Name: Eric Stearman

Dissertation Proposal Defense Meeting
Date: Wednesday, January 31, 2018
Time: 2:00pm
Location: Cognitive Ergonomics Laboratory (Coon room 252)
 
Advisor:
Professor Frank Durso, Ph.D. (Georgia Tech)
 
Dissertation Committee Members:
Professor Jamie Gorman, Ph.D. (Georgia Tech)

Professor Gabriel Radvansky, Ph.D. (Notre Dame)
Professor Paul Verhaeghen, Ph.D. (Georgia Tech)

Professor Mark Wheeler, Ph.D. (Georgia Tech)
 
Title: Situation Comprehension of Visual Scenes

Executive Summary

The current report seeks to understand how connectivity and domain knowledge influence a person’s ability to comprehend situations depicted using visual scenes.  The current report draws from the bodies of research of both language comprehension and situation awareness in defining the mechanisms and processes used in comprehension (see Durso, Rawson, & Girotto, 2007).  Three levels of representation are discussed, the surface form, the text-base/event-base, and the situation model.  Surface form representations refer to specific text/objects and their relationships.  The text-base/event-base refers to the semantic information conveyed, and the situation model uses knowledge structures stored in long-term memory to frame the situation.  The concepts of coherence and connectivity are discussed.  Coherence is thought to be achieved when an author’s intended meaning, the text itself, and a person’s interpretation of the text all align (Graesser, Singer, & Trabasso, 1994).  Connectivity can take place in different ways.  Propositional arguments offer one way in which connectivity can be achieved and lead to the development of propositional networks (Kintsch, 1988).  The author can provide a connected description using Samet and Shank’s (1984) dimensions; elaboration, coordination, and contiguity.  Or readers may make connections using Zwaan and Radvansky’s (1998) dimensions; entities, motivation, time, space, and causation.  Mental representations and the role of construction-integration in comprehension is discussed (Kintsch, 1988; Zwaan and Radvansky, 1998) in terms of the formation of propositional networks and situation models.  The role of domain knowledge in comprehension is also discussed and how it relates to the formation of situation models. 

The study proposed in the current report uses series of visual scenes of American football depicted using Xs and Os to examine the impact of connectivity and domain knowledge on comprehension of the scenes.  Series with high connectivity depict a single play unfolding, while series with low connectivity use scenes from various plays in the same series.  Domain knowledge is measured using both a self-reported measure of domain knowledge and a test designed to assess domain knowledge of various aspects of American football.  Processing time and change detection will be recorded. 

Results obtained from pilot data are discussed.  Processing time seems to decrease across scenes in the high connectivity condition while remaining constant across scenes in the low connectivity condition. For change detection, participants in the high connectivity condition were able to detect proposition and situation changes equally as well.  However, in the low connectivity condition, situation changes were detected better than proposition changes.  Domain knowledge seems to be predicative for detecting situation changes but not proposition changes.