a

1: 4 ~.~.
' t ‘P'EH.'.":.' .‘ ,. ‘
Mimi“! 2‘45“. '
I" I,. ‘Idslw‘w I h t
{In} H #8”?

li,‘,'-,3fs".’l

3': $59:

"ng . 5“

J ~41. 9‘ 1'
z. ‘. ' ', t

‘v ”5.):
*- "r. WT,
' %
r' 5‘.
Vt}:

-.N' u‘ "‘1
II‘ I n‘p‘uj
.$fi‘~u

' .t’.

x. 339‘“. 3-3:. ‘. ‘
"‘VW‘ Al". ‘4
'Yut‘ 4"..VV‘ ,5. n
'<

' ’n“:“‘l

 

‘
51'"...
1-2)

' ¢- 4 J .
45-11:

.‘-.

.;fl*'£f.-‘r‘v - .» u.

d

j _» : gr. *1 I0,‘«-.‘_:, ‘1‘“),

c'. 'u.,‘
' 1",;

D I‘
v" ~._ ’ a V
'IIF‘Jfllt"

up.

an.

. ‘|

.J ’5 ‘3‘ I:

.1 . "Y _ . 7‘. -,, :\.,‘.

.‘3‘ "NV-“N" 5353;?" “ljfii'q‘; ‘I'L‘f .. I,
‘u )" “Pr-‘49" 1""Q'y3'4'v'w‘ “ '
"~v','-.'a '5“. ‘g,".‘ " . \.I "1'" pk, Ip'
. - .z‘wn‘ m r.- .,s . .
:1 , .Ir-w ‘-,;..>

I
'I
‘I

"y— w. i—qv'?’

‘ 4..., I II
' T." \ H /| "&’#
of“ "A , ‘ . I ‘ ‘ , v \ amt
“if“:‘szi ~11},- l - t ~ . u , flaw
{w fil'. - ""'

‘4'.” D1 -

with

\H: ' ‘ 1
l 41:19. .9

 

This is to certify that the

dissertation entitled
i

PSYCHOTHERAPY OUTCOME
AND THE COURSE OF THE THERAPEUTIC ALLIANCE

presented by

Michelle Rae Klee

has been accepted towards fulfillment
of the requirements for

Ph . D . degree in PSJLchology

WW

Major professor

Norman Abeles, Ph.D.

Date em er 10 l 86

MS U is an Affirmative Action/Equal Opportunity Institution 0-12771

 

 

MSU

 

 

 

RETURNING MATERIALS:
ace in 00 rop to
remove this checkout from

 

LIBRARIES
“ your record. FINES will
be charged if 550E is
returned after the date
‘ -~—~‘,, . stamped below.
“III 2. 5M¥ ‘

J-JJLQ or}?

x.
mill.
C)

VOW/fl)
2/” ”‘

6/144 0

MAGIC é” ‘”

“twat?

 

 

PSYCHOTHERAPY OUTCOME

AND THE COURSE OF THE THERAPEUTIC ALLIANCE

BY
Michelle Rae Klee

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Psychology

1986

'K/h)

~——\’~. v

ABSTRACT

PSYCHOTHERAPY OUTCOME
AND THE COURSE OF THE THERAPEUTIC ALLIANCE

BY

Michelle Rae Klee

The failure of researchito consistently validate the
theoretical importance of the therapeutic alliance in
predicting psychotherapy outcome was hypothesized to be a
consequence of an inadequately complex approach to the
alliance research, anm approach particularly likely to
obscure the influence of therapist variables. The primary
purpose of this study was to address the complexity of the
therapeutic alliance by studying the interaction of the
patient's initial potential for establishing a
relationship, the course of patient and therapist
contributions to the alliance, and psychotherapy outcome,
defining outcome in clinically significant terms. A second
objective of the study was to assess the applicability of
the Therapeutic Alliance Rating Scale (TARS) to a
heterogeneous sample of psychotherapy cases.

Thirty-twr> adult psychotherapy' patients were
classified as having a high or low prognosis for
establishing a relationship based on clinical judge ratings
on the TARS of their contributions to the therapeutic
alliance in the first session of treatment. For each case,
judges also rated patient and therapist contributions to

the alliance for an early, middle, and late session in

Michelle Rae Klee
treatment. Patients were assigned to one of three outcome
groups, based on the application of criteria for clinically
significant improvement to scores on the SCL-90-R.

As predicted, from the first session of treatment,
patients demonstrated a potential for establishing a
therapeutic relationship that was predictive of their
capacity to contribute to a therapeutic alliance throughout
treatment. The results partially supported the hypothesis
of an increase from early to late therapy in the positive
contributions to the alliance made by patients who achieved
a reliable improvement in symptom level. Contrary to
predictions, the influence of therapist behavior on process
and outcome was no greater for low prognosis than for high
prognosis patients. No validation for the theoretical
importance of patient or therapist contributions to the
therapeutic alliance in predicting outcome was obtained.
An unpredicted interaction of phase of treatment and
outcome was found for therapist negative contributions to
the alliance.

Further research is needed before a conclusion can be
drawn about the applicability of the TARS to a

heterogeneous sample of psychotherapy cases.

To my husband Ken
and
to my family

iv

ACKNOWLEDGMENTS

I wish to acknowledge the people who assisted me in
this research. Dr. Norm Abeles was valuable as my
dissertation committee chairperson not only for his
expertise in the area of psychotherapy research, but for
his ability to strike a balance between allowing me
autonomy and providing me with guidance. Dr. Ralph Levine
gave generously'of his time and talents during the design
and data analysis phases of this research project. I am
indebted.to my other committee members,lh3 Robert Caldwell
and Dr. Bertram Karon, for their careful reading and
thoughtful comments on my work.

I would like to express my appreciation to Ann
Isenberg and Gary Gunther, who spent numerous hours
listening to and rating audiotapes of psychotherapy
sessions for this study. Finally, I would like to
acknowledge the assistance and support of‘Timothy Eaton,

with whom I collaborated in the data collection process.

TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . viii
LIST OF FIGURES. . . . . . . . . . . . . . . . . . . . x
CHAPTER
I. INTRODUCTION . . . . . . . . . . . . . . . . . 1
II. REVIEW OF THE LITERATURE. . . . . . . . . . . . 6
Concept of the Therapeutic Alliance . . . . . . 6
History . . . . . . . . . . . . . . . . . . 6
Generalization of the concept . . . . . . . 10
Core components of the therapeutic
alliance. . . . . . . . . . . . . . . . . 12
Patient contributions . . . . . . . . . . . l3
Therapist contributions . . . . . . . . . . 14
Course of the alliance. . . . . . . . 15
Research Related to the Therapeutic Alliance. . 16
Influential findings. . . . . . . . . . . . l6
Client-centered process variables . . . . . 18
Measuring the therapeutic alliance. . . . . l9
Predicting the strength of the
therapeutic alliance. . . . . . . . . . . 20
Patient variables in relationship
to outcome. . . . . . . . . . . . . . . 21
Therapist variables in relationship
to outcome. . . . . . . . . . . . . . . . 27
Course of the alliance. . . . . . . . . . . 28
Early indicators of the alliance and
outcome . . . . . . . . . . . . . 30
Phase of treatment and outcome:
Patient contributions . . . . . . . . . . 31
Phase of treatment and outcome:
Therapist contributions . . . . . . . 33
Interaction of patient predisposition
with outcome. . . . . . . . . . . . . . . 36
Conclusions from the therapeutic
alliance research . . . . . . . . . . . 40
Clinical Significance of Outcome Criteria . . . 42
The evaluation of psychotherapeutic
outcome . . . . . . . . . . . . . 42
Criteria for clinical significance. . . . . 45

vi

III.

Statement of Problem and Objectives

Hypotheses.
IV. METHOD. . . .

Participants.

Procedure . . .
Variables . . .
Instruments . .

V. RESULTS .

OBJECTIVES AND HYPOTHESES

Design and Overview of Statistical Analyses
Classification of Subjects.

Reliability . .

Hypotheses Testing Analyses
Prognosis Measure

Outcome . . . .

Phase of Treatment.

Intercorrelations Between Alliance Subscales.

Summary of Results.

VI. DISCUSSION. .

Prognosis for the Alliance.
Prognosis for the Alliance and Outcome. .
Patient Contributions to the Alliance . .

Therapist Contributions

Course of the Alliance. .

Outcome . . . .

Validation of the TARS. .

Therapeutic Alliance.

SUMMARY AND CONCLUSION. .

Applicability of the TARS

Clinically Significant Criteria for

Conclusion. . .

APPENDIX . . . .

Manova Summary Tables

LIST OF REFERENCES

vii

to the Alliance

Outcome

49

49
53

56

56
57
58
63

73

73
75
76
78
89
94
97
98
100

103

105
108
110
118
124
126
130

137
137
144
145
146

148
148

152

TABLE

LIST OF TABLES

Gross Pathology Index Data Used in
Determining Outcome Criteria. . . . . . . . . . 60

Mean Intensity Scores for the Patient
Positive Contribution Scale . . . . . . . . . . 79

Mean Intensity Scores for the Patient
Negative Contribution Scale . . . . . . . . . . 80

Summary of 2 (Prognosis) x 3 (Outcome) x 3
(Phase) Manova Results for the Patient
Positive Contribution Scale . . . . . . . . . . 82

Summary of 2 (Prognosis) x 3 (Outcome) x 3
(Phase) Manova Results for the Patient
Negative Contribution Scale . . . . . . . . . . 83

Mean Intensity Scores for the Therapist
Positive Contribution Scale . . . . . . . . . . 86

Mean Intensity Scores for the Therapist
Negative Contribution Scale . . . . . . . . . . 87

Summary of 2 (Prognosis) x 3 (Outcome) x 3
(Phase) Manova Results for the Therapist
Positive Contribution Scale . . . . . . . . . . 90

Summary of 2 (Prognosis) x 3 (Outcome) x 3
(Phase) Manova Results for the Therapist
Negative Contribution Scale . . . . . . . . . . 91

Intercorrelations Between TARS Subscales
for Each Phase of Treatment . . . . . . . . . . 99

Summary of 2 (High vs. Low Outcome) x 3
(Phase) Analysis of Variance for Patient
Positive Contribution Scale . . . . . . . . . . 148

Summary of 2 (High vs. Low Outcome) x 3

(Phase) Analysis of Variance for Patient
Negative Contribution Scale . . . . . . . . . . 149

viii

A-4.

Summary of 2 (Reliable Change vs. No Change)

x 3 (Phase) Analysis of Variance for Therapist
Positive Contribution Scale: Low Prognosis
Patients. . . . . . . . . . . . . . . . . . . . 150

Summary of 2 (Reliable change vs. No Change)

x 3 (Phase) Analysis for Variance for Therapist
Negative Contribution Scale: Low Prognosis
Patients. . . . . . . . . . . . . . . . . . . . 151

ix

FIGURE

1.

LIST OF FIGURES

Outcome x Phase interaction for Therapist
Negative Contribution Score . . . . .

96

INTRODUCTION

A relationship between two people, a relationship in
which a person in distress seeks the assistance of another
person professionally qualified to help (Tarachowy 1963h
is the unifying element in the myriad forms of individual
psychotherapy which have evolved since Sigmund Freudis
introduction of psychoanalysis. Regardless of variations
in therapeutic technique, from the exploration and
interpretation of the psychoanalyst to the executive
direction of the behavior therapist, all therapy takes
place within the context of a relationship. Though
transference and resistance have been the aspects of the
patient-therapist relationship emphasized ix: the
psychoanalytic literature (Hartley & Strupp, 1983), the
rational, positive attachment of the patient to the
therapist has been recognized as "the vehicle of success
(Freud, 1912, p. 105)" in psychotherapy since the early
writings of Freud. The idea of an undistorted, beneficial
attachment of the patient to the analyst as a prerequisite
to successful psychoanalysis has been elaborated over time
into a theoretical construct which impacts other
therapeutic modalities and is labeled the therapeutic

alliance. The therapeutic alliance can be defined as "the

2
observable ability of the therapist and patient to work
together in a realistic, collaborative relationship based
on mutual respect, liking, trust and commitment to the work
of treatment” (Foreman & Marmar, 1985, p. 922).

The theoretical importance of both patient and
therapist contributions to the therapeutic alliance as a
prerequisite to successful outcome in psychotherapy has not
been consistently borne out by research. Though the
accumulathmn of data indicates that positive patient
contributions to the alliance are associated with good
outcome, the data on patient negative contributions is more
equivocal. As in much of the psychotherapy research
literature, the hypothesiscnfa.significant influence of
therapist behavior on outcome is not well supported in the
therapeutic alliance research.

Perhaps the failure to validate empirically theory and
clinical experience stems in part from the conventional
approach of basing data analyses on the entire sample of
patients, regardless of the potential of patients for
forming a therapeutic relationship or their prognosis for
outcome. There is pueliminary evidence that specific
alliance-related behaviors may have different implications
for outcome depending upon patient predispositional
variables. Significant relationships between the course of
alliance and outcome may be eclipsed by combining the data

of patients with high and low potentials for an alliance.

3

The risk of masking relationships between process and
outcome by pooling the data for all patients is greater for
therapist than for patient variables. While individuals
with a good potential for developing a therapeutic
relationship are likely to maintain a solid alliance and to
achieve successful results, unless the therapist is
extremely untherapeutic, the influence of therapist
behavior on outcome will presumably be greater for patients
with an initially poor capacity to form an alliance and to
use therapy productively. Since patient factors appear to
account for a significantly higher proportion of variance
in outcome than therapist factors, the effect of therapist
behavior is likely to be obscured when all subjects are
pooled 1J1.a traditional correlational approach to
statistical analysis (Suh & O'Malley, 1982). In order to
elucidate the influence of therapist alliance-related
behaviors on outcome, research which addresses the
interaction of the course of the alliance across treatment
with the patient's initial potential for the formation of a
relationship with the therapist is needed.

In this study, two questions concerning the course of
the therapeutic alliance were of primary interest. What
therapist and patient contributions to the alliance are
differentially present during the course of treatment in
cases with initially similar patient potentials for a

relationship and dissimilar outcomes? Iknv does therapist

4

action influence patient alliance-related behavior among
patients with initially high and low relationship
potentials? IIt was anticipated that differentiating
patients with high and low relationship potentials would
reveal implications of therapist alliance-related
behaviors for the course of patient contributions to the
alliance and for outcome which have been obscured when all
patients are pooled. Ultimately, it was hOped that the
research would contribute toward the identification of
therapist.behaviors which facilitate the development.and
maintenance of a therapeutic alliance with patients who
begin treatment with a poor potential for a working
relationship and thus a poor prognosis for outcome.

Though there exist in the literature a number of
instruments designed to assess the therapeutic alliance,
most measures have been employed by only one group of
investigators with a single population. Additional
research is required to validate the use of the instruments
with diverse populations under varied treatment conditions.
One of the most promising measures of the alliance, the
Therapeutic Alliance Rating Scale (TARS) developed by
Marziali and her associates (Marzaili, Marmar, & Krupnick,
1981; Marziali, 1984), has been tried only in research on
brief psychotherapy with relatively high-functioning
patients, conducted by very experienced psychoanalytically-

oriented therapists. The present study provided an

5
opportunity to assess the applicability of the TARS to
research with a more heterogeneous sample of patients and
therapists engaged in treatment of varying theoretical
orientations and durations.

Psychotherapy outcome research is plagued by the
question of whether the empirical findings have any
relevance for clinical practice. Among the aspects of
research most vulnerable to criticisms by clinicians is the
overreliance upon group means and statistical tests in
measuring outcome, with little attention to
thejprobability'of benefits for the individual patient or
to the practical significance of change. It was hoped that
by the use of clinically significant outcome criteria, this
study would make a clinically relevant contribution to the

psychotherapy research literature.

REVIEW OF THE LITERATURE

Concept of the Therapeutic Alliance

History. Despite the focus in the psychoanalytic
literature on transference and resistance in the
therapeutic relationship, a realistic, positive attachment
of the patient to the therapist has been recognized as a
prerequisite to successful treatment since the inception of
psychoanalysis. In his 1913 paper, "On Beginning the
Treatment," Freud defined the first aim of treatment as
developing a rapport, attaching the patient to treatment
and to the person of the analyst. He differentiated a
conscious and unobjectionable aspect of positive
transference, which facilitates analysis, from the positive
transferencerof repressed erotic impulses, which creates
resistance to change unless it is analyzed (Freud,
1912/1958).

Subsequent conceptualizations of a treatment-enhancing
bond between patient and analyst were marked by a shift
from emphasis on a libidinal attachment to emphasis on an
alliance based on cognitive and motivational factors.
Writing from the structural perspective in psychoanalysis,
Sterba (1934) described a dissociation of the reality-

oriented part of the patient's ego to ally with the

analyst via identification, opposing the part.of the<ego
attached in transference to instinctual energy. His
formulation implied a cognitive collaboration between
patient and analyst motivated by a common goal, the
patient's health. Sterba specified the interpretation of

transference and the use of we" by the analyst as the
techniques of effecting this alliance. In his later
writings, Freud (1937) discussed a somewhat similar pact of
the analyst with the ego against the id,lnn:conceived.of
transference as the patient's strongest motive for
collaboration:hianalysis. Feniche1(l94l),following in
the direction of Sterba, postulated observing and
experiencing components of the ego; the capacity of
individual to observe and reflect upon experience was
deemed essential to forming a working relationship with the
analyst.

An emerging interest in the roles of patient ego
development and object relations in the therapeutic
relationship was evident in Zetzel's 1956 paper on
transference. She stated that a certain measure of
mature ego functioning and capacity for trust is a
prerequisite to the development of a sound therapeutic
alliance, which is essential to effective analysis.
Analysis of regression in the transference situation is
possible only if the patient possesses continuing adequate

ego strength to maintain the therapeutic alliance at an

8

adult level in the face of regression, Thus the analytic
technique was considered untenable with children and
severely disturbed individuals. Zetzel contrasted this
traditional point of view to the perspective of the object
relations theorists, for whom the distinction between the
therapeutic alliance and the transference neurosis was
judged insignificant and the preexistence of mature ego
functioning unnecessary for analysis. Regression in the
transference situation was viewed as a deepening of
analysis rather than a resistance which must be protected
against by the therapeutic alliance.

The concept of the alliance in psychoanalytic
psychotherapy has perhaps been most fully elaborated by
Greenson (1965; 1967; Greenson & Wexler, 1969), who
explicitly stated that the working alliance is as crucial
as the transference neurosis. The influence of earlier
theorists is easily seen in Greenson's definition of the
working alliance :

the relatively nonneurotic, rational rapport which the

patient has with his analyst.”.the reliable core of

the working alliance is formed by the patient's
motivation to overcome his illness, his sense of
helplessness, his conscious and rational willingness
to c00perate, and his ability to follow the
instructions and insights of the analyst. The actual
alliance is formed essentially between the patient's

reasonable ego and the analyst's analyzing ego (1967,

p 192).

Like the conceptualization by Sterba (1934), Greenson's

working alliance was a cognitive collaboration, motivated

by congruent goals of the patient and analyst and based

9
upon the patientls partial identification with the analyst.
However, Greenson recognized that the working alliance, in
addition to the dominant rational component, includes a
mixture of "unconscious motherly and fatherly components"
(1967, p. 240).

Greenson (1967) differentiated from the working
alliance and the transference a third component of the
patient-therapist relationship, the real relationship. By
the real relationship, he referred to the undistorted,
genuine relationship which is built upon the patient's
realistic perception of the analyst's consistently
therapeutic attitude and objectionable traits. The real
relationship seems to incorporate and expand upon Freudfis
original conceptualization of the patient's conscious and
unobjectionable affectionate feelings for the analyst.
According to the author, it is the consistently therapeutic
attitude of the analyst which allows the patient to develOp
the identification which is the core of the working
alliance. In Greensomfs schema, transference, the working
alliance and the real relationship, though theoretically
differentiated, overlap and influence one another.

Other authors have viewed the therapeutic alliance as
a broader phenomenon, of which the rational collaboration
of Greenson's working alliance is but one component.
Dickes (1975) conceptualized the therapeutic alliance as

being composed of five components: the patient's

10

motivation for treatment based on ego-alien symptoms:
positive transference; negative transference; the working
alliance; and the real relationship. Frieswyk, Colson and
Allen (1984) narrowly define the working alliance as the
patient's active collaboration in tune work of
psychotherapy, then discusses other element of the
therapeutic relationship which facilitate the
collaboration: the working relationship of the reasonable
ego in identification with the analyst, the real
relationship, a core of stable objects relations which
comprise the matrix transference and the specific
transference.

Generalizability of the concept. Though the concept
of the therapeutic alliance derived from the psychoanalytic
perspective, the emphasis upon relationship as a component
of the change process has by no means been limited to one
theoretical perspective. From the client-centered view of
Carl Rogers (1957), the relationship was the technique;
warmth, genuineness and empathic understanding on the part
of the therapist were perceived as the necessary and
sufficient conditions for change. Harry Stack Sullivan
(1954) and the other interpersonal theorists (eugq. Havens,
1976) viewed the therapist as a participant observer in
the relationship. Though some behaviorists are now giving
consideration to the role of the patient-therapist

interaction in the change process (e.g., Lazarus, 1974),

11
the therapeutic relationship has traditionally received
least emphasis in the behavioral school.

Bordin (1979) postulated that the working alliance is
one of the crucial factors, if not the crucial factor, in
all modes of psychotherapy, though different types of
working alliances are required for different therapies. He
designated three components of the working alliance: an
agreement upon goals, an assignment of tasks, and the
evolution of a bond. Perhaps most diverse among various
psychotherapies are the goals of treatment, ranging from
the very circumscribed behavior therapy objectives of
changing specifics acts to the comprehensive psychoanalytic
therapy goals of altering enduring patterns of thought,
feeling, and behavior. Though the tasks of psychotherapies
vary greatly, all but a few rigidly behavioral treatments
require self-observation, most of ifluun including
observation of inner experience to determine motivation for
behavior. Though the depth of the bond necessary for a
working alliance varies with the goal and tasks of
psychotherapy, some basic level of trust and attachment
must be embedded in all functional therapeutic
relationships. Bordin (1979) declared that ii: is the
strength of the alliance rather than the type<ofaalliance

that is the essential component for change.

