Changing Network Support for Drinking: Initial Findings From the
Network Support Project
Mark D. Litt, Ronald M. Kadden, Elise Kabela-Cormier, and Nancy Petry
University of Connecticut Health Center
The aim of this study was to determine whether a socially focused treatment can effect change in the
patient’s social network from one that reinforces drinking to one that reinforces sobriety. Alcohol
dependent men and women (N ? 210) recruited from the community were randomly assigned to 1 of 3
outpatient treatment conditions: network support (NS), network support ? contingency management
(NS ? CM), or case management (CaseM; a control condition). Analysis of drinking rates for 186
participants at 15 months indicated a significant interaction effect of Treatment ? Time, with both NS
conditions yielding better outcomes than the CaseM condition. Analyses of social network variables at
posttreatment indicated that the NS conditions did not reduce social support for drinking relative to the
CaseM condition but did increase behavioral and attitudinal support for abstinence as well as Alcoholics
Anonymous (AA) involvement. Both the NS variables and AA involvement variables were significantly
correlated with drinking outcomes. These findings indicate that drinkers’ social networks can be changed
by a treatment that is specifically designed to do so, and that these changes contribute to improved
Keywords: alcoholism, social support, Alcoholics Anonymous (AA), cognitive–behavioral treatment,
It has long been suggested that an effective means to forestall
relapse in treated alcoholics is to alter the reinforcers for absti-
nence and drinking behavior in their home environment (e.g.,
Bigelow, Brooner, & Silverman, 1998). One way to achieve this is
to change the person’s social network. Steinglass and Wolin
(1974), among others, noted that the social milieu of an alcoholic
serves to support the drinking of those in the network.
Longabaugh and Beattie (Beattie & Longabaugh, 1999; Long-
abaugh & Beattie, 1986) coined the term “network support for
drinking,” referring to the degree to which people in the home
environment encourage drinking. Network support for drinking
has been found to be predictive of poor outcomes in treatment-
seeking patients (Havassy, Hall, & Wasserman, 1991; Long-
abaugh, Beattie, Noel, Stout, & Malloy, 1993; Project MATCH
Research Group, 1997).
Efforts to make existing social networks less supportive of
substance use have been proposed but have not been widely
adopted (e.g., Galanter, 1986, 1999). These efforts include the
community reinforcement approach (CRA), first proposed by Hunt
and Azrin (1973) and since refined (e.g., Meyers & Miller, 2001;
Sisson & Azrin, 1989), and the United Kingdom Alcohol Treat-
ment Trial Social Behaviour and Network Therapy study (UKATT
SBNT; Copello et al., 2002; UKATT Research Team, 2005). CRA
involves reinforcing the alcoholic’s sobriety and development of
activities incompatible with alcohol use, such as participation in
recreational and social activities and employment. Although CRA
is frequently cited as one of the most efficacious approaches to
treatment (e.g., Miller, Wilbourne, & Hettema, 2003), its use is not
widespread, possibly because of the time and effort required to
implement such a comprehensive intervention. Additionally, the
complexity of the intervention makes it difficult to determine what
features of CRA make it effective.
SBNT was an effort to identify, expand, and mobilize the social
network of drinkers, particularly family, friends, and acquaintan-
ces. In the UKATT Research Team’s (2005) study, both SBNT
and a comparison motivational enhancement therapy intervention
yielded significant decreases in drinking and in drinking-related
problems at 12 months, with no differences between treatment
conditions (UKATT Research Team, 2005). The authors did not
report whether SBNT acted on the social network as planned or if
changes in social network were related to outcomes.
The clearest example of an existing social network that supports
abstinence is Alcoholics Anonymous (AA). AA is a ready-made
sobriety-supporting network and fulfills several of the conditions
required of a behavioral choice model of relapse prevention
(Tucker, Vuchinich, & Gladsjo, 1990). AA provides alternative
activities to drinking, constrains access to alcohol (at least while
Mark D. Litt, Department of Behavioral Sciences and Community
Health, University of Connecticut Health Center; Ronald M. Kadden, Elise
Kabela-Cormier, and Nancy Petry, Department of Psychiatry, University of
Connecticut Health Center.
Portions of this article were presented at the International Conference on
the Treatment of Addictive Behaviors, Santa Fe, New Mexico, February
2006. Support for this project was provided by National Institute on
Alcohol Abuse and Alcoholism Grant R01-AA12827 and in part by Na-
tional Institutes of Health General Clinical Research Center Grant M01-
RR06192. We would like to acknowledge Eileen Porter, William Blakey,
Kara Dion, Abigail Sama, Nicole Paulin, Aimee Markward, Howard Stein-
berg, Patricia Gaupp, Anna Lane, and Christine Calusine for their work in
the conduct of this study.
Correspondence concerning this article should be addressed to Mark D.
Litt, Department of Behavioral Sciences and Community Health, MC3910,
University of Connecticut Health Center, Farmington, CT 06030. E-mail:
Journal of Consulting and Clinical Psychology
2007, Vol. 75, No. 4, 542–555
Copyright 2007 by the American Psychological Association
people are attending meetings), and provides a social group that
reinforces sober behavior. Several studies support the efficacy of
AA or similar groups in reducing alcohol use. Emrick, Tonigan,
and Montgomery (1993) concluded that AA members achieve
abstinence at a higher rate than do professionally treated alcoholics
and that AA participants who are more active in the fellowship
program fare better than less active participants. The findings of
AA studies are consistent with the idea that social support for
sobriety can enhance treatment outcome. Naturalistic studies by
Kaskutas and Bond (Bond, Kaskutas, & Weisner, 2003; Kaskutas,
Bond, & Humphreys, 2002) indicate that AA effects are partly
mediated by the changes that occur in patients’ social networks,
particularly changes in network support for drinking.
A difficulty with such naturalistic studies, however, is that
patients who actually attend AA are self-selected. Thus, it is not
possible to know what the true effect of AA is. To get an under-
standing of the true value of AA, researchers would have to
manipulate attendance at AA. To some extent, this was done in
Project MATCH, in which patients assigned to twelve-step facil-
itation (TSF) were encouraged to attend AA (Project MATCH
Research Group, 1993).
Analyses of the Project MATCH data set indicated that those
with networks supportive of drinking at intake had fewer percent-
age of days abstinent and more drinks per drinking day across all
follow-ups in that study (Longabaugh, Wirtz, Zweben, & Stout,
2001). A high level of network support for drinking was also
related to lower involvement in AA.
Network support for drinking also emerged as one of the more
interesting matching variables in Project MATCH. At the 3-year
follow-up, it was found that, among patients with high network
support for drinking at intake, those who had been assigned to the
TSF intervention had better outcomes than those assigned to
motivational enhancement therapy (Project MATCH Research
Group, 1998). Although network support for drinking per se was
not altered by TSF, causal chain analyses indicated that TSF did
result in greater involvement in AA, even among those with high
network support for drinking at intake. In contrast, for patients
whose social network at intake did not support continued drinking,
AA involvement had much less impact on outcome (Longabaugh,
Wirtz, Zweben, & Stout, 1998). The implication of these findings
is that a treatment that encourages a change of social network,
from one that is supportive of drinking to one that is supportive of
sobriety (e.g., by encouraging AA participation), may be effective,
especially for those whose pretreatment environments are initially
more supportive of drinking.
In the present article, the network support treatment is de-
scribed. Unlike SBNT, network support is more reliant on AA and
other established social infrastructures to change the social net-
work. AA attendance, as well as involvement in other nondrinking
networks, is manipulated by encouragement to attend sober net-
work functions (e.g., AA meetings, nondrinking social events) and
by reinforcing attendance in treatment.
