Content uploaded by Gary Michael Burlingame
Author content
All content in this area was uploaded by Gary Michael Burlingame
Content may be subject to copyright.
The Differential Effectiveness of Group Psychotherapy:
A Meta-Analytic Perspective
Gary M. Burlingame, Addie Fuhriman, and Julie Mosier
Brigham Young University
The differential effectiveness of group psychotherapy was estimated in a meta-analysis
of 111 experimental and quasi-experimental studies published over the past 20 years.
A number of client, therapist, group, and methodological variables were examined in
an attempt to determine specific as well as generic effectiveness. Three different effect
sizes were computed: active versus wait list, active versus alternative treatment, and
pre- to posttreatment improvement rates. The active versus wait list overall effect size
(0.58) indicated that the average recipient of group treatment is better off than 72% of
untreated controls. Improvement was related to group composition, setting, and diag-
nosis. Findings are discussed within the context of what the authors have learned about
group treatment, meta-analytic studies of the extant group literature, and what remains
for future research.
Researchers’ understanding regarding the ef-
fectiveness of group psychotherapy has evolved
over the past century. Case studies and anec-
dotal reports characterized the group literature
in the first half of the 20th century, with the first
comparative studies emerging in the 1960s
(Barlow, Burlingame, & Fuhriman, 2000).
Early reviews (Pattison, 1965; Rickard, 1962;
Stotsky & Zolik, 1965) concluded that group
therapy was a helpful adjunctive treatment, al-
though little empirical evidence supported its
use as a robust independent treatment. Review-
ers in the latter part of that decade (Anderson,
1968; Mann, 1966) began to give group a heart-
ier endorsement, describing it as capable of
producing objectively measurable change in
patient attitude, personality, and behavior.
Throughout the 1970s, researchers repeatedly
concluded that group outcomes were consis-
tently superior to those of control groups (Bed-
nar & Kaul, 1978; Emrick, 1975; Lieberman,
1976; Luborsky, Singer, & Luborsky, 1975;
Meltzoff & Kornreich, 1970).
By the 1980s, the field had accumulated suf-
ficient evidence for self-reflection. For instance,
Bednar and Kaul (1978; Kaul & Bednar, 1986)
bemoaned the lack of specificity regarding ex-
planatory factors of patient improvement and
issued a clarion call for a conceptual model that
would articulate the fundamental elements in-
volved in effective group treatment. Fuhriman
and Burlingame (1990) answered this call with
a comparative framework of change mecha-
nisms operating in the group and individual
format that was subsequently applied to the
group effectiveness literature (Fuhriman & Bur-
lingame, 1994; Hoag & Burlingame, 1997;
McRoberts, Burlingame, & Hoag, 1998).
A most promising trend in the group litera-
ture of the 1990s was the development and
application of group treatment protocols that
target specific patient diagnoses, settings, and
orientations (e.g., Hoag & Burlingame, 1997;
Fettes & Peters, 1992). These developments
were in contrast to previous reviews that viewed
the group format as an all-inclusive generic
treatment consisting of heterogeneous patients
and treatment models (cf. Burlingame, Mac-
Kenzie, & Strauss, in press). An equally encour-
aging finding was the increased demand for
group treatment in clinical practice. For in-
stance, Taylor and colleagues (2001) surveyed
behavioral health maintenance organizations
Gary M. Burlingame and Julie Mosier, Department of
Psychology, Brigham Young University; Addie Fuhriman,
Planning and Assessment, Brigham Young University.
This research was supported by a grant from the College
of Family, Home and Social Science at Brigham Young
University. We are indebted to the contributions of Anthony
Tzuomas, Matthew Hoag, Chris McRoberts, and Ed Ander-
son at different stages of this project.
Correspondence concerning this article should be ad-
dressed to Gary M. Burlingame, Department of Psychology,
Brigham Young University, 238 TLRB, Provo, Utah 84602.
E-mail: gary_burlingame@byu.edu
Group Dynamics: Theory, Research, and Practice Copyright 2003 by the Educational Publishing Foundation
2003, Vol. 7, No. 1, 3–12 1089-2699/03/$12.00 DOI: 10.1037/1089-2699.7.1.3
3
(HMOs) that cover approximately 70 million
American lives. Respondents were executives
who established mental health benefit policies,
gatekeepers who actually assigned these bene-
fits, and health care providers who delivered
group treatments. Although differences of opin-
ion on a number of treatment and efficacy ques-
tions were found, there was unanimity in the
belief that HMOs would increase the use of
time-limited group treatment for specific diag-
nostic indications in the future and decrease use
of the individual format.
In summary, two factors set the stage for the
current study and bring the group format into
the limelight: the current health care climate
with its emphasis on the efficacy and cost of
services, and the maturation of the group psy-
chotherapy literature beyond demonstrating
general effectiveness to identifying effective
treatments for patients suffering from particular
disorders. These two factors lead to the next
evolutionary step—that is, a quantitative review
of the group therapy literature that is broad
enough in scope to encompass a wide range of
patient and treatment types. Given the growing
prominence of group therapy, the examination
of studies that focus on group format as the
primary treatment (rather than adjunctive) is
essential for determining its independent effect.
