Alcohol use is a complex and dynamic phenomenon, determined by a multiplicity of factors that possibly interact in a nonlinear manner. There has been very limited research so far exploring the nonlinearities in the etiology of alcohol use. The purpose of the proposed study is to design a nonlinear predictive model of alcohol use and abuse in the general population with the help of implicit and explicit alcohol expectancy measures, as well as self-reports of family history and personality style. The complex and changing relationship of negative expectancies and alcohol use will also be examined with respect to participants' drinking histories and positive alcohol expectancies. The role of implicit expectancies in predicting drinking behavior, over and above explicit questionnaires, in a large (N = 400) community population will also be studied. Participants will complete a reaction time measure of their memory associations to alcohol, as well as other self-report measures of drinking, alcohol expectancies, personality, family history, etc., and a drinking history interview. Long-term data on drinking (3-month follow up) will also be collected. Data analysis will include artificial neural network modeling and multiple regression analyses. ? ?