The National Heart, Lung, and Blood Institute (NHLBI) Twin Study provided longitudinal data on a cohort of 514 pairs of adult male twin pairs who were examined at approximate ages of 48, 57, and 63 years. Because the sample was selected from military veterans, height and weight data were also available from their induction physical examinations when they were approximately 20 years of age. From the total NHLBI Twin Study cohort, 124 monozygotic and 119 dizygotic male twin pairs had complete data available for both members of the pair at induction and three examinations spanning 43 years of adult life. Using these data, the contributions of genetics and shared and non-shared environmental factors to BMI over the 43 year period were estimated by model fitting procedures. Model fitting included both a factor decomposition of these effects as well as a developmental path model. Results from the decomposition procedure indicate significant genetic effects at each examination cycle. Fitting a developmental path model, two independent genetic contributions to the variability of BMI were found: one at, or prior to, the induction examination about age 20, and a second between ages 20 and 48. Significant non-shared environmental contributions at each examination were also indicated, but shared environmental effects were not significant. We conclude that cumulative genetic effects explain most of the tracking in obesity over time; non-shared environmental effects, although significant at each age, are relatively short-lived and make only a minor contribution to tracking.