OUP user menu


Sven Barnow, Gabriele Schultz, Michael Lucht, Ines Ulrich, Ulrich-W. Preuss, Harald-J. Freyberger
DOI: http://dx.doi.org/10.1093/alcalc/agh048 213-219 First published online: 13 April 2004


Aims: To investigate (1) whether aggressive and delinquent behaviour problems predict subsequent adolescent drinking behaviour; and (2) to what extent this association is mediated by alcohol expectancies and/or peer delinquency/substance use. Methods: 147 adolescents (approximately 15 years old) were interviewed with regard to their drinking behaviour. In addition, several self-rating questionnaires were given to gather information regarding the peers of these children. Results: As proposed by the Acquired Preparedness Model (APM), we found that behavioural problems were related to quantity and frequency of alcohol consumed, and that this relationship was mediated by alcohol expectancies. Regarding peer relations, we found positive correlations between drinking behaviour and peer delinquency/substance use, aggression/delinquency and alcohol expectancies. Furthermore, the association between behavioural problems and drinking decreased dramatically if peer delinquency/substance use was accounted for. A hierarchical regression analysis revealed that both alcohol expectancies and peer delinquency/substance use predicted alcohol consumption of adolescents at the 1-year follow-up above and beyond the effects of age, sex, family history of alcoholism and aggression/delinquency of respondents. Conclusions: Alcohol expectancies and peer delinquency/substance use are both crucial to the amount and frequency of adolescent alcohol use. They should be considered in designing prevention and intervention strategies in this age group.

(Received 5 April 2003; first review notified 18 October 2003; in revised form 30 November 2003; accepted 22 January 2004)


Numerous studies have found a positive relationship between impulsivity and alcohol consumption (Sher et al., 1991; Smith and Goldman, 1994; Barnow, 2001). Moreover, there is empirical evidence that alcohol expectancies correlate with drinking behaviour (Smith and Goldman, 1994; Sher et al., 1996; Barnow et al., 2004). The Acquired Preparedness Model (APM) describes these relationships in more detail, reflecting findings regarding the impact of impulsive behaviour (disinhibition) on alcohol expectancies and alcohol use (McCarthy et al., 2001). This model integrates the biological/genetic influence on personality and aspects of social learning. Regarding genetic studies, disinhibited individuals have been described as primarily reward-seeking (Smith and Anderson, 2001), and if punished, their resulting arousal leads them to rely on their dominant response mode, which is the active pursuit of reward (Patterson and Newman, 1993). This response style makes a number of acting out behaviours (e.g. aggressive and delinquent behaviour) more likely (Smith and Anderson, 2001). Furthermore, impulsive individuals are biased to form more positive and fewer negative expectancies if exposed to alcohol consumption and its consequences. Thus impulsivity may lead to an increased likelihood of forming positive expectancies for alcohol use overall. In this study, we used self-reported aggressive behaviour problems as an indicator of impulsivity, because several studies have shown an association between impulsivity and such problems (Cloninger et al., 1993; Coccaro and Siever, 1995). These studies regard impulsivity primarily as disposition toward disinhibition of behaviour. The correlation between aggression and measures of impulsivity ranged from 0.5 to 0.6 (Plutchik and van Praag, 1994), suggesting a moderate to close relationship between these two measures.

Children who believe that alcohol use will subsequently result in positive benefits are at greater risk for developing alcohol-related problems than those who have negative expectancies (Marlatt, 1987; Goldmann et al., 1991; Smith et al., 1995; Sher et al., 1996). Positive alcohol expectancies include beliefs that drinking will be beneficial to the individual (e.g. the beliefs that alcohol enhances sexual functioning, and increases assertiveness). Taking these findings into account, the APM suggests that impulsivity influences the formation of overly positive expectancies through alcohol's effects, which, in turn, leads to an increased risk for subsequent alcohol problems and/or higher levels of alcohol consumed. In other words, it is hypothesized that alcohol expectancies mediate disinhibitions associated with drinking. This is supported by results of several studies that found that the influence of disinhibition/impulsivity on drinking is primarily mediated through positive alcohol expectancies (Baumeister and Leary, 1995; McCarthy et al., 2001).

