Study finds visual learning subtly reshapes brain's object-processing cortex

Learning to recognize new objects produces subtle but reliable changes in the inferior temporal cortex, the brain region where high-level visual features are represented, according to research published July 8 in Nature Communications. Scientists at MIT's McGovern Institute for Brain Research and York University recorded neural activity in animals trained to discriminate objects such as elephants and chairs across sizes, angles, and backgrounds, comparing them with untrained animals. The broad pattern of IT cortex activity remained largely similar, but trained animals showed consistent, modest differences in neuronal responses to the learned objects. To test how such modest changes could support learning, the team built artificial neural networks with components mapped to the IT cortex and trained them on the same discrimination tasks using gradient descent. Only models whose IT-like stage changed in ways paralleling the biological data reproduced the animals' learning behavior. The findings suggest that visual learning does not dramatically rewrite high-level visual representations but instead tunes them in a targeted manner. The researchers say modeling these changes could eventually help predict how training reshapes perception and inform educational strategies.