Abstract
We studied the nature of motor errors in a continuous task, as well as their neural correlates and underlying mechanisms. To do this, we monitored scalp electroencephalographic (EEG) signals while participants used a manual joystick to maintain a visual inverted pendulum simulation upright. Each participant performed the task for four minutes continuously, and we found that the frequency of falling (reaching a horizontal position) decreased with time. We analyzed the EEG and joystick waveforms associated with falls, and we discovered three types of errors: Reactive Destabilization, Proactive Destabilization, and Pure Destabilization. We analyzed the EEG signals for Proactive Destabilization commands, which are defined as when the joystick command is initiated to decelerate the pendulum toward vertical but persists too long and causes a pendulum to fall back toward the fall boundary. In the EEG, we found an error-related negativity correlated to the proactive error. In the EEG, we found an error-related negativity 207.8 ms ± 17.0792 ms after the moment when the error became apparent. The latency of this event-related potential decreases as performance improved over time. The progressive change in latency suggests a learning mechanism, represented by a faster recognition of the command errors over time. The features of the EEG waveform can be used to detect a specific action that will be fed to a correction system — this can allow for early detection of mistakes in simulation training and also provide a measure to test motor skills between individuals.