W e tested the hypothesis that defects in insulin stimulation of skeletal muscle blood flow, flow dispersion, and coupling between flow and glucose uptake contribute to insulin resistance of glucose uptake in non-insulin-dependent diabetes mellitus (NIDDM). We used positron emission tomography combined with [15O]H2O and [18F]-2-deoxy--glucose and a Bayesian iterative reconstruction algorithm to quantitate mean muscle blood flow, flow heterogeneity, and their relationship to glucose uptake under normoglycemic hyperinsulinemic conditions in 10 men with NIDDM (HbA1c 8.1+/-0.5%, age 43+/-2 yr, BMI 27.3+/-0.7 kg/m2) and in 7 matched normal men. In patients with NIDDM, rates of whole body (35+/-3 vs. 44+/-3 micromol/kg body weight.min, P < 0.05) and femoral muscle (71+/-6 vs. 96+/-7 micromol/kg muscle.min, P < 0.02) glucose uptake were significantly decreased. Insulin increased mean muscle blood flow similarly in both groups, from 1.9+/-0.3 to 2.8+/-0.4 ml/100 g muscle.min in the patients with NIDDM, P < 0.01, and from 2.3+/-0.3 to 3.0+/-0.3 ml/100 g muscle.min in the normal subjects, P < 0.02. Pixel-by-pixel analysis of flow images revealed marked spatial heterogeneity of blood flow. In both groups, insulin increased absolute but not relative dispersion of flow, and insulin-stimulated but not basal blood flow colocalized with glucose uptake. These data provide the first evidence for physiological flow heterogeneity in human skeletal muscle, and demonstrate that insulin increases absolute but not relative dispersion of flow. Furthermore, insulin redirects flow to areas where it stimulates glucose uptake. In patients with NIDDM, these novel actions of insulin are intact, implying that muscle insulin resistance can be attributed to impaired cellular glucose uptake.