The development of a model to predict drinking behavior from attitudes in university students.

The major purposes of the study were: a) to examine the relationship between attitudes and self-reported levels of drinking; and b) to develop a stable prediction equation that included attitudes as a predictor of drinking behavior. A Likert-type survey was developed to measure college student's attitudes toward alcohol. The survey was administered to a sample of college students (n=1049). The coefficient alpha reliability estimate was found to be .91. Stepwise multiple regression procedures were used to ascertain the relationships between attitudes and other psychosocial constructs of drinking. The dependent variable was an alcohol consumption index. The analysis revealed that the seven variable model was the most parsimonious (R (2) =.46), and attitudes toward drinking was the strongest predictor of self-reported drinking behavior. A double cross-validation of the regression model indicated that the model was very stable, and therefore could be generalized to similar samples. The substantive findings related to students' self-reported drinking practices are also reported.