Neural mechanisms underlying stereoscopic vision

Abstract

The progressive frontalization of both eyes in mammals causes overlap of the left and right visual fields, having as a consequence a region of binocular field with single vision and stereopsis. The horizontal separation of the eyes makes the retinal images of the objects lying in this binocular field have slight horizontal and vertical differences, termed disparities. Horizontal disparities are the main cue for stereopsis. In the past decades numerous physiological studies made on monkeys, which have in many aspects a similar visual system to humans, showed that a population of visual cells are capable of encoding the amplitude and sign of horizontal disparity. Such disparity detectors were found in cortical visual areas V1, V2, V3, V3A, VP, MT (V5) and MST of monkeys and in the superior colliculus of the cat and opossum. According to their disparity tuning function, these cells were first grouped into tuned excitatory, tuned inhibitory, near and far sub-groups. Subsequent studies added two more categories, tuned near and tuned far cells. Asymmetries between left and right receptive field position, on and off regions, and intra-receptive field wiring are believed to be the neural mechanisms of disparity detection. Because horizontal disparity alone is insufficient to compute reliable stereopsis, additional information about fixation distance and angle of gaze is required. Thus, while there is unequivocal evidence of cells capable of detecting horizontal disparities, it is not known how horizontal disparity is calibrated. Sensitivity to vertical disparity and information about the vergence angle or eye position may be the source of this additional information.

Introduction

There has been among the mammalian vertebrates a constant tendency to enlarge the binocular area of the visual field at expense of overlapping both monocular fields and reducing the total space covered by both eyes together. This was achieved by a progressive eye frontalization. In sub-mammalian vertebrates there is a total decussation of the optic nerve fibers at the chiasma, and therefore both monocular inputs remain largely segregated. On the contrary, in higher mammals a complete decussation of each nasal hemiretina makes possible an early interchange of left and right eye inputs. This decussation is an efficient arrangement to bring together corresponding retinal regions of both eyes and permit a point by point matching of the visual field. Among the mammals only the carnivores, lower primates and man have developed true binocular vision. Within this group, the amount of crossing optic fibers at the chiasma proceeds parallel with the increasingly frontalization of the eyes. A recent review on depth vision in animals has been made by Pettigrew (1991).

Normal binocular vision implies fusion, a process that consists of blending together left and right images to be perceived as a single image. This seemly effortless task involves a series of complicated neural processes that are not yet well understood. Curiously, singleness of binocular vision does not confer relevant benefits, on the contrary, it represents an additional problem to solve. Thus, with the exception of stereopsis, seeing with two eyes is not much better than seeing with one.

Depth perception and stereopsis differ in that the first is based on both monocular and binocular cues, whereas the second is based on interocular retinal image differences. Depth perception is thus a broader concept than stereopsis, stereopsis being only one aspect of depth perception. The term `stereopsis' comes from the greek `stereos', which means `solid', to refer the perception of the three dimensions of the visual scene.

Under monocular vision depth can be perceived but stereopsis is not possible. Monocular form and motion alone contain robust depth cues in normal viewing conditions. There is a variety of monocular cues that provide powerful information for perception of real depth (Helmholz, 1925; Ogle, 1962b; Reading, 1983). Among the most relevant are the relative size of the objects, interposition, aerial perspective, shading, geometrical perspective, relative velocity, motion parallax and blurring. Cells of the visual system have mechanisms such as contrast sensitivity, spatial frequency filtering, motion direction and orientation sensitivity that make effective use of these cues and make them available to achieve depth perception. Although depth perception under monocular viewing can reach remarkable accuracy, it is the contribution of stereopsis which makes depth perception so highly effective in mammals.

Stereopsis is based on the interocular differences that appear between the images projected on the left and right retinas because of the horizontal separation of the two eyes. These differences are known as retinal disparities. Retinal disparities have a broad range of amplitude, and the visual system probably processes small and large disparities in different ways. Based on this idea, in 1950 Ogle postulated two kinds of stereopsis. One was termed `quantitative' or `fine stereopsis' and was related to small ranges of retinal disparities that required steady visual fixation. The other one was termed `qualitative' or `coarse stereopsis', related to large disparities and required small scanning eye movements (Bishop and Henry, 1971). Fine stereopsis requires fusion and single vision, gives a precise estimate of depth and usually operates for disparities not exceeding 2 degrees of arc (Mitchell, 1966; Fender and Julesz, 1967). Coarse stereopsis on the contrary, is based on large disparities that produce double images (Bishop and Henry, 1971). Disparities of up to 7 degrees for crossed disparities and up to 12 degrees for uncrossed disparities are believed to produce such depth information (Westheimer and Tanzman, 1956; Blakemore, 1970).

