Brain’s Hidden Learning Limits

Populations of roughly ninety simultaneously recorded neurons in a monkey’s motor cortex form the core actors, each neuron contributing one scalar firing rate that together define the moment-to-moment state of the system.

Monkeys implanted with microelectrodes in motor cortex interact with a brain-computer interface that reads activity from about ninety neurons and translates it into the position of a cursor on a screen.

The same motor-cortex neuron population that successfully controls a cursor in the movement-intention view is re-examined through an alternative projection that emphasizes differences between leftward and rightward movements.

Monkeys already proficient at controlling a cursor with a brain-computer interface are given a more demanding task that explicitly requires neural activity to approximate a time-reversed version of previously used patterns.

Human learners, whose brains share fundamental architectural principles with the monkey motor cortex studied in the Nature Neuroscience paper, face similar constraints on which patterns of neural activity are physically achievable through practice.