Familiarity Warps Memory Codes in Macaque Prefrontal Cortex

Camilo Libedinsky examines how familiarity changes sequence working memory, combining recordings from primate prefrontal cortex with recurrent neural network models. Extensive training on selected stimulus sequences improved retrieval accuracy and response speed. Population decoding and analyses of neural subspace geometry showed that well-practised sequences become more isolated in the population representation, suggesting that experience reshapes the geometry of memory representations.