Content uploaded by John Christopher Muran
Author content
All content in this area was uploaded by John Christopher Muran
Content may be subject to copyright.
American Journal of Psychotherapy
Copy of e-mail Notification zu52457
American Journal of Psychotherapy article (2457)
=====
Hello,
Attached please find the PDF of article (2006081) for the
most current issue of American Journal of Psychotherapy.
http://rapidproof.cadmus.com/RapidProof/retrieval/index.jsp
Login: Your email address
Password: ----
If you need assistance please contact me.
Pamela S. Mellick
Account Manager
Cadmus Communications, a Cenveo company
2905 Byrdhill Road
Richmond, VA 23228-5867
Tel: 804-261-3099
Fax: 804-264-0510
mellickp@cadmus.com
www.cenveo.com
Subjective and Intersubjective Analyses of
the Therapeutic Alliance in a Brief
Relational Therapy
EYAL ROZMARIN, Ph.D.* J.CHRISTOPHER MURAN, Ph.D.#
JEREMY SAFRAN, Ph.D.† BERNARD GORMAN, Ph.D.‡
JAKE NAGY, Ph.D.† ARNOLD WINSTON, M.D.#
The aim of the study described in this paper was to develop a method for
measuring the therapeutic alliance from an intersubjective perspective and to
evaluate the efficacy of the measure in predicting psychotherapy outcome. We
conducted the study using data from 22 patient–therapist dyads engaged in a
30–session protocol of a brief relational therapy. The alliance measure chosen
for this purpose was the short form of the Working Alliance Inventory. We
used the subjective patient and therapist versions of the measure and created
a correlation index representing the intersubjective congruence between
patients and therapists on their ratings of the alliance. We examined the
relations among the measures, as well as their predictive relation to an
outcome measure. The results showed significant intercorrelations among the
three alliance measures, suggesting that all captured aspects of the therapeutic
alliance. In addition, all three measures were significantly predictive of
outcome, with the correlation index appearing more powerful.
SUBJECTIVE AND INTERSUBJECTIVE ANALYSES OF THE
THERAPEUTIC ALLIANCE IN A BRIEF RELATIONAL THERAPY
The relationship between patient and therapist has been a subject of
investigation in the field of psychotherapy research since its inception.
While the traditional therapeutic relationship was conceived as a stage on
which individuals’ experience unfold, there soon emerged interest in how
this stage functions to enable and constitute the work of therapy. One
notion most investigated and discussed is that of therapeutic alliance. The
*New York University Postdoctoral Program in Psychotherapy and Psychoanalysis, Beth Israel
Medical Center; #Beth Israel Medical Center, Albert Einstein College of Medicine; †New School
University, Beth Israel Medical Center; ‡State University of New York, Beth Israel Medical
Center.
Mailing address:
270 Lafayette Street, Suite 1209, New York, NY 10012. e-mail:
erozmarin@earthlink.net
AMERICAN JOURNAL OF PSYCHOTHERAPY, Vol. 62, No. 3, 2008
1
rich4/zu5-apsy/zu5-apsy/zu500308/zu52457d08g
mcintyrn S⫽3 8/19/08 15:22 Art: 2006081
concept of therapeutic alliance formulated by Sterba (1934) was an aspect
of the patient’s position in psychotherapy. In Sterba’s thinking, for the
patient to persist in therapy he must be able to remember, even when
flooded with transference anxiety or rage, that the therapist is a benevolent
professional out to provide him with help. In this view, the alliance is
therefore the patient’s ability to transcend the dynamics of transference
and maintain a general appreciation of the therapeutic situation and its
‘real’ characteristics. Zetzel (1956) further developed the concept of
alliance; he believed that this capacity of the patient must have an affective
as well as cognitive component. Greenson (1967, 1971) differentiated
between the alliance and other aspects of the therapeutic relationship.
Bordin (1971), in a move towards a more interpersonal perspective,
expanded the concept beyond the patient’s beliefs and feelings to include
two mutual aspects of the relationship between patient and therapist:
agreement on goals and agreement on tasks. In a more recent contribution,
Safran and Muran (2000), informed by the emerging relational perspective
in psychoanalytic thinking (Aron, 1996; Benjamin, 1990; Mitchell, 1993),
refined the concept of alliance by replacing the notion of “agreement” with
that of “negotiation.” This concept puts an emphasis on the process by
which the tasks and goals of therapy develop and transform in the course
of the therapeutic endeavor.
In the field of psychotherapy research, the concept of therapeutic
alliance has received extensive attention and been cited as the most robust
predictor of treatment outcome (Horvath & Symonds, 1991; Martin,
Garske, & Davis, 2000). It has also become central to the “common
factors” argument regarding what accounts for change in psychotherapy
(e.g., Wampold, 2002). The first attempts to assess the alliance were efforts
to map and identify elements of the patient’s contribution from an
observer’s perspective. Studies including Gomes–Schwartz (1978), Hartley
and Strupp (1983), Luborsky et al. (1983), Marmar et al. (1986, 1989), and
Gaston (1990), relied on clinical judges using a variety of rating scales.
