Smoking-related data collected during 1976-87 by anonymous cross-sectional surveys of nonphysician employees of a large medical care organization were used to assess how a work site smoking ban affected employees' smoking behavior and attitudes. The smoking ban was implemented at 11 work sites at various times during 1985-86. All work sites had three or more pre-ban surveys and one or two post-ban surveys. The majority of employees reported support for the smoking ban. The ban's effects were assessed by comparing observed post-ban rates with expected rates projected from secular trends in the pre-ban data by the use of logistic regression models that included age, sex, education, and job class as covariates. The work site smoking ban had a substantial effect on the presence of smoke in the work environment, but no short-term effect on smoking prevalence or attempts to quit. The apparent effect of the ban on quantity of cigarettes smoked was assessed by a pre-ban and post-ban analysis of the 1986-87 survey data. A nonequivalent post-ban and post-ban comparison was used to estimate the secular trend. A significant reduction of 1.4 cigarettes per day (P = 0.022) was found in smokers' rate during working hours.