Background. Assessment of lifetime major depression is usually made from a single interview. Most previous studies have investigated reliability. Comparison of recall of key symptoms and longitudinal reports shows the accuracy of recall, not just reliability.

Method. At age 25, 1003 members of the Christchurch Health and Development Study cohort were asked to recall key symptoms of depression (sadness, loss of interest) up to age 21. This recall was compared with longitudinal reports at ages 15, 16, 18 and 21 years. Diagnosis was by DSM-III-R and DSM-IV criteria.

Results. Only 4% of those without previous reports recalled key symptoms. Of those with a diagnosis of depression up to age 21, 44% recalled a key symptom. Measures of severity of an episode (number of symptoms, impairment, duration, suicidality) and chronicity (years with a diagnosis, years with suicidal ideation) all strongly predicted recall. Current key symptoms increased recall, even after taking account of severity and chronicity. Being female and receiving treatment also predicted recall, although odds ratios were reduced to 1·6–1·7 when all other predictors were included. Comparison of risk factors for key symptoms showed similar results from longitudinal reports and recall. Sexual abuse, neuroticism, lack of parental attachment, gender, physical abuse and maternal depression were major risk factors in both sets of analyses.

Conclusions. Forgetting of prior episodes of depression was common. Severity, chronicity, current depression, gender and treatment predicted recall. Lifetime prevalence based on recall will be markedly underestimated but the identification of major risk factors may be relatively little impaired.