Analyses were cross-sectional, and were performed using a commercially available software program (SPSS, version 11.0; SPSS Inc, Chicago, Ill). Most analyses were performed after having excluded the 200 participants with dementia (Table 1), because dementia is a confounder in the association between MPS and function. In some analyses (Table 2), the parkinsonian sign score was stratified into subscores, based on a factor analysis.17 The subscores were the axial function subscore (speech, facial expression, posture, and axial bradykinesia; range, 0-16), the rigidity subscore (rigidity rated separately in the neck and in each limb; range, 0-20), and the tremor subscore (range, 0-4). The correlation between the parkinsonian sign score, the parkinsonian sign subscore, and the functional and performance-based test scores was assessed with the Spearman correlation coefficient (Table 2). In some analyses (Table 3), the parkinsonian sign score was stratified (0, 1, 2, 3, 4, 5, 6, and ≥7), with scores of 7 or greater collapsed into 1 stratum because of the few participants with scores of 7 or greater. The association between increasing parkinsonian sign score and functional and performance-based test scores was assessed with a test for linear trend. We examined the association between functional and performance-based test scores and the parkinsonian sign score after adjusting for other covariates using analysis of covariance models (Table 4). Covariates were chosen if they were associated with functional disability and MPS. In each of the 5 analyses of covariance models, a functional or performance-based test score was the dependent variable and covariates were parkinsonian sign score, age (in years), sex, ethnicity, education (in years), Center for Epidemiologic Studies Depression Scale score, and medical illnesses that were each coded as present vs absent by self-report (diabetes mellitus, myocardial infarction, congestive heart failure, chronic obstructive pulmonary disease, peripheral vascular disease, seizures, and arthritis). Finally, in a single general linear model in which we adjusted for the covariates previously listed, we tested whether the parkinsonian sign score was independently associated with the functional and performance-based test scores (dependent variables).