Cerebral Cortex Research Paper

‘plastic’ kind, i.e., they are modified through learning and thus incorporate knowledge about the world; and (c) lesions of the cerebral cortex often impair behavior in a realm that clearly belongs to the psychological level, such as language, orientation, and perception. This pre-eminence of the cerebral cortex in the control of complex behavior can be related to some of its anatomical and physiological peculiarities.

1. The Structural Type Of The Cortex

The term cortex defines a class of brain structures characterized by an essentially two-dimensional layout. A ‘vertical’ organization along the ‘thickness’ of the cortex is repeated almost identically throughout the ‘plane’ of the cortex. The thickness of the cerebral cortex in humans varies between about 2 and 4 mm (in the mouse 1 mm) while in the plane the cortex of one hemisphere covers an area of about 1000 cm² (in the mouse about 1 cm² ). Besides the cerebral cortex of mammals, the cerebellar cortex, the optic tectum, and many other structures of vertebrate and invertebrate brains are built according to a similar two-dimensional scheme. What they all have in common is a diffuse projection of inputs over the whole plane, which is transformed by a unitary operation, varying from cortex to cortex, into the output. In many places the plane of the ‘cortex’ represents point to point some external input space such as the visual field or the surface of the body. Also, at every point of the cortical plane, output fibers leave the cortex to reach a variety of destinations, including other parts of the cortex itself. In the cerebral cortex of mammals both input and output fiber systems are on the same side of the cortical plane, forming the so-called ‘hemispheric white matter’ (with some minor exceptions in a marginal region, where the olfactory input enters the cortex in the uppermost layer, see olfactory input O in Fig. 1). In addition to the cortico-cortical fibers, sensory input fibers and motor output fibers shown in Fig. 1, the white matter contains a diffuse system of cortical fibers reaching the basal ganglia, as well as a system of two-way connections between cortex and thalamus.

2. Layers

As in other cortices, the input from distant places reaches the cerebral cortex predominantly at a special level of the cortical thickness. Similarly, the output to other parts of the brain takes its origin from another level. This distinction of input and output levels, together with the levels where most of the internal traffic of the cortex takes place, is at the origin of the well-known laminar structure of the cerebral cortex. The most common distinction is one of six layers, numbered (usually by Roman numerals) from the top down (from the free surface to the white matter). The most characteristic features of the various layers with respect to input and output are the following: the upper layers, layers I to III, are devoted to communication between distant parts of the cortex within or between hemispheres. Layer IV is the level at which sensory input fibers terminate (relayed from the thalamus), or fibers mostly from other parts of the cortex directly relaying such inputs. Layer V sends fibers to the basal ganglia and to distant parts of the brain or to the spinal cord. Layer VI communicates with the uppermost layers, as well as with the thalamus.

All of these statements are only statistically valid: a certain amount of thalamic fibers can also reach layers I, III, and VI, and layers IV to VI also participate to a certain degree in cortico-cortical communication. In spite of this, the distinction of the layers also gains support in the appearance of histological sections through the cortex. When the neural cell bodies are stained, the layers differ both in the number and the size of the neurons they contain. They also differ in the density of fibers in myelin preparations (Fig. 2). These differences are certainly related to the different roles the layers play in cortical information handling.

The complexity of interactions both within and between layers can be appreciated by staining individual neurons in their entirety, for example with the time-honored Golgi method or with the modern techniques of intracellular injection of dyes. These methods show cortical neurons as three-dimensional devices which collect signals in the region of their dendritic trees and distribute them in the region (or regions) of their axonal terminations. The diameters of axonal and dendritic trees greatly exceed (by a factor of 10 to 100 or more) the distance between the corresponding cell bodies in the tissue. The result is an extremely dense felt of cell processes in which each neuron is interwoven with about 100,000 other neurons. Taking this into account, it follows that the borders between the layers cannot be as sharp as they sometimes appear in textbook diagrams, since they are always crossed by dendrites and axons of many neurons in adjoining, and even more distant layers (Fig. 1).

3. Types Of Neurons In The Cortex

The pattern of axonal and dendritic ramifications varies a great deal between individual neurons, depending on their localization in different layers and in different parts of the cortex. This led to the classification of a great number of neuronal types, where one overriding distinction is now accepted by most authors, that of the spiny neurons (often subsumed under the term pyramidal cells or Type I neurons), and of the spineless neurons (often subsumed under the term stellate or Type II neurons). This distinction is supported by differences in the fine structure of membrane specializations as they appear in the electron microscope both on the dendritic and axonal tree, and by electrophysiological findings. Briefly, spiny neurons receive most of their synaptic input on ‘spines,’ i.e., small processes emanating from the dendritic tree, and make ‘excitatory’ synapses onto other neurons via their axonal tree. Spineless neurons receive their synapses directly on their dendrites and ‘inhibit’ other neurons via their axonal tree. This distinction coincides with another anatomical feature: most spiny neurons have an axon which descends to the white matter and makes both, ‘short-range connections’ via local axon collaterals and ‘long-range connections’ somewhere else in the brain or—in most cases—somewhere else in the cortex (Fig. 1). In contrast, the axon of a spineless stellate cell does not enter the white matter and only contacts other neurons in its vicinity.