Overall, these studies were able to distinguish between the patient’s
affective stance towards the therapist, often labeled “therapeutic alliance,”
and his or her actual behavior, often titled “working alliance.” Subse-
quently, attention turned towards the therapist’s contribution, as well as
both therapists’ and patients’ negative contributions (Marziali, 1984; Mar-
mar, Weiss & Gaston, 1989; Marmar, Gaston, Gallagher & Thompson,
1989). Marmar et al. (1989) summarized the observer–based body of
research identifying five alliance components:
AMERICAN JOURNAL OF PSYCHOTHERAPY
2
rich4/zu5-apsy/zu5-apsy/zu500308/zu52457d08g
mcintyrn S⫽3 8/19/08 15:22 Art: 2006081
1) patient commitment,
2) patient working capacity,
3) therapist understanding and involvement,
4) patient hostile resistance, and
5) therapist negative contribution.
Another alliance research trajectory relied on alliance ratings of patients
and therapists rather than those of observers. Pioneered by Marziali
(1984), research in this field also expanded its scope to include Bordin’s
(1979) alliance factors of bond, agreement on goals, and agreement on
tasks (Marmar et al., 1989). Horvath and Greenberg (1989) developed the
Working Alliance Inventory (WAI) using the dimensions of goals, tasks
and bond. Marmar and his collaborators developed the California Psycho-
therapy Alliance Scale (CALPAS) using similar and additional factors. The
findings obtained using participants’ alliance ratings mirrored those ob-
tained using those of observers, with one significant and interesting
variation: the factor of patient hostile resistance was replaced by “disagree-
ment on goals and strategies.”
Horvath and Symonds (1991) conducted a meta–analysis of 24 studies
examining the relations between alliance and psychotherapy outcome;
both alliance and outcome were assessed using a wide range of measures
and perspectives. The results were simultaneously clear and confusing.
They concluded that the alliance is “a relatively robust variable linking
therapy process and outcome” (p. 146). They found that the alliance was
equally significant and predictive of outcome in all of the psychotherapy
methods included in the meta–analysis, providing strong support to the
idea that the therapeutic alliance is the elusive general factor responsible
for the effect of psychotherapy. However, contrary to previous research
showing high correlations between all significant alliance measures (Han-
sell, 1990; Safran & Wallner, 1990; Tichnor & Hill, 1989), Horvath and
Symonds found that different alliance measures showed substantial heter-
ogeneity in predicting outcome. Their explanation for this finding was that
different alliance measures tap into somewhat different constructs. They
further suggested that there might be “another variable involved in the
causal link between alliance and success in therapy” (p. 147). Reviewing
multiple definitions of alliance, Horvath and Symonds speculated that this
variable is related to notions of “mutuality, collaboration and engagement”
(p. 147).
Parallel to the evolution on alliance theory and research focus from
individual factors to the therapeutic relationship, psychoanalytic thinking
has been shifting from a one– to two–person psychology. In the last two
Therapeutic Alliance
3
rich4/zu5-apsy/zu5-apsy/zu500308/zu52457d08g
mcintyrn S⫽3 8/19/08 15:22 Art: 2006081
decades, two–person psychology has become a core notion of the psycho-
analytic relational school (Aron, 1996; Benjamin, 1996, 1997; Mitchell,
1993, 2001). The relational school holds that the therapeutic relationship
is more than a significant component of psychotherapy; it is its essence.
The relational perspective recasts the internal experience and dynamics of
patients and therapists, traditionally regarded as the primary focus of
clinical work, as secondary and determined in the context of their rela-
tionship. From the relational perspective, the therapeutic relationship is
the primary location of the therapeutic process and agent of therapeutic
change. Both the practice and study of psychotherapy are, in this view,
about understanding the vicissitudes of this relationship as it develops,
affects participants, and reflects who they are.
The parallel developments in psychotherapy research and psychoana-
lytic thinking point to the potential value of understanding the therapeutic
process in terms of relational phenomena. However, the field of psycho-
therapy research continues for the most part to regard and measure the
alliance from a subjective perspective: that is, from one perspective
(patient, therapist, or observer). The difficulty of measuring relational
phenomena is, first and foremost, methodological. Although we can ask
individuals to describe their perspectives on a given situation, there is no
simple way to capture what happens between individuals psychologically.
The study reported here is a preliminary attempt to do so, by describing
the relational aspects of the alliance in psychotherapy.
To achieve the goal of capturing the relational aspects of the alliance,
we first made a distinction between “subjective” and “intersubjective”
components of the alliance. We defined “subjective” as referring to what
happens inside individuals’ minds, what they perceive, register and/or
report of their experience. The subjective component is, as described
above, the one traditionally measured using self–report tools. We defined
“intersubjective” as that which occurs between individuals on the level of
their interaction. Our first goal was to attempt capturing the intersubjec-
tive component of the alliance.
We chose for this purpose a self–report measure of alliance, the
short–form of the WAI (Tracey & Kokotovic, 1989). We then applied the
distinction between subjective and intersubjective to this measure. The
WAI generates a number of traditional indices reflecting Bordin’s subcat-
egories of the alliance as well as a general alliance score. We used the
general alliance score derived from the WAI–12 to indicate subjective
experience. We then developed a correlation index measuring the congru-
ence between individuals in their ratings on the WAI–12. Our premise in
AMERICAN JOURNAL OF PSYCHOTHERAPY
4
rich4/zu5-apsy/zu5-apsy/zu500308/zu52457d08g
mcintyrn S⫽3 8/19/08 15:22 Art: 2006081
creating the correlation index was that by measuring the congruence
between individuals’ perceptions we can tap into what happens between
them in terms of ongoing agreement or disagreement in given areas, and
thus capture intersubjective phenomena directly. We examined the rela-
tions between these two types of measures and finally tested how well this
correlation index predicted outcome.
