The effect of site of starch digestion on glucose metabolism in dairy cows was studied. Four multiparous Israeli-Holstein cows in mid lactation were used in a 4 ✕ 4 Latin-square design. Average body weight of cows was 580 ± 38 kg, and average milk yield was 28 ± 3 kg/day. The cows were fitted with ruminal cannula and flexible T-cannulae in abomasum and ileum. Treatments were as follows : CON (control) : water was infused to the rumen. SR (starch-rumen) : 1·5 kg/day of maize starch solution was infused into the rumen. SA (starch-abomasum) : 1·5 kg/day of maize starch solution was infused into the abomasum. SCA (starch-casein-abomasum) : 500 g/day sodium caseinate and 1·5 kg/day of maize starch solution was infused into the abomasum. Total intake of dry matter (DM), was similar in all treatments and averaged 19·9 kg/day. Total non-structural carbohydrate (TNC) intake averaged in 6·8 kg/day. The average TNC digested in the rumen was 4·95 kg/day for CON and SR cows and 3·34 kg/day for the SA and the SCA cows. The average TNC digestion in the small intestine was 1·18 kg/day for CON and SR cows and 2·41 kg/day for the SA and SCA cows. TNC digestibility in the small intestine was highest for the SCA cows at 0·83 as compared with other treatments. Concentrations of plasma glucose and insulin were similar between treatments. No difference between treatments in total volatile fatty acid (VFA) concentration in ruminal fluids was observed. However, propionate proportion in total VFA was higher in the SR cows than in other treatments ( P < 0·04). Milk yield and composition were not affected by treatments in the present study. It was concluded that the amount of dietary protein in the small intestine has a considerable effect on TNC digestibility. Under conditions of high milk production and high rumen-by-pass TNC flow, efficiency of TNC utilization might be greater since TNC is digested in the small intestine rather than in the rumen.