Perception of distance during self-motion depends on the brain's internal model of the terrain

Abstract

The body's geometrical relationship with the terrain is important for depth perception of human and non-human terrestrial animals. Static human observers in the dark employ the brain's internal model of the terrain, the intrinsic bias, to represent the ground as an allocentric reference frame for coding distance. However, it is unknown if the same ground-based coding process operates when observers walk in a cue-impoverished environment with visible ground surface. We explored this by measuring human observers' perceived locations of dimly-lit targets after a short walk in the dark from the home-base location. We found the intrinsic bias was kept at the home-base location and not the destination-location after walking, causing distance underestimation, fitting its allocentric nature. We then measured perceived distance of dimly-lit targets from the destination-location when there were visual depth cues on the floor. We found judged locations of targets on the floor transcribed a slanted surface shifted towards the home-base location, indicating distance underestimation. This suggests, in dynamically translating observers, the brain integrates the allocentric intrinsic bias with visual depth cues to construct an allocentric ground reference frame. More broadly, our findings underscore the dynamic interaction between the internal model of the ground and external depth cues.

Fig 1. The ground-based spatial coding scheme. a. The visual system uses the intrinsic bias, an implicit curved surface representation (dashed white curve), to locate a dimly-lit target (unfilled green disc) in the dark. The target is perceived (filled green disc) at the intersection between the projection line from the eye to the target and the intrinsic bias. b. When the ground becomes visible, the intrinsic bias integrates with the visible depth cues to form a ground surface representation, which serves as a reference frame to code target location. For example, in the reduced cue condition (impoverished environment) where parallel rows of texture elements (filled red circles) on the ground provide the depth information, the visual system constructs a ground surface representation (solid red curve) from integrating the intrinsic bias with the texture cue. The surface slope of the ground surface representation is smaller than the intrinsic bias, which leads to a more accurate target localization than in the dark (a). In the reduced cue condition, the visual system is able to localize a target (unfilled green disc) using both the ground reference frame and the relative depth information between the target and the ground (e.g., relative binocular disparity).

Fig 2. Hypotheses and predictions. a. Dark-walking condition. The allocentric hypothesis predicts when the observer walks forward from the home base (blue cross) to a new location (red cross), the visual system relies on the path-integration process to keep the intrinsic bias (dashed white curve) at the home base (left figure). This causes a relative shift (backward) between the intrinsic bias and observer. The right figure shows the observer now at the new location, perceives the target (unfilled green circle) at the intersection (green disc) between the intrinsic bias anchored at the home base and the projection line from the eye to the target. b. Dark-stationary condition (same as figure 1a). A non-moving, static observer perceives the dimly lit target (unfilled green disc) at the intersection (filled green disc) between the projection line from the eye to the target and the intrinsic bias (dashed white curve). Note the perceived horizontal distance of the target is longer than that in the dark-walking condition (yellow arrows). c. Texture-walking condition. It is similar to the dark-walking condition except after walking for a short distance in the dark, the observer saw a target against a texture background. The allocentric hypothesis predicts the texture surface representation (red curve) will be compressed toward to the observer relative to the texture-stationary condition in d, and accordingly the perceived distance will be more underestimated (shorter yellow arrow bar). d. Texture-stationary condition. A non-moving, static observer perceives the dimly lit target (unfilled green disc) at the intersection (as filled green disc) between the projection line from the eye to the target and the texture surface representation (red curve). e. Left graph: The average judged target locations of the dark-stationary (red circles) and dark-walking (green triangles) conditions. The data points are fitted by red and green curves (transcribing the intrinsic bias) of the same shape but horizontally shifted. Right graph: Average judged target locations of the texture-stationary (red circles) and texture-walking (green triangles) conditions. Perceived horizontal distances were shorter in the texture-walking condition. In both graphs, the plus symbols represent the physical target locations. Error bars represent the standard errors of the mean.