Abstract
A widely reported scene perception phenomenon, boundary extension (BE), has been shown in visual memory reports for subjects recently presented an image or series of images. BE is characterized as a bias to consistently remember images as being more wide-angle than what was actually seen, possibly due to a source-monitoring error in which expectations based on scene characteristics are remembered as being part of the original sensory view (Intraub, 2011, 2012). Previous studies on BE typically present a series of images to be remembered, and then test these memories by presenting the same images again, or closer or wider-angle versions, and having subjects respond if the images are the same, closer, or wider-angle. Given that the phenomenon of interest is measured by responses concerning the perceived viewing distance in the stimuli, it is important to give attention to the variability in the scenes being presented. In previous studies, depicted main objects often exhibit extreme diversity in real world size; for example, objects vary from a candlestick to a car in Intraub and Richardson (1989). This size difference determines the perceived viewing distance from the main object in each image. For this study, we report experiments using scene stimuli presented in this traditional paradigm featuring more rigorous control of potentially confounding image features. The scenes used here depict objects of similar retinal and schematic size, a consistent background gradient across images, and the same viewing distance for each image. Manipulations of the difference between close and wide-angle image versions and other variables were performed. The results suggest that BE may not be as consistent as previous literature would suggest, yielding data that indicates no extension occurs under certain conditions using depictions of scenes.
Meeting abstract presented at VSS 2013