RATIONALE: Associations between coal mine exposures and pulmonary function have been studied in the US and abroad. Studies in new coal miners have suggested that spirometry changes may be non-linear, possibly an initial rapid decline followed by partial recovery. METHODS: We investigated the early pattern of lung function change in a cohort of 317 newly hired Chinese coal miners, compared to 132 referents. The 3year study included a pre-employment and 15 follow-up health surveys (questionnaires and spirometry), performed monthly for 3 months, then bi-monthly for 6 surveys, every 3 months for 5 surveys, and a final survey at a 6-month interval. Area sampling for total and respirable dust was done bimonthly. We used a mixed effects model to analyze longitudinal spirometry data. allowing the detection of both linear and quadratic time trends, and differences between groups in FEV1 change over the study period, controlling for age, height, pack-yrs of smoking, mean respirable dust concentration, and room temperature during testing. The model included interaction terms for both group, linear time, and group quadratic time. A spatial power law covariance structure was used, which has the property of larger correlations for proximate than distant times. RESULTS: The new miners experienced an initial rapid FEV1 decline, primarily during the first year of mining, with a leveling-off in the second year, and partial recovery during the third year. Both linear and quadratic time trends in FEV1 change were highly significant. Referents did not demonstrate rapid initial declines. CONCLUSIONS: The findings are consistent with previous studies, indicating that FEV1 change over time in new miners is non-linear. Better definition of longitudinal FEV1 patterns should assist in both design of medical monitoring programs and interpretation of results among dust-exposed workers.