Family Communication and Religiosity Related to Substance Use and Sexual Behavior in Early Adolescence: A Test for Pathways Through Self-Control and Prototype Perceptions

Acknowledgement: A version of this article was presented at the Fetzer Institute Conference on Religiosity and Substance Use, Kalamazoo, Michigan, May 2002. This research was supported by Grants P50 MH48165 and R01 MH62668 (Drs. Gibbons, Gerrard, Murry, and Brody) and Grant R21 AA13079 from the National Institute on Alcohol Abuse and Alcoholism and the Fetzer Institute (Dr. Wills). Additional funding was provided through the National Institute on Drug Abuse, Research Scientist Development Award K02-DA00252 (Dr. Wills) and the Iowa Agriculture and Home Economics Experimental Station, Project 3320 (Dr. Gibbons et al.). We thank the parents and participating adolescents for their cooperation and the staff of the Family and Community Health Study for their able assistance with the research.

This research tested predictions about variables relevant for early substance use and sexual behavior. These behaviors are significant from a health standpoint because in adolescence they may contribute to risk for accidents or sexually transmitted diseases (DiClemente, Hansen, & Ponton, 1996; Institute of Medicine, 1997). Early onset is of particular significance, because longer periods of exposure accrue if behaviors continue through adolescence (e.g., Hawkins et al., 1997; Wills et al., 2001). Hence, research is needed to test predictions about early risk and protective factors and to help professionals understand in more detail the observed correlation of substance use and sexual behavior (Leigh & Stall, 1993).

The study design was based on prior theoretical work about the origins of health-related behaviors (Gibbons & Gerrard, 1995; Wills, DuHamel, & Vaccaro, 1995) and on qualitative research with African American families (Brody, Stoneman, & Flor, 1996; Murry, 1995). The goal was to test the predictive status of constructs suggested as relevant for African American adolescents, with multivariate data from a representative community sample.

We tested three theoretical perspectives about risk and protective processes. First, parent-child communication about substance use and sex has been suggested as a protective factor, but little direct evidence has been available on this aspect of family relationships. Second, evidence has indicated religiosity as a protective factor for adolescent substance use and sexual behavior, but how these effects occur is not well understood (McCullough, Hoyt, Larson, Koenig, & Thoreson, 2000; Wallace & Williams, 1997). Third, self-control constructs and social perception constructs each have been shown to predict health-relevant behaviors (Gibbons & Gerrard, 1995; Wills, DuHamel, & Vaccaro, 1995), but there is little knowledge about the relation of self-control characteristics to social perceptions. In the following sections we discuss the constructs studied and the theoretical predictions.

It is believed that parental communication with children about drugs and sex will deter involvement in risk behavior. Family process theory suggests that open communication patterns encourage adolescents to internalize the values and norms embedded in the parents' messages (Baldwin & Baranowski, 1990; Whitaker & Miller, 2000). This is proposed to enhance psychosocial competence through promoting more direct discussions about important issues in adolescents' lives and encouraging youth to seek information from parents about health-related questions (Ary, James, & Biglan, 1999; Kafka & London, 1991).

Although parent-child communication has been widely discussed, there is less direct evidence on the role of communication in protective processes. Researchers have noted links among parent-child relationship quality, family communication about sexuality, and early sexual activity (Howard & McCabe, 1992; Murry, 1996), so it is unclear whether effects are specifically attributable to communication factors or to more general warmth and supportiveness in the parent-adolescent relationship (Kotchick, Dorsey, Miller, & Forehand, 1999; Wills & Cleary, 1996). The present research included a measure that directly indexed the level of communication between parents and adolescents about drugs and sex, and we tested its contribution to mediators and outcomes, controlling for other family characteristics.

Adolescents who attend a religious institution show lower prevalence rates for both substance use and early sexual behavior (Donahue & Benson, 1995; Wallace & Williams, 1997). This effect is most consistent for measures reflecting religious commitment and the importance placed on religion in one's life (Brody et al., 1996; Brody, Stoneman, Flor, & McCrary, 1994; Foshee & Hollinger, 1996). Religiosity has been suggested as particularly important for African Americans (Barnes, Farrell, & Banerjee, 1994; Brody et al., 1994) although significant effects have been noted in several ethnic groups (e.g., Resnick et al., 1997).

Although evidence is available on protective effects of religiosity, there is less understanding of how these effects occur (Levin, 1996; McCullough et al., 2000). Wallace and Williams (1997) suggested possible mechanisms for adolescents, including direct effects and indirect effects mediated through peer affiliations. In a large sample of Utah adolescents, Bahr, Maughan, Marcos, and Li (1998) found evidence for both a direct effect to substance use and an indirect effect mediated through peer affiliations; Brody et al. (1996) found evidence for a direct effect to externalizing problems and an indirect effect through a composite index of self-regulation. Thus, there is some evidence for both direct and indirect effects, but the available evidence concerning mechanisms for religiosity is minimal. Moreover, statistical controls for characteristics that may be correlated with religious involvement, such as family socioeconomic status or warmth of the parent-child relationship, have not always been included in analyses, and personality characteristics, such as temperament attributes, have not generally been included. In the present research we tested the effect of religiosity with control for parental education, family relationship indexes, and several indexes of adolescents' personality characteristics (Wills & Cleary, 1996; Wills, DuHamel, & Vaccaro, 1995; Wills, McNamara, & Vaccaro, 1995). We predicted that religiosity would have indirect effects, mediated through self-control and peer affiliations, and direct effects on substance use and sexual behavior.

This research considered the theoretical convergence of self-control and social perception. Self-control theory predicts that risk behavior is related to variables reflecting ability for self-regulation of cognition, emotion, and behavior (Miller & Brown, 1991; Wills & Stoolmiller, 2002). Self-control is posited to be grounded in early temperament characteristics and to develop through childhood and adolescence (Rothbart & Ahadi, 1994; Wills, Sandy, & Yaeger, 2000). Good self-control and poor self-control are conceptualized as distinct constructs, with different antecedents and consequences (J. H. Block & Block, 1980; Rothbart & Bates, 1998). Poor self-control has been suggested as a risk factor for substance use, whereas good self-control (also termed planfulness) has been suggested as a protective factor (J. Block, Block, & Keyes, 1988; Rutter et al., 1997). This perspective has been supported by data showing that self-control constructs are related to outcomes, including adolescent substance use and sexual behavior (Brody et al., 1996; Mezzich et al., 1997; Murry & Brody, 1999; Wills, Windle, & Cleary, 1998).