METHOD
PARTICIPANTS
The reported study is based on data collected from 22 cases, including
individuals (7 male, 15 female) who completed treatment in a psychother-
apy research program at a major metropolitan medical center. The mean
age of the patients involved was 42.50 (SD ⫽11.27). Eighteen patients
(82%) received DSM–IV (APA, 1994) AXIS I diagnoses, including anx-
iety, adjustment, and unipolar mood disorders. Nineteen (86%) received
AXIS II personality disorder diagnoses, including avoidant, obsessive
compulsive, dependent, and NOS. Exclusion criteria included organic
brain disorders, symptoms of psychosis, bipolar disorder, active substance
abuse, history of violent behavior, or the use of psychotropic medication
for at least one year. Treatment was provided by 13 therapists (5 male, 8
female) with a mean age of 33.50 (SD ⫽5.9). Three held Ph.D. degrees in
clinical psychology; the others were all Ph.D. candidates in the externship
or internship phases of their clinical training.
ADMISSION PROCEDURE
All patients who applied for treatment at the research program under-
went an initial phone screening. If judged appropriate individuals who met
our initial acceptance criteria, participate in an intake process, including 1)
a clinical interview, 2) completion of the Structural Clinical Interview for
DSM–IV, parts I and II (SCID–I: Spitzer et al., 1988; SCID–II: Spitzer et
al., 1987), a semi structured interview used to determine AXIS I and AXIS
II DSM–IV diagnoses, and 3) completion of several base–line / outcome
measures. Patients who did not fall within the appropriate range for brief
psychotherapy were referred for alternative treatment.
TREATMENT
The treatment provided to patients in the study involved a 30–session
protocol of brief relational therapy (BRT), (Safran & Muran, 2000) a brief
psychotherapy model informed mostly by the interpersonal and relational
schools in psychoanalysis. Brief relational therapy focuses on the relation-
ship between patient and therapist as a microcosm where the patient’s
Therapeutic Alliance
5
rich4/zu5-apsy/zu5-apsy/zu500308/zu52457d08g
mcintyrn S⫽3 8/19/08 15:22 Art: 2006081
personal and interpersonal difficulties can be explored in an emotionally
charged and immediate context. Within the therapeutic relationship, BRT
focuses on relational conflicts, or ruptures in the therapeutic alliance.
Resolving such inevitable conflicts or ruptures is critical to therapeutic
progress in BRT. Metacommunication—discussion of the therapeutic
relationship and process itself—plays an important part in the process (for
detailed description see Safran & Muran, 2000). All the therapists partic-
ipated in a weekly group–supervision session and received weekly individ-
ual supervision. Treatment adherence was routinely assessed.
ASSESSMENT MEASURES AND PROCEDURES
Therapeutic Alliance
To evaluate subjective versus intersubjective process measures of the
therapeutic alliance, we chose as our base measure the WAI–12. Through-
out the course of treatment, both patients and therapists completed the
WAI–12 after every session. The WAI–12 is a 12–item version of the
Working Alliance Inventory developed by Horvath and Greenberg (WAI:
Horvath & Greenberg, 1989; WAI–12: Tracey & Kokotovic, 1989) to
capture Bordin’s (1979) concept of the therapeutic alliance. Traditionally,
patient and therapist ratings are used separately to study the differences
between patients’ and therapist’s assessments of the alliance in relation to
outcome. Using the WAI–12 in this manner can be understood as assessing
the alliance as a subjective feeling, attitude or belief. One of our aims,
however, was to create an intersubjective measurement of the alliance. We
did this by calculating a correlation between the patient and therapist’s
individual 12 data–point ratings on the WAI–12 for each psychotherapy
session. The result was a single correlation coefficient, representing the
congruence between patient and therapist in their experience of the
alliance in that session. We believe that the congruence between patient
and therapist on their perception of the alliance can serve as an indication
for the level of agreement and actual attunement between patient and
therapist in the session and therefore represent the alliance as it occurred in
the session. It can, therefore, serve as an approximation for an intersub-
jective measure of the alliance.
Treatment Outcome
Patients and therapists were asked to complete a set of outcome
measures at several points during treatment. We assessed outcome by
measuring changes in patients’ ratings between the times of intake and
treatment termination using the following measures:
AMERICAN JOURNAL OF PSYCHOTHERAPY
6
rich4/zu5-apsy/zu5-apsy/zu500308/zu52457d08g
mcintyrn S⫽3 8/19/08 15:22 Art: 2006081
1) Global Assessment Scale ([GAS] Endicott, Spitzer, Fleiss & Cohen,
1976), a clinician–rated scale for evaluating the patient’s overall
mental health. It involves a single rating from 1 to 100, based on
clinical descriptions characterizing each 10–point interval in terms
of social and occupational functioning and levels of subjective
distress.