Studies that might explain differences in client
improvement in group treatment at an aggregate
level (e.g., theoretical model, patient diagnosis)
are noticeably absent from the literature (Bed-
nar & Kaul, 1994; Burlingame, Ellsworth, Ri-
chardson, & Cox, 2000).
In addition, the examination of specific pa-
tient, therapist, and group variables that might
contribute to the differential efficacy of group
treatment has been productive in recent reviews
(cf. Hoag & Burlingame, 1997; McRoberts et
al., 1998). Although a number of singular in-
vestigations have examined outcome in relation
to a particular diagnosis, most have not taken
into consideration important features of the
group format such as leader characteristics,
group structure, or theoretical model (Burlin-
game et al., in press). In the present study we
sought to explore, in a systematic fashion, the
relationship between improvement rates in
group psychotherapy and several treatment
(e.g., orientation, setting, dosage, group size,
and composition), therapist (e.g., experience),
client (e.g., diagnoses, chronicity of disorder,
gender, and age), and methodological (e.g., ran-
dom assignment and attrition) variables.
Method
Inclusion and Exclusion Criteria
A comprehensive literature search, including
PsycLIT and Medline databases, previous meta-
analytic studies, narrative reviews, and outcome
studies, was conducted using the search terms
group psychotherapy, group counseling, and
group therapy filtered by a health or mental
health emphasis. This search yielded 2,025 ci-
tations. Nonempirical articles were eliminated
(e.g., descriptions of treatment models or pro-
grams, case studies), as were studies that in-
cluded children or adolescents, marital and fam-
ily therapy studies, dissertations, or treatment
that did not have a prominent, interactive, ver-
bal component (e.g., structured, self-help, exer-
cise based). Only those studies that used group
therapy as a primary treatment modality were
retained. Studies were included if (a) treatment
groups met regularly with an identified therapist
and had a therapeutic purpose, (b) clients ex-
hibited an identifiable problem representative of
clients treated within a group format, (c) the
studies were in English, (d) the study was ex-
perimental or quasi-experimental in nature, and
(f) outcome statistics by which an effect size
(ES) could be calculated were reported. One
hundred eleven studies resulted.
Coding Procedures
All 111 studies were coded on client, leader-
ship, treatment, and methodological variables
that past research and theory suggest may be
related to treatment efficacy or that correlate
with an ES (e.g., Burlingame, Kircher, & Tay-
lor, 1994; Fuhriman & Burlingame, 1994; Lam-
bert & Hill, 1994). Client variables included
diagnosis, chronicity, inpatient or outpatient
status, gender, and age. Therapist characteristics
consisted of the theoretical orientation, years of
experience, and professional training (e.g., psy-
chologist or social worker). Methodological
variables included the source (self-report, ther-
apist, independent observer, significant other,
and objective indices such as physiological
readings) and content (general, personality, so-
cial adjustment, somatic, and target measures)
categories of outcome measures suggested by
4 BURLINGAME, FUHRIMAN, AND MOSIER
Lambert and Hill (1994), as well as random
assignment, use of methods to ensure treatment
integrity, number of groups per study, and at-
trition rate.
In an attempt to capture the unique dynamics
that operate within group therapy, additional
variables were included. Specifically, groups
were classified by size (small ⫽ 1–4 members,
medium ⫽ 5–12 members, and large ⬎ 12
members), membership (i.e., open vs. closed),
and composition (homogeneous vs. heteroge-
neous). The presence of pregroup training,
number and length of sessions (e.g., therapy
dosage), setting (e.g., university counseling
center, outpatient practice), and the existence of
a process focus (members encouraged to freely
interact) were also coded. Finally, because there
is some thought that the kind and degree of
structure within the group may be related to
outcome (e.g., Dies, 1994; Fuhriman & Burlin-
game, 1994), a categorization scheme was cre-
ated (outlined in Table 1) to discriminate vari-
ous levels and types of structure. Treatment
Types 1 and 2 reflect the highest degree of
structure because they are guided by a manual.
Types 3 and 4 reflect less structure, with treat-
ment guided by a general theoretical model that
does not specify session-by-session goals and
objectives.
With the exception of the treatment structure
categories, which were assigned by Gary M.
Burlingame and Addie Fuhriman, all variable
coding was conducted by clinical psychology
doctoral and advanced undergraduate psychol-
ogy students. After a semester-long training
program, coders rated in teams of two. Both
raters on a team independently rated each study.
Consensus coding or a final decision by the
senior author resolved discrepancies in indepen-
dent ratings. Average kappa values for the in-
dependent ratings, prior to consensus coding,
were good (.81).
ES Calculation
Improvement for each outcome measure re-
ported in a study was expressed as an ES using
the method suggested by Cohen (1977): ES ⫽
(M
1
– M
2
)/SD
p
, where M
1
and M
2
are the
means of the two conditions being compared
and SD
p
is the pooled standard deviation of the
two groups. Using this formula, we calculated
two comparative ES estimates to provide a met-
ric of the posttreatment gains attained by group
therapy members relative to an alternative or
wait-list comparison group. A third ES pro-
duced a pre- to posttreatment improvement
index.
When calculating individual ESs within a
study, we assigned an ES of zero to nonsignif-
icant results (e.g., a t or F value) and when
results were not reported for an outcome mea-
sure used in a study (Casey & Berman, 1985).
