Longitudinal modeling of adolescent normative beliefs and substance initiation.

Abstract

The current study investigated the effects of baseline levels of academic achievement and longitudinal trends in normative beliefs on adolescent substance initiation across a 42-month time period. Participants were 272 rural adolescents who were an average of 12.3 years old at the baseline assessment. Academic achievement positively predicted the intercept and negatively predicted the growth-trajectory of normative beliefs regarding peer substance behavior. Further, baseline academic achievement negatively predicted initial levels, as well as the growth-trajectory, of substance initiation. The discussion addresses the influence of academic achievement and normative beliefs on substance initiation and the utility of latent growth curve modeling in studying longitudinal change. In addition, implications for prevention programming are discussed.

INTRODUCTION

Research models investigating the linkages between adolescent substance use and beliefs about normative behavior generally have revealed strong, positive associations (Botvin, Griffin, Diaz, & Ifill-Williams, 2001; Botvin, Botvin, Baker, Dusenbury, & Goldberg, 1992; Ennett & Bauman, 1991; Krohn, Akers, Radosevich, & Lanza-Kaduce, 1982; Lowman, 1982; Petraitis, Flay, & Miller, 1995). However, longitudinal empirical studies of the association between beliefs about normative behavior and substance initiation behaviors have been less extensive. The focus of the current study was on the longitudinal growth-trajectories of beliefs about peer and adult normative behavior and substance initiation. Beliefs about normative use were defined as perceptions concerning levels of specific peer and adult substance use behaviors. The next section briefly reviews the relationship between normative beliefs and substance initiation.

During adolescence, beliefs related to normative behavior (e.g., substance use) undergo a transition as new information from differing groups is integrated into a personal cognitive framework. As part of that transition, adolescents progress from an almost exclusive identification with their parents to a reliance on peers. This transition combines with exposure to differing social groups and the media to influence adolescents' beliefs about normative substance use behavior. Normative beliefs are influenced by a variety of factors including the behavior of the referent group(s) with which adolescents most closely identify (Guerra, Huesmann, & Hanish, 1995). Findings from the literature are mixed concerning the influence of referent group(s) (e.g., parents and peers) on adolescents' normative beliefs (Glynn, 1981; Kandel & Lesser, 1969). Although many theories have been developed to explain the differential influence of multiple referent groups, the amount of variance in adolescent substance use accounted for by the perceived behavior of peers has been found to greatly exceed that accounted for by the perceived behavior of parents and other adults (Hansen, Graham, Sobel, Shelton, Flay, & Johnson, 1987). For example, beliefs about peer normative behavior, as compared to beliefs about adult normative behavior have been found to be more predictive of adolescent tobacco use (Botvin et al., 1992). Alternatively, peer non-use normative beliefs have been negatively associated with adolescent substance use (Johnson, 1986; Ried, Martinson, & Weaver, 1987). Research findings have documented greater overestimations of peer substance behavior among substance using adolescents compared with their non-using counterparts (Presson, Chassin, Sherman, Olshavsky, Bensenberg, & Corty, 1983).

Early theories regarding normative beliefs described a concept known as the "false consensus effect" (Ross, Greene, & House, 1977, p. 279). According to the false consensus effect, prevalence rate over-estimations predict subsequent engagement in that behavior. Two hypothesized mechanisms have been documented that might contribute to the false consensus effect: (a) the need to validate personal behavior and (b) selective exposure that fosters the perception that others are engaged in similar activities (Botvin et al., 1992). For example, adolescents who believe that most peers and adults engage in a given activity, such as substance use, might initiate that behavior in order to fulfill an expectation of normative behavior (Botvin et al., 1992).

To understand the influence of initial levels and change over time in adolescent normative beliefs on longitudinal substance initiation, it is useful to study growth-trajectories over multiple points in time (Loeber & Farrington, 1994). For the most part, however, earlier research has examined the association between adolescent normative beliefs and substance behavior with other methodologies, such as cross-sectional studies or longitudinal studies with only two measurement occasions. Although informative, cross-sectional studies do not provide a clear understanding of the initiation process because differences between users and non-users might be either a cause or a consequence of substance behavior. Longitudinal studies can identify the antecedents of substance initiation (e.g., beliefs about normative behavior) and allow for the systematic investigation of non-users who might be at-risk of becoming users. Many earlier longitudinal studies of these constructs, however, have used only two measurement occasions (e.g., baseline and follow-up) and have not evaluated change across multiple time points or the influence of initial levels on growth-trajectories. As an illustration, in a study with two measurement occasions, a positive association was found between baseline estimates of normative beliefs (i.e., peer and adult) and follow-up tobacco use (Bauman, Botvin, Botvin, & Baker, 1992). The current study aims to extend earlier investigations by examining normative beliefs (i.e., peer and adult) and substance initiation over five measurement occasions, plus the influence of initial levels and growth-trajectories of those constructs. Although previous studies have investigated the relationships between substance use and normative beliefs (i.e., both peer and adult), to our knowledge this is the only investigation to evaluate a growth-trajectory model incorporating those constructs. To accomplish the study goals, Latent Growth Curve Modeling (LGCM) was used to depict the process and development of adolescent substance initiation as a function of repeated observations over time. The LGCM methodology also allowed for the investigation of a time-invariant proximal predictor to test hypotheses concerning academic achievement, which was expected to exert an influence on both normative beliefs and Initiation.

The current study included baseline levels of academic achievement as a proximal predictor of normative beliefs and initiation because earlier research has found an association between adolescent educational difficulties and substance use (Bailey & Hubbard, 1990; Kandel, Kessler, & Margulies, 1978, 1985; Shedler & Block, 1990). Specifically, previous research provided support for the idea that academic underachievement might lead to an increased involvement in substance use (Dryfoos, 1990; Hawkins, Catalano, & Miller, 1992; Hindelang, 1973; Hundleby & Mercer, 1987; Institute of Medicine, 1994; Jessor, 1976; Robins, 1980). As an example, a cross-sectional study with young adolescents found both academic achievement and peer-related normative beliefs predicted tobacco use; academic underachievers were at-risk for tobacco initiation (Dusenbury, Kerner, Baker, Botvin, James-Ortiz, & Zauber, 1992). In a review article, Loeber and Dishion (1983) cited evidence to support the idea that adolescent academic underachievers are prone to problem behaviors, such as substance use. Academic underachievement has been found to predict both frequency and level of illicit substance use at various points during adolescence (Boyle et al., 1993; Bucholz, 1990; Dryfoos, 1990; Duncan, Duncan, Biglan, & Ary, 1998; Hawkins et al., 1992; Jessor & Jessor, 1975; O'Donnell, Hawkins, & Abbott, 1995; Paulson, Coombs, & Richardson, 1990; Smith & Fogg, 1979). Arguably, academic underachievers are less committed to educational pursuits and are more likely to engage in unconventional pursuits (e.g., substance use) and to associate with peers who engage in related behaviors. Academic underachievers might be presented with more social opportunities to engage in substance use behaviors as a result of greater exposure to reinforcements that model, condone, or approve of substance behaviors (Bourgois, 1995; Hawkins, 1997). Because research findings to date have indicated that academic underachievement and normative beliefs predict both substance use and the initiation process (Petraitis, Flay, & Miller, 1995), the current study hypothesized that those constructs would be associated with substance initiation. The current study hypothesized that adolescents who were academic achievers might be more likely to acquire accurate information about normative behavior because of attention to relevant information, as well as more likely to associate with non-using peers who are focused on academic activities. The current study examined whether academic achievement would predict both initial levels and growth-trajectories of normative beliefs (i.e., peer and adult) and substance initiation.

