Download Prof.Mower Two factor learning Theory and more Study notes Psychology of learning in PDF only on Docsity!
Psych%gia, 1962, 5, 41-48.
COMMENT ON PROF. MOWRER'S TWO-FACTOR
LEARNING THEORY
Iwataro ISlflHARA Kwansei Gakuill UnilJersity
In 1947, Prof. o. Hobart Mowrer presented a paper in which he proposed a new,
integrative learning theory named Two-Factor Theory. He arrived at it as the result of reinterpretation of the theories of Thorndike and Pavlov. It consisted of solution learn- ing (or habit formation) and sign learning. The former was essentially equivalent to the first half of the Law of Effect, which held that reward strengthens the bond between stimulus and response. However, he abandoned the second half of the Law which held that punish- ment weakens the bond, and advocated quite a new explanation. When an organism makes an instrumental response (Ri) due to a drive (Sd) and receives some punishment (Sp), fear will be conditioned to the stimuli produced by the response. Thereafter, if the organism tries to do the same response, the fear aroused by the response-produced stimuli will inhibit it. That is to say, Mowrer explained punishment by Pavlovian conditioning. In this case, re3ponse-produced (or response-correlated,) stimulus is regarded as a sign the meaning of which is fear. (Cf. Fig. 1)
Sd----Ri : reward
Sd---Ri
s Sd----Ri : s->',
s~- ==:::::c.~(rf)
Sp-Rp
Sd----Ri : s~_~ rf s -- Fig. 1. Schematic presentation of Mowrer's old two-factor theory. (1st book, p.255)
In 1956, Mowrer revised his own theory by giving the concept of so-called habit a radically new interpretation. In this present version of the two-factor learning theory, he insists that "habit" is not a strengthened bond of S-R, but the conditioning of hope to response-correlated stimuli. By this revision his diagram became symmetrical, as shown in Fig. 2. s Sd----Ri : s-_~'" s---~:.-?:(rh) Sr--Rr
. s_____________ Sd---Rl : s ~rh s
s " Sd----Ri : s _"" s---~--:"'~~(rf) s __________^ Sp-Rp Sd---- Ri : s----_rf ~ s---- Fig. 2. Schematic presentation of Mowrer's current two-factor theory. (1st book, p.256)
From that time on, Mowrer concentrated his efforts to amplifying the theory by means of comparing it with other theorie'S and applying it to every aspect of learning, from the
41
42 ISHIHARA
simple behavior of rats to the symbolic behavior of human beings. The newly published companion volumes, "Learning Theory and Behavior" (1960) and "Learning Theory and the Symbolic Processes" (1960) are the fruits not only of his energetic efforts but of his excellent creative ability. (We will call the former the first book, and the latter the second book.) Mowrer's new conception of "habit" is, thus, not only the essential part of his theory but, we should say, a serious challenge to other theories. We are going to scrutinize this new conception in this paper. Is it fully satisfactory and quite adequate or yet unsatisfactory and inadequate somewhere? First of all, we should identify the exact place of "habit" in his whole theory. In the first book, we can find the following table which summarizes his whole theory. (p. 213) Table 1. Present Version of Two-Factor Theory ·Incremental reinforcement (punishment) Primary reinforcement Secondary reinforcement Danger signal Safety signal on off (fear) (disappointment)
And there follows the explanation as below:
Decremental reinforcement (reward) Primary reinforcement Secondary reinforcement Danger signal Safety signal off on (relief) (hope)
"As previous chapters have indicated, there is now widespread agreement concerning the nature of a voidance learning, both active and passive. As suggested in Table (1), there are three ways in which actipe avoidance, and three ways in which passive avoidance, learning can occur. If an independent, environmentally produced stimulus is followed by (a) primary-drive increment, by (b) secondary-drive increment (fear type), or by (c) secondary-drive increment (disappointment type), then that stimulus will acquire the capacity to drive and keep the subject away from it. That is to say, active avoidance learning can occur on the basis of ordinary punishment (infliction of pain), threat (warning), or disappointment. And if response-correlated stimulations followed by any of these three forms of incremental reinforcement, passive avoidance learning (i.e., response inhibition) will ensue. Here theory seems to be adequate to the facts, and the facts logically exhaust the theory. "In like manner, if an independent, environmentally produced stimulus is followed (a) by primary- drive decrement, (b) by secondary-drive decrement (relief type), or (c) by secondary-drive decrement (hope type), then that stimulus will acquire the capacity to attract altd hold the subject to or near it, thus accounting
for the known forms of active approach behavior. But what happens if allY of these three forms of decremental
reinforcement is associated with response-correlated stimulation? Here is an area of the total system which is mani- festly undeveloped and unexploited. Can our surmise be correct that therein lies the possibility for an advance in the understanding of habit as radical and as far-reaching in its implications as has been the change from Thorndike's 'stamping-out' conception of punishment to the present view thereof?" (1st book, p.214) Thus Mowrer asserrs; " 'habit' are not dependent at all upon a change in the resistance or conductivity of Sd-Ri connections but, rather, involve increased conductivity between the stimuli which response Ri produces and the phenomenon of secondary reward, or 'hope.' " (1st book, p. 216) And after re-examination of the experiments by R. Buge1ski (1938) and R. C. Miles (1956), he proposed the hypothesis that "habit strength and secondary reinforcement are one and the same thing, the main difference between them being that when secondary reinforcement is connected to internal, response-produced stimuli we call it-for no very apparent reason-'habit.'''
