Identifying Media Frames and Frame Dynamics Within and Across Policy Issues Amber E. Boydstun, University of California, Davis Justin H. Gross, University of North Carolina, Chapel Hill Philip Resnik, University of Maryland, College Park Noah A. Smith, Carnegie Mellon University September 16, 2013 Abstract Framing is a central concept in political communication and a powerful political tool. Un- derstanding what frames are used to define specific issues and also what general patterns are evidenced by the evolution of frames over time is hugely important. It is also a serious challenge, thanks to the volume of text data, the dynamic nature of language, and the variance in appli- cable frames across issues (e.g., the ‘innocence’ frame of the death penalty debate is not used in discussing smoking bans). We describe a project that advances framing research methodology in two ways. First, we are developing a unified coding scheme for content analysis across issues, whereby issue-specific frames (e.g., innocence) are nested within high-level dimensions (or frame types ) that cross-cut issues (e.g., fairness); we are validating this coding scheme by applying it to news coverage of immigration, same-sex marriage, and smoking/tobacco in the United States over the course of the past twenty-three years. Second, we are developing methods for semi-automated and automated frame discovery aimed at both replicating manual coding and isolating patterns of frame evolution that might not be readily visible to human inspection. Our goal is to employ strategies heavily informed by existing work in natural language processing, but tailored to the specific needs and professional sensibilities of political communications scholars. 1 Introduction Framing—portraying an issue from one perspective to the necessary exclusion of alternative perspectives—is a central concept in political communication (see Schaffner and Sellers, 2009, Intro- duction for a nice overview, and the remainder of the edited volume for several illustrations). It is widely accepted that framing can have a significant influence on public attitudes toward important policy issues (e.g., Chong and Druckman, 2007; Nelson et al., 1997) and on the application of policy issues directly (e.g., Baumgartner et al., 2008). Understanding, for a given issue, what frames are used by politicians, the media, and the voting public to communicate about it, is an enormous chal- lenge, due to the dynamic and creative nature of language and the growing volume of data in which frames appear and develop over time. As engagement by citizens in the political discourse broadens via the widespread adoption of blogging, commenting, and other social media, scientific study of the political world requires reliable analysis of how issues are framed, ideally in real time. Yet the process by which a political scientist or communications scholar identifies the catalogue of frames in a political discourse about a particular issue (frame discovery) is complex and labor-intensive; so is the secondary process of coding instances of framing in text (frame analysis) in order to reveal patterns in frame usage. 1