Social perception theory focuses on the mental images that people hold of typical individuals who engage in risk behavior, termed prototype perceptions. It is posited that prototype perceptions develop at an early age and have implications for subsequent risk behavior (Chassin, Tetzloff, & Hershey, 1985; Gibbons & Gerrard, 1997). Having a relatively favorable prototype of engagers (i.e., people who use substances or have sex) is posited to be a vulnerability factor, whereas having less favorable perceptions of engagers (or positive perceptions of abstainers) is posited to make an individual less predisposed toward problem behavior. The status of a person's prototype perceptions may shape risk status through attitudes about the behavior and affiliation with people who are disposed to engage in the behavior, because individuals will tend to affiliate with others whom they perceive in more positive terms (Gibbons & Gerrard, 1995). The family environment is predicted to shape the development of drug and sex prototypes as parents communicate their own norms about use and their own perceptions of users (Blanton, Gibbons, Gerrard, Conger, & Smith, 1997). The social perception perspective has received support in studies of drug use and sexual behavior (Blanton et al., 2001; Gibbons & Gerrard, 1995; Gibbons, Gerrard, & Boney-McCoy, 1995).

The convergence hypothesis derives from the argument that self-control characteristics may themselves have an effect on social perceptions. Individuals who are undercontrolled may tend to perceive prototypical teen substance users (also undercontrolled) in more favorable terms, whereas well-controlled teens could have less favorable perceptions of users or more favorable perceptions of abstainers (cf. Gibbons, Gerrard, Ouellette, & Burzette, 1998; Rothbart & Ahadi, 1994; Scarr & McCartney, 1983). Thus, in addition to predicting that self-control characteristics and prototype perceptions will be related to proximal factors for substance use and sexual behavior, we also hypothesized that there will be paths from self-control constructs to social perception constructs. This perspective was tested with social perception measures involving prototypes of both abstainers and engagers (Gerrard et al., 2002).

The predictions regarding substance use and sexual behavior were tested with a community sample of African American adolescents assessed at approximately 13 years of age. For family communication about drugs and sex, we anticipated that communication would have a significant effect net of parental supportiveness and that communication would have indirect effects on outcomes through paths to perceptions of engagers and abstainers. We predicted that religiosity would have direct protective effects together with indirect effects through more conventional peer affiliations. Regarding prototype perceptions, we predicted inverse paths from parental support to prototypes of users, and we anticipated paths from self-control constructs to prototype constructs. The study addressed several methodological issues raised by Leigh and Stall (1993); we examined predictors for correlated outcomes of substance use and sexual behavior and used latent constructs so as to provide more reliable measures of constructs and allow specification of correlated errors. We evaluated the study predictions in a single analysis using structural equation modeling, testing for both direct effects and indirect effects to outcomes, with control for family demographic characteristics and adolescent temperament characteristics.

The sample consisted of African American families who had a child between the age of 10 and 11 years. Parents and target children were interviewed separately in the household by trained interviewers. If an older sibling of the target child within the age range of 12–14 years was in the household, he or she was assessed through a written questionnaire. These siblings were the participants for the present research.

The participants were 297 adolescents, with a mean age of 12.96 years (SD = 0.81). All participants were of African American ethnicity. The sample was balanced on gender, with 53% of participants being female and 47% being male. The study procedure identified a primary caregiver, the person primarily responsible for care of the children, and determined whether there was a secondary caregiver in the household. The primary caregiver was female in 93% of families and male in 7% of families. In the sample, 56% of the families had both a primary caregiver and a secondary caregiver; single-parent families were 44% of the sample. The mean age of the primary caregiver was 37.17 (SD = 8.18). The educational level of the primary caregiver, indexed on a 1–6 scale, had a mean of 3.27 (SD = 1.07), just above high school graduate.

Families were recruited from areas around Athens, Georgia, and Des Moines and Waterloo, Iowa, using sampling from census tracts. Families were enumerated through lists compiled by community coordinators (in Georgia) or through school lists (in Iowa). A letter was sent to eligible households drawn randomly from the enumeration lists, informing them about the study. The letters were followed by a visit to the household by a research staff member. The staff member explained the study procedures; invited the participation of family members, including the caregiver(s), target child, and older sibling (if appropriate); and obtained consent from the caregiver(s) and assent from children and adolescents. Interviews were completed with 72% of the eligible families; 33% of the families had a sibling in the appropriate age range.

The interviewers received initial training in a 3-day workshop that included didactic presentations about the goals and methods of the study followed by supervised practice interviews. After study initiation, quality control was maintained through regular interviewer meetings and random family probes. The privacy of research data was protected by a Certificate of Confidentiality from the U.S. Department of Health and Human Services, and both interviewers and participants were instructed about the confidentiality protections.

The participating sibling was assessed in an individual session through a written questionnaire. A trained interviewer introduced the questionnaire items and responses to the participant, emphasized the confidentiality of the data, and was available to answer any questions the participant might have about particular items. Completed questionnaires were obtained from 297 participants.

Some measures were based on previous research with adolescents and adults (Conger, Conger, Elder, & Lorenz, 1992; Gibbons & Gerrard, 1995; Wills, DuHamel, & Vaccaro, 1995), and some measures were developed for this study. Unless otherwise noted, responses to items were made on 5-point Likert scales with the anchor points not at all true for me and very true for me. Measurement structure was verified with internal consistency (Cronbach's alpha) analysis, and lower loading items were dropped from some scales. All scales were constructed such that a higher score indicates more of the named attribute.

Demographics

The participant reported on his or her gender and age in years. The primary caregiver reported on her or his gender, ethnicity, and age in years and indicated whether there was a secondary caregiver. The educational level of the caregiver(s) was reported on a continuous scale in years; this was recoded for analysis to a 6-point scale with the scale points grade school, some high school, high school graduate, some college, college graduate, and postcollege (master's/doctoral degree or other professional education).

Parent-child relationship quality

The quality of the relationship between parent and adolescent was indexed with items that asked “How satisfied are you with your relationship with [your caregiver]?” and “How happy are you with the way things are between you and [your caregiver]?” Responses were made on 5-point scales with the anchor points very unsatisfied/unhappy to very satisfied/happy. The 2-item scale had α = .80.

Parent-child communication

Items on communication between the primary caregiver and the adolescent were introduced with the lead-in statement: “During the past year, how often has [your caregiver] talked to you about …?” Items for drug use were “Drinking alcohol, “Using drugs,” and “Smoking cigarettes.” Items for sexual behavior were “Sexual intercourse (sex),” “Preventing sexually transmitted diseases (like ‘crabs,’ gonorrhea, chlamydia, herpes),” and “Birth control.” Responses were made on 4-point scales with the anchor points never and many times. Alpha was .80 for a 3-item drug communication scale and .88 for a 3-item sex communication scale.

