A simple, accurate, and reproducible method of measuring recirculation in grafts during hemodialysis is essential for improving the efficiency of dialysis. In our studies, plasma samples for plasma urea nitrogen (PUN) were taken from the arterial line of the dialyzer at blood flows (A) of 200, 300, and 400 ml/min, preceded by a 5-min period of equilibration, and at 15 s and 2 min after turning the flow down to 100 ml/min (S), the latter serving as systemic samples. Recirculation was calculated as (S - A)/(S - V). Total blood flow (Qb) through the grafts was measured by color Doppler ultrasound. We found a significant, inverse relationship between recirculation and total flow through the graft at dialyzer Qb of 400 but not 300 or 200 ml/min. The magnitude and prevalence of recirculation was always greater when the 2 min sample was used as S compared to the 15 s sample and as dialyzer Qb increased. As a qualitative, urea-independent measure of recirculation, we assayed the appearance of mannitol in the arterial line in blood drawn 15 s after initiating a mannitol push into the venous line. Blood obtained just prior to the mannitol push was used as the zero blank. Thirteen of 18 patients had a measurable, but low, level of mannitol, 5 did not, and 2 had inconsistent results in studies done on separate days. We conclude that the majority of patients receiving chronic hemodialysis have a low degree of recirculation and that methods relying on urea must be suspected of exaggerating the true degree of recirculation.