2) Target Complaints ([TC] Battle et al., 1966), an idiographic self–
report instrument developed to assess the patient’s presenting prob-
lems as they are perceived by both patient and therapist. The TC
consists of a patient target complaint (PTC) in which patients
describe three of the problems that brought them to treatment and
rate their degrees of severity on a Likert–type scale, and a therapist
target complaint (TTC), which consists of the therapists rating of
the severity of these problems. We calculated mean scores for the
PTC and the TTC.
3) Inventory of Interpersonal Problems ([IIP] Horowitz, Alden, Wig-
gins, & Pincus, 2000), a patient–rated inventory of interpersonal
functioning containing 64 items describing common interpersonal
problems and the experiences often associated with these problems.
Patients rate items for degree of severity on a 5–point Likert–type
scale. In this study, we used the overall mean score of the IIP.
These three outcome measures represent a wide range of criteria of
psychotherapy outcome, including both patient and therapist perspectives,
and encompass criteria covering 1) general functioning level, 2) perceived
complaints and goals for the therapy, and 3) interpersonal functioning.
Because all outcome measures showed, in our sample, improvement to a
significant degree (see Table 1) and because they were all highly intercor-
related (r’s ranging from .46 to .72), we combined them into a single
Table 1. OUTCOME MEASURES: MEANS, STANDARD DEVIATIONS AND
STATISTICAL SIGNIFICANCE
PRE– POST–TREATMENT
M (SD) M (SD) t df p
GAS 62.10 (4.79) 68.15 (6.63) ⫺4.42 19 ⬍.001
TTC 10.10 (1.12) 6.71 (2.14) 5.82 19 ⬍.001
PTC 9.56 (1.80) 6.64 (2.15) 6.27 19 ⬍.001
IIP 1.29 (.43) 1.02 (.51) 2.56 21 .02
GAS ⫽Global Assessment Scale; TTC ⫽Therapist Target Complaints (overall mean); PTC ⫽Patient
Target Complaints (overall mean); IIP ⫽Inventory of Interpersonal Problems (overall mean).
Therapeutic Alliance
T1
7
rich4/zu5-apsy/zu5-apsy/zu500308/zu52457d08g
mcintyrn S⫽3 8/19/08 15:22 Art: 2006081
outcome measure, using an equally weighted composite of standardized
residual gain scores derived from the differences between patients’ per-
formance on the measures at intake and termination.
RESULTS
PSYCHOMETRIC PROPERTIES OF SUBOUTCOME AND OUTCOME MEASURES
For the 22 patient–therapist dyads participating in the study, the
frequency of WAI–12 ratings per session varied as follows: 13—20 per
session for the patient–rated WAI–12 and 17—22 for the therapist–rated
WAI–12, with an average of per session WAI–12 ratings for both patients
and therapists at 19. (see Table 2)
Internal consistency for patients’ and therapists’ ratings on the measure
was calculated for three sessions (6, 16, 26), representing a session from the
start, middle, and end of the psychotherapies reported: Internal consis-
tency for patients’ and therapists’ ratings on the measure was calculated for
three sessions (6, 16, 26), representing the initial, middle and final phases
of the psychotherapies reported. We found this adequate for both patients’
and therapists’ WAI–12 ratings. Cronbach’s alpha coefficients obtained
(patient, therapist) are as follows: for session 6, .89 and .83; for session 16,
.83 and .86; for session 26, .85 and .86.
Test/retest reliability of the indices derived from the WAI–12 was
adequate. For both patient– and therapist–rated WAI–12, all indices
demonstrated high levels of test/retest reliability (patient WAI–12 r
yy
⫽
.88, therapist WAI–12 r
yy
⫽.89).
INTERCORRELATIONS AMONG SUBOUTCOME MEASURES
Having defined subjective and intersubjective measures of the alliance
on both subjective and intersubjective levels, our analysis concentrated on
two major tasks. First, we calculated intercorrelations among all three
measures. (Intercorrelations were calculated for three sessions (6, 16 and
26) representing the three psychotherapy thirds—see Table 3).
As evident in Table 3, all alliance indices are significantly correlated in
at least two of the three sessions. The correlation between the patient–
rated alliance (PWAI) and the therapist–rated alliance (TWAI) are pre-
dictably high, ranging from .52 to .65. The correlations between the two
subjective alliance measures and the alliance correlation index (CoWAI)
are also high, ranging from .42 to .72 for that between the patient alliance
and the alliance correlation index and .55 to .59 for that between the
therapist alliance and the alliance correlation index.” This pattern of
correlations suggests that the two subjective alliance indices and the
AMERICAN JOURNAL OF PSYCHOTHERAPY
T2
T3
8
rich4/zu5-apsy/zu5-apsy/zu500308/zu52457d08g
mcintyrn S⫽3 8/19/08 15:22 Art: 2006081
intersubjective alliance index used in this study overlap considerably and
to a significant extent, capture the same construct. However, the pattern
also suggests that each of these indices captures either a somewhat
different construct or different aspects of the same construct. These
intercorrelations support the proposition that the alliance, when repre-
sented through an index measuring the congruence between patients’ and
Table 2. FREQUENCY OF PATIENT AND THERAPIST RATINGS OF ALLIANCE
MEASURE BY SESSION.
