i..(e .39
. ‘ [.5731'3PV'20'?! CV»:-
. 5: iv: fit-uxlaffr191r0i3.

.I
vault. :- v;|....‘ 1!!
£039.51. 9...!!! it‘llv V
Ara-19:3: u v 10:: 1.37.335
D. I. Epic...) Pt. I
‘sliluv r - ‘ifl.
its: (:3) 25:5...‘1' .,
fa !§'.'. 3)
at! El: 41
91.3%; V
I... J {1 .o 5.. 503.9!!!
,3 "5‘! l‘plu.‘.P,-!.:
I $1.1 In} .l 7‘
$vifteuioisvf.
: . 2.33!
5.....va
"I‘v’fi [I
I? r. . 1|;
,‘U‘: 19,-“
..‘:l..:

:9;
314.1}.
{11133.}.
t.,.ln.!\.

i p!-
i. .

fifty; 1!.
3!!60t:..f.;\vls.
. via?» 0-

.35.

I?!
.18: >¢Q :- V .
It. .5... r
.i I!» ll.
31!?!» _
'1’?!

6‘:
.1113..-
1‘.

was:

17 ‘5: 5“

9
Pi... prirli. lift-57.0“
It... it. I; I‘vln‘o 3": 'f'fi‘l I
Iai'vl . «‘43.?! v I: .31.!!!
in“, 1.4.3:. XI!-

 

meets

MICHIGAN STATE UNI LIBRARI

m l mu rm n mmnWmMimi:

__I 3 1293 00899 2806

 

 

 

 

 

 

 

 

 

 

r

 

This is to certify that the

dissertation entitled

Initial and later Therapeutic Alliance
and Psychotherapy Outccme

presented by

George Y. Ankuta

has been accepted towards fulfillment
of the requirements for

Ph°D° degree in PSYCh01ogY

Mn... M

Major professor

 

Date £11 /9v

.MSU is an Affirmative Action/Equal Opportunity Insrirun'm:

042771

 

W

T LIBRARY
Michigan State
1 University

xi .1

 

 

PLACE IN RETURN BOX to remove this checkout Irom your record.
TO AVOID FINES return on or before date due.

.pAé’TE DUE DATE DUE DATE DUE
2 .

 

    

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

m

MSU Is An Affirmative Action/Equal Opportunity Institution
emana-M

 

 

 

 

Initial and Later Therapeutic Alliance
and Psychotherapy Outcome

BY

George Y. Ankuta

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of Psychology

1992

ABSTRACT

Initial and Later Therapeutic Alliance

and Psychotherapy Outcome

BY

George Y. Ankuta

Forty-four clients seen for psychotherapy as part of
the MSU Psychotherapy Research Project were rated on the
California Psychotherapy Alliance Scales (CALPAS) and the
Therapist Action Scale (TASR). Four 20 minute segments were
sampled from the session audiotapes of each case. These
samples represented segments from the third session, a
session from the second and the third quartiles of therapy,
and one session toward the end of treatment. All clients
also completed the Symptom.Checklist 90 (SCLSO-R), a
symptom.inyentory. Presenting problems included moderate to
severe depression and anxiety as well as other psychological
dysfunctions. All clients completed at least 10 sessions,
and some completed as many as 50 sessions. Clients were
predominantly female (76%) and ranged in age from.20-59

years. Therapists were graduate students in clinical

psychology with 2-7 years of prior therapy

experience. The primary theoretical orientation of the
therapists was psychodynamic although other treatment
modalities were present. Therapists and clients were blind
to the hypotheses and purposes of the study.

In this study a relationship was hypothesized between
the level of initial and later alliance, and symptom
reduction. This hypothesis was not supported.
Additionally, it was hypothesized that the quality of the
therapeutic alliance would be affected by therapist actions
related to interpretations; This hypothesis was also not
supported.

nest patients reported fewer symptoms after therapy.
It was found that patients who had a strong initial alliance
were more likely to manifest a strong alliance later in
therapy. Therapists who were more active early in therapy
were more likely to be active later in therapy. Therapeutic
alliance and therapist actions were correlated within
session, and that correlation increased over time.

Overall, findings imply that the importance of the
initial alliance in predicting outcome in psychotherapy may
need to be questioned. Evaluation of the California
Psychotherapy Alliance Scales and the Therapist Action
Scales suggested problems concerning reliability of the
measures. Further work with the Alliance and Action Scales

is also suggested to improve their reliability.

To my parents

ACKNOWLEDGMENTS

I would like to thank Dr. Norman Abeles for his support
and guidance throughout this project. His student-centered
approach and generosity with his time provided for a strong
and exciting mentor relationship. I greatly appreciated Dr.
Jack Hunter's engagement with the project as well as his
expert statistical and methodological advice. Kudos to Dr.
John Hurley for his careful and detailed reading. I thank

Dr. Jehn Mckinney for his thoughtful comments.

ii

TABLE OF CONTENTS

LIST OF TABLES ........................................ ...iv
LIST OF FIGURES.. ......................................... v
INTRODUCTION ............................................... 1
History of Therapeutic Alliance Construct ................ 1
Therapist Actions that Affect the Therapeutic Alliance
in Psychotherapy ....................................... ..9
Qualities of "Good" and Effective Therapy ................ 9
The Effect of Transference Interpretation on the
Alliance ................................................ 12
Alliance and Psychotherapy Outcome ...................... 20
The Effect of Transference Interpretation on Outcome....21
Summary. .................... .................. .......... 24
Hypotheses .............................................. 25
METHOD .................................................... 28
Participants ............................................ 28
Procedure ............................................... 29
measures ................................................ 32
CALPAS: California Psychotherapy Alliance Scales ........ 32
TAS: Therapist Action Scale ............................. 4O
SCL90-R: Symptom Checklist 90 Revised ................... 46
RESULTS......... ................. .......................50
California Psychotherapy Alliance Scale (CALPAS) ........ 50
Revision of CALPAS ...................................... 60
Therapist Action Scale Revised (TASR) ..... ..............63
Revision of the Therapist Action Scale Revised .......... 82
Hypotheses 1 and 4. ......... ............................87
Hypotheses 2 and 3 ...................................... 91
Some Related Findings ................................... 95
DISCUSSION ............................................... 100
Alliance and Outcome ................................... 102
Implications for Theory and Practice....... ....... .....116
Implications for Future Research ...................... .118
APPENDIX A - Scale Construction ....................... ...119

APPENDIX B - Some Technical Aspects of the Data Analysis.123

REFERENCES .................. . ............................ 125

iii

LIST OF TABLES

TABLE
1 California Psychotherapy Alliance Scale .............. 33
2 Therapist Action Scale - Revised ..................... 41
3 Exploratory Factor Analysis of the California
Psychotherapy Alliance Scale ......................... 51
4 Confirmatory Factor Analysis of the California
Psychotherapy Alliance Scale ......................... 53
5 California Psychotherapy Alliance Scale Interrater
Reliability .......................................... 58
6 California Psychotherapy Alliance Scale Test-retest
Reliability .......................................... 59
7 California Psychotherapy Alliance Scale Reliability
Across Time and Rater ................................ 61
8 California Psychotherapy Alliance Scale Interrater
Reliability .......................................... 62
9 Confirmatory Factor Analysis of the California

10

11
12
13
14
15
16
17
18
19
20

21

Psychotherapy Alliance Scale (CALPAS) and the

Therapist Action Scale Revised (TASR) at Four Time
Points and Pretherapy and Posttherapy Symptom

Checklist 90 Global Severity Index Scores (SCL90-R

GSI) ................................................. 64
Confirmatory Factor Analysis of the California
Psychotherapy Alliance Scale (CALPAS), the Therapist
Action Scale Revised (TASR), Support, and Patient
Negative Contribution at Four Time Points and the
Symptom.Checklist 90 Global Severity Index Scores

(SCL90-R GSI) ..................................... ...69
Exploratory Factor Analysis of the Therapist Action
Scale Revised ........................................ 75
Confirmatory Factor Analysis of the Therapist Action
Scale Revised ........................................ 76
Therapist Action Scale Revised Interrater

Reliability ......................................... .79
Therapist Action Scale Revised Test-retest

Reliability ......................................... .80
Therapist Action Scale Revised Reliability Across

Time and Rater ..................................... ..81
Therapist Action Scale Revised Interrater

Reliability ....................................... ...83
Path Analysis 1 .................................... ..89
Path Analysis 2 ...................................... 93

Analysis Variables Means and Standard Deviations.....96
The correlations between early and later alliance,

and symptoms .................................... ....106
multiple regression of posttherapy symptoms onto
initial symptoms, initial alliance, and final

alliance ............... ........ ......... . ......... ..110

iv

LIST OF FIGURES

FIGURE
1 PATH Model 1 ........................................ 85
2 PATH Model 2 ........................................ 86

3 Pretherapy SCL90-R GSI by Posttherapy SCL90-R GSI. . .97

INTRODUCTION

History of Therapeutic Alliance Construct

The relationship between the client and therapist has a
primary and central role in the process of psychotherapy.
This relationship has an affective component and a rational,
cognitive, ego function component. It is a necessary
condition for the operation of the therapeutic process.
Freud (1958/1913) wrote on the affective component of the
client therapist relationship,

It remains the first aim.of the treatment to attach him

to it and to the person of the doctor. To ensure this,

nothing need be done but to give him.time. If one
exhibits a serious interest in him, carefully clears
away resistances that crop up at the beginning and
avoids making certain mistakes, he will of himself
form.such an attachment and link the doctor up with one
of the imagos of the people by whom.he was accustomed

to be treated with affection. (pp. 139-140)

This often quoted, primary statement about the development
of the relationship between client and therapist in therapy
speaks to the stance of the therapist toward the patient.

It addresses aspects of the therapist's behaviors in the
therapy as well as the transference of positive feelings
from.past nurturing, loving relationships onto the
therapist. Vestiges of these lines of thinking are found in

virtually all subsequent theories concerning the

relationship between the client and the therapist.

Sterba (1934) discussed the role of the ego in the
analytic relationship in therapy. From a technical
perspective the analyst strives “to affect an alliance with
the ego against the powerful forces of instinct and
repression..." (p. 120). The analyst enlists a portion of
the patient's ego to collaborate with the analyst in the
project of the analysis. Sterba commented on the affective
component of the relationship that "positive transference is
needed so transitory strengthening of the ego is possible
through identification with the analyst." (p. 121). Sterba
(1934) went on to summarize.

From the onset the patient is called upon to cooperate

with the analyst against something in himself. Each

separate session gives that analyst various
opportunities of employing the term.'we', in referring
to himself and to the part of the patient's ego which

is consonant with reality. (p. 121).

The therapeutic relationship is built in therapy by the
analyst and the patient. The supportive function and the
collaboration in rational evaluation are elemental to the
relationship between the analyst and the patient's ego in
psychotherapy.

Zetzel (1956) emphasized the role of the ego in
analysis. She stated that "mature" ego functions are
necessary for a sound alliance and therefore traditional

psychoanalytic treatment. She referred to "existing and

continued functioning of adequate ego strength to maintain

therapeutic alliance at an adult level." (p. 372). She
believed as Sterba did that positive transference would help
strengthen the ego during therapy. She recommended
variations in technique, such as using greater support to
foster positive transference and strengthen defenses, when
working with borderline and psychotic patients because their
ego weaknesses and negative transference preclude the
spontaneous development of a genuine therapeutic alliance.
Rogers (1959) considered the relationship between the
client and therapist to be of primary importance in the
process of psychotherapy. He discussed six necessary and
sufficient conditions for personality change in
psychotherapy. All six directly involve aspects of the
relationship between the client and therapist without using
the term alliance. For therapeutic change to occur the
client and therapist must be in contact and have impact on
each other. The client in a state of “incongruence"- a
discrepancy between his experiences and self-image, relates
to the therapist who is in a state of “congruence". The
therapist is honest and genuine about his actual feelings
and awareness of himself in the relationship with the
client. The therapist offers the client unconditional
positive regard or acceptance of the client as he or she is.
The therapist experiences an empathic understanding of the

client's inner world which he or she communicates to the

client. The client must be able to receive the therapist's
unconditional positive regard and empathy to some degree.
When these six conditions are met the client becomes more
and more free to express feelings within the context of the
therapy. Eventually the client will become more congruent,
less defensive, and more open to experience.

Stone (1961) also placed a high importance on the
relationship between the client and the therapist in
psychotherapy. Stone viewed the primary unconscious
significance of the analyst to be the "mother of separation"
which makes the analytic relationship a recreation of the
mother-child relationship. The analytic situation is,
"separation and deprivation of primitive gratification, in
the context of verbal intimacy." (p. 107). Physical
dependence is replaced by speech. The structure of the
situation maximizes transference and the potential for
understanding. A central role of the analyst is to provide
a human context in which to convey understanding. Given the
enduring rigors of the situation, Stone believed that
certain legitimate transference gratifications should be
granted in the interest of maintaining a human context for
the analysis.

Greenson (1965) developed a most elaborate description
of the relationship between the client and therapist. He

adopted the terms "working alliance“ and "real

relationship". By "working alliance" he meant,

the relatively non-neurotic, rational rapport which
the patient has with his analyst....it centers on the
patient's ability to work in the analytic situation....
It can be seen at its clearest when a patient is in the
throes of an intense transference neurosis and yet can
still maintain an effective working relationship with
the analyst....his conscious and rational willingness
to cooperate (p. 192).

This definition focuses on the ability of the patient and
the analyst to work together constructively on the task of
analysis, based on rational ego functions.
Greenson (1965) differentiated "working alliance" from
the "real relationship" between the patient and analyst.
The term 'real' in the phrase 'real relationship' may
mean realistic, reality oriented, or undistorted as
contrasted to the term.'transference' which connotes
unrealistic, distorted, and inappropriate. (p. 217)...
There is no transference reaction, no matter how
fantastic, ‘without a germ.of truth, and there is no
realistic relationship without some trace of
transference fantasy. All patients in psychoanalytic
treatment have realistic and objective perceptions and
reactions to their analyst alongside of their
transference reactions and their working alliance. (p.
219).
This 'real' relationship includes transferential and current
emotional reactions of the patient to the analyst as opposed
to the ability to work together in the analysis.
Luborsky (1976; Luborsky, Crits-Christoph, Nintz &
Auerbach, 1988) defined the therapeutic relationship as the
I'helping alliance" - the patient's experience or belief that

the treatment or the relationship with the therapist is

helpful or potentially helpful. The helping alliance is
composed of a "Type 1" alliance and a "Type 2" alliance
(Luborsky, 1976). Type 1 alliance is based on the patient's
experience of the therapist as supportive and helpful. Type
2 alliance is based on a sense of patient and therapist
working together in a joint struggle against things
troubling the patient. The helping alliance is viewed as
arising from.positive transference, linking the therapist
with positive past imagos. The helping alliance is fostered
by a warm, friendly, and sympathetic stance on the part of
the therapist, and by the therapist's emphasis on the "we“
bond.

The necessity of a new construct called alliance has
been questioned. Brenner (1979), after reviewing Zetzel
(1956) and Greenson (1965), asserted that the concept of
the therapeutic alliance is based on nothing more the
patient's positive transference with the therapist. The
special name of therapeutic alliance and special treatment
are not necessary. Brenner (1979, p.146) commented, "when a
patient's ability to cooperate in the work of analysis is
compromised, it is often a consequence or manifestation of
transference and that it is best dealt with by correct
understanding and consistent interpretation."

Frieswyk, Colson, and Allen (1984; Frieswyk, Allen,

Colson, Coyne, Gabbard, Horwitz & Newcom, 1986) also

discussed how conceptions of the alliance are confounded.
Since the working relationship, and Greenson's "real
relationship“ or accurate perceptions of the therapist, are
influenced by transference, measures of therapeutic
alliance based on patient's perceptions (Luborsky, 1976)
will be confounded by transference. measures of alliance
that are based on patient and therapist activity will be
confounded by the therapist's technique.

Frieswyk et. al. (1984, 1986) calls for a definition of
alliance that separates it from transference and technique.
He asserted that Greenson did not distinguish between
"working relationship" and “working alliance“. Frieswyk
wanted to differentiate these terms by considering working
relationship to be aspects of the relationship that make
collaboration possible. He defined the “therapeutic
alliance“ specifically as "the patient's active
collaboration in the work of psychotherapy or
psychoanalysis." (1984, p. 460).

Although the concept of the therapeutic alliance began
in the psychoanalytic literature it was broadened to
encompass all forms of therapy by Bordin (1979). He defined
the alliance as including three features: an agreement on
goals, an agreement on tasks, and the development of
bonds.

Patients come to therapy with things in their lives

that they want to change or goals. Psychodynamic therapies
involve increasing understanding of one's inner life and how
it contributes to one's difficulties, whereas behavioral
therapies focus on changing specific acts or behaviors of
the individual. The tasks of therapy vary depending on the
patient's needs, degree of psychological mindedness, and
the mode of therapy, but usually attempts at accurate
self-observation are involved. The task in psychoanalysis
is free association while behavior therapies involve
behavioral monitoring.

The bonds or the "human“ relationship between the
patient and the therapist are expected to be more of a
factor in long term.psychodynamic therapies than in brief
behavioral therapies. Greater trust is required for the
elaborate revelation of inner experience required in dynamic
therapy. All therapies require some level of agreement on
goals and tasks, and development of bonds, which make the
concept of alliance relevant to all therapies (Bordin,

1979).

Therapist Actions that Affect the Therapeutic Alliance in

Psychotherapy

Qualities of "Good" and Effective Therapy

Lafferty, Beutler and Cargo (1989) evaluated

differences between more and less effective
psychotherapists. Pretreatment and posttreatment symptom.
level was measured with the Symptom Check List 90 Revised
(SCL90-R) (Derogatis, 1983). Therapist variables within
the treatment relationship and extra treatment variables
such as demographics were measured by posttherapy
inventories completed by the patient and the therapist. A
stepwise discriminant function analysis predicting
symptomatic improvement revealed that the best predictors of
therapist effectiveness were related to in-therapy factors
such as empathy rather then therapist attributes such as
credibility and theoretical orientation. What the therapist
does in the therapy seems to be more important than who the
therapist is, re-emphasizing the importance of therapeutic
technique.

Luborsky, McLellan, woody, O'Brien and Auerbach
(1985) conducted a study of therapist success in reducing
clients’ symptoms and its determinants. Potential factors
influencing success explored were: patient factors,
therapist factors, patient-therapist relationship factors,
and therapy factors. Three therapists using
cognitive-behavioral methods, three therapists using
supportive-expressive methods, and three therapists using
drug counseling methods, each treating about ten patients

were studied. A breakdown of patient background demographic

10

characteristics revealed no significant differences.
Therapist factors such as the therapist's level of interest
in helping patients, and the therapist's level of
psychological health did not account for differences in
patient outcome. Level of helping alliance was found to be
correlated with outcome, but did not account for all of the
outcome variance.

Luborsky et. al. (1985) also discussed results related
to therapist technical activity. A measure of purity of the
therapist's technique was developed which indicated the
extent to which the therapist, if cognitive-behavioral,
used cognitive-behavioral techniques proportional to
techniques associated with the other two forms of therapy in
this study. For cognitive-behavioral and
supportive-expressive therapy greater purity ratings were
associated with better seven month posttherapy outcome.
From.these findings, and the fact that no particular form
of therapy has generally been proven to be more effective
than another, Luborsky et. al. (1985) concluded that the
effectiveness of a particular therapy may vary depending on
how effectively the therapist executes the intended
technique, and to an even greater extent the degree to
which a helping alliance is formed. Skillfully conducted
therapy may lead to a better alliance and thus a better

OUECOM .

11

In another study of therapeutic technique O'Malley,
Foley, Rounsaville, Watkins, Sotsky, Imber, and Elkin
(1988) evaluated whether therapist technical competence in
conducting interpersonal psychotherapy of depression is
related to outcome. Skill in conducting interpersonal
psychotherapy was found to be associated with better
outcome.

In addition to the demonstrated effectiveness of
particular therapeutic techniques, there are particular
techniques of dynamic therapy which dynamic therapists are
likely to rate as part of 'good' therapy sessions. Hoyt
(1980) asked psychoanalytically trained therapists to rate
psychotherapy sessions as 'good' or 'bad'. The sessions
were also rated on the Therapist Action Scale (Hoyt,
Narmar, Horowitz, and Alvarez, 1981). The therapist
activities measured by the scale were correlated with the
good and bad ratings. Therapist activities correlated with
good ratings were: therapist emphasizing patient's
expression of thoughts and feelings, therapist discussing
patient's reactions to the therapist, and therapist
discussing the meaning of the patient's reactions. Other
therapist activities correlated with good ratings were:
discussing patterns or links between the patient's past and
present life, and exploring the patient's self-concept. A

subsequent replication study (Hoyt, Xenakis, Narmar a

12

Horowitz, 1983) demonstrated comparable results except that
the therapist activity "reactions to therapist discussed"

was not correlated with a good rating.

