VER‘BAL CONDITIONING AND TRANSFER EFFECTS IN AN INTERVIEW SETTING Thesis for the Degree of Ph. D. MICHIGAN STATE UNIVERSITY CALVIN R= KING, IR. 1968 ............... LIBRARY L5 Michigan State University THESIS This is to certify that the thesis entitled Verbal Conditioning and Transfer Effects in an Interview Setting presented by Calvin R. King, Jr. has been accepted 'towards fulfillment of the requirements for Ph.D. degree in Psychology 9W WWW Major professor Date August 9. 1968 0-169 ; Er amnme av . 1 IIOAG 8: SIIIIS’ -- ‘ 800K BINIIERV- INC. LIBRARY BINDEHS I...“.---- an.-.“- - .. ABSTRACT VERBAL CONDITIONING AND TRANSFER EFFECTS IN AN INTERVIEW SETTING By Calvin R. King, Jr. A study was designed attempting to (1) condition negative and poSitive self-references within a series of three 20—minute interviews, using reflection as the verbal reinforcer; (2) demonstrate transfer of conditioning ef- fects to a battery of five self—evaluative measures (two Semantic Differentials, two Q-Sorts, and the IPAT Anxiety Scale) administered pre— and post- interview. Seventy- five 83 were randomly placed in five experimental and con— trol groups, 15 $3 to a group. Group E1 was reinforced for positive self-references; Group E2 was reinforced for negative self-references; Group E for both positive and 3 negative self-references; Group C1 was an interview con- trol group; Group C was a test-retest, no interview con- 2 trol group. The self-evaluative test battery was admin- istered immediately before the first interview and imme— diately following the last interview, except in the case 1 Calvin R. King, Jr. of C2 controls who received the battery on a three to four day test-retest interval. Self-reference data were subjected to a nested analysis of variance design. Individual comparisons were made among cell members of the Group x Interview x Re- sponse interaction. Results confirmed that conditioning had occurred as predicted. Both positive and negative self-references were shown to be amenable to conditioning employing reflection as the reinforcer. Each self-evaluative measure was subjected to an analysis of covariance. Only changes on the IPAT Anxiety Scale reached significance. However, individual compar- isons applied to the IPAT data demonstrated that condi- tioning did not effect transfer whereas being interviewed did. Thus, all hypotheses regarding transfer were re- jected. Implications of the conditioning results were noted, with specific reference to client-centered psycho— therapy and to the concept of shaping as applied to verbal conditioning within therapy interviews. The personal and interpersonal context of verbal conditioning within in- terviews was stressed. Calvin R. King, Jr. Comments on the absence of transfer effects at- tributable to conditioning were made. Reference was made in particular to the concept of stimulus generalization and to the hypothesized desensitization rationale. It was suggested that variables affecting decreased anxiety scores were imbedded within the interpersonal context of the interviews. In essence, the study again demonstrates that verbal behavior can be modified through reinforcement in a quasi—therapeutic setting, but that the issue of trans— fer remains equivocal. VERBAL CONDITIONING AND TRANSFER EFFECTS IN AN INTERVIEW SETTING BY T\* \ Calvin R. King, Jr. A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology 1968 DEDICATION To Jeanne, who contributed in so many immeasurable and thoughtful ways to her husband's graduate career, not the least of which was her patient faith and optimism that this dissertation would be completed. ACKNOWLEDGEMENTS The author expresses his deepest appreciation to Dr. Lucy Rau Ferguson, chairlady of his dissertation and guidance committees, for her ever present support and en- couragement throughout the preparation and completion of this dissertation; and to Drs. Clarence L. Winder, Donald Grummon, and Charles Hanley for their knowledgeable com- ment and criticism. A debt of gratitude is owed them for creating an atmosphere of instructive freedom within which this dissertation was prepared. iii TABLE OF CONTENTS Page ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . iii LIST OF TABLES . . . . . . . . . . . . . . . . . . Vi LIST OF FIGURES. . . . . . . . . . . . . . . . . . vii LIST OF APPENDICES . . . . . . . . . . . . . . . . Viii INTRODUCTION . . . . . . . . . . . . . . . . . . . 1 BACKGROUND . . . . . . . . . . . . . . . . . . . . 5 PROBLEM. . . . . . . . . . . . . . . . . . . . . . 12 METHOD . . . . . . . . . . . . . . . . . . . . . . 14 Definitions. . . . . . . . . . . . . . . . . . 14 .Subjects . . . . . . . . . . . . . . . . . . . 15 Groups . . . . . . . . . . . . . . . . . . . . 16 Interviews . . . . . . . . . . . . . . . . . . 17 Criterion Measures . . . . . . . . . . . . . . 18 Procedure. . . . . . . . . . . . . . . . . . . 20 HYPOTHESES . . . . . . . . . . . . . . . . . . . . 22 Verbal Conditioning. . . . . . . . . . . . . . 22 Transfer . . . . . . . . . . . . . . . . . . . 24 RESULTS. . . . . . . . . . . . . . . . . . . . . . 28 Verbal Conditioning. . . . . . . . . . . . . . 28 Transfer . . . . . . . . . . . . . . . . . . . 39 iv Table of Contents.—-Continued Page DISCUSSION . . . . . . . . . . . . . . . . . . . . 44 Verbal Conditioning. . . . . . . . . . . . . . 44 Transfer . . . . . . . . . . . . . . . . . . . 48 SUMMARY. . . . . . . . . . . . . . . . . . . . . . 54 REFERENCES . . . . . . . . . . . . . . . . . . . . 57 APPENDICES . . . . . . . . . . . . . . . . . . . . 62 LIST OF TABLES Table Page 1. Nested Analysis of Variance for Self- Reference Data . . . . . . . . . . . . . . 31 2. Breakdown by Mean Self—References of Main and Interaction Effects for Nested Analysis of Variance Design. . . . . . . . 33 3. Duncan's New Multiple Range Test Applied to the Differences Between Selected G x I x R Mean Self-References . . . . . . . . . . . 36 4. Analyses of Covariance of Self-Evaluation Measures . . . . . . . . . . . . . . . . . 40 vi LIST OF F IGURES Figure Page 1. Experimental design. . . . . . . . . . . . . . 20 vii LIST OF APPENDICES Appendix Page A. INTERVIEW INSTRUCTIONS AND TOPICS. . . . . 63 B. SEMANTIC DIFFERENTIAL INSTRUCTIONS AND SEMANTIC DIFFERENTIAL. . . . . . . . . . 65 C. Q-SORT INSTRUCTIONS AND Q-SORTS. . . . . . 69 viii VERBAL CONDITIONING AND TRANSFER EFFECTS IN AN INTERVIEW SETTING INTRODUCTION The past decade has witnessed tremendous growth in the volume and quality of research in psychotherapy as efforts have been directed toward unraveling the "mystery" of this once rather sacrosanct subject and exposing it to rigorous investigation. Among the several strategies em- ployed to seek out and define the variables affecting psy- chotherapeutic process is the application of the principles and techniques of reinforcement learning. This is the framework of this dissertation. Krasner (1962, p. 61) states two assumptions in- volved in a reinforcement learning approach in the study of psychotherapeutic processes: (a) Psychotherapy is a lawful, predictable, and directive process which can be investigated most parsimoniously within the framework of a reinforcement theory of learning. (b) The var- iables which affect the therapy process are the same as those in other interpersonal situations which involve the reinforcement, control, manip- ulation, influencing, or redirection of human behavior. Hence, psychotherapy, regardless of the eXpressed or in— ferred orientation of the therapist, may be subjected to investigation within this framework. The major overt events in psychotherapy sessions are the words spoken by the patient and the psychothera— pist. Research shows that the psychotherapist controls the frequency of many classes of words, phrases, and other aspects of verbal expression by selective attention and responsiveness. The client verbally expresses a thought or action, feeling or attitude. The therapist may or may not respond, and if he does, he may do so in a number of ways, including a verbal response which may serve as a stimulus for a further response from the client. Assum- ing that the psychotherapist has, in his concept of therapy,' some notion as to what the client should and should not be talking about, the therapist responds in such a way as to reinforce some statements and extinguish others (Bandura, 1961, p. 150). In essence, the psychotherapist is manip— ulating the verbal behavior of his client. Granting that the therapist can modify the verbal behavior of the client during the therapy hour, associated behavioral changes outside the office should be studied. Kanfer (1961, p. 690), commenting upon this issue, notes that the ultimate goal of therapy is to return con- trol over the patient's behavior once again to the patient, after the controlling responses themselves and the behavior on which they are contingent have been modified. Continued self- reinforcement should allow the patient to pro— gressively strengthen those behavioral patterns and new perceptions which have been acquired during the therapy hours. What needs to be demonstrated, then, is that the modifica- tion of verbal behavior during the therapy hour is asso- ciated with desirable changes outside therapy, rather than merely demonstrating that the client learns to talk dif- ferently with few or no concomitant changes. However, establishing a relationship between verbal conditioning within the psychotherapy hour and favorable changes in behavior outside