Therapist Effects: Facilitative Interpersonal Skills

as a Predictor of Therapist Success

m

Timothy Anderson, Benjamin M. Ogles, and

Candace L. Patterson
Ohio University
m

Michael J. Lambert
Brigham Young University
m

David A. Vermeersch
Loma Linda University

This study examined sources of therapist effects in a sample of 25

therapists who saw 1,141 clients at a university counseling center.
Clients completed the Outcome Questionnaire-45 (OQ-45) at each
session. Therapists’ facilitative interpersonal skills (FIS) were assessed
with a performance task that measures therapists’ interpersonal skills by
rating therapist responses to video simulations of challenging client–-
therapist interactions. Therapists completed the Social Skills Inventory
(SSI) and therapist demographic data (e.g., age, theoretical orientation)
were available. To test for the presence of therapist effects and to
examine the source(s) of these effects, data were analyzed with
multilevel modeling. Of demographic predictor variables, only age
accounted for therapist effects. The analysis with age, FIS, and SSI as
predictors indicated that only FIS accounted for variance in outcomes
suggesting that a portion of the variance in outcome between therapists
is due to their ability to handle interpersonally challenging encounters with
clients. & 2009 Wiley Periodicals, Inc. J Clin Psychol 65: 755–768, 2009.

Keywords: therapist effects; therapy outcome; interpersonal skills;

effectiveness

Numerous studies demonstrate that therapist characteristics are a unique predictor of

therapy outcome (e.g., Crits-Christoph & Mintz, 1991; Dinger, Strack, Leichsenring,

Correspondence concerning this article should be addressed to: Timothy Anderson, Department of
Psychology, Ohio University, Athens, OH 45701; e-mail: andersot@ohio.edu

JOURNAL OF CLINICAL PSYCHOLOGY, Vol. 65(7), 755--768 (2009) & 2009 Wiley Periodicals, Inc.
Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/jclp.20583
756 Journal of Clinical Psychology, July 2009

Wilmers, & Schauenbirg, 2008; Luborsky, McClellan, Woody, O’Brien, & Auerbach,
1985; Okiishi, Lambert, Nielsen, & Ogles, 2003; Wampold & Bolt, 2006). Frequently
this finding has emerged in the context of clinical trials where the researchers
implemented interventions designed to eliminate nuisance variables (e.g., therapist
effects or variability) through the careful selection of therapists and rigorous training
and supervision in the provision of a manualized treatment (Wampold, 2001). For
example, Luborsky et al. (1985) found significant differences among therapists on
client outcomes despite strict adherence to manuals. Similarly, Shapiro, Firth-Cozens,
and Stiles (1989) found that one therapist in the Sheffield II psychotherapy study had
significantly better outcomes than other therapists in the study. Finally, Crits-
Christoph and Mintz (1991) found therapist effects in a meta-analysis of 10 clinical
trials. Taken together, these findings indicate that even in highly controlled
interventions where researchers attempt to suppress or control the effects of
individual therapists, client outcomes vary by participating therapists.
Advances in the application of statistical analyses (i.e., multilevel or hierarchical
modeling) have yielded mixed results about the effect of individual therapists on client
outcomes. For example, two recent re-analyses of the National Institute of Mental
Health (NIMH) Treatment of Depression Collaborative Research Program (TDCRP)
resulted in mixed findings with regards to therapist effects. Wampold and Bolt (2006)
found significant therapist effects using hierarchical linear analysis, whereas Elkin,
Falconnier, Martinovich, and Mahoney (2006) did not find therapist effects using the
same data and a similar data analytic approach. Estimation of the model and the
different treatment of outliers appear to be partly responsible for the emergence of
these differences (Elkin, Falconnier, & Martinovich, 2007; Wampold & Bolt, 2007).
Regardless, the majority of therapist effects research with clinical trials data supports
the hypothesis that individual therapists differentially impact client outcomes.
Naturalistic studies of therapist effects using hierarchical linear modeling are
relatively rare and thus little is known about the existence or quality of therapist
effects in real-world practice settings. Okiishi, Lambert, Nielsen, and Ogles (2003)
found significant variability among therapists for the outcomes of 1,841 clients seen in
a university counseling center. In further analyses on a larger version of this dataset,
Okiishi et al. (2006) found that therapist effects continued to account for variance in
client outcomes. However, attempts to identify the source of these therapist effects
were not informative. Results indicate that therapist effects were not attributable to a
variety of therapist traits, such as sex, type of training, or theoretical orientation.
Thus, the source of therapist effects in these naturalistic studies remains unknown.
Although therapist effects fairly consistently account for variance in clinical
outcomes, to a large extent these effects remain a neglected variable in
psychotherapy research (e.g., Garfield, 1997) and little is known about the sources
that underlie these effects. Generally, demographic characteristics such as therapist
age, sex, and ethnicity have failed to emerge as predictors of outcome (Beutler et al.,
2004). Conversely, research demonstrates that variables such as therapist emotional
adjustment and certain aspects of therapist personality (e.g., dominance) predict
outcome with moderate effects (e.g., Beutler et al., 2004). These empirical findings
highlight the need for studies that move beyond measuring therapists’ demographic
characteristics and general traits to include measures of therapist characteristics that
have a more solid theoretical and empirical link to client outcomes. In other words,
there is a need for therapist effects research that includes the measurement of
constructs that are grounded in findings from psychotherapy research. For example,
numerous studies support the presence of relationships between psychotherapy
Journal of Clinical Psychology DOI: 10.1002/jclp
 Therapist Effects 757

