In 2017, Di Paolo, Buhrmann, and Barandiarán proposed a list of criteria that post-cognitivist theories of learning should fulfill. In this article, we review the ecological theory of direct learning. We argue that this theory fulfills most of the criteria put forward by Di Paolo et al. and that its tools and concepts can be useful to other post-cognitivist theories of learning. Direct learning holds that improvements with practice are driven by information for learning that can be found in the dynamic organism-environment interaction. The theory formally describes information for learning as a vector field that spans a space with all the perception-action couplings that may be used to perform an action. Being located at a point of such a space means using a specific perception-action coupling. Changes in perception-action couplings due to learning can be represented as paths across the space, and can be explained with the vector field of information for learning. Previous research on direct learning considered actions that were best understood with single perception-action couplings. To conclude the article, and inspired by the criteria of Di Paolo et al., we discuss an extension of the theory to actions that are best understood with multiple perception-action couplings.