The prevalence rate of subjective memory impairment (SMI) and its value as a predictor of future depression or dementia was studied in a community sample of elderly residents in one electoral ward using the short-CARE. SMI was found to be common, occurring in 25% of subjects. Subjects with SMI were more likely to be suffering from either dementia or depression than those without the complaint, although 60% of subjects with SMI did not have evidence of either disorder. When followed up over a 2-year period, subjects with SMI were found to be at four-fold greater risk of developing future dementia and two-fold greater risk of developing a depression compared with those without SMI. The SMI scale was not found to be useful as a population screen for dementia or depression, although two of the nine items might have value as screening questions in clinical circumstances to determine those with memory complaints at risk for dementia.