The stress generation hypothesis was tested in two different longitudinal studies examining relations between weekly depression symptom ratings and stress levels in adolescents and emerging adults at varied risk for depression. The participants in Study 1 included 240 adolescents who differed with regard to their mothers' history of depressive disorders. Youth were assessed annually across 6 years (Grades 6–12). Consistent with the depression autonomy model, higher numbers of prior major depressive episodes (MDEs) were associated with weaker stress generation effects, such that higher levels of depressive symptoms predicted increases in levels of dependent stressors for adolescents with two or more prior MDEs, but depressive symptoms were not significantly related to dependent stress levels for youth with three or more prior MDEs. In Study 2, the participants were 32 remitted-depressed and 36 never-depressed young adults who completed a psychosocial stress task to determine cortisol reactivity and were reassessed for depression and stress approximately 8 months later. Stress generation effects were moderated by cortisol responses to a laboratory psychosocial stressor, such that individuals with higher cortisol responses exhibited a pattern consistent with the depression autonomy model, whereas individuals with lower cortisol responses showed a pattern more consistent with the depression sensitization model. Finally, comparing across the two samples, stress generation effects were weaker for older participants and for those with more prior MDEs. The complex, multifactorial relation between stress and depression is discussed.