12

Core cxnmponents of the therapeutic alliance. Despite

 

the variations in the types of alliance required for
different modes of therapy, and discrepancies in
definitions of the therapeutic alliance within the
psychoanalytic literature, two components seem to be
consistently recognized as essential to the therapeutic
alliance: an affective component and a cognitive-
motivational component (Bordin, 1975; Greenson, 1967;
Hartley & Strupp, 1983; Marziali, Marmar, & Krupnick,
1981). The affective component, generally identified in the
psychoanalytic literature as the real relationship, is the
mutual bond between the patient and therapist, based on
realistic, nontransferential perceptions and genuine
liking, trust and respect. The cognitive-motivational
component, identified in the psychoanalytic literature as
the working alliance, is the collaboration of the therapist
and patient in the requisite treatment tasks toward the
agreed upon goals (Hartley, 1985). While the consensus
is that transference influences the real relationship and
the working alliance, it stands in contrast to these
realistic components of the therapeutic relationship
(Frieswyk et al., 1984). Since Zetzel's 1956 paper, there
has been increasing emphasis among psychoanalytically
oriented psychotherapists on the therapeutic alliance as a

prerequisite to effective therapy.

13

Patient contributions. Given tflua contingency of

 

successful psychotherapy upon the formation of a
therapeutic alliance, *what characteristics enable a
patient to establish an alliance? Not surprisingly, this
issue has been primarily discussed in the psychoanalytic
literature, in which a variety of aspects of relatively
mature ego development are specifiedias prerequisites to
the formation of a therapeutic alliance: the capacity for
reasonably stable object relationships required to form.a1
trusting relationship; the capacity to distance oneself
from experience temporarily, observe oneself, and report
these observations; the ability to comprehend and reflect
upon the statements of the therapist; the capacity for
logical and emotional verbal communication; and the
ability to alternate between regression into fantasy and
contact with reality (Dickes, 1975; Greenson, 1967).
Though Bordin (1975) discussed variations in the nature of
the alliance required by different psychotherapies, he
designated patient characteristics commonly required across
behavioral and psychodynamic treatments which are quite
similar 13) the ego functions described in the
psychoanalytic literature.

Since such patient capacities suggest a fairly high
level of functioning, jJ: is not surprising that some
authors have questioned whether patients with major

deficits in ego development, such as those with borderline

14

and narcissistic personality disorders, are able to form an
alliance or engage in therapeutic work (Greenson, 1967;
Langs, 1982). Others, such as Kernberg (1975), Kohut and
Wolf (1978) and Masterson (1978), indicate that, with
special attention tx> the transference, a working
relationship can be established with patients with an
initially poor capacity for the formation of an alliance
and treatment.

Therapist contributions. A certain degree of
psychological health in the therapist is recognized across
theoretical persuasions as a prerequisite to effective
psychotherapy and, by implication, to the development of a
therapeutic alliance. Though the technical contributions of
the therapist toward forming an alliance vary greatly from
one mode of treatment to another, a common core of
therapist characteristics within the therapeutic situation
seems to be required. Greenson writes of the analyst's
"consistent attitude of acceptance and tolerance" (1967, p.
3) and "consistent and unwavering pursuit of insight in
dealing with any and all of the patient‘s material and
behavior" (1965, P. 210) "in an atmosphere of serious work,
straightforwardness, compassion and restraint"(1965, p.
216). Though framed within the analytic perspective,
GreensonfS'words capture the essential requirements of a
consistent attitude of respect, acceptance, and concern for

the patient and a consistent commitment to the given tasks

15

and goals of psychotherapy. The technique of the therapist
may vary greatly, from the benign neutrality of the
psychoanalytic psychotherapist tn) the self-disclosure
characteristic of some humanistic therapists, but the
therapeutic attitude toward the patient is consistent.

Course of the alliance. Despite the existence of much
literature on the therapeutic alliance, discussions of the
course of the therapeutic alliance across treatment are
infrequent. Since pretreatment patient and therapist
characteristics are viewed as determining the capacity to
form an alliance, it seems that precursors of the
therapeutic alliance must be present from the first session
of therapy. The patient's initial response to the
therapist depends upon her/his history of object relations,
level of ego development, and reactions to realistic
characteristics of the therapist. Because the bond, or real
relationship, zus more easily established” it appears
earlier in treatment than the working alliance (Dickes,
1975; Greenson, 1967). According to Greenson (1967), the
early signs of the working alliance are typically seen in
the first three to six months of analysis, after a piece of
transference-resistance has been effectively analyzed.
Obviously, the prOponents cxf brief psychoanalytic
psychotherapy (eng., Malan, 1976; Mann, 1973; Sifneos,
1972) maintain that a working alliance can be developed

much more rapidly. The real relationship has been

16
described as being predominate in the early and terminal
stages of treatment, while the working alliance develops
toward the end of early stage of treatment, but abates
periodically as the patient approaches specific areas of
conflict. As the transference which dominates the middle
phase of therapy is resolved and diminishes, the real
relationship expands (Greenson, 1967; Ticho,.Appe1baum,

Binstock, & Appelbaum, 1971).

Research Related to the Therapeutic Alliance

Influential findings. Two»major empirical findings
influenced the course of psychotherapy research toward an
emphasis on patient and therapist relationship process
variables. The first finding was the disappointing results
of' early' psychotherapy studies which focused on
pretreatment patient and therapist variables as predictors
of outcome. In a review of 166 outcome studies, Luborsky
and his associates (Luborsky, Chandlery Auerbachq Cohen &
Bachrach, 1971) found that research frequently revealed a
number of pretreatment variables to be related to outcome:
patient adequacy of personality functioning, intelligence,
motivation, anxiety, educational and social assets,
therapist attitude and experience, and patient and
therapist similarity. However, the conclusiveness of the
review was limited by methodological flaws of the studies
and inconsistent results; the factors which differentiated

research with positive results from research with negative

17

results were indiscernible. Major multivariate
psychotherapy outcome studies have consistently indicated
that, while some pretreatment variables are statistically
significant predictors of outcome, the proportion of
outcome variance accounted for is only in the 5 to 10%
range (Fiske, Cartwright, & Kirtner, 1964; Luborsky et
al., 1980; Sloan, Staples, Cristol, Yorkson, & Whipple,
1975; Strupp & Hadley, 1979). The meager results from
studies of the predictive value of pretreatment patient and
therapist variables contributed to a shift in focus to
process variables in psychotherapy research.

The second finding, which irrevocably altered the
course of psychotherapy research, was the results of the
Smith, Glass, and Miller (Smith & Glass, 1977; Smith,
Glass, & Miller, 1980) meta-analysis of 475 controlled
psychotherapy outcome studies. The authors concluded that
psychotherapy is consistently beneficial to the patient in
many ways, with no difference in the degree or type of
benefit attributable to the type of psychotherapy. This
conclusion relieved psychotherapy researchers of the burden
of proving the general efficacy of psychotherapy,
justifying the pursuit of explanations of why and with whom
various types of psychotherapy are effective (Abeles,
1985). The Smith en: a1. results have also been
interpreted by some (Shapiro & Morris, 1978) to mean that

the benefits of psychotherapy are solely due to

18

nonspecific effects, or effects that are independent of
technique, such as patient and therapist expectancy,
therapist credibility, and suggestion. Thus the Smith,
Glass, and Miller analysis stimulated increased interest
in elucidating the relationships among variables in the
process of psychotherapy and the influence of non-
specific factors. Abeles (1985) offered the term
I'unspecified" as a more accurate alternative to non-
specific, and suggested that defining the components of the
therapeutic alliance makes specific some of the unspecified
patient and therapist factors which influence psychotherapy
outcome.

Client-centered_process variables. Earlier research
from a client-centered framework suggested that patient
process variables have more influence on psychotherapy
outcome than therapist process variables. While initial
studies of the predictive value of such therapist variables
as warmth, accurate empathy, and genuineness were very
promising, (e.gH, Truax & Mitchell, 1971), more recent
research has failed to replicate these findings and the
cumulative data provides inconsistent support at best for
the relationship between the Rogerian facilitative
conditions and outcome (see review by Parloff, Waskow, &
Wolf, 1978). The results of research from the client-
centered perspective (Hi the predictive utility of patient

process variables have been more convergent. 13120 of 26

19

studies, most of which employed the Barrett-Lennard
Relationship Inventory (Barrett-Lennard, 1962), the
patient's perception of the therapist-offered relationship
was positively associated with outcome (Gurman & Razin,
19770. The patient capacity for experiencing, being able
to experience deeply and immediately and reflect upon and
report this feeling (Gendlin, 1962; Gendlin, Beebe,
Cassens, & Oberlander, 1968) has been consistently reported
to be positively associated with outcome in client-centered
therapy (see Luborsky et al., 1971 for review). The
concept of experiencing is similar to some patient
abilities defined by psychoanalytic theorists as essential
to the formation of a therapeutic alliance.

Measuring the therapeutic alliance. In the last five
years, psychoanalytic theorists have developed a number of
research instruments specifically designed to measure
aspects of the therapeutic alliance (Allen, Newsom,
Gabbard, & Coyne, 1984; Hartley & Strupp, 1983; Horvath &
Greenberg, in press,cited in Hartley, 1985; Luborsky,
Crits-Christoph, & Alexander, 1983; Marmar, Marziali,
Horowitz, & Weiss, in press; Marziali, 1984; Sachs, 1983)
and there has been a surge of research on the construct.
Many of the instruments developed to assess the alliance
have been employed by only one group of investigators with
a single population. The construct of the therapeutic

alliance has been operationalized in a variety of ways,

20
making it somewhat difficult to generalize from one study
to another. The ideal therapeutic alliance instrument
should assess separately the positive and negative
contributions of both patient and therapist to the
affective and cognitive-motivational components of the
alliance.

Predicting the strength of the alliance. Though
little research has been done on factors predicting the
strength of the therapeutic alliance, thus far support for
the influence of the pretreatment measures of patient
psychopathology on the alliance is weak. Marziali (1984)
reported depressive mood and symptoms of psychological
disturbance to be related to ratings of patient
or therapist negative contribution to the alliance only in
the first session. A significant association between
psychological health and the degree of patient involvement
in the therapeutic process was found in another study to be
largely accounted for by the correlation of psychological
health with level of interpersonal relations (Moras &
Strupp, 1982).

Firmer evidence has been produced in support of the
theoretical assumption that a more specific aspect of
patient functioning, the capacity for object relations,
influences the strength of the alliance. Moras and Strupp
(1982) found that pretherapy assessment of interpersonal

relationships based on a clinical interview predicted up to

21

25% of the variance in the patientfs activity, initiative,
and hostility during therapy. Assessing the alliance from
three perspectives, patient, therapist, and clinical judge,
Marziali (1984) reported that the pretreatment social
adjustment of the patient correlated significantly with
judge and therapist ratings of patient negative and
positive contributions to the alliance, and with patient
ratings of their own and the therapist negative
contributions. Ryan (1973) also found that pretreatment
measures of the quality of object relations predicted the
quality of alliance. Assessing process indicators of
patient ability to establish a relationship, Morgan and
colleagues (1982) showed that the level of patient
involvement in a therapeutic alliance as early as the third
and fifth sessions of treatment was predictive of patient
contributions late in treatment. Thus the available
research suggests that while the patient's degree of
psychological disturbance is of questionable influence on
the therapeutic alliance, the capacity'of the patient for
relatedness, present from the beginning of treatment and
evident in certain predispositional and process variables,
does have significant implications.

Patient variables in relationship to outcome. Though
there are inconsistencies in the literature, the cumulative
research results indicate that patient involvement in the

therapeutic alliance is predictive of treatment outcome. .A

22
number of measures relevant to the therapeutic alliance
combine positive and negative indicators into a single
dimension, a theoretically questionable practice which
complicates the interpretation of results. The original
Therapeutic Alliance Rating Scale (TARS) developed by
Marziali, Marmar, and Krupnick (1981), was designed to
measure affective and cognitive-motivational components of
the therapeutic climate“ The judge-scored Patient Total
Contribution Scale of the TARS, which incorporates both
positive and negative components of the alliance, was found
to differentiate patients with the most successful outcome
from patients with the least successful outcome. The
Patient Involvement scale from the Vanderbilt Psychotherapy
Process Scale (VPPS), which includes indicators of patient
active involvement and of negative affect toward therapist,
has been found to predict overall improvement ratings by
nonparticipant judges and therapists, and target symptom
improvement ratings by therapists (Gomez-Schwartz, 1978;
O'Malley et al., 1983). The scale does not assess
positive patient affect toward the therapist or negative
indicators of the working alliance. The Vanderbilt
Therapeutic Alliance Scale (VTAS) also combines both
positive and negative factors in each scale. In contrast to
the VPPS and the TARS, the Patient and Patient-Therapist
Interaction subscales of the VTAS (Hartley, 1978; Hartley

& Strupp, 1983) failed to discriminate between high

23
outcome, low outcome, and dropout patients.

While the selection by Marziali et.a1.(1981) of the
most and least improved patients from a larger group may
have maximized the relationship between process and
outcome, thus accounting for the discrepancy with the
results of Hartley and Strupp, the inconsistency between
findings with the VPPS and VTAS are more difficult to
explain. The 28 patients in the Hartley and Strupp study
were taken from the slightly larger pool of patients used
by Gomez-Schwartz (1978) and OPMalley et a1. (1983) and
outcome measures were the same, though used in a composite
form in the VTAS research. While O'Malley et a1. rated
only the first three sessions of treatment, process
variables were rated at a variety of points across
treatment in the other two studies. Perhaps the
inconsistent results can be accounted for by the fact that
the VTAS is more inclusive than the Patient Involvement
Scale of the VPPS and most therapeutic alliance measures,
with the Patient Scale including such components as
anxiety, defensiveness, and motivation.

From research in which positive and negative aspects
of the alliance were approached as separate dimensions, the
convergence of data indicates that ratings of positive
patient contributions to the alliance are associated with
successful outcome. Luborsky and his associates (Luborsky

et al., 1983; Morgan et al., 1982) reported that scores on

24
their Penn Helping Alliance measures, which incorporates
positive patient indicators of affective and collaborative
components of the alliance, discriminated patients with
successful and unsuccessful outcome. Sarnat (1975) found
the Quality of Alliance Scale (Ryan, 1973), an instrument
which taps some of the collaborative aspects of the
alliance but which has many methodological problems, to
predict continuation in treatment. The Patient Involvement
Scale of the VPPS, which contains many items measuring
positive collaborative behavior, was significantly
associated with outcome (Gomez-Schwartz et al., 1978;
O'Malley et al., 1983). A study using the original TARS
(Marziali et al., 1981) and a study employing the revised
TARS (Marziali, 1984) , produced evidence of a significant
association between patient positive contributions to the
alliance and outcome. In contrast to the positive results
with several different alliance measures, Horowitz et al.
(1984) reported that patient positive contributions as
measured by the California Therapeutic Alliance Scale
(CTAS) bore no relationship to outcome» The CTAS and both
versions of the TARS are very similar in form and content
to the TARS, though not all the items are identical and
wording of the items varies. The two patient positive
factors derived from a principal components analysis of
the VTAS, tapping agreement with the therapist on goals and

tasks and collaborative responsibility, also failed to

25
predict outcome (Hartley & Strupp, 1983).

Methodological differences may account for these
inconsistent findings. In the research by the Luborsky
group and in the 1981 Marziali study, indicators of the
therapeutic alliance were compared between the most and
least improved patients selected from a larger sampleiof
patients, thus maximizing the probability of differences in
therapy process between groups. Horowitz et a1. (1984) and
Marziali in her later study (1984) attempted to predict
outcome from therapeutic alliance for groups of subjects
unselected for outcome. While Horowitz assessed the
alliance using only nonparticipant judge's ratings,
Marziali measured the alliance from the perspectives of
patient, therapist and clinical judge. Marziali reported
that though patient and therapist-rated alliance scales
correlated significantly with a variety of outcome
measures, including symptom change, the judge-rated scale
correlated only with patient posttherapy evaluation,
therapist posttherapy evaluation, enui clinical evaluation
of dynamic outcome, measures not included in the Horowitz
study. Thus judge-rated measures of positive patient
contributions to the alliance appear to be predictive of
outcome as measured from only some perspectives, unless the
effect is maximized by preselecting patients on the basis
of extremes of outcome.

The data on the relationship between patient negative

26
contributions to the alliance and outcome is more equivocal
than data (”1 positive contributions. The Patient
Involvement Scale of the'VPPS, which includes many items
tapping patient negative affect toward therapist, was
reported to predict outcome (Gomez-Schwartz et al., 1978;
O'Malley et al., 1983). Using the original TARS,
researchers have found Patient Negative Contributions as
rated by clinical judges to distinguish most improved and
least improved patients, based on composite outcome
measures (Marziali et al., 1981). Patient negative
contributions to the alliance as measured by the CTAS have
been demonstrated by Horowitz and colleagues (1984) to be
negatively associated with symptom reduction. In contrast,
the Marziali 1984 study produced no relationship between
negative TARS signs of patient alliance and symptom change,
though Negative Patient Contributions correlated with
evaluations of outcome by both therapy participants and
with clinical evaluation of dynamic change. It is
difficult to account for the incongruence between the
Horowitz (1984) and Marziali (1984) findings regarding the
association between the judge-rated Patient Negative
Contribution and outcome, given the similarity of the two
scales. The Patient Resistance factor derived from the
principal components analysis of the VTAS (Hartley, 1978)
and negative ratings on the Penn Helping Alliance counting

signs measure (Luborsky et a1” 1983) have not been found

27
to relate significantly to outcome.

Therapist variables in relationship to outcome.
Theoretical assumptions about the influence on
psychotherapy outcome of therapist contributions to the
alliance have meager support at best. Research with the
Penn Helping Alliance Rating method (Morgan et al., 1982),
VTAS (Hartley & Strupp, 1983) and Vanderbilt Negative
Indicator Scale (VNIS, Sachs, 1983) has demonstrated no
association between therapist alliance-related behavior and
outcome. The therapist-offered relationship as measured by
the VPPS has been positively associated only with the
overall rating of patient improvement by the therapist
(O'Malley et al., 1983) or the therapist rating of
improvement on target complaints (Gomez-Schwartz et al.,
1978). Results from research on therapist variables with
the TARS and the CTAS are inconsistent, though the
inconsistency seems to be largely due to the perspective of
the raters of the alliance: patient, therapist, or judge.
Therapist positive and negative contributions to the
alliance as rated by nonparticipant judges were found to
bear no relationship to outcome by Marziali et al.(1981)
and Horowitz et a1. (1984). Similarly, from her 1984
study, Marziali reported no association between mean judge-
rated therapist behavior across treatment and outcome, with
the exception of a significant correlation between

Therapist Positive Contribution and patient evaluation of

28
outcome. She did find significant relationships between
therapist and patient-rated therapist behavior and symptom
improvement, patient and therapist evaluation of outcome,
and dynamic change. The significant findings were
primarily with regard to Therapist Positive Contributions.

The failure of clinical judge ratings of therapist
contributions to the therapeutic alliance to predict
outcome, while consistent with convergent evidence
throughout psychotherapy research that therapist variables
account for little of the variance in outcome, seriously
challenges theoretical assumptions. Horowitz et a1. (1984)
postulated that the use of highly experienced therapists
may preclude significant effects of therapist behavior
because the range of scores on therapist rating scales is
too narrow. Research employing a more heterogeneous group
of therapists might reveal a significant effect of
therapist alliance-related behavior on outcome.

Course of the alliance. The empirical data on the
development of the therapeutic alliance during the course
of treatment is conflicting. Marziali (1984) reported a
significant sessions effect for patient positive and
therapist positive contributions to the alliance, with
average scores for the first and third sessions of
treatment significantly lower than for the last session.
No similar trends were noted fortfluenegative scales. In

contrast to Marziali, Hartley and Strupp (1983) found that

29

the mean alliance rating on the Patient Subscale of the
VTAS decreased significantly from the first to the last
session. Research by the Luborsky group demonstrated no
significant change in the helping alliance from the
beginning to the end of treatment, and scores for
individual patients were moderately correlated from early
to late treatment.(Luborsky et a1” 1983; Morgan et a1”
1982). Gomez-Schwartz and her associates (1978) found no
significant differences in VPPS Patient Involvement
scores attributable to time sequence in treatment.

Very interesting findings were produced when Hartley
and Strupp (1983) looked at therapeutic alliance ratings on
the VTAS at five points in time across treatment, Averaged
across outcome groups, the Patient Subscale, Interaction
Subscale and Total Alliance Subscale peaked at the 25%
point in treatment. The effect approached significance (p
< .07) for the therapist subscale. Though there was no
main effect for outcome, the authors compared high and low
outcome groups at the 25% point in treatment, discovering
that the high outcome group was significantly higher on the
therapist, patient, and total alliance subscales. An
outcome by sessions analysis of variance for the principal
components of the VTAS also produced a main sessions effect
for Patient Resistance, Motivation, Responsibility, and
Anxiety. Though the authors did not conduct least

significant difference tests between sessions across

30
outcome groups, examinationiof the means suggests a peak
scores on those components at the 25% point in treatment,
with a downward trend through the remainder of treatment.

Early indicators of alliance apg_pppppmg. The

 

literature indicates that the relationship between patient
and therapist contributions to the therapeutic alliance and
outcome may vary considerably with the phase in treatment
at which the alliance is assessed. O'Malley, Suh, and
Strupp (1983) reported that the Patient Involvement Scale
of the VPPS showed little relationship to outcome in the
first session, but by the third session accounted for 19 to
28% of variance in all the outcome measures. Ratings on
the Patient Qualities and Patient-Therapist Interaction
subscales of the VNIS were significantly associated with a
composite global measure of outcome for the third, though
not the first and second, sessions of treatment (Sachs,
1983). Such findings suggest that though assessment of
patient alliance variables in the initial sessions of
treatment via the patient subscales may not be useful in
predicting the patientHs eventual improvement, failure to
develop these elements in the therapeutic relationship by
the third session may have negative implications for
outcome» Contradicting this hypothesis, Marziali (1984)
reported a number of significant associations between
patient and therapist positive alliance ratings in the

initial session and outcome. The number of significant

31
correlations did increase from session one to session
three. An analysis of the Marziali data for an outcome by
sessions effect would have been very interesting.

Phase of treatment and outcome: Patient contributions.

 

Luborsky et a1. (1983) reported a stage of treatment by
outcome interaction for a score reflecting patient positive
indicators of the alliance minus patient negative
indicators. Positive helping alliance scores of most
improved patients increased over treatment and negative
scores did not, while negative helping alliance scores of
least improved patients increased and there was little
change in positive scores. The increase reported in
positive helping alliance signs among the most improved
patients is consistent with the Marziali (1984) results
across outcome groups. Luborsky '5 subjects were the
extremes on the continuum of outcome for a large group of
patients, while Marziali used a correlational approach in
which subjects were not preselected on outcome. Since most
patients benefit from psychotherapy, it is possible that
all of Marziali‘s subjects, unlike the least improved
patients in the Luborsky study, had relatively successful
outcomes.

