A neuroscientist learns about algorithms for motor learning.

Under what conditions do people learn most effectively? This question is pertinent to several fields and to many neuropsychiatric disorders involving aberrant learning and memory. In motor neurobiology, understanding how people learn new movements may yield insight into the brain's motor-control algorithms and could help with physical training or rehabilitation.

A recent study by Maurice Smith at Harvard University in Cambridge, Massachusetts, and his colleagues suggests that the neural codes underlying motor control may help to dictate which movements are inherently more difficult to learn than others (G. C. Sing et al. Neuron 64, 575–589; 10.1016/j.neuron.2009.10.0012009).

They had volunteers grasp a robotic arm and make targeted, forward-reaching movements while the robot applied a perturbing force in a direction perpendicular to that of the reach. Volunteers learned to compensate for these perturbations most quickly when the magnitude of the disrupting force correlated positively with both arm position and velocity. Compensations involving only one of these factors took longer to learn and were learned less accurately. Even more challenging were disturbances in which the position and velocity contributions were negatively correlated to each other. Errors were systematically biased, as if the brain expected positive correlations.

The findings fit well with previous physiological recordings revealing that neural elements of motor control often encode information about limb position and velocity along positively correlated spatial directions. This aspect of the neural code for movement may impose constraints on how humans learn motor tasks and bias motor errors. More generally, many aspects of human behaviour might be shaped by underlying neural codes that affect the ease with which some behaviours are learned.

Discuss this paper at http://blogs.nature.com/nature/journalclub