Historically, debates over media violence have been a central focus of media research. Yet lacking from these debates is a meaningful discussion about the conceptualization of media violence. We argue that violence is not a monolithic construct, and is based on viewer perceptions of specific types of images and framing in media content. This idea has scholarly precedence: In 2002 and 2003, Potter and his colleagues proposed that perceptions of violence are formed as audience members make assessments about the relative levels of (in order) graphicness, realism, and justification for witnessed, on-screen violent actions. This article furthers this tri-partite conceptualization by using a binary-choice conjoint analysis to determine the role of each attribute in guiding audience perceptions of and preference for violent media in film and video games. For both media types, justification was the most central factor in shaping perceptions of violence, but realism was the most important predictor for the preference of violence.