A study by Crowder (1972) in which the Leary
Circumplex of Interpersonal Behavior was used has
relevance for the interaction of the course of patient

contribution to the therapeutic alliance across treatment

32

with outcome. Crowder assessed the occurrence of four
categories.of behavior in therapist.and.patient in early,
middle, and late therapy. Hostile-competitive and passive-
resistant behavior could be considered negative indicators
for the alliance in both patient and therapist. Support-
seeking behavior by the patient and supportive-interpretive
behavior by the therapist could be considered positive
indicators for the alliance. Though the author did not
discuss mean scores on these behaviors across treatment as
they related to outcome, he did report significant
differences between outcome groups at different points in
treatment. Successful patients were more hostile-
competitive and less passive-resistant and supportive-
interpretive than unsuccessful patients in early therapy.
By the middle of therapy, unsuccessful patients were only
more passive-resistant. The outcome groups showed no
differences by late in treatment.

(To obtain his early therapy ratings, Crowder (1972)
averaged scores for the first three sessions of treatment.
His results in combination with the findings of Gomez-
Schwartz (1978), Marziali (1984) and Sachs (1982) provide a
fairly strong argument that by the third session of
treatment, patient alliance-related behaviors have
significant implications for outcome. However, findings of
Crowder that successful patients are more hostile

competitive than unsuccessful patients early in treatment

33
is in conflict with the results of theiother researchers.
Results using the Leary Circumplex must be interpreted with
caution since the measure was not specifically designed to
assess therapeutic alliance and the behavioral categories
are likely to be overinclusive.

Phase of treatment and outcome: Therapist contributions.
Assessing the therapeutic alliance in sessions one and
three of treatment, Marziali (1984) reported some
significant relationships of therapist behavior to outcome
not found when data was averaged across all sessions.
Therapist alliance-building behavior in sessions one and
three as rated by patients, therapists and judges was
significantly related to outcome» An unexpected finding
was that clinical judges' ratings of Therapist Negative
Contributions in early sessions were significantly
positively associated with symptomatic and dynamic
improvement. The Marziali results suggest that significant
relationships between the therapist's alliance-facilitating
or inhibiting behavior at various points in treatment and
outcome may be obscured when alliance ratings are averaged
across treatment.

Some earlier psychotherapy process research employing
the Leary Interpersonal Circumplex (Leary, 1957) lends
support to the hypothesis that the relationship of
therapist alliance-related behaviors to outcome is best

studied by examining the interaction of behavior at various

34
points in treatment with outcome. Employing the Leary
Circumplex, Dietzel and Abeles (1975) studied complementary
interactions between patient and therapist in early, middle
and late phases of psychotherapy. Complementary
interactions are high probability sequences of behavior
which, "are reinforcing to both participants, contribute to
the maintenance of existing behavior patterns, reduce
anxiety, and promote increased relatedness" (p. 264).
Dietzel and Abeles (1975) found no relationship between
mean therapist complementarity across treatment and
outcome. However, in the middle phase of therapy,
successful therapists responded to their patients with a
significantly lower level of complementarity than
unsuccessful therapists. These results seem to be in
keeping with the course of the therapeutic alliance as
theorized in the psychoanalytic literature, since therapist
complementarity may be critical to establish the bond.of
the real relationship which predominates in early phases of
treatment, but some aspects of complementary behavior on
the part of the therapist would interfere with the
development of the working alliance, or cognitive-
motivational, aspect of the therapeutic alliance which
develops later. The trends discussed by Hartley and Strupp
(1983) in their research with the VTAS are congruent with
the results of Dietzel and Abeles. They reported Positive

Climate, primarily ea therapist positive contribution

35
factor, to be significantly lower for high outcome dyads
than for low outcome dyads in the median session of
treatment. The Hartley and Strupp results must be
interpreted cautiously, since comparisons between means
for outcome groups were not statistically justified due to
the absence of an outcome by sessions interaction.

Crowder (1972) found no differences between successful
and unsuccessful therapists in the middle phase of
treatment. Early in treatment, successful therapists
engaged in what would appear to be both alliance-
facilitating and alliance-inhibiting behaviors, being less
passive-resistant but more hostile-competitive than
unsuccessful therapists. Differences in therapist behavior
in the final phase of treatment were more in keeping with
the clinical literature. Therapists of successful dyads
were significantly more supportive-interpretive and less
hostile-competitive and passive-resistant than unsuccessful
therapists. .

Though results indicating that successful therapists
were more hostile-competitive in the early phase of
treatment than their unsuccessful counterparts were
unexpected, they are consistent with Marziali's (1984)
findings that judge's ratings of the therapist's negative
contributions to the alliance in early sessions were
positively correlated with symptom improvement and dynamic

change. Contradicting theoretical assumptions, the

36

empirical data suggests that perhaps some negative behavior
in response to the patient early in therapy is necessary to
challenge the patient's complacency with her/his
characteristic patterns of interaction and to facilitate
change. The level of functioning of the patient seems to
be an important consideration here. Marziali's (1984)
patients were neurotic outpatients with whom a
circumscribed focus could be maintained and Crowder%3
(1972) patients were college students presenting at a
counseling center, both groups likely to be high-
functioning. Negative therapist behavior early in
treatment could be counter-productive with less healthy
patients.

Interaction of patient predisposition with outcome.
There is preliminary evidence for an interactive effect of
patient predispositional variables and the therapeutic
alliance on outcome. The research of Horowitz and his
colleagues (1984) suggested that indicators of the alliance
have different meaning for patients depending upon level of
motivation. Among patients with low motivation, high
patient positive contribution ratings were associated with
better outcome, while high patient negative contribution
ratings were associated with poorer outcomes. Among
patients with high motivation, more positive contributions
to establishing a therapeutic alliance were associated with

poorer outcomes, (while high negative contribution ratings

37

were associated with better outcome. The authors (Horowitz
et.al”.1984) postulated that although patients with low
motivation may become too overwhelmed by negative feelings
toward the therapist to work through the reactions, the
negativity of highly motivated patients toward the
therapist may represent constructive therapeutic work
because feelings toward the therapist are actively being
addressed. The development of a therapeutic alliance may
be sufficient to sustain the involvement of an initially
poorly motivated patient in the therapeutic process.
Horowitz et a1. (1984) suggested that evidence of a
positive alliance on the part of a highly motivated patient
may represent a defense against negative feelings toward
the therapist. The interactional model employed by Horowitz
et al. revealed no influence of therapist variables on
outcome.

Using a traditional correlational approach with VPPS
data from the first three sessions of therapyy O'Malley,
Suh, and Strupp (1983) had found little relationship
between mean or session by session therapist behavior and
outcome. In an innovative approach to the psychotherapy
process research, Suh and O'Malley (1982) classified the
patients into high and IIWV prognosis groups based on
process variables, then cross classified patients according
to actual high and low outcome to create four groups.

Rather than attempt to directly predict a relationship

38
between the prognostic variable and outcome, the authors
employed a failed predictions model, hypothesizing that
cases in which the patients' outcomes were congruent with
their prognoses could be differentiated from those for whom
predictions failed by therapist behavior. When patterns of
change in therapist behavior across the first three
sessions, rather than the ratings of therapists for each
session, were evaluated for each of the four groups,
results emerged which suggest that therapist alliance-
related behaviors do have implications for outcome. For
patients with good prognoses who achieved the predicted
successful outcome, therapists tended to have an initially
positive reaction, with an increase across time in warmth
and exploration. In contrast, for patients with a high
prognosis who failed to achieve the predicted successful
outcome, therapists had an initially negative attitude
which increased across time, while therapist warmth and
exploration decreased. Therapists for low-prognosis
patients who, as predicted, luui poor outcome,
characteristically responded with eu1 initially highly
negative attitude and a decrease in warmth across
treatment. For patients who achieved a successful outcome
despite a poor prognosis, therapists exhibited an increase
in warmth and exploration over the first three sessions.
Suh and.O'Malley also found that the association between

Patient Participation.andcoutcome, which was significant

39
for session three of therapy for a variety of outcome
measures, was strengthened when change scores from session
one through three were used.

Suh and CYMalley (1982) contend that traditional
correlational analyses of psychotherapy process data are
likely'to obscure the influence of therapist variables on
outcome. "Good" psychotherapy candidates are likely to
obtain a successful treatment outcome under most
circumstances and are unlikely to elicit negative reactions
from therapists, so that therapist contributions to the
alliance are likely to have limited impact on outcome
variability with high prognosis patients. Poor
psychotherapy candidates, on the other hand, are much more
likely to be lacking the capacities for achieving high
outcome with minimal assistance and to elicit a negative
therapist response. Thus therapist behavior is postulated
to have significantly more impact on outcome among low
prognosis patients. Because patient variables apparently
account for a much larger proportion of variance in outcome
than therapist variables, the influence CHE therapist
variables on outcome is likely to be obfuscated in
traditional correlational analyses in which high and low
prognosis patients are pooled.

Taking a similar approach to the failed predictions
model of Suh and OUMalley (1982), Foreman and Marmar (1985)

studied the course of the patient's negative contributions

40
to the alliance using the CTAS. Though all six patients
in the study scored high on patient negative contributions
in session 2, half achieved significant improvement in
psychotherapy and half failed to improve. The authors
reported that the negative contributions of the improved
patients decreased substantially through the course of
treatment, while the negative contributions of unimproved
patients remained high throughout therapy. Thus, among a
group of patients whose initial contributions to the
alliance would suggest a poor prognosis for outcome,
improvement in the alliance was associated with better
outcome. The authors also noted a difference in therapist
technique between the two outcome groups. Therapists who
succeeded in promoting an improved alliance and good
outcome made interpretations specifically relatedtxbthe
alliance, while unsuccessful therapists avoided addressing
the poor alliance.
Conclusions from the therapeutic alliance research.

The frequent contradictions in the therapeutic alliance
research literature, partially a consequence of differing
measures and methodologies, also reflect the complexity of
the relationship of the alliance to process and outcome in
psychotherapy. While the convergence of data indicates
that the patientis contributions to the alliance influence
outcome, there is evidence that the implications of the

alliance-related behavior for outcome vary with certain

41
patient predispositional variables and with the timing of
their occurrence in the course of therapy. The empirical
data on therapist contributions to the alliance presents an
even more perplexing picture.‘Fheoretical assumptions about
the importance of therapist contributions to the alliance
in determining outcome have not been well-supported
empirically. However, some characteristics of the
therapeutic alliance research may be operating to obscure
therapist influence. The use of highly experienced
therapists who differ little in their behavior, thus
providing little variance in alliance ratings, may preclude
finding significant associations between therapist
alliance-related behaviors and outcome. The effect of
therapist variables is also likely to be masked by the
traditional approach of pooling data from all patients for
statistical analyses. Unless the therapist is destructive,
therapist contributions may have little effect on the
development of the therapeutic alliance or on outcome with
psychotherapy candidates who begin treatment withaigood
capacity to develop a relationship and to make use of
therapy. Therapist influence is likely to be much greater
on patients who begin treatment with poor prognosis due to
limited ability to establish an alliance. The tendency of
patient factors to account for a significantly higher
proportion of the variance in psychotherapy outcome than

therapist factors would cause therapist influence to be

42
obscured when a conventional correlational approach to
statistical analysis is used.

Thus studying the therapeutic alliance as it relates
to patient predispositional variables may be even more
critical to understanding the influence of therapist
contributions than to understanding the influence of
patient contributions. The complexity of the therapeutic
alliance as revealed by empirical results thus far dictates
a need for a complex approach to future research which
considers the interaction of patient predispositional
variables with the course of patient and therapist
contributions to the alliance across treatment as they

relate to outcome.

Clinical Significance of Outcome Criteria

The evaluation ofgpsychotherapeutic outcome. The need
for multidimensional sources and types of outcome criteria
has been a focal issue in the literature on the evaluation
of psychotherapy outcome (e4L, Bergin & Lambert, 1978;
Luborsky et alq.1971). Strupp & Hadley (1977) presented
an excellent discussion of the issue and proposed a
tripartite model of outcome criteria which has become a
prototype in the field of psychotherapy research. Under
the tripartite model, outcome is assessed from the
perspectives of society, the individual, and the mental
health professional, implying measures of observable

behavior, the individual client's sense of well-being, and

43

patient functioning as compared to theoretical standards
of psychological health.

Relative to the emphasis on the source and content of
outcome criteria, the issue of outcome measurement
methodology has been neglected by most authors. Mintz,
Luborsky, and Cristoph (1979) discussed the advantages and
disadvantages of a number of measurement methodologies
employed by psychotherapy researchers. The most frequently
used outcome measure in research on traditional
psychotherapy, probably because of its simplicity, is a
global rating of success or improvement on a single scale
(Luborsky et a1” 1971). Such global measures, however,
are particularly susceptible to bias and ambiguity as to
what is being rated, which makes generalization from
patient to patient or study to study difficuitn The gain
score from pretreatment to posttreatment, attractive in its
face validity, often presents a some statistical problems:
statistical unreliability and correlation with initial
level of symptomatology due to ceiling effects and
regression to the mean when outcomes of extreme groups are
compared. As test-retest reliability decreases, these
problems are exacerbated. Final adjustment status alone
has also been employed as the criteria of whether outcome
is adequate. A drawback to this criteria is that patients
who began treatment at a high level of functioning and

achieved no statistically reliable change are equated with

44
patients vflu: make large, positive gains in functioning in
order to obtain the final adjustment status (Mintz,
Luborsky, & Christoph, 1979).

One measurement methodology, the residual gain score,
offers the advantage of statistically compensating for the
correlation of amount of change with initial level of
functioning. The individual's simple gain score is
rescaled relative to the mean change made by others who
began treatment at the same level of functioning. The
residual gain score thus reflects the individuaidsichange
relative to the amount of change that would be predicted
based on initial level of functioning. Atdisadvantage to
the residual gain score is the complexity of interpretation
of statistical analyses, because the adjusted scores differ
so greatly from the raw data (Mintz, Luborsky, & Cristoph,
1979).

The difficulty of finding an adequate measurement
methodology for psychotherapy outcome is compounded by the
question of whether the measured benefits of psychotherapy
to a patient are clinically meaningful. Statistically
significant differences in outcome between groups are often
of little practical importance. Traditional statistical
comparisons between outcome for two or more groups of
patients are based on mean scores and provide no data as
to the proportion of patients in each group who improve,

making it difficult to use research results to estimate the

45

likelihood that a specific individual will gain from
psychotherapy (Jacobson, Follette, & Revenstorf, 1984).
There is growing recognition in the field that conventional
statistical tests of significance must be supplemented by
tests of clinical significance and by reports of the
proportion of improved patients if psychotherapy research
is to have practical implications for clinical work (eugq
Hugdahl & 0st, 1981; Kazdin & Wilson, 1978).

Q;ipg;ia_for clinical significance. Jacobson,

 

Follette, and Revenstorf (1984), suggesting that
standardized criteria for clinical significance be adopted,
proposed a two-fold criterion which could be applied across
a variety of clinical problems: the patient's posttest
level of functioning and the statistical reliability of
change. They defined a clinically significant change in
therapy as "when the patient moves from the dysfunctional
to the functional range during the course of therapy on
whatever variable is being used to measure the clinical
problem" (p. 340). The authors specified three possible
ways to operationalize the question of adequate posttest
level of functioning. The most stringent criterion would
require that measures of posttest functioning fall more
than two standard deviations, in the functional direction,
from the mean for the dysfunctional population. Less
stringent criterion would require that posttest functioning

fall within two standard deviations of the mean of the

46

functional population. Jacobson et a1. recommended a third
criterion when there is significant overlap between the
functional and dysfunctional pOpulation. The third
criterion would determine whether the posttest score would
statistically be more likely to place the patient in the
functional or the dysfunctional population. Using the
Jacobson et al. approach, the choice of criterion for
clinical improvement will depend in part upon the
availability of norms for functional and dysfunctional
population and upon the degree of overlap between the two
pOpulation distributions. The authors acknowledged that
return to normal functioning may be too demanding an
outcome criterion for some populations.

In order for change in psychotherapy to be clinically
significant, it must be of large enough magnitude to rule
out the possibility that the improvement from pretest to
posttest was due to chance. Jacobson and his colleagues
(1984) recognized a variety of possible criteria for
statistical reliability of change, but recommended the use
of a reliable change index (RC) calculated by dividing the
pre-post difference score for each patient by standard
error of measurement. The standard error of measurement is
equivalent to the spread of scores that repeated testing
would produce given that no actual change had occurred.
The probability of obtaining an RC exceedingjtlu96 if no

actual change has occurred is less than 5%. The authors

47
acknowledge a disadvantage in RC being dependent upon the
reliability of the change measure, since a small magnitude
of change can produce a large RC if the instrument is
highly reliable. However, the use of the additional
criteria of clinically adequate functioning provides a
check against this problem.

Though there is no consensus yet on criteria for
clinical significance, Jacobson et a1. (1984) provide
conventions based on sound clinical and psychometric
rationale which can be applied to a variety of clinical
problems. Research that, in addition to statistical
differences between group means, reports the proportion of
individuals in a group who achieve an acceptable level of
functioning in therapy may reveal treatment effects which
are obscured by high variability. Unlike average
improvement scores, description of the pmoportion of
improved patients in a group permits estimates of the
probability that a given patient in a clinical setting will
benefit from treatment, Adopting clinically significant
criteria for outcome seems to be a step toward bridging the
gap between psychotherapy research and clinical
application. The movement toward standardized criteria for
clinical significance (will facilitate comparisons of
efficacy of psychotherapy from one area to another and
inhibit the tendency to allow standards for successful

outcome to be eroded by adjustment to the limits of

48

therapeutic technology (Jacobson et al., 1984).

OBJECTIVES AND HYPOTHESES

Statement of Problem and Objectives

Though a positive, collaborative bond between patient
and therapist is widely recognized by psychodynamic
theorists and clinicians as a prerequisite to successful
psychotherapy, empirical support for the relationship
between the therapeutic alliance and outcome has not been
consistent. The dissonance between theory and research
results may stem in part from approaches to research which
sacrifice the complexity of the therapeutic alliance.
There are indications in the literature that comparing
alliance scores averaged across treatment obscures
variations in the course of the alliance during therapy
which influence outcome. Despite preliminary evidence that
alliance-related behavior may have different implications
for outcome depending upon patient predispositional
variables, data analyses are conventionally based on the
entire sample of patients, regardless of patient
predispositional variables. The initial capacity of the
patient for establishing a relationship seems particularly
likely to influence the significance of patient and
therapist alliance-related behavior for outcome. Since

patient factors appear to account for a significantly

49

50
higher proportion of variance in outcome than therapist
factors, the effect of therapist factors is especially
likely to be obscured when all subjects are pooled in a
traditional correlational approach to statistical analysis.

The first objective of this research was to address
the complexity of the therapeutic alliance by studying the
interaction of initial patient prognosis for establishing a
therapeutic relationship, the course of the therapeutic
alliance across treatment, and outcome. Two questions were
of central interest. How do therapist and patient
contributions to the alliance during the course of
treatment differ for cases with originally similar patient
potentials for establishing a relationship but dissimilar
outcomes? How does therapist action influence patient
alliance-related behavior among patients with initially
high and low relationship potentials? IIt was anticipated
that differentiating patients with high and low
relationship potentials would reveal implications of
therapist alliance-related behaviors for the course of
patient contributions to the alliance and for outcome which
have been eclipsed when all patients are pooled.

While a number of research instruments have been
designed to gauge the strength of the therapeutic alliance,
most measures have been applied by only one group of
investigators with a single population. The second purpose

of this study was to assess the validity of one of the more

51

promising therapeutic alliance measures, the TARS
(Marziali, 1984), for use with a heterogeneous sample of
patients and therapists engaged in a wider range of
treatment than in previous studies. The TARS has been
employed only in research on brief psychotherapy with high-
functioning patients, conducted by very experienced
psychoanalytically-oriented psychotherapists. Clients in
this research varied considerably in their degree of
psychopathology, iwith many not sufficiently high-
functioning to be apprOpriate candidates for time-limited
psychotherapy. While psychodynamic psychotherapy was the
most common treatment approach employed by therapists in
this study, a variety of theoretical perspectives were
represented. Therapists were comparable to a typical
community mental health center staff in terms of
experience, ranging from being first year practicum
students in clinical psychology to having several years of
post Masters degree clinical experience. It was hoped that
research employing therapists with a wide range of skills
would prove particularly frutiful, since the failure of
past research to reveal a significant influence of
therapist alliance-related behavior on outcome may in part
be an artifact of the limited range of scores on the
therapist scales of alliance measures.

The questionable clinical relevance of conventional

statistical approaches ‘Uo outcome ineasurement ih

52

psychotherapy research has deservedly been a target of
criticism by practicioners. The third purpose of the
research was to examine the relationship between the
therapeutic alliance and psychotherapy outcome when
outcome, rather than being determined solely by comparisons
between group means, was defined in clinically significant
terms and the prOportion of patients in each group who
benefit significantly was considered. The use of
clinically significant outcome criteria is seen as a step
toward increasing the clinical relevance of empirical work
on the therapeutic alliance.

To accomplish the specified objectives, audiotapes of
psychotherapy sessions were rated for the presence of
positive and negative patient and therapist contributions
to the therapeutic alliance, using the TARS. Patients
were classified as having a high or low potential for
establishing an alliance (Prognosis) based on TARS ratings
of patient contributions to the.alliance during the first
session of treatment. For each patient, a session from the
early, middle and late phase of treatment was also rated on
the TARS for indications of patient and therapist positive
and negative alliance-related behavior. Patients were
classified as achieving clinically significant improvement
(High Outcome), statistically reliable but not clinically
significant improvement (Medium Outcome) or no improvement

(Low Outcome) based on their scores on the Global Pathology

53

Index of the SCL-90-R.

Hypotheses

 

HI: Patient positive and negative contributions to the
therapeutic alliance in the first session of treatment
will be predictive of the patientfs ability to form and
maintain an alliance throughout treatment.

Specific prediction: High prognosis patients will have
higher Patient Positive and lower Patient Negative
Contribution scale scores averaged across early, middle
and late treatment sessions than low prognosis
patients.

52: Patients with an initially poor potential for
developing an alliance will have greater variability in
outcome than patients with a good potential for an
alliance.

Specific predictions: The variance of outcome
classification scores will be greater for low prognosis
than for high prognosis patients.

H3: Among patients who achieve significant change,
positive patient contributions will be greater in late
than in early sessions of therapy.

