of a stronger pattern, s-O-R, will, by an additional excitation of a stronger pattern, become strengthened to a degree of becoming dominant over its original pattern, s-O-r, sooner than a subliminal irradiation of a stronger pattern, S-O-r, will, by additional excitation or a weaker pattern become dominant over its original pattern, S-O-R. Specifically, all dynamic behavior modifiability may be resolved into: (a) absolute improvement of the functional state of each elementary organismic unit by stimulation; (b) decrement of the response of a reaction-pattern by its repeated isolated activation; (c) increment of a response of a more dominant pattern by its simultaneous or closely successive activation with a less dominant pattern; (d) decrement of a response of a less dominant pattern by its simultaneous or closely successive activation with a more dominant pattern; (e) mutual decrement of the responses of patterns of close degrees of dominance by their simultaneous or closely successive activations; (f) substitution of a response of a less dominant pattern by a response of a more dominant pattern and the increment of the substituted response by repeated simultaneous or closely successive activations of the two responses; (g) weakening of any modification by disuse.

A more telling criticism can be delivered against the adequacy of conditioned reflex theories of learning in general. They seem to be discredited on five grounds: (1) The specificity of nerve paths demanded is not well supported by current neurological findings. Furthermore, the concepts of spatial irradiation and concentration, advanced by Pavlov to explain the formation of temporary connections in the cortex, may be seriously questioned upon the basis of his own data. (2) Current research indicates that conditioning is a highly complex phenomenon. It has been shown by operative procedure on the spinal cord that a conditioned response can occur only if pathways for both conditioned and unconditioned excitations are intact. While such facts do not destroy the value of conditioned reflex theory, they do indicate that its supposedly diagrammatic nature is an oversimplification. (3) Complex behavior cannot be successfully analyzed into reflex units. Conditioned reflexes may be simplified habits, but all habits are not made up of simple reflex units. (4) The theory is logically in error because conditioned and unconditioned response are generally dissimilar. For example, a cat's breathing response to a shock is markedly different from that made when breathing is conditioned to a tone. After the shock alone a marked inspiration is followed by retarded and irregular expiration; after the tone alone very rapid and shallow breathing (flutter) occurs. Not even a long series of tone-shock presentations sufficed to set up a response to the tone in the form of a sharp inspiration. Even if there are examples where the response is the same throughout the experiment, there is no obvious reason for using these as an explanatory principle for all kinds of learning. Conditioned reactions appear as added and altered rather than as substituted responses, and this points to the physiological complexity of the conditioning process. (5) The theory depends almost exclusively upon the law of contiguity, and this law is insufficient to account for learning. Gestalt psychologists maintain that mere togetherness without organization will not produce learning, and that there is no conditioning or association until organization is achieved. Actual descriptions of conditioning reveal that experimenters recognize as necessary other factors than contiguity, such as a motivational excitant and a strong extraneous stimulus; but these matters are really outside conditioned reflex doctrine.

Upon the basis of the above criticisms and our present knowledge of the phenomenon, it seems reasonable to conclude that conditioning provides a sample of learning rather than a foundation for learning theory. If interpreted broadly enough, it may offer the material for a fairly adequate theory. On the other hand, to hold rigidly to the reflex itself as an explanatory principle is unjustifiable; and when we try to harmonize the conditioning theory with all we know of learning, both physiologically and psychologically, we find that our theory becomes something else than orthodox conditioned reflex doctrine.