Content uploaded by D. Russell Crane
Author content
All content in this area was uploaded by D. Russell Crane
Content may be subject to copyright.
The American Journal of Family Therapy, 33:273–287, 2005
Copyright © Taylor & Francis, Inc.
ISSN: 0192-6187 print / 1521-0383 online
DOI: 10.1080/01926180590962147
What Works for Whom: A Meta-Analytic Review
of Marital and Couples Therapy in Reference
to Marital Distress
NATHAN D. WOOD
Utah Valley State College, Orem, Utah, USA
D. RUSSELL CRANE AND G. BRUCE SCHAALJE
Brigham Young University, Provo, Utah, USA
DAVID D. LAW
Utah State University, Uintah Basin, Utah, USA
Due to methodological limitations, past meta-analytic research was
not able to identify which treatment was most effective for specific
marital distress levels. By converting pre- and post-test scores from
marital research into equivalent Dyadic Adjustment Scale scores,
it was possible to isolate mild, moderate, and severe levels of mar-
ital distress. Results show that Emotionally Focused Therapy is sig-
nificantly more effective than isolated Behavioral Marital Therapy
interventions for the treatment of moderate marital distress. Future
directions of outcome research are also discussed.
Over the years, researchers in the field of Marriage and Family Therapy
(MFT) have attempted to validate the profession’s existence through out-
come studies and meta-analyses. The results of these studies did exactly
what they were designed to do, they took a wide-angled snapshot of the
field and everyone looked good from a distance. All meta-analyses to date
have shown positive results when MFT has been compared to no treatment
(e.g., Butler & Wampler, 1999; Dunn & Schwebel, 1995; Hahlweg & Markman,
1988; Hazelrigg, Cooper, & Borduin, 1987; Shaddish, 1993, Wampler, 1982).
Originally presented as a doctoral dissertation by the first author, under the direction of
the second author. Results were also presented via poster at AAMFT, October, 2003.
Address correspondence to Nathan D. Wood, Turning Point MS-134, Utah Valley State
College, 800 West University Parkway, Orem UT, 84058. E-mail: woodna@uvsc.edu
273
274 N. D. Wood et al.
Because of these meta-analyses, we can claim that what we do matters and
that we do make a difference.
These studies then focused on comparing individual treatment models
against each other. Much like tightly focusing on a wide-angled photograph,
the results gave a blurry, but general impression of the details. Hence the
conclusions that have been reached are also general. The most common
conclusion that has been reached is that different therapeutic approaches
are essentially equal in their effectiveness. The phrase from Alice in Wonder-
land “all have won, and all must have prizes” inspired Luborsky, Singer, and
Luborsky (1975) to dub these general findings as the “Dodo bird effect.”
One reason for the Dodo bird effect in marital therapy was that all
previous meta-analyses were not methodologically designed beforehand to
analyze differences between marital therapy treatments. Previous research
also made no attempt to account for varying degrees of marital health.
Meta-analyses need to isolate marital outcome studies and account for mar-
ital distress levels to have the best chance of finding differences in marital
treatments.
Dunn and Schwebel (1995) took the first step and only included mar-
ital therapy outcome studies in their meta-analysis, but neglected distress
level. This enabled them to more accurately look at treatment effectiveness
of marital therapy by using an apparently homogeneous sample of stud-
ies. They found that no particular theoretical approach was more effective
than another, the apparent Dodo bird effect once again. However, Dunn &
Schwebel (1995) assumed that all measures of marital distress are created
equal, going as far as equating observational measures with pencil and pa-
per measures of marital distress. It is important to note that all other previous
meta-analyses have also made this assumption.
The use of multiple measures of marital distress/satisfaction in meta-
analyses could have a diluting effect on finding the most effective treatment.
To include studies with different measurements of marital satisfaction as-
sumes that all the instruments are measuring the same construct, in the same
way. Research by Crane, Allgood, Larson, and Griffin (1990) show that this
assumption is erroneous and potentially misleading.
Before the study by Crane et al. (1990) was published, couples were
considered distressed if they scored below 100 on the Marital Assessment
Test (MAT: Locke & Wallace, 1959), or below 92 (Baucom & Hoffman,
1996) or 97 (Jacobson, Schmaling, & Holtzworth-Munroe, 1987) on the
Dyadic Adjustment Scale (DAS: Spanier, 1976). Crane et al. (1990) argued
that the MAT score of 100 was comparable to DAS scores of 107, not 92
or 97.
