University of Southern California Dissertations and Theses

Dominance And Contiguity As Interactive Determinants Of Autonomic Conditioning

(USC Thesis Other)

Dominance And Contiguity As Interactive Determinants Of Autonomic Conditioning

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content DOMINANCE AND CONTIGUITY AS INTERACTIVE DETERMINANTS OF AUTONOMIC CONDITIONING by Russell Arthur Lockhart A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (Psychology) August 1965 UNIVERSITY O F S O U T H E R N C A L IFO R N IA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES. CALIFORNIA 9 0 0 0 7 This dissertation, written by ..... RussellArt.h\u?..L ...... under the direction of h.%$...Dissertation Com mittee, and approved by all its members, has been presented to and accepted by the Graduate School, in partial fulfillment of requirements for the degree of D O C T O R O F P H I L O S O P H Y Dean ACKNOWLEDGMENTS I would like to express my sincerest thanks to the members of my guidance committee, Drs. Everett Wyers, John Robb, Alfred Jacobs and Langdon Longstreth for their help, encouragement and confidence. In addition, I would like to acknowledge a special indebtedness to Drs. Ray Jordan and Gordon Mathescpn, two teachers having no direct influence on the present research, but so much of an influ ence on the author. And to my chairman and teacher these past five years, Professor William W. Grings, I express a.hope that someday I will be able to teach and inspire my students as he has taught and inspired me. Finally, I thank the National Institute of Health for equipment and subject funds made available to me through a grant awarded to Dr. Grings. TABLE OF CONTENTS Page ACKNOWLEDGMENT ....................................... ii LIST OF TABLES....................................... V LIST OF ILLUSTRATIONS.............................. vii Chapter I. INTRODUCTION ................................. 1 Degree of Conditioning and the Inter stimulus Interval The "Second Interval" Conditioned Response II. CONTIGUITY AND REINFORCEMENT IN CLASSICAL CONDITIONING THEORY ...................... 24 Contiguity Reinforcement III. DOMINANCE AND CONTIGUITY AS INTERACTIVE DETERMINANTS OF THE CLASSICALLY CONDI TIONED AUTONOMIC RESPONSE................. 66 The Influence of Razranean Theory Formulation of the Experimental Problem Predictions Generated by Experimental Question IV. METHODS AND PROCEDURES . 88 Subjects Apparatus Design Procedure Chapter V. RESULTS Page 104 The Half-Second Trace Conditioned Response Problems of Multiple Responding in Long Interval Conditioning Five-Second Trace Conditioned Responses Ten-Second Trace Conditioned Responses Discrimination as a Function of Trace Interval VI. DISCUSSION................................... 143 BIBLIOGRAPHY .......................................... 170 APPENDICES............................................ 176 LIST OF TABLES Table Page 1. CS Intensity, UCS Intensity, CS-UCS Interval and Intertrial Interval for Each of Nine Experimental Conditions (CS Intensity Value Relative to UCS Intensity Value) ........... 90 2. Discrimination Value by Dominance Condition for Each Response Definition................ 116 3. Discrimination Values for Each S During Each Response Period by Dominance Condition for 5 Second Trace Group . 117 4. Discrimination Value by Dominance Condition for Each Response Definition................ 124 5. Discrimination Values for Each S During Each Response Period by Dominance Condition for 10 Second Trace Group ......................... 130 6. Degree of GSR Discrimination (Across Dominance Conditions for Various Response Periods and Combinations of Response Periods Relevant for "UCS-Correlated" Responding .............. 145 7. Comparison of TSR and Period I Discrimination Values at First Point of Acquisition for 5 and 10 Second Trace Conditions ........... 159 8. Disparity, Reference, and Perceptual Disparity Response (PDR) Values for Each Subject . . . 181 9. Mean Magnitude of Discrimination for Three Extinction Trials and Last Acquisition Trial................................ 182 v Chapter Page 10. Comparison of "Pre-Disinhibition" and "Post- Disinhibition" Discrimination Performance for Each Trace Interval and Each Dominance Condition..................................... 183 LIST OF ILLUSTRATIONS Figure Page 1. Relative magnitude of GSR conditioning as a function of the interstimulus interval (after White and Schlosberg, 1952). Dashed line represents the response level of a sensitization control group............. 6 2. Mean magnitude of GSR conditioning as a function of the interstimulus interval (after Uchima, 196 2). Magnitude of GSR expressed in square-root-of-conductance units........................................... 8 3. Mean magnitude of GSR discrimination as a function of number of differential reinforce ments under two interstimulus interval con ditions (after Grings, Lockhart, and Dameron, 1962). 17 4. Illustration of test trial GSR pattern indicat ing the so-called "anticipatory" response and the response near the temporal point of usual UCS occurrence which is obtained in the absence of the UCS. ........................ 20 5. Comparison of two types of contiguity in a simple classical conditioning paradigm. . . 28 6. Comparison of two types of contiguity in relation to CS and-UCR characteristics in a simple classical conditioning paradigm. . . 29 7. (a) Typical trace paradigm; (b) hypothetical CS trace; (c) stimulus trace as viewed by Pavlov (1927) . 38 vii viii Figure_ Page 8. Diagrammatic representation of Hull's (1943) theory of classical conditioning............. 41 9. Diagrammatic representation of the so-called "identical stimulus" mechanism of anti cipatory responding in the classical conditioning situation...................... . 43 10. Diagrammatic representation of Mowrer's (1960) theory of classical conditioning............. 47 11. Diagrammatic representation of Pavlov's (1927) theory of classical conditioning............. 53 12. Diagrammatic representation of hypothetical dominance ratios resulting from the manipu lation of stimulus intensity ratios and degree of trace interval...................... 79 13. Measurement of response magnitude............... 101 14. Mean magnitude of discrimination as a function of dominance ratio with .5 second trace interval........................................ 106 15. Sample GSR recording during an acquisition test trial on CS+ when the UCS is not presented...................................... 109 16. Degree of GSR discrimination as a function of dominance by response period (roman numerals) for upper, middle, and lower third subgroups of Ss divided on the basis of Period I dis crimination for the 5 second trace interval. 121 17. Acquisition of GSR discrimination for each dominance condition in the 10 second trace interval group in terms of Max R......... 126 / I X t Figure , Page \ 18. Period I discrimination as a function of dominance for upper, middle and lower third subgroups of Ss divided on basis of degree of Period I discrimination for the 10 second trace interval................................. 132 19. Period IV discrimination as a function of dominance for upper, middle and lower third subgroups of Ss divided on basis of degree of Period IV discrimination for the 10 second trace interval.......................... 133 20. Discrimination as a function of dominance for maximal discrimination value (positive and negative) regardless of response interval in which the maximal discrimination value occurred " . 135 21. Degree of discrimination as a function of dominance by response period (roman numerals) for upper, middle, and lower third subgroups of Ss divided on basis of Period I dis crimination for the 10 second trace interval. 136 22. Max R discrimination for each dominance condition at each trace interval.............. 138 23. Period I discrimination for each dominance condition at each trace interval.............. 140 24. Degree of maximal discrimination without regard for response period for each dominance condition at each trace interval........................................ 142 25. Above and below median discrimination curves for each dominance condition in the 5 second trace interval group ........... 163 CHAPTER I INTRODUCTION Classical conditioning is often considered a fundamental learning process. Attempts to understand the "mechanism" of such learning usually center on the explica tion of the concepts of "contiguity" and "reinforcement." In the present study, the concern is for the nature of these variables as determining factors in the modification of autonomic responses in the classical conditioning situation. Even a cursory analysis of the literature involving contiguity and reinforcement would indicate to the reader the controversial role these concepts have played in the development of learning theory. A somewhat deeper analysis would show the reader that the meaning of both concepts is at once complex and ambiguous. Perhaps this is nowhere more clearly seen than in regard to the effects of the interstimulus interval parameter (i.e., the temporal period between CS and UCS onsets) in the classical conditioning of 2 autonomic responses. Degree of Conditioning and the Interstimulus Interval For eyeblink and other skeletally mediated responses a CS-UCS interval somewhat longer than the latency of the response has been found to produce the maximal degree of conditioning. Typically, this interval approximates one-half second. For both shorter and longer intervals the magnitude of conditioning decreases rapidly. Hilgard and Marquis (1940) interpreted this fact as evidence for a "preparatory" or "anticipatory" theory of conditioning. They reasoned that if the temporal period between CS and UCS were such that the response could anti cipate or antedate the UCS, then the response would serve a protective or defensive function. Such defensive anticipa tion would reduce the painful effect of the UCS and thus, in a general way, increase the individual's adaptive behavior.^ ^Two important questions are generated by this view of conditioning. The first concerns the mechanism by which the CR is elicited in such a way as to occur prior to the UCS. The second raises the problem of classical and instru mental conditioning as distinguishable processes. While in classical conditioning it is true that the UCS is applied to the organism without reference to its behavior, there is It should be recalled that Pavlov (1927) success fully employed much longer intervals than one-half second in his pioneering work in classical conditioning. Hilgard and Marquis (1940) interpreted this fact to mean that longer latency responses such as those mediated by the autonomic nervous system (e.g., salivation, galvanic skin response, blood volume, etc.) might indeed be better con ditioned at longer CS-UCS intervals— on the assumption that conditioning is always best when the CR may anticipate the UCS. Since the latency of autonomic responses is typically longer than one second and sometimes as much as three seconds, they predicted that the optimal CS-UCS interval for autonomic response would be longer than for the shorter- latency skeletal responses. Twelve years passed before this suggestion was put to experimental test by White and Schlosberg (1952) and no guarantee that the organism's behavior will not modify the effect of the UCS. The CR viewed in this way is instrumental in the reduction of UCS effectiveness. Both questions are intimately related and are quite tied to the question of variation in the CS-UCS interval. again two years later by Moeller (1954). Both studies concluded that the galvanic skin response (GSR) was maxi mally conditioned when the CS-UCS was one-half second— the same as for skeletal responses. These results were (and continue to be) accepted as a solution to the interstimulus interval "problem" as it relates to autonomic responses. For theorists viewing learning as a stimulus- stimulus process there was encouragement since the results readily supported the idea that the nature and characteris tics of the response conditioned is irrelevant, for what is learned is a perceptual relationship between stimulus con tingencies. ^The reason for this delayed response is an inter esting problem in itself. It is evident that in American psychology the study of "overt" behavior has always held more fascination for experimental psychologists than has the study of autonomic responses and their modification. Classical conditioning was subsumed under more general principles of reinforcement and most research concerned the effects of reinforcement in instrumental behavior (bar pressing, maze learning, etc.). Only in the past few years has there been reemergent interest in classical condition ing of autonomic responses and classical conditioning in general, perhaps in response to the degree to which classi cal conditioning has been called upon to extricate reinforcement theorists from theoretical difficulties attendant upon such problems as the "avoidance paradox" and too, because of the increased necessity for explanation by "secondary reinforcement," a classical conditioning phenomenon. 5 For those viewing learning as a stimulus-response process there was a definite problem— still not adequately resolved. Since the latency of the GSR is at least twice as long as the optimal CS-UCS interval, it is obvious that the reinforcing event antedates the response thus creating a situation which is untenable for a reinforcement theorist, i.e., for learning to occur, reinforcement must be a conse quent and therefore temporally subsequent to the response to be learned. One solution was to view autonomic responses as not conditionable at all. Such responses were seen as mere by-products of covert and unmeasured skeletal responses. Since skeletal responses are modified, according to this position, on the basis of instrumental reinforcement pro cedures, one writer allowed himself to pontificate on "conditioning as an artifact" (Smith, 1954). White and Schlosberg (1952) employed a delayed con ditioning paradigm with CS-UCS intervals of 0, .25, .50, 1,00, 2.00 and 4.00 seconds. Fig. 1 illustrates the results for the conditioned GSR. It is obvious that the maximal degree of conditioning was produced with an interval of .5 second. 6 100 80 C D rt 70 0 ft 1 60 V A o •H 50 40 20 10 2.0 3.0 0.25.50 1.0 Interstimulus Interval (Seconds) Fig. 1.- -Relative magnitude of GSR conditioning as a function of the interstimulus interval (after White and Schlosberg, 1952). Dashed line represents the response level of a sensitization control group. Moeller (1954) employed CS-UCS intervals of .25, .45, 1.00 and 2.5 seconds in a trace conditioning paradigm. Maximal conditioning was exhibited at the .45 interval. Uchima (1962) was intrigued by the fact that the poorest group in the White and Schlosberg (1954) study was not the 4 second group, but the 2 second group. Using more subjects, a substantially better design, more adequate control procedures, and a more sensitive measure of condi tioning than did White and Schlosberg (19 54), Uchima condi tioned the GSR under interstimulus intervals of 2, 3 and 4 seconds. Uchima’s results are pictured in Fig. 2. Uchima (1962) found that all three groups exhibited significant conditioning. In addition, the degree of conditioned responding in the 4 second CS-UCS interval group was signi ficantly greater than conditioned responding in the 2 and 3 second CS-UCS interval groups. Silverman (1960) reported a study in which .5 and 6 second CS-UCS intervals were employed. He found no differ ence in the amount of GSR conditioning during the acquisi tion period between these two interstimulus interval conditions. These data lead to the conclusion that more than one maximum may exist in the classical conditioning of 8 .40 .35 « eo CiJ < H O 9 > ^ .30 S> S . .25 .20 J; n 0 0 Experimental Groups Control Groups 1.0 2.0 3.0 4.0 Interstimulus Interval (Seconds) Fig. 2.— Mean magnitude of GSR conditioning as a function of the interstimulus interval (after Uchima, 1962) Magnitude of GSR expressed in square-root-of-conductance units. 9 autonomic responses. There appears to be an increasing function from 0 to .5 second, followed by a decreasing function to about 2 seconds, followed in turn by an increas ing function to a maximum point not yet empirically deter mined. Direct comparison of CS-UCS interval effects at the .5 second and 6 second points shows conditioning to be approximately equal (Silverman, 1960). The studies showing conditioning of the GSR at intervals longer than the latency of the response would seem to offer support for Hilgard and Marquis' (1940) con tention discussed earlier. However, at least two questions are raised by this view. First, if conditioning should be best when the CS- UCS interval is longer than the response, why should the GSR show any conditioning--and particularly a large degree of conditioning— at an interval almost 5 times shorter than the latency of the response? Put more simply, why should there be more than one maximum? This raises the question of whether or not the two maxima represent the same behavioral process. The second question is simply why should the second maxima be so much greater than the latency of the response? If we use Silverman's (1960) data as a guide, it is obvious that the second maximum is at least at the 6 second point, a good 4 seconds beyond the average latency of the response. This would seem to bring into question whether the response (GSR) as such or its various characteristics are importantly involved in the conditioning process. Perhaps the implications are more clearly seen if one reflects on the phenomena of long trace and delay condi tioning typified by Pavlov's (1927) early work and in this country by the work of Steckle (1933), Switzer (1934) and Rodnick (1937). The problem at issue seems to be whether the same fundamental process is operative when the CS-UCS interval is brief as opposed to when the interval is longer. At this point it should be noted that all of the preceding data were obtained through the use of the so- called "simple" conditioning paradigm. In such a paradigm, one CS is used and this single CS is reinforced. Condi tioning is usually defined in one of two ways. The first employs a control group in which the same CS and UCS are given but not in a temporally paired relationship. The second gives preconditioning presentations of CS and assesses the degree of responding. After a certain number of reinforcements responding to the CS is again measured. 11 Conditioning is defined in terms of the difference (if positive) between these two measurements. A more sensitive inferential procedure is provided by the use of a control stimulus methodology. This involves presenting two CSs to the same subject only one of which is ever paired with the UCS. Any processes other than condi tioning occurring during the experiment are assumed to be reflected in the response magnitude to both stimuli. How ever, as a result of the pairing procedure, it is assumed that there will be an added increment of responding to the CS which is paired with the UCS. Therefore, conditioning is inferred from the difference in magnitude of responding (or other appropriate dependent variable deriving from the difference score procedure) to these differentially treated stimuli. The use of such a procedure, although providing a more defensible base for inferring conditioning, must cope with the possibility that the use of two stimuli may alter the effects of the conditioning process. The use of two stimuli creates a discrimination situation and the possibil ity exists that conditioning variables operate quite differently in this more complicated situation than in the 12 more simple one.3 Hartman and Grant (1962) conditioned the eyeblink in a discrimination paradigm. Instead of finding the maxi mum degree of discrimination at .5 second, they found the maximum to occur at a 1 second CS-UCS interval (the longest interstimulus interval employed). Kimmel and Pennypacker (1963) employed four interstimulus intervals in discrimina tion conditioning of the GSR: .25, .50, 1.0 and 2.0 seconds. They found the greatest degree of discrimination with the 2.0 second CS-UCS interval.^ It is evident that the use of a discrimination paradigm produces results quite different from those ^It might be argued, however, that conditioning in any natural environment is always discriminatory in nature and would be more clearly understood by attending to dis crimination conditioning than simple conditioning. ^There is recent evidence concerning the influence of CS intensity on learning which indicates that the parameter has little or no effect when a single CS inten sity is available to the subject, but quite effective when two or more CS intensities are present in the same condi tioning context when each CS is reinforced (Grice and Hunter, 1964). Similarly, Allen and Hill (1963) demon strated that UCS intensity does not affect CR magnitude when the relationship is assessed using a between subjects design, while such a relationship is demonstrated when a within-subjects design is employed (Wickens and Harding, 1965). The latter authors argue that for the GSR the within-subject design is superior to the between-subject design. 13 obtained in simple conditioning situations. Thus, Kimmel and Pennypacker (1963) find discrimination conditioning best at 2 seconds, an interval which seems to produce little or no conditioning in the simple paradigm. (An answer to this question, however, cannot fully be elabor ated until there is a study directly comparing simple and discrimination conditioning under comparable circumstances.) Lockhart and Grings (1964) found that the degree of GSR discrimination was equal for groups conditioned under .5 and 5 second CS-UCS interval conditions. This result for the discrimination paradigm recalls Silverman's (1960) result with simple conditioning in which no difference was found between the short and long interval conditions. It would thus seem obvious that more data are required in this area before one can firmly assert that the discrimination and simple paradigms are radically different. However, the same optimism cannot be held in terms of the effects of the CS-UCS interval itself. It is widely felt and considerably promulgated in textbooks and journals that a simple function relates the degree of conditioning and the CS-UCS interval. Despite the evidence presented thus far which is contrary to this view— supporting a function with at least two maxima— the notion persists. 14 The proposition was put forward that variation in the CS-UCS interval may result in a situation allowing fundamentally different principles to operate when the interval is brief (e.g., .5 sec.) compared to when it is long (e.g., 5 sec. or longer). The extent to which this conclusion is justi fied was not made clear until Grings, Lockhart and Dameron (1962) published a study of autonomic conditioning in the mentally defective child. Interest in the role of verbal-perceptual factors in human autonomic conditioning led these investigators to a population of humans presumably on the "lower end" of what might be described as a dimension of verbal-perceptual capacity. They reasoned that if this were true, the men tally defective would be less able than the college sophomore to readily perceive and verbalize a contingent relation between CS and UCS. In other words, it was expected that mental defectives would yield learning curves with a slow and gradual development of conditioned respond ing. Grings, Lowell and Honnard (1961) had already demon strated this with nonverbalizing preschool age deaf children. Two CS-UCS intervals were used (.5 and 5 sec.) in a delay paradigm with an electrotactual UCS occurring at CS 15 offset. Two CSs were employed. One stimulus (e.g., tone) was reinforced by the electric shock (CS+), while a second stimulus (e.g., light) was never reinforced by.shock (CS-). C A 60 per cent partial reinforcement paradigm was used m which CS+ was presented 40 times and reinforced on 24 of these occasions. The CS- was also presented 40 times, but without reinforcement. The 16 nonreinforced presentations of CS+ served as "test" trials (i.e., trials on which con ditioning was to be assessed). These were scattered unsystematically throughout the acquisition period which followed an adaptation series of 6 presentations of each stimulus without pairing. CS- presentations adjacent to the test trials served as "control" trials. It was expected that discriminatory responding would develop between these stimuli as a result of their differential treatment. Discrimination conditioning was measured by the difference in magnitude of GSR to these two stimuli on test and control occasions. ^The use of a partial reinforcement paradigm in classical conditioning is rare. Pavlov (1927) reported that such a schedule retarded learning. The paradigm was employed in this study to avoid giving "too many" shocks to these patients and yet have enough trials to observe the expected gradual development of conditioned responding. 16 Figure 3 shows the magnitude of discrimination conditioning when the CS-UCS interval was .5 second. This curve increases in almost linear fashion with the number of reinforcements. Conditioning reached a statis tically significant level about midway during acquisition (point 4 on the graph). In many respects this curve parallels eyeblink discrimination obtained at this inter val . Figure 3 also shows the discrimination curve for the anticipatory response (i.e., the response which occurs in the 5 second period prior to UCS onset). The overall degree of discrimination was quite significant and significant discrimination was obtained about one-third the way through the acquisition phase (somewhat more readily than was the case with the .5 sec. CS-UCS interval). Comparison of these curves shows that to point 6 (i.e., three-fourths of the way through acquisition) there was no difference between the degree of discrimination in the shorter and longer interval groups. The reason for the dropoff in the longer interval group is a common phenomenon in long delay conditioning. Grings (1960) has shown that this dropoff is not equivalent to a decrease in learning but rather is a result of 17 o •H +» at d ■ H a •H O a •H CO CS VI o +> •H a s €> £ • - - e .5 Sec. • - — e 5 Sec. 0 - - 0 5 Sec. 5 Sec. (First Response) 5 Sec. (Second Response) / Pairs of Trials Fig. 3.— Moan magnitude of GSR discrimination as a function of number of differential reinforcements under two interstimulus interval conditions (after Grings, Lockhart, and Laaeron, 1962). 18 changing perceptions on the part of the subject. Whatever the explanation for such a phenomenon, it is apparent that quite different things are happening in the 5 second group than in the .5 second group. Earlier the possibility of two maxima in the func tion relating degree of conditioning and the CS-UCS interval was mentioned. Even this, however, is probably too simple a characterization of the underlying processes involved when there is variation in the interstimulus interval param eter. It is probably not meaningful to speak of a single CS-UCS interval function when there is such great difference in the response topography in relation to the stimulus situation. An autonomic response which occurs prior to a shock may be under the control of entirely different pro cesses than when such responding is subsequent to shock. Putting these two responses on the same continuum is not only operationally questionable, but logically indefensible. The Grings, Lockhart and Dameron (1962) study offers even more compelling data which demand such a con clusion. They appear to have found what appears to be a basic but hitherto unrecognized characteristic of the conditioning process. 19 The "Second Interval" Conditioned Response On the test trials in the aforementioned study a very definite response was observed which oftentimes looked very much like the UCR. Yet, on these trials no UCS was presented. A graphic illustration of this phenomenon is presented in Figure 4. This second response was so unexpected that the shocking circuits were checked several times before the experimenter could accept the fact that no shock was getting to the subject. The second response naturally raised the question of whether or not it too, in addition to the first or anticipatory response, was a conditioned response. For want of a better name this response was called a "second interval" response. These responses were obtained on approximately two-thirds of the test trials ^and one-third of the control trials. If this second response was a CR a second question would become evident: do the anticipatory CR and the second interval CR follow the same "laws of conditioning?" Reference again to Figure 3, page 17, shows the discrimination data concerning the second interval CR, the response which occurred where the UCR would normally occur 20 Point of U B u al UCS presentation CS No UCS presented Pig. 4.