SOURCE: "Pavlov's Theory of Conditioning," in the Psychological Review, Vol. 41, No. 2, March, 1934, pp. 199-206.

[In the following essay, Guthrie offers a critique of Pavlov's theory of the conditioned reflex.]

Pavlov's recent article, 'The Reply of a Physiologist to Psychologists,'¹ deals with two items printed some time ago in the Psychological Review, one by Lashley on 'Basic neural mechanisms in behavior,' and one by the writer, 'Conditioning as a principle of learning.' The issues raised by Pavlov deserve some further discussion because they are fundamental. My own article would have been justified if its only effect had been to persuade Pavlov to additional writing on the conditioned reflex, since by his laboratory experience he is undoubtedly its most competent exponent. However, on a number of points I remain quite unconvinced after reading his 'Reply.'

The first difference of opinion that he mentions concerns a very general issue. He says,

The psychologist takes conditioning as a principle of learning, and accepting the principle as not subject to further analysis, not requiring ultimate investigation, he endeavors to apply it to everything and to explain all the individual features of learning as one and the same process. . . . The physiologist proceeds in quite the opposite way. At every phase of his investigation he endeavors to analyze the phenomena individually and in connection with facts, determining as much as is possible of the conditions for their existence, not trusting to mere deduction or to a single hypothesis.

This characterization is substantially correct. Pavlov has been reporting many detailed experiments with resulting generalizations as numerous as the varieties of experimental procedure. It was the writer's belief that the time had arrived when an hypothesis could be set up in order to direct experimental work. The hypothesis suggested was an old and familiar one, that the phenomena of learning, when described in terms of altered movement or secretion, may be described in terms of one principle, which was called the principle of conditioning. Its statement was this: Stimuli acting at the time of a response tend on their recurrence to evoke that response. In other words, it was suggested that the outstanding characteristics of learning, which have been expressed in terms of frequency, intensity, irradiation, temporary extinction, conditioned inhibition, forgetting, forward and backward conditioning, and so on, are all derivable from a more general law, the law of simultaneous conditioning or association by contiguity in time. To this end an analysis of these various phenomena was undertaken, an analysis that was, to the writer at least, very plausible.

Pavlov's second objection concerns this analysis in so far as it applies to backward conditioning. According to him the question is: What elementary properties of the brain-mass form the basis of conditioning? Backward conditioning, practice with the conditioned stimulus following the unconditioned, has a double effect, " . . . at first, temporarily, it assists in the formation of the conditioned reflex, and then destroys it,"—becoming eventually an inhibiting stimulus.

Pavlov's explanation of this is that

.. . the cell excited by the conditioned stimulus is inhibited or comes to an inhibited state with repeated concentration on the part of the unconditioned stimulus—and the conditioned stimulus in this way meets in its cell a permanent state of inhibition.

To this there is an objection. An understanding of the phenomenon of backward conditioning can be had only by finding the conditions under which it occurs. No properties of the brain-mass have been observed; no technique for observing states of inhibition in cells has yet been suggested. An explanation in these terms is and will remain unverifiable and entirely useless for prediction. In place of this unverifiable and useless hypothesis the writer had suggested that experimental search might disclose overlapping stimuli whose presence or absence would mark the presence or absence of backward conditioning. This would be to explain backward conditioning in terms of simultaneous conditioning. No act is instantaneous, and in backward conditioning the belated cue may accompany the later part of a sustained mascular contraction. This may be the explanation for the lessening effect of backward conditioning as the interval between cue and original stimulus is increased. In the writer's experiments with backward association² with human subjects the cue could be practised before or after the original stimulus with like associative strength, and the writer is convinced that backward conditioning occurs only when, and in some measure to the degree that, there are overlapping stimuli.

Concerning remote forward conditioning, or "delayed and 'trace' reflexes," the writer had evidently not made himself clear. Pavlov says,

.. . if we grant with the author that not the bell but the centripetal flow of impulses from the motor act of listening is the true stimulus for the conditioned effect, why does that effect, in the case of delayed reflexes, nevertheless come out, not at once, but after an interval—and (furthermore) in accordance with the length of the interval between the beginning of the stimulus and the beginning of the unconditioned reflex?

For this Pavlov's explanation is two-fold.³

Many cyclic phenomena take place inside the animal's body. . . . The alimentary canal is periodically filled or emptied; and, in fact, changes in practically all the component tissues and organism are capable of influencing the cerebral hemispheres. This continuous cycle of direct and indirect influences upon the nervous activity constitutes the physiological basis for the estimation of the duration of time.

