. “WV”.-- now u“. 0‘."_‘ "'“(‘.‘\‘fifi~! OPERANT CONDITIONING OF SKIN POTENTIAL IN HUMANS WITH CONTROL .FOR SOMATIC MEDIATION Thesis for the Degree of M. A. MICHIGAN STATE UNIVERSITY MARGUERITE B. STEVENSON 1972' ”“’ LIBRARY Michigan State University ‘IHEslr ABSTRACT OPERANT CONDITIONING OF SKIN POTENTIAL IN HUMANS WITH CONTROL FOR SOMATIC MEDIATION By Marguerite B. Stevenson The long standing distinction between the domains of operant and classical conditioning has recently been challenged by demon- stration of operant conditioning of autonomic activity. Although such conditioning has been demonstrated, the possibility that the response change is mediated by a change in somatic activity has been repeatedly raised. The present study controlled for the possibility of mediation due to bilateral somatic changes by conditioning differential responding in the right and left hands. During con- tingent reinforcement sessions, subjects were reinforced for producing skin potential responses in one hand that were three times as large as the skin potential response seen in the other hand. Each subject received 10 daily, twenty-minute sessions. Four subjects received contingent reinforcement the first week and non-contingent (random) reinforcement the second week. This order was reversed for another group of four subjects. Results of the study indicated conditioning in some subjects in each of the groups. For the com- bined data for all subjects there was both a high number of criterion responses and a higher proportion of such responses during con- tingent reinforcement rather than during non-contingent reinforcement. Discussion focused on the problems and issues associated with operant conditioning of autonomic activity. OPERANT CONDITIONING OF SKIN POTENTIAL IN HUMANS WITH CONTROL FOR SOMATIC MEDIATION By |' ( \ Marguerite B? Stevenson A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Psychology 1972 To Rod ii ACKNOWLEDGEMENTS This study was supported in part by NIMH Research Grant MH-MH-18655 Biomedical Sciences Support Grant and NSF-URP Grant GY-8727 to Dr. Hiram Fitzgerald. The author would especially like to thank Dr. Fitzgerald, chairman of the thesis committee, for his advice and assistance during all phases of this study and to thank Dr. Elaine Donelson and Dr. Mark Rilling for their guidance as members of the thesis committee. Thanks are also expressed to Charles Wilson for assistance in the collection and summarization of the data as well as his advice and enthusiasm during this part of the study. Thanks also go to the department of psychology for providing access to subjects and to the subjects for their perserverance in the difficult and frustrating task. Special thanks go to my husband Rod for his encouragement and suggestions throughout the study as well as for his patience with the large amount of work involved. iii TABLE OF CONTENTS LIST OF TABLESOOOOOOO0.00...O...O....0.00000000CCOCCOOOOOOOOOCO LIST OF FIGURESOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO000......0...... INTRODUCTIONOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO..00... METHOD......................................................... Subjects.................................................. Apparatus................................................. Design and Procedure...................................... Data Reduction and Analysis............................... RESULTS........................................................ Screening Session -- Baseline Recording................... Screening Session -- Elicited Responses................... Reinforcement............................................. Criterion Responses....................................... Total Number of AC Responses.............................. Ratio..................................................... Comparisons with Baseline and Extinction Data............. Comparisons of Elicited and Spontaneous Responses......... DC Level.................................................. Magnitude of Responses.................................... Reasons................................................... Individual Data........................................... DISCUSSION..................................................... REFERENCESOOOOOOOOOOQOOOOOOOO0.00000000000000000000000000....0. iv vi viii 1 14 14 14 15 19 22 22 25 28 30 34 43 47 50 54 55 S6 58 62 66 TABLE OF CONTENTS (con't) APPENDIX A: Explanation and Instructions to the SUbject.OOOOOOOOOOOOOOOOOOCOOOOOOOOOOOOOOOOOOIOOOOOOOOOOOOO.. 73 APPENDIX B: Summaries of Statements Given at the End of Each Session in Response to Queries About How the Subject Did and How He Controlled the Slides......................... 75 APPENDIX C: Number of Total Responses and Criterion Responses (Total, Spontaneous, and Elicited) for Each Subject for EaCh seSSionCCOOOOOOOOO00.0.0.0...OOOOOOOOOOOOOOCOOOOOOOO0.0. 80 LIST OF TABLES Table Page 1 Sequence and Number of Sessions for Group I and Group IIOO0.00000000000000000000000...OOOOOOOOOOOOOO... 16 2 Number of AC Responses During Fifteen Minutes of screening seSSion...0.0.00000000000000000000000.0...... 24 3 Estimated Number of AC Responses if Screening Sessions had Continued for Twenty Minutes.............. 25 4 Magnitude of Initial Elicited Responses on the Right and Left AC Channels in mv............................. 27 5 Percentage of AC Responses that were Positive Going on Both Channels, of Criterion Magnitude in the Criterion Direction, and Greater than 2.5 mv. on the AC Channel......................................... 29 6 Analysis of Variance for the Number of Spontaneous Criterion Responses Occurring During the Ten Sessions............................................... 31 7 Analysis of Variance for the Total Number of Spontaneous Responses Occuring During the Ten Sessions............................................... 42 8 Comparisons Between the Means for Contingent Reinforcement (C) Sessions, the Means for Non- Contingent Reinforcement (NC) and the Extinction Data for (a) the Number of Criterion Responses, (b) the Number of Spontaneous Criterion Responses, (c) the Total Number of Responses, (d) the Total Number of Spontaneous Responses, and (e) the Ratio-coo. 51 9 Proportion of Criterion Elicited Responses and of Criterion Spontaneous Responses during Contingent and Non-contingent Reinforcement....................... 53 10 The Number and Proportion of Criterion Responses Emitted by Each § During Contingent and Non- contingent Reinforcement.‘OOOOOCQQCIOOOOOOOOOOOOO0.0... 61 vi Figure 10 LIST OF FIGURES The Mean Number of Spontaneous Criterion Responses By Group I and Group II §s During Contingent and Non-contingent Reinforcement Sessions............ The Mean Number of Spontaneous Criterion Responses During Contingent and Non-contingent Reinforcement seSSionSOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO The Mean Total Number of AC Responses by Group I and Group II §s During Contingent and Non-contingent Reinforcement Sessions............................... The Mean Number of Elicited AC Responses Made by Group I and Group II §s During Contingent and Non-contingent Reinforcementoo on o o. o o o o o o 0 on o o oo o o o o. The Mean Number of Elicited AC Responses Made During Contingent and Non-contingent Reinforcement Sessions............................... The Mean Number of Spontaneous AC Responses Made by Group I and Group II §s During Contingent and Non-contingent Reinforcement..................... The Mean Number of Spontaneous AC Responses Made by Group I and Group II §s During Week 1 and Week 2. (a redrawing for Figure 6)........................... The Mean Ratio (Number of Spontaneous Criterion Responses/Total Number of Spontaneous Responses) During Contingent and Non-contingent Sessions........ The Mean Ratio (Number of Spontaneous Criterion Responses/Total Number of Spontaneous Responses) for Group I and Group II §s.......................... Mean Ratio of the Magnitude of the Responses in the Designated Hand Divided by the Magnitude of the Response in the Other Hand During Contingent and Non-contingent Reinforcement Sessions................ vii Page 33 35 37 39 4O 44 45 46 48 LIST OF FIGURES (con't) Figure Page Al Responding for S #1..................................... 80 A2 Responding for §_#2..................................... 81 A3 Responding for S #3..................................... 