On the control of unstable objects: the dynamics of human stick balancing.

Objects that we control and interact with are often unstable. Riding a bicycle, balancing a tray of food, maintaining the oscillations of a hula-hoop, and even standing upright are exemplary tasks that require the control of an unstable object. Although it is difficult to characterize the physics of complex object interactions, we are adept at learning and performing these tasks in everyday life. Unstable objects require carefully assembled control mechanisms because, by definition, the object must be stabilized through the interaction between the human control and the intrinsic object dynamics. Additionally, such tasks demand extremely precise control because error can elicit abrupt and irrevocable changes in the performance (Balasubramaniam and Turvey 2004; Cluff et al. 2008). Although much is known about human motor control and object manipulation in predictable systems (when the mapping between actions and their consequences is straightforward), much less is understood about unstable object control. This is largely because the dominant research focus has been on characterizing the task and context-dependent attributes of firmly grasped, rigid object control (Imamizu et al. 2003; Milner et al. 2006). Although an extensive literature has focused on adaptation to novel mechanical loads, few studies have considered how we learn to control unstable objects. As a result, there are a number of important questions that remain in motor control research: What strategies are used to control unstable objects and how are these control mechanisms learned? Are common control processes shared between interacting motor systems? Do these control mechanisms involve predictive internal forward models? In this review article, I use a stick-balancing task as a paradigmatic example to investigate control mechanisms and skill acquisition in relation to unstable object control. The stick-balancing task imposes a complex control problem that involves