The Effect of Transference Interpretation on The Alliance
In his discussion of the helping alliance Luborsky

1976, p. 96) commented, "The strength of the helping
(alliance varies from.time to time, especially in relation
to surges in transference which may then be diminished by
the therapist's interventions." weakening in the helping
alliance caused by negative transference is expected to be
repaired by interpretation of the negative transference.
Foreman and Marmar (1985) conducted a pilot study
concerning therapist actions that might improve the
therapeutic alliance which was demonstrated to be poor in
the beginning phase of therapy. Six female patients
suffering from.post traumatic stress disorder or adjustment
disorder after the death of a parent or husband and who had
demonstrated initially poor therapeutic alliances, were
studied. Three had improved alliances at the end of
time-limited psychotherapy, and three did not. The
alliance was measured with the California Therapeutic
Alliance Rating Scale (CALTARS) (Harmer, Horowitz, weiss &

Marziali, 1986), which was an earlier version of the

13

(CALPAS) California Psychotherapy Alliance Scale used in
this study. The CALTARS was rated by blind raters whose
reliability was monitored. Therapist technical activities
were measured on a brief scale of Therapist Actions created
by Foreman and Marmar (1985), rated by one person, not
blind to process and outcome ratings.

Therapist activities were evaluated according to
standard analytic technique of interpreting: defenses,
anxiety, and the underlying feeling or impulse (Malan,
1979). The activities measured involving the
patient-therapist relationship were the extent to which the
therapist addressed: defenses, problematic feelings in
relation to the therapist, problematic relationship
patterns, problematic powerful images, problematic
vulnerable images, and the "triangle of punishment.“ The
"triangle of punishment" is a term constructed by Foreman
and Marmar (1985) which is defined as a manifestation of
"the patient's need for self-punishment to assuage the guilt
over feelings of anger or responsibility for another
person's suffering." (p. 925). The activities measured
involving the patient-other relationships were the extent to
which the therapist addressed the following: defenses,
problematic feelings in relation to the therapist,
problematic relationship patterns, problematic powerful

images, problematic vulnerable images, and the triangle of

14

punishment.

The therapist actions that were found to differentiate
best the group with improved alliance from.the unimproved
alliance group were: addressing the patient's defenses in
handling feelings in relation to the therapist and others,
addressing the triangle of punishment, addressing
problematic feelings in relation to the therapist, linking
problematic feelings in relation to the therapist with
defenses in handling these feelings.

Therapist actions that failed to distinguish improved
and unimproved alliance groups were: addressing problematic
feelings in patient-other relationships, addressing
problematic powerful images or problematic vulnerable
images. Interpretations linking the patient's reactions to
parents in the past with reactions to the therapist in the
present were found to be infrequent therapist actions.

These results lead Foreman and Marmar to conclude that
specific therapeutic approaches can improve a poor
therapeutic alliance. In unimproved therapeutic alliance
cases the therapists deviated from standard technique by
working with reality problems rather then exploring feelings
about the treatment situation. These findings were
considered consistent with current standard psychoanalytic
theory and technique.

Lansford (1986) evaluated the effects of “weakenings”

15

and repairs of the therapeutic alliance on outcome in
short-term psychotherapy. Weakenings were defined as a
negative response to therapy and the therapist, problems
with talking in therapy, or problems with termination.
Repairs were defined as the extent to which identified
weakenings were worked with and resolved by patient and
therapist. Following Bordin (1979) Lansford commented,

that the process of building an alliance and working out
weakenings in the alliance is the therapy. She found a
positive correlation between level of working alliance and
outcome (5 = .81 p,< .05). Healthier patients were found to
be more able to repair weakenings in the alliance, and
patient factors were found to be more important then
therapist activity. Interventions addressing transference
allusions did not lead to superior outcome in this study,

in contrast with Helen (1976) and Marziali (1984a).

However, it is clear that the alliance changes throughout
therapy and that changes in the alliance are integrally part
of the therapy experience.

Gaston, Harmer, Gallagher, and Thompson (1990)
studied variations in the alliance over time and the
relationship between symptom change and the alliance. A
sample of depressed older adults, 16 men and 44 women,
mean age 67 years, was randomly assigned to a behavior

therapy condition, a cognitive therapy condition, a

16

dynamic therapy condition, or a delayed treatment control
group. Therapy was from 16 to 20 sessions with Ph.D. level
therapists with at least one year of post doctoral
experience in the modality of therapy they were providing.
The Beck depression inventory and the California
psychotherapy alliance scales were completed after the 5th,
10th, 15th and final session.

Gaston et a1. (1990) found that in a multiple
regression analysis the alliance was a more powerful
predictor of outcome the later in the therapy it was
measured. However, repeated measures analysis of variance
within treatment conditions revealed no statistically
significant changes in the level of alliance over time.
These findings suggest that different individuals have
different courses to their alliances in psychotherapy. Some
may improve; some may worsen, and some may stay constant.
It is suggested that future studies attempt to identify
specific therapist interventions associated with alliance
improvement and deterioration.

A second finding of this study was that for each of the
three types of therapies the alliance scores explain large
amounts of outcome variance over and above initial symptom
level and amount of symptom change. This finding suggests
that the alliance is an active ingredient in psychotherapy

that contributes to outcome. Replication of this study

17

using independent clinical judge raters of the alliance and
patient populations other then older adults is recommended.
Gabbard, Horowitz, Frieswyk, Allen, Colson,
Newcom, and Coyne (1988) investigated the effect of
therapist actions on the therapeutic alliance with
borderline patients. In this single case experimental
design six sessions in one therapy with a borderline patent
were evaluated. The results clearly suggested that the
therapist's interpretation of the transference increased
this patient's collaboration in the treatment. The authors
speculated that transference interpretations were helpful in
this treatment due to the patient's punitive superego,
evidenced in the testing findings, which was likely
projected onto the therapist. Transference interpretations
are expected to enable the patient to see that his fears
arise from internal irrational sources rather then from real
threats from the therapist. This realization makes greater

collaboration and openness possible.

Therapist Actions that Negatively Affect the Therapeutic

Alliance

Langs (1975, p. 80) coined the term "therapeutic

misalliances" which "arise primarily out of unresolved

18

intrapsychic conflicts-- inappropriate instinctual drive
needs, and superego and ego disturbances--". Patients
frequently act in response to transference in ways that are
incongruent with a productive therapeutic alliance, and
interfere with the progress of the treatment. When the
patient's image of the therapist is negative, potential
benefits of positive identification with him/her, and
insights from interpretations are less likely. In some
cases the patient's actions play into the therapist's
countertransference reactions, and the misalliance is
perpetuated throughout the treatment, never resolved, and
treatment fails.

Langs commented that when countertransference locks the
therapist into misalliance he/she must work that out in
his/her personal therapy. The patient’s contribution to
misalliance is handled through analysis of the negative
transference to permit the therapeutic work to continue.
The patient's needs to lessen conflict and anxiety and
repeat and master past traumatic relationships must be
countered with drive toward health. Repetition must be
replaced with insight. Langs (1975, p. 103) commented, “in
pathological interactions lie the seeds of growth and
constructive change."

Finell (1987) reviewed work on the negative therapeutic

reaction in psychoanalysis. This resistance to recovery

19

often involves "a character defense rooted in negativism.
and as a chronic and deeply embedded negative transference."
(p. 487). Recovery is further threatened by the
countertransference problems these defenses promote.

She further reviewed Freud's writing on this issue,
citing a fear of recovery and negative attitudes toward the
physician as significant factors in the development of the
negative therapeutic reaction. underlying negative
transference and fear of recovery was guilt and the need for
punishment. Later in his work Freud discussed the need for
punishment as identification with a guilt-inducing parent.
Freud ultimately recommended working with these masochistic
tendencies using interpretations aimed at addressing the
need for punishment. ’

Finell commented on Freud's and Klein's shared view of
early aggression as a manifestation of the death instinct.
Any childhood deprivations confirming the aggression
projected onto the parent increases the likelihood of the
formation of an excessively punitive superego and strong
negative transference with the therapist.

Finell commented on Olinick's (1964) interpersonal
explanation of the negative therapeutic reaction. The
intense negative transference is viewed as covering positive
feelings which are feared in cases where they promote

regression to helpless symbiotic attachment to a destructive

20

mother. Promoting greater understanding of transference
reactions, and defenses against feelings toward their
therapist, is expected to lead to strengthening of the

therapeutic alliance.

Alliance and Psychotherapy Outcome

Many studies of therapy process and outcome have shown
that high levels of therapeutic alliance are correlated with
favorable outcome in psychotherapy. Luborsky (1976;
Luborsky et a1. 1985) found evidence of a helping alliance
in early sessions of individuals who demonstrated a high
level of improvement in psychotherapy, but not in those who
showed a low level of improvement. Gomes-Schwartz (1978)
found that therapy outcome was most effectively predicted by
the client's willingness and ability to become actively
involved in the therapeutic interaction'with the therapist.
Marziali, Marmar, and Erupnick (1981: Marziali, 1984b)
found that clients who maintained high levels of alliance
achieved the greatest benefits from.psychotherapy. Morgan,
Luborsky, Crits-Cristoph, Curtis, and Solomon (1982)
found that the level of helping alliance is as good or

better a predictor of successful therapy outcome than most

21

of the pretreatment measures examined in the Penn
Psychotherapy Project.

StrupD (1980), and Hartley and StruDD (1983) also
found that the therapeutic alliance- particularly the early
level of alliance, was predictive of favorable outcome.
Saltzman, Luetgert, Roth, Creaser, and Howard (1976),
Eaton, Abeles, and Gutfreund (1988), and Klee, Abeles and
Muller (1990) also found relationships between early levels
of the therapeutic alliance and outcome of therapy. The
relationship of alliance to outcome is a well established

result in the psychotherapy research literature.

The Effect of Transference Interpretation on Outcome

In one of the most major studies of the effectiveness
of transference interpretations in psychodynamic
psychotherapy outcome, Malan (1976) developed a content
analysis method that enabled him.to count interpretations
about parents, the therapist, and others, as well as
interpretations linking reactions to therapist with
reactions to a parent. He demonstrated that more frequent
transference interpretations linking reactions to_past

relationships to present relationships are related to more

22

positive outcome.

Kernberg (1976) discussed the findings of the Menninger
Foundation's psychotherapy research project. He reported
that patients with initially low ego strength, who were
treated by therapists rated to be more skillful, and
focusing more on interpretation of the transference,
demonstrated greater improvement in therapy. It was
concluded that the lower the initial ego strength of the
patients the more important it is to focus on transference
interpretation to achieve a favorable outcome. For patients
with low ego strength and borderline personality structure,
focusing on the transference, particularly negative
transference, is indicated to prevent blocks in the
therapeutic relationship, transference acting out, and
poor outcome.

Marziali (1984a) replicated Malan's finding and
corrected a major methodological flaw by rating tape
recordings of sessions rather then subjective process notes.
More favorable outcome as measured on psychodynamic scales
was associated with more frequent therapist offered
interpretations linking reactions to the therapist with the
patient's reactions to parents in the past.

Horowitz, Marmar, weiss, Dewitt, and Rosenbaum.
(1984) studied the effects of dispositional and process

variables on outcome in a sample of bereaved patients in

23

time limited dynamic psychotherapy. The dispositional
variables were motivation for dynamic psychotherapy. and
developmental level of the self-concept. The process
variables studied were therapeutic alliance and therapist
action. Neither alliance nor therapist action were related
to outcome as measured by symptom relief. There was an
interaction effect between motivation for treatment and the
therapist-parent linking transference interpretations.
Extensive transference interpretations were related to
symptomatic improvement on the SCL-90 for highly motivated
clients, but less motivated clients showed poorer outcomes.
Piper, Debbane, Bienvenu, Carufel, and Garant
(1986) studied the effects of the object focus of
transference interpretations on outcome of psychotherapy.
The transference interpretations were rated based on who was
addressed by the interpretation, mother, father, sibling,
non-specific family member, therapist-parent,
other-therapist, other-parent, etc. Multiple outcome
measures indicated a sporadic correlation with percentages
of each particular type of transference interpretation.
Nine of one hundred and seventy correlations performed were
significant, which is about what would be expected due to
chance. It may be that when the transference
interpretations are broken down into this many categories

the effects are not strong enough to demonstrate

24

significance.

Grits-Christoph, Cooper and Luborsky (1988) evaluated
the effect on outcome of the accuracy of the therapist's
interpretations based on Luborsky's core conflictual
relationship theme method. The major expectation of the
research was confirmed. A moderately strong association was
found between the accuracy of interpretations linking the
main.wishes of the client with the anticipated responses
from.others, and treatment outcome. The result held even
when the quality of the alliance and general errors in
technique were controlled. Contrary to expectation,
interpretations were not found to be more effective in the
context of a positive alliance. These findings indicate

that specific technique factors impact on outcome.

Summary

Up to now we have reviewed the history of
thetherapeutic alliance construct. The strengthening and
weakening of the alliance over the course of therapy. and
its relationship to outcome have been discussed.

The qualities of good and effective therapy have been
shown to lie in the quality with which the particular
therapeutic technique is executed by the therapist. The

effect of transference interpretation on the therapeutic

25

alliance, and therapy outcome were discussed. Therapist as
well as patient negative factors in the alliance were
explored.

In this study the effects of therapist technical
activities such as transference interpretations on the
therapeutic alliance were evaluated, because a strong
alliance is indicative of a favorable outcome. A series of

hypotheses follow.

Hypotheses

Hypothesis 1: A high level of therapeutic alliance in the
first session as measured by the California Psychotherapy
Alliance Scale (CALPAS) (Marmar & Gaston, 1989)'will
predict symptom reduction at the end of treatment as
measured by the Symptom Check List 90 Revised (SCL90-R)

(Derogatis, 1983).

Hypothesis 2: An initially poor therapeutic alliance as
measured by the California Psychotherapy Alliance Scale can
be improved by the end of treatment as a result of these
therapist technical activates measured by the Therapist
Action Scale Revised (TASR) (Hoyt, et al, 1981):

a) addressing the patient's defenses.

b) addressing the patient's guilt and expectation of

26

punishment.

c) addressing the patient's problematic feelings in
relation to the therapist.

d) linking the problematic feelings in relation to

the therapist with the patient's defenses.

Hypothesis 3: An initially good therapeutic alliance as
measured by the California Psychotherapy Alliance Scale can
be damaged by therapist technical activity measured by the
Therapist Action Scale Revised. The alliance can be
weakened by the therapist failing to:

a) address the patent's defenses.

b) address the patient's guilt and expectation of

punishment.

c) address the patient's problematic feelings in
relation to the therapist

d) link the problematic feelings in relation to the

therapist with the patient's defenses.

Hypothesis 4:

a) After the initial level of alliance is accounted for as
measured by the California Psychotherapy Alliance Scale
(CALPAS), the final level of alliance as measured by the
CALPAS will predict more outcome variance in symptom

reduction as measured by the Symptom.Check List 90 Revised

27

(SCL90-R).

b) The level of the therapeutic alliance as measured by the

CALPAS during the final session of therapy is
predictor of symptom reduction during therapy
the SCL90-R than the level of alliance during

session of therapy as measured by the CALPAS.

a better
as measured by

the first

METHOD
Participants:

Clients: The participants for this study were clients
of the Michigan State University Psychological Clinic. The
clinic is a training and research arm.of the Psychology
Department. It is a nonprofit, fee for service (income
adjusted), setting serving an outpatient, nonstudent
population in the Mid-Michigan area. The clients presenting
problems included moderate to severe depression and anxiety
as well as other psychological dysfunctions. The clients
were predominantly working and middle class. Of all clients
receiving therapy at the clinic 96% agreed to participate in
the Michigan State University (MSU) Psychotherapy Research
Project.

During an eight year period from.l976 to 1986 data were
collected on 121 adult therapy cases. One hundred cases had
pre-and-post-therapy data and session tapes needed for this
study, and had completed at least 10 sessions of therapy.
Cases with fewer then 10 sessions were not included in this
study. Fifty client-therapist dyads were randomly selected
from.these 100 cases for inclusion in the study. In 30% of

the cases included the same therapist treated two cases.

28

29

and thus in those cases the same therapist was rated twice.
'They ranged in age from 20 to 57 years with a mean of 29.8
and SQ = 8.69; 76% were female.

Therapists: The therapists were clinical psychology
graduate students at Michigan State University working at
the Psychological Clinic. The therapists included 2nd to
4th year clinic trainees as well as advanced trainees with
several years post-masters degree experience. All
therapists had completed graduate courses addressing
theories of psychotherapy and assessment courses, and all
had prior mental health experience. Overall their
experience ranged from 2-7 years. The predominant
theoretical orientation of the therapists was psychodynamic,
although other orientations to treatment were represented.
Since the study was conducted after therapy had been
completed, the therapists and clients were blind to the

hypotheses and purposes of the study.

Procedure:

During the initial intake interview at the MSU
Psychological Clinic clients were asked to participate in
the ongoing MSU Psychotherapy Research Project and informed
that their choice to participate or not participate would

not affect the services that they would receive. Of all

30

clients, 96% agreed to participate. Clients were advised
that if they chose to participate and complete the
pretherapy and posttherapy forms they would be given a
stipend equal to 10 percent of their fees paid for therapy
up to 50 dollars. Clients who agreed to participate
completed a consent form and the SCL90-R before therapy
began and after therapy was completed.

The therapists were recruited at the beginning of the
school year. All, 100% of the therapists agreed to
participate and were aware that these research audiotapes
would not be used to evaluate their personal performance and
progress as therapists. In the 4% of the cases in which the
clients chose not to participate, the therapists could not
participate.

Clients and therapists were instructed that copies of
the audio recording of their first, third, and every fifth
session thereafter would be included in the Psychotherapy
Research Tape library. Therapists were instructed to
forward these tapes to the research assistant for
duplication and filing under anonymous subject
identification code numbers. Completed cases were never
used until two years after the therapy had been completed to
further insure confidentiality.

The fifty cases selected from the MSU Psychotherapy

Research Project database were rated on the California

31

Psychotherapy Alliance Scales (CALPAS) (Marmar & Gaston,
1989) and the Therapist Action Scale (TASR) (Hoyt et al.,
1981). From the session tapes of each case four 20 minute
segments were sampled. One segment from the first session,
one segment from the third session, one segment from the
middle of therapy (40 to 60th percentile), and one segment
toward the end of treatment (80th percentile or more). At
each of these four time points sampled the segments were
randomly chosen from the beginning middle or end of the
therapy hour. The segments were placed in random order on
master data tapes and assigned code numbers. At least one
member of a team of two raters listened to each audiotaped
session segment and then scored the CALPAS. At least one
member of another team of two raters listened to each
audiotaped session segment and then rated the TASR.

CALPAS Raters: The raters were one psychodynamically
trained clinical psychologist with two years postdoctoral
experience and one psychodynamically trained clinical
psychology doctoral candidate with two years postmasters
experience. The CALPAS raters were trained for a total of 50
hours each in accord with procedures outlined in the CALPAS
Manual (Marmar a Gaston, 1989). The training involved the
following: reading and understanding the concept of alliance
as defined in the CALPAS manual, practicing rating the

alliance from excerpts of 10 precalibrated therapy sessions,

32

and making independent ratings with reliability checks.
Raters were considered proficient when their ratings
correlated .80 on the training tapes.

TASR Raters: The TASR raters were 2 psychodynamically
trained clinical psychology doctoral candidates with 1 and 6
years of postmasters experience. The TASR raters were
trained for 20 hours. The training involved: reading and
understanding the therapist actions measured by the scale
described in Hoyt et al. (1981) and in the manual (Hoyt,
1989), rating practice sessions and discussing ratings,
and making independent ratings with reliability checks.
Raters were considered proficient when their ratings

correlated .80 on the training tapes.

MGCBUI'GS 3

CALPAS: California Psychotherapy Alliance Scales

CALPAS (Marmar & Gaston, 1989) (Table 1) consisted of
30 7-point Likert type items which index aspects of the

patient therapist relationship and were rated from

33

Table 1
California Psychotherapy Alliance Scale

 

Patient Positive Contribution

 

1) Patient self-discloses thoughts and feelings.‘

2) Patient self-observes behaviors.

3) Patient explores own contribution to problems.

4) Patient experiences strong emotions.

5) Patient works actively with therapist's comments.

6) Patient deepens exploration of salient themes.

 

Patient Negative Contribution

 

7) Patient conveys an expectation of easy cure without work
on his/her part.

8) Patient acts in a hostile, attacking and critical manner
towards therapist.

9) Patient seems mistrustful and suspicious of therapist.

10) Patient engages in power struggle, attempting to
control the session.

11) Patient defies therapist's efforts to promote
self-understanding.

12) Patient holds therapist at arms length with flood of
words.

 

Patient Commitment

 

13) Patient is confident that efforts will lead to change.

14) Patient is willing to make sacrifices, for example time
and money.

15) Patient views therapy as important.
16) Patient has confidence in therapy and therapist.
17) Patient participates in therapy despite painful moments.

18) Patient is committed to go through process to
completion.

 

34

Table 1 (cont'd.)."

 

Working Strategy Consensus

 

l9) Therapy proceeds in accord with patient's ideas of
helpful change process.

20) Patient and therapist work together in a joint struggle.

21) Patient and therapist agree about the kind of changes to
make.

22) Patient and therapist share same sense about how to
proceed.

23) Patient and therapist agree on salient themes.

24) Therapist rigidly applies technique.

 

Therapist Involvement

 

25) Therapist is understanding of patient's suffering and
subjective world.

26) Therapist demonstrates non-judgemental acceptance and
positive regard.

27) Therapist demonstrates commitment to help and confidence
in treatment.

28) Therapist does not misuse treatment to serve own needs.
29) Therapist demonstrates tact and timing of interventions.

30) Therapist facilitates work on salient themes.

 

35

1 (not at all) to 7 (very much so). The items were rated by
outside judges on audiotaped segments of actual therapy
hours.

