BACKGROUND Evidence from toxicological studies indicates that the risk of respiratory diseases varies with asbestos fibre length and width. However, there is a total lack of epidemiological evidence concerning this question. METHODS Data were obtained from a cohort mortality study of 3072 workers from an asbestos textile plant which was recently updated for vital status through 2001. A previously developed job exposure matrix based on phase contrast microscopy (PCM) was modified to provide fibre size-specific exposure estimates using data from a re-analysis of samples by transmission electron microscopy (TEM). Cox proportional hazards models were fit using alternative exposure metrics for single and multiple combinations of fibre length and diameter. RESULTS TEM-based cumulative exposure estimates were found to provide stronger predictions of asbestosis and lung cancer mortality than PCM-based estimates. Cumulative exposures based on individual fibre size-specific categories were all found to be highly statistically significant predictors of lung cancer and asbestosis. Both lung cancer and asbestosis were most strongly associated with exposure to thin fibres (<0.25 microm). Longer (>10 microm) fibres were found to be the strongest predictors of lung cancer, but an inconsistent pattern with fibre length was observed for asbestosis. Cumulative exposures were highly correlated across all fibre size categories in this cohort (0.28-0.99, p values <0.001), which complicates the interpretation of the study findings. CONCLUSIONS Asbestos fibre dimension appears to be an important determinant of respiratory disease risk. Current PCM-based methods may underestimate asbestos exposures to the thinnest fibres, which were the strongest predictor of lung cancer or asbestosis mortality in this study. Additional studies are needed of other asbestos cohorts to further elucidate the role of fibre dimension and type.