Separate visual pathways: a little history

Beginning with the prize-winning work of Gasser and Erlanger (1929) on the somatosensory system, it has been recognized that different qualities within a sensory modality can be transmitted via parallel pathways that are morphologically distinct. In the somatosensory system, it was suggested that different sensations within a single cutaneous nerve (e.g., pain and temperature versus light touch and pressure) might be carried via axons of different caliber and conduction velocity. By analogy, George Bishop (1933) subsequently argued that the three groups of axons that he identified in the optic nerve of the rabbit, based on axon size and conduction latency, were evidence of parallel processing for visual qualities, although he later changed his mind and argued that axon fiber size reflected evolutionary history (Bishop, 1959). Today the idea that separate retinal ganglion cell classes transmit different sensory messages to the brain is accepted as a basic organizational principle. The issues that remain controversial concern the number of pathways that exist, homologies among pathways across species, the exact content of these pathways, and how these pathways relate to different perceptual attributes.

How we think about parallel processing is a product of several distinct approaches to the problem. Outlined below are four lines of investigation that have strongly impacted our views of parallel processing in the visual system. First, in the mid-1960s, Enroth-Cugell and Robson (1966) sparked a revolutionary shift in thinking about visual processing. Approaching the visual system from an engineering standpoint, they proposed that the visual system works as a series of spatial filters, namely, as spatial-frequency analyzers. The idea was that the visual system represents objects by tuning different cells to different ranges of spatial frequency. Enroth-Cugell and Robson (1966) used this linear systems approach to analyze the responses of ganglion cells in the cat retina. This, in turn, was followed by the application to visual psychophysics of Campbell and Robson (1968). This general approach then led to a flood of studies based on the idea that the visual system's response to any pattern could be predicted from its response to more basic components. In their original work, Enroth-Cugell and Robson (1966) subdivided concentrically organized ON- and OFF-center retinal ganglion cells in cats into two types on the basis of their spatial summation properties. Those that summed luminance changes linearly across their receptive fields were called X cells, and those that did not were called Y cells. In their report, the authors also described other features that distinguished Y from X ganglion cells including the higher conduction velocities, sensitivity to higher speeds and lower contrasts, lower spatial frequency cutoffs, larger average receptive field center sizes, and more transient responses of Y versus X cells (Enroth-Cugell and Robson, 1966). More than a decade of studies on X and Y cells followed these seminal findings. These studies showed that X and Y cells also could be distinguished based on different distributions in the retina, parallel central projections, and morphology (reviewed by Stone, 1983). From this constellation of traits, it was proposed that X cells were part of a channel to cortex that subserves high-resolution pattern vision, while Y cells were part of a channel that subserves crude form and motion vision (Stone, 1983). Also during this time, other cell types were discovered in the cat retina that were collectively called W cells. From the beginning, it was clear that unlike X and Y cells, which show relatively low within-group variability, the response properties of W cells varied widely, having perhaps little more in common than the attribute of low conduction velocity (see Kaplan, 1991; Stone, 1983, for review). Because many W cells were shown to have heavy projections to the midbrain, it was proposed that they subserved a more primitive form of vision referred to as ambient vision. X and Y cells, by contrast, provided focal vision, or vision that required cortex. Ambient vision was seen as an almost unconscious primitive ability to orient to objects and move through the environment found in all vertebrates, while focal vision was seen as the conscious vision used to identify objects typical of primates. The analysis of X, Y, and W cells in cats also led later to a similar set of investigations in primates, where both similarities and differences between cats and primates were uncovered (see Casagrande and Norton, 1991, for review). We will return to the issue of species differences in parallel processing in the next section.

The ambient/focal vision idea was actually linked to a second very influential set of investigations that also began in the 1960s. In 1969 Schneider published an important article in which he proposed that there was an anatomical separation between visual coding of the location of a stimulus and its identification. Based on behavioral/lesion work in hamsters, he argued that there were two pathways: a where pathway involving the superior colliculus and a what pathway involving the visual cortex (Schneider, 1969). The where versus what or ambient versus focal pathways were subsequently modified and described as independent pathways, one involving a pathway from colliculus to pulvinar to extrastriate cortex and the other involving the geniculostriate pathway (Casagrande et al., 1972; Diamond and Hall, 1969). The idea that these pathways were capable of independent operation was demonstrated clearly in tree shrews, in which complete removal of striate cortex and complete degeneration of the LGN does not impair discrimination of simple patterns or acuity (Ware et al., 1972, 1974). The dual pathway idea more recently has been suggested as an explanation for the blind sight exhibited by humans in the absence of visual cortex (Ungerleider and Mishkis, 1982).