Although it is common for researchers to use
more than one outcome measure in a given
study, using all ESs from a single study (or all
measures) in calculating the overall ES can un-
duly bias it toward the results found in studies
that contain a large number of outcome mea-
sures (e.g., the problem of differential weight-
ing). To avoid this problem, we calculated a
single mean ES for each study by averaging all
Table 1
Classifying Degree of Structure in Group Treatment
Type Description
1 Guiding force is the therapist or manual. Therapist or manual provides session-by-session structures.
Discussion focuses on topic or content. Members respond to specific subject matter or to specific
actions or behavioral practice.
2 Moving force is client(s) or group. Treatment is structured around the topic or content. Discussion
(as in Type 1) focuses on topic or content, with members focusing on specific subject matter or
on specific actions or behavioral practices. Groups may or may not use a manual.
3 Guiding force is the therapist. Therapist or a specific model of group therapy structures the
treatment. Discussion promotes interactive, responsive group process. Discussion focuses on the
client(s) or group as a whole—their reactions, behaviors, and feelings, with evidence of a here-
and-now orientation.
4 Moving forces are the client(s) or group. The unique social microcosm, created by the composition
of group members, structures treatment. Discussion promotes interactive, responsive group
process. Discussion focuses on the client(s) or group as a whole—their reactions, behaviors, and
feelings, with evidence of a here-and-now orientation.
5META-ANALYTIC PERSPECTIVE
of the ESs for all measures within a given study.
However, for the content and source category
estimates, an average ES was first calculated for
each category within a study (e.g., if a study
used four self-report measures, a single self-
report estimate was calculated). Finally, be-
cause larger sample sizes tend to estimate the
ES true score more reliably than studies with
small sample sizes (Hedges & Olkin, 1985),
ESs were weighted by the inverse of their
variances.
Analysis
Three distinct sets of analyses were per-
formed: group treatments compared with wait-
list control groups; within-study comparisons of
alternative treatment, as categorized by the
structure scheme (Table 1); and pre- to post-
treatment improvement. After computing indi-
vidual ESs for these three analyses, a within-
subject comparison was conducted to determine
whether improvement (i.e., ES) was related to
the aforementioned therapist, client, group,
methodology, and treatment structure variables.
Bonferroni corrections were applied to these
multiple comparisons. Additionally, because
sample size changed depending on the particu-
lar comparison being made, a Levene’s test for
equality of variance was conducted, and when
appropriate, t tests for samples with unequal
variance were reported. Finally, when the sam-
ple size was too small for the level of a variable
(e.g., only 2 out of 110 studies used nurses as
providers) or when there was too much missing
information (e.g., over half of the studies did
not report orientation), levels or variables were
removed in a particular analysis.
Results
Studies varied on who was in the group (cli-
ents), what treatment was offered (theoretical
orientation), and where the study took place
(setting). The study characteristics summarized
in Table 2 by client diagnosis, therapist charac-
teristics, theoretical orientation, and setting pro-
vide a better understanding of the sample. More
specifically, clients with medical conditions
(e.g., obesity) formed the largest part of the
sample (24%), followed by anxiety, stress, de-
pressive, and eating disorders. Individuals suf-
fering from normal stress and those with psy-
chotic or thought disorders each constituted
smaller proportions of the sample (approxi-
mately 3%).
Approximately half (47%) of the studies did
not indicate the theoretical orientation of the
provider or else stated it so unclearly that a
precise categorization was impossible. In those
studies where theoretical orientation was clearly
identified, the orientation was most often cog-
nitive–behavioral, followed by behavioral, psy-
chodynamic, and multidisciplinary therapies.
As was the case with orientation, a high per-
centage of studies failed to disclose sufficient
data regarding professional status of the leader,
making assignment to a specific category im-
possible in 35% of the studies. In those studies
where professional status was reported, the ma-
jority of providers were doctoral-level psychol-
ogists. Finally, the setting of group treatment
was most often a university counseling center,
followed by correctional institutions and outpa-
tient mental health organizations.
Group Therapy Versus Wait-List Controls
Fifty-one studies compared group therapy
with wait-list controls. No significant differ-
Table 2
Study Characteristics
Variable %
Primary diagnosis
Medical 24
Anxiety disorders 13
Stress disorders
a
11
Depressive disorders 10
Eating disorders 10
Sexual abuse 9
Criminal behavior 9
Thought disorders 3
Theoretical orientation
Cognitive–behavioral 51
Behavioral 19
Psychodynamic 15
Multidisciplinary/eclectic 15
Setting
University counseling center 52
Correctional institution 20
Outpatient mental health center 12
Therapist degree
Psychologists (PhD) 72
Master’s level (MS, MSW) 17
Note. Cumulative percentages that do not equal 100% are
reflective of studies that did not provide descriptive data on
the above characteristics.
a
Including posttraumatic stress disorder and adjustment
disorders.
6 BURLINGAME, FUHRIMAN, AND MOSIER
ences were found between outcome measures
that tapped different content domains
(F ⫽ 1.39, p ⬎ .05) or used different sources
(F ⫽ 1.23, p ⬎ .05). This may be due, in part,
to the large variability within categories (see
Table 3). Nevertheless, the average ESs were
moderate (e.g., .63 and .47), indicating that ac-
tive group treatment results in reliable improve-
ment when compared with wait-list controls.
