Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Classical Discrimination Conditioning As A Function Of Probability Of Reinforcement
(USC Thesis Other)
Classical Discrimination Conditioning As A Function Of Probability Of Reinforcement
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
CLASSICAL DISCRIMINATION CONDITIONING AS A FUNCTION OF PROBABILITY OF REINFORCEMENT by Harriet Irene Sukoneck 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) June 1971 72-3800 SUKONECK, Harriet Irene, 1945- CLASSICAL DISCRIMINATION CONDITIONING AS A FUNCTION OF PROBABILITY OF REINFORCEMENT. University of Southern California, Ph.D., 1971 Psychology, experimental University Microfilms, A XERO X Company, Ann Arbor, Michigan THIS DISSERTATION HAS BEEN MICROFLIMED EXACTLY AS RECEIVED UNIVERSITY O F S O UTHERN CALIFORNIA TH E G RADUATE SCHO OL U N IV E R S IT Y PARK LOS A NG ELES. C A L IF O R N IA 9 0 0 0 7 T his dissertation, w ritte n by Harriet Irene Sukoneck under the direction o f hex.... D issertation Com m ittee, and approved by a ll its members, has been presented to and accepted by The G radu ate School, in p a rtia l fu lfillm e n t o f require ments o f the degree of D O C T O R O F P H IL O S O P H Y T>n O Dean D ate June .1.971 PLEASE NOTE: Some Pages have indistinct print. Filmed as received. UNIVERSITY MICROFILMS TABLE OF CONTENTS Page LIST OF TABLES...........................................iv LIST OF FIGURES......................................... v Chapter I. INTRODUCTION.................................... 1 II. EMPIRICAL STUDIES............................... 12 III. PROBLEM ELABORATION............................. 19 IV. METHOD...........................................25 Subjects Apparatus Procedure Design V. RESULTS..........................................29 Responses Measured First Interval Response (FIR) — Adaptation First Interval Response (FIR) — Acquisition First Interval Response (FIR) — Extinction Second Interval Response (SIR) — Adaptation Second Interval Response (SIR) — Acquisition Second Interval Response (SIR) — Extinction Third Interval Response (TIR) — Adaptation Third Interval Response (TIR) — Extinction VI. DISCUSSION.......................................75 First Interval Response Second Interval Response Third Interval Response VII. CONCLUSION...................................... 89 APPENDIX A...............................................90 Instructions to the Subject ii Page APPENDIX B...............................................92 Description of Statistical Comparisons SELECTED BIBLIOGRAPHY....................................94 iii LIST OF TABLES Table Page 1. Summary of Analysis of Variance for FIR During Adaptation......................... 32 2. FIR: Mean CRS and Mean Differences in all Groups for Adaptation, Acquisition, and Extinction................... 33 3. Summary of Analysis of Variance for FIR During Acquisition.................... 35 4. Mean Difference (Micromhos) Between CR-l and CR2 of Groups or Combined Groups Used in Comparisons: FIR During Acquisition.............................48 5. Summary of Analysis of Variance for FIR During Extinction......................... 52 6. Mean Difference (Micromhos) Between CRj_ and CR~ of Groups or Combined Groups Used in Comparisons: FIR During Extinction............................. 53 7. Summary of Analysis of Variance for SIR During Acquisition.................... 56 8. SIR: Mean CRS and Mean Difference in all Groups for Adaptation, Acquisition and Extinction.................... 57 9. Summary of Analysis of Variance for SIR During Extinction..................... 59 10. Mean Difference (Micromhos) Between CR^ and CR? of Groups or Combined Groups Used in Comparisons: SIR During Extinction............................. 61 11. Summary of Analysis of Variance for TIR During Extinction..................... 72 12. TIR: Mean CRS and Mean Difference in all Groups for Adaptation and Extinction................................ 72 iv LIST OF FIGURES Figure Page 1. FIR During Acquisition: Trials x Stimulus Effect...............................36 2a. FIR: Mean CR^ and CR2 During Adaptation, Acquisition, and Extinction in Group 1.00-.00................. 37 2b. FIR: Mean CRi and CR2 During Adaptation, Acquisition, and Extinction in Group .80-. 00.................. 38 2c. FIR: Mean CRi and CR2 During Adaptation, Acquisition, and Extinction in Group .60-. 00.................. 39 2d. FIR: Mean CRi and CR2 During Adaptation, Acquisition, and Extinction in Group 1.00-.40................. 40 2e. FIR: Mean CRi and CR2 During Adaptation, Acquisition, and Extinction in Group 1.00-.20................. 41 2f. FIR: Mean CRi and CR2 During Adaptation, Acquisition, and Extinction in Group .70-. 10.................. 42 2g. FIR: Mean CR^ and CR2 During Adaptation, Acquisition, and Extinction in Group . 90-. 30................. 43 2h. FIR: Mean CR^ and CR2 During Adaptation, Acquisition, and Extinction in Group .80-. 20.................. 44 2i. FIR: Mean CR-^ and CR2 During Adaptation, Acquisition, and Extinction in Group .90-. 10.................. 45 3. SIR: Trials x Stimulus Effect................... 56 4a. TIR: Mean CRi and CR2 During Adaptation and Extinction in Group 1.00-.00............................ 63 v Figure Page 4b. TIR: Mean CRj and CR2 During Adaptation and Extinction in Group .80-. 00............................. 64 4c. TIR: Mean CR^ and CR2 During Adaptation and Extinction in Group .60-. 00........... 65 4d. TIR: Mean CRj and CR2 During Adaptation and Extinction in Group I. 00-. 4 0.......... 66 4e. TIR: Mean CR^ and CR? During Adaptation and Extinction in Group 1.00-.20............................ 67 4f. TIR: Mean CR^ and CR2 During Adaptation and Extinction in Group .70-. 10............................. 68 4g. TIR: Mean CRi and CR2 During Adaptation and Extinction in Group .90-.30............................. 69 4h. TIR: Mean CR^ and CR2 During Adaptation and Extinction in Group .80-.20............................. 70 4i. TIR: Mean CRi and CR2 During Adaptation and Extinction in Group .90-. 10............................. 71 vi CHAPTER I INTRODUCTION Traditionally, classical discrimination conditioning employs a procedure whereby two initially "neutral" cues come to elicit differential responses following the repetitive pairing of one cue with reinforcement (CS+) and the other with non-reinforcement CCS-). Historically, two different types of theoretical explanation have been developed to account for the phenomenon. The first theoretical account has been primarily elaborated by Hull (.1943, 1950, 1952) and Spence (1936, 1945, 1960). Within this theoretical framework, dis crimination is viewed essentially as an extension of simple conditioning processes. According to Spence ...discrimination learning is conceived of a cumulative process of building up the excitatory tendency or association between the positive stimulus cue and the response of approaching it, as com pared with the excitatory tendency of the negative stimulus cues, which receives only non-reinforcement, to evoke the response of approaching it. This process continues until the difference between the excitatory strengths of the two stimulus cues is sufficiently great to overshadow always any differences in 1 2 excitatory strength that may exist between other aspects of the situation which happen on a par ticular trial to be allied in their response evoking action with one or the other of the cue stimuli. (1945, p. 254 ) . In addition to excitatory and inhibitory tendencies and i their concomitant generalization gradients, it is assumed j that the magnitude of the inhibitory tendency is less than that of the excitatory tendency. The entire pro cess is seen as a gradual and continuous one. Harlow (1949) pointed out that the Hullians emphasized the I necessity for an historical approach to learning in terms j of past experience. They did not, however, utilize this ; approach to the degree that Harlow did. He emphasized sets or a "learning-to-learn" process. His research with ; i | monkeys and children for the most part, emphasized the | I i : t I j ! fundamental continuity in the processes of simple | ! | conditioning and discrimination. ; ! i | Another type of theory developed principally by j Lashley (Lashley, 1942; Lashley and Wade, 1946) has come to be known as the non-continuity theory. According to Lashley, prior to problem solution the organism attempts systematically to solve the problem, however, these i . | | attempts do not involve lasting associations between J response and cue. Rather the only lasting association formed is between the correct response and the cue. Further, Lashley postualted that "set" determines whether or not certain stimulus elements will elicit responses. I j Thus, differential responding depends on the perceptual j organization of the subject. This organization is the result of comparing traces of both stimuli in the discrimination situation. It is also influenced by inherent organizing tendencies which help to point out the conspicuous differences between the stimuli. From these theories and their extensions, there i ■developed some controversies. Did discrimination learning i involve a continuous or a non-continuous process? Was it a conditioning-extinction process or a relational process? j ! 1 ; Evidence for and against all sides of the question has not j i I ! seemed to resolve the controversies (e.g., Spence, 1941, | 11945; Lashley, 1942; Heron, 1942; Grice, 1949; Harlow, j j !1949; Gynther, 1957). j In 1950, Estes characterized the state of learning I ; theory as in some disorder and wanting of a more I ! quantifiable approach. He wrote I ...the necessary interplay between ; theory and experiment has been hindered, however, by the fact that none of the many current theories of learning commands general agreement among researchers. It seems likely that progress toward a common frame of reference will be slow so long as most theories are built around verbally defined hypothetical con structs which are not susceptible to unequivocal verification. While awaiting resolution of the many apparent disparities among competing theories, it may be advantageous to systematize will established empirical relationships at:a peripheral, statistical level of analysis (p. 45). Estes' 1950 paper seems to mark the beginning of a new view of learning theories characterized essentially by |stochastic and statistical or probabilistic models of jlearning. This view aimed at specifying in quantitative terms, stimulus and response elements through mathematical j i jreasoning. In addition, the concept of variability was ! j included within the definitions of stimulus and response ini order to eliminate some of the problems of ad hoc ! theorizing. Although the model approach has led to the isame sorts of problems Estes characterized in 1950, it has 1 i yielded some fruitful concepts. j ■ Estes (195 0) defined mutually exclusive classes of j jresponses in terms of objective physical criteria. Then | |given the initial probabilities of various responses in a ! given situation, the model generated prediction as to i ! changes in these probabilities as a function of changes in i i values of independent variables. Probability was 5 I operationally defined as the average frequency of occurrence1 of instances of a response class, relative to the maximum possible frequency. Stimuli were conceived by Estes to be populations of independently variable components or stimulus elements. On any trial in the learning situation, only an independent random sample (s) from the population of elements would be effective. In the model, all aspects of |the environment could be considered as an £ population. j jThe number of different populations would depend on the number of independent experimental variations. Such variables as the CS and UCS in a conditioning experiment j [ j jcould thus be conceived of as sources of stimulation and j ‘ treated as separate £ populations. In a sense, this view ! of the stimulus was not far removed from that of Lashley i although Estes treated of no particular psychological theory other than a general S-R approach. It was further j ! assumed that on each occurrence of a response, all elements i ; i of the stimulus sample not already conditioned to that j j i |response became conditioned to it. Further, the | conditioning of a stimulus element to one response j jinvariably involved the breaking of pre-existing I conditioned relations with other responses. In his initial l jpaper (Estes, 1950), Estes made no attempt to deal with |discrimination. ! ! ! I ! ; Soon after, however, Bush and Mosteller (1951a, 1951b)j ; published two articles dealing with their work on a linear ! learning model which was essentially an extension of the stimulus sampling theory. The second of these papers dealt with "A Model for Stimulus Generalization and Discrimination." This model was described in terms of mathematical set theory. It was applied to two types of discrimination problem, a generalization problem following isimple conditioning and a two-stimuli-discrimination i t problem. They stressed that in the two-stimulus discrimination situation, their model assumed that a response was always rewarded or non-rewarded to the same degree on each presentation of the two stimuli. By 1955, Bush and Mosteller had elaborated their basic mathematical system to the point where they could use it in dealing with two and three choice experiments with i ! both contingent and non-contingent reinforcement. They analyzed a variety of discrimination type problems and found extensions of their model to be highly successful in predicting the data of non-contingent situations. Data | i [analyzed included those from situations where various j 1 I [schedules of reinforcement were used following both J 1 I [discriminative stimuli. Some of these experiments [(e.g., Brunswik, 1939) will be discussed in a later section | i ;of the present paper. It is important to note, however, 'that in all cases but one the class of possible responses i yas limited to two. The exception contained a set of J j j [three possible responses. In all the experiments ! 