Abstract
The present study aims to determine the roles of color and other surface properties (e.g., texture, brightness) in rapid natural scene categorization. Six natural scene categories were adopted as stimuli. Each stimulus was in three versions: color photographs, grayscales, and line-drawings, with a resolution of 320 * 240 pixels. The line-drawings were produced by trained artists through tracing contours in the color photographs (see Walther et al., 2011). Each image was flashed for 13 ms, and then followed by 80 ms of masks and 500 ms of a blank. The stimulus onset asynchrony (SOA) between the image and the mask was 13, 27, 40 or 200 ms at random. On each trial, participants were asked to make a choice among the six categories by pressing a corresponding key, during which EEGs were recorded. Each block included 84 trials. There were 12 blocks, for a total of 1008 trials. The behavioral results showed that accuracy was lower for grayscales than for line-drawings or color photographs in each SOA, meanwhile accuracies were higher for line-drawings than for color photographs in short SOAs of 13, 26, 40 ms. For correct trials, the ERP results revealed that there were no significant differences for amplitudes of posterior P1, N1, and P2 between grayscales and color photographs, but amplitudes of anterior P2, N2, and P3b were larger for color photographs than for grayscales. That is, color does not increase the difficulty of feature analysis, but contributes to decision making later. In addition, although amplitudes of posterior P1, N1, and anterior P2 were greater for grayscales than for line-drawings, conversely, amplitudes of posterior P2 and anterior N2 and P3b were larger for line-drawings than for grayscales. That is, although other surface properties increase the difficulty of feature analysis, they impair decision making for correct classification.
Meeting abstract presented at VSS 2015