OBJECTIVE To follow up workers in the grain processing and animal feed industry five years after an initial survey, and to monitor exposures to organic dust and endotoxin and changes in prevalence of respiratory symptoms and lung function. METHODS Outcome measures in the present survey were decline in lung function over five years, rapid annual decline in forced expiratory volume in one second (FEV1) above 90 ml.s-1, and loss to follow up. RESULTS Among 140 workers included in the longitudinal analysis, annual decline in FEV1 and maximal mid-expiratory flow (MMEF) were significantly related to occupational exposure to dust and endotoxin in the grain processing and animal feed industry. Assuming a cumulative exposure over a working life of 40 years with an exposure of 5 mg.m-3, the estimated effect on the FEV1 would be a decline of 157 ml.s-1 (95% CI 13 to 300)--that is, about 4% of the group mean FEV1 and 473 ml.s-1 (95% CI 127 to 800) of the MMEF (about 12%). Workers with a dust exposure > 4 mg.m-3 or endotoxin concentrations > 20 ng.m-3 at the 1986-8 survey had significantly higher risk of rapid decline in FEV1 (odds ratio (OR) 3.3, 95% CI 1.02 to 10.3). The relations between occupational exposure and decline in lung function in this study occurred, despite the selection through the healthy worker effect that occurred as well. Increasing working years was related to decreasing annual decline in FEV1 and fewer people with rapid decline in FEV1 (OR 0.04, 95% CI 0 to 0.61 for over 20 v < 5 working years in the grain processing and animal feed industry). The presence of respiratory symptoms at baseline was a strong predictor of subsequent loss to follow up. Baseline lung function was not found to be predictive of subsequent loss to follow up. However, among workers lost to follow up the number of working years was more strongly negatively related to baseline lung function than among the workers who were studied longitudinally. CONCLUSIONS The existence of the healthy worker effect implies that an exposure-response relation in the grain processing and animal feed industry may well be underestimated. This should be taken into account when health based recommended limit values are to be developed.