The blind test-retest reliability of lifetime prevalence and age of onset of psychiatric diagnoses, based on the SADS-L interview and RDC over a three-to-five year period, was examined in 143 probands and their relatives. Reliability of lifetime prevalence of major depression was excellent; reliability of antisocial personality, panic disorder, drug abuse, GAD, depressive personality, and alcoholism was good; reliability of obsessive-compulsive disorder and phobia was acceptable but lower. The reliability of hyperthymia or cyclothymia was not acceptable. Reliability for major depression did not vary substantially by age or sex of the informant, but recall of major depression was significantly higher in the probands than in their relatives. The test-retest reliability for the age of onset of major depression and panic disorder was excellent, and for phobia, GAD and alcoholism, was acceptable. Both probands and relatives recalled the age of onset of their depression fairly accurately. However, there was a reduction in agreement over time. Recall after 3-4 yr was better than 5-6 yr. There was a tendency for older respondents to systematically increase the age of onset of their depression across the two interviews, although the increase was only a few years. Recall of age of onset did not differ significantly by sex of respondent or whether the respondent was a proband or relative. These findings are discussed in light of several available studies of reliability of lifetime prevalence of psychiatric diagnoses.