82 A4 Responding for S #4..................................... 83 A5 Responding for S #5..................................... 84 A6 Responding for §_#6..................................... 85 A7 Responding for §_#7..................................... 86 A8 Responding for §_#8..................................... 87 INTRODUCTION One support of two process theory is that operant and classical conditioning can be distinguished by the responses that they modify (Rescorla & Soloman, 1967). Two process learning theorists have traditionally said that autonomic activity could be modified only by classical conditioning, somatic activity only by operant con- ditioning. This seemed to be a reasonable distinction since classical conditioning was said to involve involuntary, reflexive actions while operant conditioning required voluntary action. To test this assumption Skinner (1938) made an early attempt to use operant procedures to modify autonomic activity. His failure to operantly condition a vasomotor response led him to conclude that such conditioning was not possible. About the same time Mowrer (1938) made an unsuccessful attempt to operantly condition the Galvanic Skin Reflex (GSR), and Konorski and Miller (1937) wrote that they did not consider autonomic activity modifiable by any means other than classical conditioning. In the fifties and early sixties a few reassertions were made that such conditioning was impossible (Keller & Schoenfeld, 1950; Kimble, 1961; Soloman & Wynne, 1954). For many years the issue was considered settled: autonomic activity could not be modified operantly. Then in 1961 Razran reported that in the Russian lab Lisina's subjects (gs) had succeeded in acquiring voluntary control of vascular activity. Subjects receiving exteroceptive feedback learned 2 vasodilation (Razran, 1961). When Lisina's work was translated into English in 1965 it was apparent that her §s had only learned to re- lax their muscles and to alter their respiratory patterns rather than to directly control autonomic activity (Lisina, 1965). But by then her study had sparked a great deal of additional work in this country. About the same time as Razran's report, Kimmel and Hill (1960) posed serious difficulty for the two factor learning theory by demonstrating operant conditioning of the GSR. These pivotal studies initiated intensive investigation of the conditionability of autonomic responses using the operant methodology. Hefferline (1962) suggested the direction of this work when he reported on the presence of interoceptors in the viscera. Following the work of Kimmel & Hill (1960) many researchers have successfully demonstrated operant conditioning of electrodermal responses in adult humans (Crider, Shapiro & Tursky, 1966; Defran, Badia & Lewis, 1969; Fowler & Kimmel, 1962; Gavalas, 1967; Greene & Nielson, 1966; Greene, 1966; Johnson, 1963; Kimmel & Kimmel, 1963; Milstead, Baer & Fuhrer, 1968; Shapiro, Crider & Tursky, 1964; Shapiro & Crider, 1967; Van Twyver & Kimmel, 1966). While the researchers just cited employed positive reinforcers, successful operant conditioning of autonomic responses is not re- stricted to this class of reinforcer. For example, several in- vestigators have successfully used aversive reinforcers to operantly modify electrodermal responses by using an avoidance design (Edelman, 1968, 1970; Grings & Carlin, 1966; Kimmel & Baxter, 1964; Kimmel, Sternthal & Strub, 1966; Kimmel &.Sternthall, 1967; Kimmel & Kimmel, 1968; Martin Dean & Shean, 1968) or a punishment design (Crider, Schwartz & Shapiro, 1968, 1970; Johnson & Schwartz, 1967; Senter & Hummel, 1965). Moreover, successful demonstrations of operant conditioning of autonomic activity have not been restricted to the use of electro- dermal responses. Operant techniques have also been employed to modify heart rate (HR) (Ascough & Sipprelle, 1968; Brener & Hothersall, 1966, 1967; Engel & Hanson, 1966; Engle & Chism, 1967; Shearn, 1962) vasomotor responding (VMR) (Snyder & Noble, 1966, 1968) salivation (Brown & Katz, 1967), and electrical activity of the brain (Bry, 1970; Kamiya, 1967, 1968, and Nowlis & Kamiya, 1970) all in human beings. Animal work has also shown modification of heart rate (DiCara & Miller, 1968a, 1969a, 1969b; Malmo, 1963) blood pressure (Plumlee, 1968, 1969) salivation (Miller & Carmona, 1967), evoked potentials (Fox & Ruddell, 1968, 1970), and the firing gffla single neuron in the precentral cortex (Fetz, 1969). Despite these apparently successful studies, evidence for operant conditioning of autonomic responses is not entirely clear. For example, some re- searchers have been unsuccessful in their attempt to operantly con- dition electrodermal activity in humans (Mandler, Preven & Kuhlman, 1962; Mednick, 1964; Stern, Boles & Dionis, 1966; Stern, 1967). Others have been unable to condition HR deceleration (Headrick, Feather & Wells, 1971; Levene, Engel & Pearson, 1968). All these studies have concentrated on demonstrating the existence of operant conditioning of autonomic activity. During a conditioning study, increases in the rate of responding could be due merely to the effect of the reinforcer as a stimulus rather than to its reinforcing effects. To control for this effect of 4 sensitization to the reinforcer, designs with control groups have been used. There are, however, two prominent difficulties with the experimental designs used: Results may be biased by (l) reinforcing control subjects only during period of non-responding or (2) by using a yoked control group. The first grOUp of studies in this area followed the experimental design of Kimmel and Hill (1960). Subjects in the experimental groups were reinforced for spontaneous autonomic activity while those in the control groups received re- inforcement contingent on non-responding. They were reinforced only during period when they did not show spontaneous autonomic activity. This had the effect of biasing downward the number of responses seen in the control group; gs could have been about to emit a response at the point the reinforcer was delivered. This reinforcer would mask the spontaneous response that might otherwise have been seen or is followed by a spontaneous response that is erroneously attributed to the reinforcer. Thus a difference seen between the experimental and control groups might be due to this bias and not to a conditioning effect. Some of the studies subject to this criticism include Fowler & Kimmel (1962), Kimmel & Kimmel (1963), Crider et al. (1966), Van Twyver and Kimmel (1966), Greene & Nielson (1966), and Schwartz & Johnson (1969). One way of overcoming this problem is to examine only extinction data where there are no reinforcers to mask possible responses. The studies by Kimmel & Hill (1960), Fowler & Kimmel (1962), Green & Nielson (1966), Gavalas (1967), and Defran, Badia & Lewis (1969), did find differences during extinction, but other studies (for example, Crider, Shapiro, & Tursky, 1966; Fowler & Kimmel, 1962) have not found such differences. A second procedure for ascertaining that increases in response rate are not due merely to sensitization is to employ a yoked con- trol group. Researchers using this control include Greene (1966), Stern (1967), Kimmel & Baxter (1964), Grings and Carlin (1966) and Kimmel et a1. (1966). Church (1964) points out that this design is also biased in favor of the experimental group. Individuals vary as to the effectiveness of the reinforcer for enhancing responsivity. If the reinforcer is an effective elicitor for the experimental g, more and more responses will be elicited as the experiment proceeds. If the experimental S is paired with his yoked control for whom the reinforcer does not elicit responses, the yoked §_can be expected to show few responses to even the larger number of reinforcements quite independent of any experimental manipulations. If, however, the reinforcer is an ineffective elicitor for the experimental §_he will give few responses and the yoked S will receive few reinforce- ments. If the reinforcer is an effective elicitor for the yoked S3 responses would have been elicited, but because of whom he is paired with, he does not receive enough reinforcement to get this effect. Therefore, less responses will show on the record of control §s even without conditioning effects. To combat this problem, many experimenters have