This paper offers an analysis of the emergence of non-contrastive segments in Optimality Theory in terms of comparative markedness. Glottal stop insertion in German serves as an example case to show that the standardly assumed theory of markedness constraints does not provide a satisfactory account of non-structure preserving phonology. Current analyses of German glottal stop insertion do not conform to the Richness of the Base Hypothesis (RotB): The segment is allowed in some environment in satisfaction of a contextual markedness constraint, but can surface as well in other positions than the loci of insertion as a correspondent of a segment in a hypothetical input. Comparative markedness captures the split nature of emergent non-contrastive segments: They are too marked to be part of the lexical inventory, but at the same time the least marked option in response to some structural requirements on surface forms.