7 analyzed, responses were always of an all-or-none variety. Following these writings in the early 1950's, Estes and Burke and Bush and Mosteller remained active in i !elaborating their models to fit specific learning I jsituations and to predict data from specific learning | ;types of experiments (Estes and Burke, 1955a, 1955b; iBurke and Estes, 1957; Estes, 1957a, 1957b; Bush and Mosteller, 1959; Bush and Estes, 1959). They were joined by a core of other mathematical model builders in the 1950's and 1960's (e.g., Atkinson, Bower and Crothers, 1965; Arrow, Karlin and Suppes, 1960), who along with them have continued to extend and apply various mathematical models to a host of learning situations. I i i One notable deviation from the original linear type models which viewed the probability of correct responses i i I ias a continuous function of successive trials with j 'reinforcement was the utilization of Markov or two-stage and i I jthree-stage discrete models (Estes, 1959b). This kind of l I imodel became necessary when it was realized that the lassumption of large stimulus populations did not always i ;hold. In such a case, i.e., where stimulus populations are small, random fluctuations in trial to trial stimulus samples are no longer negligible and the probability function for responses can no longer be treater as jrelatively linear. Another deviation, one more relevant to the problems of this paper came from Suppes (Suppes, 1959; Suppes and Rouanet, 1964) . He was the first to attempjt to analyze a linear model for simple learning with a continuum of responses rather than the discrete responses analyzed previously. Soon after, he and Rouanet (Suppes and i : Rouanet, 1964) attempted to test several stochastic i I learning models for a continuum of responses in a simple i I discrimination experiment. Of necessity, in the testing of their models they used a very simple type of idiscrimination. Essentially the subjects' task was to i i ! predict by means of a pointer, where a spot of light would appear on the circumference of a circle each time one of j : t four stimulus lights came on. Each stimulus light i corresponded with a different probability distribution of reinforcement on the circle and there was "no apparent j joverlap among the discriminative stimuli or among the four I |different reinforcement regions, and in fact each region isj | separated from the next by a significant zone of non- j ireinforcement." (p. 318). In other words the discrimination ! situation was deliberately contrived (more so than usual in j I psychology) to facilitate mathematical analyses of the | i . ! ; response data. By so doing, the discrimination situation i j i was also quite different from that employed by most other j ! i | experimenters who deal with discrimination. Suppes and I i j Rouanet concluded their elaboration of the various models i 9 by noting that "there is clearly a significant gap between theory and observation, and it appears that it will be necessary to revise the theory proposed in this paper in order to close this gap adequately" (p. 347). This last statement is strangely similar to the one by Estes (1950) quoted in this paper. Another model for discrimination learning which is more typical of the types of mathematical models that have been developed (Estes, 1959; Restle, 1957, 1958) is that of Burke and Estes (1957). It is more typical in the sense that it again applies particularly to certain especially simplified experiments and its purpose is not to provide a generally adequate account of discrimination learning. It ; i deals with discrete and voluntary responses based on large I j numbers of trials. That is, the majority of mathematical 1 models for discrimination deal with the two-choice j | probability learning situation. In this type of experiment j lone of two signal events (A^ and A2) appears on each trial, j jThe subjects' task is to predict which of two "reinforcing" ] ! events (Eq or E2) he expects to appear next. E^ and E2 jappear in accordance with a predetermined random schedule {with probabilities of tTi and TT2' T^e probabilities of I jEj and E2 (Tr 1 and7> *2) can be made to vary for different jgroups of subjects and thus serve as the independent i jvariable. In addition the proportion of times that A^ and 1 IA2 appear may also be manipulated. io ! ! 1 The mathematical models have been found to be somewhat; successful in predicting the data of particular experiments. Unfortunately, it is often the case that a model must then be altered to fit the data of different experiments. In general the models have not been able to account for errors that subjects make when predicting patterns which can be : perfectly learned. Therefore, in recent years, the role of j runs of the reinforceing and stimulus events has been examined (e.g., Gambino et al., 1967). In addition a certain amount of "psychologizing" has come back into model makeing such that models no longer necessarily remain strictly jmathematical (e.g., Lordahl, 1968; Blair and Peterson, 1968; Halpern et al., 1969). In terms of the present paper, the models make two • . ! contributions. Within their sphere of two-choice j probability learning situations, they serve to point out |the role of varying probabilities of "reinforcement". ;This variable, while almost a necessary independent j t (variable for the purposes of demonstrating mathematical i I .models, has been largely neglected in the traditional | j j iclassical discrimination conditioning studies. It has been t 'expecially neglected in conditioning studies employing i I autonomic response variables such as heart rate, blood j I ; 'volume and galvanic skin resistance. j | It is the purpose of this paper to explore GSR I jdiscrimination conditioning as a function of varying | 11 probabilities of reinforcement. Before elaborating further on the specific problem area of this paper, several empirical studies, which sought to assess the effects of varying probability of reinforcement in discriminative situations will be reviewed. Most of these studies were not done within the context of mathematical theory nro Skinnerian schedules of reinforcement. Only one deals with a continuous or graded autonomic response variable. CHAPTER II EMPIRICAL STUDIES Egon Brunswik (1939) was one of the first to feel "the need to study ambiguity experimentally". He chose to do this in the context of a choice situation by varying the relative frequency or probability of reward on the two isides of a T-maze. His subjects were five groups of rats. I I !ln his 100:0 groups, one side of the maze was always rewarded, the other side never. For this group, "...the effect on wither side was predictable with certainty." l !A second group of rats was rewarded on one side of the maze i I ;3/4 of the cases, with the other side being rewarded m the | remaining 1/4 cases. This group was labeled 75:25. In a j i ! similar manner three other groups were labeled 67:33, 1 t I 100 : 50 and 50:0. | Brunswik found that "discrimination of probabilities |takes place definitely for all of the groups except |training type 67:33". Looking at the rank order of scores j I among the groups Brunswik concluded that discrimination was l primarily an increasing function of an increasing difference! of probabilities. He also found that when the difference between probabilities was held constant ( as in comparing groups 50:0, 75:25 and 100:50) there was some influence of j jthe ratio of probabilities, although not significantly so. | 12 ! Brunswik was also quite specific in differentiating between frequency and frequency-configuratiori. ...the traditional notion of 'frequency' or of 'relative frequency' as a factor influencing learning has to be amplified to that of a 'frequency ] pattern,' of which isolated fre- j quencies are but special aspects ! applicable under limited conditions | of a non-ambiguous training (p. ^79) . jBy non-ambiguous training, Brunswik meant conventional i jl00:0 groups in which the subject can predict with cer- l tainty the outcome of any trial. With respect to ambiguity of the effects of a response, Brunswik found no indication jthat certainty of reward or non-reward as in a 100:0 group jwas a determinant of discrimination. I It is important to note that in Brunswik's experiment j a choice-point correction procedure was used such that when j 'an incorrect choice was made, the trial was continued until j the subject made the correct response. j Robbins (1969) has noted that in experiments of this j i J jsort when subjects are allowed to retrace the incorrect j maze arm, some experiments have reported probability j hatching (as in Brunswik's experiment), while others have i [reported overshooting of the matching value. Robbins jtherefore did two experiments, in each of which, two groups, 1,70:30 and 60:40 were used. Using two different correction j l i procedures in the T-maze situation, Robbins found over- j | jshooting in both cases rather than a strict matching. No | j ! [account of overshooting was given in this paper since the 14 main concern was a comparison of two different statistical models to see which was more predictive of the data. In a comparison of within- and between-subjects designs, Sadler (1968) used amount of salivation as the CR. For his three within-subjects groups, he employed a I discrimination paradigm with percentages of rewards I following the two CSs of 100:75, 100:50 and 100:0. In ;terms of procedure and response used, this experiment comes jclosest to that used in the present study. Using dogs as I subjects, auditory CSs and a food UCS, Sadler found significant discrimination only in the 100:0 group, but evidence for discrimination in the other two groups as well. A variety of discrimination type experiments in volving probability learning has been done by those interested in applications of statistical models. As noted before these ordinarily use voluntary motor responses |of an all-or-none sort. As an example, one such experiment |was done by Myers and Cruse (1968). They employed groups 'differing both in the number of presentations of one of the ; CSs and in the probability of reinforcement for the other i jcs (7r2). The probability of reinforcement for the first |CS ( 7 ? * ] , ) was held constant across groups. Thus they i ! employed nine groups with 7f'i=.85 for all groups, 1 ;TT2=-85, .50 or .15 and the proportion of CS-^ presentations ;.75, .50 or .25. i j 15 i They found that response probability on one CS was a function of the event ("reinforcement") probability on the other CS and of the relative frequencies of the two CSs. When both CSs were presented an equal number of times (.50) the probability of CR^ on a CS^ trial was lower when | TT =.50 than when 7 7“ 9=. 85 or .15. For convenience, the I 2 ^ j !terms CS and CR have been used here although this is not I the terminology used nor is there any "conditioning" truly I involved. They reported that their results supported I Atkinson's observing response model, although some of the I data such as "overshooting" and "undershooting" could not be accounted for by the model. They concluded by noting ;that to account for two-choice discrimination learning, phenomena such as overshooting, strong generalization from T2 (comparable to CS2) trials when 77*2=-50, and excellent ; discrimination when must be noted. | Although there are some similarities between the 1 !two-choice probability learning experiments and classical |conditioning which have been instrumental in forming I the basis for the present paper, the differences are too igreat to make many of these studies directly relevant. !Rather they serve as background evidence to point up the jtremendous interest in the probability of reinforcement |variable. j Recently two studies (Newman, 1967; Newman and jWoodhouse, 1969) which were concerned with quantitative ] models in the discrimination situation have gone beyond the choice situation to an eyelid conditioning paradigm. In the first study, ten groups were employed, differing in the proportion of trials on which each of the CSs was reinforced. CSs were tones and the UCS was an air puff. |The ten groups had percentages of reinforcement as follows: j j100:0, 75:0, 50:0, 25:0, 100:50, 100:75, 75:25, 75:50, j land 50:25. To assess discrimination each experimental | groups was compared to a 50:50 control group. Only the 100:0 and 75:0 groups significantly differentiated I (between the two CSs. The 50:0 groups approached the .05 ! level of significance. However, by noting Newman's graphs of percentage of CRs to CS-^ and CS2, it appears that there S was some amount of discrimination in all but the 75:25, 50:25 and 50:50 groups. Further, Newman's method of j analysis by comparison with the 50:50 group may have been i :spurious since this groups itself showed a positive ( (difference between CSs. No tests of, nor corrections for, j ■ i jthis difference are stated in the paper. Newman also | ’ j found that CR^ was directly related to TTj when T7“2=0 or *25 !but not when Tr 2=.50, although the latter case may be due | ! jto "experimental imprecision" (Newman's term). j I Newman gave a general description of his acquisition j i j idata by noting that both CRj_ and difference scores were a | j 1 j direct function of 7 T ] _ when 77* 2=0. When 77* 2>0 there was ! 