A screening questionnaire with high sensitivity for detection of Parkinson's disease would make it easier to identify undiagnosed, yet affected, family members for genetic research. We assessed the validity of a screening questionnaire developed by Duarte et al. [1995: Mov Disord 10:643-649] with reported high specificity and sensitivity for Parkinson's disease (PD). We applied the questionnaire to 78 asymptomatic members of families that had at least two people diagnosed with PD. These families were participating in a linkage study of Parkinson's disease. Examination of these 78 revealed that 53 were normal (normal controls) and 25 were classified ("undiagnosed" PD defined) as possible, probable, or clinically definite PD based on standardized criteria. We compared these results with 123 patients with clinically definite PD ("diagnosed" PD). There were significant differences among the mean scores on the questionnaire for normal controls (4.4), subjects with undiagnosed PD (9.8), and patients with diagnosed PD (42.1; p<0.000001) and a significant difference between undiagnosed PD and normals (p<0.01). The questionnaire had only 4% sensitivity for detection of parkinsonism in undiagnosed PD using the original criteria [Duarte et al., 1995]. Revising the criteria increased the sensitivity from 4 to 48% in the undiagnosed group. The positive predictive value was 39% and the negative predictive value was 72%. Prospective application of these revised criteria is necessary to confirm the improved sensitivity. However, we conclude that this screening questionnaire has inadequate sensitivity for detection of mild parkinsonism and direct examination is still critical for accurate classification for genetic studies.