In the most recent version CALPAS items were broken
into five subscales. The Patient Working Capacity- Positive
Contribution scale contained items such as “Patient deepened
exploration of salient themes“ and was thought to measure
the patient's ability to work actively and purposefully in
treatment. The Patient Working Capacity - Negative
Contribution Scale contained items such as "Patient acts in
hostile, attacking and critical manner towards therapist“
and was thought to index hostile and mistrustful attitudes
towards the therapist. The Patient Commitment scale
contained items such as “Patient views therapy as important"
and was thought to measure positive transference. working
Strategy Consensus scale contained items such as "Patient
and therapist share the same sense about how to proceed" and
was thought to measure patient-therapist agreement on
therapeutic strategy (Bordin, 1979). The Therapist
Understanding and Involvement Scale consisted of items such
as "Therapist facilitates work on salient themes" and was
thought to measure therapist's empathic understanding of the
patient's problems and active participation in the
treatment.

Marziali et a1. (1981) developed the initial version

36

of this scale. That version contained 42 items designed to
assess patient and therapist positive and negative
contributions to the alliance. Marmar et al. (1986) did
further validation work on the CALPAS, then called the
California Therapeutic Alliance Rating System (CALTARS).
CALTARS was designed to stress "the affective and
attitudinal aspects of the therapeutic climate while
excluding specific classes of therapist actions, such as
interpretation of the transference, encouraging the
expression of affect, or working on termination in order
not to confound technique with the status of the alliance“
(Marmar et al. 1986).

Marmar, Weiss, and Gaston (1989) performed an
exploratory principal components factor analysis on the
intercorrelation matrix of the items which was generated
from.the rating of 208 segments from four sessions sampled
from each of 52 client-therapist dyads. Although four
components were hypothesized, an eight component solution
emerged. Additional solutions were forced*with four, five,
six, and seven components. The four component solution did
not yield conceptually clear dimensions. The five component
solution was retained due to the interpretability of the
components, the pattern of the item loadings, and the fact
that these components accounted for 63.3% of the total

variance. These factors were: Therapist Understanding and

37

Involvement (TUI), Patient Hostile Resistance (PHR),
Patient Commitment (PC), Therapist Negative Contribution
(TNC), and Patient Working Capacity (PWC). CALPAS was
subsequently reduced to a total of 32 items after deletion
of items that did not load highly on one of the five factors

or which loaded highly on more than one factor.

Reliability; Marmar et a1. (1989) evaluated the
internal consistency and interrater reliability of the
CALTARS on a sample of 208 psychotherapy sessions from 52
predominantly female outpatients, mean age 39, who had
experienced the death of a husband or parent. The
therapists were 5 male and 4 female faculty level therapists
with an average of 9 years postdoctoral clinical experience.
The treatment was 12-session brief psychodynamic
psychotherapy. The raters of the CALTARS were advanced
clinical trainees in psychology and psychiatry.

The coefficient alphas for Therapist Understanding and
Involvement (TUI), Patient Hostile Resistance (PHR),
Patient Commitment (PC), Therapist Negative Contribution
(TNC), and Patient Working Capacity (PWC) were .94, .84,
.94, .90, and .88. The respective interrater reliabilities
of those scales (measured by pearson's r) were .78, .79,

.81, .81, and .76.

38

Validity: Marmar et al. (1989) supported the
construct validity of the CALTARS by findings that other
measures correlated with the CALTARS in a hypothesized
fashion. Higher pretreatment patient motivation scores were
associated with higher Patient Working Capacity scores.
Higher pretreatment relationship composite scores on the
Patterns of Individual Change Scale, which indicate better
pretreatment interpersonal relations, were associated with
higher scores on Patient working Capacity. Two other
confirmed predictions were that more highly educated
patients were more able to form.an alliance, and that the
absence of stressful life events was associated with higher
alliance scores.

Tichenor and Hill (1989) found support for the
convergent validity of CALPAS. CALPAS, the Vanderbilt
Therapeutic Alliance Scale (VTAS), the working Alliance
Inventory (WAI), and the Pennsylvania Helping Alliance
Rating Scale (Penn) were rated on a sample of eight
psychotherapy cases. CALPAS correlated .80 with the VTAS
and .82 with the WAI. The VTAS and WAI correlated .84.
Penn did not correlate as well with the other measures (.34
with the CALPAS and .51 with the VTAS). Tichenor and Hill
concluded that CALPAS, VTAS and WAI were satisfactory
measures of the working alliance.

Discriminant validity was demonstrated by Marmar et a1.

39

(1989). Four of the five alliance subscales were
uncorrelated with pretreatment symptomatology as measured by
the SCL90-R (5's of .09 to .12). However, Patient Working
Capacity was negatively correlated with higher levels of
pretreatment symptomatology (£_= -.29, ps.05). CALPAS
apparently addressed constructs which were different from
symptomatology except for the Patient Working Capacity Scale
which was associated with pretreatment symptomatology.

Only preliminary reliability data were available on the
version of the CALPAS used in this study because at that
time it was in an early stage of development. Postdoctoral
level clinical judges rated 30 segments after 10 hours of
training (Gaston, personal communication). Interclass
coefficients for the mean ratings of three judges were .69
for the Patient working Capacity (positive and negative)
scale, .80 for the patient commitment scale, .71 for the
Working Strategy Consensus scale, and .69 for the Therapist
Understanding and Involvement scale.

CALPAS appeared to be a reliable instrument developed
with participants similar to the type in this study. Its
use in related research such as Foreman and Marmar (1985).
left the author with the impression that it was a reasonable
approach for measuring the therapeutic alliance in the

present study.

4O

TAS: The Therapist Action Scale

Hoyt et a1. (1981) reported on a Therapist Action Scale
(Table 2) and a Patient Action Scale they developed to
assess the degree to which therapists engaged in specific
activities in dynamic psychotherapy. Behaviorally specified
therapist actions such as "Discussed patient’s reactions to
therapist" and "Linked reactions toward therapist to other
important figures“ can be rated on a Likert type scale from
0 (did not occur) to 3 (occurred with major emphasis). Its
27 items addressed many possible therapist actions issues
such as: relationships with other people, relationship with
the therapist, self-concept, stressful events, expression
of affect, goals, and termination. TAS could be applied
by the therapist, the patient, or an outside rater. In
the present study raters listened to audiotaped segments of

actual therapy hours, and then scored the TAS.

Reliability: Hoyt et al. (1981) evaluated the TAS
interrater and test-retest reliability on a sample of 25

neurotic outpatients, 20 female and 5 male, mean age 36

41

Table 2
Therapist Action Scale - Revised

 

Addressing Transference

 

1) Discussed patient's reactions to therapist.
2) Linked reactions toward therapist to parental figures.

3) Linked reactions toward therapist to other important
figures.

4) Addressed the patient's guilt and expectation of
punishment by other people.

5) Addressed the patient's guilt and expectation of
punishment by the therapist.

 

Addressing Defenses

 

6) Discussed process of patient avoiding material or
feelings.

7) Discussed process of patient avoiding material or
feelings in relation to the therapist.

8) Discussed content and meaning of material and feelings
patient was avoiding.

9) Discussed content and meaning of material and feelings
patient was avoiding in relation to the therapist.

 

Support

 

10) Conveyed confidence of a favorable therapy outcome to
patient.

11) Therapist gives explicit reassurance.

12) Expressed liking or positive regard for the patient.
13) Therapist suggests meanings of others' behavior.

14) Therapist gives explicit advice or guidance.

15) Acts to strengthen defenses (vs. stimulate insight).

 

42

"Table 2 (cont'd.)."

 

Expressive Intervention

 

l6) Encouraged or permitted expression of feelings.

17) Encouraged patient to examine meanings of his/her
thoughts, behavior, feelings.

18) Patient's feelings and perceptions are linked to
situations and feelings from the past.

 

43

years. The psychodynamically-oriented therapists were 15
postdoctoral fellows and psychiatry residents, 12 men and 3
women. Each treatment was brief time-limited dynamic
psychotherapy. The raters were five
psychodynamically-trained psychiatrists and psychologists.
The interrater reliability was assessed with Finn's 5 (Finn,
1970; Tinsley & Weiss, 1975) which includes between rater
variance in the error term. The interrater reliability had
a median of .76 and ranged from .44 to .92. The test-retest
reliability had a median of .87 and ranged from..68 to .97.

Xenakis, Hoyt, Marmar, and Horowitz (1983) studied
the reliability of TAS therapist self-ratings. Two
independent judges, a psychodynamically trained
psychiatrist and a psychodynamically trained clinical
psychologist, made TAS ratings of 27 psychotherapy
sessions. The therapists (3 psychiatrists, 4 psychologists
and one social worker, all psychodynamically trained) rated
their own tapes of these same 27 psychotherapy sessions on
the Therapist Action Scale. There was a lack of agreement
between the independent judges and therapist self-ratings
about the levels of most therapist actions. Therapist

self-reports and judge ratings were not equivalent.

Validity: TAS appeared to have face validity. Items

were behaviorally specified avoiding clinical inferences by

44

the rater such as whether "transference" was involved.
Transference was measured instead by rating operationally
defined actions. Actions such as "relationship with the
therapist discussed" were assessed for frequency and degree
of emphasis. The skillfulness of the action was not
assessed.

Criterion related validity was supported by Windholz,
Weiss, and Horowitz (1985). TAS was rated after each of 12
sessions by the therapists of 30 female clients who received
time-limited dynamic psychotherapy related to the death of a
parent. These ratings were factor analyzed. A seven factor
solution contained the following dimensions: transference
and termination, reassurance, affect expression,
clarification, relationships, meaning of bereavement, and
errors.

When graphed by session number, the levels of these
action dimensions were consistent with the hypothesized
structure of time-limited psychotherapy as outlined by Mann
(1973) and Malan (1976). Level of therapist actions
addressing affect expression, meaning of the event
(bereavement), clarification, and relationships was high
across sessions, while level of reassurance and errors were
low. The one therapist action that increased in frequency
steadily across the 12 sessions was addressing transference

and termination, which is consistent with the theory of

45

time-limited psychotherapy. There was also slight increase
in errors and reassurance in the final session. This is
consistent with theory and experience of the special
significance of the final session, which makes it
particularly difficult for the client and the therapist. It
was also possible that theoretical expectations might
influence and bias their ratings. However, one
corroborated prediction supported the construct related
validity of the scale, suggesting that it could be used to
measure levels of common therapist actions.

Construct validity was supported in a general sense by
Hoyt et al. (1983). A sample of 30 sessions were randomly
selected from a group of 25, 12-session therapies.
Independent psychodynamically trained judges rated the
sessions on the TAS and also evaluated them.as either 'good'
or 'bad'. The sessions rated 'good' were higher for
therapist actions that encouraged expression of thoughts and
feelings and exploration of patient reactions, consistent
with theoretically sound short term dynamic therapy.

TAS appeared to be a reliable instrument developed with
participants similar to the type in this study. Its use in
research such as Windholz et al. (1985) made it a reasonable
choice as a measure of therapist activities for this study.
Two supplemental items were added by the author for use in

the present study (Table 2, items 4 and 5). These items

46

were added based on Foreman and Marmar's (1985) finding that
addressing the patient's guilt and expectation of punishment
was associated with improvement in the therapeutic alliance.
Similar to other TAS factors, one item.addresses guilt and
expectation of punishment by others and the other address
guilt and expectation of punishment by the therapist. The
addition of these two items constitutes the difference
between TAS and what is referred to in the present study as

TASR.

SCL90-R: The Sygptom Check List 90 Revised

 

Derogatis' (1983) SCL90-R was used to measure the
client's symptoms before and after therapy. It is a 90 item
self-administered questionnaire composed of nine subscales
measuring nine symptom dimensions: somatization,
obsessive-compulsive, interpersonal sensitivity,
depression, anxiety, hostility, phobic anxiety, paranoid
ideation, and psychoticism, Subjects were asked the extent
to which they were distressed by: 1) headaches 2)
nervousness and shakiness inside, etc. The subjects rated
each of the 90 symptom items on a Likert type scale that
ranged from 0, (not at all) to 4, (extremely). Means were
computed for each of the nine subscales. The Global

Severity Index (GSI) was computed as the sum of all item

47

responses divided by 90. According to Derogatis (1983) the
GSI is the best single indicator of the current level of

depth of the disorder.

Reliability: Internal consistency for the nine
subscales was measured by coefficient alpha for a sample of
219 symptomatic volunteers (Derogatis, Rickels & Rock,
1976). Coefficient alpha treats within form correlations
among the items as analogous to correlations between
alternate forms, and assumes that the average correlations
among existing items would be equivalent to the correlation
among items in the hypothetical alternate form. The
coefficients obtained for this sample were satisfactory and
ranged between a low of .77 for psychoticism.to a high of
.90 for depression.

The test-retest reliability for the SCL90 was checked
on a sample of 94 psychiatric outpatients with one-week
elapsed time between testing. The test-retest reliability
coefficients ranged from .78 for hostility to .90 for phobic
anxiety (Derogatis et al., 1976). Psychopathological
symptoms would be expected to be less stable than attributes
such as intelligence, but more stable than "mood". Though
psychological symptoms can fluctuate over a period of one

week, one would not expect much change.

48

Validity: Construct validity was supported by
Prusoff, Weissman, Klerman, and Rounsaville (1980). The
SCL90-R was used in a study evaluating the utility of
Research Diagnostic Criteria for predicting differential
response to amitriptyline and/or short term interpersonal
psychotherapy. The SCL90-R was found sensitive to change
and differences in the RDC subtypes. The type and severity
of symptoms can be assessed using the SCL90-R.

Convergent validity was supported by Derogatis et al.
(1976) who compared the dimension scores of the SCL90 with
the scale scores from the MMPI. In this study 119
symptomatic volunteers were given the SCL90 and the MMPI.
The results of the study were that each dimension of the
SCL90 had its highest correlation with similar MMPI scales
except for the obsessive compulsive dimension for which
there is no directly comparable MMPI scale.

The means for the SCL90-R are available for a sample
of 1002 heterogeneous outpatients (Derogatis, 1983). The
outpatients came from centers in Johns Hopkins University,
the University of Maryland, the University of Pennsylvania
and the University of Wisconsin. There were 425 males and
577 females, approximately two thirds white, skewed
somewhat towards the lower end of the socioeconomic
spectrumt The nonpatient norm group was comprised of 974

individuals, 493 males and 480 females, eight ninths white.

49

Social class data were not available. It represented a
stratified random sample from a diverse community in a large
eastern state.

The SCL90-R appears to be a valid and reliable
instrument constructed with subjects comparable to the type
of subjects in this study. These facts combined with its
past use in a related fashion in research such as the
Derogatis et a1. (1981) study in which symptom changes were
evaluated with the SCL90-R made the SCL90-R a reasonable

choice as a symptom.measure for this study.

RESULTS

Scale evaluation will be discussed first. Exploratory
and confirmatory factor analysis and reliability analyses
will be presented for the California Psychotherapy Alliance
Scale (CALPAS) and the Therapist Action Scale (TASR). The
revision of CALPAS and TASR will be discussed. After the
scale analyses, the evaluation of the hypotheses via path

analysis and related findings will be presented.

California Psychotherapy Alliance Scale (CALPAS)

 

An exploratory principal-components factor analysis was
performed on the intercorrelation matrix of the 30 CALPAS
items. The correlations were based on a set of 184 ratings
of four sessions for each of 46 subjects. Although ratings
of 184 independent subjects would be optimal, 184
independent observations are appropriate for scale
evaluation purposes. The ratings were made by two judges
who rated 39 sessions in common for reliability analysis.

In the case where a session was rated by both judges, their
mean rating was used.

The exploratory principal-components analysis followed
by a varimax rotation yielded five factors (Table 3). This
solution contained each of the five conceptually

50

51

Table 3

Egploratogy Factor Analysis of the California Psychotherapy
Alliance Scale

 

 

Item. Varimax Factors with Items Reordered and Grouped
Number According to Their Largest Factor Loading

 

1 2 3 4 S

6 83* 27 -2 20 o
1 82* 19 -7 23 2
2 81* 19 -7 17 2
3 78* 22 -1o 16 -7
4 72* 19 2 4 -9
7

 

 

1 71* 23 -16 26 -10
5 68* 26 -31 26 1
23 55* 39 -30 41 -9
20 51* 48 -21 40 -17
21 50* 28 -29 47 -7
7 -49* -1 26 -16 21
29 17 85* 0 11 -3
28 -2 75* -7 3 8
30 38 74* -1 9 -10
25 28 70* -6 21 -22
26 28 68* -14 32 -2
24 22 65* -5 -5 -l
27 9 56* -2 40 -30
8 -3 -3 77* -20 2
9 0 -3 74* -7 -5
11 -39 -3 61* -14 12
22 43 40 -47* 35 -2
15 57 10 -13 60* -10
14 53 10 -27 59* -9
16 36 27 -44 56* 5
13 47 22 -40 55* 10
18 47 15 -40 51* -8
19 42 38 -45 49* 9
12 -6 -12 -6 1 70*
10 -11 -2 49 -10 54*

Eigenvalues
1 2 3 4 5
13.47 2.63 1.94 1.00 .65

 

Proportion of Variance

 

1 2 3 4 5
.24 .16 .11 .11 .04

 

52

interpretable hypothesized dimensions and accounted for 62%
of the total variance. This solution was used in scale
development. The structure of the CALPAS was generally
confirmed.

Five component based scales were constructed by
considering hypothesized structure, and conceptual
relevance. In confirmatory factor analysis (Table 4), the
magnitude of the items correlation with the scale relative
to the other scales as well as coefficient alpha were

considered.1

Reliability California Psychotherapy Alliance Scale

The coefficient alphas for the five scales constructed
were: .94 for Patient Positive Contribution, .77 for
Patient Negative Contribution, .92 for Patient Commitment,
.92 for werking Strategy Consensus, and .90 for Therapist
Involvement.

For purposes of checking reliability both judges rated
the California Psychotherapy Alliance Scale for a subset of
39 sessions. Reliability of the ratings was checked by
computing the Pearson g correlation coefficient between the

scores of the two judges for the 39 sessions. The

1See Appendix A for detailed item.analysis.

53

Table 4

Confirmatogy Factor Analysis of the California Psychotherapy

Alliance Scale

 

 

Scale Scale Name Items Included Standard Score
Number Coefficient Alphas
501 Patient (6.1.2.3,4,17,5) 94
Positive Contribution
502 Patient (8,9,11,10) 77
Negative Contribution
503 Patient Commitment (15.14.16.13,18) 92
504 Working Strategy (23,20,21,22,19) 92
Consensus
505 Therapist Involvement (29.28.30,25,26, 90
27,24)

506 Residual (7,12) 23

 

54

"Table 4 (cont'd.)."

 

Items & Factor Intercorrelations and Loading Matrix
Scales Communality in the Diagonal

 

6 1 2 3 4 17 5 8 9 11 10 15 14
83 79 78 76 67 72 71 ~8 ~3 ~40 ~14 65 57
79 78 78 69 62 72 73 ~13 -7 ~39 ~13 63 62
78 78 72 79 59 60 64 ~11 ~12 ~37 ~14 57 58
76 69 79 68 58 62 64 ~16 ~7 ~45 ~19 58 54
53 69 50 ~7 ~4 ~26 ~14 50 43
72 72 60 62 69 65 65 ~23 ~14 ~41 ~24 65 62
71 73 64 64 50 65 60 ~29 ~24 ~53 ~25 60 62
~8 ~13 ~11 ~16 ~7 ~23 ~29 70 61 53 46 ~27 ~38
~3 ~7 ~12 ~7 ~4 ~14 ~24 61 42 39 34 ~15 ~26

H
WOMQ‘WNHQ
01
\l
01
N
UI
\9
0|
0

11 ~40 ~39 ~37 ~45 ~26 ~41 ~53 53 39 43 44 ~37 ~44
10 ~14 ~13 ~14 ~19 ~14 ~24 ~25 46 34 44 34 ~24 ~28
15 65 63 57 58 50 65 60 ~27 ~15 ~37 ~24 67 75
14 57 62 58 54 43 62 62 ~38 ~26 ~44 ~28 75 73
16 50 48 48 50 35 52 60 ~49 ~40 ~48 ~28 66 65

13 59 59 59 58 36 52 62 ~40 ~37 ~48 ~28 65 71
18 55 52 52 54 34 61 63 ~39 ~34 ~53 ~33 66 70
23 64 63 56 61 47 65 67 ~34 ~18 ~50 ~31 60 65
20 61 63 56 56 47 65 72 ~24 ~19 ~42 ~27 66 63
21 58 56 56 57 45 57 61 ~34 ~16 ~48 ~33 57 61
22 56 56 49 46 36 58 63 ~40 ~33 ~56 ~34 53 55

19 57 61 54 52 38 58 67 ~46 ~35 ~54 ~30 64 65
29 38 30 32 31 31 35 38 ~6 ~4 ~11 ~3 29 25
28 22 16 13 19 11 19 21 ~9 ~7 ~5 ~6 12 10
30 56 42 50 52 41 43 46 ~7 0 ~22 ~13 36 36
25 48 40 43 46 42 40 43 ~14 ~10 ~19 ~20 34 36
26 48 51 42 41 34 46 49 ~21 ~14 ~25 ~17 44 42
27 32 29 25 30 25 35 29 ~13 ~8 ~15 ~22 35 37
24 33 31 25 28 26 33 33 ~3 ~6 ~13 ~4 18 25
7 ~39 ~49 ~42 ~45 ~34 ~43 ~48 21 20 36 31 ~40 ~44
12 ~9 ~8 -8 ~12 ~9 ~10 ~5 ~8 -9 8 46 ~14 ~10
501 91 89 85 82 72 80 77 ~18 ~12 ~49 ~22 72 69
502 ~24 ~27 ~27 ~32 ~18 ~38 ~48 84 65 66 58 ~38 ~50
503 69 68 66 66 48 70 74 ~47 ~37 ~56 ~34 82 85
504 71 71 65 65 51 72 78 ~42 ~29 ~59 ~37 71 74
505 53 46 44 48 40 48 50 ~14 ~9 ~21 ~16 40 40
506 ~61 ~72 ~62 ~72 ~54 ~67 ~67 17 14 54 97 ~68 ~69

 

"Table 4 (cont'd.)."

