In a six-wave longitudinal study with two cohorts (660 adolescents and 630 young adults), this study investigated the longitudinal stability of the Beck Depression Inventory II (BDI-II) using the Trait-State-Occasion (TSO) model. The results revealed that the full TSO model was the best fitting representation of the depression measured by the BDI-II for both cohorts. It was found that the trait components explained more than 50% of the state variance of the BDI-II scores in both cohorts, and the occasion-specific factor explained about 7% to 12% of variances. The occasion-specific variance was more stable (occasion stability) and somewhat higher in the younger cohort. Implications regarding the longitudinal stability of the BDI-II measure are discussed.