The role of diagnostic color in 3 dimensional scenes

Abstract

Previous research (Oliva & Schyns, 2000) suggested that when a feature (e.g. color) is diagnostic of the meaning of a scene image (e.g. forests are green), altering this feature impairs recognition. In the present study, we investigated how mandatory the processing of a semantic diagnostic feature is by studying the role that diagnostic colors play during the perception of volume in real-world scenes. Stereoscopic photographs of natural and man-made scenes, with different color diagnosticity were presented to participants for a brief duration (200 msec). Each image was shown in a normal color version, in an abnormal (or false) color version and in a grayscale version. Two sets of stereoscopic scene pictures were created: a correct 3d image (as in normal stereoscopic viewing condition) and a flat 3d image (the right image was generated by displacing the left image a constant number of pixels, producing a flat image out of the plane of the monitor). Both sets of images had the same disparity on average over the whole image. The task was to discriminate which image portrayed 3d correctly. Results indicated that the discrimination was impaired in the abnormal color condition (d' ∼ 1.3) compared to the normal color and grayscale versions (1.7 and 1.8, respectively). This effect was more pronounced when color was a strong diagnostic feature of the image (e.g. natural scenes). This study suggests that a semantic diagnostic feature may interfere with a perceptual discrimination task (such as volume estimation).