Alcoholic patients drawn from the community were randomly
assigned to one of three treatment conditions: network support
(NS), network support ? contingency management (NS ? CM),
or a case management (CaseM) control condition. We hypothe-
sized that a NS intervention would result in more adaptive change
in social networks than would CaseM, and that such changes in the
social network would predict drinking outcomes.
We also hypothesized that direct reinforcement of network
change behaviors using contingency management procedures
would lead to greater (and faster) change in the social network, and
thereby more complete and lasting reductions in alcohol consump-
tion, than in the other conditions. Finally, we tested whether a NS
intervention would be most beneficial to those whose networks
were least supportive of abstinence at intake, consistent with the
finding in Project MATCH (Project MATCH Research Group,
1998; Zywiak, Longabaugh, & Wirtz, 2002). The present article
reports on results out to the 15-month follow-up.
Recruitment occurred from October 2002 through March 2005.
Participants were recruited through newspaper and radio adver-
tisements announcing free alcohol treatment and through other
research programs at our site, which is a university medical center.
To be eligible, individuals had to be at least 18 years old, meet
Diagnostic and Statistical Manual of Mental Disorders (4th ed.;
DSM–IV; American Psychiatric Association, 1994) criteria for
alcohol dependence or abuse, and be willing to accept random
assignment to any of the three treatment conditions. Individuals
were excluded from participation if they had acute medical or
psychiatric problems requiring inpatient treatment (e.g., acute psy-
chosis or suicide/homicide risk), current dependence on drugs
(except nicotine and marijuana), intravenous drug use in the pre-
vious 3 months, reading ability below the fifth-grade level, lack of
reliable transportation to the treatment site, or excessive commut-
ing distance. Individuals were also excluded if they were already
engaged in substance abuse treatment, if they denied any drinking
in the previous 60 days, or if they had attended more than three AA
meetings in the month prior to intake.
Of the 348 men and women who responded to advertisements
and were screened, 297 were eligible according to the criteria
described. Most (31) of the 51 excluded were already engaged in
some type of treatment. The remainder were excluded because of
other drug dependence or lack of transportation and/or stable
residence. Of the 297 initially considered eligible, 87 dropped out
of the study prior to randomization because of lack of interest or
concurrent drug abuse discovered after the baseline assessment but
before randomization. The remaining 210 participants were as-
signed to the three treatment conditions: NS (n ? 69), NS ? CM
(n ? 71), and CaseM (n ? 70). A sample size of 44 per cell was
determined to be sufficient to test all hypotheses (including the
interaction of Treatment ? Network Support Change) with a
power of .83 and alpha set at .05, on the basis of effect sizes
derived from previous studies of social network changes (Project
MATCH Research Group, 1997) and studies of contingency man-
agement procedures on treatment process measures (Petry, Martin,
Cooney, & Kranzler, 2000). Given the procedures used in each
treatment, participants, therapists, and research assistants could not
be blinded as to experimental condition. A diagram showing the
flow of participants through the recruitment, treatment, and
follow-up stages of the study is shown in Figure 1.
Participants were 58% male, with a mean age of 45 years (SD ?
11.4), and they were 86% White, 8% Black, 4% Hispanic, and 2%
other. Their mean years of schooling was 13.7 (SD ? 2.1), 71%
NETWORK SUPPORT PROJECT
were employed at least part time, and 51% were living with a
spouse or partner. All met criteria for alcohol dependence (99%) or
abuse (1%), drank on a mean of 72% of days in the 3 months prior
to intake, and averaged 1.3 prior treatments for alcohol depen-
dence (SD ? 3.3).
Measures and Instruments
our research center by telephone, at which point they were
screened for eligibility with a 20-min Quick Screen.
The Structured Clinical Interview for
DSM–IV Axis I Disorders, Patient Edition, Version 2.0 (First,
Spitzer, Gibbon, & Williams, 1996) was used to determine
whether participants met inclusion/exclusion criteria for alcohol
abuse or dependence, drug dependence, and psychotic symptoms
Individuals seeking treatment contacted
in the 90 days prior to the interview. Also included in the interview
was the Slosson Oral Reading Test (Slosson, 1963).
Drinking and drug use outcome data.
and drug use data at baseline and at follow-ups using the Form-90
(Miller & DelBoca, 1994). The Form-90 is a structured interview
that combines the calendar prompts of the time-line follow-back
method (Sobell & Sobell, 1992) with drinking pattern estimation
procedures, allowing the recording of drinking for each of the
pervious 90 days. The Form-90 has good test–retest reliability and
validity for verifiable events (Tonigan, Miller, & Brown, 1997).
Verification of self-reports of drinking.
check on self-reports, we took urine samples at baseline and at
posttreatment to screen for drug use. Breathalyzer readings were
taken at intake, at every in-person follow-up, and at every treat-
ment session in all intervention conditions. Collateral reports were
We collected drinking
To provide a gross
mtgs. ? Alcoholics Anonymous meetings; Sub. ? substance; Tx ? treatment; alc. ? alcohol; CaseM ? case
management; mo ? months; NS ? network support; NS ? CM ? network support ? contingency management.
Diagram showing flow of participants through each stage of study up to 15-month follow-up. AA
LITT, KADDEN, KABELA-CORMIER, AND PETRY
obtained for one third of participants (randomly selected) and used
to verify self-reports substance use. The level of agreement be-
tween collaterals and patients regarding drinking at posttreatment
was ? ? .62 (p ? .001).
The Drinker Inventory of Conse-
quences (DrInC; Miller, Tonigan, & Longabaugh, 1995) was used
to assess problems related to drinking, including health, legal
difficulties, and social relations. The DrInC Total score had an
internal reliability ? ? .85.
Treatment process variables: Network support.
network support for drinking and for abstinence using the Impor-
tant People and Activities instrument (IPA; Clifford & Long-
abaugh, 1991). The IPA consists of a structured interview that asks
patients to identify important people in their social network, de-
fined as those people with whom they spent the most time in the
previous 12 months. For each person identified, the patient spec-
ifies the nature of the relationship (e.g., spouse, brother, friend,
coworker), the duration of the relationship, the frequency of con-
tact, the drinking behavior of each person (frequency and quan-
tity), and the person’s behavior with respect to the patient’s drink-
ing (supportive, neutral, nonsupportive of drinking, or supportive
of abstinence). Five subscales developed for use in Project
MATCH (Project MATCH Research Group, 1993) were used in
the present study as the primary social network outcomes: Social
Support for Drinking, Behavioral Support for Drinking, Attitudinal
Support for Drinking, Behavioral Support for Abstinence, and
Attitudinal Support for Abstinence.
Social Support for Drinking was comprised of two summary
variables, the mean of the drinking status (from 1 ? abstainer to
5 ? heavy drinker) of the persons named in the participant’s social
network and the mean of those persons’ reactions to the partici-
pant’s drinking (from 1 ? left the room to 5 ? encouraged
drinking). Thus, those with networks containing more heavy drink-
ers, or drinkers who were more encouraging of drinking, were
defined as having greater Social Support for Drinking. Internal
reliability of the Social Support for Drinking variable at baseline
was ? ? .72. Behavioral Support for Drinking was the proportion
of people in the participant’s social network who were classified as
heavy drinkers (i.e., modeled drinking behavior). Attitudinal Sup-
port for Drinking was calculated by taking the mean of the reac-
tions to drinking of the top four persons on the participant’s list of
important people. Internal reliability of the Attitudinal Support for
Drinking variable at baseline was ? ? .70.
Behavioral Support for Abstinence was the proportion of people
in the participant’s social network who were abstinent. Attitudinal
Support for Abstinence was calculated by taking the mean of the
reactions to not drinking of the top four persons on the partici-
pant’s list of important people (with reactions scaled from 1 ? left
the room to 5 ? encouraged nondrinking). Internal reliability of
the Attitudinal Support for Abstinence variable at baseline was
? ? .68. The intercorrelations of all five of the network support
variables were in the range of .2–.6, indicating that these variables
represented related, but not redundant, constructs.
