The accuracy and reliability of the relationship between arterial and end-tidal carbon dioxide (PETCO2 and PaCO2), expressed as the gradient, P(a-ET)CO2, was assessed with 171 comparisons in nine mechanically ventilated trauma patients. The P(a-ET)CO2 was 14 +/- 11 mm Hg. (mean +/- standard deviation.) The positive correlation between PaCO2 (44 +/- 10 mm Hg) and PETCO2 (30 +/- 10 mm Hg) for the study population (reflected by r = 0.64, p = 0.001; but r2 = 0.41) indicated statistical significance, but only 40% of the changes reflected a linear relationship. Seventy-eight percent of individual patients had significant correlations of PaCO2 and PETCO2 (p < 0.02 to p < 0.001). Changes in PETCO2 erroneously predicted the PaCO2 changes in 27% of comparisons with 15% false decreases and 12% false increases. Trends in P(a - ET)CO2 magnitude are not reliable, and concordant direction changes in PETCO2 and PaCO2 are not assured.