Static positional disparity is not the only difference between left and right retinal images. Velocity and temporal disparity between both retinal images may also occur. For instance, an object moving across the visual field towards the observer outside of the vertical meridian has images with different velocities of movement in the two eyes. These differences can be described as dynamic retinal disparities. On the other hand, temporal delays between the left and right information reaching the central nervous system are able to elicit depth perception similar to that induced by static and dynamic disparities. Although first described by Mach and Dvorak (1872)and later, in 1920, by the astronomer Max Wolf (see Howard and Rogers, 1995, pp. 535–536), this phenomenon was studied in detail in 1922 by Carl Pulfrich (Pulfrich, 1922). This effect, known as `Pulfrich effect', appears when an observer views a pendulum swinging in a frontal plane and one eye is covered with a filter that causes a delay of the visual stimulus. Under this conditions the pendulum appears to move in depth in an horizontal elliptical path.

Since retinal disparities may be static or dynamic, it has been suggested that stereopsis be static or dynamic, each having a number of properties that make them different.

Static stereopsis operates within a range of horizontal disparities of less that 0.33 degrees, its latency is large, about 250 msec, improves for targets located on the fovea, improves with exposure time, increases as the spatial frequency of the stimulus increases, improves when the horopter is determined and persists in presence of chromatic equiluminance (Ogle, 1962a; Ogle and Weil, 1958; Tyler and Torres, 1972; Regan and Beverley, 1973; Julesz et al., 1980; Mayer et al., 1982; Schor et al., 1984; Tyler and Cavanagh, 1989; Tyler, 1983, Tyler, 1990).

On the other hand, dynamic stereopsis operates over a broader range of disparities than static stereopsis, ranging from 0.1 to about 10 degrees, its latency is short (about 130 msec) with locations off the fovea producing less impairment in motion than in static stereopsis. The optimal temporal frequency is for targets that move a rate up of 3–10 cycles/sec, the spatial frequencies extend to 0.1 cycles/deg, and disparity sensitivity decreases when targets have equal red–green luminances (Ogle, 1932; Ogle, 1962a; Regan and Spekreijse, 1970; Tyler and Torres, 1972; Regan and Beverley, 1973; Norcia and Tyler, 1984; Schor et al., 1984; Norcia et al., 1985; Tyler and Cavanagh, 1989).

Experiments aimed to elucidate the neurophysiological basis of stereopsis are numerous. Barlow et al. (1967)reported for the first time the finding that some cortical units in cats were optimally excited by objects lying at different distances. Since then numerous studies were made in cat (Nikara et al., 1968; Pettigrew et al., 1968; Henry et al., 1969; Joshua and Bishop, 1970; Bishop et al., 1971; von der Heydt et al., 1978; Fischer and Krüger, 1979; Ferster, 1981), monkey (Hubel and Wiesel, 1970; Poggio and Fischer, 1977; Maunsell and Van Essen, 1983; Poggio and Talbot, 1981; Poggio et al., 1985, Poggio et al., 1988; Burkhalter and Van Essen, 1986; Hubel and Livingstone, 1987, Hubel and Livingstone, 1990; Felleman and Van Essen, 1987; Roy et al., 1992; Gonzalez et al., 1993a, Gonzalez et al., 1993b) and sheep (Clarke and Whitteridge, 1974, Clarke and Whitteridge, 1976; Clarke et al., 1976; Ramachandran et al., 1977). These studies demonstrated that cells in certain structures of the visual system have responses that are dependent on the binocular disparity of the stimuli.

Macaque monkeys seem to have a visual system very similar to that of humans (Cowey and Ellis, 1967; Farrer and Graham, 1967; DeValois and Jacobs, 1968; Harwerth et al., 1995). Behavioral studies have shown a remarkable similarity in both the oculomotor and sensory aspects of binocular vision in both species, such as sensitivity to binocular disparity (Sarmiento, 1975), binocular fusion (Harwerth et al., 1995) and ability to distinguish random dot stereograms at different apparent depths in absence of monocular cues (Bough, 1970). These observations suggest that the operating characteristics of vergence and stereopsis are virtually identical in macaque monkeys and humans.

This review will focus on the neural mechanisms which are believed to be responsible for the detection of retinal disparities, which are the basis of stereopsis. In this review, the term `disparity' indicates positional static horizontal disparity. In other instances the full term will be used.

As we shall see, observations made in the last decades, mostly from single cells recordings in cortical visual areas revealed neural mechanisms that undoubtedly underlay the visual processes leading to stereopsis. Since investigations on macaque monkeys have provided much of our knowledge of these mechanisms, most of the data referred to here comes from studies carried out on these animals. Although fluctuations in ocular position have important consequences on many aspects of binocular vision and stereopsis, the relationship between ocular movements and binocular vision is beyond the scope of this review and will not be considered in detail.