In addition, there is evidence that disinhibited individuals have a higher probability of associating with delinquent and substance using peers (Gottfredson and Hirschi, 1990; Coie et al., 1995; Barnow et al., 2002a). For instance, in a recent study from our research group we investigated whether there is an association between aggressive behaviour and membership in a substance using peer group. We found that aggressive behaviour was linked to drinking and drug usage within the peer group, and that membership in a substance using peer group was the only factor directly associated with both the quantity of alcohol consumed and alcohol problems (Jacob and Leonard, 1986; Fergusson et al., 1995; Barnow et al., 2002a,b). The peer group simplifies the acquisition and use of alcohol and drugs. Furthermore, the need for uniformity and recognition may promote the use of substances. Another factor that influences substance use within the peer group may be the perception of what is normative and normal drinking behaviour within a particular reference group, which may, in turn, influence alcohol expectancies. Taking these findings into account, it is reasonable to suggest that adolescents' association with delinquent/substance using peers functions as a mediator between impulsivity and drinking.

Summarizing the findings mentioned above there is clear evidence for the importance of alcohol expectancies and peer group characteristics (e.g. peer delinquency/substance use) in the development of drinking problems in adolescence. Both factors have also been reported to be related to impulsivity (Cloninger et al., 1988). However, to the best knowledge of the authors, it is not clear whether alcohol expectancy is the only factor mediating the association between disinhibition and drinking as suggested by the APM, or whether peer influences may be an additional mediator in this context. The purpose of this study was to test the hypothesis that alcohol expectancies mediate the association between impulsivity, as measured by the level of aggressive behaviour problems present, and alcohol consumption, using longitudinal data from a sample of 147 adolescents. Furthermore, we hypothesize that these behavioural problems are related to both peer delinquency/substance use within the peer group and to quantity/frequency of alcohol consumed. In addition, we predicted that peer delinquency/substance use would also mediate the association between aggression and drinking. Finally, we tested a modified model, which included drinking behaviour and aggressive conduct problems, with peer delinquency/substance use and alcohol expectancies as mediators. The hypothesized model is shown in Fig. 1.

Fig. 1.

The model hypothesizes that the association between impulsivity (as indexed by aggressive/delinquent behaviour problems) and alcohol consumption is mediated by peer delinquency/substance use and positive alcohol expectancies.



Data from 147 adolescents, aged 11–18 years, and their parents were analysed in the study. One hundred and twenty-six children were selected from a population based sample, and 21 were from institutions providing outpatient services for addiction problems.

The data on the adolescents from the population based sample were taken from the Study of Health in Pomerania (SHIP; John et al., 2001). The sampling procedure is described in detail elsewhere (see Barnow et al., 2002a). Briefly, SHIP is an epidemiological study in which 3748 subjects, stratified by age and sex, were evaluated between March 1997 and May 2000. For the purposes of the present study, 527 individuals between the ages of 30 and 50 were selected from among those who lived in a household with at least one child between the ages of 11 and 18. The first contact with these subjects was a mailed survey to which 315 (59.8%) family groups responded and agreed to take part in the study. From this sample, we evaluated 234 family groups that included 301 adolescents between the ages of 11 and 18 years. One hundred and eighty-five adolescents were again examined 1 year (T1) after their initial assessment (T0). Among these 185 adolescents, 38 were excluded from analysis because of missing data at T0 or T1, leaving 147 offspring from 116 families for whom all data were available for this study. Sex distribution was balanced (49.7% male) and the mean age of the sample was 15.2 (SD = 2.29) years (see Table 1).

View this table:
Table 1.