Specific prediction: There will be a significant phase
of treatment effect for the Patient Positive
Contribution scale among high and medium outcome
patients, with higher ratings on the scale for late

than for early sessions.

54

H4: There will be a significant difference between
patients who achieve clinically significant change and
patients who achieve no reliable change in the course
of patient positive and negative contributions to the
alliance.

Specific prediction: There will be a significant main
effect of outcome, or a significant Phase x Outcome
interaction, for the Patient Positive and Patient
Negative Contribution scale scores.

HS: Within patients with a poor capacity for
establishing a therapeutic alliance, therapist positive
contributions to the alliance will be positively
related to successful outcome; therapist negative
contributions will be negatively related to successful
outcome.

Specific prediction: There will be a significant
outcome effect for the Therapist Positive and Therapist
Negative Contribution scales within ltnv prognosis
patients.

H6: Therapist alliance-related behavior will be more
strongly associated anfli patient alliance-related
behavior for ltnv prognosis patients than for high

prognosis patients.

55
Specific prediction: There will be a larger
correlation between therapist and patient subscales at
each phase of treatment for low prognosis patients than

for high prognosis patients.

METHOD

Participants

Clients. The patients in the research were primarily
working and middle class adults seen in psychotherapy at
the Michigan State University Psychological Clinic who
agreed to participate in the Clinic's psychotherapy
research program. The study included the 32 treatment
cases at least 10 sessions in duration for whom complete
data was available. The number of sessions for each
treatment case range from 14 to 71, with a median of 29
sessions. There were 8 male and 24 female patient-
participants ranging in age from 20 to 57, with a median
age of 29.

Therapists. Therapists for the Psychological Clinic's
psychotherapy research project are graduate students in
clinical psychology, recruited from.the Clinic practicum
students and interns. Informed consent for the
participation of the therapists is obtained at the
beginning of each academic year.

Therapists for the study range in experience from
students in first year practicum to advanced students with
several years of post-Masters degree experience. The range

of experience of therapists is comparableeto the range of

56

57
experience typically found in a community mental health
setting. Though the predominate theoretical orientation of
the therapists is psychodynamic, including psychoanalytic
and interpersonal perspectives, other orientations to

treatment are represented.

Procedure

Clients seeking treatment at the Psychological Clinic
are routinely informed of the Clinic's psychotherapy
research project during their intake interview, and asked
if they would be willing to participate. Informed consent
of the patient-participants is obtained during the intake
interview. After intake but prior to their first meeting
with their therapist, patients agreeing to participate in
the research project completed the SCL-90. The first
session, third session, every fifth session subsequent to
the third, and the last session of therapy were audiotaped.
After termination, the SOL-90 was again completed. Before
therapy began, during the course of treatment, and at
termination, patients completed a number'of<other self—
report measures not relevant to the present study.

For each patient, audiotapes of the first session and
three other sessions spanning the course of treatment were
selected for rating. The first, middle, or last 15
minutes of each session were randomly selected to be rated
on the TARS. All audiotape segments for the study were

recorded on a master tape in random order, identified only

58
by a code number.

Judges were two advanced graduate students in clinical
psychology who met for 20 hours of training and practiced
rating approximately 40 audiotaped psychotherapy segments
not used in the study prior to rating the audiotapes for
the study; There were approximately 12 additional hours of
training during the course of data collection. Each
psychotherapy segment used in the study was rated by both
judges, who were blind to the hypotheses of the study and
the classification of each audiotape segment as to patient

prognosis, phase of treatment, and outcome.

Variables

Prognosis. For each patient, the first, middle or
last 15 minutes of the initial psychotherapy session were
randomly selected to be rated on the TARS by clinical
observers. Clients were classified as having a high or low
prognosis for establishing a good therapeutic alliance on
the basis of their scores on the Patient Positive
Contribution Scale and the Patient Negative Contribution
Scale of the TARS. Each of the 11 items on the positive
scale and the 10 items on the negative scale are rated on
a 6-point "intensity of presence" scale from 0, not
present, to 5, intensely present. In order to weight
positive and negative contributions equally, ratings of
items on each scale were summed, then divided by the number

of items in the scale, to create a mean intensity score for

59

the scale. A Prognosis score was calculated for each
patient by subtracting the mean intensity score for patient
negative contributions from the mean intensity score for
patient positive contributions. There was a possible
range of Prognosis scores from -5 to 5, with a low score
indicating a poor potential for establishing an alliance
and a high score indicating a greater potential. Though
patient alliance-related behavior is undoubtedly influenced
by therapist behavior from the beginning of treatment, it
is assumed that patient behavior in the first session is
reflective of a core capacity for object relatedness and
thus of a potential to establish a therapeutic alliance.

Phase of treatment. The third session of therapy was
selected to represent the early phase of treatment for each
patient. The middle phase of treatment was represented by
a session after 40% but before 60% of treatment has been
completed, as close as possible to the 50% point in
therapy. A session after 80% of treatment had been
completed, as close as possible to the 90% point in
therapy, was chosen to represent the late phase of
treatment.

Qppgpmg. Clients were classified as to successfulness
of outcome based on an application of the criteria for
clinical significance suggested by Jacobson et a1. (1984)
to scores on the Global Pathology Index (GPI) of the SCL-

90-R. Because there is overlap between the dysfunctional

60
and functional population distributions (”1 the Gross
Pathology Index (Derogatis, 1977). clinically significant
change was determined by whether the level of posttest
functioning indicated that the patient was statistically
more likely to be in the functional than the dysfunctional

range.

Table 1

Gross Pathology Index Data Used in Determining Outcome

 

 

Criteria

Symbol Value
x0 = mean for randlomly selected nonpatients .31
x1 = pretreatment mean for study patients 1.45
50 = standard deviation for nonpatients .31
51 = pretreatment standard deviation for study .58

patients
rxx = test-retest reliability of Gross Pathology .84
Index

 

Using the data presented in Table 1, the cutoff point
c, where probabilities of belonging to functional and
dysfunctional populations are equal, can be solved for with
the following formula.

50x1 + Slxo = .31(1.45) + .58(-3l> = .71

c = 50 + 51 .31 + .58

 

Clients with posttest GPI scores less than .71 are

61
statistically more likely to be in the functional than in
the dysfunctional population. Scores greater than .71 are
likely to be drawn from the dysfunctional distribution.
The statistical reliability of change was determined
by a reliable change index (RC) proposed by Jacobson et al.
(1984). RC is equivalent to the difference score divided

by the standard error of measurement:
RC = (x2 - X1)/SE

where x2 = the patient's posttest GPI score, x1 = that
patient's pretest GPI score, and SE = the standard error of
measurement. SE is the standard deviation of scores for
repeated tests which would be expected given that no actual

change had occurred. Based on the data from Table 1,
11/2 = .58 [1 - .8411/2 = .232

If RC is greater than t 1.96, the probability that real
change has not occurred is less than 5%.

Applying the above criteria for clinically significant
and statistically reliable outcome, patients for the
proposed study were classified into three outcome groups.
High outcome patients had achieved a statistically reliable
change (RC > 1.96) and a posttreatment level of adjustment
statistically more likely to place them in the functional
than in the dysfunctional population (GPI (.71). Medium

outcome patients had achieved a statistically reliable

62
change, but at posttest were still functioningan:a.level
which made them statistically more likely to be in the
dysfunctional population (GPI > .71). Low outcome patients
had.failedtx>achieve statistically reliable improvement
during therapy. There were 8 high outcome patients, 11
medium outcome patients, and 13 low outcome patients. With
the exception of one patient in the low outcome group with
statistically reliable deterioration, all patients had
pretreatment Gross Pathology Index scores statistically
more likely to be drawn from the dysfunctional population.

For the correlational analyses, the three outcome
classifications were assigned ordinal values corresponding
to their degree of improvement. Low outcome was assigned
the value 1, medium outcome the value 2, and high outcome
the value 3.

Therapeutic alliance. Clinical observer ratings of
psychotherapy audiotapes on the TARS were used to quantify
patient and therapist behavior congruent with descriptions
in the psychodynamic literature of positive and negative
contributions to the therapeutic alliance. Patient
positive contributions, patient negative contributions,
therapist positive contributions, and therapist negative
contributions were considered as separate dimensions of the

therapeutic alliance.

63

Instruments

 

SCL-90-R. The SCL-90-R is a self-report inventory

 

 

designed to reflect the current psychological symptom
status of psychiatric and medical patients. The 90 items
of the inventory are rated on a 5-point scale from "not at
all" to "extremely," indicating the degree to which they
have distressed the respondent. For the proposed study,
participants were instructed to rate the problems and
complaints with regard to the distress they had experienced
in the past couple of weeks, including the day of
administration. The items of the SCL-90-R contribute to
nine primary symptom dimensions: somatization, obsessive-
compulsive, interpersonal sensitivity, depression, anxiety,
hostility, phobic anxiety, paranoid ideation, and
psychoticism. Seven items of the inventory are not
included in the symptom dimensions but do contribute to the
three global indices of distress. The Global Severity
Index (GSI), the best single measure of current level of
psychological disturbance from the SCL-90-R, combines data
on the number of symptoms endorsed and the intensity of
distress experienced. The individual's style of response
is reflected in the Positive Symptom Distress Index (PSDI),
a measure of perceived intensity of distress corrected for
the number of symptoms. The total number of symptoms
which a respondent acknowledges experiencing is reflected

in the Positive Symptom Total (PSI).

64

SCL-90-R Reliability;.Adequate internal consistency

 

and test-retest reliability has been reported for the nine
primary symptom dimensions of the SCL-90-R. With a
population of symptomatic volunteers, Derogatis, Rickels,
and Rock (1976) found that coefficient alpha ranged
from .77 for the Psychoticism Scale to high of .90 for
Depression. Measures of test-retest reliability for a
heterogeneous group of psychiatric outpatients over a one
week period ranged from .78 to .90, an apprOpriate level of
reliability for syndromes of psychoPathological symptoms,
which conceptually lie between the constancy of personality
traits and the instability of mood (Derogatis, 1977L
Some evidence is available for the reliability, or
generalizability, of the primary symptom dimensions across
demographic variables. ‘Tests of factorial invariance of
the SCL-90-R have shown high levels of agreement across sex
in the structural definitions of eight of the nine symptoms
dimensions. A.moderate levelcflfagreement between males
and females was found on the composition of the dimension
of Paranoid Ideation (Derogatis & Cleary, 1977). Factorial
invariance has been demonstrated across social class and
psychiatric diagnosis for Depression, Somatization,
Hostility, Phobic Anxiety; and Obsessive-compulsive, the
dimensions which were carried over from the forerunner of
the SCL-90-R (Derogatis, Lipmen, Covi, & Rickels, 1971,

1972).

65

SCL-90-R Validity. The validity of an instrument
rests on whether the instrument succeeds in measuring what
it intends to measure. One type of validity, criterion-
related validity, is determined by the degree of
correspondence between an instrument and some conceptually
important criterion external to the instrument itself,
whether the criterion is applied prior to, concurrently
with, or after the administration of the instrument. A
study with 119 symptomatic volunteers demonstrated a high
degree of concurrent validity for the SCL-90 (Derogatis,
Rickels & Rock, 1976). In a correlational analysis of the
relationship between the SCL-90 and the clinical, content,
and cluster scales of the MMPI, the authors found that each
dimension of the SCL-90 was most highly correlated with an
MMPI scale reflecting a similar construct, except
Obsessive-compulsive, for which there was no like MMPI
scale. An investigation of the concurrent validity of the
SCL-90 with the Middlesex Hospital Questionnaire produced
correlations ranging from .36 to .74 between symptom
dimensions tapping similar constructs, with a correlation
of .92 between the global symptom scores of the two
instruments (Boleloucky & Horvath, 1974).

A number of investigations have demonstrated the
sensitivity of symptom patterns on the SCL-90 in
discriminating various clinical groups. Studying a

population of participants in a methadone maintenance

66

program, Weissman, Slobetz, Prusoff, Mezritz, andlhmvard
(1976) found the symptom profile of the SCL-90 to
discriminate clinically depressed from nondepressed
patients. Patterns of response on the "90" were reported
to discriminate between drug and placebo groups in research
on the relationship between hostility and marijuana
(Salzman, Van der Kolk, & Shader, 1976). Abelhoff and
Derogatis (1977) have shown breast cancer patients to have
a unique response pattern on the "90" which distinguishes
them from women with other types of cancer. In another
investigation with cancer patients (Craig & Abelhoff,
1974), the SOL-90 was found to be a useful screening
instrument for psychological distress among oncology
patients. One quarter of the oncology patients in the
sample produced SCL-90 profiles identical to the profiles
of psychiatric emergency patients.

The criterion-related studies cited above contribute
support for the construct validity of the SCL-90-R,
indicating ea substantial correlation between the
measurement as it is operationalized and the construct
which it is hypothesized to measure. Another aspect to the
assessment of construct validity is the investigation of
the extent to which the theoretical internal structure of
an instrument can be validated by empirically-based
analyses (Derogatis, 1977L. Factor analysiscnfthe SCL-

90-R data of 1,002 psychiatric outpatients, using both

67
procrustes and varimax procedures, revealedau1excellent
correspondence between theoretical and empirical structure
of the symptom dimensions of the SCL-90-R (Derogatis &
Cleary, 1977).

Therapeutic Alliance Rating Scale. The TARS is a 42-

 

item instrument.designed.to assess positive and negative
contributions of patient and therapist to the affective and
attitudinal (cognitive-motivational) aspects of the
therapeutic relationship. Though based in part on items
drawn from the Penn Helping Alliance methods (Luborsky,
1976; Luborsky et alu,l983) the‘Vanderbilt.Therapeutic
Alliance Scale (Hartley, 1978), and the Vanderbilt
Psychotherapy Process Scale (Gomez-Schwartz, 1978), the
TARS was intended.to exclude items pertaining to action,
technique, or specific response. The TARS to be used in
the proposed research (Marziali, 1984) is very similar in
format and content to the original Therapeutic Alliance
Scale developed.bylnarziali et al.(1981) and researched
further by Marmar and associates (Foreman & Marmar, 1985;
Horowitz et a1. 1984), though the instrument has been
partially reorganized and some items reformulated. There
are three parallel forms of the instrument, designed to be
used by therapists, patients, and trained clinical
observers of segments of psychotherapy sessions. Only the
clinical judge version of the scale will be employed in

this study.

68

The TARS has four subscales: the Patient Positive
Contribution Scale and the Therapist Positive Contribution
Scale, each with 11 items, and the Patient Negative
Contribution Scale and the Therapist Negative Contribution
Scale each with 10 items. Each item is rated on a 6-point
"intensity of presence" scale, from 0, not present, to S,
intensely present. For this study, in order to make
negative and positive scale scores comparable, a mean
intensity of presence score was calculated for each of the
four TARS subscales by summing the item ratings for the
scale, then dividing by the number of items in the scale.
Like the intensity of presence scale for each item, the
mean intensity score could range from 0, not present, to 5,
intensely present. The mean intensity scores for each
subscale will be used in the data analysis.

TARS Reliability. For TARS ratings by clinical

 

observers of 42 psychotherapy cases at six points in
treatment, internal consistency of the four subscales was

evaluated using Cronbach's alpha. For therapist positive

items, alpha = .86; for therapist negative items,
alpha = .87; for patient positive items, alpha = .93;
for patient negative items, alpha = .88. A one-way

analysis of variance in which between-judges variance was
included in the error term was used to determine
intererater reliability. The intraclass correlation

coefficients for ratings (n3 therapist positive and

69
negative items subscales ranged from .61 to .77, and for
ratings of patient positive and negative item
subscales .60 to .83 (Marmar, Marziali, Horowitz, & Weiss,
1985; Marziali, 1984).

TARS validity; The available data suggests that the
TARS as a whole, and the Patient Positive Contribution
scale in particular, has reasonably good criterion-oriented
and construct validity. Ratings by clinical judges of the
patient's contributions to the alliance are strongly
correlated with the patient contributions as perceived by
members of the dyad. Judge's ratings of patient positive
contributions correlated .56 (p < .001) (with patient
ratings and.59 (p (.001) with therapist ratings. Judge
ratings of patient negative contributions correlated .44 (p
< .01) with patient ratings and .50 (p < .001) with
therapist ratings. Correlations of therapist positive
contributions as rated by judges with ratings by members of
the therapeutic dyad were weaker but still significant
(5 = .37, p < .01 and g = .32, p< .05). There was little
consensus between patient, therapist, and clinical judge as
to the therapistfs negative contribution to the alliance
(average _r;=.06) (Marziali, 1984).

Patient's pre-therapy ratings of social adjustment
were significantly correlated va31 clinical observer
ratings of patient negative contributions to the alliance

(g ranged from .34 to .51). Convergent findings by

70

Luborsky (1983) and Moras and Strupp (1982) of significant
associations between patient social adjustment and patient
contributions to the alliance contribute support for the
concurrent validity of the TARS, Penn Helping Alliance
method, and the Patient Involvement Scale of the VPPS. The
association of a history of positive interpersonal
relationships with the ability to contribute to the
therapeutic relationship is certainly in keeping with the
theoretical importance of the patient's capacity for object
relations for the development of an alliance.

Mean Patient Positive Contribution, Patient Negative
Contribution, and Therapist Positive Contribution scores as
rated by patients, therapists and clinical observers have
been shown to predict a variety of outcome measures. The
three subscales have been correlated with patient and
therapist evaluations of outcome, with the exception of the
observer-rated therapist positive contributions and
therapist evaluation of outcome. Positive contributions to
the alliance by therapist and patient as rated by the
members of the dyad have been demonstrated to predict
symptom level at outcome, though observer ratings have
not. Observer ratings of patient contributions were
predictive of clinical evaluation of dynamic outcome. As
early in treatment as sessions one and three, patient and
therapist positive contributions to the therapeutic

alliance have been demonstrated to correlate with outcome.

71

Some question about the construct validity of the Therapist
Negative contribution scale is raised by the finding of a
positive correlation between judges' ratings on the
subscale in sessions one and three and symptomatic and
dynamic improvement. An alternate explanation is that the
finding is an artifact of the large number of correlations
in the study (Marmar et al., 1985; Marziali, 1984).

Correlational analyses to determine the independence
of subscales of the TARS provided some support for the
theoretical assumption that negative and positive
contributions t1) the alliance reflect separate dimensions
rather than opposite ends of one continuum, lWith the
exception of ratings by clinical judges of patient positive
and negative contributions (3 = -u78, 13 < .001L
Correlations between patient positive contribution ratings
and patient negative contribution ratings by patients and
therapists were significantly lower (3 = -.45 and -.60).
Therapist positive and negative contributions were most
clearly seen as separate dimensions from all three rating
perspectives(;;=-u06'U3-.38)(Marziali,l984).

A principal components analysis with varimax rotations
was carried out to examine the factor structure of the
TARS. It should be noted that the consideration given the
results of this analysis must be tempered by recognition
that there was an inadequate number of subjects (42) in

proportion to the scale items (42) to justify this

72
approach. The analysis was repeated for each of six
sessions across therapy using all the scale items.
Supporting the theoretical construction of the scales,
factor analysis revealed that the negative items were not
inverse equivalents of the positive items. Regardless of
whether the therapeutic alliance was rated by patients,
therapists or clinical observers, two factors consistently
emerged. Factor I consisted of six patient and eight
therapist positive items, including indicators of the
patient feeling helped, hopeful, and willing to examine
her/his behavior and the therapist conveying hopefulness,
support, and involvement in mutual work with the patient.
Factor II, consisting of six patient and five therapist
negative items, included indicators of patient anger,
avoidance, and resistance, and therapist criticism,

impatience, and insensitivity to patient wishes.

RESULTS

Design and Overview of Statistical Analyses

A 2 (Patient Prognosis) x 3 (Phase of Treatment, a
repeated measure) x 3 (Outcome) factorial design was
employed in the study. Data for each TARS subscale was
analyzed independently, based on the theoretical rationale
for the construction of the measure and on findings
reported by past researchers. Marziali (1984) reported
fairly low intercorrelations between the Therapist Positive
Contribution scale and the Therapist Negative Contribution
scale (3 = -.06 to —.38) when rated by patient, therapist
and external judges, indicating that the positive and
negative items tap separate dimensions of the alliance.
Intercorrelations between the Patient Positive Contribution
scale and the Patient Negative Contribution scale were
higher (_1‘_ = -.45 to -.78), suggesting that the two patient
subscales might be collapsed into a single scale. Use of
the two patient subscales, however, is justified by the
results of principal component analyses that produced
separate patient positive and patient negative factors, as
well as separate therapist positive and therapist negative
factors (Marmar et al., in press). Independent analyses of

patient positive and patient negative contributions will

73

74
also facilitate comparisons with past research,imiwhich
results have most frequently been presented separately for
the two dimensions.

Analysis of variance was the predominate statistical
approach used to analyze the data. Because the univariate
approach to the analysis of the repeated measure design is
somewhat more powerful than the multivariate approach,
especially with small sample sizes, a univariate analysis
was employed when the necessary assumptions were met. The
univariate analysis for repeated measures requires that
correlations of the dependent variable at each combination
of the within subject factors be equal, and that the
variances of the dependent variable be equal for all factor
combinations. Bartlett's test of sphericity and the Emax
test were use to test these assumptions. Inclusion of a
between subjects factor in the analysis requires that the
variance-covariance matrices for the transformed variable
for a particular effect be equal for all levels of between
subject factors. Box's M test was used to test this
assumption (Hull & Nie, 1981).

If any of these assumptions appeared to be violated, a
multivariate test of significance, Wilks Lambda, was used
to test for within-subjects factor effects. The
multivariate approach makes no assumptions about the

characteristics of the variance-covariance matrix.

Wilks lambda is based on functions of the eigenvalues

75

. -1
of the matr1x §h§e

, where §h is the matrix of the sum of
squares and cross products for the hypothesis and §e is the
matrix of the sum of squares and cross products for the
error. An eigenvalue is a measure of the relative
importance of the discriminant functions, which are linear
combinations of variables expected to differentiate one
group of subjects from another. Wilks lambda tests for the
statistical significance of the discriminating information
not already accounted for by the earlier function. The
larger the lambda, the less information not accounted for.
Wilks lambda can be transformed into an approximate 5
statistic, with the degrees of freedom depending upon the
degrees of freedom for the hypothesis, the degrees of
freedom for error, and the minimum number of dependent
variables (Hull & Nie, 1981; Klecka, 1970).

Results related to the hypotheses will be presented
first, followed by a presentation of other findings.
Results significant at the .05 level will be reported. For

the analyses of variance, tests of simple effects will be

reported where appropriate.