44 ISHIHARA
sponse-correlated stimuli but also place leaming in which emotion is conditioned to pbce- correlated stimuli are called habit. "Let the reader imagine an elevated open maze forming a perfect cross or 'X,' with each arm thereof measuring, let us say, six feet in length. Now further imagine that the four arms of the maze are made to point in the cardinal directions of the compass and that a hungry rat is placed at the end of the East arm, with food (not visible at a distance) in a little hole at the end of the North arm. After a few trials, the rat will, of course, go promptly and directly from the East arm to the North arm, and there obtain and eat food. Clearly the rat has thus acquired a 'habit,' but we cannot say whether this 'habit' represents place learning or response learning. However, the issue can be put the test in the following manner. Suppose that, instead of being released onto the maze from the end of the East arm, the animal, after training of the kind described, is released from the end of the West arm. If the 'habit' which has been acquired is a matter of response learning, the rat will, as before, turn right at the choice point, which will take it out onto the South arm of the maze. But if the 'habit' is a matter of place learning, then, at the choice point, the animal will turn left and enter, as usual, the North arm of the maze. "On the assumption that, in a situation of the kind described, the rat's initial learning involves the conditioning of hope to both place-correlated and response-correlated stimuli, a reasonable surmise would be that when tested in the manner indicated the rat would show (by hesitation and vacillation) some c01z}lict at the choice point but might very well decide in favor of the place-correlated stimuli (see also Mowrer, 1960, Chapter 7). If many rats were trained and tested as indicated, some of them might persist in turning right at the choice point (and thus moving spatially away from the goal of food); but most of them would probably act more 'sensibly' and abondon the 'correct' response in favor of the 'correct' place. This, in point of fact, has been the common empirical finding (see, for example, Galanter, 1955; Waddell, Gans, Kempner, & Williams, 1955; but see also Scharlock, 1955, and Grice, 1948, and Bugelski, 1956, pp.94ff.); and there has been some tendency to infer from such results 'that the disposition to orient toward the goal is simpler and more primitive than the disposition to make right turns (Tolman et al., 1946, p.228). "But, as is at once evident, one could easily alter the situation so that response learning would pre- dominate. Suppose, for example, that the X-maze were tublar (tunnel-like) rather than open, thus excluding all extra-maze simuli. It is now likely that place learning would be much less stable than response learning (see Aderman, 1957, 1958). As Tolman, Richie, & Kalish themselves observe, 'place-learning is silllpler than response-learning (only) in situations where there are marked extra-maze cues' (p. 229). "A strict stimulus-response psychology would seem to demand that the only form of learning which is possible is response learning; whereas a strict 'field theory' would perhaps insist that all learning is place learning. But ollr position suggests an intermediate interpretation which squares quite nicely with all the facts: Normally, in so-called habit formation, hope is conditioned both to place- and to response-produced stimuli and whether, in case of conflict, one or the other will gain the ascendancy depends upon a number of con- siderations, including the relative intensity and numerousness of each (c£ Campbell, 1954; Restle, 1957; Robinson & McGill, 1958). (1st book, pp.330-331) "It is particularly instructive to consider this issue of place versus response learning in the context of incremental rather than decremental reinforcement. Here it becomes particularly clear that it is not at all a question of either-or but of both. Who would seriously debate the question of whether a rat, as a result of going to a particular place two or three times and getting shocked there, would thereafter avoid the place or the action that took him there? Obviously, fear would become conditioned both to the place where the shock was experienced and to the stimuli produced by the anticedent response or response sequence. Having abandoned, once and for all, the bond-stamping-in and bond-stamping-out conception of habit, one sees no point to the question of whether fear gets connected, or connected more strongly, to stimuli associated with a response or with a place. Circumstances will almost certainly alter cases; and since the principle in- volved in both instances is the same, there is no theoretical problem at stake whatever. Here again, then, we see how completely revised two-factor theory dissolves the apparent differences between so-called reinforcement theory and so-called field theory." (1st book, p.332)
In this long quotation we can find a broader definition of habit in which 'habit' refers
to the situation where emotion is conditioned either to response-correlated or to extemal, independent stimuli, or to the both at the same time. Thus we have two definitions of
MOWRER'S LEARNING THEORY 45
