The magnitude of vacuum applied to the teat end can have a major effect on milking characteristics. While milking vacuum is usually measured in the milk pipeline, the teat-end vacuum during milk flow depends on the configuration of the milking unit. The objective was to establish the effect of teat-end vacuum, recorded during flow simulation, on actual milking time, milk yield, and both mean and peak milk-flow rates. Four configurations of milking units were set up to give vacuum levels of 35, 38, 40 and 42 kPa at the apex of an artificial teat during simulated milking. The experiment involved a latin square design with four groups of Friesian cows (14/group), four 2-day periods and four treatments (vacuum level). Altering the vacuum level had no significant effect on milk yield. There were no differences in milking characteristics between vacuum levels of 38 and 40 kPa. A vacuum level of 42 kPa gave a shorter milking time (P < 0.001), higher average milk-flow rate (P < 0.01) and higher peak milk-flow rate (P < 0.001) than the three lower vacuum levels. Milking time was significantly longer (P < 0.001) and peak milk-flow rate lower (P < 0.001) with a vacuum of 35 kPa compared to other vacuum levels.