Previous research demonstrated that if attention is paid to a supraliminally presented number, a subsequent quantitative estimate assimilates towards this number (the anchor effect). One explanation states that this effect is merely caused by the heightened accessibility level of the anchor value itself. Based on this numeric priming account and generalizing from subliminal priming studies, we expected a short-lived subliminal anchor effect. We presented participants subliminally with a low or high anchor value (10 or 90) and next they had to estimate the probability of an epidemic. Half of them were pressed to do this quickly. Only under time pressure, a significant anchor effect emerged. © 2005 Elsevier Inc. All rights reserved.