SESSION PWAI–12 TWAI–12
119 22
219 22
317 22
420 20
520 20
621 20
720 18
819 19
920 21
10 19 20
11 19 17
12 20 18
13 19 20
14 20 17
15 19 18
16 19 18
17 20 18
18 20 19
19 20 20
20 18 19
21 17 18
22 17 19
23 16 19
24 18 19
25 20 17
26 19 18
27 16 17
28 18 19
29 13 19
30 14 18
Average 19 19
Therapeutic Alliance
9
rich4/zu5-apsy/zu5-apsy/zu500308/zu52457d08g
mcintyrn S⫽3 8/19/08 15:22 Art: 2006081
therapists’ subjective perceptions, is a construct both related and distin-
guishable from these subjective experiences.
THE RELATIONS BETWEEN SUBOUTCOME AND OUTCOME MEASURES
Having established the pattern of intercorrelation among out subout-
come indices, we proceeded to examine the relations of these measures to
psychotherapy outcome. The method used for the latter is regression
analysis based on a generalized estimating equations approach developed
by Liang and Zeger (1986). This approach was developed for analyzing
measurements obtained at multiple time points for each participant within
a group of participants. It allows for repeated measures as well as a
possible dependence between them, while requiring independence be-
tween participants. This approach offers excellent match for the structure
of the data in the study where 22 independent dyads yielded up to 30
repeated instances of the measures used. It was implemented using a
FORTRAN program titled RMGEE (Davis, 1993) based on Liang and
Zeger’s approach. Because of the specific statistical procedure used by the
RMGEE, each process variable was evaluated independently. The regres-
sion equation used included all available instances where the measure was
obtained as predictor variable. The outcome measure (one per case) served
as criterion variable. Each such analysis provided an estimate of the
predictive power of one suboutome variable over the outcome measure.
We then contrasted this estimate with those yielded from the analyses of
others to evaluate their comparative strengths. As described above, out-
come was assessed using a composite of the GAS, PTC, TTC and IIP. Our
final analysis contrasted the overall treatment relations between the two
subjective alliance indices, the alliance correlation index and a composite
outcome measure comprised of standardized, residualized gain scores on
the IIP, TTC and PTC. As indicated in Table 4, both subjective indices
Table 3. INTERCORRELATIONS AMONG SUBOUTCOME INDICES
PWAI–12 TWAI–12 CoWAI
Session 6 .52* .49*
Session 16 .65** .44
Session 26 .59* .72**
TWAI–12
Session 6 .55*
Session 16 .31
Session 26 .59**
*p⬍.05; ** p⬍.01
AMERICAN JOURNAL OF PSYCHOTHERAPY
T4
10
rich4/zu5-apsy/zu5-apsy/zu500308/zu52457d08g
mcintyrn S⫽3 8/19/08 15:22 Art: 2006081
(PWAI & TWAI) predict outcome significantly but moderately. The
alliance correlation index (Co–WAI) is significant and large, appearing to
be a considerably more powerful outcome predictor.
DISCUSSION
The goal of the study described here was twofold: to operationalize a
distinction between the subjective and intersubjective components, or
registers of the therapeutic alliances, and to examine the effect of the
alliance as captured from these two distinct registers on the outcome of
psychotherapy. We pursued our goal using an established alliance mea-
sure—the WAI–12, utilizing the two general subjective indices tradition-
ally derived from the WAI–12, and developing a correlation index mea-
suring the congruence between patients and therapists on their rating of
the alliance in each session of their mutual work in therapy. Our belief was
that the correlation index we developed would enable us to capture the
more elusive intersubjective component of the alliance.
The findings provide significant support to the distinction between the
subjective and intersubjective components of the alliance and to the
proposition that it is possible to represent and evaluate empirically these
notions. Most notable in this context is the pattern of intercorrelations
among the three alliance indices. In our findings, all indices significantly
correlated at levels ranging from .49 to .72. This pattern established
distinct, although related, constructs of subjective and intersubjective
registers for both alliance and affiliation (see Table 3). The clear and
consistent differences between the subjective and intersubjective registers
in predicting outcome (Table 4) provide this distinction with further support.
The findings regarding the predictive power of the three alliance
indices are of particular interest. The intersubjective register, as approxi-
mated by the correlation index we developed, appears to be more pow-
Table 4. OVERALL TREATMENT RELATIONS BETWEEN THREE ALLIANCE
INDICES AND THE OUTCOME MEASURE: B COEFFICIENTS, T
VALUES AND SIGNIFICANCE LEVELS FROM REGRESSION ANALYSES
USING GENERALIZED ESTIMATING EQUATIONS.
OUTCOME PWAI TWAI CoWAI
b.31 .31 .61
t2.78 2.39 2.37
p⬍.01 ⬍.05 ⬍.05
OUTCOME ⫽Mean of standardized pretreatment to post treatment residualized gain scores on the
IIP, PTC and TTC (IIP ⫽Inventory of Interpersonal Problems; PTC ⫽Patient Target Complaints;
TTC ⫽Therapist Target Complaints).
Therapeutic Alliance
11
rich4/zu5-apsy/zu5-apsy/zu500308/zu52457d08g
mcintyrn S⫽3 8/19/08 15:22 Art: 2006081
erful than the subjective indices in predicting outcome in the psychother-
apies examined in this study. A close examination of the findings reveals a
complex picture. While there is support to the notion that the subjective
vs. intersubjective indices capture two distinct registers of the alliance,
these indices are positively correlated, particularly on the patient’s side.