process variables, such as empathy and the alliance, with clinical outcomes
(Norcross, 2002; Wampold, 2001). Therefore, a potentially promising approach
for identifying the sources of therapist effects would be to operationalize and
examine therapists’ skills in facilitating therapy processes (e.g., empathy and the
alliance) that are theoretically and empirically related to therapy outcome.
In theory, indicators of the therapist’s contribution to therapy processes can be
identified through somewhat subtle interpersonal messages (Strupp & Anderson,
1997). For example, cases in which therapists communicated subtle interpersonally
disaffiliative messages were shown to result in worse clinical outcomes than when
there was an absence of such communication (Henry, Schacht, & Strupp, 1990).
Another illustration involves the therapeutic alliance, which has commonly been
viewed as partly a therapist’s skill in facilitating a collaborative relationship with his
or her client (e.g., Safran & Muran, 2000; Norcross, 2002). Another process variable,
empathy, refers to the therapist’s skill in accurately understanding and reflecting a
client’s thoughts and emotional experience (Bohart, Elliott, Greenberg, & Watson,
2002). Because the therapist is by definition a major contributor to the facilitation of
the processes of therapy, we operationalized and examined the relationship between
therapists’ facilitative interpersonal skills (FIS; Anderson, Ogles, & Weis, 1999;
Anderson, Patterson, & Weis, 2007) and clinical outcomes in this investigation of
clients and therapists in a naturalistic setting.
The present study aims to further understanding about the sources of therapist
effects. We extended previous research (i.e., Okiishi et al., 2006; Okiishi et al., 2003)
by drawing from the same client population as these studies, but examining a
broader spectrum of therapist characteristics, which included therapists’ demo-
graphic characteristics (i.e., age, level of training, and gender) and theoretical
orientation. However, our main hypotheses centered around interpersonal skills,
which we theorized to be better reflections of the therapist qualities that would likely
influence therapeutic processes, such as the therapeutic alliance, and which
heretofore have not received significant attention as a possible explanation for
therapist effects (see above). Specifically, these variables included a self-report
measure of social skills and a performance-based measure of FIS (Anderson et al.,
1999, 2007). We examined whether FIS predicts outcome in this sample because the
sample used by Okiishi and colleagues, drawn from the same pool, found that
therapist effects accounted for variability in client outcomes but traditionally studied
therapist characteristics (e.g., demographic variables and general traits) did not
explain the therapist effects identified in their sample.
We invited therapists who participated in Okiishi and colleagues’ research to
complete a self-report social skills measure and a performance task that consisted of
responding to video-recorded analogue therapy segments. Ratings of therapists’
responses to the performance task constituted the measure of FIS. We predicted that
individual therapists would account for a significant amount of variability in
outcome and we predicted that therapists’ interpersonal skills (i.e., FIS) would
account for a significant amount of variance in client outcomes in this large sample
of clients and therapists in a naturalistic setting.

Method
Participants
Clients. An archival database of clients seen in a university counseling center was
available for this study. Clients were included in the analyses if they completed at
Journal of Clinical Psychology DOI: 10.1002/jclp
758 Journal of Clinical Psychology, July 2009

least three therapy sessions and their therapist agreed to complete the therapist
measures. Clients in the analyses were seen prior to the therapist assessment; 1,141
clients were included in the final sample.
Clients had a mean age of 23.0 years (SD 5 4.1 years; range 5 18 to 56 years).
The majority of the clients were women (62.8%) and most of the clients were
Caucasian (85.5%). All of the clients attended or were associated or employed with
the university where the counseling center was located. Clients in the sample
attended therapy for a range of 3 to 72 sessions, with a mean attendance of 9.09
sessions (SD 5 8.79).
Therapists. Thirty-two therapists working in the counseling center at a large
private university were invited to participate in the study. Twenty-eight therapists
agreed to participate in the study and completed self-report measures and an FIS
performance task. One of these therapists was excluded because of incomplete FIS
performance assessment data. Two additional therapists were excluded because data
for fewer than 10 clients were available in the archival dataset. Thus, 25 therapists
(16 men; 9 women) were initially included in the analyses. Therapists treated a mean
of 45.6 clients (ranging from 13 to 141 clients) for whom data were available in the
archival dataset. Therapists had a mean age of 43.9 years (SD 5 10.9 years) and were
predominantly Caucasian (96%). Therapists self-identified their theoretical
orientation as primarily cognitive–behavioral (CBT: n 5 8), humanistic (n 5 8),
eclectic (n 5 5), and psychodynamic (n 5 4).
With regards to level of training, the 25 therapists who completed the performance
assessment included 17 licensed doctoral level therapists; 2 postdoctoral but not fully
licensed therapists; 3 predoctoral interns, and 3 graduate trainees. Therapists had
11.5 mean years (SD 5 10.1) of clinical experience. Therapists estimated that they
spent 42.9% of their professional hours in direct clinical practice (ranging from 8%
to 70%).

