The residual bleeding vol. can be estimated in normal unselected dogs under barbital anesthesia by extrapolation from the data on mean arterial pressure after pressure has been lowered to 60 mm. Hg by controlled hemorrhage. The error in the estimation, at 0.05 probability, is not greater than 6 ml./kg., or 8%-16% of the total bleeding vol. The absolute error in the extrapolation does not appear to change with the duration of anesthesia up to 10 hrs., nor to vary with the total bleeding vol. Both bleeding vol. and mean arterial pres- sure rise during the first 3 hrs. of barbital anesthesia. Bleeding vol., but not arterial pressure, declines between the 5th and 10th hrs. of anesthesia. When animals were bled under these conditions until the residual bleeding vol. as estimated by extrapolation was the same in all cases, and were reinjected with equal vols. of blood, they had nearly uniform total bleeding vols. 4 hrs. later. If the potential bleeding vol. was estimated as the sum of the residual and the replacement vol., the vol. obtained was approx. equal to the potential vol., so long as the latter exceeded a critical value lying between 22.5 and 32.5 ml./kg. When bleeding vol. was measured earlier than 2 1/2 hrs. after the initial bleeding and replacement, the vols. obtained deviated in both directions beyond limits permitted by the error in estimating the potential vol. The partition of total bleeding vol. into a variable vol. obtainable above, and a relatively constant vol. obtainable below 60 mm. Hg mean arterial pressure, is considerably disturbed for more than 4 hrs. after partial exsanguination and replacement.