Cerebral Cortex

Abstract

The cerebral cortex is the thin, folded sheet of grey matter covering the cerebral hemispheres, the structure identified with human perception, thought, and language. A few millimetres thick yet expanded by folding into the brain's largest structure, it is built from repeating elements: neurons stacked into layers, layers assembled into vertical columns, and columns tiled into cytoarchitectonic areas. Its neocortex has six layers whose connections form a stereotyped canonical microcircuit reused across the sheet, so that the same computation serves vision in one region and language in another. Areas are linked into processing hierarchies and specialised for particular functions, an organisation established during development by the controlled migration of neurons. This article surveys the cortex's laminar and columnar architecture, its areas and functional maps, its development and folding, and the atlases now charting it, with three interactive demonstrations.

Keywords: cerebral cortex, neocortex, cortical layers, cortical column, cytoarchitecture

The cerebral cortex is the outermost sheet of neural tissue of the cerebrum, two to four millimetres thick and containing on the order of sixteen billion neurons in humans. Because its area far exceeds what would fit smoothly inside the skull, it is thrown into folds, the ridges called gyri and the grooves called sulci, so that roughly two-thirds of its surface lies hidden in the depths of the folds. Once described only in gross anatomical terms, the cortex was transformed into a mappable object by the discovery that its cellular architecture varies systematically from place to place, allowing it to be divided into discrete areas that correspond to distinct functions (Zilles & Amunts, 2010). The organising insight of modern cortical neuroscience is that this vast sheet is not a uniform mass but a mosaic of areas built from a common architectural plan, repeated with local variations across the whole surface.

What the Cerebral Cortex Is

The cerebral cortex is the layer of grey matter that forms the surface of each cerebral hemisphere, sitting above the white matter that carries its connections and enclosing the deep nuclei of the forebrain. By far its largest division is the neocortex, the six-layered sheet that makes up about ninety per cent of the human cortex and covers the frontal, parietal, temporal, and occipital lobes. A smaller, phylogenetically older allocortex, with fewer and less regular layers, includes the hippocampus and the olfactory cortex. The cortex is not an isolated processor: it is reciprocally connected with the thalamus, which relays nearly all sensory information to it and receives a massive return projection, and it drives subcortical structures including the basal ganglia and the cerebellum through loops that return to it.

What makes the cortex tractable to study is the regularity beneath its apparent complexity. Everywhere in the neocortex the same six layers appear, distinguished by the size, density, and type of their neurons, and the cells within any small patch are connected to one another in a stereotyped vertical pattern that recurs across the sheet (Douglas & Martin, 2004). This uniformity of local design, combined with systematic variation in the proportions of the layers from region to region, is what lets the cortex be parcellated into areas and what suggests that a single canonical computation, applied to different inputs, underlies functions as different as seeing an edge and planning a sentence.

Figure 1

The Six Layers of the Neocortex and the Canonical Microcircuit

Types of Cerebral Cortex

In the Medical Subject Headings (MeSH) classification the cerebral cortex is filed under the Cerebrum, the largest division of the forebrain, and the descriptor has nine narrower structures, listed below. These are regional and structural subdivisions of one continuous sheet rather than separate organs, and the grouping is partly a matter of indexing convention: MeSH is a vocabulary for cataloguing the biomedical literature, so the list reflects how work is filed rather than a single theoretical claim about how the cortex should be carved up. The subdivisions also mix different principles — the four lobes are gross anatomical territories, the neocortex and limbic lobe are architectonic and functional divisions, and the insular and sensorimotor cortices are named for their location or role — so a given patch of cortex can belong to more than one of these categories at once.