In 1982 the idea of two visual systems where versus what, took a different form. Ungerleider and Mishkin (1982) proposed that visual object identification (what) depended on connections to the temporal cortex, while object location (where) required the parietal cortex. They also argued that both areas required primary visual cortex (V1) based on their own data from other lesion studies in monkeys, as well as from human clinical studies. The cortical version of the what versus where hypothesis suggested that if the two visual systems originated subcortically, they must both pass through the LGN.

A third avenue of investigations involved efforts to define pathways based on anatomy. The advent of new technologies for tracing degenerating pathways in the 1950s and 1960s, and for anterograde and retrograde transport of tracers in the 1970s and 1980s, provided details of the connections of parallel pathways from retina to the LGN and from the LGN to the cortex in several species. For example, these studies clearly showed that different retinal ganglion cell classes projected to separate cell classes in the cat LGN and to separate layers of the LGN in all primate species examined (Casagrande and Norton, 1991). In addition, studies showed that the parallel arrangement of connections from the retina continued to V1, where X, Y, and W LGN cells in cats, and parvocellular (P) and magnocellular (M) and subsequently koniocellular (K) LGN cells in primates, were shown to project in parallel to separate layers of V1 (Fig. 31.1; see Casagrande and Kaas, 1994; Sherman and Guillery, 2001, for review).

Figure 31.1.  

LGN projection patterns to visual cortex of cats and macaque monkeys. Left: projections from the LGN to V1 in the cat. X and Y LGN cells send axons to layers 4 and 6. W cells project CO blobs in layer 3 and to layers 1 and 5a. Right: projections from the LGN to V1 in macaque monkey. M and P cells project, respectively, to layers 4Cα and 4Cβ while K cells project to the CO blobs in layer 3 and to layer 1. In addition, P cells have been proposed to send axons to layer 4A. LGN projections to layer 4A are missing in apes, humans, and some other primates. Some M and a few P cells have collateral branches terminating sparsely in layer 6. See text for details. (Left: Data from Boyd and Matsubara, 1996; Humphrey et al., 1985; Kawano, 1998; Right: modified from Casagrande and Kaas, 1994, with permission of the publisher.)

Finally, in the 1980s, a set of studies was published by Livingstone and Hubel (1988) outlining their hypothesis that different attributes such as form, color, and motion were segregated within the layers and cytochrome oxidase (CO) blob compartments of V1 (Fig. 31.2). They linked various ideas described above together in a very satisfying model. According to this model, the P retinogeniculocortical pathway (form and color) projects ultimately to the what pathway ending in the temporal lobe and the M retinogeniculocortical pathway (motion) to the where pathway in the parietal lobe. Evidence to support the links between the P pathway and form/color and the M pathway and motion came primarily from physiology and connectional anatomy. Physiological studies had shown that P LGN cells exhibit chromatic opponency and have high spatial resolution, and that M cells are not selective for wavelength but exhibit high temporal resolution (reviewed in DeYoe and Van Essen, 1988; Livingstone and Hubel, 1988). Livingstone and Hubel and others provided evidence that linked the P pathway to the CO blob and interblob compartments in cortical layer 3 with appropriate output pathways to the what hierarchy of extrastriate visual areas, as well as evidence that the M pathway connected to the where hierarchy of visual areas via connections within V1 layer 4B (Fig. 31.2). The K pathway was ignored, in part, because it did not fit well with the model and had not been studied in as much detail (Casagrande, 1994; Hendry and Reid, 2000). Nevertheless, it was already known at that time that the K LGN pathway terminated in patches that appeared to coincide with the CO blobs in V1 in macaque monkeys (Livingstone and Hubel, 1982).

Figure 31.2.  

Diagram of the functional segregation of the primate visual system. MT, middle temporal area; V4, visual area 4; LGN, lateral geniculate nucleus. (From Livingstone and Hubel, 1988, with permission of the publisher.)

The degree of synthesis provided by Livingstone and Hubel's (1988) view of the visual system had a powerful impact on current thinking about the organization of the human visual system. Because of the simplicity of the model, things that did not fit were set aside as ever more streamlined diagrams of the original appeared in textbooks. Recently, however, as more data have been gathered in a variety of primates and in other species, findings have been presented that raise questions about the model. Examples are provided in the next section.