Form and depth in global stereopsis

Three experiments were performed to investigate the role of vergence and the relationship between form and depth processes in global steropsis by comparing global and classical stereopsis. In the first experiment, the speed of stereoscopic resolution as a function of initial fixation-target distance was measured to discover the role of vergence in stereopsis with the random-dot and contoured stereograms. In the second experiment, the accuracy of form and depth discrimination as a function of fixation-target distance was measured using brief stimulus exposure (150 ms) to examine the nature of form and depth processing in global stereopsis. In the third experiment, the speed of resolving random-dot stereograms in the presence or absence of juxtaposed contoured stereograms was observed to examine the interaction of global stereopsis and classical stereopsis. The conclusions of these studies are summarized as follows: First, vergence plays a critical role in the solution of the random-dot stereograms but not in the solution of contoured stereograms. Second, performance with the contoured stereograms is better than with the random-dot stereograms in terms of both speed and accuracy. Third, in random-dot stereograms, discrimination of form is independent of and more accurate than discrimination of depth. Fourth, again for random-dot stereograms, the disparity of target relative to fixation systematically affects discrimination of form but not discrimination of depth. Fifth, a rapid reduction in reaction time over practice occurs for both types of stereograms. Finally, strong interference with the solution of the random-dot stereograms by the monocular contour occurs when the two kinds of stimuli are present simultaneously.     

Temporal properties of disparity processing revealed by dynamic random-dot stereograms

In studies of the temporal flexibility of the stereoscopic system, it has been suggested that two different processes of binocular depth perception could be responsible for the flexibility: tolerance for interocular delays and temporal integration of correlation. None has investigated the relationship between tolerance for delays and temporal integration mechanisms and none has revealed which mechanism is responsible for depth perception in dynamic random-dot stereograms. We address these questions in the present study. Across five experiments, we investigated the temporal properties of stereopsis by varying interocular correlation as a function of time in controlled ways. We presented different types of dynamic random-dot stereograms, each consisting of two pairs of alternating random-dot patterns. Our experimental results demonstrate that (i) disparities from simultaneous monocular inputs dominate those from interocular delayed inputs; (ii) stereopsis is limited by temporal properties of monocular luminance mechanisms; and (iii) depth perception in dynamic random-dot stereograms results from cross-correlation-like operation on two simultaneous monocular inputs that represent the retinal images after having been subjected to a process of monocular temporal integration of luminance.     

Temporal integration differences between crossed and uncrossed stereoscopic mechanisms

In this study, we investigated temporal integration of disparity information for crossed and uncrossed stereopsis. Across three experiments, exposure duration thresholds were measured for stereoscopic stimuli created from dynamic random-dot stereograms. In Experiment 1, an investigation of disparity detection showed that detection thresholds were equal for the crossed and uncrossed directions. In Experiment 2, an examination of duration limits for depth perception showed that critical durations were lower, and depth more veridical, for crossed depth than for uncrossed depth. In Experiment 3, an investigation of depth discrimination revealed that discrimination thresholds were lower for crossed depth than for uncrossed depth. These results suggest that crossed and uncrossed mechanisms differ in terms of their temporal integration properties at processing levels involving the computation and discrimination of depth.     

Colour inputs to random-dot stereopsis.

Recently it has been claimed by Livingstone and Hubel that, of three anatomically and functionally distinct visual channels (the magnocellular, parvocellular interblob, and blob channels), only the magnocellular channel is involved in the processing of stereoscopic depth. Since the magnocellular system shows little overt colour opponency, the reported loss of the ability to resolve random-dot stereograms defined only by colour contrast seems consistent with this view. However, Julesz observed that reversed-contrast stereograms could be fused if correlated colour information was added. In the present study, 'noise' (non-corresponding) pixels were injected into random-dot stereograms in order to increase fusion time. All six subjects tested were able to achieve stereopsis in less than three minutes when there was only correspondence in colour and not in luminance, and three when luminance contrast was completely reversed. This ability depends on information about the direction of colour contrast, not just the presence of chromatic borders. When luminance and chromatic contrast are defined in terms of signal-to-noise ratios at the photoreceptor mosaic, chromatic information plays at least as important a role in stereopsis as does luminance information, suggesting that the magnocellular channel is not uniquely involved.     

Adaptation to curvature: Eye movements or neural curvature analyzers?

