OVERVIEW

Why are neural streams worth studying?

To our mind's eye, the world appears as a combination of colors, sizes, movements, locations, and other attributes of the visual world. It is tempting to assume that every one of these attributes is handled by a separate neural population, which constitutes an information channel or stream. The observation that the visual system comprises several neuronal populations or streams suggests the possibility that each stream performs a distinct function in vision. We believe that understanding the link between the properties of the neurons in each stream and those visual functions should illuminate the computations that are performed by the visual system as it analyzes the visual world. Here I shall review the current state of our understanding of the three neuronal populations that together make up the output of the primate retina: the so-called M (magnocellular), P (parvocellular), and K (koniocellular) streams. A review is timely for several reasons: (1) Recent intracellular recordings in vitro have generated detailed information about the morphology and physiology of several types of retinal cells, including some cell types that had not been characterized before; (2) a significant new population of cells (the Koniocellular, or K cells) had been identified within the lateral geniculate nucleus (LGN), with distinctive patterns of projection into the visual cortex; (3) detailed microscopic analysis of retinal circuits has revealed new information about synaptic patterns of connections; and (4) more sophisticated stimulation techniques have shed some new light on the details of the organization and dynamics of the various neuronal populations that make up the early visual system of primates.

After a brief review of the relevant anatomical structures of the primate early visual system, I shall consider several aspects of the physiological characteristics of the three main neuronal streams and discuss their functional implications.

Cell types, visual streams, and parallel information processing

When one considers diverse neuronal populations within a neural entity such as the visual system, it is important to have a clear understanding of the taxonomical notion of cell type. In the primate visual system, the M, P, and K populations are thought to consist of three different types of neurons. The concept of a neuronal type, and its relationship to visual neuroscience, has been discussed extensively in the past (see, for example, Rodieck and Brening, 1983; Stone, 1983). Usually, one refers to a group of neurons that is distinct from other groups along one or more important dimensions, such as morphology, physiology, or connectivity. Even if along any one dimension the variations among cells form a continuous transition with a unimodal distribution (say, from small to large neurons, or from fast conducting to slow conducting), when the entire cell population is examined along multiple dimensions, one typically observes distinct clusters. This point is illustrated for two dimensions in Figure 30.1.

Figure 30.1..  

Two cell populations might have a unimodal distribution along any one dimension, but when analyzed along more than one dimension (two in this example: size and response threshold), they form two distinct clusters.

A seductive and widespread hypothesis is that different cell types serve distinct functions. If the various populations also have distinct patterns of connectivity, it is often assumed that we are looking at distinct neuronal streams or functional pathways, each devoted to the representation and analysis of a different aspect of the visual world, for which each stream is specialized. It is thought that the spatiotemporal distribution of light on the two-dimensional retinal surface is transformed by the visual system into a set of higher-order representations, or abstractions, of several aspects of the visual world, such as color, size, edge orientation, or movement. The notion of parallel representation suggests, in turn, that each neuronal type should probably cover the entire visual field, so as to avoid having regions in which one cannot see some important aspect of the world, such as blue or vertical edges or movement.

We do not know currently how many major attributes of the visual world are represented by the visual system, but it is probably safe to assume that the number is larger than 3, though probably less than 20. One might try to estimate the number of possible dimensions by considering how many neuronal types are represented at any point in the retina. Such information is exceedingly difficult to obtain and is available in only a few cases, such as the elegant work of MacNeil and Masland (1998), who found 29 types of amacrine cells at every point in the rabbit retina. The relatively large number of visual dimensions suggests that each of the three major visual streams we are discussing here (M, P, and K) must deal with more than one attribute of the visual world. We hope to determine at least some of the functions of the various streams by comparisons of the respective properties along some of the dimensions mentioned above, and much of what is to follow will be taken up by that effort.

Anatomical vs. physiological classification

In discussing neuronal streams it is useful to keep in mind the distinction between labeling neurons according to (1) where they are or who they connect with, which is usually qualitative and unambiguous; and (2) their physiological (or other) properties, which can vary in quantitative measures (e.g., conduction velocity, response time course, or chromatic opponency). The latter classification is often less certain and more controversial than the former.

Properties of the M, P, and K streams: a summary

Table 30.1 provides a concise overview of our current knowledge of the three major visual streams in the primate visual system: M, P, and K. The subsequent sections will elaborate on some of the entries in the table. We note that much more is known currently about the P and M populations than about the K population, which was discovered more recently, and which is much more difficult to study in isolation. There are several extensive reviews of the M/P pathways at various stages of the visual system (Kaplan et al., 1990; Lee, 1996; Merigan and Maunsell, 1993). Less is known about the K (koniocellular) pathway, but the reader is referred to Irvin et al. (1993), Hendry and Reid (2000), White et al. (2001), and Xu et al. (2001).