The results of an historical cohort study of mortality among individuals occupationally exposed to formaldehyde were announced in 1986 by Blair et al (JNCI 1986; 76:1071-1084). The study was a joint undertaking of the National Cancer Institute and the Formaldehyde Institute, and concluded, ". . .this large multiplant cohort study provided little evidence to suggest that formaldehyde exposure affected the mortality experience of these industrial workers." However, there were concerns by a number of workers that the design and analysis of the study had possibly masked an existing occupational hazard. Analyzing time-integrated exposure to formaldehyde without simultaneously considering length of exposure and comparing mortality of formaldehyde workers to mortality of the general population could have masked an increase in cancer risks because of the healthy worker effect. A copy of the data of the study was obtained from the principal investigator and reanalyzed. We find a significantly increased risk for all cancers and for lung cancer as a function of cumulative exposure when workers with higher levels of exposure are compared with those with little or no exposure while simultaneously considering length of exposure. When the risk ratio (RR) for lung cancer at less than or equal to 0.1 ppm cumulative exposure (CX) is taken as 1.0, the lung cancer RR for CX of 0.1 to 0.5 ppm is 1.41 (1.20 to 1.66), the RR for CX of 0.5 to 2.0 ppm is 1.73 (1.42 to 2.11), and the RR for CX greater than or equal to 2.0 is 1.70 (1.32 to 2.18). Hourly workers have a significantly higher RR than salaried workers (RR = 1.58).(ABSTRACT TRUNCATED AT 250 WORDS)