Discourse analysis is a well-established method in the humanities to analyze historical trends and knowledge figurations that are inscribed in the texts of an era. While these trends are manifested as linguistic nuances in a variety of ways, the actual discourse remains a rather abstract concept. The goal of our paper is to develop a visual representation of such a discourse. We present TexTiles, a Visual Analytics framework that allows domain specialists to visually trace historical discourses in large text corpora. The prototype allows scholars to curate a customizable corpus, analyze keywords in context, and explore the network of these context words within the corpus. Allowing for a hybrid approach between close and distant reading practices, we demonstrate the utility of our application based on a case study on the discovery of the unconscious and the mechanics of repression in the long 19th century.