Abstract
We encode the visual world retinotopically, imposing a spatial reference frame on visual information processing. However, models of brain organization generally assume that retinotopic coding is replaced by abstract, amodal codes as information propagates through from visual to memory systems. This raises a puzzle for constructive accounts of visual memory: how can mnemonic and visual information interact if they are represented in different reference frames? To address this question, participants (N=15) underwent population receptive field (pRF) mapping during fMRI. We observed retinotopic coding throughout the visual system, including the scene perception areas (occipital place area (OPA) and parahippocampal place area (PPA)). Critically, consistent with prior work, we observed robust and reliable pRFs just beyond the anterior edge of visually-responsive cortex in high-level areas previously considered amodal. A large proportion of these high-level pRFs were located immediately anterior to scene-selective visual areas OPA and PPA, in memory-responsive cortex (Steel et al, 2022). We characterized these anterior pRFs, by localizing each participant’s OPA, PPA, and place memory areas (lateral (LPMA) and ventral (VPMA)) (Steel et al, 2022). Unlike visual areas OPA and PPA that contained almost exclusively prototypical positive pRFs, we observed a striking inversion of pRF amplitude in LPMA and VPMA, such that they exhibited spatially-selective negative BOLD responses. The visual field representation of negative pRFs in mnemonic areas closely matched their perceptual counterparts’, suggesting a common reference frame between perceptual and mnemonic regions. Finally, during a visual memory task, trial-wise activity of the positive and negative pRFs within the perceptual and memory areas was negatively correlated, suggesting a competitive push-pull dynamic between these neural systems. These results suggest that retinotopic coding, a fundamental organizing principle of visual cortex, persists in high-level, mnemonic cortex previously considered amodal. This shared code may provide a robust communication system aligning these neural systems.