We investigated the mechanisms by which peripheral or portal insulin can independently alter liver glucose production. Isotopic ([3-3H]glucose) and arteriovenous difference methods were used in conscious overnight-fasted dogs. A pancreatic clamp (somatostatin plus basal insulin and basal glucagon infusions) was used to control the endocrine pancreas. After a 40-min basal period, a 180-min experimental period followed in which selective increases in peripheral (PERI group, n = 5) or portal-vein (PORT group, n = 5) insulin were induced. In control dogs (CONT group, n = 10), insulin was not increased. Glucagon levels were fixed in all studies, and basal euglycemia was maintained by peripheral glucose infusion in the two experimental groups. In the PERI group, arterial insulin rose from 36 +/- 12 to 120 +/- 12 pmol/l, while portal insulin was unaltered. In the PORT group, portal insulin rose from 108 +/- 42 to 192 +/- 42 pmol/l, while arterial insulin was unaltered. Neither arterial nor portal insulin changed from basal in the CONT group. With a selective rise in peripheral insulin, the net hepatic glucose output (NHGO; basal, 11.8 +/- 0.7 micromol x kg-1 x min-1) did not change initially (11.8 +/- 2.1 micromol x kg-1 x min-1, 30 min after the insulin increase), but eventually fell (P < 0.05 ) to 6.1 +/- 0.9 micromol x kg-1 x min-1 (last 30 min). With a selective rise in portal insulin, NHGO dropped quickly (P < 0.05) from 10.0 +/- 0.9 to 5.6 +/- 0.6 micromol x kg-1 x min-1 (30 min after the insulin increase) and eventually reached 3.1 +/- 1.1 micromol x kg-1 x min-1 (last 30 min). When insulin levels were not increased (CONT group), NHGO dropped progressively from 10.1 +/- 0.6 to 8.3 +/- 0.6 micromol x kg-1 x min-1 (last 30 min). Conclusions drawn from the net hepatic glucose balance data were confirmed by the tracer data. Net hepatic gluconeogenic substrate uptake (three carbon precursors) fell 2.0 micromol x kg-1 x min-1 in the PERI group, but rose 1.2 micromol x kg-1 x min-1 in the PORT group and 1.2 micromol x kg-1 x min-1 in the CONT group. A selective 84 pmol/l rise in arterial insulin was thus associated with a fall in NHGO of approximately 50%, which took 1 h to manifest. Conversely, a selective 84 pmol/l rise in portal insulin was associated with a 50% fall in NHGO, which occurred quickly (15 min). From the control data, it is evident that in either case approximately 30% of the fall in NHGO was due to a drift down in baseline and that 70% was due to the rise in insulin. In conclusion, an increment in portal insulin had a rapid inhibitory effect on NHGO, caused by the suppression of glycogenolysis, while an equal increment in arterial insulin produced an equally potent but slower effect that resulted from a small increase in hepatic sinusoidal insulin, from a suppression of gluconeogenic precursor uptake by the liver, and from a redirection of glycogenolytic carbon to lactate rather than glucose.