Two studies are reported which investigated the ability of people to make accurate judgments of variability in performance. In one study, a laboratory experiment in which performance was manipulated by the experimenter, it was found that significant accuracy did exist even though it was clouded by the tendency to judge Ss with high mean scores as more consistent. In the other study, persons participating in a real-life bowling league were asked to make judgments of consistency about the players on their team. There was no trace of any appropriate sensitivity to differences in variability. The halo effect involving the judged consistency of the high average players was, however, quite strong. The importance of these findings for mathematical models of decision making is briefly discussed.