We conducted a cohort analysis of longitudinal changes in pulmonary function test values in male grain elevator workers in five regions of Canada over a 6-yr period involving three observations. Pulmonary function tests were obtained approximately each 3 yr during an observation interval called a "cycle". The periods of the cycles were 1978-81 (Cycle I), 1981-84 (Cycle II), and 1984-87 (Cycle III), respectively. Data on respiratory symptoms and pulmonary function tests were collected on the same 881 workers in Cycles I and II (mean ages +/- SE: 37.0 +/- 0.48 yr at Cycle I), on the same 1,294 workers in Cycles I, II, and III (34.6 +/- 0.34 yr at Cycle I), and on the same 1,021 workers in Cycles II and III (33.9 +/- 0.38 yr at Cycle II) for a total of 3,196 workers with two or more observations over the 6-yr period. The proportion of current smokers decreased from Cycle I to Cycle III. Autoregression analysis of longitudinal changes in lung function data showed that predictors of FEV1 were height, years in the industry, smoking status, and geographic region. The mean annual losses in FEV1 (FVC) were 28.7 ml (26.4 ml), 38.4 ml (41.9 ml), and 41.7 ml (32.3 ml) for nonsmokers, ex-smokers, and current smokers, respectively. The annual loss of FEV1 (FVC) was 9.2 ml (21.1 ml) for those who were in the grain industry for less than 5 yr and 52.6 ml (60.8 ml) for those who were exposed for 20 yr or more.(ABSTRACT TRUNCATED AT 250 WORDS)