We addressed the fundamental questions of which variables underlie the control of arm movement and how they are stored in motor memory, reproduced and modified in the process of adaptation to changing load conditions. Such variables are defined differently in two major theories of motor control (internal models and threshold control). To resolve the controversy, these theories were tested (experiment 1) based on their ability to explain why active movement away from a stable posture is not opposed by stabilizing mechanisms (the posture-movement problem). The internal model theory suggests that the system counteracts the opposing forces by increasing the muscle activity in proportion to the distance from the initial posture (position-dependent EMG control). In contrast, threshold control fully excludes these opposing forces by shifting muscle activation thresholds and thus resetting the stabilizing mechanisms to a new posture. Subjects were sitting, holding the vertical handle of a double-joint manipulandum with their right hand and were facing a computer screen on which the handle and target to be reached were displayed. In response to an auditory signal, subjects quickly moved the handle from an initial position to one of two (frontal and sagittal) targets. No load was applied during the movement but in separate trials, a brief perturbation was applied to the handle by torque motors controlling the manipulandum. Perturbations were applied prior to or 3 s after movement offset, in the latter case in one of eight directions. The EMG activity of the majority of the seven recorded muscles was at zero level before movement onset and returned to zero level after movement offset. Those muscles that remained active before or after the movement could be made silent whereas previously silent muscles could be activated after a small passive displacement (several millimeters) elicited by perturbations in appropriate directions. Results showed that the activation thresholds of motoneurons of arm muscles were reset from the initial to a final position and that EMG activity was not position-dependent. These results were inconsistent with the internal model theory but confirmed the threshold control theory. Then the ability of threshold control theory to explain rapid movement adaptation to a position-dependent load was investigated (experiment 2 and 3). Subjects produced fast movement to the frontal target with and without a position-dependent load applied to the handle. Trials were organized in blocks alternating between the load and no-load condition (20 blocks in total, with randomly chosen number of five to ten trials in each). Subjects were instructed "do not correct" in experiment 2 and "correct" movement errors during the trial in experiment 3. Five threshold arm configurations underlying the movement production and adaptation were identified. When instructed "do not correct", movement precision was fully restored on average after two trials. No significant improvement was observed as the experiment progressed despite the fact that the same load condition was repeated after one block of trials. Thus, in each block, the adaptation was made anew, implying that subjects relied on short-term memory and could not recall the threshold arm configurations they specified to accurately reach the same target in the same load condition in previous blocks. When instructed to "correct" within each trial, precision was restored faster, on average after one trial. Major aspects of the production and adaptation of arm movement (including the kinematics, moveent errors, instruction-dependent behavior, and absence of position-related EMG activity) are explained in terms of threshold control.