matched Sp on the basis of the number of responses during the initial period -- on the operant levels. Johnson (1963) confirmed the feasibility of this procedure by pointing out the positive correlation between GSR operant levels and responsivity to non-contingent stimulation. An even better alternative might be to use each §_as his own control. In this way, reactivity to the re- inforcer would be perfectly matched for comparing performance during contingent and non-contingent reinforcement. The models of Black and DeToledo (1969) suggest two ways that skeletal mediation can be involved in autonomic activity. First in the case labeled parallel connection, some central change is mani- fested in both autonomic and somatic responding. For example in thinking of an uncomfortable experience a person might not only show changes in electrodermal activity and heart rate, but might also fidget in his chair. In the second case, that of overt chaining, a centrally initiated change in skeletal activity might percipitate a change in autonomic activity. For example, movement or changes in respiration can produce changes in electrodermal activity. If the model that Black and DeToledo label parallel connection is correct, the organism should show changes in addition to the change that is conditioned. For example a change in respiration or muscular activity might accompany a conditioned increase in the rate of the production of spontaneous GSRs. If it is the case that several changes accompany the change that the experimenter claims to have con- ditioned, then he cannot argue that the organism can directly control specific autonomic activity. The organism might instead be initiating some overall state of arousal which manifests itself in many ways. To control for the possibility that increased arousal could cause multiple responding, experimenters have shown that one response can be con- ditioned without effecting a second response. Kimmel and Kimmel (1967) measured GSR and vasomotor responding from two groups of §§ while con- ditioning GSR in one group and VMR in the other. Reinforcing GSRs significantly increased the frequency of spontaneous GSRs and decreased the frequency of drops in blood volume. Reinforcing VMRs increased the frequency of VMR and GSR changes. They attribute this result to the inadvertent reinforcement of simultaneous events since many GSRs were initially accompanied by VMRs. In deeply curarized rats, Miller and Banuazizi (1968) showed a similar phenomena by separately con- ditioning intestinal contraction/relaxation and HR increase/decrease. Changes in HR were not found when intestinal contractions were con- ditioned, and the rate of intestinal contraction did not change when HR was conditioned. Another study indicating specificity of responding is that of DiCara and Miller (1968b) who successfully conditioned rats to show vasodilation in one ear and vasoconstriction in the other ear. There were no changes in VMR from other body locations. Snyder and Noble (1968) reinforced vasoconstriction and did not find any systematic change in respiration rate, respiration irregularity, forearm electro- myographic recording (EMG) or HR during conditioning. In a human study, Shapiro, Tursky and Schwartz (1970) conditioned HR changes without finding accompanying changes in blood pressure. Engel and Gottlieb (1970) found the same thing with rhesus monkeys. When Plumlee (1969) conditioned blood pressure increases in monkeys, there was no change in resting pressure and even in extensively trained animals, no histo- logical indication of hypertension. 0n the other hand, DiCara and Stone (1970) have found different levels of catecholamine in rats trained in HR increase and those trained in HR decrease. This indi- cates a change resulting from conditioning which may imply mediation. Many researchers have employed masking tasks to control for possible influences of cognitive activity on (1) general arousal and therefore general somatic and autonomic activity or (2) specific autonomic responding. Gavalas (1967) had her §s read nonsense syllables from cards for one part of her study; another group of §s were asked to pronounce nonsense syllables spelled to them over an intercom. Ascough (1967) told his §s to guess about changes in their skin temperature while he was actually interested in HR changes. By contrast, three studies have suggested changes in general arousal level during conditioning indicated by changes in other auto- nomic responses. Changes in arousal level were indicated by Mandler, Preven and Kuhlman's (1962) study of GSR conditioning. They found that conductance levels rise in the control period and in the be- ginning of the reinforcement and the extinction periods. While con- ditioning HR Ascough (1967) found GSR changes which he said reflected arousal away from a balance. DiCara and Miller (1968c) recorded changes in core body temperature in rats conditioned to show vaso- dilation or vasoconstriction in the tail. Other studies have used statistical techniques to control for changes in other activity that might have accompanied the change brought about in the response under question by conditioning. The argument here is that even though there may be changes in other activity, conditioning may still be shown when these effects are statistically removed. Ascough (1967, 1968) used an analysis of covariance to control for the effect of respiration during his HR con- ditioning study. In their conditioning of salivation in dogs, Miller and Carmona (1967) wanted to correlate the number of days of training with the amount of salivation. After partialling out the effects of breathing and HR they still found significant changes in salivation over time. VanTwyver and Kimmel (1966) eliminated all GSRs that occurred within 5 seconds of a forearm EMG or respiratory irregularity. Differences between the experimental and control groups were still found. The model of overt chaining proposed by Black and DeToledo sug- gests that some subtle somatic change might initiate a change in auto- nomic activity. In this case conditioning an autonomic response might only have the effect of changing somatic activity which in turn causes the change in autonomic activity. Kendon Smith (1954) contends that neither operant ESE classical conditioning of autonomic activity can EXEE.be proved to be independent of somatic mediation. For any experiment that claims to demonstrate control of autonomic activity, he can argue that there was an undetected change in somatic activity that mediated the autonomic change. As an example of this, Obrist et a1. (1970) says that cardiac components are inextricably tied to somatic components. They report correlations between HR increase and EMGs from the chin, jaw, and mouth areas. Therefore to reinforce HR changes is to inadvertantly reinforce the subtle EMGs which can per- haps he said to "cause" the HR changes. Conditioning studies by Shean (1970) and Edelman (1970) traced the changes they found to mediators. Similarly Carroll (1970) found that motor behavior and respiration aided HR change. Some researchers have designed experiments to specifically eliminate or control for certain somatic mediation. Brener and Hothersal (1967) trained their §s to breathe at a fixed rate and amplitude. With §s breathing in this way they demonstrated successful HR conditioning. The HR changes observed could not have been attributable to changes in respiration rate or amplitude. By using §s who were controlling their breathing, and reinforcing only GSRs lO occurring in the absence of forearm EMGS, Rice (1966) demonstrated conditioning of the GSR independent of these two possible sources of mediation. Schwartz and Johnson (1969) initiated a time out period during which no‘S received reinforcement if he coughed, sighed, took deep breaths, made gross movements, or there were outside noises, or abrupt increases or decreases in heart rate. In this way no GSR that was elicited in any of these detectable ways was reinforced. One way to separate the effects of somatic and autonomic activity is to eliminate the possible confounding effects of somatic activity by the use of the drug curare which blocks virtually all muscle activity. Birk, Crider, Shapiro & Tursky (1966) demonstrated operant conditioning of skin potential (SP) in a single curarized human being. However, since curarization of a human is a complex procedure re- quiring the supervision of a skilled anesthesiologist, the technique is primarily limited to animal work. When curare is administered to animals to eliminate muscular activity and they are maintained on an artificial respiratory system, it becomes possible to demonstrate operant conditioning of autonomic activity independent of muscular or respiratory mediation. Curarization has been used particularly in studies modifying HR (Black, 1967a, 1967b; DiCara & Miller 1968a, 1969a; Hothersal & Brener, 1969; Hothersal, 1969; Miller & DiCara, 1967; Miller & Banuazizi, 1968; Trowill, 1967). It is possible that curarized animals who cannot make any movement are sending signals to their muscles to move but that curare blocks the expression of these. It could be that these movement signals also cause a change in autonomic activity -- and that these are being con- ditioned rather than the autonomic activity's being directly conditioned. 