Insomnia in elderly people has traditionally been regarded as inevitable and trivial. A longitudinal study was undertaken to examine the prevalence of sleep disturbance among elderly people in an inner London community and its association with demographic variables, depression, dementia and disability. Those aged 65 years and over living at home were interviewed using a validated and reliable semi-structured interview schedule. A total of 705 people were interviewed in 1987-88 and 524 were re-interviewed in 1990. Subjective sleep disturbance was found to be common (33% and 43%, respectively). Sleep disturbance was associated with being a woman, being unmarried, living alone, disability, and current and future depression, but not with dementia or older age. The best predictor of future depression in elderly people who were not depressed was current sleep disturbance. In the presence of current sleep disturbance, the traditional predictors of depression--being a woman, having a disability, being unmarried, living alone and being older--did not contribute further. This study has shown that sleep disorder is associated with pathology. Insomnia in elderly people requires assessment and this must be accompanied by the treatment of underlying disorders and monitoring of future health.