— Illustration of test trial GSR pattern indicating the so-called "anticipatory" response and the response near the temporal point of usual UCS occurrence which is obtained in the absence of the UCS, but without the UCS present. Notice particularly the rather high degree of conditioning which is evident at the first point on the graph (discrimination was significant at this point). It is obvious that this response shows the effects of the discrimination paradigm more readily than either the .5 second CR, or the 5 second anticipatory CR. While both the latter required several differential rein forcements before significant conditioning occurred, significant second-interval conditioning was obtained at the very first point of comparison during acquisition— in this case following only two differential reinforcements. If one reflects on this result for a moment it would seem that the subject fails to show discriminatory behavior during the period of CS delay prior to shock, but shows dramatic discrimination in the interval subsequent to that point where the UCS would have occurred. This situation led the authors to conclude that ". . . the explanation of second interval responding will require a reinterpretation of the nature of autonomic classical conditioning." It should be remembered at this point that the entire empirical substructure of classical conditioning (and much of the learning theory which it supports) is based on anticipatory responding. Grings, 22 Lockhart and Dameron (1962) have observed a response which exhibited more rapid and maximal conditioned behavior than was exhibited in the anticipatory response or in the response obtained when the CS-UCS interval was too brief to allow such responding. The possibility existed that the second interval response was idiosyncratic of the mentally defective population. A study on college sophomores (Lockhart and Grings, 1964) quickly dispelled this view. In this popu lation, administered the same pattern of stimulation, second interval responding again exhibited the most rapid and high degree of discriminatory responding. In this connection the authors noted that in terms of second interval responding there was little to differ entiate the college sophomore from the mentally defective (i.e., rapid and maximal conditioning at first point following two differential reinforcements). However, it was found that the first or anticipatory response dis crimination (and responding in the .5 sec. CS-UCS interval group) was much more rapid and greater in degree in the college student than in the mentally defective. The authors were led to speculate that the second interval response represents a more "basic" process in classical 23 conditioning than the anticipatory response. Second interval discrimination may develop prior to or be unaf fected by verbal-perceptual control, while the first response may be readily modified by such factors. What is the relationship between the inter stimulus interval and the degree of conditioned respond ing? The relatively small amount of data considered thus far would seem to invalidate the notion that the relation ship is the simple one so often described. The data support the notion that different processes are operative when the interval is long than when it is short. Moreover, and more importantly, within the same CS-UCS interval, when this interval is longer than the latency of the response, different processes seem to be in evidence. What these processes are and the degree to which it modifies our thinking about classical conditioning is still in the future. It is to this problem, however, that the present study is dedicated. This type of reasoning involves the assumption that the actual responses observed only reflect (and imperfectly so) the processes which produce them. If the responses behave differently one would seem justified in inferring that the processes which the responses reflect are different. CHAPTER II CONTIGUITY AND REINFORCEMENT IN CLASSICAL CONDITIONING THEORY If the second interval response is to modify our conception of classical conditioning, as proposed in Chapter I, the major changes are likely to be in the under standing of the concepts of "contiguity" and "reinforce ment." Therefore, and in order to grasp the full relevance of second interval responding for classical conditioning, it is necessary to analyze these concepts and the roles they play in contemporary theory. Contiguity The concept of contiguity has always played a fundamental role in theories of learning as well as in associationism, the historical antecedent of learning theories. In spite of these impressive historical and contemporary credentials, the meaning of the concept remains ambiguous. There appear to be at least two 24 25 families of meaning of the concept. For descriptive pur poses, these will be distinguished as contiguity I and contiguity II. Contiguity I. Contiguity I may be defined simply as the extent of the temporal separation between events. Examples of such usage abound in the conditioning litera ture. Kimble (1961) uses contiguity in a manner synonymous with ". . . to occur closely together in time" (p. 478). Hull (1943) uses contiguity defined as ". . . closely associated in time" (p. 80). Estes implies contiguity I when he writes ". . . that the response which occurs at any time becomes connected to any stimulus elements sampled during a short preceding interval (of the order of 0.5 sec.)" (p. 459). The implication of each of these and similar statements is that contiguity is to refer to some interval or degree of time. This durative notion is most often applied in two ways: (1) in reference to the temporal interval between stimulus events (i.e., the interstimulus interval as discussed earlier), and (2) in reference to the temporal interval between a response and stimulus event (e.g., delay of reinforcement gradient). 26 Contiguity II. In contrast to a contiguity inter preted as possessing any "degree" or "durative" character istics is the notion that contiguity means strict simultane ous occurrence of events. Estes at one time proposed to interpret contiguity as strict simultaneity, as did Guthrie (1935). However, he has abandoned his earlier views. The following statement is a typical rejection of the simultaneity view of contiguity: The precise definition of "contiguous" poses dif ficult problems. Originally, following the lead of Guthrie . . . I interpreted contiguity to mean strict temporal simultaneity. However, prolonged and strenuous attempts to develop a theory of reinforcement based on this interpretation have convinced me that no such theory can be con structed. . . . (Estes, 1959, p. 459) Very few, if any psychologists support a simultane ity view of contiguity. It will be argued here, however, that this view is the more nearly correct one. Comparison of contiguity I and II. In Figure 5 a simple classical conditioning paradigm is diagrammed. In this figure, contiguity I refers to the temporal interval between the onset of the CS and the onset of the UCS. Contiguity II is shown as that moment in time when the onset of the UCS is coincident with point X in the duration of the CS. (A subclass of contiguity I would involve the 27 extent of the temporal overlap of the CS and the UCS; a subclass of contiguity II would involve innumerable individual moments in time during which the CS would be simultaneously present with the UCS. These two subtypes, also diagrammed in Figure 5, are labelled contiguity I1 and II' respectively.) Notice that Figure 5 concerns definition of con tiguity only in terms of visible "stimulus" characteristics (i.e., onsets, offsets, points in durations, et cetera). For certain theoretical positions, contiguity involves both stimulus and response characteristics. Figure 6 com pares simultaneous and durative notions of contiguity in relation to certain CS and UCR characteristics usually thought to be critical in-S-R learning theory. There would be a contiguity I relation between CS onset and UCR onset. There would be a contiguity I* rela tion between the temporal overlap of CS and UCR. There would be a contiguity II relation between some point X in the duration of the CS and the onset of the UCR, and a II1 relation between points in time of the CS overlap with the UCR. It can be argued that the only meaningful inter pretation of contiguity is the simultaneity one. It is not 28 Contiguity 1 CS UCS 14- Contiguity 2 Contiguity 1' / ' H Contiguity 2' Pig. 5.— Comparison of two types of contiguity in a simple classical conditioning paradigm. 29 Contiguity 1' Contiguity 1 CS UCS UCR Contiguity 2 Contiguity 2' Pig. 6.— Comparison of two types of contiguity in relation to CS and UCR characteristics in a simple classical conditioning paradigm. 30 possible for an event at time t to exert an influence at time t + x except to the extent that the event which occurred at time t is "represented" at time t + x. Unless this idea is accepted one becomes committed to a doctrine of historical causation (i.e., that some event at time t can effect an event at time t + x without an intervening process--representing the event at time t— persisting through the period x). If interaction of events is to occur (and presumably this is what is necessary if conditioning is to occur) the two events (CS and UCS) must be present at the same moment of time. There simply can be no meaning to an interaction of events separated in time.l The usual retort to such a position is . this: it is known that little or no conditioning is obtained when the CS-UCS interval is 0, while a large amount of condi tioning is obtained when the CS-UCS interval is delayed by - a few tenths of a second. Such an argument clearly implies -'-The preceding is not meant to imply that any psychologist of learning actively promotes the notion of historical causation. Rather, the criticism, as will become clear, is directed at the ambiguity and impreciseness with which the concept of contiguity is used. When Estes (1959) says that a simultaneity view of contiguity cannot be held, the exact meaning of a nonsimultaneity contiguity must be specified if the notion of historical causation is not to slip through inadvertently. This is what investiga tors in the area fail to do. 31 that the terms "CS" and "UCS" are defined in terms of stimulus onsets. Since conditioning occurs when these onsets are separated in time and does not when there is strict simultaneity, it is argued that a strict simul taneity notion of contiguity cannot be accepted. However, it should always be kept in mind that learning occurs in the organism and not in the paradigm. That is, what may be defined as the CS and UCS by the paradigm (on an operational basis) may not be isomorphic with the effective CS and UCS as these are made manifest in the organism. If the CS and UCS onsets are simultaneous, it may be a mistake to infer that the neural "representations" of these events occur simultaneously. It is quite likely that the "critical" neural effect of the UCS would occur prior to critical neural effect of the CS since the former stimulus is usually physiologically "dominant" over the latter. Thus, when one considers the neural "image" of the simultaneous conditioning paradigm it is obvious that a backward dominance relation may exist, that is, a case in which the first stimulus is stronger than the second. Although the data for "backward" conditioning will not be examined here, it can be mentioned that whenever the first 32 stimulus is stronger than the second (in some sense or other) little or no conditioning occurs. Thus, when the CS and UCS onsets are contiguous the effective neural paradigm may actually be a backward conditioning one. Without further elaboration of this point at this time it may be said that successful conditioning may require the presence of a CS-correlated neural event(s) at the time of occurrence (simultaneously) of the UCS-initiated neural event(s). For such a situation to exist would require that the CS occur somewhat prior to the UCS. This does not mean that contiguity is to mean duration of such separation. Rejection of the durative interpretation of contiguity means only rejection of the notion that the actual CS is localized at the CS onset and the actual UCS at the UCS onset. It is assumed that the "actual" CS is some neural event which occurs in temporal interaction (simultaneity) with a second neural event which represents the "actual" UCS . The concept of neural trace. In short (e.g., .5 sec.) delay conditioning it would appear that few theorists see any problem in defining the CS as CS onset. However, in long delay conditioning (or long trace conditioning) 33 writers are hesitant to define the CS as simply the onset of the stimulus. It is at this point that the concept of stimulus trace is brought in. (However, the trace of the CS is almost always conceptualized.as a trace of the onset. Very little mention is ever made concerning the trace of subsequent portion of a durative stimulus. This may be a minor point, however, and will receive no further atten tion.) This "representation" of the stimulus onset is said to move forward through time towards the UCR (or UCS). The trace, however, does not remain constant, but "fades" as a direct function of the passage of time— implying some limit on the extent to which one could obtain conditioning with a trace as "the" CS before such a trace fades into the background of organismic "noise." If this trace idea is accepted it becomes a matter of definition that contiguity means simultaneity, since it is the characteristic of the trace itself at the moment of interaction with the UCS which determines whether or not conditioning will occur. If the trace fades "too much" it can no longer be contiguous with the UCS and conditioning will not occur. Therefore, a durative notion of con tiguity reduces to a simultaneity notion of contiguity. 34 If this is true, it should be evident that con tiguity as such is not and cannot be a variable. Contigu- ity is a moment or quanta of time. One cannot have more or less of a moment of time. What is important is the characteristics of the events interacting during this moment of time. When an investigator demonstrates that a particular response is conditioned best at .5 second, but not at 5 second, he cannot conclude that the critical range of contiguity ends at .5 second. What he can con clude is that under .5 second conditions, the relationship between interacting events is more conducive to condition ing than when the interval is shorter or longer. An interesting view of the conditioning process and one which is most relevant to the understanding of second interval responding emerges from this conceptual scheme. 2 The idea of a "quantum" of time is meant to imply simply that a moment of time may itself have a duration. This does not contradict a simultaneity notion of contigu ity. What it does mean is that the critical events that interact "take time" and the boundry of this quantum of time is defined by these events. Thus, for example, suppose the critical event is simply the summation of two neural impulses which then stimulate a third neuron. Suppose neural event "a" lasts for 100 msec. If at any time during this period neural event "b" occurs, neuron "c" will be stimulated. Thus, simultaneity in this instance has a "duration" in the sense that time itself is defined by the 100 msec, quantum interval. 35 Before pursuing this conceptualization further, however, it will be necessary to review how the concept of contiguity functions in modern learning theory. This will be accom plished by stating as briefly as possible the supposed "Laws" of classical conditioning as enunciated by Pavlov (1927), Hull (1943) , Mowrer (1960) , Jones (1962) and Razran (1957) . Pavlov. There seem to be at least three somewhat different conceptions of contiguity in the writings of Pavlov: (1) Hence, a first and most essential requisite for the formation of a new conditioned reflex lies in the coincidence in time of the action of any previously neutral stimulus with some definite unconditioned stimulus, (p. 27) (2) The fundamental requisite is that any external stimulus which is to become the signal in a conditioned reflex must overlap in point of time with the action of an unconditioned stimulus, (p. 26) (3) . . . it is also and equally necessary that the conditioned stimulus should begin to operate before the unconditioned stimulus comes into action, (p. 27) The phrase "coincidence in time" would seem to imply a simultaneous view of contiguity. Similarly, the phrase "overlap in point of time" could be interpreted as simul taneous occurrence of events. Superficially, it is 36 difficult to reconcile these implications with the third statement, i.e., that the CS must begin prior to the UCS. The latter statement implies a degree type of contiguity. On first analysis, then, Pavlov's treatment of contiguity is somewhat ambiguous. In order to diagram such requirements it is neces sary to define what, is meant by the concepts of CS and UCS. At least one definition of "stimulus" involves any energy change in the environment. Such a definition of stimulus clearly implies that the CS and UCS are to be equated with the onsets of the CS and UCS {since these are the points of energy change in the environment). Pavlov (1927),- for instance, states: "... we have considered only one broad group of conditioned stimuli, namely those derived from the appearance of any natural agency." A second energy change is localized at the offset. Pavlov also spoke of this energy change as useful for a CS: ". . . but the disappearance also of such an agency may become the stimulus to a conditioned reflex." If CS and UCS are defined by stimulus onsets one cannot adopt a simultaneous view of contiguity since simultaneous onset produces no conditioning. _ In a previous discussion it was shown that 37 maintaining a simultaneity view of contiguity requires reference to some process (e.g., a trace) internal to the organism which allows the effects of stimulus onsets separated in time to be simultaneously present. This particular view also receives some support from the follow ing analysis by Pavlov (1927): . . . every stimulus must leave a trace on the nervous system for a greater or less time. . . . One can speculate whether any external agent of uniform, constant strength acts on the particular analyzer of the animal, and where there is a gradual fading of the residuum, the trace of actual stimulus which has ceased; each intensity of the stimulated cell, at each separate moment of time, is an indi vidual element, differing from both all preceding and all following stages of intensity. These points can best be brought to focus in relation to the trace conditioning paradigm, that is, one in which there is no temporal overlap between the externally presented stimuli. Figure 7(a) shows such a paradigm. In Figure 7(b) the trace of the CS is shown. Point "x" on this trace is the point coincident in time with the UCS. Figure 7(c) elaborates this situation somewhat by consider ing each moment in the trace a separate "stimulus element." Instead of a smooth trace, what Pavlov (1927) seemed to have in mind was a kind of "step" function with each succeeding step of lower intensity than the preceding one. 38 CS UCS CS Trace CS UCS CS Trace (a) (b) CS UCS (e) Pig. 7.--(a) Typical trace paradigm; (b) hypo thetical CS trace; (c; stimulus trace as viewed by Pavlov (1927). 39 Only when one assumes that the actual CS is such a moment in the trace of the externally presented CS do Pavlov's statements employing contiguity become consistent. In addition, this interpretation of Pavlov coincides with the analysis of contiguity presented earlier. Hull. There are two types of contiguity evident in Hull's theorizing. His fundamental law of conditioning is stated as follows: Whenever a reaction (R) takes place in temporal contiguity with an afferent receptor impulse (s) resulting from the impact upon a receptor of a stimulus energy (S), and this conjunction is followed closely by the diminution in a need (and the associated diminution in the drive D and in the drive receptor discharges, Sq) , there will result an increment (s R) in the tendency for the stimulus on subsequent occasions to evoke that reaction. (Hull, 1943, p. 71) The use of the phrase "temporal contiguity" in this statement implies a simultaneity view in which afferent receptor discharges (s) must be present at the time of the reaction (R), otherwise there could be no "conjunction." For conditioning to occur, this "conjunction" of s-R must be followed closely (in time) by some type of event associated with need reduction (operationally identified with UCS offset). This use of the term closely seems to imply a durative view of contiguity. 40 This situation is diagrammed in Figure 8. Hull did not give the conjunction between s and R at time t (i.e., prior to UCS offset) the status of a conditioned connection. For this connection to become a "conditioned" one it must be "reinforced" by the drive reduction. In this formulation, the onset of the UCS plays no apparent critical role in the associative process, although it does elicit the response to be conditioned and, in addi tion, creates the drive, which when reduced "cements" the connection of receptor discharges to the UCR. As far as can be determined, Hull did not discuss or elaborate on these uses of the concept of contiguity. However, in 1951, Hull modified his primary reinforcement postulate to read: Whenever an effector activity (R) is closely associated with a stimulus afferent impulse or trace (s) and the conjunction is closely associ ated with the rapid diminution in the motiva tional stimulus (Sd or Sg), there will result an increment to a tendency for that stimulus to evoke that response. Close inspection of this statement shows clearly that the original phrase (i.e., "temporal contiguity") has been changed to "closely associated." Both uses of contiguity thus assume a durative interpretation of contiguity. However, by creating a degree type of 41 Point of drive reduction s-R conjunction a CS ucs UCR Pi*. 8.— Diagrammatic representation of Hull's (194-3) theory of classical conditioning. 42 contiguity between the R and the afferent impulse or trace, Hull apparently would allow the afferent trace to terminate prior to R, for if the trace were to persist to the point of R there would be no meaning to "closely associated." Whether Hull really intended this is not clear. It is, however, an untenable position. It is interesting in this context to question the mechanism by which the CR is said to become "anticipatory." Notice that in the primary law of reinforcement there is no such mechanism since the law specifically refers to a con junction consisting of s (the stimulus trace) and R (the UCR) . Thus the conjunction is dependent upon a response which occurs subsequent to the UCS. How does this "con junction" become anticipatory? Two general mechanisms have been postulated, both of which can be found in Hull's writings. First, any point in the subsident phase of a stimulus trace which is contiguous with the UCR will also occur during the recruit ment phase of the stimulus trace. This is shown in Figure 9. The symbol "x" represents the point in the stimulus trace which is contiguous with the UCR. The symbol "x1" represents this same value in the recruitment phase of the trace. Since X' will be present on trials 43 . X cs Trace CS UCS UCR Pig. 9.— Diagrammatic representation of the so- called ”identical stimulus” mechanism of anticipatory responding in the classical conditioning situation. 44 subsequent to the first reinforcement, whatever conditioned potential accruing to X will be elicited by X'. If this occurs, then the response (now CR) would antedate the UCS. This mechanism might best be termed the identical stimulus mechanism. This proposal runs into difficulty when the reinforcement occurs at the same time as the stimulus trace reaches its maximum "intensity," (i.e., at approxi mately .45 sec.). Such a point, by definition, has no "prior" occurrence and thus could not lead to anticipatory responding when the CS-UCS interval approximates .45 second. This is a difficulty because it is precisely this interval which Seems to produce "maximal" conditioning of certain skeletal responses. The second mechanism might be termed the similar stimulus mechanism. Basic to this idea is the notion of "generalization." The crux of the proposal is the assumption that each successive moment in a stimulus trace bears a certain resemblance in terms of "intensity." A response conditioned to one of these elements would then generalize along this dimension of intensive similarity such that elements occurring prior to the UCS become able to elicit the response (now CR) prior to UCS occurrence. 45 This is the meaning of the phrase in conditioning "to move forward in time." Notice that the appearance of an antici patory response is not evidence for conditioning, but for generalization of a conditioned response, and at once poses the question of whether or not it is ever possible to observe a "nongeneralized" CR. There is some indication that this problem has been recognized (Mowrer, 1960), but it has never been stated with great clarity nor squarely faced. Thorndike spoke of it as "associative shifting," Mowrer as "generalizing forward." If this is true, if what is observed as an anticipatory response is a generalized conditioned response, then it must be the case that the connection has already been formed prior to its generaliza tion. This implies that conditioning has occurred before any evidence of it is to be found in the anticipatory responding of the organism. This point should be firmly fixed in the reader's mind in reference to the possible implications of second interval responding to be elaborated later. Mowrer. Mowrer's most recent use of the concept of contiguity is quite close to this author's thinking. He uses contiguity in its strict sense: simultaneous occurrence 46 of events. Figure 10 reproduces Mowrer's conceptualization of the conditioning process. Sn may be either an external stimulus or a response-produced stimulus. Each is thought to leave a trace of itself reverberating in the organism (indicated by the wavy line). This trace is felt to diminish in intensity (indicated by the gradually diminish ing "amplitude" of the wavy line). Sitr (Mowrer's symbol for the UCS) is visualized as eliciting a total UCR (R in the figure) the conditionable component of which is sym bolized as r and is thought to be an implicit emotional response (e.g., fear). Mowrer postulates that the CS through it trace becomes strictly contiguous with the emo tional component of the UCR. This simultaneous occurrence of a stimulus event (CS trace) with a response event (r, or fear) results in the formation of a conditioned connection. To account for "anticipatory" responding, Mowrer (1960) employs the similar stimulus notion referred to earlier. Mowrer visualizes r as generalizing forward on a dimension of stimulus intensity, or better, stimulus trace intensity. This argument obviously requires the assumption that a weak intensity is sufficiently similar to a stronger intensity such that generalization may occur. It is apparent that Mowrer would agree that anticipatory Environment Sn S .-R Sit -R Pig* 10.— Diagramnatic representation of Mowrer' (I960) theory of classical conditioning. 48 responding is not evidence for conditioning per se but for the generalization of a conditioned response already formed prior to anticipatory behavior. Jones. Jones (1962) has developed a two-factor theory of classical conditioning with contiguity as one factor and reinforcement as the other. Jones interprets contiguity to mean the degree of temporal asynchrony between the CS-initiated neural event and the peripherally observed UCR. Maximal conditioning is thought to occur when the CS-initiated neural event is simultaneously present with the UCR, with lesser degrees of conditioning obtained when these events are "separated" either in a forward or backward manner. Jones uses a simultaneous notion of contiguity only in reference to the point at which conditioning would be maximal. For less than optimal conditioning, Jones inter prets contiguity in the sense of duration. This, as has been argued, confounds the idea of contiguity with the relationship and characteristics of events occurring in contiguity. Razran. In 1957, Razran combined his earlier notions of "dominance" (Razran, 19 30) with certain 49 "contiguity" ideas to form what he termed a dominance contiguity theory of the acquisition of classical condi tioning. The concept of "dominance" is related to the concept of reinforcement and will be given more detailed consideration shortly. While Razran (1957) is quite spe cific about the meaning of "neural dominance" little attention is paid to what might be termed "neural contigu ity." He speaks about the interaction of neural events as being the critical determiner of the conditioning process (providing that the interacting events— neural representa tions of CS and UCS— meet certain dominance requirements). This point of interaction of neural events may be considered the meaning of neural contiguity. This, of course, commits Razran to a simultaneity view of contiguity. Razran never comments on this interpretation directly but he does regard the lengthening of the CS-UCS interval as synonymous with the decreasing of the strength of the CS-initiated neural event. This type of reasoning has meaning only if one assumes that the critical events for conditioning occur at the point of neural interaction. It is only at such a point that the decreased intensity of the CS- initiated neural event could be reflected in the condi tioning process. Upon analysis,' then, it would seem that 50 Razran's theory requires a simultaneity concept of contiguity. Summary. Two conceptions of the meaning of con tiguity have been described. One of these involved the notion of degree of time between events, while the other involves only simultaneity of events. It was argued that in relation to the "neural image" of the conditioning paradigm only a simultaneity interpretation of contiguity is meaningful. For such meaning to accrue to this inter pretation of contiguity requires conceptualization of the critical events in conditioning as interacting neural events (i.e., neural representations of the CS and UCS). For this reason contiguity is not regarded as a true variable since one cannot vary a moment of time. The important events for conditioning occur during such moment(s) of time and it is their relationship which is critical if conditioning is to occur. The analysis of this relationship brings us to the question and problem of "reinforcement" in the classical conditioning process. 