Following Crane et al. (1990), DAS scores of 85–92 (traditionally ac-
cepted scores for mild distress) are actually more comparable to MAT scores
of 72–81 (accepted scores in the range of moderately distressed couples).
If a meta-analyst defined marital distress according to traditionally accepted
What Works for Whom 275
cutoff scores, they would be combining studies of moderately distressed cou-
ples along with those studies of mildly distressed couples.
Shaddish et al. (1995) stated “[as] long as there is such enormous diver-
sity of measurement characteristics, it is difficult to see how we will be able to
generate more coherent analyses of treatment effects over studies” (p. 354).
They assert that with a standard measure of marital distress it would be far
easier for those who review the MFT literature to tell if “differences between
behavior and nonbehavior therapy . . . would disappear when the same out-
comes were measured, or if they would still persist” (p. 354). Given all the
productive work to date, the field is still lacking understanding of which
therapeutic approach, or theory, may be the most appropriate for differing
levels of marital distress.
The first step toward identifying which treatment best fits with a corre-
sponding level of marital distress is to standardize the operationalization of
marital distress. Crane, Allgood, Larson, and Griffin (1990) made the initial
steps toward standardizing the main measures of marital satisfaction used
in marital therapy outcome research. In a sample of 302 individuals, multi-
ple regression analysis was used to estimate the linear relationship between
the DAS, MAT, and the Revised Marital Adjustment Test (RMAT; Kimmel &
Van Der Veen, 1974). This study enabled clinician and researchers to convert
scores back and forth between measures.
Later, Crane, Middleton, and Bean (2000) added the Kansas Marital Satis-
faction Scale (KMSS; Schumm, Nichols, Schectman, & Grisby, 1993; Schumm,
Paff-Bergen, Hatch, Obiorah et al., 1986) and the Revised Dyadic Adjustment
Scale (RDAS; Busby, Crane, Larson & Christensen, 1995) to the mix. It is now
possible to convert scores between the DAS, MAT, RMAT, KMSS, and RDAS.
Now marital assessment scores included in meta-analyses can be con-
verted into a single metric from which differing levels of distress can be
determined. The DAS provides the solution to the meta-analytic problem
of a standardized measure of marital satisfaction (Shadish et al., 1995). It is
now possible to explore treatment effectiveness by varying levels of marital
distress in a more meaningful way utilizing a meta-analytic format.
METHOD
Inclusionary Criteria
Studies were included in the current meta-analysis if the following crite-
ria were met: (a) The study focused specifically on conjoint treatment of
marital distress. Studies that included conjoint treatment of other disorders
were excluded from the current study. (b) Treatment involved both spouses.
(c) Measures used in the studies were one of the following: DAS, MAT, RMAT,
KMSS, or the RDAS. (d) The studies were published between 1963 and 2002.
276 N. D. Wood et al.
Procedure
The initial search for relevant literature was in electronic databases such
as PSYCHINFO, ERIC, SOCIOLOGICAL ABSTRACTS, Social Science Citation
Index (SSCI), and Dissertation Abstracts International. Key words used as
search terms in finding relevant articles included but were not limited to:
marital distress, outcome study, marital therapy, couples therapy, conjoint
therapy, marital satisfaction, and marital conflict. Once relevant literature was
identified, reference lists and bibliographies were examined to identify more
literature. Review articles and meta-analyses were also searched for literature
that fit the inclusionary criteria.
Articles were coded by advanced undergraduate and graduate students
who had a thorough understanding of statistical and methodological issues
(contact the first author for copy of the code book). Groups of 2–3 students
were initially given the same articles for coding to establish inter-rater reliabil-
ity. Consistency in the coding was validated by a visual comparison between
coding sheets. This process continued until each code sheet was identical to
the others. Once inter-rater reliability was established, each coder received
unique articles.
Non-primary literature such as unpublished dissertations were included
because they generate a more realistic picture of treatment effectiveness
(Lipsey & Wilson, 2001; Rosenthal, 1998; Wampler & Serovich, 1996). Of
the 23 studies coded for this meta-analysis, 20 peer-reviewed articles were
coded along with 3 dissertations (See Table 1 for articles included in the
analysis). From the 23 studies included, 41 treatment groups were identified.
Seven treatment groups fell within the mildly distressed range, 33 treatment
groups were classified as moderately distressed, and only one group fell into
the severely distressed range.
