Change blindness during multiple interactions with a single object

VSS 2006 Poster Abstract

In the course of repeated interactions with an object in an ongoing task, do multiple fixations on the object lead to enhanced representations in working memory? Subjects performed a task in a Virtual Reality with haptic feedback.

Subjects first placed a brick on one of four corners of a virtual tabletop. Then they were cued to pick up the brick again and move it to another corner. The correct corner depended on the brick; for instance "if the brick is tall, move it to the back of the table; if short move it to the left". These sorting instructions changed after each brick movement, and subjects moved each brick four times per trial. On some trials a saccade contingent change was made to the brick. 

Subjects missed 40% of changes, despite being told they would occur. Undetected changes were universally sorted by the old, pre-change feature, even though sorting required multiple fixations directly on the brick. Even when subjects detected changes, detection often occurred only after having made several erroneous sorting decisions. The rate of change detection did not depend on how many times the subject sorted a brick before it changed, suggesting that subjects did not enhance their internal representations of the brick when they refixated it. 

Our results suggest that visual information available in the scene is often disregarded in favor of information acquired early in a task and stored in working memory. This may be an efficient strategy for deployment of limited attentional resources in this task.

Poster (PDF 766K, Legal Paper size)

Subject eye tracking Movies (encoded in mp4, crosshair is subject's gaze, within a couple of degrees)

Return to publications

Return to Alan's vision page  

(c) 2006 Alan Robinson (robinsoncogsci.ucsd.edu)