the therapist's office is problematic. Winder (1957, p. 320), commenting on a study of the control of verbal behavior, states: One line of thought would consider psychotherapy as a complex and not always efficient condition- ing process in which the patient learns to talk differently and little else. Another line of Speculation . . . is that while there are orderly changes in verbal behavior, the essen— tial correlates are changes in the patient's perception of himself as a result of the fact that someone else spends time with him, takes what he says seriously, and shows him respect. Or, one might suppose the reorganization of verbal behavior would be accompanied by sub- stantial modifications in other behaviors, es— pecially those which are talked about. After some years of research, the issue still remains in doubt: Thus, most investigators continue to report suc- cessful conditioning, even of complex phenomena. However, evidence for generalization or transfer of such effects is equivocal (Matarazzo, 1965, p. 212). BACKGROUND The review of the literature will cover those studies which have employed the verbal conditioning strat— egy and have sought to demonstrate transfer or generaliza— tion effects as an approach to the investigation of the psychotherapy interview. Within this framework, rela- tively few studies have dealt with both issues, condition- ing and generalization, and the results of those studies generally support Matarazzo's conclusion (above). Rogers (1960), utilizing a quasi-therapy setting, hypothesized that the interviewer can produce changes in a S's self—reference verbalizations by applying simple re- inforcements and that such reinforcements can alter a 5'3 concept of himself. A series of free interviews was con- ducted with 36 introductory psychology students in which they described themselves. 33 were placed in two eXper- imental groups and one control. In one eXperimental group, 53 were conditioned for positive self—references, with reinforcement occurring via a head nod and "mmm-humm." In the other eXperimental group, Ss were similarly 5 reinforced for negative self-references. The controls were not conditioned. Prior to the experimental sessions, 55 received an adjective self—description test and a sen- tence completion test to induce "set." Pre- and post— interview measures were obtained on the Taylor Anxiety Scale and on a Q-Sort Emotional Adjustment Test. As for the experimental manipulations themselves, Rogers found that negative self-references conditioned, whereas rein- forcement prevented positive self-references from extin- guishing. On the relationship between conditioning and pre- and post-test scores, the results indicated that con- ditionability was not related to initial level of either anxiety or emotional adjustment, and that successful con- ditioning of self-references in the interview did not modify self-references outside the interview or alter scores on tests of anxiety or emotional adjustment. Thus, no transfer effects were found. Moos (1963) designed a study to determine whether established complex verbal habits could be conditioned in free interview situations, whether this conditioning ef- fect would be retained, and whether it would generalize to a different although similar situation. The 5 response categories conditioned were (1) eXpressions of independence and self—assertion, and (2) expressions of affection. 83 received eight 20-minute sessions, including a precondi- tioning session with a co—experimenter, a preconditioning session with E, four conditioning sessions with E, a post— conditioning session with E, and a postconditioning session with the co-experimenter. In the conditioning sessions, one group was reinforced via a head nod and "mmm-hmm" after each independence and assertive approach statement; a second group was likewise reinforced after each affection approach statement; the controls were reinforced every 30 seconds regardless of what they were talking about. Moos hypothesized that (l) the reinforced categories would in- crease during the conditioning sessions for the experi- mental groups but not for the controls, i.e. conditioning would occur; (2) the reinforced categories would be used more frequently in the postconditioning session with E than in the preconditioning session with E, i.e. reten- tion would occur; and (3) the reinforced categories would be used more frequently in the postconditioning session with the co-experimenter than in the preconditioning ses- sion with the co-experimenter, i.e., generalization would occur. The results demonstrated that conditioning and retention occurred. With respect to generalization, an intragroup comparison, using percentage of category used in preconditioning with co-experimenter, revealed a gen- eralization effect for the independence-assertion group but not for the affection group. However, when changes in the experimental groups were compared with the changes in the control group, no generalization was demonstrated, although differences were in the expected direction. Moos concluded that, at best, only a small generalization effect occurred. In a non-interview situation, Wimsatt and Vester (1963) designed a study to test whether the selective re- inforcement of responses to items of the Si scale of the MMPI could affect scores on the G, E, S, and A scale of Guiford-Zimmermen Temperament Survey and the Withdrawal Scale of the Psychotic Reaction Scale. All five scales were known to correlate significantly with the Si scale. Ss, who were psychiatric patients, were assigned to three groups. For one group, each Si scale item responded to in the scored direction was reinforced by "good." In the second experimental group, responses to items in the non—scored direction were similarly reinforced. The third, or control,group received no reinforcement. Although a significant conditioning effect was demonstrated, no extra- experimental effects were found. Cole (1965) assigned 53 to one of two conditions. In one, Ss received recorded verbal reinforcement while discussing negative aspects of themselves; in the other, 85 were similarly reinforced while discussing negative as- pects of a "neutral" vocational topic. As a pre- and post— test measure, Ss rated six concepts on a sematic differen- tial: "me at my best," "me as I am today," and "me at my worst," and three vocational concepts, most desirable, neutral, and least desirable, obtained from an initial in- terview. The results indicated that reinforcement produced a decrement in the response rate of the group exposed to reinforcement while discussing negative aspects of them- selves, whereas for the group reinforced while discussing negative aspects of a neutral vocational t0pic, response rate was maintained at a high level. Generalization oc~ curred only for the latter group, which showed a decline in the evaluation score on the semantic differential for the neutral vocational concept. 10 In a recent investigation by Koenig (1966), posi- tive and negative aspects of the response class, academic self-statements, were reinforced. Pre- and post- experimental scores were obtained on two scales intended to measure anxiety, and the primary goal of the study was to determine whether reinforcement of the response classes would affect postexperimental scores on the anxiety mea- sure. Specifically, it was hypothesized that reinforce- ment for positive academic self-statements would result in lowered anxiety scores while reinforcement of negative academic self-statements would result in heightened an— xiety scores. Results indicated that reinforcement sig— nificantly influenced the emission of negative academic self-statements but did not affect positive academic self- statements. Contrary to the hypothesis, reinforcement of negative academic self—statements led to a significant de— crease in the anxiety score, whereas the other condition led to only a slight reduction in anxiety. Koenig suggests that such anxiety changes resulted from the process of re- ciprocal inhibition, that is, anxiety associated with neg- ative academic self-statements became desensitized in the nonthreatening interview, leading to a reduction of aca- demically generated anxiety. ll Ullmann et al. (1961) used conditioning as the in- dependent variable in a study of the effects of reinforce- ment upon scores on the Palo Alto Group Therapy Scale. The verbal class reinforced was "emotional words," which had previously been shown to correlate significantly with group therapy ratings. 