Abstract
Crowding, the marked inability to identify shapes in peripheral vision when targets are flanked by other objects, has been widely studied. The following gross characteristics of the zone of crowding have been consistently replicated and there is growing consensus that these represent the hallmarks of crowding: (a) the spatial extent of crowding scales with eccentricity (b) the zone is elliptic with its major axis pointing towards the fovea and (c) an outward flanker is more effective at crowding than an inward flanker. In the current study, we extended these findings by undertaking detailed measurements of the crowding zone by using a single letter (‘E’) to flank a letter target (one of 10 Sloan letters) positioned at 10° in the inferior field. The flanker was placed along 12 radial directions centered on the target and at 5 flanker-target separations (1.25° to 6.25°), thus obtaining identification accuracy measures at 60 locations around the target. We found that the crowding zone is not elliptical, even though an elliptic fit can be used to provide summary statistics that are in agreement with classical findings. More interestingly, our maps (obtained from three subjects) reveal that the zone of crowding is not convex, with islands of ineffective flanking locations intruding into an otherwise crowding-inducing region. These occur mostly along the tangential direction and can sometimes lie very close to the target. There is also significant variability of the zone among subjects. Our results augment existing data sets on the crowding zone, which tested only a few cardinal directions (typically 4) or used a pair of symmetrically placed flankers (e.g. Toet & Levi, 1992) that could bias the shape toward the fovea. Most importantly our findings will help the development of computational models that can account for the shape of the zone (Nandy & Tjan, SfN 2008).