Three experiments were conducted in an attempt to discriminate between an eye-movement theory and a neural curvature analyzer theory of visual adaptation to curvature. Ninety university students served as subjects and were required to inspect stimulus lines presented as (a) curved line pairs, (b) single curved lines, (c) curved line stereograms portraying curved lines concave up, to the right, or toward the subject, or (d) random-dot stereograms portraying curved lines concave up, to the right, or toward the subject. The results of the first two experiments indicate that subjects can readily adapt to the curvature in pairs of lines of opposite curvature presented in different parts of the same or opposite retinas. These results contradict the eye-movement theory of adaptation to curvature. In the third experiment, adaptation to curvature was recorded for curved lines presented as line stereograms and random-dot stereograms. It was concluded that presently the neural curvature analyzer theory of adaptation to curvature best explains the results of these three experiments.     

Temporal constraints on visual learning: a computational model

Given a constant stream of perceptual stimuli, how can the underlying invariances associated with a given input be learned? One approach consists of using generic truths about the spatiotemporal structure of the physical world as constraints on the types of quantities learned. The learning methodology employed here embodies one such truth: that perceptually salient properties (such as stereo disparity) tend to vary smoothly over time. Unfortunately, the units of an artificial neural network tend to encode superficial image properties, such as individual grey-level pixel values, which vary rapidly over time. However, if the states of units are constrained to vary slowly, then the network is forced to learn a smoothly varying function of the training data. We implemented this temporal-smoothness constraint in a backpropagation network which learned stereo disparity from random-dot stereograms. Temporal smoothness was formalized with the use of regularization theory by modifying the standard cost function minimised during training of a network. Temporal smoothness was found to be similar to other techniques for improving generalisation, such as early stopping and weight decay. However, in contrast to these, the theoretical underpinnings of temporal smoothing are intimately related to fundamental characteristics of the physical world. Results are discussed in terms of regularization theory and the physically realistic assumptions upon which temporal smoothing is based.     

Stereoscopic illusion based on the proximity principle

A class of ambiguous random-dot stereograms were created that share the following interesting property: Although the binocular disparity forms a periodic 'sawtooth' waveform as a function of row number (the disparity is constant for a given row), these stimuli yield a monotonically increasing depth percept along the rows. The random-dot pattern of each row is periodic along the horizontal direction for the purpose of producing an ambiguous depth percept. It is this ambiguity that makes it possible for the periodic stimulus to give rise to a monotonic percept. This monotonic percept is substantially enhanced when the rows are shown in temporal sequence instead of all being displayed together. Experiments are reported which indicate that this illusion is due to the proximity, or pulling, effect in stereopsis.     

Ecologically invalid monocular texture leads to longer perceptual latencies in random-dot stereograms

Two experiments were conducted to explore Gillam and Borsting's (1988, Perception 17 603-608) report that uncorrelated monocular texture facilitates stereopsis by shortening the latency to see depth in random-dot stereograms. Experiment 1 used stereograms similar, in pattern but not disparity, to Gillam and Borsting's with monocular texture present or absent. A third condition, where monocular texture was dissimilar to the binocular panels and background, was also used. We were unable to generalize the findings of Gillam and Borsting for a depth step of 6 min of arc to a larger depth step of 24 min of arc. That is, we observed no significant difference in latencies between the conditions with monocular texture absent and present at a disparity of 24 min of arc. We found latencies to be significantly longer in the monocular-texture-different condition than the monocular-texture-absent condition, however. We account for this, ad hoc, by arguing that the monocular-texture-different stereogram depicts a rare or 'accidental' visual scenario. This account was supported by the results of experiment 2 which showed that stereograms depicting accidental views yielded longer latencies than those depicting generic views. We conclude that the ecological validity of monocular texture must also be considered when assessing the effects of monocular texture on stereopsis.     