55

 

WIDUIQhUINI-‘m

501

505
506

13

59
59
59
58
36
52
62
~40
~37
~48
~28
65
71
74
72
70
66
61
66
64
76
34
16
40
44
50
36
27
~40

85

47
~45

18

55
52
52
54
34
61

~39
~34
~53
~33
66
70
66
70
66
61
62
58
62
70
26
15
36
38
45
35
20
~47

64
~58
81
74
41
~70

23

64

20

61
63
56
56
47
65
72
~24
~19
~42
~27
66
63
62
61
62
69
65
64
73
66
57
31
61
61
64
52
48
~46
~20
73
~41
76
80
71
~84

21

58
56
56
57
45
57
61
~34
~16
~48
~33
57
61
63
66
58
81
64
68
69
64
34
21
44
49
54
39
26
~38
~5
67
~48
74
83
51
~53

22 19 29
56 57 38
56 61 30
49 54 32
46 52 31
36 38 31
58 58 35
63 67 38
~40 ~46 ~6
~33 ~35 ~4
~56 ~54 ~11
~34 ~30 ~3
53 64 29
55 65 25
65 72 38
64 76 34
62 70 26
73 69 46
73 66 57
69 64 34
76 76 45
76 66 46
45 46 75
31 38 61
48 45 76
50 48 62
54 60 67
38 44 51
36 29 61
~41 ~33 ~11
~5 2 ~19
63 67 41
~60 ~61 ~9
72 83 37
87 81 54
58 59 87
~57 ~39 ~38

28

22
16
13
19
11
19
21
-9
-7
-s
~6
12
1o
23
16
15
28
31
21
31
38
61
46
53
48
55
43
48

4
-1
21

-1o
18
35
68

3

30

56
42
50
52
41
43
46
~7

~22
~13

36
42
40
36
56

44
48
45
76
53
66
71
59
48
51
~22
~18

~16
46
61
81
~50

25

48
40
43
46
42
40
43
~14
~10
~19
~20

36
45
44
38
55
61
49

26

48
51
42
41
34
46
49

~14
~25

44
42
50
50
45
62
64
54
54
60
67
55
59
65
63
57
48
~27

54
~28
56
70
79
~46

 

"Table 4 (cont'd.).

56

 

1.1
WQU'INIbUNl-‘Ot

HHHH
emcH

HHH
mum

NNI-‘NNNN
01010100400)

NNNNOI
bQOtUIO

..4
[04

W1
”2
W3
W4
W5
”6

~39
~49
~42
~45
~34
~43
~48

21

20

36

31
~40
~44
~38
~40
~47
~46
~46
~38
~41
~33
~11

~22
~22
~27
~17
~19
18
13
~52
40
~51
~48
~22
40

12

~9
~8

~12
~9
~10
~5

~9

46
~14
~10

~8
~11
~20

~5

~19
~1
~18
~24
~9
~27
~11
13
18
~10
14

~9
~21
40

501

91
89
85
82
72
80
77
~18
~12
~49
~22
72
69
60
67
64
73
73
67
63
67
41
21
57
52
54
36
36
~52
~10
100
~37
80
82
57
~79

502

~24
~27
~27
~32
~18
~38
~48

84

65

66

58
~38
~50
~61
~57
~58
~49
~41
~48
~60
~61

~10
~16
~23
~28
~21

40
14
~37
100
~64
~61
~22
67

503
69

66
66
48
70
74
~47
~37
~56
~34
82
85
81
85
81
76
76
74
72
83
37
18
46
47
56
45
28
~51

80
~64
100

91

53
~72

504
71

65
65
51
72
78
~42
~29
~59
~37
71
74
78
79
74
88
80
83
87
81
54
35
61
63
70
51
42
~48

82
~61
91
100
72
~73

505
53

44
48
40
48
50
~14

~21
~16
40
40
51
47
41
62
71
51
58
59
87
68
81
81
79
64
63
~22
~21
57
~22
53
72
100
~54

506

~61
~72
~62
~72
~54
~67
~67

17

14

54

97
~68
~69
~46
~45
~70
~71
~84
~53
~57
~39
~38

~50
~58
~46
~55
~37

40

40
~79

67
~72
~73
~54
100

 

.12,

57

interrater reliabilities for the California Psychotherapy
Alliance Scale scales computed in this way were low (Table
5, between raters correlation matrix). The interrater

reliability for Patient Positive Contribution was g = .32,
for Patient Negative Contribution 5 = .46, for Patient I
Commitment g = .10, for Working Strategy Consensus g = .23, i

and for Therapist Involvement g = .30. Comparing the

 

between raters correlation matrix (Table 5) to the within .
rater correlation matrix, the between-raters correlations E
are much lower, but the broad pattern of correlations is
preserved. Thus, scales 1.3.4.5 correlated positively with
each other, whereas Patient Negative Contribution
correlated negatively with all other scales. Further, the
within-rater correlation matrix indicates that scales 1,3,4,
and 5 correlate highly. This pattern suggests a halo
effect, with each rater foaming an overall impression of
the alliance that influenced ratings on all scales. In
addition to the halo effect, each rater demonstrated
considerable idiosyncratic variability in their ratings
(Table 5, idiosyncratic component correlation matrix).
Test—retest stabilities for the same rater were also
low (Table 6, between times correlation matrix). The
Pearson's correlation test-retest reliability for Patient

Positive Contribution was 5 = .32, for Patient Negative

58

Table 5
California Psychotherapy Alliance Scale Interrater

Reliability

 

Different Raters at the Same Time Point

 

Rater Scale lA 2A §A 4A SA 13 23 53 £3 53

(1) ~.38 .61 .58 .28 .32 ~.36 .28 .33 .19
(2) ~.38 ~.66 ~.47 .04 .06 .46 ~.17 .03 .10

A (3) .61 ~.66 .78 .23 .01 ~.49 .10 .10 ~.05
(4) .58 ~.47 .78 .64 .04 ~.39 .09 .23 .02
(5) .28 .04 .23 .64 .13 ~.09 ~.05 .28 .30
(1) .32 .06 .01 .04 .13 ~.06 .70 .78 .61
(2) ~.36 .46 ~.49 ~.39 ~.09 ~.06 ~.10 ~.13 .01

B (3) .28 ~.17 .10 .09 ~.05 .70 ~.10 .61 .30
(4) .33 .03 .10 .23 .28 .78 ~.13 .61 .72
(5) .19 .10 ~.05 .02 .30 .61 .01 .30 .72

 

Within-Rater Correlation Matrix

 

.1 .1 .5 .1 .5

Patient Positive Contribution (1) ~.22 .65 .68 .45
Patient Negative Contribution (2) -.22 ~.38 ~.30 .02
Patient Commitment (3) .65 ~.38 .70 .27
Working Strategy Consensus (4) .68 -.30 .70 .68
Therapist Involvement (5) .45 .02 .27 .68

 

Between-Raters Correlation Matrix

 

.1 .1 .5 .1 .5

Patient Positive Contribution (1) .32 ~.15 .15 .18 .16

Patient Negative Contribution (2) ~.15 .46 ~.33 -.18 .00
Patient Commitment (3) .15 ~.33 .10 .10 ~.05
Working Strategy Consensus (4) .18 -.18 .10 .23 .15
Therapist Involvement (5) .16 .00 ~.05 .15 .30

 

Idiosyncratic Component Correlation Matrix

 

.1 .1 .5 .1 .5

Patient Positive Contribution (1) .68 ~.07 .51 .50 .28
Patient Negative Contribution (2) -.07 .54 -.05 -.12 .02
Patient Commitment (3) .51 ~.05 .90 .60 .32
werking Strategy Consensus (4) .50 ~.12 .60 .77 .53
Therapist Involvement (5) .28 .02 .32 .53 .70

 

59

Table 6

California Psychotherapy Alliance Scale Test-retest
Reliability

 

Same Rater at Different Time Points

 

 

 

Time I Time II
Time Scale 1 2 3 1 .5. 1 2. 3 1 5
(1) -.33 .72 .70 .16 .32 -.20 .33 .27 .12
(2) ~.33 -.58 ~.56 ~.18 .01 .20 .04 .06 .06
I (3) .72 -.58 .78 .17 .18 -.07 .24 .13 .11
(4) .70 -.56 .78 .51 .33 ~.05 .19 .24 .32
(5) .16 -.18 .17 .51 .26 .32 -.01 .20 .52
(1) .32 .01 .18 .33 .26 ~.12 .65 .68 .58
(2) ~.20 .20 -.07 -.05 .32 ~.12 -.37 -.44 .17
II (3) .33 .04 .24 .19 -.01 .65 -.37 .83 .49
(4) .27 .06 .13 .24 .20 .68 -.44 .83 .70

(5) .12 .06 .11 .32 .52 .58 .17 .49 .70

 

Within-Time Correlation Matrix

 

1 1 5 .1 5
Patient Positive Contribution (1) ~.23 .69 .69 .37
Patient Negative Contribution (2) ~.23 ~.48 ~.50 -.01
Patient Commitment (3) .69 ~.48 .80 .33
Working Strategy Consensus (4) .69 ~.50 .80 .61
Therapist Involvement (5) .37 ~.01 .33 .61

 

Between-Times Correlation Matrix

 

1 1 5 1 5

Patient Positive Contribution (1) .32 ~.10 .26 .30 .19
Patient Negative Contribution (2) -.10 .20 ~.02 .00 .19

Patient Commitment (3) .26 --02 .24 .16 .05
Working Strategy Consensus (4) .30 .00 .16 .24 .26
Therapist Involvement (5) .19 .19 .05 .26 .52

 

Idiosyncratic Component Correlation Matrix

 

1 1 5 1 5

Patient Positive Contribution (1) .68 -.13 .43 .39 .18
Patient Negative Contribution (2) ~.13 .80 ~.46 ~.51 ~.19
Patient Commitment (3) .43 ~.46 .76 .65 .28
werking Strategy Consensus (4) .39 -.51 .65 .76 .35
Therapist Involvement (5) .18 ~.19 .28 .35 .48

 

60

Contribution g = .20, for Patient Commitment 3,: .24, for
Working Strategy Consensus 5 = .24, and for Therapist
Involvement E = .52. Again the pattern of the correlations

was preserved from the within time correlation matrix to the
between time correlation matrix, although the relationships
are mmch stronger within time.

The between time correlation matrix with different
raters (Table 7) again showed modest correlations. The

Pearson's correlation for Patient Positive Contribution was

.g = .50, for Patient Negative Contribution 5 = .02, for
Patient Commitment“; = .34. for Working Strategy Consensus
‘5 = .33, and for Therapist Involvement g = .43. Again the

pattern of the correlations is preserved from.the within
time correlation matrix to the between time correlation
matrix, although the relationships are much stronger within

time.

Revision of CALPAS

In light of reliability problems with CALPAS, steps
were taken to improve its reliability. There has been
considerable improvement in the interrater reliability as a
result of combining the scales (Table 8). For the

California Psychotherapy Alliance Scale, the Pearson

61

Table 7
California Psychotherapy Alliance Scale Reliability Across

Time and Rater

 

Different Raters Across Two Time Points

 

Time I Time II
__Time __Sca1e 1. a 5 5. 5 1 .2. 5 1 2
(1) -.38 .77 .80 .45 .50 -.02 .31 .24 .17
(2) -.38 -.66 -.57 -.31 -.39 .02 -.34 -.27 -.21
x (3) .77 -.66 .79 .46 .43 .07 .34 .13 .08
(4) .80 -.57 .79 .66 .50 .04 .41 .33 .29
(5) .45 -.31 .46 .66 .43 .07 .24 .30 .43
(1) .50 -.39 .43 .50 .43 -.22 .63 .64 .43
(2) -.02 .02 .07 .04 .07 ~.22 -.36 -.49 .20
II (3) .31 -.34 .34 .41 .24 .63 -.36 .75 .27
(4) .24 -.27 .13 .33 .30 .64 -.49 .75 .57

(5) .17 -.21 .08 .29 .43 .43 .20 .27 .57

 

Within-Time Correlation Matrix With Different Raters

 

1 2 3 A g

.30 .70 .72 .44

Patient Positive Contribution (1)

Patient Negative Contribution (2) -.30 -.51 -.53 -.06
Patient Commitment (3) .70 -.51 .77 .37
Working Strategy Consensus (4) .72 -.53 .77 .62
Therapist Involvement (5) .44 -.06 .37 .62

 

Across-Times Correlation Matrix With Different Raters

 

1 2 3 4 5

Patient Positive Contribution (l) .50 .21 .37 .37 .30
Patient Negative Contribution (2) -.21 .02 -.14 -.12 -.07

Patient Commitment (3) .37 -.14 .34 .27 .16
Working Strategy Consensus (4) .37 -.12 .27 .33 .30
Therapist Involvement (5) .30 -.07 .16 .30 .43

 

Idiosyncratic Component Correlation Matrix

 

.1. 1 5 9. 5
Patient Positive Contribution (1) .50 -.10 .33 .35 .14
Patient Negative Contribution (2) -.10 .98 -.38 -.41 .01
Patient Commitment (3) .33 -.38 .66 .50 .21
Working Strategy Consensus (4) .35 -.41 .50 .67 .32
Therapist Involvement (5) .14 .01 .21 .32 .57

 

62

Table 8

California Psychotherapy Alliance Scale Interrater
Reliability

 

 

Rater Scale Different Raters at the Same Time Point
Correlation Matrix1

 

_1_A gs 13 _2_s

A CALPAS” (1) -0.46 0.26 -o.45
Patient Negative Contribution (2) -0.46 0.03 0.60

B CALPAS (1) 0.26 0.03 -0.09

Patient Negative Contribution (2) -0.45 0.60 -0.09

 

1Corrected for attenuation.

2CALPAS variables contain all the CALPAS scales except
Patient negative Contribution, which was separated to
function as an independent index.

 

63

correlation between raters for the patient negative
contribution scale was 5 = .60. The interrater correlation
for the combined California Psychotherapy Alliance Scale
scale was 5 = .26, which unsatisfactory.

The five CALPAS scales were averaged into a composite
at each of the four time points. A confirmatory factor
analysis was performed on these combined variables (Table
9). This factor analysis indicates that the CALPAS Patient
Negative Contribution scale does not correlate as well with
the other CALPAS scales as they do with each other nor does
it correlate well with the combined scale as a whole.

The variables were regrouped again. The California
Psychotherapy Alliance Scale scales were combined at each of
the four time points, but Patient Negative Contribution was
excluded to function as an independent scale at each time
point. A confirmatory factor analysis was performed on the
new variables (Table 10). Component scale-combined scale
correlations were higher, and component scales were more
parallel in intercorrelation. The coefficient alphas for

the combined scales were improved as well.

Therapist Actionggcale Revised (TASR)

An exploratory principal-components factor analysis was

performed on the intercorrelation matrix of the 18 item

64

Table 9

Confirmatogy Factor Analysis of the California Psychotherapy
Alliance Scale (CALPAS) and the Therapist Action Scale

Revised (TASR) at Pop; Time Points and Pretherapy and
Posttherapy Sypptom Checklist 90 Global Severity Index
Scores (SCL90-R GSI}.

 

 

Scale Scale Name Scales Included Standard Score
Number Coefficient Alpha
501 CALPAS Time 11 (1-5)‘ 88
502 CALPAS Time 2 (6-10) 85
503 CALPAS Time 3 (ll-15) 82
504 CALPAS Time 4 (16-20) 85
505 TASR Time 1.2 (21-24) 30
506 TASR Time 2 (25-28) 36
507 TASR Time 3 (29-32) 39
508 TASR Time 4 (33-36) 60
509 Symptoms3 (37-38) 38

 

1The five CALPAS scales have been averaged at each time
point to form.one scale at each time point.

2The four TASR scales have been combined at each time point
to form.one scale at each time point.

3Pretherapy and posttherapy SCL90 GSI scores were
aggregated to form one symptoms scale.

‘The following variables have been reflected:
(2,7,12,17,24,28,32,36).

65

"Table 9 (cont'd.)."

 

Scales Factor Intercorrelations and Loading Matrix
Communality in the Diagonal

1 2 3 4 5 6

58 35 75 75 54 28
35 31 6O 58 34 14
75 60 80 79 56 21
75 58 79 99 81 32
54 34 56 81 49 45
28 14 21 32 45 58
10 45 16 16 4 30
28 15 18 25 26 77
27 25 2O 34 38 82
51 37 49 59 42 33
42 34 51 49 41 47
12 16 13 3 -13 3
32 20 42 30 10 24
21 14 29 24 16 23
25 33 41 15
38 9 28 21 19
36 19 15 -15 2
38 24 26 16 14
23 10 19 12 3
23 29 45 48 8
-11 -7 17 18 5
-34 -1S -6 -2 16
-9 21 10 4 15
34 4O 54 41 19
-5 -16 O 9 10
-12 -11 10 15 13
17 32 40 35 14
22 -5 4 6 10
21 16 29 31 34

 

Patient Positive Contribution 1
Patient Negative Contribution 1
Patient Commitment 1

Working Strategy Consensus l
Therapist Involvement 1

Patient Positive Contribution 2
Patient Negative Contribution 2
Patient Commitment 2

Working Strategy Consensus 2
Therapist Involvement 2

Patient Positive Contribution 3
Patient Negative Contribution 3
Patient Commitment 3

Working Strategy Consensus 3
Therapist Involvement 3

Patient Positive Contribution 4
Patient Negative Contribution 4
Patient Commitment 4

Working Strategy Consensus 4
Therapist Involvement 4
Addressing Transference 1
Addressing Defenses 1
Expressive Intervention
Support 1

Addressing Transference
Addressing Defenses 2
Expressive Intervention
Support 2

Addressing Transference

p
Q
0

UIWHIUH HUI
ONWPUNHDONWQUQ

Addressing Defenses 3 27 -9 24 18 14 -5
Expressive Intervention -6 10 l -9 -20 -17
Support 3 -6 -3 -3 5 -3 -5

h w w M N H

Addressing Transference -29 -28 -25 -21 -12 -10

Addressing Defenses 4 -21 -22 -26 -16 -6 4
Expressive Intervention 4 2 -7 1 7 17 8
Support 4 9 -20 -14 -15 -14 -l
Pretherapy SCL90-R GSI Score -25 -2 -33 -37 -30 -7

Posttherapy SCL90—R GSI Score '10 7 -21 -7 -6 -16

-18 -35 -29 -21 -7 0
-52 -43 -35 -23

TASR Time 4
Symptoms

madambu”qummFWNHOWQQmmQUNpowQOmfiwwpowQOthNH
vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv

OOOOOOOOOUUUwUWUWUNNNNNNNNNNHHHHHHHHHHAAAAAAAAA

MUIUIUI U'IU'IU'IM U"AAAAAAAAAAAAAAAAAAAAAAAAAAAAA

CALPAS Time 1 ( 76 56 89 100 70 36
CALPAS Time 2 ( 39 37 33 45 42 76
CALPAS Time 3 ( 35 26 46 39 27 32
CALPAS Time 4 ( 13 43 25 36 23 12
TASR Time 1 ( 41 -14 28 54 44 40
TASR Time 2 ( 23 14 0 35 44 31
TASR Time 3 E 29 12 24 28 14 S
(

I
U
fi
fi

 

"Table 9 (cont'd.)."