Couples’ scores were coded and used in the analysis rather than hus-
bands’ or wives’ individual scores. Treatment approach was coded in two
ways. The first was to code the approach as it was labeled in the paper itself.
The second approach following Dunn & Schwebel (1995) was to affiliate
the main thrust of the treatment with previously established models such as
Behavioral Marital Therapy (BMT), Emotionally Focused Therapy (EFT), and
others. The approaches that did not fit solidly into a single category, such
as behavioral marital therapy with emotional validation interventions added,
were coded as “Mixed” (e.g., Halford, Sanders, & Behrens, 1993). Other stud-
ies contained only individual interventions of BMT, such as problem solving
only, or communication training only (e.g., Jacobson, 1984). These individual
interventions from BMT were coded as “BMT Components.”
Studies were coded into two groups, true experimental and quazi-
experimental designs. Following previous meta-analytic research, true ex-
perimental designs had random assignment to group, presence of a control
group, and a clear treatment protocol (Dunn & Schwebel, 1995; Carrol &
What Works for Whom 277
TABLE 1 Studies Included in the Meta-Analysis
Study Models included
Azrin et al. (1980) BMT
Baucom (1982) BMT Components, BMT
Baucom (1986) BMT, Others
Baucom, Sayers, & Sher (1990) BMT, Mixed
Beach & Broderick (1983) BMT
Bennun (1985) BMT, Others
Bornstein et al. (1983) BMT
Carlton (1978)∗EFT
Follingstad & Haynes (1981) BMT
Halford, Sanders, & Behrens (1993) BMT, Mixed
Hazelrigg, Cooper, & Borduin (1987) EFT
Huber & Milstein (1985) BMT, Others
Jacobson (1978) BMT, Mixed
Jacobson (1984) BMT Components
James (1991) Mixed
Johnson & Greenberg (1985a) BMT, EFT
Johnson & Greenberg (1985b) EFT
Goldman & Greenberg (1992) EFT
Hudson (1978)∗BMT
Liberman et al. (1976) BMT
Margolin & Weiss (1978) BMT Components
Simms (1999)∗EFT
Wilson, Bornstein, & Wilson (1988) BMT, Others
∗Studies were unpublished dissertations.
Doherty, in press; Shaddish et al., 1993). Quazi-experimental designs have
similar features to true experimental designs with the exception of random
assignment to group.
Special care was taken to ensure that each study was independent of
all other studies. The same sample may have been used for several different
studies and therefore results from these studies could not be considered inde-
pendent. If dependent studies were identified, only one study of the set was
used. Study independence is an important issue as resultant effect sizes are
to be used for traditional significance testing which holds data independence
as a primary assumption.
Preliminary Analysis
Initially, all the MAT scores in the coded literature were converted to DAS
scores (see Crane et al., 1990). No studies were found that used the RMAT,
KMSS, or RDAS. Standard deviations were converted from MAT (sMAT) to
DAS (sDAS) by the following formula:
sDAS =sMAT ∗.605
278 N. D. Wood et al.
The conversion of scores and standard deviations was possible because
of the linear relationship between MAT and DAS values (Crane et al.,
1990).
Additional computations were performed on those studies for which
only husband and wife data were coded. For those with husband and wife
scores, the two were added and divided by two. For missing data, couple’s
standard deviations were calculated by using the following formula.
Couple sd =1
4s2
h+s2
w+2rshsw
The formula takes into account the correlation between husband and
wife’s scores and was derived from statistical theory. The higher the corre-
lation coefficient used, the larger the estimate of the couple’s standard de-
viation. As standard deviations increase the resultant effect sizes calculated
decrease, or become more conservative.
Various studies have reported the correlation of husband’s and wife’s
scores to fall between .5 and .73 (Crane, Soderquist, & Frank, 1995;
Oppenheim, Wambboldt, Gavin, Renouf, & Emde, 1996). When the correla-
tion is placed in the aforementioned formula, smaller correlations will gen-
erate more liberal effect sizes while larger correlations are associated with
more conservative effect sizes. A correlation of .70 was chosen from the em-
pirically supported range of correlations in order to generate conservative
estimates.
Where studies did not report husband, wife, or couple standard devi-
ations, standard deviations were substituted by a standard deviation from a
similar sample in another study that used the same assessment instrument
(Lipsey & Wilson, 2001).
Standardized mean gain (ESsg)and mean difference (ESsm )effect sizes
were calculated and a Q-test was then performed on the data. Q-tests are
based on a Chi-square distribution with k−1degrees of freedom. Significant
values imply a heterogeneous distribution. The Q-test indicated that the both
forms of standardized effect sizes represented a heterogeneous population
of studies (p <.05).