Hypotheses

The goals of the current study were (a) to investigate individual differences in the initial levels and growth-trajectories of normative beliefs and substance initiation and (b) to examine a longitudinal model in which baseline levels of academic achievement influenced both initial levels and growth-trajectories of normative beliefs and substance initiation.

There are two hypotheses related to the first goal of the study. First, the study hypothesized significant individual differences on initial levels (i.e., intercepts) of normative beliefs and substance initiation. The second hypothesis was that there would be significant individual differences in the growth-trajectories of each construct, with an overall positive longitudinal change in normative beliefs (i.e., peer and adult) and substance initiation.

Regarding the second goal, the study evaluated a longitudinal model in which all the relevant constructs were considered. To our knowledge, this is the first study to explore systematically the pathways from academic achievement to initial levels and growth-trajectories of normative beliefs and substance initiation. Three hypotheses were tested related to this goal. The first hypothesis was that academic achievement would be inversely related to initial levels and change over time in normative beliefs (i.e., peer and adult). Second, based on research findings documented previously it was hypothesized that academic achievement would be inversely related to initial levels of initiation, as well as change over time. The third hypothesis related to this goal concerned the interrelations between initial levels and growth-trajectories for normative beliefs and substance initiation; a positive association among the initial levels and the growth-trajectories of those variables was expected. In particular, it was hypothesized that initial levels of peer normative beliefs would be positively associated with initial levels of both adult normative beliefs and initiation and that initial levels of adult normative beliefs would be positively associated with initial levels of substance initiation. A similar positive association was expected among the growth-trajectories of the study variables. Initial levels of normative beliefs were expected to positively predict the growth-trajectory of substance initiation. Finally, an inverse association between the intercept and the growth-trajectory of initiation was expected.

Figure 1 presents the full longitudinal model. The model includes both latent constructs (i.e., intercepts and slopes of peer and adult normative beliefs, plus substance initiation) and a time invariant predictor, baseline levels of academic achievement.

[FIGURE 1 OMITTED]

METHODS

Participants

Seventh grade students enrolled in 36 schools in 22 contiguous rural counties in a Midwestern state were assessed in the first wave of data collection. The 36 schools were recruited from a pool of 43 randomly selected schools in the region. Criteria for selection of the initial pool of schools included (a) 20% or more students eligible for the free and reduced-cost school lunch program, (b) district-wide school enrollment less than 1,200 students, and (c) all middle school grades (i.e., sixth through eighth) taught in one building.

A randomized block design guided the assignment of the 36 schools to experimental groups. Schools were randomly assigned to one of three experimental groups (a) the classroom-based Life Skills Training (Botvin, 2000), plus the family-focused Strengthening Families Program: For Parents and Youth 10-14 (Molgaard, Kumpfer, & Fleming, 1997), (b) the Life Skills Training only, or (c) a minimal contact control group. Following random assignment, schools were contacted and informed of the experimental group to which they had been assigned.

All seventh grade students in participating schools were recruited for involvement in the study. A total of 1,649 (1) students in the 36 schools completed the baseline pretest during the fall semester, including 488 control group students. Only control group students were selected for the study analyses. During the spring semester of seventh grade, posttest assessments were conducted with 457 (94%) control group students. Of these, 410 (90% of those posttested) completed the follow-up 1 assessment during the spring semester of eighth grade, approximately one-year after the posttest and one and one-half years after the pretest. During the spring semester of ninth grade, the follow-up 2 assessment was conducted with 365 control group students (89% of the previous wave); the follow-up 3 assessment during the spring semester of tenth grade was conducted with 318 students (87% of the previous wave). Slightly over one-half of the students were boys (51%) and the majority of participants were Caucasian (96%). To fully consider the study model without the impact of the school-based or family-focused program, the current investigation considered only control group students who provided complete assessment data for all five assessments (N = 272).

Procedures

The in-school data collection was conducted in classrooms. A passive consent procedure, that explained the assessment process and the carbon monoxide (CO) procedure, allowed parents to decline participation for their adolescent child; approximately 90% of eligible students participated in the pretest. About 40 to 45 minutes were required to complete the questionnaire. Typically, there were three to four data collectors in each classroom to coordinate assessment procedures. Students were assured that their responses to the questionnaire would be kept confidential and were allowed to decline involvement at the time of assessment. Two forms of the questionnaire were administered in each classroom to assess order effect of the individual items. Identical questions were asked in each form; only the order of items was varied. In addition, each student exhaled into a balloon, which was then connected to a CO meter to provide a CO reading. The primary purpose of this procedure was to serve as a "bogus pipeline" to strengthen validity of the smoking-related questions (Bauman, Koch, & Bryan, 1988; Botvin, Botvin, Renick, Filazzola, & Allegrante, 1984; Evans, Hansen, & Mittlemark, 1977; Murray, Hannan,& Zucker, 1989). The same assessment procedures were employed across all data collection points.

Measures

Normative beliefs. Two scales of normative beliefs were constructed. The first scale assessed beliefs concerning how many peers engaged in five substance use behaviors: smoke cigarettes, drink alcohol, smoke marijuana, use cocaine, and sniff glue. Scale reliability was a = .88 at pretest. The second scale assessed beliefs concerning how many adults engaged in the same five substance use behaviors: smoke cigarettes, drink alcohol, smoke marijuana, use cocaine, and sniff glue. Pretest scale reliability was a = .85. Response categories for both scales ranged from 1 = None through 5 = All or almost all.

Academic achievement. Academic achievement was measured with a single item. Students were asked what grades they generally achieved in school. Response categories were on a one to five scale ranging from 1 = Mostly A's through 5 = Mostly lower than D's. The scale was recoded so that higher scores reflected greater academic achievement. Previous studies have used similar one-item scales to assess academic achievement (Polk, 1975).

Substance initiation. The substance initiation scale was constructed by combining three dichotomous items regarding the lifetime use of cigarettes, alcohol, and marijuana. Participants were asked if they had ever "Smoked a cigarette," "Had a drink of alcohol," or "Smoked marijuana." Responses were coded 0 = No and 1 = Yes and summed to compute the scale. For each wave of data, responses were corrected for consistency, so that if a student answered "Yes" at any point in time, the response to the same question in subsequent data collection events also was coded "Yes." For example, if a student responded "Yes" at Wave 1 and "No" at Wave 2, the Wave 2 response was recoded to "Yes." Thus, the substance initiation scale was coded 0 = No for all substance use, 1 = Yes for use of one substance, 2 = Yes for use of two substances, and 3 = Yes for use of three substances. Test-retest reliability averaged across waves was r = .82. Because the scale was composed of dichotomous items, reliability was also assessed with the Kuder-Richardson (KR20) formula and averaged .55 across waves. Reliability and validity of self-reported substance use have been evaluated previously (Murray, O'Donnell, Schmid, & Perry, 1987; Oetting & Beauvais, 1990). In addition, a recent meta-analytic study of adolescent prevention programs (Tobler et al., 2000) determined that self-reported substance use measures were considered reliable.

