An analysis of 702 blood sugar levels during glucose tolerancetests shows that the site of blood sampling influences the blood sugar result. Capillary blood taken from the warm ear lobe gives an accurate reflection of the arterial blood sugar level. Compared to the arterial level the venous blood sugar level is unpredictable and differs from the arterial to a varying extent, e.g., from +26 mg. per 100 ml. to −4 mg. per 100 ml. at two hours in the glucose tolerance test depending on many factors. Studies of the macromethod in the AutoAnalyzer indicate that although the reproducibility is good, the volume of blood pumped by the proportioning pump is inaccurate (too small). Analysis of the micromethod shows that the reproducibility is also good, though not quite so good as the macromethod. However, the sample volume taken for analysis by the proportioning pump is accurate. Reasons are given for recommending that capillary, not venous, blood be taken, and that the ferricyanide-reducing micromethod on the AutoAnalyzer be used for blood glucose estimation.