The Cerebral Cortex

What the cortex is

The cerebral cortex is the outermost layer of the cerebrum, a continuous sheet of grey matter draped over each hemisphere. In cross-section it is strikingly thin, between about 2 and 4 millimetres, yet it is densely packed with neurons: tens of billions of cell bodies arranged in orderly layers. Because it forms the surface of the brain, the cortex is what gives the organ its familiar wrinkled appearance.

The word cortex comes from the Latin for bark, and the image is apt: like bark on a tree, it is a thin rind over a larger core. That core is white matter, the mass of fibres connecting one part of the cortex to another and to deeper structures. The cortex is not uniform. Most of it is neocortex, the six-layered tissue that dominates the human brain, but older regions with fewer layers, such as parts of the limbic border, follow a different plan.

Why the cortex is folded

The most obvious feature of the human cortex is that it is not smooth. Its surface is thrown into a pattern of ridges and grooves, and this folding is not decorative: it is a solution to a packing problem. A large, thin sheet cannot fit inside a compact skull unless it is crumpled, much as a large piece of paper must be folded to fit a small box.

The ridges are called gyri and the grooves between them sulci (a particularly deep sulcus is called a fissure). By folding the sheet this way, the brain fits far more cortex into the skull than a smooth surface would allow, roughly tripling the usable area. In humans, about two-thirds of the cortical surface is hidden from view, tucked inside the folds. Larger and deeper folds are a hallmark of brains with more cortex to accommodate.

Grey matter and white matter

The cortex is grey matter, and understanding the distinction between grey and white matter is central to understanding how it works. Grey matter is tissue dominated by neuron cell bodies, along with their dendrites and the synapses between them. This is where signals are received, combined, and processed: grey matter is where the computing happens.

Directly beneath the cortex lies white matter. This is made of axons, the long output fibres of neurons, most of them wrapped in a pale myelin sheath that gives the tissue its colour. White matter is the wiring: it carries signals from one patch of cortex to another, between the hemispheres through the corpus callosum, and down to deeper structures and the spinal cord. The cortex, then, is a thin processing sheet sitting on top of a much larger volume of connecting cable.

The six layers, and what they are for

Look closely at the neocortex and it resolves into six horizontal layers, or laminae, running parallel to the surface. Textbooks usually list them, from the surface inward, with a note on the cell types in each, and the list is duly memorised and duly forgotten. It is forgotten because a list is not an explanation. The layers only make sense once you ask a different question: not what is in each layer, but where each layer sends its axons.

Ask that question and a single organising principle appears. The cortex is a stack in which input arrives in the middle, is passed up to the superficial layers to be shared with other cortical areas, and is sent down from the deep layers to targets beneath the cortex. Everything else is detail hung on that frame.

1. Layer IV: in, from the thalamus

   The inner granular layer is the cortex's front door. Fibres arriving from the thalamus terminate here, densely and precisely. This is where information from the outside world enters the sheet, and it enters in the middle of it, not at the top.

2. Layers II and III: sideways, to other cortex

   Layer IV passes its signal upward to the superficial pyramidal cells of layers II and III, whose axons run out through the white matter to other cortical areas, in the same hemisphere and across the corpus callosum to the other. These are the layers with which the cortex talks to itself.

3. Layer V: down and out

   The large pyramidal cells of layer V send their axons out of the cortex entirely, to the striatum, the brainstem, and, from motor cortex, all the way down the spinal cord. Layer V is how a thought becomes an action. When the cortex commands anything, it commands it from here.

4. Layer VI: back to the thalamus

   The deepest layer sends a massive projection back down to the same thalamic nucleus that fed layer IV. This is the corticothalamic feedback that makes the thalamus a gate rather than a relay: corticothalamic fibres outnumber the thalamocortical fibres coming the other way. The loop is closed inside the cortex's own layering.

Layer I, the molecular layer at the very surface, holds almost no cell bodies. It is a field of dendrites and horizontally running fibres, the place where the apical tufts of pyramidal cells from the layers below reach up to be met by diffuse, modulatory input. It is less a layer of cells than a layer of contact.