85, hOSpitalized patients in group therapy, were assigned to one of three conditions: (1) positive-personal reinforcement group in which E responded to the verbal class with a head nod and "mmm-hmm;" (2) the impersonal unstructured reinforcement group in which E re- sponded by pushing a button which emitted a loud click; (3) no reinforcement group in which E made no responses. Conditions were maintained through four story-telling ses- sions. The results indicated that significant gains in the GTS ratings were obtained for the positive-personal reinforcement group, but not for the others, supporting the conclusions that one person may influence the behavior of another in a positive way, that the change can be mea- sured by an independent criterion situation, and that this change may be demonstrated to be associated with specific behavior on the part of E. PROBLEM Essentially, the study is directed toward deter— mining (1) whether a specific approach response, reflec— tion, can be employed as a verbal reinforcer of a Specific subject response class, self-references, within the frame- work of a series of three interviews; (2) whether a trans- fer effect can be obtained on a set of criterion measures gathered pre- and post— interview; and (3) whether, if a transfer effect is found, it can be shown to be related to the conditioning procedure during the interviews. Thus, there are three basic issues or questions to be investigated in the study. The first is to demon- strate verbal conditioning using the given verbal rein- forcer and subject response class to be reinforced. This is the central problem in the issue of the modification of verbal behavior. The second problem is to demonstrate a test-retest change, on a set of criterion measures, that can be shown to be the result of an intervening experience, the interviews. This is the issue broadly referred to as generalization or transfer. The third question is perhaps 12 13 more implicit--the relationship between verbal condition— ing during the interviews and transfer of effects to an extra—interview situation, in this case the criterion mea- sures. Here, it would be possible to get conditioning within the interviews but show no transfer effects. It is also conceivable that transfer would be demonstrated but not conditioning, that transfer might have more to do with talking to an attentive ear, or the "match" between E and S, or any number and combinations of pertinent but uncon— trolled variables. Hypotheses will be stated operationally in a later section. In general, it is predicted that the frequency of positive and negative sub-classes of self-reference re- sponses will be amenable to modification when reflection is applied as a reinforcer, and that the effects of rein- forcement will transfer to a set of self-evaluative mea— sures . METHOD Definitions To lend clarity to this section, two of the prin- cipal concepts to be employed will first be defined. Conditioned Response: The general response class that E will approach during the interviews for the exper- imental groups is "Self—references," defined by Rogers (1960) as ". . . a verbal response by S which describes himself in some way, tells something about him, or refers to some affect he eXperiences." This general response class is then divided into two subclasses, positive or favorable self—references, and negative or unfavorable self-references. Verbal Reinforcer: Therapists' responses in psy— chotherapy interviews have been subjected to several clas- sification schemes. One such classification is that used by Bandura et al. (1960) and by Winder et a1. (1962) in which therapists' responses are first classified as either approach or avoidance responses, and are then divided into 14 15 further subclasses. This research utilizes an approach response, defined (Mills, 1964) as "any verbalization by the therapist which seems designed to elicit from the pa- tient further expressions or elaborations of feelings, at- titudes, or actions." The particular approach response employed in this study as the verbal reinforcer is £3— flection, a response in which E repeats or restates a por- tion ofEPs verbalization of feelings, attitudes, or actions, and may use phrases of synonymous meaning. A second E response class, "exploration," was used to elicit responses in the event that S became silent for thirty or more seconds. Exploration includes remarks or questions that encourage S to further describe or express his feelings, attitudes, and actions. Subjects 53 were volunteers from introductory psychology classes. In an effort to reduce the inconvenience of "no- shows," to eliminate the disinterested and to establish superficial rapport, each potential S was contacted 16 individually and given a brief description of the proce- dure in which he would be involved. Seventy-five Ss were randomly assigned to one of five groups of 15 85 each, with the one stipulation that each group be identical in number of male and female 55. Each group, then, had six male and nine female 53. Groups Groups are defined by the experimental or control conditions employed: 1) 2) 3) 4) 5) E Positive or favorable self—references are 1: reinforced. E2: Negative or unfavorable self-references are reinforced. E3: Positive and negative self-references are reinforced. C1: No reinforcement. C2: Test—retest, no interviews. 17 Interviews Each 8 in the experimental groups (El' E E3) and 2. the interview control group (Cl) participated in a series of three 20-minute interviews scheduled over a three to four day period. Four discussion topics were randomly se- lected without replacement from a set of twelve tOpics for use in each interview (see Appendix A). Thus, over the series of interviews, each topic was covered. These topics, dealing with social and college activities, appeared con- ducive to discussion during pilot work; that is, they ap- peared to stimulate interest in the Ss or pertained to issues in which Ss might be ego-involved. Ss were instructed to discuss each topic for approximately five minutes, re- ferring especially to their own feelings, attitudes, and actions, under the guise that they were participating in a public-Opinion polling experiment. This procedure was generally well-received and appeared to reduce defensive- ness . 18 Criterion Measures A battery of five self—evaluative scales was ad- ministered prior to and immediately following the series of interviews as a criterion to determine whether the ex- perimental conditions influenced self-attitudes and level of anxiety. These scales were administered to all 83, in— cluding a test-retest control group which received the battery on a three to four day intertest-interval, four days being the limit within which the other Ss were inter- viewed. (The rationale for the use of these scales will be discussed in the hypotheses section.) These scales included: 1) The IPAT Anxiety Scale (Cattell and Scheier, 1963), which has been successfully used in other behav- ioral modification studies (for example, Paul, 1966). It was employed here to determine whether the experimental conditions would affect anxiety level. 2)- A scale of the Semantic Differential type (Osgood, et al., 1957), composed of 25 pairs of polar ad- jectives anchored at one end by an adjective judged 3) 19 to reflect favorable self-attitude and anchored at the opposite end by an adjective judged to re- flect unfavorable self-attitude. Each of the 25 polar adjective scales was rated on a seven-point scale (after Osgood's technique). The complete scale was rated twice by each S, once as he saw himself socially ("myself socially") and once as he saw himself academically ("myself as a college student"). The Standard Semantic Differential instructions were given (see Appendix B). Two Q-sorts, each composed of 26 items selected from the California Q—sort (Block, 1961) and from the Chicago Counseling Center Q-sort (Rogers and Dymond, 1954). Half of each set of items were judged to reflect favorable self—attitudes and half unfavorable self-attitudes. A seven-point, forced-distribution scale was used (distribution of items: 1-3-5-8-5-3-1), ranging from self- ratings of "very much like me" to "very much un- like men' The same self-concepts rated on the Semantic Differential technique were used here. (See Appendix C.) 20 These five self-evaluative scales were administered in random order, and Q—sort cards were shuffled after each sort. Procedure Figure 1 illustrates the experimental operations for the research (under interviews, a "+" or "—" sign in- dicates that either positive or negative self—references were reinforced, respectively). All groups except C re- ceived the battery of self-evaluative measures prior to the first interview and following the last interview. The C a three to four day 2 group received the measures on test—retest.interval. Groups Pre- Interviews Post- Test 1 2 3 Test E1 Yes Operant Condition + Condition + Yes E2 Yes Operant Condition - Condition - Yes E3 Yes Operant Condition +&— Condition +&- Yes Cl Yes Operant Operant Operant Yes C2 Yes None None None Yes Figure l.-—EXperimental design. 21 The first interview for the three experimental groups served to establish the operant level of emission of self-references. During this interview, E responded to 5 only with exploration responses. The second and third interviews served as the verbal conditioning inter- views. During these, self—references of the subclass to be conditioned were reinforced on as close to a one hun- dred percent schedule of reinforcement as possible in order to maximize the possibility of conditioning occur- ring.