Data Analyses

Study hypotheses were examined with latent growth curve modeling using MPLUS (Muthen & Muthen, 1998) to estimate the growth-trajectories with a structural equation-modeling program controlling for the clustered sample (i.e., students were clustered within schools). Latent growth curve modeling was conducted with maximum likelihood with robust standard errors estimation to evaluate relationships among the variables (see Appendix A for further detail). This method was utilized to adjust the standard errors within clusters to control for school-level effects. Overall model fit was assessed using the chi-square value and two fit indices: the likelihood ratio chi-square and the Root Mean Square Error of Approximation (RMSEA). T-tests evaluated the significance of individual model parameters.

Analyses also were conducted to determine if study drop-outs (i.e., pretest to posttest, posttest to follow-up 1, follow-up 1 to follow-up 2, and follow-up 2 to follow-up 3) differed on substance initiation. To determine whether those students who initiated substance use had higher drop-out rates than those who did not dropout of the study, an Analyses of Variance (ANOVA) was conducted. The ANOVA analyzed the substance initiation variable at each measurement occasion. Students who withdrew (e.g., dropped out of the study, moved to a different school) from the study between baseline and posttest (n = 29) had significantly higher scores on the initiation measure, F(1,478) = 18.2, p < .001. Additional students who withdrew between posttest and follow-up 1 (n = 47) also had significantly higher scores on the initiation measure, F(1,452) = 4.5, p < .05. Students who withdrew between follow-up 1 and follow-up 2 (n = 69) had significantly higher scores on the initiation measure, F(1,484) = 17.14, p < .01, as well as those who withdrew between follow-up 2 and follow-up 3 (n = 96), F(1,423) = 4.30, p < .05. Those analyses demonstrated greater attrition rates among students who reported higher scores on the substance initiation measure.

RESULTS

Univariate Longitudinal Models

Means, standard deviations, and correlations of the model variables are presented in Table 1. Using MPLUS, univariate growth curves were estimated separately for peer normative beliefs, adult normative beliefs, and substance initiation. Because the slopes were hypothesized to be linear, the values of the slope parameters were set at "0" for baseline, "1" for posttest (i.e., six months following baseline), "3" for follow-up 1(i.e., one year following posttest), "5" for follow-up 2 (i.e., two years following posttest), and "7" for follow-up 3 (i.e., three years following posttest); one unit represented a six month time frame. Because a linear change model did not fit for substance initiation, the factor loadings for follow-up 1 through follow-up 3 were allowed to be freely estimated by the program. The residuals between adjacent measurement occasions for substance initiation were allowed to correlate, based on stability estimates demonstrating an autoregressive trend, and the residuals of the variables across each time point were allowed to correlate. The univariate growth curve models for peer normative beliefs, adult normative beliefs, and substance initiation all demonstrated a very good fit to the data (see Table 2 for fit statistics and the unstandardized coefficients). Significant differences were hypothesized for the initial level and growth-trajectory of each construct. The intercepts and their respective variances were significant for all three constructs, indicating considerable differences from zero and substantial differences among adolescents on initial levels of the variables. The slopes and respective variances of peer normative beliefs and substance initiation also were significant, demonstrating longitudinal increases in those constructs; however, adult normative beliefs remained relatively stable across the study time period. As a result, partial support was found for both hypotheses related to the first study goal.

Full Longitudinal Model

The full longitudinal model was examined with the concurrent evaluation of baseline academic achievement and the growth-trajectories for peer normative beliefs, adult normative beliefs, and substance initiation. Results from the LGCM found the data fit the model well: [chi square] (86, N = 272) = 523.14,p < .01, CFI = .94, and RMSEA = .06. The model is presented in Figure 2; standardized values are presented for the coefficients.

[FIGURE 2 OMITTED]

The study hypothesis related to the inverse pathways from academic achievement to normative beliefs was partially supported. Specifically, the path coefficient from academic achievement to the slope of peer normative beliefs ([beta] = -.27,p < .01) was significant and negative; however, the path coefficient from academic achievement to the intercept of peer normative beliefs ([beta] = .21, p < .01) was significant and positive. Higher baseline academic achievement was a positive predictor of initial levels of peer normative beliefs and a negative predictor of the growth-trajectory of peer normative beliefs. The hypothesis that academic achievement would negatively predict adult normative beliefs (i.e., both intercept and growth-trajectory) was not supported. The second hypothesis was that academic achievement would negatively predict initial levels of initiation, as well as the growth-trajectory. Support was found for this hypothesis; academic achievement negatively predicted initial levels of substance initiation ([beta] = -.15, p < .01), as well as the initiation growth-trajectory ([beta] = -.11, p < .01).

The third hypothesis concerned the interrelations among the study constructs. Specifically, a positive association among the initial levels and the growth-trajectories was expected. A significant positive association was found between the intercept of peer normative beliefs and the intercept of adult normative beliefs (r = .49, p < .01). In addition, it was expected that peer and adult normative beliefs would be positively correlated with substance initiation during the same timeframe (i.e., intercepts). An examination of the correlations between the intercepts of normative beliefs (i.e., peer and adult) and substance initiation found significant associations (r = .18, p < .01 and r = .06, p < .01, respectively). Support was found for the study hypothesis concerning the correlation between the substance initiation intercept and slope; a negative association was found between initial levels and change over time. This finding suggests that individuals with lower initial initiation levels increased at a faster rate than those with initially higher levels. The hypothesis that initial levels of normative beliefs would positively predict the slope of substance initiation was partially supported. Specifically, the path from the intercept of adult normative beliefs to the slope of substance initiation was significant ([beta] = .07,p < .01); initial levels of adult normative beliefs predicted the growth-trajectory of initiation. Contrary to expectations, the pathway from the intercept of peer normative beliefs to the initiation growth-trajectory was not significant. Finally, support was found for the study hypothesis regarding the correlation between the slopes of adult normative beliefs and substance initiation (r = .06, p < .01); however, the association between the slope of peer normative beliefs and initiation was not significant.

DISCUSSION

The purpose of the current study was to evaluate differences among young adolescents in initial levels and growth-trajectories of normative beliefs and substance initiation. The hypothesized model also considered the influence of baseline academic achievement on initial levels and growth-trajectories of normative beliefs and initiation. For the most part, earlier research has examined normative beliefs and substance initiation with cross-sectional or longitudinal design methodologies. Longitudinal panel studies have focused typically on one type of change, which is change in the rank order of individuals with respect to levels of normative beliefs or substance initiation. Although informative, the limitations of longitudinal studies include a failure to focus on change over time within an individual adolescent (Rogosa, 1988; Rogosa & Willett, 1985; Stoolmiller & Bank, 1995). In contrast, the current study used methods that focused on within-subject change over time through the analysis of individual growth curves (Duncan, Duncan, Strycker, Li, & Alpert, 1999; Willett & Sayer, 1994, 1996). Results of the current study go beyond existing analyses by examining (a) associations between initial levels of normative beliefs and substance initiation and (b) growth-trajectories of normative beliefs and substance initiation.

The first goal of the study was to examine initial levels and growth-trajectories of normative beliefs and substance initiation. To begin, the hypotheses related to initial levels of normative beliefs and initiation were mostly supported. Significant individual differences were found for the intercepts of normative beliefs (i.e., peer and adult) and substance initiation. Those values indicated that initial levels were significantly different from zero and significantly different between individuals. In addition, support was found for the hypothesis that there would be differences among individual adolescents in the growth-trajectories of peer normative beliefs and substance initiation, as well as an overall positive growth-trajectory Over the 42-month time period of the study, some adolescents remained in similar trajectories of growth related to peer normative behavior and initiation, while others increased. Although the average slope for beliefs about adult normative behavior was non-significant, differences were found for the variance of the slope. Thus, meaningful differences were demonstrated between adolescents in the growth-trajectory of beliefs about adult normative behavior.

Previous research findings have demonstrated that the development of adolescent normative beliefs (i.e., peer and adult) might be influenced over time by a variety of factors. Our study evaluated the influence of academic achievement, but there are certainly many other factors that might impact normative beliefs. For example, adolescents might develop impressions about adult normative behavior through (a) discussions with their parents and other adults (e.g., teachers), (b) the quality of the relationship with their parents and other adults, and (c) the substance use behavior of parents and other adults (Brody, Flor, Hollett-Wright, & McCoy, 1998; Brody, Flor, Hollett-Wright, McCoy, & Conovan, 1999). If parents choose not to discuss with their adolescent the rules and expectations for substance use, it is possible that the behavior of other adults and/or peers might become more important in the development of substance normative beliefs. Further research is needed to explore the influence of additional factors, such as the substance behavior of other relevant adults (e.g., teachers), on the growth-trajectories of adolescents' normative beliefs. Researchers also would benefit from longitudinal investigations of the development of substance normative beliefs among youth even younger then those who participated in the current study. Such investigations could reveal pertinent information about changes in normative beliefs related to substance use, plus permit an examination of how and why normative beliefs change with age. Further investigations also might help to identify early childhood normative beliefs that are possible indicators for substance initiation during adolescence (Brody, Ge, Katz, & Arias, 2000).

The second study goal was to evaluate the full longitudinal model. An evaluation of the results from the latent growth curve model suggested two general conclusions related to the hypothesized patterns of change over time among the study variables. The first conclusion concerned the influence of baseline academic achievement on initial levels and growth-trajectories of normative beliefs and initiation. Two sets of results support the conclusion related to the pathways from academic achievement to initial levels and growth-trajectories of normative beliefs and initiation. First, evidence was found to support the hypothesis related to the pathway from academic achievement to the growth-trajectory of peer normative beliefs. Adolescents who were academic underachievers experienced a greater rate of change over time in the perception of peer normative substance behaviors. Thus, the peer normative beliefs growth-trajectory for a given adolescent varied systematically with the same individual's academic underachievement. The mechanism leading to the greater rate of change in the perception of peer normative substance behaviors is not clear. One possible explanation is that this group of adolescents might have peers who engage in substance use (i.e., tobacco, alcohol, and/or marijuana) that lead to beliefs and impressions related to the prevalence of such behaviors. Further, adolescents who are academic underachievers might be presented with and engage in social opportunities that model and condone perceptions of high levels of substance behavior among peers.

A second set of results supported the hypothesized influence of academic achievement on substance initiation. In particular, academic achievement negatively predicated initial levels and the growth-trajectory of initiation. Thus, the hypothesis that academic achievement would predict initial levels and change over time in substance initiation was supported. Consistent with previous findings (Duncan et al., 1998), study results indicated that academic underachievers reported greater initial levels, as well as faster growth-trajectories, in substance initiation. Although results from the Duncan et al. study found gender differences in substance use intercept values (i.e., higher for girl), the current study did not include gender differences. Supplemental study analyses considered gender moderation of the full model; a chi-square difference test failed to reveal gender moderation of the model. However, differences in study design methodologies might account for the variation in research results. For example, in the Duncan et al. study, substance use (i.e., tobacco, alcohol, marijuana) prevalence rates were measured, whereas the current study considered a composite index of the initiation process. In addition, the Duncan et al. study measured the influence (of a composite index of social context that included two academic achievement items (i.e., grade point average and time spent on homework) on the growth parameters of three substances (i.e., tobacco, alcohol, marijuana). In contrast, the current study considered the direct pathways from academic achievement (i.e., grades) to initial levels and growth-trajectories of substance initiation.

Although previous studies (Jesssor & Jessor, 1977) have examined academic underachievement as a problem behavior by itself, other research findings (Kandel et al., 1978; Smith & Fogg, 1978) have suggested the importance of academic underachievement as an antecedent of problem behaviors, such as substance initiation. Results in the current study support the idea that academic underachievement is a precursor to the problem behavior of substance initiation. Although initiation of substance use might contribute to academic underachievement, it seems likely that adolescents who do not perform well academically might be at a greater risk for association with deviant peers and for initiation of substance behaviors. While analyses in the current study support and extend previous findings (Bauman et al., 1992; Duncan et al., 1998; Leventhal, Glynn, & Fleming, 1987) that have documented the influence of academic achievement in the development of adolescent substance initiation, further research is needed to explore the exact mechanism by which academic achievement might influence the initial levels and growth-trajectories of initiation.

The second study conclusion concerned the interrelations between normative beliefs and substance initiation. Strong positive correlations were found between initial levels of peer and adult normative beliefs. Because the findings in the current study were correlational, further research is necessary to determine the causal direction of the influence. Although the association could reflect causal connections between beliefs about peer normative behavior and adult normative behavior, it also might reflect the influence of common predictors (Sher & Wood, 1997). The issue that requires further exploration concerns the possibility that adolescents who overestimate peer and adult normative behavior do so because a large number of peers and adults in their lives are engaged in substance behaviors. That is, differences in beliefs about behaviors within peer groups, families, and communities might foster the impression among adolescents that the majority of people engage in similar behaviors. Consistent with findings from previous research, positive associations were found between both peer and adult normative beliefs and initiation (i.e., intercepts), as well as between the growth-trajectories of adult normative beliefs and substance inflation. Contrary to findings from previous research, the current study found a positive association between the growth-trajectory of adult normative beliefs and initiation. However, differences in study methodologies might account for such variation. To our knowledge, this is the first study to examine the growth-trajectory of normative beliefs (i.e., both peer and adult) and substance initiation.

Implications for Substance Abuse Prevention

Findings from the current study suggest several important implications for prevention programming. First, the current study found evidence that beliefs concerning peer normative behavior had a long-term influence on substance initiation. Adolescents who believed a large proportion of their peers engaged in substance behavior were more likely to initiate substance use during early and middle adolescence. The study findings also provided evidence that normative education components incorporated into prevention education programming can result in lower levels of initiation of various substances. That is to say, adolescents who report perceptions of lower substance behavior among both peers and adults can effectively apply these beliefs to counterbalance a range of internal and external pressures to initiate substance use. Based on the current analyses, these internal pressures include a wide range of perceptions concerning substance behaviors among peers and adults, including the use of tobacco, marijuana, and alcohol. A variety of potentially destructive behaviors that can interfere with health and well being during adulthood can be forecasted by substance initiation during adolescence. Given the mounting evidence that has accumulated from evaluation studies of several preventive interventions, normative education (i.e., modifying beliefs concerning peer and adult substance use) appears to be a worthwhile programming component. Indeed, the current findings suggest that we should continue providing adolescents with appropriate and relevant information concerning normative beliefs, particularly those related to peer behavior.

