Four hundred and eighty-six textile workers in three cotton mills and one wool/synthetic mill were studied for symptoms and functional effects of workroom exposure to dust. Byssinosis was found in 5.7% of 386 cotton workers, with an apparent threshold level of 0.5 mg cotton dust/m3 of air. Mean post-shift functional declines were greater in workers exposed to greater than or equal to 0.2 mg/m3. Workers with byssinosis were unequally distributed, however, with respect to job category and mill; and these variables, rather than current dust exposure levels, accounted for the observed distribution of byssinosis prevalence rates. Variation in biological potency of different samples of cotton dust could be responsible for 'mill effect', the residual variation in response rates by mill after controlling for variation due to dust exposure. A number of other potential influencing variables that are likely to be distributed unequally by mill should also be considered. Mill effect should be assessed in large-scale studies of byssinosis, most of which have analysed biological response rates by combining mill and other variables to examine first-order effects of dust dosage. In such analyses, much of the observed variability may be due to factors other than dust dosage.