Given that some couple data and standard deviations had been calcu-
lated (i.e., converted from MAT to DAS scores) rather than directly measured,
it was hypothesized that there might be a systematic difference between di-
rectly measured data versus calculated data causing the heterogeneity. Mea-
surement error could be one explanation of the difference between DAS
scores and converted DAS scores. Measurement error is the difference be-
tween a subject’s “true score,” or the score they would receive under perfect
conditions, and their actual score. Additional error was possibly introduced
when a score with existing measurement error was converted from the MAT
scale to the DAS scale.
What Works for Whom 279
With these two potential sources of error in converted DAS scores,
additional heterogeneity tests were performed on effect sizes generated
from original DAS scores verses effect sizes from converted DAS scores.
Heterogeneity was still present in both groups suggesting that the het-
erogeneity present was not from converting scores to the DAS scale.
While the conversion of scores to the DAS scale did not explain the
heterogeneity, there is still the potential for error to influence the final
results.
Possible error in the converted scores was controlled for statistically by
including a dummy variable in regression analysis. The dummy variable in
this case was the assessment instrument administered in the study, (0 =DAS,
1=MAT). The presence of this variable (“Assessment”) in a regression model
eliminated the possibility of any difference found between treatment models
to be explained by the conversion of scores.
Where there were no apparent fixed effects, or systematic differences
between the studies that contributed to the sample heterogeneity, it could
be safely assumed that random effects were present. Random effects are
unknown variables that can contribute to the overall variability in effect sizes
beyond sampling error (Lipsey & Wilson, 2001).
Therefore, a mixed model weighted linear regression was utilized in
this study. The mixed model aspect took into account the random effects
and adjusted the effect sizes accordingly. This was in addition to weighting
each effect size by the inverse variance to take into account the sample size
of each individual effect size.
RESULTS
Of the treatment groups found in the initial analysis, 17% (n =7) scored in the
mildly distressed range (DAS 96–107). Eighty percent (n =33) of the groups
included in this meta-analysis were moderately distressed groups whose pre-
test DAS scores ranged from 80–95.9. Only one group fell below the DAS
cutoff for severe marital distress (DAS <80). Due to the low sample size of
the mildly and severely distressed groups, the main focus of the results and
discussion will be on moderately distressed couples.
Standardized effect sizes do not lend themselves to intuitive interpreta-
tion. They refer directly to the pooled standard deviation of reference group
(e.g., control group in the case of mean differences and pre-test/post-test
scores in reference to treatment gains). For example, a mean difference ef-
fect size of .50 means the treatment group performed one half of a pooled
standard deviation better than the control group. A mean difference effect
size of one means the treatment group performed one pooled standard de-
viation above controls.
280 N. D. Wood et al.
Mean Gain
Mean gain effect sizes (ESsg)refer to standardized pre-post test differences of
each treatment group. Moderate distress at pre-test translates into DAS scores
between 80 and 95.99. Each treatment modality included in this analysis had
at least four effect sizes available for analysis. The Standardized Mean Gain
ESsg for moderately distressed groups was used at the dependent measure in
Table 2.
Table 2 shows a regression comparing all theoretical approaches to
BMT Components after controlling for the assessment instrument used at
pretest. The variable “constant” represents BMT Components. The choice of
the comparison group was arbitrary, any theoretical model could have been
the comparison group. Coefficients for the theoretical models represent the
magnitude of the ESsg that needs to be added to, or subtracted from, the ESsg
of BMT Components.
Emotionally Focused Therapy had the highest ESsg while BMT Compo-
nents had the lowest ESsg. These data did not show whether or not each
approach was effective. It shows that there was not any specific approach
that was significantly more predictive of treatment gains than BMT Compo-
nents. This was also accounting for the influence of the other treatments, and
the marital assessment used at pretest.
