We assessed the relationship between bronchial hyperresponsiveness (BHR) and the onset of wheezing 5 years later, by epidemiological analysis of 194 working men without asthma or wheezing at the first examination. In 1985/ 1986 and 1990/1991, subjects answered a British Medical Research Council questionnaire and performed lung function measurements and methacholine challenge tests (total dose 6 mg). BHR was measured in three ways: (1) FEV1 fall > or = 20% (PD20+); (2) the two-point response slope expressed as percentage decline of FEV1/dose, and (3) a four-parameter model: FEV1 at dose (d)/ prechallenge FEV1 = ONE-k(d-delta)+a, where 'k' is the slope of the relative variation of FEV1 with the dose, 'delta' the threshold dose, and 'alpha' a shape factor. In the 13 new wheezers, the mean values of the two-point slope and of k were significantly increased, and the proportion of reactors was almost threefold (the latter was not statistically significant). Among nonsmokers, delta was significantly lower in new wheezers than in the others, whereas the slope and k had similar mean values. Among smokers, new wheezers had increased mean values for the slope and k, and an increased proportion of reactors, whereas delta was not decreased. Thus, BHR was a significant predictor of wheezing, independent of the method of analysis. Moreover, the model distinguished between two components of bronchial response: wheezing was predicted by sensitivity (delta) in nonsmokers, and by reactivity (k) in smokers.