Abstract
Objects in our peripheral vision are difficult to identify in clutter. This deleterious integration, known as crowding, is strongest when objects are similar in appearance (e.g. all green) and weak when the target object is distinct (e.g. red amongst green). At present, it is not known how this selectivity applies to more complex objects composed of multiple parts. For instance, can a whole object be saved from crowding if only one of its parts differs from the flankers? We examined this using object parts that were both relevant and irrelevant to the required task.
Our objects were composed of three parts: an outer circle, a near-horizontal bisecting line, and a central dot (all matched for surface area and luminance). Observers judged the orientation of the target bisecting line (±10deg. of horizontal) at 10deg. eccentricity, either in isolation or with four abutting flankers. Objects were either wholly coloured (red or green) or selectively coloured with one of the three parts rendered red/green and the remainder in grey. The colouring of the target either matched or differed from the flankers.
In all conditions, performance was poor when target and flanker colours matched, and greatly improved when the target was differentially coloured. That is, colour differences reduced crowding regardless of whether the difference applied to the whole object or just individual parts. This was true even for the inner-dot and outer-circle parts, which were entirely uninformative for the orientation task. Because positional uncertainty was low in our experiments, the colour difference was not simply a cue to the target location. Rather, our results suggest that when any part of a target differs from surrounding objects, the whole target is saved from crowding. The release from crowding may thus involve high-level mechanisms that are tuned for objects rather than individual feature dimensions.
Meeting abstract presented at VSS 2013