Client Measures
Outcome Questionnaire-45. The Outcome Questionnaire-45 (OQ-45; Lambert
et al., 2004) is a 45-item self-report general symptom measure with a total score that
is the sum of three subscales that measure the subjective, interpersonal, and social
role components of global psychological distress. Response options range from 0
(never) to 4 (almost always). This measure has good internal consistency (alphas
range from .70 to .93) and test-retest reliability over 3 weeks of .84 (Ogles, 1996). The
OQ was administered routinely at the beginning of each therapy session at this
counseling center. A cutoff score of 64 or above is used to indicate a significant
likelihood that the participant’s score belongs to a clinical sample and reliable
change is set at 14 OQ points (Lambert et al., 2004).

Therapist Measures
Social Skills Inventory. The Social Skills Inventory (SSI; Riggio, 1986) is a
90-item self-report questionnaire that assesses self-reported social skills. Items are
scored using a 5-point Likert scale, with response options ranging from 1 5 not at all
like me to 5 5 exactly like me. The SSI measures skills in expressivity, sensitivity, and
control in verbal (social) and nonverbal (emotional) domains. The SSI yields a global
score and six subscales with 15 items each, though subscale scores are highly
correlated with each other and the total score. Thus, for the sake of parsimony, only
Journal of Clinical Psychology DOI: 10.1002/jclp
 Therapist Effects 759

the global score was used in this study. Coefficient alphas range from .62 to .87 for
each of the subscales and test-retest correlations range from .81 to .96 for a 2-week
interval (Riggio, 1989). In the present study, the SSI was administered to therapists
after they completed treatment with the clients in this study.
Facilitative Interpersonal Skills (FIS) Performance Task. The Facilitative
Interpersonal Skills (FIS) Performance Task (Anderson et al., 2007) is designed to
elicit responses that are indicants of a person’s ability to perceive, understand, and
communicate a wide range of interpersonal messages, as well as a person’s ability to
persuade others with personal problems to apply suggested solutions to their
problems and abandon maladaptive patterns. A performance task was designed as a
means of measuring therapists’ abilities to respond to challenging interpersonal
situations in a therapy setting. The development and use of a performance task, as
opposed to self-report measures of FIS, is advantageous in that this task has a high
level of ecological face validity.
Four problematic therapy process segments were selected from the videotaped
archives of a study that focused on problematic interpersonal interactions between
patients and therapists (Strupp, 1993). In addition, unique interpersonal patient
styles were selected to represent a range of interpersonal patterns, including (a) a
confrontational and angry patient (‘‘You can’t help me’’); (b) a passive, silent, and
withdrawn patient (‘‘I don’t know what to talk about’’); (c) a confused and yielding
patient (only the therapist’s opinion matters); and (d) a controlling and blaming
patient (implies that others, including the therapist, are not worthy of him). Thus,
two cases were designed to include patients who were highly self-focused, negative,
and self-effacing, and the remaining two cases were designed to be highly other-
focused, friendly, but highly dependent clients. Two brief segments (approximately
2 minutes each) were selected for each problematic patient–therapist interaction, and
hence the final performance task consists of eight of these brief segments. Actors
were hired to re-enact the eight scenarios. These actors memorized the transcripts
from the sessions and were coached on how to capture the interpersonal style of the
patients they were enacting. Using therapy transcripts, actors re-enacted the
scenarios and practiced for multiple sessions before the re-enactments were video-
recorded.
Therapists from the present study were presented with these eight brief situations
and were prompted to respond to the patient–actors (who were filmed directly facing
the camera) at predefined moments ‘‘as if’’ they were the therapist in the situation.
The video clips were presented via a computer program that allowed therapists to
make their responses in the privacy of their offices at the counseling center.
Therapists were asked to leave the audio recorder running to capture the therapists’
initial, nonpracticed responses to the video scenarios.
FIS ratings and scoring. FIS item content was selected from the clinical and
research literature (e.g., Norcross, 2002) on common therapist interpersonal skills
and facilitative conditions. Specifically, the 10 FIS items included ratings of verbal
fluency, emotional expression, persuasiveness, hopefulness, warmth, empathy,
alliance-bond capacity, and problem focus. Each item was scored using a 5-point
Likert type scale, where ratings of 1 or 2 represent deficiencies in the skill, 3 indicates
a neutral level, and ratings of 4 and 5 denote proficiencies of the skill being rated.
Two licensed PhD research clinicians rated each of the eight recorded responses
for each therapist. Raters were provided with a manual for rating the FIS items.
After studying the manual, the two raters (one of which developed the manual) met
Journal of Clinical Psychology DOI: 10.1002/jclp
760 Journal of Clinical Psychology, July 2009