Religiosity

Religiosity was indexed with the item “How important is your religion to you?” Responses were made on a 4-point scale with the anchor points not at all important and very important.” This item tapped a dimension identified in previous studies of religion and problem behavior (Brody et al., 1996; Wallace & Williams, 1997).

Temperament

Temperament characteristics of the participants were assessed through the Dimensions of Temperament Survey (Windle & Lerner, 1986) and the Emotionality, Activity, and Sociability Inventory (Buss & Plomin, 1984). Protective temperament dimensions were indexed with a 6-item scale for task orientation (e.g., “I keep working at a task until it's finished,” α = .85) and a 5-item scale for positive emotionality (e.g., “My mood is generally cheerful,” α = .86). Difficult temperament dimensions were indexed with a 6-item scale for activity level (e.g., “I can't stay still for long,” α = .85) and a 5-item scale for negative emotionality (e.g., “I get upset easily,” α = .79).

Self-control

Self-control constructs were indexed with items from the Kendall-Wilcox inventory as adapted for research with adolescents (Kendall & Williams, 1982; Wills, Vaccaro, & McNamara, 1994). A 7-item scale on good self-control (α = .79) had items such as “When I promise to do something, you can count on me to do it”; “When I have to wait in line, I do it patiently”; and “I prefer to concentrate on one thing at a time.” A 10-item scale on poor self-control (α = .79) had items such as “I have to have everything right away,” “I have to be reminded several times to do things,” and “I like to switch from one thing to another.”

Risk-taking tendency

The tendency to enjoy taking risks was indexed with an adaptation from Eysenck and Eysenck's (1977) inventory. The 6-item scale (α = .83) had items such as “I enjoy taking risks,” “I could do something most people consider dangerous (like driving a car fast),” and “I would do almost anything for a dare.”

Prototypes of substance users

Measures for prototypes of users (Blanton et al., 1997; Gibbons & Gerrard, 1995) were introduced with the lead-in statement “Take a moment to think about the type of kid your age who [smokes]. We are not thinking about anyone in particular, just your image of kids who [smoke].” Following were seven items with the adjective descriptor stem “How [xxx] are they?” Each item had a 4-point response scale keyed to the descriptor not at all [xxx] to very [xxx]. The descriptors were “popular,” “careless,” “smart,” “cool,” “attractive (good-looking),” “immature (childish),” and “dull (boring).” Prototype measures were obtained for smokers (as described, α = .73), drinkers (The type of kid your age who uses alcohol regularly, α = .78), and drug users (The type of person your age who uses drugs, α = .83). Prototype measures were scored as 6-item scales, dropping the recoded item “immature” (cf. Gibbons & Gerrard, 1995). These were scored such that a higher score meant a more favorable perception of substance users.

Prototypes of sex engagers

Perceptions of prototypical sex engagers were indexed with procedures similar to those described above. A measure for the sexual activity prototype had the stem “Please tell us what you think boys/girls your age who have sex regularly are like…” A scale for the pregnancy prototype had the stem “Please tell us what you think [boys your age who get a girl pregnant]/[girls your age who get pregnant] are like …” A 6-item scale for sexually active prototype had α = .76, and a 6-item scale for pregnancy prototype had α = .78. These were scored so that a higher score meant a more favorable perception of engagers.

Prototypes of abstainers

Perceptions of prototypical abstainers were indexed with procedures similar to those described above but with nonengagers as the target. The stem for the drug abstainer prototype measure was “Please tell us what you think the type of kid your age who decides they are not going to drink alcohol, smoke cigarettes, or use drugs at all is like.” The stem for the sexual abstainer prototype measure was “Please tell us what you think [boys/girls] your age who choose not to have sex at all are like…” The adjective descriptors were the same as described above. A 6-item scale for the drug abstainer prototype had α = .86, and a 6-item scale for the sex abstainer prototype had α = .80. These were scored such that a higher score meant a more favorable perception of abstainers.

Resistance efficacy

A scale for substance resistance efficacy (Hays & Ellickson, 1990) had the lead-in statement “Suppose you were with a group of friends and some of them were [smoking]. There are some extra [cigarettes] there that you could have if you wanted. How willing would you be to say no and not [smoke any cigarettes]?” Responses were made on 1–5 scales with the anchor points not at all willing and very willing. Items were administered for smoking (as described), alcohol (You were with a group of friends and there was some alcohol there that you could have if you wanted) and drugs (You were with a group of friends and there were some drugs you could have if you wanted), with similar responses. A 3-item scale had α = .72, with a higher score indicating more resistance efficacy.

Friends' substance use

Items on peer substance use were introduced with the stem “During the past 12 months, how many of your friends have …” Responses were made on 5-point scales with the anchor points none of them and all of them. The items asked how many friends had “Smoked cigarettes,” “Drank alcohol,” “Drank a lot of alcohol (3 drinks or more),” and “Used illegal drugs.” A 4-item scale for friends' substance use had α = .90.

Friends' sexual behavior

Sexual behavior among the participant's peer group was indexed with a procedure similar to that described above. The item asked “During the past 12 months, how many of your friends have had sex?” A 5-point response scale was used with the anchor points none of them and all of them.

Criterion constructs: Adolescent's substance use

Items were introduced with the stem “We'd like to know whether you have done any of these things during the past 12 months. This is personal and confidential, no one will know how you answered these questions. In the past 12 months, how often have you …” Responses were made on 5-point scales with the anchor points never to 6 or more times. The items were “Smoked cigarettes,” “Drunk alcohol,” “Drunk a lot of alcohol (3 or more drinks at one time),” and “Used illegal drugs.” A 4-item scale for adolescent substance use had α = .85.

Adolescent's sexual behavior

The participant's sexual behavior was indexed with a procedure similar to that described above. The items were “Had sex,” “Had sex without birth control,” and “Had sex without a condom.” A 3-item scale for adolescent sexual behavior had α = .83. Methodological data showing good validity for self-reports of substance use and sexual behavior obtained under confidential conditions have been discussed in several places (Gerrard, Gibbons, & Bushman, 1996; Murray & Perry, 1987; Patrick et al., 1994; Wills & Cleary, 1997).

Prevalence rates for adolescents' substance use and sexual behavior are presented in Table 1. A large majority of the sample had not engaged in either type of behavior: 72%–90% for substance use indices and 75%–89% for sexual behavior indices. There were varying degrees of substance use above this, with 3%–7% of the participants indicating some experimentation with substances and 2%–6% indicating use four times or more often. These rates are approximately half the prevalences found for 13-year-olds in general-population studies (Johnston, O'Malley, & Bachman, 1995; Wills, McNamara, Vaccaro, & Hirky, 1996), reflecting the lower rates of substance use typically observed among African American adolescents (Bachman et al., 1991; Vaccaro & Wills, 1998). Overall, 25% of the participants reported having had sexual intercourse during the previous year. These encounters were mostly protected ones, but 11% of the sample reported having had unprotected sex once or more.