Classification of Supjects

Subjects were classified according to prognosis for
forming and maintaining a therapeutic alliance based on the
Prognosis score, calculated by subtracting the mean
intensity rating on the Patient Negative Contribution scale

from the mean intensity rating on the Patient Positive

76

Contribution scale for the first session of treatment.
There was a possible range of Prognosis scores from -5 to
5, with. low scores indicating a poorer potential for
establishing an alliance and high scores indicating a
greater potential. Patient scores ranged from -.1 to 2.5,
with a mean of 1.3, a standard deviation of .7, and a
median of 1.4. Because of a roughly bi-modal distribution,
with prognosis scores clustering around 1.0 to 1.1 and
around 1.4 to 1.5, a mean split was used as the cut-off for
placement in the high or low prognosis group. It is not
possible to compare this distribution of scores to past
findings, since the results of previous research with the
TARS and the closely related CTAS have been presented only
in terms of correlations.

Classification of subjects according to prognosis and
outcome resulted in six groups. There were 9 subjects in
the low prognosis-low outcome group: 3 subjects in the low
prognosis-medium outcome group; 2 subjects in the low
prognosis-high outcome group; 4 subjects in the high
prognosis-low outcome group; 8 subjects in the high
prognosis-medium outcome group; and 6 subjects in the high

prognosis-high outcome group.

Reliability

 

All study segments were rated by both judges, with the
exception of two segments on which ratings were missing for

one judge. The subscale score, rather than the subscale

77

item, was considered to be the unit of reliability, since
data analysis was based on the subscale scores. Interrater
reliability was calculated on all study segments with
complete data. Because the same two judges rated all
segments, and the mean of the judges' ratings was used for
the data analysis, judges were considered to be fixed
effects. Results of the data analyses were not affected
by judge mean differences. A two-way mixed effects
analysis of variance was used to obtain the intraclass
correlation coefficient, since judges were considered as
fixed effects. The form of intraclass correlation
coefficient obtained from this analysis is equivalent to
Cronbach's (1951) alpha. The reliability index is
interpreted in this case as a measure of rater consistency
rather than rater agreement (Shrout & Fleiss, 1979).

The average intraclass correlations coefficient for
the mean rating of each subscale was .59 for the Patient
Positive Contribution scale,.70 fortfluaPatient Negative
Contribution scale, .46 for the Therapist Positive
Contribution scale, and .57 for the Therapist Negative
Contribution scale. The average reliabilities for the
Therapist Positive and Therapist Negative Contribution
scales were attenuated by the low intraclass correlation
coefficients for the middle phase of treatment. A Judge x
Patient interaction seems a probable explanation for the

low coefficients, but in the absence of repeated ratings of

78
each patient by each judge at each phase of treatment, the
interaction components and the error components cannot be
estimated separately from the analysis of variance.
Limited variance in the rating matrix contributed to the
difficulty in achieving adequate interrater reliability for

the Therapist Positive Contribution scale.

Hypotheses Testing Analyses

Distribution of patient contribution scores. In

 

Tables 2&nui3, the mean intensity scores for the Patient
Postive Contribution scale and Patient Negative
Contribution Scale within outcome and prognosis group are
presented for each phase of treatment. The reader will be
asked to refer to these tables as the results of the
analyses of the patient subscales are discussed.

Hypothesis one. Hypothesis 1 stated that the positive
and negative contributions to the therapeutic alliance made
by the patient in the first session of treatment would be
predictive of the patientds ability to form and maintain a
therapeutic alliance throughout treatment. The data
supported this hypothesis. A two (Prognosis) x 3 (Phase, a
repeated measure) x 3 (Outcome) analysis of variance
produced significant main effects of prognosis for the
Patient Positive Contributions scale, 3(1, 26) = 7.26,
p <.02, and fortfluaPatient.Negative Contribution scale,
5(1, 26) = 6.24, p < .02. Patients who showed a low

prognosis for developing a therapeutic alliance in the

79

Table 2

Mean Intensity Scores for the Patient Positive Contribution

 

 

 

 

Scale
Phase of Treatment
Outcome Early Middle Late
Low prognosis patients
Low
M 2.53 2.62 2.62
SQ 0.59 0.31 0.50
Medium
M 2.21 2.26 2.76
SQ 0.49 0.45 0.17
High
M 2.43 2.61 2.54
SD 0.61 0.42 0.06
High prognosis patients
Low
M 2.95 3.01 2.65
SQ 0.41 0.39 0.60
Medium
M 2.74 2.87 2.82
SE 0.38 0.41 0.45
High
M 2.74 2.82 3.07
SQ 0.36 0.29 0.28

 

80
Table 3

Mean Intensity Scores for the Patient Negative Contribution

 

 

 

 

Scale
Phase of Treatment
Outcome p Early Middle Late
Low prognosis patients
Low 9
M 1.34 1.51 1 66
S9 0 20 0.33 0 70
Medium 3
M 1.53 2.22 1 65
S2 0 33 0.50 0 33
High 2
M 1.85 2.32 1 70
SD 0 92 1.31 0 35
High prognosis patients
Low 4
M 1.14 1.22 1.57
SQ 0.17 ' 0.36 0.56
Medium 8
M 1.41 1.26 1.61
SQ 0.53 0.27 0.71
High 6
M 1.24 1.32 1.15
§_ 0.26 0.28 0.28

 

81

first session of therapy made more negative contributions
and fewer positive contributions to the alliance throughout
treatment than patients with initially high prognoses for
develOping an alliance. Results of the 2 x 3 x 3 analysis
of variance for the Patient Positive and Negative
Contribution scales are presented in Tables 4 and 5. A
multivariate approach was used to test the within subjects
effects for the Patient Negative Contribution scale data
because the assumptions of compound symmetry were violated.
The univariate approach was used for the Patient Positive
Contribution Scale data.

Hypothesis Two. Hypothesis 2 stated that patients

 

with an initially poor prognosis for developing an alliance
would have greater variability in outcome than patients
with a high prognosis. An E-test failed to provide
evidence that the variance in outcome classification was

2 = .575) and low

significantly different for high (g
prognosis (p2 = .577) patients, §(13, 17)= 1.00, n.s. F-
tests comparing the variance in outcome as measured by
posttreatment Gross Pathology Index scores on the SCL-90-R
also failed to reach significance, §(13, 17) = .95, n.s.
(High prognosis, §2 = .104; low prognosis, s2 = .099).
There was a trend toward greater variance in the pre- to
posttreatment difference scores on the Gross Pathology
Index for low prognosis patients (§2==.523) than for high

prognosis patients (§2==.24IJ, {(13, 17) =2.17, p:<.10.

Table 4

Summary of 2 (Prognosis) x 3

(Outcome) x 3

Results for the Patient Positive Contribution Scale

(Phase) Manova

 

 

Source of variance g; Mean square 3 p
Prognosis 1 2.176 7.26 .02
Outcome 2 .110 (1.00 .70
Prognosis by outcome 2 .028 (1.00 .91
Error between 26 .300

Phase 2 .142 (1.00 .32
Prognosis by phase 2 .035 (1.00 .75
Outcome by phase 4 .150 1.22 .32
Prognosis by outcome by 4 .102 (1.00 .51

phase
Error within 52 .123

 

*

p<.05

.‘l'lll. Illlll‘lill‘l

 

 

 

83

Table 5
Summary of 2 (Prognosis) x 3 (Outcome) x 3 (Phase) Manova

Results for the Patient Negative Contribution Scale

 

*

Between subjects factor effects

 

Source of Variance df Mean square E p
Prognosis 1 2.074 6.24 .02*
Outcome 2 .437 1.31 .29
Prognosis by outcome 2 .404 1.22 .31
Error between 26 .332

 

Multivariate test for within-subjects factor effects

 

Effect Wilks Hypothesis Error 3 p
Lambda df g;
Phase .844 2 25 2.32 .12
Prognosis by phase .838 2 25 2.42 .11
Outcome by phase .840 4 50 1.14 .35
Prognosis by outcome .870 4 50 <1.00 .47
by phase

 

p < .05

84
While patients with an initially poor prognosis for
developing an alliance showed some tendency to be more
varied than high prognosis patients in the amount of change
they achieved during treatment, they were no more varied in
their final outcome.

Hypothesis Three. Hypothesis 3 stated that, among

 

patients who achieve a significant change in psychotherapy,
positive patient contributions will be greater in late than
in early sessions of treatment. Hypothesis 3 was partially
supported. Though the main effect of phase of treatment
for the pooled patient positive data of high and medium
outcome patients failed to reach significance, {(2,136) =
2.71, p < .10, a contrast between the early and late phases
of treatment produced a significant difference, 5(1, 18) =
5.33, p <.05. Patients who achieved a statistically
reliable change showed a significant increase from early in
treatment (M = 2.9) to late in treatment (M = 3.2) in the
intensity of presence of their positive contributions to
the therapeutic alliance.

Hypothesis Four. Hypothesis 4 stated that there would
be a significant difference in the course of patient
positive and negative contributions to the therapeutic
alliance between patients who achieve clinically
significant change and patients who achieve no reliable
change. No support for Hypothesis 4 was found. The 2 (Low

versus High Outcome) x 3 (Phase) analysis of variance

85

produced no significant Outcome x Phase interaction, 5(2,
38) = 1.08, n.s., and no main effect of outcome, E (1, 19)
< L0, n.su for the Patient Positive Contribution Scale.
No Outcome x Phase interaction, _F_'_(2, 18) = 1.49, n.s., and
no main effect of outcome, F(1, 19) = 1.11, n.su (was
obtained for the Patient Negative Contribution scale.
Patients who achieved a statistically reliable, clinically
significant change during psychotherapy and patients who
achieved no reliable change did not differ in their
positive and negative contributions to the therapeutic
alliance during the course of treatment. The Outcome x
Phase analysis of variance results for the Patient Positive
and Negative Contribution scales are presented in full in
Tables A-1 and A-2 of the Appendix.

Distribution of therapist contribution scores. In

 

Tables 6 and 7, the mean intensity scores for the Therapist
Patient Postive Contribution scale and Therapist Negative
Contribution Scale within outcome and prognosis group are
presented for each phase of treatment. The reader will be
asked to refer to these tables as the results of the
analyses of the therapist subscales data are discussed.

Hypothesis Five. Hypothesis 5 stated that within

 

patients with a: poor capacity for establishing a
therapeutic alliance, ‘therapist positive contributions to
the alliance would be positively related to successful

outcome; therapist negative contributions would be

86
Table 6
Mean Intensity Scores for the Therapist Positive

Contribution Scale

 

Phase of Treatment

 

Outcome p Early Middle Late

 

Low prognosis patients

 

Low 9
M 2.37 2.46 2.28
SD 0.38 0.28 0.45
Medium 3
M 2.23 2.64 2.41
SQ 0.58 0.12 0.57
High 2
M 2.27 2.66 2.61
SD 0.13 0.16 0.03
High prognosis patients
Low 4
M 2.55 2.72 2.37
SQ 0.16 0.09 0.38
Medium 8
M 2.39 2.58 2.51
SD 0.38 0.34 0.39
High 6
M 2.42 2.56 2.52
SD 0.57 0.48 0.29

 

87
Table 7
Mean Intensity Scores for the Therapist Negative

Contribution Scale

 

 

Phase of Treatment

 

Outcome p Early Middle Late

 

Low prognosis patients

 

Low 9
M 1.21 1.22 1.77
SQ 0.28 0.24 0.73
Medium 3
M 1.12 1.53 1.63
SQ 0.32 0.63 0.79
High 2
M 2.02 1.62 1.85
SQ 0.32 0.18 1.41
High prognosis patients
Low 4
M 0.91 0.95 1.55
SQ 0.11 0.17 0.49
Medium 8
M 1.45 1.20 1.12
SQ 0.72 0.39 0.27
High 6
M 1.22 1.31 1.02
SD 0.35 0.32 0.07

 

88

negatively related to successful outcome. No support for
Hypothesis 5 was obtained. Because of the small number of
low prognosis patients Q1==14), the data for mediunl(p =
3) and high outcome patients (p = 2) was pooled to form a
reliable change group for this analysis. Because the
assumption of compound symmetry was violated, the
multivariate approach was used in analyzing the within
subject effects for the Therapist Negative Contribution
scale. The 2 (No Reliable Change versus Reliable Change) x
3 (Phase) analyses of variance produced no significant
main effect of reliable change and no interactive effect of
reliable change with phase of treatment for therapist
positive (Reliable change: _F_(1, 12) < 1.0, n.s.; Reliable
Change x Phase: F(2, 24) = 1H3, ILS.) or therapist
negative contributions (Reliable change: 3(1, 12) = 1.16,
ILS.; Reliable change x Phase: §(2, 11) < 1.0, n.sJ.
The analysis of variance results are presented in full in
Tables A-3 and A-4 of the Appendix.

Hypothesis Six. Hypothesis 6 stated that therapist

 

alliance-related behavior would be more strongly associated
with patient alliance-related behavior for low prognosis
patients than for high prognosis patients. No support was
obtained for Hypothesis 6. For each prognosis group,
Pearson product-moment correlation coefficients were
calculated for each pair of subscale scores for each phase

of treatment. Fisher's Z-transformations of the

89
product-moment correlations coefficients were then used to
test the hypothesis that the relationship between patient
and therapist subscales at each point in time was stronger
for low prognosis patients than for high prognosis
patients. None of the correlations between patient and
therapist subscales were significantly higher for low

prognosis patients than for high prognosis patients.

Prognosis Measure

 

Prognosis for the alliance. Data for each TARS

 

subscale was subjected to a 2 (Prognosis) x 3 (Outcome) x 3
(Phase of Treatment) analysis of variance. Results of the
analyses for the Patient Positive Contribution scale are
presented in Table 4, the Patient Negative Contribution
scale in Table 5, the Therapist Positive Contribution scale
in Table 8, and the Therapist Negative Contribution scale
in Table 9. The validity of the Prognosis score as a
predictor of the patient's ability to establish and
maintain a therapeutic alliance throughout treatment was
supported by the finding of a main effect of prognosis on
patient positive and patient negative contributions to the
therapeutic alliance across treatment, as<discussed with
regardianypothesis.L. Patients whose contributions to
the alliance in the first session of treatment placed them
in the high prognosis group made significantly more
positive contributions and fewer negative contributions to

the alliance than low prognosis patients throughout

90

Table 8

Summary of 2 (Prognosis) x 3 (Outcome) x 3

Results for the Therapist Positive Contribution Scale

(Phase) Manova

 

 

Source of Variance d; Mean square p
Prognosis 1 .246 1.05 .32
Outcome 2 .010 (1.00 96
Prognosis by outcome 2 .053 (1.00 .80
Error between 26 .235

Phase 2 .297 2.95 .06
Prognosis by phase 2 .005 (1.00 .95
Outcome by phase 4 .079 (1.00 54
Prognosis by outcome by 4 .026 (1.00 .91

phase
Error within 52 .101

 

91
Table 9
Summary of 2 (Prognosis) x 3 (Outcome) x 3 (Phase) Manova

Results for the Therapist Negative Contribution Scale

 

Between subjects factor effects

 

Source of variance g; Mean square 3 p
prognosis 1 1.588 6.06 .02*
Outcome 2 .208 (1.00 .46
Prognosis by outcome 2 .330 1.26 .30
Error between 26 .262

 

Multivariate tests for within-subjects factor effects

 

Effect Wilks Hypothesis Error 3 p
Lambda d; d;
Phase .925 2 25 1.02 .38
Prognosis by phase .837 2 25 2.43 .11
Outcome by phase .822 4 50 1.29 .29
Prognosis by outcome .757 4 50 1.87 .13
by phase

 

*

p<.05

92
treatment. A main effect of prognosis was also obtained
for the Therapist Negative Contribution scale (EU, 26) =
6.06, p ( .02) , with therapists showing a greater tendency
to make negative contributions to the alliance when working

with patients with an initially low potential for

establishing a relationship (M = 1.5) than when working
with patients with a high potential (M = 1.2) for
establishing a: relationship. Therapist positive

contributions to the alliance appeared to be uninfluenced
by the patient's potential for establishing a: working
relationship.

Prognosis and symptom level. When the dual criteria
for clinically significant improvement in reported symptom
level were applied. with. low prognosis patients,
approximately 14% were found to meet both criteria,
achieving a statistically reliable change in symptom level
which placed them in the normal range of functioning; 21%
achieved statistically reliable change but remained outside
the normal range; 64% failed to achieve a statistically
reliable change. Among high prognosis patients,
approximately 33% achieved clinically significant
improvement; 44% achieved a statistically reliable change
but failed to reach a normal level of functioning; 22%
achieved no reliable change. A chi-square test of the
relatedness between time prognosis classification and

outcome failed to reach the criteria for statistical

93
significance , M2 = 5.77, p < .06.

When the relationship between the Prognosis score and
outcome was assessed.by'alconventional one-tailed p-test
between means for the two prognosis groups on pre- to
posttreatment difference scores on the SCL—90-R Gross
Pathology Index, a significant difference was obtained, p
(30) = 2.03, p < .05. The decrease in the level of
reported symptoms from pue- to posttreatment was
significantly greater for high prognosis (M = 0.77)
patients than for low prognosis (M = 0.34) patients.
However, correlational tests revealed rub significant
relationship between the continuous variable of Prognosis
score and outcome classification (g = .34, p (.23) or
change in symptomatology as indicated by the SCL-90-R (g
= .11, p < .28). No associations were found between the
Patient Positive Contribution scale scores or the Patient
Negative Contribution Scale scores for session 1 and
session 3 and outcome classification or the SCL-90-R change
score. The 2 (Prognosis) x 3 (Phase of Treatment) x 3
(Outcome) analyses of variance, presented.in Tables 4, 5,
8, and 9, showed no interactive effect of prognosis and
outcome for any TARS subscale.

Patient potential for establishing a therapeutic
alliance as demonstrated in the first session of therapy
was unrelated to pretreatment symptom level, 5 = .07,

p < .35.

94
Outcome

Patient improvement. Before treatment, this sample

 

had levels of symptoms comparable with other outpatient
samples in psychotherapy research. The mean gross
pathology score on the SCL-90-R for this sample at intake
(M = 1.45, S9 = .58)iwas somewhat higher than the mean
gross pathology score (M = 1.25, S2 = .39) reported by
Derogatis et al. (1976) for a sample of symptomatic
outpatients in a validation study of the SCL-90-R.

Because a disparity in variances precluded the use of
a parametric test between pre- and posttreatment gross
pathology index means, the Wilcoxon.matched-pairs ranked
signs test was used to determine whether psychotherapy was
effective in reducing the level of symptoms for the sample
as a whole. The Wilcoxon test revealed a significant
improvement in symptom level as measured by the SCL-90-R
gross pathology index, 5 = -4.10, p < .001. After
treatment, there was significantly less variance in the
level of symptoms than before treatment, 5(31, 31) = 3.29,
p < .001.

Of the 32 patients in the study, 59% achieved
statistically reliable change, as defined earlier. Only
25% of the sample met both criteria for a clinically
significant improvement, achieving reliable change and a
posttreatment level of' symptomatology statistically more

likely 1x3 place them 1J1 the functional than the

95

dysfunctional range. Two patients were assigned to the low
outcome group despite posttreatment SCL-90-R scores which
fell in the functional distribution, because they failed to
achieve a statistically reliable change.

The level of symptoms initially reported by the
patient was highly predictive of outcome. The greater the
pretreatment level of symptoms endorsed, the more decrease

in symptoms the patient was likely to achieve by

termination, g = .86, p < .001, and the higher the
patient's outcome classification was likely to be, p = .42,
p < .01.

Interaction with other factors. 2 (Prognosis) x 3
(Outcome) x 3 (Phase) analyses of variance on data for all
subjects, presented in Tables 4,5L.8, and.9, produced no
main or interactive effect of outcome for any subscale. 'mo
assess differences in the course of contributions to the
alliance between extreme outcome groups, subjects who
achieved no statistically reliable change and subjects who
achieved clinically significant change, data for each TARS
subscale was subjected to a 2 (Low versus High Outcome) x 3
(Phase of Treatment) analysis of variance. The only
significant effect obtained was an Outcome x Phase
interaction for the Therapist Negative Contribution scale,
2(2, l8)= 3.91, p < .05. A graph of the interaction is
presented in Figure 1.

Though therapists appeared to make substantially more

96
negative contributions to the alliance with.high outcome
patients in the early and middle phases of therapy, but
more negative contributions with low outcome patients in
the late phase, tests for a simple effect of outcome within
each phase revealed a statistically significant difference
between high (M = 1.39) and low (M = 1.13) outcome patients
only in the middle of treatment, _F_(1, 19) = 4.25, p ( .053.
Tests for the simple effect of phase within outcome
revealed a significant effect for phase within low outcome
patients , 3(2, 18) = 5.74, p < .02. A contrast of the
early (M = 1.13) and middle (M = 1.11) phases with the late

(M = 1.70) phase of treatment within low outcome patients

 

 

 

 

0 2.0 1)-

h

8 -- 0 Low outcome

“:3 13--

2 «- I High outcome

3

.9 1.6 --

E -

8

o 1.41- I—

.5 ..

g» 1.2 --

z ._ ca

g 1.0..

as

.C

l- l L 1

l T I
Early Middle Late
Phase of Treatment
Eiggge 1. Outcome x Phase interaction for Therapist

Negative Contribution score.

97
revealed a: highly significant increase 1J1 therapist
negative contributions at the end of treatment, F(1,

19) = 10.60, E < .005.

Phase of Treatment

As can be seen from Tables 4, 5, 8, and 9, the main
effect of phase of treatment failed to reach statistical
significance for any TARS subscale, though the effect for
the Therapist Positive Contribution scale just missed the
.05 level of significance, E (2, 52) = 2.95, p < .06, and
there was weak trend for the Patient Negative Contribution
scale, 3 (2, 25) = 2.32, p < .12. A contrast between the
mean Therapist Positive Contribution score for the middle
phase of treatment (M = 2.57) and the mean (2.41) of
Therapist Positive Contribution scores for the early (M =
2.39) and late (M = 2.43) phases of treatment revealed that
therapists made significantLy more positive contributions
in the middle phase of treatment, 3(1, 31) = 5.25, p ( .05.
The results of this contrast must be interpreted with
caution since the main effect of phase of treatment fell
short of the established significance level, and because
the reliability coefficient for the Therapist Positive
Contribution scale was low. Patient negative contributions
increased from the early (M = 1.36) to the middle (M =

1.49) to the late (M = 1.54) phase of treatment.