11 To test this, DiCara and Miller (1969b) initially conditioned HR in- creases in curarized rats. Then they observed the animals in the same situation after the effects of curare had worn off. If central messages to move had accompanied the HR changes, movement could have been expected when the curare wore off. They observed some initial movement, but this tapered off. Miller (p. 442, 1969) reports DiCara's observation that rats recovering from curare "that had been trained, through the avoidance or escape reward, to increase their heart rate were more likely to squirm, squeal, deficate and show other responses indicating emotionally than were those that had been trained to reduce their heart rate." These indicate that some signals for somatic changes may have accompanied the conditioning and this sug- gests that conditioning of the autonomic activity might not have been direct. The studies with curarized animals have either used an avoidance design or they have relied on electrical stimulation of the brain as a positive reinforcer. Unfortunately such designs make it difficult to compare the animal work with human work using more common reinforcers. There is, however, one very good animal study by DiCara and Miller (1968b) which argues against the necessity for somatic mediation in operant conditioning of autonomic activity. Using curarized and artificially respirated rats they conditioned vasodilation in one ear and vasoconstriction in the other. If the observed conditioning had been an artifact caused by a change in arousal level; vasomotor activity of the tail, body temperature and heart rate might have been expected to show changes - but no changes were observed in these indicators. If the animal controlled the right and left sides of the body 12 independently, in addition to the differences found in the ears, other differences in the two sides of the body might have been expected. However there was no change in either the right or left forepaw. This indicates extremely specific control of autonomic activity. In another study comparing responding in the right and left sides of the body, Wyatt and Tursky (1969) used human beings to show that somatic changes similarly effect both sides of the body. For example, when they delivered a shock to the left side of the body they did not find a larger skin potential response on that side than on the right side. They found similar responding on both the stimulated and unstimulated sides. However they found that most gs consistantly gave somewhat larger responses in one hand, for §s the right hand showed somewhat larger responses regardless of the side of the body stimulated by shock. Pilot work for the present study has indicated that somatic changes such as changes in respiration or movements produce similar changes in skin potential in both hands -- generally with a somewhat larger re- sponse in the right hand. Thus the question of operant conditioning of autonomic activity has by no means been answered. Not only did early studies use biased designs, but they also failed to address the question of possible mediation. Studies that have tried to see if change in other activity accompanies the change in the conditioned response have had mixed re- sults. Researchers statistically controlling for this can only control for the selected activity that they record. When experiments are designed to control for possible sources of somatic mediation, they have to limit their conclusions to those sources they controlled. Animal studies have eliminated somatic activity 137 curare but are open to l3 criticism because differently conditioned animals show different activity as the curare wears off. The animals have not learned to directly control only specific autonomic activity. In addition, the animal studies use techniques that are not practical for applications to human beings. Therefore the present study was designed to demonstrate operant conditioning of autonomic activity independent of somatic mediation. The present study follows a design similar to that of DiCara and Miller (1968b). It attempts to differentially condition electrodermal activity in the two hands; responses of a much larger magnitude in one hand than in the other hand were reinforced. Because responses elicited by somatic changes are of similar magnitude in both hands (wyatt and TUrsky, 1969), conditioning dissimilar responses in the two hands would demonstrate operant conditioning independent of somatic mediation. For this within §_design, half the gs received contingent reinforcement for their first sessions and non-contingent reinforcement for their later sessions; for the other half the order was reversed. In this way, S's performance during contingent reinforcement and non-contingent reinforce- ment could be compared. Because the operant level for responses of this type (much larger in one hand than the other) was very low, it was necessary to have several reinforcing sessions and to use a powerful reinforcer. Since recording GSR involves introducing a small current to the body and since introducing currents to both sides of the body simultaneously may interfere with the activity of the heart, the SP response was used to enable recording from both sides of the body with- out introducing a current. METHOD Subjects Twenty-eight male undergraduate students, enrolled in intro- ductory psychology courses at Michigan State University received extra course credit for their participation in the screening session experi- ment. Of these, 8 met the criteria and received a greater amount of extra credit for their longer participation in the study. Apparatus SP was recorded both AC and DC with Beckman biopotential elec- trodes placed at active site on §fs thenar eminance and a referant site on the inside of §fs arm 1 inch below the elbow. Attachments were made on the right and left sides of the body. A ground was placed about 5 inches below the left referant electrode. The AC record was used for determining criterion responses while the DC record served as a backup by recording all SP changes. Since movement artifacts are more clear on the DC record than on the AC record, this helped eliminate from the analysis all movement associated responding. The AC signal was recorded on two Grass Wide Band AC EEG Pre-amplifiers (7PSA) having time constants of .45 seconds and a sensitivity of l mv/cm. For DC recording, the signals from the same electrodes were also fed into two Grass low-level DC Pre-amplifiers (7P1A). For reinforcement, slides of female nudes were shown to male §§° This is the same reinforcement that Schwartz and Johnson (1969) used to successfully condition GSR. l4 15 This reinforcer has also been successfully used by Lovibond (1963) and DeFran, Radin and Lewis (1969).:1 The reinforcing stimuli were 55 chromatic slides of the female nudes taken from a popular magazine. The 10 in. x 10 in. image was shown by rear projection on a 2 ft. x 2 ft. ground glass screen placed 3 feet in front of S, The three- second duration of the slides was controlled manually by advancing the projector to a black slide which projected no image. Design and Procedure The first session for all §s was a screening session used to obtain operant levels for responding and to obtain a record of elicited responses. These sessions took place during the two weeks prior to the beginning of the experiment. Responding during these sessions was used as a basis for selecting §s for participation in the main experiment. The eight E? that were selected were then randomly assigned to Group I or Group II. The two groups differed in the order of the contingent and non-contingent reinforcement sessions. Table 1 indicates the sequence of the sessions. Insert Table 1 about here Group I had the following sequence: (a) four contingent rein- forcement sessions (C), (b) one extinction session (E), (c) four non- contingent reinforcement sessions (NC), and (d) one E session. The sequence for Group II was: (a) four NC sessions, (b) one B session, (c) four C sessions, and (d) one E session. Pilot work indicated little change in performance over subsequent extinction sessions. 