51 Reinforcement The concept of reinforcement has received much more extensive treatment by learning theorists than has the concept of contiguity. While the concept of contigu ity is surrounded by an aura of "being understood," only controversy surrounds the concept of reinforcement. There are three main points of controversy: (1) What is reinforced, stimuli or responses? (2) What is it that reinforces, stimulus occurrence or response occurrence or change in some internal motivational state? (3) Does all learning require reinforcement (e.g., does conditioning require only contiguity, or both contiguity and reinforcement)? These questions have been the basis for polemical debate for many years which has led some writers to ques tion their meaningfulness (e.g., Kendler, 1952). Certainly there is no agreement, nor is there likely to be for some time to come. The following remarks, then, are not to be considered as a clarification but rather as an illustra tion of the confusion. Pavlov. Pavlov is so often described as a pure contiguity theorist that it is necessary to point out that 52 this is simply not true. While much of Pavlov's writing concerning basic requirements for conditioning dealt with and stressed temporal relations between stimuli, Pavlov never felt that these were the necessary and sufficient conditions for learning to occur. One of the little- mentioned, but quite critical requirements for the occur rence of conditioning was that the UCS-initiated reaction be "physiologically more powerful" than the CS-initiated reaction. The UCS was required to dominate the CS. If the term "reinforcement" is to be applied to Pavlov's theory it would be in relation to this dominance requirement. That is, the strength of the UCS provides the means whereby ". . . external stimuli . . . transmitted to a definite center . . . can be diverted and made to follow another route." For Pavlov, responses are not reinforced; rather, it is the reinforcing "power" of the UCS which "forces" new nervous connections creating the situation whereby the CS gains "access" to the center of UCS reception. Pavlov's theory is illustrated in Figure 11. For Pavlov, the connection which is conditioned is a new neural pathway between the afferent CS center and the afferent UCS center. The connection is in no way to be interpreted as a Conditioned connection UCR CS Neural Level External Events CS UCS UCR Pig. 11.— Diagrammatic representation of Pavlov' (1927) theory of classical conditioning. connection between the CS center and the motor center of the UCR or of the peripheral UCR as Pavlov's theory is so often characterized (e.g., Kimble, 1961). Pavlov was strictly a stimulus-stimulus theorist in terms of what he considered was the learned connection. The UCS, as the reinforcing agent, was said to create a "dominant focus" within the brain which attracted less dominant points of excitation to it as a magnet might attract iron filings. Since the CS afferent focus is nonrandomly excited just prior to the UCS afferent point, it would be primarily the CS excitation which is attracted to the UCS focus than other points of excitatory focus which would be randomly occurring (or at least not contingently related to the CS). Reinforcement for Pavlov might best be thought of as a "forcing" mechanism: the effect of the UCS is to "forge" new connections by repeatedly pulling CS initiated excitation to itself. Hull. Hull's concept of reinforcement is a hedonistic one. Whatever response is occurring near the point of a reduction in an internal drive (motivation) state will be connected to any stimulus or stimulus trace present at that time. In reference to classical 55 conditioning, the UCS serves two roles: (1) initiation of a heightened drive state (e.g., the onset of a painful stimulus), and (2) termination of the drive state (e.g., the offset of the painful stimulus). As noted before, two types of contiguity were described by Hull (1943). One is a contiguity (type II) between the stimulus trace and the UCR; the second a contiguity (type I) between this "conjunction" and the reduction of the drive state. Hull's diagram (1943) of the Pavlovian paradigm indicates clearly that the point of reinforcement (e.g., shock offset) must occur subsequent to the UCR if condi tioning is to occur. The example employed is one where this is true by definition, i.e., an escape response which is obviously followed by the reduction in drive- receptor impulses contingent upon shock. One problem with the Hullean interpretation of reinforcement involves the relation of point of reinforce ment with long latency autonomic responses. A GSR to shock, for example, has a peripheral latency no shorter than one second. When a CS-UCS interval of .5 second is employed the entire trial including the termination of shock has occurred prior to the response. Yet, condition ing occurs readily and to a large degree. As discussed 56 earlier, such a situation is untenable for a reinforcement theorist. Another line of experimentation which indicates the difficulties with defining the reinforcing event as dependent upon shock-termination is a series of studies exploring the effect of varying UCS duration. Presumably, a long UCS would create an unfavorable contiguity situation leading to poor conditioning. However, the studies involv ing autonomic responses (particularly heart rate) have usually concluded that UCS duration has a motivational effect, but little or no influence on learning. Still another series of arguments cast some doubt on this theory as originally stated. If the UCS serves as both drive initiator and drive reducer how is it possible for avoidance learning to occur and persist in spite of the fact that no UCS is administered when the animal responds prior to the point of UCS onset? This problem has been labelled the avoidance "paradox." The "solution" of this paradox has led to theoretical developments involving conditioned implicit responses of an emotional nature (e.g., fear) which create the necessary motivation for the overt avoidance behavior which in turn terminates the implicit emotional state. Reinforcement under this 57 interpretation (i.e., secondary reinforcement) is identi fied with the reduction of internal emotional responding. Theoretical developments along this line indicate the increasing importance of the concept of classical condition ing (i.e., the means by which such emotional responses are learned). This is a turnabout from Hull's theorizing in which classical conditioning was subsumed under instru mental learning. Now it appears that instrumental learning is more properly conceptualized as being dependent upon classically conditioned internal response states (e.g., Mowrer, 1960). Mowrer. The central point of Mowrer1s most recent theorizing is that classical conditioning applies only to internal emotional responses most meningfully identified with the autonomic nervous system. The responses thus conditioned serve as both motivation and reinforcement for overt (skeletally mediated) behavior. Mowrer believes that overt behavior as such is not learned but only modified by the nature of the classically conditioned emotional responses. In the most important respects, Spence (19 56) adopts a similar view. Thus, for Mowrer, reinforcement has little meaning for classical conditioning. In this sense, 58 one could classify Mowrer as a "pure" contiguity theorist. Jones. Jones' (1962) two-factor theory of classi cal aversive conditioning defines reinforcement as . . some feature of the UCS presentation . . . most likely its onset or offset." She cites the oft-mentioned Mowrer and Soloman (1954) study showing that the UCS onset is the critical event in classical conditioning, but points out that when the UCS is of short duration the assumption of either onset or offset as the critical variable leads to the same prediction. Of interest at this point is the concept of "gradient of reinforcement" postulated by Jones. The optimum point of reinforcement would be correlated with the onset of the UCS spreading in decreasing manner to both sides of this point. She then postulates an inter action of the contiguity and reinforcement gradients. However, before the reinforcement can have an effect on the CR, the CR must of necessity occur prior to the onset of the UCS. She is led to postulate that "the initial connection would appear to depend upon the operation of the contiguity principle alone." What Jones seems to be saying is that the classical 59 conditioned connection is formed solely on the basis of the contiguity variable but is then modified or influenced in some manner by the reinforcement gradient. In fact, Jones hypothesizes that late in training the contiguity variable would have a minimal effect on the conditioned connection while the reinforcement variable would have a maximal effect. This formulation may be another way of saying that early in training classical conditioning occurs, which, when the subject begins to respond in an anticipa tory manner, is modified into an instrumental conditioning situation as we have discussed earlier. Razran. As noted earlier, Razran's concept of reinforcement is to be found in the Pavlovian idea of dominance. The central postulate in Razran's (1957) revitalized Pavlovian theory is "neural dominance." Most simply, neural dominance means the extent to which the UCS-initiated neural event (symbolized as UCSn) exceeds in neural strength the CS-initiated neural event (CSn). As a determining variable, neural dominance is expressed as a ratio of the strengths of the two critical neural events (UCSn/CSn). Subsidiary concepts involving this ratio are 60 termed (a) minimal dominance, (b) over dominance, and (c) optimal dominance. Minimal dominance is a threshold like concept which requires the ratio to exceed unity by a certain degree before conditioning will occur. Over dominance, also a threshold-like concept, implies that the degree of conditioning will decrease if the dominance ratio is too great. Optimal dominance (or "favorable ratio") implies that there exists some dominance value (for every absolute stimulus intensity) at which conditioning will be maximal. A final concept, neural action level, implies that the extent of conditioning also varies as a function of the absolute intensities of CSn and UCSn, i.e., these values must exceed, individually, some minimal amount before they would be able to enter into an inter action relation. The UCSn/CSn ratio is operationally defined in Razran's system by another ratio. The neural strength of CSn is defined in terms of the intensity of the externally applied CS (CSe). The neural strength of UCSn is defined by the magnitude of the externally observed response to the UCS (UCRe). The critical ratio (UCSn/CSn) is then defined as UCRmagnitude/CSintensity. According to Razran, neither contiguity nor 61 dominance are sufficient in themselves to produce condi tioning. The fact of conditioning is seen as an inter action of dominantly related events at the moment of contiguity. Thus, conditioning is essentially a function of the relation between neural strengths of neural events occurring together at a particular moment in time at some as yet unspecified locus (or locii) in the nervous system.^ Summary. A number of conceptions of reinforcement have been described. Pavlov (1927) viewed the UCS as eliciting a physiologically powerful event which forces or forges new neural connections such that the CS-initiated neural center becomes functionally connected with the UCS neural center. This idea is continued in the theorizing of Razran (1957) where it is felt that conditioning is a Razran has been criticized for his neurologizmg. Diamond, Balvin and Diamond (1963) write: "Razran's emphasis on empiricism and on 'neural events' places him under obligations which he does not meet" (p. 269). Spe cifically, however, the criticism is pointed at Razran's failure to include in his neural model a concept of "neural inhibition" which the authors feel is necessary of any neural model unless the theory is to be nothing more than an "anachronism." It should be recognized, however, that theories become anachronistic only when they fail to predict, not when they fail to include one concept or another. The question is whether or not Razran's theory is predictive of behavior in the classical conditioning situation. 62 product of the degree to which the UCS neural event "dominates" the CS-initiated neural event at the moment of contiguity. Razran does not view the UCS as forcing neural connections but rather the neural connections result from an intensity interaction of these events at a critical time. Jones (1962) visualizes the UCS as creating a reinforcement gradient which operates on a conditioned response only after it has been established through the contiguity of a CS-correlated event with the UCR. Mowrer (1960) proposes that classical conditioning occurs when a CS or its trace is contiguous with an emotional component of the UCR. The critical stimulus event is the onset of the UCS but only in the sense of eliciting the conditionable component of the total unconditioned reaction. Hull (1943) viewed classical conditioning as a subclass of instrumental learning and thus committed him self to a drive-reduction view of reinforcement operative in the classical conditioning situation. For Hull, this meant identifying the reinforcement with the events attendent to the offset of the UCS. Such a view encounters serious difficulties in relation to the question of reinforcement of long latency responses and with avoidance 63 behavior. Synthesis. It would appear that no single theory described is adequate to explicate the mechanism of classi cal conditioning. However, it may be profitable at this time to attempt a general statement of the conditioning process combining various features of each theory. The present author feels that conceptualizing the conditioning process in terms of observable stimuli and responses is no longer profitable. Such objectivity has long since rooted mentalism from its hold on psychological theorizing. This point is made manifestly clear when the phenomenon of trace conditioning is analyzed. Such conditioning requires a concept of an internal process (at present unobservable). It is~felt that the critical determiner of classical condi tioning is the nature of the relationship between the neural representation of the CS and UCS at the moment or moments during which they interact. The present writer feels, along with Pavlov and Razran, that conditioning is dependent upon the intensity relation of these interacting neural events, i.e., a dominance relationship. Once such conditioning is established (the number of trials not being an issue here) it would appear that 64 some type of generalization mechanism operates which results in the CR being elicited prior to the UCS. It is at this point, when the CR anticipates the UCS, that the paradigm no longer represents a classical conditioning paradigm, but rather undergoes a transformation into an instrumental paradigm. That is, when the CR antedates the UCS, this response will modify and influence the effect of the UCS on the organism. Three principle statements can serve as a summary of the content and intent of the present analysis: (1) classical conditioning occurs prior to its observation in anticipatory response; (2) conditioned responding general izes along a dimension of stimulus (trace) intensity; (3) the occurrence of anticipatory responding alters the situation in such a way that the subject's behavior becomes instrumental in modifying the effect of the UCS. Is the second interval response "the" conditioned response? This chapter began with the assertion that the phenomenon of second interval responding would importantly modify our understanding of the concepts of contiguity and reinforcement as these apply to the classical conditioning situation. 65 There seems to be an idea running through the theoretical statements examined in this chapter which implies that the first instance of a conditioned connec tion would not be present anticipatory to the UCS. Only gradually (or at least subsequently) is evidence for con ditioned responding presented in the form of anticipatory responding. The central question seems to involve whether or not the second interval response as described in the previous chapter reflects such pre-anticipatory conditioned responding. CHAPTER III DOMINANCE AND CONTIGUITY AS INTERACTIVE DETERMINANTS OF THE CLASSICALLY CONDITIONED AUTONOMIC RESPONSE The Influence of Razranean Theory Before detailing the exact nature of the experi mental question and the experimental procedures employed to obtain information concerning the problem two major influences of Razranean theory should be made explicit. Evolutionary levels of learning. One influence involves the concept of "levels" of learning. Razran (1955) postulates that in the evolutionary development of animal structure and organization there occurs a concomi tant evolutionary development of the qualitative types of behavioral modification. For example, at the lowest level (unicellular organization) only sensitization and habitua tion are considered influential in terms of modifying animal behavior. At a low level of multicellular organiza tion classical conditioning becomes possible. At a higher 66 67 level, perhaps with the advent of a truly encephalized nervous system, operant-instrumental modification becomes possible. At still higher levels, perhaps with the advent of cortical control, conceptual learning becomes possible. Each higher level of organization brings with it both the accumulated bases of modification present in the older levels as well as the newly emergent modificatory possibil ity . There remains, however, a clear separation of the "level" at which these processes are effective even in the highest form of animal organization. For example, in the human being Razran holds that classical conditioning is possible only at an "unaware" level, meaning that classical conditioning Operates perhaps exclusively on the vegetative responses of the organism. The conditioning of the internal organs and autonomic responses allows Razran to conceptually relate such modification to concepts like "unaware" and "unconscious" (Razran, 1961). Operant and instrumental conditioning is conceptualized as involving "voluntary" components and is thus largely an "aware" or "conscious" form of modification. Conceptual learning requires the subject to be sensitive to the "meaning dimension" of a stimulus which is generally considered a conscious-type process. The study referred to earlier (Grings, Lockhart and Dameron, 1962) on the mentally defective children was motivated in part by such a conceptualization. Without the influence of verbal-perceptual factors, conditioning was thought to proceed on the basis of unaware processes. Razran (1955) argues that the term classical conditioning should be reserved only as a designation for the situation where the subject is unaware of the contingent relation between the CS and UCS. The prototype of such unaware learning is seen in the conditioning of interceptive responses (Razran, 1961). Both the CS and UCS are such as to be not perceivable by the subject thereby excluding a perception of the relation between them. Yet, condition ing occurs. This is perhaps the "purest" case of classical conditioning. When the person becomes "aware" of the contingent relation between the CS and UCS a number of processes are brought to bear on the paradigm in such a manner that the subject "acts on the paradigm" rather than the "paradigm acting on the subject." Under these circumstances, many of the critical variables of classical conditioning are likely to be less critical in determining the subject's 69 subsequent responding than are his attitudes and expecta tions developed as a result of verbal-perceptual inter action with the classical conditioning paradigm. The present writer likewise considers classical conditioning to be best studied under "unaware" conditions. However, of equal importance, if not more so, is the manner in which the subjects' behavior is modified in the classi cal conditioning situation when no attempt is made to control the subject's level of awareness. The question is a generally difficult one because of the difficulties encountered in trying to operationalize a concept such as "unaware." The point to be made here is simply that the problem is recognized as a relevant and important one in any study ojf classical conditioning employing, verbalizing organisms. Dominance-contiguity theory of classical condition ing. A second and more important influence of Razran's thinking on the present study has been his "dominance- contiguity" theory of the acquisition of classical condi tioning. One might think that a systematic theory concern ing a fundamental topic such as classical conditioning reported in 46 pages of a major psychological journal and 70 buffeted by more than 600 empirical references might occasion some response from the psychological community, especially among those who call themselves learning theor ists. In many quarters, Razran's contribution was not even recognized as being a theoretical statement. For example, Lawrence (1958) characterizes Razran theory as a "classi fied bibliography of over 600 Russian studies on salivary conditioning in dogs." Razran's theory is primarily a restatement and refinement of certain Pavlovian ideas. Yet, unlike the latter, the former has received no critical attention, nor has it proved a stimulus for empirical research. Not only has Razran's contribution been ignored on the theoretical side (e.g., the theory is not mentioned in a recent text book of conditioning, Kimble, 1961), but there does not exist a single experimental attempt to directly test the implications of Razran's 1958 dominance-contiguity model.^ 1-Two "exceptions" to this statement should be men tioned. First, Kimmel (1958) in a study on the effect of CS intensity on autonomic conditioning, determined for each £ what he termed a "CR/UCR" ratio, with the CR value being the "average response to the CS on preliminary trials" and the UCR value being the "average magnitude of response to the US on sensitization trials." He then defined five ratio intervals from 0 to 1.00 and constructed a contingency table representing those Ss within each interval who met the criteria for conditioning and those who did not. These data 71 While the present study, to be elaborated below, is primarily intended as an interpretive framework for the analysis of second interval responding, the author feels that the analysis requires a view of conditioning not too different from that proposed by Razran. Thus, the present study might also be interpreted as providing an empirical test of Razran's dominance-contiguity theory of classical conditioning. are reproduced below: Number of Ss CR/UR ratio learning not learning .81-1.00 2 4 .61- .80 6 8 .41- .60 10 8 .21- .40 5 1 0- .20 4 0 It can be seen that more than one-third of the Ss showing signs of learning were characterized by CR/UCR ratios falling in the middle range of values. This result was regarded as support for Razran's argument that condi tioning would increase up to a certain ratio and decrease thereafter. The second exception is a study by Passey (19 59) who attempted to test Razran's (19 30) formulation of "favorable ratios of excitation." His results, using the conditioned eyelid response, failed to support Razran's theory. The most ready interpretation of this negative result is that CS intensity and CS/UCS intensity ratios have not proved effec tive variables in the conditioning of skeletal responses, particularly when the study employs a between groups design. It is conceivable that the dominance variable effects only the conditioning of autonomic and not skeletal responses. 72 Formulation of the Experimental Problem Previous arguments have emphasized the critical importance of the neural events related to the onset of the UCS. It is felt that any stimulus which is (nonrandom- ly) neurally represented at that moment in time will inter act with the neural representation of the UCS and, depending upon the neural ratio of dominance, will become more or less conditioned. The greater the difference in the neural strength of the two events (or "event classes")— providing always that the second is stronger than the first— at this moment of contiguity the greater will be the conditioning. The "connection" inferred from conditioned respond ing is seen as one between neural systems (although how this is accomplished in actual fact remains unknown— but not unknowable). It is not felt that the connection is between one neural system and some peripheral response system (or central motor system). It has never been clear how an effector discharge can get connected to anything (except, perhaps through afferent feedback). The response itself is simply (or complexly) the by-product of neural discharge of an efferent system under the influence of an afferent system with the neural determinants occurring 73 prior to the response observed. Because of this the author cannot accept the usual statement that the actual connec tion is between the afferent representation of the CS and the peripherally occurring response. For example, Jones (1962) writes: "the occurrence of a response at an effector is the event whose temporal relations to other events is crucial in establishing learned connections." Perhaps this formulation confuses two things: (1) the justifiable use of the term "connection" meaning only stimulus-response correlation (i.e., that a particular stimulus is followed by a particular response), and (2) the justifiable use of the term "connection" meaning the actual nature of the physical connection within the organism (i.e., that a par- * ticular stimulus causes a particular response because neural system A gets connected to neural system B). Jones (1962) confuses these two meanings. But what is gained by implicating the nervous system? Is anything added beyond the connotations aroused by the term "neural" which could not be adduced by sticking strictly to observable stimulus-response correlations? At this point one is accustomed to discussing the so-called problem of reductionism. The present author feels that only pseudo-issues are involved in the problem of 74 reduction. The question of the neural basis of condition ing is a matter which will be settled on the basis of discovery, not by polemics or dogmatic pronouncements on the nature of psychological science. The author sees no fundamental difference between conceptualizing the process of conditioning as one involv ing neural dominance or one involving "fear." The matter is largely one of personal preference. If the concept of neural dominance is to have any meaning beyond its inter vening variable nature (i.e., "neural dominance" is not necessarily any more an hypothetical construct than is "fear") than the translation of external observable events (e.g., CS, UCS) into internal nonobservable events (e.g., CS-initiated "neural event") must be a careful one. This translation in most theories, if it is done at all, is almost too gross to be meaningful in terms of the actual functioning of a nervous system. Nor is the translation made simply by the use of the term "neural." It should be kept in mind that although the present problem is formulated in a context of neural terms and neural events each concept is defined on the basis of some observable or manipulable aspect of the external 75 environment. For this reason, the neural "super structure" is largely unnecessary, for the experimental problem could be formulated quite adequately without it. The experimental question. The experimental ques tion central to the present investigation may be phrased quite simply: is the degree of autonomic conditioning a function of the degree of neural dominance at the moment of neural contiguity? Three aspects of this formulation require special elaboration: (a) moment of neural contigu ity; (b) degree of neural dominance; and (c) degree of autonomic conditioning. Neural contiguity. Defining the "moment of neural contiguity" as the point in time of UCSn-CSn interaction requires one to view the adequate test for conditioning in somewhat altered form. Typically, the presence and degree of conditioning are inferred from the fact and characteris tics of the so-called "anticipatory" response, i.e., the response which occurs prior to the occurrence of the UCS. If one assumes, as is assumed here, that the "true" condi tioned connection is made during the moment of contiguity (most immediately related to the temporal point of UCSn which is most immediately definable in terms of the 76 temporal point of UCSe onset), then the test for such a conditioned connection must be made at this point or in such a way as to include this measurement possibility. This has not always been possible in classical conditioning experiments since the usual study employs 100 per cent reinforcement. Partial reinforcement is considered detri mental to the development of a classically conditioned response. However, as outlined in Chapter I and intimated in Chapter II, a number of recent studies employing a partial reinforcement paradigm have appeared. The major result of such studies has been the observation of the so-called "second interval response." It is a response which occurs during the normal UCR period in the absence of a UCS. The analysis of this response in Chapter II led to the postula tion that this second response may represent the "true" conditioned response, the observation of which would be required if the notion of neural contiguity is correct. That is, the neural CS occurs in temporal contiguity with the neural UCS. If the latter is omitted, the CSn is present and if conditioning has occurred the conditioned response (representing or reflecting the neural connection) should be observed at the point of CSn. 77 Therefore, a critical procedural detail of the present study involves the necessity of partial reinforce ment and particular attention to responding which occurs at or near the point where the UCS would have occurred on non reinforced "test" trials. Neural dominance. As indicated earlier, the matter of critical concern in a classical conditioning trial is the intensive relationship (i.e., dominance) between neural events at the moment of their interaction (i.e., contiguity). There are two ways in which the dominance relation may be manipulated. A direct procedure involves the intensity relation of the externally applied stimulation. The dominance ratio in such a procedure is defined simply in reference to the physical intensities of the CS and UCS (in terms of voltage in the present study). The major assumption involved in such a direct procedure is that the neural strength of the critical neural events varies lawfully with the physical intensity of the stimuli. The nature of this relationship need not be specified for the purposes of the present investigation. An indirect procedure does not Involve the variation 78 in administered stimulus intensities but rather the degree of temporal separation of the onsets of the externally applied stimulation. This procedure requires the assump tion that increasing the interstimulus interval produces a reduction in the strength of the neural trace of the initial stimulus. For example, if CS and UCS are the same physical intensity and are separated in time by a 10 second interval, the neural representation of the CS at the point of UCS onset will be considerably reduced in effective intensity in comparison with the strength of the neural trace at the point of CS onset. Employing both procedures in an interacting manner leads to an experiment in which there is both variation in the intensive relationship between external stimuli and in the temporal separation of these events. In the present experiment the physical intensity relationship between the CS and UCS was varied at three levels (i.e., values of CS intensity representing 1/3, 2/3 and 3/3 the value of a maximally intense UCS) as was the degree of temporal separation of stimulus onsets (i.e., values of .5, 5 and 10 sec.). Figure 12 presents a diagrammatic representation of nine experimental subgroups formed by the interaction of Ratio 3/3 0 3 .5 Sec, 7M V' F I // 5 Sec. 10 Sec. Ratio 3/2 \ \ 7. .5 Sec. 5 Sec. 10 Sec. Ratio 3/1 n m EL .5 Sec. 5 Sec. 10 Sec. Pig. 12.