Level of involvement in AA was also considered an important
process variable. The AA Involvement Questionnaire (Tonigan,
Connors, & Miller, 1996) is a 16-item self-report inventory that
measures lifetime and recent attendance and involvement in AA
(e.g., acting as a sponsor). The AA Involvement Questionnaire
used here had a range from 0 to 5 and a reliability of ? ? .73.
Self-reports of number of AA meetings attended were also used as
Initial evaluation, informed consent, and identification of col-
Prospective participants were first evaluated through the
Quick Screen procedure and were either excluded (and referred
elsewhere for treatment appropriate to their needs) or scheduled
for an intake interview with a research assistant. The final decision
about eligibility was made at the intake interview, after completion
of the Structured Clinical Interview for DSM–IV Axis I Disorders
(to determine alcohol dependence or the other exclusionary diag-
noses). Those who were eligible and agreed to be randomly as-
signed to treatment reviewed and signed the informed consent
form and completed the intake assessment. One third of the par-
ticipants were asked to give the name of a collateral (usually a
spouse or close friend) who would be able to verify the partici-
pant’s level of drinking and his/her location.
Assignment to treatment.
We randomly assigned those who
agreed to participate to treatment using a computerized urn ran-
domization procedure (Stout, Wirtz, Carbonari, & Del Boca, 1994)
that balanced the three groups for gender, age, ethnicity, alcohol
dependence, and lifetime involvement with AA. Participants were
informed of their treatment assignment by a research assistant at
intake and subsequently were contacted by their therapist to sched-
ule their first treatment appointment.
Data collection procedures.
assistants conducted the pretreatment and follow-up research as-
sessments. In-person follow-up interviews were conducted at
Months 3 (posttreatment), 9, 15, 21, and 27. Assessments at
Months 6, 12, 18, and 24 were conducted by telephone. Partici-
pants were compensated $40 for attending the initial intake assess-
ment, $50 for each in-person follow-up assessment, and $20 for
the telephone follow-ups. At the 9-month and 15-month follow-
ups, 79% and 76% of the interviews were conducted in person,
respectively. Results presented in this article cover only through
the 15-month time period (1-year posttreatment).
Trained bachelor-level research
Treatment was conducted in 12 weekly 60-min outpatient ses-
sions, employing detailed therapist manuals. Any participant with
a breathalyzer reading above .05 was asked to wait until the level
declined or to return at a later date. Precautions were taken to
prevent those who were legally intoxicated or impaired from
driving a car, although this occurrence was rare.
The CaseM intervention was based on that used in the
Marijuana Treatment Project (Steinberg et al., 2002) and was
intended to provide an active control condition. During treatment,
problems were explored in several domains: psychiatric, interper-
sonal (family, childcare, and other social issues), medical, employ-
ment, educational, financial, housing, legal, and transportation.
The therapist and participant used a problem checklist to identify
problems that could be barriers to abstinence. Examples of goals
included contacting a psychiatrist for depression, finding a job, or
finding a new place to live.
After goals were selected, the participant and therapist identified
resources to address them, using a comprehensive guide to local
NETWORK SUPPORT PROJECT
services. The role of the therapist was to explore the relationship
between identified problems and drinking, monitor progress to-
ward goal attainment, and support participants’ efforts to reach
their CaseM goals. Efforts were made to minimize overlap with the
NS and NS ? CM treatments by avoiding explicit recommenda-
tions regarding social support or skills development. Attendance at
AA was neither encouraged nor discouraged for CaseM partici-
The NS intervention consisted of twelve one-
hour sessions intended to help the patient change his or her social
support network to be more supportive of abstinence and less
supportive of drinking. Because AA is a ubiquitous source of
social support and one that is already tapped by most treatment
services, it was thought that encouraging attendance at AA might
be an efficient way to quickly engage patients in a supportive
abstinence-oriented social network. Thus, the NS intervention was
based on the TSF treatment created for Project MATCH (Nowin-
ski, Baker, & Carroll, 1992). The program consisted of six core
sessions, plus six elective sessions that were chosen by the thera-
pist and the patient together. Core topics included Program Intro-
duction; Acceptance (of alcoholism as a problem); Surrender (giv-
ing up the idea of managing without help); Getting Active (start
changing the social network); People, Places, and Things (stimulus
control of drinking); and Termination.
Although the core topics were based on topics presented in
Project MATCH TSF, the emphasis was shifted to changing one’s
overall social support network. AA-specific philosophy and a
focus on a higher power were downplayed, and attendance at AA
was presented as a way to avoid drinking, make new acquaintan-
ces, and derive enjoyment (reinforcement) from activities other
than drinking. If a patient was adamantly opposed to attending AA,
the emphasis on AA was dropped. In all cases, other social
networks were also explored.
Sessions 2–11 began with a review of the Adherence Checklist,
a record of treatment-relevant activities, including AA meetings
and other nondrinking social activities engaged in. If a slip was
reported, the therapist worked with the participant to find a way to
use his/her social network to help avoid future slips (e.g., calling
an AA sponsor or other nondrinking friend, asking the nondrinking
spouse to help monitor activities). The session then proceeded to
new material, consisting of a core topic or an elective session.
Elective topics included the following: Genograms (importance of
family drinking); Enabling; Taking Moral Inventories; Sober Liv-
ing; HALT (dangers of being Hungry, Angry, Lonely, Tired);
Assertiveness Training; Increasing Pleasant Activities; and con-
joint sessions with the participant’s spouse or partner.
The end of the session entailed a discussion of Recovery Tasks
(homework). These took the form of going to AA meetings and
exploring ways to change one’s network of support. These tasks
included Education (obtaining information about a course at a
community college); Employment (e.g., searching for and apply-
ing for a job in a nondrinking environment); Family (e.g., family
outing or family therapy); and Social/Recreational (e.g., going to a
gym, reestablishing contact with nondrinking friends and rela-
NS ? CM.
Participants in this condition received the same NS
treatment as described above. In addition, reinforcements were
provided contingent upon completion of assigned tasks between
sessions. Verification of completion of tasks consisted of receipts
or signed slips with names and phone numbers listed for confir-
mation. The form of verification was specified each week on the
The contingency management portion of this condition was
adapted from Iguchi, Belding, Morrel, and Lamb (1997) and Petry
et al. (2000), and it used a fishbowl-drawing procedure for deter-
mining amount of reinforcement. Participants earned drawings
from the fish bowl if they offered verification of completed as-
signments related to developing nondrinking social support net-
works. Each drawing provided the opportunity to accumulate slips
that could be redeemed for services or merchandise.
The drawing fish bowl contained 500 slips of paper. Half of
these slips were nonwinning slips that read, “Sorry, try again.” Of
the winning slips, 199 were for small prizes (e.g., $1 coupons for
local merchants, $1 fast food gift certificates, or bus tokens). Of
the winning slips, 50 specified prizes worth approximately $20 in
value, such as radios, art and craft supplies, and gift certificates.
One of the 500 slips could be redeemed for a prize worth about
$100 (e.g., a hand-held television, a portable stereo CD player) or
five $20 prizes. Participants were made aware of their chances of
picking a winning slip on each trial, and they were encouraged to
make suggestions for prizes in all three categories.
Participants earned at least one drawing per verified activity
completed. If they accomplished three activities in a given week,
the number of drawings earned for each activity escalated by one
for each successive week that they accomplished three activities.