16 Table 1 Sequence and Number of Sessions for Group I and Group II Week One Week Two Number of Number of Number of Number of Reinforcement Extinction Reinforcement Extinction Sessions Sessions Sessions Sessions Experimental 4 4 Group I Contingent 1 Non-contingent 1 Experimental 4 4 Group II Non-contingent l Contingent l 17 Subjects were seated in a dimly lighted sound attenuated room with an ambiant noise level of 51 db. and a temperature of approxi- mately 70°F. An intercom allowed communication between the experi— mental room and the adjacent recording room. Before attaching the electrodes and proceeding with the screening session, an explanation of the experiment was given to S and his permission was obtained. (See Appendix A). Subjects were instructed to remain alert and to re- frain from moving during the experiment. Only a male experimenter had contact with S until after the last session. After two minutes for machine adjustment, the beginning of the experiment was announced over the intercom and approximately fifteen minutes of baseline recording was made. Elicited responses were then obtained by delivering one re- inforcement and asking § over the intercom to do a series of things: hold his breath; move his feet; make a fist with the right then the left hand; move his head back and forth; breathe in and out rapidly; think emotional thoughts; relax as completely as possible; and to tense his whole body. §_waited until § had ceased responding to each stimulus before requesting the next action. The recording from the AC channel was used to determine the number of criterion responses. The criterion was production of a negative wave SP in one hand three or more times as large as the negative wave SP in the other hand (this is a 1:3 ratio). A large ratio would be most desirable because it would be most likely to be detectable by g, However, if too large a ratio was chosen, almost no §s would have a sufficient number of criterion responses for conditioning to be possible. The ratio was chosen on the basis of the pilot work. It was the largest ratio such that more than a third of those gs screened had a sufficiently 18 high operant level. In addition, the larger responses had to be of a magnitude of .3 mv. or greater.1 very small responses may not be mediated centrally (VanTwyver & Kimmel, 1966) and thus may not be amenable to operant modification. Using only the negative wave com- ponent of SP makes the criterion definition consistent for all §s and only eliminates a very small number of monophasic positive responses. Only §$ showing seven criterion responses with the larger response in the same hand during the fifteen minutes of baseline recording were continued in the study. This was done because some studies have in- dicated that at least thirty reinforcements are necessary for this type of conditioning (Milstead, Baer & Fuhrer, 1968; Crider, et al., 1966). If S continued to give at least seven criterion responses for each of the four conditioning sessions, he would receive close to thirty reinforcers. Subjects whose elicited responses consistently met the criteria were eliminated since criterion responses could not be considered as independent of somatic mediation in that case. Subjects selected in this way had five afternoon sessions a week for two weeks. Each S was instructed only that the recordings would be made during the baseline session and that he should see what he could do to view as many slides as possible. Non-informative instruc- tions were selected because Engle and Hanson (1966) found that all four 'Ss who correctly inferred that HR changes were being reinforced were among the five non-learners. In addition, for a somatic response, Hefferline (1962) found poorer performance among §s without such 1Because of the filers used (3% amp hi freq on the driver amplifier and .15% amp lo freq on the AC preamplifier) responses are somewhat attenuated. Therefore a response recorded as .3 mv. is actually some- what larger. Since the attenuation is constant, the criterion magnitude is consistent although not exactly .3 mv. l9 instructions. Greene and Nielson (1966) found that §s who scored high on Mandler's Autonomic Perceptibility Questionnaire showed less in- crease in GSRs than §s with less autonomic awareness. Each session consisted of twenty minutes of recording, then S was asked how he thought he did, and what ideas he had about how he might be controlling the slides. During the contingent reinforcement sessions, reinforcement was delivered when it was the consensus of the two experimenters that a criterion response had occurred. When a criterion response occurred a slide was shown for three seconds, then there was a ten second blackout period. Since responses during this blackout period were probably elicited by the reinforcement, they were not reinforced even if they met the criterion. Blackout periods of this length were used by Crider, et a1. (1966). During the extinction sessions, no reinforcement was delivered. During the non-contingent reinforcement sessions, reinforcements were delivered at random intervals and followed by the same blackout. The total number of non- contingent reinforcers delivered for each session was approximately one and one half times the number of criterion responses that occurred during the initial baseline session. This meant that more reinforcers were delivered during the non-contingent sessions than during the con- tingent sessions. This should give an upward bias to the number of criterion responses during the NC period and provide a conservative measure with which to compare the C periods. Data Reduction and Analysis To reduce the eighty experimental records to manageable form, the magnitude of all responses were determined. Only responses that met the criteria were considered in the total number of responses. 20 Responses that were associated with sound from the experimental chamber that might have indicated movement or some bodily change such as a cough were eliminated since these were clearly elicited by something beyond the experimenter’s control. Responses where neither recording channel indicated a change of .3 mv. or more were not included in the total because these small responses may not be detectable to S, Positive wave SP responses were also eliminated because these may be controlled by a different mechanism than negative wave SPs. As an exception, responses that had a positive component less than the .3 mv. detectable level were included as long as the responses on the other side of the body was negative going and at least .3 mv. The responses to the "begin" and "end" announcements were not included in the total group of responses. Responses included in the total that also met the criterion of being three or more times as large in the designated hand as in the other hand and of being of minimum magnitude (.3 mv.) in at least one hand were called criterion responses. All responses were classified as either spontaneous or elicited. Elicited responses were defined as all responses occurring within thirteen seconds of the onset of the reinforcer or within thirteen seconds of the peak on the AC channel of a criterion response. This thirteen seconds included the three seconds during which the slide was exposed and the ten second blackout period. Those responses that were not elicited in either of the above ways were called spontaneous. These spontaneous responses were therefore not elicited in any way detectable to the experimenter. To look at the relationship between the number of spontaneous responses and total responding, a ratio was made of