Humans rely on two strategies to walk to a goal: (1) Optic flow strategy: null the visual angle between the heading specified by optic flow and the visual direction of the goal; (2) Egocentric direction strategy: null the angle between the locomotor axis and the egocentric direction of the goal. Optic flow dominates in environments with sufficient visual surface structure. In the 1960's, Held proposed that optic flow might also drive prism adaptation during walking.

We test this hypothesis by adapting participants to displays in which the optic flow pattern was displaced from the walking direction by 10 deg to the right, and then testing them with normal flow. Participants walked to a target in an immersive virtual environment (the 12 m x 12 m VENLab) while wearing a head-mounted display and head tracker. Two worlds were used in both adaptation and test phases: (a) a lone target line (minimal optic flow), or (b) textured surfaces and posts (rich optic flow). This created a 2x2 design with four groups of participants.

In the Line world, results show gradual adaptation, with gradually reduced heading error and straighter paths over 36 trials. In the Textured world there was immediate behavioral adaptation in the first few trials. When both adapted and tested in the Line world, participants showed a small negative after-effect, such that heading errors and paths deviated in the opposite direction. But when adapted in the Textured world and tested in the Line world, there was a large negative after-affect. In contrast, when tested in the Textured world, both after-effects were immediately abolished.

The results indicate that optic flow serves as the teaching signal that adapts the visual-motor mapping between egocentric direction and walking direction. Optic flow thus dominates both the guidance of walking and visual-locomotor adaptation in visually structured environments.