“Other” approaches came close to being significantly different from BMT
Components (p =.07) requiring that an additional .69 be added to the ESsg of
BMT Components in order for BMT Components to be equal with “Other ap-
proaches.” The assessment instrument used at pretest significantly predicted
ESsg scores. If the MAT was used in a study, the effect size of any treatment
model would need to be reduced by .71 (p <0.01). The significance of the
“Assessment” variable may be a reflection of measurement error introduced
TABLE 2 Summary of Random Effects Regression Analysis for Treatment
of Moderate Marital Distress, Mean Gains (DAS =80–95.9, n =33)
Variable B SE B Beta
Constant (BMT Comp) 1.3626∗∗ 0.2852 0.0000
BMT 0.4284 0.2849 0.2750
EFT 0.5476 0.3798 0.2647
Mixed 0.0936 0.3744 0.0465
Other 0.6958 0.3848 0.2847
Assessment −.7140∗∗ 0.2551 −0.4674
Constant =Isolated components of BMT; BMT =Behavioral Marital Therapy;
EFT =Emotionally Focused Therapy; Mixed =Approaches that combined
elements of EFT and BMT; Others =Cognitive Behavioral Therapy, “Systemic
Therapy”; Assessment =Categorical variable coded 0 =DAS, 1 =MAT; Depen-
dent Variable =ESsg; Method of Moments Random Effects Variance Component,
v=.2344; Q =17.0665, p <.01 (Overall Model); R2=.2983; ∗p<.05; ∗∗p<.01.
What Works for Whom 281
TABLE 3 Summary of Random Effects Regression Analysis for
Treatment of Moderate Marital Distress, Mean Difference (DAS =
80–95.9, n =33)
Variable B SE B Beta
Constant (EFT) 1.0079 0.5477 0.0000
BMT −0.6577 0.4836 −0.3587
BMT Comp −1.1513∗0.5549 −0.5410
Mixed −0.8884 0.5092 −0.3848
Other −0.4048 0.7703 −0.0958
Protocol 1.0667 0.5827 0.3221
Assessment −0.7691∗0.3826 −0.4103
Constant =Emotionally Focused Therapy; BMT =Behavioral Marital
Therapy; Mixed =Approaches that combined elements of EFT and BMT;
BMT Comp =Isolated components of BMT; Others =Cognitive Behav-
ioral Therapy, “Systemic Therapy”; Protocol =Categorical variable to
indicate a clear treatment protocol; Assessment =Categorical variable
coded 0 =DAS, 1 =MAT; Dependent Variable =ESsm; Method of Mo-
ments Random Effects Variance Component, v =.25747; Q =18.6671,
p<.01 (Overall Model) R2=.4922; ∗p<.05, ∗∗p<.01.
by converting MAT scores into DAS scores. It may also be indicative of the
assessment instrument used in the studies.
Mean Difference
Mean difference effect sizes (ESsm)are the effect sizes commonly reported
in other meta-analyses (i.e., Cohen’s d). Table 3 shows the mixed model
regression results for treatment of moderate marital distress with ESsm as
the dependent variable. The overall model significantly predicted 49% of
the variance in treatment vs. control comparisons (Q =18.67, p <.05). The
constant in Table 3 represents EFT. It is important to note the presence of the
“protocol” variable (p =.067). This was a dichotomous variable indicating
whether or not there was a clear treatment protocol (e.g., treatment manual)
that was followed in the study. The coefficients returned for “Assessment” and
“Protocol” remained the same no matter the comparison group. Therefore it
can be said that when clear protocols are followed, an additional 1.067 can
be added to the effect size of any treatment used.
The assessment instrument used at pretest (“Assessment”) was also sig-
nificantly predictive of ESsm (p <0.05). This indicates that if the MAT was
used in the study, the effect size of any theoretical approach needs to be
adjusted by −.77. This may or may not be a statistical anomaly because of
the conversion of MAT scores into DAS scores. In depth explanations behind
the significance of the “Assessment“ variable are present in the discussion
section.
Once the variables “protocol” and “test” have been controlled for, EFT
performed significantly better than BMT Components in reference to their
282 N. D. Wood et al.
respective control groups. All things being equal, effects sizes from BMT
Components need to be reduced by 1.15 in comparison to EFT treatments
(p <0.05). Mixed models approached being significantly different (i.e.,
worse) than EFT (p =0.081).
Analysis of mean gain and mean difference effect sizes for mild distress
show no significant differences between treatment approaches. No statistical
procedures were able to be performed with severely distressed couples as
there was only one group.
DISCUSSION
Limitations
One of the fundamental weaknesses in this study was the small number
of studies available for consideration. The relatively low number of groups
included in the analysis lowers the statistical power of the current results and
increases the odds of saying there are no differences in treatments when in
fact there may be. Saying that there are no differences in treatment in this
situation is not the same as claiming the Dodo bird effect. In fact, in some
instances, the results of this study throw doubt onto the Dodo bird effect.
