Psychology of Addictive Behaviors © 2010 American Psychological Association

2010, Vol. 24, No. 3, 386 –396 0893-164X/10/$12.00 DOI: 10.1037/a0019801

Parental Alcohol Involvement and Adolescent Alcohol Expectancies

Predict Alcohol Involvement in Male Adolescents

James A. Cranford, Robert A. Zucker, Hiram E. Fitzgerald

Jennifer M. Jester, Leon I. Puttler Michigan State University
University of Michigan

Current models of adolescent drinking behavior hypothesize that alcohol expectancies mediate the
effects of other proximal and distal risk factors. This longitudinal study tested the hypothesis that
the effects of parental alcohol involvement on their children’s drinking behavior in mid-adolescence
are mediated by the children’s alcohol expectancies in early adolescence. A sample of 148 initially
9 –11 year old boys and their parents from a high-risk population and a contrast group of community
families completed measures of drinking behavior and alcohol expectancies over a 6-year interval.
We analyzed data from middle childhood (M age ⫽ 10.4 years), early adolescence (M age ⫽ 13.5
years), and mid-adolescence (M age ⫽ 16.5 years). The sample was restricted only to adolescents
who had begun to drink by mid-adolescence. Results from zero-inflated Poisson regression analyses
showed that 1) maternal drinking during their children’s middle childhood predicted number of
drinking days in middle adolescence; 2) negative and positive alcohol expectancies in early
adolescence predicted odds of any intoxication in middle adolescence; and 3) paternal alcoholism
during their children’s middle childhood and adolescents’ alcohol expectancies in early adolescence
predicted frequency of intoxication in middle adolescence. Contrary to predictions, child alcohol
expectancies did not mediate the effects of parental alcohol involvement in this high-risk sample.
Different aspects of parental alcohol involvement, along with early adolescent alcohol expectancies,
independently predicted adolescent drinking behavior in middle adolescence. Alternative pathways
for the influence of maternal and paternal alcohol involvement and implications for expectancy
models of adolescent drinking behavior were discussed.

Keywords: adolescent alcohol expectancies, adolescent drinking behavior, parental drinking, parental
influence.

Does exposure to parental drinking influence the drinking be- identified as a priority topic (Kraemer, Stice, Kazdin, Offord, &
havior of their children once drinking onset has occurred? If so, are Kupfer, 2001). In this paper, we examine the longitudinal effects
the effects mediated by childhood alcohol expectancies (AEs) of parental alcohol involvement and adolescent AEs on subsequent
preceding the child’s drinking? A probabilistic-developmental alcohol involvement among adolescents who had begun drinking
model of risk (Zucker, 1994; Zucker, Fitzgerald, & Moses, 1995; by mid-adolescence. Below, we review the literature on parental
Zucker, Donovan, Masten, Mattson, & Moss, 2008) emphasizes alcohol involvement and adolescent AEs on later drinking behav-
the cumulation of risk factors over time for the developmental ior.
course of problem drinking, and the specification of how various
risk factors work together to produce negative outcomes has been
Effects of Parental Alcohol Involvement on Adolescent
Drinking Behavior

James A. Cranford, Robert A. Zucker, Jennifer M. Jester, and Leon I. An extensive body of evidence showed that parental substance
Puttler, Addiction Research Center, Department of Psychiatry, University use and family history of substance use are predictive of adoles-
of Michigan; Hiram E. Fitzgerald, University Outreach and Engagement, cent substance use (for reviews, see Ellis, Zucker, & Fitzgerald,
Kellogg Center, Michigan State University. 1997; Hawkins, Catalano, & Miller, 1992; Scheier, 2001; Sher,
This work was supported by Grants T32 AA07477 and Grant R37 Grekin, & Williams, 2005; Wills & Yaeger, 2003; Windle, 1996).
AA07065 from the National Institute on Alcohol Abuse and Alcoholism to For example, Sher, Walitzer, Wood, and Brent (1991) examined
Robert A. Zucker, PhD. We thank the families who participate in the the concurrent effects of parental alcoholism on adolescent alcohol
Michigan Longitudinal Study and Susan K. Refior, whose sustained work involvement in a large sample of college freshmen (M age ⫽ 18.2
with the families in this study has been a major contributor to the study’s
years). They found that children of alcoholics (COAs) reported
continuation. We also thank the editor and the anonymous reviewers for
helpful comments on previous drafts of this manuscript.
higher levels of alcohol involvement (e.g., quantity-frequency of
Correspondence concerning this article should be addressed to James A. consumption, heavy drinking, and alcohol dependence symptoms)
Cranford, PhD, Addiction Research Center (UMARC), Rachel Upjohn compared to non-COAs, although the extent to which alcoholic
Building, 4250 Plymouth Rd., Department of Psychiatry, University of parents were actually drinking during the child’s earlier years was
Michigan, Ann Arbor, MI 48109-2700. E-mail: [email protected] not evaluated. Similar findings have been obtained with samples of

386
 PARENT DRINKING AND ADOLESCENT ALCOHOL EXPECTANCIES 387

younger adolescents. For example, working with a sample of Christiansen, Smith, Roehling, & Goldman, 1989; Newcomb,
alcoholic and nonalcoholic families from the community, Chassin, Chou, Bentler, & Huba, 1988; Smith, Goldman, Greenbaum, &
Rogosch, and Barrera (1991) examined the concurrent effects of Christiansen, 1995). Similar findings have been obtained from
parental alcoholism on the alcohol involvement of the adolescent longitudinal studies of college student samples (Darkes, Green-
offspring (M age ⫽ 12.7). Results showed that paternal but not baum, & Goldman, 2004; Goldman, Greenbaum, & Darkes, 1997;
maternal alcoholism predicted greater adolescent alcohol involve- Kidorf, Sherman, Johnson, & Bigelow, 1995; Sher, Wood, Wood,
ment, and this effect was stronger among the older adolescents. & Raskin, 1996). In addition, there is longitudinal evidence that
Other evidence in addition to the Chassin et al. (1991) study AEs predict the transition from nonproblem to problem drinking
indicated that the association between parental and offspring al- (Christiansen et al., 1989), and a recent study of high school
cohol involvement differs depending on the gender of the parent students (M age ⫽ 16.2) found that tension-reduction AEs were
and/or the child. However, the nature of the relationship has not concurrently associated with frequency of drunkenness (Catanzaro
been consistently replicated. Thus, Zhang, Welte, and Wieczorek & Laurent, 2004).
(1999) found that paternal but not maternal drinking had a direct
concurrent effect on adolescent boys’ drinking (age range 16 –19). Child and Adolescent AEs as Mediators of the Effects
By contrast, Ohannessian et al. (2004) found that maternal sub-
of Parental Drinking on Adolescent Drinking
stance use consequences were more consistently related to con-
current adolescent psychopathology (including alcohol depen- To this point, our review has indicated some linkages between
dence, depression, and conduct disorder) than paternal substance a) parental and adolescent drinking, and b) adolescent AEs and
use consequences. Furthermore, parental effects are not always adolescent drinking. Models of AEs emphasize their role as me-
observed; in another cross-sectional study, Cooper, Peirce, and diators of the effects of more distal risk factors (Goldman et al.,
Tidwell (1995) found no consistent associations between maternal 1999; Petraitis, Flay, & Miller, 1995). Tests of social learning
or paternal drinking and adolescent substance use. Similarly, Yu theory explanations of adolescent alcohol use have shown that
(2003) showed that parental alcohol use was related to lifetime but exposure to parents who use alcohol has a direct relationship to
not current alcohol use of their adolescent children (age range AEs (Zucker, Kincaid, Fitzgerald, & Bingham, 1995), which in
15–18). turn predicts alcohol involvement (Petraitis et al). The hypothe-
Although smaller in number, longitudinal studies have yielded sized mediational effects of AEs have been elaborated by a number
similar results. White, Johnson, and Buyske (2000) followed ad- of alcohol researchers (Goldman et al., 1999; Scheier & Botvin,
olescents across four waves from ages 15 to 28. Paternal and 1997; Sher et al., 1991).
maternal drinking were predictive of a heavy drinking trajectory However, direct tests of the mediational role of AEs, based on
among sons and daughters. Wills, Sandy, Yaeger, and Shinar the assumption that AEs are among the most proximal correlates of
(2001) assessed 1,269 adolescents in 6th, 7th, and 8th grade and drinking behavior, have yielded conflicting findings. For example,
found that parental substance use (alcohol and tobacco use as the cross-sectional study by Sher et al. (1991) found that the effect
reported by the child) predicted higher adolescent substance use (a of family history of paternal alcoholism on college students’
latent variable comprised of alcohol, tobacco, and marijuana use) alcohol involvement was mediated by the students’ own level of
in grade 6, but did not predict increases in adolescent use over behavioral under-control as well as by their (positive) AE level.
time. In one of the few longitudinal studies utilizing a high-risk Ouellette, Gerrard, Gibbons, and Reis-Bergan (1999) followed
sample, Chassin, Curran, Hussong, and Colder (1996) found that parents and their offspring over 4 years and showed that the effect
paternal and maternal alcoholism predicted initial levels of sub- of parental drinking (average of maternal and paternal consump-
stance use (alcohol and illicit drug use) among girls and boys, but tion) on adolescent alcohol consumption was mediated by adoles-
only paternal alcoholism and adolescent male gender predicted cent AEs. By contrast, a longitudinal study by Colder, Chassin,
increases in substance use over a 3-year interval. Although the Stice, and Curran (1997) found that parental alcoholism had a
effects of paternal alcoholism were partially mediated by fathers’ direct effect on increases in adolescent heavy drinking that was not
monitoring and adolescents’ stress, negative affect, and associa- mediated by the adolescents’ AEs.
tions with substance-using peers, these hypothesized mediators did
not fully account for the paternal alcoholism effect. The Present Study

