How Central Is the Alliance in Psychotherapy?
A Multilevel Longitudinal Meta-Analysis
Christoph Flu ¨ckiger
University of Bern and University of Wisconsin—Madison
A. C. Del Re
University of Wisconsin—Madison and Veterans Affairs Long
Beach Healthcare System, Long Beach, California
Bruce E. Wampold
University of Wisconsin—Madison and Research Institute,
Modum Bad Psychiatric Center, Vikersund, Norway
Dianne Symonds
Kwantlen Polytechnic University
Adam O. Horvath
Simon Fraser University
Prior meta-analyses have found a moderate but robust relationship between alliance and outcome across a broad
spectrum of treatments, presenting concerns, contexts, and measurements. However, there continues to be a lively
debate about the therapeutic role of the alliance, particularly in treatments that are tested using randomized clinical
trial (RCT) designs. The purpose of this present study was to examine whether research design, type of treatment,
or author’s allegiance variables, alone or in combination, moderate the relationship between alliance and outcome.
Multilevel longitudinal analysis was used to investigate the following moderators of the alliance– outcome corre-
lation: (a) research design (RCT or other), (b) use of disorder-specific manuals, (c) specificity of outcomes, (d)
cognitive and/or behavioral therapy (CBT) or other types of treatments, (e) researcher allegiance, and (f) time of
alliance assessment. RCT, disorder-specific manual use, specificity of primary and secondary outcomes, and CBT
did not moderate the alliance– outcome correlation. Early alliance– outcome correlations were slightly higher in
studies conducted by investigators with specific interest in alliance than were those in studies conducted by
researchers without such an allegiance. Over the course of therapy, these initial differences disappeared. Apart from
this trend, none of the variables previously proposed as potential moderators or mediators of the alliance– outcome
relation, alone or in combination, were found to have a mediating impact.
Keywords: therapeutic alliance, alliance– outcome correlations, multilevel longitudinal meta-analysis
There is a growing body of evidence that the quality of the
therapeutic alliance is linked to the success of treatment across
a broad section of clients, treatments, and identified problems.
The relation between alliance and outcome is modest, approx-
imately 7% of the variance, but this link has proven to be robust
across four meta-analyses conducted over the past 20 years
(Horvath & Bedi, 2002; Horvath, Del Re, Flu ¨ckiger, & Sy-
monds, 2011a, 2011b; Horvath & Symonds,1991; Martin, Gar-
ske, & Davis, 2000). Moreover, though the correlational link is
not very large, it is greater than the relation reported between
other treatment variables such as therapist adherence to treat-
ment manual, competence and outcome (Webb, DeRubeis, &
Barber, 2010).
The relation between alliance and outcome appears to be ubiq-
uitous across different types of treatments, but there is an ongoing
controversy about the relative importance of the alliance, particu-
larly in treatments that are tested using a randomized clinical trial
(RCT) design (e.g., Crits-Christoph, Wilson, & Hollon, 2005;
DeRubeis, Brotman, & Gibbons, 2005; Strunk, Brotman, & De-
Rubeis, 2010). The clinical aspects/implications of this debate are
important: Advocates of specific factors, for example Siev, Hup-
pert, and Chambless (2009), have argued that the alliance is
relatively unimportant in some therapies, whereas it is a specific
treatment technique in others (e.g., in relationship-based thera-
pies). On the other side of the debate, the argument is made that the
alliance is a measure of the therapists’ and clients’ mutual engage-
ment in the work of therapy regardless of the specific treatments or
problems involved and thus represents an important component for
This article was published Online First October 10, 2011.
Christoph Flu ¨ckiger, Department of Clinical Psychology and Psychother-
apy, University of Bern, Bern, Switzerland, and Department of Counseling
Psychology, University of Wisconsin—Madison; A. C. Del Re, Department of
Counseling Psychology, University of Wisconsin—Madison, and Veterans
Affairs Long Beach Healthcare System, Long Beach, California; Bruce E.
Wampold, Department of Counseling Psychology, University of Wisconsin—
Madison, and Research Institute, Modum Bad Psychiatric Center, Vikersund,
Norway; Dianne Symonds, Department of Nursing Science, Kwantlen Poly-
technic University, Surrey, British Columbia, Canada; and Adam O. Horvath,
Faculty of Education and Department of Psychology, Simon Fraser University,
Burnaby, British Columbia, Canada.
This article is based on the data set presented by Horvath, Del Re,
Flu ¨ckiger, and Symonds (2011b), Alliance in Individual Psychotherapy, in
J. C. Norcross (Ed.), Psychotherapy relationships that work: Evidence-
based responsiveness (2nd ed.). New York, NY: Oxford University Press.
This research was supported by Swiss Science National Foundation Grant
PA00P1-124102 awarded to Christoph Flu ¨ckiger.
Correspondence concerning this article should be addressed to Christoph
Flu ¨ckiger, University of Bern, Department of Clinical Psychology and
Psychotherapy, Gesellschaftsstrasse 49, CH-3012 Bern, Switzerland.
E-mail: christoph.flueckiger@psy.unibe.ch
Journal of Counseling Psychology © 2011 American Psychological Association
2012, Vol. 59, No. 1, 10 –17 0022-0167/11/$12.00 DOI: 10.1037/a0025749
10