Touch and Pain

The puzzle pain sets

Take the standard shorthand seriously for a moment. Pain, on this account, is a signal that travels from injured tissue up a nerve to the brain, and its intensity reports the extent of the injury. It is the account most people carry in their heads, it is the one most clinical language quietly assumes, and it is intuitive because it matches the ordinary case: you cut your finger, it hurts, the worse the cut the worse the hurt.

Now put three well-documented observations beside it. First, people who sustain catastrophic injuries in circumstances of extreme urgency, in battle, in road accidents, in sport, quite frequently report at the time that they felt little or nothing, and the reports are consistent enough that clinicians have a name for the phenomenon. Second, a very large number of people live with severe, disabling, entirely genuine pain in the absence of any injury that scanning or examination can identify. Third, in the other direction, extensive tissue damage can be silently present and painless, as in the neuropathic foot ulcers that people with diabetes can develop and not notice.

Any one of these might be explained away. Together they are fatal to the shorthand. Pain and tissue damage are correlated, but the correlation is loose, and a loose correlation is not what you would expect from a measuring instrument. If pain were a readout of damage, it would track damage. It does not track damage reliably. So the question that organises this entire page is: what does it track?

The answer, worked out over the second half of the twentieth century and now the consensus position in pain science, is that pain is not a message the body sends but a conclusion the brain reaches. The brain takes nociceptive input, that is, incoming signals about tissue threat, and it weighs that input alongside context, expectation, attention, prior experience, and the current demands on the organism, and it produces a protective output. That output is pain. It is real, it is measurable, it is emphatically not imaginary. But it is manufactured centrally, and once you understand that, several otherwise inexplicable facts fall into line at once.

Getting to that claim honestly requires building up from the periphery, and the cleanest place to start is not pain but touch, because touch is the case where the simple story is more or less correct, and seeing exactly where and how pain departs from it is what makes the departure convincing.

Touch: four receptors, four questions

Press a fingertip against a surface and you learn a great deal very quickly: whether it is smooth or rough, whether it is moving under your finger, whether it is vibrating, how hard you are pressing, whether your skin is being stretched. These are different physical quantities, and the skin does not measure them with a single all-purpose sensor. It measures them with four distinct types of encapsulated mechanoreceptor, each of which is a nerve ending wrapped in a different arrangement of tissue, and the wrapping is what determines what the ending can detect.

Look at what has been achieved by this division of labour. Four receptor types, differing on just two axes, receptive field size and adaptation rate, between them cover the space of mechanical questions the hand needs to ask. Small field plus slow adaptation gives you sustained fine detail. Small field plus fast adaptation gives you the onset of local events. Large field plus fast adaptation gives you vibration. Large field plus slow adaptation gives you stretch. No single receptor could do all four, because the physical requirements conflict: a receptor built to hold a signal for sustained pressure is a receptor that cannot follow a 200 hertz vibration. The system solves the conflict by building four receptors and letting the brain read them together.

Two further classes of ending complete the somatosensory periphery. Proprioceptors, principally the muscle spindles in the belly of a muscle and the Golgi tendon organs at its junction with tendon, report muscle length and muscle tension, and from these the brain reconstructs limb position; they are described in detail on the motor system page, where their role in reflexes matters most. And nociceptors, which are not encapsulated at all but simply free nerve endings, respond to stimuli that damage or threaten tissue. That absence of a capsule is not an incidental detail. The elaborate wrappings of the mechanoreceptors exist to make them selective, to filter the world so that only one kind of question gets through. A free nerve ending has no filter, and that is precisely the design you want in a damage detector: something that does not care what kind of insult it is, only that there is one.

Why you stop feeling your clothes

Adaptation rate deserves a section of its own, because it explains one of the most familiar and least examined facts of everyday experience, and because the explanation reveals a design principle that runs through the whole nervous system.

When you put on a shirt, you feel it. Thirty seconds later, you do not. The shirt has not moved and the pressure on your skin has not changed. What has changed is that the rapidly-adapting receptors, the Meissner and Pacinian corpuscles, which fired vigorously at the moment the fabric first touched and moved across the skin, have fallen silent. They have not been switched off by any central command. They simply cannot signal a constant. Their structure ensures that a maintained stimulus produces no maintained response.

The slowly-adapting receptors provide the counterweight. Merkel discs and Ruffini endings do keep firing during a maintained stimulus, which is why you can, if you choose to attend to it, feel the weight of a book resting in your hand for as long as it rests there. So the system is not blind to steady states. It simply demotes them, and the demotion is done at the very first stage, in the skin, before a single signal has been sent to the brain. Filtering at the periphery is cheap. Filtering centrally would mean transmitting and then discarding, which is exactly the waste that a limited nervous system cannot afford.

