Doch is one of the German discourse particles that has been widely discussed in the literature (a.o. Grosz, 2014; Egg and Zimmermann, 2012; Karagjosova, 2009). There are two main challenges of providing a meaning for it: first, doch can be prosodically realized either unstressed (roughly as you know) or stressed (roughly after all) with the two realizations providing related yet different meanings. While some accounts focus on only one of the two realizations (a.o. Rieser, 2015; Grosz, 2014), other accounts pursue a unified account for both realizations, treating the two as one lexical item (a.o. Rojas-Esponda, 2014; Egg and Zimmermann, 2012). Second, one has to account for its distribution. While the occurrence of doch in declaratives can straightforwardly be explained in most accounts, its discourse initial use and its occurrence in other sentence types is often absent from the discussion of doch. I follow the line of previous accounts that treat the two prosodic realizations as a single lexical item and propose a new unified account in which both the unstressed and the stressed realization of doch conventionally convey the speaker’s belief as well as her belief about the addressee’s belief at a past time, attributing the difference in meaning contribution of the two realizations to the contribution of stress.