Spiny neurons are the great majority (about 85 percent) of all neurons in the cerebral cortex. This category largely coincides with that of pyramidal cells of older classifications, characterized by a bipartite dendritic tree, with several ‘basal dendrites’ distributed around the cell body, and one ‘apical dendrite,’ ascending vertically through the cortex and ramifying in upper layers (Fig. 1). There are, however, also spiny neurons without an apical dendrite, sometimes termed ‘spiny stellate cells,’ especially as recipients of primary sensory input in layer IV of primary sensory areas. Disregarding a subpopulation of these which does not project to the white matter, the basic connectivity of cortical neurons can be described as follows (Braitenberg and Schuz 1998).

4. The Basic Connectivity Of The Cerebral Cortex

The ‘skeleton cortex’ (Fig. 1) consists of a large number of pyramidal cells (of the order of 10 in humans), distributed throughout all layers of the cortex, and producing excitatory synapses which for the most part again contact other pyramidal cells. There is considerable divergence and convergence built into this system, since every pyramidal cell communicates with many thousands of other pyramidal cells both in its input and in its output. Thus, ‘diffuse excitatory feedback’ is the pre-eminent feature of the cerebral cortex. Only a small percentage of pyramidal cells project to other parts of the brain and only a few percent of the synapses in the cortex come from neurons from other parts of the brain. By far the greatest part of synaptic traffic in the cortex is internal and only involves cortical neurons.

The inhibitory stellate cells are diffusely distributed in the network of pyramidal cells. In some cases their inhibitory synapses have a strong grip on a relatively small number of cortical neurons in their vicinity (pyramidal and stellate). In other cases their inhibitory synapses are distributed more profusely, making the distinction of several classes of stellate cells possible according to the pattern of their ramification (basket cells, chandelier cells, double bouquet cells, etc.; see Peters and Jones 1984). They, too, receive their input mainly from cortical pyramidal cells. In the regions where primary sensory input (e.g., visual) enters the cortex, they may also be directly contacted by the incoming fibers.

Although the inhibitory interneurons in some cases are certainly involved in the computation which takes place in the cortical network, their role may be considered as ancillary with respect to that of the pyramidal cells. They may perhaps put a brake on neuronal activity when it threatens to explode in a runaway reaction, as may be expected in a network of elements, such as the pyramidal cells, all exciting each other. Indeed, this braking action occasionally fails, as evidenced by epileptic fits, one of the commonest forms of functional derailment in the cortex.

The comparison with other parts of the brain shows that a network of mainly excitatory connections is indeed a striking peculiarity of the cerebral cortex. In the other major parts of the brain the majority of neurons are either connected into an inhibitory network (basal ganglia) or they do not form a network among themselves but relay some input to some output in a feed forward manner (cerebellar cortex, thalamus).

5. The Basic Function

The question of what could possibly be the advantage of an immense network of interconnected excitatory neurons whose increase in size has accompanied mammalian evolution up to the crowning event of human culture, has received various tentative answers. The most convincing interpretation of cortical structure is based on the observation that the synapses between pyramidal cells, the majority of all synapses in the cortex, are of the special kind residing on dendritic spines, and on the supposition that spine-synapses are ‘plastic,’ i.e., modifiable by ‘learning.’ There is no definitive proof of this supposition, but enough indirect evidence to make it plausible (see review by Horner 1993), and, moreover, alternative explanations for the role of spines are less convincing. Be this as it may, there is little doubt that, of all parts of the brain, the cerebral cortex is the one most concerned with the acquisition of knowledge. Suffice it to say that lesions of the cerebral cortex impair complex acquired capacities, such as language.

