Objectives: We performed a prospective study to determine whether weight gain predicts future ultrasonographically detected fatty liver (USFL) in a lean adult population. Methods: Among 15 347 Korean male workers, aged 30-59 years, who participated in a health check-up programme in 2002, a USFL-free cohort of 4246 nondiabetic men was followed until September 2007. Alcohol consumption was assessed by a questionnaire. Weight change for each subject was calculated as the difference between baseline and subsequent measurements. Biochemical tests for liver and metabolic function were done. The primary outcome was ultrasound-diagnosed fatty liver. A standard Cox proportional hazards model and time-dependent Cox model were performed. Results: During 16 829.7 person-years of follow-up, 622 participants developed USFL. After adjusting for age, the period from visit 1 to visit 2, BMI, HDL-C, triglyceride, uric acid, alanine aminotransferase, and HOMA-IR, the risk for USFL increased with increasing quartiles of weight change (p for trend <0.001). This association remained significant when weight change and covariates, except age and the period from visit 1 to visit 2, were modelled as time-dependent variables. Subjects in the fourth quartile (weight gain ≥2.3 kg) were at significantly elevated risk for USFL (adjusted hazard ratio (aHR), 1.26; 95% CI, 1.01 to 1.58). These associations did not change, even in normal weight men with a baseline BMI between 18.5 and 22.9 kg/m2 (n=2186). Conclusion: Weight gain per se appears to increase the risk for developing USFL. Thus, avoiding weight gain, even among lean adult individuals, can be helpful in preventing this disease.