— Diagrammatic representation of hypothetical dominance ratios resulting from the manipulation of stimulus intensity ratios and degreeof trace interval. 80 direct and indirect means of varying the dominance ratio. The vertical bars represent the intensity values of the CS and the UCS. The horizontal dimension represents time between CS and UCS onset. The hatched vertical bar is intended to represent the "effective" intensity of CSn at the moment of contiguity. The value of this intensity, of course, is hypothetical. The vertical bars indicate that the CS begins and terminates prior to the UCS onset. The paradigm outlined in Figure 12 is therefore a trace paradigm. Such a para digm is employed in this study for a number of reasons. First, the author is generally interested in the so-called phenomenon of trace conditioning and feels that this phenomenon has never been adequately studied. Second, the analysis of conditioning presented in previous pages has implicated the concept of stimulus trace in understanding the basic mechanisms of conditioning. The concept of stimulus trace would seemingly be more meaningfully studied in a trace than in a delay paradigm. Third, the notion of decreasing CSn as a function of variation in the inter stimulus interval is more easily conceptualized in a trace than in a delay paradigm. Most importantly, however, is that the occurrence of "second interval" responding in a 81 trace situation would be difficult to explain in terms other than the analysis of conditioning outlined in previ ous chapters. Still another reason for the trace paradigm is that only in such a paradigm is it easy to quantify the direct ratio of stimulus intensities. To directly quantify the ratio a somewhat unorthodox conditioning paradigm was developed: both the CS and the CJCS were electrotactual stimuli. It should be obvious that the quantification of a stimulus intensity ratio is greatly simplified if the two stimuli occur in the same sensory modality. It should be noted here that a stimulus definition of the dominance ratio is quite different from that employed by Razran (1957). He defines the critical ratio in terms of CS intensity and UCR magnitude. Simply regarded, it is difficult to see what a ratio between a response magnitude and a stimulus intensity would mean. It is therefore necessary to question Razran's logic on this point. Razran, unlike Pavlov, is a stimulus-response oriented theorist. The "connection" to be explained is 2 between the CS and UCR. For this reason, the characteris tics of the UCR are likely to assume considerable impor tance. By observing that CR efficacy increases with UCR magnitude, he is able to postulate that when UCR magnitude data are available, it becomes superfluous to inquire as to the characteristics of the UCS. This approach represents a bias which may have detrimental effects. For example, the critical ratio is considered the determining factor in the formation of the conditioned response. Yet, the very definition of the critical ratio involves a response characteristic which, in many instances, cannot be determined prior to a condi tioning trial. It thus becomes logically impossible to manipulate the critical ratio as an independent variable. As defined, the critical ratio has only an' ad hoc status. Another point involves Razran's apparent oblivious ness to the possibility of instrumental modification of the effect of the UCS in the classical conditioning situation. This statement is true only for Razran*s theory at the empirical level. It is clear from Razran's postu lated neural interaction hypotheses that the "connection" is that between an afferent representation of the CS and an afferent representation of the UCS. In Razran*s model there is no concern for external response events. 83 That is, once the conditioned response is developed, the UCR is not likely to remain unaffected. While it is true that the administration of the UCS is independent of the behavior of the subject, it is not possible to say that the organismic effect of the UCS is independent of the sub ject's ongoing behavior. The decrease in magnitude of the UCR in eyeblink conditioning trials observed by Kimble and Ost (1961) reflects just this possibility. Obviously a variable which is considered a determining variable involv ing UCR magnitude as a necessary component of the defini tion of the variable would be affected by such a phenomenon. On the basis of Razran's proposed ratio one could conclude that conditioning should decrease since the dominance ratio (UCRmagnitude/CSintensity) decreases. This is simply not the case. For these reasons the present writer cannot accept the definition of UCSn in terms of UCR magnitude. However, rejection of this definition has no functional effect within the neural model proposed by Razran, since, as we have seen, no provision is made in the neural conceptuali zation for the importance of UCR magnitude. When information on UCS intensity is lacking, or when the CS and UCS are presented in different sensory modalities such that the ratio of stimulus intensities would be difficult to determine, then information concern ing UCR magnitude might be useful. However, testing of the theory requires operationalizing the concept of dominance as a determining variable capable of independent variation. This cannot be done in an ad hoc manner. Therefore, in this research, the ratio of UCSn/CSn will be defined in terms of UCSintensity/CSintensity. This is accomplished by use of a trace paradigm in which both CS and UCS are electrotactual and with the intensity defined in terms of voltage. Degree of conditioning. In order to properly con trol for the effects of sensitization (i.e., increasing the magnitude of the UCR to the CS as a function of UCS occurrence in the same contextual situation as the CS in either paired or unpaired temporal relation) a discrimina tion paradigm was employed. As noted earlier the discrimination procedure involves the use of two stimuli only one of which is paired with the UCS (in this study on a partial reinforcement basis). It is assumed that all non-conditioning forms of response modification will be reflected in responding to 85 the nonpaired stimulus and the paired stimulus. Only the latter, however, will reflect response modification result ing from stimulus pairing. Thus, the degree of conditioning will be inferred from a difference score derived from response magnitudes to differentially treated stimuli. The dependent variable, then, is discrimination, rather than response acquisition as in simple conditioning. Predictions Generated by Experimental Question Second interval responding. The first prediction generated by the experimental question and the attendent theoretical analysis outlined in Chapter II, is the fact of second interval responding. The theory predicts that a i response should be observed on nonreinforced test trials at a time when the UCS would have occurred and that such a response will occur more often and in greater degree to the stimulus with a reinforcement history than to the stimulus which has never been reinforced. A second prediction is that the magnitude of this "time-specific" responding should vary in some meaningful way with the dominance relation under which it is condi tioned. No precise predictions can be made concerning the 86 ordered relations of dominance and degree of conditioning in the present study because the dominance values prevail ing at the moment of contiguity are hypothetical. For example, it is difficult to know whether the dominance relation is greater in a group with CS and UCS of equal intensity but separated by 10 seconds than in a group in which the CS is one-third the intensity of the UCS but separated only by 5 seconds, or .5 second. The only pre diction, then, is that the relation between dominance conditions and degree of conditioning will be nonrandom. First interval responding. The theory requires that evidence of anticipatory conditioning be subsequent to evidence for second interval conditioning if it is assumed that the former is some type of "generalization" of the latter. (It is possible, of course, that there is no such relationship between these two types of respond ing, that the mechanism of each is quite different.) Therefore, it is predicted that second interval discrimina tion will occur prior to first interval discrimination. This prediction, however, is only a tentative one. The first point of test-control comparison at which discrimina tion conditioning can be measured occurs after two 87 differential reinforcements. There is already evidence (Lockhart and Grings, 1964) that at this point both first and second interval discrimination is significant, Assess ment of this prediction requires a somewhat different design (i.e. , test for discrimination after one differen tial reinforcement, or a situation where first response discrimination and second response discrimination develop more slowly, as might be the case with an animal). There fore, no precise prediction should be made concerning the relationship between first and second interval responding. In summary, then, it is predicted that (1) second interval responding will occur in trace conditioning; (2) that the magnitude of such responding should be non- randomly related to the degree of dominance; and (3) first interval discrimination should occur subsequent to second interval discrimination. The half-second CR. Since second interval respond ing does not occur when the CS-UCS interval is .5 second, the main prediction regarding the degree of conditioning under .5 second trace conditions is that there will be an increasing degree of conditioning as a function of increasing dominance. CHAPTER IV METHODS AND PROCEDURES Subjects Eighty-one volunteer students attending introduc tory psychology classes at either the University of Southern California or at Pasadena City College served as subjects (Ss). All were solicited by written invitation to partici pate in a "psychological experiment1 1 and all agreed to submit themselves to electric shock. Each S^ was paid $1.00 for his participation.^ Apparatus All stimuli employed were electrotactual and pro duced by two Grass (Model S-5) stimulators. One delivered 1The author gratefully acknowledges the financial support supplied by Professor William W. Grings through his grant (M3916) awarded by the National Institute of Health. 88 89 the conditioned stimuli and the second the unconditioned stimulus. All shocks were dc and administered to S through .75-inch diameter silver electrodes contained within a pair of lucite cups filled with a salt-based electrode paste and attached to the calf of S's legs by means of adjustable rubber straps. Two pairs of such electrodes were attached to S_'s dominant leg (one on the medial sur face, the other on the lateral surface), and one pair to the nondominant leg (lateral surface). The former place ments received the conditioned stimuli, the later the unconditioned stimulus. Stimulus durations and interstimulus intervals were controlled by Hunter electronic timers. All stimulus durations were .3 second. Table 1 presents information concerning CS intensity, UCS intensity, interstimulus trace interval, and approximate intertrial intervals for each of nine ..experimental groups,.« , Stimulus intensities were manipulated in terms of voltage output of the stimulator. (Preliminary work indi cated that under the conditions of the present electrode arrangement, voltage and not current was the critical parameter of the shock stimulus.) 90 TABLE 1 CS INTENSITY, UCS INTENSITY, CS-UCS INTERVAL AND INTERTRIAL INTERVAL FOR EACH OF NINE EXPERIMENTAL CONDITIONS (CS INTENSITY VALUE RELATIVE TO UCS INTENSITY VALUE) Group CS Intensity UCS Intensity CS-UCS Interval Intertrial Interval 1 3 3 0.5 20-40 2 2 3 5.0 30-50 3 2 3 10.0 40-60 4 2 3 0.5 20-40 5 2 3 5.0 30-50 6 2 3 10.0 40-60 7 1 3 0.5 20-40 8 1 3 5.0 30-50 9 1 3 10.0 40-60 91 The GSR was obtained as a dc resistance change through 1/2 in. by 5/8 in. silver electrodes bent to the contour of S's first and third fingers of the nondominant hand, coated with electrode paste, and taped firmly to the finger tips. The GSR was continuously recorded on a two- channel Dynograph amplifier-recorder (Offner Type 542) which received input from a Darrow-type bridge circuit. Design Independent variation. Two independent variables, each varied at three levels, produced a 3 x 3 factorial arrangement of nine experimental subgroups (refer to Table 1). One variable (dominance) involved the ratio of UCS to CS intensity in voltage units. The UCS/CS ratios employed were 3/3, 3/2 and 3/1 where "3" represents the UCS intensity (a maximal intensity shock). The second variable (contiguity) involved the temporal interval between CS and UCS onsets. The inter stimulus intervals employed were .8, 5.3 and 10.3 (with trace intervals of .5, 5.0 and 10.0). Nine Ss were assigned to each of the nine groups in the following manner. Each experimental condition was 92 assigned a number (1-9). This set of numbers was then permuted in random blocks of nine until 81 cases were obtained (Cochran and Cox, 1954). Each experimental con dition thus occurred in random order in each set of nine Ss run in consecutive order. The 81 cases were assigned consecutive subject numbers (1-81). The experimental session consisted of six phases: (1) shock work-up trials; (2) familiarization trials; (3) acquisition trials; (4) disparity; (5) extinction; (6) disinhibition. Because the latter three phases con cerned peripheral and secondary considerations, further description is not essential for the primary experimental aims. Appendix A describes these phases and contains the results obtained during these periods. Preceding these six-phases, S was given a general introductory statement (see Appendix B). General instructions. Following the greeting and — } seating of S, and attachment of stimulus and response electrodes, S was given a set of typewritten instructions in two parts. The first (see Appendix B) was a detailed description of the GSR and a warning to S to keep his overt movements to an absolute minimum. The second part (see 93 Appendix C) described S's task in setting the level of shock intensity. Shock workup procedure. The shock workup procedure was designed to determine S's minimal and maximal threshold to electrotactual stimulation. This was accomplished by presenting subliminal shock voltages followed by a step wise increase (in 2 volt steps) until S reported feeling "something." The shock was then increased in larger steps (5 volts) until reported that the shock was "painful." The maximal shock intensity for each ! 3 was defined as 5 volts in excess of the voltage reported by S^ as painful. This value was used as the UCS and stimulus ratios were defined on the basis of this value. For example, if S_ reported that 55 v. was "painful" the UCS intensity was set at 60 v. If S^ were in the 3/3 group, the CS intensity would also be 60 v.; if in the 3/2 group CS intensity would be 40 v.; if in the 3/1 group CS intensity would be 20 v. This workup procedure was repeated at each electrode site in counterbalanced order across all £3s. CS and UCS intensi ties for each S^ by experimental group may be found in Appendix D. Acquisition. A total of 40 acquisition trials was 94 presented to each S. Of these, half consisted of a par tially (60 per cent) reinforced conditioned stimulus (CS+) and half of a never reinforced conditioned stimulus (CS-). Of the 20 CS+ trials, 12 were reinforced by the UCS. The 8 nonreinforced presentations of CS+ were desig nated as test trials. These occurred on ordinal trial numbers 6, 9, 15, 20, 26, 29, 35 and 39. Alternately adjacent presentations of CS- were designated as control trials. These occurred on ordinal trial numbers 5, 10, 14, 19, 27, 30, 34 and 40. In each block of 10 trials there were 3 reinforced presentations of CS+, 2 nonreinforced presentations of CS+, and 5 nonreinforced presentations of CS-. Dependent variation. It is necessary at this point to explicitly note an unusual feature of the paradigm employed in the present study. It is usually considered a desirable feature of a conditioning situation for the CS to be a "neutral" stimulus. In its strictest form, such a requirement demands that the CS, prior to CS-UCS pairing, elicit no response. A somewhat less restrictive require ment (and a more usual one) is that the CS prior to pairing must not elicit the UCR or any component thereof. Thus, a 95 t o n e m a y e l i c i t w h a t i s d e s c r i b e d a s a n o r i e n t a t i o n r e a c t i o n ( l o o k i n g a t t h e s o u r c e o f t h e t o n e , c h a n g e s i n E E G , G S R , d i g i t a l a n d t e m p o r a l b l o o d v o l u m e , e t c . ) , b u t m u s t n o t e l i c i t s a l i v a t i o n , i f t h e t o n e i s t o b e s u b s e q u e n t l y p a i r e d w i t h m e a t p o w d e r . W h e n c o n d i t i o n i n g c e r t a i n r e s p o n s e s , h o w e v e r , s u c h a c r i t e r i o n c a n n o t b e s o e a s i l y m e t . T h e G S R i s a p r i m e e x a m p l e o f s u c h a r e s p o n s e . T h e G S R i s s e n s i t i v e t o a l m o s t a n y s t i m u l a t i o n o r c h a n g e s i n s t i m u l a t i o n . T h e r e f o r e , i f o n e w i s h e s t o c o n d i t i o n t h e G S R s p e c i a l a t t e m p t s m u s t b e m a d e t o d i s t i n g u i s h b e t w e e n a n o r i e n t i n g G S R a n d a c o n d i t i o n e d G S R . T h i s b e c o m e s a m a j o r p r o b l e m i f i t i s a s s u m e d t h a t o r i e n t i n g r e s p o n s e s t h e m s e l v e s m a y b e c o m e c o n d i t i o n e d r e s p o n s e s . S o m e w r i t e r s h a v e a t t e m p t e d t o s e p a r a t e O R s f r o m C R s o n t h e b a s i s o f t h e l a t e n c y o f G S R r e a c t i o n s ( S t e w a r t , S t e r n , W i n o k u r a n d F r e d m a n , 1 9 6 1 * L e o n a r d a n d W i n o k u r , 1 9 6 3 ; M c D o n a l d a n d J o h n s o n , 1 9 6 5 ) . U s i n g l o n g d e l a y c o n d i t i o n i n g s i t u a t i o n s ( u s u a l l y l o n g e r t h a n 5 s e c . ) t h e s e i n v e s t i g a t o r s h a v e f o u n d t h a t f o l l o w i n g t h e i n i t i a l r e s p o n s e t o t h e C S t h e r e o c c u r s a s e c o n d r e s p o n s e s o m e w h a t p r i o r t o o r j u s t o v e r l a p p i n g t h e p o i n t o f u s u a l U C S o c c u r r e n c e . T h e f o r m e r r e s p o n s e h a s b e e n t e r m e d a n 96 orienting reaction and the latter a true anticipatory response. They have found that the frequency of the anticipatory response increases more from adaptation to acquisition than does the orienting response, and that more anticipatory responses are given to a reinforced than a nonreinforced cue while this was not true of the orient ing response. They conclude that the first response is "only" an orienting reaction and that the second response is the "true" conditioned response. Lockhart and Grings (196 3) have argued that such a conclusion is not warranted. Frequency of responding is only one possible dependent variable and may not be the most adequate one to reflect modification of a sensitive response system like GSR. For example, if one considers the magnitude rather than the frequency of the so-called orienting response both simple and discrimination condi tioning has been demonstrated (e.g., Grings, Lockhart and Dameron, 1962). For this reason (among others), some investigators prefer the discrimination paradigm and the magnitude of GSR as the dependent variable. It has been argued (Grings, Lockhar-t and Dameron, 1962; Wickens and Harding, 1965) that such an arrangement is considerably more sensitive to the 97 effects of conditioning than other arrangements. This reasoning has particular relevance for the present study. The use of electric shock as both CS and UCS carries with it special problems. First, both stimuli are "adequate" in the sense of eliciting the UCR. In those groups in which the CS and UCS are the same intensity it is obvious that the UCR is elicited "twice." When the CS is of lesser intensity, the UCR is simply elicited to a lesser degree. When the two stimuli are close together there is the possibility of some type of response interaction. At first glance, then, such a paradigm would seem to violate one of the first requirements of a conditioning paradigm, i.e., the neutrality of the CS. Such a requirement, however, is a relative one. It has been argued earlier that the critical point is the intensive relation between the CS and UCS at a particular point, or points, in time. If this notion is valid then it matters little whether the two stimuli involved elicit the same response or not. Conditioning occurs if the second stimulus is "stronger" than the first at the critical point of contiguity. However, and this is the crucial point in this discussion, demonstration of such conditioning requires special attention in terms of the definition of the 98 dependent variable from which such conditioning is to be inferred. The dependent variable in the present study is the magnitude of discriminatory responding. It should be care fully noted that the CS+ and the CS- are the same physical intensity. It is assumed that whatever differences develop in the responding to these two stimuli must be a result of the differential treatment of the two stimuli. AH' non conditioning modification is assumed to be equal in degree to both stimuli. This would apply to operant GSR level, sensitized spontaneous responses, and "pure" (i.e., non conditioned) orienting reactions, and pure stimulus inten sity effects. Thus, if the responding to CS-is "subtracted" from the responding to CS+, these effects do not enter into the value of the difference obtained. It is presumed that such a difference represents the magnitude of conditioning. It is for this reason that the dependent variable in the present study is based upon the difference in magnitude of responding to CS+ and CS-. Responses measured. Because of the nature of the experimental variation involving CS-UCS interval, a number of responses were usually obtained during a single trial 99 (in the longer interval groups). This necessitated develop ment of criteria concerning the length of time on a particu lar trial during which responses would be measured. This period of time was different for each of the three CS-UCS interval conditions. For the .5 second trace group any change in resistance (exceeding 100 ohms) occurring within 5.3 second of CS onset was counted as a response (excluding such changes if they occurred prior to 1.0 second following stimulus onset). Invariably, only one response was found to occur within these limits on either test or control trials. For the 5 second trace group, any change in resistance (in excess of 100 ohms) occurring between 1.0 and 10.3 second following CS onset was counted as a response. Not infrequently, two or more responses were found to occur during this period. For the 10 second trace group, any change in resistance (in excess of 100 ohms) occurring between 1.0 and 15.3 second following CS onset was counted as a response. Again, more often than not, more than a single response was observed within this interval. Response measurement. The magnitude of response 100 was obtained in the following manner:. (1) the base resis tance of the first response during the measurement interval was determined; (2) the change in resistance of this first response and all responses during the measurement interval were measured in reference to (i.e., as changes from) this base level. This applied to all responses during a single trial except in those very few instances when succeeding responses during the interval occurred as resistance changes from a base level lower than that characterizing the first response. In these cases, the response was measured in terms of its own basal level at response onset. These points should be made clear by referring to Figure 13. Changes in resistance were transformed to square- root-of-conductance-units. This was accomplished in the following manner: (1) reciprocals of the base resistance and the resistance level at the point of maximal change were determined and each value is multiplied by 10^; (2) the difference between these two values was obtained; (3) the square root of this difference was then computed. In addition to magnitude values, both latency of and recruitment values of all responses were determined. Latencies were measured in terms of the number of seconds 101 — ►Magnitude of response GSR Reference level Reference level CS n. UCS Figure 13*— Measurement of response magnitudes. 