A common cortical basis for variations in visual crowding

Abstract

AbstractPeripheral vision is limited by crowding, the disruptive effect of clutter on object recognition. Crowding varies markedly around the visual field, with e.g. stronger performance decrements in the upper vs. lower visual field. Crowding also changes object appearance – target and flanker objects appear more similar (assimilation) in some instances and dissimilar (repulsion) in others. Here we examined whether these performance and appearance effects co-vary, and in turn whether a common cortical factor could drive all of these effects. Participants judged the orientation of a target Gabor with and without flankers in 3 experiments. The first placed a flanker in either the ipsilateral or (the more cortically distant) contralateral hemifield. Although crowding was observed, flanker location had no effect. We next measured recognition at a range of eccentricities in the upper and lower field, observing that both threshold elevation (performance) and assimilative errors (appearance) were higher in the upper vs. lower field. Similarly, flankers on the radial axis around fixation produced high threshold elevation and assimilation, while tangential flankers gave lower elevation and repulsion errors. This common pattern of variations in performance and appearance is well described by a population-coding model of crowding that varies the weighted combination of target vs. flanker population responses. We further demonstrate that neither the cortical distance between elements nor receptive-field size variations can account for the observed variations. Instead, using a series of models we show that the common factor could bereceptive field overlap– the intermixing of the spatial distribution of target/flanker responses. That is, crowding is strong (with high threshold elevation and assimilation) when the degree of overlap in the spatial distribution of population responses is high, and reduced (with low threshold elevation and repulsion) when these responses are separable.