The group treatment versus wait list ESs were
examined on the aforementioned client, leader-
ship, treatment, and methodological variables.
Significant differences emerged on three. First,
members in homogeneous groups (n ⫽ 71,
ES ⫽ 0.56) attained more improvement than
members in heterogeneous groups (n ⫽ 8,
ES ⫽ 0.25), t(77) ⫽ 2.40, p ⬍ .05. Second,
outpatient groups (n ⫽ 72, ES ⫽ 0.55) outper-
formed inpatient groups (n ⫽ 6, ES ⫽ 0.20),
t(76) ⫽ –2.49, p ⬍ .05. Third, differences were
found in patient gender. Specifically, members
of mixed-gender groups (n ⫽ 41, ES ⫽ 0.66)
posted higher gains than all-male (n ⫽ 8,
ES ⫽ 0.41), t(47) ⫽ –2.06, p ⬍ .05, and all-
female groups (n ⫽ 23, ES ⫽ 0.39), t(62) ⫽
–1.97, p ⬍ .05. However, there were no signif-
icant differences in improvement between the
all-male and all-female groups.
No significant differences were found in de-
gree of improvement examined by patient diag-
nosis, group size, dosage, random assignment,
treatment integrity, or attrition. The exploratory
variable on group structure (Table 1) was not
related to patient improvement (Type 1 ⫽ .57,
Type 2 ⫽ .37, Type 3 ⫽ .35, and Type 4 ⫽ .51).
However, the majority (i.e., 52 of 78) of treat-
ment groups were classified as Type 1, making
this an unbalanced comparison. Differences be-
tween groups with (n ⫽ 14, ES ⫽ 0.39) and
without pregroup training (n ⫽ 65, ES ⫽ 0.57)
approached the conventional level of statistical
significance, t(77) ⫽ 1.84, p ⫽ .07, although the
imbalance numbers in this comparison make the
impact of conducting such pregroup interviews
less clear. Finally, lack of reported informa-
tion in the studies made it impossible to analyze
ESs on the treatment setting, theoretical orien-
tation, chronicity (acute vs. chronic patients),
open versus closed group membership, and age
variables.
Alternative Group Treatments
Fifty comparisons were found that directly
contrasted two or more of the four structural
classifications outlined in Table 1. In most
cases, the number of studies representing a spe-
cific contrast (e.g., Treatment 2 vs. Treatment 3)
was too small to make an average ES meaning-
ful. Using a standard of at least four studies per
comparison, a sufficient number of studies (n ⫽
30) existed for four comparisons (see Table 4).
However, none of these comparisons attained
statistical significance.
The small number of studies in each compar-
ison virtually eliminated all analyses of client,
leadership, treatment, and methodological vari-
ables. However, there were enough Type 1 ver-
sus Type 2 comparisons in which study authors
reported gender composition to test for this ef-
fect. Mixed-gender groups (n ⫽ 7, ES ⫽ 0.42)
demonstrated significantly more improvement
than all-male groups (n ⫽ 7, ES ⫽ 0.00) and
approached significance when compared with
all-female groups (n ⫽ 3, ES ⫽ 0.00): F(2,
45) ⫽ 4.28, p ⬍ .05, and F(1, 12) ⫽ 3.29, p ⫽
.09, respectively. No significant differences
were found in this comparison on the variables
of random assignment, level of experience,
treatment integrity, or attrition rate. Insufficient
data were available on group size, chronicity,
age, pregroup training, composition, member-
ship, and inpatient versus outpatient.
Table 3
Group Versus Wait-List Control Effect Sizes (ESs)
Ordered by Content and Source
Variable n ES 95% CI
a
Content
General 40 0.57 0.29–0.65
Target 58 0.52 0.37–0.69
Personality 26 0.71 0.37–0.77
Social adjustment 8 1.01 ⫺0.07–2.08
Somatic 13 0.35 0.13–0.52
Total 145 0.63 0.21–0.94
Source
Self-report 72 0.56 0.41–0.64
Significant other 6 0.81 0.48–1.16
Therapist 3 0.35 0.46–2.50
Independent rater 15 0.73 0.03–1.07
Objective 15 0.31 0.05–0.51
Total 111 0.47 0.29–1.12
Note. Totals greater than 51 reflect multiple measures
used within individual studies. CI ⫽ confidence interval.
a
A range that includes 0 suggests no reliable change (effect
size).
7META-ANALYTIC PERSPECTIVE
Pre- to Posttreatment Change
Comparisons
Pre- to posttreatment ESs were calculated on
111 active group treatment and 51 wait-list con-
trol groups (see Table 5). The overall ES for
active treatment groups was .71, which demon-
strates that statistically significant pre- to post-
treatment improvement took place in these pa-
tients. The aggregate patient improvement for
the wait-list controls did not statistically differ
from zero, indicating that no reliable improve-
ment took place. This indirect comparison of
improvement complements the aforementioned
active versus wait-list comparison, providing an
index of absolute patient improvement.
