GENERATING PREDICTIVE INFERENCES WHILE VIEWING A MOVIE

This study investigated conditions that enable viewers to predict futu re events while viewing a movie. Predictions about future events can b e supported by visual, auditory, and discourse information. These info rmation sources serve to emphasize or introduce the individually neces sary, and jointly sufficient, conditions that support predictive infer ence processes. They may also introduce new, anomalous information tha t can be understood in terms of future story events. In Experiment 1, viewers were instructed to generate predictions while watching the Jam es Bond movie, Moonraker (Broccoli & Gilbert, 1979). The presence of s upport through visual and discourse modes increased the likelihood tha t participants would generate a specific prediction. Furthermore, the likelihood of generating a given prediction increased with multiple so urces of visual and discourse modes of support. In Experiment 2, parti cipants provided think-aloud protocols at different locations in the F ilm, but were not instructed to generate predictions. As in Experiment 1, viewers made predictions when the sources were available, even tho ugh they were not specifically instructed to do so. Furthermore, the l ikelihood that viewers in Experiment 2 would generate the same predict ions as in Experiment 1 also increased as a function of the amount of support. The data are interpreted in terms of pragmatic and causal fac tors in comprehension.