Abstract
Two-tone images, such as the Dalmatian Dog image by R. C. James, have edges whose source can be highly ambiguous. It is known that prior object knowledge can disambiguate these edges, and be critical to detecting an object in two-tone images. But can prior knowledge also influence how well the visual system can identify visual objects in these images?
Using a one-back task, we measured the ability of human subjects to identify faces in the same set of two-tone images before and after viewing a different set of full-color images of the same faces. Each image consisted of an unfamiliar face rendered against an outdoor background. The face identity, background, viewpoint and direction of illumination varied systematically across the images. No image was presented twice in two-tone and/or full-color form, so that the subjects had to generalize across the images in order to perform the task.
Prior to viewing the full-color images, the subjects were unable to reliably identify the faces in the two-tone images (mean correct one-back response, 5%). After viewing the full-color images, the subjects performed significantly better using the same set of two-tone images as before (71% correct; p ≪ 0.05, binomial proportions test). When the faces in the full-color images were different from those in the two-tone images, the performance showed little change before vs. after viewing the full-color images (7% vs. 9% correct; p [[gt]] 0.05). Thus, the improvement in the identification performance was attributable to gaining familiarity with specific faces.
Several brain regions near the superior temporal sulcus and in the frontal cortex showed significantly larger BOLD responses to two-tone images after viewing the corresponding full-color faces than before the viewing. No increased response was evident anywhere in the visual cortex using the same contrast upon viewing irrelevant full-color faces.
This work was supported by ONR grant N00014-05-1-0124 and NEI grant R01 EY015261 to DK and NIH grant T32 HD00715 to SKT. The 3T scanner at the University of Minnesota Center for Magnetic Resonance Research is supported by BTRR P41 008079 and by the MIND Institute. Office facilities were provided by the Center for Cognitive Sciences at the University of Minnesota.