Participants could thus earn up to two drawings per activity com-
pleted in Week 2 (total ? six draws for Week 2), up to three in
Week 3 (total ? nine draws for Week 3), and so on. During the 12
weeks of study participation, they could earn a maximum of 234
Failure to attend a scheduled treatment session without giving
notice, or failure to provide verification for activity completion,
resulted in loss of drawing opportunities for that week, and the
number of drawings earned was reset back to one drawing per
activity verified. The number of earned drawings was restored to
the prior level after verification of three completed activities at
three successive sessions. Participants were encouraged to provide
honest reports of alcohol use and were assured that drawings
would be earned as long as contracted activities were accom-
plished and verified, regardless of alcohol use.
Detailed, step-by-step outlines were provided for each treatment
session. Therapists were required to follow the session outline and
check off areas covered as the session proceeded. The requirement
to adhere to the session outlines and to record progress reduced the
likelihood of extraneous or tangential discussions and ensured that
required material would be covered. Additionally, each treatment
session was audiotaped. The Project Coordinator (PC) listened to
all session tapes for the therapist training cases (36 tapes per
therapist) and provided weekly supervision. Thereafter, the PC
listened to 33% of session tape recordings and independently
confirmed that session outlines were followed and that elements of
treatment were being delivered. The PC also verified that elements
unique to one treatment were not employed in the other treatments.
Supervision was provided biweekly and covered both clinical
issues and treatment adherence.
LITT, KADDEN, KABELA-CORMIER, AND PETRY
Multiple chi-square tests and one-way analyses of variance were
used to verify that the three treatment groups were equivalent with
respect to the balancing variables: gender, age, ethnicity, alcohol
dependence, and current involvement with AA, as well as other
relevant sample variables (education level, marital status, employ-
ment, and DrInC Total score). There were no significant between-
groups differences on any variable.
Of the 210 participants randomized to treatment, 17 (8%)
dropped out of the study prior to completion of treatment and
provided no posttreatment data. By the 15-month follow-up, an-
other 7 had dropped out, and the remaining participants were
distributed among the treatment conditions as follows: CaseM
(n ? 64), NS (n ? 63), and NS ? CM (n ? 59). Analyses
indicated no differences in any patient characteristic by treatment
condition among those who completed the 15-month follow-up.
These 186 participants provided the data used in the intention-to-
treat analyses described below.
Adherence to Treatment
We evaluated session attendance and adherence to treatment
assignments using one-way analysis of variance. Over all three
conditions, participants attended 8.7 sessions of 12 (SD ? 4.2).
There were no significant differences in number of sessions at-
tended as a function of treatment condition: CaseM, 9.2 sessions
(SD ? 3.8); NS, 8.3 sessions (SD ? 4.4); and NS ? CM, 8.6
sessions (SD ? 4.5); F(2, 190) ? 0.69, p ? .50.
In the two NS conditions, therapists judged whether each as-
signed activity was completed. Patients in the NS condition com-
pleted 83.2% of assignments, and those in NS ? CM completed
90.2% (Z ? 1.17, p ? .25). Patients in the NS ? CM condition
earned on average 56 draws and redeemed $250.00 worth of
prizes. No adherence calculation could be made for the CaseM
condition because of the looser behavioral requirements in that
Treatment Effects on Outcome
The primary drinking outcome variables derived from the
Form-90 were Proportion of Days Abstinent (PDA) and Continu-
ous Abstinence for the 90-day period prior to each follow-up. The
PDA data were arcsine transformed to decrease the inherent skew-
ness of proportion data (Winer, 1971). The DrInC Total score was
the primary psychosocial dependent variable examined.
Linear mixed modeling (Proc MIXED; SAS Institute, 1999) was
used to analyze the transformed PDA outcome variable over time
as a function of treatment condition. Main effects for Treatment
and Time (from baseline to 15 months) were examined, as well as
two planned contrasts comparing (a) both NS conditions versus the
CaseM condition (i.e., CaseM vs. NS, NS ? CM), and (b) NS
versus NS ? CM (i.e., NS vs. NS ? CM). The mixed modeling
procedure was used because it employs maximum likelihood es-
timation to calculate parameter estimates and thus allowed us to
take advantage of all data collected. In this analysis, Treatment
condition was treated as a fixed effect. Both Time and intercept
were included as random effects. An unstructured covariance
structure was adopted on the basis of accepted fit criteria (?2
restricted log-likelihood, Akaike Information Criterion; Judge,
Griffiths, Hill, Lutkepohl, & Lee, 1985).
The analysis of PDA (transformed) through 15 months indicated
no main effect for Treatment, F(2, 931) ? 1.60, p ? .05; a
significant effect for Time, F(5, 931) ? 123.66, p ? .001; and a
significant Treatment ? Time interaction, F(10, 934) ? 2.13, p ?
.05. Analysis of planned Time ? Treatment contrasts indicated
that the two NS conditions yielded significantly greater PDA than
did the CaseM control condition at follow-ups, Fcontrast(1, 934) ?
3.15, p ? .05, but that the two NS conditions did not differ from
each other. The treatment effect contrast NS, NS ? CM versus
CaseM had a moderate effect size (average d ? 0.41). Pretreat-
ment, posttreatment, and follow-up levels of PDA (detransformed)
by treatment condition are shown in Panel A of Figure 2 to
illustrate the magnitude of the treatment effects.
Figure 2, Panel B shows the levels of continuous abstinence
reported for the 90 days prior to each follow-up period. As seen in
the figure, abstinence rates in the NS condition reached 40% at 15
months. A generalized estimating equations (Proc GENMOD;
SAS Institute, 1999) model was used to analyze the effect of
treatment condition on continuous abstinence prior to each of the
follow-up points. Pretreatment PDA (transformed) served as a
covariate. The analysis yielded no significant main effect for
Treatment condition, or for Time, and there was no Treatment ?
Time interaction. An a posteriori contrast comparing the NS con-
ditions versus CaseM, however, was significant, ?2(1, N ? 210) ?
4.27, p ? .05. This contrast had a moderate effect size of h ? .31.
The two NS conditions did not differ from each other.
Figure 2, Panel C, shows the mean DrInC negative conse-
quences scores by treatment condition for each of the follow-up
points in which the measure was administered. As can be seen in
the figure, reports of negative consequences declined over time for
patients in all treatments. A linear mixed model analysis was used
to examine differences in DrInC Total score as a function of
treatment condition over the 15 months of follow-up. PDA at
baseline (transformed) served as a covariate. Results indicated no
main effect for Treatment, F(2, 483) ? 1.33, p ? .05; a significant
effect for Time, F(3, 483) ? 102.89, p ? .001; and no significant
Treatment ? Time interaction, F(6, 483) ? 1.25, p ? .05. Planned
Time ? Treatment contrasts also failed to show systematic differ-
ences in DrInC scores attributable to Treatment condition.
Treatment Effects on Social Network Change
Linear mixed models with planned contrasts were used to de-
termine the effect of treatment over time on the five IPA subscales.
As seen in Table 1, no main effects for Treatment emerged in the
analyses of any of the social network variables tested. Effects for
Time, and/or for Treatment ? Time, were found for four of the
variables: Social Support for Drinking, Behavioral Support for
Drinking, Behavioral Support for Abstinence, and Attitudinal Sup-
port for Abstinence. Examination of least-squared means sug-
gested that Social Support for Drinking decreased significantly in
all treatment conditions from baseline to posttreatment and re-
mained low at the 15-month follow-up point.
Three variables showed Treatment ? Time interactions. Scores
on Behavioral Support for Drinking decreased more steeply in the
two NS conditions from pre- to posttreatment, whereupon scores in
NETWORK SUPPORT PROJECT
all treatments tended to increase somewhat by 15 months (see
Figure 3, Panel A). The two variables assessing support for absti-
nence increased from pre- to posttreatment differentially by Treat-
ment and remained elevated through the follow-ups (see Figure 3,
Panels B and C). The Treatment Contrasts ? Time Analyses (see
Table 1) revealed that support for abstinence increased among
those in the two NS conditions but not among those treated in
CaseM. There was no difference between the NS and NS ? CM
management; Net Support ? network support; NS ? ContM ? network support ? contingency management;
Tx ? treatment; Mo ? months; DrInC ? Drinker Inventory of Consequences.