Second, the strength of the association between normative beliefs related to both peer and adult substance behavior presents a significant rationale for multifaceted prevention programming to reduce or delay substance initiation. This is especially important given the research findings pointing to the diverse array of risk factors that promote substance initiation (Hawkins, Catalano, & Miller, 1992). One implication of the study findings is that prevention programs that address normative education might have important deterrent effects, not only in terms of the initiation of tobacco and alcohol, but also in terms of marijuana use. Public health efforts that reduce, delay, or prevent early experimentation with alcohol or tobacco may significantly reduce regular marijuana use in later adolescence and adulthood (Kandel, Yamaguchi, & Chen, 1992).

Finally, study findings indicated that academic underachievement was a precursor of the growth-trajectories of beliefs concerning peer normative behavior and initiation. Those findings suggest that academic achievement was a precursor of both constructs and makes a case for increased efforts to improve academic skills in an attempt to prevent and/or delay substance initiation.

Limitations

In relating conclusions from the current investigation to previous research results, differences in study samples and methodologies (e.g., sample type and size), plus varying assessment methodologies, must be kept in mind. Generalizations of the study findings to samples in different stages of development and with different demographic characteristics must be made with caution. A further limitation concerns the self-report nature of the study. Findings were based on adolescent self-reported behaviors and beliefs. However, self-reported measures related to substance use, normative beliefs, and academic achievement are common; the reliability and validity of similar assessments of adolescent behaviors and beliefs have been demonstrated (Botvin, Baker, Dusenbury, Botvin, & Diaz, 1995; Botvin et al., 2001). The study sample also was limited due to attrition; adolescent prone to be substance users were underrepresented in the study sample. In addition, the present study combined three different substances to represent an overall measure of the initiation process. As noted earlier, differences might exist among the various substances and previous research findings typically have linked normative beliefs to specific substances (i.e., tobacco and alcohol use). Given the strong association between the use of tobacco, alcohol, and marijuana among adolescents, however, the present study focused on an examination of the hypothesized pathways for the general problem of adolescent substance initiation (Han, McGue, & Iacono, 1999; McNeal & Hansen, 1999; Sobeck, Abbey, Agius, Clinton, & Harrison, 2000).

Although the sample was predominately Caucasian, it is representative of the Midwestern region in which the study was conducted. However, that does limit generalizability of study findings to other groups, such as urban minority youth. Despite the noted limitations, the study provided evidence to suggest the influence of academic achievement on initial levels and the growth-trajectories of normative beliefs (i.e., peer) and substance initiation, as well as the influence of adult normative beliefs on the initiation process. In conclusion, the current investigation stimulated questions for further research to investigate the pathways between the study constructs, including other proximal variables (e.g., parental substance use behavior) that might be associated with young adolescent sub.... stance Initiation.

APPENDIX A

Latent Growth Curve Modeling

The latent growth curve estimation began by describing change over time for each individual in the study. Conceptually, this was done by fitting a regression line (i.e., a growth curve) that linked the outcome variable (i.e., measurement of normative beliefs and substance initiation) to time for each individual. The individual regression prediction equations were then summarized to obtain an average intercept and average slope for all individuals, each with a variance. The goal was then to explain why some individuals had higher initial levels (i.e., intercepts) than others and why some individuals had steeper slopes (i.e., rates of change) (Wickrama, Lorenz, & Conger, 1997).

The process to estimate growth curves for each individual p(p = 1, 2 ... n) was done by writing and estimating equations in which the observed scores at each time point ([y.sub.t1], [y.sub.t2] and [y.sub.t3]) were linked to time ([t.sub.1], [t.sub.2] and [t.sub.3]) using the two parameters of an intercept (i.e., level), [[pi].sub.0] and a slope (i.e., rate of change), [[pi].sub.1] as symbolized in the following' equations:

[y.sub.1p] = 1 * [[pi].sub.0p] + [t.sub.1] * [[pi].sub.1p] + [[epsilon].sub.11]

[y.sub.2p] = 1 * [[pi].sub.0p] + [t.sub.2] * [[pi].sub.1p] + [[epsilon].sub.22]

[y.sub.3p] = 1 * [[pi].sub.0p] + [t.sub.3] * [[pi].sub.1p] + [[epsilon].sub.33]

(In the above equations, [[epsilon].sub.11], [[epsilon].sub.22] and [[epsilon].sub.33] were the residuals at each time point.)

In the current study, time-varying y was either normative beliefs or substance initiation. For example, for each individual, equations linked the three measurements of normative beliefs or substance initiation to the time points [t.sub.1], [t.sub.2] and [t.sub.3] using an intercept, [[pi],sub.0], and a slope, [[pi].sub.1].

Three equations corresponded to the structural equations with latent and observed measurements. In a structural equation framework, [[pi].sub.0] and [[pi].sub.1] were considered as two latent variables measured by three observed indicators [y.sub.t1], [y.sub.t2] and [y.sub.t3]. The coefficients of [[pi].sub.0] in the three equations (always 1) were equal to the factor loadings of the three indicators (i.e., measurements) for 'level' latent variables. The coefficients of [[pi].sub.1] for the three equations ([y.sub.t1], [y.sub.t2] and [y.sub.t3]) were equal to factor loadings of the three measurements for the 'rate of change' latent variables. If [t.sub.1], [t.sub.2] and [t.sub.3] were 0, 1, and 2 respectively, change would be a linear increase across the three time points and the level (i.e., intercept) would be defined by the measurement of y at the first time point, the point at which the factor loading would be equal to 0.

Intercepts and growth-trajectories (i.e., level and rate of change) were expected to differ from individual to individual (for a sample of individuals, there will be as many as n different rates of change over time and n different levels of behaviors). Although each individual intercept and growth-trajectory varied in level and rate of change, they were aggregated so that, for the entire sample, there was an average (i.e., mean) intercept with a variance and an average (i.e., mean) growth-trajectory, also with a variance. The mean and variance of the individual-level intercept parameters identified the overall average of the specific individual levels and variability of levels (i.e., dispersion) across individuals. The mean for the growth-trajectory described the overall average change in behaviors of individuals over time, while the population variance for the change parameter reflected differences in the rates of change across individuals. Significant variance in the growth parameters demonstrated the heterogeneity of the individual-specific growth-trajectories. This is the variability predictors of change should explain in the analysis of change. Also, predictors can explain variability of the individual intercepts.

