Exposure to metalworking fluids has been linked to modest cross-shift reductions in FEV1 and occupational asthma. To identify responsible agents, we measured personal exposures to thoracic particulate (TP), viable plus nonviable thoracic bacteria (BAC), and vapor phase nicotine (VPN) (as a surrogate for tobacco particulate) among 83 machinists exposed to soluble oils and 46 dry assemblers working in an automotive transmission machining plant using biocides infrequently. The participants completed interviews and performed pre- and postshift spirometry on Monday and Thursday of the same week in each of three rounds of data collection (June 1992, January 1993, June 1993). Generalized estimating equations were used to combine information across rounds in multiple regression models of cross-shift and cross-week changes in forced expiratory volume, I second (FEV1) and forced vital capacity (FVC). Mean seniority was 19 years among machinists. Mean personal TP levels were 0.41 mg/m3 in machinists and 0.13 mg/m3 in assemblers. Six of the 83 machinists and none of the 46 assemblers experienced a greater than 19% cross-shift decrement in FEV1 or FVC at least once (p = .07). In regression models using either TP or BAC, among subjects with lower baseline (Monday preshift) FEV1/FVC ratios, increasing exposure was significantly associated with increasing cross-shift decrements in FEV1 and FVC in linear models, and with increased likelihood of a 10% or greater cross-shift decrement in FEV1 or FVC in logistic models. Adjustment of TP for VPN did not affect models significantly. We conclude that clinically important cross-shift decrements in pulmonary function are associated with exposure to metalworking fluid aerosols within a high-seniority population.