Effects of Child and Adolescent AEs on Adolescent Taken together, the available evidence remains unclear with
Drinking Behavior respect to the hypothesis that adolescent AEs mediate the effects of
their parents’ alcohol involvement on their own drinking during
AEs are another potent risk factor for adolescent alcohol in- adolescence. Although evidence suggests linkages between a) ma-
volvement. The available evidence indicates that alcohol expect- ternal and paternal alcohol involvement and their adolescent chil-
ancies are “among the strongest predictors of drinking, even after dren’s drinking and b) the adolescent’s AEs on their own drinking,
other variables are controlled” (Goldman, Del Boca, & Darkes, longitudinal studies have not produced consistent support for the
1999, p. 219). Previous research using school-based samples mediational hypothesis. A major challenge to understanding these
showed that AEs predicted drinking onset among adolescents relationships is that both drinking and nondrinking adolescents
(Bauman, Fisher, Bryan, & Chenoweth, 1985; Killen et al., 1996), have typically been included in the same analyses even though
and results from several longitudinal studies found that adolescent evidence has indicated that AEs change substantially as a function
AEs predicted increases in alcohol consumption over time (e.g., of the transition from nondrinking to drinking (see Christiansen,
Aas, Leigh, Anderssen, & Jakobsen, 1998; Bauman et al., 1985; Goldman, & Inn, 1982; Schell, Martino, Ellickson, Collins, &
388 CRANFORD, ZUCKER, JESTER, PUTTLER, AND FITZGERALD

McCaffrey, 2005). Such changes have been linked to the concrete pleted the T3 protocol, 57.1% (N ⫽ 148) had begun drinking. We
experience of drinking and the exposure to alcohol’s pharmaco- analyzed data from these 148 boys and their parents at T1, T2, and
dynamic effects (Aas et al., 1998). Thus, inclusion of both groups T3. Based on responses to the question “How old were you the first
creates a confounded predictor (the expectancies). In this paper, time you ever took a drink? Do not count the times you were given
we examine the possible mediational role of childhood AEs in a ‘sip’ by an adult,” the mean (SD) of drinking onset was 13.5 (2.5)
explaining the association between parental alcohol involvement years old (range ⫽ 5 to 17 years old). Although the sample of 148
and adolescent drinking behavior once drinking has begun (i.e., boys included 5 male siblings of the male target children (MTCs),
after the transition has occurred). Using longitudinal data, we all participants were treated as independent based on an intraclass
tested the hypothesis that the association between parental drink- correlation of 0. At baseline, T2, and T3, average ages for children
ing and adolescent drinking is mediated by adolescents’ AEs. (with standard deviations in parentheses) were 10.4 (.9), 13.5 (.9),
and 16.5 (1.0) years.
Method Parents at T1 were 148 mothers and fathers whose mean (SD)
ages were 36.6 (4.0), and 39.2 (5.0) years, respectively. Couples
had been married for an average of 11.4 years. Both couple
Participants
members had completed about 2 years of education beyond high
The present work is from the ongoing Michigan Longitudinal school: mean total education years for mothers, M (SD) ⫽ 14.2
Study (MLS; Zucker et al., 2000), a prospective study that is (1.9); for fathers, M ⫽ 14.9 (2.3), and median family income was
following a community sample of initially intact families with high $40,000. All families were Caucasian because less than 4% of the
levels of substance use/abuse, along with a community contrast population we sampled from was non-Caucasian. Given our sam-
sample of families drawn from the same neighborhoods, but with- ple size, this precluded effective analyses of race and ethnic
out the high substance abuse profile. The long term focus of the differences.
project is the emergence and development of substance abuse and
problems in the children, and the patterns of stability and change Measures
in drug involvement among the parents.
A community-based, but high alcohol-involved, sample of ini- Parental lifetime alcohol use disorder (AUD) and alcohol
tially intact families was recruited by identifying fathers on the involvement at T1. DSM-IV alcoholism diagnosis for both par-
basis of a drunk driving conviction with a high blood alcohol level ents was assessed using several measures, including the Short
(0.15 percent if a first conviction, 0.12 percent if not the first). Michigan Alcoholism Screening Test (SMAST; Selzer, Vinokur,
Families were required to have at least one son in the 3–5 year age & van Rooijen, 1975), the Drinking and Drug History Question-
range, and daughters in the 3–11 year age range were recruited naire (DDH; Zucker, 1991), and the Diagnostic Interview Sched-
when present. Presence of fetal alcohol syndrome was ruled out by ule—Version IV (DIS-IV; Robins, Helzer, Croughan, & Ratcliff,
study exclusionary criteria. Both biological parents were required 1981). The SMAST is a 13-item self-report screening inventory
to be living with the child at the time of recruitment, and mothers’ that assesses alcohol problems. The DDH contains a series of
substance use status was free to vary. A contrast/control group of questions asking about alcohol and other drug use and alcohol-
families who resided in the same neighborhoods as the drunk related consequences over the past 6 months. The DIS-IV is a
driver families but had no substance abuse history for either parent structured diagnostic interview that collects extensive information
was also recruited. A second subset of families with a father who about physical, alcohol- and drug-related symptoms, and other
also had an alcohol use disorder was uncovered and recruited psychiatric symptoms. Trained clinicians used data from all three
during the community canvass for controls (Zucker et al., 2000). sources of data to create a best-estimate diagnosis (Leckman,
After initial recruitment and assessment, individuals participated in Sholomskas, Thompson, Belanger, & Weisman, 1982) of a life-
multi-session assessments every 3 years. Data collection was com- time alcohol use disorder (abuse or dependence) for both parents.
pleted by professional staff, graduate students, and carefully DIS data were used as the base supplemented by the DDH and
trained and supervised undergraduates. SMAST data, guided by the principle that when a symptom was
Participants for the present study were biological fathers, moth- admitted, even from only one source, it probably was present. To
ers, and sons who completed relevant measures at child ages 9 –11 evaluate the reliability of this pooled diagnosis, two raters inde-
(middle childhood), 12–14 (early adolescence), and 15–17 (mid- pendently diagnosed a series of 26 protocols. Agreement as eval-
adolescence). For ease of presentation, we refer to these time uated by kappa was .81, indicating acceptable reliability. In this
points as baseline (T1), T2, and T3 (although the MLS designation subsample of N ⫽ 148 drinking boys, 33.1% (n ⫽ 49) of the
is T3, T4, and T5). A total of 259 MLS families completed the mothers and 76.4% (n ⫽ 113) of the fathers had a lifetime AUD.
study protocol at T3. This total included a subset of girls poten- In terms of family risk status, 23.7% (n ⫽ 35) of the families were
tially available for the study. However, the sample of girls avail- control families; 18.9% (n ⫽ 28) were families from the commu-
able for this study (n ⫽ 21) was not large enough to conduct nity in which the father had an AUD; and 57.4% (n ⫽ 85) were
meaningful analyses by gender, and because substantial sex dif- families in which the father had a drunk driving conviction.
ferences are known to exist for many drinking indicators, we Less severe forms of parental alcohol involvement (e.g., fre-
limited our analyses to boys. Also, because the present work is quency of alcohol consumption; Ary, Tildesley, Hops, & Andrews,
focused on individual differences in later adolescence, when drink- 1993) have also shown longitudinal associations with adolescent
ing has to a large extent been initiated, we analyzed data only from drinking. Accordingly, we included a measure of average number
families where the boys had begun drinking by T3 (when they of drinking days per month in the last 6 months from the DDH
were an average of 16.5 years old). Of the 259 boys who com- (Zucker, 1991). The average number of drinking days per month
 PARENT DRINKING AND ADOLESCENT ALCOHOL EXPECTANCIES 389