Effects of treatment on drinking outcomes and on drinking consequences. Case Mgmnt ? case
LITT, KADDEN, KABELA-CORMIER, AND PETRY
conditions. For both of the abstinence support variables, the effect
size of the NS, NS ? CM versus CaseM contrasts was approxi-
mately d ? 0.70.
Examination of the social networks confirmed the findings
regarding increase in support for abstinence in the NS conditions.
Networks were examined for number of Heavy Drinking Friends
(e.g., those who drank at least everyday) and Nondrinking Friends.
Friends were people with whom the patient associated at least once
per week. Linear mixed model analyses indicated that the number
of Heavy Drinking Friends remained constant (at about 1.3) in all
conditions over all time periods. The number of Abstinent Friends,
however, showed a Treatment ? Time effect, F(4, 207) ? 2.39,
p ? .05, such that number of Abstinent Friends increased over
time for those in the NS conditions (from 1.2 at baseline to 1.5 at
15 months) versus the CaseM condition, which showed no in-
crease, Fcontrast(1, 207) ? 2.10, p ? .05.
Given the emphasis on AA in the NS interventions, a linear
mixed model like that just described was performed on the base-
line, posttreatment, and 15-month follow-up AA Involvement
scores. Examination of pretreatment and follow-up least squared
means indicated that AA Involvement actually decreased among
the CaseM participants (from 2.2 to 1.0), whereas AA Involvement
increased in the two NS conditions (from about 2.0 to 2.5). The
analysis showed a significant main effect for Treatment, F(2,
328) ? 5.28, p ? .01; no Time effect; and a significant Treat-
ment ? Time interaction, F(4, 328) ? 5.80, p ? .001. Time ?
Treatment contrasts indicated that the two NS conditions yielded
significantly greater AA Involvement scores at posttreatment than
did the CaseM condition, F(1, 328) ? 6.75, p ? .01, but the two
NS conditions did not differ from one another. The effect size for
the NS, NS ? CM versus CaseM contrast was moderate to large,
d ? 0.68.
We analyzed AA attendance using a linear mixed model, like
those described above. Results indicated a main effect for Treat-
ment, F(2, 330) ? 3.37, p ? .05; a main effect for Time, F(2,
330) ? 24.96, p ? .001; and a significant Treatment ? Time
interaction, F(4, 330) ? 2.35, p ? .05. AA attendance in the NS
conditions averaged over 25 meetings in the prior 90 days at the
15-month point, versus just over 8 for CaseM participants, result-
ing in a significant contrast of NS, NS ? CM versus CaseM ?
Time, F(1, 330) ? 8.61, p ? .01. The effect size of this contrast
was d ? 0.80. The two NS conditions did not differ from each
An additional analysis examined whether patients attended AA
at all during the treatment period as a function of treatment
condition. Of the CaseM participants, 18% attended at least one
meeting during the treatment period, as compared with 67% of NS
patients and 56% of NS ? CM patients. A logistic regression
analysis that used attendance (yes–no) as the outcome, and that
controlled for lifetime AA attendance, indicated that those in either
of the NS conditions were over 7 times more likely to attend AA
than those in CaseM (B ? 2.04, SE ? 0.39, Wald ?2? 26.80, p ?
.001, odds ratio ? 7.71). There were no differences between the
two NS conditions.
As noted above, not all patients attended AA, even though AA
attendance was encouraged at every NS session. Of the more than
3,100 network support activities completed by patients, 26% were
accounted for either by AA attendance or AA-related activity (e.g.,
arranging a sponsor). The rest of the activities completed consisted
of Family Activities (e.g., family outings, church, activities with
children; 27%), Other Social (e.g., going to gym, walks or lunches
with nondrinking friends, activities with church groups; 30%), or
Individual-Nondrinking (completing job applications or going on
job interviews, attending classes; 17%).
Network Change and Treatment Outcome
A series of partial correlations was performed to evaluate the
influence of change in each of the network support variables on
outcome at posttreatment through 15 months (see Table 2). For
each partial correlation, the baseline value of both the support
variable and the outcome variable was controlled for. As the
correlations indicate, change in network support for drinking
tended to be only weakly associated with treatment outcome.
However, changes in support for abstinence, and involvement with
AA, were significantly related to improvements in abstinence at
Results of Repeated Measures Mixed Model Analyses of Variance on Network Support Variables at Posttreatment and 15-Month
(dfs ? 2, 206)
(dfs ? 1, 206)
Time ? Treatment
(dfs ? 2, 206)
Time ? Treatment Contrast Results
(dfs ? 1, 206)
NS, NS ? CM vs. CaseMNS vs. NS ? CM
Social Support for Drinking
(range ? 2–10)
Behavioral Support for Drinking
(range ? 0–1)
Attitudinal Support for Drinking
(range ? 1–7)
Behavioral Support for Abstinence
(range ? 0–1)
Attitudinal Support for Abstinence
(range ? 1–7)
*p ? .05.
The values shown are F values for each effect. NS ? network support; NS ? CM ? network support ? contingency management; CaseM ? case
***p ? .001.
NETWORK SUPPORT PROJECT
posttreatment and follow-up. Of the network change variables
examined, only increase in AA attendance was consistently asso-
ciated with decreases in DrInC scores.
To determine whether network change was mediating the rela-
tionship between treatment and outcome, we examined a series of
hierarchical linear models, with PDA (transformed) through 15
months as the dependent variable. As before, time and intercept
were added as random effects, and an unstructured covariance
structure was adopted.
In the first model, baseline values of all IPA network support
variables that showed change from pre- to posttreatment were
entered (i.e., Behavioral Support for Drinking, Behavioral Support
People and Activities instrument [IPA]). Case Mgmnt ? case management; Net Support ? network support;
NS ? ContM ? network support ? contingency management; Tx ? treatment; Mo ? months.
Effects of treatment on selected measures of social network change (measured with the Important
LITT, KADDEN, KABELA-CORMIER, AND PETRY
for Abstinence, and Attitudinal Support for Abstinence). In addi-
tion, AA attendance at baseline was also entered. In the second
model, all variables from Model 1 were entered, plus a term
representing the treatment contrast (CaseM vs. NS, NS ? CM)
was entered. In the third model, all variables from Model 2 were
entered, as well as posttreatment values of the network support
variables and AA attendance at posttreatment. (Because levels of
network support did not generally increase substantially after the
posttreatment period, the increases from pre - to posttreatment
were used to represent the effects of changes in network support.)
If the addition of the network support variables resulted in the loss
of predictive power of the Treatment condition variable, it would
indicate that network support was a mediator of the relationship
between treatment and outcome (Baron & Kenny, 1986). Results
of these analyses are seen in Table 3.
Examination of Table 3 indicates that of the variables entered in
Model 1, only Time was a significant predictor of PDA outcome.
In Model 2, the Treatment Contrast ? Time interaction was
significant. However, when posttreatment values of the network
support variables were added in Model 3, the Treatment Con-
trast ? Time interaction term was no longer significant, though the
support-for-abstinence variables were significant. This result sug-
gests that changes in Behavioral Support for Abstinence and in AA
attendance from pre- to posttreatment were true mediators of the
NS, NS ? CM versus CaseM treatment effect.
Finally, we also examined the change in the nature of the social
network and its relationship to outcomes. Patients (37 of 186 total
available patients) were classed as very successful if they reported
abstinence on at least 90% of days at all follow-ups. Of the very
successful patients, the number of nondrinking close friends in the
Partial Correlations of Posttreatment Network Support Variables With Treatment Outcome Variables
Posttreatment network support
PDA Continuous AbstinenceDrInC Total score
Social Support for Drinking
Behavioral Support for Drinking
Attitudinal Support for Drinking
Behavioral Support for Abstinence
Attitudinal Support for Abstinence
AA meetings attended last 90 days
Consequences; mo ? months; AA ? Alcoholics Anonymous.
aDegree of freedom (df) ? 179.