Cortical Layers and Columns

The defining feature of the neocortex is its division into six horizontal layers, numbered I to VI from the pial surface inward, each with a characteristic population of neurons. Layer I, the molecular layer, is nearly cell-free and consists mostly of the dendrites and axons of cells lying deeper. Layers II and III, the external granular and pyramidal layers, contain small and medium pyramidal neurons that are the principal source of connections to other cortical areas. Layer IV, the internal granular layer, is packed with small stellate cells and is the main recipient of input arriving from the thalamus. Layer V, the internal pyramidal layer, holds the largest pyramidal cells, whose axons leave the cortex for subcortical targets such as the brainstem, spinal cord, and striatum. Layer VI, the multiform layer, sends a projection back to the thalamus, closing the loop (Douglas & Martin, 2004).

These layers are not wired at random. Tracing the connections within a small patch reveals a stereotyped circuit: thalamic afferents excite layer IV, which drives the superficial layers II and III, which in turn drive layer V, while layers V and VI provide the output and feedback. This recurring pattern, the canonical microcircuit, appears with only modest variation throughout the neocortex and is the strongest evidence that the cortex applies one basic operation to whatever signals reach it (Douglas & Martin, 2004). Cutting across the layers is a second, vertical organisation. Recording electrodes passed straight down through the cortex encounter cells that share the same response properties — the same receptive-field location, the same preferred stimulus — leading to the proposal that the cortex is organised into vertical columns, each a small group of interconnected neurons spanning all six layers and acting as a basic functional unit (Mountcastle, 1997). The demonstration below lets each layer be selected in turn to show its cells and its place in the circuit.

Cytoarchitecture and Cortical Areas

Although the six-layered plan is universal, the relative thickness and cell composition of the layers vary systematically across the sheet, and this variation is what allows the cortex to be divided into areas. The study of these differences, cytoarchitecture, was placed on a rigorous footing by Korbinian Brodmann, who in 1909 divided the human cortex into fifty-two areas on the basis of their cellular structure alone; his numbered map remains the common language of cortical localisation more than a century later (Zilles & Amunts, 2010). Brodmann's areas are not arbitrary: many coincide with functional boundaries, so that area 17 is the primary visual cortex and area 4 the primary motor cortex, a correspondence that vindicated the idea that structure predicts function.

Cortical areas fall into recognisable architectural types according to how prominent their granular layer IV is. Primary sensory areas are granular or koniocortex, with a thick, dense layer IV suited to receiving heavy thalamic input. The primary motor cortex is at the opposite extreme, agranular, with layer IV so reduced as to be almost absent and a correspondingly enlarged output layer V containing the giant Betz cells. Between these lie the homotypical areas of the association cortex, in which all six layers are clearly present and balanced. Modern work has replaced Brodmann's hand-drawn boundaries with quantitative, observer-independent maps: probabilistic cytoarchitectonic atlases delineate each area statistically across many brains and register it to a common space, capturing the variability between individuals that a single idealised map cannot (Amunts et al., 2020). The demonstration below builds the laminar profile of each architectural type and shows how the balance of layers shifts from sensory to motor cortex.

Functional Organization

Cortical areas do not operate in isolation but are linked into processing hierarchies. In the visual system, an anatomical analysis of the connections between areas revealed a hierarchy of some dozen levels, from the primary visual cortex through progressively higher areas, in which forward connections carry signals up the hierarchy and reciprocal feedback connections carry them down (Felleman & Van Essen, 1991). This layered arrangement is the structural basis for the idea that perception is built up in stages, each area extracting more abstract features from the outputs of those below it. In influential accounts of predictive coding, the feedforward connections are held to carry the residual error between an area's input and the predictions descending through the feedback connections, an interpretation that maps the two directions of flow onto the distinct superficial and deep layers of the canonical microcircuit (Bastos et al., 2012).

At the top of the sensory hierarchies the cortex becomes strikingly specialised. Functional imaging has revealed regions of the association cortex that respond far more strongly to one category of stimulus than to any other — a region of the fusiform gyrus selective for faces, an adjacent region for places — showing that particular patches of cortex are dedicated to particular perceptual and cognitive functions (Kanwisher, 2010). This functional specialisation extends across the cortex: distinct areas support language, the maintenance of information in working memory, the control of attention, and the executive processes of the prefrontal cortex. The same canonical microcircuit, repeated across the sheet, thus gives rise to a mosaic of functionally distinct regions, their specialisation set by the inputs they receive and their place in the network rather than by any difference in their basic design.

