Magno-, Parvo-, Koniocellular Pathways
Magno-, parvo-, and koniocellular pathways are the three visual pathways in primates. These pathways are established at the level of the lateral geniculate nucleus (LGN) of the thalamus. They are formed of morphologically distinct cellular layers that receive information from different types of retinal ganglion cells and project to different layers in the primary visual cortex.
The LGN layers of each of the three visual pathways have a specific cytoarchitectonic structure. The names of the pathways are derived from these structural characteristics. Magnocellular (M) cells have relatively large bodies (lat. Magnus: large) and are found in the lowest two layers (layers 1 and 2) of the LGN. Parvocellular (P) cells have smaller bodies (lat. Parvus: small) and are found in the top four layers of the LGN (layers 3, 4, 5, and 6). Koniocellular (K) cells (gr. Konios: dust) are very small and form six thin layers. The first K layer is positioned ventral to the first M layer, while the others are intermediate to the six LGN layers. The M cell axons terminate primarily in layer 4Cα of the primary visual cortex (V1) while the P cell axons terminate primarily in layers 4Cβ and 4A of V1 . The six K layers are further subdivided on the basis of their projections . The two dorsal K layers project into layer 1 of the primary visual cortex. The two intermediate K layers receive information from small bistratified retinal ganglion cells and project it into layers 3 and 4A of the V1. The two ventral K layers project into the superior colliculus.
Structural differences between the three pathways emerge at the level of the retina as each pathway receives projections from a somewhat different group of retinal ganglion cells. Although retinal ganglion cells that ultimately project to the P layers are heterogeneous, midget ganglion cells form the bulk of the P pathway. The receptive fields of P cells which receive inputs from midget ganglion cells oppose inputs from Long- (L) and Middle-wavelength (M) sensitive cones (L-M; reddish-greenish dimension). On the other hand, the M pathway is based on inputs from parasol ganglion cells. Receptive fields of these cells combine inputs from L and M cones (L + M; black-white dimension). The two intermediate K layers receive projections from small bistratified cells and from several other cell types with Short-wavelength (S) cone input. The receptive fields of such K cells oppose the excitatory input from S cones with the inhibitory input from L and M cones (S-(L + M); bluish-yellowish dimension). The distinction between the processing of luminance signals (L + M) and the processing of two types of chromatic signals (L-M and S-(L + M)) by geniculate cells had led to their conceptualization as the three orthogonal visual mechanisms. Such a conceptualization is implemented in the DKL space, a physiological color space based on the seminal study of macaque LGN cell response properties conducted by Derrington, Krauskopf, and Lennie in 1984 .
Based on his observations of three different types of axons in the rabbit optic nerve Bishop speculated in 1933 that visual information is likely to be processed along morphologically distinct pathways that operate in parallel . An in-depth historical overview of research into parallel visual pathways is provided in the review by Casagrande and Xu . By the end of the 1980s, pathway-tracing studies on primates confirmed Bishop’s speculation, establishing that different retinal ganglion cell classes projected to separate layers in the LGN, which subsequently projected to separate cortical layers. Numerous physiological studies uncovered marked differences in spatial, temporal, and contrast sensitivities between various cell types that were found in the retina, LGN, and the primary visual cortex. Early studies of primate LGN cell properties attempted to relate their findings to extensive research performed on cats since the 1950s. By the 1970s, cat retinal ganglion cells had been classified into X, Y, and W types based on their physiological properties and responsiveness to different types of stimuli: for example, X cells, which were the most numerous, were responsive to spatial detail, and the less numerous Y cells were responsive to motion. The W cells were found to have heterogeneous properties and were the least understood. Researchers linked different LGN layers to X, Y, and W cell types on the basis of observed similarities in their spatiotemporal response properties and sensitivities to various visual attributes. At the same time, attempts were also made