*p ? .05.
Baseline values of network support outcome variables are partialled. PDA ? Proportion of Days Abstinent; DrInC ? Drinker Inventory of
bdf ? 164.
cdf ? 163.
**p ? .01.
***p ? .001.
Results of Multiple Hierarchical Linear Regression Analyses Showing Influence of Network
Support and Treatment Condition Variables on PDA Outcome Over 15 Months Posttreatment
ModelVariables in model
1IPA Behavioral Support for Drinking—Baseline
IPA Behavioral Support for Abstinence—Baseline
IPA Attitudinal Support for Abstinence—Baseline
AA attendance lifetime
All baseline variables entered
Treatment contrast (NS, NS ? CM vs. CaseM)
Treatment Contrast ? Time
All baseline variables entered
Treatment Contrast (NS, NS ? CM vs. CaseM)
Treatment Contrast ? Time
Behavioral Support for Drinking—Posttreatment
Behavioral Support for Abstinence—Posttreatment
Attitudinal Support for Abstinence—Posttreatment
AA attendance past 90 days—Posttreatment
Alcoholics Anonymous; NS ? network support; NS ? CM ? network support ? case management; CaseM ?
*p ? .05.
PDA ? Proportion of Days Abstinent; IPA ? Important People and Activities instrument; AA ?
***p ? .001.
NETWORK SUPPORT PROJECT
social network increased from a mean of 1.3 (SD ? 1.1) at baseline
to a mean of 1.9 at 15 months (SD ? 1.0). The number of heavy
drinking close fiends among these patients remained steady (M ?
approximately 1.7, SD ? 1.1). We used logistic regression anal-
yses to predict success status using baseline and posttreatment
numbers of heavy drinking and nondrinking friends in the social
network. Change in number of heavy drinkers in the social net-
work was not predictive of success status. However, change in the
number of nondrinkers in the social network from baseline to
posttreatment was a significant predictor of success status (B ?
0.22, SE ? 0.18, Wald ?2? 2.02, p ? .05), such that an increase
of one nondrinking friend in the social network translated into a
27% increase in the probability of being very successful by our
Tests of Differential Efficacy of Network Support
To determine whether network support treatment was differen-
tially effective for those who were higher in baseline Support for
Drinking (or lower in Support for Abstinence) at intake, we ex-
amined linear mixed model analyses of PDA (transformed) over
time. In each of these analyses, the following variables were
included in the model: the Social Support variable (Social Support
for Drinking or Behavioral Support for Abstinence); a term rep-
resenting the primary Treatment contrast (CaseM vs. NS, NS ?
CM); the Time variable; and the interaction terms of Support ?
Treatment, and Support ? Treatment ? Time. In each of the
analyses, the interaction terms involving the Support Variable ?
Treatment Contrast failed to reach significance (p values ranged
from .16 to .94). These results suggest that the NS intervention was
not differentially effective for those higher in initial social support
Network Support Treatment and Gender Effects
Women were well-represented in our patient sample. We were
therefore able to analyze treatment effects on outcome by gender.
A linear mixed model regression analysis was conducted in which
PDA (transformed) from baseline to the 15 month follow-up point
constituted the repeated dependent variable. Terms entered were
patient gender, Treatment condition, Time in months, and the
interactions of each of these. Subject and Time were treated as
random effects. Gender was treated as a fixed effect. An autore-
gressive covariance matrix was used to model the repeated data.
In this analysis, the main effect for Treatment remained signif-
icant, F(2, 919) ? 3.31, p ? .05, as did the Time effect, F(5,
919) ? 108.67, p ? .001. However, a main effect for Gender also
emerged, F(1, 919) ? 5.29, p ? .05, such that, overall, men were
seen to have better outcomes than women. A Gender ? Treatment
interaction, F(2, 919) ? 5.70, p ? .01, indicated that women fared
better than men did in CaseM but less well than men in the NS
conditions, by a margin of about 10%. We obtained similar results
with an analysis of repeated abstinence outcomes using a gener-
alized estimating equations model. Significant Time and Treat-
ment effects emerged, as did a main effect for Gender, ?2(1, N ?
210) ? 10.39, p ? .001, and an interaction of Gender ? Treat-
ment, ?2(2, N ? 210) ? 6.29, p ? .05.
One possibility for the relatively poorer performance of women
in the NS conditions was that women might have already had
sufficiently supportive networks and that an intervention seeking
to change that might have been disruptive. A multivariate analysis
of variance was therefore performed in which the network support
variables measured at baseline were examined as a function of
gender. Results indicated no overall significant differences in
social network supportiveness between men and women at this
baseline point (Wilks’ ? ? .96, p ? .17). Examination of network
support variables from baseline through 15-month follow-up, how-
ever, did show a systematic difference in mixed model regression
analyses. Throughout the trial, women reported significantly
higher Social Support for Drinking than did men, regardless of
treatment condition, Fgender(1, 203) ? 5.25, p ? .05.
The results of this study indicate that a treatment specifically
designed to change the social network can effect beneficial
changes. Although behavioral support for drinking decreased in all
conditions from pre- to posttreatment, only the NS treatment
conditions appeared to result in increased support for abstinence.
Contrary to expectations, the NS conditions did not yield decreases
in Social Support for Drinking relative to the CaseM condition.
Social Support for Drinking was a composite variable made up of
the number of drinking friends and their reaction to drinking
(ranging from nonsupportive to supportive). Examination of the
number of people in each patient’s network indicated that the
number of drinking friends remained constant, but that the number
of abstinent people in the network increased from pretreatment
through the follow-ups in the NS Conditions. In retrospect this
result might have been expected; the NS treatments were focused
on expanding abstinence-oriented relationships but did not specif-
ically discourage established relationships with drinking friends.
This result is somewhat at odds with those of Humphreys and
Noke (1997), who noted that social network size tends to remain
constant but that composition changes. In the present study, how-
ever, the abstinence-based network had to increase by only one
person to make a significant difference.
As expected, attendance at, and involvement with, AA played a
large role in support for abstinence. Posttreatment AA variables
correlated with posttreatment Support for Abstinence variables in
the range of .23–.33 (ps ? .001). This would suggest that, in many
instances, AA was indeed being used as an additional social
network. (That is, numbers of abstinent people in the social net-
work increased as AA involvement increased.) However, it should
also be noted that, despite our urgings and incentives to attend AA,
39% of NS patients never attended an AA meeting during the
treatment period. For those patients who reported increases in
Support for Abstinence and did not attend AA, the NS treatments
may have operated by effecting changes in other areas (new
friendships, new activities). Indeed, the NS participants who never
attended AA tended to fare slightly less well than AA attendees in
the later follow-ups but not significantly so. For example, at 15
months AA attendees in the NS conditions were reporting 70%
days abstinent versus 62% for nonattendees.
As hypothesized, the NS conditions resulted in better posttreat-
ment outcomes than did the CaseM condition. The CaseM condi-
tion was intended to control for many of the aspects of treatment
LITT, KADDEN, KABELA-CORMIER, AND PETRY
that have been termed “nonspecific” but are perhaps better termed
“unspecified.” These factors include therapist attention and sup-
port, the addition of structure to one’s weekly schedule, and the
effect of simply being in a clinical trial. The CaseM condition
resulted in roughly a 25% increase in PDA, a 5%–10% increase in
continuous abstinence, and a 10-point drop in DrInC scores. The
NS conditions resulted in more than a doubling of those effects.
Much of that improvement can be attributed to improvements in
Social Network variables.
