Patients with chronic obstructive pulmonary disease (COPD) usually have limited exercise tolerance owing to low ventilatory capacity. Because metabolic acidosis induced by exercise increases ventilatory drive, decreasing the hydrogen ion stimulus may improve exercise capacity. However, in those with mechanical limitation to ventilation or chemoreceptor insensitivity, identifying metabolic acidosis may be difficult using gas exchange methods that depend on the ventilatory response to the acidosis. We compared a modification of a gas exchange method (V-slope) for determining the lactate (anaerobic) threshold (AT), which is independent of ventilatory response with a method using the change in blood standard bicarbonate (HCO3-) level in COPD and normal subjects during cycle incremental exercise. In 43 normal subjects, the VO2 at which metabolic acidosis was identified using the two method correlated (r = 0.75), although mean values differed. In 22 patients with moderately severe to severe COPD, eight who had a change in standard HCO3- less than 2.0 mEq/L between rest and 2 min of recovery from exercise (group 1) were contrasted with 14 whose blood standard HCO3- fell by greater than 2.5 mEq/L (group 2). Mean VC was higher and FEV1/VC was lower in group 2, but mean FEV1, maximal voluntary ventilation, and diffusing capacity for carbon monoxide were not different. The degree of obstruction did not correlate strongly with the degree of exercise metabolic acidosis. The AT determined by the V-slope method was compared with that from standard HCO3-; good correlation between these methods was found (r = 0.98), although mean values were different. The V-slope method predicted metabolic acidosis in 10/14 who had a fall in HCO3- more than 2.5 mEq/L. A significant proportion of patients with COPD seem to develop metabolic acidosis during exercise. The V-slope gas exchange method may be useful in selecting those patients with COPD who develop exercise metabolic acidosis and might therefore benefit from exercise training.