The relationships between pre- to posttreat-
ment ESs and the client, leadership, treatment,
and methodological variables were also exam-
ined. The most robust finding was the associa-
tion between ES and diagnosis, F(2,
155) ⫽ 2.07, p ⬍ .05. Nearly three fourths
(71%) of the client diagnosis categories demon-
strated reliable improvement (Table 5). Those
that did not (i.e., outpatient, substance abuse,
thought disorder, and criminal behavior) typi-
cally had smaller sample sizes, which in turn
reduced the statistical power of the analysis.
Reliable differences were also found between
diagnostic categories. For instance, depressed
patients benefited significantly more from group
treatment than did those with medical condi-
tions, stress-related disorders, no diagnoses, and
neurotic disorders (Table 5). Eating-disordered
clients improved significantly more from group
therapy than did those with medical or stress-
related conditions, the sexually abused, outpa-
tients, inpatients, neurotics, normals, and those
in correctional institutions.
Pre- to posttreatment ESs for clients who
remained on wait list revealed that those in
correctional institutions significantly deterio-
rated when left untreated (ES ⫽ –0.33, p ⬍
.05), whereas those with eating disorders, gen-
Table 4
Aggregate and Individual Effect Sizes (ESs) for
Alternative Group Treatments
Comparison n Mean ES
Type 1 vs. Type 2 15 0.19
Type 1 vs. Type 3 6 0.36
Type 1 vs. Type 4 5 0.40
Type 2 vs. Type 4 4 0.00
Note. Types 1–4 are defined in Table 1. Positive ESs
signify an advantage for the first group over the second.
Table 5
Pre- to Posttreatment Effect Sizes (ESs) for Various Patient Types and
Diagnoses
Patient category
Treatment Wait list
n ES n ES
Depression 25 1.10* 3 0.43
Eating disorder 12 1.38* 7 0.00
Personality disorder 1 0.91*
Substance abuse 2 0.90
Thought disorder 4 0.64
Anxiety disorder 10 0.84** 4 0.20
Inpatient 5 0.66**
a
1 2.24**
Outpatient 5 0.53
a
3 0.00
Sexually abused 5 0.69*
a
3 0.00
Stress 7 0.50*
a,b
5 0.12
Neurotic 4 0.45*
a,b
2 0.00
Medical conditions 23 0.49*
a,b
13 0.00
Criminal behavior 4 0.49
a
1 ⫺0.33*
Normal (DSM–IV V-codes) 4 0.37*
a,b
2 0.28
Total or average 111 0.71** 51 0.21
Note. DSM–IV ⫽ Diagnostic and Statistical Manual of Mental Disorders (4th ed.; American
Psychiatric Association, 1994).
a
Significantly lower ES than that achieved by eating-disordered clients (p ⬍ .05).
b
Sig
-
nificantly lower ES than that achieved by depressed clients (p ⬍ .05).
* p ⬍ .05. ** p ⬍ .01.
8 BURLINGAME, FUHRIMAN, AND MOSIER
eral outpatients, the sexually abused, the neu-
rotic, and those with medical conditions re-
mained unchanged (ES ⫽ 0.00). Although cli-
ents with depressive or anxiety disorders and
those described as neurotic or normal appeared
to improve (i.e., there was a positive ES), this
improvement did not reach statistical signifi-
cance. The only diagnostic category on the
wait-list condition that demonstrated significant
improvement, inpatient, is suspect, because it
is 9 times higher than the average wait-list ES
and was based on a single study.
Once again, members of homogeneous
groups (n ⫽ 104, ES ⫽ 0.82) posted more
improvement than their counterparts in hetero-
geneous groups (n ⫽ 13, ES ⫽ 0.42),
t(28) ⫽ 3.43, p ⬍ .01 (unequal variance). Set-
ting was related to improvement, with greater
gains found for outpatients when compared
with inpatients, t(18) ⫽ 3.64, p ⬍ .01 (unequal
variance). A modest effect for theoretical orien-
tation was found, F(5, 68) ⫽ 2.86, p ⬍ .05, with
behavioral treatments outperforming eclectic
models (ES ⫽ 1.02 and 0.40, respectively).
Finally, no significant relationships were found
on pregroup training, membership, therapist ex-
perience, client gender or age, random assign-
ment, treatment integrity and structure (Table
1), or attrition rate. Insufficient information was
available to test group size and chronicity.
Discussion
The study characteristics, although not the
focus of the research, do lend some insight as to
the settings in which empirical analysis most
often is undertaken, who is providing the treat-
ment, and what orientation they espouse. An
exhaustive search of the literature and rather
stringent criteria for study acceptance revealed
that the settings wherein this activity is staged
were most often university counseling centers
(52%), followed by correctional institutions
(20%) and outpatient mental health organiza-
tions (12%). Professionals conducting group
treatment largely came from the ranks of doc-
toral-level psychologists (72%), and cognitive–
behavioral was by far the theoretical orientation
most used (51%), versus behavioral (19%), psy-
chodynamic (15%), and eclectic (15%).
These facts raise a number of interesting
questions. For example, are these settings pro-
portionately accurate as to where group therapy
occurs? Are doctoral-level psychologists really
the major deliverers of group treatment? And
what drives the interest in understanding group
therapy effectiveness? Contributing to these
questions are the results indicating the number
of studies that had small sample sizes, the result
of which limits the ability to understand the
effectiveness of treatment with specific diag-
noses. Perhaps what is clear from these ques-
tions and the study features is the need for
clinicians and researchers across various set-
tings (including private practice and inpatient)
to collaborate in their efforts to understand not
only the efficacy of group treatment in general
but the implications for individual diagnoses
and process variables as well.