Demographic characteristics, % of a positive FH of alcoholism and distribution of behavioural problems of the sample (n = 147)

Age (mean/SD)15.2 (2.29)
Sex (% male)49.7
Number of children in the household (mean/SD)1.4 (0.59)
Child lives with
    % both parents86.4
    % biological mother only13.6
School degree of parents
    Highschool (9 years)2.0
    Highschool (10 years)65.3
    Upper Secondary school (13 years) graduate30.6
Quantity of consumption (equals 9 g absolute alcohol)
    per typical day (mean/SD)1.7 (3.99)
    per typical week (mean/SD)3.4 (9.02)
% positive family history of alcoholism (FHalc+)27.2
Behavioural problems (aggression/delinquency) mean/SD
    Raw value10.9 (6.43)
    t-values50.6 (9.24)

Selection analyses showed that the 147 adolescents included in this study differed from the 38 excluded adolescents in terms of whether they lived with only one or both parents and with respect to family history of alcoholism (FHalc). The children excluded from the study were more likely to be living with one parent only (28 vs. 14%; d.f. = 1, χ2 = 5.31; P = 0.022) and were more often from families where at least one parent had an alcohol use disorder (AUD) (51 vs. 27%, χ2 = 6.86; d.f. = 1, P = 0.008).

Levels of drinking experience found in the present sample are consistent with data reported by other researchers. For example, 27.2% of adolescents in our sample had a positive FHalc, which is comparable to an evaluation of the prevalence of alcohol use, abuse, and dependence among 3021 offspring between 14–24 years of age living in the Munich area in Germany (Early Developmental Stages of Psychopathology study, EDSP; Holly and Wittchen, 1998), in which the authors found 22.5% with a positive FHalc. Regarding the prevalence of AUD among the adolescents, we found 6.1% of respondents had a diagnosis of alcohol misuse and 2.7% with alcohol dependence, whereas Lieb et al. (2002) reported about 5% with alcohol misuse and 2% with alcohol dependence in the 14–17 year olds in the EDSP study. Results of a survey of adolescents between 12 and 15 years of age revealed 30% who reported that they drink alcohol weekly (BZgA, 2001), while 27.3% of our sample stated that they consume alcohol at this frequency. Taking these data into account, it is reasonable to conclude that the sample studied is representative with regard to drinking behaviour, while they may be (to some degree) positively selected regarding demographic characteristics (e.g. more adolescents of the sample lived with both parents).

Instruments at T0 (initial assessment)

Behavioural problems were determined using the Youth Self Report (YSR, German version; Döpfner et al., 1995). Second-order principal factor analyses have revealed two broad groupings of the syndromes, labelled as emotional problems and behavioural problems. In this study, we used the behavioural broadband scale ‘externalizing symptoms’, which contains the subscales of aggression (α = 0.84 for boys and α = 0.82 for girls) and delinquency (α = 0.70 and α = 0.77 for boys and girls, respectively). Transforming raw values into t-values by using standardized German norms revealed a mean t-value of 50.6 and a standard deviation of 9.24, suggesting that the distribution of behavioural problems in our sample does not differ significantly from the distribution in the general population.

To obtain information on peer delinquency/substance use, we used a questionnaire that was developed by our own research group. Adolescents were asked five questions about friends who use alcohol at least once per week, peers who take drugs, and whether peers were involved in criminal or violent behaviour. All items are answered with no or yes (coded 0 or 1, respectively), the sum of which produces a score between 0 and 5. The internal consistency (coefficient alpha) of this scale in the present sample was only α = 0.57. In an effort to improve the alpha level, we calculated several reliability analyses with different numbers of items. This did not, however, lead to an improvement of internal consistency of the scale. Thus, the included five items reflect the best solution, although alpha is only moderate.

Instruments at T1 (1-year follow-up)

Alcohol consumption and presence of AUD were obtained from both the subject and parent via the Semi-Structured Interview for the Genetics of Alcoholism (SSAGA; Bucholz et al., 1994). The interview was carried out by trained graduate students in Psychology, who were supervised by the first author. Children and parents were all interviewed in their homes. The individual questions are based on well-validated items of other research instruments, such as the Diagnostic Interview Schedule (DIS; Helzer and Robins, 1988). The SSAGA records drinking amount and frequencies, tolerance and withdrawal, and alcohol problem behaviours, along with social, physical, and cognitive effects of alcohol use. Cross-centre studies of the SSAGA indicate good reliability regarding alcohol use disorders, with test-retest agreement (Kappas) for the DSM-III-R diagnoses of 0.87–0.89 for dependence and 0.57–0.74 for misuse (the range reflects the results of different studies). The alcohol frequency/quantity measure was defined as the average amount of alcohol consumed (in standard drinks) per occasion in a typical week during the prior 12 months. The following criteria were used to assign the family history of alcoholism status (FHalc) of the adolescents: FHalc+: biological mother or father met the criteria of the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, International Version (DSM-IV; APA, 1994) for alcohol abuse or alcohol dependence; FHalc–: neither father, stepfather or mother had an AUD. Using the methods described above, 27.2% of the adolescents had a positive FHalc.