In terms of engineering, the network of cortical pyramidal cells could be likened to a giant ‘associative memory,’ a device which connects together more strongly (through modifiable synapses) the neurons that are often active at the same time. Thus, events of the outside world which tend to present themselves together will be represented in the brain by neurons tied together by strong synapses. The idea that the cortex incorporates knowledge by repeating, in its synaptic connectivity, the structure of reality was spelled out in two well-known theories. In 1949, D. O. Hebb proposed ‘cell assemblies’ as the units of cognitive operations in the brain. These are ensembles of cortical neurons connected to each other by excitatory synapses which have been strengthened in the course of a learning process. Such cell assemblies may be thought of as representing ‘objects’ of the world. Due to activity reverberating among its neurons, such an assembly stays active for some time once it has been activated, a property which perhaps is at the basis of ‘short-term memory.’ Also, a cell assembly may become active in its entirety even if only a subset of its neurons is activated initially, in a way reminiscent of the phenomenon of ‘pattern completion’ well known to perceptual psychologists.

‘Synfire chains’ (Abeles 1991) are the other theoretical proposal based on associative memory. Quite compatible with the anatomy of the cortical network and with the physiological properties of single neurons, it is possible to imagine sets of neurons, each set when activated in synchrony, activating another such set, and this in turn another one, etc., forming long chains of activity propagating through the cortex with great selectivity and temporal precision. This scheme explains how events displayed in time (e.g., the words of a language, musical themes, complex movements, etc.) are incorporated in memory. The temporal precision postulated by this theory has been impressively verified by correlation studies on spike activity of different cortical neurons (Abeles and Prut 1996).

Between Hebbian cell assemblies representing ‘objects’ of the world, and Abelesian synfire chains representing ‘events,’ there may be transitional forms of cortical activity, all based on the idea of modifiable synapses embodying the notion of synchronous events, or of events occurring in succession, acquired through a process often called ‘Hebbian learning.’

6. Cortical Areas

From the ‘macroscopic layout’ of the cortical network it is also possible to gain insights into its organization. It has been known for some time that restricted lesions of the cortex produce different symptoms according to their localization. This led to the definition of ‘cortical areas,’ distinct regions of the cortex a few centimeters across (in humans) whose reality has since been confirmed by many detailed electrophysiological studies. There are ‘sensory areas’ (visual, acoustic, somatosensory, olfactory), ‘motor areas,’ and ‘association areas.’ The latter in part have been recognized as secondary and tertiary stations in the elaboration of primary sensory input, as in the case of special areas for the extraction of motion (area V5) or color (area V4) in the visual scene (Zeki 1993), or in the case of a special area (Wernicke’s area) fed by acoustic input in the context of language, distinct from other areas concerned with acoustic perception. Evidently, besides the various sensory inputs reaching the cortex in separate regions, and the long efferent (e.g., corticospinal) axons emanating from other regions, it is the context in which the cortex operates locally that defines the areas.

Apart from the different effects of lesions, cortical areas were also delimited on the basis of subtle, and sometimes also quite evident, differences in the appearance of their layers and in the degree of myelination (‘cortical architectonics’). For example, a small-celled layer IV is particularly well developed in primary sensory areas while it cannot be discerned at all in the primary motor area (Area 4). The primary motor area sticks out because of a population of very large cells in layer V, the cortico-spinal neurons. Large pyramidal cells in layer III are found in some association areas, including the speech centers. Two heavy bands of horizontal myelinated fibers (the ‘stripes of Baillarger’) are evident in some areas and only one such band in others (Fig. 2). Overall myelination tends to decrease towards higher association areas. Such structural differences were at the basis of ‘cortical maps,’ the best known of which, the map by K. Brodmann from 1909, distinguishes about 50 areas. Although this large number was met with skepticism at first, it is remarkable that many of the structural distinctions were later shown to correspond to different properties of neurons when cortical mapping was undertaken anew by microelectrode analysis. More recent maps tend to assume even higher numbers of areas.

Evidently, the neuronal network of the cortex, although built according to common principles throughout its extent, is subject to variations that adapt it locally to different kinds of input and to different kinds of computation, which the input is subjected to. The variations which appear in the Nissl (cell body) picture and in the myelin preparations (cyto and myeloarchitectonics) are just two aspects of the underlying variations of the basic network, and indeed the two were shown by Hellwig (1993) to be related to each other by a set of simple rules.

