Abstract
In current theories of scene perception, the construct of scene gist plays a critical role in guiding attention, facilitating object recognition, and influencing memory. Yet an important and frequently overlooked issue in research and theory on scene gist is the relative contributions of central versus peripheral vision in its acquisition. The current study investigated this issue, while testing how well the conclusions of Larson and Loschky (2009) generalize to more realistic viewing conditions. Larson and Loschky used 27° x 27° images (i.e., maximum retinal eccentricity = 13.5°--near peripheral vision), whereas the current study used a 180° (H) x 40° (V) screen (i.e., maximum eccentricity = 90°--the extreme periphery). We similarly used the Window/Scotoma paradigm, comparing perception of full scenes centered at fixation versus scenes viewed through a “Window” (circular region centered at fixation, with no imagery outside the circle), or a “Scotoma” (the inverse--a blank circular region centered at fixation, with normal imagery outside the circle). Window radii varied from 1-24°, Scotoma radii varied from 10-70°, with both compared to the full image. We flashed images for 33 ms and ensured central fixation using an eyetracker. Stimuli were 512 full color panoramic photographs in 8 basic level scene categories. The results confirmed the main findings of Larson and Loschky (2009) but more dramatically. First, peripheral vision was more important for scene gist than central vision. Removing all information ≤ 30° eccentricity was equal to viewing the entire image, but showing only information within the central 5° eccentricity was significantly worse. Second, although central vision was less important, it was far more efficient than peripheral vision in terms of scene gist extracted per image pixel viewed. Window and Scotoma regression slopes for non-asymptotic proportion accuracy as a function of proportion of image shown were 9.55 versus 1.05, respectively.
Meeting abstract presented at VSS 2015