For peer behavior measures, a moderate proportion of participants (20%–41%) reported having a few friends who used substances or had sex (31%), and a smaller proportion reported friendship networks with more extensive involvement. For example, 7%–12% of the participants reported that most or all of their friends used tobacco, alcohol, or other drugs, and 20% reported a comparable level for sex. These friendship networks can include older peers, which may account for the higher exposure levels reported for friends' use.

Absolute levels for resistance efficacy were generally high, with the majority of the sample reporting that they would resist opportunities for substance use. However, approximately 20% of the sample reported limited resistance to opportunities for use. This would include persons who had already used and those who were susceptible for onset.

Descriptive statistics (see Table 2) indicated that the measures were normally distributed for the most part. Scores for importance of religion were shifted toward the higher end of the scale, with the skewness value negative by convention, and scores for risk taking tendency were shifted toward lower values. The prototype measures were not strongly skewed, and means indicated that a proportion of the sample had somewhat positive perceptions of prototype substance users and sex engagers. Means for the abstainer prototypes indicated that a moderate proportion of the sample held reasonably favorable perceptions of abstainers, a finding that is possible because favorability of perceptions for users and abstainers were not strongly correlated. Descriptives for the composite substance use score and sexual behavior score were more skewed, because the majority of the sample reported relatively low levels, and skewness issues were addressed in the data analyses.

A confirmatory analysis tested the measurement model and indicated the correlations among the constructs. Substance user prototype was specified as a latent construct measured by indicators of smoker prototype, drinker prototype, and drug user prototype. Sex engager prototype was specified as a latent construct with indicators of sexual activity prototype and pregnancy prototype. Adolescent substance use was specified as a latent construct measured by indicators of cigarette smoking, alcohol use, heavy drinking, and illicit drug use. Adolescent sexual behavior was specified as a latent construct measured by indicators of sex during past year, sex without birth control, and sex without a condom. The other model constructs were analyzed as manifest variables, each measured by a single indicator.

We performed the confirmatory analysis in MPlus (version 2.02; Muthén & Muthén, 1998) using the maximum likelihood method with the expectation maximization algorithm for missing data. The confirmatory model had reasonable fit to the data, χ2(165, N = 295) = 259.70, root-mean-square error of approximation < .05 (.044), confidence interval = .034–.054. The measurement model was satisfactory, as indicated by high loadings of indicators on constructs. Correlations among the variables in the confirmatory model are presented in Table 3. In most cases, the predictors had significant correlations with the criterion variables of adolescent substance use, sexual behavior, or both.

The patterning of correlations was consistent with theoretical predictions. Self-control and risk-taking tendency had moderate correlations with the prototype scales, and prototypes (except for the drug abstainer prototype measure) had moderate to strong correlations with both resistance efficacy and peer behavior. The abstainer prototype measures had relatively low correlations with the user/engager prototype measures, which supported including them as separate constructs. There were substantial correlations between indexes for substance use and sexual behavior; this was true within the prototype measures, the peer measures, and the criterion measures of adolescent substance use and sexual behavior (cf. Leigh & Stall, 1993).

To test the hypothesized relationships of the constructs, we performed structural modeling analysis with exogenous variables, those not predicted by any prior construct in the model, and endogenous variables, those that could be predicted by a prior construct in the model. Parent-adolescent relationship, parent-adolescent communication, and religiosity were specified as exogenous together with temperament characteristics, gender, and parental education, to control for any correlations with these variables.

The endogenous variables were specified according to the hypotheses outlined previously. Self-control constructs and risk-taking tendency were specified as the first endogenous variables, with covariances of their residual terms, and the prototype constructs were specified subsequent to these, also with covariances of their residual terms. The proximal factors (resistance efficacy, peer substance use, and peer sexual behavior) were specified subsequent to these, with residual covariances. The two criterion constructs—adolescent substance use and sex behavior—were specified with a covariance of their residual terms.

We analyzed the structural model in MPlus using the EM algorithm for missing data (Little & Rubin, 1987). An initial model was estimated with a minimal set of paths based on theory and prior research; two measures were tested in initial models but were dropped from subsequent analysis. Additional structural coefficients were added to the model if they had p < .05, and correlated errors were added if they had p < .001. Although skewness does not affect estimates of structural coefficients, it may affect standard errors (cf. Chou & Bentler, 1995; West, Finch, & Curran, 1995), so the initial model was replicated with robust estimates of standard errors, and one nonsignificant path was dropped. The other paths in the model were all significant (p < .05) with robust estimates. The final model, including three correlated error terms, had reasonable fit to the data, χ2(248, N = 295) = 392.73, root-mean-square error of approximation = .044 (confidence interval: .036 –.052). The model is presented in Figure 1. Included in the model but omitted from the figure for graphical simplicity are covariances of exogenous variables (included in Table 3), measurement model parameters (presented in Table 4), and residual covariances (presented in Table 5). Prototype perceptions of sexual abstainers were statistically distinct from perceptions of substance users or sex engagers. Prototype perceptions for substance users and sex engagers had a substantial correlation but were retained as separate constructs to test for differential effects, because the residual correlation for adolescent substance use and sexual behavior was moderate (r = .33). Prior variables in the model accounted for 6%–16% of the variance in prototype perceptions, and prior variables in the model accounted for 11%–22% of the variance in proximal factors. Together, the variables in the model accounted for 40% of the variance in adolescent substance use and 40% of the variance in adolescent sexual behavior.

The results indicated several significant relationships of exogenous variables to mediators and criterion constructs; paths have positive coefficients unless otherwise noted. (Note that all the results reported here are independent effects.) Parent-adolescent relationship had a path to good self-control and an inverse path to perceptions of sex engagers as well as a direct effect to higher resistance efficacy. Parent-adolescent communication had an inverse path to prototype of substance users. Religiosity had inverse direct effects on adolescent substance use and sexual behavior. Temperament dimensions had paths to good self-control and poor self-control as well as a path from difficult temperament to risk taking tendency. Demographic characteristics also had significant effects: Male gender was positively related to poor self-control, positively related to risk taking, and inversely related to perceptions of abstainers. Parental education had an inverse direct effect on adolescent sexual behavior.

Results for self-control were partly consistent with the prediction. Self-control constructs were related to prototypes of abstainers, although they were not related to prototypes of users. Individuals with higher good self-control had more positive perceptions of sexual abstainers, and individuals with higher poor self-control had less positive perceptions of abstainers. Good self-control also had a direct effect on resistance efficacy. Risk taking had paths to prototypes of substance users and sex engagers and had direct effects on both criterion constructs; thus, risk taking was indicated as an independent factor for problem behavior.

