The association between exposure to dust and pulmonary function was studied by longitudinal and cross sectional analyses in a group of United States underground coal miners beginning work in or after 1970. Quantitative estimates of exposure to respirable coal mine dust were derived from air samples taken periodically over the entire study period. The cohort included 977 miners examined both in round 2 (R2) (1972-5) and round 4 (R4) (1985-8) of the National Study of Coal Workers' Pneumoconiosis. Multiple linear regression models were developed for both cross sectional (pulmonary function at R2 and R4) and longitudinal (change in pulmonary function between R2 and R4) analyses with exposure partitioned into pre-R2 and post-R2 periods and controlled for covariates including smoking history. The results indicate a rapid initial (at R2) loss of FVC and FEV1 in association with cumulative exposure of the order of 30 ml per mg/m3-years. Between R2 and R4 (about 13 years) no additional loss of function related to dust exposure was detected although the percentage of predicted FVC and FEV1 did decline over the period. After some 15 years since first exposure (at R4), a statistically significant association of cumulative exposure with FEV1 of about -5.9 ml per mg/m3-years was found. These results indicate a significant non-linear effect of exposure to dust on pulmonary function at dust concentrations present after regulations took effect. The initial responses in both the FVC and FEV1 are consistent with inflammation of the small airways in response to exposure to dust.