Alcohol outcome expectancies were assessed using the AEQ-3 (Alcohol Expectancies Questionnaire, Third Revision; (George et al., 1995), which is a modified version of Rohsenow's 40-item Alcohol Effects Questionnaire, Second Version (AEQ-2; Rohsenow, 1983). Confirmatory factor equation models revealed six positive (global positive, social and physical pleasure, social expressiveness, sexual enhancement) and two negative expectancy subscales (power and aggression, and cognitive and physical impairment). All items are answered using a six-point Likert-type scale (agree strongly to disagree strongly). The internal consistencies of all scales were found to be good, with alphas between 0.83 and 0.93. Because of the high intercorrelation of the subscales, the authors suggest that a positive expectancy score is computed using the six positive subscales, and/or a negative score using the two negative subscales. In this study only, the sum score of the positive subscales was used because it has not been demonstrated that the two negative subscales constitute separate factors (Leigh, 1989). The internal consistency (coefficient alpha) of the global positive expectancy scale in the present sample was α = 0.95.

Statistical considerations

Baron and Kenny (1986) propose that support for a mediational effect is dependent upon three criteria: (1) significant direct relations between independent (behavioural problems) and dependent variables (quantity/frequency of alcohol consumed); (2) significant relations between independent variables and putative mediators (alcohol expectancies and peer delinquency); and (3) a reduction in the direct path from the independent to the dependent variables when analysed in conjunction with indirect (mediated) effect. Taking these suggestions into account, we explored the mediational role of alcohol expectancies and peer delinquency/substance use by examining the degree to which consideration of both factors reduced the correlation between externalizing symptoms and drinking behaviour. If the behavioural problem effect was reduced by the inclusion of alcohol expectancies or peer delinquency/substance use, this would then suggest a mediation effect.

In the second step, we used several hierarchical regression analysis (HRA) to determine the contribution of aggression/delinquency, alcohol expectancies and peer delinquency/substance use for the T1 alcohol quantity and frequency measure by controlling for age, sex and FHalc. Specifically, the following order of entry of variables was used in the multiple regression analyses: first, age, sex and FHalc were entered; second, the hypothesized mediator was included (alcohol expectancies or peer delinquency/substance use); third, adolescents' behavioural problem score was incorporated. With respect to the three criteria proposed by Baron and Kenny (1986), it is assumed that the mediators (alcohol expectancies or peer delinquency) are significantly predictive of alcohol consumption, while the behavioural problems measure no longer predicts later alcohol consumption when included after the mediator. The significance of the relative increase in R2 following variable set entry was evaluated using SPSS software version 11.5.


Table 1 shows demographic characteristics, percent of FHalc, mean number of drinks consumed in a typical day, and raw and t-values of the aggression/delinquency scale to give the reader the possibility of inspecting our results. For instance, 86.4% of adolescents lived with both parents, and 27.2% came from families where at least one parent had an alcohol use disorder.

The correlation matrix using all variables is shown in Table 2. This Table reveals that all independent variables were significantly correlated with quantity/frequency of alcohol consumed. Furthermore, alcohol expectancies and peer delinquency/substance use were significantly related to each other, as well as to the aggression/delinquency measure.

View this table:
Table 2.