17 either no discrimination or CRs were a joint function of 77" i and 7T2* He suggested that any model which would describe the necessary conditions for differentiation would have to state that at least one CS have 77 * value of zero. In analyzing the extinction data, Newman concluded that a partial reinforcement effect (PRE) existed between groups (i.e., when 7T1 and 7^ are averaged within a group) for all but the 1.00-.00, 1.00-.25, 1.00-.50, and 1.00-.75 groups. Within groups (i.e., between stimuli) no PRE was obtained, with the tendency being in the opposite direction to PRE for seven of the ten groups. In a later study (Newman and Woodhouse, 196 9) a similar arrangement to that of Newman's first study was used. Here, however, standard differential training (771=1 .00 and77~ 2=-00) was given prior to shifting 2 to a value greater than zero. Control groups received similar training to the experimental groups of the 1967 study except that acquisition was preceeded by random unpaired trials of CSi, CS2, and the UCS. It was found that standard pre-training did facilitate the more difficult discriminations and the degree of differentiation among groups was as follows: (1.00-.00)>(1.00-.25)>(1.00-.50*1.00-.75sl.00-1.00),p<.01. It is of interest that the performance of the control groups paralleled the performance of their corresponding experimental groups, although that of the control groups I 18 was attenuated. Discrimination was significant in the 100:25 control group, a result not found in the earlier study. In addition, the graphic representation of the 100:50 groups seemed to show some discrimination. The studies reviewed all deal generally with jvarying the proportion of reinforcement in a discriminative ^situation. Similar studies are noticeably lacking the ! GSR conditioning literature. The present paper will attempt to alleviate this lack, by using probability of reinforce ment as an independent variable in a classical discrimination conditioning study of the galvanic skin response (GSR). CHAPTER III PROBLEM ELABORATION GSR discrimination experiments typically investigate cognitive variables such as instructions (e.g., Baer and Fuhrer, 1968), parameters of intensity such as CS and UCS intensity (e.g., Zeiner, 1968), and parameters of time such as inter-stimulus interval, CS and UCS duration, inter-trial interval, etc. (e.g., Hunt, 1967; Cermak and Wickens, 1969). These variables are of basic interest for the process of discrimination conditioning. In the probability learning situations discussed earlier, another variable of special interest involves the ■relative frequency of stimulation. Thus, the probability or relative frequency of reinforcement and the relative frequency of the cue stimulus presentations have been jmanipulated and investigated. An interest in this type of I jvariable and its effects on discrimination has been i Hacking, however, in the area of GSR discrimination {conditioning. Although there are no readily apparent reasons for this lack in GSR experiments, the research i ! in other areas such as probability learning situations, shows this kind of variable to be important to the process i |of discrimination. I ! I ... _......... I?.........-......... -..-.... ----- ; 20 I The purpose of this dissertation will be to examine the role of probability of reinforcement in a classical discrimination conditioning study of the GSR. jProbability of reinforcement means the relative frequency of presentations of the UCS following each of the two conditioned stimuli. Various problems will be considered. One problem I involves the question of whether, as is traditional in GSR studies, one of the conditioned stimuli must have a probability of zero reinforcement in order for discrimina tion to occur. Another problem concerns the effectiveness i |of various schedules of reinforcement. Typically in a discrimination experiment, the positive CS may be followed by the UCS on all trials or only on a partial schedule ranging from 50% to 100%. Often the specific percentage of reinforced stimulus presentations depends on the number of "test" trials chosen by the experimenter. The number of I "test" trials in turn seems to be arbitrary, often chosen I !so as to be most amenable to future analyses (i.e., i |comparisons of CR+ and CR-). The question to be answered i in this study asks which schedules of reinforcement lead to i jgreatest differential responding. | Previous research (outlined earlier) suggests that there is some combination of reinforcement probabilities for the two discriminative stimuli which leads to no jdifferential response. Which combinations do not lead to 21 discrimination seems to depend on the task, species, response measure and number of trials. In addition, Newman's (1967) results suggest that discrimination may be related to regularity of UCS presentation following at least one of the CSs. Regularity is here defined as the case where probability ( ) equals .00 or 1.00 and complete uniformity of reinforcement or non-reinforcement exists. On the other hand, Brunswik's (1939) experiment suggests that discrimination may be a function primarily of the difference between CS^ and CS2 in terms of numbers of reinforcements. One purpose of this study is to see if either of these suggestions, in the form of hypotheses, can help to account for discrimination conditioning of the GSR. One hypothesis, which will be called the difference hypothesis, suggests that discrimination is a function of the difference in relative numbers of UCSs following the two conditioned stimuli. Thus, if two stimuli, CS^ and CS2 have probabilities, an<^ 2' being followed by the UCS, the amount of discrimination will increase as 1 minus 2 increases. In a sense, this hypothesis merely invokes principles of simple conditioning. Thus, it is akin to the Hullian view of discrimination. To put it simply, the more a stimulus-response sequence is followed by reinforcement, the greater the conditioning. Given two stimuli and their corresponding responses, the 22 one reinforced relatively more often will lead to greater response strength. A second hypothesis, which will be called the certainty hypothesis, suggests that discrimination is a function of regularity of reinforcement associated with each CS. It is only where probability -equals .00 or 1.00, that a subject can ever be certain that he will or will i I not receive the UCS on any one trial. For all other (probabilities, there is no way for the subject to discover whether a particular trial will include the UCS. The assumption is that it is less difficult to learn about the (difference between two stimuli when at least one of them I ihas a regular pattern of reinforcement. An extension of (this hypothesis, to be called the zero-regularity (assumption, might consider that where the regular probability equals .00, discrimination will be easier or (greater than where the regular probability is 1.00. Once (the case where the regular probability equals .00 is I (learned, the CS associated with that probability produces jan expectation of "safety". It comes to signal the |occurrence of no shock. It is assumed that the subject i (comes to have a kind of "relaxation" response to this i I 'stimulus. Where77 = 1. 00 , the subject is exerting effort (following the CS with regular reinforcement in addition (to exerting effort in connection with the "irregularly 23 reinforced" stimulus-response sequence. This "effort" state may be equated to arousal in which case there may be greater generalization (actually sensitization) between the stimulus with regular probability and the stimulus with irregular probability. That is, the subject is sensitized to respond to all stimuli when 7 r f or the regular stimulus equals 1.00. However, when 77*for the regular stimulus equals .00, sensitization due to arousal does not occur and the subject may respond to each of the two conditioned stimuli in a more discriminatory manner. Thus differential responding in the case where the regular probability is 1.00 may be less than where the regular probability is .00. In order to examine the independent variable of probability of reinforcement in this study, various groups differing in 7T\ and 77“ 2 will be employed in a discrimination conditioning paradigm. The groups will then be compared in terms of magnitude of differential response. Prob abilities, 7T]_ and 7T2 will be manipulated to yield the following experimental groups: 1.00-.00, 1.00-.20, 1.00- .40, .90-.30, .80-.20, .90-.10, .70-.10, .60-.00, and .80-.00. The uncertainty hypothesis might be tested, for example, by comparing groups .80-.20 and .60-.00 or groups .90-. 10 and .80-. 00. That is, with the ( 7 -7T*2) 24 difference held constant the regularity principle can be examined. The zero-regularity assumption might be examined by looking at a comparison of the 1.00-.40 group to the .60-.00 group, for example. The difference hypothesis might be examined by comparing groups such as .80-.00 to .60-.00, where regularity of 7 equal but the groups differ in terms of the CITi - difference. If the difference hypothesis holds, one would expect the .80-.00 group to show greater differential responding than the .60-.00 group. Similarly the 1.00-.20 group should show greater discrimination than the 1.00-.40 group. Finally the .80-.00 group would be expected to show greater discrimination than the 1.00-.20 group, according to the zero-regularity assumption. CHAPTER IV METHOD Subjects Subjects were 20 females and 70 males. Most were students taking an introductory psychology course at the University of Southern California. These subjects participated in return for credit in their classes. The rest of the subjects were volunteers paid $2.00 for the one-hour session. All were recruited from the campus area. Apparatus The conditioned stimuli were a blue light and a yellow light of 4 seconds duration presented by a Grason- Stadler (Model E 4580) multiple stimulus projector, placed about three feet in front of the subject. The uncon ditioned stimulus was a .5 second DC shock from a Grass S-5 stimulator delivered to the volar surface of the subject's right forearm through two circular silver electrodes 17mm. in diameter. Stimulus duration and interstimulus interval (4 seconds) were controlled by two Hunter 111C timers. Intertrial intervals of 30, 35, 40, and 45 seconds were controlled by a punched tape feeding through a Gerbrands programmer. 25 26 Skin resistance was recorded through 1/2 by 5/8 inch silver electrodes, coated with NASA formula electrode paste and taped to the first and third fingers of the subject's left hand. Resistance changes picked up by the GSR recording electrodes were fed through a Darrow-type Wheatstone bridge circuit to an Offner type R polygraph. Current through the electrodes was 45 microamperes. Procedure Subjects were randomly assigned to experimental groups as they came to the experiment with the restric tion that the proportion of males to females be equal across groups and presentation orders. When a subject arrived for the experiment, he was seated in a comfortable chair. The experiment was briefly explained to him (see Appendix A). GSR and shock electrodes were attached to the subject. A shock work-up procedure to find the level of shock to be used for that subject was explained. The experimenter then left the room to administer the shock work-up. E returned to the room and instrucred the subject about the experiment. Subjects were told that they would see different colored lights on the small screen before them and that these lights might sometimes be followed by shock. E then left the room and the experiment began. Each S received 8 adaptation trials (4 of each CS) 27 with neither CS followed by shock. Without interruption, acquisition trials were then begun. Each CS was presented 10 times with CS^ and CS2 followed by shock according to the group that £ was in. Finally 20 extinction trials were given (10 presentations of each CS) in which neither CS was followed by shock. For all trials', CS duration and inter stimulus interval were 4 seconds. The duration of the UCS was .5 seconds. Intertrial intervals varied randomly between 30, 35, 40, and 45 seconds from CS onset to CS onset. Design Each of nine groups of ten subjects received 24 presentations of the conditioned stimuli. Stimuli were presented in a random order with the restriction that each stimulus not appear more than twice consecutively. Stimulus presentations were classified into three stages: adaptation, acquisition and extinction. During adaptation each stimulus was presented 4 times, during acquisition, 10 times and during extinction, 10 times. Groups were differentiated during acquisition according to the relative probability of reinforcement following each CE}7T^ and 7^. These probabilities were varied across groups during acquisition to yield the following experimental groups: 1.00-.00, .80-.00, .60-.00, 1.00-.40, 1.00-.20, .90-.10, .80-.20, .70-.10, .90-.30. 