Investigative Ophthalmology & Visual Science, 37:213. 1996. [ARVO 1996; Ft. Lauderdale, FL.]

Purpose. To determine the nature of the representations formed when viewing images of real-world scenes, and to determine the role of attention in building and maintaining these representations.

Methods. We employed a "flicker" technique, in which an original and a modified image (each of duration 240 ms) continually alternated, with a blank field (duration 80 ms) between each display. Images were all of real-world scenes. One of three kinds of change (appearance/disappearance, color, or translation) was made to an object or region in each scene. Changes were large and easily seen under normal conditions. Subjects viewed the flicker display, and pressed a key when they noticed the change.

Results. Changes in scenes were extremely difficult to notice under flicker conditions: reaction times for some scenes were as high as 20 seconds, even though the changes were large, continually repeated, and expected by the subjects. Verbal cues improved performance considerably, showing that the effect was not due to reduced visibility. It was also easy to spot changes in those objects mentioned in brief verbal descriptions of the scene.

Conclusions. In order to perceive that an item is changing, attention must be directed to it while the change is taking place. Attention is not only drawn by localized motion signals, but can also be attracted on the basis of high-level "interest". Consequently, the representation of a dynamic scene is accurate only in those aspects of greatest importance to the observer.