Prototypes of substance users had paths to peer substance use and (less) resistance efficacy, and prototypes of sexually active teens had a path to peer sexual behavior. In contrast, sexual abstainer prototype had a path to resistance efficacy and an inverse direct effect on adolescents' sexual behavior. Thus, pathways for prototypes to criterion constructs indicated both direct and indirect effects.

Effects for the proximal factors were consistent with prior research (e.g., Brody et al., 1996; Gibbons & Gerrard, 1995; Wills et al., 1998). Resistance efficacy had a moderate inverse path to adolescent substance use. Peer substance use had a strong path to adolescent substance use (but not to sexual behavior), whereas peer sexual behavior had a strong path to adolescent sexual behavior (but not to substance use). Thus, although peers' substance use and sexual behavior were correlated, there was evidence for differential effects.

This research was designed to test hypotheses about protective and risk processes among a representative sample of African American adolescents. Predictions derived from three theoretical perspectives were generally supported, and the results showed several different types of pathways to outcomes. Religiosity had protective direct effects, whereas parental variables primarily had protective indirect effects. Self-control constructs were related to perceptions of abstainers, whereas perceptions of engagers were related to a different construct: risk-taking tendency. Resistance efficacy was an important proximal pathway for the operation of protective processes, whereas peer affiliations were important proximal pathways for risk processes. These findings were independent of family structure and parental education. Although temperament attributes of the adolescent were correlated with religiosity and with parent-child relationship variables (Rothbart & Ahadi, 1994; Wallace & Williams, 1997), the findings also are independent of these correlations.

The results support the utility of self-control theory and social perception theory for understanding early risk behavior (Gibbons & Gerrard, 1997; Wills et al., 2000). The effects of self-control constructs were entirely indirect ones, mediated through perceptions of abstainers. Effects of prototype measures were more complex. Some involved pathways through resistance efficacy, with favorable prototypes of abstainers contributing to higher efficacy and favorable prototypes of users contributing to lower efficacy; some effects involved pathways through social affiliations, with individuals having favorable prototypes of engagers being more likely to affiliate with peers who were sexually active and using substances. Thus, self-control characteristics are linked to perceptions of other persons in a manner that has implications for risk behavior, and social perception processes involve both social mechanisms and efficacy mechanisms.

The results bear on theoretical issues raised by Leigh and Stall (1993) about the correlation of substance use and sexual behavior. There was a significant correlation of these behaviors at a young age, when levels of substance use and sexual behavior are relatively low, and the findings show that such behavior occurs within a matrix of individual attributes and social network characteristics. There was a substantial correlation between peers' substance use and sexual behavior, and the findings suggest that affiliation with particular peers was shaped through a series of pathways involving self-control characteristics and social perceptions. Thus, a number of psychosocial factors contribute to the correlation of sexual behavior and substance use in early adolescence.

Regarding Leigh and Stall's (1993) suggestion that personality characteristics, such as risk taking, may be a confounding factor, the present results show that a confounding explanation is in some sense correct but also is not a complete explanation. Risk-taking tendency is clearly rooted in temperament characteristics but also operates through effects on social perceptions that lead to affiliation with particular types of peers. So, risk taking is a confounder in the sense that it has direct effects on substance use and sexual behavior, but focusing on this interpretation would miss the role of social variables in setting the stage for risk behavior through a particular network of peer affiliations. Also, the results indicated that the risk status of male gender was accounted for by paths from gender to self-control and social perceptions, so the present approach indeed showed several pathways through which effects of gender occur, a possibility Leigh and Stall suggested.

Some limitations of the present study could be noted. The measures were relatively simple ones, and further research could use more extensive assessments of religiosity, self-control, and resistance efficacy for various type of problem behaviors (Hill & Wood, 1999; Wills et al., 2001; Wills et al., 2000). The present cross-sectional data do not provide definitive demonstrations of the directionality of relations between variables. Reciprocal effects are possible and can be investigated in further research with multiwave designs and latent growth approaches (e.g., Wills & Cleary, 1999). Finally, research with Asian, Hispanic, and Caucasian adolescents would be desirable to help extend the generality of the findings.

The results gave some support for parent-child communication as a unique protective construct, showing that communication about drugs and sex is related to more unfavorable protototypes of substance users. This result was obtained independent of the correlation of communication with relationship quality and parental education. This suggests the significance of this conceptual domain for prevention research.

The measure of parent-child relationship quality had three independent effects. Parental supportiveness contributed to protective processes because of a pathway through good self-control, a pathway through unfavorable prototypes of sex engagers, and a direct effect on higher resistance efficacy. Thus, parent-child relationship appears to have generalized effects on self-regulation and social perceptions. The effects on self-control and resistance efficacy may reflect modeling of well-controlled behavior by parents (cf. Rothbart & Ahadi, 1994; Wills, Blechman, & McNamara, 1996). The effect on prototypes may reflect a process in which parents clearly communicate their norms about substance use (cf. Brody, Flor, Hollett-Wright, & McCoy, 1998). Thus, parental supportiveness may have protective effects through several processes (Wills & Cleary, 1996; Wills & Filer, 2000).

The inverse effects of religiosity on substance use and sexual behavior indicate one mechanism through which religiosity may contribute to better health (cf. McCullough et al., 2000). Religiosity did not have direct effects on peer affiliations, as found in a study of Utah adolescents (Bahr et al., 1998). The present model included several constructs not included in the model tested by Bahr et al. (1998), so model specification may be a factor in the differing results. It is also possible that peer variables are less relevant among African American adolescents, as suggested in some studies (e.g., Landrine, Richardson, Klonoff, & Flay, 1994).

The direct effects of religiosity we observed are notable because they occurred in a model that included self-control constructs and social factors, hence they must reflect another type of protective process. It has been proposed that religious involvement may deter risk behavior in part because some religious institutions hold norms contrary to use (Hadaway, Elifson, & Petersen, 1984; Lorch & Hughes, 1985). Other possible pathways—for example, through attitudes that are less tolerant of deviant behavior—were not tested in the present research but have been suggested in some articles (e.g., Wallace & Williams, 1997). Research to directly investigate such processes would be informative.

Theoretical discussions on the protective status of planfulness suggest this occurs in part through linkages to the social world that individuals inhabit (Rutter et al., 1997; Wills et al., 2000). Individuals who carefully think about future consequences of behavior may come to regard favorably others who do the same, and perceptions of other persons may affect decisions about one's own behavior. The ability to anticipate situations could also lead to avoidance of situations (or peers) that are expected to be problematic. This is consistent with the present finding that good self-control was related to more positive perceptions of abstainers. Perceptions of abstainers also had other effects, reflected in the path from prototype of sexual abstainers to higher resistance efficacy. Thus, the linkage of self-control and social perception theory has implications for prevention research.