the number of spontaneous criterion 21 responses divided by the total number of spontaneous responses. This gives the proportion of the spontaneous responses that met the criterion. An analysis of variance was used to determine the effects of the experiment on these variables. In each of these analyses, some of these four factors were used: (1) Group -- was §_in Group I or Group II. (2) Contingency -- during that week was § receiving contingent or non-contingent reinforcement. (3) Sessions -- this refers to the session of the week regardless of the contingency of reinforcement. (4) Time -- this refers to the time periods within the sessions. To see if there might have been some effects within the contingent sessions or within the non-contingent sessions, simple effects were tested. This analysis served to indicate if there were significant differences within the contingent sessions considered by themselves and within the non-contingent sessions by themselves. All eight §s completed their ten experimental sessions. For one ‘S scheduling difficulty arose which necessitated running two sessions in one day -- one in the morning and one in the afternoon. All others had one session at the same time each weekday for two weeks. On three separate occasions the slide projector failed to operate correctly and the reinforcing slides were exposed for much longer than three seconds. Since these sessions cannot be appropriately included in the analysis, means from the other §s in that condition on that day were used as data points for these gs in the analysis. RESULTS Screening Session - Baseline Recording Twenty-eightlgs had screening sessions before eight were selected who had sufficiently high operant levels. One additional § had a sufficiently high operant level, but was unable to find time to participate in the study. One of the selected gs showed only six criterion responses during twenty-two minutes of baseline recording. He was included because such a high proportion of his small number of total responses met the criterion. For the two §s whose baseline re- cording periods were only 10 minutes long, the number of responses that might have occurred during the next 5 minutes was assigned to the third 5 minute block of time. This method of estimating was chosen since there did not seem to be consistent increases or de- creases over time for the other Ss. The same method of adjustment was used to estimate the number of responses that would have occurred if their first contingent reinforcement session had continued for 20 minutes instead of 10. So that performance during the 15 minute screening sessions and the 20 minute experimental sessions could be compared, an estimate was made of the number of responses that might have occurred if the screening session had extended 5 additional minutes. The number of responses during the first 15 minutes was multiplied by 4/3 to obtain this estimate. 22 23 For the eight Ss that were selected, performance during the screening sessions was used to determine which hand showed larger responses during criterion responding. Four of the gs showed larger responses in the right hand, 4 showed larger ones in the left hand. The number of criterion responses during the fifteen minutes varied from 3 to 31 with a mean of 18.75 (§2_= 10.138). For two Group I subjects, no criterion responses in the criterion direction were emitted during the first ten minutes of their first contingent session, while a large number of criterion responses in the opposite direction were emitted. For this reason, the criterion direction was changed for one subject from right to left and for another from left to right at that point. For these two subjects, the number of criterion responses during the first ten minutes of their C session was used for baseline data. The number of AC responses made by each § during the screening session can be seen in Table 2. The total number of responses and the criterion responses are separately divided into elicited and spontaneous responses. The number of positive going responses varied Insert Table 2 about here from none to 14 -- these responses are indicated in Table 2 but have not been included in the total number of responses. Table 3 gives the estimated number of responses if the screening session had con- tinued for 20 minutes. These estimated numbers can be compared directly to the data from the twenty minute experimental sessions. 24 Table 2 Number of AC Responses During Fifteen Minutes of Screening Session Number Number Number Number Number Criterion Total Positive Criterion Total Subject Responses Responses Responses Elic. Spon. Elic. Spon. 1 ll 94 0 3 8 ll 83 2 3 27 4 O 3 0 27 3 26 62 14 7 19 13 50 4 10 28 l l 9 l 27 5 31 80 2 9 22 23 57 6 18 74 O 3 15 16 58 7* 30 73 6 O 30 21 52 8* 21 50 l 8 13 12 38 Total 150 488 28 31 119 97 392 Mean 18.75 61 3.5 3.8 14.8 12.1 49 *Tabled data estimated by taking 3/2 of the number of responses occurring during the ten minutes of baseline recording. 25 Insert Table 3 about here Screening Session - Elicited Responses The initial AC responses elicited by asking §_to perform certain actions can be seen in Table 4. These initial elicited responses can be compared with (l) the first §_responses elicited by each stimulus (2) the data from the 15 minutes of baseline recording (3) the responses during all 8 experimental sessions. In this way it is possible to see if the elicited responses differ from those not elicited. Elicited responses were more often positive going than Insert Table 4 about here non-elicited responses. 36.04% of the initial elicited responses and 28.50% of the first 5 responses elicited by each stimuli were posi- tive going while only 5.42% of the responses in the 15 minutes of baseline recording and 6.75% of the responses from the 8 experimental sessions were positive going. This indicates a difference in the typology of the responding for elicited and non-elicited responses. These comparisons can be seen in Table 5. The percentage of criterion responses occurring in these conditions is also presented in Table 5. A smaller percentage of the elicited responses meet the criterion than do the non-elicited responses. This is especially true of the initial elicited responses where only 1.16% of the total responses were criterion responses. By contrast, 13.9% of the total responses 26 Table 3 Estimated Number of AC Responses if Screening Sessions had Continued for Twenty Minutes* Number Number Number Number Number Criterion Total Positive Criterion Total Subject Responses Responses Responses Elic. Spon. Elie. Spon. l 15 126 0 4 ll 15 111 2 4 36 5 O 4 O 36 3 35 84 19 10 25 17 67 4 13 37 l 1 l2 1 36 5 41 107 3 12 29 31 76 6 24 98 O 4 20 21 77 7 40 97 8 O 40 28 69 8 28 67 1 ll l7 16 51 Total 200 651 37 41 158 129 523 Mean 25 81.4 4.6 5.1 19.7 16.1 65.4 *Tabled data estimated by taking 4/3 of the number of responses occurring during the fifteen minute screening session. 27 m.oI :.mI u.m+ m.m+ A.~+ m.A+ m.~+ ~.A+ ~.AI m. I ~.AI A. I n.~+ m.m+ seen mason s.~I A.~I A.mI m.~+ o. I a. I a. I a: swan meou m.nI :.mI o.o+ m.n+ m.m+ o.m+ o.A+ m. + m. I A. I Amcowuoao chAA A.~I m.AI a.m+ A.m+ «.3. m.¢+ N. I A. I o. I m. I ~.AI :. I :.~+ m.A+ sAoAamu onuwoun Hum- Mom”!- NOJ+ Com...- m. .l HON-T Mo I m. In No -l w. .l m. .l COM-T CON... ““8 gOE m.oAI A.mI 0.8+ ~.k+ :. I A. I a.m+ m.:+ a. I o.AI m.m+ o.A+ o.m+ m.o+ ~.:+ AmAm AmmA ~.AI A.~I m.o+ m.o+ m. I m. I ~.:+ o.o+ o. + ~.~+ m. I m. I «.mI ~.m+ m.m+ A.~+ AmAu AewAu o.mI m.nI A.m+ o.oI m.AI o. I N.MI m.~+ m. + o. I o.m+ an :.A+ m.:+ m.aI m.AI use“ 0>OB A.mI N.AI o.mI o. + m.AI m. I m.~+ m.~+ A. I m. I o.m+ :.~+ m.o+ N.m+ m.AI ~.AI use Apmmun m.A+ A.N+ m.AI o.MI N.AI m.AI A.~I m. I a. I o. I o.AI o.AI o.~I m. I o.AI ~.AI Aummun cAoA m.A+ o.m+ m.~I «.mI m.~I :.AI A.AI o.AI m.mI m.NI m.NI o.~I +m.mI +m.~I moAAm A m A m A m A m A m A A A A A m can: m A pomnnsm .