OBJECTIVE Animal and in vitro studies indicate that a decrease in beta-cell insulin secretion, and thus a decrease in tonic alpha-cell inhibition by intraislet insulin, may be an important factor for the increase in glucagon secretion during hypoglycemia. However, in humans this role of decreased intraislet insulin is still unclear. METHODS We studied glucagon responses to hypoglycemia in 14 nondiabetic subjects on two separate occasions. On both occasions, insulin was infused from 0 to 120 min to induce hypoglycemia. On one occasion, somatostatin was infused from -60 to 60 min to suppress insulin secretion, so that the decrement in intraislet insulin during the final 60 min of hypoglycemia would be reduced. On the other occasion, subjects received an infusion of normal saline instead of the somatostatin. RESULTS During the 2nd h of the insulin infusion, when somatostatin or saline was no longer being infused, plasma glucose ( approximately 2.6 mmol/l) and insulin levels ( approximately 570 pmol/l) were comparable in both sets of experiments (both P > 0.4). In the saline experiments, insulin secretion remained unchanged from baseline (-90 to -60 min) before insulin infusion and decreased from 1.20 +/- 0.12 to 0.16 +/- 0.04 pmol . kg(-1) . min(-1) during insulin infusion (P < 0.001). However, in the somatostatin experiments, insulin secretion decreased from 1.18 +/- 0.12 pmol . kg(-1) . min(-1) at baseline to 0.25 +/- 0.09 pmol . kg(-1) . min(-1) before insulin infusion so that it did not decrease further during insulin infusion (-0.12 +/- 0.10 pmol . kg(-1) . min(-1), P = 0.26) indicating the complete lack of a decrement in intraislet insulin during hypoglycemia. This was associated with approximately 30% lower plasma glucagon concentrations (109 +/- 7 vs. 136 +/- 9 pg/ml, P < 0.006) and increments in plasma glucagon above baseline (41 +/- 8 vs. 67 +/- 11 pg/ml, P < 0.008) during the last 15 min of the hypoglycemic clamp. In contrast, increases in plasma growth hormone were approximately 70% greater during hypoglycemia after somatostatin infusion (P < 0.007), suggesting that to some extent the increases in plasma glucagon might have reflected a rebound in glucagon secretion. CONCLUSIONS These results provide direct support for the intraislet insulin hypothesis in humans. However, the exact extent to which a decrement in intraislet insulin accounts for the glucagon responses to hypoglycemia remains to be established.