This research investigate how neurons in the macaque visual cortex respond to different types of imagery across the ventral stream. Using generative models, researchers compared neuronal preferences for abstract textures versus photorealistic objects in the V1, V4, and PIT areas. The study utilized a closed-loop optimization process, allowing individual neurons to guide the synthesis of images that maximize their firing rates. Results showed that while early visual areas align more closely with texture-based patterns, the posterior inferotemporal cortex responds dynamically to both textures and object motifs. This suggests that high-level neurons prioritize shared local features rather than complete object templates. Ultimately, the findings provide new insights into the hierarchical organization of vision and the limitations of current artificial intelligence models.

References:

• Wang B, Ponce C R. Neuronal tuning aligns dynamically with object and texture manifolds across the visual hierarchy[J]. Nature Neuroscience, 2026: 1-12.