Abstract
Evidence exists suggesting that our ability to integrate local features to form a coherent percept of a global pattern is degraded in peripheral vision. If this deficiency in feature integration generalizes to letter identification, then we would expect that more features are required for identifying letters in peripheral than in foveal vision. Because crowding has been attributed to an erroneous feature integration process, we would also expect that more features are required for identifying crowded letters than for single letters. To test our predictions, we used a set of 26 lowercase letters that were constructed of Gaussian patches, with each patch representing a “letter feature” and could be individually turned on or off. We measured the accuracy of identifying such letters when they were presented singly, or when flanked by two nearby letters, at the fovea and 10° lower field for three observers. The percentage of patches within the target and its flanking letters was varied from trial to trial, and ranged between 20 and 100%. Letter sizes were 1, 2 and 4× the acuity-threshold for a given condition. As expected, accuracy of letter identification increased with the percentage of patches present in the letters. The threshold number of patches that yielded 50% correct of letter identification decreased for larger letters. The drop was larger for identifying single than for crowded letters; consistent with the explanation that crowding is due to erroneous feature integration. However, contrary to our prediction, for any given nominal letter size, the threshold numbers of patches required to identify letters (single or crowded) were virtually identical at the fovea and 10° eccentricity. This finding suggests that as long as the difference in visual resolution is taken into account, feature integration for letter identification can be as good in the periphery as in the fovea.
Supported by NIH grant R01-EY12810.