Jan W. Kurzawski¹ (), Brenda Qiu², Denis G. Pelli¹, Jonathan Winawer¹; ¹New York University, ²University of Washington

Crowding is the failure to recognize an object in the presence of other objects. Crowding distance is the minimum object spacing for recognition, and varies widely across healthy adults. The biological bases of this variation are unknown. Previously (VSS 2021), we reported that in a group of 26 participants, the square of crowding distance (i.e., “crowding area”) correlates inversely with the surface area of hV4, but not V1, V2, or V3. Those results were exploratory and therefore potentially subject to selection bias. METHOD. Here, in a confirmatory study, we doubled the sample size and added reliability measures of crowding distance and map surface area. In 50 observers, we measured crowding on a letter recognition task and surface area from a retinotopic fMRI experiment. RESULTS. Reliability was high for both measures. Test-retest measures of the Bouma factor, b = (crowding distance)/eccentricity, correlated highly across two sessions (r = 0.94). Surface area estimates of visual maps, based on boundaries drawn independently by two researchers, were also highly correlated (r = 0.94, 0.88, 0.74, 0.73 for V1, V2, V3, hV4, respectively). These reliability checks confirm that we can accurately measure individual differences. We then replicated our prior analyses and confirmed that the square of crowding distance scales inversely with the surface area of hV4 (R2 = 0.43) but not V1, V2, V3 (R2 < 0). CONCLUSION. From the relationship between crowding distance and surface area, we estimate a recognition-threshold letter spacing of 1.5 mm on the hV4 map. That spacing is conserved across observers, despite the two-fold variations in crowding distance and visual map size. Our finding constrains the brain locus of crowding, a key step in a computational account of object recognition.

Acknowledgements: Support from NIH P30EY013079.