This investigation evaluated cancer incidence among workers at a petrochemical research facility in Illinois. A cluster of brain cancer and other intracranial tumors had occurred at the facility before the study began. The subjects were 5641 people who had worked at the facility from 1970 through 1996 and who had lived in Illinois at any time between 1986 and 1997. Data on cancer cases came primarily from the Illinois State Cancer Registry. Analyses compared the 1986-to-1997 cancer incidence rates of employees with Illinois general population rates. Subjects had 18% fewer than expected total cancers (125 observed/153 expected cases; standardized incidence ratio [SIR], 82; 95% confidence interval [CI], 68 to 98), which was primarily attributed to a large deficit of lung cancer (10/26; SIR, 39; CI, 19 to 72). Brain cancer was increased in the overall study group (6/2.7; SIR, 222; CI, 81 to 484). This excess was restricted to white men who were scientists or technicians for one of the three companies at the facility (6/0.8; SIR, 750; CI, 275 to 1633); all cases in this group had worked in the "500 building complex" (6/0.6; SIR, 968; CI, 355 to 2106). Subjects also had an increased incidence of thyroid cancer (7/2.6; SIR, 265; CI, 106 to 546) that was not concentrated in particular occupational or building groups. The brain cancer incidence patterns indicated that an unidentified occupational exposure might have been responsible for the excess. Chance, socioeconomic factors leading to better case detection in facility employees than in the general population, and confounding by potential nonoccupational risk factors are plausible explanations of the observed increase in thyroid cancer.