Stability of recall of DSM-III diagnoses was assessed at two interviews 2 years apart in a sample of 150 offspring, aged 6 to 23 years, at high and low risk for major depression. Stability of recall was good for major depression with the use of DSM-III criteria and fair for major depression with the use of "strict" criteria (based on 4 weeks' duration of illness and an impairment in a major social role). Stability of recall was good for substance abuse and conduct disorder. Stability of recall was generally poor for anxiety disorder, regardless of subtype. For all major disorders except anxiety disorder, the difference in reported age at onset between the two interviews was small (less than 1 year) and not statistically significant. The most important correlates of stability of reports of major depression were previous psychiatric treatment and dysthymia and poor social functioning at the initial interview. This is the first study to evaluate long-term recall of DSM-III lifetime diagnoses in a nonreferred sample of children, adolescents, and young adults.