for 2 days to discuss and practice ratings with sample responses, none of which
included therapist responses from the present study. Then from two separate
locations, the PhD research clinicians rated all participating therapists’ responses to
each of the 10 items. The mean score for the two raters on each item were then
summed to obtain one FIS performance rating for each therapist. Hence, possible
scores ranged from 5 to 50 and the mean total FIS rating was 29.8 (SD 5 4.52) and
ranged from 19.8 to 36.8. Replicating previous research involving the FIS rating
system (e.g., Anderson, Crowley, & Carson, 2001), each of the FIS items had
acceptable interrater reliabilities (all were r4.70).

FIS in previous research. In a prior research study (Anderson et al., 2001),

therapists’ FIS served as an independent variable. Doctoral students from a range of
graduate programs (e.g., psychology, biology, history) were the therapists in this
prior study on FIS. These trained (psychology doctoral students) and untrained
(doctoral students in disciplines that do not relate to clinical therapeutic work)
therapists were selected based on their performance on the FIS task and based on
their social skills, empathy, and sociability. This screening process allowed for a
group of high-FIS therapists and a group of low-FIS therapists to participate in the
study. Therapists met with clients for seven sessions. Results indicate that clients of
high-FIS therapists had significantly better outcomes (OQ-45; Lambert et al., 2004)
relative to clients seen by low-FIS therapists. This difference was evident at both
termination and at 3 months posttreatment.

Procedures
Outcome data were collected at a university counseling center and portions of the
outcome data presented in this sample have been published in other studies (e.g.,
Okiishi et al., 2003, 2006). However, the assessment of therapists’ FIS occurred
specifically for this study. Clients receiving services at the center were instructed to
complete the OQ-45 prior to each session. Data collected at the same time as the
OQ-45 included the client identification number, date of session, and the client’s
therapist.
Therapists were assessed with the self-report FIS and the FIS performance task,
both of which were collected in the privacy of an onsite research office (the
researcher was not present). To protect therapist identities, a nontherapist research
assistant used randomly assigned therapist identification numbers so that identifying
individual therapists would not be possible when viewing the data set. After
therapists completed their assessment, they placed their questionnaires and tape-
recorded responses to the performance analysis in an envelope, which they were
instructed to seal and sign before returning the envelope to the onsite researcher.
Client OQ session data for this study occurred over a 45-month period. All
therapist assessments were collected over a 4-month period, beginning in the 40th
month and ending in the 44th month of client OQ data collection (most being
collected in the 40th month). More specifically, 76.6% of OQ session data were
collected prior to the date of the assessment, 23.1% of OQ data were collected after
the assessment, and 0.3% occurred on the therapist-reported date of the assessment.
The majority of clients (63.1%; SD 5 28.2) had completed all of their sessions by the
time the therapist completed their individualized FIS assessment. At the extremes,
one therapist had not completed therapy with any clients at the center while four
therapists had terminated all clients prior to the FIS assessment.
Journal of Clinical Psychology DOI: 10.1002/jclp
 Therapist Effects 761

Once therapist assessments were completed, all therapist data were sent to a
second university site. The two raters of the therapist performance task were also
located at different university sites, neither of which was where the client data was
collected. Thus, raters were blind to the identities of the therapists being rated and
client outcome data was forwarded only after completion of the therapist ratings.

Results
Plan of Analysis
Data were analyzed using hierarchical linear modeling (HLM). Hierarchical linear
modeling has a number of advantages over other multivariate methods of analysis
(Bryk & Raudenbush, 1992; Singer, 1998) and is especially useful for naturalistic
data in which there are variable lengths of data collection and missing data. An
initial HLM was examined using an unconditional model with sessions nested within
clients and clients nested within therapists. Although the first session intake was
sometimes conducted by a different therapist, this intake was attributed to the
treating therapist. After examining the unconditional model, a second analysis was
conducted taking into account therapist variables (level of training, primary
theoretical orientation, type of training, and gender). This analysis focused on the
perennial question of therapist variables that may influence outcome (Beutler et al.,
2004). A final analysis examined the influence of therapist facilitative interpersonal
skills on client outcome. These analyses were followed with further exploration of
differences in outcome among therapists and their relationship to therapist
characteristics.

