Transcutaneous and arterial carbon dioxide were measured simultaneously in 57 children (age range 10 days to 14.3 years) undergoing intensive care. All were haemodynamically stable at the time of study. Mean calibration time with 5 and 10% carbon dioxide was 43 (range 38-58) minutes and mean arterialisation time was 10.5 (range 3-30) minutes. Duplicate hourly arterial samples over a four hour period showed that transcutaneous: arterial carbon dioxide correlation was independent of electrode temperatures over the range 42-44 degrees C and was independent of child age. One and a half hours after electrode placement transcutaneous carbon dioxide (kPa) = arterial carbon dioxide X 1.41 + 0.02. Use of a simple ratio (arterial carbon dioxide = transcutaneous carbon dioxide/1.40) resulted in a mean estimate of arterial carbon dioxide within 3% of the actual value with coefficients of variation of 11 and 15% at 1.5 and 4.5 hours, respectively. For prediction of arterial carbon dioxide 95% confidence limits around the regression mean rose from +/- 1.04 kPa (7.8 mm Hg) at 1.5 hours to +/- 1.56 (11.7 mm Hg) at 4.5 hours. Baseline drift was more than +/- 0.67 kPa (5 mm Hg) during 22% of the studies. For a drift of less than 0.67 kPa 95% confidence limits were 0.87 and 1.12 kPa, respectively. Arterial carbon dioxide can be estimated with clinically acceptable tolerances from surface electrodes operating at temperatures between 42-44 degrees C. Improved electrode stability and speed of calibration should make this a valuable monitoring technique.