Additional study characteristics are revealed
by the methodological variables. The majority
of the studies (n ⫽ 58) tapped content through
focused target symptoms, followed by general
outcome (n ⫽ 40) and personality (n ⫽ 26)
measures. In the case of the source of outcome
data, the vast majority used self-report measures
(n ⫽ 72), followed by independent rater and
objective (n ⫽ 15 each) instruments. Taken
together, it appears that the group efficacy lit-
erature encompasses a broad range of content
areas but is more limited regarding the sources
from which it gathers its data. Questions follow
these characteristics: What is the relationship
(regarding validity and reliability) among the
varying source measures? At what point do
multi-informant data yield more accurate re-
sults? It is encouraging that researchers are us-
ing multiple measures that will add precision to
the literature base. Nevertheless, there could be
a conservation of effort if our understanding of
the validity of sources were better.
These source and content variables were not
significantly related to the effectiveness of treat-
ment. When we found that the level of improve-
ment achieved by group treatment did not vary
significantly depending on the source of the
outcome measures, we were surprised given
that earlier reviews reported that significant-
other ratings reliably exceeded physiological
measures (Lambert & Hill, 1994; Shapiro &
Shapiro, 1982) and that therapist ratings pro-
duced larger ESs than self-report and signifi-
cant-other ratings (Lambert & Hill, 1994). Ad-
ditionally, in the present study, self-report mea-
sures were not reliably higher than independent
observer ratings, as has been suggested by past
reviewers (Miller & Berman, 1983).
9META-ANALYTIC PERSPECTIVE
The lack of relationship between the content
of the outcome measure and level of improve-
ment parallels results of Robinson, Berman, and
Neimeyer (1990) but counters the conclusions
of Lambert and Hill (1994) and Shapiro and
Shapiro (1982). For instance, in contrast to the
finding that targeted outcome measures yield
the highest ESs (Lambert & Hill, 1994), we
found that social adjustment measures yielded
the highest levels of improvement for group
treatment. Although this is compatible with
Yalom’s (1995) social microcosm theory, it is
best considered tentative given that the estimate
was based on only eight comparisons.
Group Therapy Versus Wait-List Controls
The findings of this meta-analytic study pro-
vide a quantitative estimate of the effectiveness
of group therapy taken from a large number of
investigations wherein group was the predomi-
nant focus of treatment. The fact that the aver-
age active group treatment client was better off
than the untreated controls (average ES ⫽ 0.58)
provides quantitative support for group treat-
ment as an independently efficacious treatment.
The meaning of this finding can be better un-
derstood when compared with ES estimates de-
rived from the general psychotherapy literature.
Lambert and Bergin (1994) gave an estimate of
75% (ES ⫽ 0.82) for the effectiveness of active
psychotherapies when contrasted with no-treat-
ment controls.
Significant differences between the therapy
groups and the wait-list controls occurred
within variables that are driven by client char-
acteristics and combinations of other variables.
Specifically, members in homogeneous, outpa-
tient, and mixed-gender groups attained more
improvement than did their counterparts. This
result draws our attention to the composition of
the group—the way that groups are comprised.
This effect highlighting group formation is
helpful to the literature, particularly given the
fact that composition is one of the least studied
factors in all the group empirical history. Given
the importance of client variables in both indi-
vidual and group formats, it is peculiar that
more attention has not been given to the mix of
client factors operating in the group. In addition,
the finding regarding homogeneity of problem
focus lends strength to the current trend in
group treatment, that being the conduct of short-
term, topic-focused groups (e.g., cognitive–be-
havioral, eating disorder, grief–loss). The fact
that no differences between the treatment and
control groups occurred among patient diag-
noses, group size, dosage, random assignment,
treatment integrity, or attrition may, in fact,
illustrate our lack of knowledge concerning the
interaction of patient, setting, and other treat-
ment variables.
Alternative Group Treatment
Least meaningful of the analyses was the
comparison of differing levels of structure being
applied in each of the groups, given either an
insufficient number of studies available for
comparison or a lack of significance found.
However, comparisons between Type 1 and
Type 2 groups (the more highly structured
groups) were possible, and, similar to the com-
parison between treatment and controls, mixed-
gender groups significantly outperformed all-
male groups and approached significance in
outperforming all-female groups. Once again,
composition appears important even in highly
structured, more psychoeducational groups in
which the interaction is driven by the therapist,
a manual, or a specific topic. The relationship
between these gender composition features and
the amount and kind of interaction remains to be
seen but would be important to know, given
what is already known about the effect of “ther-
apeutic” interaction. Nevertheless, this finding
must be considered in light of the small sample
of groups that were composed of the same gen-
der, as it appears that the vast majority of
groups conducted are not segregated by gender.
The lack of detail or consideration of such com-
position and format factors in the literature and,
hence, in the analyses of this study (i.e., patient-
by-process variables) is of concern.
