LOCAL AND GLOBAL VISUAL PROCESSING IN THE MOUSE BRAIN: FROM RECEPTIVE FIELDS TO POPULATION CODES

Every day, we interact with thousands of objects, each composed of local features, such as edges and textures, and global features, including overall shape and structure. Object recognition relies on both the processing of fine details and their integration into coherent representations, referred to as local and global processing. Along the ventral visual stream, information flows from the primary visual cortex (V1) to higher-order visual areas (HVAs), where receptive field (RF) size increases. However, it remains unclear whether combining information from neurons with spatially distributed RFs enhances or impairs object discrimination. We addressed this question by analyzing two-photon calcium imaging data recorded simultaneously from V1 and HVAs (LM, AL, RL) while head-fixed mice passively viewed 3D moving objects undergoing transformations that modulated local and global features. Using a linear classifier, we decoded object identity from neuronal subpopulations with varying spatial distributions of RF centers. In V1, LM, and RL, object discriminability decreased as the spatial spread of RFs increased, whereas AL maintained stable decoding performance across spatial scales. To determine whether this scale-dependence could be due to local connectivity or feedback connections, we generated a control model. Object stimuli were convolved with V1-like filters obtained through Independent Component Analysis to generate simulated neural responses. The model reproduced the decreased discriminability with increasing spatial spread, suggesting that stimulus statistics partially account for the observed effects. Our findings reveal that different visual areas employ distinct strategies for integrating spatial information during object recognition, shedding light on computational principles underlying hierarchical visual processing.