Now the regional variation, which the older accounts present as a curiosity, becomes a prediction. If layer IV is the thalamic input layer, then an area whose business is receiving a torrent of thalamic input must have a fat layer IV. And so it does: in the primary visual cortex layer IV is so thick and so distinctly striped that the region can be identified with the naked eye in a stained section. If layer V is the output layer, then an area whose business is issuing commands must have a fat layer V and has little use for layer IV. And so it is: the primary motor cortex has a layer V crowded with the giant pyramidal cells that project to the spinal cord, and a layer IV so thin that the region is described as agranular.

The layer thicknesses are not arbitrary regional decoration. They are a readout of what a piece of cortex is wired to do. A patch that mostly listens is built one way; a patch that mostly speaks is built another. This is the whole reason the next two sections work.

The cortical column

The layers describe the cortex horizontally. There is a second organisation running at right angles to them, vertically, through the depth of the sheet, and it is often asserted rather than explained: neurons in a narrow vertical cylinder of cortex tend to respond to the same feature and to act as a unit. Why should they?

The answer is in the wiring, and it follows from the stack described above. A given bundle of thalamic fibres terminates in a small patch of layer IV. From there the signal is passed vertically, up into layers II and III and down into layers V and VI, within roughly the same column of tissue. Connections running vertically within a column are dense; connections running sideways to neighbouring columns are comparatively sparse. So the cells stacked above and below one another are all listening to the same input, and they are wired far more tightly to each other than to their neighbours a fraction of a millimetre away.

A group of cells that shares an input and is densely interconnected will behave as a unit. It could hardly do otherwise. The column is therefore not a module that anatomy decreed; it is what dense vertical wiring produces. This matters because it tells you what the cortex is: an array of small, near-identical processors, each receiving a slice of input and passing its verdict up, out, and down, tiled across the folded sheet. That is why the same six-layered tissue can do vision at the back of the head and planning at the front. The circuit is the same. Only its inputs and its targets differ.

Brodmann: structure predicts function

In 1909 the German anatomist Korbinian Brodmann published a map that divided the human cortex into numbered areas, running from 1 to 52 but describing only 43 of them in humans. The gaps are telling rather than sloppy: numbers 12 to 16 and 48 to 51 are areas he could make out in other mammals and could not find in us, and he left the numbers empty rather than force the human brain to fill them. His only evidence was cytoarchitecture: the size, density, and layering of the cells as they appeared in a stained slice under a microscope. He was not measuring what the areas did. In most cases nobody knew what they did. He was reading the tissue and drawing a boundary wherever the layering changed.

The map should not have worked. It did. Brodmann's area 17, defined by the striking stripe in its enormous layer IV, turned out to be the primary visual cortex, the destination of the fibres from the lateral geniculate nucleus. His area 4, defined by giant cells in layer V and almost no layer IV, turned out to be the primary motor cortex, the origin of the fibres that reach the spinal cord. Areas 41 and 42 turned out to be auditory cortex, areas 3, 1 and 2 the somatosensory strip. A century of stimulation studies, lesion studies, and functional imaging has refined the boundaries but has not overturned the scheme, and neuroscientists still cite Brodmann numbers as a matter of routine.

Take that seriously and it is not a piece of trivia about a dead anatomist. It is the strongest evidence there is for the six-layer logic set out above. The microscopic structure of a patch of cortex predicts its function because the layer thicknesses record its connections, and a region's connections are what it is for. Brodmann was, without knowing it, reading the wiring diagram off the cell counts.

Sensory, motor, and association areas

Although the cortex is built from the same basic layered tissue everywhere, different regions specialise in different jobs. Broadly, cortical areas fall into three functional classes, and this division is a useful spine for understanding what the cortex does.

The balance of these areas is telling. In the human brain, the sensory and motor areas together occupy only a modest fraction of the cortex; the great majority is association cortex. It is this expansion of integrative tissue, more than any single specialised area, that most distinguishes the human cortex and underlies its capacity for abstract thought and language.