* The interview control group (C1) continued to re- ceive exploration responses; at no time were self- references approached for these Ss. *The author wishes to thank Dr. Wm. Stellwagen for his ad- vice on the schedule of reinforcement. HYPOTHESES Verbal Conditioning It is hypothesized that conditioning of the S verbal response class, positive or negative self—reference, will occur over a series of interviews when, contingent upon and immediately following the emission of a self- reference of the subclass to be approached, E responds with a reflection of S's self—reference. Essentially, this is a Skinnerian approach to the conditioning of verbal behavior (see Skinner, 1957). It assumes, basically, that verbal behavior can be modified through a reinforcement procedure. Additionally, it as- sumes that verbal behavior can be divided into response classes for purposes of investigation, and that the re- sponse class chosen for investigation may occur ". . . with some ascertainable frequency prior to the introduc- tion of any specific set or sets of operations by the ex— perimenter (Greenspoon, 1962)." The specific set of oper— ations is the application of the verbal reinforcement, in 22 23 this study the E response class "reflection," although it seems erroneous to refer to a specific E response as a re- inforcer until it has been shown to function as one. Self-reference was selected as the S response class to be manipulated (the dependent variable) because, firSt of all, it has been shown to be manipulable in other verbal conditioning studies (Rogers, 1960, for example); secondly, since a self-reference has S as its object, it was assumed that it would already be operant within S's speech, espe— cially within the context of the interviews and their topics. Reflection was selected as the independent variable because of its frequent use in psychotherapy interviews, eSpecially of the "client—centered" type (Rogers, 1951, 452-455). Secondly, E wanted to avoid the use of the simple social reinforcers such as "good," "um-hmm," or a head nod, and employ a verbal response that might be more analogous to a "natural" interview. A final reason for selecting these two variables is that an element of "bare faced" empiricism still exists in attempting to demonstrate verbal conditioning employing the two response classes used here. 24 Transfer The term transfer is used in this study to refer to changes in the test-retest "criterion" measures which can be attributed to the intervening experience of the interviews in general and the experimental manipulations (conditioning) in particular. A. Self-evaluative measures: Semantic Differentials and Q-Sorts: At an initial glance, this might first be consid- ered a question of response generalization. This assumes that the responses emitted and reinforced during the con- ditioning phase bear a similarity to the responses elicited by the self-evaluative measures, a point stressed by Kras- ner (1962). It is postulated here that self-references, having S as the object, tap the same behavioral domain as do the responses to the self-evaluative measures, namely, S's self-perceptions or his self-concept. From a response generalization vieWpoint, then, it would be predicted that an increase in favorable or positive self—references via re- inforcement during the interviews would generalize to an in- creased favorableness of self—report on the evaluative measures. Similarly, an increase in unfavorable or 25 negative self-references in the interviews would generalize to an increased unfavorable self-report. However, the author would like to suggest that transfer of effects from the conditioning of negative self- references to the self—evaluative measures is £95 a matter of response generalization. When S is responding with a negative self-reference, it may be considered that he is discussing an aspect of himself that is an expression of an event or events in his history which are associated with anxiety. Now, the essence of reflection lies not so much in the technique itself as it does in conveying from E to S, in a nonthreatening manner, a sense of understanding and acceptance as opposed to anxiety-generating disapproval. Thus, it is reasoned, the condition of reflection of nega- tive self-references would tend to desensitize the anxiety associated with the negative self-reference. This would occur through a process of reciprocal inhibition (Wolpe, 1958) in which, in this case, it is conceptualized that a feeling of being understood, accepted, and unthreatened is incompatible with a feeling of anxiety. Finally, since the anxiety associated negative self-reference would tend to devaluate a person's self-concept, desensitizing the 26 anxiety should eventually lead to a more favorable self- evaluation. (Admittedly, this bit of theorizing is novel if for no other reason than that it spans two "schools" of psychotherapy, namely, client—centered and behavior therapy.) With respect to the above comments, the following hypotheses are offered: 1) 2) 3) It is predicted that favorableness of self-report will increase in the E1 group (conditioning of positive self-references). This is predicated on the basis of response generalization. It is predicted that favorableness of self-report will increase in the E2 group (conditioning of negative self-references). This is formulated on the basis of desensitization. It is predicted that favorableness of self-report will increase in the E3 group (conditioning of both positive and negative self-references). This hypothesis is established on the basis of both re- sponse generalization and desensitization. Here, it might also be speculated that positive changes 27 in self—evaluation would be greater for this group than for the other groups. 4) It is predicted that there will be no change in self-evaluation for the C1 group (interview con- trol) since there is no systematic attempt to man- ipulate any variable in this group. 5) It is predicted that there will be no change in self-evaluation in the C2 group (test-retest, no interviews). This is essentially a test-retest reliability group. B. Self-evaluative measure: Anxiety Scale: Continuing with the rationale for desensitization presented above, 53 reinforced via reflection of negative self—references should experience a reduction in anxiety associated with the events underlying and prompting their negative self-reference. Consequently, it is predicted that in the two groups (E2 and E3) in which negative self- references are reflected, a decrease in measured anxiety will occur. 'As for the other groups, no change in mea- sured anxiety is predicted. RESULTS Verbal Conditioning The initial step in the analysis of the interview data was to demonstrate that positive and negative self- references could be reliably coded from taped interviews. There were three twenty-minute interviews for each of the 60 55, or a total of 60 hours of recorded interviews to code. Because E represented one of the judges, an effort was made to eliminate examiner bias by coding interviews selected randomly so that E did not know to which group or interview a particular interview segment being coded belonged until the final data were reassembled by groups and interviews. A second judge was trained on pilot tapes to code self-references, and then coded a stratified random sample of twelve interview segments from each of the four groups without knowledge of the group and interview from which each tape segment was selected. A Pearson product- moment correlation coefficient was then computed between the codings of E and the codings of the second judge for 28 29 positive and negative self-references from the 48 twice— judged interview segments. This yielded an r = 0.86 which was considered a satisfactory indication of the reliability of coding self-reference responses. The self-reference data, in the form of raw number of self-references, were then subjected to a nested anal- ysis of variance design (see Edwards, 1968) in which 85 are nested within groups within response subclasses within interviews. This design permitted the following signi- ficance tests: 1) Groups (G): Differences among the four groups across interviews and response subclasses. 2) Interviews (I): Differences among the three in- terviews across groups and response subclasses. 3) Responses (R): Differences between the two re- Sponse subclasses (positive and negative self- references) across groups and interviews. 4) Groups x Interviews Interaction (G x I): Differ- ences among group by interview entries across re- sponse subclasses. 3O 5) Groups x Responses Interaction (G x R): Differ- ences among group by response subclass entries across interviews. 6) Interviews x Responses Interaction (I x R): Dif- ferences among interview by response subclass entries across groups. 7) Groups x Interviews x Responses Interaction (G x I x R): Differences among group by inter- view by response subclass entries. Table 1 summarizes the nested analysis of variance design for the self-reference data. It should be noted that the error (a) mean square (based on the sum of squares between $3 nested within groups) is used for testing the significance of the group mean square, while the error (b) mean square (residual error term based on the pooled in- teraction sum of squares involving interactions of Ss and combinations of I and R) is used for testing the signifi- cance of all other mean squares. The F—test for Groups (G) is insignificant, meaning that summing across interviews and responses, there are no significant differences among the group means. 