28 Half of the subjects in each experimental group received one order of stimulus presentations (Order A) while the rest received a different order (Order B). The purpose of this was to minimize order effects. Within order A and order B, shock trials were randomly assigned following the appropriate probabilities. CHAPTER V RESULTS Response Measured Several electrodermal measures, temporally differen tiated, were defined as the dependent variables. The first to be considered was the largest decrease in skin resistance occurring between 1.0 and 3.25 seconds after CS onset. It is termed the first interval response (FIR). The second measure was the largest decrease in skin resistance to occur between 3.25 and 5.25 seconds after CS onset. This is termed the second interval response (SIR). The largest decrease in resistance to occur between 5.25 and 7.0 seconds after CS onset was termed the third interval response (TIR). This response was only analyzed for the adaptation and extinction periods when both CSs were not followed by shock. The FIR and SIR were recorded and analyzed for every trial during adaptation, acquisition, and extinction. All resistance change scores have been converted to square root of change-in-conductance scores and are expressed in micromhos. In all groups and for all analyses, CRi represents the conditioned response following the stimulus with the higher probability of reinforcement (CS^) within 29 30 a group. Although analysis of the FIR is most meaningful in terms of stable results as are usually found in discrimination studies, the SIR and TIR were analyzed in a subsidiary attempt to explore possible applications of the difference and uncertainty hypotheses to these responses. It should be noted that the SIR is a somewhat arbitrarily defined response since in each GSR study its temporal limits are conditional on the particular ISI used. The TIR as noted previously also presents problems in discrimination studies since it can only be compared on adjacent CS^ and CS£ trials, leaving the choice and number of CS test trials to be determined by the experimenter. That is, test trials will vary from experiment to experiment and except for instances of experimental replication different results may be due to different paradigms. The TIR can be conveniently studied in terms of comparing CR-^ and CR2 during extinc tion but this type of comparison tells nothing about the development and acquisition of discrimination. In addition whether or not there are test trials during acquisition (for CS2) will determine differential extinction across groups. That is, where 7^ 1•0° various effects of partial reinforcement may differen tially influence extinction. Groups differing in numbers 31 of test trials for CS2 would thus be expected to show differential courses of extinction. Such would be the case particularly in this study since, in effect, each group differed in the number of "test trials" (CS2 trials not followed by the UCS). It should also be noted that what the TIR, measured in the period of UCS omission, represents has been open to question. This response may be nothing more than an OR to the absence of the UCS, i.e., a response to change in stimulus conditions. If this is the case, then in the present study, groups will differ with respect to degree of change in stimulus conditions and comparisons in terms of conditioned discriminations may not be as meaningful as other comparisons. First Interval Response (FIR) — Adaptation The adaptation data were analyzed through an analysis of variance (AOV) with trials, stimuli and groups as the main independent variables. Results are summarized in Table 1. This analysis was done to determine whether or not groups and stimuli were equivalent with regard to magnitude of responses before acquisition. The trials effect was significant at the .001 level (F3 243=15•80) with response magnitudes decreasing over trials. The stimulus effect was significant at the .025 level (F^ sl=5.93) with mean CR2 magnitude =.77 and mean 32 TABLE 1 SUMMARY OF ANALYSIS OF VARIANCE FOR FIR DURING ADAPTATION Source SS df MS F P G 6.13 8 .77 .65 s/ G 94.96 81 1.17 T 5.51 3 1.84 15.80 .001 GT 2.74 24 .11 .98 Ts/G 28. 26 243 .12 C .75 1 .75 5.93 .025 GC 1.17 8 .15 1.16 Cs/G 10. 20 81 .13 TC .37 3 .12 1.08 GTC 2. 24 24 .09 .81 TCs/G 27.93 243 .11 Note: For all AOV's, G=Group Effect, T=Trial Effect C=Stimulus Effect, s=Subjects CR^ magnitude = .70 (see Table 2). This small difference was significant due to an extremely small error variance. While it was to be expected that there would be no signi ficant difference between responses to the two CSs during adaptation, the difference obtained represents a bias 33 TABLE 2 FIR: MEAN CRS AND MEAN DIFFERENCES IN ALL GROUPS FOR ADAPTATION, ACQUISITION, AND EXTINCTION Adaptation Acquisition Extinction Group CR1 CR2 dif CEi cr2 dif CR1 cr2 dif o 0 • 1 o o • .75 .79 -.04 .96 .76 .20 .74 .70 .04 • 00 0 1 • o o .51 .72 -.21 .82 .67 .20 .66 .43 .23 o o • 1 o VO • .73 .70 .03 .89 .74 .15 .92 . 63 .29 0 • 1 o o • .77 .90 -.13 1.00 .91 .08 .76 .66 .10 .00-.20 .63 .63 .00 .78 .71 .07 .56 .53 .03 0 1 • t —■ O .83 . 82 .01 .98 .88 .10 .85 .78 .07 .90-.30 .58 .64 -.06 .81 .79 . 02 .77 .72 .05 .80-.20 .74 . 79 -.05 1.00 .96 .04 1.08 .95 .13 .90-.10 .78 .94 -.16 1.05 .82 .13 .69 .58 .11 Mean .70 .77 -.07 .92 .80 .12 .78 .67 .11 34 against eventual discrimination where the expectation is that CR-^ will be greater than CR2- In effect then, such discrimination that later occurs is in opposition to the initial CR bias. First Interval Response (FIR) — Acquisition Acquisition data were initially analyzed using an analysis of variance with trials, stimuli and groups as the main independent variables. Results are summarized in Table 3. The trials effect was significant at the .001 level (F9^729=7.14) with CR magnitudes increasing to an optimal level at the third presentation of each CS and then decreasing. The main stimulus effect was significant at the .001 level (Fg 729=44.89) reflecting discrimination across all groups with mean CR^=.92 and mean CR2=.80 (see Table 2). The trials x stimulus effect was signif icant at the .01 level (Fg 729=2.8!) and is presented graphically in Figure 1. Finally, the groups x stimulus effect approached the .05 level of significance (F8,81=1* 92, .10<p<.05). In general, the effect here represents good discrimination in groups 1.00-.00, .90-.10, .80-.00, and .60-.00 with less discrimination in the other groups (see Table 2). The groups x stimulus x trials effect is significant at the .05 level (F72,729=1•20) and is represented graphically in Fig. 2a-2i. 35 TABLE 3 SUMMARY OF ANALYSIS OF VARIANCE FOR FIR DURING ACQUISITION Source SS df MS F p G 13.27 8 s/G 244.00 81 T 10.29 9 GT 10.78 72 Ts/G 116.85 729 C 6.08 1 GC 2.08 8 Cs/G 10.98 81 TC 3.06 9 GTC 10.47 72 TCs/G 88.12 729 1.66 .55 3. 01 1.14 7.14 .001 .15 .93 .16 6.08 44.89 .001 .26 1.92 .10 .14 .34 2.81 .01 .15 1.20 .05 .12 1.10- 1.00' .90- N/a C .80 .70' .60 / r \ \ \ CR 1234567 89 10 TRIALS Figure 1. FIR DURING ACQUISITION: TRIALS X STIMULUS EFFECT LO a \ * t o o A e o u V •H £ a • r 4 P 5 w o 1.20 1.10 1.00 90 80 70 60 CR 0f=-1.00} cRr &r= .ool 50 1-2 3-4 5-6 7-8 9-10 Acquisition 1-2 3-4 Adapt. Extinction TRIALS O J •>] Figure 2a. FIR: Mean CR and CR2 during adaptation, acquisition, and extinction in group 1.00-.00. GSR i n micromhos 1.00 80 70 50 40 30 — CR. ( 7 7 ^ = .80) cr; t ot=.oo) 1-2 3-4 Adapt. 1-2 3-4 5-6 7-8 9-10 Acquisition 1-2 3-4 5-6 7-8 9-10 Extinction TRIALS Figure 2b. FIR: Mean CRj and CRo during adaptation, acquisition, and extinction in group .80-.00. Co CO GSR i n micromhos VAC 1.10 - 90 - 80 70 60 50 1-2 3-4 Adapt. Extinction Acquisition TRIALS .60) .00) Figure 2c. FIR: Mean CRj and CR, during adaptation, acquisition, and extension in group .66-.00. u> so GSR i n micromhos Vac 1.20 1.10 1.00 90 70 60 50 CR., Q7"=l. 00] CR^ 0T= .40) 1-2 Extinction Acquisition Adapt TRIALS o Figure 2d. FIR: Mean CR^ and CR during adaptation, acquisition, and extinction in group 1.00-.40. GSR i n micromhos V a c 1.00 90 40 — CR CR 30 1-2 3-4 Adapt, Extinction Acquisition TRIALS 1.00) .20) Figure 2e FIR: Mean CR and CR during adaptation, acquisition, and extinction in^-group l?00-.20. GSR i n micromhos 1.10 1.00 .90 .80 .60 .50 CRn (TT=.70) CR? (7F=. 10) .40 1-2 3-4 5-6 7-8 9-10 Extinction 1-2 3-4 Adapt. Acquisition TRIALS Figure 2f. FIR: Mean CR.^ and CR2 during adaptation, acquisition, extinction in group .70-.10 GSR i n micromhos 1.00 80 .70 .60 50 CR ( 7 7 = . 90) CR^T (77=. 30) .40 Extinction Acquisition Adapt TRIALS Figure 2g. FIR: Mean CR^ and CR2 during adaptation, acquisition, and extinction in group .90-.30. GSR i n micromhos 8 1,30 1.20 1.10 1.00 90 80 70 — CR-, (7T= CR„ (JT= 60 1-2 3-4 5-6 7-8 9-10 Acquisition 1-2 3-4 Adapt. Extinction TRIALS .801 ,201 Figure 2h. FIR: Mean CR^ and CR2 during adaptation, acquisition, and extinction in group .80-.20. GSR i n micromhos 1.30 1.20 1.10 1.00 80 60 50 1-2 3-4 Adapt. Extinction Acquisition .90) . 10) TRIALS Figure 2i. FIR: Mean CRj and CR2 during adaptation, acquisition,.and extinction in group .90-.10. ( j i 46 While the results of the analysis of variance are in the predicted direction showing discrimination over all groups combined and differences in discrimination among groups, it was felt that further analyses were necessary in order to test the specific hypotheses of this paper. Planned comparisons were decided upon before the data were collected. Since assumptions about homogeneity of variance did not always hold among all groups, non-parametric comparison tests were employed in a more detailed analysis. All comparisons were planned except for those specifically noted as "post hoc". For the first interval response (FIR) during acquisition (and for all responses over both acquisition and extinction later on) trials were combined into three groups representing mean CRs (FIRs here) for early (trials 2 through 4), middle (trials 5 through 7), and late (trials 8 through 10) trial blocks. Trials were divided into blocks in this manner in an attempt to note differential development of discrimination among the nine groups. It was felt that if some groups began to discriminate or stopped discriminating before other groups, a division into early, middle and late blocks would permit inspection of the phenomenon as well as permit the two hypotheses to be tested during different periods. 47 Within each trial block a Mann-Whitney U test was carried out to compare various groups as a test of the hypotheses presented at the beginning of this paper. Two of these comparisons test the difference hypotheses and two test the uncertainty hypothesis. All U tests are one-tailed Comparison 1 compares groups in which (7ri-7T2=-80) to groups in which (iri-Tr2=. 60) where no groups have a "regularly reinforced" CS, i.e., where neither ^ nor 2 equals .00 or 1.00. This tests the difference hypothesis. Specifically it compares group .90-.10 to groups .90-.30, .80-. 20, and .70-.10 combined via a Mann-Whitney U test. For the early trial block Uio,30=77, .10. For the middle trial block Uiof30=82j ^ 05, while for the late trial block U^g^30=135,p >.10. As can be seen in Table 4 for comparison 1, group .90-.10 showed significantly greater discrimination than the other groups combined during the early and middle trial blocks, thus supporting the difference hypothesis for these periods. Comparison 2 compares groups in which (Tr^-tT)3.80) to groups in which (iri-ir2= .60) where all groups have one "regularly reinforced" CS, i.e., whereTr^ or7T 2 equals .00 or 1.00 respectively. This again tests the difference hypothesis. Specifically, groups 1.00-.40 and .60-.00 48 combined are compared to groups .80-.00 and 1.00-.20 combined via a Mann-Whitney U test. For the early trial block, U20, 20=-*-81• • For middle trial block, U20,20=189,p>.05. In the late trial block, U2Q 20=171, p> .05. Thus for all periods there is no significant difference in the means of the groups being compared. TABLE 4 MEAN DIFFERENCE (MICROMHOS) BETWEEN AND CR2 OF GROUPS OR COMBINED GROUPS USED IN COMPARISONS: FIR DURING ACQUISITION Trial Block Comparison __________________ Early Middle Late 1 .90-.10 • . 