Learning new skills can be hard: it takes time and, above all, practice. Only through a gradual process of trial and error do we acquire the ability to walk, speak, or play the trumpet. This process, called motor skill learning, involves two structures within the brain: an ‘actor’, which adjusts the behaviour of the animal, and a ‘critic’, which monitors the effect of the adjustments and compares it with the desired outcome. The interaction between these two regions results in a reinforcement of the adjustments made by the actor that the critic thinks will take the behaviour closer toward the desired outcome, until practice has indeed made perfect.

A central yet untested assumption in the current model of motor skill learning is that the gradual changes implemented by the actor are an important part of this learning, and that the critic monitors progress by feedback received from the movements that are generated as a result of these changes. A groundbreaking study recently published by Jonathan Charlesworth, Timothy Warren and Michael Brainard in Nature disputes this assumption.

Bengalese finches can be trained to modify the pitch of their song, a gradual process of motor skill learning requiring the anterior forebrain pathway (AFP) as the actor, and dopaminergic neurons within the basal ganglia as the critic. The authors of the study decided to test whether the adjustments implemented by the AFP actor were required in the training process by blocking the output of the AFP to the motor system with a drug.

As expected from previous studies, the finches did not gradually change their song in response to the training regime when the drug was applied. However, much to their surprise, when the authors removed the drug, the birds immediately produced a perfectly modified pitch. The finches therefore skipped all the steps that they normally take during the training process, and instantly went from novices to experts. These findings suggest that the step-by-step improvements normally seen during training are not required for learning.

However, when the team disrupted neuronal activity within the AFP during training, they found that not only was there no gradual improvement but also the birds did not modify their song after the drug was removed. Inactivating the AFP actor therefore interferes with learning, suggesting that activity within the circuit is required during training.

These findings show that Bengalese finches can modify their song even without gradual improvements, suggesting that the critic does not monitor the consequences of the actor's adjustments. So how does the critic know whether to tell the actor it is improving? The authors speculate that it is not the animal's movements that the critic looks at but the underlying brain activity instead. The critic probably receives a representation of the activity of the motor system, a so-called ‘efference copy’, which it uses to identify successful adjustments.

This study takes a huge leap towards identifying the neural mechanisms underlying motor skill learning, and will direct many future experiments in the field. Just don't think it will help you to become the next Miles Davis: you'll still have to practice.

J. D.
T. L.
M. S.
Covert skill learning in a cortical-basal ganglia circuit