Unconditional Model
An examination of the unconditional model with sessions nested within clients and
clients nested within therapists indicated that on average clients started treatment
with an OQ-45 score of 68.24 (SE 5 0.68) and that the average slope across
therapists was approximately one OQ-45 point of improvement per session,
avg. 5 .96 (SE 5 0.10). There were no significant differences among therapists in
terms of initial OQ-45 scores (intercepts) prior to treatment onset, p4.05. However,
there were differences among therapists in terms of average outcomes (slope) across
clients, w2(24) 5 42.80, p 5 .011. Estimated intercepts and slopes for the 25
participating therapists are displayed in Table 1. As can be seen, one therapist
(therapist I) was a significant outlier with a slope across 50 clients that was
significantly lower than all other therapists. Because this case happened to be such a
substantial outlier, further analyses were conducted without this therapist. An
additional qualitative assessment of this unique therapist is included in the
discussion.

Traditional Therapist Variables

The next HLM analysis of therapist data was conducted to see if any of the four
traditional therapist variables (age, sex, theoretical orientation, and percentage of
work time conducting therapy) might account for differences in outcomes
among therapists. The results of this analysis indicated that the sex, theoretical
orientation (CBT, humanistic, eclectic, dynamic), and percentage of time conducting
treatment did not significantly account for variation in outcomes among therapists
Journal of Clinical Psychology DOI: 10.1002/jclp
762 Journal of Clinical Psychology, July 2009

Table 1
Therapist Hierarchical Linear Modeling (HLM) Estimated Intercepts and Slopes Across Clients
Therapist Estimated intercept Estimated slope (Empirical Bayes)

A 68.18 1.29
B 68.21 1.15
C 68.22 1.08
D 68.20 1.19
E 68.24 0.95
F 68.23 1.03
G 68.18 1.29
H 68.26 0.85
I 68.37 0.27
J 68.23 1.03
K 68.25 0.93
L 68.19 1.25
M 68.27 0.82
N 68.21 1.12
O 68.24 1.00
P 68.29 0.67
Q 68.25 0.92
R 68.27 0.83
S 68.27 0.82
T 68.22 1.09
U 68.27 0.83
V 68.28 0.75
W 68.22 1.10
X 68.24 0.97
Y 68.27 0.79

Table 2
Hierarchical Linear Modeling (HLM) With Age, Sex, Percentage of Work Time Conducting
Treatment, and Theoretical Orientation as Predictors of Outcome (Slope)
Predictor Coefficient SE T-Ratio

Age 0.02 0.01 2.77

Sex 0.06 0.13 0.46
Percentage of time conducting treatment 0.00 0.00 0.74
Theoretical orientation 0.01 0.08 0.17
po.05.

(see Table 2). In contrast, the age of the therapist did account for differences in
outcome, with older therapists having better outcomes than younger therapists.

FIS as a Predictor
To examine the variable of interest, FIS, a third HLM analysis was conducted with
age, self-report of social skills (SSI), and the FIS performance assessment ratings as
predictors of outcome (see Table 3). Age was included because it was a significant
predictor in the first stage of the analysis. As can be seen in Table 3, FIS was the lone
predictor of variation in outcome slope among therapists. In the context of SSI and
FIS, age was no longer a significant predictor of therapist slope.
Journal of Clinical Psychology DOI: 10.1002/jclp
 Therapist Effects 763

Table 3
Hierarchical Linear Modeling (HLM) With Age, SSI, and FIS as Predictors of Outcome
(Slope)
Predictor Coefficient SE T-Ratio

Age 0.01 0.01 1.70

SSI 0.00 0.00 0.66
FIS 0.03 0.01 2.16

Note: SSI 5 Social Skills Inventory; FIS 5 facilitative interpersonal skills. po.05.

Figure 1. Scatterplot of facilitative interpersonal skills (FIS) by outcome slope.

Additional Analyses
To further explore the relationship between therapist characteristics and client
outcome in this naturalistic setting, we examined the scatter plot of, and calculated
the simple correlation between, the FIS total score for each therapist and the HLM
estimated outcome slope for each therapist’s caseload of clients. Although this
analysis is somewhat counterintuitive to the HLM design in that the hierarchical
nature of the data is lost when aggregating at the therapist level, it does allow for
another perspective on the relationship between slopes and FIS. As can be seen in
Figure 1, therapists with higher facilitative interpersonal skills had clients with
greater change rates (slopes). The simple correlation between therapist FIS sum and
outcome slope was significant, r 5 .47 (n 5 24).
Because age initially emerged as a significant predictor of outcome, only to be
displaced by FIS, we were interested in other associations with age in our dataset.
Most notable was that age and FIS were significantly associated, r(27) 5 .45, po.02.
In addition, because the OQ data were collected over a prolonged period, it seemed
possible that older therapists may have had a disproportionately large number of
cases completed prior to the assessment (e.g., OQ scores could have been influenced
Journal of Clinical Psychology DOI: 10.1002/jclp
764 Journal of Clinical Psychology, July 2009

by the FIS assessment). Thus, we explored whether older therapists may have had a
greater number of completed cases prior to their FIS assessments. Therapist mean
percentage of cases completed prior to the FIS assessment was not significantly
correlated with demographic variables, including therapist age, r(27) 5 .24, ns. Also,
the percentage of completed cases prior to the therapist assessment was not
associated with the mean estimated slope of OQ scores from the above HLM
analysis, r(27) 5 .11, ns. Although not all therapists estimated their hours of clinical
experience, there was, not surprisingly, a strong and significant relationships with
age for those who did, r(21) 5 .81, po.001.

