This paper is concerned with the form and interpretation of the colon in German; for example, Ada hat zwei Äpfel gekauft: Sie hat einen Boskop und einen Elstar gekauft (‘Ada has bought two apples: she has bought a Belle de Boskoop and an Elstar’). First, I argue against the syntax-based approach by Bredel (2008, 2011), according to which the colon yields a dislocation of either the left-handed colon construction or the right-handed colon extension. Specifically, the approach cannot account for the intuition that the colon identifies the colon construction as an announcement that is satisfied by the extension. Second, I argue in favor of a lexicon-based alternative approach according to which the colon is a general lexical marker for discourse-structural subordination, namely, it marks the colon construction as subordinating the extension. This approach has the following advantages: (i) It accounts for the observation that the colon is compatible with subordinating discourse relations (e. g., elaboration and explanation), but incompatible with coordinating ones (e. g., narration, parallel, contrast). (ii) Following Jasinskaja & Karagjosova (2021), subordinating discourse units are defined by their communicative goal being incomplete without the subordinate unit. This derives the announcement effect of the colon for free. (iii) The lexicon-based approach can be implemented in terms of standard compositional semantics. This facilitates a systematic analysis of examples where the colon breaks into the incremental process of structure building; for example, Ben hat gekauft: einen Boskop und einen Elstar ‘Ben has bought: a Belle de Boskoop and an Elstar’. Two overarching results are noteworthy: the analysis suggests that formal semantics can be applied fruitfully to graphematics. In turn, semantic-pragmatic research can profit from integrating graphematics. Specifically, to date, no general lexical marker for subordination has been found; the given analysis suggests that this lexical gap is filled by the colon.