Association areas are not a single uniform region but several. The prefrontal association cortex at the front supports planning, judgement, and self-control; the parietal-temporal-occipital association cortex, where those three lobes meet, binds together vision, hearing, and touch into a unified perception of the world; and the association cortex of the temporal lobe helps with recognising objects and faces and with attaching meaning to what is perceived. Between them these regions turn raw sensation and simple movement into the rich, flexible behaviour of a thinking animal.

Cortical maps and the homunculus

One of the most elegant findings in neuroscience is that the sensory and motor areas are organised as orderly maps of the body. In the primary somatosensory cortex, which handles touch, and the primary motor cortex, which drives movement, the body is laid out in a continuous strip: adjacent parts of the body are represented by adjacent patches of cortex, so the map preserves the layout of the body it serves.

We know the layout of these maps in the human brain largely because of Wilder Penfield, the neurosurgeon who, while operating on conscious epilepsy patients under local anaesthetic in the 1930s and after, stimulated points on the exposed cortex with a small electrode and asked the patient what they felt, or watched which muscle twitched. Stimulate here and the patient reports a tingling in the thumb; a centimetre along and it is the lip. Doing this systematically, across many patients, produced the map, and the figure drawn from it.

These maps are not drawn to scale. The amount of cortex devoted to a body part reflects not its size but its importance for fine sensation or control. The hands, lips, and face take up disproportionately large areas, while the trunk and legs take up relatively little. If the body were redrawn in the proportions the cortex assigns it, the result would be a distorted figure with huge hands and lips and a small torso, an image known as the cortical homunculus, meaning "little man".

You can verify the principle on yourself, without any equipment. Have someone touch you with two pencil points, held a centimetre apart, on a fingertip: you feel two distinct points. Do the same on your back and you feel one. The distance at which two points can still be told apart, the two-point discrimination threshold, is a few millimetres on a fingertip and several centimetres on the trunk. That difference begins in the skin, which is packed with touch receptors on the fingertip and sparsely supplied on the back, and it is carried faithfully upward: a densely innervated patch of skin sends more fibres, and more fibres claim more cortex. The homunculus is what the receptor density looks like once it has been drawn on the brain.

Maps are not fixed

The claim just made, that cortical territory is allocated by required precision rather than by body size, sounds tidy but it needs testing. If the allocation really tracks how much a part is used and how finely it must be resolved, then the map should be able to change when the use changes. It does, and the evidence is some of the most direct in the whole field.

Train a finger intensively, as a string player trains the fingering hand across years of practice, and the cortical representation of those fingers expands. Take a sensory input away, and the territory does not lie idle: it is annexed. In people who have lost a hand, the patch of somatosensory cortex that used to receive input from that hand comes, over months, to respond to touch on neighbouring parts of the map instead, and the face is a common neighbour. This has a striking experiential consequence. Stroke the cheek of such a person and they may feel the touch not only on the cheek but in the missing fingers, because the cortex reporting the sensation is the cortex that used to be the hand. The phantom limb, felt so vividly that it can seem to be clenched or aching, is not an illusion produced by a stump. It is a map that has been redrawn while the labels stayed where they were.

This is why the homunculus should be read as a snapshot rather than a blueprint. The cortex is the same six-layered tissue everywhere; what a given patch does is set by what is wired into it, and wiring, within limits, moves. For the mechanisms by which it moves, see neuroplasticity.

Hemispheres and lateralisation

The cortex is split into two hemispheres, left and right, joined by the corpus callosum. In general each hemisphere handles sensation and movement for the opposite side of the body. Beyond this crossing, some functions are handled more by one hemisphere than the other, a property called lateralisation. In most people, for example, the main language areas lie in the left hemisphere, while aspects of spatial attention and the processing of faces lean toward the right.

Lateralisation is genuine, but it is routinely misunderstood. The popular notion that individuals are logical left-brained or creative right-brained personalities has no scientific basis. Lateralisation describes where certain processing tends to be concentrated across the population; it is not a claim that a person uses one hemisphere more than the other, and it does not divide personality between the sides.

Understood correctly, the two hemispheres are best thought of as a single cooperative system with a modest division of labour, not as two rival minds. The steady traffic across the corpus callosum keeps them working as one.

Related topics