Mechanisms of Association Between Paternal Alcoholism and Abuse of Alcohol and Other Illicit Drugs Among Adolescents

Abstract

The current study examines the effect of paternal alcohol problems on adolescent use of alcohol and other illicit drugs as a function of maternal communication, as well as adolescent social and coping skills (N = 145). Structural equation modeling (SEM) analyses indicated that adolescents with a paternal history of alcohol problems reported higher levels of problematic communication with their mothers than did adolescents with fathers who were light drinkers or who were non-drinkers. Moreover, problematic maternal communication functioned as a partial mediator for both adolescent negative social skills and self-blame coping skills. Adolescents who reported using self-blame coping skills more frequently also tended to use alcohol more frequently. Findings suggest that these high-risk adolescents might need more intensive and targeted social and psychological services in their schools and communities.

INTRODUCTION

Substance use by adolescents remains a critical and prevalent health problem in the United States (Bukstein & Winters, 2004). Recent national data from the 2009 Monitoring the Future Survey indicates that the proportion of students involved in any consumption of alcohol in the past 30 days was 14.9%, 30.4%, and 43.5% in grades 8, 10, and 12, respectively. The proportion of students involved in any use of illicit drugs (including marijuana) in the past 30 days was 8.1%, 17.8%, and 23.3% in grades 8, 10, and 12, respectively.

The prevalence of alcohol and other drug (AOD) use is even higher among certain high-risk subgroups of adolescents. One of these high-risk groups is children of alcoholics (COAs). Recent epidemiological data (2009 National Survey on Drug Use and Health based on data from 2002–2007) indicated that almost 7.3 million children lived with a parent who was dependent on or abused alcohol. Other epidemiological data (Ramisetty-Mikler & Caetano, 2004) indicate that the number of children in the United States under the age of 18 exposed to parental alcohol use problems is higher, 13.6 million, which represents 19% of all children in the United States. An especially significant behavioral consequence of COA status is the heightened risk for substance use problems (Chalder, Elgar, & Bennett, 2006; Hoffmann & Cerbone, 2002; Latendresse et al., 2008; Rubio-Stipec, Bird, Canino, Bravo, & Algria, 1991). COA status has been repeatedly associated with increased risk for substance use disorders (Ahmadi, Arabi, & Mansouri, 2003; Brook et al., 2003; Chassin, Curran, Hussong, & Colder, 1992; Peleg-Oren, Rahav, & Teichman, 2008), low self-esteem and feelings of guilt and loneliness (Johnson & Pandina, 1991; Ritter, Stewart, Bernet, Coe, & Brown, 2002; Zanoti-Jeronymo & Pimenta Carvalho, 2005), problems in cognitive development (Carle & Chassin, 2004; Moss, Vanyukov, Majumder, Kirisci, & Tarter, 1995; Mylant, Ide, Cuevas, & Meehan, 2002; Noll, Zucker, Fitzgerald, & Curtis, 1992), and deficits in verbal abilities and academic performance (Noll et al., 1992; Sher, Walitzer, Wood, & Brent, 1991). Less-researched problems in social development associated with COA status include difficulty maintaining normal social contacts, which has been theorized to be a consequence of unreliability in parent-child relationships (Christozov & Toteva, 1989; Elkins, McGue, Malone, & Iacono, 2004; Jones & Houts, 1992; Nurco, Blatchley, Hanlon, & O’Grady, 1999). Sher’s model (1991) proposed that parental alcoholism increases risk for alcohol use disorders, among other risks, through poor parenting that leads to conduct problems and emotional and behavioral problems and thus to high risk for alcohol use. While both genetic and non-genetic influences appear to contribute to the risk for substance use problems among COAs, the exact mechanisms of the relationship between COA status and substance use problems is not well understood (Clark, Cornelius, Wood, & Vanyukov, 2004; Johnson, Van Den Bree, & Pickens, 1996). The current study examines components of Sher’s model among adolescents in an attempt to explore key variables (i.e., maternal communication, coping skills, and social skills) previously established as influential in underage drinking and drug abuse problems.

