A neural population mechanism for rapid learning

This article is a preprint and has not been peer-reviewedLong-term learning of language, mathematics, and motor skills likely requires plastic changes in the cortex, but behavior often requires faster changes, sometimes based even on single errors. Here, we show evidence of one mechanism by which the brain can rapidly develop new motor output, seemingly without altering the functional connectivity between or within cortical areas. We recorded simultaneously from hundreds of neurons in the premotor (PMd) and primary motor (M1) cortices, and computed models relating these neural populations throughout adaptation to reaching movement perturbations. We found a signature of learning in the ″null subspace″ of PMd with respect to M1. Earlier experiments have shown that null subspace activity allows the motor cortex to alter preparatory activity without directly influencing M1. In our experiments, the null subspace planning activity evolved with the adaptation, yet the ″potent mapping″ that captures information sent to M1 was preserved. Our results illustrate a population-level mechanism within the motor cortices to adjust the output from one brain area to its downstream structures that could be exploited throughout the brain for rapid, on-line behavioral adaptation. [ENG]The authors would like to thank Dr. Sara Solla for her helpful discussions in preparing these analyses. M.G.P. received funding from NIH NINDST32 HD07418 and NIH NINDSF31 NS092356. J.A.G. received funding from the European Commission (FP7-PEOPLE-2013-IOF-627384). This project was additionally funded by NIH NINDSNS053603 and NIH NINDSNS074044.Peer reviewe