Development, Folding, and Evolution

The orderly map of the adult cortex is laid down during development by the controlled production and placement of neurons. Cortical neurons are generated in a proliferative zone lining the ventricle and migrate outward along radial glial fibres to their final positions, with each wave of neurons passing those already in place so that the layers are built from the inside out (Rakic, 2009). The radial-unit hypothesis proposes that the number of these radial columns sets the cortical surface area while the number of neurons within each column sets its thickness, so that the great expansion of the cortex over evolution reflects mainly an increase in the number of columns (Rakic, 1988). The layout of areas is prefigured early, in a molecular protomap of gradients across the proliferative sheet that biases each region toward its eventual identity, which is then refined by the inputs the region later receives (Cadwell et al., 2019).

The same expansion that enlarged the cortical sheet forced it to fold, and the pattern of folding is not incidental. One influential proposal holds that the gyri and sulci arise from mechanical tension along axons in the white matter: strongly interconnected regions are pulled together, drawing the cortex between them outward into a gyrus, while weakly connected regions drift apart into a sulcus, so that the folding pattern reflects the underlying wiring and helps minimise the total length of connections (Van Essen, 1997). Folding is also what makes the cortex so much larger than the skull that contains it, a relationship the worked example below quantifies. The demonstration accompanying it shows how the degree of folding determines what fraction of the cortex lies buried in the walls of the sulci.

Worked Example

The degree of cortical folding is summarised by the gyrification index, defined as the ratio of the total cortical surface area, including the walls buried in the sulci, to the exposed surface area visible on the outside: GI = A_total / A_exposed. A smooth (lissencephalic) brain would have a gyrification index of 1, because nothing is hidden; the human cortex has an index of roughly 2.5. The fraction of the cortex that is buried follows directly: f_buried = (A_total − A_exposed) / A_total = 1 − 1/GI.

Take a human cortex with a total surface area of A_total = 2000 cm² and a gyrification index of GI = 2.5, the values the third demonstration uses. The exposed area is A_exposed = 2000 / 2.5 = 800 cm², so only 800 cm² of the sheet is visible on the outer surface. The buried area is therefore 2000 − 800 = 1200 cm², and the buried fraction is f_buried = 1 − 1/2.5 = 1 − 0.4 = 0.6. Fully sixty per cent of the cortex lies out of sight in the depths of the folds, which is why the exposed surface of a brain so badly understates its true extent, and why unfolding the cortex computationally — as modern surface-based atlases do — reveals so much more area than the visible convexities suggest. If folding increased to GI = 3.0, the buried fraction would rise to 1 − 1/3 = 0.667, and if the brain were smooth at GI = 1.0 it would fall to zero; the demonstration lets the index be varied across this range and reports the exposed and buried areas at each setting.

Discussion

The cerebral cortex illustrates how a single architectural motif, repeated and locally tuned, can give rise to the full range of human mental life. The motif is the six-layered canonical microcircuit, wired the same way everywhere and applying one basic operation to whatever the thalamus and other cortical areas deliver to it (Douglas & Martin, 2004). Stacked vertically, the circuit forms the columns that Mountcastle proposed as the cortex's elementary functional units (Mountcastle, 1997); tiled horizontally with systematic variation in its layers, it forms the cytoarchitectonic areas that Brodmann first mapped and that structure predicts function so reliably (Zilles & Amunts, 2010). Wiring these areas into hierarchies with matched feedforward and feedback connections yields the staged, expectation-guided processing that characterises perception and cognition (Felleman & Van Essen, 1991), and allowing regions to specialise produces the dedicated modules for faces, places, language, and control that functional imaging reveals (Kanwisher, 2010). Development explains how the map is built, through the radial migration and molecular patterning that set each area's size and identity (Rakic, 1988; Cadwell et al., 2019), and the mechanics of folding explain why the finished sheet is packed into the skull the way it is (Van Essen, 1997). Understood this way, the cortex is neither a uniform sheet nor a patchwork of unrelated organs, but one design elaborated into a mosaic.