To date, there have been several investigations examining the impact of both paternal and maternal alcohol and other drug use on COA functioning (for a review, see Ellis, Zucker, & Fitzgerald, 1997; Harter, 2000; Latendresse et al., 2008). Open communication with the parents, parenting support, and control have been found to be negatively associated with emotional distress (Operario, Tschann, Flores, & Bridges, 2006), psychological maladjustment (Suchman, Rounsaille, Decoste, & Luthar, 2007), and adolescent drinking (Foxcroft & Lowe, 1991). In the majority of studies, paternal alcohol problems affected the functioning of the mother as well as the children and demonstrated strong associations with unhealthy adolescent functioning (Jacob & Johnson, 1997; Pollock, Schneider, Gabrielli, & Goodwin, 1987). Studies have found that paternal alcoholism is a stronger predictor of adolescent substance use problems than is maternal alcoholism (Brook et al., 2003; Chassin et al., 1996). Moreover, research has suggested that maternal influences (i.e., affection, open communication, and competence) may mitigate the negative effects of paternal alcoholism on adolescents (McCord, 1991; Moos, Finney, & Cronkite, 1990; West & Prinz, 1987). Maternal open communication reflects a mother’s general tendency to be aware of the child’s needs, have open communication, express approval, and direct positive behaviors toward the child (Zhou et al., 2002).

Conversely, problematic communication may heighten the risk for substance use problems in adolescents (Cohen, Richardson, & LaBree, 1994; Foxcroft & Lowe, 1991; Kandel & Davis, 1996; Ruchkin, Koposov, Eisemann, & Huggolf, 2002). A recent longitudinal study found that problematic maternal communication with children of alcoholic fathers was predictive of lower child self-regulation at early age (Eiden, Colder, Edwards, & Leonard, 2009). While, based on the literature, parental support and open communication might mediate the effect of parental alcoholism on offspring substance use (King & Chassin, 2004), there is a dearth of empirical research on this topic. In addition, most of the literature examined parental communication without specifying maternal or paternal. The current study extends the literature by investigating whether maternal communication mediates the relation between paternal alcoholism and substance use among adolescents.

Moreover, learning dysfunctional coping behaviors is one of the more serious consequences of growing up with parental alcoholism (Fox & Gilbert, 1994). COAs rely more heavily on escape/avoidance coping than do non-COAs across a variety of stressful situations (Stratton, 1998). As adolescents, COAs with substance abuse problems reported engaging in significantly more avoidance stress-coping and significantly less temptation-coping than demographically matched adolescents without substance use problems (Wagner, Myers, & McIninch, 1999). In other studies, emotion-focused, self-blame relief-oriented coping stress was related to adolescent substance use whereas problem-focused coping was unrelated to substance use (Labouvie, 1986, 1987; Wagner, 1993; Wagner et al., 1999; Wills, 1986; Wills, Vaccaro, McNamara, & Hirky, 1996).

Little is known about the social skills of COAs. Norton (2000) found two specific family patterns that appear to promote positive social skills among COAs: (1) father’s frequent presence in the home, coupled with positive family interaction, and (2) high family income, which was associated with positive socialization experiences outside the nuclear family. However, Segrin and Menees (1996) did not find any significant relations between parental alcohol abuse and children’s social skills. Hussong, Zucker, Wong, Fitzgerald, and Puttler (2005) explained the disparate findings by suggesting that the association between social competence and adolescent AOD (alcohol and other drug) outcomes operates independently from COA status. Adolescents who were more socially confident, assertive, and had better communication skills reported less drinking and smoking. Adolescents with poor social competence may turn to alcohol use and smoking as they perceive it to provide important social benefits (Griffin, Epstein, Botvin, & Spoth, 2001).

The current study builds on the few available studies by using a multivariate model including problematic maternal communication, adolescent coping skills, and social skills as mediators of the association between paternal history of alcoholism and adolescent alcohol and other drug (AOD) use. We hypothesized that adolescents with paternal alcohol problems, problematic maternal communication, poorer social skills, and a heavier reliance on self-blame coping skills will demonstrate greater AOD use problems (see Figure 1 for proposed model).

FIGURE 1.

Hypothesized model.

Notes. Rectangles represent observed variables. The straight lines with arrows from observed variables represent presumed causal pathways. Father Alcohol Use = adolescent’s report of father’s level of alcohol use.

METHODS

Participants, Procedures, and Study Description

Given the heightened liability for substance use problems among adolescents with parents who have a history of substance use problems, COA status is common among adolescents who present for school-based interventions. Therefore, the data for the current study were culled from the baseline assessment of a National Institute on Alcohol Abuse and Alcoholism (NIAAA)-funded randomized clinical trial (R01 AA10246; PI: Wagner) examining the efficacy of a brief 10-session school-based group intervention (Teen Intervention Project [TIP]) for underage drinking (Wagner & Macgowan, 2006).