to relate the newly found geniculate visual pathways to the ventral or “what” cortical stream, which was proposed to subserve the processing of objects’ visual attributes, and the dorsal or “where” cortical stream, which was proposed to subserve the processing of objects’ locations. These streams were proposed on the basis of findings from lesion studies on monkeys conducted by Ungerleider and Mishkin in the early 1980s . In 1988, Livingstone and Wiesel proposed an integrative model of retinogeniculocortical pathways , linking the M pathway to the processing of motion due to the transient responses and high achromatic sensitivity of its neurons and the P pathway to the processing of form and color due to its neurons’ sustained response to chromatic contrast and weak response to achromatic contrast. Livingstone and Hubel’s model epitomized the modular approach that was popular at the time, relating discrete brain structures to cognitive or perceptual functions. The koniocellular (K) pathway was ignored by Livingstone and Hubel’s model. At that time, the K layers in primates were considered to be too thin to contribute substantially to any single cortical module. Between the 1950s and 1970s, the K pathway was generally studied in bushbabies and lorises, which are nocturnal prosimian primates . Subsequent studies confirmed that K cells in these species were quite similar to those in other primates. Since then, a range of studies was performed on the primate K pathway, revealing diverse functional properties of K cells, including their role in color vision. At the same time, much more has become known of retinal processing, with 20 different types of retinal ganglion cells identified . Cortical projections of M, P, and K cells have also been mapped more extensively, as will be discussed in the next section. These latest findings have brought into contention the actual number of parallel visual pathways and the degree to which they should be perceived as functionally distinct.
Numerous neuroanatomical and neurophysiological studies have investigated the properties of the three parallel pathways. As noted in the historical overview, continuous attempts have been made to relate the pathways to the processing of different visual properties: motion processing, form processing, luminance and color processing, and so on. As evidence accumulated it became more and more clear that simple and direct associations between physiological responses and perceptual properties are difficult to establish . In fact, M, P, and K pathways can only tentatively be treated as singular entities: for example, the M pathway contains cells that project information into dorsal cortical areas subserving motion processing, which fits Livingstone and Hubel’s model, but it also contains cells that project information into ventral areas subserving object representation .
The most robust and the least disputed evidence for differential processing between M, P, and K pathways comes from primate color vision research. Strong links have been established between the M pathway and the luminance (L + M) channel. The chromatic L-M channel, which computes the difference between L and M cone signals providing a reddish-greenish dimension of color, maps on to the cells of the P pathway. Furthermore, the L-M cone-opponent mechanism provides signals that can be transmitted in combination with luminance information . The view that such “multiplexing” through the L-M channel underlies achromatic vision was based on the fact that P cells are much more numerous than M cells and was initially strengthened by observations of poor spatial acuity of M cells, although recent data showed much smaller differences in M and P cell receptive field center sizes than originally found. In fact, there was an opposing view that strict subcortical segregation of luminance and L-M signals would be beneficial, since it would optimize signal transmission in a noisy, bandwith-limited system such as primate vision .
Unlike P cells, K cells that receive S-cone input from small bistratified cells do not generally contribute to luminance processing. Their receptive fields are larger than those of P cells, compromising their usefulness for high-acuity spatial vision. However, S-cone sensitive K cells may still contribute to functions other than just color appearance. They project directly to motion-sensitive cortical area (MT/V5), and their signal is also amplified cortically, being weakly present in many cells from V1 onward. Projection of S-cone signals into area MT/V5 could underlie their contribution to motion processing .