This raises an important question regarding the nature of the relations
between subjective experience and intersubjective congruence. We may
hypothesize that intersubjective congruence depends on positive subjective
experience, that is, that patients and therapists are more likely to think and
feel alike when they feel good about each other. We might also hypothesize
that positive subjective feelings are the result of intersubjective congru-
ence, that is, that when patients and therapists see eye to eye (even if this
congruence is not explicit), they tend to feel good about each other.
Finally, we argue that the subjective and intersubjective are two facets of
the same phenomenon and are not causally linked. Indeed, while the
findings indicate patterns of cooccurrence of certain subjective and inter-
subjective aspects of the relationships studied, they do not indicate any
particular causal link between them. What they do suggest strongly, taken
in their entirety, is that while subjective experience is highly relevant to the
process and outcome of psychotherapy, the intersubjective congruence
between patients and therapists is a more strongly and directly related to
therapeutic success.
While the findings obtained are strong, we would like to point out that
the reader might want to regard them with some caution. A major
reservation is related to the inherent difficulty in defining and operation-
alizing the highly complex concepts of “subjective” and “intersubjective.”
The strategy developed for this study centered on a conceptualization of
“subjective” as the articulated content of one’s mind and “intersubjective”
as congruence between subjects about such articulated contents. It was
further based on a premise that such congruence can be captured empir-
ically through the statistical procedure of correlation. This solution, al-
though not optimal, was necessiated by the nature of the currently
available definitions and measures of alliance. Alliance is inherently a
relational construct, but tradition defines and measures it exclusively as a
subjective one. We, therefore, developed a correlational approach that can,
in our opinion, offer an inferential approximation of intersubjective phe-
nomena. Until it is possible to measure intersubjective process more
directly, the findings obtained should be regarded as instructive, yet
tentative.
The clarity of the findings is further challenged by some of the specific
AMERICAN JOURNAL OF PSYCHOTHERAPY
12
rich4/zu5-apsy/zu5-apsy/zu500308/zu52457d08g
mcintyrn S⫽3 8/19/08 15:22 Art: 2006081
characteristics of the WAI–12. The WAI–12 is from the point of view of
subjective vs. intersubjective, a measure of a dual nature: it requires
respondents to make assessments of both subjective experience and of
their mutual relationship. Although individuals’ assessment of their rela-
tionships is still subjective, the WAI–12 does not focus solely on individ-
uals’ feelings and attitudes but rather asks respondents to make judgments
about intersubjective reality. This dual focus might in fact explain part of
the overlap between the subjective and correlation indices derived from
the WAI–12 in predicting outcome. It also limits the extent to which it can
be used to fully differentiate between the subjective and intersubjective
components of the alliance.
In addition, to the extent that the WAI–12 is a subjective measure, it is
also biased since those items involving the assessment of subjective expe-
rience ask both patient and therapist to focus on the patient. And while the
patient’s WAI–12 provides information about the patient’s subjective
experience, the therapist’s form does not provide similar information
about the therapist. Some previous research (Horvath & Greenberg, 1989;
Tracey & Kokotovic, 1989) suggests that all elements of the alliance
measured by the WAI–12 are highly correlated and can, therefore, be
justifiably used together. The findings obtained in this study cannot
support nor reject this conclusion.
Reservation should be made about the method we used to examine the
relations between the alliance and outcome measures. The specific nature
of the data used in this study—suboutcome and outcome measures,
provided by two participants in a joint process—necessitated an analysis
method that allowed for the potential of codependence between observa-
tions while assuming independence among participants. Liang and Zeger’s
(1986) generalized estimating equations approach (and the analysis pro-
gram based on it—RMGEE—Davis, 1993) suited our purposes. The
approach was developed for analyzing measurements obtained at multiple
time points for each participant within a group of participants, allowing for
repeated measures as well as a possible dependence among these measures,
while requiring independence between participants. However, this ap-
proach presented us with one major shortcoming: it renders it impossible
to judge the significance of the differences between the coefficients gen-
erated when using different predictor variables (in our case, the subout-
come measures of alliance) on the criterion variable (in our case, the
outcome measure). The reason is precisely because Liang and Zeger’s
approach is based on generalized estimating equations. In this method, the
sample size, and the degrees of freedom are an unknown estimate, and
Therapeutic Alliance
13
rich4/zu5-apsy/zu5-apsy/zu500308/zu52457d08g
mcintyrn S⫽3 8/19/08 15:22 Art: 2006081
therefore, it is not possible to calculate significance. While we claim all
coefficients generated in our analysis are significant and the findings
appear to present a picture in which the alliance correlation index is a
considerably more powerful predictor of outcome, we do not have a
significance test that we can apply to the difference to this and the power
of the other two measures.
Further research may address some limitations of the current study.
Further research should experiment with splitting the WAI–12 into two
sub–scales, one including the six subjective oriented items, the other the
intersubjective subset of items in the area of alliance measurement. The
constructs measured by such split indices would be more easily defined
and interpreted. Their varying relationships with each other, other rela-
tionship measures, and outcome could serve to clarify some of the findings
presented here, particularly regarding the distinction and respective power
of the subjective and intersubjective registers.
