We wished to identify the physiologic abnormalities that distinguish severely breathless (SB) patients with chronic airflow limitation (CAL) from mildly breathless (MB) patients. Thirty-seven patients with stable, advanced CAL (FEV1 = 38 +/- 10 percent predicted, mean +/- SD) were separated into two distinct groups, SB and MB, solely on the basis of their baseline dyspnea index (BDI). BDI ratings in SB (n = 17) and MB (n = 20) patients were 2.5 +/- 1.5 and 8.5 +/- 1.5 (mean +/- SD), respectively (p less than 0.001). Groups were compared with respect to pulmonary function, breathing pattern parameters, arterial blood gases (ABGs), and responses to progressive exercise. Steady-state gas-exchange parameters were measured in a subgroup of 16 patients during exercise. There were no significant intergroup differences in dynamic flows, plethysmographic lung volumes, ABGs, resting ventilation, or breathing pattern parameters. However, the SB group had significantly lower single-breath diffusing capacities for carbon monoxide (Dco) (by an average of 50 percent, p less than 0.001), together with significantly higher resting ventilatory equivalents for carbon dioxide (VE/VCO2) (by 17 percent, p less than 0.01) and dead space to tidal volume ratios (by 11 percent, p less than 0.05). Ventilatory responses for a given metabolic load were, on average, 33 percent higher (p less than 0.05) in the SB group reflecting greater ventilation-perfusion inhomogeneity and wasted ventilation. The SB subgroup (n = 7), in contrast to the MB subgroup (n = 9), demonstrated significantly (p less than 0.01) greater O2 desaturation during exercise; PaO2 decreased in SB and MB at peak exercise by -13 +/- 7 mm Hg and -4 +/- 2 mm Hg (mean +/- SD), respectively. Stepwise regression analysis selected DCO and VE/VCO2 as the only predictors of breathlessness in this group, accounting for 52 percent of the variance in BDI (F-ratio = 18.49, p less than 0.001). Although the origin of breathlessness is multifactorial, variation in its intensity among patients with comparable levels of airflow limitation can be accounted for, in part, by underlying pathophysiologic differences. Severely breathless patients were characterized by lower resting diffusing capacities and accelerated ventilatory responses to exercise.