Information-Theoretic Measures of Reader Enjoyment

Abstract

Two studies based on an information theory model of reader enjoyment investigated the role of syntactic and semantic unpredictability in determining readers' evaluations of journalistic prose. In each study, reader enjoyment ratings for a set of articles reporting a single news event were compared with cloze procedure results in which function-word and content-word responses were analyzed separately using entropy and cloze scoring techniques. Both studies revealed a statistically significant correlation between function-word predictability and reader enjoyment. In addition, a strong correlation between content-word unpredictability and reader enjoyment in one study supported the notion that readers prefer texts that are characterized by a high degree of semantic unpredictability.