In order to evaluate the human carcinogenicity of formaldehyde, we conducted a proportionate mortality study of garment workers engaged in the production of shirts from formaldehyde-treated cloth. This study included three plants, and was based upon 256 deaths identified from a death-benefit insurance fund. No deaths due to nasal cancer were observed, and the mortality from respiratory cancer (11 cases, PMR = 95) was slightly less than expected. Statistically significant (p less than .05) elevations in proportionate mortality were observed for malignant neoplasms of the "buccal cavity" (three cases, PMR = 750), for "biliary passages and liver" (four cases, PMR = 313) and for "other lymphatic and hematopoietic sites" (four cases, PMR = 400). A proportionate cancer mortality (PCMR) analysis also was conducted, and cancer of the "buccal cavity" (three cases, PCMR = 682), and other "lymphatic and hematopoietic sites" (four cases, PCMR = 342) were still significantly elevated. The observed excesses in cancer mortality were primarily experienced by white females, who made up the major portion of the workforce, and workers with more than 10 years of latency and duration of exposure, a criterion for inclusion for most workers in the study group. The neoplasms observed were not equally distributed among the three facilities included in the study. Because of the small number of deaths involved and the lack of consistency with other studies, we believe that these findings should be viewed cautiously, pending the outcome of more definitive studies.