Resource allocation in neural networks for motor control

Multiplicative noise plays an important part of a non-predictive control mechanism for stick balancing at the fingertip. However, intentionally-directed movements are also used in stick balancing, particularly by beginners. The interplay between intentional and non-predictive control mechanisms for stick balancing was assessed using two dual task paradigms: the subject was asked to either move one of their legs rhythmically or to imagine moving their leg while balancing a stick (55.4 cm, 35 g) at their fingertip. Performance was measured by determining the stick survival function, i.e. the fraction of trials (total $\ge 25)$ for which the stick remained balanced at time t as a function of t. Performance was increased by concurrent rhythmic leg movements (50{\%} survival time shifted from 8-9s to 15s in a typical subject). Imagined movements resulted in a similar improvement (50{\%} survival time of 20s for the above subject) suggesting that this enhancement is not simply related to mechanical vibrations of the fingertip induced by leg movement. These observations emphasize the importance of the development of mathematical models for neural control of skilled motor movements that take into resource allocation of limited resources, such as intention.