Abstract
In peripheral vision, similarity grouping influences what information can be accessed: observers can better identify a target when flankers share a feature distinct from that of the target. For example, one can more readily identify a target when flankers have opposite sign of contrast. One possible explanation of this "relief from crowding" is that feature pooling occurs only within perceptual groups, presumed to lead to better representation of targets that do not group with the flankers. The similarity effect, however, may not solely derive from better representation of an oddball target. A confounding factor remains: peripheral vision is characterized by location uncertainty. Any cue to target location would aid decision-making and identification. In most studies, the item with the distinctive feature is always the target, providing a 100% reliable cue. To understand how much of similarity effects can be attributed to target location cueing vs. improved target representation, we varied the reliability of the distinctive feature as a cue to the target. In the cue-reliable block, the distinctive feature is always associated with the target. In the cue-unreliable block, either the target or one of the flankers has the distinctive feature with equal probability. As expected, we observed improved performance with dissimilar compared to similar flankers in contrast polarity (M=20%,n=4,p=.002), orientation (M=13%,n=8,p=.001) and spatial frequency (M=16%,n=8,p=.003). Moreover, for contrast polarity, performance on distinctive target trials was significantly higher in cue-reliable blocks (72%) compared to cue-unreliable blocks (59%) (p=.002). However, such distinction was not found for orientation (reliable=66% vs. unreliable=59%, p=.38) or spatial frequency (reliable=67% vs. unreliable=66%, p=.86). Our results show that distinctive contrast polarity reduces target location uncertainty, aiding discriminability of oddball peripheral targets. Other similarity effects may instead derive more from better representation of an oddball target rather than from cueing effects.
Meeting abstract presented at VSS 2018