D-mannoheptulose (MH) administration induced a decreased serum insulin concentration in fed and starved mice, and a transient hyperglycaemia in fed mice, but not in starved ones. The liver glycogen concentration was decreased in starved controls and in fed mice treated with MH. Differences in the capacity for rapid hepatic glycogenolysis may have contributed to the different blood glucose responses in fed and starved mice. The hyperglycaemia in fed mice was unaffected by pre-treatment with L-leucine, or p-hydroxymercuribenzoate (PMB), but was abolished by pre-treatment with tolbutamide, and by post-treatment with insulin. Treatment of fed mice with MH before alloxan caused a marked "initial" hyperglycaemia but no second hyperglycaemia, and thus no development of alloxan diabetes. In starved mice injected with MH before alloxan there was an inhibition of the initial hyperglycaemia, but occurrence of a "second" hyperglycaemia, suggesting an absence of protection against the development of alloxan diabetes. The data show that alloxan diabetes may develop in the absence of an "initial" hyperglycaemia and a triphase blood glucose response. The hyperglycaemic action of MH in fed mice is believed to underlie the protection against alloxan toxicity.