The large-scale mail questionnaire is a useful tool in epidemiological investigation and will probably come into wider use in the search for an environmental cause of Parkinson's disease (PD). To determine the validity of mail questionnaires in patients with PD, we administered a 17-item questionnaire by in-person interview, as a standard, to 68 patients (and/or a relative when necessary) and compared the results with the same questionnaire mailed at least 1 month before or after the interview. Questions in three formats requested recall of the clinical course and past environmental factors. Each patient also completed a multiple-choice physical self-assessment (a modification of four items on the Columbia Scale) immediately before seeing the neurologist, who completed the same form after the examination. Percent of patients with zero discordance between mail and interview responses averaged 52% for the nine fill-in-a-year items, 53% for the three list-generation items. Kappa statistics for the five multiple-choice items, which each offered four choices, averaged 0.67 (range 0.40-0.89). Kappa for the physical examination items, each rated on a 0-3 scale, was finger-tap 0.12, gait 0.34, tremor 0.35, and chorea 0.20. Patients' ratings tended to be more severe than neurologists' ratings. We conclude that mail surveys in PD should either be avoided or rigorously pretested for validity.