Smoking was examined in relation to Parkinson's disease (PD) in a population-based study in northern Manhattan (New York City) because of its putative "protective effect." Using a case-control design, information on smoking and associated behaviors was obtained in structured interviews after standard diagnostic evaluations in both cases and controls. The overall prevalence of smoking in the population was 43.7%, decreasing to 37% after age 85. Smoking was most frequent in men, Blacks, and in both cases and controls using alcohol once per week or more. Cases had quit smoking more often than controls (87 vs. 64%), and had smoked for significantly fewer years (31 vs. 41 yrs; p < 0.05 for both). The age-at-onset for smokers with PD was similar to age-at-onset for nonsmokers with PD. The odds ratio (OR) for a history of smoking associated with PD was 1.1 (95% CI 0.7-1.8). No protective gradient was associated with heavier smoking patterns. However, the odds that patients with PD were still smoking at the time of the interview were significantly less than those for controls (OR = 0.2; 95% CI 0.1-0.5). These results do not support the hypothesis that smoking protects against PD; rather they strongly imply the converse, that PD reduces smoking.