The present study reviews problems in the political learning literature, including ambiguous causality and a lack of specificity in linking communication content to learning outcomes. As a partial solution, our study of media and discussion influence incorporates both manipulated and observed aspects of mass and interpersonal communication. Results indicate that beyond (and often more important than) experimental manipulations, selection processes in news use and variations in the content of political discussions within exposure conditions matter for political knowledge. However, findings vary in predictable ways depending on the form of knowledge—overall factual knowledge, issue-specific knowledge, or knowledge structure density. These results suggest that the process of political learning via communication is more complex than it is often treated empirically.