AA involvement (including attendance) was a large part of the
improvement in social networks, and it was also a large contributor
to outcomes. As can be seen in Table 2, increases in AA involve-
ment variables were the strongest predictors of outcomes, and the
only significant predictors of improvement in drinking conse-
quences. As has been frequently noted (e.g., Moos & Moos, 2005),
AA attendance is associated with improved outcomes in virtually
all cases. Those who choose to attend AA, however, remain a
self-selected subgroup of patients. The 61% overall attendance rate
in the NS conditions versus the 18% in CaseM suggests that
participants can be “encouraged to self-select.” In any case, AA in
this study seems to have acted as described by Humphreys, Man-
kowski, Moos, and Finney (1999), who reported that AA operated
by expanding social networks.
The hypothesis that those who had social networks most sup-
portive of drinking would most benefit from NS treatment, as
found in Project MATCH (Project MATCH Research Group,
1998), failed to be supported. One possibility for this is that the
Project MATCH sample may have been more severe or more
socially isolated than the current sample. However, an examination
of the characteristics of the current sample indicates that it is quite
comparable with the Project MATCH outpatient sample in terms
of demographics and treatment history. Similarly, in terms of
Social Support for Drinking, the Project MATCH sample reported
a mean score of 6.2 (SD ? 1.08) at intake versus a mean of 5.8
(SD ? 1.06) for the current sample. Thus, it seems unlikely that
sample differences account for the differences in findings.
Another possibility seems more likely. The Project MATCH
findings (Project MATCH Research Group, 1998) were not dis-
covered until 3 years after treatment. It may take a considerable
amount of time for alcohol dependent patients to adopt a different
view of what their social network should look like. The Network
Support Project will continue to follow the current sample for
another year, during which further social network changes may
A notable, and consistent, finding in the current study was the
failure of contingency management to yield either more therapy-
related behavior change or improved outcomes. This failure to find
an effect may be related to the target for reinforcement. In most
studies of contingency management for drug use, for example, the
target for reinforcement is the delivery of a negative urine sample
(Petry, 2000), thus directly reinforcing abstinence. Reinforcing
abstinence directly may engender better substance use outcomes
than reinforcing alternate behavior patterns, a finding that was also
noted in a recently completed study of contingency management
for illicit substance abusers (Petry et al., 2006).
The lack of beneficial effects of the contingency management
procedure in the present study may also relate to characteristics of
the population being treated or to the treatment itself. In this study,
patients completed over 83% of their weekly assignments even
when no reinforcement was provided (i.e., in the NS alone condi-
tion). This high level of adherence may have resulted in a ceiling
effect, leaving little room for improvement for contingency man-
agement. Indeed, the NS ? CM condition in this study did not
significantly improve compliance with weekly activities. In con-
trast, among illicit substance dependent patients, compliance levels
with activities tend to be substantially lower, with only about 66%
of activities completed even when reinforcement for completion is
provided (Petry et al., 2006; Petry, Tedford, & Martin, 2001) .
The present study has some limitations. Most notably, there was
no active coping skills-based treatment against which to compare
the NS treatments. Consequently, there is no way to determine
whether NS is superior to other treatments, such as cognitive–
behavioral treatment. It should be noted, however, that the CaseM
condition in this study was manualized and provided the same
degree of attention and task involvement as the other active treat-
ments. Given the levels of attendance and satisfaction seen in this
study (the mean satisfaction ratings ranged from 4.5 to 4.7 of 5 for
all treatments), it appears that CaseM was seen as an active and
viable treatment, and as such was a useful control condition.
We do not yet know what aspects of “support for abstinence”
are truly responsible for improvements in outcomes. One possibil-
ity is that simply spending time with abstinent persons takes time
away from drinking. However, the fact that associations with
persons supporting abstinence and AA attendance remained ele-
vated throughout the follow-up period indicates that patients may
find contact with abstinence-oriented individuals reinforcing.
Although we were able to achieve an impressive level of at least
minimal AA involvement, over one third of our patients refused to
attend AA. Thus, the self-selection problem in the interpretation of
AA outcomes is still with us to some extent. Additional explora-
tion of the differences between those who attend AA and those
who do not, and the strategies they each use to control drinking,
will be needed.
Finally, the analysis of outcomes by gender actually showed that
women fared less well in the NS conditions than did the men. To
say that this was unexpected is an understatement. The result is
certainly at odds with findings like those by Davis and Jason
(2005), who found that, among those in a recovery community,
women, and not men, used their social networks to increase their
sense of self-efficacy for abstinence, and that social support for
abstinence mediated the relationship between treatment and out-
come. The finding that baseline support variables were equivalent
between men and women suggests that it was not the case that the
NS treatments were disruptive of adaptive support networks al-
ready in place for women. To the contrary, women tended to have
less adaptive networks over time than did the men. It may be that
the NS conditions encouraged these women to rely more on these
less adaptive networks.
In summary, the present study indicates that a manualized
treatment focused specifically on changing the social environment
can indeed alter the social network of alcohol dependent patients,
and that adaptive social network changes are predictive of drinking
outcome, at least in the short term. It remains to be seen whether
this particular treatment approach will in the long run be better
suited to those with impoverished social networks, as would be
predicted from previous research.
NETWORK SUPPORT PROJECT
American Psychiatric Association. (1994). Diagnostic and statistical man-
ual of mental disorders (4th ed.). Washington, DC: Author.
Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable
distinction in social psychological research: Conceptual, strategic, and
statistical considerations. Journal of Personality and Social Psychology,
Beattie, M. C., & Longabaugh, R. (1999). General and alcohol-specific
social support following treatment. Addictive Behaviors, 24, 593–606.
Bigelow, G. E., Brooner, R. K., & Silverman, K. (1998). Competing
motivations: Drug reinforcement vs. non-drug reinforcement. Journal of
Psychopharmacology, 12, 8–14.
Bond, J., Kaskutas, L. A., & Weisner, C. (2003). The persistent influence
of social networks and Alcoholics Anonymous on abstinence. Journal of
Studies on Alcohol, 64, 579–588.
Clifford, P. R., & Longabaugh, R. (1991). Manual for the Important People
and Activities instrument. Providence, RI: Brown University, Center for
Alcohol and Addiction Studies.
Copello, A., Orford, J., Hodgson, R., Tober, G., Barrett, C., & the UKATT
Research Team. (2002). Social behaviour and network therapy: Basic
principles and early experiences. Addictive Behaviors, 27, 345–366.
Davis, M. I., & Jason, L. A. (2005). Sex differences in social support and
self-efficacy within a recovery community. American Journal of Com-
munity Psychology, 36, 259–274.
Emrick, C. D., Tonigan, J. S., & Montgomery, H. (1993). Alcoholics
Anonymous: What is currently known? In B. S. McCrady & W. R.
Miller (Eds.), Research on Alcoholics Anonymous: Opportunities and
alternatives (pp. 41–76). Piscataway, NJ: Rutgers Center of Alcohol
First, M. B., Spitzer, R. L., Gibbon, M., & Williams, J. B. W. (1996).
Structured Clinical Interview for DSM–IV Axis I Disorders—Patient
Edition (SCID-I/P, Version 2.0). New York: New York State Psychiatric
Institute, Biometrics Research Department.
Galanter, M. (1986). Social network therapy for cocaine dependence.
Advances in Alcohol and Substance Abuse, 6, 159–175.
Galanter, M. (1999). Network therapy for alcohol and drug abuse. New
York: Guilford Press.
Havassy, B. E., Hall, S. M., & Wasserman, D. A. (1991). Social support
and relapse: Commonalties among alcoholics, opiate users, and cigarette
smokers. Addictive Behaviors, 16, 235–246.
Humphreys, K., Mankowski, E., Moos, R., & Finney, J. (1999). Enhanced
friendship networks and active coping mediate the effect of self-help
groups on substance abuse. Annals of Behavioral Medicine, 21, 54–60.
