Abstract Certain patterns of directional syncretism, in particular bidirectional syncretism, have been argued to necessitate the power of directional rules that create a form dependency between two cells in a paradigm. The existence of such patterns, divergent bidirectional syncretism in particular, has been claimed to be fatal for the approach to syncretism adopted in Distributed Morphology that holds that (directional) syncretism always involves a retreat to less specific and therefore less marked exponents. In this paper, we will demonstrate that this is not the case. Apparently challenging cases of directional syncretism can be adequately handled on the view that impoverishment may have two outcomes: deletion of features or insertion of contextually unmarked values. Once this view is adopted, bidirectionality is no longer a problem. Furthermore, we argue that this view allows us to still maintain the claim that syncretism is universally markedness decreasing if markedness is defined over insertion contexts rather than exponents.