ABSTRACT
In 2 longitudinal studies of negative life events and depressive symptoms in adolescents (N = 708) and in children (N = 508), latent trait-state-error structural equation models tested both the stress generation hypothesis and the stress exposure hypothesis. Results strongly suggested that self-reports of depressive symptoms reflect the influence of a perfectly stable trait factor as well as a less stable state factor. Support emerged for both the stress generation model and the stress exposure model. When the state depression factor was modeled as predicting stress, support for the stress generation model appeared to increase with age. When the trait depression factor was modeled as the predictor of stress, support for the stress generation model did not vary with the child's age. In both models, support for the stress exposure remained relatively constant across age.