was lower for mothers (M ⫽ 4.1, SD ⫽ 5.6) than fathers (M ⫽ 8.6, variables often exceed this restriction by having larger variances
SD ⫽ 8.5), paired t(135) ⫽ ⫺6.2, p ⬍ .01. than means (i.e., overdispersion; Horton, Kim, & Saitz, 2007).
Alcohol expectancies at T2. AEs were assessed with the Although the present sample consisted only of those who had
Beverage Opinion Questionnaire (BOQ; Fitzgerald, Zucker, & started drinking, a substantial percentage of participants reported
Noll, 1990) which was administered to participants starting when that they did not drink or experience any drunken episodes in the
they were between the ages of 6 and 8 years and then again at T1, past 6 months (29.1% and 34.5% of the sample, respectively).
T2, and T3. This 25-item questionnaire assesses negative (5 items) Count variables with large numbers of zeroes are referred to as
and positive (20 items) expectancies for alcohol and, as a buffer, “zero-inflated” (Karaszia & van Dulmen, 2008). Accordingly,
also includes 30 expectancy questions about soft drinks. The BOQ zero-inflated Poisson (ZIP) regression analysis (Lambert, 1992)
is based on the adolescent version of the Alcohol Expectancy was used to examine predictors of drinking behaviors. For ZIP
Questionnaire (AEQ-A; Christiansen et al., 1982; Brown, Chris- models, two regression equations are estimated simultaneously: 1)
tiansen, & Goldman, 1987) and the adult version of the Alcohol a logistic regression model is used to predict whether or not a
Expectancy Questionnaire (AEQ; Brown, Goldman, Inn, & Ander- given behavior occurs (i.e. membership in an “always zero” versus
son, 1980). The original version of the AEQ-A was developed for a “not always zero” latent group; Karazsia & van Dulmen, 2008),
use with adolescents ages 12–19 (Christiansen et al., 1982) and and 2) a Poisson regression model is used to predict the number of
consists of 90 items. To reduce participant burden we sought to times a given behavior occurs (Atkins & Gallop, 2007; Muthen &
reduce the number of items for inclusion in the BOQ. Further, Muthen, 2007). Estimating ZIP models thus allows for the possi-
because we began asking about AEs when participants were be- bility that different variables may predict whether or not someone
tween the ages of 6 and 8 years old, we selected those items that drinks and how much or how often someone drinks (Atkins &
seemed likely to be most comprehensible when read to children by Gallop, 2007). All ZIP models were estimated with the Mplus
the interviewers. With these concerns in mind, we selected 23 software package using maximum likelihood estimation with ro-
items from the AEQ-A. In addition, we selected two items from bust standard errors (Muthen & Muthen, 2007). For ease of inter-
the adult AEQ that focused on sleep. Each statement concerning pretation of the logistic regression results, we report the reciprocal
alcohol is in the format “Drinking beer or wine would. . .”, fol- odds ratio for each predictor so that they represent the odds of
lowed by a phrase indicating an expectancy for alcohol, e.g., being in the nonzero class, i.e. the odds of the occurrence of each
“Drinking beer or wine would make me feel good” (positive drinking behavior.
expectancy), “Drinking beer or wine would make me feel angry”
Missing data. Because the pairwise sample sizes for the vari-
(negative expectancy).1 Adolescents were asked to respond to each
ables in our models ranged from n ⫽ 87 to n ⫽ 148, we used
item on a four-point scale (1 ⫽ agree completely, 2 ⫽ somewhat
multiple imputation (MI; Rubin, 1987; Sinharay, Stern, & Russell,
agree, 3 ⫽ somewhat disagree, 4 ⫽ completely disagree). Inspec-
2001) to impute missing data. In MI, each missing value is re-
tion of the item distributions showed that few participants selected
placed by m ⬎ 1 simulated values, resulting in m complete data
the “somewhat agree” or “somewhat disagree” response options.
sets (Schafer, 1997; Schafer & Graham, 2002). These m data sets
Thus, we collapsed response options to create binary versions of
each item, where 0 ⫽ disagree completely or somewhat disagree
and 1 ⫽ agree completely or somewhat agree. Items were summed 1
Brown et al. (1987) noted that, compared to the adult version of the
to create negative and positive expectancies scores for each par- AEQ, for the AEQ-A “statements are worded more generally to accom-
ticipant. At T2, the positive AEs scale had an alpha of .88 (M ⫽ modate adolescents who have had little or no experience with alcohol” (p.
2.4, SD ⫽ 3.6), the negative AEs scale had an alpha of .74 (M ⫽ 485). Specifically, “the Adult AEQ involves statements regarding the
1.9, SD ⫽ 1.4), and the two scales were moderately correlated, r ⫽ effects of alcohol on the respondent, whereas the Adolescent AEQ focuses
.44, p ⬍ .01. on the effect of alcohol on people in general” (p. 488). For example, the
Adolescent alcohol involvement at T3. For the adolescent adult AEQ item “Drinking makes me feel good” was modified for the
AEQ-A to “Drinking alcohol makes a person feel good and happy.” We
version of the DDH, participants were asked about average drink-
agree with Brown et al. that such modifications likely make the AEQ-A
ing days per month (drinking days) over the past 6 months at T3
items more applicable to the entire adolescent population, including “ad-
(M ⫽ 2.8, SD ⫽ 3.8).2 In other words, adolescents reported olescents who have not yet had direct or personal experience with alcohol”
drinking on approximately 18 days during the past 6 months. (p. 489). Because the MLS by design recruited a sample of families that
Adolescent participants at T3 were also asked to indicate how was likely to have extensive experience with alcohol, we decided to retain
many times during the past 6 months they had gotten drunk or very the original wording of the items such that they referred to the adolescent.
high from drinking alcohol. Scores on this variable ranged from 0 Recognizing that even in a high-risk sample not all adolescents will have
(not at all) to 52 (about twice a week) with a mean (SD) of 9.8 consumed alcohol, each item was framed as a pure expectancy, rather than
(20.4) episodes of drunkenness during the past 6 months. as a putative effect of alcohol involvement. Returning to the earlier
example, the AEQ item “Drinking makes me feel good” was modified for
the AEQ-A to “Drinking alcohol makes a person feel good and happy,” and
Analytic Plan we in turn modified this item for the BOQ to “Drinking alcohol would
make me feel good.”
Correlation and regression analyses were used to test the study 2
Three cases were identified as outliers (i.e., more than 3 standard
hypotheses. Our two dependent variables (number of drinking days deviations above the mean; Stevens, 1998) on this variable. All analyses
and number of drunken episodes in the past 6 months) were count were conducted with and without these three outlier cases. The results did
variables. Count variables can sometimes be modeled as Poisson not differ for the two sets of analyses. Because descriptive statistics were
variables, but the Poisson distribution is restricted to a single unduly influenced by these outlier cases, all descriptive statistics are
parameter for the mean and the variance, and alcohol-related count reported with the outlier cases excluded.
390 CRANFORD, ZUCKER, JESTER, PUTTLER, AND FITZGERALD

