We investigated the variability of quantitative duplex Doppler measurements of portal flow. Measurements were validated in vitro using a flow phantom. The measured flow Q (ml/min), is related to the actual phantom output P (ml/min) according to the following formula: Q = 1.08 (P + 44) (r = 0.998). To estimate inter- and intra-observer variance, 38 subjects without portal hypertension were examined in two groups. Two observers examined the first group of subjects (n = 19), from a routine daily ultrasound schedule. Significant differences were found in mean +/- S.D. portal flow (692 +/- 182 ml/min vs. 613 +/- 185 ml/min, p = 0.04) and mean +/- S.D. velocity (15.3 +/- 3.9 cm/s vs. 13.2 +/- 2.6 cm/s, p = 0.01). The combined inter- and intra-observer coefficient of variation (S.D.) was 24% (158 ml/min), 9% (0.92 mm) and 24% (3.4 cm/s) for portal flow, diameter and velocity respectively. Non-systematic components of variance were the largest. Patient characteristics, age, sex, height, weight and body surface area did not influence measurement variations. In the second group of healthy volunteers (n = 19), where variance in measurements over 3 consecutive days was comparable to the combined variance in the first group, the non-systematic variance component was also the largest. We conclude that quantitative duplex Doppler measurements of portal venous flow are mainly subject to non-systematic variability. A coefficient of variation of 24% can be expected in diagnostic measurements in a single patient. Examination by a single observer is advisable. The value of this technique lies in the analysis of pathophysiological mechanisms in portal flow changes in large groups of subjects.