102 (to nearest tenth) from CS onset to response onset. Recruitment values were obtained by measuring the elapsed time in seconds (to the nearest tenth) from response onset to response maxima. Procedure Ss were brought into a sound-proof experimental chamber and seated at a table directly in front of a parti tion separating from all stimulus production and response recording equipment. All watches and other forms of jewelry were removed from S^. All critical skin surfaces were then prepared for electrode placement. This consisted of: (1) scrubbing the skin with acetone followed by (2) swabbing with alcohol. When the skin area was dry, the electrodes (filled with electrode paste) were applied firmly. Ss were given the typewritten instructions. While read the instructions, E calibrated the equipment and took preliminary resistance measurements of all electrode sites. Following these preliminary steps, S was again told the procedure to be followed in setting the intensity of the shock. Following the completion of the shock intensity settings, S was given a third set of instructions (see Appendix E) which told S to relax and that while the light was out not to move. All light during the experimental session subsequent to these instructions emanated from a 100 W bulb in E's chamber. During the entire session, an electric fan was in operation about 3 feet to the left and 3 feet above S and served as a masking device. In addition, a tape recording of white noise played during the entire experimental session from E's chamber. The fan and noise effectively removed the possibility of £ hearing any of the equipment sounds associated with stimulus occurrence. Following the instructions to "relax" the trials described in the design section were administered to S. Following the completion of the experimental trials electrodes were removed from the S. During this time, each S _ was asked whether or not he was able to "know when any of the shocks were coming?" Depending upon his answer he was roughly classified as being "aware" or "unaware." S was paid and warned not to discuss the details of the experimental session with anyone. CHAPTER V RESULTS The Half-Second Trace Conditioned Response Response measured. As noted previously, only one response to a CS is obtained when the CS-UCS interval is less than the latency of the GSR. Therefore, it is meaning ful to speak of the half-second trace CR as a singular phenomenon. It should be kept in mind that the CR in the present study is considered a discriminatory response as measured by the difference in magnitude of responding to CS+ and CS-. In addition, all values are in terms of "adjusted" magnitudes. The adjustment procedure consisted of the following steps: (1) computation of the average magnitude of response to CS+ and CS- (separately for two presentations of each stimulus) prior to acquisition (i.e., during the "familiarization" phase); (2) subtracting these values from test (in the case of CS+) and control (in the case of CS-) response magnitudes. The discrimination score 104 105 is then computed as a difference between adjusted magni tudes of response. The effect of dominance. Figure 14 presents the mean magnitude of discrimination for each dominance condi tion. Nonparametric‘ d analysis showed that significant discrimination occurred only in the 3/1 dominance condition (Wilcoxon T = 0; N = 9; p .01). A one-way analysis of variance (Kruskal-Wallis) indicated that the difference between discrimination means was quite significant (H = 13.41; df = 2; p .01). The above discrimination values were based on the entire 8 test and control trial differences during acquisi tion. A previous study (Lockhart ahd Grings, 1964) found that significant discrimination may be obtained' quite early in the acquisition phase. Likewise, in the present study, significant discrimination was obtained by the first point of test-control comparison, a point following two differen tial reinforcements, but only in the 3/1 dominance group (mean discrimination = .69; T = 3; N = 9; p = .02). Mean ■^•Preliminary analysis of the data indicated that inhomogeniety of variance between groups would preclude parametric analysis. For this reason all analyses were by nonparametric methods. Mean Magnitude o f Discrimination 106 •8 .7 .6 .5 .4- .3 .2 .1 0 - .1 - .2 3/3 3/2 3/1 Dominance Ratio Fig. 14.— Mean magnitude of discrimination as a function of dominance ratio with *5 sec. trace interval 107 values of discrimination in the 3/3 and 3/2 groups (.01 and .20 respectively) were insignificant. Mean test and control responses (and discrimina tion scores) for overall acquisition trials may be found in Appendix F. Problems of Multiple Responding in Long Interval Conditioning Previous reports (Grings, Lockhart and Dameron, 1962; Lockhart and Grings, 1964) indicate that a single "function" relating the degree of discrimination condition ing to the interstimulus interval probably does not exist for the simple reason that when the time between the CS and the UCS is extended beyond the latency of the response more and varied kinds of responding becomes possible. When the interstimulus interval is on the order of a half-second, the CS-UCS combination terminates prior to response occurrence and as previously noted, only one response is produced. When the interval is longer (e.g., 5 sec.) more than one response is usually found. There is typically a response prior to the UCS (the so-called "anticipatory" response) as well as the response to the 108 UCS (i.e. , the UCR) . When the interval is extended even further (e.g., 10 sec.) the possibility of observing several responses develops. Therefore, when long inter vals are employed it is necessary to develop a logic for multiple response measurement. A logic for multiple response measurement. For .5 second interstimulus interval condition only one response occurs. On test and control trials it is this response which is measured. In the longer interval condi tions a number of possibilities present themselves. Consider Figure 15, a sample response pattern obtained on a test trial when the UCS is not presented. The latency of each response is given in parentheses. (This is an actual response record: test trial number 3, subject number 13, 10 second, 3/2 dominance condition.) The primary question is whether or not each of these responses is in some way or another a CR (or a component of a CR). Criteria are needed for distinguish ing these responses from responses under the influence of nonconditioning variables such as sensitization and spon taneous responding. As indicated previously, and again here for 109 GSR CS UCS \ 'No UCS presented Fig. 15,---Sample GSR recording during an acquisi tion test trial on CS+ when the UCS is not presented. (13.1) (1. 8) (6.5) (10.5) Reference Level SI 110 emphasis, the method employed in the present study for distinguishing between conditioned and nonconditioned components of the GSR involves the use of a "control" stimulus. Responses to this never-reinforced cue (CS-) are interpreted as representing the sum total of all non conditioning types of response modification operating within the context of the conditioning situation. Response to the "test" stimulus (i.e., a nonreinforced presentation of CS+) in excess of responding to the control stimulus is assumed to reflect the effect of the pairing procedure, i.e., conditioning.^ In attempting to develop a procedure for analyzing multiple responding, this test-control difference procedure has played a central role. The basic idea is simple: if any response (or responses) to a test stimulus, regardless of where it may occur in the temporal duration of a trial, can be shown to be consistently more frequent or consis tently greater in magnitude than "similar" responding to 2 It is recognized that responding to CS- may very well contain components of conditioned responding which "generalize" from CS+ to CS-. If this is the case, then the difference in responding to these two stimuli is an underestimate of the amount of conditioning to CS+. Ill the control stimulus, then there are grounds for inferring that such a response (or response magnitude) is a CR (or component of a CR). In this study, each response on every test and control trial during acquisition was measured for both the 5 and 10 second trace groups. For the 5 second group responses were measured in an interval extending from 1.0 to 10.3 second following CS onset. For the 10 second group, responses were measured in the interval extending from 1.0 to 15.3 second following CS onset. All responses beginning within this period were measured in terms of magnitude of deflection. Thus, if a response in the 10 second trace condition began at 14.8 second following CS onset, its magnitude was determined at the point of maxi mum deflection even though this point may have occurred outside the indicated latency boundries. The first step in the assessment of conditioning under these long trace conditions involved the definition of maximal responding. The maximal response (Max R) was defined simply as the sum of all individual response magnitudes occurring within the intervals noted above. Presumably, if trace conditioning occurs at these longer intervals, the Max R to CS+ should exceed Max R to CS-. 112 Components of the Max R CR. The response measur ing interval was divided into what appeared to be meaning ful subintervals. For the 5 second group four such sub intervals were defined. (Five subintervals were employed in the 10 second group and will be considered when the results for this group are presented.) The first such subinterval, termed Period I, was an interval extending from 1.0 to 2.8 second following CS onset. It is in this interval that the response to the electrotactual CS occurred. This interval was considered meaningful because of the question of whether or not trace conditioning could be reflected in a response which is quite temporally removed from the point of "reinforce ment" and at the same time itself an unconditioned response. Since the response in Period I is "stimulus bound" there can be no question of this response "moving towards" the point of reinforcement such as would be required for the demonstration of an inhibition of delay phenomenon. Anticipatory responding, as well as inhibi tion of delay, might well show itself in a quite different form, i.e., in terms of a response (in addition to that elicited by the CS) which occurs prior to the UCS. 113 Therefore, a second period was defined. Period II encom passed the interval 2.8 to 5.3 second following CS onset. (It might be mentioned at this point, that it is during this interval that Stewart et al. (1961) imply that "true" GSR conditioning should be found.) Previous demonstrations of the so-called "second interval" response led to the definition of a Period III response. This was any response occurring within the interval 5.3 to 7.8 second following CS onset. In previous studies demonstrating second interval responding the response measurement interval was 5 second following the point where the UCS would have occurred.' A 2.5 second measuring interval was employed here for two reasons: (1) the measurement of the second interval response might be more refined if it is true that the response represents a temporally correlated process with the point in time of UCS occurrence, and (2) the 5 second period employed previously could still be used here since responses occurr ing in the interval 7.8 to 10.3 second were also measured. These were termed Period IV responses. To obtain the degree of second interval responding as previously defined involves only the simple addition of discrimination values obtained for Periods III and IV. 114 Combinations of response intervals. In addition to the subintervals described above, other response definitions involved various combinations of the sub intervals as follows: 1. First Interval Response (FIR): defined as the sum of all individual responses occurr ing during the CS period prior to the point of usual UCS onset, or, alternatively, the sum of Period I and Period II. 2. Time Specific Response (TSR): defined as the sum of all individual responses occurring during Period II and Period III, i.e., a 5 second period "surrounding" the point of usual UCS onset. 3. Second Interval Response (SIR): defined as the sum of responses occurring during Period III and Period IV. The use of sums of response magnitudes in the definitions of these various responses rather than simply the "largest" response (or some other selective criteria which would ignore some responses) was thought to make the most adequate use of the available data. And, once again, the important matter is the difference between test and control responding during these subintervals or combinations of subintervals. 115 Five-Second Trace Conditioned Responses Table 2 presents the mean magnitude of discrimina tion (based on all test-control differences during acquisi tion) by dominance condition for each type of response definition described above. In addition, the table con tains the average value of discrimination for each response definition across dominance conditions. Within each dominance condition no significant degree of discrimination was obtained for any response. In general, all discrimination means were quite small and only a very few even approached significance. However, when the dominance groups were combined, significant dis crimination was obtained in terms of both TSR and SIR. Further analysis of discrimination and dominance. Since little or no significant discrimination was obtained within the individual dominance conditions there is little meaning to a comparison between dominance groups. How ever, a question can be raised as to why discrimination was not obtained since previous studies have shown quite rapid and significant conditioning employing a 5 second CS-UCS interval. Table 3 presents the mean discrimination values by 116 TABLE 2 DISCRIMINATION VALUE BY DOMINANCE CONDITION FOR EACH RESPONSE DEFINITION Dominance Condition Type of CR 3/3 3/2 3/1 Average Max R .64 .25 .19 .36 FIR .34 .02 -.06 .10 TSR .38 .10 .08 .19* SIR .30 .22 .24 .2 0* Period I .13 -.0 2 - . 0 2 .03 Period II .22 .04 -.03 .07 Period III .17 .06 .11 .11 Period IV .13 .17 .13 .14 *Sign Test: x = 5, N = 25; p = .002 117 TABLE 3 DISCRIMINATION VALUES FOR EACH s during EACH RESPONSE PERIOD BY DOMINANCE CONDITION FOR 5 SECOND TRACE GROUP Dominance Response Period Ratio 3/3 3/2 3/1 I II III IV .87 - . 1 1 .47 .27 .12 - . 2 1 -.08 .14 .22 .19 . 3.2 .11 -.38 .03 .2 0 - .17 .40 .00 .00 .00 -.15 1.80 -.08 .03 -.15 -.67 .50 .01 .23 .25 . 82 .63 -.02 .70 -.62 -.23 .37 .24 - . 0 2 .09 -.90 .01 -.08 .00 .30 -.18 .31 .00 .65 .29 . 35 .66 -.49 -.41 -.31 -.01 .66 -.15 . 30 -.15 -.31 .61 -.55 .66 .18 -.30 .57 -.30 -.64 .29 -.06 .56 -1.47 - . 0 1 .01 - . 0 1 .06 .00 .00 .00 .53 .24 .09 .65 -.05 -.41 .61 .08 .47 .04 .31 .07 -1.12 .14 .34 -.09 .35 .00 .01 .29 .45 .00 .05 .17 .57 -.28 -.41 .03 118 response period for each S in all dominance conditions (similar data for both test and control responses may be found in Appendix G). Although there are a priori grounds for the occurrence of zero discrimination and positive discrimination little ground exists for expecting negative discrimination. Yet, fully one-third of the dis crimination values in Table 2 are negative, and two-thirds of the subjects show one or more negative discrimination scores. Why these negative differences occur is not clear, although negative discrimination has been reported previously (Hilgard, Jones and Kaplan, 1951). The presence of negative discrimination values creates a genuinely difficult situation to analyze. If negative discrimination is an individual difference vari able, then the failure to find significant discrimination conditioning in the 5 second group may simply be a result of combining together Ss who are discriminating in opposite directions yielding a group average approximat ing zero. Suppose, for example, we wished to compare the performance of a group conditioned under a 3/1 dominance condition and a .5 second CS-UCS trace interval with a comparable dominance group conditioned with a 5 second 119 CS-UCS trace interval. In the former group, no discrimina tion value was negative. In the latter group, we are faced not only with negative discrimination scores, but with the additional problem of what response or combina tion of responses should enter into such a comparison. Suppose that in order to avoid the first difficulty we compare only the top five discrimination values in each group; and, to "avoid" the second problem, suppose we choose as a basis for comparison the Max R CR in the 5 second group. The mean discrimination of the best five discriminators in the .5 second group was 1.02 and the mean Max R discrimination of the best five discriminators in the 5 second group was .79. While the former mean is larger it is not statistically so (U = 10; 2 = 5; p = . 345) . To further complicate matters, suppose that we select as the basis of comparison the largest degree of (positive) discrimination shown by each subject in the 5 second group regardless of what temporal interval such discrimination occurred. When this is done, quite signi ficant discrimination is obtained for each dominance condition in the 5 second trace group. This would lead to the conclusion that all Ss in the 5 second group show at 120 least some degree of discrimination at some point within the temporal interval of the trial. Careful inspection of Table 2, page 116, reveals that one-third of the sample discriminated best during Period I, about one-tenth during Period II, one-third during Period III and about one-fifth during Period IV. (To confuse things even further, it would appear that significant negative discrimination also occurs in each group if the period in which the lowest degree of discrimination is disregarded.) An additional factor is that negative discrimina tion in one period may be followed in the very next period by quite positive discrimination. As an example of this type of possible "dynamic" relationship, consider Figure 16. The question involved here is the degree to which Period I discrimination is related to discrimination per formance in subsequent periods. The figure presents the mean magnitude of discrimination by response period for upper, middle and lower thirds of discrimination values (based on Period I discrimination) for each dominance condition. Ss showing relatively large degree of Period I discrimination show little or no discrimination at later points (except in the 3/3 dominance group). Those failing to discriminate in Period I (the middle third group) do Mean Magnitude of Discrimination .6 .5 .4 .3 .2 .1 0 -.1 ■.2 -.3 -.4 -.5 -.6 •.7 - . 8 Upper Third Middle Third lower Third J 1 -----1 _____i — I II III IV 3/3 o-yo / II III IV 3/2 » / / 0 II III IV 3/1 Pig. lb•— Degree of GSR discrimination as a function of dominance by response period (roman numerals) for upper, middle, and lower third subgroups of _Ss divided on the basis of Period I discrimination for the 5 sec."trace interval. fO 12 2 not discriminate at any other point. Ss showing negative discrimination in Period I do not tend to show large dis crimination values at subsequent points. However, the change in discrimination from Period I to subsequent periods is quite large. Thus, between Period I and Period II, discrimination increases by mean values of .6 6 , .58 and .76 for the 3/3, 3/2 and 3/1 dominance conditions respectively. Thus, for Ss discriminating negatively during Period I, subsequent discrimination behavior is markedly different (in terms of change) than for Ss discriminating positively during Period I. Since this is true across dominance conditions, it would appear that an individual difference variable may be involved. In addition, it must be concluded that one cannot predict with any accuracy during which period S^ will exhibit maximal discrimination performance between a reinforced and nonreinforced stimulus. This situation seems to point to a complicated interaction of dominance condition and an individual difference variable when the interval between the CS and UCS is relatively long. 123 Ten-Second Trace Conditioned Responses Table 4 presents the mean magnitude of discrimina tion (based on all test-control differences during acquisi tion) by dominance condition for each type of response definition described below: 1. Period I: responses occurring between 1.0 and 2.8 second following CS onset. 2. Period II: responses occurring between 2.8 and 7.8 second following CS onset. 3. Period III: responses occurring between 7.8 and 10.3 second following CS onset. 4. Period IV: responses occurring between 10.3 and 12.8 second following CS onset. 5. Period V: responses occurring between 12.8 and 15.3 second following CS onset. 6 . Max R: sum of all responses occurring within each of the above periods. 7. FIR: sum of all responses occurring in Periods I, II and III. 8 . TSR: sum of all responses occuring in Periods III and IV. 9. SIR: sum of all responses occurring in Periods IV and V. In addition, the table contains the average value of dis crimination for each response definition across dominance conditions. Quite in contrast to the 5 second trace data, 124 TABLE 4 DISCRIMINATION VALUE BY DOMINANCE CONDITION FOR EACH RESPONSE DEFINITION Type of CR Dominance 3/3 Condition 3/2 3/1 Average Max R 1.14** 1.48* .95* 1.19*** FIR . 89** 1 .0 2 * .58* .83*** TSR .58** .41* 4 7** .48*** SIR .25 .46* .37* .36*** Period I .13 .65* .53** e 4 4 *** Period II .55* .21 -.18 .19 Period III .20 .16 .23 .20 Period IV . 38** .25* .24* # 29*** Period V - . 1 2 .21 .13 .07 P .05 (Wilcoxon T) P .01 (Wilcoxon T) P .001 (Sign Test x) significant 10 second trace conditioning was obtained in each dominance condition for a number of response defini tions. Across dominance conditions, all combinatory responses showed very significant conditioning; both Period I and Period IV discrimination was very significant. Max R CR. Max R discrimination was significant in each dominance condition (Wilcoxon T = 1, 4, and 4 for the 3/3, 3/2, and 3/1 dominance conditions respectively with an N of 9 in each case). Although the relation between degree of discrimination and dominance is an inverted U function, no significant difference was found between the three groups. Analysis of discrimination at the first point during the acquisition periods showed that only in the 3/3 dominance condition was discrimination significant at this point (mean discrimination = 1.41, Wilcoxon T = 3; N = 9; p = .02). Mean discrimination values for the 3/2 and 3/1 dominance conditions at this point were .41 and .22 respectively. Figure 17 shows the trial by trial discrimination scores for each dominance condition. The main result here seems to be that discrimination occurred first in the 3/3 dominance condition, while discrimination in the 3/2 and Mean Magnitude of GSR Discrimination 126 3/3 3/2 3/1 2.4 2.2 2.0 1.8 1.6 1.2 1 2 3 Fairs of Acquisition Trials Fig. 17.— Acquisition of GSR discrimination for each dominance condition in the 10 sec. trace interval group in terms of MaxR. 127 3/1 dominance conditions occurred more slowly but reached a generally higher level of discrimination. FIR CR. The mean magnitude of FIR discrimination was significant for each dominance condition (T = 2, 4 and 6 ; N = 9). As with Max R discrimination, the dominance condition did not produce a statistically significant dif ference between groups. TSR CR. TSR discrimination was significant in each dominance condition (T=0, 6, 1; N=9). Again, there was no reliable difference between dominance condi tions . SIR CR. Significant SIR discrimination on an overall basis was obtained in the 3/2 and 3/1 dominance conditions (T = 3 and 4 respectively; N = 9). In this case, the difference between groups is highly significant (H = 15.54; p .001). Period I CR. Significant Period I discrimination occurred in the 3/2 and 3/1 dominance conditions (T = 3 and 2 respectively; N = 9; p .02 and p .01) but not in the 3/3 dominance condition. Difference in discrimination between groups was significant (H = 7.65; p .05). 128 Period I discrimination was also significant at the first point of test-control comparison for the 3/2 (mean dis crimination = .48) and the 3/1 (mean discrimination = .40) groups but not for the 3/3 dominance condition (mean discrimination = .1 1). Period II CR. In contrast to Period I discrimina tion, Period II discrimination was significant only in the 3/3 dominance condition (T = 5, N = 9; p .05). However, the difference between dominance groups failed to reach significance (H - 4.34; p .20). Period III CR. Period III was the 2.5 second period just prior to point of usual UCS occurrence. How ever, no discrimination was shown during this interval for any dominance condition, or with dominance conditions com bined . Period IV CR. Period IV encompasses the 2.5 second interval just following the point of usual UCS occur rence. It is during this interval that the phenomenon r ‘ of "second interval responding" was predicted. Although the absolute values of discrimination occurring during this interval are relatively small, significant Period IV 129 discrimination was obtained in each dominance condition (T = 0, 5 and 6;N=9;p .01; p .05; p=.05 for the 3/3, 3/2 and 3/1 dominance conditions respectively). The difference between dominance conditions was not significant. Period IV discrimination was not significant in any domin ance condition at the first point of test. Period V CR. No significant Period V discrimina tion occurred in any dominance group. Further analysis of discrimination and dominance. Table 5 presents the mean discrimination values by response period for each S in all dominance conditions (similar data for test and control responses separately may be found in Appendix H). As with the 5 seconds S s, negative discrimination values seem to be scattered unsys tematically throughout response periods and dominance conditions. While only 25 per cent of the total number of discrimination values were negative, 80 per cent of the Ss in the entire sample showed at least one negative discrimination value in one of the five response periods. Thus, again, negative discrimination must be considered. However, negative discrimination in the 10 second trace 130 TABLE 5 DISCRIMINATION VALUES FOR EACH S DURING EACH RESPONSE PERIOD BY DOMINANCE CONDITION FOR 10 SECOND TRACE GROUP Dominance Condition Response Period I II III IV V 3/3 . 30 - . 0 2 .00 .45 -.38 .71 1.08 -.83 .86 -.14 .24 1.53 .38 .25 -.55 -.31 .67 .68 .86 .22 -.28 .04 .70 .12 -.42 . 30 .97 . 82 .69 .16 -.35 .00 .00 .00 .00 .27 - . 0 1 .25 .00 .00 . 31 .72 ~ - 17 .17 .00 ■ 3/2 1 . 0 0 . 10 .15 .39 .24 2 .09 .35 .00 .75 .37 .42 .48 .01 -.03 . 30 - . 6 8 -.07 -.03 -.07 -.07 .71 -.24 .04 .29 .16 . 32 .88 1.45 .47 1 . 0 1 .97 .00 .00 .06 - . 2 0 .68 .90 -.06 . 35 .00 . 31 -.47 -.09 .05 .10 3/1 .95 .14 .00 .40 .00 -.29 .21 - . 2 2 .80 .13 1 . 0 2 -.63 .41 .44 .76 .44 -.27 1 . 1 2 .17 .26 1 . 1 2 -.26 .24 .28 .12 . 58 - . 1 2 .53 -.31 .42 .13 .81 -.06 .21 -.08 . 45 -1.24 - . 1 1 .02 .01 . 39 .04 .12 .16 -.48 131 group did not preclude the observation of significant posi tive discrimination as was the case in the 5 second trace group. If, in each dominance condition, the sample is divided into upper, middle and lower thirds based on the degree of Period I discrimination, the curves in Figure 18 are obtained. These curves indicate that within homogene ous groups of Period I discriminators (i.e., negative discriminators, nondiscriminators, and positive discrimina tors) , the dominance variable functions in a relatively consistent manner. There is an increase in discrimination from the 3/3 to the 3/2 dominance condition followed by a slight decrease to the 3/1 dominance condition. The same data in terms of Period IV responding (see Figure 19) indicates that again the dominance variable functions con sistently between homogeneous discriminators. (However, in the curve for the upper group, the difference between dominance conditions is significant: H = 6.27; p .03.) In comparing Figures 18 and 19 it should be noticed that the 3/3 dominance group is characterized by the highest level of Period IV discrimination and the lowest level of Period I discrimination. 132 a o •H 4* c d 0 ■H * •H © « w cs Vi o « T3 0 +» • H 0 M 0 £ § £ 1.4 1.2 1.0 .8 .6 .4 .2 0 -.2 -Upper Third _-Middle Third Q Lower Third / - o 0 I I 3/3 3/2 Dominance Ratio 3/1 Fig. 18 .— Period I discrimination as a function of dominance for upper, middle and lower third subgroups of Ss divided on basis of degree of Period I discrimination for the 10 sec. trace interval. 133 o •H -P £ o TO CO CS <H O TO T3 TO +> TO s § TO £ .8 .7 .6 .5 .4- . .3 .2 .1 0 Upper Third Middle Third Lower Third o o 0 — — o I I I 3/3 3/2 3/1 Dominance Ratio Fig. 19.--Period IV discrimination as a function of dominance for upper, middle and lower third subgroups of Ss divided on basis of degree of Period IV discrimination 7or the 10 sec. trace interval. 134 Figure 20 shows the degree of discrimination as a function of dominance condition irrespective of response interval. In this case, the degree of discrimination is quite significant (all values positive) but very nearly equal in terms of the dominance variable. Likewise, when the degree of negative discrimination was measured for each S regardless of response interval, the degree of negative discrimination was significant for each dominance condition. Figure 21 compares interperiod discrimination values for upper, middle and lower subgroups for each dominance condition (with division based on degree of Period I discrimination). In each dominance condition the curve for the lower level discriminators is quite different from that of the upper level discriminators. This is best seen in the 3/1 dominance condition where the two curves are almost reciprocal (except during Period IV). As with the 5 second Ss, the manner in which S discriminates in Period I seems to be uniquely related to the manner of discrimination in subsequent response periods. This indi cates, as before, some type of dynamic relationship pos sibly relating to an individual difference variable. 135 A O •H +> a s U o 0 9 o € ) tJ 3 -P •H £ t a f i a s S § « ) £ 1.0 .8 .6 .4- .2 0 -.2 -.4- .Positive Discrimination (All £3s) Negative Discrimination (All Ss) 0 1 3/3 3/2 3/1 Dominance Ratio Fig. 20 .— Discrimination as a function of dominance for maximal discrimination value (positive and negative) regardless of response interval in which the maximal discrimination value occurred. 136 ft o •H +> as ft L i O m ■ H O r r J 3 +3 •H g C ) S g « £ 1.2 1.0 8 6 4 2 0 2 Til I I II III w 3/3 Upper Third Middle Third Lower Third & / / * • i I I 0 1 I I IIIU IV V 3/2 Dominance Ratio I II HIIV V 3/1 Fig. 21.— Degree of discrimination as a function of dominance by response period (roman numerals) for upper, middle, and lower third subgroups of Ss divided on basis of Period I discrimination for the 10*”sec. trace interval. 137 Discrimination as a Function of Trace Interval Figure 22 compares the degree of Max R discrimina tion for each dominance condition as a function of the temporal separation of the CS and UCS. The comparison of degree of discrimination as a function of trace interval is a complicated one, since in the .5 second trace groups only one response is obtained, while in the longer trace intervals multiple responding occurs. Figure 22 provides useful information if our question concerns the total degree of discrimination (i.e., difference between test and control responses). It has been argued that whatever dif ferences occur between test and control responses must be due to the pairing operation and thus reflect the total degree of discriminatory responding. For this reason, there is probably more justification for comparing the Max R discrimination value across time variation than any other type of response definition. It can be seen that when both the CS and UCS are of equal intensity, the degree of discrimination increases as the trace interval increases. The difference between these discrimination values is highly significant (H = 10.69; p .01). When the CS is two-thirds as intense as Magnitude of GSR Discrimination 138 3/3 3/2 3/1 1.8 1.6 1.4 1.2 1.0 8 0 — X .5 5.0 10.0 Trace Interval (Seconds) Pig. 22.