The best example of this was in the analysis of mean gains (ESsg)in
the treatment of moderate distress. Table 2 shows obvious differences be-
tween full theoretical approaches and isolated BMT Components. The lack
of additional statistically significant differences in treatment models may be
aType II error showing non-significant results when there may actually be
adifference present.
More replications need to be performed on EFT, Mixed, CMT, and CBMT
before any differences can be determined in terms of treatment gains. The
low sample size is due to two main reasons. The first was the narrow focus
of this study. The second major reason of smaller sample size is the lack of
established outcome measures in the field.
The relative lack of replication, besides that of BMT and EFT, may be
more of a comment on the state of academia and what is considered pub-
lishable material. Replications rarely show anything new and therefore may
not be as readily published. Beyond these systemic barriers, clinical outcome
research is expensive and very time consuming which also add to the rela-
tive lack of pure replication. For the field of Marriage and Family Therapy to
thrive, there must be more sophistication and standardization in the way we
measure and treat marital and family issues.
The best example of an approach that was excluded from this analysis
is Insight Oriented Marital Therapy (IOMT; Snyder & Wills, 1989). Studies of
IOMT were excluded due to the fact that the dependent measures used in
their study could not be converted to DAS scores. It is wise to explore the im-
pact of treatment on different aspects of relational and individual functioning.
What Works for Whom 283
However, to exclude established measurements of marital distress makes it
difficult to gain a better understanding of treatment through meta-analysis.
Ideally, the same measure needs to be used across studies. Statistical conver-
sions are possible, as shown by this study. However, conversions between
scores at the scale of meta-analysis can add unwanted variability that may
not be necessarily present if all studies used the same instrument.
The work of Crane et al. (1990) started the process of attempting to
empirically show where the cutoffs were for all three distress levels rather
than statistical speculation. If the field is to use a common metric such as
the DAS, or another instrument, clear criteria must be set forth to distinguish
between distressed and non-distressed populations.
The best example of the importance of setting criteria came in the rel-
ative absence of severely distressed couples. This may not be as much the
fault of researchers as it is the population. Attrition rates may be higher
with severely distressed couples. Their motives in coming to therapy may
also be more disguised. For example, one partner may already be emo-
tionally divorced from the relationship, affairs, or decreased commitment to
the marriage may all effect attrition and what is considered a “successful
outcome.”
An additional confounding factor with severely distress couples is how
the field of marriage and family therapy defines successful outcomes, or clin-
ical significance. In instances of physical, emotional, and/or sexual abuse,
a successful outcome may be the termination of the relationship. Unfortu-
nately, these types of outcomes are difficult to measure, especially if the
abuse was not disclosed in the first place.
More research needs to be done to identify clear cutoffs for mild, mod-
erate, and severe distress. The results from this study clearly show that once
distress level is identified, the Dodo bird hypothesis can be tentatively put on
the endangered species list. It is in no way extinct yet, as more research must
be done. Specifically, more replication needs to be performed with distress
level included in the design.
Are All Assessments Created Equal?
The argument for a single measurement of marital distress is buoyed by the
statistically significant presence of “Assessment” variable included in the anal-
yses. This variable was included to control for any measurement or statistical
errors in converting scores from MAT to DAS. The fact that it was statistically
significant in every regression model for moderately distressed couples raises
questions.
It simply may be a reflection of error introduced by the conversion
of scores and nothing more. Another explanation may relate to the year
of the study used. The DAS has been used much more in recent marital
research than the MAT. It could be argued that recent research is more
284 N. D. Wood et al.
methodologically stringent and would reflect higher effect sizes on average
that past research that used the MAT.
Yet, the impact of the “Assessment” variable in many analyses required
the effect size of some theoretical models be reduced by half. This would
imply that the DAS is more predictive of higher effect sizes than the MAT.
Following this conclusion would call into question previous meta-analytic
work that included effect sizes based on DAS scores because the overall
average effect size generated from the meta-analysis would be higher than
otherwise would be expected if there were no differences between measures.
Another assessment implication of the current study is who is considered
to be mildly distressed. The mildly distressed group in this the study would
be considered to be an “enrichment” population as their pre-test scores are
close to (at least a 96 on the DAS) or greater than the traditionally accepted
cutoff score (i.e., DAS =97). Yet there was no indication from the authors
of those studies that they were studying an enrichment population.
