A cohort mortality study was conducted among workers from two plants in the diatomaceous earth mining and processing industry in California. Diatomaceous earth consists of the skeletal remains of diatoms. Exposure to amorphous (non-crystalline) and crystalline silica in the form of quartz results from open pit mining and exposure to crystalline silica (principally cristobalite) occurs in the processing of the material. Lung cancer and non-malignant respiratory diseases have been the health outcomes of greatest concern. The main study cohort included 2570 white men (533 Hispanic and 2017 non-Hispanic workers) who were employed for at least 12 months cumulative service in the industry and who had worked for at least one day during the follow up period, 1942-87. Vital status was ascertained for 91% of the cohort and death certificate information was retrieved for 591 of 628 (94%) identified deaths. The all causes combined standardised mortality ratio (SMR) was slightly increased (SMR = 1.12; 628 observed) compared with rates among US white males. The principal contributors to this excess were increased risks from lung cancer (SMR = 1.43; 59 observed) and non-malignant respiratory disease (NMRD) excluding infectious diseases and pneumonia (SMR = 2.59; 56 observed). The excess of lung cancer persisted when local county rates were used for comparison (SMR = 1.59). Internal rate comparisons by Poisson regression analysis were conducted to assess potential dose-response relations for lung cancer and NMRDs. Mortality trends were examined in relation to duration of employment in dust exposed jobs and with respect to an index of cumulative exposure to crystalline silica. The crystalline silica index was a semiquantitative measure that combined information on duration of exposure, differences in exposure intensity between jobs and calendar periods, the crystalline content of the various product mixes, and the use of respiratory protection devices. Increasing gradients of risk were detected for lung cancer and NMRD with both exposure indices. The relative risk trends for lung cancer and NMRD with crystalline silica exposure lagged 15 years were respectively: 1.00, 1.19, 1.37, and 2.74, and 1.00, 1.13, 1.58, and 2.71. Based on a review of available but limited data on cigarette smoking in the cohort and from application of indirect methods for assessing confounding variables, it seems unlikely that smoking habits could account for all of the association between exposure to dust and lung cancer. The intense and poorly controlled dust exposures encountered before the 1950s were probably the most aetiologically significant contributors to risks from lung cancer and NMRDs. The absence of an excess of lung cancer among workers hired since 1960, and the finding of no deaths attributed to pneumoconiosis as an underlying cause of death among workers hired since 1950 indicate that exposure reductions in the industry during the past 40 years have been successful in reducing excess risks to workers. Further mortality follow up of the cohort and the analysis of radiographic data will be needed to determine conclusively the long term patterns of disease risks in this industry.