Objectives : This study was performed using a longitudinal approach to explore the predictors for current smoking among male high school students. Methods : Baseline data was collected in May 2004 through a self-administrated questionnaire completed by 607 male students in a technical high school in Daegu city, Korea. Subsequently, their smoking behaviors were followed one year after. Among the 544 followed participants, data for 439 non-smokers in the first year was used in longitudinal analysis. Current smokers were defined as those respondents who had smoked one or more cigarettes within the 30 days preceding the survey. Several potential predictors for smoking were investigated including smoking history(never, experimental, former smoker), sociodemographic factors, environmental factors, attitudes toward smoking, and behavioral factors. Logistic regression was used to predict smoking with SPSS ver. 12.0. Results : According to multiple logistic regression analysis, those students who were more likely to smoke after one year were former smokers(OR: 2.12, 95% CI=1.01-4.44), current drinkers(OR: 2.55, 95% CI=1.33-4.89), who had four or five smokers among five best friends(OR: 3.43. 95% CI=1.14-10.30). In addition, those who had smokers among family members besides parents or siblings(OR: 1.66, 95% CI=0.92-2.98), exhibited a high level of subjective stress(OR: 1.77, 95% CI=0.96-3.26), or had a very good relationship with friends(OR: 1.93, 95% CI=0.99-3.75) were also more likely to smoke albeit with marginal statistical significance(p<0.1). Conclusions : A smoking prevention program aimed at high school students may be more effective with due consideration of the predictors highlighted in this study. However, further studies with larger sample size and various target populations are necessary to find potential predictors not found in this study but suggested in other longitudinal studies.