Clinical Implications
Clinical conclusions for mildly distressed couples are very tentative due to
the low sample size. The data show that any treatment would be helpful to
some extent. More research needs to be done once a clear definition of what
constitutes a mildly distressed couple.
Treatment for moderately distressed couples should have a clear treat-
ment plan that is followed. While EFT is showing significant results with mod-
erately distressed couples, other full treatment models were also significant.
Treatment that delivered only isolated components of BMT had treatment
gains substantially lower than other full model treatments. “Other” treatment
models (CMT, CBMT, and BMT Group) also approached being significantly
different from BMT Components (p =0.07). This was after controlling for the
assessment instrument used in the studies. More specifically it can be stated
that holding all other things equal, “Other” treatment approaches were as-
sociated with an additional 9-point jump on DAS scores from pre-test to
post-test beyond the treatment gains of BMT Components.
The difference between EFT and BMT Components in reference to con-
trol groups is intuitive as EFT offers a comprehensive treatment plan where
the BMT Component studies were focused on a single technique or interven-
tion. A group of interventions tied together have more potential for benefit
than a single intervention. This would be especially true for a moderately
distressed population. It is difficult to say whether including more studies in
the moderately distressed groups would confirm or disprove the differences
between EFT and the Mixed approaches.
The evidence of a clear treatment protocol approaching significance
in the treatment of moderate marital distress is more intuitive. The results
What Works for Whom 285
suggest that all other things being equal, a clear plan of treatment and follow
through with that plan, is associated with a 17-point DAS difference between
treatment and control groups at post-test.
FINAL THOUGHTS
This study was unique in many ways. It was the first to standardize outcome
measures across studies. The presence of a standard dependent measure
across studies made it possible to compare treatment gains of the treatment
groups across multiple studies. Treatment gains are more reflective of treat-
ment effectiveness as they focus solely on the magnitude of change of the
treatment group. Until the field as a whole adopts a standard dependent mea-
sure, these results represent the best attempt at comparing realized treatment
gains.
This study was also first in attempting to measure treatment effectiveness
by differing levels of marital distress. The data from this study more conclu-
sively showed that mild, moderate, and severely distressed couples should
not be aggregated. Differences in treatment approaches appeared once dis-
tress levels were operationalized and isolated. Through these data, the initial
attempts to match presenting complaint to treatment have been initiated.
In reference to what works for whom, any intervention for mildly dis-
tressed couples is better than no intervention. No one intervention stood
out against any other for mild distress. Moderately distressed couples should
receive a full treatment model rather than isolated components or interven-
tions. Emotionally Focused Therapy stood out in terms of treatment versus
control comparisons when compared to isolated BMT Components. “Other”
approaches stood out in terms of treatment gains when compared to isolated
BMT Components. Only BMT was used to treat a severely distressed group,
but to great success.
REFERENCES
Busby, D. M., Crane, D. R., Larson, J. H., & Christensen, C. (1995). A revision of
the Dyadic Adjustment Scale for use with distressed and nondistressed couples:
Construct hierarchy and multidimensional scales. Journal of Marital and Family
Therapy, 21, 289–308.
Butler, M. H., & Wampler, K. S. (1999). A meta-analytical update of research on
the couple communication program. American Journal of Family Therapy, 27,
223–237.
Carrol, J. S., & Doherty, W. J. (2003). Evaluating the Effectiveness of Premarital Pre-
vention Programs: A Meta-Analytic Review of Outcome Research. Family Rela-
tions: Interdisciplinary Journal of Applied Family Studies, 52, 105–118.
Crane, D. R., Allgood, S. M., Larson, J. H., & Griffin, W. (1990). Assessing marital qual-
ity with distressed and nondistressed couples: A comparison and equivalency
286 N. D. Wood et al.
table for three frequently used measures. Journal of Marriage and the Family,
52, 87–93.
Crane, D. R., Middleton, K. & Bean, R. (2000). Establishing criterion scores for
the Kansas Marital Satisfaction Scale and the Revised Dyadic Adjustment Scale.
American Journal of Family Therapy, 28, 53–60.
Crane, D. R., Soderquist, J. N., & Frank, R. L. (1995). Predicting divorce at marital
therapy intake: A preliminary model. American Journal of Family Therapy,23(3),
227–236.
Dunn, R. L., & Schwebel, A. I. (1995). Meta-analytic review of marital therapy out-
come research. Journal of Family Psychology, 9, 58–68.