Thus, in addition to the simple description of change, growth curve modeling permitted an explanation of the systematic inter-individual differences in both the level of behavior and the rate of change. When the change parameter (i.e., rate of change in normative beliefs or substance initiation) covaried significantly with a predictor variable (e.g., academic achievement), differences in the rate of behavior change across individuals were said to be systematic (Willet & Sayer, 1994). Thus, the behavior growth parameters of individuals ([[pi].sub.0p] or [[pi].sub.1p]) is represented by the following equations using academic achievement as a predictor variable:

1. For level of [p.sup.th] individual:

[[pi].sub.0p] = [[alpha].sub.[pi]op]+ [[beta].sub.1] (academic achievement) + [[zeta].sub.1]

2. For rate of change of [p.sub.th] individual:

[[pi].sub.1p] = [[alpha].sub.[pi]1p]+ [[beta].sub.2] (academic achievement) + [[zeta].sub.2]

In the above equations, [[alpha].sub.[pi]op] and [[alpha].sub.[pi]1p] represent average levels of the prediction equations [[beta].sub.1] and [[beta].sub.2] are path coefficients that link the change parameters of substance initiation and predictor variables, and [[zeta].sub.1] and [[zeta].sub.2] are disturbances. More importantly, if the predictor variable is time-varying (i.e., normative beliefs), growth-trajectories can also be estimated for the predictor variable and the slope parameter of one time-varying variable (i.e., substance initiation) can be predicted by the intercept (i.e., level) parameter of the other. In our study, the equation predicting the slope parameter of substance initiation for [p.sub.th] person is symbolized as follows:

[(Rate of change in substance initiation).sub.p] = [[alpha].sub.[pi].sub.1p] + [[beta].sub.2] [(level of peer normative beliefs).sub.p] + [[beta].sub.3] [(level of peer normative beliefs).sub.p] + [zeta]

Table 1. Means, Standard Deviations, and Correlations for the Model
Variables

                                 Beliefs about Peer
                                   Normative Use

Variable           T1        T2        T3        T4        T5

Beliefs about
Peer
Nominative Use
  Time 1         1.00       .41 **    .30 **    .22 **    .12 *
  Time 2                   1.00       .42 **    .39 **    .27 **
  Time 3                             1.00       .45 **    .37 **
  Time 4                                       1.00       .48 **
  Time 5                                                 1.00

Beliefs about
Adult
Normative Use
  Time 1
  Time 2
  Time 3
  Time 4
  Time 5

Substance Use
Initiation
  Time 1
  Time 2
  Time 3
  Time 4
  Time 5

Academic
Achievement

Means            1.88      2.00      2.15      2.26      2.41

Standard
Deviation         .67       .63       .62       .57       .51

                                Beliefs about Adult
                                   Normative Use

Variable           T1        T2        T3        T4        T5

Beliefs about
Peer
Nominative Use
  Time 1          .42 **    .19 **    .10       .17 **    .05
  Time 2          .32 **    .49 **    .23 **    .28 **    .14 **
  Time 3          .26 **    .31 **    .50 **    .37 **    .23 **
  Time 4          .23 **    .40 **    .30 **    .51 **    .32 **
  Time 5          .16 *     .44 **    .37 **    .40 **    .65 **

Beliefs about
Adult
Normative Use
  Time 1         1.00       .48 **    .37 **    .31 **    .25 **
  Time 2                   1.00       .53 **    .42 **    .44 *
  Time 3                             1.00       .36 **    .37 **
  Time 4                                       1.00       .45 **
  Time 5                                                 1.00

Substance Use
Initiation
  Time 1
  Time 2
  Time 3
  Time 4
  Time 5

Academic
Achievement

Means            2.75      2.77      2.79      2.90      2.73

Standard
Deviation         .68       .63       .58       .58       .50

                                    Substance
                                    Initiation

Variable           T1        T2        T3        T4        T5

Beliefs about
Peer
Nominative Use
  Time 1          .08       .05       .02       .04      -.02
  Time 2          .08       .12 *     .12 *     .09       .07
  Time 3          .09       .15 **    .24 **    .24 *     .20 *
  Time 4          .20 **    .14 *     .21 **    .26 **    .20 **
  Time 5          .11 *     .15       .15       .16 **    .22 **

Beliefs about
Adult
Normative Use
  Time 1          .06       .09       .10       .11       .09
  Time 2          .05       .05       .11 *     .07       .11
  Time 3          .09       .15 **    .17 **    .13 *     .15 **
  Time 4          .08       .05       .14 *     .15 **    .15 *
  Time 5          .02       .05       .06       .05       .16 **

Substance Use
Initiation
  Time 1         1.00       .73 **    .62 **    .54 **    .53 **
  Time 2                   1.00       .80 **    .71 **    .68 **
  Time 3                             1.00       .88 **    .79 **
  Time 4                                       1.00       .87 **
  Time 5                                                 1.00

Academic
Achievement

Means             .56       .85      1.15      1.29      1.50

Standard
Deviation         .69       .81       .89       .89       .92

                  Academic
                 Achievement

Variable

Beliefs about
Peer
Nominative Use
  Time 1            .14 **
  Time 2            .04
  Time 3            .03
  Time 4            .04
  Time 5           -.09

Beliefs about
Adult
Normative Use
  Time 1            .00
  Time 2           -.11 *
  Time 3           -.18 *
  Time 4           -.03
  Time 5           -.15 **

Substance Use
Initiation
  Time 1           -.20 **
  Time 2           -.25 **
  Time 3           -.31 **
  Time 4           -.22 **
  Time 5           -.25 **

Academic           1.00
Achievement

Means              4.16

Standard
Deviation           .80

* p < .05. ** p < .01.

Table 2. Growth Curves Estimates for Beliefs about Peer Normative
Use, Beliefs about Adult Normative Use, and Substance Initiation

                                             Intercept (a,b)

                                             M          Variance

Beliefs about Peer Normative Use            1.90           .19
                                          (56.87)        (6.95)

Beliefs about Adult Normative Use           2.75           .25
                                          (77.06)        (8.23)

Substance Initiation                         .54           .46
                                          (13.12)       (11.64)

                                                Slope (b)

                                             M          Variance

Beliefs about Peer Normative Use            .07            .01
                                         (12.27)         (4.52)

Beliefs about Adult Normative Use          -.01            .01
                                         (-1.33)         (3.94)

Substance Initiation                        .27 (c)        .07
                                         (11.16)         (6.25)

                                        [chi square]      CFI

Beliefs about Peer Normative Use           12.47          .99
                                          (10 df)

Beliefs about Adult Normative Use          19.45          .97
                                          (10 df)

Substance Initiation                       92.89          .94
                                         (8 df) (d)

Note: N = 272

(a) For the intercepts [[lambda].sub.11] = [[lambda].sub.21] =
[[lambda].sub.31] = [[lambda].sub.41] = [[lambda].sub.51] 1.0
for all models.

(b) T-values are in parentheses.

(c) [[lambda].sub.32], [[lambda].sub.42], and [[lambda].sub.52],
were freely estimated, rather than constrained to linearity, as
the Substance Initiation slope was not linear.

(d) Residual variance for substance initiation was set to .01
to rectify a Heywood case (Bollen, 1989).

Footnote 

(1) A total of 1,673 students completed the baseline pretest assessment. However, students who subsequently moved from a school in one condition to a school in another condition were excluded from the sample summary, because these students could not be assigned to one of the experimental conditions.

