Information on the size and shape of about 8500 maximum expiratory flow-volume (MEFV) curves was related multivariately to respiratory symptoms to construct a sensitive measure of airflow limitation. The data were obtained in an epidemiological follow-up study of a normal population in The Netherlands. Using non-linear canonical correlation analysis, thirteen variables from each curve yielded two uncorrelated variates which were optimally related to two variates simultaneously derived from all respiratory data. Curves with different size and shape may yield the same probability for respiratory symptoms. In this context the scores of individuals on the scales of the two curve-derived variates do not have to be corrected for body or lung size. Using FEV1 and standing height about half as much of the variability in respiratory symptoms is explained as when using the whole MEFV-curve. Moreover, the distribution of the scores for the MEFV-curves allows the differentiation between three patterns of airflow limitation graphically. These three types suggest different pathophysiological mechanisms related to airways resistance and lung elastic recoil and may be used to characterize individual curves. One type relates to subjects with symptoms of 'asthma', one to airflow limitation as found among people with dyspnoea and elderly subjects and another to bronchitic symptoms. In females, the association of the MEFV-curve with 'asthma' is stronger than in males but much weaker with bronchitic symptoms. Curves found to be more related to 'asthma' or bronchitic symptoms, but not to dyspnoea, are more prevalent among current than among never smokers in males.