Implicit and explicit use of scripted constraints in lip-reading

Publisher Website
Google Scholar
Abstract
The interaction between typicality, abstraction and temporal order within scripts, and different contextual conditions, was specified in a lip-reading model of predictive sentence comprehension. The model suggests both an implicit and an explicit mode of script processing, emphasising either hierarchical or temporal representations within scripts. In each of five experiments, subjects lip-read 48 sentences from three scripts. Different levels of abstraction (basic vs low-level), typicality (typical vs atypical) and temporal order (early vs late) were always embedded in each sentence. To assess a baseline, Experiment 1 measured “pure” lip-reading in the absence of scripted context. In Experiment 2, the script header was present. Experiment 3 correctly or falsely primed the script activation along each organisational dimension. In Experiment 4, the subjects received correct or false script-scene information, either before or after sentence presentation. In Experiment 5, the subjects received either basic or specific scene context before sentence presentation. The results confirmed the predictions in that: (a) there were minor dimensional effects when the script heading was absent; (b) typicality and abstraction constituted the main organisers within scripts, and typical, basic-level information occurring late was always implicitly activated; (c) the temporal structure at atypical, low levels of abstraction was the only dimension that was explicitly activated (i.e. sensitive to false cueing); (d) pre-exposed correct script-scene context enhanced lip-reading relative to the other conditions; and (e) specific scene context optimises a vertical “upwards” activation for basic and typical representations. It was concluded that the predictive script model of lip-reading received strong support, and that a “weak” hierarchical view of scripts is compatible with implicit processing, whereas a “strong” temporal view is compatible with explicit script processing.

This publication has 39 references indexed in Scilit: