Abstract
Categorization and naming of a scene is faster for scene images presented in color than for otherwise identical images presented in greyscale. This especially holds for natural scenes, for which appropriate color information is deemed ‘diagnostic’. It has been proposed that color specifically facilitates the detection of the gist of such a scene. Therefore, we hypothesized that color information might impair change detection when only details (though obvious in retrospect) of an image are changed.
Participants viewed a set of scene-image pairs in a change-detection flicker paradigm. The images were presented in alternating fashion (200 ms presentation duration) interleaved by a 200 ms blank interval. The participants' task was to press a button as soon as they discovered which part of the scene had changed. Subsequently, they were asked to verbally describe the change. Each image pair was presented in both greyscale and in color, but participants never viewed the same scene twice.
The distribution of reaction times across participants and images was shifted between color and greyscale versions of the image pairs. Changes in greyscale image pairs were generally perceived faster than the same changes in color image pairs for both natural and manmade scenes. In addition, changes in manmade scenes appeared to be detected faster than those in natural scenes. However, this latter finding may bear solely upon the difference in complexity of images of both scene types.
The results suggest that indeed color information is especially important for the encoding of the gist of a scene. When color information is absent, a more detailed representation of the scene may be constructed, hence the superior change detection observed.
Supported by the Netherlands Organization for Scientific Research (NWO)