Our prediction that self-control would be related to perceptions of substance users was not supported. Instead, prototypes of substance users and sexually active teens were positively related to risk taking, a dispositional construct linked to novelty seeking and sensation seeking (Wills et al., 2001; Wills et al., 1994Zuckerman, 1994). Individuals high in risk-taking tendency seem to regard users as more attractive and popular, suggesting that individuals with similar temperamental profiles may come to select similar “social niches” (Scarr & McCartney, 1983). Risk taking also had direct effects on drug use and sexual behavior, so the effect of risk taking is not entirely social in nature and may involve a direct preference for risky and dangerous behavior (Maggs, Almeida, & Galambos, 1995).

Implications for prevention are indicated, because correlated substance use and sexual behavior have contributions from several domains of variables, including dispositional variables, such as temperament and risk taking, and family variables, such as supportiveness and communication. These factors operate through shaping self-control characteristics and social images of abstainers or users. The effects of these risk and protective processes are transmitted in part through efficacies and peer affiliations and in part through direct effects (for religiosity and parental education). Hence, the results indicate a range of distal and proximal processes that may be targeted in translational prevention research. For prevention programs, attention to increasing communication between parents and children about risk behavior is one implication of the findings (cf. Brody et al., 1998; Spoth, Redmond, & Shin, 1999). Methods to improve self-control skills and influence perceptions of substance users are also indicated as potential goals for school- and family-based programs. Finally, we note that boys in this sample were at greater risk because they had less self-control, more risk-taking tendency, and more unfavorable perceptions of abstainers. Attention to these gender differences may be a salient issue for prevention programs.

1  A prior analysis tested whether indicators of substance use and sexual behavior had appropriately high loadings on underlying latent constructs. This was established, and the confirmatory model for the criterion constructs had a comparative fit index of .98. This is consistent with previous research on measurement models for adolescent problem behavior (e.g., Hays, Widaman, DiMatteo, & Stacy, 1987; Needle, Su, & Lavee, 1989; Newcomb & Bentler, 1988).

2  Family structure was excluded from the confirmatory model because it was not correlated with the criterion variables. Two cases were excluded for extensive missing data, so the analytic sample had 295 cases. An analysis based on listwise deletion had 230 cases; results were quite similar for analyses with and without missing-data imputation.

3  We performed a nested analysis to test dimensionality of the self-control constructs, using three random parcels constructed from the scale items for each of the constructs. A nested analysis compared a model that specified these as indicators of two constructs (good self-control and poor self-control) with a model that specified the six indicators as measures of a single construct, with positive loadings for one set of indicators and negative loadings for the other set. The difference chi-square (1 df) was 250.42 (p < .0001), indicating marked superiority for the two-factor model. This is consistent with two-factor models for self-control latent constructs in other studies (Wills et al., 2001; Wills & Stoolmiller, 2002).

4  Participant's age was dropped because it had effects on peer use but no effects on other variables, and including age did not affect other results. The measure for drug abstainer prototype was correlated with the self-control measures and resistance efficacy but was omitted from the structural model because it did not have any unique effects on subsequent variables net of the sexual abstainer prototype. It is possible that the compound nature of this measure (asking for an image of people who did not use cigarettes or alcohol or drugs) was too complex for the developmental level of the participants, resulting in lower validity. Further research could use more articulated measures, with prototypes obtained for different types of nonusers.

5  The model had 118 estimated parameters and so is below the 4:1 ratio of cases:parameters generally recommended for structural modeling. To test the generality of the findings, we performed a bootstrapping analysis in which 1,000 replications were derived by randomly drawing from the data with replacement. The structural model (see Figure 1) was run on each of the replicate samples, and parameter estimates were averaged across the analyses. Specified structural coefficients were all significant in the bootstrap analysis, and other parameters in the model fell within the confidence intervals of the averaged coefficients from the bootstrap analysis in almost all cases. Thus, there is reasonable evidence for the stability of the parameters in the present analysis (cf. Latimer, Newcomb, Winters, & Stinchfield, 2000).

Ary, D. V., James, L., & Biglan, A. (1999). Parent-daughter discussions to discourage tobacco use: Feasibility and content. Adolescence, 34, 275–282.

Bachman, J. G., Wallace, J. M., O'Malley, P. M., Johnston, L. D., Kurth, C. L., & Neighbors, H. W. (1991). Ethnic differences in smoking, drinking, and drug use among American high school seniors, 1976–89. American Journal of Public Health, 81, 372–377.

Bahr, S. J., Maughan, S. L., Marcos, A. C., & Li, B. (1998). Family, religiosity, and the risk of adolescent drug use. Journal of Marriage and the Family, 60, 979–992.

Baldwin, S. E., & Baranowski, M. V. (1990). Family interactions and sex education in the home. Adolescence, 25, 573–582.

Barnes, G. M., Farrell, M. P., & Banerjee, S. (1994). Family influences on alcohol abuse and other problem behaviors among Black and White adolescents in a general population sample. Journal of Research on Adolescence, 4, 183–201.

Blanton, H., Gibbons, F. X., Gerrard, M., Conger, K. J., & Smith, G. E. (1997). The role of family and peers in the development of prototypes associated with health risks. Journal of Family Psychology, 11, 1–18.

Blanton, H., VandenEijnden, R. J. J. M., Buunk, B. P., Gibbons, F. X., Gerrard, M., & Bakker, A. (2001). Social images in the prediction and promotion of condom use. Journal of Applied Social Psychology, 31, 274–295.

Block, J., Block, J., & Keyes, S. (1988). Foretelling drug use in adolescence: Early personality and environmental precursors. Child Development, 59, 336–355.

Block, J. H., & Block, J. (1980). The role of ego-control and ego-resiliency in the organization of behavior. In W. A.Collins (Ed.), Minnesota Symposium on Child Psychology (Vol. 13, pp. 39–101). Hillsdale, NJ: Erlbaum.

Brody, G. H., Flor, D. L., Hollett-Wright, N., & McCoy, J. K. (1998). Children's development of alcohol use norms: Contributions of parent and sibling norms, children's temperaments, and parent-child discussions. Journal of Family Psychology, 12, 209–219.

Brody, G. H., Stoneman, Z., & Flor, D. (1996). Parental religiosity, family processes, and youth competence in rural, two-parent African-American families. Developmental Psychology, 32, 696–706.

Brody, G. H., Stoneman, Z., Flor, D., & McCrary, C. (1994). Religion's role in organizing family relationships in rural, two-parent African American families. Journal of Marriage and the Family, 56, 878–888.