>8 cw maoccmao u< puma cam map :0 momnoamom wouwowam HMHHHGH 3 oHnt AAwAm mo owswwawmz 28 were criterion responses in the experimental sessions. Table 5 also shows that a higher percentage of elicited responses were of greater magnitude than were non-elicited responses. Large responses were defined as those that were 2.5 mv. or larger or at least one hand. Thus the typical elicited responses was positive going, larger, and less likely to meet the criterion than were other responses. Insert Table 5 about here Reinforcement To bias conservatively the number of reinforcements delivered during non-contingent reinforcement, a number of reinforcements approximately one and one-half times the number of criterion responses that occurred during the 15 minutes of baseline recording were delivered for each non-contingent reinforcement session. In this way, more reinforcers (mean = 24.06, §2_= 12.36) were delivered during non-contingent reinforcement sessions than during most con- tingent reinforcement sessions (mean = 8.22, §2_= 8.17). Therefore results seen are unlikely to be due merely to sensitization to the reinforcer. For contingent reinforcement, two experimenters watched the record and determined visually if each response was of criterion magnitude and if a sufficient time had elapsed since the previous reinforcement so that the response could be considered spontaneous. When there was consensus between the experimenters, reinforcement was delivered manually. The number of contingent reinforcements delivered per session varied from O to 24 with a mean of 8.22 (SD = 8.17). 29 Table 5 Percentage of AC Responses that were Positive Going on Both Channels, of Criterion Magnitude in the Criterion Direction, and Greater than 2.5 mv. on the AC Channel.* First Five Fifteen Min. Initial Responses Baseline All Eight Percent of Elicited Elicited By Recording From Experimental Responses in: Responses Each Stimuli Screening Sess. Sessions That Were 36.04 28.5 5.42 6.75 Positive Going Of Criterion Magnitude in the Criterion 1.16 7.45 29.06 13.90 Direction Greater than 2.5 mv. on at least One Hand (AC) and 12.79 17.10 2.32 5.67 Negative Going * Only for the calculation of these percentages were positive going responses included in the total number of responses. 30 Because of the instantaneous nature of the judgment about whether or not a response met the criterion, there were errors in re- inforcement. Most of these consisted of reinforcing a response which was slightly under the criterion level or of failing to reinforce a response slightly over the criterion level. Over all the contingent reinforcement sessions, 43 responses were reinforced that were not criterion responses, and 70 spontaneous criterion responses that should have been reinforced were not. A total of 199 reinforcements were correctly delivered. The total number of correct contingent reinforcers that each S received varied from 5 to 56 with a mean of 24.87 (SD = 16.72). Criterion Responses If conditioning had occurred, more criterion responses should have been seen during contingent reinforcement than during non-con- tingent reinforcement. Since only spontaneous criterion responses were reinforced, primarily the number of spontaneous responses was expected to increase. Table 6 contains the analysis of variance for the number of spontaneous criterion responses occurring during the ten sessions. It can be seen that there is a significant group effect. Insert Table 6 about here Subjects who had the contingent sessions first (Group I) had a lower level of criterion responses during their sessions than §s who began with non-contingent reinforcement. During their first two contingent sessions they had more criterion responses than during their first Analysis of Variance for the Number of Spontaneous Criterion Responses Occurring During the Ten Sessions Table 6 SS df MS F Contingency 68.45 1 68.45 1.11 Group 414.05 1 414.05 6.72 Sessions 199.17 4 49.79 —- Contingency X Group 115.20 1 115.20 1.87 Contingency X Sessions 267.98 4 66.99 1.08 Group X Sessions 110.82 4 27.70 -- Cont. X Group X Sessions 159.13 4 39.78 -- Within cell 3699.22 60 61.65 -- Total 5034.02 79 -- -- *Significant at .05 level -II ‘:‘“II I 32 two non-contingent reinforcement sessions. Performance during the 3rd and 4th acquisition sessions and the extinction session was similar regardless of the contingency/non-contingency for this group. Subjects who began the experiment with their non-contingent reinforce- ment sessions (Group II) showed a higher level of criterion responses during their sessions than the Ss in Group I. For the Group II Ss, there were a larger number of criterion responses in their first contingent session than in the first non-contingent session. But the third session contained more criterion responses under non-contingent reinforcement than under contingent reinforcement. Contingent and non-contingent reinforcement produced similar results in the other sessions. These effects can be seen in Figure 1. Although §s were Insert Figure 1 about here randomly assigned to Group I and Group 11, there was a significant difference between the number of spontaneous criterion responses for the groups during the screening session. This initial difference explains the difference between groups that was seen during the experimental sessions. Combining group one and two §s to look at the effect of con- tingency gives a slight interaction. During contingent reinforcement the number of criterion responses decreases fairly steadily across sessions. The test for linear trend here was almost significant (F(1,60) = 3.23, P< .10). During non-contingent reinforcement the number of criterion responses starts much lower, but rises above the 33 16 " NON—CONTINGENT 0—. CONTINGENT 0—-..) 14 I. 12 ., 10 .. 3.. MEAN NUMBER OF RESPONSES .p I l I l I i ‘r . SESSIONS 1 2 3 4 extinction Figure l. The mean number of spontaneous criterion responses by group I and group II_§s during contingent and non-contingent reinforcement sessions. 34 number that there were during the third contingent reinforcement session. The number then drops back. This can be seen in Figure 2. Insert Figure 2 about here During conditioning, it would have been expected that the number of spontaneous criterion responses would increase. The decrease that was seen may indicate a decrease in §fs arousal or total responding -- this will be discussed later. Findings for the total number of criterion responses (including those elicited) were similar. The differences between groups was also significant (F(1,60) = 6.78, p‘(.05). An analysis of variance for simple effects of sessions (p>>.05), and Newman Keul's analysis did not indicate any significant effects (p ) .05) . Total Number of AC Responses An increase in the total number of responses may accompany the increase in criterion responses seen as a result of conditioning. This would indicate less specificity of control than if the total number of responses were not affected. First, the total number of responses (elicited and spontaneous, criterion and not) will be examined. Group II gs gave a number of responses during their first 3 non-contingent sessions far larger than they gave in their first 3 contingent sessions or Group I gave in their contingent or non-contingent sessions. This indicates that the highest level of responding was ob- tained during the first week by subjects who received non-contingent reinforcement then. This shows up as a slight 35 14' CONTINGENT 12It lO-- 8‘. 64b (0 m (D Z 8 E NON- CONTINGENT 84 l I- x: g 22 «I. L I TL I h? SESSIONS l 2 3 4 extinction Figure 2. The mean number of spontaneous criterion responses during contingent and non-contingent reinforcement sessions. 36 interaction between contingency of the reinforcer and group of the subject (F(l,60)= 3.14, p<(.lO). It may be that the larger number of reinforcers delivered there kept responding at a higher level which constrasted with the lower levels of the second week and of the con- tingent reinforcement. It cannot be explained as a §_effect since groups did not differ during screening. This relation can be seen in Figure 3. When the extinction sessions were excluded from the Insert Figure 3 about here analysis there was a non-significant trend for the contingency of reinforcement (F(l,48) = 3.65, p.(.lO). MOre responses were given during non-contingent reinforcement than during contingent reinforce- ment for the conditioning sessions. This is probably attributable to the increased stimulation of more slides. Analysis of simple effects for contingency of reinforcement did not yield significant results although both F values were greater than one. For contingent reinforcement this reflects the increase in responding seen during extinction; for non-contingent reinforcement this is primarily a result of decreased responding during extinction. Newman Keuls analysis did not indicate any significant differences between individual sessions. To examine the effect of the increased number of reinforcers delivered, the number of elicited responses were examined. There were significantly more elicited responses during non-contingent reinforce- ment than during contingent reinforcement (F(1,48) = 28.23, p<.Ol). 