This is his first suggestion. The second is as follows:

Although prolonged for a significant length of time, the conditioned stimulus remains one and the same; but for the central nervous system (and it is especially necessary to think of the cerebral hemispheres) it is distinctly different in different periods of its course. This comes out particularly clearly with olfactory stimuli, which we sense at first very keenly, and then quickly as weaker and weaker, even if they remain objectively constant. Obviously the state of the stimulated cortical cell under the influence of an external stimulus undergoes successive changes and in the case of delayed reflexes only the state of the cell near the time of the addition of the unconditioned reflex acts as a signal for the conditioned stimulus.

The writer had made his suggestion expressly to account for the fact that the delayed reflex was elicited not at once but after an interval. The 'clock' by which this timing occurred was taken to be regular changes in the movement-produced stimuli following the cue. Pavlov's first suggestion would be in entire agreement with the writer's views if he emphasized not cyclic visceral changes but skeletal muscle changes. The visceral cycles to which Pavlov refers are more or less independent of the training situation. The skeletal muscles are in much closer touch with the external world. In practice there is no guarantee that the salivary flow would be regularly accompanied by the same phase of any visceral cycle, whereas this would be possible in skeletal movements initiated by the bell. Listening, for instance, is not a sudden explosive act, but a series of acts, one movement leading to another, that to a third, and so on. Just as an orchestra player keeps time through an interval by counting or by 'beating time,' so it is possible that the timing of the delayed and trace reflexes may be accomplished by characteristic movement series. It is true that such movement series might not be regularly the same, but it is also true that this variation may account for the variations observable in the response. Pavlov cites another fact which the writer believes is aptly explained in terms of these movement-produced conditioners.

If the conditioned reflex be formed first with the short interval of a few seconds between the beginning of the conditioned and of the unconditioned stimuli, and then suddenly that interval be increased to a few minutes—then the conditioned effect, hitherto quick to come forth, will promptly go on to gradual but utter disappearance. And then, on continuing the experiment, there appears for a considerable lapse of time a period of absence of all conditioned effects. Only then does the conditioned reflex appear anew, at first just the moment preceding the addition of the unconditioned stimulus. Thence it grows gradually and recedes somewhat from the time of appearance of the unconditioned stimulus.

The writer's suggestion is that it has been necessary to condition the reflex to a new point in the movementproduced series, and that it has been 'extinguished' as a response to the original point.

Of Pavlov's second suggestion, that a constant stimulus has diminishing effects on the central nervous system, Adrian has demonstrated instances of such sense-organ adaptation to mean the gradual cessation of afferent impulses. Changes in the cortical cells are not the occasion of sense-organ adaptation. Now the effect of cutting off afferent impulses from a sense organ is to diminish the contraction of muscles and so to change actively the movement-produced pattern of stimulation. Here is positive stimulation which may serve as a delayed conditioner and we are not forced to invoke cortical cell changes invented solely for the purpose of explaining delayed reaction. Furthermore, movement-produced stimuli are essentially observable, though with difficulty, and cortical cell changes of the sort invoked by Pavlov are not. To assert that delayed reflexes take place under certain unobservable conditions is to be safe but futile.

So much for remote conditioning. To the writer's analysis of temporary extinction in terms of simultaneous conditioning of inhibiting responses there are offered similar objections. From Pavlov's article.

First of all the author takes a stand against us by saying that it is not the brevity of the interval between repetitions of the non-reinforced conditioned reflexes that contributes to extinction of these reflexes, but the number of repetitions. But this is absolutely untrue.

The writer's inference was made from the following table in Conditioned Reflexes, page 53.

With an interval of 2 minutes, extinction was obtained in 15 minutes.

With an interval of 4 minutes, extinction was obtained in 20 minutes.

With an interval of 8 minutes, extinction was obtained in 54 minutes.

With an interval of 16 minutes, extinction was incomplete in 2 hours.

With an interval of 2 minutes, extinction occurred in 18 minutes.

The rate of extinction depends, according to Pavlov, on the nature of the dog, the extent to which the reflex is established, the intensity of the unconditioned reflex underlying the conditioned one which is undergoing experimental extinction, the length of pause between successive repetitions of the stimulus without reinforcement, and the number of times the given reflex has been subjected to extinction in the same animal. The writer would agree to all of these but the one in italics, for which he would substitute conditioned inhibition. From the table there seems to be a regular increase in the time required to extinguish with the increase of interval. But it will be noticed that application every two minutes for fifteen minutes, application every four minutes for twenty minutes, application every eight minutes for fifty-four minutes, and application every sixteen minutes for two hours (with incomplete extinction) involve respectively 7.5 applications, 5 applications, 6.75 applications, and 7.5 applications, which are certainly approximately equal. The reason why Pavlov calls the writer's suggestion 'absolutely untrue' is the new fact offered by Pavlov that one application prolonged from three to six minutes gives an uninterrupted extinction. To this the writer would suggest sense organ adaptation, or exhaustion of the receptor, or the emotional inhibition resulting from the withholding of food described by Winsor,⁴ as the explanation, and plead that his version of Pavlov's table, while not absolutely true was approximately true and not 'absolutely untrue.'

