Abstract
In crowding, target perception deteriorates in the presence of clutter. Crowding is usually explained by pooling models where higher level neurons pool features from both a target and "informationless" flanking elements. Here, we show that such models fail to explain a large body of findings on pattern recognition, thereby undermining the philosophy of this approach. For example, observers judged the offset of a vernier presented in peripheral vision. When the vernier was flanked by eight aligned verniers on each side, strong crowding occurred, as expected. Next, we presented the vernier and the flankers as in the previous condition and, in addition, an aligned vernier at the same location as the target vernier. Crowding did not increase as one might have expected from adding an "informationless" element. Quite to the contrary, crowding strongly decreased. We argue that crowding can be explained by pattern processing and grouping. The aligned vernier complements the two arrays of flanking verniers, by creating a regular pattern of equally spaced, identical elements. Since the aligned vernier groups with the flankers, the target vernier does not group with the flankers anymore, and crowding is weak. When the aligned vernier was longer than both the vernier and flankers, no reduction of crowding occurred because, as we argue, the length difference prohibits the aligned vernier from completing the pattern of the flankers. It is the "good" pattern that matters. When we presented only one flanker to the left and right of the vernier, crowding was as strong as with eight flankers. However, crowding remained strong when we added the aligned vernier. We show by computer simulations how pattern recognition and crowding can be explained by recurrent processing.
Meeting abstract presented at VSS 2015