7. Columns

Beyond the confirmation of the different nature of individual cortical areas, microelectrode neurophysiology in many cases revealed an even finer mapping within areas, the so-called ‘columns.’ In the primary visual area V-1 (Brodmann’s area 17) certain properties to which neurons are tuned, such as responding to stimuli from the right or left eye, or to vertical, horizontal, or oblique stripes, periodically recur over the surface of the cortex at distances of about 0.5 or 1 mm (Hubel 1988). The term ‘column’ was chosen for these small compartments within the areas since the locally specific properties of neurons in one column tend to be similar for neurons in all, or in several layers. Columns (or slabs) of neurons with similar properties, but different in a systematic way from column to column have also been described in secondary visual areas, in the somatosensory cortex, and in the primary acoustic cortex. In many cases, these functionally defined columns can be attributed to the way the input is organized: inputs from different sources tend to enter the cortex in alternating bundles (e.g., right eye vs. left eye in the visual cortex, fibers from the same or the opposite hemisphere in the auditory and prefrontal cortex), thus imposing periodical inputs onto a largely homogeneous looking intracortical network. However, periodicity can also be found in the pattern of axonal arborizations of pyramidal cells within cortical areas where they preferably project to columns with similar functional properties (Malach et al. 1997, Yoshioka et al. 1996). Occasionally, columns may also be visible in the arrangement of cell bodies and dendritic trees (whisker representation in rodents), or as the ‘cytochrome oxidase blobs’ of the visual cortex. However, neither areas nor columns can be regarded as separate entities; the fiber felt is continuous across the borders between them.

8. Connections Between Areas

The reality of cortical areas is also evident in the pattern of connections between them. This can be studied (in animals) by injecting certain dyes locally into the cortex, which are then transported in axons either in the direction away from the cell body (‘anterograde transport’) or towards it (‘retrograde transport’) depending on the particular dye chosen. The overall picture is the following (Young et al. 1995): individual areas are connected to several other areas, but not to all. Areas can differ considerably with respect to the number of areas they are connected to. Some primary sensory areas are connected to only a few other areas; higher association areas can be connected to more than one third of all cortical areas. The majority of the connections between areas are reciprocal. In the monkey cortex, if an area A projects to another area B, there is also a projection from B to A in 82 percent of the cases. Areas that are located close to each other in the cortex, often related to the same sensory modality, are more likely to be connected, but there are also connections between areas quite far apart. Most areas are connected to their symmetrical partners in the other hemisphere. The exceptions are in parts of the visual and somatosensory areas, which do not contribute fibers to the corpus callosum, the main interhemispheric fiber bundle.

The strong internal connectedness of the cortex in each hemisphere is reflected also in the large volume of the hemispheric white substance (the cerebellar hemispheres by comparison have a very scanty white substance, containing only afferent and efferent fibers). In the human brain and in the brains of other large mammals, within the white matter underlying the cortex it is possible to discern about six large bundles connecting distant parts of the cortex, such as the occipito-frontal fascicle or the arcuate fascicle between the motor and the sensory speech areas. The rest of the white matter is composed of shorter fibers between neighboring areas, and to a lesser extent, of commissural fibers, connecting both hemispheres via the corpus callosum, and of the fibers connecting the cortex, both ways, with subcortical centers. The fact that the volume occupied by the long cortico-cortical bundles is a small fraction of the total volume of the white substance suggests that there is a hierarchical organization, with a great amount of preprocessing between neighboring areas, often within one modality, preceding the global integration.

Bibliography:

1. Abeles M 1991 Corticonics: Neural Circuits of the Cerebral Cortex. Cambridge University Press, Cambridge, UK
2. Abeles M, Prut Y 1996 Spatio-temporal firing patterns in the frontal cortex of behaving monkeys. Journal of Physiology 90: 249–50
3. Braitenberg V, Schuz A 1998 Cortex: Statistics and Geometry of Neuronal Connectivity, 2nd edn. Anatomy of the Cortex: Statistics and Geometry 1991 Springer, Berlin, Heidelberg, New York
4. Hebb D O 1949 1961 Organization of Behavior. A Neuropsychological Theory, 2nd edn. Wiley & Sons, New York
5. Hellwig B 1993 How the myelin picture of the human cerebral cortex can be computed from cytoarchitectural data. A bridge between von Economo and Vogt. Journal of Brain Research 34(3): 387–402
6. Horner C H 1993 Plasticity of the dendritic spine. Progress in Neurobiology 41: 281–321
7. Hubel D H 1988 Eye, Brain and Vision. The Scientific American Library, New York
8. Malach R, Schirman T D, Harel M, Tootell R B H, Malonek D 1997 Organization of intrinsic connections in owl monkey area MT. Cerebral Cortex 7: 386–93
9. Peters A, Jones E G (eds.) 1984 Cerebral Cortex, Cellular Components of the Cerebral Cortex. Plenum Press, New York, London, Vol. 1
10. Yoshioka T, Blasdel G G, Levitt J B, Lund J S 1996 Relation between patterns of intrinsic lateral connectivity, ocular dominance, and cytochrome oxidase-reactive regions in macaque monkey striate cortex. Cerebral Cortex 6: 297–310
11. Young M P, Scannell J W, Burns G 1995 The Analysis of Cortical Connectivity. Springer, Berlin, Heidelberg, New York
12. Zeki S 1993 A Vision of the Brain. Blackwell Scientific Publications, London, Edinburgh