Using unpublished data from five completed prevalence surveys of Parkinson's disease (PD), we investigated case ascertainment uncertainties that potentially have a direct effect on prevalence. These uncertainties arise from the choice of diagnostic criteria, the choice of screening method, and the amount of information lost because of nonresponse. The surveys were conducted in Argentina, India, China, Italy, and the Netherlands. Our analyses consisted of simple comparisons of prevalence results, positive predictive values (a screening measure), and nonresponse percentages. We found that (a) prevalence comparisons between surveys have diminished value if the surveys used different diagnostic criteria for PD; (b) screening performance may be affected adversely if symptom questions are answered by one family member for the entire family living together rather than by each family member individually; and (c) nonresponse from refusal or unavailability does not necessarily lead to bias, but special caution may be appropriate with prevalence results pertaining to elderly women.