Learning to Learn: Metaplasticity Improves the Temporal Precision of Motor Learning

In order to facilitate motor learning, the brain is thought to implement a supervised learning process whereby errors in performance are incrementally compensated for through the adjustment of the neural synapses that were responsible for causing the error. While artificial neural networks can easily perform these adjustments, the brain faces the difficult problem of receiving reports about performance errors at a nontrivial delay, thus requiring that rules governing the induction of synaptic plasticity are in place to ensure that only the synapses that caused the error are adjusted. Recently, our lab has shown that the cerebellum, the brain region that supports motor learning, solves this problem for oculomotor learning by selectively adjusting the strengths of synapses that were active 120ms prior to the report of an error – the same amount of time necessary for error feedback to reach the cerebellum. Recent findings from our lab further suggest that this temporal rule for the induction of plastic changes is not genetically hard-coded, but rather that the brain must actively learn it through experience of the 120ms visual feedback delay.
Here, we report two fundamental findings. First, we show that this metaplasticity, or the tuning of the synaptic plasticity rules, affects the timing of oculomotor learning. Using a well-characterized form of oculomotor learning, we show that without visual experience, and thus without experience of the visual feedback delay, animals adjust eye movements at a delay relative to animals with normal visual experience. Second, we provide first evidence for adult metaplasticity using this behavioral correlate. We show both that dark-reared animals are able to significantly improve the timing of oculomotor learning after adulthood rehabilitation to visual experience, and that eliminating adulthood visual experience in normally-reared animals impairs the timing of learning.