Experienced drug takers and clinicians report that cocaine causes panic attacks. This claim is supported by laboratory evidence on the pharmacologic activity of the drug. In this paper, the authors have used an epidemiologic strategy to examine the suspected cocaine-panic association, with interview data from 5,896 adult household residents sampled in the early 1980s and followed prospectively for a collaborative multisite study of mental disorders in five US metropolitan areas: New Haven, Connecticut; Baltimore, Maryland; St. Louis, Missouri; Durham, North Carolina; and Los Angeles, California. The risk of panic attacks was observed to be greater for identified cocaine users in this sample, as compared with subjects who did not use cocaine during the follow-up interval. The cocaine-panic association remained strong after statistical adjustment for preexisting psychiatric conditions, use of alcohol and marijuana, and suspected sociodemographic risk factors for panic attacks. The risk was greatest among cocaine users who reported no marijuana use during the follow-up interval (estimated relative risk = 13.0, 95% confidence interval: 2.24-75.8). The study also identified other determinants for panic attack, including sex, marital status, employment status, job prestige, major depression, and heavy drinking.