66

 

\OQ‘IOUIQUNH

7 8
10 28
45 15
16 18
16 25

4 26
30 77
3O 59
59 79
53 80
29 33

6 27
15 -4

O 21
-4 6
-9 3
15 20
11 2
19 13

3 0

-13 3

2 6

8 39
15 33
13 22

-11 3
-3O 9
16 32
40 22
16 35
-17 4
-13 -10
-13 -5
-9 -7
-3 -3
-6 7
-9 4

2 -13

6 -1
24 29
55 89

2 15
10 11
27 72
10 43

-17 15
-12 O
8 -14

-4 21

28 66

58 81

52

30
17

-12

17 18
9 4
36 38
19 24
15 26
-15 16
2 14
11 19
2 13

7 21
19 21
8 43
31 16
28 37
28 43
-7 32
24 7O
15 53
53 92
49 91
-1 46
-7 O
-32 -25
-28 -14
19 31
O 18

4 12
-12 15
13 29
-2 9
-7 2
23 19
18 18
-29 4
-18 10
-6 25
-6 -41
-16 -10
-2 -1
16 27
11 24
25 48
38 96
-34 -6
3 49
20 31
-27 -1
-18 -11

23

 

67

 

"Table 9 (cont'd.)."
21 22 23 24 25 26 27 28 29 30 31 32 33 34
l 10 4 11 32 -9 11 35 -2 30 27 -6 -6 -29 -21
2 -ll -34 -9 34 -5 -12 17 22 21 -9 10 -3 -28 -22
3 -7 -15 21 40 -16 -11 32 -5 16 24 1 -3 -25 -26
4 17 -6 10 54 0 10 40 4 29 18 -9 5 -21 -16
5 18 -2 4 41 9 15 35 6 31 14 -20 -3 -12 -6
6 5 16 15 19 10 13 14 10 34 -5 -l7 -5 -10 4
7 2 8 15 13 -11 -30 16 40 16 -17 -13 -13 -9 -3
8 6 39 33 22 3 9 32 22 35 4 -10 -5 -7 -3
9 17 10 10 32 16 1 18 29 38 -5 -19 -3 7 14
10 22 -1 -5 32 9 -4 10 -1 45 21 -12 -17 0 -8
11 8 -3 2 41 19 13 26 12 23 28 -2 0 -5 6
12 -7 -10 -4 -3 -5 3 0 17 -34 -31 -2 18 -7 4
13 13 6 10 34 2 7 42 3 -5 14 22 10 -6 0
l4 7 -4 -10 15 11 13 18 4 -5 9 8 15 9 12
15 13 -8 -17 27 16 4 23 -7 23 33 2 17 25 16
16 13 -22 -16 32 24 12 19 28 24 2 9 21 21 23
17 -7 -32 -28 19 0 4 -12 13 -2 -7 23 18 -29 -18
18 0 -25 -14 31 18 12 15 29 9 2 19 18 4 10
19 8 -25 -16 26 25 19 8 19 0 0 18 14 7 15
20 29 -15 -11 32 15 15 31 5 29 27 -2 -2 24 25
21 11 10 7 16 29 4 23 11 14 12 14 4 7 17
22 10 27 52 -18 -12 24 32 -6 15 26 16 -13 13 -10
23 7 52 38 -10 -2 -4 40 -3 3 11 18 3 12 1
24 16 -18 -10 l 9 1 35 23 —4 0 -10 -4 -7 13
25 29 -12 -2 9 34 45 -9 19 13 —4 -16 8 10 6
26 4 24 -4 1 45 41 9 4 4 9 -9 7 -7 9
27 23 32 40 35 -9 9 0 7 -1 18 28 9 16 8
28 11 -6 -3 23 19 4 7 6 21 -3 -2 16 13 15
29 14 15 3 -4 13 4 -1 21 11 34 2 3 10 -14
30 12 26 11 0 -4 9 18 -3 34 49 27 3 14 12
31 14 16 18 -10 -16 -9 28 -2 2 27 14 13 13 6
32 4 -l3 3 -4 8 7 9 16 3 3 13 2 23 21
33 7 13 12 -7 10 -7 16 13 10 14 13 23 67 69
34 17 -10 1 13 6 9 8 15 -14 12 6 21 69 71
35 14 -2 6 2 19 -4 16 28 33 12 19 24 38 22
36 2 -10 0 -3 -11 -15 -1 4 2 11 5 14 6 25
37 -7 -1 -21 -31 23 7 -27 -15 15 -15 5 -12 -1 -11
38 -7 4 -17 -3 1 3 13 7 -1 -17 0 -3 20 9
501 7 -l4 9 51 -6 3 41 7 33 19 -7 -2 -29 -23
502 15 20 18 32 7 -3 25 27 46 -1 -19 -12 -5 l
503 9 -5 -6 32 12 12 31 8 0 15 8 17 4 10
504 12 -33 -23 38 23 17 17 25 16 6 18 19 7 15
505 32 51 64 -8 17 18 94 18 20 36 28 -7 19 15
506 45 25 20 45 58 66 4 23 24 13 0 26 21 25
507 29 28 23 -12 0 7 34 21 33 73 36 14 38 16
508 18 -4 9 2 11 -8 18 28 14 22 19 37 82 86
509 -14 3 -37 -33 23 10 -14 -7 14 -31 5 -15 19 -2

 

"Table 9 (cont'd.)."

68

 

\OOQOUIQWNH

35

2
-7

17

38 501 502 503

-10
7
-21
-7
-6
-16

76
56
89

504 505 506 507

13 41 23 29
43 -14 14 12
25 28 O 24
36 54 35 28
23 44 44 14
12 40 31 5
10 27 10 -17
11 72 43 15
23 50 42 7
34 35 9 23
45 34 47 32
15 -18 10 -31
41 46 36 26
52 6 30 17
47 11 24 48

-4 54 39
-4 97 49
54 97 41
39 49 41
13 18 32 60
4 -58 7 -18

 

 

69

Table 10

Confirmatogy Factor Analysis of the California Psychotherapy
Alliance Scale (CALPAS), the Therapist Action Scale Revised
(TASR), SupportI and Patient Negative Contribution at Four
Time Points and the Sypptom Checklist 90 Global Severity
Index Scores (SCL90-R GSI).

 

 

Scale Scale Name Scales Included Coefficient
Number Alpha

501 CALPAS Time 11 (1.3-5) .90
502 CALPAS Time 2 (6.8-10) .86
503 CALPAS Time 3 (11.13-15) .86
504 CALPAS Time 4 (16.18-20) .89
505 TASR Time 1’ (21-23) .48
506 TASR Time 2 (25-27) .35
507 TASR Time 3 (29-31) .44
508 TASR Time 4 (33-35) .70
509 Pretherapy SCL90-R GSI Score (37)3 *
510 Posttherapy SCL90-R GSI Score (38) *
511 Patient Negative Contribution 1 (2)‘ *
512 Patient Negative Contribution 2 (7) *
513 Patient Negative Contribution 3 (12) *
514 Patient Negative Contribution 4 (17) *
515 Support Time 1 (24) *
516 Support Time 2 (28) *
517 Support Time 3 (32) *
518 Support Time 4 (36) *

 

1CALPAS variables at times 1-4 contain all scales except
Patient Negative Contribution, which was separated to
function as an independent index.

2TASR variables at times 1-4 contain all TASR scales except
Support, which was separated to form.an independent index.

3In the Factor Loadings and Intercorrelations Matrix, all
scales corresponding one to one with clusters, i.e., 37 and
509, 38 and 510, etc. were deleted.

‘The following variables have been reflected:
(2,7,12,17,24,28,32,36)

* Constrained to be 1 since only one subscale composes the
scale.

 

“Table 10 (cont'd.)."

7O

 

 

Scales Factor Intercorrelations and Loading Matrix
Communality in the Diagonal
1 3 4 5 67"

Patient Positive Contribution l (1) 64 75 75 54 28 28
Patient Commitment 1 (3) 75 70 79 56 21 18
Working Strategy Consensus 1 (4) 75 79 96 81 32 25
Therapist Involvement 1 (5) 54 56 81 54 45 26
Patient Positive Contribution 2 (6) 28 21 32 45 70 77
Patient Commitment 2 (8) 28 18 25 26 77 68
WOrking Strategy Consensus 2 (9) 27 20 34 38 82 80
Therapist Involvement 2 (10) 51 49 59 42 33 33
Patient Positive Contribution 3 (11) 42 51 49 41 47 27
Patient Commitment 3 (13) 32 42 30 10 24 21
working Strategy Consensus 3 (14) 21 29 24 16 23 6
Therapist Involvement 3 (15) 17 25 33 41 15 3
Patient Positive Contribution 4 (16) 4 9 28 21 19 20
Patient Commitment 4 (18) 4 24 26 16 14 13
Working Strategy Consensus 4 (19) —3 10 19 12 3 0
Therapist Involvement 4 (20) 32 29 45 48 8 3
Addressing Transference l (21) 10 -7 17 18 5 6
Addressing Defenses l (22) 4 -15 -6 -2 16 39
Expressive Intervention l (23) 11 21 10 4 15 33
Addressing Transference 2 (25) -9 -16 0 9 10 3
Addressing Defenses 2 (26) ll -11 10 15 13 9
Expressive Intervention 2 (27) 35 32 40 35 14 32
Addressing Transference 3 (29) 30 16 29 31 34 35
Addressing Defenses 3 (30) 27 24 18 14 -5 4
Expressive Intervention 3 (31) -6 1 -9 -20 -17-10
Addressing Transference 4 (33) -29 -25 -21 -12 -10 -7
Addressing Defenses 4 (34) -21 -26 -16 -6 4 -3
Expressive Intervention 4 (35) 2 1 7 17 8 7
CALPAS Time 1 (501) 80 84 99 73 38 29
CALPAS Time 2 (502) 43 34 48 48 84 82
CALPAS Time 3 (503) 36 47 44 35 35 18
CALPAS Time 4 (504) 11 22 36 30 13 11
TASR Time 1 (505) 16 -1 13 13 23 50
TASR Time 2 (506) 27 3 35 43 27 32
TASR Time 3 (507) 35 28 26 17 8 19
TASR Time 4 (508) -23 -24 -15 -l 1 -1
Pretherapy SCL90-R GSI Score (509) -25 -33 -37 —30 -7-13
Posttherapy SCL90-R GSI Score (510) -10 -21 -7 -6 -16 -1
Patient Negative Contribution 1 (511) 35 60 58 34 14 15
Patient Negative Contribution 2 (512) 10 16 16 4 30 59
Patient Negative Contribution 3 (513) 12 13 3 -13 3 -4
Patient Negative Contribution 4 (514) 9 19 15 -15 2 2
Support Time 1 (515) 32 40 54 41 19 22
Support Time 2 (516) -2 -5 4 6 10 22
Support Time 3 (517) -6 -3 5 -3 -5 -5
Support Time 4 (518) 9 -14 -15 -14 -l 4

 

71

"Table 10 (cont'd.)."

 

9 10 11 13 14 15 16 18 19 20 21 22

‘OQOSUIDUOH
w
0
h
N
h
p
H
O
p
O)
h
H
N
H
H
0‘
H
N
Q
Q
..a
m
I
N

513 1 -29 28 44 58 -8 12 16 4 -10 -7 -10
514 7 19 8 28 28 -7 24 53 49 -1 -7 -32
515 32 32 41 34 15 27 32 31 26 32 16 -18
516 29 -1 12 3 4 -7 28 29 19 5 11 -6
517 -3 -17 0 10 15 17 21 18 14 -2 4 -13
518 5 -16 -19 -7 -19 -15 -20 -41 -36 -2 2 -10

 

"Table 10 (cont'd.).

72

 

26

13

-4
13

27

29 30
30 27
16 24
29 18
31 14
34 -5
35 4
38 -5
45 21
23 28
-5 14
-5 9
23 33
24 2
9 2
0 0
29 27
14 12
15 26
3 11
13 -4
4 9
-1 18
13 34
34 71
2 27
10 14
-14 12
33 12
32 25
49 5
11 27
19 9
21 32
11 16
33 90
14 18
15 -15
-1 -17
21 -9
16 -17
-34 ~31
-2 -7
-4 0
21 -3
3 3
2 11

-2
5

33 34
-29 -21
-25 -26
-21 -16
-12 -6
-10 4

-7 -3

7 14
O -8
-5 6
-6 O
9 12
25 16
21 23
4 10
7 15
24 25
7 17

13 -10

12 1

1O 6

-7 9

16 8

10 -14

14 12

13 6

86 69

69 50

38 22
-26 -21

-3 2

7 11

17 22

21 5

13 17

25 2

95 70

-1 -11

2O 9
-28 -22

-9 -3

-7 4
-29 -18

-7 13

13 15

23 21

6 25

35 501 502 503 504

80
84
99
73
38
29
35
60
55
34
27
35
19
21

-10

43 36
34 47
48 44
48 35
84 35
82 18
102 30
46 27
43 75
26 82
19 94
22 61
28 49
22 50
12 44
23 56
16 13
20 -3
17 -5
12 16
6 12
24 35
49 11
5 27
-19 10
-3 7
2 11
20 24
52 48
35

35
26 61
34 3
30 45
24 32
9 21
-7 -8
-5 -11
29 25
54 -2
-9 39
1O 19
34 38
19 4
—10 14
-3 -19

505

16
-1
13
13
23
50
24
10
4
19
-5
-8
-16
-25
-21

506
27

 

 

73

"Table 10 (cont'd.)."

 

507 508 509 510 511 512 513 514 515 516 517 518

1 35 -23 -25 -10 35 10 12 9 32 -2 -6 9
3 28 -24 -33 -21 6O 16 13 19 4O -5 -3 -14
4 26 -15 -37 -7 58 16 3 15 54 4 5 -15
5 17 -1 -30 -6 34 4 -13 -15 41 6 -3 -14
6 8 1 -7 -16 14 30 3 2 19 10 -5 -1
8 19 -1 -13 -1 15 59 -4 2 22 22 -5 4
9 10 24 -2 -2 25 53 1 7 32 29 -3 5
10 36 6 -1 4 37 29 -29 19 32 -1 -17 -16
11 34 9 -2 -25 34 6 28 8 41 12 O -19
13 21 O -21 -18 20 O 44 28 34 3 10 -7
14 8 18 -10 -2 14 -4 58 28 15 4 15 -19
15 39 38 6 12 9 -9 -8 -7 27 -7 17 -15
16 24 40 -1 18 38 15 12 24 32 28 21 -20
18 20 19 -10 -1 38 19 16 53 31 29 18 -41
19 12 25 O 12 23 3 4 49 26 19 14 -36
20 37 45 -11 27 23 -13 -10 -1 32 5 -2 -2
21 27 19 -7 -7 -11 2 -7 -7 16 11 4 2
22 39 0 -1 4 -34 8 -1O -32 -18 -6 -13 -1O
23 22 10 -21 -17 -9 15 -4 -28 -10 -3 3 0
25 -5 18 23 1 -5 -11 -5 O 9 19 8
26 2 -1 7 3 -12 -3O 3 4 1 4 7 -15
27 3O 20 -27 13 17 16 0 -12 35 7 9
29 33 14 15 -1 21 16 -34 -2 -4 21 3
3O 90 18 -15 -17 -9 -17 -31 -7 O -3 3 11
31 24 18 5 0 10 -13 -2 23 -10 -2 13 5
33 25 95 -1 20 -28 -9 -7 -29 -7 13 23 6
34 2 7O -11 9 -22 -3 4 -18 13 15 21 25
35 43 37 10 9 -7 -6 -5 -6 2 28 24 2
501 31 -19 -37 -13 56 14 5 9 50 1 -2 -10
502 24 9 -7 -5 29 54 -9 10 34 19 -10 -3
503 32 21 -8 -11 25 -2 39 19 38 4 14 -19
504 28 4O -7 17 37 7 7 38 37 24 16 -30
505 57 19 -19 -13 -34 16 -14 -43 -8 2 -4 -5
506 20 26 2 12 0 -18 -1 -6 32 21 17 -19

507 35 4 -13 15 -9 -45 9 -9 11 13 12
508 35 -1 19 -28 -9 -4 -26 4 28 33 16
509 4 -1 24 -2 2 -16 -16 -31 -15 -12 -10
510 -13 19 24 7 6 -13 -2 -3 7 -3 5
511 15 -28 -2 7 45 16 36 34 22 -3 -20
512 -9 -9 2 6 45 15 11 13 40 -13 -9
513 -45 -4 -16 -13 16 15 31 -3 17 18 0
514 9 -26 -16 -2 36 11 31 19 13 18 -6
515 -9 4 -31 -3 34 13 -3 19 23 -4 -3
516 11 28 -15 7 22 4O 17 13 23 16 4
517 13 33 -12 -3 -3 -13 18 18 -4 16 14

518 12 16 -10 5 -20 -9 O -6 -3 4 14

 

74

Therapist Action Scale. The correlations were based on a
set of 184 ratings of four sessions for each of 46 subjects.
The ratings were made by two judges who rated 45 sessions in
common for reliability analysis. In the case where a
session was rated by both judges the mean of their ratings
was used.

The exploratory principal-components analysis followed
by a varimax rotation yielded three components which
accounted for 37% of the total variance. The three factor
solution contained the four hypothesized dimensions, once
the first component was split (Table 11). The three factor
solution was used in scale development. The structure of
the Therapist Action Scale Revised is generally confirmed.

Four component based scales were constructed by
considering hypothesized structure and conceptual relevance.
In confirmatory factor analysis (Table 12), the magnitude
of item correlations with the scale relative to the other
scales as well as the pertinent coefficient alpha were

considered.

Reliability of the Therapist Action Scale Revised

For purposes of checking interrater reliability both

judges rated a subset of 45 sessions. The interrater

75

Table 11

Exploratogy Factor Analysis of the Therapist Action

Scale Revised

 

Item. varimax Factors With Items Reordered and Grouped
Number According to Their Largest Factor Loading

 

 

 

1 2 3
9 78* 5 1
7 77* 0 -5
1 73* 6 5
3 43* 10 0
S 38* -4 -8
2 18* 4 3
4 9* -4 -2
8 38 70* 5
6 48 62* 6
17 3 54* -21
16 4 51* 1
18 -5 42* —8
13 -13 38* 6
11 -2 3 76*
15 -13 -5 70*
12 5 -6 68*
14 -19 4 61*
10 19 -8 53*
Eigenvalues
1 2 3

2.95 2.25 1.44

 

Proportion of Variance

 

.14 .10 .13

 

76

Table 12

Confirmatogy Factor Analysis of the Therapist Action Scale

Revised

 

 

Scale Scale Name Items Standard Score
Number Included Coefficient Alphas
501 Addressing Transference (1.3.5) 65
502 Addressing Defenses (6.8.7.9) 78
503 Expressive Intervention (13.16.17.18) 57
504 Support (11.12.14.15) 78
505 Residual (2.4.10) -14

 

Items & Factor Intercorrelations and Loading Matrix
Scales Communality in the Diagonal

 

1 3 S 6 8 7 9 13 16 17 18 11 12 14

1 58 46 4O 30 29 50 51 -10 13 11 3 8 13 -16

3 46 35 28 19 19 22 19 8 15 0 1 2 0 -5

5 40 28 27 9 8 19 19 -5 -1 6 2 -6 -3 -7

6 30 19 9 60 75 34 44 12 24 25 15 4 O -4

8 29 19 8 75 40 28 29 14 26 26 26 5 -7 3

7 50 22 19 34 28 41 72 -1 -2 8 -2 -9 -3 -16

9 51 19 19 44 29 72 51 -8 5 10 -3 -2 0 -9
13 -10 8 -5 12 14 -1 -8 17 19 2O 26 7 -1 12
16 13 15 -l 24 26 -2 5 19 26 44 14 4 10 -1O
17 11 0 6 25 26 8 10 20 44 44 27 -15 -15 -11
18 3 1 2 15 26 -2 -3 26 14 27 18 -3 -5 -3
11 8 2 -6 4 5 -9 -2 7 4 -15 -3 63 54 47
12 13 0 -3 O -7 -3 O -1 10 -15 -5 54 35 30
14 -16 -5 -7 -4 3 -16 -9 12 -10 -11 -3 47 30 38
15 -8 -8 -5 -1O -1O -7 -5 7 O -12 -6 56 43 53
2 36 24 -6 4 5 1 3 -9 6 -6 21 7 7 -4

4 11 -5 20 -4 -4 5 9 1 0 8 -2 3 3 1
10 14 2 -7 15 8 12 13 -6 -4 -19 -13 37 45 17
501 77 58 51 31 30 48 47 -4 14 9 3 2 6 -14
502 58 29 20 77 63 64 72 6 19 25 13 -1 -3 -9
503 8 12 1 38 47 1 2 4O 51 67 42 -4 -5 -6
504 -1 -4 -7 -4 -3 -13 -6 9 1 -19 -6 8O 59 61
505 37 13 4 9 6 11 15 -8 1 -11 4 28 33 8

 

77

 

"Table 12 (cont'd.)."

15 2 4 10 501 502 503 504 505
l -8 36 11 14 77 58 8 -1 37
3 -8 24 -5 2 58 29 12 -4 l3
5 -5 -6 20 -7 51 20 1 -7 4
6 -10 4 -4 15 31 77 38 -4 9
8 -10 5 -4 8 30 63 47 -3 6
7 -7 1 5 12 48 64 1 -13 11
9 -5 3 9 13 47 72 2 -6 15
13 7 -9 1 -6 -4 6 40 9 -8
16 0 6 0 -4 14 19 51 1 1
17 -12 -6 8 -19 9 25 67 -19 -11
18 -6 21 -2 -l3 3 13 42 -6 4
11 56 7 3 37 2 -l -4 80 28
12 43 7 3 45 6 -3 -5 59 33
14 53 -4 1 17 -14 -9 -6 61 8
15 58 -6 -5 28 -11 -12 -6 76 10
2 -6 100 -4 -3 29 5 6 1 56
4 -5 -4 100 -6 14 2 4 1 54
10 28 -3 -6 100 5 18 -21 46 55
501 -11 29 14 5 57 11 -6 29
502 -12 5 2 18 57 32 -9 15
503 -6 6 4 -21 11 32 -7 -7
504 76 1 l 46 -6 -9 -7 29

505 10 S6 54 55 29 15 -7 29

 

78

reliability was checked by Pearson's g correlation
coefficient between the scores of the two judges for the 45
sessions. The interrater reliabilities for the Therapist
Action Scale Revised scales computed in this way were
moderate (Table 10, between raters correlation matrix).

The interrater reliability for Addressing Transference was 5
= .62, for Addressing Defenses 5 = .44, for Expressive
Intervention g = .47, and for Support 5 = .50. Comparing
the between rater correlation matrix (Table 13) to the
within rater correlation matrix. the correlations are
comparable in pattern and magnitude. The coefficient alphas
for the scales range from .57 to .78, which represents
adequate internal consistency.

