Figure 3. Obstacle avoidance paradigm.

(a) Setup for the obstacle avoidance experiments, where virtual obstacles could obstruct movements to on-screen targets. (b) Illustration of obstacle-free 1-target and 2-target trials, similar to Figure 1b. (c) Diagram of left-side obstacle-present 1-target trials. The obstacle in Expt 2a was designed to elicit rightward movement deflections for left 1-target trials. The obstacle in Expt 2b, which included an identical protrusion towards the midline, was designed to promote a leftward deflection for left 1-target target trials. (d) Predictions during uncertainty. Left panels: motor averaging (MA; purple arrow) would average the plans for the obstacle-obstructed left and unobstructed right 1-target trial movements. It thus predicts initial movement directions on 2-target trials with opposite deflections from the midline for Expts 2a and 2b (away from the obstacle in Expt 2a, but towards the obstacle in Expt 2b, where the safety margin around the obstacle would be reduced). Right panels: performance optimization (PO; green) would promote task success by balancing the costs of safely avoiding the obstacle and rapid target acquisition. It thus predicts initial movement directions on 2-target trials to be consistently deflected away from the obstacle. Thus, PO and MA make similar predictions in Expt 2a, but opposite predictions in Expt 2b. Note that (c) and (d) depict a left-side obstacle condition, but that the experiments were balanced within participants, with an equal number of trials for left- and right-side obstacles. A right-side obstacle condition would lead to mirror-opposite predictions for both MA and PO.