How owls structure visual information

Recent studies on perceptual organization in humans claim that the ability to represent a visual scene as a set of coherent surfaces is of central importance for visual cognition. We examined whether this surface representation hypothesis generalizes to a non-mammalian species, the barn owl (Tyto alba). Discrimination transfer combined with random-dot stimuli provided the appropriate means for a series of two behavioural experiments with the specific aims of (1) obtaining psychophysical measurements of figure–ground segmentation in the owl, and (2) determining the nature of the information involved. In experiment 1, two owls were trained to indicate the presence or absence of a central planar surface (figure) among a larger region of random dots (ground) based on differences in texture. Without additional training, the owls could make the same discrimination when figure and ground had reversed luminance, or were camouflaged by the use of uniformly textured random-dot stereograms. In the latter case, the figure stands out in depth from the ground when positional differences of the figure in two retinal images are combined (binocular disparity). In experiment 2, two new owls were trained to distinguish three-dimensional objects from holes using random-dot kinematograms. These birds could make the same discrimination when information on surface segmentation was unexpectedly switched from relative motion to half-occlusion. In the latter case, stereograms were used that provide the impression of stratified surfaces to humans by giving unpairable image features to the eyes. The ability to use image features such as texture, binocular disparity, relative motion, and half-occlusion interchangeably to determine figure–ground relationships suggests that in owls, as in humans, the structuring of the visual scene critically depends on how indirect image information (depth order, occlusion contours) is allocated between different surfaces. Electronic Publication     

Texture discrimination does not occur at the cyclopean retina

The ability to segregate texture patterns at the cyclopean retina was tested with random-dot stereograms. When fused, patterns displayed arrays of texture elements which varied either in their form or in apparent depth. If elements of different form appeared at similar disparity in the random-dot stereograms, they did not provide the visual impression of distinct texture areas, although individually they could be easily discriminated. When texture elements differed in apparent depth rather than in form, segregation of different areas was readily achieved. These results restrict the possible site in the visual system for texture discrimination.     

Motion parallax as an independent cue for depth perception.

The perspective transformations of the retinal image, produced by either the movement of an observer or the movement of objects in the visual world, were found to produce a reliable, consistent, and unambiguous impression of relative depth in the absence of all other cues to depth and distance. The stimulus displays consisted of computer-generated random-dot patterns that could be transformed by each movement of the observer or the display oscilloscope to simulate the relative movement information produced by a three-dimensional surface. Using a stereoscopic matching task, the second experiment showed that the perceived depth from parallax transformations is in close agreement with the degree of relative image displacement, as well as producing a compelling impression of three-dimensionality not unlike that found with random-dot stereograms.     

Random-dot stereograms of real objects: observation on stereo faces and moulds

A simple technique for producing random-dot stereograms of real objects is described. It is then shown that stereograms of inside-out faces (moulds) cease to be perceptually reversible only when the presentation is truly cyclopean. Further possible applications of the technique are outlined.     

Conditions under which stereopsis and motion perception are blind

We describe modified random-dot stereograms in which the corresponding elements differ from non-corresponding elements in colour, size, and luminance. Despite these visible differences between the elements, depth perception collapses when the spatially integrated luminous flux is similar for the corresponding and non-corresponding elements. Our results suggest that a low-pass spatial filter precedes the mechanism that recognises disparity. A similar phenomenon is observed for the perception of coherent motion in random-dot kinematograms. Our modified stereograms and kinematograms may find other uses when experimenters wish to study the contribution of colour to visual processes and require a method of eliminating edge artifacts.     

Object recognition contributions to figure-ground organization: Operations on outlines and subjective contours

In previous research, replicated here, we found that some object recognition processes influence figure-ground organization. We have proposed that these object recognition processes operate on edges (or contours)detected early in visual processing, rather than on regions. Consistent with this proposal, influences from object recognition on figure-ground organization were previously observed in both pictures and stereograms depicting regions of different luminance, but not in randomdot stereograms, where edges arise late in processing (Peterson & Gibson, 1993). In the present experiments, we examined whether or not two other types of contours—outlines and subjective contours—enable object recognition influences on figure-ground organization. For both types of contours we observed a pattern of effects similar to that originally obtained with luminance edges. The results of these experiments are valuable for distinguishing between alternative views of the mechanisms mediating object recognition influences on figure-ground organization. In addition, in both Experiments 1 and 2, fixated regions were seen as figure longer than nonfixated regions, suggesting that fixation location must be included among the variables relevant to figure-ground organization.     