References

Bailey, S. & Hubbard, R. L. (1990). Developmental variation in the context of marijuana initiation among adolescents. Journal of Health and Social Behavior, 31, 58-70.

Bauman, K., Botvin, G., Botvin, E., & Baker, E. (1992). Normative expectations and the behavior of significant others: An integration of traditions in research on adolescents' cigarette smoking. Psychological Reporta, 71, 568-570.

Bauman, K. E., Koch, G. G., & Bryan, E. S. (1988). Validity of self-reports of adolescent cigarette smoking. International Journal of the Addictions, 17, 1131-1136.

Bollen, K. (1989). Structural equations with latent variables. New York: Wiley.

Botvin, G. J. (1996). Life Skills Training: Promoting health and personal development. Princeton, NJ: Princeton Health Press.

Botvin, G. (2000). Life Skills Training. Princeton, NJ: Princeton Health Press.

Botvin, G., Baker, E., Dusenbury, L., Botvin, E., & Diaz, T. (1995). Long-term follow-up results of a randomized drug abuse prevention trial in a White middle-class population. Journal of the American Medical Association, 273, 1106-1112.

Botvin, G., Botvin, E., Baker, E., Dusenbury, L., & Goldberg, C. (1992). The false consensus effect: Predicting adolescents' tobacco use from normative expectations. Psychological Report, 70, 171-178.

Botvin, E. M., Botvin, G. J., Renick, N., Filazzola, A. D., & Allegrante, J. P. (1984). Adolescents' self-reports of tobacco, alcohol, and marijuana use: Examining the comparability of video tape, cartoon, and verbal bogus pipeline procedures. Psychological Report, 55, 379-386.

Botvin, G. J., Griffin, K., Diaz, T., & Ifill-Williams, M. (2001). Drug abuse prevention among minority adolescents: Posttest and one-year follow-up of a school-based preventive intervention. Prevention Science, 2(1), 1-13.

Bourgois, P. (1995). In search of respect: Selling, crack in El Barrio. New York: Cambridge University Press.

Boyle, M. H., Offord, D. R., Racine, Y., Sanford, M., Szatmari, P., & Fleming, J. F. (1993). Predicting substance use in early adolescence based on parent and teacher assessment of childhood psychiatric disorder. Results from the Ontario Child Health Study follow-up. Journal of Child Psychology and Psychiatry, 38, 535, 544.

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., Flor, D. L., Hollett-Wright, N., McCoy J. K., & Donovan, J. (1999). Parent-child relationship, child temperament profiles, and children's alcohol use norms. Journal of Studies on Alcohol, (Suppl. 13), 45-51.

Brody, G. H., Ge, X., Katz, J., & Arias, I. (2000). A longitudinal analysis of internalization of parental alcohol-use norms and adolescent alcohol use. Applied Developmental Science, 4(2), 71-79.

Bucholz, K. K. (1990). A review of correlates of alcohol use and alcohol problems in adolescence. In M. Galanter (Ed.), Recent developments in alcoholism: Vol. 8. Combined alcohol and other drug dependence (pp. 111-123). New York: Plenum.

Dryfoos, J. (1990). Adolescents at risk: Prevalence and prevention. New York: Oxford University Press.

Duncan, S. C., Duncan, T. E., Biglan, A., & Ary, D. (1998). Contributions of the social context to the development of adolescent substance use: A multivariate latent growth modeling approach. Drug and Alcohol Dependence, 50, 57-71.

Duncan, T. E., Duncan, S. C., Strycker, L.A., Li, F., & Alpert, A. (1999). An introduction to latent variable growth curve modeling: Concepts, issues, and applications. Mahwah, NJ: Lawrence Erlbaum Associates.

Dusenbury, L., Kerner, J., Baker, E., Botvin, G., James-Ortiz, S., & Zauber, A. (1992). Predictors of smoking prevalence among New York Latino youth. American Journal of Public Health, 82(1), 55-58.

Ennett, S. & Bauman, K. (1991). Mediators in the relationship between parental and peer characteristics and beer drinking in early adolescents. Journal of Applied Social Psychology, 21 (20), 1699-1711.

Evans, R. I., Hansen, W. B., & Mittlemark, M. B. (1977). Increasing the validity of self-reports of smoking behavior in children. Journal of Applied Psychology, 62, 521-523.

Glynn, T. (1981). From family to peer: A review of transitions of influence among drug-using youth. Journal of Youth and Adolescence, 10(5), 363-383.

Guerra, N., Huesmann, L., & Hanish, L. (1995). The role of normative beliefs in children's social behavior. In N. Eisenberg (Ed.), Social development: Review of personality and social psychology. Vol. 15 (pp. 140-158). Thousand Oaks, CA: Sage Publications, Inc.

Han, C., McGue, M. K., & Iacono, W. G. (1999). Lifetime tobacco, alcohol and other substance use in adolescent Minnesota twins: Univariate and multivariate behavioral genetic analyses. Addiction, 94, 981-993.

Hansen, W. B., Graham, J. W., Sobel, J. L., Shelton, D. R., Flay, B. R., & Johnson, C. A. (1987). The consistency of peer and parent influences on tobacco, alcohol, and marijuana use among young adolescents. Journal of Behavioral Medicine, 10, 559-579.

Hawkins, D. (1997). Academic performance and school success: Sources and consequences, in R. Weissberg, T. Gullotta, R. Hampton, B. Ryan, & G. Adams (Eds.), Enhancing children's wellness. Thousand Oaks, CA: Sage Publications.

Hawkins, D., Catalano, R., & Miller, J. (1992). Risk and protective factors for alcohol and other drug problems in adolescence and early adulthood: Implications for substance abuse prevention. Psychological Bulletin, 112, 64-105.

Hindelang, M. (1973). Causes of delinquency: A partial replication and extension. Social Problems, 20, 471-487.

Hundlbey, J. & Mercer, G. (1987). Family and friends as social environments and their relationship to young adolescents' use of alcohol, tobacco, and marijuana. Journal of Clinical Psychology, 44, 125-134.

Institute of Medicine. (1994). Reducing risks for mental disorders: Frontiers for preventive intervention research. Washington, PC: National Academy Press.

Jessor, R. (1976). Predicting time of onset of marijuana use: A developmental study of high school youth. Journal of Consulting and Clinical Psychology, 44, 125-134.

Jessor, R. & Jessor, S. (1975). Adolescent development and the onset of drinking: A longitudinal study. Journal of Studies on Alcohol, 36, 27-51.

Jessor, R., & Jessor, S. L. (1977). Problem behavior and psychosocial development: A longitudinal study of youth. New York: Academic Press.

Johnson, K. A. (1986). Informal control networks and adolescent orientations toward alcohol use. Adolescence, 21(84), 767-784.

Kandel, D. B., Kessler, R. C., & Margulies, R. Z. (1978). Antecedents of adolescent initiation into stages of drug use: A developmental analysis. In D. B. Kandel (Ed.), Longitudinal research on drug use: Empirical findings and methodological issues (pp. 73-99). Washington, PC: Hemisphere.

Kandel, D. B., Kessler, R. C., & Margulies, R. Z. (1985). Antecedents of adolescent initiation into stages of drug use: A developmental analysis. Annual Progress in Child Psychiatry and Child Development, 61, 646-676.

