Retinotopic Scaffolding of High-Level Vision

Functional specialization within high-level vision is reflected in the topographic organization of the ventral temporal cortex (VTC). The presence and consistent locations of small areas responding selectively to particular visual categories – such as faces and scenes – has led to proposals of innate domain-specific modules. However, such proposals do not easily explain other aspects of an apparently multi-scale topographic organization of high-level visual features in VTC. Computational models have recently accounted for the presence of domain-selective areas and other facets of topographic organization from a basic optimization process with local topographic pressures, such as locally constrained connectivity, but fail to account for the consistent location of category-selectivity. In the current work, we extend a recent computational model to demonstrate how this consistency may emerge from wiring constraints external to VTC, focusing on the role of retinotopically organized early visual representations. After training several random initializations of the model, we find consistent global topographic organization, with face- and scene-selectivity emerging on opposite ends of a medial-lateral gradient corresponding to eccentricity bias, similar to human VTC. As in human VTC, the eccentricity-biased topography persists across viewing sizes under sufficiently broad viewing bias distributions, suggesting that it is a learned bias for efficiently organizing representations proximal to the most useful inputs, rather than merely an explicit retinotopic code. Abolishing the retinotopic constraint abolishes topographic consistency, but not topographic organization. Our work suggests that the organization of high-level visual cortex may emerge from domain-relevant interactions between viewing biases and task demands with an innate retinotopic scaffold. More generally, we suggest that both local and global connectivity constraints interact with representation learning to produce mature cortical organization: local constraints pressure the system towards smooth organization, whereas long-range constraints encourage a consistent global layout.