Evaluating the reliability of self-assessed disability in patients with Parkinson's disease (PD) is important for therapeutic trials and epidemiologic surveys. This is the first study to examine the interrater reliability between physician and patient in the historical section of the Unified Parkinson's Disease Rating Scale (UPDRS). Thirty consecutive subjects with idiopathic PD self-administered the historical section of a modified UPDRS. This instrument was then readministered to these 30 subjects by a neurologist. Interobserver reliability was assessed with the weighted kappa statistic (kw). The kw for each of the 17 items in the historical section of the UPDRS ranged from 0.63 to 1.0 (moderate to excellent agreement). Total kw = 0.83 (95% confidence interval = 0.79-0.87). There were no correlations between kw and age, disease duration, dose of levodopa, Hoehn and Yahr score, Schwab and England score, or modified Mini-Mental State score. Nondemented subjects with PD may reliably assess their level of disability by self-administering the historical section of the UPDRS. This has important implications for the reliable use of self-administered disability instruments for clinical research.