are analyzed using standard analytic methods, and the results are paternal drinking frequency. No statistically significant correla-
combined to obtain parameter estimates and standard errors that tions between parental alcohol involvement and adolescents’ pos-
take into account missing data uncertainty (Sinharay et al., 2001). itive AEs were observed, but there was a weak direct association
MI is based on the assumption that the data are missing at random between paternal AUD and adolescents’ negative AEs 3 years later
(MAR; Sinharay et al., 2001). Since this assumption is not testable, at T2. Paternal AUD and maternal drinking frequency were sig-
we included several variables in the imputation model that could nificantly associated with both measures of adolescent alcohol
potentially be linked to the missingness of the imputed variables involvement. Maternal AUD was also positively associated with
(Schafer, 1997; Sinharay et al., 2001). adolescents’ drinking frequency (but not frequency of drunken-
The pattern of missing data appeared to be arbitrary, and so we ness), and paternal drinking frequency was positively associated
used the Markov Chain Monte Carlo (MCMC) imputation method with adolescents’ frequency of drunkenness (but not drinking
(Schafer, 1997). For the variables in the models we tested, the rate frequency). Adolescents’ positive AEs at T2 were positively re-
of missing information (␭) ranged from a low of .15 to a high of lated to frequency of drunkenness (but not drinking frequency) at
.48. Results from a simulation study (Graham, Olchowski, & T3. Adolescents’ negative AEs at T2 were not significantly asso-
Gilreath, 2007) showed with m ⫽ 10 imputations: 1) the power to ciated with either measure of adolescent alcohol involvement.
detect a small effect size when ␭ ⫽ .50 showed a decrease of only Baron and Kenny (1986) outline methods for testing media-
about 3%, compared to simulations with m ⫽ 100 imputations; and tional hypotheses. These steps include 1) establishing an associa-
2) the relative efficiency of a given parameter estimate when ␭ ⫽ tion between the predictor and the outcome variable; 2) establish-
.50 is .96 (compared to simulations with m ⫽ 100 imputations). ing an association between the predictor and the putative mediator
Accordingly, we used SAS PROC MI (SAS, 2004) to create m ⫽ variable; 3) establishing an association between mediator and the
10 imputed data sets. We then used PROC CORR to conduct outcome variable when the predictor is statistically controlled; and
correlational analyses on the m ⫽ 10 imputed data sets, and PROC 4) showing that the association between the predictor and the
MIANALYZE to combine the results from analyses of the m ⫽ 10 outcome is reduced in magnitude when the mediator variable is
data sets. As noted earlier, for the ZIP models, we used the Mplus
entered into the regression equation. Although demonstration of a
program with multiple imputation and maximum likelihood esti-
relationship between the predictor and the outcome is not always
mation with robust standard errors (Muthen & Muthen, 2007).
required (Shrout & Bolger, 2002), an association between the
Attrition analyses. For mothers and fathers, we compared T1
predictor and the putative mediator variable is necessary to estab-
responders and nonresponders on the measure of drinking fre-
lish mediation. However, as seen in Table 1, only one of the
quency at T3, and no significant differences between responders
measures of parental alcohol involvement was associated with
and nonresponders were observed. For adolescents, we compared
adolescent AEs: paternal AUD showed a direct association with
T2 responders and nonresponders on the measures of positive and
negative AEs. Further, negative AEs were not associated with
negative AEs at T3, and no significant differences between re-
either measure of adolescent alcohol involvement. These findings
sponders and nonresponders were observed. These results indicate
do not support the hypothesis that adolescent AEs mediate the
that any nonresponse bias was minimal.
effects of parental alcohol involvement on adolescent drinking.
We then examined the effects of T1 parental alcohol involve-
Results ment and T2 adolescent AEs as independent predictors of adoles-
Correlations between all study variables are presented in Table 1. cent drinking at T3. ZIP regression analyses were conducted using
Maternal and paternal measures of AUD and drinking frequency the Mplus statistical software program (Muthen & Muthen, 2007).
were moderately correlated (rs ranged from .22 to .33). Interest- Results are presented in Table 2. We first tested the effects of
ingly, there was no significant association between paternal AUD lifetime paternal and maternal AUD, T1 paternal and maternal
and maternal drinking frequency, or between maternal AUD and alcohol involvement, and T2 adolescent AEs as predictors of

Table 1
Zero-Order Correlations Between Study Variables

Variable 1. 2. 3. 4. 5. 6. 7. 8.

1. Maternal lifetime ALC DX ⫺

2. Paternal lifetime ALC DX .22ⴱⴱ ⫺
3. Maternal drinking days per month in last 6 months T1 .32ⴱⴱ .04 ⫺
4. Paternal drinking days per month in last 6 months T1 .11 .33ⴱⴱ .27ⴱⴱ ⫺
5. Adolescent positive AEs at T2 .13 .12 ⫺.06 ⫺.16 ⫺
6. Adolescent negative AEs at T2 ⫺.14 .20ⴱ ⫺.11 ⫺.10 .44ⴱⴱ ⫺
7. Drinking days per month in last 6 months at T3 .25ⴱⴱ .17ⴱ .23ⴱ .11 .14 ⫺.02 ⫺
8. Frequency of drunkenness in last 6 months at T3 .13 .22ⴱⴱ .27ⴱ .26ⴱⴱ .22ⴱ .09 .56ⴱⴱ ⫺
M .33 .76 4.1 8.6 2.4 1.9 2.8 9.8
SD .47 .43 5.6 8.5 3.6 1.4 3.8 20.4

Note. N ⫽ 148 male adolescent drinkers and their parents. ALC DX ⫽ DSM-IV lifetime alcoholism diagnosis coded 0 ⫽ No ALC DX and 1 ⫽ ALC
DX. AEs ⫽ alcohol expectancies.
ⴱ
p ⬍ .05. ⴱⴱ p ⬍ .01.
 PARENT DRINKING AND ADOLESCENT ALCOHOL EXPECTANCIES 391

Table 2
Zero-Inflated Poisson Regression Analysis of Longitudinal Predictors of Adolescents’ Average Number of Drinking Days per Month
and Number of Times Intoxicated in Past 6 Months

Average drinking days in past 6 months at T3 Number of times intoxicated in past 6 months at T3

Any drinking days Number of drinking Any intoxication Number of times intoxicated
Predictors exp(␥)1 days exp(␤)2 exp(␥)1 exp(␤)2

Age 2.05ⴱ 1.19 2.11ⴱ 1.19

T2 negative expectancies 0.63 0.99 0.61ⴱ 0.99
T2 positive expectancies 1.19 1.01 1.39ⴱ 1.07ⴱ
T1 maternal AUD 0.63 1.61 0.48 0.94
T1 maternal drinking3 1.01 1.03ⴱ 1.07 1.03
T1 paternal AUD 1.38 1.61 2.43 3.35ⴱ
T1 paternal drinking3 1.05 0.99 1.04 1.03
1
Exponentiated coefficient from logistic regression component of ZIP model.
2
Exponentiated coefficient from Poisson regression component of ZIP model.
3
Average drinking days per month in the past 6 months.
ⴱ
p ⬍ .05.

average drinking days in the past 6 months at T3. Because there For example, at the average level of positive AEs at T2 (2.4), the
was variation in age within waves, we controlled for adolescents’ expected number of times intoxicated is 8.3. At one unit above the
age at T3 in all analyses. For the logistic regression of the binary average level of positive AEs at T2 (3.4), the expected number of
part of the dependent variable, age was the only significant pre- times intoxicated is 8.9. We also used procedures outlined by Long
dictor, and every 1 unit increase in age resulted in a 2.05 increase (1997) to calculate the additive change in number of times intox-
in the odds of drinking on any days in the past 6 months. For the icated for adolescents as a function of father’s lifetime AUD. For
Poisson regression of the count part of the dependent variable, the adolescents with a non-AUD father, the expected number of times
only significant predictor was T1 maternal drinking. To gain intoxicated in the past 6 months is 3.3. By contrast, for adolescents
perspective on the meaning of the Poisson coefficients, we used with an AUD father, the expected number of times intoxicated in
procedures outlined by Long (1997, p. 229). For each 1-unit the past 6 months is 11.0, holding all other variables constant.
increase in mothers’ average drinking days per month, adoles- Thus, while lifetime paternal AUD and T2 positive expectancies
cents’ drinking days per month increased by a factor of 1.033, an were independently associated with increases in the number of
increase of 3.3%, when all other predictors were statistically times intoxicated in the past 6 months at T3, the effects of paternal
controlled. Although significant, this appears to be a relatively lifetime AUD were particularly strong in magnitude.
small effect. For example, at the average level of mothers’ average
drinking days per month (4.3), the expected number of adoles- Discussion
cents’ drinking days is 3.2. At one standard deviation above the
average level of mothers’ average drinking days per month (10.7), This study tested the hypothesis that AEs mediate the effects of
the expected number of adolescents’ drinking days is 4.0. parental alcohol involvement on adolescent drinking behavior. In
Next, we tested the effects of lifetime paternal and maternal partial support of our hypotheses, we found that two aspects of
AUD, T1 paternal and maternal alcohol involvement, and T2 parental alcohol involvement (i.e., paternal lifetime AUD and
adolescent AEs as predictors of frequency of intoxication in the maternal average drinking days per month) during middle child-
past 6 months at T3. As seen in Table 2, for the logistic regression hood (T1) predicted some dimensions of mid-adolescent drinking
of the binary part of the dependent variable, age and negative and (T3). Contrary to our hypothesis, results showed that the effects of
positive AEs were significant predictors of any intoxication in the parental alcohol involvement were not mediated by adolescent
past 6 months. Increases in age and positive expectancies resulted AEs. Rather, parental drinking and positive and negative adoles-
in higher odds of any intoxication, and increases in negative cent AEs had independent longitudinal associations with adoles-
expectancies resulted in lower odds of any intoxication. None of cent drinking behavior.
the maternal or paternal alcohol involvement variables were asso-
ciated with any intoxication in the past 6 months. For the Poisson Parental Alcohol Involvement and Adolescents’
regression of the count part of the dependent variable, lifetime Drinking Behaviors
paternal AUD and T2 positive expectancies were significantly
associated with the frequency of intoxication in the past 6 months. Our findings with respect to the effects of parental alcohol
We again used procedures outlined by Long (1997) to gain involvement on adolescents’ drinking behaviors are consistent
perspective on the meaning of the Poisson coefficients. For each with a long line of work indicating that parents have profound
1-unit increase in positive AEs at T2, adolescents’ number of times effects on the drinking behaviors of their children (e.g., Fitzgerald,
intoxicated increased by a factor of exp(.069) ⫽ 1.07, an increase Davies, & Zucker, 2002; Jacob & Johnson, 1997; Wills & Yaeger,
of 7.0%, when all other predictors were statistically controlled. 2003; Windle, 1996; Zucker et al., 2000, 2008). Our findings are
Although significant, this appears to be a relatively small effect. unique, however, in showing that different aspects of paternal and
392 CRANFORD, ZUCKER, JESTER, PUTTLER, AND FITZGERALD

maternal alcohol involvement are longitudinally associated with terized by a biological history of alcoholism, might be consider-
different aspects of their sons’ alcohol involvement 6 years later. able. In the absence of a direct measure of parental modeling, a
The finding that mothers’ but not fathers’ drinking behavior was drinking-modeling explanation for the present findings related to
predictive of subsequent alcohol involvement in their sons is paternal AUD cannot be ruled out.
consistent with the work of Brook and her colleagues (Brook,
Whiteman, Gordon, & Cohen, 1986), who found that aspects of the Adolescent AEs and Drinking Behaviors
mother-child relationship were stronger protective factors for ad-
olescent drug use than were similar aspects of the father-child Our results also showed that adolescents’ negative and positive
relationship. Brook et al. speculated that mothers have more in- AEs were longitudinally associated with higher odds of any intox-
fluence on child-rearing practices than do fathers. Our findings ication 3 years later, and positive AEs further predicted frequency
indicate that this influence extends to the domain of alcohol of drunkenness, independently of parental alcohol involvement.
involvement, at least in terms of adolescents’ average drinking These findings replicate previous work in showing that AEs in
days per month (also see Christiansen & Goldman, 1983). Related early adolescence are longitudinally associated with drinking later
to this point, the greater amount of time spent with mothers versus in adolescence (Reese, Chassin, & Molina, 1994; Smith et al.,
fathers may lead adolescents to more closely model their own 1995), and more generally with previous results showing that
drinking behavior after that of their mothers. This may in part positive and negative expectancies are predictive of alcohol in-
explain why maternal drinking behavior, but not maternal AUD, volvement (Goldman & Darkes, 2004). Further, the differential
predicted their son’s drinking behavior (Ohannessian & Hessel- effects of negative and positive AEs on the occurrence and fre-
brock, 2004). quency of intoxication are consistent with evidence reported by
By contrast, paternal alcoholism (but not paternal drinking be- Leigh and Stacy (2004), who suggested that “negative expectancy
havior) was predictive of sons’ alcohol involvement, and this predicts abstention while positive expectancy predicts amount of
effect was limited to frequency of intoxication. Paternal alcohol- drinking among those who drink” (p. 224) (also see Chen, Grube,
ism is associated with a wide range of parenting variables, includ- & Madden, 1994).
ing less parental discipline (King & Chassin, 2004), lower levels of Results are also consistent with the hypothesis advanced by Sher
parental monitoring (Chassin, Pillow, Curran, Molina, & Barrera, and Gotham (1999) that AEs are developmentally specific risk
1993; Chassin et al., 1996), and higher levels of child abuse and factors for alcohol involvement. Alcohol schemas emerge as early
neglect (Richter & Richter, 2001); all of these variables are sep- as age 3 (Zucker et al., 1995), and evidence indicated a shift in AEs
arately associated with earlier and heavier drinking among off- from more negative to more positive during the period from
spring. Our findings are thus consistent with recent work showing middle childhood to early adolescence (grades 6 to 9; Dunn &
that COAs continue to show elevated levels of heavy drinking Goldman, 1998, 2000; cf. Spiegler, 1983). Positive expectancies
even when their fathers’ alcoholism has remitted (DeLucia, Belz, may better predict alcohol involvement than negative expectancies
& Chassin, 2001). In addition, parental alcoholism confers height- among younger participants, but the effects of negative AEs be-
ened genetic risk among some COAs (Zucker et al., 2008). The come stronger as a function of age (Leigh & Stacy, 2004). The
combination of socialization and genetic risk may explain the current results add to this literature by showing that positive
relatively large magnitude of the effect of paternal alcoholism on expectancies in middle adolescence have a stronger association to
adolescents’ intoxication. risky drinking than to overall frequency of drinking behavior.
With respect to the null findings for paternal drinking behavior,
we note that paternal as compared to maternal drinking is more Adolescent AEs as Mediators of the Effects of Parental
likely to occur on a sporadic basis for antisocial alcoholics (Jacob Alcohol Involvement
& Leonard, 1988), a group that has a substantial representation in
the current sample of alcoholics. Thus, exposure to father drinking, Our findings did not support the hypothesis that the effects of
especially during late preadolescence, would have been less avail- parental alcohol involvement are mediated by AEs (Chassin et al.,
able to the children than exposure to mother drinking. Last, to 1996). However, it is important to consider some aspects of the
some degree the alcoholic fathers’ drinking behavior was damp- current study that limited our ability to draw firm conclusions
ened during the earlier years of the study as a result of conviction about the mediation hypothesis. Our sample was by design limited
for drunk driving. Such convictions sometimes required attendance to Caucasian males, many of whom were living in high-risk
at alcohol education classes and produced a dampening effect on families, and this limits the generality of the findings. Also, our
fathers’ consumption which would distort the relationship between relatively small sample size was underpowered to detect mediation
their own undampened drinking and their children’s alcohol in- when the associations between a) the independent variable and the
volvement. This process was not in operation for the mothers. dependent variable, and b) the mediator and the dependent variable
The observation that paternal alcoholism— but not paternal are in the small to moderate range (Fritz & MacKinnon, 2007).
drinking behavior—was predictive of sons’ alcohol involvement is Our own, as well as other, work suggests a complex relationship
not readily attributable to a drinking-modeling explanation (see between parental drinking behavior and AEs in children and ado-
Sher et al., 2005). However, exposure to parental modeling was lescents. As noted earlier, several studies have found evidence for
not directly measured in this study. Brown, Tate, Vik, Haas, & linkages between paternal alcohol involvement and their children’s
Aarons (1999) showed that degree of exposure to an alcohol- AEs (e.g., Brown et al., 1999). By contrast, Kraus, Smith, and
abusing family member mediated the association between parental Ratner (1994) found no cross-sectional associations between AEs
alcoholism and positive AEs, and noted that variation in exposure among 268 children in grades 2 through 4 and maternal and
to alcohol-abusing family members, even within families charac- paternal drinking attitudes, parental problem drinking, and family
 PARENT DRINKING AND ADOLESCENT ALCOHOL EXPECTANCIES 393