Gurman, A. S., & Kniskern, D. P. (1978). Research on marital and family therapy:
Program, perspective, and prospect. In S. Garfield & A. Bergin (Eds.), Handbook
of psychotherapy and behavior change; An empirical analysis (pp. 820–821).
New York: Wiley.
Hahlweg K., & Markman, H. J. (1988). Effectiveness of behavioral marital therapy:
Empirical status of behavioral techniques in preventing and alleviating marital
distress. Journal of Consulting and Clinical Psychology, 56, 440–447.
Hazelrigg, M. D., Cooper, H. M., & Borduin, C. M. (1987). Evaluating the effectiveness
of family therapies; An integrative review and analysis. Psychological Bulletin,
101, 388–395.
Jacobson, N. S., Follette, W. C., & Revenstorf, D. (1984). Psychotherapy outcome
research: Methods for reporting variability and evaluating clinical significance.
Behavior Therapy, 15, 336–352.
Jacobson, N. S., Follette, W. C., Revenstorf, D., Boucom, D. H., Hahlweg, K., &
Margolin, G. (1984). Variability in outcome and clinical significance of behavioral
marital therapy: A reanalysis of outcome data. Journal of Consulting and Clinical
Psychology, 52, 497–504.
Jacobson, N. S., & Truax, P. (1991). Clinical significance: A statistical approach to
defining meaningful change in psychotherapy research. Journal of Consulting
and Clinical Psychology, 59, 12–19.
Lipsey, M. W., & Wilson, D. B. (2001). Practical meta-analysis. Thousand Oaks, CA:
Sage.
Locke, H. J., & Wallace, K. M. (1959). Short marital adjustment and prediction
tests: Their reliability and validity. Marriage and Family Living, 21, 251–
255.
Luborksy, L., Singer, B., & Luborksy, L. (1975). Comparative studies of psychother-
apies: Is it true that “Everybody has won and all must have prizes?” Clinical
Psychology: Science and Practice, 2, 106–109.
Oppenheim, D., Wambboldt, F., Gavin, L. A., Renouf, A. G., & Emde, R. N. (1996).
Couples’ co-construction of the story of their child’s birth: Associations with
marital adaptation. Journal of Narrative and Life History, 6, 1–21.
Schumm, W. R., Nichols, C. W., Schectman, K. L., & Grigsby, C. C. (1983). Charac-
teristics of responses to the Kansas Marital Satisfaction Scale by a sample of 84
married mothers. Psychological Reports, 53, 567–572.
Schumm, W. R., Paff-Bergen, L. A., Hatch, R. C., Obiorah, F. C. et al. (1986). Concur-
rent and discriminant validity of the Kansas Marital Satisfaction Scale. Journal
of Marriage & the Family, 48, 381–387.
What Works for Whom 287
Shaddish, W. R. (2002). The effects of marriage and family interventions: What can we
learn from meta-analyses? In D. Sprenkle (Ed.), Effectiveness research in mar-
riage and family therapy.Washington, D.C.: American Association for Marriage
and Family Therapy.
Shaddish, W. R. (1992). Do family and marital psychotherapies change what people
do? A meta-analysis of behavioral outcomes. In T. D. Cook, H. M. Cooper, D. S.
Cordray, H. Hartmann, L. V. Hedges, R. J. Light, T. A. Louis, & F. Mosteller (Eds.),
Meta-analysis for explanation: A casebook (pp. 129–208). New York: Sage.
Shaddish, W. R., Montgomery, L. M., Wilson, P., Wilson, M. R., Bright, I., & Okwum-
abua, T. (1993). Effects of family and marital psychotherapies: A meta-analysis.
Journal of Consulting and Clinical Psychology, 61, 992–1002.
Shadish, W. R., Ragsdale, K., Glaser, R. R. & Montgomery, L. M. (1995). The effi-
cacy and effectiveness of marital and family therapy: A perspective from meta-
analysis. Journal of Marital and Family Therapy, 21, 345–360.
Spanier, G. B. (1976). Measuring dyadic adjustment: New scales for assessing the
quality of marriage and similar dyads. Journal of Marriage and the Family, 38,
15–28.
Wampler, K. S. (1982). The effectiveness of the Minnesota Couple Communication
Program: A review of research. Journal of Marital and Family Therapy, 8, 345–
355.
Wampler, K. S., & Serovich, J. M. (1996). Meta-analysis in family therapy research. In
D. H. Sprenkle, & S. M. Moon (Eds.), Research methods in family therapy (pp.
286–307). New York: Guilford.