30* .13 .90-.30, 0 1 —1 1 0 0 CM 1 O 0 0 .06 . 02 .03 2 1.00-.20, .80-.00 .18 .13 .17 1.00-.40, .60-.00 .12 .04 .24 3 .90-.10 .32 .30 .13 1.00-.20, • 0 0 0 1 • 0 0 .18 .13 .17 4 1.00-.40, .60-.00 .12 .04 *24* .80-.20, .90-.30, .70-.10 .06 .02 .03 5 .80-.00, .90-.10, 1.00-.20 . 23* .60-.00, 1.00-.40, .90-.30, .70-.10, .80-.20 6 O O • 1 O CO • 1.00-.20, .60-.00, .15 1.00-.40 0 1 —1 « 1 0 O ' * • .90-.30, .70-.10, .12 .80-.20 * Comparison significant at .05 level ** Comparison significant at .01 level 49 Comparison 3 compares "regularly reinforced" to "irregularly reinforced" groups with the eiri-w2) difference held constant at .80. This tests the uncertainty hypothesis. It compares groups .80-.00 and 1.00-.20 combined to groups .90-.10 through a Mann-Whitney U test. The early trial block yields U10,20=82' P^* 05* For the middle trial block Uio,20=83' p>.05, and for the late block, U^g 20=97,p>.05. This comparison yields results which do not support the uncertainty hypothesis. Comparison 4 compares "regularly reinforced" to "irregularly reinforced" groups with the difference held constant at .60. This also tests the uncertainty hypothesis. It compares groups .60-.00 and 1.00-.40 combined to groups .80-.20, .90-.30, and .70-.10 combined using the Mann-Whitney U. For the early trial block, U2Q .10>p>.05. For the middle trial block U2o, 30=2^1' P ^ an<3 f°r t^ ie late trial block u20 3o=-*-98, p £ . 05. Thus in the early and late trial blocks the regularly reinforced groups combined show significantly greater mean discrimination than the irregularly reinforced groups (.80-.20, .90-.30, and .70-.10) combined. Descriptive data used in the preceeding Mann-Whitney U tests can be seen in Table 4. In addition a description 50 of the four comparisons can be found in Appendix B for later reference. Two additional comparisons suggested by inspection of the date (i.e., post hoc comparisons) were also done. One was an overall test of the difference hypothesis in which all groups were (^-^2=* 60) were compared with all groups where (7Tx-W*2=* 8°) for the early trial block (since initial discrimination seems to occur in this block). Thus the comparison was between groups .80-.00, .90-.10, and 1.00-.20 combined and groups .60-.00, 1.00-.40, .90-.30, .70-.10, and .80-.20 combined. The Mann Whitney t^g ^0=1385,p<.001 reflecting significantly greater discrimination in groups where ( 7 T^-7 T2=. 80) . The second post hoc comparison was an overall test of the uncertainty hypothesis in which all "regularly reinforced" groups are compared to all "irregularly reinforced" groups. Specifically, groups .80-. 00, 1.00-.20, .60-. 00, and 1.00- .40 combined were compared to groups .90-.10, .90-.30, .70-.10, and .80-.20 combined. The Mann-Whitney U4q^4q=881 10 reflecting no significant difference between groups due to uncertainty or "regularity" vs "irregularity". Descriptive data for these comparisons are given in Table 4 Although the post hoc comparison which tests the difference hypothesis may appear to be confounded with the uncertainty principle due to the particular groups 51 being compared, it yields a significant difference while the second comparison, an actual test of the uncertainty principle is not significant. This suggests that, as tested, the difference hypothesis is supported while the uncertainty hypothesis is not, for discrimination involving the FIR in early acquisition trials. First Interval Response (FIR) — Extinction An overall AOV with groups, trials and stimuli as the main effects showed the main stimulus effect to be significant (F^, 8=28 . 41,p4. 001) with mean CR^=.78 and mean CR2=.66. A summary of this analysis is shown in Table 5. The main trials effect, significant at the .005 level (F9 729=3 • 36), shows response magnitude reaching its optimal level on the second presentation of each CS in extinction and then decreasing over trials. No other effects reached acceptable significance levels although the groups x stimuli effect approached the .05 significance level (Fg g^=l.92,p <.10). In general, this last effect reflects good overall discrimination (resistance to extinction) in groups 1.00-.00, .80-.00, .60-.00 and 1.00-.40 with groups .90-.10, .90-.30 , .70-.10 and 1.00-.20 showing some discrimination only until the middle period of extinction. Figures 2a through 2i present these results graphically. 52 TABLE 5 SUMMARY OF ANALYSIS OF VARIANCE FOR FIR DURING EXTINCTION Source SS df MS F P G 33.54 8 4.19 J 1 VO 1 00 • s/G 392.99 81 4.85 T 5.90 9 .66 3.36 .001 GT 16.43 72 .23 1.17 Ts/G 142.25 729 .20 C 5.81 1 5.81 28.41 .001 GC 3.14 8 .39 1.92 .10 Cs/G 16.58 81 .20 TC 1.34 9 .15 .95 GTC 10.73 72 .15 .95 TCs/G 113.92 729 .16 As with acquisition data, extinction trials were com bined into early (trials 2 through 4), middle (trials 5 through 7), and late (trials 8 through 10) trial blocks and the same specific comparisons were done to test the difference and uncertainty hypotheses within each block. This division into blocks was designed to examine the course of extinction more precisely. Table 6 shows the mean difference between CR^ and CR2 for groups used in the four comparisons. For comparison 1, a test of the difference hypothesis, group .90-.10 was compared to groups .90-. 30, .70-.10 and .80-.20 combined. During the early trial 53 block, U^0 3q=140, p > .10. For the middle trial block, U1Q 3q=141, p> .10 and for the late trial block, Uio . 10>p>.05. During the late block, group .90-.10 showed greater discrimination (approached significance) than the other groups combined, thus supporting the difference hypothesis. The difference was significant at just above the .05 level of significance. TAELE 6 MEAN DIFFERENCE (MICROMHOS) BETWEEN CR. AND CR2 OF GROUPS OR COMBINED GROUPS USED IN COMPARISONS: FIR DURING EXTINCTION TRIAL BLOCK Comparison ___________________ Early Middle Late 1 .90-.10 .12 .00 .23 .90-.30, • 00 0 1 • N J O • 0 1 • f —1 o .12 .00 .07 2 1.00-.20, .80-.00 •12* *04* .16 1.00-.40, .60-.00 .18 .17* .20 3 .90-.10 .12 .00 .23 1.00-.20, • 0 0 0 1 0 o o .12 .04 .16 4 1.00-.40, .60-.00 .18 •17* .20 .80-.20, .90-.30, .70-.10 .12 .00 .07 ♦comparison significant at .05 level ♦♦comparison significant at .01 level Comparison 2, which also tests the difference hypothesis (compares groups 1.00-.40 and .60-.00 combined 54 to groups .80-.20, .90-.30, and .70-.10 combined) produced in the early trial block, U2Q 2q=129.5, For the middle trial block 20=136, p<.05. For the late trial block U20 ,20=-^6, p >. 05. Here in the early and middle trial blocks, groups 1.00-.40 and .60-.00 combined show greater discrimination than the 1.00-.20 and .80-.00 groups combined, a result in conflict with the difference hypothesis. However, this discrepancy may be due a true inapplicability of either hypothesis during extinction when various effects such as the partial reinforcement effect might be expected to influence differential responding. This result will be discussed in greater detail in a later section of this paper. Comparison 3, a test of the uncertainty hypothesis, which compares group .90-.10 to groups 1.00-.20 and .80-.00 combined, yields during the early trial block U10 20=^ ' P^ ‘05 ’ For the ^"^le trial block U1Q 2q=93, p>.05. For the late trial block U^q 2q=78.5,p>.05. Here for all blocks there is no significant difference in discrimination between groups compared. Comparison 4 which again tests the uncertainty hypothesis (compares groups 1.00-.40 and .60-.00 combined to groups .80-. 20, .9 0-. 30, and .70-.10 combined) produces in the early trial block, U^q,3q=273,p>. 05. In the middle trial block, U^q,30=2® ^ 05 . In the late trial 55 block, U-lq^3q=237, P> .05. Here in the middle trial block groups 1.00-.40 and .60-.00 combined show greater dis crimination than groups .80-.20, .90-.30, and .70-.10 combined thus, thus supporting the uncertainty hypothesis. Second Interval Response (SIR) — Adaptation An analysis of variance over all adaptation trials with groups, stimuli and trials as the main independent variables yielded no significant effects. This analysis served as a check on equivalence of groups in terms of responses prior to acquisition. Second Interval Response (SIR) — Acquisition During acquisition the same type of analysis of variance yielded significant trials, stimulus and trials x stimulus effects. The AOV is summarized in Table 7. The trials effect, significant at the .001 level ^F9 7 29=4 34^ reflects the increasing of response magnitudes to an optimum on the eighth presentation of each CS. The stimulus effect (F^ 0^=15 . 71 001) suggests that discrimination across all groups combined did occur with CRj=.21 and CR2=.14. The trials x stimuli effect was significant at the .025 level (F9f729=2•44) and is graphically presented in Figure 3. Mean CRs and differences between CRs for all groups during acquisition 56 TABLE 7 SUMMARY OP ANALYSIS OF VARIANCE FOR SIR DURING ACQUISITION Source SS df Ms F P G 6.19 8 .77 s/G 48.89 81 .60 1.28 T 4.12 9 .46 4.34 .001 GT 5.37 72 .08 .71 Ts/G 76.90 729 .11 C 2.41 1 2.41 15.71 .001 GC 1.13 8 .14 .92 Cs/G 12.41 81 .15 TC 2. 25 9 .25 2.44 .025 GTC 4.50 72 .06 . 61 TCs/G 74.60 729 .10 35 30 25 20 .15 10 05 10 TRIALS Figure 3. SIR: Trials by stimulus effect. 57 as well as adaptation and extinction are presented in Table 8. TABLE 8 SIR: MEAN CRS AND MEAN DIFFERENCES IN ALL GROUPS FOR ADAPTATION, ACQUISITION, AND EXTINCTION Adaptation Acquisition Extinction Group CRX cr2 Dif CRX cr2 Dif CRq cr2 Dif 1 . 0 0 - . 0 0 .08 .05 .03 .17 .02 .15 . 25 .12 .13 o o • 1 o 00 • .07 .07 .00 .16 . 07 .09 .43 .10 . 33 o o • 1 o VO • .09 .12 -.03 .24 .14 .01 .44 .14 .30 I -1 • o 0 1 • o .17 .18 -.01 . 20 .21 .01 .38 .27 .11 1.00-.20 .06 .11 -.05 .21 .10 .11 . 33 .27 .06 0 1 • O .13 .13 .00 .14 .10 .04 . 22 .19 .03 .90-.30 .08 .15 -.07 .17 .14 .03 .35 . 28 .07 .80-.20 .15 .23 -.08 .27 .23 .04 .55 .46 . 09 .90-.10 .21 .23 -.02 .33 .22 .11 . 52 .32 .18 Mean .12 .14 -.02 .21 .14 .07 .39 .24 .15 Specific comparisons for SIR data during acquisition were done in the same way as for FIR data. Since the overall groups x stimulus effect of the analysis of variance had not approached significance, it was not expected that any of the comparisons would yield significant results. However, the comparisons were carried out since they had been planned prior to collection of data. As expected, none of these comparisons 58 yielded significant results. (See Appendix for description of the four comparisons.) For comparison 1, a test of the difference hypothesis, U^q 3Q=120,p>.05 in the early trial block, U10 30=12^,P^ *°5 the trial block, and Uio 30=137,p>.05 in the late trial block. For comparison 2, a test of the difference hypothesis, U20 P> .05 in the early trial block, U20 20=-*-53, p>.05 in the middle trial block, and U2o,20=193f p>.05 in the late trial block. For comparison 3, a test of the uncertainty hypothesis, in the early trial block U^q 20=93• in the middle trial block, 20=9® *5' ' and ■ ' ■ n the late trial block, U = 81,p>.05. X U fa U For comparison 4, a test of the uncertainty hypothesis, in the early trial block gQ=268 , p> . 05 , in the middle trial block, U20,30=271, ?> .05, and in the late trial block, I^q 2q=291 • * Second Interval Response (SIR) — Extinction During extinction second interval responses were analyzed through a general AOV with trials, groups, and stimuli as the independent variables. The stimulus effect was significant at the .001 level (F-^ ^ g^=38. 80) with mean CR^.39 and CR2=.24. A gradual decrease in response magnitude over extinction trials was reflected 59 in a significant trials effect (Fg ^29=3*22f .05). The groups x stimuli effect was significant at the .05 level (Fgfgi=2.31). Descriptive data for this last interaction are presented in Table 8. In general, groups .60-.00, .80-.00, and .90-.10 showed greater discrimination than the other groups and all groups showed a difference between CR^ and CR2 in the expected direction except for group 1.00-.40. A summary of the analysis of variance is in Table 9. TABLE 9 SUMMARY OF ANALYSIS OF VARIANCE FOR THE SIR DURING EXTINCTION Source SS df Ms F P G 16.32 8 2.14 1.08 s/G 152.94 81 1.89 T 4.54 9 .50 3.22 .005 GT 10.81 72 .15 .96 Ts/G 114.29 729 .16 C 9.20 1 9.20 38.80 .001 GC 4.39 8 .55 2.31 .05 Cs/G 19.21 81 .24 TC 1.35 9 .15 .94 GTC 13.17 72 .18 1.14 TCs/G 116.91 729 .16 More specific comparisons among groups were done using the Mann-Whitney U and the four comparisons pre viously explained (see Appendix). Again trials were combined into early, middle and late trial blocks. For comparison 1, which tests the difference 60 hypothesis, (compares group .90-.10 to groups .90-.30, .80-.20, and .70-.10 combined), in the early trial block U10,30=150 ' .05? in the middle trial block, U]_o,30=^1,5' p<,.05; in the late trial block, U^q , 30=97 ' P ^ • 05 • Th^t is, in the middle and late trial blocks there is greater discrimination in the .90-.10 group than in the others combined. This result supports the difference hypothesis. For comparison 2, which tests the difference hypothesis (compares groups .80-.00 and 1.00-.20 combined to groups 1.00-.40 and .60-. 