Buss, A., & Plomin, R. (1984). Temperament: Early developing personality traits. Hillsdale, NJ: Erlbaum.

Chassin, L. A., Tetzloff, C., & Hershey, M. (1985). Self-image and social-image factors in adolescent alcohol use. Journal of Studies on Alcohol, 46, 39–47.

Chou, C.-P., & Bentler, P. M. (1995). Estimates and tests in structural equation modeling. In R. H.Hoyle (Ed.), Structural equation modeling: Concepts, issues, and applications (pp. 37–54). Thousand Oaks, CA: Sage.

Conger, R., Conger, K., Elder, G., & Lorenz, F. (1992). A family process model of economic hardship and adjustment of adolescent boys. Child Development, 63, 526–541.

DiClemente, R., Hansen, W., & Ponton, L. (Eds.). (1996). Handbook of adolescent health risk behavior. New York: Plenum.

Donahue, M. J., & Benson, P. L. (1995). Religion and the well-being of adolescents. Journal of Social Issues, 51, 145–160.

Eysenck, S. B. G., & Eysenck, H. J. (1977). The place of impulsiveness and venturesomeness in a dimensional system of personality description. British Journal of Social and Clinical Psychology, 16, 57–68.

Foshee, V. A., & Hollinger, B. R. (1996). Maternal religiosity, adolescent social bonding, and adolescent alcohol use. Journal of Early Adolescence, 16, 451–468.

Gerrard, M., Gibbons, F. X., & Bushman, B. J. (1996). Relation between perceived vulnerability to HIV and precautionary behavior. Psychological Bulletin, 119, 390–409.

Gerrard, M., Gibbons, F. X., Reis-Bergan, M., Trudeau, L., Vande Lune, L., & Buunk, B. P. (2002). Health cognitions and adolescent alcohol consumption: Inhibitory effects of drinker and non-drinker prototypes. Health Psychology, 21, 601–609.

Gibbons, F. X., & Gerrard, M. (1995). Predicting young adults' health risk behavior. Journal of Personality and Social Psychology, 69, 505–517.

Gibbons, F. X., & Gerrard, M. (1997). Health images and their effects on health behavior. In B.Buunk & F. X.Gibbons (Eds.), Health, coping, and well-being: Perspectives from social comparison theory (pp. 63–94). Mahwah, NJ: Erlbaum.

Gibbons, F. X., Gerrard, M., & Boney-McCoy, S. (1995). Prototype perceptions predict pregnancy prevention. Personality and Social Psychology Bulletin, 21, 85–93.

Gibbons, F. X., Gerrard, M., Ouellette, J. A., & Burzette, R. (1998). Cognitive antecedents to adolescent health risk. Psychology and Health, 13, 319–339.

Hadaway, C., Elifson, K., & Petersen, D. (1984). Religious involvement and drug use among urban adolescents. Journal for the Scientific Study of Religion, 23, 109–128.

Hawkins, J. D., Graham, J. W., Maguin, E., Abbott, R., Hill, K. G., & Catalano, R. F. (1997). Exploring the effects of age of alcohol use initiation and psychosocial risk factors on subsequent alcohol misuse. Journal of Studies on Alcohol, 58, 28–290.

Hays, R. D., & Ellickson, P. L. (1990). Adolescents' beliefs about pro-drug pressures and resistance self-efficacy. Journal of Applied Social Psychology, 20, 321–340.

Hays, R. D., Widaman, K. F., DiMatteo, M. R., & Stacy, A. W. (1987). Structural equation models of drug use. Journal of Personality and Social Psychology, 52, 134–144.

Hill, P. C., & Wood, R. W., Jr. (Eds.). (1999). Measures of religiosity. Birmingham, AL: Religious Education Press.

Howard, M., & McCabe, J. A. (1992). An information and skills approach for younger teens: Postponing Sexual Involvement Program. In B. C.Miller & J. J.Card (Eds.), Preventing adolescent pregnancy (pp. 83–109). Newbury Park, CA: Sage.

Institute of Medicine. (1997). The hidden epidemic: Confronting sexually transmitted diseases. Washington, DC: National Academy Press.

Johnston, L. D., O'Malley, P. M., & Bachman, J. G. (1995). National survey results on drug use from the Monitoring the Future study, 1975–1994: Volume 1. Secondary school students. Rockville, MD: National Institute on Drug Abuse.

Kafka, R. R., & London, P. (1991). Communication in relationships and adolescent substance use: The influence of parents and friends. Adolescence, 26, 587–598.

Kendall, P. C., & Williams, C. L. (1982). Assessing the cognitive and behavioral components of children's self-management. In P.Karoly & F. H.Kanfer (Eds.), Self-management and behavior change (pp. 240–284). New York: Pergamon Press.

Kotchick, B. A., Dorsey, S., Miller, K., & Forehand, R. (1999). Adolescent sexual behavior in minority single-parent families. Journal of Family Psychology, 13, 93–102.

Landrine, H., Richardson, J., Klonoff, E., & Flay, B. (1994). Cultural diversity in the predictors of adolescent cigarette smoking. Journal of Behavioral Medicine, 17, 331–346.

Latimer, W., Newcomb, M., Winters, K., & Stinchfield, R. (2000). Adolescent substance abuse treatment outcome. Journal of Consulting and Clinical Psychology, 68, 684–696.

Leigh, B. C., & Stall, R. (1993). Substance use and risky sexual behavior for exposure to HIV. American Psychologist, 48, 1035–1045.

Levin, J. S. (1996). How religion influences morbidity and health: Reflections on natural history, salutogenesis and host resistance. Social Science and Medicine, 43, 849–864.

Little, R. J. A., & Rubin, D. B. (1987). Statistical analysis with missing data. New York: Wiley.

Lorch, B. R., & Hughes, R. H. (1985). Religion and youth substance use. Journal of Religion and Health, 26, 9–36.

Maggs, J. L., Almeida, D. M., & Galambos, N. L. (1995). The paradoxical meaning of risk behavior for young adolescents. Journal of Early Adolescence, 15, 344–362.

McCullough, M. E., Hoyt, W. T., Larson, D. B., Koenig, H. G., & Thoresen, C. (2000). Religious involvement and mortality: A review. Health Psychology, 19, 211–222.

Mezzich, A. C., Tarter, R. E., Giancola, P. R., Lu, S., Kirisci, L., & Parks, S. (1997). Substance use and risky sexual behavior in female adolescents. Drug and Alcohol Dependence, 44, 157–166.

