Over the past decade, the numbers of longitudinal studies of adolescent alcohol use have grown dramatically. Accompanying this growth in longitudinal research has been an explosion in statistical methods for analysis of longitudinal data. The primary purpose of this study is to develop and apply marginal, latent variable, and transition models to an investigation of alcohol behavior using data from a longitudinal study of 584 rural adolescents. Data will be analyzed from three time periods: 8th, 9th, and 10th grades. In addition, analyses will be conducted to explore gender and race/ethnic specific patterns of alcohol behavior and associated problems. Longitudinal studies of adolescents rely on repeated measures of both alcohol outcomes and predictors (i.e., peer drinking) to assess underlying constructs for analysis. It is not clear whether standard questions on alcohol use, attitudes about alcohol, or problems associated with drinking have the same meaning for adolescents of varying developmental stages, gender or racial/ethnic groups and whether information about drinking is remembered and retrieved in a similar manner. A second objective of this study is to explore how adolescents at varying ages, who are members of diverging racial/ethnic and gender groups, understand and process questions about their alcohol use and the predictors of use. This information will also be used to refine standard questions about alcohol use and its predictors. Two sets of cognitive interviews will be conducted with 72 adolescents from rural site in Salisbury, Maryland stratified by three categories of adolescence (12 to 14 years, 15-17 years, and 18-20 years), gender, and race/ethnicity (African American, European American). Trained interviewers will conduct individual interviews using questions taken from the original longitudinal study and from other studies of alcohol. The information about how questions are understood and processed will be categorized and subgroups compared. The resulting data will be used to help inform the findings from the longitudinal modeling, and to suggest alternative models. Questions will be revised and evaluated in the second set of interviews.