1.2.3  Layout of the Subcortical Pathways

In addition to the visual pathway to LGN and striate cortex, which receives the great majority of retinal ganglion cell axons, there also are subcortical pathways, formed by optic nerve fibers projecting to pretectal nuclei and to the pulvinar. In primates, the projections to the pregeniculate nucleus and pulvinar are thought to be of minor importance, and may be thought of as anatomical remnants: In lower mammals, ablation of striate cortex at birth allows these projections to greatly increase in density, leading to the development of crude functional vision, but similar experiments in newborn monkeys show neither the proliferation of projections nor appreciable acquisition of visual function [14, 15]. Other projections, however, in particular those to the pretectal nucleus of the optic tract (NOT) and the terminal nuclei (TN) of the accessory optic system, have been demonstrated to play an important role in the rapid control of eye position through vestibulo-ocular reflex, saccades, and sustained fixation [33, 43].

Detailed studies of anatomy and physiology of the primate eye movement system over the last several decades in awake, trained animal models, have shown that the NOT receives information on “retinal slip,” i.e., generalized displacement of the retinal image [58]. This retinal slip signal is encoded as a velocity signal, and serves as input to the neural integrator in the nucleus prepositus hypoglossi [50]. Pathways between the NOT and primary visual cortex (as well as multiple similar projections between cortical and subcortical structures) are also known to exist, and have been shown to compensate in part for lesions to the NOT or its retinal input [23, 31].

1.3  An Overview of Human Visual Function

The anatomy and physiology of the visual system presented above can help us understand many of the properties of normal vision, and some of the vision defects experienced by patients with blinding eye diseases. We will briefly discuss the aspects most pertinent in understanding the requirements for neural visual prostheses.

1.3.1  Roles of Central (Foveal) Vision

Central visual function is more than just the utilization of the denser packing of photoreceptors in the central retinal area and the higher density of ganglion cells per photoreceptor in this area. These properties of the retina would account for basic properties such as good two-point resolution, but they would not explain why foveal vision is superior to peripheral vision in many other ways. The following major areas of foveal specialization should be considered.

Spatial integration tasks, e.g., hyperacuity. Normally sighted observers have the ability to resolve small deviations in alignment of parallel or abutting lines, small angular differences and displacements, all on a scale well below 1 arcmin, the spacing of foveal photoreceptors. Such “hyperacuities” apparently rest on the ability of foveal projections in cortical areas to combine the precise positional coding of earlier stages in the visual system over increasing distances, using feedback and tuning mechanisms that have been honed by years of experience. The notions of learning and tuning are supported by the lack of hyperacuity in subjects with inherited abnormalities of foveal development and eye movements [64] or with developmental deficits [9], and by the gradual acquisition of hyperacuity performance throughout childhood [69].

Stereopsis. Combination of the signals from corresponding locations in the two ­retinas, in a highly systematic fashion, is required for perception of depth in stationary three-dimensional scenes. This function takes place at and beyond the V1 cortical level [55]. Fusion of the two retinal images on an object of interest defines a curved plane, the horopter, formed by the collection of points being imaged at exactly corresponding locations on the retinas of the two eyes. Finely-tuned disparity neurons detect left-right eye correspondence of retinal locations for points slightly in front of (crossed disparity) or beyond (uncrossed disparity) the horopter, with resolution on the order of arc seconds, similar to that seen in hyperacuity task performance.

Complex pattern recognition and discrimination tasks, e.g., face recognition. Beyond the ability to make precise visual judgments enabled by the high resolution of foveal vision, normally-sighted observers acquire great skill at memorizing, recognizing, and discriminating among patterns, varying from feature discrimination in the natural environment, such as recognizing human faces, to the processing of complex man-made forms and objects, such as reading text or maps. These capabilities require both high-level visual processing skills and cognitive brain functions such as leaning and memory. It is not necessarily true that these specialized skills cannot be acquired in peripheral vision: Certainly, a person with a central scotoma (blind area) due to macular degeneration can read, if given text with enough magnification and contrast [26, 40]. Nonetheless, these skills appear to depend critically on specialization during early phases of development, and functions such as reading, that once were linked to foveal visual function, can only partially, and with great effort, be taken over by extrafoveal vision, as if the task of vision itself has to be re-learned [28]. On the other hand, children with poorly developed foveal vision, such as those with albinism or aniridia, can learn to proficiently read and recognize patterns or faces, if given adequate magnification, and the same intensive exposure as their normally-sighted peers [30].

Visuomotor integration tasks, e.g., handwriting. These tasks are very similar to the pattern recognition tasks described above, in that they require complex visual ­processing and memory functions, but moreover they require integration with proprioceptive and motor functions distributed across many different brain areas. Some of these tasks may depend less critically on foveal function, but inasmuch as they are based on skills learned during early development, their execution often proves difficult when foveal vision becomes impaired later in life [44].

Note that all the skills referred to as specific for foveal vision involve the ability to see fine detail, combined with extensive learning throughout the critical period of development.

1.3.2  Roles of Peripheral Vision

The role of peripheral vision in performing daily activities is often underappreciated. Most human-designed visual tasks rely on the perception of detailed shapes, but there are important exceptions: Noticing traffic off to the side while driving and keeping track of fellow and opponent players during team sports require continuous processing of events and objects throughout the visual field. Similarly, observing wildlife and other outdoor activities require the use of our entire visual field. In almost all cases, these visual tasks require us to perceive motion and other changes, and it should come as no surprise that the evolution of the vertebrate visual system has favored use of the periphery for precisely these functions [6]. On the other hand, our attention tends to be focused on objects and events in central vision, whereas school children with severely impaired central vision appear to use their peripheral vision much more efficiently than normally-sighted individuals [65].

One of the surprising aspects of peripheral vision is how much of it can be lost before a person becomes aware of the change. Thus disorders such as glaucoma and retinitis pigmentosa may go undetected well beyond the point where irreversible damage to cells in the peripheral retina has occurred [13, 42].

1.3.3  Roles of Dark-Adapted Vision

As was mentioned above, cones are unable to function effectively at light levels below 0.003 cd/m². At these low illumination levels, rod photoreceptors continue to be effective, by virtue of the high gain, multi-stage phototransduction cascade in the rod outer segment. On the other hand, at intensities above 3 cd/m² rod function is actively suppressed. Rods are not distributed evenly throughout the retina: The center of the retina, with a diameter of approximately 5°, forms a rod-free zone, and the highest density of secondary retinal cells receiving rod signals is situated between at eccentricities between 5 and 10°, as evidenced by the common experience that a dim object at night is best observed by intentionally looking slightly away from the object.

Dark-adapted vision differs from daytime vision in two important respects, both related to the need for maximum sensitivity, i.e., the detection of a very small stimulus signal in the ongoing background of visual noise (spontaneous activity of retinal cells). Ganglion cells in the dark-adapted retina integrate signals from a much larger number of photoreceptors than in the light-adapted retina [60], and the time course over which this integration takes place is significantly extended [7].