Kandel, D. B. & Lesser, G. (1969). Parental and peer influences on educational plans of adolescents. American Sociological Review, 34(2), 213-223.

Kandel, D. B., Yamaguchi, K., & Chen, K. (1992). Stages of progression in drug involvement from adolescence to adulthood: Further evidence for the gateway theory. Journal of Studies on Alcohol, 53, 447-457.

Krohn, M., Akers, R., Radosevich, M., & Lanza-Kaduce, L. (1982). Norm qualities and adolescent drinking and drug behavior: The effects of norm quality and reference group on using and abusing alcohol and marijuana. Journal of Drug Issues, 12, 343-359.

Leventhal, H., Glynn, K., & Fleming, R. (1987). Is the smoking decision an 'Informed Choice'? Journal of the American Medical Association, 257(24), 3373-3376.

Loeber, R. & Dishion, T. (1983). Early predictors of male delinquency: A review. Psychological Bulletin, 94(1), 68-99.

Loeber, R. & Farrington, D. (1994). Problems and solutions in longitudinal and experimental treatment studies of child psychopathology and delinquency. Journal of Consulting and Clinical Psychology, 62(5), 887-900.

Lowman, C. (1982). Parental dimensions in teenage drinking. Alcohol Health and Research World, 6(4), 58-63.

McNeal, R. B. & Hansen, W. B. (1999). Developmental patterns associated with the onset of drug use: Changes in postulated mediators during adolescence. Journal of Drug Issues, 29, 381-400.

Molgaard, V., Kumpfer, K., & Fleming, B. (1997). The strengthening families program: For parents and youth (pp. 10-14). Ames, IA: Iowa State University Extension.

Murray, D. M., Hannan, P. J., & Zucker D. M. (1989). Analysis issues in school-based health promotion studies. Health Education Quarterly, 16, 315-320.

Murray, D., O'Donnell, C., Schmid, L., & Perry, C. (1987). The validity of smoking self-reports by adolescents: A re-examination of the bogus pipeline procedures. Addictive Behaviors, 12, 7-15.

Muthen, L. & Muthen, B. (1998). Mplus: The comprehensive modeling program for applied researchers user's guide. Muthen & Muthen: Los Angeles.

O'Donnell, J., Hawkins, J. D., & Abbott, R. D. (1995). Predicting serious delinquency and substance use among aggressive boys. Journal of Consulting and Clinical Psychology, 63, 529-537.

Oetting, E. & Beauvais, F. (1990). Adolescent drug use: Findings of national and local surveys. Journal of Consulting and Clinical Psychology, 58, 385-394.

Paulson, M. J., Coombs, R. H., & Richardson, M. A. (1990). School performance, academic aspirations, and drug use among children and adolescents. Journal of Drug Education, 20(4), 289-303.

Petraitis, J., Flay, B., & Miller, T. (1995). Reviewing theories of adolescent substance use: Organizing pieces of the puzzle. Psychological Bulletin, 117, 67-86.

Polk, K. (1975). Schools and the delinquency experience. Criminal Justice and Behavior, 2, 315-338.

Presson, C. C., Chassin, L., Sherman, S. J., Olshavsky, R., Bensenberg, M., & Corty, E. (1983). Predictors of intentions to smoke: Age, sex, race, and regional differences. International Journal of the Addictions, 19, 503-519.

Ried, L. D., Martinson, O. B., & Weaver, L. C. (1987). Factors associated with the drug use of fifth through eighth grade students. Journal of Drug Education, 17(2), 149-161.

Robins, L. (1980). The natural history of drug abuse. Acta Psychiatrica Scandinavia, 62(Suppl. 284), 7-20.

Rogosa, D. (1988). Myths about longitudinal research. In K. W.. Schaie, R. T. Campbell, W. Meredith, & S. C. Rawlings (Eds.), Methodological issues in aging research, Vol. 4 (pp. 63-105). New York: Springer.

Rogosa, D. & Willett, J. B. (1985). Understanding correlates of change by modeling individual differences in growth. Psychometrika, 50, 203-228.

Ross, L., Greene, D., & House, P. (1977). The "False Consensus Effect": An egocentric bias in social perception and attribution processes. Journal of Experimental Social Psychology, 13, 279-301.

Shedler, J. & Block. J. (1990). Adolescent drug use and psychological health. American Psychologist, 45, 612-630.

Sher, K. J. & Wood, R K. (1997). Methodological issues in conducting prospective research on alcohol-related behavior: A report from the field. In K. J. Bryant & M. Windle (Eds.), The science of prevention: Methodological advances from alcohol and substance abuse research (pp. 3-41). Washington, DC: American Psychological Association.

Smith, G. M. & Fogg, C. R (1978). Psychological predictors of early use, late use, and nonuse of marijuana among teenage students. In D. B Kandel (Ed.), Longitudinal research on drug use: Empirical findings and methodological issues (pp. 101-113). Washington, DC: Hemisphere Publishing Corporation.

Smith, G. M. & Fogg, C. R (1979). Psychological antecedents of teenage drug use. Research in Community and Mental Health, 1, 87-102.

Sobeck, J., Abbey, A., Agius, E., Clinton, M., & Harrison, K. (2000). Predicting early adolescent substance use: Do risk factors differ depending on age of onset? Journal of Substance Abuse, 11, 89-102.

Stoolmiller, M. & Bank, L. (1995). Autoregressive effects in structural equation models: We see some problems. In J. M. Gottman (Ed.), The analysis of change (pp. 261-276). Mahwah, NJ: Lawrence Erlbaum Associates.

Tobler, N., Roona, M., Ochshorn, R, Marshall, D., Streke, A., & Stackpole, K. (2000). School-based adolescent drug prevention programs: 1998 Meta-analysis. Journal of Primary Prevention, 20(4), 275-336.

Wickrama, K. A. S., Lorenz, E O., & Conger, R. D. (1997). Parental support and adolescent physical health status: A latent growth-curve analysis. Journal of Health and Social Behavior, 38(2), 149-163.

Willett, J B., & Sayer, A. G. (1994). Using covariance structure analysis to detect correlates and predictors of change. Psychological Bulletin, 116, 363-381.

Willett, J B. & Sayer, A. G. (1996). Cross-domain analyses of change over time: Combining growth modeling and covariance structure analysis. In G. A. Marcoulides & R. E. Schumacker (Eds.), Advanced structural equation modeling: Issues and techniques (pp. 125-157). Hillsdale, NJ: Lawrence Erlbaum Associates.

Author Note

Catherine J. Lillehoj, Linda Trudeau, Richard Spoth, Partnerships in Prevention Science Institute, Iowa State University.

Work on this paper was supported by research grants OA 070 29-01 and DA 10815-01 from the National Institute on Drug Abuse.

Correspondence concerning this article should be addressed to Catherine J. Lillehoj at the Partnerships in Prevention Science Institute, Iowa State University Research Park, Building 2, Suite 500, 2625 North Loop Drive, Iowa State University, Ames, Iowa 50010-8296. E-mail: [e-mail]

Catherine J. Lillehoj, Linda Trudeau, Richard Spoth

Partnerships in Prevention Science Institute

Iowa State University