Test-retest reliabilities for the same rater computed
with Pearson's correlation were low (Table 14. between
times correlation matrix). Test-retest reliability for
Addressing Transference was g = .16, for Addressing
Defenses g = .13. for Expressive Intervention g = .29. and
for Support g = .36. Again the pattern of the correlations
is preserved from the within time correlation matrix to the
between time correlation matrix. and the magnitude of the
correlations is comparable.

The between time correlation matrix with different
raters (Table 15) again showed correlations near zero.
Pearson's g for Addressing Transference was (5,: -.03). and
for Addressing Defenses (g = .12, and for Expressive

Intervention (g = .01), and for Support (5 = .01).

Table 13

79

Therapist Action Scale Revised Interrater Reliability

 

Different Raters at the Same Time Point

 

 

 

 

 

 

 

Rater Scale 1A 2A 3A AA 13 gs 33 43
(l) .34 07 -.05 .62 .36 .44 -.22

A (2) 34 16 -.09 .42 .44 .38 -.11
(3) 07 .16 -.11 -.04 .18 .47 -.03

(4) - 05 -.09 - 11 .02 -.02 -.17 .50

(1) 62 .42 - 04 .02 .58 .23 -.15

B (2) .36 .44 .18 -.02 .58 .39 -.16
(3) .44 .38 .47 -.17 .23 .39 -.32

(4) —.22 -.11 -.03 .50 -.15 -.16 -.32

Within-Rater Correlation Matrix

1111

Addressing Transference (1) .46 .15 -.10
Addressing Defenses (2) .46 .28 -.13
Expressive Intervention (3) .15 .28 -.22

Support (4) -.10 -.13 -.22

Between-Raters Correlation Matrix

111.4.

Addressing Transference (1) .62 .39 .20 -.10
Addressing Defenses (2) .39 .44 .28 -.07
Expressive Intervention (3) .20 .28 .47 -.10
Support (4) -.10 -.07 -.10 .50
Idiosyncratic Component Correlation Matrix
1114.

Addressing Transference (1) .38 .07 -.05 -.00
Addressing Defenses (2) .07 .56 -.00 -.06
Expressive Intervention (3) -.05 -.00 .53 -.11
Support (4) -.00 -.06 -.11 .50

 

80

Table 14

Therapist Action Scale Revised Test-getest Reliability

 

 

Same Rater at Different Time Points

 

Time I Time II

Lime __S¢ale 1 1 5 5 1 1 5 1
(1) .48 -.10 -.27 .16 .08 -.06 -.14

I (2) .48 .13 .09 .18 .13 .03 -.04
(3) -.10 .13 .00 .03 .32 .29 .01
(4) -.27 .09 .00 -.23 -.04 .00 .36
(l) .16 .18 .03 —.23 .37 .04 -.03

II (2) .08 .13 .32 -.04 .37 .22 -.07
(3) -.06 .03 .29 .00 .04 .22 -.23
(4) -.14 -.04 .01 .36 -.03 -.07 -.23

 

Within-Time Correlation Matrix

 

1 1 5 .4.

-‘U!’ “‘7" ..r. “-l

 

 

 

 

 

Addressing Transference (l) .43 -.03 -.15
Addressing Defenses (2) .43 .18 .01
Expressive Intervention (3) -.03 .18 —.12
Support (4) -.15 .01 -.12

Between-Times Correlation Matrix

1 1 1 1

Addressing Transference (1) .16 .13 -.02 -.19
Addressing Defenses (2) .13 .13 .18 -.04
Expressive Intervention (3) -.02 .18 .29 .01
Support (4) -.19 -.04 .01 .36
Idiosyncratic Component Correlation Matrix

1 1 5 5

Addressing Transference (1) .84 .30 -.02 .03
Addressing Defenses (2) .30 .87 .00 .05
Expressive Intervention (3) -.02 .00 .71 -.12

Support

(4)

.03 .05 -.12 .64

 

81

Table 15
Therapist Action Scale Revised Reliability Across Time and

Rater

 

Different Raters Across Two Time Points

 

Time I Time II

Time Scale 1. .2 .3 .4 .1 .2 .3 .4
(1) -.01 -.18 -.18 -.03 -.09 .01 .12

I (2) -.01 .40 .15 .16 .12 -.08 .01
(3) -.18 .40 -.17 .28 .00 .01 -.07
(4) -.18 .15 -.17 -.11 -.01 -.16 .01
(l) -.03 .16 .28 -.11 .47 .14 -.12

II (2) -.09 .12 .00 -.01 .47 .16 -.09
(3) .01 -.08 .01 -.16 .14 .16 -.26
(4) .12 .01 -.07 .01 -.12 -.09 -.26

 

Within-Time Correlation Matrix With Different Raters

 

1 1 5 5

Addressing Transference (1) .23 -.02 -.15
Addressing Defenses (2) .23 .28 .03
Expressive Intervention (3) -.02 .28 -.22
Support (4) -.15 .03 -.22

 

Across-Times Correlation Matrix With Different Raters

 

1 1 5 5

Addressing Transference (1) -.03 .03 .15 .00
Addressing Defenses (2) .03 .12 -.04 .00
Expressive Intervention (3) .15 -.04 .01 -.12
Support (4) .00 .00 -.12 .01

 

Idiosyncratic Component Correlation Matrix

 

1 1 5 5.

Addressing Transference (1) 1.03 .19 -.16 -.16
Addressing Defenses (2) .19 .88 .32 .03
Expressive Intervention (3) -.16 .32 .99 -.10

Support (4) -.16 .03 -.10 .99

 

82

Reliability of the Revision of the Therapist Action Scale
Revised (TASR)

Steps were taken to improve the reliability of the
TASR. There has been considerable improvement in the
interrater reliability as a result of combining the scales
(Table 16). For the Therapist Action Scale Revised (TASR)
the Pearson correlation for the TASR combined variable
between raters was.£ = .98, and for the Support scale.£ =

.64.

Revision of the Therapist Action Scale Revised

Each of the four Therapist Action Scale Revised scales
were combined together into one large scale at each of the
four time points measured. For comparison, the pretherapy
and posttherapy SCL90-R Global Severity Index scores were
combined into one symptoms measure.

A confirmatory factor analysis was performed on these
combined variables (Table 9). For the Therapist Action
Scale Revised, the Support scale does not correlate as well
with the other Therapist Action Scale Revised scales as they
do with each other nor does it correlate well with the

combined scale as a whole. There is room for

83

Table 16
Therapist Action Scale Revised Interrater Reliability

 

Rater Scale Different Raters at the Same Time Point
Correlation Matrix1

 

1A 1A 1a 1a

A TASR“ (1) -o.41 0.98 0.15
Support (2) -0.41 -0.21 0.64
TASR (1) 0.98 -0.21 -0.18
Support (2) -0.15 0.64 -0.18

 

1Corrected for attenuation.

2TASR variables contain all TASR scales except Support,
which was separated to form an independent index.

84

improvement in the combined scale alphas as well.

The variables were regrouped again. The Therapist
Action Scale Revised scales were combined at each time
point, but the support variable was split out to function
as an independent variable at each time point. The
pretherapy and posttherapy SCL90-R Global Severity Index
Scores were separated to function as independent scales.

A confirmatory factor analysis was performed on the new
variables (Table 10). Component scale-combined scale
correlations were higher. and component scales were more
parallel in intercorrelation. The coefficient alphas for

the combined scales were improved as well.

Path analysis was conducted on the combined variables
in Table 14. The time 2 data was dropped because of lower
reliability in the Therapist Action Scale Revised. alpha =
.35 compared with .48, .44, and .70 for the other times,
and component-combined scale correlation was erratic (.40,
1.0, 0.0). The time 2 data failed to follow a pattern that
the other time points shared involving the combined CALPAS
and TASR variables being more highly correlated with the
immediately following time point than with the more distant
points (Table 10). The two path models in Figure 1 and 2
were tested and will be discussed in relation to the

hypotheses of the study.

Figure 1

85

Path Model 1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2 1 e 1e
3 ...___. 7 11
4 ——I e 12
5 .L__.. 9 —- 13

 

 

 

 

 

 

 

 

 

 

 

14

 

 

 

Number

Variable Name

 

OQVO‘thP-JH

Pretherapy SCL90R Global Severity
CALPAS Time 1

TASR Time 1

Patient Negative Contribution Time
Support Time 1

CALPAS Time 3

TASR Time 3

Patient Negative Contribution Time
Support Time 3

CALPAS Time 4

TASR Time 4

Patient Negative Contribution Time
Support Time 4

Posttherapy SCL9OR Global Severity

Index Score

3

4

Index Score

 

86

Figure 2

Path Model 2

:4
In
.(4

(A
W“
by!
3.91

/ \

V"
f
\4
)2):

l
r
r’
b
r
4
>

 

 

 

 

 

 

 

 

 

 

Number Variable Name
1 Pretherapy SCL9OR Global Severity Index Score
2 CALPAS Time 1
3 TASR Time 1
4 Patient Negative Contribution Time 1
5 Support Time 1
6 CALPAS Time 3
7 TASR Time 3
8 Patient Negative Contribution Time 3
9 Support Time 3
10 CALPAS Time 4
11 TASR Time 4
12 Patient Negative Contribution Time 4
13 Support Time 4
14 Posttherapy SCL9OR Global Severity Index Score

 

87

Hypotheses 1 and 4

Hypothesis 1 stated that a high level of therapeutic
alliance in the first session as measured by CALPAS would
predict symptom reduction at the end of treatment on the
SCL90-R. Hypothesis 4a stated that after the initial level
of alliance is accounted for, the final level of alliance
will predict more outcome variance in SCL90-R symptom
reduction. Hypothesis 4b stated that CALPAS level of the
therapeutic alliance during the final session of therapy
would better predict symptom reductions during therapy than
would that level during the first session of therapy. Both
hypotheses are rejected.

All correlations of posttherapy SCL90-R Global Severity
Index (GSI) with other variables were low (see Table 17).
Due to these correlations and the low reliability of CALPAS,
the most parsimonious reaction is to reject the hypotheses.
With regard to hypothesis 1, CALPAS session one alliance
scores correlated only -.13 with posttherapy SCL90-R GSI.
Concerning hypotheses 4a and 4b, the initial alliance
correlated -.13 with GSI while the alliance measured by the
CALPAS at time 4 correlated .17 with GSI. Considering these
meager correlations and the change in sign as well as the
poor reliability of CALPAS, these

correlations are within the error of measurement.

88

Consistent with this evaluation of the correlations, the
Path coefficients in model 1 (Table 17, Figure 1)
predicting posttherapy SCL90-R GSI were low; The
coefficient for pretherapy GSI predicting posttherapy GSI
was only .26, and the coefficient for alliance at time 4
measured by CALPAS predicting Posttherapy SCL90-R GSI score

was .20, which was both small and not in the predicted

direction.

89

Table 1? Path Analysis 1
Number Variable Name

 

 

Reliability
1 Pretherapy SCL90R Global Severity Index Score 1.00
2 CALPAS Time 1 0.90
3 TASR Time 1 0.48
4 Patient Negative Contribution Time 1 1.00
5 Support Time 1 1.00
6 CALPAS Time 3 0.86
7 TASR Time 3 0.44
8 Patient Negative Contribution Time 3 1.00
9 Support Time 3 1.00
10 CALPAS Time 4 0.89
11 TASR Time 4 0.70
12 Patient Negative Contribution Time 4 1.00
13 Support Time 4 1.00
14 Posttherapy SCL90R Global Severity Index Score 1.00

 

Original Correlations
1 2 3 4 65* 6 T73 8 9 1o 11 123313 14

 

1 -37 -19 2 31 -8 4 16 12 -7 -1 16 10 24
2 -37 12 -56 -50 48 31 -5 2 30 -19 -9 10 -13
3 -19 12 34 8 3 57 14 4 -18 19 43 5 -13
4 ‘2 -56 34 34 -25 -15 16 -3 -37 28 36 -20 -7
5 31 -50 8 34 -38 9 -3 -4 -37 -4 19 -3 3
6 -8 48 3 -25 -38 32 -39 -14 61 21 -19 19 -11
7 4 31 57 -15 9 32 45 -13 28 35 -9 -12 -13
8 16 -5 14 16 -3 -39 45 18 -7 4 31 0 13
9 12 2 4 -3 -4 -14 -13 18 -16 -33 18 14 3
10 -7 30 -18 -37 -37 61 28 -7 -16 40 -38 30 17
11 -1 -19 19 28 -4 21 35 4 -33 40 26 -16 19
12 16 -9 43 36 19 -19 -9 31 18 -38 26 -6 2
13 10 10 5 -20 -3 19 -12 0 14 30 -16 -6 -5

14 24 -13 -13 -7 3 -11 -13 13 3 17 19 2 -5

 

Path Coefficients

 

 

1 **2* 3 4 5 6 7 a 9 10 11 12 15"12‘
1 o o o o o o o o o o o o o o
2 -37 o o o o o o o o o o o o o
3 -19 o o o o o o o o o o o o o
4 2 o o o o o o o o o o o o o
5 31 o o o o o o o o o o o o o
6 o 43 o o o o o o o o o o o o
7 o o 57 o o o o o o o o o o o
8 o o o 16 o o o o o o o o o o
9 o o o o -4 o o o o o o o o o
10 o o o o o 61 o o o o o o o o
11 o o o o o o 35 o o o o o o o
12 o o o o o o o 31 o o o o o o
13 o o o o o o o o 14 o o o o o
14 26 o o o o o o o o 20 9 2 -12 o

 

90

"Table 17 (cont'd.).“

 

Reproduced Correlations

 

 

 

1 2 3 4 5 6 7 8 9 10 11 12 13 14
1 100 -37 -19 2 31 -18 -11 0 -1 -11 -4 0 0 24
2 -37 100 7 -l -11 48 4 0 0 29 1 0 0 -4
3 -19 7 100 0 -6 3 57 0 0 2 20 0 0 -3
4 2 -1 0 100 1 0 0 16 0 0 0 5 0 1
5 31 -ll -6 l 100 -6 -3 0 -4 -3 -1 0 -1 7
6 -18 48 3 0 -6 100 2 0 0 61 1 0 0 7
7 -11 4 57 0 -3 2 100 0 0 l 35 0 0 0
8 0 0 0 l6 0 0 0 100 0 0 0 31 0 1
9 -1 0 0 0 -4 0 0 0 100 0 0 0 14 -2
10 -11 29 2 0 -3 61 1 0 0 100 0 0 0 17
11 -4 1 20 0 -1 1 35 0 0 0 100 0 0 8
12 0 0 0 5 0 0 0 31 0 0 0 100 0 2
13 0 0 0 0 -1 0 0 0 14 0 0 0 100 -12
14 24 -4 -3 1 7 7 0 l -2 l7 8 2 -12 100
Errors: (Actual - Reproduced)
1 2 3 4 S 6 7 8 9 10 11 12 13 14
l 0 0 0 0 0 10 15 16 13 4 3 16 10 0
2 0 0 5 -55 -39 0 27 -5 2 1 -20 -9 10 -9
3 0 5 0 34 14 0 0 14 4 -20 -1 43 5 —10
4 0 -55 34 0 33 -25 -15 0 -3 -37 28 31 -20 -8
5 0 -39 14 33 0 -32 12 -3 0 -34 -3 l9 -2 -4
6 10 0 0 -25 -32 0 30 -39 -14 0 20 -19 19 -18
7 15 27 0 -15 12 30 0 45 -13 27 0 -9 -12 -13
8 16 -5 14 0 -3 -39 45 0 l8 -7 4 0 0 12
9 13 2 4 -3 0 -14 -13 18 0 -16 -33 18 0 5
10 4 1 -20 -37 -34 0 27 -7 -l6 0 40 -38 30 0
11 3 -20 -1 28 -3 20 0 4 -33 40 0 26 -16 11
12 16 -9 43 31 l9 -l9 -9 0 18 -38 26 0 -6 0
13 10 10 5 -20 -2 19 -12 0 0 30 -16 -6 0 7
l4 0 -9 -10 -8 -4 ~18 -13 12 5 0 11 0 7 0

 

Note: The sum of squared errors in the lower triangle is
3.405053
Note: The analysis for the model as a whole is

The overall chisquare is 50.31
The degrees of Freedom are 74

91

Hypotheses 2 and 3

Hypothesis 2 stated that an initially poor therapeutic
alliance as measured by CALPAS can be improved by the end of
treatment as a result of these therapist technical activates
measured by TASR:

a) addressing the patient's defenses.

b) addressing the patient's guilt and expectation of

punishment.

c) addressing the patient's problematic feelings in

relation to the therapist.

d) linking the problematic feelings in relation to the

therapist with the patient's defenses.

Hypothesis 3 stated that an initially good therapeutic
alliance as measured by CALPAS can be damaged by therapist
technical activity measured by TASR. The alliance can be
weakened by the therapist failing to:

a) address the patient's defenses.

b) address the patient's guilt and expectation of

punishment.

c) address the patient's problematic feelings in

relation to the therapist

d) link the problematic feelings in relation to the

therapist with the patients defenses.
The hypotheses were rejected. These hypotheses taken

together essentially state that this group of therapist

92

activities strengthens and maintains the therapeutic
alliance. Testing them together avoids the substantial
reduction in power involved in partitioning an already small
sample. In addition, the therapist actions were combined
as discussed above to improve the reliability. The
resulting groups of therapist actions included: a) (TASR)-
addressing transference, addressing the patient's defenses,
making interventions that encourage patient expression, b)
(Support)- providing support. The results that follow are
based on the combined scales, and on not dividing the data
into initially high and low alliance groups.

Path analysis 2 (Table 18) indicates that the alliance
at time 3 measured by CALPAS was not affected by therapist
actions at time 1 measured by TASR. The path coefficient
for variable 3 onto variable 6 is -3. Similarly, path
analysis 2 indicates that the alliance at time 4 was not
affected by therapist actions at time 3. The path
coefficient for variable 7 onto variable 10 is -10.

Comparing path analysis model 1 (Table 17, Figure 1)
with path analysis model 2 (Table 18, Figure 2), path
analysis model 1 contains less error, which suggests it is

a better model. Path analysis 1 has pretherapy SCL90-R

Table 18 Path Analysis 2

93

 

 

 

 

 

 

Number Variable Name
Reliability
1 Pretherapy SCL90R Global Severity Index Score 1.00
2 CALPAS Time 1 0.90
3 TASR Time 1 0.48
4 Patient Negative Contribution Time 1 1.00
5 Support Time 1 1.00
6 CALPAS Time 3 0.86
7 TASR Time 3 0.44
8 Patient Negative Contribution Time 3 1.00
9 Support Time 3 1.00
10 CALPAS Time 4 0.89
11 TASR Time 4 0.70
12 Patient Negative Contribution Time 4 1.00
13 Support Time 4 1.00
14 Posttherapy SCL90R Global Severity Index Score 1.00
Original Correlations
1 **2 3 4 5 7 8 9 10 11 12 13 14
l -37 -19 2 31 -8 4 16 12 -7 -1 16 10 24
2 -37 12 -56 -50 48 31 -5 2 30 -19 -9 10 -13
3 -19 12 34 8 3 57 14 4 -l8 19 43 5 -l3
4 2 -56 34 34 -25 -15 16 -3 -37 28 36 -20 -7
5 31 -50 8 34 -38 9 -3 -4 -37 -4 l9 -3 3
6 -8 48 3 -25 -38 32 -39 -14 61 21 -19 19 —11
7 4 31 57 -15 9 32 45 -13 28 35 -9 -12 -13
8 16 -5 14 16 -3 -39 45 18 -7 4 31 0 13
9 12 2 4 -3 -4 -14 -13 18 -16 —33 18 14 3
10 -7 30 -18 -37 -37 61 28 -7 -16 40 -38 30 17
11 -1 -19 19 28 -4 21 35 4 -33 40 26 -l6 19
12 16 -9 43 36 19 -19 -9 31 18 -38 26 -6 2
13 10 10 5 -20 -3 19 -12 0 14 30 -16 -6 -5
14 24 -13 -13 -7 3 -ll -13 13 3 17 19 2 -5
Path Coefficients
1 2 3 4 *5 6 7 a 9 1o 11 12 13 14
l 0 0 0 01 0 0 0 0 0 0 0 0 0 0
2 -37 0 0 0 0 0 0 0 0 0 0 0 0 0
3 -19 0 0 0 0 0 0 0 0 0 0 0 0 0
4 2 0 0 0 0 0 0 0 0 0 0 0 0 0
5 31 0 0 0 0 0 0 0 0 0 0 0 0 0
6 0 42 -3 6 -19 0 0 0 0 0 0 0 0 0
7 0 13 66 -38 23 0 0 0 0 0 0 0 0 0
8 0 -4 11 14 -10 0 0 0 0 0 0 0 0 0
9 0 -5 7 -7 -5 0 0 0 0 0 0 0 0 0
10 0 0 0 0 0 73 -10 28 -12 0 0 0 0 0
ll 0 0 0 0 0 7 30 —2 -28 0 0 0 0 0
12 0 0 0 0 0 13 -35 51 6 0 0 0 0 0
13 0 0 0 0 0 48 -43 36 9 0 0 0 0 0
14 26 0 0 0 0 0 0 0 0 20 9 2 -12 0

 

"Table 18 (cont'd.)."