Participants were referred to TIP primarily by teachers and school counselors. A small proportion (<10%) were self-referred or parent-referred. Once referred, active parental consent for participation was obtained, using procedures approved by the university’s Institutional Review Board. Students whose parents consented to have their child involved in the program met during school hours with the clinical assessor, who explained the program and sought student assent. Once consented, participants were administered a baseline assessment which included measures of demographics, substance use, social functioning, and familial functioning. Analyses for the present study were conducted with data collected during the baseline assessment.

Inclusion criteria for the current study were the following: (1) at least 6 occasions of alcohol use in the 6-month period prior to enrollment in the school-based intervention and (2) 6 or more in-person contacts per month with their fathers (this ensured sufficient contact with fathers for meaningful paternal influence). From a total of 214 students who participated in the original study, 145 (68%) students met these additional inclusion criteria.

Measures

Paternal drinking status

Participants reported on their father’s alcohol use patterns using a revised version of the Social Support Questionnaire, a brief measure developed by Richter, Brown, and Mott (1991) that gathers information on the substance use patterns of family members and peers. The adolescent was asked, “How would you describe your father’s Alcohol Use Pattern: 1. heavy drinker; 2. light drinker; or 3. non-drinker.” As established by Richter and colleagues (1991), respondents were given anchors for categorizing alcohol use as follows: (1) heavy drinkers drank frequently and exhibited problems in their life due to their drinking; (2) light drinkers drank infrequently or socially and had not experienced problems due to their drinking; and (3) non-drinkers did not drink alcohol at all. Based on the adolescent’s report, study participants were classified into three groups with regard to their paternal drinking status: (1) fathers who are heavy drinkers; (2) fathers who are light drinkers; and (3) non-drinking fathers. Paternal drinking status was reflected in the model by using dummy variables with dummy coding.

Maternal communication

The Parent-Adolescent Communication Scale (PACS; Barnes & Olson, 1982), a child self-report measure, was used to evaluate maternal-adolescent communication. The PACS has good reliability and validity (Barnes & Olson, 1982; Olson et al., 1983) and has been used in many previous studies of parent-adolescent communication (e.g., Bhusan & Shirali, 1992; Knight, Tein, Shell, & Roosa, 1992). For the purpose of this study, we used the “Problems in Maternal-Adolescent Communication” 10-item subscale of the PACS, in which participants rate their communication with their mother when dealing with problems (e.g., “my mother insults me when she is angry with me”). The responses were rated on a 5-point scale, from “strongly disagree” (1) to “strongly agree” (5). For the current study the alpha reliability coefficient for the “Problems in Maternal-Adolescent Communication” scale was .73.

Social skills

The Teenage Inventory of Social Skills (TISS; Inderbitzen & Foster, 1992) was used to evaluate social skills. TISS includes two subscales: a 20-item Positive Behavior Scale (e.g., “I help other classmates with their homework when they ask me for help”) and a 20-item Negative Behavior Scale (e.g., “I ignore classmates when they tell me to stop doing something”). Inderbitzen and Foster (1992) documented high internal consistency reliability (.88 for both positive and negative scales) and good two-week test-retest reliability (.90 for the positive scale and .72 for the negative scale). For the present study, a preliminary limited information confirmatory factor analysis was conducted using one latent variable (social skills) with two indicators (i.e., Positive Behavior Scale and Negative Behavior Scale). The responses were rated on a 6-point scale, from “not at all” (1) to “totally” (6). However, good fit indices were not obtained for this type of factor structure. After inspecting the correlational matrix between the two scales, it was found that even though they were statistically significantly correlated (r = −.19, p < .05) with each other, the correlation was not large and they were also each differentially correlated with other variables in the larger model. Therefore, it was decided to include separate manifest variables of the Positive Behavior Scale and Negative Behavior Scale in the model and correlate their residual terms.

Coping skills

The Revised Ways of Coping Checklist (RWCCL; Vitaliano, Russo, Carr, Mauiro, & Becker, 1985), designed to assess strategies utilized by an individual in response to stressful situations, was used to evaluate stress coping. In the current study, we used the Self-Blame scale to represent emotion-focused relief-oriented coping (e.g., “In a situation that you are alone and feeling emotionally bad, how much have you criticized or lectured yourself?”). The responses were rated on a 4-point scale, from “never use” (1) to “regularly use” (4). Previous studies have demonstrated that the RWCCL displays good construct and concurrent validity, both with adult (Vitaliano et al., 1985) and adolescent (Wagner, 1993) samples. Previous estimates of internal consistency (coefficient alpha) range from .71 to .88 (Vitaliano et al., 1985; Wagner, 1993). In the current study, internal consistency of the Self-Blame scale was adequate (∞ = .73).

