In order to assess hepatic glycogen stores in patients with noninsulin dependent diabetes mellitus (NIDDM) after a 3-day fast, the incremental glucose response to 1.0 mg iv glucagon (glucose area under the curve, glucoseAUC) was assessed in 19 obese diabetic subjects after an overnight (14 h) fast and again after a 3-day (64 h) fast. Results were compared to those of lean (n = 6) and obese (n = 15) nondiabetic subjects. During the fast, plasma glucose fell significantly in the lean (4.9 +/- 0.2 to 3.9 +/- 0.2 mmol/L), obese (5.1 +/- 0.1 to 4.2 +/- 0.2 mmol/L), and diabetic (14.7 +/- 0.7 to 10.3 +/- 1.0 mmol/L) subjects. However, in contrast to the fall in glucoseAUC observed in the lean (92.4 +/- 15.4 to 39.9 +/- 8.1 mmol min-1 L-1, P less than 0.02) and obese (64.4 +/- 11.1 to 48.4 +/- 9.4 mmol min-1 L-1) subjects, the glucoseAUC increased in diabetic subjects from 81.6 +/- 8.6 to 103.9 +/- 8.8 mmol min-1 L-1 during the fast, and was significantly greater than that of either the lean (P less than 0.001) or obese (P less than 0.001) nondiabetic subjects after the 64-h fast. Evidence that the glucose response to glucagon after a 64-h fast represents glycogenolysis and not gluconeogenesis was provided by studies in 10 additional subjects (5 obese nondiabetic subjects and 5 patients with NIDDM). Overall hepatic glucose output calculated from glucose kinetic data [( 3-3H]glucose) increased in diabetic and nondiabetic subjects during the first 30 min after glucagon administration and fell progressively thereafter. However, no increase in alanine gluconeogenesis (14C-alanine incorporation into glucose) was observed after glucagon administration in either subject group. The paradoxical accumulation of glycogen in the patients with NIDDM during the fast occurred despite basal rates of hepatic glucose output on the third day of the fast which were greater than those of obese nondiabetic subjects (9.0 +/- 1.2 vs. 5.6 +/- 0.5 mumol kg-1 min-1, P less than 0.05). A glycogen sparing action of increased gluconeogenesis is proposed as the explanation for the preservation of liver glycogen in patients with NIDDM.