Development of visual attention: a stereoscopic view

Five-year-olds, 8-year-olds, and adults (20 each) tachistoscopically viewed random-dot stereograms containing either uniform depth or an elevated disk whose perimeter fell 1.5 degrees, 2.5 degrees, or 4 degrees beyond fixation. When subjects were forewarned as to the size of the disk to look for (precued trials), signal detection improved (d' increased) for all three groups, and 5-year-olds improved the most. Without forewarning (self-cued trials), the 5-year-olds performed more poorly than the other two groups, which performed alike. Although young children are able to allocate their visual attention selectively, they do not appear to deploy it as quickly or as effectively as their elders when the focus of attention is theirs to choose.     

Disparity detection in anticorrelated stereograms.

Recent physiological observations in which stimuli with opposite contrast signs in the two eyes have been used (anticorrelated stereograms) show that these stimuli evoke responses in primary visual cortex which are the reverse of responses to correlated stimuli. Psychophysical investigations reveal no such reversals: reversed-contrast bars with crossed disparities are seen in front of those with uncrossed disparities. For anticorrelated random-dot stereograms human subjects perceive no depth at all, except at low dot densities. However, these human studies were carried out with stimuli that differed in several ways from those used in physiological studies. We therefore reexamined psychophysical responses using stimuli similar to those used for physiological recordings. Our results confirm the previous findings: there is no evidence of a reversed depth sensation for bar stereograms (crossed disparities are never seen behind uncrossed disparities), and subjects are unable to detect depth in anticorrelated random-dot stereograms at the densities used for the physiological recordings. The discrepancy between the psychophysical data and the responses of single neurons in primary visual cortex suggests that further processing outside area V1 is necessary to provide the signals that produce the sensation of stereoscopic depth.     

Structure from stereo by associative learning of the constraints

A computational model of structure from stereo that develops smoothness constraints naturally by associative learning of a large number of example mappings from disparity data to surface depth data is proposed. Banks of disparity-selective graded response units at all spatial locations in the visual field were the input data. These cells responded to matches of luminance change at convergent, divergent, or zero offsets in the left and right 'retina' samples. Surfaces were created by means of a pseudo-Markov process. From these surfaces, shaded marked and ummarked surfaces were created, along with random-dot versions of the same surfaces. Learning of these example shaded and shaded marked surfaces allowed the system to solve stereo mappings both for the surfaces it had learned and for surfaces it had not learned but which had been created by the same pseudo-Markov process. Further, the model was able to solve some random-dot versions of the surfaces when the surfaces had been learned as shaded marked surfaces.     

The effect of control and predictability in learned helplessness

The present 2 x 2 study deals with the influence of controllability and predictability of an aversive noise stimulus on a subsequent learning task. Eighty-four subjects participated in two experiments. In correspondence with the concept of learned helplessness, controllability was shown to be the dominant factor in the first experiment. In the second experiment, a modified test task was used in which both factors were shown to act in specific ways: The main influence of controllability is upon response measures (latency, omission), whereas unpredictability retardates learning of new predictive connections. The results are discussed in terms of contingency learning.     

Learning categories by making predictions: An investigation of indirect category learning

Categories are learned in many ways, but studies of category learning have generally focused on classification learning. This focus may limit the understanding of categorization processes. Two experiments were conducted in which participants learned categories of animals by predicting how much food each animal would eat. We refer to this as indirect category learning, because the task andthe feedback were not directly related to category membership, yet category learning was necessary for good performance in the task. In the first experiment, we compared the performance of participants who learned the categories indirectly with the performance of participants who first learned to classify the objects. In the second experiment, we replicated the basic findings and examined attention to different features during the learning task. In both experiments, participants who learned in the prediction-only condition displayed a broader distribution of attention than participants who learned in the classification-and-prediction condition did. Some participants in the prediction-only group learned the family resemblance structure of the categories, even when a perfect criterial attribute was present. In contrast, participants who first learned to classify the objects tended to learn the criterial attribute.     

Global stereopsis in rhesus monkeys

In three separate experiments an attempt was made to demonstrate global stereopsis in two rhesus monkeys by using random dot stereograms projected and viewed through polarizing filters. Although both animals learned a number of discriminations, control tests showed that both were perceiving non-depth cues such as monocular identification of minute pattern differences or brightness differences caused by reflections of polarized light. In a final experiment red/green anaglyph forms of the stereograms were viewed through red/green filters. Both monkeys, together with a third experimentally naive animal, showed incontrovertible evidence of prompt discrimination based on stereopsis. This paper makes a number of recommendations about the use of random dot stereograms to demonstrate global stereopsis in animals.