— MaxR discrimination for each dominance condition at each trace interval. 139 the UCS, the degree of discrimination is equal for the .5 and 5 second trace groups but significantly larger when the trace is extended to 10 second (H = 6.76; p .05). When the CS is one-third the intensity of the CS, condi tioning is obtained at the .5 second trace interval, decreases to a nonsignificant level at the 5 second trace interval, and increases again to a significant level as the trace is extended to 10 second. The difference between these values approaches statistical significance (H - 4.10; p .20 .10). Another justifiable comparison across trace inter vals is Period I discrimination. This follows from the fact that since the CS was electrotactual a response was almost invariably elicited during Period I. This is true for each dominance condition and for each trace condition. The degree of discrimination as a function of trace inter val for each dominance condition is presented in Figure 23. When the CS and UCS are of equal intensity, no Period I discrimination occurs at any trace interval. When the CS is two-thirds the value of the UCS significant condition ing is obtained only when the trace is 10 second in dura tion. When the CS is one-third the intensity of the UCS, significant conditioning is obtained both when the trace is Mean Magnitude of GSR Discrimination 140 3/3 3/2 3/1 9 ♦ 8 7 6 .5 .4 - 0 3 2 1 0 -.1 -.2 r. .5 5*0 10.0 Trace Interval (Seconds) Pig. 23.— Period I discrimination for each dominance condition at each trace interval. 141 .5 second and 10 second but not when it is 5-second. This latter case is the most puzzling one. Since it was not found possible to predict during which response period would exhibit maximal discrimina tion, it may be permissible to compare discrimination values across the time interval using a mean discrimination value based upon the largest degree of discrimination shown by each S irrespective of what response interval this value was found. Such data are shown in Figure 24. When the CS and UCS are equally intense, significant discrimination occurs only in the 5 and 10 second trace groups (T = 0, N = 9 in each case; p .01). When the CS is two-thirds as intense as UCS, significant conditioning occurs again only in the 5 and 10 second trace groups (T = 1, N = 9 in each case; p .01). When the CS is one-third as intense as the UCS conditioning is obtained at each trace interval (T = 0 in each case; N = 9; p .01) . Mean Magnitude of GSR Discrimination 142 .9 .8 .7 .6 5 3 .2 1 3/3 3/2 3/1 0 -.1 .2 5 5.0 10.0 Trace Interval (Seconds) Fig. 24.— Degree of maximal discrimination without regard for response period for each dominance condition at each trace interval. CHAPTER VI DISCUSSION The present study was designed to evaluate the effects of stimulus intensity ratios (i.e., dominance) on the degree of GSR discrimination conditioning. The logic of the experiment was predicated (but not dependent) upon a theoretical analysis of the mechanism of conditioning involving the notions of neural dominance and neural contiguity. The critical event in classical conditioning, according to this interpretation, occurs at a moment in time (i.e., neural contiguity) when neural events interact in a dominance relationship (i.e., where the UCS-initiated neural event is neurally dominant over the CS-initiated neural event). Three general predictions were generated from this analysis of classical autonomic conditioning: (1 ) that second interval responding would be observed in the trace paradigm; (2 ) that the magnitude of conditioning (dis crimination) should be nonrandomly related to the degree of 143 144 dominance; and (3) that first interval discrimination should occur subsequent to second interval discrimination. These predictions concerned the longer trace conditions. For the very short trace condition, it was predicted that the magnitude of discrimination should be an increasing function of the degree of dominance. Do second-interval responses occur in trace con ditioning? The first question is simply whether or not "second interval responses" occur in the relatively long trace intervals employed in the present study. The "theory" predicts responding correlated in a temporal sense with the point of UCS occurrence. Therefore, when the interval is 5 second such UCS-correlated responding might be found in Period II or in Period III. Or, it could be demonstrated by measuring responses occurring within both of these intervals (TSR), or only in the 5 second period following the usual point of UCS occurrence (SIR). Likewise, in the 10 second trace condition, such responding might appear in Period III or Period IV, or in terms of some combination of subperiods (as in TSR and SIR). Table 6 produces certain values representing the magnitude of discrimination (across the dominance variable) indicated by these various 145 TABLE 6 DEGREE OF GSR DISCRIMINATION (ACROSS DOMINANCE CONDITIONS FOR VARIOUS RESPONSE PERIODS AND COMBINATIONS OF RESPONSE PERIODS RELEVANT ■ FOR "UCS-CORRELATED" RESPONDING Response Period Trace 5" Interval 1 0 " 5" 1 0 " Period II Period III .07 .20 Period III Period IV .11 .29** TSR TSR .19* .48** SIR SIR .2 0* .36** *p .002 **p .001 146 responses for both the 5 and 10 second trace conditions. For the 5 second trace condition, significant UCS-correlated responding is found only when measured in terms of TSR or SIR. In the 10 second trace condition, significant UCS-correlated responding is found for these response definitions as well as for Period IV responding (the 2.5 second period immediately following the temporal point of the UCS). The data in Table 6 allow the conclusion that UCS- correlated discrimination does indeed occur in trace condi tioning. In addition, it is apparent that such discrimina tion occurs to a-greater degree in the 10 second trace interval than in the 5 second trace interval. This is particularly true in terms of a Period Ill-Period IV com parison of discrimination means (z = 2.24; p .025; based on Mann-Whitney U test for large samples). It should also be recalled at this point that significant Period IV dis crimination was obtained in each dominance group of the 10 second interval condition, while no Period III value was significant in the 5 second interval condition. The fact that significant trace conditioning of a UCS-correlated response occurs better under a longer than a shorter trace would seem to constitute a major problem of interpretation 147 for any extant theory of classical conditioning. Is discrimination conditioning a function of dominance? Results of the present investigation offer an unqualified "yes" in answer to this question in terms of short trace conditioning (e.g., 5 second). Although the degree of discrimination was a linear function of increas ing dominance, significant conditioning was obtained only when the CS was one-third as intense as the UCS.-1- Perhaps the basic implication arising from the .5 second trace data is that variables such as CS-UCS interval simply cannot be considered apart or separate from the ^The effect of CS intensity in the present study supports the previous results of Kimmel (1959) in which conditioning increased as the CS became less intensively similar to the UCS. The dominant theoretical attitude (e.g., Kimble, 1961) is that CS intensity is a motivational variable (affecting drive strength) influencing performance but not learning. A dominance theory would predict that as long as CS was above threshold, increasing the intensity of CS in relation to a standard UCS intensity would result in decreased conditioning. In the present study, CS intensity in the 3/1 conditioning group was near threshold while CS intensity in the 3/2 group was moderately annoying but not painful. The results of these two groups would seem to offer strong support for a dominance conception and strong negative evidence for a motivational interpretation of CS intensity effects, particularly since the motivational effects are probably constant for CS+ and CS-. The differ ence between responses to these two stimuli represent a learning and not a motivational state. 148 intensive relation between CS and UCS. Contrary to popular dogma, the .5 second interval produces conditioning only when the dominance variable is favorable. And when this variable is favorable, discrimination occurs readily (i.e., after two differential reinforcements). This fact provides additional support for a similar finding by Lockhart and Grings (1964) in the more typical classical conditioning situation (i.e., tone as CS, shock as UCS). In discussing such rapid discrimination, the above authors argued that the actual response process observed (e.g., GSR) is not importantly involved in the discrimina tion process, but rather reflects a perceptual discrimina tion. Another way to put this is that the GSR is not to be considered a conditioned response, but rather a response system which is sensitive to perceptual processes (responses) which are developed in the classical conditioning situation. Still another way of putting it is that the conditioned con nection, whatever its nature may be, is most likely formed in the nervous system and is more or less successfully reflected in peripheral responding and overt behavior. While it might be concluded that the data for the .5 second group supports the conclusion that a certain degree of intensive difference between CS and UCS is 149 required for discrimination conditioning to occur, there is a much different interpretation which is also consistent with the results. It can be argued that the intensive difference is required only for the reflection of dis crimination in overt autonomic behavior, there being no such requirement for S to make the discrimination that one stimulus is followed by a second stimulus while another is not. In fact, this supposition is probably more nearly correct since a number of Ss in each dominance condition were able to verbally report which leg was stimulated twice in "quick" succession (information gained by informal verbal report following experimental session). This.situation, if studied in more detail and more directly than was done in the present study, would seem to bring into question the status of certain types of response modification in a full analysis of the conditioning process. 2 This is the question of the relation between per ception and learning (see Razran, 1955). Razran argues that if S is aware or can perceive the relation between CS and UCS the laws of conditioning are abrogated. He would restrict the term classical conditioning to those cases in which higher level perceptual processes are not involved. The major problem for conditioning theory is the extent to which S's overt behavior (e.g., GSR discrimination) is not isomorphic or reflective of S's covert perceptual behavior (e.g., perception of fact of double stimulation). The main question: can such perceptual behavior be considered a conditioned response? 150 The point is easily seen in eyeblink conditioning. Many differential reinforcements are required before differen tial eyeblink responding is obtained. Yet, if one asks S after a single differential reinforcement, "Which stimulus * was followed by the puff?" most Ss would be able to answer correctly. Any analysis of classical conditioning at the human level must be able to determine the effects of such perceptual discrimination on peripheral response discrimin ation performance. And, too, such a state of affairs may not be strictly human. Salivary discrimination in the dog may take a number of differential reinforcements in the same manner as the eyeblink in the man. Although it is not easy to ask the dog which stimulus was followed by meat powder, conditions could be arranged in such a way that the animal could answer such a question. For example, after a single differential reinforcement in which a white circle is followed by food and a black circle not, the dog could be freed from his restraining device and allowed to choose between the black and white circle. It would not be too surprising if the animal chose the previously reinforced stimulus more often than chance would allow. When two stimuli (e.g., CS and UCS) are presented at a level perceptible by S, the kinds of resulting 151 perceptual responses may not reflect themselves in the response system of more immediate concern to the experi menter. In making inferences from the latter, one must assume (a) relative independence of the response process from perceptual factors (as might be the case with intero ceptive conditioning), or (b) determine what effects the perceptual factors have on the response process. In either case, it is folly to ignore the perceptual processes which may develop in the conditioning situation. The role of dominance in long trace conditioning is relatively clear in theory. The trace of the CS diminishes in "neural strength" as a function of time resulting in a greater degree of dominance at the point of neural contigu ity than would occur if the trace interval were shorter. Thus, the greater the trace (within limits) the greater the degree of conditioning. Such a prediction is not readily derived from any nondominance theory of condition ing. The degree to which this prediction is confirmed diminishes in equal degree the relevance of current theories of classical conditioning--at least in terms of trace con ditioning of autonomic responses. Because of the multiple responding occurring during trace intervals longer than .5 second it was not a simple matter to define the CR. In fact, the notion of a single CR is probably quite illusory, the indication being that discrimination may occur at many temporal points dur ing the trace interval prior to the point of UCS occurrence, or after such a point when the UCS is omitted. It is undoubtedly an oversimplification to assume that the only effect of the CS over time is one of decreasing neural representation of the CS. Although such a postulate is quite relevant in interpreting the stimulus basis for "second interval type" responses, the various responses occurring during different periods of the trace as well as the complicated interrelationships between such responses argue against such a simple characterization. More likely is the assumption that with increasing temporal intervals between CS and UCS events a variety of behavioral pro cesses occur some of which may reflect conditioning and others which may reflect quite nonconditioned processes. This assumption gains support from the data when it is remembered that for any S it was not possible to predict at what point during the temporal interval of the trial maximal discrimination would occur. Some Ss conditioned best in terms of responses temporally near the point of usual UCS occurrence, while others exhibited best 153 discrimination at points quite temporally removed from this point. In addition, many Ss gave signs of "negative" dis crimination and in some cases such reversed discrimination was significant. In large part these various "complica tions" were ascribed to some type of individual difference. This point will be discussed in more detail at a later point. There are several ways of attempting to show the relationship between discrimination and dominance when the trace interval is relatively long. The first is simply to compare discrimination values for different direct dominance ratios. For the .5 second group we have seen that the relationship is an increase in discrimination with increas ing dominance. For the 5 second trace group, no such rela tion was found for any response definition (see Table 4). In fact, within each dominance condition, no significant discrimination was obtained for any response definition. For the 10 second trace group (see Table 5) significant conditioning was found within each dominance condition for a number of response definitions but in most cases the value of discrimination did not vary with the degree of direct dominance ratio. This was not true, however, for Period I discrimination, which was an increasing function 154 of dominance, at least from the 3/3 to the 3/2 dominance condition. Thus, for the longer trace intervals, the directly manipulated dominance ratio (i.e., in terms of CS and UCS intensity) does not effect the degree of dis crimination (except for Period I responding as indicated above). The second general way in which the relation between dominance and discrimination can be assessed is to compare the magnitude of discrimination within each direct dominance condition across the trace interval. The main problem here is what response, or combinations of response, are justifi ably compared across time variation. In terms of Max R discrimination (see Figure 22) the magnitude of discrimina tion increases with increasing degree of trace interval. This was particularly clear when the CS and UCS were the same intensity. This relationship was confused at other dominance levels by the failure of the 5 second trace groups to discriminate. When only the first response was compared, much the same relation was found. When the maximal discrimina tion value (regardless of response periods) was employed significant conditioning was found in all 5 second domin ance conditions and the degree of discrimination was an 155 increasing function of the degree of trace interval for both 3/3 and 3/2 dominance groups. A U curve was found in the 3/1 dominance groups. These results, coupled with the finding'that a significantly greater degree of "UCS-correlated" discrimin ation occurred in the 10 second than in the 5 second trace interval condition, tend to confirm the hypothesis that when dominance is indirectly manipulated (i.e., by lengthening the trace interval between CS and UCS) the degree of discrimination (as measured by various response definitions) is an increasing function of dominance. No reinforcement-based theory of conditioning (or cognitive theories either) would predict that conditioning would be an increasing function of the degree of temporal separation of CS and UCS. While it is true that many of the results obtained are more readily interpreted within a dominance-based theoretical framework, even this position is not wholly satisfactory. As explained earlier, dominance has meaning only at the point of neural interaction, a point which in the present study was separated from the Period I response by 10 second in the longest interval trace condition. Yet, the greatest degree of discrimina tory responding was shown under these conditions--and 156 significantly so after only two differential reinforcements. While rapid second-interval discrimination may be implicit in a dominance conception of conditioning, such rapid dis crimination as evidence in Period I is not. Increasing Period I discrimination as a function of increasing the CS-UCS trace interval constitutes a major problem for any theory of classical conditioning, and is perhaps the major result of the present study. What is the relationship between pre- and post-UCS discrimination? The third question concerns the relation ship between "anticipatory" or pre-UCS discrimination and post-UCS or "second interval" discrimination. The present study was not designed to explicitly test this relation, although the dominance theory of conditioning would gen erally consider anticipatory responding a generalized form of second interval responding (as would most other theories of classical conditioning discussed earlier). A possible explanation of this result lies in Hull's postulated "identical stimulus" mechanism of anti cipatory responding. If a particular trace point during the subsident phase of the trace is the "actual" CS, then this same trace intensity value must occur in the recruitment phase of the trace. Since discrimination was significant only in Periods I and IV in the 10 second trace groups this type of explanation cannot be dis counted . 157 The first opportunity to observe a relationship between pre- and post-UCS discrimination behavior followed two differential reinforcements. An earlier conditioning study (Lockhart and Grings, 1964) had shown significant pre- and post-UCS discrimination at this first point of comparison. In the present study, post-UCS discrimination was not significant at this early point in either the 5 or 10 second trace conditions, while pre-UCS discrimination (i.e., Period I), was significant at this point in the 5 second 3/3 dominance condition and in the 10 second 3/2 and 3/1 dominance conditions. This is the first study in which attention has been focused on the so-called second interval response in which this response did not show either the most rapid discrimination or the largest discrimination. It may be important to note here that even though Period I discrimination occurs so rapidly there may be no basis for inferring a rapid "perceptual" discrimination as was discussed in reference to the .5 second trace condition. In contrast to those Ss, very few Ss in either the 5 second trace and only 3 Ss in the 10 second trace condition were able to verbalize which side of the CS leg was followed by stimulation of the other leg. Ss were not able, according to their verbalization, to predict when the "strong" shock 158 would occur. If these verbalizations reflect the true status of the perceptual state, the nature of Period I discrimination, and particularly its rapidity, becomes a tantalizing problem for conditioning theory. There are further ways of looking at the data which complicate matters even further. If, for example, one examines the Period I discrimination at the first point of test-control comparison in those Ss showing the largest degree of TSR discrimination at the same point, an interesting problem emerges. The relevant data are shown in Table 7. The first point of interest is that (except for the 3/3 dominance condition) the degree of TSR discrimina tion is superior in the 5 second group. It should be remembered that these groups are based only on an N of 3 and statistical tests are not possible. This relation reverses itself, however, after two differential reinforce ments . The second point of interest is that in each case the degree of TSR discrimination exceeds by a large degree the value of Period I discrimination on the same trial. It is apparent that Ss discriminating at the first point in terms of TSR do not discriminate in terms of Period I 159 TABLE 7 COMPARISON OF TSR AND PERIOD I DISCRIMINATION VALUES AT FIRST POINT OF ACQUISITION FOR 5 and 10 SECOND TRACE CONDITIONS (Based on 3 best TSR discriminators in each group) Dominance Condition 3/3 3/2 3/1 Average 5 Second TSR 1.31 1.22 1.79 1.44 Period I . 39 -.28 .37 . 16 10 Second TSR 1.98 .70 .28 .99 Period I -.01 .10 .06 .05 160 responding. This would seem to imply a relative independ ence of a "UCS-correlated" discriminatory response and a "CS-correlated" discriminatory response. This implication was confirmed when the degrees of TSR and Period I dis crimination were correlated using the data of all Ss within each trace group. For the 5 second group the correlation was .09 (Spearman rho) and .07 (Spearman rho) for the 10 second group. Thus there are at least some grounds for inferring that Period I discrimination may be a function of processes independent of those operating to produce "UCS- correlated" discrimination. GSR discrimination and individual differences. Although one of Pavlov's keenest interest revolved around difference in conditioning behavior between his dogs, there has been little tendency in American psychology to continue his stress on typology. Such differences are usually treated as inherent variability in the behavior under study on the assumption that the experimental variables of inter est are sufficiently powerful to "overcome" whatever basic differences may exist between individuals at the time of behavioral sampling. Although the classical research orientation wherein individual differences are ignored 161 still commands the greater attention, current data of con ditioning seem to be forcing a more intensive study of individual differences than at any previous point in Ameri can psychology. This is not the appropriate place to review the extensiveness of these data but it may be well to indicate a few brief examples. The problem of "voluntary" and "nonvoluntary" responders in eyeblink conditioning has become a major problem area since Spence and Ross (1959) presented evidence indicating that the "laws" governing the acquisition of the eyeblink CR are different for the two types of responders. There have been numerous suggestions relating the degree of autonomic conditionability to various aspects of autonomic responding (e.g., responsivity). A recent monograph (Turner and Soloman, 1962) was based on the analysis of those Ss (an unidentified number in any conditioning experiment) who "fail to learn." Grings and Lockhart (1963) demonstrated that individual differences in degree of responding to a disparity experience are predic tive of performance in a subsequent extinction series. As a final example, there is an extensive literature relating personality variables to conditionability (e.g., Eysenck, 1965). 162 It is important to recognize that even a partial analysis of the classical conditioning of autonomic responses (or any other type of behavior) may ultimately depend upon the extent to which the investigator is willing to examine the variability in his data as a source of information. It is no longer sufficient to declare, some what benignly, that behavior is inherently variable and thus must be treated as a backdrop to "chance." If one assumes that behavior is lawful then implication dictates that variability in data means lack of control which in • turn indicates a lack of full understanding of the lawful relationships operating— hidden as they may be--in the data. Throughout the analysis of the results of the present experiment appeal was made to some type of individ ual difference to account for certain complexities in the data. This was most pronounced for the 5 second trace groups where the presence of this "individual difference" seemed to preclude demonstrating discrimination on any kind of group basis. If, for example, each dominance group in the 5 second trace condition is divided at the median in terms of degree of Period I discrimination, the curves in Figure 2 5 are obtained. The above-median-discriminators 163 Ft O •H +» c d S3 •i-t S • r - 4 o 0 0 CO c ! > «H o 4 > T3 3 +* •H c d S § « s .5 .4 .3 .2 .1 0 -.1 -.2 -.3 -.4 H Above Median Discriminators Below Median Discriminators \ o — — o 3/3 3/2 3/1 Dominance Ratio Fig. 25.--Above and below median discrimination curves for each dominance condition in the 5 sec. trace interval group. discriminate significantly but are not affected by the pre vailing dominance ratio. The below-median discriminators, however, are effected by the dominance ratio: the degree of discrimination becomes increasingly negative. (This finding would seem to rule out the possibility that the results are due to measurement error.) When these two subgroups are combined, of course, the overall degree of discrimination becomes almost nil. The overall or group curve is misleading because it masks the lack of a func tional relationship between discrimination and dominance when discrimination is positive, and hides a functional relationship between negative discrimination and dominance in negative discriminators. Why Ss discriminate negatively is unknown. The failure to obtain group discrimination in the 5 second trace groups, while obtaining good discrimina tion in the 1 0 second group may have been the result of a peculiar interaction between the 5 second trace interval and the individual difference variable of negative dis crimination. If conditioning is to serve as a major explanatory schema for behavior it is ultimately necessary to base such a schema on the validity of classical conditioning concepts as they operate within the individual. The 165 continued success of operant conditioning can be laid main ly to the control of necessary stimulus variables within the individual and not the group. Demonstrating any phenomenon on the basis of group data is only a first step and not the last. And certain features of the present data indicate that the group approach may even be a misstep. The author feels that it is more important, for example, to determine why one S discriminates positively and another negatively than it is to determine why the average of the two represents no discrimination at all. Concluding remarks. It seems appropriate at this point to conclude this discussion with a brief summary of the "positive" statements which the data support as well as to outline the future steps which should be taken to con tinue the efforts initiated here. The following statements then may be taken to represent the contribution of the present experiment to the field of autonomic conditioning: 1. Discrimination conditioning of GSR is easily and readily obtained in a trace conditioning paradigm in which both CS and UCS are electrotactual stimuli. 2. The degree of discrimination in trace condi tioning increases as the degree of trace is increased from .5 to 10 second particularly when the CS and UCS are of equal intensity. 