Humphreys, K., & Noke, J. M. (1997). The influence of posttreatment
mutual help group participation on the friendship networks of substance
abuse patients. American Journal of Community Psychology, 25, 1–16.
Hunt, G. M., & Azrin, N. H. (1973). A community reinforcement approach
to alcoholism. Behavior Research and Therapy, 11, 91–104.
Iguchi, M. Y., Belding, M. A., Morrel, A. R., & Lamb, R. J. (1997).
Reinforcing operants other than abstinence in drug abuse treatment: An
effective alternative for reducing drug use. Journal of Consulting and
Clinical Psychology, 65, 421–428.
Judge, G. G., Griffiths, W. E., Hill, R. C., Lutkepohl, H., & Lee, T.-C.
(1985). The theory and practice of econometrics. New York: Wiley.
Kaskutas, L. A., Bond, J., & Humphreys, K. (2002). Social networks as
mediators of the effect of Alcoholics Anonymous. Addictions, 97, 891–
Longabaugh, R., & Beattie, M. C. (1986). Social investment, environmen-
tal support, and treatment outcomes of alcoholics. Alcohol Health and
Research World, 10, 64–66.
Longabaugh, R., Beattie, M. C., Noel, N. E., Stout, R. L., & Malloy, P.
(1993). The effect of social investment on treatment outcome. Journal of
Studies on Alcohol, 54, 465–478.
Longabaugh, R., Wirtz, P. W., Zweben, A., & Stout, R. L. (1998). Network
support for drinking, Alcoholics Anonymous and long-term matching
effects. Addiction, 93, 1313–1333.
Longabaugh, R., Wirtz, P. W., Zweben, A., & Stout, R. L. (2001). Network
support for drinking. In R. H. Longabaugh & P. W. Wirtz (Eds.), Project
MATCH: A priori matching hypotheses, results, and mediating mecha-
nisms (NIAAA Project MATCH Monograph Series, Vol. 8, pp. 260–
275). Rockville, MD: U.S. Government Printing Office.
Meyers, R. J., & Miller, W. R. (Eds.). (2001). A community reinforcement
approach to addiction treatment. Cambridge, England: Cambridge Uni-
Miller, W. R., & DelBoca, F. K. (1994). Measurement of drinking behavior
using the Form-90 family of instruments. Journal of Studies of Alcohol,
Miller, W. R., Tonigan, J. S., & Longabaugh, R. (1995). The Drinker
Inventory of Consequences (DrInC) (DHHS Publication No. 95–3911).
Rockville, MD: U.S. Government Printing Office.
Miller, W. R., Wilbourne, P. L., & Hettema, J. E. (2003). What works? A
summary of alcohol treatment outcome research. In R. K. Hester &
W. R. Miller (Eds.), Handbook of alcoholism treatment approaches:
Effective alternatives (3rd ed., pp. 13–63). Boston: Allyn & Bacon.
Moos, R. H., & Moos, B. S. (2005). Paths of entry into Alcoholics
Anonymous: Consequences for participation and remission. Alcoholism:
Clinical & Experimental Research, 29, 1858–1868.
Nowinski, J., Baker, S., & Carroll, K. (1992). Twelve-step facilitation
therapy manual: A clinical research guide for therapists treating indi-
viduals with alcohol abuse and dependence (NIAAA Project MATCH
Monograph, Vol. 1, DHHS Publication No. [ADM] 92–1893). Wash-
ington, DC: U.S. Government Printing Office.
Petry, N. M. (2000). A comprehensive guide to the application of contin-
gency management procedures in clinical settings. Drug & Alcohol
Dependence, 58, 9–25.
Petry, N. M., Alessi, S. M., Carroll, K. M., Hanson, T., MacKinnon, S.,
Rounsaville, B., & Sierra, S. (2006). Contingency management treat-
ments: Reinforcing abstinence versus adherence with goal-related activ-
ities. Journal of Consulting and Clinical Psychology, 74, 592–601
Petry, N. M., Martin, B., Cooney, J. L., & Kranzler, H. R. (2000). Give
them prizes, and they will come: Contingency management for treatment
of alcohol dependence. Journal of Clinical and Consulting Psychology,
Petry, N. M., Tedford, J., & Martin, B. (2001). Reinforcing compliance
with non-drug related activities. Journal of Substance Abuse Treatment,
Project MATCH Research Group. (1993). Project MATCH (Matching
Alcoholism Treatment to Client Heterogeneity): Rationale and methods
for a multisite clinical trial matching patients to alcoholism treatment.
Alcoholism: Clinical & Experimental Research, 17, 1130–1145.
Project MATCH Research Group. (1997). Matching alcoholism treatments
to client heterogeneity: Project MATCH posttreatment drinking out-
comes. Journal of Studies on Alcohol, 58, 7–29.
Project MATCH Research Group. (1998). Matching alcoholism treatment
to client heterogeneity: Project MATCH three-year drinking outcomes.
Alcoholism: Experimental & Clinical Research, 22, 1300–1311.
SAS Institute. (1999). SAS/STAT software: Changes and enhancements
through V7 and V8 [Computer software]. Cary, NC: Author.
Sisson, R. W., & Azrin, N. (1989). The community reinforcement ap-
proach. In R. Hester & W. R. Miller (Eds.), Handbook of alcoholism
treatment approaches: Effective alternatives (pp. 242–258). New York:
Slosson, R. L. (1963). Slosson Oral Reading Test manual. East Aurora,
NY: Slosson Educational Publications.
Sobell, L. C., & Sobell, M. B. (1992). Timeline follow-back: A technique
for assessing self-reported alcohol consumption. In R. Litten & J. Allen
(Eds.), Measuring alcohol consumption (pp. 41–71). Totowa, NJ: Hu-
LITT, KADDEN, KABELA-CORMIER, AND PETRY
Steinberg, K. L., Roffman, R. A., Carroll, K. M., Kabela, E., Kadden, R.,
Miller, M., et al.. (2002). Tailoring cannabis dependence treatment for a
diverse population. Addiction, 97(Suppl. 1), 135–142.
Steinglass, P., & Wolin, S. (1974). Explorations of a systems approach to
alcoholism. Archives of General Psychiatry, 31, 527–532.
Stout, R. L., Wirtz, P. W., Carbonari, J. P., & Del Boca, F. K. (1994).
Ensuring balanced distribution of prognostic factors in treatment out-
come research. Journal of Studies on Alcohol, 12, 70–75.
Tonigan, J. S., Connors, G. J., & Miller, W. R. (1996). Alcoholics Anon-
ymous Involvement (AAI) Scale: Reliability and norms. Psychology of
Addictive Behaviors, 10, 75–80.
Tonigan, J. S., Miller, W. R., & Brown, J. M. (1997). The reliability of
Form 90: An instrument for assessing alcohol treatment outcome. Jour-
nal of Studies on Alcohol, 58, 358–364.
Tucker, J. A., Vuchinich, R. E., & Gladsjo, J. A. (1990). Environmental
influences on relapse in substance use disorders. International Journal of
the Addictions, 25, 1017–1050.
UKATT Research Team. (2005). Effectiveness of treatment for alcohol
problems: Findings of the randomised United Kingdom Alcohol Treat-
ment Trial (UKATT). British Medical Journal, 331, 541–545.
Winer, B. J. (1971). Statistical principles in experimental design. New
Zywiak, W. H., Longabaugh, R., & Wirtz, P. W. (2002). Decomposing the
relationships between pretreatment social network characteristics and
alcohol treatment outcome. Journal of Studies on Alcohol, 63, 114–121.
Received December 15, 2006
Revision received April 27, 2007
Accepted May 10, 2007 ?
NETWORK SUPPORT PROJECT