Several designs of personal samplers were tested for use to collect cotton dust. The IOM personal inhalable-dust sampler was selected because: (1) collection of the whole inhalable fraction was preferred, since all inhaled sizes are under suspicion as contributing to respiratory symptoms in cotton exposure; (2) this sampler is well characterized; and (3) it was found to be practicable in the environments examined. Gauze shields to exclude 'fly' from the personal sampler were tried, but were rejected mainly because measurement of the whole inhalable fraction by a validated sampler was felt to be more appropriate. A range of processes at a representative selection of mills was assessed by a hygiene team, and classified as 'clean' or 'dirty' in terms of present standards of control. This classification agreed well with subsequent measurements using the present method, which uses a large static sampler. A personal sampling survey then showed that in about two-thirds of 'clean' processes personal exposure of at least 80% of those employed was less than about 2-2.5 mg m-3. Only one-tenth of 'dirty' processes met this standard. Personal exposure correlates poorly with the present static method, as expected, but comparison of the results suggested that a mean background level of 0.5 mg m-3 would correspond to a median personal exposure of about 2.2 mg m-3. Side-by-side measurements by the background method differed by less than 0.15 mg m-3 on about 95% of occasions. Niven et al. (to be published) have compared the IOM head used in this study with the Manchester University sampler previously used by Cinkotai et al. [Ann. occup. Hyg. 32, 103-113 (1988)] to derive a relationship between personal exposure and prevalence of byssinotic symptoms in spinners. According to Cinkotai et al.'s results the concentrations of 2-2.5 mg m-3 discussed would correspond to a prevalence of 3-5%. However, this prevalence probably reflects higher exposures in the past.