31 Table 1 Nested Analysis of Variance for Self—Reference Data Source SS df MS F G 75.631 3 25.210 1.357 NS Error (a) 1040.552 56 18.581 I 117.505 2 58.753 27.506 ** R 40.670 1 40.670 19.040 ** G x I 56.358 6 9.393 4.397 ** G x R 20.720 3 6.907 3.234 * I x R 1.907 2 0.954 0.447 NS G x I x R 55.612 6 9.269 4.339 * Error (b) 597.976 280 2.136 Total 2006.931 359 **Significant at .01 level *Significant at .05 level The F-test for Interviews (I) is significant, mean- ing that summing across groups and responses, there are significant differences among the three interviews in mean self-references. The F-test for Responses (R) is significant, mean- ing that summing across interviews and groups, there is a 32 significant difference between positive and negative self- references, with a greater number of the former than of the latter. The Groups x Interviews (G x I) interaction is significant, meaning that significant differences exist among mean self-references (both positive and negative combined) within interviews within groups. The Groups x Responses (G x R) interaction is also significant, meaning that significant differences exist between mean positive and mean negative self-references among groups, summing across interviews. The Interviews x Responses (I x R) interaction is insignificant, meaning that no significant differences exist between mean positive and mean negative self- references within interviews summing across groups. Finally, the Groups x Interviews x Responses (G x I x R) interaction is significant, meaningpthat there are significant differences among the means for positive and for negative self-references within interviews within groups. Table 2 provides a breakdown by mean self-references of the group, interview, response, and all the interaction 33 Table 2 Breakdown by Mean Self-References of Main and Interaction Effects for Nested Analysis of Variance Design Groups E1 E2 E3 C1 21.200 21.133 20.667 14.667 Interviews 1 2 3 4.967 6.717 7.733 Responses + _ 10.717 8.700 Groups X Interviews Groups 1 5.333 4.867 4.533 5.133 Interviews 2 7.133 7.333 7.533 4.867 3 8.733 8.933 8.600 4.667 34 Table 2.--Continued Groups X Responses ~Groups El E2 E3 C Responses + 12.600 10.533 11.400 8.333 — 8.600 10.600 9.267 6.333 Interviews X Repponses Interviews 1 2 3 Responses + 2.717 3.750 4.250 - 2.250 2.967 3.483 Groups X Interviews X Responses Interviews 1 2 Responses + - + - + - E1 2.533 2.800 4.333 2.800 5.733 3.000 E2 2.933 1.933 3.800 3.533 3.800 5.133 Groups E3 2.667 1.867 4.000 3.533 4.733 3.867 C 2.733 2.400 2.867 2.000 2.733 1.933 35 effects. For purposes of further analysis of the data, however, the Groups x Interviews x Responses (G x I x R) interaction is referred to, for it is the most important of the F-tests in terms of interpreting the data in light of the specific hypotheses about conditioning. If the G x I x R interaction is envisioned as a 24 cell design (see Table 2), with each cell containing the mean responses for a specific response subclass for a specific interview for a specific group, the salience of this interaction be- comes clearer. It can then be stated that among those 24 cells significant differences exist. To test each Specific hypothesis, it was necessary to make individual comparisons among the G x I x R cell means. The procedure used was Duncan's new multiple range test (See Edwards, 1968, ch. 8), selecting only those cell means that were relevant to the hypotheses. Table 3 sum- marizes the data employed in the multiple range test; Spe- cific cell means used are referred to by group, interview, R+ denotes the cell mean for group 1 and response. Hence 3111 E interview 1, positive self-references. ll It was first necessary to demonstrate that no sig- nificant differences existed between first and third I36 .ucououuwv haucuowwwcmwo uoc one used 08am usu >n pouoouuopcs mauve m x H x 0 03» >c< oucou“uufiv “HUCUUMWMgvfiD 0H” “CH." ”BMW 0;“ “Q ”muoumumvcg U0“ WCMUE m x H x O 03“ \AC¢ .q .x .8 .H .m .o .m .m .a .o .m .< uuuuu mmn.m +mMHHm .q m N . Hmoo.H ooo.o ..... mma.m .m H m x mamm.H ooo.H oov.o ..... mms.v +mMHmm .n 05mm.a oom.H oo~.H oom.o ..... som.m -mmHmm .H mesm.a ooo.m oov.~ ooo.~ va.H ..... mms.~ +mMHHo .m moom.H ooo.m ooe.~ ooo.~ vma.a ooo.o ..... mms.~ +mHHHo .o mmvm.a soo.m oos.~ ooo.~ oo~.H ooo.o ooo.o ..... soo.m +mszm .E samm.a oo~.m ooo.~ oo~.~ emm.H oo~.o oo~.o «ma.o It's- mmm.~ +mHHHm .m maom.a mmm.m mms.~ mmm.~ sos.a mmm.o «mm.o so~.o mma.o ..... oov.~ -mHHHo .o oasv.a oom.m oo~.m oom.~ «mm.H oom.o oom.o ems.o oom.o sov.o ..... mmo.H -mMHHo .o omme.a oom.m oo~.m oom.~ vmm.H oom.o oom.o ems.o oom.o smv.o ooo.o ..... mmm.H -mHHNm .m mesm.a som.m oo~.m oom.~ ooo.~ omm.o oom.o oom.o moo.o mmm.o woo.o ooo.o ..... som.a -mszm .< Ho.oua museum mms.w mma.m mms.v som.m mms.~ mms.~ smm.~ mmm.~ oov.~ mmm.H ans.” som.a ucmonuucmnm +mMH -mMHNm +mMHm Immamm +mmnao +mHHHo +mHHmm +mHHHm -mHHH ImMHao -mHHNm ImHHmm scam: ammuuonm .a .x .s .H .m .o .u .u .o .o .m ..¢ moucmummozluaom cam: m x H x 0 pwuuoaom coo3u0m meocuuemwfio mnu ou pawanmt umma omcsm mamwuasz 3oz m.cmu::o m OHQMB 37 interviews for the control group (C1) for both positive and negative self-references. Comparisons between C111R+ and C113R+, and between ClllR- and ClI3R- showed this to be the case. Stated specifically, the hypothesis predicted that in group E , where positive self-references were reflected, 1 conditioning would occur across interviews. To test this hypothesis, first the ElIlR+ and the E113R+ cell means were compared and found to differ Significantly. Then, to "double-check" for conditioning, the first and third inter- views for positive self-references in the E group were 1 compared respectively with the first and third interviews for positive self-references in the control group (C1)' It was expected that there would be no differences in first interview comparisons, Since the first interview served to establish Operant level of responding. However, differ— ences between third interview comparisons were expected if conditioning of positive references had occurred in the E1 group. Both of these expectations were borne out by the data analysis. Thus, conditioning of positive self- references occurred. 38 The hypothesis also predicted that in group E2, where negative self-references were reflected, condition- ing of those responses would occur. As above, a compar- ison was first made between the first and the third in- terviews for negative self—references which obtained sig- nificance. Similarly, comparisons were made between the E2 and C1 groups for first and third interviews respec- tively, for negative self-references. No difference was found for the first interview comparison, but the third interview comparison obtained Significance. Thus, the prediction of conditioning of negative Self-references was supported. As for the third group, the hypothesis predicted that both positive and negative self—references would be conditioned Since both were reflected during the second and third interviews. Comparisons were made between the E I R+ and the E I R+ cell means, and between the E I R- 3 1 3 3 3 1 and the E313R- cell means, both of which yielded signifi— cant results. Comparisons for both response subclasses were made between E3 and Cl first interviews and between E3 and C1 third interviews, the latter of which obtained Significance. Again, the hypothesis was supported. 39 Thus,the conclusion is reached that conditioning occurred as predicted, that when positive or negative self- references are approached with reflection, reflection acts as a verbal reinforcer. Transfer The data of each of the five self-evaluative mea- sures were subjected to an analysis of covariance design (Edwards, 1968, ch. 16). The results of these analyses are summarized in Table 4. In these analyses, the pre- interview raw test scores constituted the X measures (or supplementary measures); the post-interview raw test scores constituted the Y measures (or dependent measures). An- alysis of covariance was used because it was assumed that the X and Y measures were substantially correlated; co- variance analysis, then, takes into account the regression of the Y measures on the X measures and yields a lower error estimate than would be obtained from an analysis of variance. It is immediately apparent that, of the five mea- sures employed, the only one to yield Significant results 40 Table 4 Analyses of Covariance of Self—Evaluation Measures Source Treatments Error Total IPAT A SS 274.231 1599.068 1873.299 nxiety Scale .gg MS F 4 68.558 2.958* pg 23.175 73 *Significant at the 0.05 level. Semantic Differential (Social) Source SS ‘gf MS F Treatments 182.116 4 45.529 0.631 N.S. Error 4982.469 .gg 72.710 Total 5164.585 73 Semantic Differential (Collegg) Source SS 'Qf MS F Treatments 54.497 4 13.624 0.148 N.S. Error 6352.721 .gg 92.068 Total 6407.218 73 QfSort (Social) Source SS .gg MS F . Treatments 20.219 4 5.055 0.770 N.S. Error 453.035 §2_ 6.566 Total 473.254 73 41 Table 4.--Continued. QeSort (Collegp) Source SS .gg MS F Treatments 16.324 4 4.081 0.745 N.S. Error 378.217 §2_ 5.481 Total 394.541 73 was the IPAT Anxiety Scale. An inspection of the data of the other four measures reveals very few, and unsystematic, changes between scores on the first and scores on the sec— ond administration of those measures. Thus, the hypotheses predicting changes in the direction of more favorable self-A reports for the experimental (conditioned) Ss are rejected. The analysis of covariance for the data yielded by the IPAT Anxiety Scale resulted in an F = 2.958 which, for 4 and 69 degrees of freedom, indicates that the adjusted treatment mean square is significant at the .05 level. From this analysis it may be concluded that Significant differences exist among the adjusted treatment means. Tests of comparisons were then made among those adjusted treatment means (Edwards, 1968, p. 341). Taking the largest adjusted mean, which is for the C group 2 42 (test-retest control, no interviews), and the second largest adjusted mean, which is for the E group (rein- 3 forcement of both positive and negative references), a t-test was computed for the difference between these two means. This yielded a t = 2.311 for 69 degrees of free— dom which is significant at the .025 level. It may then be reasoned that differences between the adjusted mean for and C would also C and the adjusted means for E E2. 1 2 1' reach significance, for each of these means when compared with the C2 adjusted mean yield differences greater than that between the C2 and E3 means used in the t-test. As a further check on the data, a t-test was computed on the difference between the largest and smallest adjusted means 1, 3, and Cl' that IS, between E3 and El' This yielded a t = 0.670 with 69 degrees of freedom which among E E E 2! is insignificant. Thus, there are no Significant dif- or C ad— E3' 1 ferences between any combination of E E 1' 2' justed means. It may then be concluded from this analysis of the data of the IPAT Anxiety Scale that all four of the groups which participated in the interviews differ Significantly from the test-retest, no interview control group (C2). 43 Each of the four interview groups showed a decrease in measured anxiety while the C group Showed an increase 2 in measured anxiety on inspection of the data. Consi- dering the fact that there are no significant differences among the four interview groups themselves, it may further be concluded that the factors accounting for the observed differences between the interview groups and the no- interview group are to be found in the interviews but not in the modification of verbal behavior. DISCUSS ION Verbal Conditioning The analysis of the Self—reference data confirms the hypothesis which predicted that when a reflection is made contingent upon and immediately following a self- reference of the particular response subclass approached, the frequency of emission of such responses will increase. Reflection did, in fact, act as a reinforcer of both pos- itive and negative self-references, demonstrating that self—references is a modifiable verbal response class when reflection is utilized as a verbal reinforcement. These results have implications for the study of psychotherapy process. In a general sense, it lends sup- port to the claim made by behaviorally-oriented psycho- therapists that verbal modification can and does occur within psychotherapy sessions. More specifically, it has implications for the "client—centered" approach to psycho- therapy in which reflection is often employed as a ther- apist response. Traditionally, client-centered 44 45 psychotherapy does not eXplicitly deal with the issue of verbal (or, more generally, behavioral) manipulation, but the results of this study suggest that the specific tech- nique of reflection may act to modify, through reinforce- ment, the verbal eXpressions of the client. Other classes of E and S verbal behavior employed in a client—centered or other "types" of psychotherapy could be explored. For example, recent studies by Truax (1966; 1968) suggest "the presence of significant differential reinforcement effects imbedded in the transactions of client-centered psychotherapy (1966, p. 7)," when the use of accurate empathy, nonpos- Sessive warmth, and genuineness are viewed as reinforcers. Another implication generated by the verbal con— ditioning data of this study is that verbal reinforcement in psychotherapeutic and other types of interviews is akin to what experimental psychologists refer to as Shaping. Sidman (1962, p. 172) defines shaping as a technique used to establish a desired behavior "by reinforcing succes- sively closer approximations to the behavior with which the experimenter ultimately wants to work." In a review of conditioning studies, Krasner (1966, p. 297) commented on shaping in psychotherapy interviews in his remark that 46 "one seemingly clear implication from such verbal condi- tioning studies is that the therapist shapes the behavior of his patient to his own biases." While it might pre— sent problems to attempt to demonstrate that the therapist reinforces "successively closer approximations" to the behavior he hOpes to establish, nevertheless it does not seem unreasonable to assume that the therapist has some predetermined ideas about what the client should and should not be talking and that, consequently, he selectively re- sponds to the client's verbalizations in such a way as to "shape" the verbal behavior of the client through rein- forcement. When the verbal conditioning results, per se, of this study are compared with those of other studies Sim— ilarly involving conditioning of positive and negative self-references, both agreement and disagreement are found. Sarason and Ganzer (1962) successfully conditioned both positive and negative self—references, using "um-hmm," "I see," "yeah," or "uh-huh" as reinforcers; Adams and Hoffman (1960), who did not differentiate between positive and negative self-references, successfully conditioned the general response class of self—references, applying 47 attention (raising head and looking at S) and "mm-hmm," as reinforcement. These two investigations, and parti- cularly the first, lend support to the findings reported here. However, as was previously noted, Rogers (1960) and Koenig (1966), attempting to modify positive and neg— ative self-references and academic self—statements re— spectively, successfully conditioned negative responses but not positive responses. In addition, Rogers (1960) found that reinforcement prevented positive self-references from extinguishing. Finally, Sarason and Ganzer (1963), attempting to condition only negative self-references, re— ported successful results. Thus, it may be concluded that negative self-references constitute a modifiable response class, whereas conditioning of positive self—references is equivocal. With respect to the Skinnerian reinforcement strategy employed in the above studies, the results are not fully in accord with predictions generated by reinforc- ment theory, although those found by this author are. Precisely why this should be the case is unknown. In terms of psychotherapy, it is perhaps advantageous that negative self-references are amenable to reinforcement, especially if it is assumed that negative self-references are verbal 48 expressions of the troublesome behavior that prompted seeking psychotherapy in the first place. Transfer The hypotheses predicting changes in the direction of more favorable self-report for the experimental groups on the two Semantic Differential scales and the two Q- Sorts were rejected by the data analysis. This is most Simply interpreted as indicating that insofar as these specific measures are concerned, the effects of the exper- imental manipulations did not transfer to the extra— interview Situation. However, this begs the question of why no transfer effects were found. The first suggestion is Simply that, in an interview Situation, conditioning produces changes only in verbal behavior or, as has already been remarked, the SS learn to talk differently and little else (Winder, 1957). A second possible source of failure to obtain transfer may lie in the insensitivity of the measuring instruments to detect changes, especially when any hypothesized changes are likely to be small considering that only three brief interviews were employed as the 49 intervening experience. Either or both of these two in~ terpretations would fit the data. However, there is a methodological reason which may underly the absence of transfer effects in this study (as well as in others). The concept of response general- ization was presented earlier as a theoretical rationale for part of the predicted transfer effects. However, if generalization of responses is sought, it can be reasoned that there should be some similarity in the stimulus char- acteristics of the learning and the transfer phases of the experiment, or as Drennen (1963) conceives it, trans- fer may be assessed as a function of Similarity of cues in the learning gpg the transfer task situations. Within the context of a study such as the one presented here, then, it may be reasoned that not only Should the criterion measures of transfer elicit responses similar to those re- inforced during the interviews (response generalization), but moreover, the stimuli eliciting those responses Should also bear some similarity to the experimental stimulus conditions. To illustrate this point, two research inves- tigations are cited: Lovaas (1961) succeeded in increasing children's non-verbal aggressive behavior by conditioning their verbal 50 aggressive behavior. The operant level of play behavior with two toys, a "striking-doll" (aggressive) toy and a ball-toy (nonaggressive) was determined prior to condi- tioning. Then Lovaas conditioned verbal aggressive re- sponses to a "bad doll" in one group of SS and verbal nonaggressive responses to a "good doll" in another group, using trinkets for reinforcement. When again presented with the "striking-doll" toy and the ball-toy, the group conditioned for verbal aggressive behavior responded at a significantly higher level to the "striking-doll" toy than did the children conditioned for nonaggressive verbal responses. D'Zurilla (1966) first rated SS participation in a classroom setting. Then experimental Ss were assigned to treatment discussion groups in which E responded to S verbalizations with persuasion, praise, or neutrality, de- pending upon the treatment conditions. A control group remained within the regular classroom. Following a series of group (treatment) discussion sessions, treatment SS re- turned to the classroom where their verbal participation was rated. Those SS who had been involved in the treat- ment conditions participated at a significantly greater 51 level after returning to the classroom, even after a de— crement in participation among control SS was considered. These two studies serve to emphasize the point that when considerable similarity exists in stimulus prop- erties between the treatment and transfer phases, transfer effects are more likely to occur. A point Should also be raised regarding the pre- dictions of transfer based on the author's theorizing on the issue of desensitization. In the absence of transfer effects, of course, it remains conceivable that desensiti- zation occurred but not sufficiently to effect a change on the self-report measures, or the latter were lacking in sensitivity to such effects. However, it is also possible that the negative self—references elicited were not suf- ficiently "anxiety laden" to be affected by desensitization, if it was occurring; or, a feeling of understanding and acceptance conveyed by the reinforcer reflection was not "incompatible" with anxiety. Finally, the fact that there were significantly fewer negative self-references emitted overall than positive self-references may have contributed to the absence of effects by virtue of there having been limited opportunity for desensitization to occur, if it was occurring. 52 Only one of the self—evaluative measures, the IPAT Anxiety Scale, yielded data which obtained significance on the analysis of covariance. However, tests of comparisons among the adjusted means demonstrated that the significant differences were to be found between each of the interview groups (including the interview controls as well as the conditioned groups) gpd_the no-interview controls. An in- spection of the data indicated that each of the interview groups decreased in measured anxiety, whereas the no- interview control group increased in measured anxiety. Thus, the hypotheses regarding transfer effects on the IPAT Anxiety Scale were rejected, for it can not be con- cluded that conditioning contributed to the changes on the anxiety scale. It can only be interpreted as meaning that the variables influencing a decrease in anxiety were em- bedded within the context of the interviews, for example, within the interpersonal context of the situation. Although transfer of the effects of conditioning to the extra-interview measures was not demonstrated in this study, it Should not necessarily be concluded that transfer does not occur. The writer believes, however, that a potentially more fruitful view of the role of verbal 53 conditioning is to keep in mind that it occurs within the interpersonal context of psychotherapy. Application of reinforcement can no more be divorced from the person ap- plying the reinforcer than the conditioned responses can be divorced from the person expressing those responses. Moreover, overriding separate considerations of the ther- apist and the client is the fact that they are interacting with one another, and that this interpersonal climate of psychotherapy may be the paramount factor in establishing behavior change and its transfer to the extra-interview life of the client. S UMMARY A study was designed attempting to (1) condition negative and positive self—references within a series of three 20-minute interviews. using reflection as the verbal reinforcer; (2) demonstrate transfer of conditioning ef- fects to a battery of five self-evaluative measures (two Semantic Differentials. two Q—Sorts. and the IPAT Anxiety Scale) administered pre- and postwinterview. Seventy-five SS were randomly placed in five experimental and control groups. 15 Ss to a group. Group E1 was reinforced for positive self—references; Group E was reinforced for neg— 2 ative self-references; Group E for both positive and neg— 3 ative self-references; Group C1 was an interview control group; Group C2 was a test-retest. no interview control group. The self-evaluative test battery was administered immediately before the first interview and immediately following the last interview. except in the case of C2 controls who received the battery on a three to four day test-retest interval. 54 55 Self-reference data were subjected to a nested analysis of variance design. Individual comparisons were made among cell members of the Groups x Interviews x Re- sponses interaction. Results confirmed that conditioning had occurred as predicted. Both positive and negative self-references were shown to be amenable to conditioning employing reflection as the reinforcer. Each self-evaluative measure was subjected to an analysis of covariance. Only changes on the IPAT Anxiety Scale reached significance. However, individual compar- isons applied to the IPAT data demonstrated that condition— ing did not effect transfer wheras being interviewed did. Thus, all hypotheses regarding transfer were rejected. Implications of the conditioning results were noted, with specific reference to client—centered psycho- therapy and to the concept of Shaping as applied to verbal conditioning within therapy interviews. The personal and interpersonal context of verbal conditioning within inter- views was stressed. Comments on the absence of transfer effects attrib- utable to conditioning were made. Reference was made in particular to the concept of stimulus generalization and 56 to the hypothesized desensitization rationale. It was suggested that variables affecting decreased anxiety scores were imbedded within the interpersonal context of the interviews. In essence, the study again demonstrates that verbal behavior can be modified through reinforcement in a quasi-therapeutic setting, but that the issue of trans- fer remains equivocal. REFERENCES REFERENCES Adams, J. S. & Hoffman, B. The frequency of self-reference statements as a function of generalized reinforce- ment. Journal of Abnormal and Social Psychology, 1960, Q, 384-389. ‘ Bandura, A. Psychotherapy as a learning process. Psycho- logical Bulletin, 1961, §§, 143-159. Bandura, A., Lipsher, D. H., & Miller, Paula E. Psycho- therapists' approach-avoidance reactions to pa- tients' expressions of hostility. Journal of Con- sultinggPsychology, 1960, 23, 1-8. Block, J. The grsort method inypersonality assessment and psychiatric research. Springfield, 111., Charles C. Thomas, 1961. Cattell, R. B. & Scheier, I. H. Handbook for the IPAT anxiety scale questionnaire. Champaign, 111., Institute for Personality and Ability Testing, 1963. Cole, C. W. Effects of verbal stimuli in a counseling analogue. Journal of Counseling Psychology, 1965, 12, 408-413. Drennen, W. T. Transfer of the effects of verbal condi- tioning. Journal of Abnormal and Social Psycholggy, 1963. 29, 619-622. D'Zurillo, T. J. Persuasion and praise as techniques for modifying verbal behavior in a "real life" group setting. Journal of Abnormal Psychology, 1966, 11, 369-376. Edwards, A. L. Experimental design in psychological re- search. New York, Holt, Rinehart and‘Winston, 1968 (3rd edition). 58 59 Greenspoon, J. Verbal conditioning and clinical psychology. In A. J. Bachrach (Ed.), Experimental foundations of clinical psychology. New YOrk, Basic Books, 1962. Kanfer, F. H. Comments on learning in psychotherapy. Psy- chological Reports, 1961, 2, 681-699. Koenig, K. P. Verbal behavior and personality change. Journal of Personality and Social Psychologx. 1966, g, 223-227. Krasner, L. Behavior modification research and the role of the therapist. In L. A. Gottschalk and A. H. Auer- bach (Eds.), Methods of research in psychotherapy. New York, Appleton-Century-Crofts, 1966. Krasner, L. The therapist as a social reinforcement machine. In H. H. Strupp and L. Luborsky (Eds.), Research in psychotherapy. Proceedings of a Conference, Chapel Hill, N.C., May 17-20, 1961. Washington, D.C., American Psychological Association, 1962. Lovaas, 0. I. Interaction between verbal and nonverbal be- havior. Child Development, 1961, 33, 329-336. Matarazzo, J. D. Psychotherapeutic processes. In P. R. Farnsworth (Ed.), Annual Review of ngchologx. 1965, lg. 181-224. Mills, D. H. Liking as a therapist variable in the psycho— therapeutic interaction. Unpublished doctoral dis— sertation, Michigan State University, 1964. Moos, R. H. The retention and generalization of Operant conditioning effects in an interview situation. Journal of Abnormal and Social Psychology. 1963, fig, 52-58. Osgood, C. E., Suci, G. J. & Tannenbaum, P. H. The mea- surement of meaning. Urbana, University of Illi— nois Press, 1957. Paul, G. L. Insight vs. desensitization in psychotherapy. Stanford, Stanford University Press, 1966. 60 Rogers, C. R. Client-centered therapy. Boston, Houghton Mifflin, 1951. Rogers, C. R. & Dymond, Rosalind (Eds.). Psychotherapy and personality change. Chicago, University of Chicago Press, 1954. Rogers, J. M. Operant conditioning in a quasi-therapy setting. Journal of Abnormal and Social Psycho- lpgy, 1960, pp, 247-252. Sarason, I. G. & Ganzer, V. J. Anxiety, reinforcement, and experimental instructions in a free verbali- zation Situation. Journal of Abnormal and Social Psychology. 1962, p5, 300-307. Sarason, I. G. & Ganzer, V. J. Effects of test anxiety and reinforcement history on verbal behavior. Journal of Abnormal and Social Psychology, 1963, £1, 513-519. Sidman, M. Operant techniques. In A. J. Bachrach (Ed.), Experimental foundations of clinical psychology. New York, Basic Books, 1962. Skinner, B. F. Verbal behavior. New York, Appleton- Century—Crofts, 1957. Truax, C. B. Reinforcement and nonreinforcement and Ro- gerian psychotherapy. Journal of Abnormal Psy- ChOIQSY’ 1966' 2.1:; 1-9. Truax, C. B. Therapist interpersonal reinforcement of client self-exploration and therapeutic outcome in group psychotherapy. Journal of Counseling Psychology. 1968, lg, 225-331. Ullmann, L. P., Krasner, L. & Collins, Beverly J. Modi- fication of verbal behavior through verbal condi- tioning: effects in group therapy. Journal of Abnormal and Social Psychology, 1961, Q2, 128-132. 61 Wimsatt, W. R. & Vester, N. D. Extraexperimental effects Winder, Winder, Wolpe, in verbal conditioning. Journal of Consulting Psychology. 1963, g1, 400-404. C. L. Psychotherapy. In P. R. Farnsworth (Ed.), Annual Review of PsycholoQY. 1957, 8, 309-330. C. L., Ahmad, F. 2., Bandura, A. & Rau, Lucy C. Dependency of patients, psychotherapists' re— sponses, and aspects of psychotherapy. Journal of ConsultinngsycholoQY. 1962, gs, 129-134. J. Psychothergpy by reciprocal inhibition. Stan- ford, Stanford University Press, 1958. APPEND ICES APPENDIX A INTERVIEW INSTRUCTIONS AND TOPICS Instructions During this series of interviews I want you to discuss your Opinions, your attitudes and feelings with reference to a variety of different tOpics which will be of interest to you.. In a sense, I'd like you to "open up" about your Opinions on these topics--anything you say will, of course, be kept in strict confidentiality, and you won't be identified to anyone but me. Just say what you think or feel as an individual. There are twelve tapics altogether, and I'll se- lect four at random from each interview. As you can see, the interviews will be taped, but try to be "natural" as you discuss the topics. Do you have any questions before we begin? TOpics l. Ybur family--how it has influenced your attitudes and feelings about yourself and others, about college, goals, standards. 2. Impersonality of college 1ife--your opinions about this issue; if you feel that college life is imper- sonal, why, how does it affect you, what do you do about it; if not, why you have found it a personal experience. 3.' Adjustment to college--your opinions and reactions about making the transition to college: what, in terms of your experience, makes it easy of difficult. 63 10. 11. 12. 64 The student as an activist in political, social, and moral issueS--what would you say about his role here; what is your position; why. Public and personal goals--what influences students' goals; what hOpeS and aspirations have you set for yourself. Study habits and exam preparations--what have you learned about this from your experiences. College rules and regulations--your opinions and feel- ings about such matters as late hours for coeds, visiting in dorm rooms between men and women, for ex- ample. Dating habitS--your Opinion about dating patterns, behavior, marriage in college. Values and standards of college students--changing, revising, develOping personal values; what influ- ences them. Grading systems and examinationS--how do you feel about the way you're graded and tested; why. Social organizations--what Opinions do you have about, for example, fraternities and sororities as Opposed to remaining independent; what does joining a sOcial club mean to you. Relationship between students and instructors (and advisors)--what are your Opinions in terms Of your own experiences; for example, do you find that stu- dents can get personal assistance. APPENDIX B SEMANTIC DIFFERENTIAL INSTRUCTIONS AND SEMANTIC DIFFERENTIAL Semantic Differential Instructions The purpose of this study is to measure the mean- ipg of the concept (ME) to various people by having them judge the concept against a series of descriptive scales. In taking this test, please make your judgments on the basis of what the concept means pg ypg. Here is how you are to use those scales: If you feel that the concept at the top of the page is very closely related to one end Of the scale, you should place your checkmark as follows: fair X unfair fair : : : : : X unfair If you feel that the concept is ggite closely re- lated to one or the other end of the scale (but not ex- tremely), you should place your checkmark as follows: strong : X : 3. : : : weak or strong : : : : : X : weak If the concept seems only slightly related to one side as Opposed to the other (but is not really neutral), then you Should check as follows: 65 66 active : : X : : : : passive or active : : : : X : : passive The direction toward which you check, Of course, depends upon which of the two sides of the scale seem most characteristic of the concept you're judging. If you consider the concept to be neutral on the scale, both sides of the scale equally associated with the concept, or if the scale is completely irrelevant, unrelated to the concept, then you should place your checkmark in the middle space: safe : : X : : : dangerous Important: 1) Place your checkmarks in the middle Of spaces, not on the boundaries: . this not this : : X :___: : X: 2) Be sure you check every scale for the concept-- do not omit any. 3) Never put more than one checkmark on a single scale. 4) Make each item a separate and independent judgment. Don't puzzle over individual items. It is your first im- pressions, the immediate "feelings" about the items, that we want. On the other hand, please do not be careless, because we want your true impressions. Thank you. precise tasty happy withdrawing independent worthless kind strong meek excitable constricted sensitive clear sharp awful submissive cold open unfocused slow .0 Semantic 67 Differential U. 0‘ O. vague distasteful sad outgoing dependent valuable cruel weak aggressive calm expressive indifferent hazy dull nice self-assertive hot closed focused fast constrained content tense adaptable incongruent 68 0. free frustrated relaxed inflexible congruent APPENDIX C Q-SORT INSTRUCTIONS AND Q-SORTS Q-Sort Instructions Here is a set of 26 cards, each with a statement characteristic Of peOple typed on it. You are to sort these statements into a row Of 7 categories as you see in front of you. The number on each Of the 7 categories is the number Of statements that you will eventually place in each category. It might be convenient to first sort the cards into three piles, those you deem charac- teristic of you in one pile, those uncharacteristic Of you in another pile, and the remaining statements in a third pile. Then pick the statement you feel is the most characteristic Of you and place it in this category marked "most characteristic." Similarly, select the statement you feel is least characteristic (or most uncharacteristic) and place it in this pile. Work on through the categories, placing 3 statements you feel are "fairly characteristic" Of you here, 3 statements that are "fairly uncharacter- istic" of you here; then 5 statements in the "somewhat characteristic" category and 5 in the "somewhat uncharac- teristic" category, and 8 in the middle, or essentially "neutral" category. Since the instructions are somewhat complicated, take your time, but please try to be honest about your- self. Feel free to ask questions. Now sort this set of cards with reference to how you see yourself as a (college student) (socially). 69 70 Q-Sort "Myself Socially" Favorable 1. I can usually live comfortably with the people around me. 2. I behave in a giving way toward others. 3. I have a warm relationship with others. 4. I value my independence and autonomy. 5. My behavior is consistent with my standards. 6. I have a sense of humor. 7. I am dependable and responsible. 8. I have an attractive personality. 9. I am socially perceptive. 10. My personality is attractive to the Opposite sex. 11. I am assertive. 12. I am satisfied with myself. 13. I am the master of my environment. Unfavorable 1. I have few values and standards of my own. 2. I am a poor mixer. 3. I am inconsiderate Of others. 71 4. I put on a false front. 5. I often feel humiliated. 6. All you have to do is just insist with me and I'll give in. 7. I place the blame for my troubles on other people. 8. I keep a distance between myself and other people. 9. I feel cheated by life. 10. I feel insecure within myself. 11. I am afraid Of a full-fledged disagreement with a person. 12. I am no one. Nothing seems to be me. 13. I am afraid Of what other people will think of me. Q-Sort "Myself as a Collegg_Student" Favorable 1. I value intellectual matters. 2. I am critical, skeptical, not easily impressed. 3. I am able to see the heart of important problems. 72 4. I am self-reliant. 5. I am intelligent. 6. I feel adequate. 7. I am productive, I get things done. 8. I have initiative. 9. I am a competitive person. 10. I feel relaxed and little really bothers me. 11. I am a hard worker. 12. I can usually make up my mind and stick to it. 13. I am a rational person. Unfavorable 1. I usually feel driven. 2. I need somebody else to push me through on things. 3. I feel unsure Of my abilities. 4. I have a horror Of failing in anything I want to accomplish. 5. I am dissatisfied with myself. 6. I am an irresponsible person. 7. I can't seem to make up my mind one way or another. 8. I am disorganized. 9. My decisions are not my own. 73 10. I am confused. 11. I am a failure as a student. 12. I give up easily. 13. I am reluctant to commit myself to a course of action. uni! ‘ 4. u I. . I I r..llld!i1{l"l‘ 'u' 91.. I. .‘T .