Adolescent AOD use

Timeline Followback Interview (TLFB; Sobell & Sobell, 1992, 1996; Sobell, Sobell, & Ward, 1980) measured adolescent AOD use. The TLFB is a widely used and well-known research and clinical assessment tool with acceptable reliability and validity for various groups of individuals with substance use problems. Substance-consumption information was collected using a calendar format to provide temporal cues (e.g., holidays, special occurrences) to assist in recall of days when substances were used. With regard to alcohol, daily-use information was collected in number of standard drinks per day and frequency of alcohol use. In the current study, we used frequency of alcohol use for the 30 days prior to assessment as our primary substance use variable. The correlation between 30-day frequency of alcohol use and total number of drinks for the same period of time (30 days) was substantial (r = .68, p < .01). We conducted parallel analyses for both variables and the results were unchanged. All analyses reported are for the frequency of alcohol use variable. Daily drug use was recorded as positive or negative only, as there is currently no standard way of recording drug use.

Statistical Analysis

In the first step of the statistical analyses, we used descriptive statistics to assess the demographic variables, the key variables of interest, maternal communication, social skills, coping skills, and AOD use for the past 30 days. Second, we conducted structural equation modeling (SEM) to represent the hypothesized sequential links among paternal drinking status, problematic maternal communication, positive and negative social skills and self-blame coping skills and the AOD use outcomes. Third, in order to test for possible demographic confounds, we conducted additional exploratory modeling including those variables as covariates for all of the endogenous variables.

We used the logic of the joint significance test as a basis for stating mediation (MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002). This is a superior test of mediation relative to the traditional approach of Baron and Kenny (1986). The logic is that if all path coefficients in a mediated chain are statistically significant, then mediation is declared.

RESULTS

Descriptive Analyses

The mean age of the 145 adolescents in our study was 15.36 years old (1.51 SD), 56.6% were male, 68% were non-Hispanic white, 22% were Hispanic, 7% were black, and 5% were “other.” Table 1 presents means and standard deviations for all of the continuous variables included in the model for each category of Paternal Drinking Status. As such, 33% reported being children of fathers who are heavy drinkers (N = 47), 37% children of light-drinking fathers (N = 54), and 30% children of non-drinking fathers (N = 44).

TABLE 1.

Descriptives for the Continuous Variables in the Model by Paternal Drinking Status

	Heavy Drinkers N = 47		Light Drinkers N = 54		Non-Drinkers N = 44
	Mean	SD	Mean	SD	Mean	SD
Maternal Communication
Problematic communication	1.34	.48	1.52	.50	1.50	.51
Social Skills
Positive	81.06	15.39	83.28	15.61	74.46	15.87
Negative	56.96	15.68	52.44	15.34	53.11	16.92
Coping Skills
Self-blame	6.72	2.16	6.52	2.09	6.26	2.53
Frequency of Use (prev. 30 days)
Alcohol	2.66	3.21	2.19	2.70	1.30	2.03
Illicit drugs	5.89	8.68	5.11	7.99	4.24	7.18

Primary Analyses

The primary analytic approach involved the use of structural equation modeling (SEM) using AMOS version 17. The model (Figure 1) was statistically over-identified. There were no missing data. Prior to analysis, the data for the covariance matrix were evaluated for multivariate outliers by examining leverage indices for each individual and defining an outlier as a leverage score four times greater than the mean leverage. No outliers were observed. Multivariate normality was evaluated using Mardia’s index. The multivariate test was statistically significant (p < 0.05) indicating the presence of non-normality. Examination of univariate indices of skewness and kurtosis revealed only one absolute skewness value above 1.80 for the frequency of drug use variable (2.13) and two absolute values of kurtosis greater than 2.0 (3.51 and 3.85 for two of the measures of substance use). To accommodate the non-normality that was present, the model was evaluated using bootstrapping with 2,000 bootstrap replicates. The reported parameter estimates and standard errors are from the bootstrapped analyses.

