This dissertation aims to highlight the theoretical importance and predictive power of audience interpretations within media effects research through an experiment comparing four variable sets -- message features, audience attributes, audience states, and audience interpretations -- in terms of their contribution to three commonly assessed outcomes of media exposure: attitudes, beliefs, and behavioral intentions. Within this study, interpretations refer to the meanings that audiences construct from media content (e.g., perceptions and evaluations of characters and behaviors). In other words, interpretations make up the "effective stimulus" within a viewer's mind. According to this definition, interpretation variables are conceptually distinct from message features (inherent elements within a media message), audience attributes (demographics and stable traits), and audience states (temporary conditions experienced during media exposure). The predictive power of these four variable sets was tested in the context of television depictions of cocaine use. As such, the specific outcomes of interest included attitudes about cocaine, cocaine effect expectancies, and intentions to use cocaine.

A total of 311 undergraduate students were randomly assigned to view one of three televisions episodes (The Wire, Girls, or Entourage) in which main characters were shown using cocaine. The episodes were chosen because they depict diverse portrayals of cocaine use in terms of key variables such as consequences, character status, and humor. Considering their differences in terms of message features, the treatment conditions represented the message variable set within the study. Variables for the other three sets were assessed via pre-test and post-test questionnaires. The specific factors included in each set were selected based on past findings suggesting that they were likely to influence the outcomes of focus.

The study's results revealed that interpretation variables were overwhelmingly and consistently the strongest predictors of all three outcome variables. Perhaps most notably, audience interpretations explained between 10% and 23% more variance than the treatment conditions to which participants were assigned. The other types of variables also were found to be useful predictors of the three outcomes. Findings related to these variable sets aligned with the existing media effects literature, which pointed to particular attributes (e.g., sensation seeking personality), states (e.g., emotional reactions), and message features (e.g., character status) likely to predict viewers' responses to television depictions of drug use. For example, findings related to the effect of the treatment condition reinforced the relevance of the message features of consequences, humor, and character status, and supported the premise of social cognitive theory. The three outcome measures varied across treatment conditions in expected ways -- with statistically significant differences observed for attitude and belief outcomes. Specifically, participants who viewed the Girls episode (featuring relatable characters, humor, and positive outcomes) reported more positive attitudes about cocaine and more positive effect expectancies than participants who viewed The Wire episode (featuring criminal characters, no humor, and extremely negative consequences).

By illustrating how distinct variable sets contribute to different types of media outcomes, this dissertation lays the groundwork for a new "phase" of media effects experiments that accounts for audience interpretations in addition to message features, audience attributes, and audience states. Importantly, research conducted during this interpretation phase would involve (1) more comprehensive applications of prominent media theories (e.g., social cognitive theory and priming), (2) experimental designs that account for participant interpretations, and (3) receiver-oriented approaches to content analysis. Based on the findings reported here and elsewhere in the literature, it is expected that such research would result in stronger predictive power, larger effect sizes, and most importantly, a more complex and complete understanding of the process of media influence.