We conducted a cross-sectional survey to estimate the prevalence of tuberculosis infection among health care workers at two downtown Montreal hospitals. Participants completed questionnaires, then underwent two-step tuberculin testing. Records of previous tuberculin tests and BCG vaccinations were reviewed. Charts of all tuberculosis patients admitted in 1992-93 were also reviewed. Air changes and direction of air flow in patient care areas were measured using tracer gas techniques and smoke tubes. Of 619 eligible workers, 522 participated (84%). 196 (38%) were tuberculin reactors; 23 (4%) had documented conversions. Inadequate ventilation and delays in diagnosis were identified at both hospitals. Comparing clinical with nonclinical personnel, the adjusted odds of a significant initial tuberculin reaction were 2.6 (95% confidence interval 1.3, 5.2), of a documented conversion 13.6 (1.4, 132), and of a booster reaction 0.9 (0.2, 3.6). Initial tuberculin reactivity was associated with male gender (p = 0.008), BCG vaccination (p = 0.0001), foreign birth (p = 0.007), age (p < 0.0001), and occupation (p = 0.02); conversion with male gender (p = 0.001) and occupation (p = 0.01); and boosting with older age (p = 0.02) and BCG vaccination (p = 0.001). Among clinical personnel at two hospitals, the prevalence of significant tuberculin reactions and of documented conversions was unexpectedly high.