Miller, W. R., & Brown, J. M. (1991). Self-regulation as a conceptual basis for the prevention of addictive behaviours. In N.Heather, W. R.Miller, & J.Greeley (Eds.), Self-control and the addictive behaviours (pp. 3–79). Sydney, Australia: Maxwell Macmillan.

Murray, D. M., & Perry, C. L. (1987). Measurement of substance use among adolescents. Addictive Behaviors, 12, 225–233.

Murry, V. M. (1995). Pregnancy resolution decisions among African-American and Hispanic adolescent females. Youth and Society, 26, 325–350.

Murry, V. M. (1996). An ecological analysis of coital timing among middle-class African American adolescent females. Journal of Adolescent Research, 11, 261–279.

Murry, V. M., & Brody, G. H. (1999). Self-regulation and self-worth of Black children reared in economically stressed, rural, single mother-headed families. Journal of Family Issues, 20, 458–484.

Muthén, L., & Muthén, B. (1998). Mplus: User's guide. Los Angeles, CA: Muthén & Muthén.

Needle, R., Su, S., & Lavee, Y. (1989). A comparison of the empirical utility of three composite measures of adolescent drug involvement. Addictive Behaviors, 14, 429–441.

Newcomb, M. D., & Bentler, P. M. (1988). Impact of adolescent drug use and social support on problems of young adults. Journal of Abnormal Psychology, 97, 64–75.

Patrick, D. L., Cheadle, A., Thompson, D. C., Diehr, P., Koepsell, T., & Kinne, S. (1994). The validity of self-reported smoking: A review and meta-analysis. American Journal of Public Health, 84, 1086–1093.

Resnick, M. D., Bearman, P. S., Blum, R. W., Bauman, K. E., Harris, K. M., Jones, J., et al. (1997). Findings from the National Longitudinal Study on Adolescent Health. Journal of the American Medical Association, 278, 823–832.

Rothbart, M. K., & Ahadi, S. A. (1994). Temperament and the development of personality. Journal of Abnormal Psychology, 103, 55–66.

Rothbart, M. K., & Bates, J. E. (1998). Temperament. In W.Damon (Series Ed.) & N.Eisenberg (Vol. Ed.), Handbook of child psychology: Vol. 3. Social, emotional, and personality development (5th ed., (pp. 105–176). New York: Wiley.

Rutter, M., Dunn, J., Plomin, R., Simonoff, E., Pickles, A., Maughan, B., et al. (1997). Integrating nature and nurture: Implications of person-environment correlations and interactions. Development and Psychopathology, 9, 335–364.

Scarr, S., & McCartney, K. (1983). How people make their own environments: A theory of genotype → environment effects. Child Development, 54, 424–435.

Spoth, R., Redmond, C., & Shin, C. (1999). Direct and indirect parenting outcomes of family-focused preventive interventions: Extending a public health-oriented research base. Journal of Consulting and Clinical Psychology, 66, 385–399.

Vaccaro, D., & Wills, T. A. (1998). Stress-coping factors in adolescent substance use: Ethnic and gender differences. Journal of Drug Education, 28, 257–282.

Wallace, J. M., & Williams, D. R. (1997). Religion and adolescent health-compromising behavior. In J.Schulenberg, J.Maggs, & K.Hurrelmann (Eds.), Health risks and developmental transitions during adolescence (pp. 444–468). New York: Cambridge University Press.

West, S. G., Finch, J. F., & Curran, P. J. (1995). Structural equation models with nonnormal variables. In R. H.Hoyle (Ed.), Structural equation modeling: Concepts, issues, and applications (pp. 56–75). Thousand Oaks, CA: Sage.

Whitaker, D. J., & Miller, K. S. (2000). Parent-adolescent discussions about sex: Impact on peer influences of sexual behavior. Journal of Adolescent Research, 15, 251–274.

Wills, T. A., Blechman, E. A., & McNamara, G. (1996). Family support, coping and competence. In E. M.Hetherington & E. A.Blechman (Eds.), Stress, coping, and resiliency in children and the family (pp. 107–133). Mahwah, NJ: Erlbaum.

Wills, T. A., & Cleary, S. D. (1996). How are social support effects mediated: A test for parental support and adolescent substance use. Journal of Personality and Social Psychology, 71, 937–952.

Wills, T. A., & Cleary, S. D. (1997). The validity of self-reports of smoking in a school sample of urban adolescents. American Journal of Public Health, 87, 56–61.

Wills, T. A., & Cleary, S. D. (1999). Peer and adolescent substance use among 6th–9th graders: Latent-growth analyses. Health Psychology, 18, 453–463.

Wills, T. A., Cleary, S. D., Filer, M., Shinar, O., Mariani, J., & Spera, K. (2001). Temperament related to early-onset substance use: Test of a developmental model. Prevention Science, 2, 145–163.

Wills, T. A., DuHamel, K., & Vaccaro, D. (1995). Activity and mood temperament as predictors of adolescent substance use: Test of a self-regulation mediational model. Journal of Personality and Social Psychology, 68, 901–916.

Wills, T. A., & Filer, M. (2000). Social networks and social support. In A.Baum & T.Revenson (Eds.), Handbook of health psychology (pp. 209–234). Mahwah, NJ: Erlbaum.

Wills, T. A., McNamara, G., & Vaccaro, D. (1995). Parental education related to adolescent stress-coping and substance use. Health Psychology, 14, 464–478.

Wills, T. A., McNamara, G., Vaccaro, D., & Hirky, A. E. (1996). Escalated substance use: A longitudinal grouping analysis. Journal of Abnormal Psychology, 105, 166–180.

Wills, T. A., Sandy, J. M., & Yaeger, A. (2000). Temperament and adolescent substance use: An epigenetic approach. Journal of Personality, 68, 1127–1152.

Wills, T. A., & Stoolmiller, M. (2002). Role of self-control in escalation of substance use: A time-varying analysis. Journal of Consulting and Clinical Psychology, 70, 986–997.

Wills, T. A., Vaccaro, D., & McNamara, G. (1994). Novelty seeking and related constructs as predictors of adolescent substance use. Journal of Substance Abuse, 6, 1–20.

Wills, T. A., Windle, M., & Cleary, S. D. (1998). Temperament and novelty-seeking in adolescent substance use. Journal of Personality and Social Psychology, 74, 387–406.

Windle, M., & Lerner, R. M. (1986). The Revised Dimensions of Temperament Survey. Journal of Adolescent Research, 1, 213–229.

Zuckerman, M. (1994). Impulsive unsocialized sensation seeking. In J. E.Bates & T. D.Wachs (Eds.), Temperament: Individual differences at the interface of biology and behavior (pp. 219–255). Washington, DC: American Psychological Association.

Submitted: June 28, 2002 Revised: November 14, 2002 Accepted: November 21, 2002