Abstract The reliability of three methods commonly used for blood P co 2 measurements (CO 2-electrode, whole-blood C CO 2 /pH method, Astrup method) was evaluated by analysis of human blood samples tonometered at known CO 2 pressures. P CO 2 , was in the physiological region (38–53 mm Hg); S O 2 ranged from 93% to 16%. The COa-electrode yielded the most satisfactory results (smallest scat- ter, no directional error). The whole-blood C CO 2 /pH method was likewise devoid of directional error, but its scatter was appreciably larger. The Astrup method showed, along with an intermediate scatter, a desaturation-dependent systematic error. This error could be attributed to two factors affecting the desaturation correction: 1. the presence of small but not negligible amounts of non-oxygenable Hb in the samples (COHb+metHb), and 2. an overestimation of the Haldane effect in terms of BB change.