Abstract
Visual crowding dramatically reduces the perception of word form and meaning in peripheral vision. Previous research demonstrates that crowded letters and shapes can be “demasked” when the same form is simultaneously presented at the fovea, provided the central and peripheral figures are similar in orientation and contrast (Geiger, Lettvin, Perception, 1985). The present study explored whether a briefly presented, but visible, identical or semantically related foveal prime word would “uncrowd” flanked peripheral words compared to an unrelated prime condition. We tested eighteen observers across three target eccentricities (fovea, 4°, and 8°) and three prime-target relationships (unrelated, identical, or semantically related). Using a dual-task design, participants first discriminated target words from non-words in a 2 AFC peripheral lexical decision task (LDT). In the second task, observers gave a letter-by-letter account of the target, providing a more detailed report of the deleterious effects of crowding. Across all eccentricities, LDT percent correct and discriminability (d′) improved modestly, but not significantly, on identical and semantically related trials compared to the unrelated prime condition. However, the letter-by-letter analysis revealed that identical and semantically related word trials reduced crowding and reaction times. Subjects correctly reported all letters of the crowded word more often in both the identical and semantically related conditions (F(2,34)= 15.6, p<0.01). Further, on correct LDT trials, reaction times were reduced for both identical and related word pairs across all eccentricities (F(2,34)= 12.2, p<0.01). Taken together, these results suggest that LDT does not provide the most sensitive measure of semantic influences on word crowding. More importantly, these results demonstrate that semantic priming at the fovea can reduce peripheral word crowding and facilitate reaction times. This adds to a growing body of research demonstrating that the resolution of vision in crowded scenes can be modified by learning and prior experience.
Supported by NIH grants T32-EY007043-31 and R01-EY012810.