history of AUD (also see Brown, Creamer, & Stetson, 1987; observed between parental AUD and adolescent AEs. In addition,
Henderson, Goldman, Coovert, & Carnevalla, 1994; Miller, Smith, our use of 3-year assessment intervals, combined with the rela-
& Goldman, 1990). Reasons for variability in the association tively small sample size, might have reduced the likelihood of
between parental alcohol involvement and adolescent AEs include detecting some of the hypothesized mediational processes. Further
sample heterogeneity and use of different measures of alcohol research using designs that combine shorter (e.g., daily) and longer
involvement and expectancies (e.g., Sher et al., 1991). Another time lags will clarify the status of the expectancy mediation
important difference relates to the time lag between longitudinal hypothesis.
assessments. Collins and colleagues (Collins & Graham, 2002)
noted that the association between two variables can change dra-
matically across different measurement intervals and highlighted Conclusions and Implications
the importance of temporal design— defined as “the timing, fre-
quency, and spacing of observations in a longitudinal study” Despite these limitations, the present study has several impor-
(Collins, 2006, p.508)—for longitudinal studies of developmental tant strengths. Results build on our earlier work showing that
processes (see Handley & Chassin, 2009). Greater attention to alcohol schemas form as early as age 3 (Zucker et al., 1995), and
temporal design will clarify the status of mediational hypotheses our design allowed us to follow children and their parents from
about parental alcohol involvement and adolescent AEs (also see middle childhood to middle adolescence, which covers the critical
Sher et al., 1996). period during which adolescents first begin experimenting with
alcohol and other drugs (Zucker, 2006). Also, by tracking children
Adolescent AEs and Parental Alcohol Involvement as who transitioned from nondrinkers in middle childhood to drinkers
in middle adolescence, we were able to confirm that AEs precede
Independent Risk Factors for Adolescent Alcohol
the development of risky drinking. Intervention studies have
Involvement
shown that AEs are amenable to experimental manipulation, and
The current findings are most consistent with the hypothesis that challenges to expectancies predict reductions in alcohol consump-
parental alcohol involvement and AEs represent independent risk tion among males (e.g., Dunn, Lau, & Cruz, 2000). The present
factors for subsequent alcohol involvement among adolescents results highlight the potential utility of challenges targeting posi-
(e.g., Mann, Chassin, & Sher, 1987). This pattern of results rep- tive AEs for reducing risky drinking behaviors.
resents a conceptual replication of previous work showing unique Our findings are particularly important in light of recent evi-
longitudinal associations between parental alcohol involvement dence that family influences on adolescent substance use are more
and AEs and subsequent alcohol involvement in adolescents (e.g., pervasive than peer and neighborhood influences (Ennett et al.,
Reese et al., 1994). However, to our knowledge, ours is the first 2008) and persist through late adolescence (Wood, Read, Mitchell,
study to find unique longitudinal effects of maternal drinking and & Brand, 2004). An important avenue for further work is identi-
paternal AUD (assessed in middle childhood) and positive and fication of how different risk and protective factors at different
negative AEs (assessed in early adolescence) on different dimen- levels influence one another (Buu et al., 2009). Also, while ex-
sions of alcohol involvement (assessed in middle adolescence). pectancies are clearly important precursors of alcohol involve-
Thus, at least in this sample, the evidence suggests that a) cogni- ment, other cognitive constructs (e.g., substance use intentions;
tive factors may be of greater importance than parental factors in Anderson, Smith, & Fischer, 2003) and personality constructs
terms of whether or not an adolescent decides to engage in risk (e.g., resilience; Lee & Cranford, 2008) should also be considered.
drinking; and b) paternal AUD may be of greater importance than It is important to note that these results were obtained across a
cognitive factors in terms of the frequency of engaging in risk particularly crucial period in adolescent development. Parents and
drinking.
adolescents were assessed when offspring were in middle child-
hood (ages 9 –11, M age ⫽ 10.6 years), early adolescence (ages
Limitations 12–14, M age ⫽ 13.5 years), and mid-adolescence (ages 15–17, M
The present study has several limitations. First, the sample was age ⫽ 16.5 years). Evidence showed that positive AEs increase
limited to adolescent boys, and there is some evidence that the over this period (Dunn & Goldman, 1998), particularly among
effects of parental alcohol involvement and AEs may be different adolescents exposed to peer and parental drinking (Cumsille,
for adolescent girls (see Pastor & Evans, 2003). Second, the Sayer, & Graham, 2000). Yet, in a recent review, Windle et al.
sample was limited to Caucasians, and evidence showed that (2008, p. S285) asserted that “Unfortunately, we do not yet have
ethnicity moderates some associations between expectancies and longitudinal data mapping the progression of expectancy endorse-
alcohol involvement (Chartier, Hesselbrock, & Hesselbrock, ment and its prediction of subsequent drinking among children 10
2009). Third, we relied on adolescents’ self-reports of alcohol use. to 12 years of age. This gap in the literature is an important one
Although self-report measures of alcohol involvement seem to that needs to be rectified.” Windle et al. noted that this period
have adequate reliability and validity (Babor, Steinberg, Anton, & usually involves the transition to middle school and adolescence;
Del Boca, 2000), there are numerous factors that influence the as such, “this transition may become a turning point for some
validity of self-reports, including age and forgetting (Brener, Billy, children, and their developmental trajectories may become char-
& Grady, 2003; Del Boca & Darkes, 2003), and these factors may acterized by maladaptive features.” The current study addresses
have a stronger effect on self-report in younger adolescents. this gap in the literature and demonstrates the unique longitudinal
Fourth, the restricted range/class of families in the “low risk” effects of maternal and paternal alcohol involvement and adoles-
group may have contributed to the nonsignificant correlations we cent AEs for specific dimensions of underage drinking.
394 CRANFORD, ZUCKER, JESTER, PUTTLER, AND FITZGERALD

References use: A test of three mediating mechanisms. Journal of Abnormal Psy-

chology, 102, 3–19.
Aas, H. N., Leigh, B. C., Anderssen, N., & Jakobsen, R. (1998). Two-year Chassin, L., Rogosch, F., & Barrera, M. (1991). Substance use and symp-
longitudinal study of alcohol expectancies and drinking among Norwe- tomatology among adolescent children of alcoholics. Journal of Abnor-
gian adolescents. Addiction, 93, 373–384. mal Psychology, 100, 449 – 463.
Anderson, K. G., Smith, G. T., & Fischer, S. F. (2003). Women and Chen, M. J., Grube, J. W., & Madden, P. A. (1994). Alcohol expectancies
acquired preparedness: Personality and learning implications for alcohol and adolescent drinking: Differential prediction of frequency, quantity,
use. Journal of Studies on Alcohol, 64, 384 –392. and intoxication. Addictive Behaviors, 19, 521–529.
Ary, D. V., Tildesley, E., Hops, H., & Andrews, J. A. (1993). The influence Christiansen, B. A., & Goldman, M. S. (1983). Alcohol-related expectan-
of parent, sibling, and peer modeling and attitudes on adolescent use of cies versus demographic/background variables in the prediction of ad-
alcohol. International Journal of the Addictions, 28, 853– 880. olescent drinking. Journal of Consulting and Clinical Psychology, 51,
Atkins, D. C., & Gallop, R. J. (2007). Rethinking how family researchers 249 –257.
model infrequent outcomes: A tutorial on count regression and zero- Christiansen, B. A., Goldman, M. S., & Inn, A. (1982). Development of
inflated models. Journal of Family Psychology, 21, 726 –735. alcohol-related expectancies in adolescents: Separating pharmacological
Babor, T. F., Steinberg, K., Anton, R., & Del Boca, F. (2000). Talk is from social-learning influences. Journal of Consulting and Clinical
cheap: Measuring drinking outcomes in clinical trials. Journal of Studies Psychology, 50, 336 –344.
on Alcohol, 61, 55– 63. Christiansen, B. A., Smith, G. T., Roehling, P. V., & Goldman, M. S.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable (1989). Using alcohol expectancies to predict adolescent drinking be-
distinction in social psychological research: Conceptual, strategic, and havior after one year. Journal of Consulting and Clinical Psychology,
statistical considerations. Journal of Personality and Social Psychology,
57, 93–99.
51, 1173–1182.
Colder, C. R., Chassin, L., Stice, E. M., & Curran, P. J. (1997). Alcohol
Bauman, K. E., Fisher, L. A., Bryan, E. S., & Chenoweth, R. L. (1985).
expectancies as potential mediators of parent alcoholism effects on the
Relationship between subjective expected utility and behavior: A lon-
development of adolescent heavy drinking. Journal of Research on
gitudinal study of adolescent drinking behavior. Journal of Studies on
Adolescence, 7, 349 –374.
Alcohol, 46, 32–38.
Collins, L. M. (2006). Analysis of longitudinal data: The integration of
Brener, N. D., Billy, J. O. G., & Grady, W. R. (2003). Assessment of
theoretical model, temporal design, and statistical model. Annual Review
factors affecting the validity of self-reported health-risk behavior among
of Psychology, 57, 505–528.
adolescents: Evidence from the scientific literature. Journal of Adoles-
Collins, L. M., & Graham, J. W. (2002). The effect of the timing and
cent Health, 33, 436 – 457.
spacing of observations in longitudinal studies of tobacco and other drug
Brook, J. S., Whiteman, M., Gordon, A. S., & Cohen, P. (1986). Some
use: Temporal design considerations. Drug and Alcohol Dependence,
models and mechanisms for explaining the impact of maternal and
68, S85–S96.
adolescent characteristics on adolescent stage of drug use. Developmen-
Cooper, M. L., Peirce, R. S., & Tidwell, M. C. O. (1995). Parental drinking
tal Psychology, 22, 460 – 467.
problems and adolescent offspring substance use: Moderating effects of
Brown, S. A., Christiansen, B. A., & Goldman, M. S. (1987). The Alcohol
demographic and familial factors. Psychology of Addictive Behaviors, 9,
Expectancy Questionnaire: An instrument for the assessment of adoles-
cent and adult alcohol expectancies. Journal of Studies on Alcohol, 48, 36 –52.
483– 491. Cumsille, P. E., Sayer, A. G., & Graham, J. W. (2000). Perceived exposure
Brown, S. A., Creamer, V. A., & Stetson, B. A. (1987). Adolescent alcohol to peer and adult drinking as predictors of growth in positive alcohol
expectancies in relation to personal and parental drinking patterns. expectancies during adolescence. Journal of Consulting and Clinical
Journal of Abnormal Psychology, 96, 117–121. Psychology, 68, 531–536.
Brown, S. A., Goldman, M. S., Inn, A., & Anderson, L. R. (1980). Darkes, J., Greenbaum, P. E., & Goldman, M. S. (2004). Alcohol expect-
Expectations of reinforcement from alcohol: Their domain and relation ancy mediation of biopsychosocial risk: Complex patterns of mediation.
to drinking patterns. Journal of Consulting and Clinical Psychology, 48, Experimental and Clinical Psychopharmacology, 12, 27–38.
419 – 426. Del Boca, F., & Darkes, J. (2003). The validity of self-reports of alcohol
Brown, S. A., Tate, S. R., Vik, P. W., Haas, A. L., & Aarons, G. A. (1999). consumption: State of the science and challenges for research. Addiction,
Modeling of alcohol use mediates the effect of family history of alco- 98, 1–12.
holism on adolescent alcohol expectancies. Experimental and Clinical DeLucia, C., Belz, A., & Chassin, L. (2001). Do adolescent symptomatol-
Psychopharmacology, 7, 20 –27. ogy and family environment vary over time with fluctuations in paternal
Buu, A., DiPiazza, C., Wang, J., Puttler, L. I., Fitzgerald, H. E., & Zucker, alcohol impairment? Developmental Psychology, 37, 207–216.
R. A. (2009). Parent, family, and neighborhood effects on the develop- Dunn, M. E., & Goldman, M. S. (1998). Age and drinking-related differ-
ment of child substance use and other psychopathology from preschool ences in the memory organization of alcohol expectances in 3rd-, 6th-,
to the start of adulthood. Journal of Studies on Alcohol and Drugs, 70, 9th-, and 12th-grade children. Journal of Consulting and Clinical Psy-
489 – 498. chology, 66, 579 –585.
Catanzaro, S. J., & Laurent, J. (2004). Perceived family support, negative Dunn, M. E., Lau, C., & Cruz, I. Y. (2000). Changes in activation of
mood regulation expectancies, coping, and adolescent alcohol use: Ev- alcohol expectancies in memory in relation to changes in alcohol use
idence of mediation and moderation effects. Addictive Behaviors, 29, after participation in an expectancy challenge program. Experimental
1779 –1797. and Clinical Psychopharmacology, 8, 566 –575.
Chartier, K. G., Hesselbrock, M. N., & Hesselbrock, V. M. (2009). Eth- Ellis, D. A., Zucker, R. A., & Fitzgerald, H. E. (1997). The role of family
nicity and adolescent pathways to alcohol use. Journal of Studies on influences in development and risk. Alcohol Health and Research
Alcohol and Drugs, 70, 337–345. World, 21, 218 –226.
Chassin, L., Curran, P. J., Hussong, A. M., & Colder, C. R. (1996). The Ennett, S. T., Foshee, V. A., Bauman, K. E., Hussong, A., Cai, L., Luz, H.,
relation of parent alcoholism to adolescent substance use: A longitudinal . . . DuRant, R. (2008). The social ecology of adolescent alcohol misuse.
follow-up study. Journal of Abnormal Psychology, 105, 70 – 80. Child Development, 79, 1777–1791.
Chassin, L., Pillow, D. R., Curran, P. J., Molina, B. S. G., & Barrera, M. Fitzgerald, H. E., Davies, W. H., & Zucker, R. A. (2002). Growing up in
(1993). Relation of parental alcoholism to early adolescent substance an alcoholic family: Structuring pathways for risk aggregation and
 PARENT DRINKING AND ADOLESCENT ALCOHOL EXPECTANCIES 395

