The errors in the measurement of blood gas analysis are loss of CO2 by exposure to atmospheric air, effects of the anticoagulant itself, temperature differences between the experimental subject and the measuring electrode, and metabolic changes which occur between sampling and measurement. The errors due to these metabolic processes may be minimized by transferring samples from the subject to the measuring electrode as quickly as possible. This is net always feasible, The effect of delay in estimation was studied in 22 patients who were taken elective operation or respiratory care with arterial line in situ. The syringes were stored at 0 -4degrees C (refrigerator) and 20-24degrees C(room temperature), and samples fur analysis were taken at intervals through 3 hours. We obtained the following results. 1) The partial pressure of oxygen fell significantly by 20 minutes at 0-4degrees C and 10 minutes at 20-24degrees C. 2) Oxygen disappeared from blood in the group with PO2 above 150 mmHg. The rate of disappearance was 18 mmHg/hr at 0-4degrees C and 42 mmHg/hr at 20-24degrees C. 3) PaCO2 increased significantly by 180 minutes at 0-4degrees C and 10 minutes at 20-24degrees C. 4) The rate of PaCO2 rise was 0.6 mmHg/hr at 0-4degrees C and 1.2 mmHg/hr at 20-24 degrees C. 5) Blood pH decreased significantly by 60 minutes at 0-4degrees C and 20 minutes at 20-24degrees C. 6) Bicarbonate, arterial O2 saturation and O2 content were not changed. So we highly recommend that blond gas analysis should be performed as soon as possible after sampling, especially within 10 minutes.