Bovine crystalline insulin, mixed with an absorption enhancer, was loaded by hand into gelatin capsules, which were then coated with an azopolymer designed to deliver the insulin in the upper colon. In 34 experiments with 14 pancreatectomized mongrel dogs of both sexes, the coated capsules were administered orally after a pre-dose period of 1 h. The dogs had cannulae in the portal vein, hepatic vein and femoral artery and Doppler flow probes on the portal vein and hepatic artery. Insulin and food were withdrawn the day before an experiment. Responses measured were plasma glucose, plasma insulin, hepatic glucose production rate, hepatic plasma flow rate and plasma glucagon-like immunoactivity (GLI). Control experiments, with capsules without insulin, produced small changes from 'pre-dose' values. Insulin-containing capsules, without the azopolymer coating, resulted in some early changes consistent with upper gastrointestinal absorption. Single oral doses (66 to 400 nmol/kg) of insulin in completely coated capsules produced peaks of portal plasma insulin and transient decreases in plasma glucose, hepatic glucose production, hepatic plasma flow and plasma GLI. The changes usually began 1·5–2 h after administration of a single dose, and lasted for up to 3 h, but were not significantly related to the dose of insulin. Multiple oral doses of insulin, given at 1·5-h intervals, resulted in multiple peaks of plasma insulin, a continuing dose-dependent fall in plasma glucose to near-euglycaemia with the highest dose, and profound decreases in hepatic glucose production and plasma GLI. These data demonstrate that insulin absorbed from the gastrointestinal tract causes changes in glucose metabolism in the diabetic dog that are consistent with the action of insulin primarily on the liver and that repeated oral doses are necessary to correct the hyperglycaemia. Journal of Endocrinology (1991) 131, 267–278