00 combined), U20 f 20=^-77 • 5' p A .05 in the early block, U20t29=181, p> .05 in the middle block and U20 20=^63' P>*°5 i - n the late trial block. Here, there is a reversal of expected results in terms of the difference hypothesis. That is, groups 1.00-.4 0 and .60-.00 combined show greater discrimination than groups .80-.00 and 1.00-.20 combined, in the early trial block. For comparison 3 which tests the uncertainty hy pothesis (compares groups .90-.10 to groups .80-.00 and 1.00-.20 combined), U1Q 2o=:99'P^*05 in the early block, ”10,20=94.5,p>.05 the middle block and ”io,20=83' p>.05 in the late block. There are no significant differences between the groups compared in the three trial blocks. For comparison 4 which again tests the uncertainty 61 hypothesis (compares groups .60-.00 and 1.00-.40 combined to groups .80-.20, .90-.30 and .70-.10 combined), U20,30=250, p y. 05 in the early block, I^q ^q=229, p>.05 in the middle block and U2o/3o=193, .05 in the late block. Here the uncertainty principle is supported late in extinction. Means of groups involved in these comparisons are presented in Table 10. TABLE 10 MEAN DIFFERENCE (MICROMHOS) BETWEEN CRX AND CR2 OF GROUPS OR COMBINED GROUPS USED IN COMPARISONS: SIR DURING EXTINCTION TRIAL BLOCK Comparison _____________________________ Early Middle Late 1 .90-.10 .15 • I2* .26 .90-.30, O 1 —I 1 O • O CM 1 O CO .10 -.03 .07 2 .80-.00, 1.00-.20 . 15 * .24 .16 1.00-.40, .60-.00 .19 .15 .28 3 .90-.10 .15 .12 .26 .80-.00, 1.00-.20 .15 .24 .16 4 . 60—.00, 1.00-.40 .19 .15 .28 . 80-.20, .90-.30, .70-.10 .10 -.03 .07 * comparison significant at .05 level Third Interval Response (TIR) — Adaptation An overall analysis of variance showed only the trials 62 effect to be significant (F3^243=3•' P ^ •°5)• Response magnitude decreased over trials during adaptation. Third Interval Response (TIR) — Extinction An overall AOV showed both the trials and the three-way interaction effects to be significant. Response magnitude decreases over trials (Fg 729=3•74,p<.001) . The groups x trials x stimuli effect was significant at about the .05 level (F72,729=1 • 37) and i - s represented graph ically in Figures 4a through 4i. On the first extinction trial (test trial) group 1.00-.00 clearly showed discrimination in the expected direction as did groups .80-.00, .60-.00, 1.00-.40 and .90-.10. Following trial 1 in extinction this discrimination broke down rapidly (by the second presentation of each CS) in all groups, but group .90-.10 which continued to show good differential responding until the seventh presentation of each CS during extinction. A summary of the AOV can be found in Table 11. Means and mean differences for each group are presented in Table 12. A test of the difference and uncertainty hypotheses for the TIR was done using post hoc comparisons of the Mann-Whitney U type. This test was done only for the first trial during extinction which can be considered as equivalent to a traditional conditioning "test" trial during which the UCS is omitted. m o ■§ o n u ■H g d ■H P S CO u > .60- .50- .40- .30- .20- .10- .00- A \ \ \ \ \ \ \ \ \ " s . ( I T CR2 ( I T 1-2 3-4 1-2 3-4 5-6 7-8 Adapt. Extinction 9-10 TRIALS Figure 4a. TIR: Mean CR-^ and CR2 during adaptation and extinction in group 1.00-.00. = 1.00) = .00) GSR i n micromhos Vac .60- .50- .40- .30- .20- .10- . 00- Extinction Adapt. CRX (7T= .80) CR2 (7T= .00) TRIALS Figure 4b. TIR: Mean CR and CR during adaptation and extinction in group .80-.00. < T \ & GSR i n micromhos V/AC .50- .40- .30- .20- .10- .00- r S ' CR-l ( TT= .60) CR2 (J7= .00) 1-2 3-4 1-2 3-4 5-6 7-8 9-10 Adapt. Extinction TRIALS Figure 4c. TIR: Mean CR^ and CR2 during adaptation and extinction in group .60-.00. C T v U1 GSR i n microirtfios .50- .40- .30- .20- .10- ,00- \ N > \ \ cr-l (jr cr2 ( i t 1-2 3-4 1-2 3-4 5-6 7-8 9-10 Adapt. Extinction TRIALS Figure 4d. TIR: Mean CR.^ and CR2 during adaptation and extinction in group 1.00-.40. = 1.00) = .40) cn GSR i n micromhos .50- (j .40- k .30- .20- .10- .00- 3-4 1 Adapt. Extinction CR ( I f = 1.00) CRl (7f= .20) TRIALS Figure 4e. TIR: Mean CR.^ and CR2 during adaptation and extinction in group 1.00-.20. .50- 0 < > to o ■ a o > - l u •H £ a ps C D o .40- .30- .20- ,10- .00- \ \ \ 1-2 3-4 1-2 3-4 5-6 7-8 Adapt. Extinction TRIALS - 9-10 CR (JT= .70) CR^ (JT= .10) Figure 4f. TIR: Mean CR1 and CR_ during adaptation and extinction in group .70-.10. cr. ; 0 0 .50- 1 0 5 0 ) O iC e u 0 •H e c •H P S IQ CD .40- .30- .20- .10- .00- 1-2 3-4 5-6 7-8 9-10 CR-j CR, Adapt. Extinction TRIALS Figure 4g. TIR: Mean CR^ and CR2 during adaptation and extinction in group .90-.30. = .90) = .30) O v VO GSR i n micromhos Va c .50- 40- .30- .20- 10- .00- 1-2 3-4 1-2 3-4 5-6 7-8 9-10 Adapt. Extinction TRIALS Figure 4h. TIR: Mean CR^ and CR during adaptation and extinction in group .80-.20. i t i = .80) = .20) i S O IiU IO J D T U I Uf HSD .40- .30- .20- 3-4 5-6 7-8 9-10 4 1 Adapt. Extinction = .90) = .10) TRIALS Figure 4i. TIR: Mean CR.^ and CR2 during adaptation and extinction in group .90-.10. TABLE 11 72 SUMMARY OF ANALYSIS OF VARIANCE FOR TIR DURING EXTINCTION Source SS df MS F P G 2.22 8 .28 .47 s/G 47. 91 81 .59 T 4.43 9 .49 3.74 .001 GT 9.01 72 .13 .95 Ts/G 96.03 729 .13 C .29 1 .29 2.26 GC .80 8 .10 .79 Cs/G 10.27 81 .13 TC 1.08 9 .12 1.08 GTC 10.99 72 .15 1.37 .05 TCs/G 80.97 729 .11 TABLE 12 TIR: .MEAN CRS AND MEAN DIFFERENCE IN ALL GROUPS FOR ADAPTATION AND EXTINCTION Group Adaptation Extinction cr- l cr2 Dif f CRn O . cr2 Diff 1.00-.00 . 23 .25 -.02 .17 .14 .03 .80-.00 .08 .13 -.05 .20 .13 .07 .60-.00 .08 .07 .01 .20 .15 . 05 1.00-.40 .18 .17 .01 .21 .19 .03 1.00-.20 .07 .06 . 01 .23 .23 .00 .70-.10 .21 .19 .02 .25 .19 .06 .90-.30 .20 .13 .07 .14 .22 -.08 .80-.20 .37 .22 .15 . 26 .23 . 03 .90-.10 .33 .40 -.07 .29 .24 . 05 Mean .20 .18 . 02 . 22 .19 .03 73 For comparison 1, which tests the difference hypothesis (compares group .90-.10 to groups .90-.30, .80-.20, and .70-.10) U^qf3o=76, p-.01. Here the .90-.10 group showed discrimination in the expected direction while the other groups combined did not. The difference between the groups compared was significant. For comparison 2, a test of the difference hypothesis, (compares groups 1.00-.40 and .60-.00 combined to groups .80-. 00 and 1.00-.20 combined) U20, 20=m • .01. Here groups 1.00-.40 and .60-.00 combined showed dis crimination in the expected direction while other groups did not. The difference is significant but according to the difference hypothesis one would expect the .80-.00 and 1.00-.20 groups to show discrimination and to show it to a greater degree than the 1.00-.40 and .60-.00 groups. Therefore the results are contrary to the difference hypothesis, if indeed the difference hypothesis is meaningful for the TIR at this period of conditioning. These results will be discussed in a later section of this paper. Comparison 3, a test of the uncertainty hypothesis, (compares group .90-.10 to groups .80-.00 and 1.00-.20 combined) yielded U=53, .025. According to the difference hypothesis one would expect the .80-.00 and 1.00-.20 groups to show greater discrimination than 74 group .90-.10. Thus the results were contrary to the difference hypothesis and also significant. Group .90-.10 showed discrimination in the expected direction while the other groups did not. Comparison 4, a test of the uncertainty hypothesis (compares groups .60-.00 and 1.00-.40 combined to groups .80-.20, .90-.30 and .70-.10 combined) yielded U20,30=178' 4^.01. Here, the results support the uncertainty hypothesis, but since groups .80-.20, .90-.30 and .70-.10 combined do not show discrimination in the expected direction, further discussion is necessary and will be done in a later section of this paper. In general for the TIR, there was not evidence of either discrimination within groups nor discrimination among all groups combined. Therefore, further analysis was not considered truly appropriate as was the case with the first interval response, for example. CHAPTER VI DISCUSSION In any discussion of the conditioned galvanic response, more than one variable may be defined. The most often used and temporally stable is what is here been called the first interval response (FIR). In most GSR studies where the inter-stimulus interval is at least one second, the response will be exhibited. This is not true for what is here called the second interval response (SIR). That is, the ISI may not be long enough to allow observation of the SIR and too, the point at which the SIR begins will vary both with the length of the ISI and with the experimenter's definition. As for the third interval response (TIR), this can only be measured on so-called "test trials" which are again arbitrarily chosen by the experimenter. This leaves only extinction for a trial by trial analysis and so the development of the TIR is largely unknown. The FIR, on the other hand, has the advantage of being readily measured and assessed during all phases of the experiment with long ISIs: adaptation, acquisition, and extinction. The FIR has alternatively been viewed as an orienting response (OR) or as a kind of "preparatory" response which "prepares" the subject for receipt of the 75 76 UCS. In this context, it is especially suitable for study in discrimination paradigms since it can appear on both CS^ and CS2 trials and provides a ready comparison of the effects of the two conditioned stimuli on GSR activity. For these reasons the primary focus of this discussion will be on the FIR. Nevertheless, SIR and TIR results were analyzed and an attempt shall be made to meaningfully discuss the results. It should be born in mind that the study was not designed to optimally explore these results. First Interval Response The results of this study indicate clearly that discrimination occurred across all groups combined and that the groups differed in degree of discrimination. In all groups the mean response to the CS;^ was greater in magnitude than that to CS2, a result which takes on greater significance when it is realized that during adaptation the response to CS2 was usually greater than that to CS-p Thus for discrimination to occur an initial bias to respond with greater magnitude to CS2 had to be overcome. This bias during adaptation was itself unexpected. The only plausible interpretation for it rests on the colors chosen for the CS lights. Originally it was thought that the lights had no intrinsic characteristics that would contribute to any 77 differential response during adaptation. Since the yellow light used as CS2 yielded larger responses during adap tation, it can only be assumed that yellow had some connotatively arousing effect not possessed to the same degree by the color blue. Perhaps the subjects when told they were in a "shock" experiment related yellow, a color used in other contexts to signify danger, with shock. This is especially feasible in light of the instructions (see Appendix A) which were given before adaptation trials began. Thus before any shocks were administered in the acquisition series, subjects did not know which of the two lights would be followed by shock or if both of them would be. It might have seemed more "logical" to them that yellow should signify shock to come. Over all groups, the trials x stimulus effect (Figure 1) shows the CR^ increasing to a greater magnitude than the CR2 and remaining at this level until late in the series. The CR2, on the other hand, shows a decrease in magnitude by trial 4 after an initial increase on trials 2 and 3. This result is consistent with past studies and is typical of GSR discrimination curves in general. It is noteworthy because in the present study some groups were not typical of those used in other studies and yet the same kind of conditioning curve was obtained. This may be interpreted as indicating that there is a much 78 greater latitude in discrimination studies than is usually believed. That is, as long as one stimulus has a higher probability of being followed by shock than the other (i.e., 7T1-7T2= .60 at least), some differential responding occurs regardless of the probabilities and regularity of reinforcement. On the other hand although the groups x trials x stimulus effect was insignificant, it is also apparent (see figures 2a through 2i) that the development of the differential response differs among groups and this difference is less easily quantified. By development is meant the differential acquisition among groups (and the underlying — physiological and cognitive — basis for it) of the conditioned discrimination. That is, some groups acquire the response on earlier trials than others. Some extinguish during conditioning; others do not. Such differences are not easily observed, much less understood or explained, with use of only "traditional" groups in discrimination studies. For example, groups .80-.00 and .60-.00 show greatest discrimination in the first half of the acquisition series while group 1.00-.20 shows greatest discrimination (comparable to that in the .80-.00 and .60-.00 groups) on trials 4 through 6, as does group 1.00-.40 although to a lesser extent. On the other hand what is not immediately apparent from inspec 79 tion of single graphs and means alone is that the 1.00-.40 and .60-.00 groups when combined show less discrimination in the early and middle trials than do groups 1.00-.20 and .80-.00 combined. This result is reversed, however, for the late trial block. While not significant by itself this comparison does reflect a trend which is significant in comparison 1 and which also tests the difference hypothesis. During acquisition, groups 1.00-.00 and .9 0-.10 exhibited the greatest and most consistent (over trials) discrimination. Groups .90-.30, 1.00-.40, .80-.20, and .70-. 