This paper uses revealed preference methods to estimate demand for non-instrumental information in entertainment. We apply and extend the theory presented in Ely, Frankel, and Kamenica (2015, JPE) to conduct an empirical analysis that examines the effect of suspense and surprise on consumer demand. We first introduce alternative definitions of suspense and surprise using the theory of mutual information, and prove that suspense is in fact expected surprise. We then estimate the impact of suspense and surprise on television viewership using play-by-play and high-temporal frequency television ratings data from the National Basketball Association (NBA). Our primary results suggest that a one standard deviation increase in suspense increases viewership by 2.53% - 2.91%, while surprise has no impact. We also estimate within-game impacts of (i) absolute score differential and (ii) absolute score differential with respect to the point spread on viewership. These findings have important implications for entertainment media companies, including leagues and television broadcasters, and advertisers.