OBJECTIVE: To study the determinants of overweight tracking from childhood to adolescence of Hat Yai schoolchildren. DESIGN: A longitudinal study. SETTING: Primary and secondary schools of Hat Yai municipality, southern Thailand. SUBJECTS: 2252 schoolchildren recruited in 1992 and follow-up for 5 y. MEASUREMENTS: Child's annual body mass index (BMI, kgam 2 ) from 1992 to 1997; parental BMIs, parental income, and family history of diseases by a questionnaire completed by parents in 1992. RESULTS: Prevalence of overweight of males using the 85th percentile of the U S First National Health and Nutritional Examination Survey reference for age and sex as a cut-off point increased from 12.4% in 1992 to 21% in 1997, whereas that of females went down from 15.2 to 12.6. At the end of the ®fth year, 11.8% of children remained overweight while 4.5% became overweight. Comparing to the non-overweights, the risk for becoming an overweight adolescent of an overweight boy was 8.2 (95% con®dence interval (CI) 6, 11.2) whereas that of the overweight girls was 20 (95% CI 12.4, 32.3). The generalized estimating equations model predicted an increase in child BMI associated with having a father or a mother with high BMI, a family history of obesity, a monthly income greater than 5000 baht, and a lower level of exercise than their peers. Secular increase in BMI was also observed. CONCLUSION: Predictors of overweight tracking found in this study would be useful to select children at risk for preventive intervention.