98
Intercorrelations Between Alliance Subscales

Intercorrelations between the TARS subscales for each
phase of treatment are presented in Table 10. A two-tailed
test of statistical significance was applied to each
correlation coefficient since no explicit hypotheses about
relationships between subscales for the total sample had
been made. The average correlation coefficient across the
phases of treatment was calculated for each subscale. The
hypothesis of no relationship between subscales was
rejected for the relationships between the Patient Positive
and Patient Negative Contribution scales, average 5 =
-.77, p < .001, the Patient and Therapist Positive
Contribution scales, average 5 = .56, p ( .001, the Patient
Positive and Therapist Negative Contribution scales,
average £==-n44, p (.01, and the Patient and Therapist
Negative Contribution scales, average 5 = .60, p ( .001.
The null hypothesis was accepted for the relationships
between the Patient Negative and Therapist Positive
Contribution scales, average 5 = -.18, nAL, and the
Therapist Positive anui Negative Contribution scales,
averagqu==.27, n.s.

To facilitate comparisons with reported results from
past research, four-session total mean scores were
generated for each TARS subscale, and correlations between
mean scores were obtained. As with the average

correlations, the null hypothesis of no relationship had to

99
Table 10
Intercorrelations Between TARS Subscales for Each Phase of

Treatment (n = 32)

 

Subscale 2 3 4

 

Session 1

 

 

 

 

1.Patient positive -.51** .71*** -.44:*

2. Patient negative __ -.23 .59

3. Therapist positive __ -.15

4. Therapist negative __
Early phase

1. Patient positive -.47** .61*** -.32***

2. Patient negative __ -.25 .60

3. Therapist positive __ -.31

4. Therapist negative __

Middle phase

1. Patient positive -.76*** .37* -.50::

2. Patient negative __ -.03 .58

3. Therapist positive __ -.27

4. Therapist negative __
Late phase

1. Patient positive -.77*** .49** -.51::*

2. Patient negative __ -.20 .63*

3. Therapist positive __ -.35

4. Therapist negative __

* ** ***

2 < .05; p < .01; p < .001

100
be rejected for the relationships between the Patient
Positive and Negative Contribution scales, £,= -.76, p
< .001, the Patient and Therapist Positive Contribution
scales, 5 = .56, p ( .01, the Patient Positive Positive and
Therapist.Negative Contribution scales, £==-u55, p (.01.
and the Patient and Therapist Negative Contribution scales,
g = .66, p < .001. The null hypothesis was accepted for
the relationships between the Patient Negative and
Therapist Positive Contribution scales, £_= -u21, rns., and

the Therapist Positive and Negative Contribution scales,

5 = -.27, n.s.

Summary of Results

Hypothesis testing analyses. The analysis of the data
provided support for Hypotheses 1 and 3. Patients who
showed low prognosis for developing an alliance, as
indicated by their positive and negative contributions to
the therapeutic alliance in the first session of treatment,
made more negative contributions and fewer positive
contributions to the alliance throughout treatment than
patients with a low prognosis. Among patients who achieved
a significant change in psychotherapy, patient positive
contributions to the alliance increased significantly from
early to late in treatment.

No support was obtained for the other hypotheses.
Patients who began treatment with a poor prognosis for

developing an alliance were not more variable in outcome

101
than patients with a high prognosis. There was no evidence
that patients who achieve a statistically reliable,
clinically significant improvement during psychotherapy
differed from patients who achieve no reliable change in
the positive and negative contributions they made to the
alliance during the course of treatment. Therapist
positive and negative contributions to the alliance
appeared to be unrelated to the outcome of therapy with
patients having an initially poor potential for forming an
alliance. Therapist alliance-related behavior was not more
strongly associated with patient-alliance related behavior
for patients with a low prognosis for the alliance than for
patients with a high prognosis.

cher findings. A relationship between patient

 

prognosis for the alliance and therapist alliance-related
behavior was found, with therapists having a greater
tendency to make negative contributions to the alliance
throughout treatment with low prognosis patients than with
high prognosis patients. Prognosis for forming an alliance
showed no relationship to the pretreatment level of
symptomatology or to therapy outcome. The patient's
pretreatment level of symptomatology was predictive of
outcome.

One unexpected, interesting finding was the
significant Outcome x Phase interaction. for the Therapist

Negative Contribution scale, when the data from only high

102

and low outcome patients was analyzed. In the early and
middle phase of therapy, therapists made more negative
contributions to the alliance with high than with low
outcome patients. In the final phase of therapy, however,
therapists negative contributions to the alliance with low
outcome patients increased significantly, while
contributions with high outcome patients decreased.

Intercorrelations between TARS subscales revealed a
reciprocity between patient and therapist alliance-related
behaviors, with the exception of patient negative and
therapist positive contributions. There was a strong
negative correlation between patient positive and patient
negative contributions to the therapeutic alliance, but no
significant relationship between therapist positive and

negative contributions.

DISCUSSION

Despite theoretical assertions that a therapeutic
alliance is a prerequisite to successful outcome in
psychotherapy, the research has not consistently validated
the importance of patient and therapist contributions to
the alliance in predicting the outcome of treatment. The
convergence of the data appears to indicate that patient
positive contributions are associated with successful
outcome, but empirical support for the influence of patient
negative contributions and therapist contributions to the
alliance is equivocal.

It was hypothesized that the failure of research to
consistently validate theory might be a consequence of an
inadequately complex approach to research on the
therapeutic alliance. There is preliminary evidence that
alliance-related behaviors may have different implications
for outcome depending upon patient predispositional
variables and upon the phase of treatment in which they
occur. The influence of therapist behavior on outcome
seems particularly likely to be obscured in the
conventional statistical approach of pooling the data for
all patients without regard for their predispositional

variables, since patient factors appear to account for a

103

104
significantly higher proportion of variance in outcome than
therapist factors.

The primary purpose of this study was to explore the
course of patient and therapist contributions to the
therapeutic alliance as they relate to the patient's
initial potential for establishing a relationship and to
treatment outcome. The influence of therapist alliance-
related behaviors on patient contributions to the alliance
and on outcome with patients demonstrating initially poor
relationship potentials was of particular interest. A
second objective of the study was to bridge the gap between
clinical practice and research by employing clinically
significant criteria for successful psychotherapy outcome,
rather than relying on comparisons of group means. The
third purpose of the study was to assess the applicability
of the TARS to research with a heterogeneous sample of
patients and therapists engaged in treatment of varying
theoretical orientations and durations.

To accomplish these objectives, psychotherapy patients
were classified as having high or low potentials for
establishing a relationship based on their contributions to
the therapeutic alliance in the first session of treatment,
as rated on the TARS by external judges from audiotaped
segments of the session. For each treatment case, judges
also rated patient and therapist contributions to the

alliance from segments of an early, middle, and late

105
session in treatment. Patients were assigned to one of
three outcome groups, based on criteria of having achieved
a statistically reliable change in symptom level and having
achieved by termination a level of adjustment statistically

likely to place them within the normal range.

Prognosis for the Alliance

Hypothesis one. As predicted, patients rated as
having an initially low prognosis for forming an alliance
made more negative (contributions and fewer positive
contributions to the alliance throughout the course of
treatment than patients rated as having an initially high
prognosis. The results are consistent with past findings
that pretreatment assessments of the patient's capacity for
interpersonal relationships are predictive of contributions
to the alliance during treatment (Marziali, 1984; Moras &
Strupp, 1982; Ryan” 1973) and that patient contributions
to the alliance early in therapy are predictive of
contributions late in therapy (Morgan et al., 1982). The
data suggest that patients possess a core capacity for
relatedness, reflected in pretreatment measures and in
interpersonal behavior from tflma first session of
psychotherapy, that has significant implications for the
ability to form and maintain a therapeutic alliance
throughout treatment. The results are consistent with
psychoanalytic theory, in which the capacity of the patient

for reasonably stable object relationships is specified as

106
a prerequisite to forming a therapeutic alliance.

A comparisons of this study with past studies in
regard to first session indicators of patient prognosis for
forming an alliance is not possible because the
distribution of TARS/CTAS scores has not been reported by
past researchers. It would be interesting to know where
patients in this study fell in their first session positive
and negative contributions to the alliance relative to
norms for a typical outpatient psychotherapy population.

QEEEE-£§1§X§EE_£§§BAE§° Patient prognosis for
developing an alliance was also predictive of the
therapist's negative contributions to the alliance
throughout treatment, but unrelated to therapist positive
contributions. The results suggest that a patient with an
initially poor capacity to relate is likely to elicit
negative responses from the therapist, perhaps negative
countertransference, throughout treatment. The absence of
results for positive therapist contributions suggest the
possibility that positive contributions are part of a "good
therapist" stance that, for the majority of therapists, is
not easily discomposed by a difficult client. This
hypothesis is further supported by the failure to find a
significant correlation between patient negative
contributions and therapist positive contributions, while
therapist negative contributions are highly correlated with

patient negative contributions. If therapist positive

107
contributions are relatively stable regardless of patient
behavior, it seems likely that therapist negative
contributions will have more impact on the process of
therapy.

The failure to find an association between the
pretreatment Gross Pathology Index score of the SCL-90-R
and the Prognosis score or the Patient Negative
Contribution scale score from session one is incongruent
with Marzialids (1984) report that pretreatment symptom
level and depressive mood were associated with patient
negative contributions in the first session of treatment.
However, past researchers employing the TARS/CTAS have
found no significant associations between mean Patient
Positive or Patient Negative Contribution scale scores and
pretreatment symptom level (Marmar et al., in press;
Marziali, 1984). Though.the incongruencezof.Marziali's
results for the Patient Negative Contribution scale is a
problematic issue, the bulk of the data suggests that
ratings of patient behavioa'on.the‘TARS subscales are not
measures of symptomatic distress. Pretreatment symptom
level and initial capacity for object relatedness may both
have implications for psychotherapy process and outcome,
but they reflect differential aspects of psychological
functioning and should be treated as distinct patient

predispositional variables.

108
Prognosis for the Alliance and Outcome

HypppM§§i§_£gp. The hypothesis that "poor"
psychotherapy candidates, candidates with a low potential
for establishing a therapeutic alliance, would have more
varied outcome than "good" psychotherapy candidates was not
supported. There was a trend only for greater variance for
low prognosis patients in the SCL-90-R change score. The
results for Hypothesis 2 appear to invalidate the
underlying assumption that "good" psychotherapy candidates
typically benefit from treatment, barring any major failing
on the part of the therapist, while "poor" psychotherapy
candidates have a range of outcomes reflecting the
influence of therapist skill, since "poor" candidates lack
the resources to achieve gains with minimal assistance, but
are able to respond to effective interventions by the
therapist.

Capacity for relatedness as a predictor of outcome. A
premise more basic to Hypothesis 2, the premise that
patients with high and low prognoses for forming an
alliance are "good" and "bad" psychotherapy candidates in
the conventional sense of potential for outcome, appears to
be questionable. Use of the dual criteria for clinically
significant outcome revealed that only 14% of patients with
a low prognosis for the alliance, compared to 33% of the
high prognosis patients, achieved clinically significant

improvement. While only 22% of the high prognosis patients

109

failed to achieve El statistically reliable change in
symptoms, 64% of the low prognosis patients failed to
achieve a reliable symptom change. These figures suggest
that low prognosis patients had a much lower probability of
a successful outcome, but the chi-square test of a
relationship between prognosis grouping and outcome
classification fell short of statistical significance. A
p-test indicated that high prognosis patients achieved
significantly more symptom change than low pmegnosis
patients, but the more accurate correlational analyses
indicated no significant relationship between prognosis
score and change in symptom level, or between prognosis
score and outcome classification.

The failure to find a solid relationship between
patient contributions to the alliance in session one and
outcome is consistent with the results of O'Malley, Suh,
and Strupp (1983) and Sachs (1983), but inconsistent with
the findings of Marziali (1984), who reported a significant
relationship between session one patient positive
contributions and some outcome variables. She did not
indicate whether symptom level was among these variables.
Neither Marziali (1981) nor Horowitz et a1. (1984) found a
relationship between pretreatment measures of interpersonal
functioning and symptom change.

The results of the present study in combination with

past research suggest that though patients' initial

110
capacity for relatedness are predictive (Hf their
contributions to the alliance throughout treatment, they
are not necessarily predictive of treatment outcome as
reflected in symptom change. Contrary to theoretical
assumptions, the data provide no evidence that a good
initial capacity for relatedness distinguishes a "good"
psychotherapy candidate vflu: will likely benefit from
treatment, at least with regard to symptom change. The
negative results raise the possibility that initial
capacity for relatedness has implications for outcome only
in interaction with other variables not examined in this
study. The failure to find greater variance in outcome
among patients with a low prognosis for a therapeutic
alliance does not controvert the assumption that therapist
behavior will more strongly influence the final level of
adjustment for patients with an initially low prognosis for
outcome than for patients with an initially high prognosis

for outcome.

Patient Contributions to the Alliance

Hyppphesis three. As predicted, the positive

 

contributions of patients who achieved a statistically
reliable change increased significantly from early to late
treatment. When data for the low outcome patients was
included in the analysis, this difference between phases
was not obtained. The results are consistent with the

work of Luborsky et al. (1983), who reported an increase in

111

positive contributions for successful patients only, and
with the pattern reported for all patients by Marziali
(1984). Given that Marziali's study was conducted with
initially fairly high-functioning patients and very skilled
therapists, it is possible that all patients in her study
achieved a relatively successful outcome. The failure of
Gomez-Schwartz etial.(1983) to find a similar trend for
patient positive contributions may be attributed to her
approach of pooling data for all outcome groups. The
Hartley and Strupp (1983) findings of a decrease in patient
contributions from the beginning to the end of treatment,
as measured by the VTAS, are more difficult to reconcile
with other results. Perhaps they can be attributed to the
overinclusiveness of the patient subscale of the VTAS,
which encompasses anxiety, defensiveness, motivation, and
positive and negative contributions in a single scale.
Neither in the present study nor in the Marziali (1984)
study was a difference found between early and late
sessions in patient negative contributions.

The results of this study add further support for the
assumption that, among patients who benefit from therapy,
there is an increase from the beginning to the end of
treatment in the strength of the bond to the therapist and
in active involvement in the collaborative work of therapy.
A similar increase does not seem to occur in less

successful treatments.

112

Hypothesis four. Based on the assumption that a more
complex approach to research might clarify some of the
contradictions in the empirical literature regarding the
relationship between the therapeutic alliance and
psychotherapy outcome, patient contributions to the
alliance were studied at different phases of treatment.
Contrary to the prediction of Hypothesis 4, there was no
evidence that patients who achieved a clinically
significant improvement in symptom level differed from
unsuccessful patients in their positive and negative
contributions to the alliance in early, middle, or late
therapy. No support was found for a simple relationship
between outcome classification and positive or negative
contributions collapsed over phase of treatment.

The failure to find a simple or complex relationship
between patient contributions 11) the therapeutic alliance
and psychotherapy outcome challenges the theoretical
assumption that the formation of an alliance is a
prerequisite to successful psychotherapy. It also
contradicts the convergence of empirical evidence that
patient positive contributions to the alliance are
predictive of outcome. A more careful examination of the
empirical literature reveals, however, that though clinical
judges'ratings of patient positive contributions to the
alliance have been reported to predict a number of outcome

measures, they have not often been found to predict

113
patient-reported symptom level.

In her 1984 research employing the TARS, Marziali
reported that patient and therapist ratings of patient
positive contributions to the alliance correlated with a
variety of outcome measures, including symptom change, but
external judge ratings correlated only with patient and
therapist posttherapy evaluations and clinical evaluations
of dynamic outcome. Horowitz et al. (1984) failed to find
a significant direct relationship between judge ratings on
the CTAS of patient positive contributions to the alliance
and decrease in total symptoms as measured by the SCL-90-R.
In earlier research, however, Marziali et al. (1981) found
that judge-rated scores combining patient positive and
negative contributions differentiated the five most
improved from the five least improved patients, selected
from a group of 25. The contradictions between the Marziali
et al. (1981) results and the results of other research
with the TARS/CTAS may in part be attributable to the
Marziali (et al., 1981) selection procedure, which was
likely to maximize the outcome effect. Outcome groups were
less extreme in the present study, and patients were
unselected for outcome in the Horwitz et al. (1984) and
Marziali (1984) investigations. More important, however,
seems to be the outcome criteria employed. Unlike the
present study and other TARS/CTAS research, the Marziali et

al. (1981) investigation based outcome classification on a

114

composite of several self-report and judge measures. No
assessment was made of the relationship between patient
contributions and outcome measured by symptom level alone.

Similar patterns have been found in research with
other alliance-related measures. Associations have been
reported between the Patient Involvement Scale of the VPPS
and a variety of outcome measures, such as therapist and
clinical observer rating of improvement, but not self-
reported symptom change as measured by the MMPI (Gomez-
Schwartz et al., 1978; O'Malley et al., 1983). Hartley
and Strupp (1983) found no relationship between the VTAS
patient subscale, which incorporates positive and negative
behaviors, and outcome classification based on a composite
measure which included MMPI change score. Luborsky and his
associates (Luborsky et al., 1983; Morgan et al., 1982)
dud report a significant association between patient
positive contributions to the therapeutic alliance and
outcome. In their work, the probability of obtaining an
outcome effect was maximized by the selection of extreme
outcome groups from a large group of subjects, based on a
composite score of patient and clinical observer ratings on
a variety of scales. The accumulation of the data seems to
suggest that, though patient positive contributions to the
therapeutic alliance as rated by external judges are
predictive of some measures of psychotherapy outcome, they

are not related to change in patient-reported symptom

115
level.

The failure to find evidence of a relationship between
patient negative contributions to the alliance and outcome
is generally consistent with results of past research,
since empirical support for the relationship is equivocal
at best. Marziali (1984), Hartley (1978), and the Luborsky
group (Luborsky et al., 1983) reported that their measures
of patient negative contributions to the alliance did not
predict outcome. The inconsistency between the Marziali
et al.(1981) finding that patient negative contributions
as measured by the TARS did predict outcome, and the
negative results from this and other research, may be
accounted for by Marziali's (1981) use of extreme outcome
groups and by the use of different outcome measures.
However, the report by Horowitz et al. (1984) of a
significant inverse relationship between patient negative
contributions and total symptom reduction as measured by
the SCL-90-R is very difficult to reconcile with the
failure in this study and the Marziali et al. (1984) study
to find a significant relationship between the same
variables. It is possible that the Patient Negative
Contribution Scale of the TARS was given different
interpretations by raters in the various studies. Or
perhaps patient negative contributions ix) the therapeutic
relationship have different implications for patients

presenting with grief reactions, as in the Horowitz study,

116
than for patients with heterogeneous presenting problems.

The conclusion that patient positive contributions to
the therapeutic alliance as rated by clinical judges do not
predict improvement in self-reported symptom level, though
they may predict other outcome criteria, certainly adds
weight to the argument for multidimensional outcome
criteria. However, it also calls into question the
theoretical assumption that a therapeutic alliance is a
prerequisite for successful psychotherapy, since a decrease
in symptoms experienced by the patient is certainly one
essential criteria for successful outcome. Confidence in
conventional beliefs about the importance of the
therapeutic alliance for treatment outcome is further
eroded by the paucity of support for a relationship between
patient negative contributions and outcome.

One possible explanation of the failure of empirical
studiestxbcorroborate theoretical assumptions about the
importance of patient contributions to the alliance for
improvement in symptom level is that the implications of
alliance-related behavior for outcome can be understood
only in interaction with other variables. An attempt was
made in this study to determine whether patients with
similar initial capacities for relatedness, as indicated by
first session contributions to the alliance, but with
dissimilar outcomes, were differentiated by their pattern

of contributions to the alliance during treatment. The

117

data provided no support for the hypothesis that.patient
contributions ix) the alliance have different implications
for outcome depending upon the patient's predisposition for
establishing an alliance. Horowitz et al. (1984) reported
that patient motivation, however, interacted with both
patient positive and patient negative contributions to the
alliance to account for a significant proportion of the
variance in symptom improvement. Patient positive
contributions were positively related to outcome at low
levels of motivation, but shifted direction and became
negative as motivation increased. Patient negative
contributions were negatively related to outcome at lower
values of motivation, but shifted from negative to positive
as motivation increased. Though the patient's initial
capacity for relatedness does not appear to interact with
contributions to the alliance to predict outcome, other
patient predispositional variables do.

It seems that if, as psychoanalytic theory postulates,
the formation of a therapeutic alliance is a prerequisite
to successful psychotherapy outcome, the relationship
between patient contributions to the alliance and outcome
is neither simple nor direct. Perhaps patient
contributions to the alliance predict a significant
proportion of the variance in symptom improvement only in

interaction with other variables.

118

Therapist Contributions to the Alliance

 

Hypothesis five. ‘Following the assumption of Suh
and O'Malley (1982) that the influence of therapist
variables on psychotherapy outcome for "poor" psychotherapy
candidates tends to be obscured when the data for all
patients is pooled, patients in this study were designated
as having a high or low prognosis for forming a therapeutic
alliance, and the therapist contribution data was analyzed
separately for low prognosis patients. Contrary to the
predictions of Hypothesis 5, no evidence was found for a
relationship between therapist positive or negative
contributions to the alliance and psychotherapy outcome in
the treatment of It”: prognosis patients. Though the
failure to find a relationship between therapist alliance-
related behaviors and outcome is consistent with most past
research, it had been anticipated that the technique of
differentiating high and ltnv prognosis patients would
elucidate relationships which had been obscured in the
past. Two factors seem to have been operating to undermine
the effectiveness of this approach.

The first factor was the assumption that patients with
low prognoses for establishing an alliance were necessarily
"poor" psychotherapy candidates, and patients with high
prognoses were "good" candidates. As discussed earlier
with regard to Hypothesis 2, the data from this

investigation, contradicting theory, clinical experience,

119

and some past research, provided no evidence that a good
potential for forming a therapeutic alliance distinguishes
a "good" psychotherapy candidate who will benefit from
treatment, at least as reflected by symptom improvement.
Thus the absence of support for Hypothesis 5 does not rule
out the possibility that therapist alliance-related
behavior will influence outcome for patients with an
initially poor prognosis for benefiting from treatment.

The second factor is the size of the sample. The
number of low prognosis patients was so small that only an
extremely strong relationship would have registered
statistical significance. The "failed predictions"
patients who were of greatest interest, low prognosis
patients who achieved clinically significant improvement,
numbered only two, preventing meaningful generalizations
about therapist contributions with these patients and
minimizing the possibility of statistically significant
results. The small number of patients seems to have
precluded a valid test of the hypothesis that therapist
alliance-related behavior influences outcome among patients
who begin treatment with a poor capacity for relatedness.