Discussion
A substantial body of research suggests that the variance in therapy outcomes is
explained in part by individual therapists. In a meta-analysis of therapist effects,
Crits-Christoph and Mintz (1991) found that approximately 9% of the variance in
therapy outcomes was due to differences among therapists. However, little is known
about the specific characteristics that underlie these therapist effects. The current
study sought to understand not only the impact of traditional therapist variables
(e.g., age, theoretical orientation, sex of the therapist) on clinical outcomes, but also
to understand the role of therapists’ facilitative interpersonal skills (FIS) on
outcome. To investigate these characteristics, HLM analyses were conducted on a
subset of the sample from Okiishi and colleagues’ (2003, 2006) studies. Therapists in
the current study’s sample completed self-report measures and a performance task
designed to measure one’s level of facilitative interpersonal skills.
As predicted, there were considerable differences among therapists with regards to
average outcomes across clients. In other words, therapist effects were present in the
current study. To investigate the potential sources for these effects, the traditional
therapist variables of therapist age, therapist sex, percentage of work time
conducting treatment, and therapist theoretical orientation were analyzed with
HLM analysis. Of these four variables, only the age of the therapist accounted for
differences in therapy outcome, specifically, that older therapists produced superior
outcomes. This was an unexpected finding because most of the previous literature on
therapist characteristics has not found age to be a predictor of outcome. (e.g.,
Beutler et al., 2004). However, when the therapist’s age, self-reported social skills,
and FIS were examined with HLM analysis, age no longer predicted outcome.
It seems reasonable to infer that age serves as an indicator of the accumulation of
clinical experience needed to master the interpersonal qualities inherent in FIS. We
believe that the collection of interpersonal qualities needed to optimize the
therapeutic relationship, as measured in the current study through the use of
vignettes that are therapeutically challenging, likely requires considerable effort and
more practice than typically afforded in standard clinical training (e.g., see Strupp &
Anderson, 1997). Indeed, our study found that age and FIS were significantly
correlated. Even though age no longer predicted outcome when accounting for FIS
in our hierarchical analysis, we suspect that the relationship of experience, practice,
and life experience deserve future study as significant aspects of constructs such as
FIS, which attempt to test the limits of therapists’ interpersonal abilities. An
important finding from this study is that FIS emerged as a significant predictor of
outcome (slope). This finding indicates that the therapist characteristic of facilitative
interpersonal skills (e.g., emotional expression and persuasiveness) represents a
quality of the therapist that impacts clinical outcomes.
Journal of Clinical Psychology DOI: 10.1002/jclp
 Therapist Effects 765

Correlational analyses were conducted in an effort to better understand the

relationship between FIS and outcome. These analyses further illuminated the
importance of FIS in terms of clients’ outcomes. Results indicated that larger rates of
improvement were reported by clients whose therapists had higher levels of
facilitative interpersonal skills, as compared to clients of therapists with lower levels
of facilitative interpersonal skills. The finding that therapists’ interpersonal skills
predicted outcome is congruent with the literature surrounding empirically
supported relationships (Norcross, 2002). In essence, this study sheds light on one
of the factors that underlie therapist effects and the findings from this study support
the notion that the person of the therapist is a key ingredient of psychotherapy.
Although the finding regarding therapists’ facilitative interpersonal skills is the most
striking result from this study, analyses yielded additional interesting results.
The unconditional HLM analysis of the entire sample indicated that therapists’
clients did not differ substantially from one another in terms of their initial symptom
severity (measured by the OQ-45 prior to intake). One strength of this study is that
clients completed the OQ-45 at each session, so that session-by-session change was
available for analyses. Across therapists, clients in the current study improved an
average of approximately 1-point per session. This rate of change indicates that 14
sessions were needed for clients, as a group, to make reliable change. Thus, clients in
the naturalistic setting had change rates that were consistent with findings from
Lambert, Hansen, and Finch’s (2001) study on the number of sessions needed for
clients to make reliable change in therapy.
One noteworthy finding relates to a unique therapist that was treated as an outlier
in the current study. The slope across clients for this therapist revealed a much lower
rate of change in client symptoms. Of the total 25 therapists in the present study, this
one distinct therapist represents 4% of the sample of therapists. It could be said that
this therapist is similar to therapists identified in other studies who have significantly
poorer outcomes (Bergin & Suinn, 1975). This finding has important implications for
future studies. Not only should researchers examine therapist effects in terms of
therapist factors that enhance outcome, such as facilitative interpersonal skills, but
also researchers must examine therapist factors that lead to or are related to client
deterioration. Both sets of factors have the potential to lead to methods for
improving the effectiveness/efficacy of psychotherapy.