In addition, further investigation should focus on examining whether
the findings obtained here are repeatable in other conditions. These may
include larger samples and samples including both successfully terminated
therapies and prematurely terminated (dropouts) therapies. They may also
include samples using other treatment modalities. In emphasizing the
therapeutic relationship, BRT is a treatment method where alliance issues
and experiences may be particularly prominent. While the popularity of
the construct of alliance derives stable performance across treatment
circumstances and modalities, the results of this study should be compared
to results obtained in samples where other therapy methods were used.
This study attempted to expand both the conceptualization and means for
measuring the alliance as an intersubjective construct. However, we tested
our expanded definition and methods on only one treatment modality.
Our sample size, 22 therapy dyads, is limited in size and did not include
dropouts. For our results to have general theoretical and clinical value,
their performance should be evaluated using larger and more varied
samples as well as different treatment modalities.
Finally, the results presented here provide only a partial picture of the
suggested subjective and intersubjective aspects of the alliance and thera-
peutic relationship. Achieving further understanding in this area requires
an expansion of our research methods. In this study, we examined the
alliance and therapeutic relationship from a macro–aggregate perspective,
focusing on large–scale patterns. This type of research may provide a
general view of the factors interacting in creating the phenomenon, but it
is at a disadvantage in providing insight into how the smaller pieces come
AMERICAN JOURNAL OF PSYCHOTHERAPY
14
rich4/zu5-apsy/zu5-apsy/zu500308/zu52457d08g
mcintyrn S⫽3 8/19/08 15:22 Art: 2006081
together. The smaller pieces may count the most, particularly when
applying theoretical understanding to clinical practice. Further study in
this area should, therefore, include focusing on smaller time units, observ-
ing the unfolding therapeutic process in addition to relying on participants
retrospective self–reporting, obtaining from patients and therapists more
detailed and/or less structured reporting, and finally, integrating the
findings of research with the immense body of knowledge gathered in
traditional clinical practice and theory.
Acknowledgement: The research was supported in part by a grant from the National Institute of
Mental Health MH50246. It was presented at the annual meeting of the Society for Psychotherapy
Research, Snowbird, Utah, 1998.
The authors wish to acknowledge the contributions of all the clinical supervisors and research
assistants, as well as the therapists and patients who participated.
REFERENCES
American Psychiatric Association. (1987). Diagnostic and statistical manual of mental disorders (3
rd
ed., rev.). Washington, DC: Author.
Aron, L. (1996). A meeting of minds: Mutuality in psychoanalysis. Hillsdale, NJ: The Analytic Press.
Battle, C.C., Imber, S.D., Hoehn–Saric, R., Stone, A.R., Bash, E.R., & Frank, J.D. (1966). Target
complaints as criteria for improvement. American Journal of Psychotherapy, 20, 184–192.
Benjamin, J. (1996). Like subjects, love objects: Essays on recognition and sexual difference. New Haven,
CT: Yale University Press.
Benjamin, J. (1997). The shadow of the other: Intersubjectivity and gender in psychoanalysis. New York:
Routledge.
Bordin, E.S. (1979). The generalizability of the psychoanalytic concept of working alliance. Psycho-
therapy, 16, 252–260.
Derogatis, L.R. (1977). SCL–90–R: Administration, scoring and procedures manual I. Baltimore:
Clinical Psychometric Research.
Eaton, T.T., Abeles, N., & Gutfreund, M.J. (1988). Therapeutic alliance oucome: Impact of treatment
length and pretreatment symptomatology. Psychotherapy, 25, 536–542.
Endicott, J., Spitzer, R.L., Feliss, J.L, & Cohen, J. (1976). The global assessment scale: A procedure for
measuring overall severity of psychiatric disturbance. Achieves of General Psychiatry, 33,
766–771.
Gaston, L. (1990). The concept of alliance and its role in psychotherapy: Theoretical and empirical
considerations. Psychotherapy, 27, 143–153.
Gomez–Schwartz, B. (1978). Effective ingredients in psychotherapy: Prediction of outcome form
process variables. Journal of Counseling and Clinical psychology, 46, 1023–1035.
Greenson, R. (1967). The technique and practice of psychoanalysis. New York: International Universities
Press.
Greenson, R., (1971). The real relationship between the patient and the psychoanalyst. In M. Kanzer
(Ed.), The unconscious today (pp. 213–232). New York: International Universities Press.
Hartley, D. (1985). Research on the therapeutic alliance in psychotherapy. In American Psychiatric
Association (Ed.), Psychiatric Updates, 4, 532–549.
Hartley, D., & Strupp, H. (1983). The therapeutic alliance: Its relationship to outcome in brief
psychotherapy. In J. Masling (Ed.), Empirical studies of psychoanalytic theories, 1, 1–27.
Horvath, A.O., & Symonds, B.D. (1989). Development and validation of the Working Alliance
Inventory. Journal of Counseling Psychology, 36, 223–233.
Horvath, A.O., & Symonds, B.D. (1991). Relation between working alliance and outcome in
psychotherapy: A meta–analysis. Journal of Consulting and Clinical Psychology, 38, 139–149.
Johnson, M.E., (1988). Construct validation of the therapeutic alliance. Paper presented at the annual
meeting of the Society of Psychotherapy Research, Santa Fe, New Mexico.
