BACKGROUND In 1997 a Working Group of the International Agency for Research on Cancer changed an earlier classification of crystalline silica as a human carcinogen from Group 2A to Group 1, though commenting that the carcinogenicity might vary with industrial circumstances and depend on additional factors affecting biological activity, including the distribution of its polymorphs. OBJECTIVE We aimed to determine whether pure quartz exposure uncomplicated by the presence of other contaminating carcinogens, as experienced by workers in the production of high-grade industrial sand, was causally related to an increased risk of lung cancer. METHODS A cohort of 2670 men employed before 1980 for 3 years or more in one of nine North American sand-producing plants and a large associated office complex was selected for study. Of the cohort, 2644 (99%) were traced through 1994, and certificated cause of death ascertained for 1025 (99%) of the 1039 men known to have died. Standardised mortality ratios (SMRs) were calculated for the main causes of death, using both US and state or provincial male mortality rates for reference. RESULTS The main analyses of deaths, 20 or more years after first employment against regional rates, gave the following SMRs: all causes 109, lung cancer 139, other malignancies 98, non-malignant respiratory disease 161, and nephritis/nephrosis 244. There were, in total, 37 deaths from silicosis or silico-tuberculosis, with one or more death at least in all nine production plants. Analyses failed to show any relation between lung cancer risk and duration of employment. The increased SMR for lung cancer was wholly due to high rates in four plants in two states, whereas no increase was found in the remainder of the cohort. CONCLUSIONS In the absence of information on smoking histories and risk in relation to estimated exposure, the increased SMR for lung cancer (139), although statistically significant, cannot be attributed confidently to crystalline silica. An answer to the question of attributability must await the findings of the nested case-control study, in which level of exposure and smoking habits were ascertained for cases and matched controls. The strong indication in this cohort of excess mortality from non-malignant renal disease deserves further investigation.