Current Directions

Contemporary cortical neuroscience is dominated by the drive to chart the sheet completely and at ever finer resolution. Combining several imaging modalities — cortical thickness, myelin content, resting-state function, and task activation — the Human Connectome Project produced a parcellation dividing each hemisphere into 180 areas, nearly half of them not previously described, and did so in a way that could be applied automatically to new individuals (Glasser et al., 2016). In parallel, the return to classical cytoarchitecture with modern tools has yielded three-dimensional probabilistic atlases such as the Julich-Brain, which map each area statistically across many post-mortem brains and register it to a reference space, so that the boundaries carry an explicit measure of their variability between people (Amunts et al., 2020). A third front descends to the level of cell types: large-scale single-cell transcriptomic and physiological surveys have begun to compile a census of the cortex's constituent neurons and glia, cataloguing the cell types that make up a cortical area and asking how their proportions differ across areas and species (BRAIN Initiative Cell Census Network, 2021). Together these efforts are converging on a multi-scale description of the cortex — from cell types through microcircuits to areas and their connections — that the classical maps could only sketch.

Common Misconceptions

The cerebral cortex is where thinking happens, separate from the lower brain.

The cortex is densely and reciprocally connected with the thalamus, basal ganglia, and cerebellum, and depends on those loops to function; cognition is a property of the whole system, not of an isolated sheet (Felleman & Van Essen, 1991).

The folds of the brain exist to increase its surface area.

Increased area is a consequence of folding, but the pattern of folds is thought to be shaped by mechanical tension along axons, drawing strongly connected regions together and thereby also shortening the wiring — not by a simple drive to maximise area (Van Essen, 1997).

Each cortical area is a fundamentally different kind of tissue.

Areas differ in the relative thickness and cell composition of their layers, but they share the same six-layered plan and the same canonical microcircuit; their functional differences arise largely from their inputs and connections, not from a different basic design (Douglas & Martin, 2004).

Glossary

Agranular cortex.

Cortex with a reduced or absent granular layer IV, typical of the primary motor cortex, where the output layer V is correspondingly enlarged.

Allocortex.

The phylogenetically older cortex with fewer than six layers, including the hippocampus and olfactory cortex; contrasted with the six-layered neocortex.

Brodmann area.

One of the numbered cytoarchitectonic regions into which Brodmann divided the cortex in 1909; many coincide with functional boundaries, such as area 17 with the primary visual cortex.

Canonical microcircuit.

The stereotyped pattern of connections among the cortical layers — thalamic input to layer 4, processing in the superficial layers, output from the deep layers — repeated across the neocortex.

Cortical column.

A vertical group of interconnected neurons spanning all six layers and sharing response properties, proposed by Mountcastle as the basic functional unit of the cortex.

Cytoarchitecture.

The study of the cellular structure of tissue; in the cortex, the regional variation in the size, density, and arrangement of neurons across the layers, used to define areas.

Gyrification index.

The ratio of the total cortical surface area to the exposed surface area; a measure of folding, equal to 1 for a smooth brain and about 2.5 for the human cortex.

Gyrus.

A ridge of the folded cortical surface, raised between the grooves; the plural is gyri.

Koniocortex.

Granular cortex with a thick, densely packed layer IV, characteristic of primary sensory areas that receive heavy thalamic input.

Neocortex.

The six-layered cortex that makes up about ninety per cent of the human cortex and carries out its sensory, motor, and associative functions.

Predictive coding.

A theory in which each cortical level predicts the activity of the level below and the feedforward connections carry only the residual prediction error, mapped onto the superficial and deep layers of the canonical microcircuit.

Protomap.

The early molecular pattern of gradients across the proliferative sheet that biases each region toward its eventual areal identity before it receives its inputs.

Pyramidal neuron.

The principal excitatory projection neuron of the cortex, named for its triangular cell body, concentrated in layers III and V and providing most cortical output.

