Population receptive field size in (un)crowding: to isolate or to combine?

Traditional models posit that visual processing is local and feedforward. In this vein, crowding is explained to be the result of pooling of target and flanker features in early visual areas. Specifically, it has been suggested that when the target and flankers fall within the same receptive field, their features are pooled together, resulting in an irreversible loss of information and degraded performance. A recent fMRI study investigated the relation between crowding strength and population receptive field (pRF) sizes (He et al., 2019). The authors reported that stronger crowding coincides with larger pRF sizes in V2, suggesting that pooling occurs in V2. Here, we investigated pRF size in crowding, no crowding and uncrowding, in which adding flankers leads to improvements in target discrimination, contrary to what traditional models of crowding predict. We replicated previous findings of increased pRF size in crowding as compared to no crowding, with pronounced differences throughout early visual areas (V1 to V4, not only V2). Surprisingly, uncrowding coincides with the smallest pRF size, even compared to the no crowding condition, in which behavioral performance is best. Thus, pRF size is modulated by global context and the relationship between pRF size and behavioral performance is non-monotonic. Our findings suggest that pRF size is modulated by top-down feedback, which goes against the classic, purely feedforward models of crowding. We propose that pRFs are the means by which the brain either combines or isolates the target and the flankers – reflected in behavioral performance as crowding and uncrowding, respectively. [We would like to thank our funding sources: Swiss National Science Foundation (176153, http://p3.snf.ch/Project-176153; NCCR Synapsy, project grant numbers 32003B_135679, 32003B_159780, 324730_192755 and CRSK-3_190185), Leenaards Foundation, ROGER DE SPOELBERCH and Partridge Foundations.]