166 3. When the trace interval is quite brief, the degree of discrimination is a direct function of the degree of dominance. 4. When the trace interval is relatively long, maximal discrimination may appear in the form of differential responding at various temporal points throughout the trace interval and beyond, with the predominant points of maximal discrimination associated with CS occurrence and the temporal point of UCS occurrence when the UCS is not presented. 5. There was evidence that CS intensity is not an independent motivational non-learning fac tor in the conditioning process, but rather in concert with its intensive relation to the UCS determines the degree of conditioning, par ticularly when the trace interval is quite short. 6 . As the trace interval is lengthened from .5 to 5 and 10 second there occurs an interaction with the temporal interval and some type of individual difference factor which results in "negative" discrimination in many S_s. 7. Perceptual awareness of the contingent rela tionship between CS and UCS is more prominent when the trace interval is brief than when it is long. 8 . There is evidence that the discrimination per formance in terms of autonomic responding imperfectly reflects the degree of perceptual discrimination, calling into question the use of the former to define the latter and requir ing the investigation of the relationship between central and peripheral response pro cesses which occur in the conditioning situation. 9. The results of the present investigation are most amenable to a dominance interpretation of classical conditioning, although certain fea tures of the data (e.g., rapid Period I 167 discrimination in the 10 second trace group) cannot be explained in terms of any current conception of the conditioning process. It is important to indicate the broad outline of steps which must be taken to elucidate and confirm these conclusions. Because of the uniqueness of the paradigm employed in the present study, there is no possible way of assessing the reliability or generality of the data obtained. Therefore, it will be necessary to utilize the particular paradigm of CS and UCS as electrotactual stimuli in a wider variety of situations, with other responses and with other animals than the human. The interval between CS and UCS must be extended beyond 10 second to determine the limits of the degree of discrimination with increasing trace. It will be necessary to use shock as CS and UCS in the simple nondiscrimination conditioning situation. In addition to studying the applicability of the double-shock paradigm in a wider variety of situations is the necessity of studying the possible variables responsible for the individual differences found in the present study. This concern could take several forms from examining response variables such as responsivity, activation level, degree of spontaneous reactivity, et cetera, to the examination of various personality factors such as degree of manifest anxiety and intraversion-extraversion, et cetera. Another line of additional,study would be the more direct examination of the role of perceptual factors in the conditioning paradigm. Such an approach would perhaps manipulate the situation in terms of instructional param eters, or attempt to eliminate perceptual activity by dis guising the situation. In addition, it would be necessary, perhaps, to attempt the same paradigm at the animal level and at the interoceptive level. The latter may be even more amenable to a dominance conceptualization of condi tioning than exteroceptive conditioning. The present experiment began as an attempt to test certain notions concerning the classical conditioning of autonomic behavior in human Ss. The idea that condi tioning is a function of dominance received substantial empirical support from the data, while at the same time certain features of the data could not be readily inter preted from a dominance point of view. The idea that anticipatory conditioning represents a generalization of temporal-specific conditioning occurring near the point of neural contiguity could not be adequately tested by the present data. Indeed, the indication was that these two 169 response processes are independent. Perceptual influences may be the key to understanding the influence of individual differences which were evident in the data, particularly when the trace is long. The results indicate that the conditioning process in the human is an amazingly complex thing. The sorting out of the variables responsible for the variability found between subjects and within subjects is an.enormous task but one which must be undertaken if the phenomena of con ditioning are to be understood and if they are to serve as a basic framework or foundation for theories of learning. The present study should be considered only a first step in an eventual focus on the complexity of the conditioning process. No longer can this form of learning be explained away as "simple," "stupid," or "mechanical." It is none of these. But what it is— that is the problem with which the present experiment began and with which the present experi ment ends. B I B L I O G R A P H Y BIBLIOGRAPHY Cochran, W. G. and Cox, G. M. Experimental designs. New York: Wiley, 1957. Diamond, S., Balvin, R. S. and Diamond, F. R. Inhibition and choice. New York: Harper and Row, 19 63. Estes, W. K. "The Statistical Approach to Learning Theory." In S. Koch (Ed.) Psychology: A study of a science. New York: McGraw-Hill Book Company, Inc., 1959. Eysenck, H. J. Extraversion and the acquisition of eye- blink and GSR conditioned responses. Psychol. Bull., 1965, 63_, 258-271. Grice, G. R. and Hunter, J. J. Stimulus intensity effects depend upon the type of experimental design. Psychol. Rev., 1964, 71, 247-256. Grings, W. W. Preparatory set variables in the classical conditioning of autonomic responses. Psychol. Rev., 1960, 6_7, 242-252. Grings, W. W. and Lockhart, R. A. Effects of "anxiety- lessening" instructions and differential set develop ment on the extinction of GSR. J. exp. Psychol., 1963, 6 6 , 292-299. Grings, W. W., Lockhart, R. A. and Dameron, L. E. Condi tioning autonomic responses of mentally subnormal individuals. Psychol. Monogr. , 1962, 16_, (39, Whole No. 558). Grings, W. W., Lowell, E. L. and Honnard, R. R. GSR con ditioning with deaf children. J. comp, physiol. Psychol., 1961, 54, 143-148. 171 172 - Guthrie, E. R. The psychology of learning. New York: Harper, 1935. Hartman, T. F. and Grant, D. A. Differential eyelid con ditioning as a function of the CS-UCS interval. J. exp. Psychol., 1962, £4, 131-136. Hilgard, E. R. and Marquis, D. G. Conditioning and learn ing. New York: Appleton-Century-Crofts, Inc., 1940. Hilgard, E. R. , Jones, L. V. , and Kaplan, S'. J. Condi tioned discrimination as related to anxiety. J. exp. Psychol. , 1951, 42^, 94-99. Hull, C. L. Principles of behavior. New York: Appleton- Century-Crof ts , Inc., 1943. ___ . Essentials of behavior. New Haven: Yale, 1951. Jones, J. E. Contiguity and reinforcement in relation to CS-UCS intervals in classical aversive conditioning. Psychol. Rev., 1962, 69_f 176-186. Kendler, H. H. "What is learned?"— A theoretical blind alley. Psychol. Rev., 1952, 59^, 269-277. Kimble, G. A. Conditioning and learning, a revision of Hilgard and Marquis. New York: Appleton-Century- Crofts, Inc., 1961. Kimble, G. A. and Ost, J. W. P. A conditioned inhibitory process in eyelid conditioning. J. exp. Psychol. , 1961, 61, 150-156. Kimmel, H. D. Amount of conditioning and intensity of conditioned stimulus. Unpublished doctoral disserta tion, University of Southern California, 1958. Kimmel, H. D. and Pennypacker, H. S. Differential GSR conditioning as a function of the CS-UCS interval. J. exp. Psychol., 1963, 65_, 559-563. Lawrence, D. H. "Learning." In P. R. Farnsworth and Q. McNemar (Eds.) Annual Review of Psychology. Palo Alto: Annual Reviews, Inc., 1958. 173 Leonard, C. and Winokur, G. Conditioning versus sensitiza tion in the galvanic skin response. J. comp, physiol. Psychol., 1963, 5£, 169-170. Lockhart, R. A. and Grings, W. W. Comments on "An analysis of GSR conditioning." Psychol. Rev., 1963, 70, 562- 564. ________ . Interstimulus interval effects in GSR discrimin ation conditioning. J. exp. Psychol. , 1964, 61_r 209- 214. Moeller, G. The CS-UCS interval in GSR conditioning. J. exp. Psychol., 1954, 48, 162-166. McDonald, D. G. and Johnson, L. C. A re-analysis of GSR conditioning. U. S. Navy Medical Neuropsychiatric Research Unit Report 64-6, 1964. Mowrer, O. H. Learning theory and behavior. New York: John Wiley and Sons, Inc., 1960. Mowrer, O. H. and Soloman, L. N. Contiguity vs. drive- reduction in conditioned fear: temporal variations in conditioned and unconditioned stimulus. Amer. J. Psychol., 1954, 67, 15-25. Passey, G. E. On Razran's favorable ratios of excitation. J. Psychol., 1959, £8 , 341-346. Pavlov, I. P. Conditioned reflexes. New York: Oxford University Press, 1927. Razran, G. Theory of conditioning and of related phenomena. Psychol. Rev., 1930, 31_, 225-243. Conditioning and perception. Psychol. Rev., 1955, £2, 83-95. ________ . The dominance-contiguity theory of the acquisi tion of classical conditioning. Psychol. Bull., 1957, 54, 1-46. 174 Razran, G. The observable unconscious and the inferable conscious in current Soviet Psychophysiology: intero ceptive conditioning, semantic conditioning, and the orienting reflex. Psychol. Rev., 1961, 81-145. Rodnick, E. H. Characteristics of delayed and trace conditioned responses. J. exp. Psychol., 1937, 20, 409-425. Silverman, R. E. Eliminating a conditioned GSR by the reduction of experimental anxiety. J. exp. Psychol., 1960, 59. 122-125. Smith, K. Conditioning as an artifact. Psychol. Rev., 1954, 61, 617-225. Spence, K. Behavior theory and conditioning. New Haven: Yale University Press, 1956. Spence, K. W. and Ross, L. E. A methodological study of the form and latency of eyelid responses in condition ing. J. exp. Psychol., 1959, 5jJ, 376-381. Steckle, L. E. A trace conditioning of the galvanic reflex. J. gen. Psychol., 1933, 9_, 475-480. Stewart, M. A., Stern, J. A., Winokur, G., and Fredman, S. An analysis of GSR conditioning. Psychol. Rev., 1961, 6J3, 60-67. Switzer, St. C. A. Anticipatory and inhibitory character istics of delayed conditioned reactions. J. exp. Psychol. , 1934, 17_, 603-620. Turner, L. H. and Solomon, R. L. Human traumatic avoidance learning: theory and experiments on the operant- respondent distinction and failures to learn. Psychol. Monogr. , 1962, 16_ {40, Whole No. 559). Uchima, A. The conditioning of galvanic skin response as a function of the time interval between the conditioned stimulus and the unconditioned stimulus. Unpublished Master's thesis, University of Southern California, 1962, 175 White, C. T. and Schlosberg, H. Degree of conditioning as a function of the period of delay. J. exp. Psychol. , 1952, 43_, 357-362. Wickens, D. D. and Harding, G. B. The effect of UCS strength on GSR conditioning: a within subject design. J. exp. Psychol., 1965, in press. Wickens, D. D., Allen, C. K., and Hill, F. A. Effect of instructions and UCS strength on extinction of the conditioned GSR. J. exp. Psychol., 1963, 6 6 , 235-244. APPENDICES APPENDIX A Disparity The perceptual disparity response. Grings (1960) has defined the perceptual disparity response (PDR) as a "difference in magnitude of response between situations where the receipt of stimulation is in accord with past experience and where such receipt of stimulation is not in accord with past experience." Based on such a concept, Grings offered confirmatory data for the hypothesis that the degree of PDR varies with the strength of the percep tual set, i.e., the greater an expectation, the greater the disruptive influence of a change in the conditions. Disparity and the half-second trace CR. In the .5 second trace group, disparity was created by pairing the previously nonreinforced cue (CS-) at the normal CS- UCS trace interval (.5 sec.) with the UCS. On the just preceding trials, CS- was paired with the UCS. This latter"trial was employed as a "reference" trial. PDR was defined as the difference in UCR magnitude on dispar ity and reference occasions. A simple prediction based on Grings1 (1960) hypothesis is that the degree of PDR would vary with the degree of discrimination conditioning. For the .5 second trace group, this would mean that PDR should increase as a function of dominance. Reference to Table 8 shows quite clearly that this is not the case. No PDR value reached significance. Examination of the individual dis parity values revealed that more than one-half of the group (17 of 27) responded more to the reference stimulus than to the disparity stimulus. For the .5 second condi tions, then, there appears to be no relation whatever between degree of discrimination and magnitude of PDR. 177 178 Disparity and the five-second trace condition. In the 5 second trace group disparity was produced by presenting the UCS at a point 5 second following the usual point of UCS occurrence (on the CS- trial). This should be termed a "temporal" disparity. Reference again to Table 8 shows again, as with the .5 second trace group, that no disparity value was significant. This could have been predicted, however, since no discrimination was shown in the 5 second trace group. About one-half of the sample again exhibited "negative" disparity. Disparity and the ten-second trace condition. In the 10 second trace group, disparity was produced by pre senting the UCS at a point 5 second prior to its usual time of occurrence. Again, this is a temporal disparity. Reference once again to Table 8 shows that the mean PDR is positive only in the 3/2 dominance condition but not significant. However, significant "negative" disparity occurred in the 3/1 dominance group (T = 3; N = 9; p = .02). Over half of the sample exhibited larger reference than disparity responses. Comment. It should be kept in mind that the present study represents a radical departure from the typical conditioning study and from any study in which the PDR has been found. Therefore, these results should not be readily interpreted as negative evidence for the con cept of PDR since the latter may be characterized by cer tain boundary conditions which the present study does not . meet. And for these reasons, no further analysis of the data concerning disparity will be presented. Extinction Extinction period and half-second discrimination. Extinction trials began immediately following the disparity experience. Table 9 shows the degree of discrimination performance on the last acquisition test control point (prior to disparity) and the degree of discrimination maintained between test and control responding during the three presentations of the CS- and CS- during the extinc tion period for each dominance conditioning separately. For both the 3/3 and 3/2 dominance condition there is an 179 increase in discrimination during extinction in comparison with the last point in acquisition. For the 3/1 group, the level of discrimination performance remains approximately equal. Extinction period and five-second discrimination. Table 9 also shows the extinction data for the 5 second trace groups for each dominance condition. In general, it is evident that discriminatory responding is not only reduced (as in the 3/3 group) but in many cases becomes negative (as in the 3/2 and 3/1 groups). Extinction period and ten-second discrimination. Table 9 also shows the extinction data for the 10 second trace groups for each dominance condition. As with the 5 second trace group, in comparison with the last point during acquisition, discriminatory responding is greatly reduced (and in several cases negative). Comment. The effect of the extinction trials is apparently different depending upon the degree of trace interval. Under half-second conditions extinction dis crimination increased in those groups showing no overall acquisition conditioning behavior (3/3 and 3/2 conditions) and remained the same for the group showing significant acquisition behavior (3/1 condition). In the 5 second trace group and the 1 0 second trace group the degree of extinction discrimination decreased markedly. There are two factors which discourage further analysis of the extinction data. The first is that the extinction data follow the disparity experience. Grings and Lockhart (196 3) have shown that the degree of respond ing in the disparity situation may play a large role in determining responding to future changed conditions (e.g., extinction). However, since the disparity data themselves were quite ambiguous in the present study attempts at further analysis would seem to carry one far afield of the intent of the present study and will be left to a future occasion. The second reason, of course, is again the idiosyncratic nature of the present investigation. Until further studies of the type described herein are per formed there is no way to assess the reliability or generality of these findings. 180 "Disinhibition" Following the sixth trial of extinction an uncued presentation of the UCS was administered. In a general way, the occurrence of an uncued UCS was considered a "nov el" stimulus and thus capable of serving the typical dis- inhibition function. During extinction, this would mean a reinstatement of the conditioned response, or, in this case, a reinstatement of discriminatory responding. How ever, since the extinction data were so complicated in the present study it is hardly realistic to expect a readily interpretable effect of the "disinhibitory" experience. For this reason, the "disinhibitory" data is simply presented without evaluative comment. The data in Table 10 are comprised of the degree of discrimination at the point just preceding the disinhibitory UCS and the degree of discrimination just following and the degree of difference between these two values. £f ■— 8e*e S 6 * S 81 ' 90't» XZ‘— 9 Z " t " 9 O " t - e>B t = — 3 A v e 3-— 8 -X S 3•X 3 e * — G 6-3 X 9 “ 3 86 * — 6 9*G X z .*3 S 3 ■— Z . ^“9 3 3■9 xe * SX ■9 9^ *9 3 O * — G 3 G •G z . e *— e z . “ G 9 e*e 9 c * 8 X ”G ^ S*G GO* S 6■ *3 8 6“3 9 3 *— x 0 - C s z .*3 3S * 3 8*3 G*G G 3 * — ^X *G X 6“3 0 3 ■— 0 8*3 0 9■3 x 0 * — X 8*3 08 ■3 s^ * x — 6 ^ 9 f ' O•S 0 s*x — Z . 8"e z . e■3 S O * X 9 3 ■G XG • s s * 98 ■V O ^“S 9 0•x — X S*e S •3 6 X * 9 Z .•S S 6"S so* 3 8“ Z . 8 " ^ 0• X 0*e s 0 •e S6* — G 3 - 8 3 “G Z . 3 * — X S •G 3•G ^ 0 * z . s* 3 e s•3 9 1 - * 6 G■S S 8•S 3 S ' " G S*S S 6“S 61 * — Z . X * G 86*3 Z _ O * 09*3 Z . 9*3 OX* Ol^"3 O S * 3 3 S • 6 G- X 6 - Z , S * 8 6-G S S - Z . G ■— 6 3•G 3 6•3 3 O• G X •G SX •G O 9 * 3 9■3 3 3 ~ e O • 3 9 3 3 O •G X X ~ 6 Z .■3 0 6-3 e> * — O 8■3 Z . G•3 ^ 3~— S Z .~3 X S•3 8 3 * 9 O•G t ^ e•G Sfr* Z . 6 ■ X 3 ^■3 X 6 - X 8 Z .•3 6 9“ X 0• — 6 8•X o z .*X 66*- 66 * O O O O 3 X * — G 3~G XX *G O X 3■G X 3 *G 9 f? “— 6 X “3 G Z .•X 3 O• X 6•3 G 6*3 S 9 * 9 3■G X 6■G Z . G•— 8 G “X X 0 *X 6 6* X — 6 9•3 o z ." XX * — S 6 •X ^ 8 “X 9 G*— ^ Z .•3 8 G•3 f ” G* — 8 V - frU * 9 X " — ^ 9*3 8 f? •3 Z . I* ■ X 8■G 8 3•fr SO* X 6 * 3 96*3 30“— 38*8 08*8 SX* 99*3 X8*3 ^£>t=> — zx 9 t r* 3 9*3 80 “G O 3 ■— S 8“3 S 9~3 3 X•— X 0■3 6 S ■ X s 0 ■ X 6-3 9 6*3 S ^* 6 8’ ^ G•S S 3- SX •X O ^-X G X • — G G ■X O 3*X G G• G O’3 9 G-3 9* G S•3 Z . X •G G O* — S 3 ■3 3 3*3 z . 0■— 3 Z .~3 S 9*3 G O~— Z . 3 *G ^ 3*G X X * — 9 f r-3 S G*3 SS •— S 3-G O Z .*3 O 9-X 9 6•G 9 S~S OX * — 3 O * 3 6*G 6 O *— 8 G*3 6 3 * 3 9 G* S S*3 X 6*3 03 ■ — T? fr ~G z~G ^e •— OG •3 9 6•X 6 6•— f? 8~ S 8~G 03 • X ^*e 3 9*G 0 z .■— 9 Z .■G 9 O■G X 0~— GX *X 3 X •X xe • X Z .•G 3 O • « > ■ X 0*X 9 9•S Z . 9 • 9 ^ O•— O S-3 9 ^ -•3 Had - 3 : ^ *3 ; - dsxa Had • 3 : ^hi ’dsxa h i a d *g:aH * dsya _ _ z/c c/e j-oargns Hova h o h sa:mr^z\ (Had) asNOdsan j,iHvasia T : \ z f i J L c a ; a : o H i 3 : « 3 : a r s i ^ z 'aDNanaaan ' Aj,iHVdsia 8 a a a v i , 182 TABLE 9 MEAN MAGNITUDE OF DISCRIMINATION FOR THREE EXTINCTION TRIALS AND LAST ACQUISITION TRIAL Last Acquisition Extinction Trials Trial 1 2 3 .5" 3/3 -.36 .15 . 51 .08 3/2 •46 .59 1.18 .82 3/1 1.14 1 . 0 2 1.09 1.19 5" 3/3 1 . 0 1 1.05 .83 .02 3/2 .16 -.50 -.26 .72 3/1 .01 -.74 -1 . 0 2 .32 1 0 " 3/3 .50 .55 -1.37 -.28 3/2 2.89 .59 .91 - . 1 0 3/1 2.15 -.16 .51 .34 183 TABLE 10 COMPARISON OF "PRE-DISINHIBITION" AND "POST- DISINHIBITION" DISCRIMINATION PERFORMANCE FOR EACH TRACE INTERVAL AND EACH DOMINANCE CONDITION Pre- disinhibition discrimination Post- aisinhibition discrimination Differ ence Half-Second Trace 3/3 .08 .01 -.07 3/2 .82 1.19 .37 3/1 1.19 1.14 -.05 Five-Second Trace {Max R) 3/3 .02 .78 .76 3/2 .72 .31 -.41 3/1 . 32 -.95 -1.27 Ten-Second Trace (Max R) 3/3 -.28 -.63 -.35 3/2 - . 1 0 2.18 2.28 3/1 .34 .41 .07 APPENDIX B General Instructions The surface of the skin on the body has some interesting electrical properties. One of these is the fact that if you put two metal electrodes at different places on the body there will be a continual change in electrical potential evident between these two electrodes. It is as if the skin were a very tiny battery, the voltage of which fluctuates with the internal workings of the body. We have discovered, for example, that concentrating on an arithmetic problem leads to a different pattern of elec trical waves between the two fingers than does simply sitting quietly with the eyes closed. Perhaps you have been acquainted with this phenomenon because of the fact that it has been used in the so-called "lie detector." This is because there is reason to believe that the electrical responses are different when an individual is trying to misrepresent information than when he is telling the truth. If you are interested in the physics involved in this particular measurement it may be informative to know that these changes may be recorded as electrical potentials or as changes in the resistance of the skin to the passage of a minute current. The results obtained both ways are similar. We shall measure the resistance between an elec trode on one finger and an electrode on another finger. The current passing through your hand will be too small for you to detect. Any changes in the amount will be recorded on electrical amplifying and recording equipment which is located behind the partition. At the conclusion of the experiment we will be very happy to show you the equipment and the type of record which is being made. There are certain actions on your part which might cause disturbances on the record. Therefore, your 184 185 cooperation in avoiding these will be appreciated. Try not to move the hand on which the electrodes are placed any more than is necessary to keep a comfortable position. Particularly, do not push the electrodes or flex the fingers excessively. Coughing and deep breathing should also be avoided as much as possible. However, there is no need for you to be uncomfortable--just relax. APPENDIX C Shock Setting Instructions You will recall that you would receive occasional electric shocks during this experiment. It is now your task to help me set an intensity value which is appropriate for you. This is necessary because each person has a dif ferent threshold and tolerance for shock stimulation. Pairs of electrodes are attached to your legs. The shock will be delivered through these electrodes. We will begin with an intensity that you cannot feel. The shock intensity will then be increased in gradual steps. I want you to tell me as soon as you feel the shock in your leg. At this low value it should feel like a slight "tingle." After your threshold is determined the shock will continue to increase until a point is reached beyond which you don't wish to continue. The same procedure will be followed for each electrode site. 186 APPENDIX D MEAN VOLTAGE VALUE FOR CS AND UCS STIMULI FOR EACH EXPERIMENTAL CONDITION Dominance Condition 3/3 3/2 3/1 .5 Trace CS 68.9 49 . 8 24.6 UCS 68.9 71.6 73.9 5 Trace CS 72.8 50 .1 24 .4 UCS 72.8 76 .1 73.4 10 Trace CS 68.3 53.7 25.6 UCS 68.3 80 .0 73.9 187 APPENDIX E Final Instructions For the rest of the experimental session I would like you simply to sit back and relax. There is nothing special for you to do. I am interested simply in the way in which your galvanic skin response reacts to occasional shock stimulation. Remember--do not move your fingers or hands excessively. Try to keep your legs in a comfortable posi tion without too much movement. There is nothing for you to do except relax. Try now to assume as comfortable position as possible and try to keep movement to a minimum. I will turn the lights down now and we will begin. 188 APPENDIX F MEAN ADJUSTED TEST AND CONTROL TRIAL RESPONSES FOR EACH SUBJECT IN EACH DOMINANCE CONDITION FOR THE .5 SECOND TRACE INTERVAL Group Subject CS+ CS- Difference 3/3 5 -1.58 - . 8 6 -.72 12 1.14 .13 1 . 0 1 23 .64 1 . 2 2 i • U1 00 36 .29 .51 - . 2 2 45 .52 .26 .26 51 .14 .34 - . 2 0 61 - . 2 1 -.36 .15 67 .08 -.16 .24 81 .39 .17 .22 3/2 2 .67 . 32 .35 14 -.93 -2 . 1 0 1.17 21 .35 .01 .34 30 .33 .33 .00 44 .21 -.80 1 . 0 1 54 .12 .24 - . 1 2 59 .34 .24 .10 71 -.25 -.37 .12 76 -.77 -.48 -.29 3/1 4 1.07 .39 00 • 10 1. 1 1 -.26 1.37 27 -.36 -.85 .49 28 .39 .08 . 31 40 .15 -.35 .50 48 -.15 ' -.50 .35 57 .40 .01 .39 64 1.90 -.36 2.26 75 .21 -.30 .51 189 APPENDIX G MEAN TRIAL BY TRIAL TEST AND CONTROL RESPONDING FOR EACH DOMINANCE CONDITION IN THE .5 SECOND TRACE GROUP Trial Comparison 1 2 3 4 5 6 7 8 3/3 CS+ cs- (aiff.) -.13 -.14 ( .0 1) -.16 .09 (-.25) - . 1 1 -.22 (.1 1) -.23 .30 (-.53) -.03 .26 (-.29) -.14 .20 (-.34) + .12 .30 (-.18) -.06 .30 (-.36) 3/2 CS+ CS- (diff.) .11 -.09 (.2 0) .14 -.18 (.32) .06 .03 (.03) .21 -.08 (.29) -.06 -.43 (.37) .01 -.34 (.35) .23 -.29 (.52) .09 -.37 (.46) 3/1 CS+ CS- (diff.) .53 -.16 (.69) .53 -.20 (.73) .40 -.10 (.50) .49 -.17 (.6 6) .52 -.22 (.74) .80 -.15 (.95) .82 -.37 (1.19) .69 -.45 (1.14) rl 0 i f t o m o i t o s M n i o i P c o i n m o H s f o i o c o P « * • • * • • • • C HH 1 I II 0 1 1 u 01 31 p o i ' f c o i o H i n H o o i 01 rl « 1 0 1 ^ l O M t s rl'0, ls ls Z \ (1 ) 0 01 E i 01 H rl ft 1 1 1 01 M s P (1 H ft 4 1 D •n M ' l o m M o o i n o Q 0 P HOIOIOI^COIOCO o ft 3 H C 3 01 ft 31 01 ft 0 rl Q ft 0 03 10.03 0 1 0 3 in 03 0 0 1 \ ^ p C O O M O O I O O I ' I ' p • • • • • • • • * Q c 1 1 1 1 H ft Z 0 1 H 0 u 31 01U 31 X “ 01 H ft p ft r l r l N M O r l C O i n O l Q in in 01 4 1 or>coo\C3iooivooa Z E l \ 4 ) X • ••«•« t • • 31 Z 01 E i rfj I I rl rl | | ft * H S 1 ft ft < M X U P < 31 0 a b 4 ) 33 •n r i i o o i o i o i o i m o i f fl P lb p r l (S 0 |3 f< fin ifi t' 0 tfl 3 ft 01 h a z u O C 0 61 H 0 o i u i c a o i o i r ' r i o i o i Q ft P i n o i d o i o o i ' f o i r ' Z O P • * . i i i i i . <j ft 3 1 r| r| r| rl rl 0 1 ti U 01 31 E h P o i o i ^ i n o i t ' o i n i n Z 01 ( 1 1 o o i c o o o i i n H H i n < \ 0 ) « • * • » • • • ! I 0) H 01 H rl 01 rl 01 rl 1 P 0 0 ) •n o i i n i n o i H M D a n 1 p H o i o i p . ^ i n o r ' 3 01 1 0 0 1 1 0 < * 0 0 0 1 e c u 1 o n n o 0 0 ' ' ' 0 • • ' r t oi oi h co in oi oi f l 10 ^ H ^ f 0 0 0 • • • '000 0 I s a n onnoho 01 01 M r l r l H 0 0 1 1 ' 05 • t i i i • » » < h r l 0] 0 < # C O 0 1 co o oi h n o 0 0 1 • 0 ' ' ' • H H co co m ^ o * io h ro sr >j co co Q 0 • ' ' ' ' • i rl c*MsinMooino riorninnnoiom oo in to o o co oi O f l d < ( « O' . . . . .000 . H ini'>aiacoinriM’ i q H O l W ' J H H H O I O l o ............................... H rl rl 05 n o oi n oi oi io in 'CCOOICO 01 rl 01 rl i . . • o . ■ . . rl rl r| oi .-w io in in co n 0 0 i n r > o ooi^oi 1 i i . o . . ■ . rl 01 rl rl o i i n m o i r i M o c o i 1 r lOICO^'finiDt' in M O 01 0 C D 03 co 00 co in i' ^ in oi 1 < t i Q < i i i H H 03 m m o in m o oi in o H co 03 o in . . . . o • • * . rl rl rl 01 0 3 i n i n O l r l M O C O ^ H o i n ^ ' t i n c o c ' m o o n n ^ ' c c o i n c o Minn^mcooion' • •••••••* H rl rl I I I I I | I I I OI'J OIOO IIO O IOO I O l - t f r l i n O I M O H O « i t • i « • i » 01 rl rl I rl | I I I I r'Momt'iooino rlOJOlOl^CDlOOO M0C0l0C0in03O3f I ' o m ^ o i c o o ^ j H • ••••••»• I I I I I I I H I orar'air'riooio ^o\oioi3fo^oim • • « • • • • • * I I I rl I rl | I I eif'^iomoorimoi o i o M o i o o i H i n o . i i . . • . . . I I rl | I 01 01 01 0) ^ 0 I s s f o i i n o i o i H o o i o <f rl . I I I . . . • . I I rl II I oiininniHMoco* H o m m ' i n i o r * APPENDIX X ME AN TEST AND CONTROL MAX R, TSR, SIR AIMD RERIOD X RESPONSES EO F ? . EACH SUBJECT XN XO SECOND TRACE GROUF* 3/3 3/2 3/X S ut> j ect Test Control S ut> ject Te s t Control Subject Tes t Control 8 1.60 1.25 6 MAX R 3.35 1-47 3 1.49 O 18 4-91 3.23 1 3 3.35 — .21 11 1.70 1.07 2 O 4.49 2.64 2 4 1.85 .67 19 .71 — 1.29 31 5.25 3.13 3 4 . 12 1.04 3 3 3.28 1.56 4 2 3.98 3.22 3 9 1.62 .66 3 8 1.28 — .22 5 3 4.50 1.58 5 O 5.57 1.44 5 2 3.12 2.02 6 2 — .06 .29 5 5 6.84 .70 6 3 3.08 2.07 “ 7 2 - 4 7 — .04 7 O 2.26 - 39 6 6 4.37 5.25 7 9 .93 — . IO 78 . 49 .59 7 3 1.07 1.08 8 - 4 5 o 6 TS R .54 O 3 -40 O 1 8 1.27 1.24 13 .96 . 2 1 11 1.14 .56 2 O 1.24 .61 2 4 .37 . 39 19 1.39 . 54 3 1 2.22 .68 34 .40 . 50 3 3 1.49 .20 4 2 1.98 1 . 16 3 9 .63 . 30 3 8 . 52 O 3 3 2.47 .96 50 2.94 1.02 5 2 _ 52 - 30 6 2 O O 5 5 .06 O 6 3 1.02 - 8 7 7 2 .25 O 7 O - 2 9 O 6 6 1.51 1.60 79 O O 7 8 .46 .50 7 3 .63 - 35 8 .45 - 3 8 6 SIR .63 O 3 .40 O X 8 1.34 .62 1 3 1.33 - 2 1 11 1.32 . 39 2 O .61 - 9 1 24 .66 - 39 19 1.51 . 31 3 1 1.08 O 3 4 . 5 2 .66 3 3 .87 .44 4 2 1.70 1.40 3 9 .61 . 16 3 8 . 40 O 5 3 1.55 .70 5 O 2.25 .77 5 2 r .60 .49 6 2 O O 5 5 .06 . 2 : O 6 3 . 8 1 .68 7 2 O O 7 O . 35 O 6 6 1 . 14 1.11 7 9 . 17 O 7 8 .32 - 1 7 7 3 .32 .65 8 1 . 15 . 87 6 BERIOD X 2-16 1-16 3 - 9 5 O 1 8 .43 — .28 1 3 1.11 — .98 11 — 1.04 — .75 2 0 1.12 .88 2 4 . 70 .28 19 — . 81 — 1 .83 31 . io - 41 34 — .84 — . 16 3 3 1.30 . 86 4 2 .57 . 85 39 .83 . 12 3 8 .63 — . 49 5 3 -07 — .23 5 O .03 — . 29 5 2 .88 . 30 6 2 — . 06 .29 5 5 .07 — .90 6 3 .71 .58 7 2 .23 — .04 70 . 7 8 . IO 6 6 .45 O 7 9 - 2 1 — . IO 7 8 — .07 — .38 7 3 - 2 9 — . io

