Abstract
This study examined the role of common factors in residential cognitive therapy (RCT) and residential interpersonal
therapy (RIPT) for social phobia. Eighty social phobic patients were randomized to lo weeks of RCT or RIPT. Patients and
their individual therapists completed process and suboutcome measures weekly. The ratings were examined using linear
mixed models. Most patient-rated process variables showed U-shaped (quadratic) patterns over the course of treatment.
Therapist-rated alliance increased linearly. Therapist-rated first-week alliance and empathy predicted improvement in social
role security overthe course of therapy. The weekly fluctuations in common process predicted subsequent fluctuations in
suboutcomes in seven of 10 possible cases, whereas suboutcomes predicted process in four cases. The results support the
causal role of common factors.
ーーーーー
Psychotherapies consist of both specific and com-
mon factors, of which both, one, or neither may be
remedial. Common-factor models of psychotherapy
assume that what bothers people seeking therapy
(psychiatric symptoms and demoralization thatin-
cludes a negative view ofthe self and interpersonal
difficulties) are changed primarily or even exclusively
through common factors such as the therapeutic
alliance (Wampold, 2001). Contextual models differ
from common-factor models primarily in thatthey
assume that specific factors (e.g., cognitive restruc-
turing in cognitive therapy [CT]) are also required in
orderforthe common factors to be effective (Frank
& Frank,1991). The present study examines the role
of common factors in cognitive and interpersonal
therapy for social phobia. We have selected factors
that are emphasized in common-factor and contex-
tual models: alliance, empathy, and expectations of
treatment outcome (e.g., Frank & Frank,1991).
The therapeutic alliance can be understood as
consisting ofthree elements: agreement between the
patient and therapist on the tasks oftherapy, agree-
ment on the goals oftherapy, and the bond between
patient and therapist(Bordin,1979). The alliance is
also relevantin group therapy as the bond and
teamwork among the group members (Tschuschke
& Dies,1994). Empirically, both U-shaped (high-
low-high) quadratic and linear growth patterns
during the course oftherapy have been found
(Fitzpatrick,Iwakabe, & Stalikas, 2005). Most
studies showing a positive relationship between the
alliance and outcome (Orlinsky, Rennestad, & Will-
utzki, 2004) have measured these constructs in
overlapping time periods, precluding inferences
abouttemporal and causalrelationships between
them (Feeley, DeRubeis, & Gelfand,1999). How-
ever, using mixed-effects growth-curve analyses,
Klein et al.(2003)found thatthe early alliance
predicted subsequentimprovementin depressive
symptoms in a large sample of depressed patients.
Therapist empathy may be defined as the thera-
pist’s willingness and capacity to understand the
client’s thoughts,feelings, and concerns (Rogers,
1980). Most studies of various types of psychother-
apy have supported the hypothesis that empathy is
related to outcome (Orlinsky et al., 2004), but again
overlapping time periods in most studies preclude
inferences regarding the causal direction between
empathy and outcome. Using a structural equation
modeling approach. Burns and Nolen-Hoeksema
(1992)found results consistent with a causalinflu-
ence oftherapist empathy on outcome in cognitive-
behavioraltherapy of depression.
Frank and Frank (1991) asserted that a loss of
hope characterizes persons seeking help for emo-
tional problems and thatthe powerto install
optimism and expectations thattherapy will help
was a crucial common factor across psychotherapies.
This hypothesis has been supported in several
correlational studies (Orlinsky et al., 2004).In a
cognitive-behavioral group therapy of social phobia,
Safren, Heimberg, and Juster(1997)found that
expectations measured after a presentation ofthe
cognitive-behavioral model of social phobia in the
first session predicted outcome.
With respectto the relationship between the
common factors emphasized by Frank and Frank
(1991),therapeutic alliance has been shown to
mediate the influence of expectations on outcome
among depressed patients (Meyer et al., 2002).
In most studies of common process,the process
variables have been assessed once ortwice in early
therapy. The initiallevel of, orinitial change in,the
process variables have then been related to outcome
during the entire course oftherapy. This approach
can potentially sufferfrom ambiguities concerning
the time precedence ofthe process variable. Further-
more, a multitude of stable external, personality, and
measurementfactors may explain the covanance
between alliance and outcome. Standard regression
has typically been used to analyze the data. Un-
fortunately,this method does not separate the
predictors’relationship to the initiallevel of
the outcome measure from its influence on change
scores. Furthermore,it cannot accommodate long-
itudinal correlated data.
Moreover,the concept of process implied in this
approach is limited in scope. Therapeutic process
might be better denoted as specific events and
experiences occurring in and between therapy ses-
sions and having ratherimmediate effects on out-
come (Orlinsky et al., 2004). Forinstance, changes
in expectations during therapy may well have rapid
effects on symptoms that could be observed after
days ratherthan weeks or months. Studying the
influence of specific events on outcome usually
require statistical approaches that accommodate
longitudinal data that are correlated overtime.
To address both aspects of process, we measured
process and outcome repeatedly throughouttherapy.
We then modeled the effects ofthe initial and the
ongoing process on subsequent suboutcomes. This
allowed us to study both week-to-week changes and
the effects of early process that might extend overthe
course oftherapy. The weekly suboutcome measures
used in this study were selected to reflectthe overall
aims ofthe therapies: social anxiety reduction and
improved socialrole security. They were also con-
sistent with the types of outcomes emphasized in
contextual models:restoration of morale,including
an improved self-view and improved interpersonal
functioning, as well as the alleviation of symptoms.
The suboutcome measures were also correlated with
the overall outcome measures to ensure thatthey
were related to treatment outcome. The treatments
were delivered in a residential setting. Although this
is notthe typical venue of most psychotherapeutic
treatments,the concept of common factors in
psychotherapy would imply thattheir effects would
be presentregardless of setting.
We have reported the overall outcome results in
our sample of social phobic patients in another study
(Borge et al., 2008).In brief,the patients in both
conditions@residential cognitive therapy (RCT)
and residential interpersonal psychotherapy
(RIPT)@showed robustimprovements from pre-
to posttreatment and continued to improve in the 1-
yearfollow-up period. The number of clinically
significantly improved patients was 25 (31%) at
posttreatment and 38 (48%) atfollow-up. There
were no differences in outcome between RCT and
RIPT from pre-to posttreatment orfrom pretreat-
mentto follow-up on the overall outcome measures.
However, RCT patients improved more during
treatment on one oftwo weekly outcome measures
(see Method section). Patients rated RCT and RIPT
as equally credible.
The purpose ofthis study was to examine the
effects ofthe selected common factors in these
therapies. Specifically, we soughtto address the
following questions:
What were the change patterns ofthe common
process variables during the course oftherapy?
Were the changes in the common process
variables more influenced by one ofthe treat-
ments?
Which ofthe common process variables pre-
dicted treatment outcome from theirlevels in
early therapy?
Did the week-to-week fluctuations in the
common factors impact subsequent weekly
outcomes?
Did the common process variables have the
same effects in both treatments?
Did weekly fluctuations in the alliance mediate
the relationship between weekly fluctuations in
expectations and outcome?
Did the week-to-week fluctuations in weekly
outcome measures impact subsequent weekly
process?