theory-driven intervention. In R. J. McMahon & R. D. Peters (Eds.), The Leckman, J. F., Sholomskas, D., Thompson, W. D., Belanger, A., &
effects of parental dysfunction on children (pp. 127–146). New York: Weissman, M. M. (1982). Best estimate of lifetime psychiatric diagno-
Kluwer. sis: A methodological study. Archives of General Psychiatry, 39, 879 –
Fitzgerald, H. E., Zucker, R. A., & Noll, R. B. (1990). The Beverage 883.
Opinion Questionnaire. Unpublished manuscript, Michigan State Uni- Lee, H. H., & Cranford, J. A. (2008). Does resilience moderate the
versity, East Lansing, MI. associations between parental problem drinking and adolescents’ inter-
Fritz, M. S., & MacKinnon, D. P. (2007). Required sample size to detect nalizing and externalizing behaviors? A study of Korean adolescents.
the mediated effect. Psychological Science, 18, 233–239. Drug and Alcohol Dependence, 96, 213–221.
Goldman, M. S., & Darkes, J. (2004). Alcohol expectancy multiaxial Leigh, B. C., & Stacy, A. W. (2004). Alcohol expectancies and drinking in
assessment: A memory network-based approach. Psychological Assess- different age groups. Addiction, 99, 215–227.
ment, 16, 4 –15. Long, J. S. (1997). Regression models for categorical and limited depen-
Goldman, M. S., Del Boca, F. K., & Darkes, J. (1999). Alcohol expectancy dent variables. Thousand Oaks, CA: Sage.
theory: The application of cognitive neuroscience. In K. E. Leonard & Mann, L. M., Chassin, L., & Sher, K. J. (1987). Alcohol expectancies and
H. T. Blane (Eds.), Psychological theories of drinking and alcoholism the risk for alcoholism. Journal of Consulting and Clinical Psychology,
(2nd ed., pp. 203–246). New York: Guilford. 55(3), 411– 417.
Goldman, M. S., Greenbaum, P. E., & Darkes, J. (1997). A confirmatory Miller, P. M., Smith, G. T., & Goldman, M. S. (1990). Emergence of
test of hierarchical expectancy structure and predictive power: Discrimi- alcohol expectancies in childhood: A possible critical period. Journal of
nant validation of the Alcohol Expectancy Questionnaire. Psychological Studies on Alcohol, 51, 343–349.
Assessment, 9, 145–157. Muthen, L. K., & Muthen, B. O. (2007). Mplus user’s guide (5th ed.). Los
Graham, J. W., Olchowski, A. E., & Gilreath, T. D. (2007). How many Angeles, CA: Muthen & Muthen.
imputations are really needed? Some practical clarifications of multiple Newcomb, M. D., Chou, C. P., Bentler, P. M., & Huba, G. J. (1988).
imputation theory. Prevention Science, 8, 206 –213. Cognitive motivations for drug use among adolescents: Longitudinal
Handley, E. D., & Chassin, L. (2009). Intergenerational transmission of tests of gender differences and predictors of change in drug use. Journal
alcohol expectancies in a high-risk sample. Journal of Studies on Alco- of Counseling Psychology, 35, 426 – 438.
hol and Drugs, 70, 675– 682. Ohannessian, C. M., & Hesselbrock, V. M. (2004). Do alcohol expectan-
Hawkins, J. D., Catalano, R. F., & Miller, J. Y. (1992). Risk and protective cies moderate the relationship between parental alcoholism and adult
factors for alcohol and other drug problems in adolescence and early drinking behaviors? Addictive Behaviors, 29, 901–909.
adulthood: Implications for substance abuse prevention. Psychological Ohannessian, C. M., Hesselbrock, V. M., Kramer, J., Bucholz, K. K.,
Bulletin, 112, 64 –105. Schuckit, M. A., Kuperman, S., & Nurnberger, J. I., Jr. (2004). Parental
Henderson, M. J., Goldman, M. S., Coovert, M. D., & Carnevalla, N. substance use consequences and adolescent psychopathology. Journal of
(1994). Covariance structure models of expectancy. Journal of Studies Studies on Alcohol, 65, 725–730.
on Alcohol, 55, 315–326. Ouellette, J. A., Gerrard, M., Gibbons, F. X., & Reis-Bergan, M. (1999).
Horton, N. J., Kim, E., & Saitz, R. (2007). A cautionary note regarding Parents, peers, and prototypes: Antecedents of adolescent alcohol ex-
count models of alcohol consumption in randomized clinical trials. BMC pectancies, alcohol consumption, and alcohol-related life problems in
Medical Research Methodology, 7, 9. rural youth. Psychology of Addictive Behaviors, 13, 183–197.
Jacob, T., & Johnson, S. (1997). Parenting influences on the development Pastor, A. D., & Evans, S. M. (2003). Alcohol outcome expectancies and
of alcohol abuse and dependence. Alcohol Health and Research World, risk for alcohol use problems in women with and without a family
21, 204 –209. history of alcoholism. Drug and Alcohol Dependence, 70, 201–214.
Jacob, T., & Leonard, K. E. (1988). Alcoholic-spouse interaction as a Petraitis, J., Flay, B. R., & Miller, T. Q. (1995). Reviewing theories of
function of alcoholism subtype and alcohol consumption interaction. adolescent substance use: Organizing pieces in the puzzle. Psychological
Journal of Abnormal Psychology, 97, 231–237. Bulletin, 117, 67– 86.
Karazsia, B. T., & van Dulmen, M. H. M. (2008). Regression models for Reese, F. L., Chassin, L., & Molina, B. S. G. (1994). Alcohol expectancies
count data: Illustrations using longitudinal predictors of childhood in- in early adolescents: Predicting drinking behavior from alcohol expect-
jury. Journal of Pediatric Psychology, 33, 1076 –1084. ancies and parental alcoholism. Journal of Studies on Alcohol, 55,
Kidorf, M., Sherman, M. F., Johnson, J. G., & Bigelow, G. E. (1995). 276 –284.
Alcohol expectancies and changes in beer consumption of first-year Richter, L., & Richter, D. M. (2001). Exposure to parental tobacco and
college students. Addictive Behaviors, 20, 225–231. alcohol use: Effects on children’s health and development. American
Killen, J. D., Hayward, C., Wilson, D. M., Haydel, K. F., Robinson, T. N., Journal of Orthopsychiatry, 71, 182–203.
Taylor, C. B., . . .Varady, A. (1996). Predicting onset of drinking in a Robins, L. N., Helzer, J. E., Croughan, J., & Ratcliff, K. S. (1981). National
community sample of adolescents: The role of expectancy and temper- Institute of Mental Health Diagnostic Interview Schedule: Its history,
ament. Addictive Behaviors, 21, 473– 480. characteristics, and validity. Archives of General Psychiatry, 38, 381–
King, K. M., & Chassin, L. (2004). Mediating and moderated effects of 389.
adolescent behavioral undercontrol and parenting in the prediction of Rubin, D. B. (1987). Multiple imputation for nonresponse in surveys. New
drug use disorders in emerging adulthood. Psychology of Addictive York, NY: John Wiley & Sons, Inc.
Behaviors, 18, 239 –249. SAS Institute Inc. (2004). SAS/STAT 9.1 user’s guide. Cary, NC: SAS
Kraemer, H. C., Stice, E., Kazdin, A. E., Offord, D. R., & Kupfer, D. J. Institute Inc.
(2001). How do risk factors work together? Mediators, moderators, and Schaefer, J. L. (1997). Analysis of incomplete multivariate data. New
independent, overlapping, and proxy risk factors. American Journal of York, NY: Chapman & Hall.
Psychiatry, 158, 848 – 856. Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the state
Kraus, D., Smith, G. T., & Ratner, H. H. (1994). Modifying alcohol-related of the art. Psychological Methods, 7, 147–177.
expectancies in grade-school children. Journal of Studies on Alcohol, 55, Scheier, L. M. (2001). Etiologic studies of adolescent drug use: A com-
535–542. pendium of data resources and their implications for prevention. Journal
Lambert, D. (1992). Zero-inflated Poisson regression, with an application of Primary Prevention, 22, 125–168.
to defects in manufacturing. Technometrics, 34, 1–14. Scheier, L. M., & Botvin, G. J. (1997). Expectancies as mediators of the
396 CRANFORD, ZUCKER, JESTER, PUTTLER, AND FITZGERALD

