Test-retest reliability of onset date and type of initial symptoms of Parkinson's disease (PD) was assessed in 45 patients with PD who were drawn from a community-dwelling cohort in New York City. Patients were the sole information providers on both rating occasions. Reliability of symptom-onset data was found to be high, even when the interval between test and retest was up to 3 years and when the information was elicited by different neurologists. However, these patients were less consistent about the type of early symptoms. When interpreting their own disease history, patients with PD may attach more importance to when their disease began than to how it began.