Abstract
Movies are ubiquitous and rapidly understood, but how does this occur? Voluminous reading research has investigated eye-movement/comprehension relationships, but do these relationships hold for film comprehension? We hypothesized film viewers' narrative event models would guide their attention while watching films. To test this, we manipulated the presence/absence of prior film context and measured resulting differences in film comprehension and eye-movements. We presented participants with one of two versions of the opening scene of Orson Welles' (1958) "Touch of Evil." In the Context condition, the clip opens with a bomb placed in a car trunk, the unknowing owners driving the car down the street, a couple (the protagonists) and numerous other people walking by the car on the street, and then the couple kissing with the car off-screen. The No-context condition was the same, but did not show the bomb placed in the car. The critical 3-minute portion of the clip was identical for both conditions, which differed only in seeing the bomb. In Exp 1, participants watched one of the two clips and then were asked what would happen next. We hypothesized that only Context condition viewers would mention an explosion. The results confirmed this hypothesis, with Context condition viewers far more likely to predict an explosion (76.2%) than the No-context viewers (8.3%), thus establishing a clear comprehension difference between the viewing conditions. Exp 2 was the same, but included eye-tracking. We hypothesized that viewers in the Context condition would fixate bomb-relevant details (the car and its trunk when on-screen) more than the No-context condition, showing the influence of film comprehension on eye-movements. Preliminary qualitative analyses support the predicted eye-movement differences between conditions, which are currently being quantified using dynamic region of interest analyses. Overall, preliminary results indicate that film viewers' comprehension does indeed influence their attention and eye-movements.
Meeting abstract presented at VSS 2014