10 showed the poorest discrimination. By looking at the results of all four comparisons during acquisition a pattern is suggested which supports both the difference and uncertainty hypotheses but at different points in acquisition. Thus the trend is for the difference hypothesis to exert influence early in acquisition and the uncertainty hypothesis to exert influence, although to a lesser extent, during late acquisition. The greater overall influence of the difference principle during acquisition is also supported by the results of compar isons 5 and 6. During extinction there is again discrimination across all groups combined and differences in discrimi nation among groups. Again the mean response to CS-^ 80 is greater than the mean response to CS2 in each group. Individual comparisons suggest that the difference hypothesis is not supported during extinction. In comparison 2, for example, there is a reversal of what obtained during acquisition. Now the 1.00-.40 and .60-.00 groups combined show greater discrimination (or less resistance to extinction of the discrimination) than the 1.00-.20 and .80-.00 groups combined. This reversal suggests that an effect loosely akin to the PRE in simple conditioning studies may be operating here. In simple conditioning studies the PRE is demonstrated as greater resistance to extinction of those groups which in acquisition received the lower probability schedule of reinforcement. Here the groups where the probability difference . 60) is lower showed greater resis tance to extinction of the discrimination. Since the PRE with regard to discrimination has been little investigated previously and since the present study is not designed to assess such effects, further discussion of such a loosely related effect in discrimination is tenuous. In this context, however, it should be noted that in Newman's (1967) study he concluded from his results that his discrimination hypothesis (similar to the difference hypothesis here) was not applicable within groups during extinction. As here, the difference 81 hypothesis was not supported during extinction. However, he did find a between groups PRE (as he termed it) such 7T1-TT2 that as (---2---) acquisition increased, resistance to extinction (defined in terms of (CR1^—^2) ) decreased. In addition he found that in seven of his ten groups resis tance to extinction was greater for the CRs to the CS with the higher probability value. In the present study a similar effect was obtained in that in all groups resistance to extinction was greater for CR^ than for CR2. Although appeal to the PRE is not applicable here (if anything, the reverse occurs), the analogy to the PRE is applicable. Thus, there is some evidence to suggest that the difference hypothesis operates, showing a reversal of acquisition results in somewhat the same manner that the PRE accounts for resistance to extinction in low reinforcement probability groups during simple conditioning. It should be clear, that, by way of analogy, a phenomenon is described rather than explained just as the PRE describes an event rather than serves as explanation. An elaboration of the mechanisms underlying the phenomenon here described is necessarily contingent on further study devoted to that purpose. The significance of the FIR results both during acquisition and extinction are not easily assessed through the statistical analyses conventionally used. Many 82 of the important differences appear to be qualitative rather than quantitative. In fact the independent variable itself while quantitative in appearance does not readily lend itself to true quantitative analyses. The comparisons here utilized are not to be taken each by itself as significant or not significant and then interpreted. Rather it is the trends suggested by all analyses together and inspection of the data themselves which is of importance. Thus the overall conclusion that can be drawn from the FIR data is that both the difference principle and the uncertainty principle influence the differential GSR but that the difference principle operates to a greater extent than the uncertainty principle. In addition, that in all groups which have been used in past studies (i.e., 1.00-.00, .80-.00, and .60-.00) are not the only groups yielding optimal discrimination; discrimination is good in other groups such as the .90-.10 group as well. Still other groups (e.g., .85-.25 or .80-.15) may aid in better understanding the nature and development over trials of differential responding (discrimination). Although the "traditional" groups in the past have yielded informative, reliable data and are representative of the discrimination process, they are not inclusive of all paradigms for discrimination, either within or 83 without the laboratory setting. The results of the present study suggest that the initial attempt here made to explore various reinforcement contingencies should be further explored in future studies in an attempt to understand the development and nature of conditioned discrimination. Second Interval Response The SIR is often considered an "anticipatory" response for receipt of the UCS. It seems to be time locked to onset of UCS rather than CS onset. SIR discrimination did occur over all groups combined as evidenced by the significant stimulus and stimulus x trials effects. There were however, no significant statistics with regard to differences among groups during acquisition. As opposed to the FIR, this may represent a slower development of the SIR over trials. Typically the SIR is a response small in magnitude and often absent during adaptation and early acquisition trials. Thus the idea that it takes longer to develop than for example, the FIR is quite plausible. Especially in groups .80-.00 and .60-.00 where an effect aking to a PRE might be expected to occur, one sees greater discrimination during extinction than during acquisition, again supporting the notion of slow development (it takes longer to appear than the FIR, for example). This is true 84 especially if one notes the explanation of PRE which sees it as a non-discrimination between the end of the acquisition period and the beginning of the extinction period. During extinction there is a significant groups x stimulus effect and in addition three of the comparisons which test the difference hypothesis yield significant U’s. In comparison 2 where "regularly reinforced" stimuli are compared, the T^-Tf2=. 60 groups show greater differ ential responding than the "TT^— T T 2= - 80 groups. Again this result can be elaborated in the same manner as the similar results found for the FIR. This appeal to an effect loosely analogous to the PRE is again post-hoc and tentative. It suggests a possible mechanism that can not be fully assessed in this study since the hypotheses of the study are intended to explore acquisition of discrimination; not extinction. Nevertheless, the data are available and preliminary notation of the extinction data may be useful with regard to future studies designed specifically to assess discrimination during extinction when probability is the independent variable. It is interesting to note that the course of extinction is not linear. That is, within any comparison group or groups combined there is not a gradual decrease in differential responding. Rather in all groups but 85 two there is a fluctuation in differential response such that the mean difference between CR^ and CR^ is greatest by the end of extinction, I.e., by the last trial block. \ This perhaps represents as adjustment by subjects to the changed conditions of extinction vs. acquisition and then a "remembering" of the original significance of the two CSs. It is as if the Ss at the beginning of the extinction period experience a disorientation due to the change in the stimulus conditions and thus have to reorient. Then they either believe that what appears to be the case actually is, i.e., "no more shock" or due to the intermittent nature of the reinforcement in acquisition they continue to expect shock. At least the CS once again contains its original meaning of "shock to come perhaps". In a sense the CS-^ is no longer just a light but a light "plus". It is a light plus all the conno tations derived from the experimental manipulation itself. Again there is only one significant result among the comparisons that test the uncertainity hypothesis. While an isolated significant result out of many tests is not conclusive of anything, it does suggest that the uncertainty principle exerts some influence even in late extinction trials. It can be concluded from the SIR results that discrimination occurs across all groups combined, to some 86 extent. However, since there were not between groups differences during acquisition, neither the difference nor uncertainty hypotheses can in a real sense be supported. On the other hand, during extinction, there is greater evidence of discrimination across and within groups with both hypotheses being partially supported the difference hypothesis in a manner loosely akin to PRE. The significance, however, of the two hypotheses at this late point (extinction) may be different from their significance during acquisition Third Interval Response For the TIR, discrimination over all groups combined was not obtained (i.e., not statistically sig nificant) even in the early trial block. Reference to Figures 4a through 4i makes it apparent that the groups with poor or no discrimination on early extinction trials are groups .70-.10, .90-.30, .80-.20, and 1.00-.20. It is clear that in terms of both the difference and uncer tainty hypotheses, all of these groups but the 1.00-.20 would be expected to show poor discrimination. That the 1.00-.20 group also shows poor discrimination at the beginning of extinction can not be accounted for by the two hypotheses. Further,most post-hoc explanations would not suffice because they would call into play the performance of the .80-.00, .90-.10, or 1.00-.40 groups 87 in terms of the difference and uncertainty hypotheses respectively. These latter groups however, did perform as expected and so the performance of the 1.00-.20 group must be explained by itself. Yet an explanation of this group alone seems quite inappropriate and so the con clusion must be drawn that this group's performance is due to chance factors and that on replication such a result would not obtain. The poor performance (i.e., lack of significant discrimination) among all groups during extinction may be explained by referring to the nature of the extinction process itself. That is, only the first trial of extinction can truly be considered comparable to a "test trial" since following this trial, contingencies change. If the height of conditioned discrimination of the TIR is reached somewhere during acquisition, then response decrement may begin at least in some groups long before nominal extinction is initiated. Therefore, what is recorded and measured on the first trial of extinction may from the subjects1 point of view be akin to a later extinction trial, in which case not much discrimination is to be expected. This is especially likely in consideration of the fact that GSR conditioning usually reaches its maximum somewhere before the tenth trial in conditioning. The graphs for the FIR do show 88 this and one can, by analogy, assume that a similar process is occurring with regard to the TIR. This process is not directly measurable in the present study, however, since optimizing the paradigm for study of the FIR makes it impossible to also optimally study the TIR during acquisition. With regard to later trials in the extinction period it should be noted that since contingencies have changed and since response decrement may have started during the acquisition period, lack of differential response during these late trials perhaps should have been expected. That is, following the first trial of extinction the subject may have to relearn "what-follows-what" and upon relearning responds according to these new contingencies, rather than those learned during acquisition. The performance of the .90-.10 group which seems to show the greatest resistance to extinction of all groups is of interest. The data suggests, that in terms of experiments dealing with resistance to extinction, a .90-.10 group may be most fruitful for study. In an attempt to understand the nature of discrimination conditioning, this never-used group suggests that un certainty as a principle, given a minimum difference, may be of great use. CHAPTER VII CONCLUSION The results of this study support both the difference and uncertainty hypothesis proposed and explained in the beginning of this paper. There is some evidence to suggest that at least in acquisition and particularly for the FIR, the difference principle seems more potent in determining discrimination performance. The effect of either of these principles has been assessed through different periods within both acquisition and extinction and it has been found that each may operate to a different extent at different points throughout these periods. Further, they have varying effects on the three dependent variables utilized. That the .90-.10 group for all responses and all periods shows good discrimination — at times better than other more conventional groups — suggests that here is a long overlooked experimental arrangement that may prove advantageous for further study of discrimination con ditioning. Further studies using varying probabilities of reinforcement may be of benefit in assessing the nature and development of classical discrimination conditioning. 