Behavioral Consequences of Velocity Commands: Brownian Processes in Human Motor Tasks

Abstract

Human postural control exhibits stochastic behavior resembling a random walk, which implies unbounded growth of position variance. Here we show that unbounded position variance is indeed observed in a common task, turning a crank. In postural tasks, intermittent control has been proposed to limit position variance. We further show that when position variance is unbounded, it displays a distinctive signature of underlying Brownian motion, a low-frequency power spectrum that declines at -20 dB/decade but exhibits flattening in bounded tasks. Remarkably, a descriptive model of a control system competent to account for all of these data must feature a forward-path velocity command corrupted by stationary noise. This provides independent behavioral support for the theory that the brain encodes motion commands in terms of velocity.