The ability of readers to represent protagonists' changing emotions was explored. In Experiment 1, subjects read stories in a cumulative version (all the events biased the same emotion) or in a shifting version (the first part biased an emotion and the second part suggested another one). An emotional sentence (e.g. Ann felt proud of her decision) placed at two different loci was either consistent or inconsistent with the first part of the story. In the cumulative context, inconsistent targets were read slower in any locus. In the shifting context, inconsistent sentences were read slower in the first locus, and faster in the second locus, indicating that readers had updated the protagonist's emotion. In Experiments 2 and 3 subjects read versions of the stories with a first part that biased one of two alternative emotions (e.g. proud or guilty), followed by neutral sentences and then the target sentence. The consistency effect was obtained again, demonstrating that the representation of the initial emotion was highly accessible beyond the local context. The results indicate that readers built updatable mental models of protagonists' emotions.
Manuel de Vega (Wed,) studied this question.