This same logic, that neural systems encode change and difference rather than absolute level, recurs at every scale of the brain: in retinal ganglion cells that respond to contrast rather than to brightness, in cortical neurons that habituate to repeated stimuli, in the reward system's coding of prediction error rather than reward itself. Touch is the place where you can feel it happening on your own skin.

Two pathways, two crossing points

The signals now have to get from the skin to the cortex, and here the somatosensory system does something that looks, at first sight, wasteful. It does not carry them all together. It sorts them into two entirely separate ascending pathways, which run in different parts of the spinal cord, use different neurons, and cross the midline at different places. Understanding why this matters is the single most valuable piece of clinical reasoning in the whole of sensory neurology, so it is worth tracing both routes carefully.

Both pathways end up in the same thalamic nucleus and the same cortical area. Both are three-neuron chains. Both convey information about the same patch of skin. The only substantial architectural difference between them is where they cross the midline, and if that difference had no consequences it would be a curiosity of embryology and nothing more.

It has an enormous consequence, and the next section is entirely about it.

Brown-Sequard: the anatomy is the diagnosis

Consider a lesion that damages one half of the spinal cord, the left half, say, at the level of the mid-thoracic cord. This can happen from a stab wound, from a tumour compressing one side, from a bullet, or from a plaque of demyelination. It is called a spinal cord hemisection, and the clinical picture it produces is named Brown-Sequard syndrome.

Do not memorise the syndrome. Derive it. You already have everything you need.

1. What happens to fine touch and proprioception from the left leg?

   These signals entered the cord on the left and turned upward on the left, because the dorsal column pathway does not cross until the medulla. At the mid-thoracic level, therefore, they are still on the left. The left-sided lesion cuts them. Result: loss of fine touch, vibration, and proprioception in the left leg, that is, on the same side as the lesion, below its level.

2. What happens to pain and temperature from the left leg?

   These signals entered the cord on the left, synapsed at once, and crossed within a segment or two, so by the time they are ascending through the mid-thoracic cord they are already on the right. The left-sided lesion misses them entirely. Result: pain and temperature from the left leg are preserved.

3. What happens to pain and temperature from the right leg?

   These entered on the right, crossed immediately, and are now ascending on the left, which is exactly where the lesion is. Result: loss of pain and temperature in the right leg, that is, on the side opposite to the lesion, below its level.

4. And the motor tract?

   The corticospinal tract, which carries the movement command, crossed high up in the medulla and is descending on the left. The left lesion cuts it. Result: weakness of the left leg, on the same side as the lesion. See the motor system for why that tract crosses where it does.

This is what neurologists mean when they say that the anatomy is the diagnosis. Nobody has to look inside the cord. The pattern of what is lost and what is spared reports, unambiguously, both the level of the lesion, which is given by the highest segment at which sensation is abnormal, and its side, which is given by the side of the weakness and the proprioceptive loss. The clinical examination is a functional dissection.

The same reasoning, applied at higher levels, produces the crossed findings that localise a brainstem lesion, which are set out on the brainstem and cerebellum page. The principle is identical: pathways cross at known places, so the pattern of the deficit reports the position of the damage relative to those places. This is the central logic of clinical neuroanatomy, and Brown-Sequard is its purest demonstration.

One further prediction falls out of the anatomy, and it is a good test of whether the derivation has been understood. Because the spinothalamic fibres cross one or two segments above the level at which they enter, a lesion at a given spinal level will produce a contralateral pain and temperature loss beginning not at that level but a segment or two below it. That small offset is exactly what is observed. A model that predicts a detail as fiddly as that, and is right, is a model to trust.

The map, and why it is distorted

Both pathways deliver to the same destination: the primary somatosensory cortex, a strip of cortex on the postcentral gyrus, immediately behind the central sulcus, at the very front of the parietal lobe. And they deliver in an orderly way. Neighbouring patches of skin project to neighbouring patches of cortex, so that the body surface is laid out across the cortical strip as a map.

The map is systematically distorted, and the distortion is the interesting part. It was charted directly in the 1930s and 1940s by the neurosurgeon Wilder Penfield, who, operating on conscious epilepsy patients under local anaesthetic, stimulated points along the postcentral gyrus with a small electrode and asked them what they felt. Patients reported tingling, pressure, or a sense of touch in a specific body part, and by moving the electrode systematically Penfield built up a point-by-point map of which cortical territory served which piece of skin. The cartoon of the resulting figure, with its grotesque hands and lips and its shrivelled trunk, is the somatosensory homunculus, published with Rasmussen in 1950.

The distortion is not arbitrary and it is not a distortion of body size. Cortical territory is allocated according to innervation density, that is, according to how many receptors are packed into a given patch of skin. The fingertips and the lips carry an enormous density of mechanoreceptors; the skin of the back carries very few. So the fingertips and lips command a vast area of cortex and the back commands almost none, regardless of how many square centimetres of actual body surface each occupies.