Hypothesis six. No evidence was found for a greater

 

influence of therapist alliance-related behavior on patient
alliance-related behavior among low prognosis patients than
among high prognosis patients. For both groups, patient

and therapist alliance-related behavior tended to be highly

120

correlated at each phase of therapy, but no more so for low
prognosis patients than for high prognosis patients. The
absence of support for Hypothesis 6 further discredits the
assumption that patients with an initially low prognosis
for forming an alliance will be more susceptible to the
influence of therapist behavior than high prognosis
patients.

Outcome x Phase interaction. Unpredicted but very
interesting was the Outcome ){ Phase of Treatment
interaction for the Therapist Negative Contributions scale
score when only high and low outcome patients were included
in the analysis. In the early and middle phases of
treatment, therapists were more likely to make negative
contributions to the alliance when working with patients
who eventually achieved clinically significant improvement
than when working with patients who failed to achieve
reliable change. However, by the end of treatment, the
pattern had reversed. Therapists were making more negative
contributions with low outcome patients than with high
outcome patients.

The pattern of the Outcome x Phase effect is counter
to theoretical assumptions that negative therapist
contributions to the alliance inhibit the patient's
progress in therapy, but ix: is not inconsistent with
results reported by some other investigators. Marziali

(1984) found that therapist negative contributions to the

121

alliance in the first and third session of treatment were
positively associated with symptomatic and dynamic
improvement. Dietzel and Abeles (1975) reported that
successful therapists responded to their clients with a
significantly lower level of complementarity during the
middle phase of therapy than unsuccessful therapists.
Crowder (1972) noted more hostile-competitive behavior
among successful therapists than among unsuccessful
therapists in the early phase of therapy. Hartley and
Strupp (1983) found Positive Climate, which is primarily a
therapist contribution factor, to be significantly lower
for high outcome dyads than for low outcome dyads in the
median session of treatment. The accumulated data suggests
that that perhaps some therapist negative behavior in
response to the client early in therapy is necessary to
challenge tflua client's characteristic patterns of
interaction and to facilitate change. The apparently
negative behaviors may actually be necessary to engaging
the patient in the cognitive-motivational aspects of the
therapeutic alliance. By the end of treatment, therapists
may no longer see a need to challenge patients who are
making significant gains, whereas the increase in negative
contributions with unsuccessful patients may reflect
therapist frustration with the patientis failure to make
progress.

A factor which may confound the interpretation of the

122
Phase x Outcome effect for the Therapist Negative
Contributions scale is the strong relationship between
outcome and pretreatment level of symptomatology. High
outcome patients were significantly higher (§(19) = 3.18, p
< .005) in pretreatment symptom level than low outcome
patients. It may be that more negative therapist behavior
is elicited early in treatment by the eventual high outcome
patients because they are more disturbed than the low
outcome patients. By the end of treatment, high outcome
patients are no longer eliciting such negative therapist
behavior because they have reached a "normal" level of
functioning. What is interesting is that these initially
low functioning patients make progress in therapy despite
the theoretically alliance-inhibiting behavior of their
therapists. It may be that therapist positive
contributions, for which there was no Phase x Outcome
interaction, were consistent enough to develop and sustain
the emotional bond of the alliance with these initially
difficult patients despite the negative therapist
contributions. Or, as suggested earlier, what appear to be
negative therapist contributions to the alliance may
actually facilitate change by disrupting the patient's
characteristic patterns of interaction. Though the low
outcome patients initially elicit less negative therapist
behavior because they are less disturbed, by the end of

treatment the therapist may be quite frustrated because

123
these patients with a seemingly good initial potential to
use therapy have accomplished little.

Since outcomezfor the Crowder (1972) and Dietzel and
Abeles (1975) studies were based on clinicianfis ratings of
pre- and posttherapy MMPI profiles, it is possible that
there was a similar correlation between initial level of
pathology and outcome classification. Hartley and Strupp
(1983) also employed gain scores to obtain their outcome
classification, with no mention of control for initial
level. The findings by Marziali (1984), of a positive
association between early negative therapist contributions
and decrease in symptoms, however, cannot be accounted for
by the effect of initial symptom level on outcome, because
she controlled for this effect.

There was little evidence in the literature to support
theoretical assumptions about the importance of therapist
behavior for psychotherapy outcome, but a number of aspects
of the present study were intended to maximize the
probability of teasing out the influence of therapist
variables. The technique of differentiating high and low
prognosis patients in order to avoid masking the influence
of therapist contributionscnioutcome with low prognosis
patients failed tx: produce results. The approach of
looking at the therapeutic alliance more complexly did
prove to be fruitful. Though no simple relationship was

found between therapist contributions to the alliance and

124
outcome, therapist negative contributions had different
implications for outcome depending upon the phase of
treatment in which they occurred. The use of therapists
with a wider range of skill than typically found in past
research may also have strengthened the effect of therapist
negative contributions. Both lMarziali (personal
communication) and‘Horowitz et a1.(1984) mentioned the
lack of variability of ratings of therapist variables as a
factor interfering with obtaining significant results.
Though distributions of scores from the previous studies
are not available for comparison, distributions of scores
for this study suggest that lack of variability was not a
problem with the therapist negative contribution score,
which was equivalent to the patient negative contribution
score in variance and range. A lack of variability in the
therapist positive contribution score may have inhibited
the likelihood of obtaining statistically significant
results. Perhaps fairly consistent positive contributions
are a given with most therapists, but negative
contributions to the alliance are more easily made by
unskilled therapist. or therapists in whom

countertransference reactions have been triggered.

Course of the Alliance

 

For the sample of patients as a whole, the effect of
phase of treatment did not reach significance for any TARS

subscale, though there was a strong trend for the Therapist

125

Positive Contribution Scale and.ai weak trend for the
Patient Negative Contribution scale. Therapist positive
contributions to the alliance were significantly higher in
the middle of treatment than in the early'or late phases.
It may be that therapists increase their positive
contributions from the beginning through the middle of
therapy as they strive to build a solid alliance. By the
end of treatment the alliance should be well established,
and perhaps fewer positive contributions on the part of the
therapist are required to maintain it. Interpretations of
the phase of treatment effect for the Therapist Positive
Contribution scale should be made with caution, given the
low interrater reliability for the scale. Reliability was
particularly problematic for the middle phase of treatment.

Contrary to what might be expected, patient negative
contributions increase over the course of therapy. It may
be that the negative patient behaviors scored on the TARS
overlap with indicators of negative transference, which
could be more likely to occur as the patient becomes
increasingly involved in the therapeutic process, though
negative transference would be expected to be resolved by
the end of treatment. Because the Psychological Clinic
where the treatments were conducted operates on an academic
calendar, it is possible that termination for many of these
patients was dictated by the clinic calendar rather than

the achievement of the treatment goals, leaving negative

126
transference not only unresolved, but heightened by the
circumstances of termination. The trend for patient
negative contributions is compatible with the Hartley and
Strupp (1983) finding of a decrease over time on the
Patient subscale of the VTAS, which incorporates negative
and positive patient behaviors but is scored in the
positive direction. Phase of treatment effects have not
been reported for patient negative contributions to the

alliance by other researchers.

Outcome

Comparisons of pre- and posttreatment mean symptom
levels indicated that psychotherapy was beneficial for the
sample as a whole. As would be expected if treatment was
effective, the variance in symptom scores decreased
significantly from pre- to posttreatment, as patients moved
toward a more functional level of adjustment. Applying the
dual criteria for clinically significant outcome study
permits a more meaningful statement about the effectiveness
of therapy for individual patients in this study. A
statistically reliable improvement in symptom level was
achieved by 59% of the patients, though only 25% of the
total sample also reachedealevel of adjustment likely to
place them in the functional range.

The dual criteria for psychotherapy outcome employed
in this study was intended to avert some of the

methodological difficulties of common approaches to outcome

127

measurement in psychotherapy research: global, subjective
ratings which defy meaningful generalization of findings;
statistical unreliability of gain scores; equating
patients with similar final adjustments but very different
degrees of change; and statistically significant but
clinically irrelevant differences between.group means on
outcome measures. The patients identified as having a high
outcome in this study had achieved statistically reliable
change in symptom level during treatment and had reached a
final level of adjustment more likely to place them in the
functional than in the dysfunctional range. Thus high
outcome patients had clearly achieved clinically
significant improvement.

Correlation with pretreatment symptom level. The
situations for low outcome patients, however, is more
ambiguous, due to the correlation of pretreatment symptom
level with outcome. While past researchers have reported
that pretreatment symptom level accounted for only 5 to 10%
of the variance in outcome, pretreatment symptom level in
the present study accounted for 74% of the variance in the
change in symptom level and 18% of the variance in outcome
classification. The higher the initial level of reported
symptomatology, the more change the patient was likely to
achieve during therapy and the greater the probability of
being classified as having a clinically significant

improvement.

128

The correlation between pretreatment level and outcome
can be attributed in part to the greater influence of
regression to the mean on change scores of patients with
initially high levels of symptoms, and to a ceiling effect.
Patients who began treatment with a relatively low level of
symptoms, though within the dysfunctional range, could make
considerable gains by termination but be classified as
having a low outcome because the SCL-90-R is not
sufficiently sensitive to register changes at the healthy
end of the continuum of psychological functioning. In the
present study, only two patients reached a level of
adjustment in the functional range but were classified as
having low outcome because their changes scores were not
statistically reliable. Reanalyses of tflua data excluding
these two patients did not produce significantly different
results from the original analyses. If a substantial
prOportion of the low outcome patients in a study had
reached.a functional level of adjustment, the pattern of
relationship between psychotherapy process and outcome
could be quite distorted. Even in the present research,
the interpretation of results must be made with an
awareness that high outcome patients are also initially low
functioning patients. It seems possible that the strong
correlation between pretreatment symptom level and outcome,
which has been controlled in past research with the TARS,

may have obscured any relationship between the therapeutic

129
alliance scales and outcome.

Selecting only patients with comparably high
pretreatment symptom levels could avert the problem of
outcome correlating with pretreatment symptom level. With
initially higher functioning patients, an outcome measure
more sensitive than the SCL-90-R to changes at the healthy
end of the psychological continuum might be employed.

Need for Multidigensional criteria. Though the

 

approach to the measurement of outcome taken in this study
represented an improvement in methodology over many past
studies, it was limited by the use of a single dimension of
outcome. Patient positive contributions to the alliance as
rated.by«clinical judges have rarely been found to relate
to symptom change, but associations with other types of
outcome criteria have been found. Perhaps the
probability of finding a relationship between the
therapeutic alliance and outcome could have been maximized
by employing an outcome measure which, like the process
measure, reflected the perspective of a mental health
professional. The ideal would be to assess outcome from
the perspectives of society, the mental health
professional, and the patient, with the dual criteria of
statistically reliable change and final adjustment within
the functional range applied to each perspective. An
absence of reliable measures with the required norms

available could make such ideal criteria difficult to

130

implement.

Validation Of the TARS

Reliability; One purpose of the study was t0«assess

 

the applicability of the TARS for use with a heterogeneous
sample of clients and therapists engaged in a wider range
of treatment than in previous studies. Establishing
adequate interrater reliability with our sample of cases,
which was much more divergent than the samples of short-
term psychodynamic psychotherapies with neurotic patients
used in previous TARS/CTAS research, was difficult.
Clinical judges in this study received significantly more
hours of training than the 12 hours typically provided by
past researchers. jyet intraclass correlation coefficients
for the subscales were considerably lower than the
coefficients presented by Marziali and her associates
(Marziali, 1984; Marziali et al., 1981). This might
suggest that, like the Vanderbilt Negative Indicators Scale
(Suh, O'Malley, & Strupp, 1982), the TARS may be most
applicable to therapy conducted from the psychoanalytic
framework in which it was conceived. However, reliability
coefficients from two other investigations, also done on
short-tern: psychodynamic jpsychotherapy' with :neurotic
patients, are more comparable to the reliability
coefficients from this study than the Marziali studies.
Horowitz et al. (1984) reported intraclass correlation

coefficients for Patient and Therapist Negative

131

Contributionscof.62 and.51.respectively, and indicated
that the unspecified reliability levels for positive
subscales were sufficiently lower to be deemed
unacceptable. The higher range of reliabilities reported
by Horowitz et al. were fairly comparable to coefficients
for this study, which ranged from .46 to .70. Though a
second reliability analysis of the same data produced
intraclass correlation coefficients ranging from .65 to
.76, the use of a two-way analysis of variance which
Shrout and Fleiss (1978) indicate is inappropriate to the
Horowitz et al. method of data collection seems to
invalidate the second set of reliability coefficients
reported. A third reliability study with a small sample
produced coefficients ranging from..50-Uo.72 (Marmar et
al., in press).

The important factor distinguishing the Marziali
(1984; Marziali et al., 1981) research from other
investigations with significantly lower reliability
coefficients seems to be the level of clinical experience
of the judges. Marziali employed experienced clinicians,
while advanced trainees in psychology, psychiatry, and
social work served as judges in this study and the other
two studies with lower reliabilities. Some discrepancy in
the experience levels and theoretical orientations of the
clinical judges in this study may have also contributed to

the difficulty in establishing reliability. One judge was

132
less clinically experienced and less psychoanalytically
oriented than the other judge.

Average reliability coefficients for the Therapist
Positive and Negative Contribution scales for this study
were diminished considerably by poor reliabilities for the
middle session of therapy, when there appeared to be a
patient by judge interaction influencing the ratings.
Limited variance in the rating matrix contributed to the
difficulty in establishing reliability for the‘Therapist
Positive Contribution Scale.

With the exception of the coefficient for the Patient
Negative Contributions scale, the interrater reliabilities
obtained in this study are lower than desirable according
to text-book standards. However, the intraclass
correlations coefficients are comparable to the
coefficients reported by previous researchers using the
TARS/CTAS with advanced clinical trainees as judges. The
power of the test for reliability in the present study is
strengthened by the sample size of 126 segments, each rated
by two raters.

It appears that the sophistication of clinical
judgments required in applying the TARS/CTAS makes it very
difficult to establish adequate interrater reliability on
the instrument, thoughtfluause of seasoned clinicians as
judges maximizes reliability. It may also be that our

classical standards 1km: research methodology are

133
unrealistic within the context of psychotherapy research.
In an article on methodological strategies in psychiatric
clinical research, Kraemer (1981) contended that test-
retest reliabilities of 80% and above should be considered
near perfect, 60%-80% as satisfactory, and 40%-60% as
acceptable but possibly improvable.

The comparability of intraclass correlation
coefficients for this study with coefficients for other
student-judged studies suggests that the TARS can be
applied as reliably to psychotherapy of varying durations
conducted from diverse theoretical perspectives with a wide
range of patient as it can be applied to time-limited
psychodynamic psychotherapy with patients showing neurotic
level responses to stress. Given the discrepancies between
our reliabilities and the reliabilities obtained in the
Marziali (1984; Marziali en: a1” 1981) studies,
additional research with seasoned clinicians applying the
TARS to acdiverse sample of psychotherapy cases would be
recommended to confirm the equivalent reliability.

Intercorrelations of subscales. The patterns of

 

association between subscales found were identical to the
patterns reported in the Marziali (1984) research.
Therapist and Patient Positive Contribution scales and
Therapist and Patient Negative Contribution scales had a
strong positive correlation, while Patient Positive and

Negative Contribution scales showed a: strong inverse

134

relationship. In contrast, Therapist Positive and Negative
scales were not significantly correlated, suggesting that
they represented two separate dimensions. This similarity
of relationships between subscales in the Marziali (1984)
study and the present study strengthens the case for the
applicability of the TARS to divergent dyadic therapeutic
relationships. The discrepancies between the patterns of
subscale intercorrelations reported in the 1981 Marziali et
al. study and the patterns found in the present research
and the 1984 Marziali research may be due to the selection
of extreme outcome groups in the earlier Marziali study.

Validity. lDiscriminant validity for the TARS with the
varied sample of psychotherapy cases in this study was
supported by the failure to find a significant correlation
between prognosis score and pretreatment symptom level,
indicating that, consistent with conceptualization, the
patient contribution scales are not merely measures of
symptomatic distress.

The question of whether the TARS has predictive
validity for a psychotherapy with a heterogeneous sample of
patients and therapists remains unresolved. The dearth of
support obtained 131 this study for ea significant
relationship between the TARS subscales, with the exception
of the Therapist Negative Contribution scale, and outcome
suggests the possibility that the scale may be most

applicable 1x) psychotherapy’ conducted. within the

135
theoretical framework from whichtfluaTARS was conceived.
Though the originators intended the TARS to assess the
collaborative aspect CHE any dyadic therapeutic
relationship, the measure has been applied only to time-
limited psychoanalytically-oriented psychotherapy in the
research published thus far. At the Michigan State
University Psychological Clinic, interpersonal, ego
psychoanalytic, Rogerian, and cognitive-behavioral
approaches to treatment are represented. Bordin
(1979) contended that different kinds of therapeutic
alliances are required in psychotherapy conducted from
different theoretical perspectives. It may be that the
TARS taps aspects of the alliance, particularly cognitive-
motivational aspects, that are prerequisites for successful
psychoanalytically oriented therapy, but less relevant for
other modes of treatment. The use of symptom level as the
only dimension of outcome in this research, when past
investigators have generally failed to find a relationship
between the TARS subscales and symptom improvement, makes
it difficult to compare the predictive validity of the TARS
for this heterogeneous sample of cases to the predictive
validity for the time-limited, psychoanalytically-oriented
treatments of neurotic patients to which it has previously
been applied. While the predictive validity of the TARS
with a heterogeneous population needs to be assessed with

multidimensional outcome criteria, it seems that validity

136
of clinical judge ratings on any of the TARS scales for
predicting symptomatic improvement of psychotherapy

patients remains in question.

SUMMARY AND CONCLUSION

The objectives of this research were threefold. The
first objective was in: address the complexity of the
therapeutic alliance by studying the interaction of the
patient's initial prognosis for establishing a therapeutic
relationship, the course of contributions to the
therapeutic alliance, and outcome. The more complex
approach was intended to elucidate some of the
relationships between contributions to the alliance,
particularly therapist contributions to the alliance, and
outcome, which had been obscured in past research. The
second.purpose:of the study was to assess the validity of
the TARS for use with a more heterogeneous sample of
therapist and patients engaged in a wider range of
treatments than in previous studies. The third objective
of the research was to examine the relationship between the
therapeutic alliance and outcome when outcome, rather than
determined solely by comparisons between groups means or by

correlations, was defined in clinically significant terms.

Therapeutic Alliance
The results of this study indicate that, from the

first session of therapy, patients demonstrate a potential

137

138

for establishing a therapeutic relationship that is
predictive of their capacity to contribute to a therapeutic
alliance throughout treatment. This finding is consistent
with the theoretical assumption that patient possession,
prior to beginning psychotherapy, of certain aspects of
relatively mature ego development is a prerequisite to the
formation of a therapeutic alliance. The patient's
predisposition for establishing a relationship appears to
further influence the strength of the therapeutic alliance
because patients with a low potential for an alliance tend
to elicit more negative contributions to the alliance from
therapists throughout treatment. As predicted, there was
an increase from early to late therapy in the positive
contributions to the alliance made by patients who achieved
a reliable improvement in symptoms.

A theoretical assumption not strongly supported in
this study is that a good patient potential for forming a
trusting bond and working in a collaborative way with the
therapist also implies a good potential for benefiting from
psychotherapy. Though patients with a low prognosis for
forming an alliance were more likely to fail to achieve a
statistically reliable change in symptoms and less likely
to make a clinically significant improvement than high
prognosis patients, patient prognosis for the alliance did
not account for a significant proportion of the variance in

symptom.improvement.

139

The failure to find a strong relationship between
patient prognosis for forming an alliance and outcome is
part of a more general failure to validate the theoretical
importance of patient contributions to the alliance for
psychotherapy outcome. Despite efforts to take a more
complex approach to the therapeutic alliance, considering
the influence of’a patient predispositional variable and
the vicissitudes of the alliance in different phases of
therapy, no simple or complex relationship between patient
alliance—related behavior and outcome was found. Though
these negative results are inconsistent with the
convergence of data indicating that patient positive
contributions to the alliance predict some measures of
outcome, they contribute to the growing evidence that
clinical judge ratings of patient contributions to the
therapeutic alliance are unrelated to improvement in
patient-reported symptom level.

The approach of differentiating patients with high and
low relationship potentials, hoping to uncover an influence
of therapist alliance-related behavior on low potential
patients which had been obscured in past research when all
patients were pooled, did not prove fruitful. There was no
evidence that patient alliance-related behavior correlated
more strongly with therapist alliance-related behavior
among low prognosis patients than among high prognosis

patients. While no support was obtained for the hypothesis

140

that therapist contributions to the alliance influence
outcome for patients with a low initial capacity to relate,
it seems that the small number of low prognosis subjects
did not allow a valid test of the hypothesis. As
discussed earlier, the assumption that a poor potential for
establishing a therapeutic alliance was indicative of a
poor potential for benefiting from psychotherapy was not
well founded. Differentiating patients as having a high or
low prognosis based on a different predispositional
variable, a variable actually related to outcome, might
have revealed some influence of therapist contributions to
the alliance on outcome for patients lacking the initial
resources to benefit from treatment with minimal assistance
from the therapist.

The more complex approach of studying the therapeutic
alliance at different phases of treatment did reveal a
relationship between therapist negative contributions to
the alliance and outcome which collapsing the alliance
ratings over the phases of treatment would have masked.
The significant results contradict theoretical assumptions,
however, suggesting that therapist negative contributions
in the early and middle phases of treatment, rather than
inhibiting progress, facilitate change by challenging the
patient's characteristic patterns of relating. The
approach.of employing a sample of therapists with atvide

range of skills was effective in revealing a relationship

141

between therapist behavior and outcome, unlike past studies
ma which therapist skill levels have typically been
homogeneous and scores on therapist alliance subscales have
varied little. The results of the study also suggest that
therapist positive contributions to the alliance are fairly
stable, but therapist negative contributions are vulnerable
to the influence of difficult clients.

This research produced no support for the theoretical
assumption that "the observable ability of the patient and
therapist to work together in a realistic, collaborative
relationship based on mutual respect, liking, and
commitment to the work of treatment" (Foreman & Marmar,
1985, p. 922) is predictive of psychotherapy outcome. One
possible explanation is; that therapeutic alliance as
conceptualized and operationalized in the TARS has
implications for outcome only for psychotherapy conducted
from ant ego psychoanalytic framework, not for a.
heterogeneous sample of cases. A review of past research,
however, indicates that clinical judge ratings of patient
contributions to the alliance on the TARS/CTAS in
psychoanalytically—oriented treatments have been found to
predict a number of outcome measures, but rarely has
symptom improvement been one of those measures. The
pattern is similar for other instruments designed to
measure the therapeutic alliance. Evidence for an

influence of therapist contributions to the alliance, like

142
most therapist variables, on symptom improvement or any
outcome measure is rare.