Radial migration.

The outward movement of newly generated neurons along radial glial fibres to their layer, building the cortex from the inside out so that later neurons pass earlier ones.

Radial-unit hypothesis.

Rakic's proposal that the cortex is built from radial columns of neurons produced at the ventricle, with their number setting surface area and the neurons within each setting thickness.

Sulcus.

A groove of the folded cortical surface, sunk between the ridges; the walls of the sulci hold most of the cortex's area. The plural is sulci.

Key Researchers

Katrin Amunts. Director at the Institute of Neuroscience and Medicine at Forschungszentrum Jülich and at the C. and O. Vogt Institute of Brain Research in Düsseldorf; she leads the Julich-Brain probabilistic cytoarchitectonic atlas and helped direct the Human Brain Project. Wikipedia - Faculty

Korbinian Brodmann (1868–1918). German neuropsychiatrist working in the Vogt laboratory in Berlin; his 1909 division of the cortex into cytoarchitectonic areas remains the standard reference for cortical localisation. Wikipedia

David C. Van Essen. Professor of Neuroscience at Washington University in St. Louis; he mapped the hierarchical organisation of the primate visual cortex, proposed the tension-based theory of cortical folding, and led the Human Connectome Project. Wikipedia - Faculty

Nancy Kanwisher. Professor in the Department of Brain and Cognitive Sciences and the McGovern Institute at the Massachusetts Institute of Technology; she discovered category-selective cortical regions such as the fusiform face area, and received the Kavli Prize in Neuroscience in 2024. Wikipedia - Faculty

Vernon B. Mountcastle (1918–2015). Neuroscientist at the Johns Hopkins University School of Medicine; his discovery of the columnar organisation of the cortex established the column as its candidate basic functional unit. Wikipedia

Pasko Rakic. Professor of Neuroscience at the Yale School of Medicine; his radial-unit hypothesis and protomap model explain how the number of cortical neurons and the layout of areas are set during development. Wikipedia - Faculty

Frequently Asked Questions

What is the cerebral cortex?
The cerebral cortex is the thin, folded outer sheet of grey matter covering the cerebral hemispheres, containing on the order of sixteen billion neurons and responsible for perception, voluntary movement, language, and higher cognition (Zilles & Amunts, 2010).

What is the difference between the cortex and the neocortex?
The neocortex is the six-layered part of the cortex that makes up about ninety per cent of the human cortical sheet; the rest is the older allocortex, such as the hippocampus, which has fewer and less regular layers (Douglas & Martin, 2004).

How many layers does the cerebral cortex have?
The neocortex has six layers, numbered I to VI from the surface inward, distinguished by their neuron types; thalamic input arrives mainly at layer IV and output leaves from the deeper layers V and VI (Douglas & Martin, 2004).

What is a cortical column?
A cortical column is a vertical group of interconnected neurons spanning all six layers that share response properties, proposed by Mountcastle as the elementary functional unit of the cortex (Mountcastle, 1997).

What are Brodmann areas?
Brodmann areas are the fifty-two cytoarchitectonic regions into which Brodmann divided the cortex in 1909 on the basis of cell structure; many coincide with functional boundaries and the numbering is still used today (Zilles & Amunts, 2010).

Why is the cerebral cortex folded?
Folding packs a large surface area into the skull, and the specific pattern of folds is thought to arise from mechanical tension along axons, which draws strongly connected regions together and shortens the brain's wiring (Van Essen, 1997).

How does the cerebral cortex develop?
Cortical neurons are generated near the ventricle and migrate outward along radial glia to build the layers from the inside out; the number of radial columns sets the cortex's surface area and an early molecular protomap biases each region toward its areal identity (Rakic, 1988; Cadwell et al., 2019).

Are different parts of the cortex specialised for different functions?
Yes; areas are linked into processing hierarchies and specialised for particular functions, from primary sensory maps to regions of association cortex selective for faces, places, or language (Kanwisher, 2010).

References