While it is not possible to fully equate geniculate pathways with psychophysically identified luminance and chromatic channels, a certain degree of functional overlap between them is evident and has been used in many studies that rely on luminance or chromatic contrast to attempt to isolate or bias visual processing toward different pathways. As M pathway dysfunctions have been hypothesized to underlie schizophrenia, autism spectrum disorders, dyscalculia, and dyslexia, the ability to isolate this pathway through carefully designed stimuli is of keen interest to clinical neuroscience. Studies in this field do not only compare responses to luminance stimuli with those to isoluminant stimuli but also often rely on other well-known processing differences between parallel visual pathways. Fast temporal frequency, low luminance contrast, and low spatial frequency are generally thought to bias processing toward the M pathway, while low temporal frequency, high chromatic contrast, and high spatial frequency are assumed to bias processing toward the P pathway. It is paramount to be careful when interpreting the findings of such studies. For example, besides transient response neurons, the M pathway also contains some neurons with sustained responses. Similarly, while contrast sensitivity functions for luminance and chromatic stimuli do indeed differ, the majority of spatial frequencies that lead to larger excitation in one pathway happen to concurrently cause some excitation in the other pathway. Therefore, the level of bias to one or the other pathway is very difficult to quantify. Finally, responsiveness of M, P, and K pathways to various visual features cannot be assumed to translate directly into perceptual sensitivity for these features. For example, P cells with receptive fields in the parafovea are sensitive to a wide range of temporal frequencies, but human observer performance reveals that the sensitivity of the visual system does not match the sensitivity that is present in this geniculate circuitry . Thus, some of the sensitivity evident in P cells is not utilized at subsequent stages of processing that drive behavior.
Isolating subcortical processing streams or biasing processing toward a certain pathway is ultimately complicated by the fact that the separation of signals from different visual pathways is not maintained in the cortex. Receptive fields in V1 are larger than in the LGN, indicating that each V1 cell receives input from several LGN cells. S-cone signals become more prominent in the cortex, with weak S-cone signals being found in many V1 neurons. Finally, LGN cells with chromatic sensitivity show preferential responses to stimuli that align with the S-(L + M) and L-M mechanisms, but V1 cells have much more widely distributed color preferences. This indicates that S-(L + M), L-M, and L + M signals start interacting in the cortex. In fact, most of the early visual cortex contains neurons that are tuned to both color and luminance, while also being sensitive to certain spatial properties, e.g., orientation .
Magno-, Parvo-, and Koniocellular pathways are established at the level of the LGN. They receive inputs from different retinal ganglion cells and project to different layers of the V1, leading to parallel processing of visual information. Parallel visual pathways have been extensively researched both in terms of their structure and in terms of their functional properties. Geniculate pathways are of interest for color scientists as the retinal projections they receive differ in chromoluminance content: while the M pathway is solely devoted to the processing of luminance information, the P and K pathways also subserve chromatic processing. Other differences between pathways concern contrast sensitivity and spatial and temporal processing, with the M pathway being preferentially responsive to high temporal but low spatial frequencies and to low luminance contrast. In spite of these recognized differences, studies that attempted to relate these pathways to specific visual functions have met with limited success. Such research is complicated by the fact that data on physiological responses of M, P, and K cells does not always correspond to behavioral data obtained from human observers under the same conditions. This is at least partly caused by the fact that visual signals undergo much further processing in the cortex, as signals from different pathways get recombined from primary visual cortex onward. There is growing evidence that contributions of M, P, and K pathways to various perceptual functions are not as segregated as was once thought. Recent research has also revealed that subcortically there are likely to be more than just three parallel visual pathways, based on the evidence that the processing in the retina itself is already very complex and driven by a wide array of cells with different inputs and connectivity.
- 1.Casagrande, V.A., Xu, X.: Parallel visual pathways: a comparative perspective. In: Werner, J.S., Chalupa, L.S. (eds.) The New Visual Neurosciences, pp. 494–506. MIT Press, Cambridge (2004)Google Scholar
- 4.Bishop, G.H.: Fiber groups in the optic nerves. Am. J. Physiol. 106, 460–470 (1933)Google Scholar
- 5.Ungerleider, L.G., Mishkin, M.: Two cortical visual systems. In: Goodale, M.A., Mansfield, R.J.W. (eds.) Analysis of Visual Behavior, pp. 549–586. MIT Press, Cambridge (1982)Google Scholar
- 8.Kaplan, E.: The M, P and K pathways of the primate visual system revisited. In: Werner, J.S., Chalupa, L.S. (eds.) The New Visual Neurosciences. MIT Press, Cambridge (2012)Google Scholar
- 9.Stockman, A., Brainard, D.H.: Color vision mechanisms. In: Bass, M. (ed.) OSA Handbook of Optics, 3rd edn, pp. 11.1–11.104. McGraw-Hill, New York (2010)Google Scholar