Identification of soft tissue sarcomas (STSs) in epidemiologic mortality studies is complicated by nosologic coding rules that require that STSs arising in a visceral organ must be coded in the International Classification of Diseases (ICD) category for that organ, rather than in the ICD category for malignant neoplasms of connective tissue. Moreover, prior studies have shown poor agreement between diagnoses recorded on death certificates compared with those in hospital records for these tumors. We reviewed deaths from STS among workers in a registry of 6,716 dioxin-exposed workers at the National Institute for Occupational Safety and Health (NIOSH) and in a NIOSH cohort mortality study of 10,240 workers exposed to chlorinated naphthalenes. We identified 19 subjects with STSs. Of these, 17 (89%) were identifiable by reading the entries on selected death certificates, and two (11%) were found only by reviewing medical records of cases coded to ICD categories likely to have contained STS. Of the 17 STSs identified from death certificates, only nine (53%) had been coded as underlying cause of death to the ICD category "malignant neoplasms of soft and connective tissue." Medical records were obtained for 14 of the 17 cases (82%), and in each case, the STS diagnosis was verified. Tissue blocks from tumors were available for review in nine of the 17 cases identified from death certificates, and the diagnosis of STS was verified in seven (78%). Nosologic rules reduce the sensitivity of cohort mortality studies to detect excesses of STS.(ABSTRACT TRUNCATED AT 250 WORDS)