89 APPENDIX A Instructions to the subject The experiment I am running involves the recording of your galvanic skin responses while you are receiving different kinds of stimulation such as colored lights and electric shock. Before we start, I'd like you to read this paper and sign it if you wish. ( £ 3 was given a subject consent form to sign.) The galvanic skin response or GSR is a response of the autonomic nervous system. It is an involuntary response. Much of the response is due to the activity of the small sweat glands which are found mostly in the palms and soles. In order to record the response I will place two electrodes on your fingers. You will also be receiving some brief electric shocks through these electrodes (E points to another pair of electrodes) which I'll place on your right arm. (E attaches all electrodes.) Because people vary in their sensitivity to the same shock, we let each person set the level of shock for himself•before the experiment begins. This is how we will do it. I will begin by giving you a brief low intensity shock to your right arm. It will probably be too low for 90 91 you to feel. I'll gradually raise the intensity a little at a time. I'd like you to let me know when you first feel a shock by raising your right hand. From there I'll continue to raise the intensity a little at a time and I want you to let me know when it gets to a level you consider annoying or uncomfortable. It should be a negative kind of stimulus but it should not be painful. When we get to that level, raise your hand a second time and that will be the level at which it will remain for the experiment. (E leaves room to administer shock work-up.) The last shock you felt is the level we will keep it at during the experiment. During the experiment, you will see different colored lights presented many times on this screen. They may sometimes be followed by shock. All you have to do during the experiment is pay attention to the screen so that you don't miss seeing any of the lights. It is very important that you pay attention. At the end of the experiment I'll have some questions to ask you about what happened during the experiment. APPENDIX B Description of Statistical Comparisons Comparison 1 compares groups in which .80) to groups in which (7T^-^2=*60) where no groups have a "regularly reinforced" CS, i.e., where neither 7T[ nor 7 7 " 2 equals .00 or 1.00. This tests the difference hypothesis. Specifically it compares group .90-.10 to groups .90-. 30, .80-.20, and .70-.10 combined via a Mann-Whitney U test. Comparison 2 compares groups in which .80) to groups in which Or±-7T2=. 60) where all groups have one "regularly reinforced" CS, i.e., where7Tj or #2 equals .00 or 1.00 respectively. This again tests the difference hypothesis. Specifically, groups 1,00-.40 and .60-.00 combined are compared to groups .80-.00 and 1.00-.20 combined via a Mann-Whitney U test. Comparison 3 compares "regularly reinforced" to "irregularly reinforced" groups with the difference held constant at .80. This tests the uncertainty hypothesis. It compares groups .80-.00 and 1.00-.20 combined to groups .90-.10 through a Mann- Whitney U test. Comparison 4 compares "regularly reinforced" to "Irregularly reinforced" groups with the 92 93 difference held constant at .60. This also tests the uncertainty hypothesis. It compares groups .60-.00 and 1.00-.40 combined to groups .80-. 20, .90-. 30, and .70-.10 combined using the Mann-Whitney U test. SELECTED BIBLIOGRAPHY Arrow, K. J., Karlin, S., and Suppes, P., Mathematical Methods in the Social Sciences, 1959. Stanford: Stanford University Press, 1960. Atkinson, R. C., Bower, G. H., and Crothers, E. J., An Introduction to Mathematical Learning Theory. New York: John Wiley and Sons, 1965. Baer, P. E., and Fuhrer, M. J., Cognitive processes during differential trace and delay conditioning of the GSR. J. Exp. Psychol., 1968, 78(1), 81-88. Blair, W. C., Jr., and Peterson, C. R., Decision-theoretic model for probability learning. Proceedings 76th Annual Convention, APA, 1968, 3, 51-52. Brunswik, E., Probability as a determinant of rat behavior. J. Exp. Psychol., 1939, 25, 175-197. Burke, C. J., and Estes, W. L., A component model for stimulus variables in discrimination learning. Psychometrika, 1957, 74(5), 410-419. Bush, R. R., and Estes, W. K., (Eds.) Studies in Mathematical Learning Theory. Stanford: Stanford University Press, 1959. Bush, R. R., and Mosteller, F., A mathematical model for simple learning, Psychol. Rev., 1951, 58, 313-323. Bush, R. R., and Mosteller, F., A model for stimulus generalization and discrimination. Psychol. Rev., 1951, 58, 413-423. Bush, R. R., and Mosteller, F., Stochastic Models for Learning. New York: John Wiley and Sons, 19 55. Cermak, R., and Wickens, D. D., Interstimulus interval and CS duration effects in differential conditioning J. Exp. Psychol., 1969, 79(2), 233-235. Estes, W. K., Component and pattern models with Markovian interpretations. Studies in Mathematical Learning Theory. Edited by R. R. Bush and W. K. Estes. Stanford: Stanford University Press, 1959. 94 95 Estes, W. K., Of models and men. American Psychologist. 1957, 12, 609-617 Estes, W. K., Theory of learning with constant, variable or contingent probabilities of reinforcement. Psychometrika, 1957, 22(2), 113-132. Estes, W. K., Toward a statistical theory of learning. Psychol. Rev., 1950, 57, 94-107. Estes, W. K., and Burke, C. J., Application of a statis tical model to simple discrimination learning in human subjects. J. Exp. Psychol., 1955, 50(2), 81-88. Gambino, B., and Myers, J., Role of event runs in probability learning. Psychol. Rev., 1967, 74(5), 410-419. Grice, G. R., Visual discrimination learning with simultaneous and successive presentation of stimuli J. Comp. Physiol. Psychol., 1949, 42, 365-373. Gynther, M. D., Differential eyelid conditioning as a function of stimulus similarity and strength of response to the CS. J. Exp. Psychol., 1957, 53, 408-416. Halpern, Lantz, and Schwartz , Cue properties of the event run in choice discrimination learning. J. Exp. Psychol., 1969, 80(2), 237-242. Harlow, H. F., The formation of learning sets. Psychol. Rev., 1949, 56, 51-65. Heron, W. T., The effects of a differential rate of reinforcement of responses to two levers. J. Comp. Physiol. Psychol., 1942, 33, 87-96. Hilgard, E. R., and Bower, G. H. (Eds.) Theories of Learning. New York: Appleton-Century-Crofts, 1966 Hull, C. L., Principles of Behavior. New York: Appleton-Century-Crofts, 1943. Hull, C. L., Simple qualitative discrimination learning. Psychol. Rev., 1950, 57, 303-313. Hull, C. L., A Behavior System: An Introduction to Behavior Theory Concerning the Individual Organism. 96 New Haven: Yale University Press, 1952. Hunt, D. P., Differential GSR conditioning at extended interstimulus intervals. J. Exp. Psychol., 196 6, 20-25. Kimble, G. A., Conditioning and Learning. New York: Appleton-Century-Crofts, 1961. Lashley, K. S., An examination of the "continuity theory" as applied to discriminative learning. J. Gen. Psychol., 1942, 26, 241-265. Lashley, K. S., and Wade, M., The Pavlovian theory of generalization. Psychol. Rev., 1946, 53, 72-87. Lordahl, D., Pattern perception and use of hypotheses in prediction of binary events. Proceedings 76th Annual Convention, APA, 1968, 3, 49-50. Myers, J. L., and Cruse, D., Two-choice discrimination learning as a function of stimulus and event prob abilities. J. Exp. Psychol., 1968, 77(3), 453-459. Newman, F. L., Differential eyelid conditioning as a function of the probability of reinforcement. J. Exp. Psychol. 1967, 75(3), 412-417. Newman, F. L., and Woodhouse, J., Differential eyelid conditioning: Establishing differential responding prior to varying the probability of reinforcement. J. Exp. Psychol., 1969, 80(1), 146-149. Restle, F., Theory of selective learning with probable reinforcements. Psychol. Rev., 1957, 64, 182-191. Restle, F., Toward a quantitative description of learning set data. Psychol. Rev., 1958, 65, 77-91. Robbins, J., Probability learning in the correction T-maze under noncontingent reinforeement schedules. Dissert. Abstracts, 1969, 29(11-B), 4411. Sadler, E. W., A within-and-between-subjects comparison of partial reinforcement in classical salivary conditioning. J. Comp. Physiol. Psychol., 1968, 66 (3) , 695-698. Spence, K. W., An experimental test of the continuity and non-continuity theories of discrimination learning. 97 J. Exp. Psychol., 1945, 35, 253-266. Spence, K. W., Behavior Theory and Learning; Selected Papers. Englewood Cliffs: Prentice-Hall, 1960. Spence, K. W., Failure of transposition in size-discrim- ination of chimpanzees. American J. Psychol., 1941, 54, 223-229. Spence, K. W. , The nature of discrimination learning in animals. Psychol. Rev., 1936, 43, 427-449. Suppes, P., A linear model for a continuum of responses. Studies in Mathematical Learning Theory. Edited by R. R. Bush and W. K. Estes. Stanford: Stanford University Press, 1959. Suppes, P., and Rouanet, H., A simple discrimination experiment with a continuum of responses. Studies in Mathematical Psychology. Edited by R. C. Atkinson. Stanford: Stanford University Press, 19 64, Zeiner, A., Second interval discrimination of the GSR as a function of UCS intensity and trace and delay conditioning paradigms. J. Exp. Psychol., 1968, 78 (2), 276-280.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Ucr Diminution As A Function Of Isi
PDF
The Effects Of Reinforcement Pattern, Intertrial Interval Regularity, And adaptation On The Orienting Response And Classical Gsr Discrimination Conditioning
PDF
The Effect Of Discriminability On The Partial Reinforcement Effect In Human Gsr Conditioning
PDF
The Modification Of Partial Reinforcement Effect As A Consequence Of Electrical Stimulation Of The Caudate Nucleus In Cats
PDF
Human Performance As A Function Of The Joint Effects Of Drive And Incentive Motivation
PDF
Dominance And Contiguity As Interactive Determinants Of Autonomic Conditioning
PDF
Human Gsr Classical Conditioning And Awareness Of The Cs-Ucs Relation
PDF
The Discriminant Function As A Model For Diagnostic Problem-Solving
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 Effects Of Prior Part-Experiences On Visual Form Perception In The Albino Rat
PDF
Reward Expectancy Strength As Related To The Magnitude Of Frustration In Children
PDF
Habituation Of The Multiple Unit Discharge Response To White Noise Stimulation In The Unanesthetized Rabbit
PDF
The role of cues in the arousal of anxiety
PDF
Effects Of Task Performance Upon The Acoustic Reflex
PDF
The Effects Of Ether And Electroconvulsive Shock On One Trial Appetitive And Adversive Learning
PDF
Clinical Implications Of Verbal Learning Without Awareness: An Experimental Study
PDF
Age Differences In Serial Reaction Time As A Function Of Stimulus Complexity Under Conditions Of Noise And Muscular Tension
PDF
Effects Of Success And Failure On Impulsivity And Distractibility Of Three Types Of Educationally Handicapped Children
PDF
The Effects Of Two Types Of Experimenter Intervention And Schedules Of Reinforcement On Verbal Operant Conditioning Of Affective Self-References
Asset Metadata
Creator
Sukoneck, Harriet Irene
(author)
Core Title
Classical Discrimination Conditioning As A Function Of Probability Of Reinforcement
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
), Lewis, Donald J. (
committee member
), Smith, Robert A. (
committee member
)
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c18-544402
Unique identifier
UC11363017
Identifier
7203800.pdf (filename),usctheses-c18-544402 (legacy record id)
Legacy Identifier
7203800
Dmrecord
544402
Document Type
Dissertation
Rights
Sukoneck, Harriet Irene
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
Tags
psychology, experimental