A retrospective exposure assessment study in the automotive parts industry conducted in conjunction with a cancer mortality and respiratory morbidity study [Kennedy et al. (1989): Am J Ind Med 15:627-641; Eisen et al. (1992): Am J Ind Med 22:809-824; Tolbert et al. (1992): Scand J Work Environ Health 18:351-360] describes exposure to different types of machining fluids and selected components that may contribute to the conditions investigated. A dataset of 394 industrial hygiene measurements made between 1958 and 1987 was used to estimate past machining fluid levels using a linear statistical model. The effects of different plants, machining fluid types, machining operations, and time periods were examined in the model. Separate analyses examined the effects of different sampling and analysis methods and other measurement variables. Machining fluid levels prior to 1970 were generally two to five times higher than subsequent measurements. The arithmetic mean exposure of all measurements taken before 1970 was 5.42 mg/m3. Arithmetic means for different subgroups ranged from 0.59 to 20.28 mg/m3, depending upon plant, machining fluid, and operation. The arithmetic mean exposure after 1980 was 1.82 mg/m3 with subgroups ranging from 0.45 to 2.79 mg/m3. Changes in exposure levels generally corresponded with reported changes in plant environments such as installation of enclosures and local exhaust ventilation on machines.