Abstract
After viewing a picture of a scene, people remember having seen a wider-angle view than was originally presented. This phenomenon, known as "boundary extension" (Intraub & Richardson, 1989), is very robust across viewing contexts, but few studies have attempted to quantify the magnitude of this effect. In this study, we investigate the magnitude of boundary extension produced by images at various levels of zoom, and examine the roles of foreground objects and background textures in this effect. We ran two boundary extension experiments using two different image sets. Training sets were created by cropping the images so that the central object would fill a particular proportion of the image height (ranging from 45% to 90%). Participants viewed a stream of training images with a particular level of zoom and were asked to remember the images in as much detail as possible. At the end of the sequence, participants were shown the same images again in an interactive window, and they were asked to zoom in or out on each test image to recreate the view which they had previously seen. We obtained a consistent boundary extension effect: across image sets and conditions, people generated wider-angle views in which the central object was about 5% smaller than it had been in the original image. We ran another experiment in which, rather than zooming the entire image to produce the remembered view, participants were asked to manipulate the central object or the background independently. When asked to resize the central object, participants mimicked boundary extension by making the central object about 6% smaller than in the original. However, they showed the opposite effect when manipulating the background: they chose a view in which the background texture was about 6% larger than it had been in the original.
Meeting abstract presented at VSS 2012