The purpose of this study was to determine the predictor variables for breathlessness and to investigate the criteria of reliability and responsiveness for measuring breathlessness during progressive, incremental exercise on the cycle ergometer. We studied a heterogeneous group of patients with stable asthma (mean +/- SEM age, 46 +/- 4 yr) for four visits at weekly intervals. Predictor variables were determined at the first visit. Nine independent physiologic variables were obtained at each minute during exercise; the Borg rating of breathlessness (range 0 to 10) was used as the dependent variable. The regression model relating the physiologic parameters to the Borg rating of breathlessness was highly significant (model F = 43.4; p = 0.0001). Backward elimination selected the strongest predictors of the Borg rating: peak inspiratory flow (VI); tidal volume (VT)/FVC; frequency of respiration (f); and peak inspiratory mouth pressures (Pm). These four variables explained 63% of the variance in the rating of dyspnea. Each of the four variables exhibited a linear relationship with the Borg rating. Test-retest reliability was assessed by comparing results at the first and second visits. Individual slopes (except for VT/FVC) and intercepts for the four predictor variables versus Borg ratings were highly reliable. The slope for work intensity (watts) and Borg ratings, but not the intercept, was highly reliable. Responsiveness was evaluated by randomly administering inhaled methacholine or inhaled metaproterenol, alternately, at the third and fourth visits to induce acute changes in lung function before exercise.(ABSTRACT TRUNCATED AT 250 WORDS)