Therapeutic Alliance
15
rich4/zu5-apsy/zu5-apsy/zu500308/zu52457d08g
mcintyrn S⫽3 8/19/08 15:22 Art: 2006081
Lambert, M., Shapiro, D., & Bergin, A. (1986). The effectiveness of psychotherapy. In S. Garfield &
A. Bergin (Eds.) Handbook of psychotherapy and behavior change (pp. 157–212). New York:
Wiley.
Liang, K.Y., & Zeger, S.L. (1986). Longitudinal data analysis using generalized linear models.
Biometrika, 73, 13–22.
Luborsky, L., Crits–Cristoph, P., Alexander, L., Margolis, M., & Cohen, M. (1983). Two helping
alliance methods for predicting outcome e of psychotherapy: A counting signs vs. a global
rating methods. Journal of Nervous and Mental Disease, 171, 480–491.
Luborsky, L., McLellan, T., Woody, G., O’Brian, C., & Auerbach, A. (1985). Therapist success and
its determinants. Archives of General Psychiatry, 42, 602–611.
Marmar, C.R., Horowitz, M.J., Weiss, D.S., & Marziali, E. (1986). Development in therapeutic rating
system. In L.S. Greeberg & W.F. Winsof (Eds.), The psychotherapeutic process: A research
handbook (pp. 367–390). New York: Guilford.
Marmar, C.R., Gaston, L., Gallagher, D., & Thompson, L.A. (1989). Alliance and outcome in late life
depression. Journal of Nervous and Mental Disease, 177, 464–472.
Marmar, C.R., Weiss, D.S., & Gaston, L. (1989). Towards the validation of the California Therapeutic
Alliance Rating System. Journal of the Consulting and Clinical Psychology, 1, 46–52.
Marziali, E. (1984). Three viewpoints on the threpautic alliance: Similarities, differences and associ-
ations with psychotherapy outcome. Journal of Nervous and Mental Disease, 172, 417–423.
Mitchell, S.A. (1988). Relational concepts in psychoanalysis. Cambridge, MA: Harvard University Press.
Morgan, R., Luborsky, L., Crits–Cristoph, P., Curits, H., & Salomon, J. (1982). Predicting the outcome
of psychotherapy with the Penn helping alliance rating method. Archives of General Psychiatry,
39, 397–402.
Muran, J.C., Gorman, B.S., Safran, J.D., Twining, L., Samstag, L.W., & Winston, A. (1995). Linking
in–session change to overall outcome in short–term cognitive therapy. Journal of Consulting and
Clinical Psychology, 63, 651–657.
Muran, J.C., Samstag, L.W., Jilton, R., Batchelder, S., & Winston, A. (1997). Development of a
suboutcome strategy to measure interpersonal process in psychotherapy form an observer
perspective. Journal of Clinical Psychology, 53, 405–420.
Muran, J.C., & Safran J.D. (1998). Negotiating the therapeutic alliance in brief psychotherapy. In J.D.
Safran & J.C. Muran (Eds.), The therapeutic alliance in brief psychotherapy. Washington, DC:
APA Books.
Safran, J.D., Rice, L.N., & Greenberg, L.S. (1988). Integrating psychotherapy research and practice:
Modeling the change process. Psychotherapy, 25, 1–17.
Safran, J.D., & Walner, L.K. (1991). The relative predictive validity of two therapeutic alliance
measures in cognitive therapy. Psychological Assessment, 3, 188–195.
Safran, J.D., & Muran, J.C. (1994). Towards a working alliance between research and practice. In P.F.
Talley, H.H. Strupp & S.F. Butler (Eds.), Psychotherapy research and practice: Bridging the gap
(pp. 206–226)⬎New York: Basic Books.
Safran, J.D., Muran, J.C., & Samstag, L.W. (1994). Resolving therapeutic alliance ruptures: A task
analytic investigation. In A.O. Horvath & L.S. Greenberg (Eds.), The working alliance: Theory
research and practice. New York: Wiley.
Safran, J.D., & Muran, J.C. (1996). The resolution of ruptures in the therapeutic alliance. Journal of
Consulting and Clinical Psychology, 64, 447–458.
Safran, J.D., & Muran, J.C. (2000). Negotiating the therapeutic alliance: A relational treatment manual.
New York: Guildford Press.
Sterba, R. (1934). The fate of the ego in analytic therapy. International Journal of Psychoanalysis, 15,
117–126.
Strupp, H.H. (1993). The Venderbilt psychotherapy studies: Synopsis. Journal of Consulting and
Clinical Psychology, 61, 431–433.
Tichenor, V., & Hill, C.E. (1989). A comparison of six measures of working alliance. Psychotherapy,
26, 195–199.
Zeger, S.L, & Liang, K.Y. (1986). Longitudinal data analysis for discrete and continuous outcomes.
Biometrics, 42, 121–130.
Zetzel, E. (1956). Current concepts of transference. International Journal of Psychoanalysis, 37,
369–375.
AMERICAN JOURNAL OF PSYCHOTHERAPY
16
rich4/zu5-apsy/zu5-apsy/zu500308/zu52457d08g
mcintyrn S⫽3 8/19/08 15:22 Art: 2006081
JOBNAME: AUTHOR QUERIES PAGE: 1 SESS: 3 OUTPUT: Tue Aug 19 10:43:05 2008
/rich4/zu5⫺apsy/zu5⫺apsy/zu500308/zu52457d08g
AUTHOR QUERIES
AUTHOR PLEASE ANSWER ALL QUERIES 1