Intercorrelation matrix of all variables

Behavioural problemsAlcohol expectanciesPeer delinquency/substance useAmount of alcohol consumed
Behavioural problems1.0000.262**0.321**0.194*
Alcohol expectancies1.0000.334**0.373**
Peer delinquency/substance use1.0000.447**
Amount of alcohol consumed1.000

Using partial correlation analysis, associations between behavioural problems and quantity/frequency of alcohol consumed at T1 decreased significantly from r = 0.194 (P = 0.018, see Table 1) to r = 0.107 (P = 0.144), after partialling out the effect of alcohol expectancies, and to 0.060 (P = 0.471) when the effect of peer delinquency/substance use was accounted for. This suggests that alcohol expectancies as well as peer delinquency/substance use mediate the association between aggressive and delinquent behaviour problems and quantity of drinking.

Tables 3 and 4 show the results of the HRA distinguishing between alcohol expectancies and peer delinquency/substance use. The HRA revealed that both hypothesized mediators predicted alcohol consumption of adolescents at the T1 above and beyond the effects of age, gender, and FHalc; while behavioural problems no longer predicted later quantity/frequency of alcohol consumed when entered after the mediators, suggesting a mediational effect of both factors.

View this table:
Table 3.

Test of mediation: alcohol expectancies

Order of variable entry dependent variable: quantity/frequency of alcohol consumptionBetaR2Increase in R2F for change in R2
Step 10.2190.21913.354**
Step 20.2690.0509.763**
    +Alcohol expectancies0.252**
Step 30.2660.0040.790
    ++Behavioural problems broadband scale (aggression/delinquency)0.067
View this table:
Table 4.

Test of mediation: peer substance use

Order of variable entry dependent variable: quantity/frequency of alcohol consumptionBetaR2Increase in R2F for change in R2
Step 10.2190.21913.354**
Step 20.2510.0336.189*
    +peer substance use0.200*
Step 30.2560.0040.784
    ++Behavioural problems broadband scale (aggression/delinquency)0.090

Finally, we examined whether the simultaneous inclusion of both mediators in the model accounted for incremental variance of alcohol consumed at the 1-year follow-up. Table 5 shows that the inclusion of alcohol expectancies, as well as peer delinquency/substance use, significantly predicted alcohol consumption of adolescents above and beyond the effects of age, sex, and FHalc.

View this table:
Table 5.

Test of hypothesized model (see Figure 1): Hierarchical Multiple Regression Analyses for the alcohol quantity and frequency measure at one-year-follow-up

Order of variable entry dependent variable: quantity/frequency of alcohol consumptionBetaR2Increase in R2F for change in R2
Step 10.2120.21219.42**
Step 20.2240.0112.10
    +Behavioural problems broadband scale (aggression/delinquency)0.109
Step 3†0.2660.0438.25**
    +Alcohol expectancies0.237**
Step 4†0.3150.0489.96**
    +Peer delinquency/substance use0.258**
  • * P < 0.05.

  • ** P < 0.01.

  • Interaction effect: Beta = 0.27; R2 = 0.282; Increase in R2 = 0.052; F for change in R2 = 10.12; P = 0.002.


The purpose of this study was to identify the relationship between impulsivity, as measured by the extent of aggressive and delinquent behaviour problems, and the frequency/quantity of alcohol consumption. In addition, we hypothesized that alcohol expectancies and peer delinquency/substance use mediate this relationship. This hypothesis was based upon the findings of the APM, which assumes that the association between impulsive behaviour and drinking behaviour is mediated by positive alcohol expectancies. The APM proceeds from the premise that impulsive children tend to develop positive alcohol expectancies when alcohol is consumed by a family member, due to their sensitivity to reward cues. Subsequently, these positive alcohol expectancies serve as a significant predictor of high alcohol consumption and alcohol problems in adolescence. However, the APM does not account for findings that show that impulsive children tend to be attracted more often to delinquent peer groups in which substance use is prevalent (Barnow et al., 2002a), and that membership in substance using peer groups is associated with both higher quantity of alcohol consumption and alcohol problems (Needle et al., 1986; Steinberg et al., 1994). Hence, it is reasonable to assume that membership in a delinquent substance-using peer group may mediate the relationship between impulsive behaviour and drinking problems.