habit, broader and narrower. Mowrer's terminology is sometimes rather loose like this, and embarrasses his readers. The present writer infers that Mowrer prefers the narrower definition to the broader one. This inference comes from the discussion with him and the contents of his books. For example we can find the following sentences in the early part of the chapter from which we quoted that long paragraphs. "Thus far, major attention has been given to the changes in behavior that occur when emotions get conditioned to response-correlated stimuli. The reason for this is that our principal objective has been to develop (as a counterpart to the 'new' theory of punishment) a new, and improved, way of thinking about the pheno- menon of 'habit.' " (1st book, p. 307) Habit, of course, consists of responses. Therefore, any habit theory shOlJd contain an explanation of response. It is natural that Mowrer searched for the cause ef habit not in the external world but in responses themselves. As shown in Figure 2, if an organism receives reward (Sr) when he does an instrumental response (Ri) due to stimulus drive (Sd), hope (rh) which is a component of the total response (Rr) will be conditioned to the stimuli (sss ... ) which are produced by Ri. After that if he does the same Ri at least implicitly, the response-produced stimuli (sss ... ) will evoke hope and the latter will facilitate that response. This is the essense of Mowrer's habit theory. Let us call attention to the fact that in this diagram (Fig. 2) no external stimulus is indicated in particular. This does suggest the character of the theory. As a matter of fact, external stimulus is suggested in Sd which can be external as well as internal. If external stimulus is conditioned to emotion, "it may motivate any of a variety of oven behavior, depending upon what the organism's prior experience has been in dealing with this drive." (2nd book, p. 68) That is to say, the emotion which gets conditioned to an external stimulus motivates a behavior as Sd, while the emotion which gets conditioned to response-produced stimuli regulates or guides the behavior as a feedback stimulus. Thus, the role of external stimulus in learning may be different from
the role of response-produced stimulus. Nevertheless, Mowrer says, "the principle involved
in both instances is the same," as cited above. Naturally the principle should be the same. But how it can be the same? In this respect, we should scrutinize the important role of the feedback system in his theory. The present version of two-factor theory could be attained, we might say, by the effort of Mowrer to introduce feedback theory into his theory. This can be inferred from the fact that the present version appeared two years after the publication of his paper, "Ego psychology, cybernetics, and learning theory" in 1954. Mowrer divided feedback into two phases or aspects, positive and negative. If hope is attached to response-correlated stimuli, it will facilitate the response, by feeding back positively. And if fear is attached to response-correlated stimuli, it will feed back negatively and inhibit the response. Thus, emotion, hope or fear, can control and guide instrumental responses, and the responses can be performed effectively. Mowrer says, "the first 'effect'
of a bit of behavior on the part of living organisms is to stimulate, to inform, to feed back
knowledge to the subject about the direction, extent, spread, and general nature of the behavior itself." (2nd book, p. 269) Moreover, there is another kind of feedback. Let
us quote again. "There then follows a second, subsequent effect in the nature of success or
MOWRER'S LEARNING THEORY 47
bilities which occurs in cOllnectioll with response selection (cf. Jeffress, 1951; Adrian et al.,
1954); and here sYl/lbolic behavior seems to play a crucial role." (2nd book, p. 286)
In these sentences, "image" refers to the image of "a particular course of behavior." In other words, it is the image of a group of response-correlated stimuli, and image of external stimuli is excluded or neglected. The real intention of his explanation is, maybe, to let the sign which necessarily follows a response precede it, and give the sign or its sub- stitute the possibility to initiate the response. Even if this resolution were correct, it might not be a welcome guest to his own theory. Because it will force the theory to start from the top in spite of his effort to begin with an objective base of behavior. Because it assumes that only imagery which is one of the highest mental functions can initiate responses, and without it any response can not be started. He argued: "In stressing the primacy of cognition, W oodworth seems to be reverting to prebehavioristic psychology, with its emphasis upon consciousness and con- scious states, per se. If one starts with an objective, behavioral base and then, as occasion
demands, introduces affectivity and then imagery, one has, it seems, a far sounder and more