Limitations
This study contains several limitations that must be addressed. The primary
limitations of this study include the following: (a) selection and timing of therapist
assessments; (b) outcome was assessed with a single measure, as opposed to multiple
measures; and (3) limitations typically associated with naturalistic studies. The
details and implications of these limitations are described below.
As described earlier, this study’s sample is composed of a subset of clients and
therapists from the sample of therapists and clients that participated in Okiishi and
colleagues’ (2006, 2003) studies. Therapists from these previous studies were invited
to participate in the current study and their participation required that they complete
self-report measures and a performance assessment of facilitative interpersonal skills.
To reduce possible biases, two independent raters that were located at two separate
universities from the data collection site completed ratings of the FIS performance
task. Although this rating procedure reflects a strength of this research, it is unclear
whether the volunteer selection process affected the results of this study. In addition,
Journal of Clinical Psychology DOI: 10.1002/jclp
766 Journal of Clinical Psychology, July 2009

therapists completed the self-report measures and the FIS performance task after
their clients terminated from therapy. It seems likely that social and facilitative
interpersonal skills represent stable constructs; thus, the timing of measurement may
not have impacted the study’s results. However, because the stability of these
constructs has not been formally researched, we cannot make definitive conclusions
about the effects that the timing of assessment had on the findings from this
investigation.
The second limitation of this study is the fact that clinical outcomes were
measured by only one self-report instrument (OQ-45). Unfortunately, no single
outcome measure can depict a complete picture of a client’s gains or declines. To
fully capture therapy outcomes, a wide range of measures with multiple target areas
and sources are necessary. Nevertheless, the OQ-45 is an outcome instrument that is
highly correlated with a variety of other self-report instruments typically used in
outcome research and with psychometric properties that allowed for repeated
assessments of outcome over the course of numerous sessions.
Lastly, inherent in the present research is the set of limitations found in any study
of psychotherapy (i.e., an imbalance of internal and external validity). The lack of
control in naturalistic studies leads to decreased internal validity. For example, the
current study did not have a control condition or random assignment. However, the
strength of this study lies in its external validity. Because this study did not have
stringent criteria for inclusion/exclusion of clients and the setting was not highly
controlled, its findings may be generalized to other clinical service-oriented settings
that treat similar clinical samples.

Conclusion
There is variability in psychotherapy outcome that can be linked to individual
therapists and it appears that a portion of this variability can be linked specifically to
therapists’ interpersonal skills rather than to other variables. Because therapists’
facilitative interpersonal skills explain a significant proportion of therapist variability
in client outcomes, future research in this area has the potential to improve both
clinical trials research and evidence-based practice in naturalistic settings. By
selecting therapists who have equivalent FIS scores, clinical trials researchers may be
better able to control therapist variability and thus, be better able to determine the
effects of the specific techniques or treatments being studied. Further examination
and consideration of therapists’ FIS have the potential to enhance client outcomes in
real-world practice settings, such as the counseling center where our data was
collected. The FIS construct is broad-based and is more similar to general therapist
competencies than to specific techniques. Our findings suggest that emphasis should
be placed on facilitative interpersonal skills similar to those examined in this study in
the selection and training of therapists. In addition, from the point of view of the
consumers of services it would be an advantage to know beforehand which therapists
are high in such skills. Future research will help clinicians understand the situations
in which the therapists’ interpersonal skills influence the progress of therapy and
whether interpersonal skills reflect a stable or dynamic characteristic that may be
influenced by therapeutic process and interactions with different clients. Addition-
ally, further research in this area would help clinical supervisors understand and
identify the relative importance of competencies for therapist training. The current
study reaffirms the notion that many psychotherapies, as routinely practiced, are
Journal of Clinical Psychology DOI: 10.1002/jclp
 Therapist Effects 767

evidence-based when delivered by therapists who can offer high levels of

interpersonal skills on a performance-based measure.