In addition, we conducted a power analysis to determine the possibility of a Type II error for statistically nonsignificant path coefficients. A limited information approach was used to obtain a rough approximation of statistical power (Jaccard & Wan, 1996). The power analysis was based on an alpha level of 0.05 and a two-tailed test using three predictors for a multiple regression analysis. The sample size of 145 yielded an acceptable power coefficient of .85.

The Bollen-Stine p value for the proposed model was statistically significant (p < 0.001), suggesting poor model fit. The predominant source of ill fit was the assumption of uncorrelated residuals for negative coping skills and negative social skills. Since numerous other variables outside the conceptual framework of this model might serve as common causes for each of these constructs (e.g., self-esteem, peer relationships), it was conceptually justified to correlate their residual terms.

The revised model with correlated error was re-fit to the data. The Bollen-Stine bootstrapped chi-square test yielded a p value greater than .319. The root mean square error of approximation (RMSEA) was .04. The comparative fit index (CFI) was .98. The standardized root mean square residual (RMR) was 0.04. All indices suggest an overall good-fitting model. More focused analyses revealed no absolute standardized residuals greater than 1.96 and no theoretically meaningful modification indices of notable size. No offending estimates were evident. Figure 2 presents the parameter estimates for the final model. Both unstandardized and standardized path coefficients are presented with standardized coefficients in parentheses. All residuals are in standardized metrics. Correlations between the exogenous variables are omitted for purposes of figure clarity. Key results are addressed in the following sections.

FIGURE 2.

Final SEM model.

Notes. The values along the pathways are unstandardized path coefficients. The values inside the parentheses are the standardized path coefficients. The values in the circles represent standardized error variances (residual terms). ^TTrending toward significant (p < .08). ^aPaternal Drinking Status dummy-coded variables, reference group is Non-User Fathers. *p < .05.

Problematic maternal communication

The participants with fathers who are heavy drinkers reported problematic communication with their mothers statistically significantly more frequently than both the group with fathers who are light drinkers and the group with fathers who are non-drinkers. Specifically, the mean difference on problematic maternal communication between the group with fathers who are heavy drinkers and the group with fathers who are light drinkers was 2.81 (p < .05, 95% CI = .81 to 4.83) and between the group with fathers who are heavy drinkers and the non-drinkers group was 2.55 (p < .05, 95% CI = −4.73 to −.37). The mean difference on problematic maternal communication between the group with fathers who are non-drinkers and the group with light drinker fathers was not statistically significant. This pattern of significant differences maintained when a Holm-modified Bonferroni procedure was used to control the family-wise error rate at .95.

Adolescents’ social and coping skills

Results indicated that the mean difference on positive social skills between the group with light drinker fathers and the group with fathers who are non-drinkers was statistically significant (mean difference = 7.72, p < .05, 95% CI = −12.90 to −21.3), indicating that the group with light drinker fathers reported greater positive social skills than the group with fathers who are non-drinkers. The two other contrasts were not statistically significantly different. In addition, those who reported more problematic maternal communication also tended to report both higher negative social skills and self-blame coping skills. Specifically, for every one unit that problematic maternal communication goes up, negative social skills increases by .61 units and self-blame coping skills increase by .09 units respectively, holding all other variables in the model constant. Therefore, when applying the joint significance test (MacKinnon, Lockwood, Hoffman, West, & Sheets, 2002), there is evidence that problematic maternal communication functions as a partial mediator between paternal drinking status and both negative social skills and self-blame coping skills.

Substance use outcomes

Those who reported self-blame coping skills more frequently were more likely to use alcohol (unstandardized path coefficient = .25). In addition, a few other pathways that were related to the frequency of alcohol and of drugs use were trending toward significance (p < .08). Negative social skills was related to both frequency of drug use (unstandardized path coefficient = .08) and of alcohol use (unstandardized path coefficient = .03). Also, self-blame was related to frequency of drug use (unstandardized path coefficient = .48).

Supplemental Analyses

Although we did not expect any differences by gender, age, and/or ethnicity, we included these variables as covariates for all endogenous variables in order to avoid specification error. All significant paths remained significant even after controlling for these demographic covariates. In addition, the model was tested including maternal alcohol use as a covariate for all of the endogenous variables to check if the previously observed pattern held. Again, all significant paths remained significant.

DISCUSSION

Using a sample of 145 adolescents participating in an NIAAA-funded randomized clinical trial (RCT) in a high school-based underage-drinking intervention, the current study revealed that adolescents with a paternal history of alcohol use problems reported higher problematic communication with their mothers than adolescents with fathers who were light drinkers and non-drinkers. Moreover, problematic maternal communication functioned as a partial mediator for both adolescent negative social skills and self-blame coping skills. Self-blame coping skills predicted frequency of adolescent alcohol abuse.

