This research study investigates how learning a task transforms the way neurons in the macaque visual cortex process and share information. Contrary to the classic belief that the brain improves efficiency by reducing overlapping signals, the authors discovered that information redundancy actually increases as animals become more proficient at a task. By recording neural activity over several weeks, they observed that neurons began to share more task-relevant data through feedback and internal inference rather than operating as independent sensors. This shift suggests that sensory processing is a bidirectional inference process where prior expectations and incoming evidence are integrated across the neural population. Ultimately, the study reveals that the brain prioritizes distributing information across many neurons to support robust decision-making rather than simply eliminating noise or redundancy.

References:

• Liu S, Pletenev A, Haefner R M, et al. Task learning increases information redundancy of neural responses in macaque visual cortex[J]. Science, 2026, 391(6789): 1029-1035.