The authors assessed the limits of reliable history-taking in depressed elderly patients (N = 20) with some cognitive impairment. Each subject and an informant was interviewed with structured instruments by two trained raters. An expert panel formed consensus judgments after reviewing information reported by the patients, the informants, and each of the clinical raters. Intraclass correlation between the two raters was 0.99 for the duration of depressive episodes and 0.88 for age at onset. The raters agreed on the duration of major depressive episodes in 85% of cases and on age at onset in 80% of cases. The duration of previous depressive episodes and age at depression onset cannot always be determined reliably even when informants and structured interviews are used. Greater difficulties may be encountered in patients with minor depression or chronic intermittent depression and early-onset depression. Clinicians should obtain history from as many reliable sources as possible and critically evaluate this information while considering the entire clinical picture. The aggregate kappa statistic can provide a clinically meaningful way of assessing interrater reliability of psychopathological constructs for which several definitions are used.