Pre- to Posttreatment Change
Comparisons
The pre- to posttreatment change compari-
sons begin by underscoring the overall effec-
tiveness of group therapy. Improvement did,
indeed, take place, thus confirming that group
therapy works. In these studies, waiting for
treatment did not affect patients’ improvement
in a positive manner. In fact, in the case of
clients in incarcerated groups, clients in the
wait-list controls deteriorated, providing a cau-
10 BURLINGAME, FUHRIMAN, AND MOSIER
tion to random assignment to waiting lists or no
treatment for some clientele. Not only do these
pre- to posttreatment results give us information
regarding the general efficacy of group treat-
ment and reconfirm the conclusions from the
active versus wait-list comparisons, they are
also encouraging to the field because they pro-
vide effectiveness of group-by-individual diag-
noses. Specifically, three fourths of the classi-
fied patient diagnoses demonstrated reliable im-
provement, thus helping refine our knowledge
about what works with whom. Regrettably,
there was no improvement with outpatient, sub-
stance abuse, thought disorder, or criminal be-
havior. However, an understanding of the re-
sults regarding these disorders is limited by
smaller sample sizes, once again reminding us
of the need for tracking outcomes on specific
diagnoses rather than merely on the aggregate
of group treatment. The fact that of all variables
studied (i.e., client, leadership, treatment, and
methodology), those relating to the patient are
the most robust is congruent with and extends
the findings of Gorey and Cryns’s (1991) meta-
analysis, in which improvement was found to be
most related to client variables.
Further specification of treatment change was
revealed among the client variables in the reli-
able differences that occurred between several
diagnostic categories. Patients suffering from
depression and eating disorders indicated more
improvement than did those with other disor-
ders. It is not apparent why such differences
occurred; nonetheless, it remains important to
consider that there are differential effects of
treatment on the disorders treated with group
therapy and to further the pursuit of why such is
true.
Finally, the results indicate that clients in
homogeneous groups outperformed those in
groups with mixed symptoms, outpatients im-
proved more than inpatients, and those in
groups guided by a behavioral orientation did
better than those in groups of an eclectic orien-
tation. It is difficult to account for such findings,
particularly when viewed individually, but a
collective examination of them does offer one
reasonable response. Groups guided by a behav-
ioral orientation tend to be composed of clients
who suffer from the same or similar problems
and use a manual- or therapist-driven approach
(i.e., illustrated by the number of studies using
the two most structured group formats). A
working assumption herein is that the acquisi-
tion of information and the practice of relevant
behaviors are prerequisite to treatment gains.
These factors also coincide, more characteristi-
cally, with outpatient versus inpatient groups.
Hence, it may well be that the combination of
these three features—homogeneity, outpatient
composition, and behavioral orientation—is as-
sociated with overall greater client improve-
ment in treatment groups.
The meta-analytic data from this study con-
firm the general and selected diagnostic effec-
tiveness of group treatment, and in a day when
group treatment is on the rise, this indeed is
encouraging. Nonetheless, this is also a time
when there is increased interest on the part of
service recipients and public and private agen-
cies for assurance that treatment will be effec-
tive, regardless of where it takes place. Such
assurance can come only from valid, reliable,
and clearly communicated results from current
clinical practice. The day is long gone for the
field to rely on information emanating from
isolated, fragmented empirical studies. The call
herein is for increased collaboration among cli-
nicians and researchers and for the conduct of
empirical work in all of the various treatment
settings.
References
American Psychiatric Association. (1994). Diagnos-
tic and statistical manual of mental disorders (4th
ed.). Washington, DC: Author.
Anderson, A. (1968). Group counseling. Review of
Educational Research, 33, 209–226.
Barlow, S., Burlingame, G., & Fuhriman, A. (2000).
Therapeutic application of groups: From Pratt’s
“thought control classes” to modern group psycho-
therapy. Journal of Group Dynamics, 4, 115–134.
Bednar, R. L., & Kaul, T. J. (1978). Experiential
group research: Current perspectives. In S. Gar-
field & A. Bergin (Eds.), Handbook of psychother-
apy and behavior change (2nd ed., pp. 769–815).
New York: Wiley.
Bednar, R. L., & Kaul, T. J. (1994). Experiential
group research: Can the cannon fire? In A. E.
Bergin & S. L. Garfield (Eds.), Handbook of psy-
chotherapy and behavior change (4th ed., pp. 631–
663). New York: Wiley.
Burlingame, G. M., Ellsworth, J. R., Richardson,
E. J., & Cox, J. C. (2000). How effective is group
psychotherapy: A review and synthesis of treat-
ment outcomes. Ricerche Sui Gruppi, 10, 7–35.
Burlingame, G., & Kircher, J., & Taylor, S. (1994).
Methodological considerations in group psycho-
therapy research: Past, present, and future prac-
11META-ANALYTIC PERSPECTIVE
tices. In A. Fuhriman & G. Burlingame (Eds.),
Handbook of group psychotherapy (pp. 41–80).
New York: Wiley.
Burlingame, G. M., MacKenzie, K. R., & Strauss, B.
(in press). Small group treatment: Evidence for
effectiveness and mechanisms of change. In M. J.
Lambert (Ed.), Handbook of psychotherapy and
behavior change (5th ed.). New York: Wiley.