94

 

Reproduced Correlations

 

 

 

1 2 3 4 5 6 7 8 9 10 11 12 13 14

1 100 -37 -19 2 31 -21 -11 -4 -1 -15 -4 -1 -7 24

2 -37 100 7 -1 -11 44 16 -2 -4 31 9 -1 14 -4

3 -19 7 100 0 -6 1 66 11 7 -3 18 -17 -23 -2

4 2 -1 0 100 1 5 -38 14 -7 12 -10 21 23 0

5 31 -11 -6 1 100 -23 18 -11 -5 -21 5 -15 -23 7

6 -21 44 1 5 -23 100 -1 2 -1 74 7 14 49 4

7 —11 16 66 -38 18 -1 100 -1 6 -11 28 -35 -43 2

8 -4 -2 11 14 -11 2 -1 100 0 29 -2 51 38 1

9 -1 -4 7 -7 -5 -1 6 0 100 -13 -26 4 6 -6

10 -15 31 -3 12 -21 74 -11 29 -13 100 5 28 50 11
11 -4 9 18 -10 5 7 28 -2 -26 5 100 -12 -12 10
12 -l -1 -17 21 -15 14 -35 51 4 28 -12 100 41 2
13 -7 14 -23 23 -23 49 -43 38 6 50 -12 41 100 -4
14 24 -4 -2 0 7 4 2 1 -6 11 10 2 -4 100

Errors: (Actual - Reproduced)

1 2 3 4 5 6 7 8 9 10 11 12 13 14

1 0 0 0 0 0 13 15 20 13 8 3 17 17 0
2 0 0 5 -55 -39 4 15 -3 6 -1 -28 -8 -4 -9
3 0 S 0 34 14 2 -9 3 -3 -15 1 60 28 -11
4 0 -55 34 0 33 -30 23 2 4 -49 38 15 -43 -7
5 0 -39 14 33 0 -15 -9 8 1 -16 -9 34 20 -4
6 13 4 2 -30 -15 0 33 -41 -l3 -l3 14 -33 -30 -15
7 15 15 -9 23 -9 33 0 46 -19 39 7 26 31 -15
8 20 -3 3 2 8 -41 46 0 18 -36 6 -20 -38 12
9 l3 6 -3 4 1 -13 -19 18 0 -3 -7 l4 8 9
10 8 -l -15 -49 -16 -13 39 -36 -3 0 35 -66 -20 6
11 3 -28 1 38 -9 l4 7 6 -7 35 0 38 -4 9
12 17 -8 60 15 34 -33 26 -20 14 -66 38 0 -47 0
13 17 -4 28 -43 20 -30 31 -38 8 -20 -4 -47 0 -1
l4 0 -9 -11 -7 -4 -15 -15 12 9 6 9 0 -l 0

Note: The sum of squared errors

4.947172

in the lower triangle is

Note: The analysis for the model as a whole is

The overall chisquare is

63.38

The degrees of Freedom are 50

95

Global Severity Index score predicting each of the four
analysis variables at time 1: then each variable predicts
its counterpart at time 3: then time 3 variables predict
time 4: then the time 4 variables predict posttherapy
SCL90—R Global Severity Index score. Path analysis model 2
contains these relationships plus all the causal arrows
between the variables at each time point. Overall, the
mismatching of the path coefficients at the two time periods
(see Appendix B for technical details). combined with the
lower error in model 1, supports model 1 over model 2.
Accepting model 1 over model 2 indicates that the
variables are causally independent over time. This fails to
support hypothesis 2 and 3. Again, the near 0 path
coefficients for variable 3 onto 6 and 7 onto 10 (Table 18)
specifically fail to support hypothesis 2 and 3 as well as
demonstrating the superiority of model 1 over model 2, and

the causal independence of these particular variables.

Some Related Findings

While the particular hypotheses concerning the
therapeutic alliance and therapist actions were not
supported, note that most patients in this sample improved

in psychotherapy (Table 19, Figure 3). Pretherapy SCL90-R

96

Table 19

Analysis Variables Means and Standggd Devi_atiogs_

 

 

 

 

Variables Time 1 Timez Time 3 Time 4
Alliance1 .M 5.21 4.9 5.43 5.52
(CALPAS) §p .76 .66 .48 .52
Patient Negative M. 1.84 1.9 1.57 1.77
Contribution SQ. .34 .75 .57 .79
Therapist Actions2 M 1.46 1.6 1.56 1.58
(TASR) SE .27 .32 .31 .35
Support .M 1.28 1.3 1.40 1.63
SE. .34 .48 .54 .71
N 44 44 44 44
Pretherapy Posttherapy
Global Severity M l . 36 . 84
Index Score .SQ .60 .48
(SCL90R) N 44 44

 

1CALPAS variables at times 1-4 contain all scales averaged

together except patient Negative Contribution,

which was

separated to function as an independent index.

2The TASR variables at times 1-4 contain all scales averaged

together except Support,
independent index.

which was separated to form an

97

Figure 3

re__tb__eggp,y__5CL9DR GSI sv Pogttherapy 501.906: GSI

 

 

 

 

 

1 1 1 1 1 1 1 1 1 1
1
F.
o
s 2.1: 1
t
t
h
e
r 1.754
a
p 1 1
y 1
S 1.4+ 1
C
L 1
9
0 1 1
R 1.05: 1 1 1
1 1 1 1
G 1
S 1 1 1 1 1
I 1
.7: 1
S 1
c 1 1 1 1 1
o 1 1
r 1 1 1
e .35- 11 1 1
1 1 1
1
I I 1 TI I I I I I l
.75 1.25 1.75 2.25 2.75
.5 1 1.5 2 2.5

Pretherapy SCL9GR GSI Score

98

symptom level was reduced from a mean of 1.36 to a
posttherapy mean of .84. This represents a change of 1.08
standard deviations on the posttherpy SCL90-R distribution
of scores.

A high alliance at time 1 indicates that the alliance
will most likely be high at time 3 and 4 (Table 17),
because the path coefficient of variable 2 onto 6 is .48,
and the path coefficient of variable 6 onto 10 is .61.
Overall, there is a slight tendency for the alliance to get
stronger over time (Table 19). Likewise, high Patient
Negative Contribution at time 1 indicates. that for at
least some patients, Patient Negative Contribution will be
high at time 3 and time 4 (Table 17), because the path
coefficients of variable 4 onto variable 8 is .16, and the
path coefficient of variable 8 onto variable 12 is .31.
High therapist action at time 1 indicates that therapist
actions at time 3 and 4 will most likely be high (Table 17),
because the path coefficients of variables 3 onto 7 is .57,
and the path coefficient of variable 7 onto 11 is .35.

While correlations between different variables over
time are not substantial, the correlations within sessions
are high (Table 17, original correlations). Therapist
Support correlates inconsistently within session as well as
over time. However, Patient Negative Contribution,
California Psychotherapy Alliance Scale. and Therapist

Action Scale Revised are correlated in the same pattern

99

within each time point.

Alliance as measured by the California Psychotherapy
Alliance Scale (CALPAS) and therapist action as measured by
the Therapist Action Scale Revised (TASR) are correlated
within time and that correlation tended to increase over
time (Table 17, original correlations). At time 1 CALPAS
and TASR are correlated .12; at time 2 CALPAS and TASR are
correlated .22; at time 3 CALPAS and TASR are correlated
.32, and at time 4 CALPAS and TASR are correlated .40.
Similarly, Patient Negative contribution and TASR are
positively correlated across each time point (.34, .12, .45,
.26), even though CALPAS and Patient Negative contribution
are negatively correlated within times (-.56, -.46, -.39,

-.38).

DISCUSSION

This study has yielded a considerable amount of
information on the structure of the California Psychotherapy
Alliance Scale. Exploratory and confirmatory factor
analysis identified five factors which turned out to be
highly correlated with each other in this sample. Marmar.
Gaston, Gallagher and Thompson (1989) found similar
interscale correlations and they also treated the data by
combining the scales. Their intercorrelation may be the
consequance of raters having formed one overall impression
of the therapeutic alliance that influanced their ratings of
all items or a scoring manual that failed to fully clarify
the criteria for rating discrete attributes. There are also
good theoretical reasons to expect these intercorrelations.
High Patient Positive Contribution scores seem.likely when
both Patient Commitment, and Therapist Involvement are
high, and there is also high WOrking Strategy Consensus.
For well-trained therapists. such a level of Therapist
Involvement should be a given provided the patient is
active.

While refinement of the rater's manual and their more
intensive training might be helpful, there is also good
reason to question whether these subscales can differentiate

group data since most often they can be expected to vary

100

101

together. Future researchers should be cautious about
anticipating differences on the California Psychotherapy
Alliance Scale subscales. In fact, this study demonstrates
that the greatest reliability is obtained by aggregating
four of the CALPAS subscales into a single Patient Positive
Contribution index while retaining the Patient Negative
Contribution Scale.

Tables 5, 6, and 7 clarify how quickly relationships
between variables are dissolved when correlations are
compared over time or between raters, and how they are
diminished when correlations are compared between raters and
across time. To some extent. this suggests future
researchers need to take steps to insure reliability by
means of more intensive rater training. and more extensive
manual documentation. On the other hand. these results
suggest that there are many state-like aspects of the
alliance and fluctuation is the norm.

Reliability problems with the California Psychotherapy
Alliance Scale obscure the findings of this study. This
study has demonstrated some problems with the reliability of
the California Psychotherapy Alliance Scale. Previously,
CALPAS was used primarily by its developers. Given that the
variables of the study are correlated within sessions but
not across times. we can only speculate about the extent to

which correlations are obscured by the low reliability of

102

the California Psychotherapy Alliance Scale.

This study has provided further information on the
Therapist Action Scale Revised as well. While four discrete
subscales emerged, and there was less intercorrelation
between subscales. the internal consistency of the
subscales (measured by Cronbach's alpha) was unsatisfactory.
While subscales demonstrated only modest interrater
reliability, it may be desirable to include more items
required in order to establish internal consistency
(subscales now contain only 3-4 items). Although combining
subscales yielded more reliable scales for analysis, this
method does not permit investigation of what appear to be
conceptually discrete subdimensions (Addressing Transference

vs. Addressing Defenses).

Alliance and Outcome

Freud (1913/1958) hypothesized the primary and central
importance of the relationship between the client and
therapist or the therapeutic alliance in the process of
psychotherapy. He viewed the establishment and maintenance
of the alliance as a primary goal of treatment. Freud wrote
about the affective component or capability to form.a
positive transference that the client brings to treatment.

Sterba (1934) and Zetzel (1956) stressed the analytic ego

103

functions which were viewed as another component that the
client must bring to treatment.

Luborsky (1976) developed the first operational
translation of the alliance. Luborsky (1990) has been
evaluating empirically the proposition that a therapeutic
alliance must develop in order for a patient to benefit from
dynamic psychotherapy. Gaston (1990), in a theoretical
review article, has recapitulated the theories of the roles
played by the alliance in outcome that have been
investigated up to now. The alliance can be therapeutic in
and of itself. The alliance may be a prerequisite for
therapist interventions to be effective. The alliance may
interact with therapist interventions in determining
successful outcome. None of these supositions were
supported by the data of this study.

The initial alliance was viewed as particularly
predictive because the alliance was considered at least a
necessary condition for favorable outcome. Its presence
initially therefore at least assures that favorable outcome
is possible. Those who do not initially demonstrate an
alliance may or may not develop an alliance as therapy
progresses. Gaston et. al. (1990) found that the alliance
was a better predictor of outcome when it was measured later
in the therapy because more of the participants who

eventually develop an alliance are included in the

104

prediction. This finding is in direct contrast to the
findings that will be presented from this study, which
indicate that higher final alliance is associated with
GREATER posttherapy symptoms.

Hypotheses l and 4 will be discussed together.
Hypothesis 1 states that a high level of therapeutic
alliance in the first session as measured by the California
Psychotherapy Alliance Scale will predict symptom reduction
at the end of treatment on the SCL90-R. Hypothesis 4a
states that once one accounts for the initial level of
alliance, the final level of alliance will predict more
outcome variance in SCL90-R symptom.reduction. Hypothesis
4b states that the level of the therapeutic alliance as
measured by the California Psychotherapy Alliance Scale
during the final session of therapy is a better predictor of
symptom reduction during therapy than the level of alliance
during the first session of therapy. Both hypotheses were
rejected.

Initial therapeutic alliance and the final therapeutic
alliance are not highly enough correlated with lower
posttherapy symptom level to be statistically significant
with N = 44. However, Horvath and Symonds (1991) conducted
a meta-analysis of 24 studies done on 20 distinct data sets
and found a moderate and reliable relationship between the

therapeutic alliance and therapy outcome. The effect size

105

was based on correlation (Pearson) between alliance and
symptom outcome. The average effect size, when alliance
was rated by an outside rater and outcome was rated by the
client, was.£ = .21 with the standard deviation being.§p =
.14. The result of the present study is that initial
alliance is correlated .13 in magnitude1 with relatively
lower posttherapy symptoms. Since N = 44, the standard
error for this correlation is .15. The 95% confidence
interval is .13 :_.29, which includes the .21 found in the
meta-analysis. Thus the .13 found here does not differ by
more than sampling error from the .21 of the meta-analysis.

Certain key results have been included in the prior
presentation. These results have been highlighted in Table
20. Consider first the correlation between initial alliance
and posttherapy symptoms: the correlation often used to
assess outcome. This correlation is -.13 in the present
study and -.21 in the Horvath and Symonds (1991)
meta-analysis. This modest correlation has been interpreted
in past studies as suggesting a relationship between initial
alliance and outcome. we will show that this interpretation
can be questioned.

Consider the correlation between initial alliance and
pretherapy symptoms; a correlation of -.37. This
1In this study higher initial alliance is associated with

lower posttherapy symptoms so the sign of the
correlation is negative.

106

Table 20

The correlations between early and later alliance. and
sygptoms.

 

yrs 3 FA ATS

Before Therapy Symptoms (BTS) 1.00 -.37 -.07 .24
Initial Alliance (IA) -.37 1.00 .30 -.13
91:11.1 1111...... (FA) -.07 .30 1.00 .17

After Therapy Symptoms (ATS) .24 -.13 .17 1.00

 

107

suggests that patients who enter at a more disturbed level
have more trouble establishing rapport with the therapist.
But this means that patients with a poor initial alliance
not only have more symptoms after therapy but began*with
substantially more symptoms. Indeed. the fact that
patients with lesser alliance finish with more symptoms
could simply mean that they were initially more disturbed
and may NOT mean that they responded more poorly to therapy.

The key question is this: If patients were matched for
level of initial symptoms, how well would initial alliance
predict outcome? The statistical answer to this question is
the partial correlation between initial alliance and
posttherapy symptoms with pretherapy symptoms held constant.
This partial correlation is -.05. This partial correlation
is not statistically significant. This near 0 partial
correlation implies that the only reason that initial
alliance predicts symptoms after therapy is because it is
highly correlated with symptoms before therapy. This result
suggests that there is actually no relationship between
initial alliance as measured by this scale and subsequent
response to therapy.

This research also provides data on the relationship
between the later alliance and outcome. In particular. the
correlation between final alliance and posttherapy symptoms

is +.17. This is in the opposite direction to that

108

predicted: those with a high final alliance have more
symptoms than those with a poor final alliance. Before we
consider this an outcome relationship, we must conceptually
match the patients for pretherapy symptoms. The partial
correlation between final alliance and posttherapy symptoms
with initial symptoms held constant is +.19. That is. even
matched for initial symptom level, those with a strong final
alliance describe more symptoms than those with a weak final
alliance. The partial correlation of +.l9 can be compared
with the simple correlation of +.17.

Consider these two facts in conjunction: (a) Those with
a strong initial alliance are those with FEWER symptoms
before therapy, while (b) those with a strong alliance at
the end of therapy describe MORE symptoms than those with a
weak final alliance. This suggests that an increase in
alliance is associated with the admission of more symptoms.
Consider patients matched on both pretherapy symptoms and on
initial alliance. Individual differences within such a
matched group measure change in alliance and change in
symptoms. That is, the correlation within such a matched
group is the correlation between change in symptoms and
change in alliance. Statistically, this is the partial
correlation between final alliance and posttherapy symptoms
with both initial symptoms and initial alliance held

constant. That partial correlation is +.22. Thus in this

109

data there is a strong correlation between change in
alliance and change in stated symptoms. Given the fact that
the average patient increased in alliance and decreased in
symptoms, this means that those patients whose alliance
increased the most were those who described the smallest
decrease in symptoms.

This interpretation of the change in posttherapy
symptoms can also be checked by doing a multiple regression
of posttherapy symptoms onto initial symptoms, initial
alliance. and final alliance. The standardized regression
equation is:

(After Therapy Symptoms = .21 Before Therapy Symptoms -
.12 Initial Alliance + .22 Final Alliance) (see table 21)
In this equation, the regression weight for initial alliance
is negative while the weight for final alliance is positive.
This shows that outcome, as defined by decreased symptoms,
was best predicted by the change in alliance: the greater

the increase in alliance, the less the decrease in
symptoms. That is, for the patients in this study,
increased alliance led to greater description of symptoms.

There may be state-like and trait-like aspects to the

therapeutic alliance which are influenced by transference

110

Table 21

Multiple regression of posttherapy sypptoms onto initial
sypptoms. initial alliance, and final alliance.

 

 

Variables Mean Standard Deviation
Y1 = Pretherapy Symptoms 1.36 .60
x1 = Initial CALPAS 5.21 .76
x2 = Final CALPAS 5.52 .52
Y2 = Posttherapy Symptoms .84 .48

 

Raw Score Regression Equation

 

Y

2

.17 Y1 -

.07 X1 +

20 X2

.125

 

Consider various relatively common pairs of groups
(All combinations of the independent variables
either 1 SD above or 1 SD below the mean)

 

Yl

.76
.76

.76
.76

1.96
1.96

1.96
1.96

X1

4.45
4.45

5.97
5.97

4.45
4.45

5.97
5.97

X2

5.00
6.04

5.00
6.04

5.00
6.04

5.00
6.04

Y2

.96
1.17

.94
1.15

1.16
1.37

1.15
1.36

CALPAS
CALPAS

CALPAS
CALPAS

CALPAS
CALPAS

CALPAS
CALPAS

up a little
up more. higher symptoms

down
up, higher symptoms

up a little
up more, higher symptoms

down
up. higher symptoms

 

111

and resistance. In some clients a deeply entrenched
characterological negativism.and hostility may prevent any
type of alliance formation. However. this study suggests
that there are also clients who can form.a relationship with
time. Low initial alliance may be indicative of
moretransitory transference phenomenon and resistance based
on the material being discussed. which can be overcome in
therapy. Similarly, some individuals evidencing higher
initial negative contribution can utilize these expressions
of hostility to improve in therapy (the correlation of
initial patient negative contribution and posttherapy
symptoms was -.07). It is also possible that in some cases
high initial alliance does not guarantee a decrease in
symptom descriptions. The high initial alliance may be
indicative of idealized transference and obligatory
compliance which is not the same as an enduring commitment
to treatment. On the other hand. high final alliance may
indicate dependence on the therapist and regressive
transference, suggesting continued need for treatment.
Hypotheses 2 and 3 will be discussed together.
Hypothesis 2 states that an initially poor therapeutic
alliance as measured by the California Psychotherapy
Alliance Scale can be improved by the end of treatment as a
result of therapist activates measured by the Therapist

Action Scale Revised. Hypothesis 3 states that an initially

112

good therapeutic alliance as measured by the California
Psychotherapy Alliance Scale can be damaged by the absence
of certain therapist activity measured by the Therapist
Action Scale Revised. These hypotheses taken together
essentially state that the following group of therapist
actions strengthen and maintain the therapeutic alliance: a)
addressing transference, addressing the patient’s defenses,
making interventions that encourage patient expression, b)
providing support‘. Both hypotheses were rejected"
Therapist actions were not associated with the
strengthening or weakening of the alliance over time.
Therapeutic alliance and therapist actions were causally
independent over time. This finding is in direct
contradiction to Forman and Marmar (1985) who found, in a
pilot study of six individuals. that addressing
transference and addressing defenses was associated with
improvement in the therapeutic alliance. This study's
finding is consistent with Lansford's (1986) results, also
based on only six individuals, which indicated patient

factors to be more important than therapist activity in the

1As discussed in the Results section. grouping of
therapist actions was collapsed to these broader scales
to improve reliability.

113

repair of weakened alliances. In a single case study,
Gabbard et. al. (1988) found that the alliance with a
borderline patient was improved with transference
interpretations. Studies with one or six subjects are less
persuasive than studies with greater numbers of subjects.
In this regard, the finding of this study based on 44
individuals clearly fails to support the hypothesis that
therapist actions improved and sustained the alliance over
time.

A patient with a high alliance at time 1 will most
likely have a high alliance at time 3 and time 4.
Considering this finding in connection with the failure of
therapist action to impact on the alliance over time
supports the notion that the formation of the alliance rests
largely with the capacity of the patient. Piper, Azim,
Joyce. McCallum, Nixon, and Segal (1991) found that the
lifelong quality of the patient's object relations was a
better predictor of the therapeutic alliance and therapy
outcome than measures of recent interpersonal functioning.
This is to say that patients enter therapy with a certain
trait-like capacity to form a relationship with the
therapist that will facilitate therapy. and this capacity
is not malleable by the therapist. Patients may likewise
have a trait-like capacity for change from initial to final

level of alliance. Considering the level of

114

pathology that would be indicated by lifelong impaired
object relations (e.g. borderline personality, narcissistic
personality, psychosis), it is not surprising that the
quality of relationship capability was notfound malleable,
particularly in one year of psychotherapy.

