The visual system's intrinsic bias influences space perception in the impoverished environment

Abstract

A dimly lit target in the intermediate distance in the dark is judged at the intersection between the target's projection line from the eye to its physical location and an implicit slanted surface, which is the visual system's intrinsic bias. We hypothesize that the intrinsic bias also contributes to perceptual space in the impoverished environment. We first showed that a target viewed against sparse texture elements delineating the horizontal ground surface in the dark is localized along an implicit slanted surface that is less slanted than that of the intrinsic bias, reflecting the weighted integration of the weak texture information and intrinsic bias. We also showed that while the judged egocentric locations are similar between 0.15- to 5-s exposure durations, the judged precision improves with duration. Furthermore, the precision for the judged target angular declination does not vary with the physical angular declination and is better than the precision of the eye-to-target distance. Second, we used both action and perceptual tasks to directly reveal the perceived surface slant. Confirming our hypothesis, we found that an L-shaped target on the horizontal ground with sparse texture information is perceived with a slant that is less than that of the intrinsic bias.

Figure 1

Space perception based on the visual system's intrinsic bias and external depth information. (a) A dimly lit target in an otherwise dark environment is perceived at the intersection between its projection line and the visual system's intrinsic bias (slanted curve). (b) A dimly lit L-shaped target in the dark is perceived as slanted along the intrinsic bias. (c) The sequential surface integration process (SSIP) constructs the ground surface representation by integrating the texture information on the ground. It is predicted that when the texture information is insufficient (reduced cue condition), the ground surface representation will be a slanted surface due to the influence of the intrinsic bias.

Figure 2

Predictions of space perception in a reduced cue condition where a 2×3 array of dimly lit texture elements delineates the horizontal ground in the dark. (a) The SSIP hypothesis predicts the ground surface representation that acts as a reference frame for space perception is slanted. Thus a dimly lit target is perceived as if located on the slanted surface. (b) Similarly, the SSIP hypothesis predicts a horizontal L-shaped target being perceived as slanted. (c) Opposing the SSIP hypothesis, a horizontal compression hypothesis predicts that the ground surface representation is horizontally compressed toward the observer. One consequence is that the subjective angular declination is larger than the physical angular declination (α′> α)

Figure 3

Illustration of the stimuli used in Experiment 1. (a) Side view of an observer viewing a target against a parallel-texture background. (b) and (c) Top view of the parallel-texture and convergent-texture background conditions, respectively.

Figure 4

The average perceived target locations in Experiment 1 in (a) the dark, (b) parallel-texture, and (c) convergent-texture conditions. Note that in all three conditions, the perceived locations of the dimly lit targets with the three different stimulus durations largely cluster together even as they are deviated from the physical target locations.

Figure 5

The average perceived target locations in Experiment 1. Each symbol represents the average of three perceived locations with different stimulus durations. (a) Targets were located 0.5m above the horizontal ground. (b) Targets were located on the horizontal ground. The average eye position is depicted on the top of the y-axis. It is linked to the physical target locations by the dash projection lines. The error bars are plotted in polar coordinates with the ones pointing to the average eye position being the mean standard errors of the perceived eye-to-target distance, whereas the orthogonal error bars represent the standard errors of angular declination scaled according to the perceived eye-to-target distance. See text for details.

Figure 6

The average judged angular declinations for all the six target locations in the three background conditions. All have slopes close to unity, although the data for the convergent-texture condition are systematically shifted downward.

Figure 7

The precisions of performance in Experiment 1 based on inter-trial variabilities (standard deviations). (a) Standard deviation of judged angular declination as a function of stimulus exposure duration. (b) Standard deviation of eye-to-target distance as a function of stimulus exposure duration. In both (a) and (b), each data point represents the average of six target locations. (c) Standard deviation of judged angular declination as a function of the physical target angular declination. (d) Ratio of the standard deviation of the judged eye-to-target distance to the judged eye-to-target distance as a function of the physical eye-to-target distance. In both (c) and (d), each symbol depicts the average of each target location for the three test conditions. See text for details.

Figure 8

Illustration of the two stimulus conditions used in Experiment 2. (a) Viewing the L-shaped target in the dark. (b) Viewing the L-shaped target in the presence of a parallel-texture background in the dark.

Figure 9

Actual physical aspect ratios of the matched L-shaped targets in the dark and parallel-texture conditions.

Figure 10

Comparison of the derived surface slant based on three different tasks: (a) the aspect ratio; (b) palm-board/palm orientation; and (c) walking/palm-orientation.