Arterial PO2 measurement during exercise is an important part in the evaluation of pulmonary disease but requires an intra-arterial cannula. However, in clinical work it would be preferable to assess PO2 non-invasively. To evaluate such a technique, simultaneous measurements of transcutaneous PO2 (tcPO2) and arterial PO2 (PaO2), sampled from an indwelling arterial radial cannula, were made before, during and after a fatigue or symptom-limited bicycle exercise test in 16 patients referred to hospital because of dyspnoea. In total 181 paired measurements were made. Mean values (range) of PaO2 and tcPO2 were 11.2 kPa (5-16) and 9.5 (5-13.3), respectively. The correlation coefficient between PaO2 and TcPO2 was only 0.36 (P < 10(-5)). By normalizing the values of tcPO2 and PaO2 to corresponding values at supine rest before exercise, the correlation coefficient increased to 0.80 (P < 10(-6). Using PaO2 as golden standard, tcPO2 described the trend in pO2 during exercise reasonably well in all cases and this information is often sufficient for assessing the degree of pulmonary insufficiency. Thus, transcutaneous blood gas monitoring during exercise is useful for clinical evaluation of pulmonary disease, but a single arterial blood sample at rest before exercise is recommended for baseline correlation.