In this study it is clear that most patients who rated
high on Patient Negative contribution at time 1 were still
high on Patient Negative Contribution at time 3 and time 4.
This further supports the notion that the capability to form
a relationship or not form a relationship lies with the
patient. It also supports Finell's (1987) contention that
there is a negative therapeutic reaction rooted in
characterological negativism and deeply rooted negative
transference, which one could theorize goes back to Freud's
notion of identification with a punitive parent. However,
these individuals have all stayed in therapy and experienced
symptom reduction. In fact, some individuals demonstrating
higher initial negative contribution can utilize these
expressions of hostility to improve in therapy (the
correlation of initial patient negative contribution and
posttherapy symptoms was -.07). The therapist's willingness
to continue to work with these types of individuals may yet
be fruitful. This study may be a snapshot of a longer term
change process for these individuals.

Within individual sessions there is a relationship

115

between therapeutic alliance and therapist action, and this
relationship gets stronger over time. Either therapist
actions influence patient behavior within sessions. and/or
patient behavior influences therapist action within
sessions. Following the patient centered model of the
alliance, when the patient is working and producing
material or being hostile, the therapist can provide a
range of responses. It is also possible that in the
individual session therapist actions facilitate the
development of the alliance. The increasing relationship
over time between therapist actions and the alliance
indicates that the therapist calibrates actions to the
client's demonstrated capabilities better over time. Given
the overall favorable outcome in symptom reduction for this
sample, this calibration between the client and the
therapist appears to be constructive.

One positive implication of this study is that clients
who stay in therapy improve symptomatically over time. It
is also reassuring from a person-centered perspective that
to a substantial degree. the capacity to form.an alliance
and change in therapy lies with the client. It is also
reassuring that continuing to work with initially hostile

clients may be fruitful as well.

116

Ipplications for Theogy and Practice

This study suggests that the importance of the initial
alliance in predicting outcome in psychotherapy may need to
be questioned. A strong initial alliance does not guarantee
benefit from therapy in a population where all patients
appear to have symptom reduction. The absence of a strong
initial alliance did not preclude benefits from
psychotherapy.

The view of effective therapy demonstrated by this
study is the following. Forming a strong initial alliance
is not associated with symptomatic improvement, because
this association is influenced by initial symptomatology.
The partial correlation between initial alliance and
posttherapy symptoms with initial symptoms partialed out was
-.05. In the final stage of therapy it is those who have a
weaker alliance who describe less symptoms (correlation of
initial alliance with posttherapy symptoms is -.13 whereas
correlation of final alliance with.posttherapy symptoms is
.17). This interpretation of the change in posttherapy
symptoms can also be checked by doing a multiple regression
of posttherapy symptoms onto initial symptoms. initial
alliance. and final alliance. The standardized regression
equation is:

ATS = .21 BTS - .12 IA + .22 FA (see Table 21)

117

One theory of improvement is supported by this data. First
the alliance must be established. then those patients who
are confronted with alternative views of themselves and
their problems by the therapist to the point at which the
alliance actually decreases are those who improve. However,
therapists do not appear to be accurate judges of when the
alliance is weak or strong. The correlation between the two
therapist-judges alliance ratings was .26 for patient
positive contribution and .60 for patient negative
contribution. This difficulty assessing the alliance
suggests that therapists should simply plan to do as much as
possible to build an alliance during the first two sessions
of therapy and then switch to a more confrontational
strategy.

Rather than a continual building of the alliance
throughout therapy it is giving the patient an alternative
view that is effective. Solely building the alliance does
not appear to be effective and in fact the data suggests
that this may not contribute to patient statements
concerning symptoms. The therapist must not reinforce the
patient's pathology but rather must ultimately present the
patient with an alternative view, which the patient may not
want to hear. This presentation may be to the detriment of

the alliance, yet may lead to symptom.reduction.

118

Ipplications for Future Research

An investigation of the speculations above is worth
considering. The degree to which therapists confront
patients with a different view of themselves and their
problemm could be measured and the relationship of the
degree of this action to alliance and outcome could be
evaluated.

Investigation of therapy process variables with
specialized treatments of specific disorders such as
systematic desensitization of phobias, sex and marital
therapy procedures etc. would help in evaluating the
importance of the alliance. It may be that there are
particular techniques that are more directly related to
symptomatic improvement and the course of the alliance.
because the therapist actions assessed in this study did not
demonstrate a direct relationship to improvement or the
alliance. Comparing the role of the alliance in treatment
by more and less experienced therapists could be useful,
because therapist skill may impact on the effectiveness of
technique. Although this study stopped short of assessing
the impact on the alliance of the accuracy and theoretical
integrity of the therapist's interventions, investigations

along these lines in the future would also be useful.

APPENDIX A

Scale Construction

APPENDIX A

Scale Construction

California Psychotherapy Alliance Scale

The Patient Positive Contribution scale items (1-6)
(Table 1) along with item.(17) loaded on the first factor of
the exploratory factor analysis (Table 3), which accounted
for 24% of the total variance. Judging from the content
item.(17) appears to belong there, and its high correlation
with that scale (.80 as opposed to .70 with its hypothesized
scale) further clarifies its placement. The items correlate
from .77 to .91 with the scale in confirmatory factor
analysis (Table 4). The coefficient alpha for this scale is
.94.

The Patient Negative Contribution scale items (8.9.11)
(Table 1) and item (22) loaded on factor three of the
exploratory factor analysis (Table 3), which accounted for
16% of the total variance. Item (10). ‘which is
hypothesized to be in this scale. loaded highest with this
factor .49 after the null fifth factor, which included only
items (10.12). The items (8,9,10,11) correlate from .58 to
.84 with this scale in confirmatory factor analysis (Table

4). The coefficient alpha is .77.

119

120

The Patient Commitment Items (13.14.15.16,18) (Table 1)
loaded on factor 4 of the exploratory factor analysis (Table
3). which accounted for 11% of the total
variance. The items (13.14.15.16,18) correlate from .81 to
.85 with this scale in confirmatory factor analysis (Table
4). The coefficient alpha is .92.

The Working Strategy Consensus scale items (19-23)
(Table l) loaded moderately on the four meaningful factors
of the exploratory factor analysis (Table 3). However. the
items (19-23) correlate from .81 to .88 with this scale in
confirmatory factor analysis (Table 4). which is higher
than their correlation with any of the other scales. The
coefficient alpha is .92.

The Therapist Involvement Scale items (25-30) along
with item 24 load on factor 2 of the exploratory factor
analysis (Table 3), which explains 16% of the total
variance. Judging from the content item.(24) belongs with
the therapist items. The items (25-30. and 24) correlate
from..63 to .87 with this scale in confirmatory factor
analysis (Table 4). The coefficient alpha is .90. Thus the
structure of the California Psychotherapy Alliance Scale is

basically confirmed.

121

Therapist Action Scale Revised

The Addressing Transference Scale items (1-5) (Table 2)
loaded along with items (7 and 9) on the first factor of the
exploratory factor analysis (Table 11). which accounted for
14% of the total variance. Items (7 and 9) are moved back
to their hypothesized scale Addressing Defenses. Items (2
and 4) are dropped because they correlate poorly with the
other items on the scale (1.3.5). and poorly with the scale
score. Items (1,3,5) correlate from .51 to .77 with the
scale in confirmatory factor analysis (Table 12). The
coefficient alpha for this scale is .65.

The Addressing Defenses items were split between factor
1 and factor 2. Because these items belong together
conceptually and are well correlated with one another they
were put together on the Addressing Defenses Scale. The
items correlate from .63 to .77 with this scale in
confirmatory factor analysis (Table 12). The coefficient
alpha for this scale is .78.

The Expressive Intervention items (16-18) along with
item.13, loaded on factor 2 of the exploratory factor
analysis (Table 11). Because of its higher loading on this
factor and because it conceptually seems to belong there it
was moved to this scale. The items (16-28 and 13) correlate

from .40 to .67 with the scale. The coefficient alpha for

122

this scale is .57.

The Support items (11.12.14.15) (Table 2) loaded on
factor 3 of the exploratory factor analysis (Table 11),
which explained 13% of the total variance. The items
correlate from .59 to .80 with the scale (Table 11). The
coefficient alpha for this scale is .78. The structure of

the Therapist Action Scale is basically confirmed.

APPENDIX B

Some Technical Aspects of the Data Analysis

APPENDIX B

Some Technical Aspects of the Data Analysis

In terms of most theories involving longitudinal data.
the block of variables 6 to 9 correlated with variables 2 to
5 and the block of variables 10 to 13 correlated with
variables 6 to 9 should differ only by sampling error (Table
17). The best estimate of both correlation blocks would be
to average the two. Variable 2 onto variable 6 correlates
.42 which matches up with variable 6 onto variable 10
correlated .73. However, variable 4 onto variable 7
correlates -.38 which does not match up with variable 8 onto
variable 11 correlated —.02. An even worse match is
variable 5 on to variable 7 correlated .23 which does not
match up with variable 9 onto variable 11 correlated -.28.
Variable 2.3.4, and 5 onto 9 are correlated near 0, while
variables 6,7, and 8 onto 13 are correlated high (.48 -.43
.36). These fluctuations are the result of multicolinearity
resulting from variables 6 and 8 being correlated -.39.
variables 6 and 7 being correlated .32, and variables 7 and
8 being correlated .45. Overall, the mismatching of the

path coefficients at the two time periods, combined with

123

124

the lower error in model 1 supports model 1 over model 2.
Overall, the mismatching of the path coefficients at the
two time periods, combined with the lower error in model 1

supports model 1 over model 2.

REFERENCES

Bordin. 3.8. (1979). The generalizability of the
psychoanalytic concept of the working alliance.
Psychotherapy: Theogyg Research and Practice. 1§j3),
252-260.

Brenner. C. (1979). working alliance. therapeutic alliance.
and transference. Journal of the American
Psychoanalytic Association, 21, 137-158.

Crits-Christoph, P., Cooper. A.. & Luborsky, L. (1988). The
accuracy of therapists' interpretations and the outcome
of dynamic psychotherapy. Journal of Consulting and
Clinical Psychology,.§§(4), 490-495.

Derogatis. L. (1983). SCL-90-R administrationI scoring &
procedures manual-II. Towson: Clinical Psychometric

Research.

Derogatis, L., Myer, J., a King. R. (1981). Psychotherapy in
individuals with sexual dysfunction. American JOurnal
p§.Psychiat£y..1;§, 757-763.

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

Eaton. T., Abeles, N. & Gutfreund, M. J. (1988). Therapeutic

125

126

alliance and outcome: Impact of treatment length and
pretreatment symptomatology. Psychotherapy,.2§(4),
536-542.

Finell, J. S. (1987). A challenge to psychoanalysis: A
review of the negative therapeutic reaction.

Psychoanalytic Review. 11(4). 487-515.
Finn. R. H. (1970). A note on estimating the reliability of

categorical data. Educational and Psychological
Measurment. 39. 71-76.

Foreman. S. A.. & Marmar, C. R. (1985). Therapist actions
that address initially poor therapeutic alliances in
psychotherapy. American Journal of Psychiatgy. 152j8),
922-926.

Freud, S. (1958/1913). On beginning the treatment. In J.

Strachey. (Ed.). The standard edition of the copplete
psychological works of Sigmu_£d_ Freud. (Vol. ;2_, pp.

121-144). London: Hogarth Press.

Frieswyk, S. H.. Allen, J. G.. Colson, D. B., Coyne, L.,
Gabbard. G. O., Horwitz, L., & Newsom. G. (1986).
Therapeutic alliance: Its place as a process and
outcome varriable in dynamic psychotherapy research.
Journal of Consulting and Clinical Psychology,.§g.
32-38.

Frieswyk, S. H.. Colson, D. B., 6 Allen, J. G. (1984).

Conceptualizing the alliance from a psychoanalytic

127

perspective. Psychotherapy,.gl. 460-464.
Gabbard. G. O.. Horwitz, L., Frieswyk, S., Allen, J. G..

Colson, D. B., Newsom. G.. & Coyne. L. (1988). The

effect of therapist interventions on the therapeutic

alliance with borderline patients. JOurnal of the

American Psychoanalygic Association. 39(3),

697-727.

Gaston, L. (1990). The concept of the alliance and its role

in psychotherapy: Theoretical and empirical
consderations. Psychotherapy. 31(2), 143-153.

Gaston, L., Marmar, C.. Gallagher, D., 5 Thompson. L. W.
(1990). Variations in the alliance over time:
Antecedent or consequence of symptomatic change?.
Manuscript Submitted for Publication.

Gaston, L., Marmar, C. R.. Thompson, L. W4, & Gallagher, D.
(1988). Relation of patient pretreatment

characteristics with the therapeutic alliance in

diverse psychotherapies. JOurnal of Consulting and
Clinical Psychology,.§§(4), 483-489.

Gomes-Schwartz, B. (1978). Effective ingredients in

psychotherapy: Prediction of outcome from.process

variables. Journal of Consulting and Clinical

Psychology,.g§. 1023-1035.
Greenson. R. (1965). The working alliance and the

transference neurosis. Psychoanalysis Qparterly..35(2),

128

155-181.

Hartley, D. E., & StruDD. H. (1983). The therapeutic
alliance: Its relationship to outcome in brief
psychotherapy. In J. Masling (Ed.), gppirical Studies
.2; Psychoanalysis. (pp. 1-37). Hillsdale. NJ: The
analytic press.

Horowitz, M. J., Marmar, M. J., Weiss. D. S., DeWitt. K. N.,
& Rosenbaum. R. (1984). Brief psychotherapy of
bereavement reactions, the relationship of process to
outcome. Archives of General Psychiatgy..£1, 438-448.

Horvath. A. O. a Symonds. D. (1991). Relation between
working alliance and outcome in psychotherapy: A
meta-analysis. JOurnal of Counseling Psychologyy.3§(2),
139-149.

Hoyt, M. (1989). Therapist action scale manual. Unpublished
mauscript, Langley Porter Institute. University of
California.

Hoyt, M. (1980). Therapist and patient actions in'Good'
psychotherapy sissions. Archives of Genereal
Psychiatgy,.;1, 159-161.

Hoyt, M. F.. Marmar, C. R.. Horowitz. M. J., & Alvarez, W.F.
(1981). The therapist action scale and the patient
action scale: Instruments for the assessment of
activities during dynamic psychotherapy. Psychotherapy:
TheogyI Research and Practice, 18(1): 109-116.

129

Hoyt, M. F.. Xenakis. S. N., Marmar. C. R. & Horowitz, M. J.
(1983). Therapists' actions that influence their
perceptions of "Good" psychotherapy sissions. Journal
of Nervous and Mental Disease, 111(7). 400-404.

Rernberg. O. F. (1976). Some methodological and strategic
issues in psychotherapy research: Research implications
of the Menninger Foundation's psychotherapy research

project (pp. 23-38). In R. L. Spitzer & D. F.
Klein (Eds.). Evaluation of Psychological
Therapies. Baltimore: The Jones Hopkins University
Press.

Klee, M. R.. Abeles, N. & Muller, R. T. (1990). Therapeutic
alliance: Early indicators, course, and outcome.
Psychotherapy. 21(2), 166-174.

Lafferty, P., Beutler, L. E. & Crago, M. (1989). Differences
between.more and less effective psychotherapists: A
study of select treatment variables. JOurnal of

Consulting and Clinical Psychology,.§1, 76-80.
Langs. R. J. (1975). Therapeutic misalliances. International

JOurnal of Psychoanalypic Psychotherapy,.g, 77-105.
Lansford, E. (1986). Weakenings and repairs of the working
alliance in short-term psychotherapy. Professional

Psychology Research and Practice..11(4). 364-366.
Luborsky, L. (1976). Helping alliances in psychotherapy. In

J. L. Claghorn (Ed.), Successful Psychotherapy (DD.

130

92-118). New York: Brunner/Mazel.

Luborsky, L., Barber. J. P., & Crits-Christoph. P. (1990).
Theory-based research for understanding the process of
dynamic psychotherapy. Journal of Consulting and
Clinical Psychology,.§§(3). 281-287.

Luborsky, L., Crits-Christoph. P., Mintz. J., & Auerbach. A.
(1988). Who Will Benefit from.Psychotherapy?. New'York:
Basic Books Inc.

Luborsky, L., McLellan. T., Woody. G. E., O'Brien, C. P. &
Auerbach. A. (1985). Therapist success and its

determinants. Archives of General Psychiatgy,.gg,
602-611.

Malan, D. H. (1976). Toward the Validation of Qypgpic
Psychotherapy. New York: Plenum.Medical Book Company.
Malan, D. H. (1979). Individual psychotherapy and the

science of psychodypgpics. London: Butterworths.
Mann. J. (1973). Time Limited Psychotherapy. Cambridge:

Harvard University Press.

Marmar. C. R.. Horowitz, M. J., Weiss. D. S., & Marziali, E.
(1986). The development of the therapeutic alliance
rating system, In L. S. Greenberg & W. M. Pinsof
(Eds.), The psychotherapeutic process: A research
handbook (pp. 367-390). New York: The Guilford Press.

Marmar. C.. & Gaston, L. (1989). California Psychotherapy
Alliance Scales (CALPAS) manual. Unpublished

131

manuscript. University of California, San Francisco.
Marmar. C. R.. Gaston, L., Gallagher, D., 6 Thompson. L. W.
(1989). Alliance and outcome in late-life depression.

The Journal of Nervous and Mental Disease. 177(8),

 

464-472.

Marmar, C. R.. Weiss, D. S.. & Gaston. L. (1989). Toward the
validation of the california therapeutic alliance
rating systems Psychological Assessment..1(l), 46-52.

Marziali, E. (1984a). Prediction of outcome of brief
psychotherapy from therapist interpretive
interventions. Archives of General Psychiatgy, 41,
301-304.

Marziali, E. (1984b). Three viewpoints on the therapeutic
alliance similarities. differences, and associations
with psychotherapy outcome. The Journal of Nervous and
Mental Disease. 112(7). 417-423.

Marziali, E., Marmar. C.. & Erupnick, J. (1981). Therapeutic
alliance scales: Development and relationship to
psychotherapy outcome. American Journal of Psychiatgyy
1§§(3). 361-364.

Morgan, R.. Luborsky. L., Crits-Christoph, P., Curtis, H.. 5
Solomon. J. (1982). Predicting the outcomes of
psychotherapy by the Penn helping alliance rating

method. Archives of General Psychiatgy,.32. 397-402.
O'Malley, S. S., Foley. S. H.. Rounsaville, B. J., Watkins.

132

J. T., Sotsky, S., Imber. S. D., & Elkin, I. (1988).
Therapist competence and patient outcome in
interpersonal psychotherapy of depression. Journal of

Consulting and Clinical Psychologyg.§§. 496-501.
Olinick. S. L. (1964). The negative therapeutic reaction.

International JOurnal of Psychoanalysis..1§. 540-548.

Piper, W. E., Debbane, E. G.. Bienvenu, J., Carufel, F. &
Garant, J. (1986). Relationships between the object
focus of the therapist interpretations and outcome in
short-term individual psychotherapy. British JOurnal of
Medical Psychology, 13, 1-11.

Piper, W. E., Azim, H. F.. Joyce, A. S., McCallum. M.,
Nixon, G. W., & Segal, P. S. (1991). Quality of object

relations versus interpersonal functioning as
predictors of therapeutic alliance and
psychotherapy outcome. :22 Journal of Nervous and
Mental Disease. 111(7). 432-438.

Prusoff, B., Weismann. M., Klerman. G. L., & Rounsaville. B.
J. (1980). Research diagnostic criteria subtypes of
depression: Their role as predictors of differential
response to psychotherapy and drug treatment. Archives
.2; General Psychiatgy,.11, 796-801.

Rogers. C. (1959). A theory of therapy, personality. and
interpersonal relationships. as developed in the

client-centered framework. In S. Roch (Ed.).

133

Psychology:.1_g§pdy o; science. formulations of the
pgggon and the socia1 context (pp. 184-256). New York:
McGraw-Hill.

Saltzman, C.. Luetgert, M. J., Roth. C. H.. Creaser. J. &
Howard. L. (1976). Formulation of a therapeutic
relationship: Experiences during the initial phase of
psychotherapy as predictors of treatment duration and
outcome. Journal of Consulting and Clinical Psychology.
.$£(4). 546-555.

Sterba, R. (1934). The fate of the ego in analytic therapy.
International Journal of Psychoanalysis,.1§, 117-126.

Stone, L. (1961). The Psychoanalytic Situation. New York:
International Universities Press.

Strupp, H. (1980). Success and failure in time-limited
psychotherapy; Archives of General Psychiatgy,.11,
708-716.

Titchenor, V., a Hill, C. E. (1989). A comparison of six
measures of working alliance. Psychotherapy 26(2)
195-199.

Tinsley, H. E. & weiss, D. J. (1975). Interrater reliability
and agreement of subjective judgments. Journal of
Counseling Psychology..13(4) 358-376.

Windholz, M. J., Weiss. D. S.. & Horowitz, M. J. (1985). An
empirical study of the natural history of time-limited

psychotherapy for stress response syndromes.

134

Psychotherapy,.1, 547-554.
Xenakis. S. N., Hoyt, M. F.. Marmar. C. R.. & Horowitz. M.

J. (1983). Reliability of self-reports by therapists
using the therapist action scale. Psychotherapy:
Theogy. Research and Practice, 2Q, 314-320.

Zetzel, E. R. (1956). Current concepts of transference.

International Journal of Psychoanalysis,.11, 369-375.

"I7111‘1111111‘11‘11