This dissertation develops a framework for systematically and quantitatively analyzing narratives in mass political discourse. First, “open narratives” are defined as distinct from the traditional “closed” sense of a narrative, which has a beginning, middle, and end. Open narratives are co-constructed, rhythmic processes of selecting events and interpreting them as related instances of an established story-type. Drawing widely on narrative theory from different disciplines, it is shown how open narratives can be detected and represented in terms of ensembles of narrative phrases—charged labels and figures of speech that represent archetypal story elements. Second, an unsupervised natural language processing method to detect open narratives from news article data is introduced. Third, the method is used to quantitatively compare the volume of open narratives constructed between rightwing and center-left US news publishers between 2015-2024, finding a substantial advantage in the volume of narratives constructed by rightwing publishers for many of the most important issues in the US media ecosystem.