A Study on the Cognitive Process of Stereopsis in Small Uncrossed Disparity

Static random-dot stereograms (RDS) were used as stimuli to investigate the uncrossed disparity in 15 normal subjects. The response of every subject was recorded with different disparities of 3.27 arc min, 6.54 arc min, 8.18 arc min, 11.45 arc min, 14.72 arc min, 17.99 arc min, 21.26 arc min and 24.53 arc min. The results showed that the human visual system was the most sensitive to stereo images at disparity of 8.18 arc min. Disparity of 21.26 arc min had significant differences with other small disparities in reaction times, as supported the viewpoint that it was reasonable to limit the fine disparity in 20 arc min.

Decoding conjunctions of direction-of-motion and binocular disparity from human visual cortex

Motion and binocular disparity are two features in our environment that share a common correspondence problem. Decades of psychophysical research dedicated to understanding stereopsis suggest that these features interact early in human visual processing to disambiguate depth. Single-unit recordings in the monkey also provide evidence for the joint encoding of motion and disparity across much of the dorsal visual stream. Here, we used functional MRI and multivariate pattern analysis to examine where in the human brain conjunctions of motion and disparity are encoded. Subjects sequentially viewed two stimuli that could be distinguished only by their conjunctions of motion and disparity. Specifically, each stimulus contained the same feature information (leftward and rightward motion and crossed and uncrossed disparity) but differed exclusively in the way these features were paired. Our results revealed that a linear classifier could accurately decode which stimulus a subject was viewing based on voxel activation patterns throughout the dorsal visual areas and as early as V2. This decoding success was conditional on some voxels being individually sensitive to the unique conjunctions comprising each stimulus, thus a classifier could not rely on independent information about motion and binocular disparity to distinguish these conjunctions. This study expands on evidence that disparity and motion interact at many levels of human visual processing, particularly within the dorsal stream. It also lends support to the idea that stereopsis is subserved by early mechanisms also tuned to direction of motion.

Influences of Motion and Depth on Brightness Induction: An Illusory Transparency Effect?

Two experiments were performed to investigate whether motion and binocular disparity influence brightness induction, and whether the effects of motion and binocular disparity, if any, interact with each other. In order to introduce motion, textured backgrounds were used as the inducing field. The results showed that motion and/or crossed disparity reduce brightness induction, whereas uncrossed disparity increases it. The effect of motion and the effect of disparity are independent of each other and additive, which suggests that, to the extent that brightness induction reflects segmentation of objects, motion and binocular disparity serve independently to segment objects from their background. The difference between the effects of crossed and uncrossed disparity can be explained by what we call ‘illusory transparency’.

Disparity Sensitivity of Frontal Eye Field Neurons

Information about depth is necessary to generate saccades to visual stimuli located in three-dimensional space. To determine whether monkey frontal eye field (FEF) neurons play a role in the visuo-motor processes underlying this behavior, we studied their visual responses to stimuli at different disparities. Disparity sensitivity was tested from 3° of crossed disparity (near) to 3° degrees of uncrossed disparity (far). The responses of about two thirds of FEF visual and visuo-movement neurons were sensitive to disparity and showed a broad tuning in depth for near or far disparities. Early phasic and late tonic visual responses often displayed different disparity sensitivity. These findings provide evidence of depth-related signals in FEF and suggest a role for FEF in the control of disconjugate as well as conjugate eye movements.

Role of Spatial and Temporal Coincidence in Depth Organization

The linking of spatial information is essential for coherent space perception. A study is reported of the contribution of temporal and spatial alignment for the linkage of spatial elements in terms of depth perception. Stereo half-images were generated on the left and right halves of a large-screen video monitor and viewed through a mirror stereoscope. The half-images portrayed a black vertically oriented bar with two brackets immediately flanking this bar and placed in crossed or uncrossed disparity relative to the bar. A pair of thin white ‘bridging lines' could appear on the black bar, always at zero disparity. Brackets and bridging lines could be flickered either in phase or out of phase. Observers judged whether the brackets appeared in front of or behind the black bar, with disparity varied. Compared to conditions when the bridging lines were absent, depth judgments were markedly biased toward “in front” when bridging lines and brackets flashed in temporal phase; this bias was much reduced when the bridging lines and brackets flashed out of phase. This biasing effect also depended on spatial offset of lines and brackets. However, perception was uninfluenced by the lateral separation between object and brackets.

Testing the Hypothesis of Labelled Detectors

The hypothesis of labelled detectors (or ‘lines’) is the present-day version of the basic Müller - Helmholtz doctrine. Müller's dictum of specific energy of nerves stated: “the same internal cause excites (…) in each sense the sensation peculiar to it”. Helmholtz made ‘the cause’ external to the body and postulated that all knowledge about the world thus comes through the senses. The key word is specificity. The strong version of the hypothesis must treat detection - identification as a single task: a stimulus would be identified whenever it is detected. The weak version requires only that we identify a specific mechanism by which both detection and identification are achieved, even though the latter may require additional processing. In the general case, the strong version (with its ludicrous ‘grandmother cell’ as the neural substrate) finds little support. Detection and recognition of complex shapes (letters, faces, etc) aside, even discrimination between simple increments and decrements of luminance is difficult to attribute directly to a specific mechanism (in this case, the activity in either ON or OFF systems, respectively). This is demonstrated by experiment 1 reported here. However, perception of relative depth seems to conform to the strong version of the hypothesis, as experiment 2, also reported here, indicates. Thus, at least some specific neural mechanisms (in this case, probably the crossed and uncrossed disparity detectors) may be indeed linked directly to perception.

The Extrinsic/Intrinsic Classification of Motion Signals is a High-Level Process

It has been proposed that the 2-D motion signals elicited by the bar endings of a barber-pole stimulus disambiguate 1-D motion signals with a variable strength which depends on depth (Shimojo et al, 1989 Vision Research29 619 – 626): these signals would be ‘abolished’ when they are extrinsic (ie the moving grating is behind the plane of the background containing the aperture), whereas they would be given full strength when they are intrinsic (ie the plane of the grating is in front of the background). These authors have suggested that the intrinsic/extrinsic classification is an early process. However, the very long duration (2300 ms) used in their study suggests other interpretations. Therefore, we tried to test whether the barber-pole illusion could be abolished with a shorter duration when the grating had an uncrossed disparity relative to the aperture plane, as initially described in the above-mentioned study (our 30 observers had to adjust an arrow to indicate the perceived direction of the grating). In accordance with our prediction, we could not replicate their finding with a duration of 400 ms. Surprisingly, increasing the duration up to 2300 ms was not sufficient to obtain a large bias towards 1-D signals. To understand this unexpected result, we tried to isolate the relevant difference between the initial study of Shimojo et al and our. We found that the main determinant of the suppression of the barber-pole illusion was the experimental procedure: when our observers had to assess the perceived direction of the barber-pole by choosing between horizontal and vertical, as in the initial study, the results did show a much larger bias towards 1-D signals. We suggest therefore that the extrinsic/intrinsic classification is a high-level process which can be influenced by the observer's expectations.