effects of social influences and alcohol knowledge on adolescent alcohol Developmental processes and mechanisms between 10 and 15 years of
use: A prospective analysis. Psychology of Addictive Behaviors, 11, age. Pediatrics, 121, S273–S289.
48 – 64. Wood, M. D., Read, J. P., Mitchell, R. E., & Brand, N. H. (2004). Do
Schell, T. L., Martino, S. C., Ellickson, P. L., Collins, R. L., & McCaffrey, parents still matter? Parent and peer influences on alcohol involvement
D. (2005). Measuring developmental changes in alcohol expectancies. among recent high school graduates. Psychology of Addictive Behaviors,
Psychology of Addictive Behaviors, 19, 217–220. 18, 19 –30.
Selzer, M. L., Vinokur, A., & van Rooijen, L. (1975). A self-administered Yu, J. (2003). The association between parental alcohol-related behaviors
Short Michigan Alcoholism Screening Test (SMAST). Journal of Stud- and children’s drinking. Drug and Alcohol Dependence, 69, 253–262.
ies on Alcohol, 36, 117–126. Zhang, L., Welte, J. W., & Wieczorek, W. F. (1999). The influence of
Sher, K. J., & Gotham, H. J. (1999). Pathological alcohol involvement: A parental drinking and closeness on adolescent drinking. Journal of
developmental disorder of young adulthood. Development and Psycho- Studies on Alcohol, 60, 245–251.
pathology, 11, 933–956. Zucker, R. A. (1991). Scaling the developmental momentum of alcoholic
Sher, K. J., Grekin, E. R., & Williams, N. A. (2005). The development of process via the lifetime alcohol problems score. Alcohol and Alcoholism,
alcohol use disorders. Annual Review of Clinical Psychology, 1, 493– 26 (Supplement No. 1), 505–510.
523. Zucker, R. A. (1994). Pathways to alcohol problems and alcoholism: A
Sher, K. J., Walitzer, K. S., Wood, P. K., & Brent, E. E. (1991). Charac- developmental account of the evidence for multiple alcoholisms and for
teristics of children of alcoholics: Putative risk factors, substance use and contextual contributions to risk. In R. A. Zucker, J. Howard & G. M.
abuse, and psychopathology. Journal of Abnormal Psychology, 100, Boyd (Eds.), The development of alcohol problems: Exploring the bio-
427– 448. psychosocial matrix of risk (NIAAA Research Monograph No. 26, pp.
Sher, K. J., Wood, M. D., Wood, P. K., & Raskin, G. (1996). Alcohol 255–289). Rockville, MD: U.S. Department of Health and Human
outcome expectancies and alcohol use: A latent variable cross-lagged Services.
panel study. Journal of Abnormal Psychology, 105, 561–574. Zucker, R. A. (2006). Alcohol use and the alcohol use disorders: A
Shrout, P. E., & Bolger, N. (2002). Mediation in experimental and non- developmental-biopsychosocial systems formulation covering the life
experimental studies: New procedures and recommendations. Psycho- course. In D. Cicchetti & D. J. Cohen (Eds.), Developmental psychopa-
logical Methods, 7, 422– 445. thology, Vol 3: Risk, disorder, and adaptation (2nd ed., pp. 620 – 656).
Sinharay, S., Stern, H. S., & Russell, D. (2001). The use of multiple Hoboken, NJ: John Wiley & Sons, Inc.
imputation for the analysis of missing data. Psychological Methods, 6, Zucker, R. A., Donovan, J. E., Masten, A. S., Mattson, M. E., & Moss,
317–329. H. B. (2008). Early developmental processes and the continuity of risk
Smith, G. T., Goldman, M. S., Greenbaum, P. E., & Christiansen, B. A. for underage drinking and problem drinking. Pediatrics, 121, S252–
(1995). Expectancy for social facilitation from drinking: The divergent S272.
paths of high-expectancy and low-expectancy adolescents. Journal of Zucker, R. A., Fitzgerald, H. E., & Moses, H. D. (1995). Emergence of
Abnormal Psychology, 104, 32– 40. alcohol problems and the several alcoholisms: A developmental per-
Spiegler, D. L. (1983). Children’s attitudes toward alcohol. Journal of spective on etiologic theory and life course trajectory. In D. Cicchetti &
Studies on Alcohol, 44, 545–552. D. J. Cohen (Eds.), Developmental psychopathology: Risk, disorder, and
Stevens, J. (1996). Applied multivariate statistics for the social sciences adaptation (Vol. 2, pp. 677–711). New York: John Wiley & Sons.
(3rd ed.). Mahwah, NJ: Erlbaum. Zucker, R. A., Fitzgerald, H. E., Refior, S. K., Puttler, L. I., Pallas, D. M.,
White, H. R., Johnson, V., & Buyske, S. (2000). Parental modeling and & Ellis, D. A. (2000). The clinical and social ecology of childhood for
parenting behavior effects on offspring alcohol and cigarette use: A children of alcoholics: Description of a study and implications for a
growth curve analysis. Journal of Substance Abuse, 12, 287–310. differentiated social policy. In H. E. Fitzgerald, B. M. Lester & B. S.
Wills, T. A., Sandy, J. M., Yaeger, A., & Shinar, O. (2001). Family risk Zuckerman (Eds.), Children of addiction: Research, health, and public
factors and adolescent substance use: Moderation effects for tempera- policy issues (pp. 109 –141). New York: Routledge Falmer Press.
ment dimensions. Developmental Psychology, 37, 283–297. Zucker, R. A., Kincaid, S. B., Fitzgerald, H. E., & Bingham, C. R. (1995).
Wills, T. A., & Yaeger, A. M. (2003). Family factors and adolescent Alcohol schema acquisition in preschoolers: Differences between chil-
substance use: Models and mechanisms. Current Directions in Psycho- dren of alcoholics and children of nonalcoholics. Alcoholism: Clinical
logical Science, 12, 222–226. and Experimental Research, 19, 1011–1017.
Windle, M. (1996). Effect of parental drinking on adolescents. Alcohol
Health and Research World, 20, 181–184. Received September 21, 2009
Windle, M., Spear, L. P., Fuligni, A. J., Angold, A., Brown, J. D., Pine, D., Revision received March 26, 2010
. . .Dahl, R. E. (2008). Transitions into underage and problem drinking: Accepted April 5, 2010 䡲