A decreased level of glucose-6-phosphate dehydrogenase might result from decreased rate of synthesis, synthesis of an enzyme of lower catalytic efficiency, increased lability, or a combined mechanism. To test the hypothesis of increased lability, the rate of decline of the enzyme in vivo was measured in three groups of individuals, controls, Gd(-),A-males, and Gd(-), Mediterranean males, by the slope of decline of activity in fractions containing erythrocytes of progressively increasing mean age. These fractions were obtained by ultracentrifugation on a discontinuous density gradient of erythrocyte suspensions free of contaminating platelets and leukocytes. The rate of in vivo decline of pyruvate kinase (another age-dependent enzyme) was also measured and found very similar in the three groups. The in vivo decline of glucose-6-phosphate dehydrogenase was found to follow an exponential rate, with a half-life of 62 days for controls and 13 days for Gd(-),A- erythrocytes. The activity in normal reticulocytes was estimated at 9.7 U and in Gd(-),A- reticulocytes at 8.8 U. These estimates were confirmed by direct measurements in reticulocytes isolated from patients with extreme reticulocytosis. In Gd(-),Mediterranean erythrocytes activity could be demonstrated only in reticulocytes, which were estimated to average 1.4 U. The rate of decline is so extreme that no activity could be detected in mature erythrocytes. These data suggest that the glucose-6-phosphate dehydrogenase deficiency of both the Gd(A-) and the GdMediterranean variant results from different degrees of in vivo instability of the abnormal enzyme.