Disentangling decision uncertainty and motor noise in curved movement trajectories

TL;DR

The study finds that most movement curvature during reaching tasks is due to motor noise, not decision uncertainty.

Contribution

The paper introduces a model that disentangles decision uncertainty and motor noise in curved movement trajectories.

Findings

Most reaches were unaffected by decision uncertainty and explained by motor noise.

A mixture model with baseline and choice components best explained curvature distributions.

Decision uncertainty's role may be overstated if unrelated variability is not considered.

Abstract

When a manual reaching target is selected from a number of alternatives, decision uncertainty can often result in curvature of movement trajectories toward a nonchosen alternative. This curvature in the two-dimensional object plane is typically attributed to competitive interactions between different movement goals. Several models of action selection assume an explicit link between the momentary position of the hand and the state of the underlying decision process. Under this assumption, tracking the position of the hand can be used to infer the temporal evolution of the decision. However, even without a selection requirement, movements show variable amounts of curvature due to motor noise. We assessed the relative contributions of decision uncertainty and motor noise to the variability in curvature in naturalistic reach-to-grasp actions. Participants had to pick up one of two blocks (the brighter/dimmer block) and we manipulated decision uncertainty by varying the luminance difference between the two blocks. Single target baseline reaches were included to model the variability in curvature without a choice requirement. We assessed to what extent this baseline model can account for the curvature distributions observed under choice conditions, and tested several modifications of the model to capture any effects of decision uncertainty. The best model of the curvature distributions under choice conditions involved a mixture of the baseline component along with a separate choice component. The weight of this choice component and analysis of the likelihood of observed reaches under the choice/baseline components, suggest that the majority of reaches were unaffected by decision uncertainty and were compatible with the natural variability in movement trajectories due to motor noise. Unless the variability induced by factors unrelated to the decision process is adequately accounted for, the role of decision uncertainty may be overstated when it is inferred from reach trajectories.

Figures