Though evidence from past research that patient
contributions to the alliance predict outcome measures
other than symptom improvement adds weight to the argument
for multidimensional outcome criteria, the failure in this
and other studies to find a significant relationship
between clinical judge ratings of patient or therapist
contributions to the alliance and symptom improvement
presents a serious challenge to theoretical assumptions
about the role of the therapeutic alliance. Symptom level
is only one aspect of psychotherapy outcome, but it is
certainly a critical aspect.

The accumulation of data thus far indicates that if,
as theorized, indicators of patient and therapist
involvement in a therapeutic alliance are predictive of
outcome in psychotherapy, the relationship is neither
simple nor direct. Judge ratings of contributions to the
alliance may predict some outcome measures but not others.
Patient or therapist ratings of the alliance may predict
outcome measures that judge ratings fail to predict. The
implications of therapist negative contributions to the
alliance for outcome vary depending upon when in treatment
they occur. Some patient predispositional variables
interact with contributions to the alliance to predict

outcome, though the patient's initial capacity for forming

143
a therapeutic relationship does not appear to have a
significant influence.

Inconsistent research results force a recognition of
the possibility that a positive bond between patient and
therapist, marked by a mutual commitment to the
collaborative work of therapy, may not be a prerequisite to
successful outcome in psychotherapy. Clinical experience,
however, supports the probability that the empirical
inconsistencies reflect a relationship between the
therapeutic alliance and psychotherapy outcome that is
highly complex. It seems likely that implications of
patient and therapist contributions to the alliance may be
understood only in interaction with a a number of other
variables.

The power of this research was undermined by the
sample size which, though not atypical for psychotherapy
research, was relatively small. A relationship between
variables would have reached statistical significance only
if it was quite strong. Given the probable complexity of
interactions between contributions to the therapeutic
alliance and other variables in predicting outcome, large
samples of psychotherapy cases may be necessary to extract
patterns of relationships between variables if traditional
statistical approaches are to be taken to psychotherapy
research. Another alternative might be to take a more

microscopic approach to a small number of cases, perhaps

144
doing a session-by-session analysis to elucidate the
patterns of interactions between variables within each
case. Though not productive in the present study, the
failed predictions approach seems to offer much promise for
identifying the aspects of process which distinguish

SUCCGSSfUl from unsuccessful treatment cases.

Applicability of the TARS
Only a somewhat tentative conclusion that the TARS is
applicable to a diverse sample of patients and therapists
engaged in treatments of varying durations and theoretical
orientations can be drawn. Though interrater reliability on
the TARS proved difficult to achieve with this
heterogeneous sample of cases, reliability coefficients
were equivalent to coefficients of other studies in which
the judges were also advanced clinical trainees. Patterns
of correlations between subscales of the TARS were similar
to patterns obtained in past research. While first session
indicators of patient potential for forming a therapeutic
alliance were predictive of patient contributions to the
alliance throughout treatment. no predictive validity of
the TARS for symptom improvement was found. This finding
would raise serious questions about the applicability of
the TARS with a heterogeneous sample of psychotherapy
cases, except that it is congruent with most results of
earlier TARS/CTAS investigations <rf time-limited,

psychoanalytically-oriented psychotherapy with neurotic

145
patients, and of results with other therapeutic alliance
'instruments with other populations. It seems that further
research with the TARS on a diverse sample of treatment
cases with multidimensional outcome criteria and with more
seasoned clinicians serving as judges is needed before a
conclusion can be drawn about the applicability of the

scale to any dyadic therapeutic relationship.

Clinically Significant Criteria for Outcome

 

Applying the dual outcome criteria of a
statistically reliable improvement and a final adjustment
within the normal range of functioning avoided a number of
methodological problems of common approaches to«outcome
measurement, and identified a group of high outcome
patients who had clearly achieved an improvement in symptom
level with practical implications for their lives.
Posttreatment, they reported no more psychological symptoms
than a random sample of non-patients. Due to a strong
correlation of outcome with pretreatment symptom level, the
failure of all low outcome patients to achieve a clinically
meaningful change was more equivocal. In future research,
some control for pretreatment symptom level should be
included. Ideally, the clinically significant criteria
would be also be applied to multiple outcome measures,
though results could In: difficult to interpret since
different measures would be likely to diverge in their

classifications of patients.

146

Reporting the proportion of patients in each prognosis
group who achieved clinically significant improvement and
who achieved no reliable change provided information about
the effects of treatment for individual patients lost in
the statement that high prognosis patients had a
statistically significantly higher mean change score than
low prognosis patients. If prognosis for the alliance had
shown better predictive validity for outcome, the results
of this study could have been used to estimate, based on
first session contributions to the alliance, the
likelihood that a specific individual would achieve a
cflinically significant improvement in psychotherapy.
Reporting the proportion of clinically improved patients
seems a very useful adjunct to traditional statistical

approaches.

Conclusion.

 

The results of the study seem.to speak most strongly
to the need for a multidimensional approach to both process
and outcome variables in the study of the therapeutic
alliance. The relationship between the therapeutic
alliance and outcome seems likely to be mediated by the
interplay of alliance-related behavior with a variety of
variables, including patient predispositions, therapeutic
technique, and the sequence of events in psychotherapy.
The probability of clarifying the relationship between

psychotherapy process and outcome would be increased by use

147
of more robust, multidimensional outcome measures, applying
the dual criteria for clinically significant improvement to
each dimension. If patient and therapist contributions to
the therapeutic alliance do have significant implications
for psychotherapy outcome, as theory and clinical
experience attest, it seems that the relationship is highly
complex and will only be understood with an equally complex

approach to research.

APPENDIX

APPENDIX

Manova Summary Tables
Table A-1
Summary of 2 (High vs. Low Outcome) x 3 (Phase) Analysis

of Variance for Patient Positive Contribution Scale

 

 

Source of variance g; Mean square E p
Outcome 1 .188 (1.00 .46
Error between 19 .325
Phase 2 .053 (1.00 .67
Outcome by phase 2 .142 (1.00 .35
Error within 38 .131

 

148

149

Table A-2

Summary of 2 (High vs. Low Outcome) x 3 (Phase) Analysis

of Variance for Patient Negative Contribution Scale

 

Between subjects factors effects

 

 

 

 

Source of variance g; Mean square 5 p
Outcome 1 1.116 (1.00 .87
Error between 19 38.303
Within subjects factors effects
Effect Wilks Hypothesis Error 5 p
Lambda d: pg
Phase .796 2 18 2.30 .13
Outcome by phase .858 2 18 1.49 .25

 

150
Table A-3
Summary of 2 (Reliable Change vs. No Change) x 3 (Phase)
Analysis of Variance for Therapist Positive Contribution

Scale: Low Prognosis Patients

 

 

Source of variance a; Mean square 3 p
Change 1 .073 (1.00 .58
Error between 12 .227

Phase 2 .166 1.92 .17
Change by phase 2 .112 1.30 .29

Error within 24 .086

 

151
Table A-4

Summary of 2 (Reliable Change vs. No Change) x 3 (Phase)

 

Analysis of Variance for Therapist Negative Contribution

Scale: Low Prognosis Patients

 

Between subjects factor effects

 

 

 

 

Source of variance a; Mean square 3 p
Change 1 .355 1.16 .30
Error between 12 .306
Within subjects factor effects
Effect Wilks Hypothesis Error 5 p
Lambda a; a;
Phase .703 2 11 2.33 .15

Change by phase .953 2 11 (1.00 .77

 

LI ST OF REFERENCES

LIST OF REFERENCES

Abeles, N. (1985, October). Egogpsychology in 1985.
Paper presented an: the Missouri Psychological
Association Meeting, Columbia, MO.

 

Abelhoff, M.IL4 & Derogatis,Ih R. (1977). Psychological
aspects of management of primary and metastatic breast
cancer. Proceeding_of the International Conference on
Breast Cancer. New York: A. R. List.

Allen, J. G., Newsom, G. E., Gabbard, G. 0., & Coyne, L.
(1984). Scales to assess the therapeutic alliance from a
psychoanalytic perspective. Bulletin of the Menninger
Clinic, 48, 383-400.

Barrett-Lennard, G. T. (1962). Dimensions of therapist
response as causal factors in therapeutic change.

Psychological Monographs, lg (43, Whole No. 562).

Bergin, A. E., & Lambert, M. J. (1978). The evaluation of
therapeutic outcomes. In A. E. Bergin & S. L. Garfield
(EdsJ, Handbook of psychotherapy and behavior change:
An empirical analysis (2nd edJ. New York: John Wiley &
Sons.

Bordin, E. S. (1979). The generalizability of the psycho-
analytic concept of the working alliance. Psychotherapy;
Theory, Research, and Practice, _§_, 467-481.

Boleloucky, Z., & Horvath, M. (1974). The SCL-90 rating
scale: First experience with the Czech version in healthy

male scientific workers. Acta Nerva Superior,}u 115-
116.

Craig, T. J.,&Abelhoff,M. (1974). Psychiatric symptoma-
tology among hospitalized cancer patients. American
Journal of Psychiatry, 131, 1323-1327.

Cronbach, L. J. (1951). Coefficient alpha and the internal
structure of tests. Psychometrika, Ag, 297-334.

Crowder, J. E. (1972). Relationship between therapist and

client interpersonal behaviors and psychotherapy outcome.
Journal of Counseling Psychology, 19, 68-75.

152

153

Derogatis, L. R. (1977). The SCL-9O scoring, administra-
pign and procedures manual I (for the revised versiofi).
Baltimore: John Hopkins University School of Medicine,
Clinical Psychometrics Research Unit.

Derogatis, L.IL, & Cleary, P.1L (1977). Confirmation of
the dimensional structure of the SCL-90: A study in
construct validation. Journal of Clinical Psychology,

3;, 981-985.

Derogatis, L. R., Lipman, R. S., Covi, L., & Rickels, K.
(1971). Neurotic symptom dimensions as perceived by
psychiatrists and patients of various social classes.
Archives of General Psychiatry, 34, 454-464.

Derogatis, L. R., Lipman, R. S., Covi, L., & Rickels, K.
(1972). Factorial invariance of symptom dimensions in
anxious and depressive neuroses. Archives of General Psy-
chiatryLQl, 659-665.

Derogatis, L. R., Rickels, K., & Rock, A. F. (1976). The
SCL-90 and the MMPI: A step in the validation of a new
self-report scale. British Journal of Psychiatry, 128,
280-289.

Dickes, R. (1975). Technical considerations of the thera-
peutic and working alliances. International Journal of
Psychoanalytic Psychotherapy, 4, 1—24.

Dietzel, C. S., & Abeles, N. (1975). Client-therapist
complementarity and therapeutic outcome. Journal of Coun-
seling Psychology, 33, 264-272.

Fenichel, O. (1941). Problems ofjpaychoanalytic technique.
New York: Psychological Quarterly.

Fiske, D. W., Cartwright, D. S., & Kirtner, W. L. (1964).
Are psychotherapeutic changes predictable? Journal of
Abnormal and Social Psychology, Q2, 418-426.

Foreman, S. A., & Marmar, C. R. (1985). Therapist action
that addresses initially poor therapeutic alliances in
psychotherapy. American Journal of Psychiatry, 142(8),
922-926.

Freud, S. (1958). The dynamics of transference. In J.
Strachey (Ed. and Trans.), The standard edition of the
complete works of Sigmund Frewd (Vol.12, pp.99-108).
London:Hogarth Press. (Original‘work published 1912)

154

Freud, S. (1958). On beginning the treatment: Further rec-
ommendations on the technique of psychoanalysis. In J.
Strachey (Ed. and Trans.), The standard edition of the
complete works of Sigmund Freud (Vol. 12, pp. 123-144).
London: Hogarth Press. (Original work published 1913)

Freud, S. (1964). .Analysis terminable and interminable.
In J. Strachey (Ed.andTrans.)flThe standard edition of
the complete works of Sigmund Freud (Vol. 23, pp. 209-
253). London: Hogarth Press. (Original work published
1937)

Frieswyk, S. H., Colson, D. B., & Allen, J. G. (1984).
Conceptualizing the therapeutic alliance from a psycho-
analytic perspective. Psychotherapy,.glhn, 460-464.

Gendlin, E. T. (1962). Experiencing and the creation of
meaning. New York: Free Press of Glencoe.

Gendlin, E. T., Beebe, J., Cassens, J., & Oberlander, M.
(1968). Focusing ability in psychotherapy, personality
and creativity. In J.M.Shlien,H.F.Hunt, J. D.
Matarazzo, & C. Savage (Eds.), Research in psychotherapy
(Vol. 3, pp. 217-256). Washington, DC: American Psycho-
logical Association.

Gomez-Schwartz, B. (1978). Effective ingredients in psycho-
therapy: prediction of outcome from process variables.
Journal of Consulting and Clinical Psychology,4§, 1023-
1035.

Greenson,R.R.(1965). The working alliance and the
transference neurosis. Psychoanalytic:guarterly, 33,
155-181.

Greenson, R. R. (1967). The technique and practice of
psychoanalysis. New York: International Universities
Press.

Greenson, R. R., & Wexler, M. (1969). The non-transference
relationship in the psychoanalytic situation. Interna-
tional Journal of Psychoanalysis, 22, 27-40.

Gurman, A. S., 8: Razin, A. M. (1977). Effective psycho-
therapy: A handbook of research. New York: Pergamon
Press.

Hartley, D. E. (1978). Therapeutic alliance and the success
of brief individual psychotherapy. Unpublished doctoral
dissertation, Vanderbilt University, Nashville.

155

Hartley, D. E. (1985). Research on the therapeutic alliance
in psychotherapy. In R. E. Hales 8: A. J. Frances (Eds.),
American Psychiatric Association Annual Review (Vol. 4,
pp. 532-549). Washington, DC: American Psychiatric Press.

Hartley,IL E”.& Strupp,!L (1983). Therapeutic alliance:
a contribution to outcome in brief psychotherapy. In
J. Masling (Ed.), Empirical studies of psychoanalytic
theory (pp. 1-37). Hillsdale, NJ: Earlbaum.

HavensgLJl976). 'Participant Observation. New York:
Jason Aronson.

Horowitz, M. J., Marmar, C. R., Weiss, D. S., DeWitt, K. N.,
& Rosenbaum,lL (1984). Brief dynamic psychotherapy of
bereavement reactions: The relationship of process to
outcome. Mgchives of General Psychiatry, 2;, 438-448.

Hugdahl, K. & Ost, L. (1981). On the difference between
statistical and clinical significance. Behavioral
Assessment, ;, 289-295.

Hull, C. H., & Nie, N. H. (1981). SPSS update 7-9: New
procedures and facilities for releases 7-9. New York:
McGraw-Hill.

Jacobson,N.S.,Follette,W.C.,&Revenstorf,D. (1984).
Psychotherapy outcome research: Methods for reporting
variability and evaluating clinical significance. Behav-
ior Therapy, géJ 336-352.

Kazdin, A. E., & Wilson, G. T. (1978). lEvaluation of behav-
ior therapy: Issues, evidence and research strategies.
Cambridge: Ballinger.

Kernberg,<341975). Borderline conditions and pathological
narcissism. New York: Jason Aronson.

Klecka, W. R. (1970). Discriminant analysis. In N. H. Nie,
C. H. Hull, J. G. Jenkins, K. Steinbrenner, & D. H. Bent
(Eds.), Statistical package for the social scienceg
(2nd ed.). New York: McGraw-Hill.

Kohut,H.,& Wolf,E.S. (1978). The disorders of the self
and their treatment: An outline. International Journal
of Psychoanalysis, g2, 413-425.

Langs, R. (1982). Psychotherapy: A basic text. New York:
Jason‘Aronson.

Lazarus, A. A. (1974). .A multimodal behavioral treatment of
depression. Behavior Therapy, 5, 549-554.

156

Leary/T41957). Inteppersonal djagnopjs of personality.
New York: Ronald Press.

Luborsky, L. (1976). Helping alliances in psychotherapy.
In J. L. Claghorn (Ed.), Successful psychotherapy. New
York: Bruner/Mazel.

Luborsky,L. L.,Chandler,M.,Auerbach,A.,Cohen,J., &
Bachrach, H. M. (1971). Factors influencing the outcome
of psychotherapy: A review of quantitative research.
Psychological Bulletin, 15, 145-185.

Luborsky, L., Crits-Cristoph, P., Alexander, L., Margolis,
M., & Cohen, M. (1983). Two helping alliance methods for
predicting outcomes of psychotherapy: A counting signs
versus a global ratings method. Journal of Nervous and
Mental Disease, Ali, 480-491.

Luborsky,L.L.,Mintz,J.,Auerbach, A., Cristoph, P.,
Bachrach,H.,Todd,T.,Johnson,M., Cohen, M., &
O'Brien, C. P. (1980). Predicting the outcome of
psychotherapy: Findings of the Penn Psychotherapy

Project.Archives of General Psychiatry, 31, 471-481.

Mann,J.(1973). Time-limited psychotherapy. Boston:
Harvard University Press.

Malan, D.IL(1976). The frontiers of brief psychotherapy.
NewYork: Plenum.

Marmar, C. R., Marziali, E., Horowitz, M. J., & Weiss, D. S.
(in press). The development of the therapeutic alliance
rating system. In In S.Greenberg & W.Pinsof(EdsJ,
The psychotherapeutic_process:.A research handbook.New
York:GuilfordPress.

Marziali, E.(1984). ‘Three viewpoints on the therapeutic
alliance; Similarities, differences, and associations
with psychotherapy outcome. The Jougnal of Nervous and
Mental Disease, 172(7), 417-423.

Marziali, E., Marmar, C., & Krupnick, J. (1981). Thera-
peutic alliance scales: Development and relationship to
psychotherapy outcome. .American Journal of Psychiatry,
lggg, 361-364.

Masterson, J. F. (1978). The borderline adult: Therapeutic
alliance and transference. American Journal of
Psychiatry, 135, 437-441.

157

Mintz,J.,Luborsky,L.L.,&Cristoph,P. (1979M
Measuring the outcomes of psychotherapy-findings of the
Penn Psychotherapy Project. Journal of Consulting_and
Clinical Psychology, 31, 319-334.

Moras, K., & Strupp, H. H. (1982). Pretherapy inter-
personal relations, patients' alliance, and outcome in
briefpsychotherapy. Archives of Genepal_§§ychiatry,
g2, 405-409.

Morgan, R., Luborsky, L., Crits-Christoph, P., Curtis, H.,
& Solomon, J. (1982). Predicting the outcomes
of psychotherapy by the Penn Helping Alliance Rating
Method. Archives of General Psychiatry, 32, 397-402.

O'Malley, S. S., Suh, C. S., & Strupp, H. H. (1983). The
Vanderbilt Psychotherapy Process Scale: A report on the
scale development and a process-outcome study.

Journal of Clinical and Consulting Psychology, ii, 581-
586.

Parloff,M.B.,Waskow,I.E.,& Wolfe,B. E. (1978).
Research on therapist variables in relation to process
and outcome. In S.L~Garfield& A. E.Bergin (EdsJ
Handbook of psychotherapy and behavior change: An
empirical analysis (2nd ed.). New York: Wiley.

Rogers, C. R. (1957). The necessary and sufficient
conditions of therapeutic personality change. Journal of
Consulting Psychology, 2;, 95-103.

Ryan,EJL (1973). The capacity of the patient to enter
an elementary therapeutic relationship in the initial
.psychotherapy interviewu Unpublished doctoral
dissertation, University of Michigan.

Sachs, J. S. (1983). Negative factors in brief psycho-
therapy: An empirical assessment. Journal of Consulting
and Clinical Paychology, pl, 557-564.

Salzman.C.,‘Van der Kolk, B. A., & Shader, R. I. (1976).
Marijauna and hostility in a small group setting.
American Journal of Psychiatry, 133, 1029-1033.

Sarnat, J. E. (1975). .A comparison ofjpsychoanalytic and
client-centered measures of initial in-therapy patient
participation. Unpublished doctoral dissertation,
University of Michigan.

158

Shapiro, A. K., & Morris, L. A. (1978). The placebo
effect in medical and psychological therapies. In S. L.
Garfield & A. E. Bergin (Eds.),Handbook of_psychotherapy
and behavior change: An empirical analysis (2nd ed.).
New York: John Wiley. ‘

Sifneos,PJ& (1972). Short-term psychotherapy and
emotional crisis. Cambridge, Ma.: Harvard University
Press.

Sloan, R. B., Staples, F. R., Cristol, A. H., Yorkson, N.
& Whipple, K. (1975). Psychotherapy versus behavior
therapy. Cambridge: Harvard University Press.

Smith, M. L., & Glass, G. V. (1977). Meta-analysis of
psychotherapy outcome studies. American Psychologist,
132, 752-760.

Smith,M.L.,Glass,G.V.,&Miller,D. (1980). Benefits
of psychotherapy. Baltimore: John HOpkins University
Press.

Sterba, R. (1934). The fate of the ego in psychoanalytic
therapy. International Journal of Psychoanalysis, 1934,
lg, 117-126.

Strupp,IL H”.& Hadley,EL W. (1979). Specific versus
nonspecific factors in psychotherapy: A controlled
study of outcome. Archives of General Psychiatry, pp,
1125-1136.

Suh, C. S., & O'Malley, S. S. (1982, June). file
identification of facilitative therapist factors: Method-
ological consideration and research findings of a stggy.
Paper presented at the annual meeting of the
Society for Psychotherapy Research, Smuggler's Notch,
Vermont.

Sullivan,H.S. (1954). The psychiatric interview. New
York: W. W. Norton.

Tarachow,S.(1963). .An intpoduction togpsychotherapy.
NewYork: International Universities Press.

Ticho,lEq Appelbaum, A” Binstock,iN.A.,& Appelbaum, S.
(1971). Panel on recent advances in psychoanalysis
and psychotherapy. Bulletin of the Menninger Clinic, 35,
447-460.

159

Traux, C. B., & Mitchell, K. M. (1971). Research on
certain therapist interpersonal skills in relation to
process and outcome. In A. E.Bergin & S.L. Garfield

(EdsJ, Handbook of psychotherapy and behavior change:
An empirical analysis. New York: Wiley.

Zetzel,E.(1956). Current concepts of transference.
International Journal of Psychoanalysis, 31, 369-376.