37 lqu- O———--. NON- CONTINGENT 0-— — —o CONTINGENT I 100 'I 90. I GROUP II 80 .I 70.- GROUP I MEAN NUMBER OF RESPONSES I l A v v SESSIONS l 2 3 4 extinction Figure 3. The mean total number of AC responses by group I and group II_§s during contingent and non-contingent reinforcement sessions. 38 This effect stands to reason because there were more reinforcers delivered which could potentially elicit responses during non-con- tingent reinforcement. There was also a small but not significant interaction between this contingency of reinforcement and st group (F(1 48) = 3.70, p .10). The interaction can be seen in Figure 4. 3 Insert Figure 4 about here Since there were no reinforcers delivered during extinction, responses called elicited only include those occurring within thirteen seconds of a criterion response. For this reason, far fewer elicited responses were seen during extinction. When the analysis of variance is per- formed on all sessions - including extinction sessions -- there is a significant sessions effect (F = 3.90, p(.01) which indicates (4,60) that far fewer elicited responses were seen during extinction. A Newman Keuls analysis showed this significant difference between the extinction session and each of the contingent reinforcement sessions (p .05). There is also a significant interaction between the con- tingency of the reinforcer and sessions (F(4,60) = 3.72, p .05) which is primarily contributed to by the difference during extinction. These effects can be seen in Figure 5. Within the non-contingent week, the simple effect for sessions was significant (F(4,60) = 6.84, p .01). This reflects the rise to the middle sessions and the drop for extinction. The simple effect for sessions was not significant within Insert Figure 5 about here 39 30-- 25P 20m H 9‘ I MEAN NUMBER OF RESPONSES 5 ‘13 l I WEEK contihgent non-contingent Figure 4. The mean number of elicited AC responses made by group I and group II §s during contingent and non-contingent reinforcement. 4O NON—CONTINGENT 30‘- 254. 20 II m \ m z \ c> g \ 5L9 K\ h CONTINGENT \\ o ._ .05). Comparisons with Baseline and Extinction Data To compare the contingent conditioning sessions with baseline data from the screening session, first a two~way analysis of variance was run. The effect of the sessions (baseline, 4 contingent, extinc- tion) and group were determined, then orthogonal comparisons were made to compare the baseline data with the 4 conditioning sessions. There were significantly more criterion responses and spontaneous criterion responses during the screening session than during the contingent re- inforcement sessions (for criterion responses, F(1,36) = 8.26, pI(.Ol; for spontaneous criterion responses, F(l,36) = 7.89, p(.Ol). This decrease in the desired response tends to argue against conditioning, but it could still be possible to show that an operant conditioning procedure slows the normal decrease in criterion responding. None of the measures of total responding showed a significant change from the screening session to the contingent reinforcement sessions. Similarly there was no significant difference for the ratio measure of the .MEAN RATIO 48 .35 ‘- .30 :r- .25 ‘b .20 3' GROUP I GROUP II 0' \ up I I V 1 SESSIONS l 2 3 4 extinction ‘- I v Figure 9. The mean ratio (the number of spontaneous criterion responses / number of spontaneous responses) for group I and group II gs. 49 production of criterion responses. To compare performance during the reinforcement sessions with performance during the extinction sessions, orthogonal comparisons were made. These comparisons contrasted performance during the four reinforcement sessions (either contingent or non-contingent) with per- formance during the extinction session which followed. This was done for each of the variables previously analyzed. For the number of criterion responses (total or spontaneous) the number did not differ for the contingent sessions and extinction or for the non-contingent sessions and extinction. There were, however, more criterion re- sponses on the average during contingent reinforcement than during non-contingent reinforcement or extinction. This can be seen in Table 8a, 8b. For the contingent reinforcement week, there is an al- most significantly higher total number of responses during extinction than during the reinforcement sessions (F(l,60) = 3.98, p<(.lO). This may reflect some frustration or increased arousal with withdrawal of the reinforcer after S has learned the task. When only spontaneous responses are considered, the effect reaches significance (F(1,60) = 7.55, p(.Ol). This increase in total re- sponding during extinction has the effect of depressing the ratio. However, during contingent reinforcement the ratio is above what it is during non-contingent reinforcement. Therefore the ratios are signifi- cantly different for the contingent reinforcement sessions than for the extinction that follows (F(1,60) = 4.68, p<(.05). These can be seen in Tables 8c, d, e. There is an increased ratio during conditioning which drops during extinction -- the drop being primarily due to an increase in total responding. 50 Insert Table 8 about here Comparisons of Elicited and Spontaneous Responses The responses elicited by the reinforcer were compared to the spontaneous responses. While it was not expected that no elicited responses would meet the criterion, it was expected that there would be no more elicited responses than spontaneous responses meeting the criterion and that there might be less. This is central to the argument that the criterion responses are not elicited by some somatic change. For if more elicited responses met the criterion, it could not be claimed that the criterion responses were independent of somatic mediation. To test this, the proportion of criterion elicited responses and of criterion spontaneous responses was calculated separately for contingent and non-contingent reinforcement sessions. Extinction sessions were not included since there was no reinforcement to elicited responses. Results of this can be seen in Table 9. Insert Table 9 about here An analysis of variance was performed for these proportions of criterion responses. There was a significantly higher proportion of criterion responses within the spontaneous responses than within the elicited responses (F(1,120) = 5.84, p( .05). This was as expected and indicates that elicited responses are less likely to be criterion responses than Spontaneous responses are. The effect for the con- tingency of the reinforcer was also significant :- \‘ .I I \ I I I h .. -'0‘ \ -" I \ \I I 0 di- A II \ 1 , \ I \ 3,,‘0. \ /’ ‘OD——‘d ba’ ‘D I ‘~ - -" -'» . I .L I : i I I SESSION sc ncl nc2 nc3 nc4 e c1 c2 c3 c4 e Figure A4. Responding for S_# 4. 84 total 140 ” responding ---- spontaneous responding 130 1b . 0 total '0 elicited ‘4 spontaneous 120 o llOJ. 1004p 901- 80.. 70 II- 604» 50+' 40 ‘D R z: \ cn \ g I m 30 qr a ’\ \ In \ \‘\ ' <320 \\b—-—";O,“ g b” \‘q‘ I, \ 2“ II \\ \ I \ 510 "' q\ \\\ N, \ b. I”! \\ \ \ \ \‘ I / \\ \‘ \ ” Xn‘\\ ’/ \\ Q“ 2E”’D\‘--O’/ “‘Zi—vbu \m’5 . t ‘ 4r j : I t ; 332:0— SESSION sc c1 c2 c c4 e ncl nc2 nc3 nc4 e Figure A5. Responding for_§ # 5. 85 total responding ‘--- spontaneous responding O 0 total Ia elicited ‘4 spontaneous lOO -- T 90 1. 80 .p 70 in 60 .. 50 n- m40 ., m g 8 0:30 'P 5 $20 ‘F’ \ \ E ‘ \ 210 .. \ ‘fx _ - SESSION sc cl c2 c3 c4 e ncl nc2 nc3 nc4 e Figure A6. Responding for S # 6. 100« 80 ‘ 7O . 50 40 . L0 0 NUMBERNQF RESPONSES <3 [.1 O SESSION nc c1 Figure A7. 86 c2 c3 c4 e Responding for S # 7. total responding ----spontaneous responding . 0 total [a elicited ‘|‘§ spontaneous -- 0‘0- ... - «mo- ‘ A .. ._. ncl nc2 nc3 nc4 e 100 n 90 <- 80 4t- 70.“ 60." so -. N U) -I-\ O O O l l 1 r V I I NUMBER OF RESPONSES .13 H O 5 JD I SESSION sc cl Figure A8. 87 c2 c3 c4 e Responding for S # 8. total responding “--- spontaneous responding .0 total In elicited ‘ A spontaneous ncl nc2 nc3 nc4 e llllllllllllILHHHIHIIIIH 6 0 4 5 7 1 3 mumm 0 umumumu