Abstract
I investigated the hypothesis that performance on a visually guided action improves with practice (in part) because actors learn to exploit more reliable information in optic flow. Participants performed a simulated braking task, using a joystick to decelerate to a stop before a target. Information is available in optic flow that specifies how hard to decelerate, and is invariant across changes in target size and approach speed. However, observers (especially novices) may rely on a less reliable non-invariant, which would result in biases as target size and approach speed are varied. Participants completed eleven blocks of 25 trials. Target size and approach speed were manipulated within each block. Participants exhibited a weak size bias that quickly disappeared with practice, and a strong speed bias that diminished but did not disappear. In a control condition in which the textured ground plane was absent, the speed bias was greater and did not change with practice. The results suggest that participants quickly became attuned to a size-invariant optical variable, and learned to use more reliable information in optic flow (global optic flow rate of the ground plane) to perform the task across changes in speed. I conclude that actors do not always rely on a single optical invariant, contrary to most existing theories of continuously controlled visually guided actions. The reliability of different optical variables depends on task and environmental constraints. The ability to flexibly tune to different optical variables may underlie actors' ability to adapt to changes in such constraints.