A simple discriminant function in four variates (area of mill, average number of cigarettes currently smoked per day, a complex of dyspnoea, cough, and wheezing grades, and occupation) was derived to discriminate between non-byssinotic (grade o) and byssinotic (grades I and II) preparers. This function diagnosed 14% of non-byssinotics as byssinotic and 35% of byssinotics as non-byssinotic, giving a total misclassification rate (24·5%) that agreed closely with the theoretically expected rate (26·4%). Similar misclassification rates were found in a randomly selected check sample of preparers. The function did not diagnose accurately preparers with early byssinosis (grade ½); 64% of such workers would have been diagnosed as normal.