These results are consistent with previous studies, which have demonstrated that there is substantially elevated risk for alcohol use and alcohol use problems among COAs (Barnow, Schuckit, Lucht, John, & Freyberger, 2002). The familial nature of alcohol use disorders is likely to reflect both the influence of genetic and psychosocial factors. Having a family history of alcohol use has been connected to an increased psychosocial risk for children (Barnard & McKeganey, 2004; Chassin, Pillow, Curran, Molina, & Barrera, 1993; Locke & Newcomb, 2004; Sher et al., 1991), and results from the current study confirm these previous findings. However, the current study expands upon previous research by examining maternal communication, and social and self-blame coping skills as mediators between paternal alcohol drinking and adolescent AOD use.

Parents who use praise, show affection, and open communication are modeling a positive behavior which deemphasizes aggression and punitive actions. Our findings that problematic maternal communication functioned as a partial mediator for both adolescent negative social skills and self-blame coping skills, and that self-blame coping skills predicted frequency of adolescent alcohol abuse, are consistent with the finding that problematic communication may heighten the risk for substance use problems in adolescents (Cohen et al., 1994; Foxcroft & Lowe, 1991; Kandel & Davis, 1996; Ruchkin et al., 2002). However, this form of mediation differs from the findings of King and Chassin (2004), who found that the effects of parental alcoholism on the development of drug use disorders were partially mediated by behavioral undercontrol and parental discipline. When undercontrol and discipline were added to King and Chassin’s mediational models, the differences between COAs and non-COAs became marginally significant.

Maternal communication is an important variable in the development of a secure attachment style between the mother and the child (Ainsworth, Belhar, Waters, & Wall, 1978). Secure attachment has been found to predict adaptive social and emotional functioning, while insecure attachment has been linked to poor social and emotional functioning. One potentially critical pathway by which paternal alcoholism may affect adolescent adjustment is through its effect on mothers. Paternal alcoholism may deplete maternal resources typically allocated to childrearing, with the net effect being a decrease in maternal communication. The continuing test of such mechanisms of association is an important topic for future research.

Results from the current study showed that adolescents with non-drinker fathers reported lower levels of positive social skills than adolescents with fathers who were light drinkers. This finding may reflect the widespread use of alcohol in social situations among adults; arguably, an aspect of adult social competence is non-problematic drinking in social situations. At least for some abstinent adults, reasons for not drinking may include religious prohibitions, being in recovery from alcohol use problems, or other types of health concern, each of which may be associated with atypical or uncommon socialization practices. This is an interesting and important finding, and it deserves further study because of its implications for adolescent development.

Although the current study adds to our understanding of mechanisms of association between paternal alcoholism and AOD use among adolescents, some limitations exist. First, the data were cross-sectional and do not permit causal inference. Second, using self-report scales with regard to adolescents’ perception of parental drinking may be subject to under- or overreport-ing response biases among teenagers (Roosa, Michaels, Groppenbacher, & Gerrsten, 1993). At the same time, a strength of the study with regard to self-report measures of adolescent functioning is that they are based on adolescents’ own perceptions of their social and emotional environments. These perceptions may be even more important than what actually occurs in predicting adolescent functioning (i.e., adolescents’ perceptions are reality, at least as far as the teen is concerned). Second, our sample of referred substance-involved individuals may not be representative of the general population of adolescents who are COA or AOD users. Finally, as a secondary analysis this study specifically focused on maternal communication, only one of numerous parenting variables—such as supervision, discipline, and control—that may influence adolescent behavior. In addition, it was not possible to ascertain from the data the reason for abstinence in the group of fathers who were non-drinkers. Future research should attempt to examine the possible differences between fathers who have never drunk and those who are recovering alcoholics.

Despite these limitations, the current study further demonstrated, as Sher (1991) suggested, that paternal alcoholism may affect offspring outcomes through many diverse pathways. Future research should examine the mechanisms of association between paternal alcoholism and AOD use among adolescents by means of a longitudinal study in order to ascertain (1) the degree of stability or change in the relations among these variables and (2) their influence on the adolescent’s AOD use at various stages of development. Given the findings of the current study, important avenues for future research should include attempts to identify additional personal characteristics (e.g., perceived stress) and family characteristics (e.g., parenting style) that may contribute to adolescents’ AOD use.