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Reward Expectancy Strength As Related To The Magnitude Of Frustration In Children

PDF

Human Performance As A Function Of The Joint Effects Of Drive And Incentive Motivation

PDF

Inhibitory-excitatory interaction of conditioned autonomic responses through interoceptive and exteroceptive conditioning

PDF

The Effects Of Prior Part-Experiences On Visual Form Perception In The Albino Rat

PDF

Conflict And Generalized Conflict In Verbal And Motor Responses

PDF

The Effects Of Reinforcement Pattern, Intertrial Interval Regularity, And adaptation On The Orienting Response And Classical Gsr Discrimination Conditioning

PDF

The role of cues in the arousal of anxiety

PDF

The Effects Of Ether And Electroconvulsive Shock On One Trial Appetitive And Adversive Learning

PDF

The Visual Fusion Threshold (Vft) Test As A Measure Of Perceptual Efficiency In Kindergarten And First Grade, And As A Possible Predictor Of Later Reading Retardation

PDF

Classical Discrimination Conditioning As A Function Of Probability Of Reinforcement

PDF

An Effect Of Verbal Conditioning On Nonverbal Behavior

PDF

The Modification Of Partial Reinforcement Effect As A Consequence Of Electrical Stimulation Of The Caudate Nucleus In Cats

PDF

Autonomic and cognitive indices of semantic conditioning and generalization

PDF

Transfer Of The Partial Reinforcement Extinction Effect Across Tasks In Normal And Retarded Boys

PDF

The Effect Of Discriminability On The Partial Reinforcement Effect In Human Gsr Conditioning

PDF

A Cross-Cultural Study Of The Use Of Art Forms

PDF

A Philosophical Critique Of F. R. Tennant'S Empirical Approach To Theism In The Light Of Current Science

PDF

Human Gsr Classical Conditioning And Awareness Of The Cs-Ucs Relation

PDF

The Development And Pilot Study Of A College Curriculum Unit Which Emphasizes Structure And Inquiry For Instruction On Constitutional Rights

PDF

A Preliminary Study Of The Dimensions Of Future Time Perspective

Asset Metadata

Creator Lockhart, Russell Arthur (author)

Core Title Dominance And Contiguity As Interactive Determinants Of Autonomic Conditioning

Degree Doctor of Philosophy

Degree Program Psychology

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag OAI-PMH Harvest,psychology, experimental

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Grings, William W. (committee chair), Jacobs, Alfred (committee member), Longstreth, Langdon E. (committee member), Robb, J. Wesley (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c18-186907

Unique identifier UC11359762

Identifier 6603821.pdf (filename),usctheses-c18-186907 (legacy record id)

Legacy Identifier 6603821.pdf

Dmrecord 186907

Document Type Dissertation

Rights Lockhart, Russell Arthur

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA