V r fl. Lhfiéfifi;v;“'.3 .1...‘ ‘ ., “r $3.455” ‘1’, 313:.“ 2;.1351? ,{311 15%;. H «M ‘ . ‘r‘ui‘..\gw~l.3K""‘ w»: 131 524$?“ L V 1; a: 1?. .2 w "’ ' * ~ 43;. W “am; 3%.. . «=1 3:11.551 , ~ .. _ .. 5% QEL ' iv . -..L A“ -s-M:-.;. a... . . '~ . ‘33:: 9.1 .‘qug 2 l3, ,Cn’figx: ad “:1,“ nx .1133} k13“;fl°fi¢n ‘ fl ‘ Tunmn. My“? s..,L . A: 4|» .7 UL‘M \ i ‘91.: . . ‘4'}- . 1 “ii? ,1 *5: . .0. _ F” 2;? .3: .. fiifi . My .. o gas, «1qu 13.1.“, $ Wig? m w fiw%.§uw-mww Wm ”5-219 \ 1'.» _w: . a...» «gm-X‘ :‘aiak I .13 ~w u .. u Md -. _1.\_& 2L3; .. wuw "2:31: 5 $3,511 3 T. ‘- _‘ . Qdfiwxflm ‘. 5‘ '. ‘7 _~- , 1.1. ¥ , fiat? yaw g5“ a“ . 3} J'Fsun-I _ 1"»... ‘ C 56“ r \ _,, {v kt? “NE-m :""“- *‘ , '1 ~ 33% 1m» .éfiifl‘g. A ~ 7 -<.,:« .. . cm: -~‘ 121%? ‘ . , ‘ V ‘ .. .a “ ' v” w mmsta »- ~ 2 FL?" «.33.: L. V1 .. «l.“~- 5" ‘ v. - ~d.-... .‘1 _ ‘,__‘,‘ .1 v... s. ., H.-.» 3: n M. . ~1fit.’4.l~-h4 ~ . '1 \«n - 44"- s. ‘ § 4 1‘“ ‘ 4x 1.- 4:»; ':~'.‘L fhdk' Lifitd‘h. Ju. .1.~ in s..u-I'»ul.o.~i .. _ .0... , ' ' ‘ ‘ s \v» 1 u ‘. u. ¢ . ‘ 1 W 4 -173... _ 12‘..- ..::17 w ‘ -,. ;:1‘VA' "‘ 1 . -‘g. . '1.-- raw-mwéfil‘ ' ‘. .o r” ’y‘us‘i < 5: IV RSITY LIBRARIES 1 11.1111 111'111‘1111111 111 3 1293 00882 4082 l MICHIGAN STATE \ l This is to certify that the dissertation entitled Therapist Interventions: Their Relation to Therapeutic Alliance and Outcome“’ in Dynamic Psychotherapy presented by Mary Janice Gutfreund has been accepted towards fulfillment of the requirements for Ph - 0- degree in Psychology flim/ \' Major professor MS U i: an Affirmative Action/Equal Opportunity Institution 0- 12771 LIBRARY Michigan State University PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before due due. DWUE DATE DUE DATE DUE l—ll::|l__l HCLJ CHI |- II II I l——ll l MSU Ie An Affirmative Action/Equal Opportunity Institution ‘ CMMW1 THERAPIST INTERVENTIONS: THEIR RELATION TO THERAPEUTIC ALLIANCE AND OUTCOME IN DYNAMIC PSYCEOTHERAPY BY Mary Janice Gutfreund A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology 1992 fa k\ as “y d\ 7 6 ABSTRACT THERAPIST INTERVENTIONS: THEIR RELATION TO THERAPEUTIC ALLIANCE AND OUTCOME IN DYNAMIC PSYCHOTHERAPY BY Mary Janice Gutfreund The purpose of this study was to test causal models of relationships among symptoms, therapist interventions, patient alliance, and outcome in a sample of 46 cases seen at a university training clinic (median 29 sessions). For each case, four 20-minute samples, selected from the first and third sessions, and sessions in the middle and late phases of treament, were rated on the California Psychotherapy Alliance Scales (CALPAS) (Marmar & Gaston, 1989), and a scale of therapist interventions based on the Therapist Actions Scale (Hoyt, Marmar, Horowitz, & Alvarez, 1981) by teams of two raters each. Outcome was assessed through standardized gain scores on the SCL—90 GSI (Derogatis, 1977) and five items addressing dynamic outcome from the Post-Therapy Therapist Questionnaire (Strupp, Fox, & Lesser, 1969). Confirmatory factor analysis of the CALPAS revealed a three-factor structure, a positive and negative patient factor and a positive therapist factor. Data were analyzed using a path analytic strategy. Results from the path analyses were the following: level of symptom severity predicted positive alliance; transference interventions (first session) predicted early patient positive alliance (third session) and dynamic outcome, but were predictive of poorer symptom outcome. Patient positive alliance predicted symptom outcome only; patient negative alliance predicted worse dynamic outcome. Earlier alliance was not predictive of later alliance, and therapist interventions, with the exception of transference, were not predictive of other phenomena. Symptomatic and dynamic outcome scores were not significantly related. Post-hoe examination of bivariate correlations revealed some significant relations among variables which changed as therapy progressed. The discussion of these results included both methodological and theoretical considerations. It was felt that these findings were consistent with earlier studies; however, low inter-rater reliability probably attenuated what are usually weak positive results (e.g., alliance-outcome) in other studies. Recommendations for further research were made. For Larry ACKNOWLEDGEMENTS As I was nearing the completion of this work, my thoughts turned to the things I had thought I might be when I "grew up". Being a psychologist was not among them, but they all involved attending graduate school. In this context I see the completion of this dissertation as the fulfillment of a lifelong dream. It was not accomplished in a vaccuum: many people from both my personal and professional life nurtured and encouraged me. I would first like to thank Dr. Norman Abeles, my dissertation chair, mentor, colleague and friend. He has always been generous with his time, knowledge, and support. I have benefited much from his advice and his keen editorial pen. I take full responsibility for any six-line sentences remaining in this work! I thank Dr. Bertram Karon, who has taught me to put patients first and not be afraid to look upon the "Establishment" with skepticism. I thank Dr. Ralph Levine, who gave freely of his vast statistical knowledge; and Dr. Robert Caldwell, whose first advice as a supervisor, to think of saying goodbye when saying hello, has stayed with me. Additionally, I have benefited much from discussions with colleagues: in particular Louise Gaston from McGill V vi University; John Clarkin and Steve Hurt from New York Hospital, and the research group at the University of Michigan Psychological Clinic: Bob Hatcher, Alex Barends, Kim Leary, Julia Davies, Jenny Stuart, Jim Hansell, Marci Gittleman, and Cheryl-Lynn Podolski; and my friend and partner in dissertating, Gary Gunther. More thanks to my long-suffering raters: Tim Eaton, Richard Myer, Gary Gunther and George Ankuta; the unsung heroes of psychotherapy research! I am fortunate to have a list of friends and family that is too long to mention personally, but I would in particular like to thank: my parents, who fueled me with their love of education and justice; my in-laws for their support, understanding, and a place to stay in East Lansing, (and the laser printer!); friends and colleagues who put up with periods of grumpiness, antisocial-ness, and reminded me to laugh. I am grateful for the friendship of Karen Cruise and Ellen Luborsky, two psychologists who have been full- time graduate students and full-time single mothers, whose energy and enthusiasm seem boundless. And, to my dear husband Larry, lover, friend, and intellectual soulmate, who has been with me right along; I could think of no one better to share this with. TABLE OP CONTENTS LIST 0’ TABLES I O O O O O O O O O O O O O O O O O O O O O O O O O I O O O O O O O O O O O O O O 0 LIST or FIGURES O ...... O O O O O O O O .......... O O O O O O O O O O O O O O O 0 INTRODUCTION 0 C O O O I O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O O 0 LITERATURE REVIEW ....................................... Psychoanalytic Conceptualizations of the Therapeutic Alliance: A Brief Overview ....................... Empirical Research on the Therapeutic Alliance ..... Therapeutic Alliance and Outcome ................... Therapeutic Alliance and Pretreatment Characteristics .................................... Level of interpersonal functioning ..... ....... . Motivation ..................................... Coping ......................................... Expectations for treatment ..................... Pretreatment symptomatology .................... Therapist Influence on the Therapeutic Alliance .... Transference Interventions and Psychotherapeutic Process ........................................... Supportive Interventions and Psychotherapeutic Process OOOOOOOOOOOOOO...00.0.0...00.000.000.000... HYPOTHESES ............................................. Path Model #1: Alliance Changes Early in Therapy .. Path Model #2: Alliance, Transference Interventions and Dynamic Outcome ............................... Path Model #3: Alliance, Interventions, and Outcome ....................................... METHODS ................................................ Subjects ........................................... Selection of Cases for Analysis .................... Characteristics of the Study Sample ................ Data Collection .................................... Therapists ......................................... Raters ............................................. Measures ........................................... Therapeutic alliance ........................... Swmflmcmmwe.u.u.u.u.u.u.n.u.u.u.u Dynamic outcome ................................ Therapist interventions ........................ vii ix xi 12 17 27 27 29 30 30 30 31 43 49 53 57 59 61 63 63 63 64 65 66 66 66 66 72 74 75 viii RESULTS ................................................ Measurement Model Analysis of CALPAS Subscales ..... Summary Statistics for CALPAS ...................... Reliability of CALPAS .............................. Measurement Model Analysis of TAS Subscales ........ Summary Statistics for TAS ......................... Reliability of TAS ................................. Outcome Measures ................................... SCL-go 0......OOOOOOOOOOOOOOOOO00.000.000.000... Post-Therapy Therapist Questionnaire ........... Path Model #1: Alliance and Interventions Early in Treatment OOOOOOOOOOOOOOOOOOOOOOOOOIOOOOOCOOOCOOOOO Path Model #2: Alliance, Interventions, and Outcome ....................................... Post-Hoc Analyses: The Role of Supportive Interventions .......................... Supportive interventions and outcome .......... DISCSUSSION ............................................ General Review of Purpose and Findings ............. Transference Interventions, Alliance, and Outcome .. Alliance, transference, and dynamic outcome .... Alliance, transference, and symptom outcome .... Conclusions .................................... Pretreatment Symptomatology and Alliance ........... Alliance and Outcome ............................... Other Therapist Interventions, Alliance, and Outcome ....................................... Confrontation and/or clarification of defensive attitudes ........................... Supportive interventions ....................... Study Limitations .................................. Sample considerations .......................... Measures of pretreatment patient characteristics ............................... Measures of process variables (1) CALPAS ....... (2) TAS Measures of outcome ............................ Use of path analytic strategies ................ Recommendations for Further Research ............... Final Summary ...................................... APPMIX 0.00.0000...000......OOOOOOOOOOOOOO...0.0.0.... LIST OP REFERENCES 80 8O 84 84 87 88 93 93 94 95 104 119 119 130 130 131 131 133 134 136 137 139 139 140 141 142 142 143 145 146 146 147 149 150 162 LIST OF TABLES Table Page 1 Item-Factor and Factor-Factor Intercorrelation matrix: Final CALPAS (N=184) 0.00.00.00.00... ..... O 83 2 CALPAS Reliability: Intraclass Correlation Coefficients .............................. ........ 86 3 CALPAS Reliability: Finn's r Calculations ......... 87 4 Item-Factor and Factor-Factor Intercorrelation matrix: TAS (N=184) 00......OOOOOOOOOOOOOOOOOOOOOO 89 5 TAS Reliability: Intraclass Correlation coeffiCients O.......OOOOOOOOOOOOI....... ....... .0. 92 6 TAS Reliability: Finn's r Calculations ............ 93 7 Post-Therapy Therapist Questionnaire: Final - List of Dynamic Outcome Items ..................... 95 8 Correlation Matrix for Path Model #1 ............... 97 9 Matrices of Reproduced Correlations and Errors in Prediction: Path Model #1 ..................... 99 10 Significance Testing of Path Coefficients for PathMOde1’1(N=46) ....0.........OOOOOOOOOOOOOOOO 101 11 Correlation Matrix for Path Model #ZA, Dynamic outcome 0.00.0.00.........OOOOOOOO.......OIOOOOOOOO 108 12 Matrices of Reproduced Correlations and Errors in Prediction: Path Model #ZA, Dynamic Outcome ... 110 13 Significance Testing of Path Coefficients for PathMOdel ’ZA 0.0.000........OOOOOOOOOOO0.0.000... 111 14 Correlation Matrix for Path Model #28, Symptom outcome ......OIOOOIIOOOO......OOOIOOOOOIOOOOOO.... 114 ix 15 16 17 18 19 20 21 22 Matrices of Reproduced Correlations and Errors in Prediction: Path Model #23, Symptom Outcome ... 116 Significance Testing of Path Coefficients for PathMOdel #28 ......OOOOOOOOOOOOOOOOO0...... ...... 117 Correlation Matrix for Post Hoc Path Model predicting Dynamic Outcome from Supportive Interventions ......OOOOOOOOOO......OOO0...... ..... 121 Matrices of Reproduced Correlations and Errors in Prediction: Post Hoc Path Model predicting Dynamic Outcome from Supportive Interventions ..... 122 Significance Testing of Path Coefficients for Post Hoc Path Model predicting Dynamic Outcome from Supportive Interventions ..................... 125 Correlation Matrix for Post Hoc Path Model predicting Symptom Outcome from Supportive Interventions 0............OOOOOOOOOOOOOOOOO ....... 126 Matrices of Reproduced Correlations and Errors in Prediction: Post Hoc Path Model predicting Symptom Outcome from Supportive Interventions ..... 128 Significance Testing of Path Coefficients for Post Hoc Path Model predicting Symptom Outcome from Supportive Interventions ..................... 129 LIST 0? FIGURES Figure Page 1 Path Model #1: Alliance Changes Early in Therapy .. 58 2 Path Model #2: Alliance, Transference Interventions, and Dynamic outcome ......OOOOOOOOOOOOO00.0.0000... 60 3 Path Model #3: Alliance, Interventions, and outcome 00......0.........OOOOOOOOOOOOOOOOOOOO..... 62 4a Positive Alliance: 95% Confidence Intervals AroundMean ........OOOOOOOOOOO......OOOOOOOO00.0.0 85 4b Patient Working Capacity-Negative Aspects: 95% Confidence Intervals Around Mean ...... ........ 85 5a Transferencee Interventions: 95% Confidence Intervals Around Mean .............. 90 5a Transference Interventions: 95% Confidence Intervals Around Mean ..... ...... ... 90 5b Confront/Clarify Defensive Attitudes: 95% Confidence Intervals Around Mean .............. 90 5c Supportive Interventions: 95% Confidence Intervals Around Mean .............. 90 5d Expressive Interventions: 95% Confidence Intervals Around Mean ......... ..... 90 6 Path Model #1: Alliance Changes Early in Therapy ... 98 7 Path Model #ZA: Alliance, Interventions, and Dynamic outcome 0.0IO.....OOOOOOOOOOOOIOOOOOOO0.... 106 8 Path Model #28: Alliance, Interventions, and Symptom Outcome ................................... 115 9 Post Hoc Path Model predicting Dynamic Outcome from Supportive Interventions ..................... 124 10 Post Hoc Path Model predicting Symptom Outcome from Supportive Interventions ..................... 127 xi INTRODUCTION The therapeutic alliance, the aspect of the patient- therapist relationship that permits therapeutic work to take place, has been cited by psychoanalytic theorists as crucial to therapeutic success. A stable alliance is necessary for the patient to accept the interpetations of the analyst (e.g., Freud, 1913; Zetzel, 1956). Although most agree that a positive alliance is necessary for positive outcome in therapy, there has been little agreement to both its nature and the role each participant plays. A few state that it is merely a form of transference (e.g. Brenner, 1979), others a patient disposition (e.g., Frieswyk, Colson, & Allen, 1984), and others a collaboration (Bordin, 1979). Even those who see alliance as a patient characteristic, however, note that the therapist has a role in building and maintaining a stable therapeutic alliance. Empirical research on the therapeutic alliance has demonstrated a moderate relationship between alliance and positive outcome (e.g., Horvath & Symonds, 1991). However, the strength of alliance-outcome relationship varies with both method of measurement and type of outcome. Patient and therapist ratings of alliance, particularly the patient's, have been the best predictors of outcome. When observers 1 2 rate the alliance, usually only findings of a relationship between patient alliance and outcome is detected. This discrepancy among perspectives may be due to: the increased information participants have about their therapy, making them better predictors of outcome; bias on the part of participants (as outcome is often symptom change, which is rated by patients), or bias on the part of observers (who are therapists themselves) who may have difficulty rating other therapists. While these reasons must be taken into account, this study investigates a third perspective: whether or not the difficulty in measuring the therapist contribution to the therapeutic alliance and outcome is due to the absence of consideration of technique in most alliance measures. A review of the theoretical and empirical literature on dynamic psychotherapy suggests this may be the case. The purpose of this study will be to determine whether one can demonstrate a causal relationship between therapist interventions and patient alliance factors in dynamic psychotherapy. We seek to investigate whether therapist strategies which aim at identifying and interpreting negative alliance or supportive strategies which seek to suppress it improve therapeutic alliance, thus addressing a controversy among practitioners of psychodynamic psychotherapy (e.g. Luborsky, 1984). In addition, we seek to investigate the relationship of therapist interventions 3 to outcome, particularly the role of transference interventions. These phenomena will be studied using path analytic models. A path analytic strategy, using correlations corrected for unreliability, has been chosen in order to test a time-sequence model. In this model, alliance influences earlier interventions which in turn influence later alliance. We also wish to investigate relationships among patient characteristics, alliance, therapist interventions, and outcome. LITERATURE REVIEW c ana 1 Co on tual zations t 0 her t o W Nearly every paper on the therapeutic alliance begins with a discussion of Freud's conceptualization of the relationship between analyst and patient. Issues such as the nature of the therapeutic alliance (e.g., alliance as an individual versus a joint creation), the analyzability of patients who could not form stable alliances, and techniques for facilitating alliances, were all first addressed by Freud. In spite of later innovations in theory these issues as Freud stated them are still subjects of great interest in psychoanalysis and psychodynamic psychotherapy. Freud considered the establishment of a stable relationship between analyst and patient crucial for analytic success. It formed the background upon which interpretations could be made and heard: When do we begin our disclosures to the patient? The answer to this can only be: not until a dependable transference, a well-developed rapport, is established in the patient. The first aim of the treatment consists in attaching him to the treatment and the person of the physician. To ensure this one need do nothing but allow him time. If one devotes serious interest in him, clears away carefully the first resistances that arise and avoids certain mistakes, such an attachment develops in the patient of itself, and 4 5 the physician becomes linked up with one of the imagos of those persons from whom he was used to receive kindness. It is certainly possible to forefit this primary success if one takes up from the start any standpoint other than that of understanding, such as a moralizing attitude, perhaps,or if one behaves as the representative or advocate of some third person, maybe the husband or wife, or so on. (Freud, 1913, p. 139-140) Freud, particularly in his earlier writings, seemed to place the responsibility for the formation and maintenance of the therapeutic alliance with the analyst. This was communicated to the patient through interest, nonjudgmental listening, restraint, and clearing away of initial resistances. In spite of this, however, Freud also felt that in order for analysis to occur, there needed to be a portion of the patient's psychic structure that could regard the analyst as a good and stable object. Freud felt this positive transference was unanalyzable, and separate from the unconscious erotic and hostile transference which was the source of resistance. He speculated that those who could not form such a positive relationship under these conditions (such as individuals with severe character disorders or psychoses) were unsuitable for analysis. In his later writings, Freud (1937) re-emphasized the role of transference as primary motivator in a patient's commitment to analysis. A variant of this most extreme position has a minority of contemporary psychoanalytic theorists as adherents, whose most articulate spokesperson is Brenner (1979). In contrast to psychoanalytic theorists 6 (who are reviewed below) who draw on the developments of ego psychology and object relations psychology to develop a concept of therapeutic alliance, Brenner and his colleagues maintain that an emphasis on an alliance concept obscures the transference-based influence on all the patient's behavior in an analysis. Richard Sterba (1934) developed a theoretical model for the basis of analytic "rapport" within Freud's structural theory and was the first widely read analyst to write about the therapeutic relationship. In Sterba's view, analysis was possible because part of the patient's ego, based on reality, allied with the analyst against repression and the id. This took place on the basis of: a certain amount of positive transference, on the basis of which a transitory strengthening of the ego takes place through identification with the analyst. This identification is produced by the analyst...each separate session gives the analyst various opportunities of employing the term 'we' in referring to himself and to the part of the patient's ego which is consonant with reality Sterba, 1934, p. 120). This identification with the analyst through a dissociation of ego processes permitted the patient to examine his/her impulses and conflicts on the basis of adult reality. Through the use of interpretation, the portion of the ego tied up in instinctual conflict and defense decreased. This permitted increasing identification with the analyst in his/her reality function, and therapeutic change thus occured. Although not placed explicitly within the 7 terminology of therapeutic alliance, this implied that the working relationship can be seen as separate from the transference neurosis, which forms the basis of resistance. The therapeutic alliance, according to Sterba, may therefore be seen as a conflict-free ego function. Fenichel (1941), with a view similar to Sterba, postulated that the ego had both observing and experiencing aspects. The ability to use both these faculties was essential to the successful formation of a relationship sufficient to carry out analytic work. Greenson (1965) continued Sterba's and Fenichel's attempts to differentiate aspects of the analytic relationship. Greenson maintained the analytic relationship was composed of three parts: the working alliance, the transference neurosis, and the real relationship. He saw the working alliance much as Sterba saw it -- a split in the ego of the patient which permits both the observation and experience of regression. But, in his model, the working alliance and the transference neurosis were not always well-differentiated, as often aspects of the working alliance required analysis. All of this was well within traditional psychoanalytic theory. But Greenson added a third dimension -- the real relationship, the portion of the analytic relationship based on the real qualities of the analyst. As with all "real relationships," the real relationship often grows and changes during the course of the analysis, and can form a realistic basis for the working 8 alliance. He discussed this in reference to a clinical example: In addition to these transference reactions, however, Mr. C. also indicates some realistic awareness of me as a person to whom he is relating. He knows that I like him, that I keep trying to understand him, and that I am persistent and patient. He is also aware that I can be fooled, I can be wrong, and at times harsh. Yet he senses I have a good grasp of his underlying feelings and impulses, I must resemble him in some way; I am warm, not weak, and also not afraid of choosing words which get to the heart of the matter. Furthermore, he also realizes that psychoanalysis has no absolute standards for right and wrong. I submit that these are not distortions, but accurate perceptions and judgments based on his observations of me and my work during the 18 months of treatment. They coexist with the transference reactions and do not negate them. (Greenson, 1971, pp. 225). Zetzel's classic paper on transference and therapeutic alliance in the analysis of more disturbed patients (1956) is notable for her skillful contrast of the classical and the object relations view of the relationship of transference and alliance. She used this contrast to outline the reasons she explicitly advocated the use of ego- supportive interventions to bolster the positive alliance in more disturbed patients. Zetzel herself held the classical view that a therapeutic alliance needed to be established before interpretation of transference phenomena could be made. Without this relationship, the patient was susceptible to excessive regression in the analytic relationship, which she saw as a form of resistance. The ability to maintain a therapeutic relationship somewhat 9 separate from the transference neurosis is based on the presence of at least some mature ego functions. Zetzel was one of the first to advocate explicit use of supportive interventions in analyses of more disturbed patients with the aim of increasing ego strength. This represented a departure from traditional notions of the analyst never providing "gratification" (that is, never offering anything except interpretations and comments prepatory to them). Eissler (1953), writing about the same time, also created a theoretical basis for the use of what he termed "parameters", interventions which were aimed at bolstering the therapeutic relationship in more disturbed patients. Greenson (1967) writes of the use of "nonanalytic" interventions to help patients overwhelmed by affect to recover ego functioning, but he did not refer to the use of these interventions to bolster alliance. In contrast, as Zetzel reviewed it, (cf. Klein, 1965) the object relations view was that there is no clear separation between transference as part of the therapeutic alliance, and transference as transference neurosis. Transference is a form of object relationship, and even in its most adaptive form linked to the unconscious; furthermore, the structure of the ego is determined by its external and internal objects. Interpretation of the transference brings about change by changing the nature of object relationships within the ego. Therefore, preanalytic ego strength was not a prerequisite for analysis and 10 formation of a working relationship, and supportive interventions would not necessarily be alliance-building. Other modern psychoanalytic theorists have written on the concept of the therapeutic alliance as a collaborative process that includes transference as well as reality-based elements. Dickes (1975) considered alliance to be "all the elements favorable to the progress of therapy" (p. 1). Among these would be included the motivation for treatment to relieve the patient's suffering, the transference (both positive and negative), and the real, rational relationship between analyst and patient. The working alliance, he maintained, is only one portion of the therapeutic alliance. Hatcher and Hansell (1990) have conceived of the therapeutic alliance as a real entity, but having a shifting relationship to transference and therefore not easily separable. Sandler, Dare, and Holder (1973) wrote that the therapeutic alliance is based on "the patient's conscious or unconscious wish to co-operate and accept the therapist's aid in overcoming internal difficulties" (Sandler, Dare, & Holder, 1973, p. 30). They differentiated this from the wish to get better, which would likely be based on the patient's hopes for gratification, not insight. The most explicit advocate of the collaborative theory of therapeutic alliance is Bordin. Bordin, while coming originally from a psychoanalytic tradition, asserted that therapeutic alliance (or working alliance) is common to all types of therapy. He further asserted that the strength of 11 the working alliance is most important to positive therapy outcome, but different therapeutic approaches made different demands on patient and therapist (Bordin, 1979). Three features were characteristic of this working alliance: (1) an agreement on goals, as the examination of the role of the patient's childhood events in present adaptation in psychoanalytic therapy or target behaviors in behavior therapy; (2) an assignment of tasks, as homework in cognitive therapy or free association in psychoanalytic therapy; and (3) the development of bonds appropriate to a specific therapy, as the strength of bond for long-term intensive therapy might differ from that needed for short- term behavioral treatment. While the major thrust in alliance theory has been on its collaborative aspects, others have renewed their focus on therapeutic alliance as a patient variable. The most articulate spokespersons for this position have been Frieswyk and his colleagues. Frieswyk, Colson, and Allen (1984) defined the therapeutic alliance as the degree of the patient's active collaboration in the work of analysis. Other factors, such as the alliance with the analyst's ego, the real relationship, and the level of object relations which the patient brings to treatment are seen as contribgrgry but not part of the therapeutic alliance per se. Their original idea was to render the therapeutic alliance more amenable to investigation, but clearly their 12 theoretical position that alliance is something that the patient brings. In summary, the history of the therapeutic alliance concept in psychoanalysis contains an unresolved debate as to its nature, although all attest to its necessity to the success of analytic therapy. They differ to which alliance is wedded to transference; whether it is a patient variable, a patient variable facilitated by the therapist, or a collaborative process; or whether it even exists as all. All of these, of course, have different technical implications for the conduct of psychoanalysis and psychoanalytic psychotherapy. Empirical researchers have taken upon themselves the task of defining the nature of the therapeutic alliance, its course in treatment, and relation (if any) to outcome. as a ch o t e The ut l anc :-; ~ - 1 ;s r- a: ;-;, c - -19. -.. R-1;t~t ’_°c’s::4 As discussed in the previous section, the therapeutic alliance as a concept grew out of psychoanalysis. However, the effort to study the therapeutic relationship -- even to consider the therapeutic relationship as a relationship per se -- was begun by practitioners of client-centered therapy. Although these theorists (e.g. Rogers, 1957) did not use the phrase "therapeutic alliance," they engaged in the study of process variables thought essential to therapeutic success, such as empathy or level of respect for the other. 13 Later researchers have been engaged in an effort to conceptualize the therapeutic alliance itself in such a way that it can be measured accurately and interpreted meaningfully. Most of these investigators have been from the psychoanalytic and client-centered tradition, but many of these workers have taken Bordin's (1979) theoretical position that the therapeutic alliance exists and can be measured in all types of therapies. Two teams of investigators, one with a client-centered orientation, the other with both client-centered and psychoanalytic orientations, developed scales to study aspects of the therapeutic relationship thought to be related to successful outcome. Barrett-Lennard developed the Relationship Inventory (Barrett-Lennard, 1962), a scale which measures relationship aspects thought to be curative in client-centered therapy (Level of Regard, Empathic Understanding, Congruence, Unconditionality, and Willingness to be Known). He found that patients whose therapists who rated them more highly on these scales had better outcomes. Another scale, the Vanderbilt Psychotherapy Process Scale (VPPS) was designed to measure process variables distinguishing psychoanalytic, experiential, and alternate (nonprofessional helpers) modalities of therapy. Seven factor-derived scales appeared to measure three process dimensions: Exploratory Processes, Patient Involvement, and Therapist-Offered Relationship. Patient Involvement and Therapist-Offered Relationship have been found to be related 14 to outcome (Hartley & Strupp, 1983). (These studies are reviewed in the section on therapeutic alliance and outcome.) Other groups of investigators engaged in the study of psychodynamic psychotherapy have developed scales to measure the therapeutic alliance directly from session data. Despite a shared perspective, each group has a somewhat different focus in their conceptualization of what they believe to be the most salient aspects of the therapeutic alliance, paralleling the controversies about the nature of the alliance that were reviewed in the section on psychoanalytic theory. Some have focused on patient factors; others on patient and therapist factors; still others on factors unique to the interaction. Most seek to exclude considerations of technique in alliance formation, but the system from the Vanderbilt group includes it explicitly. Luborsky and his colleagues (Luborsky, 1976) originally conceptualized the helping alliance, their term for the therapeutic alliance, as having two components: Type 1, which consists of the patient feeling s/he has received help from the therapist, and Type 2, where the patient feels s/he is working in collaboration with the therapist against what is troubling him/her. The original scale measured patient behaviors only; therapist behaviors, labeled "facilitating behaviors," were added to a later revision and were rated separately. Like many researchers, Luborsky's group have 15 taken the theoretical position that the helping relationship exists separate from transference and from psychotherapeutic technique. A more explicit measurement of alliance as a patient factor has been attempted by the Menninger group (Allen, Newsom, Gabbard, & Coyne, 1984; Frieswyk, Allen, Coyne, Gabbard, Horwitz, & Newsom, 1986; Frieswyk, Colson, & Allen, 1984), who also sought to differentiate alliance from both transference and from therapist technique. To that end, the group measured only the patient's collaboration, that is, "the extent to which the patient makes active use of the treatment as a resource for constructive change" (Frieswyk et al., 1986). Separate scales measuring "transference- based dimensions" of trust, acceptance, and affect expression have also been developed. These elements, which the authors labeled "mediating variables", are hypothesized to be those which contribute to the formation and maintenance of the alliance. In the following therapeutic alliance systems, the alliance is conceived as containing both therapist and patient factors. Marziali, Marmar, and Krupnick (1981) developed the first version of the Therapeutic Alliance Rating System (TARS) (later revisions known as the California Therapeutic Alliance Rating System, or CALTARS). This scale was specifically designed to exclude items involving technique, action, and specific response, and to include items focusing on the "affective, attitudinal l6 aspects" of the therapeutic relationship. Both positive and negative factors for therapist and patient were included. A factor analytic study done with the CALTARS (Marmar, Weiss, & Gaston, 1989) found five factors, which they named: Therapist Understanding and Involvement, Patient Hostile Resistance, Patient Commitment, Therapist Negative Contribution, and Patient Working Capacity. Based on the results of the previous study and in an effort to broaden the applicability of the CALTARS to other modalities of psychotherapy, the Marmar group has developed the California Psychotherapy Alliance Scale (CALPAS), which seeks to measure four theoretical dimensions of therapeutic alliance: Therapist Understanding and Involvement, Patient Working Capacity, Patient Commitment, and Working Strategy Consensus, the latter being a scale of collaboration and a direct outgrowth of Bordin's (1979) theory. The scale comes in a therapist, patient, and rater version. The Vanderbilt group (Hartley & Strupp, 1983; O'Malley, Suh, & Strupp, 1983) developed the Vanderbilt Therapeutic Alliance Scale (VTAS). In their system, in contrast to the others, the therapeutic alliance was conceived as the interaction of both relationship and technical factors rather than a product of relationship elements alone. Although they originally proposed three a priori relationship factors for the alliance composed of therapist items, patient items, and interaction items, factor analytic techniques grouped and redistributed these items into six 17 factors, each related to specific patient and therapist factors. WWW Before the development of formal scales to measure the therapeutic alliance, investigators studied relationships between process variables related to alliance and outcome. Many of these variables were derived from client-centered theory; but as can be seen from the following they have some robustness as predictors of patient outcome, particularly as rated by participants. Strupp, Fox, and Lesser (1969) used post-therapy questionnaires to assess therapists' perceived warmth and understanding, and quality of relationship. These were related to patients', therapists', and judges' evaluation of positive outcome. Saltzman, Luetgert, Roth, Crease, and Howard (1976) measured helping relationships in client-centered therapy through the use of self-report forms. By the third session, ratings of clients' felt level of improvement correlated significantly with ratings of feelings of being understood by the therapist and seeing the therapeutic relationship as unique. Therapists' assessment of change was related to clients' ratings of felt respect, understanding, openness, security, movement, sense of continuity, and expression of affect related to treatment. For therapists, level of respect for the patients correlated with their assessment of change. First session ratings had little predictive significance. 18 Three studies using the Vanderbilt Psychotherapy Process Scale (VPPS), an observer-rated scale, showed relationships between process variables and positive outcome. Gomes-Schwartz (1978) demonstrated differences in successful and unsuccessful cases of college student patients undergoing either brief psychodynamic or experiential therapy. Level of Patient Involvement was related to outcome as measured by clinicians and therapists, and therapists' ratings of target complaints, whereas Therapist-Offered Relationship was only related to therapist ratings of outcome target complaints. This relationship was similar across therapy modalities. Using the same sample and the VPPS, O'Malley, Suh, and Strupp (1983) found that the predictive association of the VPPS and outcome went from virtually none in the first session to a consistent association in the third. Overall therapist ratings of outcome were predicted by Patient Involvement, Exploratory Processes, and Therapist-Offered Relationship at a statistically significant level. Clinician and patient ratings of improvement were also predicted by ratings of Patient Involvement. A third study used the VPPS in a sample of adult outpatients in psychodynamic psychotherapy (Windholz & Silberschatz, 1988). Patient Involvement as measured from a session at the middle of treatment (session eight) was related to therapist ratings of patient outcome and decreased scores on the Global Assessment Scale (GAS). 19 Therapist-Offered Relationship was related to decreased scores on the GAS and reduction in target complaints. Patient and evaluator ratings of outcome were not related to these outcome dimensions, although evaluator ratings of global change and reduction in target complaints approached statistical significance. The Luborsky group (Luborsky, 1976) demonstrated a positive relationship between therapeutic alliance and outcome. In their initial study, they compared the ten most-improved and ten least-improved patients in the Penn Psychotherapy Project. Those patients seen for more than 25 sessions were selected for study, resulting in seven in the most-improved category and eight in the least-improved. Four 20-minute transcript excerpts from each patient were rated by external judges for signs of a helping alliance. Both positive and negative signs were counted. They found that helping relationships developed in six of seven most improved patients and in none of the least-improved. Patients who developed helping alliances did so quite early, by the third to fifth session. These were Type 1 alliances (the patient's sense that s/he is being helped); Type 2 alliances (the patient's sense of working jointly with the therapist) developed only toward the end of treatment and in only two patients. A later study utilized a global rating form rather than counting signs form of the Helping Alliance Scale (Morgan, Luborsky, Crits-Christoph, Curtis, & Solomon, 1982). Level 20 of alliance correlated significantly with outcome as measured by a composite of pre-and post-therapy measures including the Health-Sickness Rating Scale and the MMPI (Residual Gain), and with ratings of change reported by both therapist and patient (Rated Benefits). Contrary to predictions, the two types of helping alliances were not differentially predictive, but a trend toward an increase in Type 2 helping alliance toward the end of therapy was noted in the most improved patients. Surprisingly, observer measures of patient insight and resistance were not correlated with outcome. A further study with the Helping Alliance Scale (Luborsky, McLellan, Woody, O'Brien, & Auerbach, 1985) found ratings of helping alliance significantly related to outcome in brief (3-24 sessions) psychotherapy of methadone-maintained, drug-dependent patients. Hartley and Strupp (1983), using the Vanderbilt Therapeutic Alliance Scale (VTAS), studied the individuals from the Vanderbilt Psychotherapy Project. They were divided into three categories: high outcome, low outcome, and premature terminators (less than five sessions). Scores on the original subscales which defined therapist actions, patient actions, or interactions on the empirically-derived scales distinguished outcome groups, although there was more variance within groups than between them. However, by using the empirically-derived subscales, differences in the pattern of the therapeutic alliance across time among 21 outcome groups were discovered. In the dropout group, Positive Climate as well as Therapist Intrusiveness were significantly higher in the last session, while in the high outcome group, only Positive Climate was significantly higher. Additionally, in the high outcome group, all indices of therapeutic alliance peaked to about the 25% point in treatment, and trailed off as therapy progressed. For the low outcome group, there was a slight decrease in all variables at the 25% point. At the 25% point, high outcome patients tended to be less resistant, more motivated, and more anxious. End-of—treatment outcome effects reached significance, but there was a trend for low outcome patients to score higher on the Resistance and Anxiety scales. The authors believed the drop in therapeutic alliance scores was the failure of study therapists to deal with termination issues in a brief (ZS-session) therapy. They felt that the increase in Therapist Involvement scores seen in the therapists in the dropout group may have been in response to the slight trend of the patients to be more defensive and less involved in treatment. Several studies using the Therapeutic Alliance Rating System (TARS/CALTARS) (Marmar, Marziali, & Krupnick, 1981) demonstrated a relationship between alliance and outcome. The original validation sample was selected from a sample of 25 patients treated in 12-session brief therapies for reactions to severe life stress, such as the death of a 22 parent or spouse. Ten subjects, five with good outcomes and five with poor outcomes, were selected for study. Raters based their ratings on listening to audiotapes of the second, fifth, eighth, and eleventh sessions of each therapy. Intercorrelations between scales in this study were such that Therapist and Patient Positive and Negative scales were collapsed into a Therapist Total Contribution Scale and a Patient Total Contribution Scale. While the Therapist scale did not distinguish between outcome groups, the Patient scale did. Patients who formed poor therapeutic alliances had poor treatment outcomes as measured by both self-report and by clinical judges. The authors felt the study results indicated that the patient's negative disposition seemed to determine the course of treatment and that the therapist's efforts to offer a positive relationship with the patient were not helpful in this regard. A second study (Marziali, 1984a) compared therapeutic alliance from the viewpoint of therapist, patient, and clinical judge and its relation to outcome in an effort to provide further validation of the observer-rated measures. Forty-two patients who met Malan's criteria for suitability for brief psychotherapy and who completed a 20-session treatment were chosen for study. Therapist and patient completed therapeutic alliance measures immediately following sessions 1, 3, 5, 10, 15, and 20; these sessions were also rated by judges. Outcome was measured by patient 23 self-report scales, therapist scales, and clinician ratings. Therapists' and patients' alliance responses were associated with positive therapeutic change. Patient and therapist ratings of positive alliance were associated with decreased symptoms. Judges' ratings of patient positive alliance were associated with patient and therapist evaluations of change and clinical evaluations of dynamic outcome. Averaged rater responses on the Patient Positive and Therapist Positive Alliance contributions were significantly lower in the first and third sessions in contrast to the final session, although a significant relationship between alliance and outcome was established. The relationship between participants' ratings of therapist alliance and outcome constituted the major new finding of this study, contrasting with the first study in which only patient alliance was associated with positive outcome. Raters may have had less sensitivity to the effects of the therapist on the therapy relationship, which may have influenced the decision to collapse scales in the first study. In this study therapist and patient positive and negative alliance factors were relatively non- correlated, supporting theory that predicts they are to a degree independent of each other. This inconsistency in results using the CALTARS prompted the Marmar group to conduct studies which might further delineate the relationship between therapeutic alliance and outcome (Marmar, Weiss, & Gaston, 1989). A 24 sample of 52 partcipants who underwent a 12-session therapy for pathological grief was studied. Ratings of therapeutic alliance for sessions 2, 5, 8, and 11 were obtained from trained observer-judges and averaged. Again, only patient factors affected outcome: Patient Working Capacity was found to be related to increased interpersonal functioning at termination. In a further study with this same group of patients, Horowitz, Marmar, Weiss, DeWitt, and Rosenbaum (1984) found the only zero-order correlation to be association between patient negative alliance and lessened symptom change. The patient's positive alliance was not directly predictive of outcome but only in interaction with other variables. Two other studies using the observer form of the CALTARS drew opposite conclusions concerning the relationship between alliance and outcome. Klee (1986) found, also using the observer version of the CALTARS, studying a sample of 32 outpatients selected from cases seen at a university-based clinic, that alliance factors were not directly related to outcome but interacted with both prognosis (as measured by alliance measured in the first 10% of treatment) and phase of treatment. In contrast, Eaton, Abeles, and Gutfreund (1988) studied 40 cases of dynamic psychotherapy selected from a general outpatient sample. Alliance levels were established early in treatment and remained relatively stable throughout therapy. For the entire sample, positive patient alliance was associated with 25 decreases in symptomatology as reported by the patient. When patients were grouped by length of treatment, however, some differences emerged, particularly in the relationship between alliance and participant ratings of outcome. There was a trend for therapist positive alliance to be lower in the short length group (less than 20 sessions). Therapist ratings of positive outcome were associated with high scores on Patient and Therapist Negative Alliance and negatively associated with Patient Positive Alliance in the long length group (greater then 40 sessions). In the medium—length group scores on the Patient Positive Alliance scale were associated with positive outcome as rated by both participants. In an attempt to study therapeutic alliance comparatively, Marmar and his colleagues studied elderly depressed outpatients who underwent behavioral, cognitive, or brief dynamic therapy (Marmar, Gaston, Gallagher, & Thompson, 1989). Ratings were made using the therapist and patient versions of the California Psychotherapy Alliance Scale (CALPAS), the revision of the CALTARS specifically designed to be applicable to a broad range of therapies. Therapist's ratings of Patient Commitment and Patient Working Capacity were related to decreased scores on the Beck Depression Inventory and the Hamilton Rating Scale for Depression. When outcomes of the three different modalities were compared, alliance-outcome relationships were strongest in the cognitive therapy condition. In this sample, 26 alliance and outcome was not related for brief dynamic therapy. This suggests that different treatment modalities may possess different relationship between alliance and outcome. In summary, these studies suggest therapeutic alliance is generally associated with positive outcome. This relationship appears to be absent at the very beginning of treatment but emerges very early; most studies demonstrated measurable effects at the third session. The findings of the studies vary largely with the method used to study alliance and outcome; observer-rated measures tend to find relationships between patient alliance and outcome only, particularly for global outcome and dynamic factors; participants's ratings of alliance are more predictive of their assessment of outcome, including symptomatic outcome. This conclusion was similar to that drawn by Horvath and Symonds (1991) from their meta-analysis, which demonstrated a moderate but consistent relationship between positive therapeutic alliance and outcome, with the strongest relationships shown by patient ratings. The latter suggests that the perspective of the rater is an important factor to consider in evaluating these results. It could be that participants are more accurate observers of the therapeutic process; or they could share a bias which also distorts their judgment of outcome. Observer-raters may escape this but perhaps at the cost of some information. The near- absence of a relationship between observer-rated therapist 27 alliance and outcome may reflect a bias on part of raters (who, by the nature of the scales, must also be clinicians) which render the judgment of therapists' alliance less reliable; or it could suggest that something other than therapist alliance influences the outcome of therapy, particularly dynamic psychotherapy. Additionally, the relationship between alliance and outcome may be mediated by other factors. Some of these are considered in the next section. e ti 1 a co n P et e 0 ha act c nggl 9; interpersonal fungtiorigg. Traditional clinical lore maintains that successful psychotherapy is dependent upon a certain history of attainment one or more satisfactory interpersonal relationships prior to therapy. Freud's concept of "unanalyzable" positive transference centers on this principle. Hence, it would follow that pretreatment interpersonal functioning should have significant effects on the therapeutic alliance. Empirical research is mostly supportive of this finding; however, other factors, such as type of alliance variable, and sample, seem to affect this relationship. Moras and Strupp (1982) found clinicians' judgments of pretherapy interpersonal functioning was related to level of patients' involvement in psychotherapy for college students. Similarly, clinicians' ratings on the Health-Sickness Rating Scale was related to the formation of a strong helping 28 alliance (Morgan, Luborsky, Crits-Christoph, Curtis, & Solomon, 1982). Piper, deCarufel, and Szkumelak (1985) noted, in a group of 21 prescreened outpatients who were seen in brief dynamic therapy, that quality of defensive processes and object choice were related to process measures. Quality of defensive processes (based on Vaillant's hierarchy of defenses -- see Vaillant, 1977) and of object choice were strongly predictive of processes favorable to the development of a therapeutic alliance (measured here as judges' evaluation of revealing of private material and degree of understanding the therapist's interventions, as rated by the patient) averaged across therapy. These variables were also associated with positive outcome. In a sample of patients seeking psychotherapy for bereavement reactions, higher relationship composite scores on the Patterns of Interpersonal Change Scales (PICS) measured pre-treatment were related to higher scores on the Patient Working Capacity Scale of the CALTARS (Marmar, Weiss, & Gaston, 1989). However, three further studies, using subjects from the same sample found no association between PICS interpersonal functioning and the four main subscales of the CALTARS. Two of these studies, however, demonstrated a relationship between a scale measuring developmental level of the self- concept, an object relations scale (a correlate, presumably, of quality of interpersonal relationships) and Patient Positive Alliance (Horowitz et al., 1984; Marmar et al., 29 1986). In a sample of elderly patients seeking psychotherapy for depression, no relationship was found between interpersonal functioning as measured by the Young Loneliness Inventory and alliance as measured by the CALPAS (Gaston, Marmar, Thompson, & Gallagher, 1988). The authors felt that sample factors accounted for the lack of relationship, that is, elderly patients seeking treatment for depression may not demonstrate the same range of difficulties in interpersonal relationships as younger patients. This is similar, the authors pointed out, to the lack of association between these variables generally found in the sample of bereaved patients who do not present for psychotherapy because of interpersonal or character problems. They speculated different samples may yield different relationships among these variables. ngriggrigrr Motivation for psychotherapy appears to affect the therapeutic alliance and its relation to outcome. Marmar et al. (1986) demonstrated a positive relationship between pretreatment motivation and outcome. A more detailed analysis by Marmar, Weiss, and Gaston (1989) showed motivation related to the Patient Working Capacity on the CALTARS. Horowitz et a1. (1984), using the same sample, took a somewhat different approach using multiple regression. They regressed level of motivation, alliance, and an interaction term composed of these two variables on symptomatic outcome (SCL-90). They found that Patient Positive Alliance was positively related to outcome at low 30 levels of motivation. As motivation increased, the relationship between Patient Positive Alliance and outcome went from positive to negative. A similar finding was discovered with the Patient Negative Alliance scale. ggpiggr Gaston, Marmar, Thompson, & Gallagher (1988) studied the relationship between defensiveness as measured by Avoidance Coping Strategies and therapeutic alliance. Patients high in defensiveness had lower scores on the scales Patient Working Capacity and Patient Commitment, showing less commitment and less ability to engage in the self-reflection necessary for psychotheraputic work. figpggtgtiogs gor treatgggt, Gaston, Marmar, Gallagher, and Thompson (1989) studied the relationship between patients' expectations for therapy and the therapeutic alliance. Decreased therapist ratings of Patient Commitment (on the CALPAS patient version) were associated with the patient's feelings of being helped by insight and support in brief dynamic therapy. This contrasted with cognitive therapy where the relationship was positive. r e 3 tom tolo Many studies of therapeutic alliance have found no relationship between pretreatment symptomatology and therapeutic alliance (Hartley & Strupp, 1983; Marmar et al., 1986; Marmar, Gaston, Thompson, & Gallagher 1988; Marmar, Marziali, & Krupnick, 1981; Piper et al., 1985). However, this is not a uniform finding. Marmar, Weiss, and Gaston (1989) found 31 Total Pathology Scores on the SCL-90 were associated with decreased scores on Patient Working Capacity of the CALTARS. An association was also found between level of stress from recent life events and decreased Patient Working Capacity and Patient Commitment. Eaton, Abeles, and Gutfreund (1988) also found therapeutic alliance to be adversely affected by pretreatment symptomatology. Luborsky et a1. (1983) found a similar association using the Helping Alliance Questionnaire. They also demonstrated a relationship between a clinician-rated measure of somatic symptoms on the Klein Somatic Scale and lower scores on the Helping Alliance Scales. There are two possible explanations that may account for the divergent findings. It may be that the disorganization caused by intense affect (one source of which may be multiple life stressors) impairs alliance formation, particularly a capacity to engage in the self- reflection and exploration required by dynamic therapy. Another explanation is suggested from a study by Rosenbaum, Horowitz, and Wilner (1986). Patients who reported increased symptomatology on the SCL-90 were perceived by their evaluators as being more difficult, and therefore may present more of a challenge to the therapist. er u As discussed previously, much research to date on the therapeutic alliance has shown patient factors to have a 32 direct influence on the alliance. Fewer studies have concluded similarly for therapist factors. However, the preponderance of psychoanalytic (as well as client-centered) theory has emphasized the therapist's role in creating and maintaining a therapeutic alliance, while taking patient factors into account. This role of the therapist has been emphasized particularly for more disturbed patients. The discrepancy between theory and empirical research will be considered in the following section. While most alliance systems attempt to eliminate technical considerations, the following review of the literature on the role of the therapist suggests that technical factors must also be considered in understanding the therapist's contribution to the therapeutic alliance. Luborsky et a1. (1986) analyzed results from several outcome studies such as the Penn Psychotherapy Project, the Vanderbilt Psychotherapy Project and found that individual therapist effects accounted for significant amounts of the variance in outcome. Additionally, this difference held across individual patients. He concluded that who one's therapist was did make a difference in outcome no matter what the treatment modality. Additionally, good therapists worked equally well with a variety of patients, not just certain ones. Sachs (1983) found that level of errors in technique in psychodynamic psychotherapy was significantly related to outcome. In the Menninger Psychotherapy Project (Kernberg et al., 1972), therapist skilfullness, rated by 33 judges, was associated with outcome. This was particularly true for patients with low ego strength. Luborsky, McLellan, Woody, O'Brien, and Auerbach (1985) demonstrated that personal qualities of the therapist significantly affected alliance and outcome. Qualities such as “interest in helping patients" and "unusually psychologically healthy" and "very capable and skillful therapist" were highly related to helping alliance (r=.74) but not to outcome. However, helping alliance itself was related to treatment outcome (r=.65). The authors drew the following conclusion: "The therapist's ability to form an alliance is probably the most crucial determinant of his effectiveness...(authors' emphasis)... On the other hand, there is some evidence that the helping alliance is also influenced by the predisposition that the patient brings to treatment... On balance then, the helping alliance appears to be an interactive product of therapist and patient qualities." (Luborsky, McLellan, Woody, O'Brien, & Auerbach, 1985, p. 610). Five studies have examined the role of therapist interventions on the therapeutic alliance. O'Malley et a1. (1983) examined the relationship of process to outcome in patients designated as high and low prognosis based on first session scores on the Patient Participation scale of the Vanderbilt Psychotherapy Process Scale. High prognosis patients were defined as those individuals who had high scores on this scale at the first session, and low prognosis 34 as those who had low scores on this scale. Scores on therapist factors -- Warmth and Exploration -- increased over the first three sessions for patients with high outcome and decreased for patients with low outcome, regardless of prognosis. For patients with low outcome there were decreases in Therapist Warmth and increases in Negative Therapist Attitude. For low prognosis-high outcome cases, there was a significant increase in Patient Participation. Positive outcome across measures was associated with both the absolute level of Patient Participation and rate of increase; positive outcome was also associated with increases in Therapist Warmth and Exploration over the first three sessions. Foreman and Marmar (1985) examined videotapes of six patients in therapy who demonstrated initially high levels of negative alliance; three subsequently had lowered negative alliances and positive outcomes and three did not. The authors made up a list of potential actions to observe from "general psychoanalytic interpretive techniques... identifying the patient's defense against anxiety or about an underlying feeling or impulse, as well as interpreting defense before impulse...the patient's conflict should be interpreted in relation to others in his everyday life, in relationship to the therapist (in the here-and-now) and in relation to the parents (usually in the past)." (Foreman & Marmar, 1985, p. 925). Therapist actions that distinguished between the two groups were: 1) addressing defenses the 35 patient used to deal with feelings in relationship to the therapist and to others; 2) addressing problematic feelings in relationship to the therapist; 3) linking the patient's problematic feelings about the therapist with the patient's defenses; and 4) addressing the "triangle of punishment," that is, the patient's need for self-punishment to assuage the guilt over anger or responsibility for the patient's suffering. The authors noted that substantial interpretive work was done in all six cases, but only those interventions that addressed difficulties in the relationship with the therapist brought about results and those that avoided it did not. A second study utilized a new scale, the Alliance Building Action Scale (ALBAS) (Gaston, Marmar, & Ring, 1988) developed from the results of the previous study and with the addition of items taken from the literature on the difficult patient in cognitive therapy. This scale was then used to assess the relationship between therapist actions and alliance in cognitive therapy of depression in a sample of elderly depressed outpatients. Five patients were selected -- three with improved alliances and better outcomes, two with continuing problematic alliances (defined as continuing high levels of patient negative alliance). In contrast to the previous study, therapists' addressing of the patients' interpersonal relationship difficulties was related to increased alliance and improved outcome. The patient's problems in relationship to the 36 therapist were rarely addressed in either group. The authors were surprised at the latter finding and suggest further research with more diverse patient samples is necessary to explain it. Gabbard, Horwitz, Frieswyk, Allen, Colson, Newsom, & Coyne (1988) studied the effect of therapist interventions in the therapy of one hospitalized borderline patient. They selected six widely-spaced sessions from over three hundred completed at the time of the study. Therapeutic alliance was measured by shifts in the patient's collaboration with the therapist using the Frieswyk et a1. (1984) scale. Therapist interventions were rated using Gill's Process Coding Categories. Shifts in collaboration were associated with therapist interventions which focused on the therapeutic relationship, particularly transference interpretations. In 11 of 13 instances there was a shift toward increased collaboration. In the two instances where a downward shift occured, subjective examination of the clinical material revealed possible technical difficulties in the timing and depth of the interpretation which may have accounted for the shift away from the therapist. The authors suggest the use of this technique with other borderline patients and those in other diagnostic categories to determine if this phenomenon is true across patients. Luborsky (1984) has identified what he terms "therapist facilitative behaviors" which have been shown to be highly correlated with the presence of a helping alliance and have 37 a low to moderate correlation with outcome (Luborsky, McLellan, Woody, O'Brien, & Auerbach (1985); Luborsky, Crits-Christoph, Mintz, & Auerbach, 1988) As can be seen from the following list of therapist facilitative behaviors, Luborsky advocates bolstering the helping alliance through positive support: Helping Alliance 1: (the patient's sense s/he is being helped): 1) Convey through words and manner support for the patient's wish to achieve the goals of therapy; 2) Convey a sense of understanding and acceptance of the patient; 3) Develop a liking for the patient; 4) Help the patient maintain vital defenses and activities which bolster the level of functioning; 5) Communicate a realistically helpful attitude that the therapy goals are likely to be achieved; 6) Recognition, on appropriate occasions, that the patient has made some progress toward the goals; 7) Encouraging some patients to express themselves on some occasions; Helping Alliance 2 (the patient's sense of working jointly with the therapist): 1) Encourage a "we bond"; 2) Convey respect for the patient; 38 3) Convey recognition of the patient's growing ability to do what the therapist does in using the basic tools of treatment; 4) Refer to experiences that the patient and therapist have been through together; 5) Engage in a joint search for understanding (Luborsky, 1984, pp. 82-89, partial). This list of behaviors is quite similar to those in the Therapist Positive Alliance Scale on the CALTARS, with two exceptions: the prescription of specific measures, and the emphasis on support of vital existing defenses. The latter is explicitly a supportive technique. Luborsky's technique for strengthening the therapeutic alliance emphasises maintaining and supporting areas of competence, and directly supporting the positive alliance rather than confronting the negative alliance. Luborsky states that such alliance- strengthening techniques are needed either when an otherwise well-functioning patient is overwhelmed by stress or anxiety, or with severely disturbed patients who show low anxiety tolerance and poor control over impulses. This view contrasts with that of Foreman and Marmar who suggest confrontation of difficulties in the therapeutic relationship is essential for improvements in the alliance. Luborsky prefers to view this confrontation not as alliance- bolstering but the work of the therapy itself, enabled by the presence of a positive helping alliance. Here it can be 39 seen that Luborsky takes the classical psychoanalytic view that a stable alliance is necessary for therapeutic work (that is, interpretation of transference) to take place, and like Zetzel advocates the use of supportive techiques. An explanation of the differences among these studies may be partially explained by Luborsky's understanding of the effect of the presence of positive and negative helping alliances on treatment. He noted that the presence of a positive helping alliance was most predictive of therapeutic success. Patients who had successful outcomes in psychotherapy possessed both positive and negative helping alliance signs. The presence of negative helping alliance signs in the absence of positive ones was most predictive of therapeutic failure. He would, therefore, be more likely to focus on building the more positive aspects of the helping alliance first. One may conceptualize the therapist's role in fostering the therapeutic alliance as containing two aspects; the provision of an atmosphere conducive to developing a positive attachment, and a way to identify and discuss what patient attitudes may be preventing the formation of a stable therapeutic alliance. The relevant issue involved is: how does one reduce disruption in the therapeutic relationship, particularly that due to negative affects? Does one consider it a manner of ego weakness and act in a supportive manner? Does one treat it through clarification and interpretation? This dilemma is theoretically difficult 40 to separate from analysis of both resistance and transference; indeed some would identify it as a subset of resistance. A brief review of relevant analytic literature follows. Greenson (1967) in his work on standard psychoanalytic technique, enumerates the classical view of the analysis of resistance and transference in the analysis of neurotic patients. He points out that resistance, an effort to avoid a painful affect such as anxiety, guilt, or shame, is present at all points in a psychoanalysis. This painful affect is a result of a traumatic event where the ego was overwhelmed by conflict between reality and an instinctual impulse. Analysis of resistance involves these steps: confrontation, clarification, interpretation, and working through. Greenson suggests that beginning work on resistances is in association with elaborating the nature of the resistance (e.g., discussion of sexual issues in treatment cannot take place until the patient's difficulties about talking about sex are addressed). These resistances usually involve transference phenomena but Greenson would not advocate interpretation until it is both apparent as a resistance and when its genetic origins become apparent. Gill and Muslin (1976) suggest that transference is present from the beginning of treatment, and that transference manifested in the here—and-now must be addressed early in treatment. A failure to do so, particularly with negative transference, may lead to a 41 distortion of the therapeutic relationship and/or premature termination. They also state that it is not necesary to know genetic origins of transference in order to work meaningfully with it. They advocate an interpretive approach to early alliance difficulties, presumably caused by negative transference phenomena. These theorists have written about their analytic work with neurotic patients. More disturbed patients, such as patients with borderline personality, present for therapy with more negative transference and a disinclination to form an alliance with the therapist around the tasks of therapy. Kernberg (1986), from his work with borderline patients, suggests that transference be confronted and interpreted as soon as it becomes clear and presents as resistance. He maintains that interpretation has an ego-strengthening effect and opens the way for feelings of being understood and accepted by the therapist. Kernberg also advocates interpretation of idealizing transference as well. However, he also notes that psychoanalysis proper does not work with these patients as they are unable to withstand the regression inherent in the analytic situation. They require the structure of expressive psychotherapy, which includes the introduction of support by means of a treatment contract as well as occasional non-analytic "parameters" which provide structure to the treatment. He does eschew the use of supportive manuevers and cautions the therapist to 42 analyze with the patient any non-analytic intervention for its transference-inspired meaning to the patient. Masterson (1978), in his approach to borderline patients, maintains that setting parameters of treatment, that is, maintenance of a set hour and a reasonable fee and the confrontation, within session, with reality factors, builds a stable therapeutic alliance. This strategy has clear supportive implications. Only then, when the therapist is experienced as a real object, can interpretive work on the patient's central issues take place. Horwitz (1985) maintains that the divergent styles of technique -- interpretation versus relationship-building and noninterpretation of negative transference -- may represent treatment techniques applicable with different types of borderline patients. Those who are symbiotic in their tie with the therapist and rigidly defend against it may not be amenable to classical transference interpretation, while those who alternate between extremes of distance and closeness may be able to withstand the rigors of transference analysis. Examination of this brief review illustrates the controversy among practitioners of psychoanalysis and psychodynamic psychotherapy about the relative roles of support versus interpretation of negative affects toward the therapist as a method to improve the therapeutic alliance. Luborsky, and to a certain extent Masterson, suggests that it is the supportive aspects that are most necessary to 43 building the alliance in fragile patients and that negative factors be best left to the latter stages of treatment. In other words, only a stable alliance forms the groundwork for effective interpretations. Horwitz maintains that type of severe pathology may influence selection of supportive versus interpretive technique. Kernberg, on the other hand, maintains that interpretation of negative transference is crucial to the development of the therapeutic alliance even in severely disturbed patients and support is only a necessary adjunct. Thus Kernberg advocates a method more along the lines of psychoanalytic technique for patients with neurotic symptoms and "character neuroses." However, it is of note that all of these authors do maintain that the therapeutic alliance contains aspects of transference, so that working with alliance of necessity involves working with transference. ns eren e nte re at and s t era ut c Proces In classical psychoanalytic theory, interpretation is the mechanism through which therapeutic change takes place. The most powerful interpretation is the transference interpretation, in which behavior and emotions experienced by the patient toward the analyst are understood and interpreted as a re-enactment of unresolved conflicts in the patient's past, most importantly, to the patient's parental figures. All analytic theorists, including such non- 44 Freudian analysts as Kohut (1977) recognize the central importance of interpretation in producing change. There has been some research on the effects of interpretations on the therapeutic process. Garduk and Haggard (1972) found that psychoanalytic interpretations (that is, an explanation of relationships between conscious and unconscious phenomena) increased affect expression, production of transference-based material, understanding and insight, and defensive and oppositional associations. Interpretations decreased production of conscious (secondary process) material. Luborsky, Bachrach, Graff, Pulver, and Christoph (1979) examined the immediate effects of transference interpretations and outcome. In their sampling from the analyses of three patients, who varied in severity of psychological problems, they found that reactions to transference interpretations varied with level of pathology. Patient A consistently responded to interpretations with an increase in resistance. Patient B reacted with increased involvement with the analyst, increased transference, and a gradual decrease in resistance. Patient C, the most healthy of the three, reacted with increased affect, involvement, understanding, and transference. Resistance remained approximately the same. Patients B and C had positive outcomes, the latter the most positive. In their discussion, the authors noted that Patient A showed only negative helping alliance signs, while Patients B and C 45 showed both positive and negative helping alliance signs, implying that patients benefit from transference interpretations only in the presence of a positive helping relationship. The Menninger Psychotherapy Project (Kernberg et al., 1972) concluded that initial ego strength and the related quality of interpersonal relationships modified the relationship between supportiveness and outcome of psychotherapy in nonpsychotic patients. Patients with initial high scores on both factors benefited from both supportive psychotherapy and psychoanalysis, with psychoanalysis achieving the most benefit. Patients with low initial scores (many of these patients would be considered to have psychopathology in the borderline personality spectrum) benefited most from a supportive- expressive approach with a focus on the transference with use of hospitalization if necessary to control acting out. They benefited less from both supportive psychotherapy and psychoanalysis. The investigators concluded that interpretation of the transference, particularly the negative transference as it impedes the psychotherapeutic relationship, was necessary to strengthen the therapeutic bond, and support necessary to control transference acting- out and stem regression (that is, formation of a psychotic transference). Piper, Debbane, DeCarufel, and Bienvenu (1987) found that total therapist interpretive activity was related to therapist ratings of positive outcome. 46 Malan (1976), in his classic study of brief psychotherapy, found that the proportion of therapist- parent/sibling (T/P) interpretations was associated with more successful dynamic outcomes. The proportion of T/P interpretations increased up to a point, and then leveled off. In sessions 1-5, T/P consists of only 4.35% of the interpretations made. This percentage peaks in the block of 16-20 sessions, where T/P interpretations peak at 9.27%. The prOportion steadily decreases thereafter. It is of note that the total number of interpretations decreases steadily as therapy progresses. Marziali (1984b) replicated Malan's study using the audiotaped sessions of 25 patients treated in brief dynamic therapy (20 sessions). Patients were selected using Malan's criteria for suitability for brief dynamic therapy. Three months after termination, follow-up interviews were conducted by another clinician who rated the patient on a revised version of Malan's Global Outcome Scale. At that time, patients completed the Derogatis Behavior Symptom Index. Presence of T/P and T/P/O (therapist-patient-other) interpretations were significantly associated with global dynamic outcome. There was overlap between measures of dynamic change and symptomatic change but dynamic change clearly measured improvement separate from the symptomatic dimension. Other investigators have taken a different approach to the study of transference interpretation. One of these is 47 the Core Conflictual Relationship Theme (CCRT) (Luborsky, Crits-Christoph, Mintz, & Auerbach 1988) defined as a repeating "relationship episode" which contains elements of: a) a wish, need, or intention toward another person; b) the other person's response; and c) the response of the self. This approach to the transference is succinctly characterized as a recognition that "the transference is the reactivation in the here-and-now of internalized object relations...unconscious intrapsychic conflicts always involve the relationship of an aspect of the self relating to a significant object" (Kernberg in Luborsky, Crits- Christoph, Mintz, & Auerbach, 1988, p. xiii). In a pilot study relating accuracy of interpretation according to the CCRT and patient resistance using the data from the Penn Psychotherapy Project, it was found that accurate interpretation decreases resistance in patients with high outcome and increases it in patients with poorer outcomes (Grits-Christoph, Schuller, & Connolly, 1988). When observing general levels of resistance, therapies with high accuracy on the Wish and Response from Other aspects of the CCRT produce higher resistance in the form of vagueness and doubting, where those with high accuracy in the Response from Self aspect were associated with decreased vagueness and doubting following interpretations. The authors interpret these findings in the following manner: patients confronted with their wishes and responses from others are unsettled and therefore think in more vague and doubting 48 ways; accurate interpretation of affective states make the patient feel understood and therefore more comfortable. In using the same data, it was found that the accuracy of the therapist in interpreting the Wish and Response from Others was significantly predictive of outcome. This finding, surprisingly, was not related to or interactive with the strength of the helping alliance, nor was it related to level of errors in technique. The authors state that caution should be taken with these findings, however. They noted that there was a significant lack of both poor alliances and severely disturbed patients in their sample. They also noted that the general accuracy of interventions was low and reliability of rating errors in technique was also low. On the other hand, the relationship between accuracy of interpretation and outcome appeared as early as the fifth session. This finding was similar to that of Silberschatz, Fretter, and Curtis (1986) who found a significant correlation between interpretations consistent with a patient's unconscious "plan" (cognitive schema) and psychotherapy outcome. However, the views of the Luborsky group and the Silberschatz group differ somewhat to the nature of the phenomenon they have been studying. While Silberschatz et a1. differentiate their "plan" concept from transference, the Luborsky group sees, as mentioned above, the CCRT as a manifestation of transference. They cite as evidence their findings, in addition to the accuracy factor, that reductions in the pervasiveness of relationship 49 conflicts as measured by the CCRT are related to outcome. They also relate the CCRT to Freud's 10 concepts of transference: the presence of one main pattern, the specificity and distinctiveness for each person, its prominence in erotic relationships, that part of the pattern is unconscious, is consistent over time, changes slightly over time, becomes evident in relation to the therapist, its resemblance to early relationships with parental objects, the resemblance of in-treatment patterns to out-of-treatment patterns of behavior, and the presence of positive and negative components to the pattern. erv tion a d P o h 0 Although it was originally believed that supportive techniques exerted only a stabilizing function, primarily through the use of suppression, recent theoretical developments and research suggest that supportive techniques also have an insight-producing and a structure-building effect. Developmental psychoanalytic theorists such as Blanck and Blanck (1974) maintain that supportive techniques such as ego support, where the therapist helps the patient engage his/her own strengths (that is, their highest functional level), build autonomous ego functions in the patient. They further point out such techniques are increasingly inappropriate for neurotic patients, as they can perform these functions themselves. Deficit theorists such as Kohut (1977), would also concur, maintaining that 50 so-called "supportive" interventions were essential in resolving earlier developmental failure and enabling the patient to then approach oedipal conflicts. Kernberg (1985), who advocates supportive psychotherapy only when psychoanalysis or expressive psychotherapy is unfeasible or in character-disordered patients with antisocial features, notes that supportive psychotherapy does promote psychotheraeputic change, even though the material for psychotherapeutic work is conscious and preconscious. A recent review of technique in supportive psychotherapy notes: Even the most supportive treatment has the objectives of increasing the patient's awareness of the relationship between his behavior and the responses of other people, his ability to sort out cause-and-effect relationships, and his appreciation on a manifest level of the connection between past and current patterns." (Winston, Pinsker, & McCullough, 1986, p. 113) Two empirical studies have addressed the relationship among pretreatment factors, therapist interventions, and outcome. Horowitz et al. (1984) found, in his sample of patients undergoing psychotherapy for bereavement reactions, that therapist interpretive work focusing on meanings of the stressful event, and transference interpretations benefited patients with high pretreatment motivation and high developmental level of self-concept and did not benefit those scoring low on these measures. Emphasis on 51 stabilization of the self-concept, a supportive intervention, benefited patients in the opposite manner. Jones, Cumming, and Horowitz (1988), used the Psychotherapy Process Q-Sort (composed of items related to psychotherapy process) to study the same sample of patients mentioned above. For patients with high pretreatment disturbance levels (high symptom level, high experience of stress, poor psychological functioning) benefited most from interventions that were of a supportive nature. The patients expressed dependency wishes quite readily in therapy and appeared to desire a restorative relationship with the therapist. Patients relatively low in pretreatment disturbance benefited most from treatment which focused on the patient's affective experience and relationship to the therapist. The expression of dependency wishes vis-a-vis the therapist was much more conflicted. Only the item "Patient achieves new understanding or insight" was predictive of outcome regardless of pretreatment disturbance. The authors note that "successful therapy with more severely disturbed patients had a more external focus, one aimed away from emotional conflicts and personal meanings of experience and toward a more reality-oriented construction of the patient's problem. The constellation of findings clearly represents an anxiety-suppressive or supportive psychotherapy." (Jones, Cumming & Horowitz, 1988, p. 52). 52 In contrast, this was precisely the focus for the less- disturbed patients. In attempting to explain these data, the authors suggest they show that therapist actions strongly define the character of interactions in psychotherapy. They feel the skillful clinician fosters the therapeutic alliance and eventual outcome by the careful selection and execution of proper psychotherapeutic technique. This application is tailored to the patient's needs, level of psychopathology, external circumstances, as well as to the style of the therapist. In summary, supportive techniques appear to play a role in psychotherapeutic change, particularly with more disturbed patients. How they work is not as clear as for the transference interpretation. Some maintain they suppress anxiety and strengthen reality testing; some that they provide the necessary structure for psychotherapeutic work; and others maintain that they, in and of themselves, build psychic structure and foster psychological development. HYPOTHEBEB The major hypothesis in this study is that therapist interventions significantly affect both alliance and outcome in psychotherapy. Part of the influence of therapist interventions on outcome relates to the impact of the therapist's contribution to the therapeutic alliance. In 1 particular we are interested how the therapist's interventions affect the patient's ability to form a therapeutic alliance. Although the therapeutic alliance may be thought of as a joint endeavor, patient and therapist contribute to it differently. The patient's contribution to the alliance is measured directly -- the patient's capacity to work in treatment, a patient's commitment, and negative and/or problematic attitudes that might directly impede therapeutic work. These have been shown to predict outcome. These ratings of patient alliance will be used in this study. However, as noted in the literature review, studies utilizing clinical raters have not found consistent relationships between therapist alliance and outcome. This may be due in part to the difficulty judges have in rating therapists' attitudes; but it may also be that specific therapist interventions are the most important contribution 53 54 to the success of psychotherapy, especially dynamic psychotherapy. The therapist must convey a sense of warmth, empathy, and serving the patient's interest. But the technical aspect of the therapist's activity -- interventions directed at furthering the work of therapy -- are unique to the therapist and the therapeutic situation. Although warmth, empathy, and understanding can be demonstrated by the therapist's demeanor, they can be demonstrated through accurate interpretations, providing appropriate support and directing self-exploration of salient conflicts in a manner that takes into account both the patient's capacity and inclination to do such work. Much study concerning the role of therapist interventions for furthering the work of therapy and the therapeutic alliance has been theoretical. The notable exceptions have been: Malan's (1976) study demonstrating a relationship between transference interpertations and dynamic outcome, and its replication by Marziali (1984a); Foreman and Marmar (1985) and Flasher (1988) on interventions concerning negative attitudes early in therapy improving alliance and outcome. A replication of these findings is one of the aims of this study. This latter finding, the clarification and discussion of negative attitudes toward treatment, has also been described by Kernberg (see the earlier review of his work). Other writers on the therapeutic relationship, most notably Freud (1912), Zetzel (1956), Greenson (1967), and Luborsky (1984) 55 have written on the importance of support and clarification for patients with poor ego strength (i.e., less ability to work in therapy, less positive alliance). This study will test the hypothesis that supportive interventions will improve the alliance in such patients, and be ineffective for patients who already have a high positive alliance. Included in the study there will be attempts at replication of findings of other studies of alliance and outcome, such as the association of pretreatment symptoms and alliance and association of alliance and outcome. The following hypotheses of the effect of patient factors, therapist interventions on alliance and outcome will be addressed: 1) Patients' level of psychopathology adversely affects their ability to form a therapeutic alliance. 2) Patients who present for treatment with low positive alliance will benefit from supportive interventions -- that is, their alliance will improve with more support. However, those patients who come to therapy demonstrating a capacity to work and a commitment to therapy will find supportive interventions either unhelpful or detrimental to their positive alliance. Therefore, there is an interaction between positive alliance and supportive interventions. 3) The major effect of supportive interventions on outcome is through their interaction with positive alliance. One of these effects is to help the patient hear and understand interventions dealing with the transference. 56 4) Patients presenting for therapy with negative and/or defensive attitudes toward therapy (negative alliance) will benefit from interventions aimed at clarification and exploration of these attitudes. This will be associated with decreased defensive attitudes. 5) Interventions relating to transference (here broadly defined as interventions including the therapist— patient relationship as well as the traditional therapist- parent-other link) will be associated with increased positive alliance and increased outcome. Outcome measures related to dynamic change will be most affected. 6) Interventions relating to transference will be associated with positive outcome in conjunction with the use of other therapist interventions. a) Patients with low positive alliance will benefit from supportive interventions which will be related to the incidence of transference interventions. 7) Expressive interventions (aimed at increased exploration of thoughts and feelings) will positively effect therapeutic outcome. These hypotheses will be tested with three path models. A path analytic approach was chosen for two reasons. Clinical theory is at bottom a theory of causation; path analysis permits an assessment of causality. Also, clinical theory about change is complex, and involves the interaction of many factors; again, path analysis permits the testing of complex models of change. Scales will be analyzed using the 57 multiple groups confirmatory factor analytic technique (Hunter, undated) which will ensure both scale homogeneity and correct for attenuation of correlations due to unreliability of the measures. gata Made; £1: Alliance Changes Early ia Therapy The first of these, shown in Figure 1, presents a model for the relationship among pretherapy patient symptoms, initial patient alliance, therapist interventions and alliance in the third session of therapy. This path model, and all others subsequent to it, displays both the observed variables (the actual items administered, represented here by squares) and the latent variables (the underlying concepts thought to be measured by the observed variables), here illustrated by ellipses). Dark arrows represent causal paths between latent variables. Going from left to right: Initial psychopathology, as demonstrated by the General Severity Index, (or GSI), leads to decreased capacity to work in therapy (Patient Working Capacity-Positive Aspects, or PWC-POS) as demonstrated in previous studies and in other work with this sample. On an experimental basis, we also will link symptomatology to negative alliance (Patient Working Capacity-Negative Aspects, or PWC-NEG). Initial alliance variables are intercorrelated as has been shown in previous work. Initial alliance is also associated with third session alliance. The relationship between initial alliance and third session alliance is affected by therapist interventions. Positive 58 m ocooB ”mosoaom ouo mBnoEo> cotomno i BE: 58.— mmosonb E boom momcono oocozz mam x mozm Boom m ooiomno o5 «Emomfioo mEofi mom: 3. Book. momma—o mBnoEo> unoaB "mosoovm oso meoEo> oo>sombo omlqom Hmom Boom oozombo 05 mfimomEoo mEofi mom: 2 Boom. Boom moo 23m manna—o 33223 ocooB Hobos—om 93 $33.23 ooiombo oEoooso poo .ocoscozooo. 625:: 5* 3.62 5on— ,/, . m 953m / j, m<fi§t§o cm.— a»; Clue flanged n—flth—S 00:06::00 NWG “ED” one: 11! 858.5! H Esau—=3 H on 0.55..— asoauom v n N a q d u q H 1 on.— 1 n.— 1 n».— N A453 coo—33083:. 25.8235 .32: p.598 B2535 oosovcsoo N09 "3:. or. ensur— cso: IXI 95.55! H nus—Ewe! H anon-now ‘ 1') N c-a o.— 725 863.823:— 3:08.23...— 91 Expressive Interventions scale. As shown in Figures 5a-d, Segment 4 (Late Session) is significantly higher than Segment 1 (Session 1) on the Transference Interventions subscale; Segment 4 is significantly higher than Segment 1 on the Confrontation and/or Clarification of Defensive Attitudes subscale; and Segment 4 is significantly higher than Segments 1, 2, and 3 on the Supportive Interventions subscale. i o A As with the CALPAS, a subsample of segments (45, or 25%) were rated on the TAS by both judges. Reliability of a single rating was estimated using the intraclass correlation coefficient recommended by Fliess (1986). and for the mean rating (Shrout 8 Fliess, 1979). These are listed in Table 5. As with the CALPAS, TAS many subscale ratings were quite low, but also more variable. The problem with low subscale variability that attenuated CALPAS reliability ratings was present here as well. Also, particularly for the first hour, therapists made few interventions, lowering internal consistency, and therefore also reliability (Lahey, Downey, 8 Saal, 1983). As with the CALPAS, ratings were performed over an extended period of time. Reliabilities for single sessions varied widely as well, ranging from 0.00 for Supportive Interventions and 0.81 for Transference Interventions. The 92 Table 5 TAS Reliability Intraclass Correlation Coefficients Single Mean Scale Rating Rating All Segments (N=45) Transference Interventions .47 .65 Contront/Clarify Defensive Attitudes .28 .46 Supportive Interventions .47 .67 Expressive Interventions .34 .69 surprising result for Supportive Interventions is misleading and is due to the extremely low mean and standard deviation of this scale in the first session where this reliability was found. Examination of the raw data revealed that raters agreed on 4 of 15 segments used to estimate reliability for this session; no ratings were more than 0.60 apart. For reasons similar to the CALPAS, a Finn's r was calculated for each of the TAS subscales. These single- session ratings are shown in Table 6a. These reliabilities were a considerable improvement over the ICC estimates of reliability as well, although a bit lower, particularly for Expressive Interventions. They fall within the range of acceptable reliability suggested by Kraemer (1981). 93 Table 6 - TAS Reliability Finn's r Calculations Single Scale Rating All Segments (N=45) Transference Interventions .90 Contront/Clarify Defensive Attitudes .72 Supportive Interventions .84 Expressive Interventions .64 Qataoaa geasurea SCQ-90. It was discovered that a small amount of data (165 out of 8280 data points, approx. 2%) were missing from the study sample. To complete the data set, the mean of the subscale scores for each item was substituted for the missing values. Each subject had at least some original data, and many subjects had only a few data points missing. As the SCL-90 has been frequently used and has demonstrated reliability and validity, no formal tests of the measurement model were made here. The General Severity Index (GSI), used in this study, had a pretherapy mean of 1.33 (SD 0.61) and a mean of 0.84 (SD 0.47) at termination. A paired-samples t-test showed these means to be significantly different (t=4.96, 45 df, p < .001, r=.27). A standardized 94 gain score was calculated for the GSI and will be used in data analysis. 29at:Ih2rs22_The:snist.929§tiennsirgi A factor analysis procedure was used to obtain psychometric characteristics of the questions selected to estimate dynamic outcome from the Post-Therapy Therapist Questionnaire. A principal components analysis with oblique rotation was performed. Five factors emerged, including a factor which contained most of the dynamic outcome factors. One original question ("Degree of basic personality change") loaded highly on several factors, and was therefore eliminated. A confirmatory oblique multiple groups factor analysis was then performed on the data, which confirmed the model. As the therapist estimated the degree of improvement with the patient's baseline, a standardized gain score was calculated as well. The mean score on the dynamic change items at the beginning of treatment was 4.61 (SD 1.12) and at termination was 5.57 (SD 1.37). These means were significantly different at the less than .001 level (t=5.28, 45 df, r=.53) by a paired-sample t-test. Table 7 shows the items used to calculate dynamic change. 95 Table 7 Post-Therapy Therapist Questionnaire Final List of Dynamic Outcome Items 1) Ego strength (before and after treatment). 2) Capacity for insight (before and after treatment). 3) Adjustment (before and after treatment). 4) Motivation for psychotherapy (before and after treatment). 5) Prognosis (before and after treatment). t ° ce n rve t o 1119829118 The first path model is a test of the following hypotheses: (1) Patients' level of psychopathology adversely affects their ability to form an alliance; (2) Patients who present for treatment with low positive alliance will benefit from supportive interventions —- that is, their (positive) alliance will improve with more support. However, those patients who come to therapy demonstrating a capacity to work and a commitment to therapy will find supportive interventions either unhelpful or detrimental to their positive alliance. Therefore, there is an interaction of alliance and supportive interventions. (3) Patients presenting for therapy with negative and/or defensive attitudes toward therapy (negative alliance) will benefit from interventions aimed at clarification and confrontation of defensive attitudes. This will be associated with decreased defensive attitudes. 96 Because of the changes necessitated by the revision of the CALPAS scales, the path model was revised to fit the new data. These revisions can be seen in the revised path model shown in Figure 6. For reference, the measurement model is included in the figure. To obtain a correlation matrix corrected for attenuation for the path analysis, a confirmatory oblique multiple groups factor analysis was performed on the scales used in the first path model: the General Severity Index (GSI) of the SCL-90; from the CALPAS, Patient Positive Alliance (POS) for sessions 1 and 3; Patient Working Capacity-Negative Aspects (NEG) for Sessions 1 and 3; Patient Commitment for Sessions 1 and 3; from the TAS, Transference (TRA) for Session 1; Supportive Interventions (SUP) for Session 1; Clarification and/or Confrontation of Defensive Attitudes (DEF) for Sessionl; and an interaction term composed of POS and SUP. This analysis revealed item 3 of the Transference Interventions scale of the TAS (Linked reactions toward therapist to other important figures) had a negative loading on its own factor for this subset of the data; it was eliminated for this analysis. The factor- factor intercorrelation matrix corrected for attenuation for Path Model 1 is shown in Table 8. Table 9 shows the matrix of correlations among variables estimated by the path equations and the resultant error matrix. Figure 6 shows the path model with path 97 Table 8 Correlation Matrix for Path Model #1 GSI P01 NEl TRA DEF SUP POP P02 NE2 GSI 1.00 .20 .30 -.06 -.03 .35 .21 .32 -.02 P01 .20 1.00 -.44 .10 .24 .04 .64 -.02 -.10 NEl .30 -.44 1.00 -.08 .03 .35 -.03 .03 .03 TRA -.06 .10 -.08 1.00 .31 .28 .24 .22 -.08 DEF -.03 .24 .03 .31 1.00 .24 .51 .18 -.20 SUP .35 .04 .35 .28 .24 1.00 -.02 .69 .18 POP .21 .64 -.03 .24 .51 .69 1.00 -.06 -.18 P02 .32 -.02 .03 .22 .18 .18 -.06 1.00 -.20 NE2 -.02 -.10 .03 -.08 -.20 -.02 -.18 -.20 1.00 GSI: General Severity Index P01: Patient Positive Alliance, Session 1 NE1: Patient Working Capacity, Negative Aspects, Session 1 TRA: Transference Interventions DEF: Clarification and/or Confrontation of Defensive Attitudes SUP: Supportive Interventions POP: Interaction of Positive Alliance and Supportive Interventions P02: Patient Positive Alliance, Session 3 NE2: Patient Working Capacity, Negative Aspects, Session 3 m .20. Due to this chi-square, the null hypothesis that the two matrices are the same can not be rejected, confirming the model is a good fit for the data. The significance of the individual paths was tested using the computer program REGRESS (Gerbing 8 Hunter, 1988) which calculates the significance of path coefficients by testing the beta weights generated by the path equations that have been corrected for attenuation caused by error of measurement. A significance level of 0.10 was employed for the analysis of beta weights to minimize the possibility of a Type II error. Due to the experimental nature of this work, and low power due to sample size and low reliability on some of the scales the probability of a Type II error is elevated, necessitating this adjustment. The results of these tests are displayed in Table 10. The results suggested by the significance tests listed in Table 10 are the following: 1) Pretreatment symptomatology is predictive of alliance, but only of Positive Alliance, and in the opposite direction (that is, high pretreatment symptomatology is 101 Table 10 Significance Testing of Path Coefficients for Path Model #1 (N=46) De .e,. '- -e. '. ta ’0- '21 1.. :,_- e. Path: GSI + P01 + POP = P02 Multiple R: .41 Shrunken R: .34 R Squared: .17 Standard Error Variable Beta of Beta t Significance GSI .37 .15 2.41 p < .01 P01 .12 .21 0.52 ns POP -.33 .20 -1.64 p = .06 Path: GSI + P01 + SUP + POP = P02 Multiple R: .48 Shrunken R: .40 R Squared: .23 Standard Error Variable Beta of Beta t Significance GSI .14 .25 0.54 ns P01 .56 .60 0.93 ns SUP .80 .74 1.07 ns POP -.99 .88 -1.12 ns Path: GSI + P01 = P02 Multiple R: .34 Shrunken R: .26 R Squared: .11 Standard Error Variable Beta of Beta t Significance GSI .33 .15 2.25 p < .01 POl -.09 .16 —0.53 ns 102 Table 10 (cont.) Significance Testing of Path Coefficients for Path Model #1 (N=46) O. 0 p. L. ...: O I Path: GSI + P01 + TRA = P02 Multiple R: .42 Shrunken R: .34 R Squared: .18 Standard Error Variable Beta of Beta t Significance GSI .36 .15 2.32 p < .01 P01 .12 .17 0.70 ns TRA .25 .19 1.31 p = .10 Path: P01 == P02 Pearson r (corrected for attenuation) = .02, ns GSI: General Severity Index PO1: Patient Positive Alliance, Session 1 TRA: Transference Interventions SUP: Supportive Interventions POP: Interaction of Positive Alliance and Supportive Interventions P02: Patient Positive Alliance, Session 3 103 Table 10 (cont.) Significance Testing of Path Coefficients for Path Model #1 (N846) - ~72~ 1 '1 z: '; 2 .- .in- :-—c ‘ .;-; v WWI Path: GSI + P01 + NE1 = NE2 Multiple R: .10, ns All beta weights not significant at the 0.10 level, one- tailed test. Path: GSI + NE1 = NE2 Multiple R: .04, ns All beta weights not significant at the 0.10 level, one- tailed test. Path: GSI + P01 + DEF = NE2 Multiple R: .20 ns All beta weights not significant at the 0.10 level, one- tailed test. Path: NE1 == NE2 Pearson r (corrected for attenuation) = .03, ns GSI: General Severity Index P01: Patient Positive Alliance, Session 1 NE1: Patient Working Capacity, Negative Aspects, Session 1 DEF: Clarification and/or Confrontation of Defensive Attitudes P02: Patient Positive Alliance, Session 3 NE2: Patient Working Capacity, Negative Aspects, Session 3 104 associated with greater initial positive alliance). Thus Hypothesis 1, that symptoms adversely affect positive therapeutic alliance, is rejected. While the zero-order correlation was significant beteween GSI and PWC-NEG, no relationship was demonstrated in any of the path equations for PWC-NEG . 2) The interaction between positive alliance and supportive interventions in the first session is not a significant predictor of alliance in the third session; Hypothesis 2 is also rejected. 3) First session alliance is not predictive of third session alliance. 4) Alliance does not significantly predict therapist interventions. All paths suggesting such a relationship were not significant. 5) Therapist interventions have some predictive power of alliance. Transference interventions were positively associated with increased third session positive alliance. However, the hypothesis that therapist's confronting and/or clarifying defensive attitudes would be associated with lower negative alliance in the third session (Hypothesis 3) was disconfirmed. a c s O c This model is a test of the following hypotheses: 3) The major effect of supportive interventions on outcome is through their interaction with positive alliance. 105 One of these effects is to help the patient hear and understand interventions dealing with the transference. 5) Interventions relating to transference (here broadly defined as interventions including the therapist- patient relationship as well as the traditional therapist- parent-other link) are associated with increased positive alliance and increased outcome. Outcome measures related to dynamic change will be most affected. 6) Interventions relating to transference will be associated with positive outcome in conjunction with the use of other therapist interventions. a) Patients with low positive alliance will benefit from supportive interventions which will be related to the incidence of transference interventions. b) Patients with high levels of negative alliance will benefit from interventions addressing defensive attitudes which will be related to the incidence of transference interventions. c) Transference interventions are related to positive outcome. 7) Expressive interventions (aimed at increased exploration of thoughts and feelings) will positively effect therapeutic outcome. The analysis of these hypotheses set out in the Hypotheses section was modified somewhat. Due to the nonsignificant correlation of therapist-rated dynamic outcome and patient-rated symptom outcome (r=.09, ns) 106 meoCo> cozomno BS wcwmomaoo mEofi mom: m< Bnoe oEoooso Baocma coo .mcofisotoog .oosozz «film _Booz Boom m<.~. Ema Boom 70/ N. 25me m .10. This chi-square suggests that the model is is a good fit to the data. Figure 7 shows the path model with path coefficients. (Note again that Supportive Interventions is included in this path analysis for statistical purposes, and is not part of the theoretical model.) Significance testing of the path coefficients was performed using the REGRESS program. The results can be seen in Table 13. Significance testing could not be performed on the path POS - SUP - POP - DYN due to the very high intercorrelation of the SUP and POP scales; it is not included in the table. Based on the results displayed in this table, the following conclusions can be drawn concerning the relationship between alliance, therapist interventions, and dynamic outcome: 1) Transference interpretations are clearly predictive of positive dynamic outcome, confirming Hypothesis 5. 2) Hypothesis 6a, which posits that supportive interventions interacting with positive alliance will 110 Table 12 Matrices of Reproduced Correlations and Errors in Prediction Path Model #2A, Dynamic Outcome Reproduced correlations: POS NEG POP DEF SUP EXP TRA DYN POS 1.00 -.32 .29 -.06 -.14 -.30 .29 .16 NEG -.32 1.00 -.09 .20 .04 .10 -.10 -.28 POP .29 -.09 1.00 -.02 1.00 -.09 -.16 -.13 DEF -.06 .20 -.02 1.00 .01 .02 -.08 -.10 SUP -.14 .04 1.00 .01 1.00 .04 -.30 -.21 EXP -.30 .10 -.09 .02 .04 1.00 -.09 -.36 TRA .29 -.10 -.16 -.08 -.32 -.09 1.00 .73 DYN .16 -.28 -.13 -.10 -.23 -.36 .73 1.00 Errors: (Actual - reproduced) POS NEG POP DEF SUP EXP TRA DYN PCS .00 .00 .00 .02 .00 .00 .00 .00 NEG .00 .00 .24 .00 .27 -.15 .03 .08 POP .00 .24 .00 -.28 .00 -.59 .02 .23 DEF .02 .00 -.28 .00 -.34 .25 .08 .02 SUP .00 .27 .00 -.34 .00 -.71 .11 .26 EXP .00 -.15 -.59 .25 -.71 .00 .23 .20 TRA .00 .03 .02 .08 .13 .23 .00 -.09 DYN .00 .08 .23 .02 .26 .20 -.09 .00 Sum of squared errors in the lower triangle = 1.52 Chi-Square = 21.50, 15 df, p > .10 111 Table 13 Significance Testing of Path Coefficients for Path Model fZA Path: P08 + EXP = DYN Multiple R: .20, us All beta weights not significant at the 0.10 level, one- tailed test. Path: P08 + TRA DYN Multiple R: .64 Shrunken R: .62 R Squared: .41 Standard Error Variable Beta of Beta t Significance PCS -.03 .18 0.15 ns TRA .65 .26 2.50 p < .01 Path: P08 + POP + TRA = DYN Multiple R: .67 Shrunken R: .65 R Squared: .45 Standard Error Variable Beta of Beta t Significance P08 -.11 .20 0.54 ns POP .23 .19 1.22 p = .11 TRA .70 .17 4.05 p < .001 112 Table 13 (cont.) Significance Testing of Path Coefficients for Path Model #2A Path: POS == DYN Pearson r = .16, t = 1.07, p =.15 Path: NEG + DEF + TRA = DYN Multiple R: .67 Shrunken R: .65 R Squared: .45 Standard Error Variable Beta of Beta t Significance NEG -.15 .19 -0.80 ns DEF -.05 .18 -0.27 ns TRA .63 .14 4.47 p < .001 Path: NEG == DYN Pearson r = -.20, t = -1.25, p =.10 POS: Patient Positive Alliance NEG: Patient Working Capacity, Negative Aspects POP: Interaction of Positive Alliance and Supportive Interventions DEF: Clarification and/or Confrontation of Defensive Attitudes EXP: Expressive Interventions TRA: Transference Interventions SUP: Supportive Interventions DYN: Dynamic Outcome 113 positively link with transference-based interpretations, could not be tested and therefore could not be confirmed. 3) Interventions addressing defensive attidudes are not related to negative alliance, nor are they associated with increased dynamic outcome; Hypothesis 6a is rejected. 4) Expressive interventions do not significantly predict dynamic outcome; Hypothesis 7 is rejected. 5) Negative, but not positive alliance predicts dynamic outcome. As stated before, the factor-factor intercorrelation matrix for Path Model #2A is displayed in Table 14. Table 15 shows the matrix of correlations among variables estimated by the path equations, the resultant error matrix, the sum of squared errors in the matrix, and the chi-square for the model. (Supportive Interventions is included in this path analysis for statistical purposes; it is not part of the theoretical model.) The chi-square for this model was 19.61, 15 df, p > .10. This chi-square suggests that the model is a good fit to the data. Figure 8 shows the path model with path coefficients. Significance testing of the path coefficients was performed using the REGRESS program. The results can be seen in Table 16. Again, significance testing could not be performed on the path POS - SUP - POP - GSI due to the very high intercorrelation of the SUP and POP scales; it is not included in the table. 114 Table 14 Correlation Matrix for Path Model #ZB, Symptom Outcome POS NEG POP DEF SUP EXP TRA GSI POS 1.00 -.32 .29 -.04 -.14 -.30 .29 .16 NEG -.32 1.00 .15 .20 .31 -.05 -.07 -.11 POP .29 .15 1.00 -.30 1.00 -.68 -.14 .12 DEF -.04 .20 -.30 1.00 -.33 .27 .00 -.15 SUP -.14 .31 1.00 -.33 1.00 -.67 -.19 .02 EXP -.30 -.05 -.68 .27 -.67 1.00 .14 .07 TRA .29 -.07 -.14 .00 -.19 .14 1.00 -.19 GSI .16 -.11 .12 -.15 .02 .07 -.19 1.00 POS: Patient Positive Alliance NEG: Patient Working Capacity, Negative Aspects POP: Interaction of Positive Alliance and Supportive Interventions DEF: Clarification and/or Confrontation of Defensive Attitudes EXP: Expressive Interventions TRA: Transference Interventions GSI: Symptom Outcome 115 mBnoCo> oo>somno B: mammomaoo mEoB mom: m< Boom. m5. Ema Boom .0! omlqom Hmo Boom i . a v oEoobso Eoomamm .mcofisozoos .ooso2:< Ammo“ Eco: noon m 2de 9:38 8225 28m i cm. omz mmf mo Tl , moi om.l ow mxm i . m5“. mDmxmom mod. m<.r m oo>somno B: uEmoQEoo mEofi mom: m< Boob oEoobflo BEoEmQ coo boommsm BB6: zoom oonlomom a 25mg ,_ m: 2.35 ES 28m 8226 £68 . v 5:. «4m. a e i . A ohob m5. XPI «mom m<fi mQ Boom an. Boom mxm Boom Boom mom Boom : : 11111111 Table 19 Significance Test Predicting Dynami Va Path: NEG + SUP Multiple R: .69 R Squared: .48 Variable Beta NEG -.22 SUP .23 TRA .67 Path: NEG + DEF Multiple R: .20, All beta weights tailed test. POS: Patient Pos 125 ing of Path Coefficients for Post Hoc Model c Outcome From Supportive Interventions e' o e + TRA = DYN Shrunken R: .66 Standard Error of Beta t Significance .21 -1.08 ns .23 1.00 ns .16 4.22 p < .001 = DYN ns not significant at the 0.10 level, one- itive Alliance NEG: Patient Working Capacity, Negative Aspects DEF: Clarificati Attitudes SUP: Supportive EXP: Expressive TRA: Transferenc DYN: Dynamic Out on and/or Confrontation of Defensive Interventions Interventions e Interventions come 126 Table 20 Correlation Matrix for Post Hoc Path Model predicting Symptom Outcome from Supportive Interventions POS NEG DEF SUP EXP TRA GSI POS 1.00 -.32 -.04 -.14 -.30 .29 .16 NEG -.32 1.00 .20 .31 -.05 -.07 -.11 DEF -.04 .20 1.00 -.33 .27 .00 -.15 SUP -.14 .31 -.33 1.00 -.67 -.19 .02 EXP -.30 -.05 .27 -.67 1.00 .14 .07 TRA .29 -.07 .00 -.19 .14 1.00 -.19 GSI .16 -.11 -.15 .02 .07 -.19 1.00 POS: Patient Positive Alliance NEG: Patient Working Capacity, Negative Aspects DEF: Clarification/Confrontation of Defensive Attitudes SUP: Supportive Interventions EXP: Expressive Interventions TRA: Transference Interventions GSI: Symptom Outcome 127 BmU om Iqom Hmm Boom meoEo> oozomno B3 wswmoaaoo mEofi mom: m< Boob oSoooso Eoomobmm moo boommzm BB6: noon oomlomom m4; Ema Boom S 658$ 255 8225 28m m4; 3:. Boom E . 9 mm. m N . II 6/ S. on. cm. 9 1 t 82 on: moo on: m<._. mod. mDm muoucH m>HuumquO «ONO ocOHucm>umucH m>Huu0OO5O «OOO umosuHqu m>HmcoO0O no OOOHHOOHuHuuHo\ocoHvoucouucoo «OOO ocoHucm>umucH vouonOloucoumwmcoua «OOH OOH OHl HM kH NM Om MO OM le HNl NHl NNl OHl oo M m OM OH m mH ONO OHl OOH Ol HHl ml Ol ONl M oo Om OO Nk HO Ml Ol Ol ml kl kl ml ODO HM Ol OOH Oo HH OH oN H OH NHl kl OHl kl VO OO Mk Ok o ON km Ono kH HHl mo OOH MH OH H oHl O kHl m mHl M OM Ho Oo mN oo MO Mk dam NM ml HH MH NH MH OM ON NHl Ol Ol HHl Ol MN Hl Ml NH OH M N OH O Om Ol OH OH MH «kN oo mH OHl OHl M Ol Nl MN Hl M HN N OH MH oH m lhb ONl ON H OM oo «mm mm OHl NHl OHl kHl mHl kN k O ON ml cl HH mH O OM M H OHl ON mH OH O kl HH Nl m o O Ol Ol k Ol Ml MHl HH le oo OH O NHl OHl OHl kl «OH oH mo oN mM k OH OH NH Nl H MH O O HNl Om NHl kHl Ol OHl NHl HH oH «NM OM Om Mo Nl OHl Ol Ol kl Ol OHl NH O NHl OO kl m Ol M OHl Nl mo OM «oo oo mm OHl Ml Nl Ol N Ml OH OH O NNl Nk OHl OHl HHl Ol kHl m oN Om oo «Nm km mHl OHl Ol OHl HHl OHl HHl MH m OHl Hm. kl M Ol Nl mHl o mM Mo mm km cmO O OHl Ol ol o Hl M O oo Ml oO OM MN MN kN O k Nl OHl mHl O Ho MM HM mk o OH HM O M Ol OO Ho Hl Hl k Ol OH OHl Ml OHl OHl MM coo Ok OM N NN Om m O m Ol Mk oo Ml M O Ol OH Ol Nl Ol Ol HM Ok «om mo N OH Hm k O OM ml Ok mN NH HN on. k NH Ol Ol -MH1 ol mk OM mo «mm M oH mN o O OH kl o Oo OH N ml Ol Nl kl N HHl o o N N M MN ON OM N a m kl ON MO M OH cl Ml H Ol Ml OHl Hl OH NN OH OH ON «co ko M O mH ml km Mk N MH HH MHl MH OHl OH HHl M HM Om Hm ON OM ko «Mm H 9 OOO MOO NOO HOO OH OH mH HH O NH OH MH O O m k v N M H .vauze wee uouoom EmuH Ode aflquumucH Od QHQOB 159 Table A7 TAS Items Discarded and Retained from Pactor Analyses WW 1) Therapist suggests meanings of others' behavior. (SUP) WW 1) Discussed process of patient avoiding material and/or feelings in relation to the therapist. (DE?) 2) Discussed content and meaning of material and/or feelings in relation to the therapist. (DEF) 3) Conveyed confidence of favorable therapy outcome for the patient. (SUP) 4) Patient's feelings and perceptions are linked to situations and feelings from the past. (EXP) mas-2mg; WM). 1) Discussed patient's reaction to therapist. 2) Linked reactions toward therapist to parental figures. 3) Linked reactions toward therapist to other important figures. WW 1) Discussed process of patient avoiding material or feelings. 2) Discussed content and meaning of material and feelings patient was avoiding. W1 1) Therapist gives explicit reassurance. 2) Expressed liking or positive regard for the patient. 3) Therapist gives explicit advice or guidance. 4) Acts to strengthen defenses (vs. stimulate insight). WW 1) Encouraged or permitted expression of feelings. 2) Encouraged patient to examine meanings of his/her thoughts, behavior, or feelings. DEF: Confrontations/Clarifications of Defensive Attitudes SUP: Supportive Interventions EXP: Expressive Interventions 160 Table A8 Items for Observed Variables in Revised Path Analyses WW1 1) Patient self-discloses thoughts and feelings. 2) Patient self-observes behaviors. 3) Patient explores own contribution to problems. 4) Patient experiences strong emotions. 5) Patient works actively with therapist's comments. 6) Patient deepens exploration of salient themes. 7) Patient views therapy as important. 8) Patient participates in therapy despite painful moments. ' ‘ ' .- . . b . - A -xl“ t - f w. 1, 2' A. ' '_ ' 1 , L .A ‘i-_i 1,‘ 1) Patient acts in hostile, attacking and critical manner towards therapist. 2) Patient seems mistrustful and suspicious of therapist. 3) Patient engages in power struggle, attempting to control session. 4) Patient defies therapist's efforts to promote self-understanding. 8 re I an O t. on A8 1) Discussed patient's reaction to therapist. 2) Linked reactions toward therapist to parental figures. 3) Linked reactions toward therapist to other important figures. f1 t 1) Discussed process of patient avoiding material or feelings. 2) Discussed content and meaning of material and feelings patient was avoiding. a e AS 1) Therapist gives explicit reassurance. 2) Expressed liking or positive regard for the patient. 3) Therapist gives explicit advice or guidance. 4) Acts to strengthen defenses (vs. stimulate insight). 161 Table A8 (cont.) Items for Observed Variables in Path Analyses WW 1) Encouraged or permitted expression of feelings. 2) Encouraged patient to examine meanings of his/her thoughts, behavior, or feelings. W Interaction of POS and SUP §££11_Q§11_!!!2£211 GSI from SCL-9O 925221! : o - e 1) Ego strength (before and after treatment). 2) Capacity for insight (before and after treatment). 3) Adjustment (before and after treatment). 4) Motivation for psychotherapy (before and after treatment). 5) Prognosis (before and after treatment). Wan-222W Standardized gain score from the GSI pre-and post-therapy LIST OF RBYBRBNCBB REFERENCES Allen, J., Newsom, G., Gabbard, G., & Coyne, L. (1984). Scales to assess the therapeutic alliance from a psychoanalytic perspective. anlletin_2f_the_nenninger M 383-400. Barrett-Lennard, G. T. (1962). Dimensions of therapist response as causal factors in therapeutic change. 2sxcnglegica1_u2n2gzaphsi léi 1-36. Blanck, 6., & Blanck, R. (1974). 3 ° ang_p;ag§1ggy New York: Columbia University Press. Bordin, E. (1979). The generalization of the psychoanalytic concept of the working alliance. Egyghgthgyapyy 16. 252-260. Brenner, C. (1979) Working alliance, therapeutic alliance, and transference. o rn he a Esxcheanalxtic_A§sgciatieni alifiuppiii 137-157- Crits-Christoph, P., Schuller, R., & Connolly, M. B. (1988). ac u o e 's s' ' te on patients' levels of resistance in dynamig psychotherapy, Unpublished manuscript, Department of Psychiatry, University of Pennsylvania. Derogatis, L. R. (1977) - ' 5' Baltimore, MD: Johns Hopkins University School of Medicine. Derogatis, L. R., & Cleary, P. A. (1977). Confirmation of the dimensional structure of the SCL-90: A study in construct validation. l9nrna1_2f_§1inical_2§xchelegxi ;;L 981-989. Derogatis, L. R., Lipman, R. S., Covi, L., & Rickels, K. (1971). Neurotic symptom dimensions as perceived by psychiatrists and patients of various social classes. Archixes_2f_§en_ra1_2sxcnia§rxi 211 454-464- 162 163 Derogatis, L. R., Rickels, K., & Rock, A. F. (1976). The SOL-90 and the MMPI: A step in the validation of a new self-report scale. figitlsn Journal of Psychiatny. lggy 280-289. Dickes, R. (1975). Technical considerations of the therapeutic and working alliances. Ingannanlgnal Jgngnal of Psyghganalyglc Psychothenapy. l. 1-24. Eaton, T. T., Abeles, N., & Gutfreund, M. J. (1988). Therapeutic alliance and outcome: Impact of treatment length and pretreatment symptomatology. angnggnanapyy zfiy 536-542. Eissler, K. R. (1953). The effect of the structure of the ego on psychoanalytic technique. J_nnnal_gfi_gna nne zlgan Psychoanalytic ns sagglatlon, 11 104-143. Evenson, R. C., Holland, R. A., Mehta, S., & Yasin, F. (1980). Factor analysis of the Symptom Checklist-90. Psycnolggical Reponts. 56, 695-699. Fenichel, O. (1941). Pro . New York: Psychoanalytic Quarterly. Finn, R. H. (1970). A note on estimating the reliability of categorical data. W W 3.9... 71-76- Flasher, L. V. (1988). Negative factors in short—term psychotherapy: Focus on therapist interventions. s t ' r cts t t ona §§(2-§), 2779. Fleiss, J. L. (1986). ' axp_;lnannay New York: John Wiley. Foreman, S., & Marmar, C. (1985L Therapist actions that address initially poor alliances in psychotherapy. WM 142.; 922- 926- Freud, S. (1913). On beginning the treatment. In Ina '11.. ‘ ’; he 00 7.: o _ 11!: - °. '1 ° !‘ Signund Ezang. vol, 12- tr. J. Strachey, pp. 123-144. Freud, S. (1937). Transference. In Ina_§;angazg_fi_1glgn_gfi !‘ cup - - '= Cl. ..' . .o s . _7-m ,d e . lg, tr. J. Strachey, pp. 97- -108. Frieswyk, S., Colson, D., & Allen, J. (1984). Conceptualizing the therapeutic alliance from a psychoanalytic perspective. Baygngtnaxapyy 21. 460- 464. 164 Frieswyk, S., Allen, J. Colson, D., Coyne, L., Gabbard, C., Horwitz, L., 8 Newsom, G. (1986). Therapeutic alliance: Its place as a process and outcome variable in dynamic psychotherapy research. Jgnznal_gf 92nsn11ing_ang_slinical_2sxchologxi 55, 32-38. Gabbard, G. 0., Horwitz, L., Frieswyk, S., Allen, J. G., Colson, D. B., Newsom, G., 8 Coyne, L. (1988). The effect of therapist interventions on the therapeutic alliance with borderline patients. Jgnynal_g£_§na Amer12an_2sxchoanalxtic_Association1 311 697-727- Garduk, E. L., 8 Haggard, E. A. (1972). Immediate effects on patients of psychoanalytic interpretations. Psychological_1sssesi 11 1-85- Gaston, L., Marmar, C., Gallagher, D., 8 Thompson, L. W. (1989). Impact of confirming patient expectations of change processes in behavioral, cognitive, and brief dynamic psychotherapy. Egygnggnanapyy 2§(3). 296-302. Gaston, L., Marmar, C., Thompson, L. W., 8 Gallagher, D. (1988) Relationship of pretreatment characteristics to the therapeutic alliance in diverse therapies. gannnal of_Q11nica1_and__2n§ulting_£§22hologxi 561111 483-489- Gaston, L., Marmar, C., 8 Ring, J. M. (1988) Engaging_gna f u t tie t co 1 .t er ctio s ' h ° Unpublished manuscript, University of California at San Francisco. Gerbing, D. W. 8 Hunter, J. E. (1988). IIAN; A statistical o ra IT '3 w t c elatio data in21sQing_mn1tiple_gro_ps_factor_analxsi_1_ [Computer program and manual. ] East Lansing, MI: Michigan State University. Gill, M. M., 8 Muslin, H. L. (1976). Early interpretation of transference. Journal_of_the_3nerican Bsxchoanalxtic__Associationi 211 779-794- Gomes-Schwartz, B. (1978). Effective ingredients in psychotherapy: Prediction of outcome from process variables. na in C ' EfixgthQQXL $61 1023'1035- Greenson, R. R. (1965). The working alliance and the transference neurosis. Esygnoanalytlc Qnantarly. 35, 155-181. Greenson. R- R- (1967)- The_technigse_ang_practice_gf angnganalyalal New York: International Universities Press. 165 Greenson, R. R. (1971). The "real" relationship between the patient and the psychoanalyst. In M. Xanzer (Ed.) Ina u (pp. 213-232). New York: International Universities Press. Hartley, D., 8 Strupp, H. (1983). The therapeutic alliance: Its relationship to outcome in brief psychotherapy. In J. Masling (ed). ' , (pp. 1-35). Hillsdale, NJ: Analytic Press. Hatcher, R., 8 Hansell, J. (1990). gnanapanglg_alllangay Paper presented at the let Annual Meeting of the Society for Psychotherapy Research, Wintergreen, VA. Horowitz, M., Marmar, C., Weiss, D., DeWitt, J., 8 Rosenbaum, R. (1984). Brief psychotherapy of bereavement reactions. s 51‘ 438-448. Horvath, A. O., 8 Symonds, B. D. (1991). Relationship between working alliance and outcome in psychotherapy: A meta-analysis. 19nrna1_9f_Qgsnseling_£§xchologxi §§(g). 139-149. Horwitz, L. (1974). ' ' ' t o s . New York: Jason Aronson. Horwitz, L. (1985). Divergent views on the treatment of borderline patients. u t' e ' e ' 'c 52. 525-545. Hoyt, M. F., Marmar, C. R., Horowitz, M. J., 8 Alvarez, W. F. (1981). The Therapist Action Scale and the Patient Action Scale: Instruments for the assessment of activities during dynamic psychotherapy. Psychotherapxi 131 109-116- Hunter- J- E. (undated). Qp11g2e_msltiple_grosps_factor analxsis1_path_analxsis1 Unpublished manuscript- Michigan State University. Hunter, J. E. (1988). Bfiggfigga [Computer program.) East Lansing, MI: Michigan State University. Hunter, J. E. (1989). ° 8 5 es nggzamy [Computer program.] East Lansing, MI: Michigan State University. Jones. E. E. (1985). u 0 he s t P cess Q:§QI§L Upublished manuscript, University of California at Berkeley. 166 Jones, E. E., Cumming, J. D., 8 Horowitz, M. J. (1988). Another look at the nonspecific hypothesis of therapeutic effectiveness. on n s a d Qlinica1_2sxchologxi 561 48- -55. Jones, E. E., 8 Pulos, S. M. (1987). anghgthaxanyl Unpublished manuscript, University of California at Berkeley. Kernberg. 0- (1985). Sexere_personalitx_disorders1 New Haven, CT: Yale University Press. Kernberg, 0., Burstein, E., Coyne, L., Applebaum, A., Horwitz, L., 8 Voih, H. (1972). Psychotherapy and psychoanalysis: Final report of the Menninger Foundation's Psychotherapy Research Project. Ballagln of_the_Menninger_§linici 161 1-275- Klee. M- (1986)- Esx2h2therapx_9stoome_and_the_course_of the_therspeutic_alliancei Unpublished doctoral dissertation, Michigan State University, East Lansing, MI. Klein, M. (1975). The origins of transference. In R. Langs (Ed-) Beag1ngs_1n_psychoanalytic_technigue (pp- 9-15)- New York: Jason Aronson. Kohut. H. (1977). The_restoration_of_1he_self1 New York: Jason M01180!) . Kraemer, H. C. (1981). Coping strategies in psychiatric clinical research. ' Baxchologxi 121111 309-319- Lahey, M. A., Downey, R. G., 8 Saal, F. E. (1983). Intraclass correlation coefficients: There' s more than meets the eye- Rsxchological_flulletini 231111 586- -595 Luborsky, L. (1976). Helping alliances in psychotherapy. In J. L. Claghorn (Ed.) (pp. 92-116). New York: Brunner/Mazel. Luborsky- L- (1984)- Erinciples_of_psxchoanalxtic paygnggnazapyy New York: Basic Books. Luborsky, L., Bachrach, H., Graff, H., Pulver, S., 8 Cristoph, P. (1979). Preconditions and consequences of transference interpretations. ur f Ne us Menta1_niseasei 1611 391-401- 167 Luborsky, L., Crits-Cristoph, P., Alexander, L., Margolis, M., 8 Cohen, M. (1983). Two helping alliance methods for predicting outcomes of psychotherapy. gannnal_afi Neryous_and_uental_niseasei 1111 480-491. Luborsky, L., Crits-Cristoph, P., McLellan, A., Woody, G., Piper, W., Liberman, B., Imber, S., 8 Pilkonis, P. (1986). Do therapists vary much in their success? Amer12an_l9urnal_of_9rthops¥chiatrxi 661 501-512. Luborsky, L., Crits-Cristoph, P. Mintz, J., 8 Auerbach, A. (1988). Eho_w1ll_benefit_fr2m_psxchotheraexz_ 2redictins_therapeutic_outcomes1 New York: Basic Books. Luborsky, L., McLellan, A. T., Woody, G. E., O'Brien, C. P., 8 Auerbach, A. (1985). Therapist success and its determinants. Arch12es_of_Qeneral_Esxch1atrri 311 471- 481. Malan, D. H. (1976) he t ' s ch er . New York: Plenum. Marmar, C., Gaston, L., Gallagher, D., 8 Thompson, L. W. (1989). Alliance and outcome in late-life depression. ggaznal 9f Nezvgus ang Mennal Qiaaasa, 127(§)- 464-472. Marmar, C., Horowitz, M., Weiss, D., 8 Marziali, E. (1986). The development of the the Theraeputic Alliance Rating System. In L. Greenberg and W. Pinsof (Eds.), Ina 2sxch2therapeut1c_2rocess1__A_research_handpook_ (pp. 367-390). New York: The Guilford Press. Marmar, C., Weiss, D. S., 8 Gaston, L. (1989). Toward the validation of the California Therapeutic Alliance Rating System. angultlng ang Clinical Es ygnglggy, l(l). 46-52. Marziali, E. (1984a). Three viewpoints of the therapeutic alliance: Similarities, differences, and associations with psychotherapy outcome. ggurnal 9f Nenvgna and Mental_nisease1 1121 417-423. Marziali, E. (1984b). Prediction of outcome of brief psychotherapy from therapist interpretive interventions. Archixes_of_§eneral_£sxchiatrxi 111 301- 304. Marziali, E., Marmar, C., 8 Krupnick, J., (1981). Therapeutic alliance scales: Development and relationship to psychotherapy outcome. nnanigan Journal_of_£sxchiatrxi 1131 361-364. 168 Therapeutic Masterson, J. F. (1978). The borderline adult: American_lournal_ef alliance and transference. E§¥£h1§3121.1l§1 437-441- (1990)- Taking Messer, S.B., Tishby, 0., 8 Spillman, A. ch thera : P tie nte t s riousl 0 es s n ' '5 Manuscript submitted for publication. (1979). Measuring Mintz, J., Luborsky, L., 8 Christoph, P. the outcomes of psychotherapy: Findings of the Penn Psychotherapy Project. 52, 319-334. Moras, K., 8 Strupp, H. (1982). Pretherapy interpersonal relations, patients' alliance, and outcome in brief ' 12‘ 405-409. therapy. c 'v ene a 8 Morgan, R., Luborsky, L., Crits-Cristoph, P., Curt1s, H., (1982). Predicting the outcomes of Solomon, J. psychotherapy by the Penn Helping Alliance Rating Method. 32‘ 397-402. & Suh C. S., 8 Strupp, H. H. (1983). The A report on O'Malley, S. S., , Vanderbilt Psychotherapy Process Scale: study. Journal scale development and a process-outcome of_Q2nsu1t1ng_and.§11n12a1_2sxcholos¥1 511111 581-586. , (1985). Piper, W. E., deCarufel, F. L., Szrumelak N. Patient predictors and outcome in short-term ind1v1dual e t 'se se psychotherapy. o e vo s 111, 726-735. Piper, W. E., Debbane, E. G., deCarufel, F. L., 8 Bienvenu, A system for differentiating therapist J. p. (1987). interpretations from other interventions. finlla§1n_a§ th__Menninser_211nici 511611 532- -550. Rogers, C. R. (1957). The necessary and sufficient condit1ons of therapeutic personality change. Journal of Qonsulging Psycnolggy. zly 95-103. ' (1986). Rosenbaum, R.L., Horowitz, M. J., and Wilner, N. Clinician assessments of patient difficulty. Earshotherapxi.211 417-424. ' at'e en e Ryan, E. R. (1973). a a e t’c e t'onsh ' th me ta psychotherapx_1nterxiew1 Unpublished doctoral dissertation, University of Michigan, Ann Arbor. 169 Sachs, J. S. (1983). Negative factors in brief psychotherapy: An empirical assessment. Jgnrnal_gf Qonaul11ng_and_£lin1211_zsxchologxi 51(5), 557-564. Saltzman, C., Luetgert, M. J., Roth, C. H., Creaser, J., 8 Howard, L. (1976). Formation of a therapeutic relationship: Experiences during the initial phase of psychotherapy as predictors of treatment duration and outcome. EfixgthQQXL $51 545-555- Sandler, J., Dare, C., 8 Holder, A. (1973L Ina_narlanr_ang l‘ .la a ' 9‘ 313‘: 0 1 9‘ !' 1:, 0 00 (pp. 27-121). New York: International Universities Press. Shrout, P. E., 8 Fleiss, J. L. (1979). Intraclass correlation coefficients: Uses in assessing rater reliability. Psychological_fiulletini 36121. 420-428. Silberschatz, G., Fretter, P., 8 Curtis, J. (1986). How do interpretations influence the process of psychotherapy? I2urnal_of_gonsulting_and_clinical_£§xcholosxi 511 646- 652. Sterba, R. (1934). The fate of the ego in analytic therapy. In1erna112na1_J_urna1_of_£_xch2:An_lxs1_1_1§1 117-126. Strupp, H. H., Fox, R. E., 8 Lesser, K. (1969). Eatlents x1ew_their_psxch21her1221 Baltimore: Johns Hopkins University Press. Strupp, H. H., 8 Hadley, S. W. (1979). Specific vs. non- specific factors in psychotherapy: A controlled study of outcome. Arcnivea at General Esygnlarry. 36I 1125- 1136. Strupp, H. H., Wallach, M. S., Wogan, M., 8 Jenkins, J. W. (1962) Psychotherapists' assessments of former patients. a vo e 's e 37 222-230. Tinsley, H. E. A., 8 Weiss, D. J. (1975). Interrater reliability and agreement of subjective judgments. 12urna1_2f_§oun§eling_2sxchologxi 111 358-376. Vaillant, G. (1977). AQQEL§£123_£Q_11£§1 Boston: Little, Brown. Windholz, M. J., 8 Silberschatz, G. (1988). Vanderbilt Psychotherapy Process Scale: A replication with adult outpatients. a Co 5 t d 'n' al E§¥§h919911 §§1 56-60. Wf 170 Windholz, M. J., Weiss, D. S., 8 Horowitz, M. J. (1985). An empirical study of the natural history of time-limited psychotherapy for stress response syndromes. 22‘ 547-554. Winston, A., Pinsker, H., 8 McCullough, L. (1986). A review of SUPportive psychotherapy. WW Baroniatrxi 111 1106-1114. Xenakis, S. N., Hoyt, M. F., Marmar, C. R., 8 Horowitz, M. J. (1983). Reliability of self-reports by therapists using the Therapist Action Scale. angngrnarapyy zgy 314-320. Zetzel, E. (1956). Current concepts of transference. - 31, 369-376. 1111111111