© 2020, Imprint Academic. All rights reserved. This paper addresses the cognitive basis of anticipatory action. It does so by taking up what we call the acuity problem: the problem of explaining how skilled action seems, on the one hand, to be executed and unfold automatically and reflexively and, on the other hand, to involve anticipation of context-sensitive and constantly changing conditions in performance. The acuity problem invites two contemporary forms of reply, which we label non-inferential enactivism and Helmholtzian inference, respectively. We advance a third avenue for replying to the acuity problem, which takes active inference under the free energy principle as its theoretical starting point. This third way is, we contend, preferable to the other two across a number of important theoretical dimensions.