This study was designed to standardise a progressive exercise test for the assessment of change in carbon monoxide transfer coefficient (KCO) with exercise and to examine the variation between subjects and the reproducibility within subjects. Normal subjects exercised on a bicycle ergometer while ventilation, heart rate, and expired gas concentrations were recorded continuously. Preliminary studies showed that reduction of the breath hold time to six seconds made measurements of KCO during heavy exercise more comfortable without affecting the result. When KCO was measured immediately after exercise it was lower than when measured during exercise. KCO was measured in 50 normal subjects at rest and at three different work loads maintained for three minutes with a pause of five minutes between each. The relationships between KCO and both oxygen consumption and work load were linear in all subjects but the relationship between KCO and heart rate was distorted by high resting heart rates in some subjects. The mean slope of the relationship between KCO and oxygen consumption (VO2) was steeper in women than in men (mean slopes 0.627 and 0.348 mmol min-1 kPa-1 l-1 per 1 min-1 respectively), and the same was true for the relationship between KCO and work rate. The heart rate rose more steeply in relation to VO2 in women, so that the relationship of KCO to heart rate was similar in men and women (mean slope 0.01 mmol min-1 kPa-1 l-1 per beat min-1). Repeat studies on five occasions in five individuals gave coefficients of variation for the slopes of the relationships between KCO and VO2, work rate, and heart rate of 5-10%.