CapiFlow (CF), a new fully computerized system for the measurement of capillary blood velocity (CBV) was compared to manual frame by frame analysis (a) in a model system, and (b) in finger nailfold capillaries recorded on video tape. In the model the overall agreement between the two methods was very good (figure 1), with no significant differences being noted between the two sets of results and the calculated velocities. However, when comparing frame by frame and CapiFlow directly, CapiFlow read on average 4.50 +/- 5.21% higher than frame by frame analysis (figure 2). The in vivo results obtained by the two methods showed similar dynamic changes although some differences between the overall mean CBVs were noted (capillary 1, manual 0.13 +/- 0.59 mm s-1 versus CF 0.12 +/- 0.02 mm s-1, (mean +/- SD), p = 0.354; capillary 2, manual 0.66 +/- 0.23 mm s-1 versus CF 0.47 +/- 0.09 mm s-1, p < 0.001; capillary 3, manual 2.53 +/- 0.73 mm s-1 versus CF 2.35 +/- 0.34 mm s-1, p = 0.062). Further analyses established the optimum settings of delta limit and cross correlation. Investigations into the effects of changes in window size, window distance or video settings on CBV results obtained by CapiFlow, indicated that only settings radically different from the optimum had a significant effect on the results obtained.