Today’s studies of networked discussions may be divided into theory-driven and data-driven, but both lines of research neglect the role of contextual knowledge in assessment of real-world public discourse. As scholars note, without context, data lose meaning and value; however, there is a striking vacuum of scholarly discussion on how to delineate the relevant context for network discussion studies, as well as what procedures of its description in academic publications should be employed. As a mediator between theories and data-driven results, context has a potential of eliminating the opposition between theory- and data-driven research designs. In an attempt to conceptualize context, we suggest to adapt the long-term experience of cognitive linguistics and critical discourse analysis for developing rigorous procedures of selection, assessment, and explicit description of relevant context(s). We bring attention to the paradox that, in online discussion studies, scholars extract sociologically relevant conclusions from the data of non-sociological nature (that is, either text or network structures), and argue it might be fruitful for selection of appropriate contextual background. After meta-reviewing the conceptual papers on online discussion research and using our own experience in such studies of over 7 years, we suggest three types of contexts for network discussions: cognitive, platform-technological, and media/communicative contexts – that need to be taken into account in network discussion studies.