References
Anderson, T., Crowley, M.E., & Carson, K.L. (2001, August). Therapist interpersonal skills
(but not training) influences outcome and alliance. Paper presented at the 109th annual
convention of the American Psychological Association, San Francisco, CA.
Anderson, T., Ogles, B.M., & Weis, A. (1999). Creative use of interpersonal skills in building a
therapeutic alliance. Journal of Constructivist Psychology, 12, 313–330.
Anderson, T., Patterson, C.L., & Weis, A.C. (2007). Facilitative Interpersonal Skills
Performance Analysis Rating Method. Unpublished coding manual, Department of
Psychology, Ohio University, Athens, OH.
Bergin, A.E., & Suinn, R.M. (1975). Individual psychotherapy and behavior therapy. Annual
Review of Psychology, 26, 509–556.
Beutler, L.E., Malik, M., Alimohamed, S., Harwood, T.M., Talebi, H., Noble, S., et al. (2004).
Therapist variables. In M.J. Lambert (Ed.), Bergin and Garfield’s handbook of
psychotherapy and behavior change (pp. 227–306). New York: Wiley.
Bohart, A.C., Elliott, R., Greenberg, L.S., & Watson, J.C. (2002). Empathy. In J.C. Norcross
(Ed.), Psychotherapy relationships that work (pp. 89–108). Oxford: Oxford University
Press.
Bryk, A.S., & Raudenbush, S.W. (1992). Hierarchical linear models: Applications and data
analysis methods. Thousand Oaks, CA: Sage.
Crits-Christoph, P., & Mintz, J. (1991). Implications of therapist effects for the design and
analysis of comparative studies of psychotherapies. Journal of Consulting and Clinical
Psychology, 59, 20–26.
Dinger, U., Strack, M., Leichsenring, F., Wilmers, F., & Schauenbirg, H. (2008). Therapist
effects on outcome and alliance in inpatient psychotherapy. Journal of Clinical Psychology,
64(3), 344–354.
Elkin, I., Falconnier, L., & Martinovich, Z. (2007). Misrepresentations in Wampold and Bolt’s
critique of Elkin, Falconnier, Martinovich, and Mahoney’s study of therapist effects.
Psychotherapy Research, 17(2), 253–256.
Elkin, I., Falconnier, L., Martinovich, Z., & Mahoney, C. (2006). Therapist effects in the
National Institute of Mental Health Treatment of Depression Collaborative Research
Program. Psychotherapy Research, 16, 144–160.
Garfield, S.L. (1997). The therapist as a neglected variable in psychotherapy research. Clinical
Psychology: Science and Practice, 4(1), 40–43.
Henry, W.P., Schacht, T.E., & Strupp, H.H. (1990). Patient and therapist introject,
interpersonal process and differential psychotherapy outcome. Journal of Consulting and
Clinical Psychology, 58, 768–774.
Lambert, M.J., Hansen, N.B., & Finch, A.E. (2001). Patient-focused research: Using patient
outcome data to enhance treatment effects. Journal of Consulting and Clinical Psychology,
69, 159–172.
Lambert, M.J., Morton, J.J., Hatfield, D., Harmon, C., Hamilton, S., Reid, R.C., et al. (2004).
Administration and Scoring Manual for the Outcome Questionnaire - 45. Orem, UT:
American Professional Credentialing Services.
Luborsky, L., McClellan, T., Woody, G., O’Brien, C., & Auerbach, A. (1985). Therapist
success and its determinants. Archives of General Psychiatry, 42, 602–611.
Norcross, J.C. (2002). Empirically supported therapy relationships. In J.C. Norcross (Ed.),
Psychotherapy relationships that work (pp. 3–16). Oxford: Oxford University Press.
Ogles, B.M. (1996). Assessing outcome in practice. Journal of Mental Health, 5(1), 35–46.
Journal of Clinical Psychology DOI: 10.1002/jclp
768 Journal of Clinical Psychology, July 2009

Okiishi, J., Lambert, M.J., Eggett, D., Nielsen, L., Dayton, D.D., & Vermeersch, D.A. (2006).
An analysis of therapist treatment effects: Toward providing feedback to individual
therapists on their clients’ psychotherapy outcome. Journal of Clinical Psychology, 62(9),
1157–1172.
Okiishi, J., Lambert, M.J., Nielsen, S.L., & Ogles, B.M. (2003). Waiting for supershrink: An
empirical analysis of therapist effects. Clinical Psychology & Psychotherapy, 10(6),
361–373.
Riggio, R.E. (1986). Assessment of basic social skills. Journal of Personality and Social
Psychology, 51(3), 649–660.
Riggio, R.E. (1989). Manual for the Social Skills Inventory. Palo Alto, CA: Consulting
Psychologists Press.
Safran, J.D., & Muran, J.C. (2000). Negotiating the therapeutic alliance: A relational
treatment guide. New York: Guilford Press.
Shapiro, D.A., Firth-Cozens, J., & Stiles, W.B. (1989). The question of therapists’ differential
effectiveness: A Sheffield Psychotherapy Project addendum. British Journal of Psychiatry,
154, 383–385.
Singer, J.D. (1998). Using SAS PROC MIXED to fit multilevel models, hierarchical models,
and individual growth models. Journal of Education and Behavioral Statistics, 23(4),
323–355.
Strupp, H.H. (1993). The Vanderbilt Psychotherapy Studies: Synopsis. Journal of Consulting
and Clinical Psychology, 61(3), 431–433.
Strupp, H.H. & Anderson, T. (1997). On the limitations of therapy manuals. Clinical
Psychology: Science and Practice, 4(1), 76–82.
Wampold, B.E. (2001). The great psychotherapy debate. London: Erlbaum.
Wampold, B.E., & Bolt, D.M. (2006). Therapist effects: Clever ways to make them (and
everything else) disappear. Psychotherapy Research, 16, 184–187.
Wampold, B.E., & Bolt, D.M. (2007). Appropriate estimation of therapist effects: One more
time. Psychotherapy Research, 17(2), 256–257.

Journal of Clinical Psychology DOI: 10.1002/jclp