We examined the ability of two transcutaneous devices (Fastrac, Sensormedics Corporation, Yorba Linda, California, U.S.A. and Hewlett Packard M1018A, Hewlett Packard Component Monitoring System, Hewlett Packard, North Hollywood, U.S.A.) to measure arterial PCO2 and PO2 in neonates. Thirty-seven neonates had transcutaneous oxygen measured with the Hewlett Packard (HPO2 group), 38 neonates had transcutaneous carbon dioxide measured with the Hewlett Packard (HPCO2 group) and the Fastrac was used on 27 neonates (FTCO2 group). Both devices were operated with electrode temperatures of 43.5 degrees C although an additional ten subjects were studied using the Fastrac with an electrode temperature of 43.0 degrees C. The mean differences (transcutaneous--arterial) and upper and lower limits of agreement were calculated for each group. For the HPO2 group they were 3.78 mmHg (-12.23 to 19.80 mmHg), for the HPCO2 group they were 0.40 mmHg (-4.50 to 5.30 mmHg) and for the FTCO2 they were -0.96 mmHg (-7.85 to 5.92 mmHg). For the Fastrac group at an electrode temperature of 43.0 degrees C the mean difference and limits of agreement were -1.00 mmHg and -4.58 mmHg to 2.58 mmHg. The average sensitivity and specificity for both machines for the detection of hypocarbia were 82% and 92% respectively while for hypercarbia they were 90% and 94% respectively. For hypoxaemia, the sensitivity and specificity were 40% and 94% while for hyperoxaemia the sensitivity and specificity were 83% and 97%. We conclude that both machines provide a useful supplement to arterial PCO2 measurements and the Fastrac performs better at 43.0 degrees C. The measurement of PO2 is less accurate but is still of clinical use.