The requirement for D-glucose in T-cell-mediated cytolysis was studied using mouse spleen cells sensitized against alloantigens in vitro. Glucose was required for cytolysis: (a) cytolysis proceeded in a simple buffered salt solution containing Ca++ and Mg++ (low phosphate-buffered saline, LPBS) in the presence but not in the absence of added glucose; (b) 2-deoxy-D-glucose blocked cytolysis. The block by this agent was overcome by excess glucose added as late as 40 min after the inhibitor. This block was not due to inhibition of NADP reduction, since 2-deoxy-D-glucose failed to interfere with the rate of CO2 production by the pentose cycle which we found to be of significant activity in sensitized spleen cells; (c) dialyzed fetal bovine serum (DFBS) in LPBS supported cytolysis in the absence of added glucose. However, 2-deoxy-D-glucose was also inhibitory under these conditions, suggesting that carbohydrate was required here as well. Further results supported the conclusion that DFBS was not acting as a direct source of the required carbohydrate. The relationship between cytolysis, glucose requirement, and provision of energy was studied. As little as 0.1 mM D-glucose in LPBS supported cytolysis. At this glucose concentration, there was no measurable accumulation of lactate in sensitized spleen cells, but Krebs cycle activity was detectable. In 3 mM glucose or above, the range covered by standard tissue culture media, anaerobic glycolysis became a major source of energy in sensitized spleen cells. Consequently, it appears that in standard tissue culture medium, effector cells can generate sufficient energy for cytolysis either by aerobic or anaerobic metabolism. However, the addition of an energy source alone in the absence of glucose was insufficient to support cytolysis in LPBS. Pyruvate in LPBS did not support cytolysis but was shown to be a good substrate for aerobic metabolism in sensitized spleen cells. Glycogenic amino acids and glycerol also failed to support cytolysis. The stage of cytolysis at which glucose is required was investigated. Glucose was necessary for the calcium-dependent lethal hit phase, but not for the cytochalasin A-blockable recognition stage, nor for 51Cr release from injured target cells. Models for the lethal hit process are discussed, which are compatible with the observed requirement for certain hexoses unrelated to their capacity to serve as sources of energy.