Integration of 3D structure from disparity into biological motion perception independent of depth awareness

Abstract

Images projected onto the retinas of our two eyes come from slightly different directions in the real world, constituting binocular disparity that serves as an important source for depth perception - the ability to see the world in three dimensions. It remains unclear whether the integration of disparity cues into visual perception depends on the conscious representation of stereoscopic depth. Here we report evidence that, even without inducing discernible perceptual representations, the disparity-defined depth information could still modulate the visual processing of 3D objects in depth-irrelevant aspects. Specifically, observers who could not discriminate disparity-defined in-depth facing orientations of biological motions (i.e., approaching vs. receding) due to an excessive perceptual bias nevertheless exhibited a robust perceptual asymmetry in response to the indistinguishable facing orientations, similar to those who could consciously discriminate such 3D information. These results clearly demonstrate that the visual processing of biological motion engages the disparity cues independent of observers' depth awareness. The extraction and utilization of binocular depth signals thus can be dissociable from the conscious representation of 3D structure in high-level visual perception.

Figure 1. Schematic representations of 3D biological motion stimuli and results of Experiment 1.

(A) During each trial, a pair of point-light walker sequences with horizontal binocular disparity was presented to the left and right eyes of the observer through a mirror stereoscope. (B) A series of point-light walkers, either facing toward (T-trial) or away (A-trial) from the observer defined by stereoscopic cues were displayed in random order. Observers were divided into experimental and control groups according to their perceived depth information. (C) The upper panel shows the contrast of accuracy for in-depth facing orientation judgment between the two groups, with the physical in-depth facing orientations combined or separated. Observers in the experimental group correctly identified the in-depth rotating direction of point-light sphere (lower left panel), yet their performance for discriminating the depth-reversed point-light walkers was at chance level (lower right panel). Error bars indicate one SEM.

Figure 2. Experimental conditions and results of the left-right walking direction discrimination experiment.

(A) Walking directions of stereo point-light walkers, defined by disparity cues (Purple - toward; Green - away), deviated slightly from the observer’s line of sight (L - Left; R - Right). (B) Left-right discrimination thresholds of toward and away conditions plotted for individual observers (the experimental group) and as group averages (the experimental and the control groups). The threshold of the toward condition was significantly lower compared with that of the away condition. Error bars indicate one SEM. **p<.01.

Figure 3. Illustration of a single frame of a sample stimulus used in the biological motion detection experiment and results of the experiment.

(A) An upright point-light walker (target) embedded in dynamic noise dots. Blue lines are for illustration purpose here and were not shown in the actual experiment. (B) The accuracy of the detection task was significantly higher for the toward condition (purple) than for the away condition (green). Exp_LM+ or Exp_LM-: average performance of the experimental group, with congruent local motion (i.e., noise made from scrambled targets) or incongruent local motion (i.e., noise made from scrambled walkers rendered with disparity signals opposite to those of the target). Ctrl: average performance of the control group, with local motion conditions combined. Error bars indicate one SEM. **p<.01.