Abstract
Knock-out is a new form of visual masking, distinguished by robustness to temporal onset and considerable indifference to the visual appearance of the mask (Cramer & Rensink, VSS, 2012). Indeed, some localized masks are as effective as their larger-scale counterparts. We investigate here the extent to which knock-out is sensitive to grouping.
As before, observers detected a change in an alternating display containing an array of line segments, one of which changed orientation by 45° on half the trials. The duration of each display was 60 ms and the inter-stimulus interval (ISI) was 420 ms. When a mask was present, it appeared at the location of each item for 100 ms during the ISI, with a stimulus onset asynchrony (SOA) of 220 ms. Twelve observers were tested for each type of mask. Masks varied along overall spatial extent, and ease with which their constituents could form separate groups.
Results showed that knock-out depended on spatial extent: masks that were bounded (i.e., terminated before they reached the edge of the screen) had a greater effect than those that were unbounded (i.e., appeared to extend off-screen). Grouping was also important: if the elements of an extended mask could form a group that was was bounded, substantial impairment resulted. Interestingly, this impairment did not rely on the elements being in one-to-one correspondence with the targets: performance suffered even when mask elements did not overlap any targets.
These results suggest that the masking responsible for knock-out involves at least two processes: one that acts locally, and one that involves larger-scale structures that are largely irrelevant if not part of the same "framework" or "region" of the scene. They also suggest that knock-out itself is a useful way to explore the nature of these larger-scale structures, and the kinds of grouping processes involved.
Meeting abstract presented at VSS 2013