One of the strengths of this study is that it considers not only multiple relevant risk factors, some of which are analysed on a longitudinal basis, but also uses a complex measure of drinking behaviour that include quantity and the frequency of alcohol consumption. This accounts for the fact that drinking patterns differ among individuals who consume the same total quantity in a typical week. For example, with regard to the hazardous consequences of alcohol consumption, it is necessary to distinguish whether an individual consumes a large amount of alcohol once or twice per week, or drinks smaller quantities five times per week. One author, for example, studied the relationship between drinking behaviour and health disorders in 18 973 individuals from the Dutch population (San Jose et al., 2000). They noted that persons who drank relatively large quantities of alcohol (six drinks or more) only a few days during the week, reported more health problems than those who drank relatively small amounts (one or two drinks) almost every day of the week. Another advantage of our study is that we controlled for parental drinking behaviour, which was obtained directly from interviews of the parents.

Taking the hypotheses stated in the introduction into account, the results of correlation analyses and hierarchical regressions show that (1) there was a significant correlation between behavioural problems and frequency/quantity of alcohol consumed, and (2) this association was mediated by alcohol expectancies, as well as peer delinquency/substance use. In light of these findings, the question arises as to what extent alcohol expectancies are associated with peer delinquency/substance use. Regarding this issue, we found a significant correlation between the two factors, although it is not possible to state with certainty whether membership in a delinquent/substance-using peer group fosters positive alcohol expectancies or whether adolescents with positive alcohol expectancies tend instead to seek out delinquent/substance-using peer groups. The result of the hierarchical regression analysis that includes both mediators show, however, that alcohol expectancies, as well as peer delinquency/substance use, exhibit a significant increase in explained variance with respect to frequency/quantity of alcohol consumption above and beyond the effects attributable to age, sex, FHalc and aggressive behaviour problems. Thus, it can be assumed that the two factors influence adolescent drinking behaviour independently of one another.

Altogether, the results of our study confirm the assumptions of the APM. In addition, our findings support the integration of peer influences into the APM. The results of the study can be interpreted as supporting our hypotheses that both cognitions and peer influences are involved in the development of drinking, and that both mediate impulsivity's relationship to drinking. Therefore, we conclude that both aspects should be taken into account in interventive and preventive measures. In this context, concepts applied in psychological education, such as the modification of positive alcohol expectancies in favour of more realistic and possibly negative long-term consequences of alcohol consumption, as well as measures designed to help integrate problem youths into a ‘positive peer group’ may be of significance. For example, the results of an international survey indicate that membership in delinquent/substance-using peer groups is an important predictor of quantity/frequency of substance consumption (perhaps because this kind of peer group supports positive alcohol expectancies and facilitates access to alcohol and drugs), while on the other hand, adolescents who were members of peer groups characterized by negative attitudes toward drinking reported lower alcohol use and fewer alcohol problems (McArdle et al., 2000). Furthermore, it has been shown that the increased availability of activities involving sports in an urban district resulted in a significant decline in problem behaviour in general (Jones and Offord, 1989).

As with any study, the findings of this study need to be considered in light of its limitations. It was not possible to determine whether positive alcohol expectancies prompt adolescents to join delinquent/substance-using peer groups, or whether, conversely, membership in such delinquent/substance-using peer groups promotes such expectancies. In addition, alcohol expectancies were identified only for T1, thus no longitudinal data were available for this measure. Further studies with a longitudinal design and based on larger samples will be necessary in order to achieve clarity with respect to the direction and chronological progression of the individual relationships. Other critical aspects include the small sample size and the fact that separate analyses were not performed for boys and girls, although the effect of sex was taken into account in the hierarchical regression analyses. Finally, impulsive behaviour and disinhibition were not assessed directly, but were instead operationalized on the basis of the extent of behavioural problems such as aggression and delinquency.


The work is part of the Community Medicine Research net (CMR) of the University of Greifswald, Germany, which is funded by the Federal Ministry of Education and Research (grant no. ZZ9603), the Ministry of Cultural Affairs as well as the Social Ministry of the Federal State of Mecklenburg-West Pomerania. The CMR gathers several research projects that are sharing data of the population-based Study of Health in Pomerania (SHIP; http://www.medizin.uni-greifswald.de/cm).


View Abstract