Precision can be challenging in human movement. Nikolai Bernstein noted that humans exhibit trial-to-trial movement variability when executing repetitive tasks. This variability has both positive and negative consequences and wholly depends on the task. For instance, movement variability can be a beneficial element in human performance as it allows the motor system to accidentally discover the space for better solutions to a task. Conversely, excessive movement variability can be detrimental when a high degree of precision is involved. It is commonly thought that the sources of movement variability stem from redundant degrees of freedom and noise in the motor system. Here, we attempt to experimentally manipulate movement distributions, both the central tendency and its associated variability, within several tasks by imposing discrete spaces of high cost and low cost. We attempt to show how actiontolerance influences motion variability as well as how formation of boundaries guide movements. In addition, we will show how a new model of movement control can explain adaptation to these tasks and other classic motor behavioral tasks through learning “boundaries” of movement, thereby unifying movement variability with movement adaptation. Two studies were conducted to reshape movement variability through two different sensory distortions. Then, we proposed and developed a planning model that can account for these and other movement experiments.

The first study tasked 18 nondisabled subjects and 10 stroke survivors with intercepting a projectile launched towards their chest over 600 trials. The subjects were equally divided into two groups;​ 9 nondisabled subjects and 5 stroke survivors belonged to a control group, while the other 9 nondisabled subjects and 5 stroke survivors belonged to a treatment group. The treatment group received forces that pushed subjects away from an invisible region whenever they were outside the region in trials 201-400. The control group received no such perturbation throughout the experiment. The region was box-shaped, with infinite height and very narrow depth along the anterior-posterior axis, the axis along which the projectile was launched. The nondisabled subjects, as a group, reshaped their movements within the box, and distributed their movements more uniformly. Some stroke survivors also exhibited this capability, however as a group, they failed to reshape their distributions within the box.

The second study tasked 9 nondisabled subjects with intercepting a projectile launched towards their chest over 600 trials, like the first study. However, they experienced a visual distortion instead of a force whenever they were outside the box-shaped region. For this distortion, the cursor was displaced from their actual hand position away from the hand position as they were outside the box, making interception of the projectile difficult. Compared to controls, the treatment group reshaped their movements to be more within the box and exhibited aftereffects similar to the haptic group.

We designed a new movement control model that could explain how exploration and avoidance can result in the formation of “safety-margins”, boundaries within which the nervous system chooses to operate, and avoid states beyond. We assessed the capability of this model in simulating targeted reaching and its associated variability. Furthermore, we tested this model on more complex tasks that involved adaptation to visual motor rotation, the aforementioned limit-push experiment, and even obstacle avoidance, showing the model is translatable across tasks. Potential applications of this endeavor could serve as a generic robotic controller that can adapt to changing environments and situations, as well as learn without supervision. Additionally, this model could be used as a training simulator to predict how distributions can be reshaped based on placement of high-cost regions. The model could also identify parameter values and predict how well individuals might operate and learn in new tasks. This is a new simple approach to motor planning that can predict coordinated movements in human motion tasks as well as robotics applications.