Motor learning: changes in the structure of variability in a redundant task.

Although variability is a fundamental and ubiquitous feature of movement in all biological systems, skilled performance is typically associated with a low level of variability and, implicitly, random noise. Hence, during practice performance variability undergoes changes leading to an overall reduction. However, learning manifests itself through more than just a reduction of random noise. To better understand the processes underlying acquisition and control of movements we show how the examination of variability and its changes with practice provides a suitable window to shed light on this phenomenon. We present one route into this problem that is particularly suited for tasks with redundant degrees of freedom: task performance is parsed into execution and result variables that are related by some function which provides a set of equivalent executions for a given result. Variability over repeated performances is analyzed with a view to this solution manifold. We present a method that parses the structure of variability into four conceptually motivated components and review three methods that are currently used in motor control research. Their advantages and limitations are discussed.