The estimation of blood flow in arteries is important in the study of the (patho)physiology of the circulatory system. However, non-invasive techniques using pulsed Doppler ultrasound have so far shown potentially large errors. The aim of this study was to evaluate the precision and variability of a newly developed non-invasive ultrasonic system based on time domain processing, CVI (Philips). In vitro measurement of blood flow was conducted on a phantom with an elastic silicon tube. Both constant and pulsatile flow were tested at increasing flow levels, and pulsatile flow measurements were performed at three different pulse rates at each flow level. Furthermore, fixed and hand-held probes were compared. In vivo measurements of blood flow were conducted on the common carotid artery of seven volunteers. The intra- and interobserver variability was evaluated. In vitro measurements with fixed transducer and continuous flow showed an inaccuracy of 3.5%. Pulsatile flow showed no difference between measurements at the three different pulse rates (P = 0.69), and the overall inaccuracy of pulsatile measurements was 2.5%. Pulsatile flow was more accurately measured than continuous flow (P < 0.01). No difference was seen between fixed and hand-held transducers (P = 0.26). In vivo, the intra-observer variability was 25 ml min(-1) (CV 7.8%), and the interobserver variability was 35 ml min(-1) (CV 10.9%). CVI is a promising non-invasive new technique for flow measurements in arteries that gives correct data with good reproducibility in vitro, as well as in healthy common carotid arteries in vivo.