Casey, R. J., & Berman, J. S. (1985). The outcome of
psychotherapy with children. Psychological Bulle-
tin, 98, 388–400.
Cohen, J. (1977). Statistical power analysis for the
behavioral sciences (Rev. ed.). New York: Aca-
demic Press.
Dies, R. (1994). Therapist variables in group psycho-
therapy research. In A. Fuhriman & G. M. Burlin-
game (Eds.), Handbook of group psychotherapy:
An empirical and clinical synthesis (pp. 114–154).
New York: Wiley.
Emrick, C. (1975). A review of psychologically ori-
ented treatment of alcoholism. Journal for the
Study of Alcoholism, 36, 88–108.
Fettes, P. A., & Peters, J. M. (1992). A meta-analysis
of group treatments for bulimia nervosa. Interna-
tional Journal of Eating Disorders, 11(2), 97–110.
Fuhriman, A., & Burlingame, G. (1990). Consistency
of matter: A comparative analysis of individual
and group process variables. Counseling Psychol-
ogist, 18(1), 6–63.
Fuhriman, A., & Burlingame, G. M. (1994). Group
psychotherapy: Research and practice. In A. Fu-
hriman & G. M. Burlingame (Eds.), Handbook of
group psychotherapy: An empirical and clinical
synthesis (pp. 3–40). New York: Wiley.
Gorey, K. M., & Cryns, A. G. (1991). Group work as
interventive modality with the older depressed cli-
ent: A meta-analytic review. Journal of Geronto-
logical Social Work, 16(1–2), 137–157.
Hedges, L. V., & Olkin, I. (1985). Statistical methods
for meta-analysis. New York: Academic Press.
Hoag, M. J., & Burlingame, G. M. (1997). Evaluating
the effectiveness of child and adolescent group
treatment: A meta-analytic review. Journal of
Clinical Child Psychology, 26, 234–246.
Kaul, T. J., & Bednar, R. L. (1986). Experiential
group research: Results, questions, and sugges-
tions. In S. L. Garfield & A. E. Bergin (Eds.),
Handbook of psychotherapy and behavior change:
An empirical analysis (3rd ed., pp. 671–714). New
York: Wiley.
Lambert, M. J., & Bergin, A. E. (1994). The effec-
tiveness of psychotherapy. In A. E. Bergin & S. L.
Garfield (Eds.), Handbook of psychotherapy and
behavior change (4th ed., pp. 143–189). New
York: Wiley.
Lambert, M. J., & Hill, C. E. (1994). Assessing
psychotherapy outcomes and processes. In A. E.
Bergin & S. L. Garfield (Eds.), Handbook of psy-
chotherapy and behavior change (4th ed., pp. 143–
189). New York: Wiley.
Lieberman, M. (1976). Change induction in small
groups. Annual Review of Psychology, 27, 217–
250.
Luborsky, L., Singer, B., & Luborsky, L. (1975).
Comparative studies of psychotherapy. Archives of
General Psychiatry, 32, 995–1008.
Mann, J. (1966). Evaluation of group psychotherapy.
In J. Moreno (Ed.), The international handbook of
group psychotherapy (129–148). New York: Phil-
osophical Library.
McRoberts, C., Burlingame, G. M., & Hoag, M. J.
(1998). Comparative efficacy of individual and
group psychotherapy: A meta-analytic perspective.
Group Dynamics, 2, 101–117.
Meltzoff, J., & Kornreich, R. (1970). Research in
psychotherapy. New York: Atherton Press.
Miller, R. C., & Berman, J. S. (1983). The efficacy of
cognitive behavior therapies: A quantitative re-
view of the research evidence. Psychological Bul-
letin, 94, 39–53.
Pattison, E. (1965). Evaluation studies of group psy-
chotherapy. International Journal of Group Psy-
chotherapy, 15, 382–397.
Rickard, H. (1962). Selected group psychotherapy
evaluation studies. Journal of General Psychol-
ogy, 67, 35–50.
Robinson, L. A., Berman, J. S., & Neimeyer, R. A.
(1990). Psychotherapy for the treatment of depres-
sion: A comprehensive review of controlled out-
come research. Psychological Bulletin, 108, 30–
49.
Shapiro, D. A., & Shapiro, D. (1982). Meta-analysis
of comparative therapy outcome studies: A repli-
cation and refinement. Psychological Bulletin, 92,
581–604.
Stotsky, B., & Zolik, E. (1965). Group psychotherapy
with psychotics. International Journal of Group
Psychotherapy, 15, 321–344.
Taylor, N. T., Burlingame, G. M., Fuhriman, A.,
Kristensen, K. B., Johansen, J., & Dahl, D. (2001).
A survey of mental health care provider and man-
aged care organization attitudes toward, familiarity
with, and use of group interventions. International
Journal of Group Psychotherapy, 51, 243–263.
Yalom, I. D. (1995). The theory & practice of group
psychotherapy (4th ed.). New York: Basic Books.
Received April 30, 2002
Revision received October 16, 2002
Accepted October 16, 2002 䡲
12 BURLINGAME, FUHRIMAN, AND MOSIER
A preview of this full-text is provided by American Psychological Association.
Content available from Group Dynamics Theory Research and Practice
This content is subject to copyright. Terms and conditions apply.
