Abstract. Systemic hyperinsulinaemia induces vasodi latation in human skeletal muscle. This effect is gra dual in onset, and at low insulin levels not maximal until at least 3 h. To investigate whether the vasodi lator response to insulin results from a direct vascular effect, we infused insulin directly into the cannulated brachial artery (perfused forearm technique) in a total of 30 experiments in 20 healthy, lean, normotensive volunteers. Local, intra-arterial, infusion of insulin (180 min, O O m U dL " 1 forearm volumemin" 1, « = 15, forearm venous insulin concentration approximately 540pm olL -1) induced a gradual increase in forearm blood flow (FBF; venous occlusion plethysmography) from 1-86 ± 0*17 to 3*64 ± 0,6 4 m L d L " 1 min~l after 180 min (a n o v a P < 0-001), Percentage increases in FBF after 60, 120 and 180 min averaged 14*4 ± 5-9, 59*4 ± 25*5 and 124*6 ± 51*2% respectively. Fore arm glucose uptake increased from 0*24 ± 0-05 to a maximum of 1*98 ± 0-28 /.¿moldL~l min (P < 0 001). Furthermore, insulin infusion increased forearm lac tate release and potassium uptake. In 10 out of these 15 individuals, the forearm glucose uptake was further increased in a second, separate, repeat experiment with concomitant intra-arterial infusion of glucose 5% (O^mLdL™ 1 m in-1), resulting in forearm venous glucose concentrations of approximately 15 mmol L"*1. This combined infusion achieved a simi lar vasodilator response to the infusion of insulin alone. The individual vascular responses of the two paired experiments showed a strong correlation (r = 0*87, P < 0*01). In five subjects time and vehicle control experiments were performed, showing no changes in FB F or metabolism during the 180 min. W e conclude that the slow vasodilator response to insulin (as observed during systemic infusion) can, at least partly, be explained by a direct vascular effect of insulin. Insulin-mediated skeletal muscle glucose uptake precedes this effect, but seems not to be an im portant determinant of the vasodilator response to insulin.