systematic conceptual scheme than if one instead st,lrts 'at the top.' " (1st book, p. 303) He stresses the role of response-correlated stimuli in this problem as just described. He sets forth his views to the effect that if one is deprived of any response-correlated stimuli he will not be able to perform, even to start any behavior. (2nd book, pp.283-5) This may be true, but, we might say that without any external stimuli it is quite impossible to start and perform any behavior. We have pointed out above that response-correlated stimuli can not afford sufficient condition to guide behavior correctly and that they should be sup- plied with external, independent stimuli to compose a constellation of stimulation. This' constellation, in cooperation with motivation, may, it seems, resolve the problem of re-' spouse initiation. Every learning may be a sign learning. A response-correlated stimulus as a sign will acquire its meaning by learning. But the meaning can not be determined by response only, because every response is performed in an environment and whether the response is right or wrong depends upon its environment. In other words, the meaning of re- sponse-correlated stimuli is relative to the meaning of external, response-independent stimuli. In a shuttle box, for example, a rat will run about in order to escape the shock. When he runs into the next room and escapes the shock, the fear reduction or hope will be condi- tionen both to the response-correlated stimuli and to the external stimuli which come from the inside of the next room. The running to the wrong direction will not be re- warded, although it is the same response as the running to the right direction. Therefore the feedback signals which guide and control his rUlllling should consist of both response- correlated and external stimuli. Similarly, when an animal is about to start his response, the constellation of both kinds of stimuli will make him select a particular course of behavi or. He will be able to select a correct response by taking external stimuli into account. If he wants to initiate a response only by response-correlated stimuli, he will be embarrassed because there is not yet any response at all. And if he wants to do so by the image of response-correlated stimuli, again he will be embarrassed when he is on different starting points from the one where he has learned to start from, or when the situation is changed in
48 ISHIHARA
some degree. Mowrer's concept of response-correlated or produced stimulus is much broader than Guthrie's. It implies not only kinesthetic and tactile stimuli, but also "visual, olfactory, and possibly other associated forms of stimulation." (1st book, p.246, footnote) For instance, if one turns to the left and sees a bookshelf, the shelf is taken for a response-correlated stimulus, because it can be received as a result of his response, i.e., turning to the left. While, if he hears a sound of buzzer from somewhere, the sound is taken for an external stimulus, because it does not depend upon his response. By this broad definition, it seems to be possible for Mowrer to be released from some of the comments above mentioned, but not from all of them. Moreover, this distinction between response-correlated and external stimuli is not clear as a matter of fact, even though it is clear in its logic. The distinction is, we are afraid, inconvenient in scientific research, because it is rather subjective than objective. Moreover, we had better limit the range of response-correlated stimuli in the stimuli which are the direct results of responses. Because it is important for an organism to make sure of the distinction between the stimuli which are the direct results of his re- sponses and another stimuli which exist already and are just found by him. That is to say, he should learn whether or not one thing belongs to himself or to the outer world. If he fails in this distinction, he will not be able to do any adaptive bell?vior. In the above example, the bookshelf which a man found by turning his head should be classified as an external stimulus. What difference can be recognized between the bookshelf and the buzzer? He mav perceive both as the objects which he should respond to if need be. The problem of response initiation is, as stated above, hard to resolve in Mowrer's theory. We have tried to modify it in order to avoid this crux, leaving the whole frame- work intact. In his companion volumes, Mowrer is trying to synthesize the both ex- tremities of learning theories, i.e., response learning and place learning, but we caunot help receiving the impression that his theory is still in favor of response learning. By stressing the role of external stimuli in learning, it is hoped that we will be able to approach a real synthesis and to give 'habit' an appropriate explanation. MS. received XII 20, 61. This paper was prepared when I was staying at the University of Illinois. Dr. Mowrer kindly take time off his busy work once a week to give me and Mr. Yasuo Maki of Kyoto University the chances to discuss with him. And he encouraged me to write my comment on his learning theory in a paper. I'm glad to express my hearty thanks for his kind instruction. Iwataro Ishihara C5jjlt'fj-:.t&[.) 1915- ) graduated from Tohoku University in 1944; 1948-1950 Instruc- tor, 1951-1956 Assistant Professor, 1957- Professor of Psychology, Kwansei Gakuin University; 1960 Doctor of Letters; 1960-1961 Visiting Scholar, University of Illinois; Specialized field, psychology of verballearning and psycholinguistics.