The writer's contention is that temporary extinction differs only from conditioned inhibition in that it is inhibition conditioned on a temporary set, posture, or proprioceptive pattern. When this set is broken up by extraneous stimuli it no longer is present to act as an inhibitor. This is an hypothesis open to verification. Pavlov's objection to it is that

. . . the most important part of these impulses (proprioceptive) proceed to the lower divisions of the brain, and . . . under usual circumstances absolutely do not make themselves known to the cerebral hemispheres but serve only for the selfregulation and greater precision of movements, such as, for example, the continuously occurring cardiac and respiratory movements.

To the writer this seems not only a rather broad statement, but a statement which would, since Pavlov believes conditioning to be the exclusive function of the hemispheres, deny the possibility of acquiring any motor skills based on the conditioning of one movement on the stimulus of another, and deny that serial motor habits conform to the law of the conditioned response.

Pavlov's own explanation of temporary extinction is that it is

. . . the manifestation, according to law, of the most important properties of the cortical cells, as the most reactive of all cells of the organism, when they remain at work for a greater or less period of time—even if generally a short one—without a satisfying accompaniment for the fundamental innate reflexes; for, the chief physiological role of excitation of these cells is to serve as signals in place of the special stimuli of the latter reflexes. As the most reactive cells, they quickly become fatigued from work and go on not to an inactive state but to inhibition, which probably not only assists in their rest but also hastens their recovery. But when they are accompanied by unconditioned stimuli, then these stimuli .. . at once, and so to speak by way of protection, inhibit them and thus contribute to their recovery.

Since no observations have been made of states of these cells supposed to have gone 'on strike' and no observations are likely to be made, this hypothesis would appear to be not only unverifiable, but unassailable and at the same time of rather little value in the prediction of behavior.

Two more differences remain. Pavlov says,

In regard to the fact of gradual, intensification of the conditioned effect during the process of its formation, it is necessary to state that in this case it is the gradual removal of extraneous stimuli which disturbs the formation of the reflex, and not the opposite, namely—the author's view which consists in attributing to these stimuli an ever-growing rôle in creating the conditions for the effect. During our first experiments often 50 to 100 or more repetitions of the procedure were required in order to develop a complete conditioned reflex, but now 10 to 20 times are sufficient, and often much fewer.

The writer had evidently failed to make his meaning here clear because there is on this point no difference of opinion. In fact he had suggested that perfect control of the experimental conditions, which would include the proprioceptive situation as well as the exteroceptive, might well result in a certain conditioned reflex with one repetition.

The final issue is this: The writer had suggested that the phenomenon of irradiation, involving the effectiveness of receptors which were near, but distinct from the practised conditioning receptors, could be explained by the fact that neighboring receptors, by virtue of their nearness, called out responses (avoidance of a touch, for instance) which were substantially identical, and that these approximately identical movements furnished the effective conditioners of the saliva flow. According to Pavlov this is an irradiation of stimulation spreading over a definite part of the cortex. Pavlov's objection to the writer's explanation is that in many cases the phenomenon is observed 'without any trace of orientation.' To this the writer suggests that in such cases it is possible that closer observation would find such a trace.

It is also possible that it would not. As was said at the beginning of this article the writer was putting forward an hypothesis which still has for him a certain plausibility. It is only observation and experiment that can say the last word. It remains his opinion that this hypothesis is at least not inconsistent with any of Pavlov's facts.

It is evident that the differences between us do not concern Pavlov's laboratory findings. These have been of immense service to psychology. Our differences concern a strong tendency found throughout Pavlov's reports to interpret the facts in terms of supposititious events and states in the cortex. These events and states are not the characteristics of nerve conduction described by such men as Adrian and directly demonstrable in the laboratory. They are, on the contrary, quite unlike any of the properties of nervous tissue that have been so directly demonstrated, and they are of such a nature that they can not be observed by any available technique. A proper and scientific theory of learning must describe the phenomena of learning, and also describe the conditions under which these phenomena appear. Both the phenomena and the conditions determining them must be clearly defined and observable. Pavlov's work has satisfied these requirements in so far as laboratory practice is concerned, but his excursions into interpretation and theory are not so fortunate.

If further experiment should establish the law of association by contiguity in time, it will not be the first occasion in the history of science that a variegated lot of experimental results have been found on analysis to be cared for by one descriptive generalization.

NOTES

¹ I. P. Pavlov, PSYCHOL. REV., 1932, 39, 91-126. All quotations will be from this article where not otherwise indicated.

² Association as a function of time interval, PSYCHOL. REV., 1933, 40, pp. 355-367.

³ Conditioned reflexes, Oxford Univ. Press, 1927, p. 42.

⁴ A. L. Windsor, Observations on the nature and mechanism of secretory inhibition, PSYCHOL. REV., 1930, 37, 399-411.