You can verify this on yourself, and the fact that it is checkable is what makes it worth stating. The measurement is called two-point discrimination: the smallest separation at which two simultaneous points of contact are still felt as two points rather than one. Touched with the two points of a bent paperclip, a fingertip can distinguish separations of only a few millimetres. On the skin of the back, the two points must be several centimetres apart before they stop feeling like a single touch. That is a difference of more than an order of magnitude in spatial resolution, measured with a paperclip, and it is a direct psychophysical readout of the innervation density that the cortical map reflects.

Two honest caveats. First, the neat homunculus figure is a simplification of a messier reality: the map is not a single continuous body, it contains discontinuities, and there are in fact several body maps in the somatosensory region rather than one. Second, the map is not fixed. It reorganises with use and with injury, which is a fact that matters enormously for pain, as the section on phantom limbs will show. The cerebral cortex page treats the map and its plasticity in more depth, and neuroplasticity covers the mechanism.

Nociception is not pain

Everything above has been about touch, and about pain only in so far as pain travels a pathway. Now the argument turns, and the turn depends on a distinction that is easy to state and surprisingly hard to internalise, because ordinary language runs the two words together.

Notice what that distinction does to the shorthand. If pain were the arrival at the brain of a pain signal, then the signal would be the pain, and the amount of signal would set the amount of pain. But there is no pain signal. There is a nociceptive signal, and what the brain does with it is a further question, with a further answer, and that answer is where all the variability lives.

What the brain does with it, roughly, is this. Nociceptive input arrives, along with a great deal of other information: what you can see, what you expect, what happened last time, whether you are in danger, whether you are attending to the body part, what the injury means for what you are trying to do. The brain integrates all of it, across a distributed set of regions including the somatosensory cortices, the insula, the anterior cingulate cortex, the thalamus, and the prefrontal cortex, and it arrives at an assessment: how much protective action does this situation demand? Pain is the output of that assessment. It is not a measurement. It is a decision about threat.

That is why there is no single "pain centre" in the brain, despite decades of searching for one. It is also why the soldier's story is not a paradox. In a situation where escaping the battlefield is the overwhelming priority, a brain that produced disabling pain and stopped the body moving would be making a catastrophic error. The brain suppresses the pain because, in that context, pain is not useful. And it has the machinery to do exactly that, as the section on descending modulation will show. The suppression is not stoicism or shock in some vague sense. It is a specific neural mechanism doing exactly what it evolved to do.

And it is why the reverse case, severe pain with no injury, is not a paradox either. If pain is a protective output rather than a damage readout, then a system that has learned, wrongly, that a body part is under threat will produce pain about that body part, and the pain will be exactly as real as any other pain, because there is no other kind. The rest of this page traces the machinery that makes both of those things possible.

Two fibres, two waves of hurt

Before the brain can decide anything, the signal has to get there, and the peripheral half of the nociceptive system has a property you can demonstrate on yourself without equipment.

Nociceptors send their signals into the cord along two classes of fibre, and the difference between them is myelination.

Now, the demonstration. Stub your toe hard and pay attention to the sequence. There is a first pain: sharp, immediate, clearly located there, in that toe, and it fades. And then, a beat later, a second, quite different pain arrives: a deep, spreading, sickening ache that is far harder to localise and that lasts much longer. Everyone has felt this. Almost nobody notices that it is strange.

It is strange because a single injury has produced two temporally separated sensations of two distinct qualities. If one kind of fibre carried pain, that would be impossible. The explanation is simply arithmetic. The two signals left the toe at the same instant, but one travelled at perhaps fifteen metres per second and the other at perhaps one. Over a metre or so of leg, that difference is enough to separate their arrival by a substantial fraction of a second, which is comfortably long enough to be felt as two events.

The physical reason for the speed difference is myelination, and it is worth being precise about it. A myelin sheath forces the action potential to jump from one node of Ranvier to the next rather than propagating continuously along the membrane, which is enormously faster. An unmyelinated axon has no nodes and must regenerate the impulse at every point along its length. The mechanism is set out on the action potential page. Two fibre types, one insulated and one not, produce two conduction speeds, and two conduction speeds produce two pains. The experience follows from the physics.

The gate in the dorsal horn

Nociceptive fibres enter the cord and synapse in the dorsal horn. If the dorsal horn were merely a relay, a place where the first-order neuron hands its signal to the second-order neuron and the signal continues on its way, then nothing interesting could happen there, and the strength of the pain signal reaching the brain would be simply the strength of the signal that arrived from the periphery.

In 1965, Ronald Melzack and Patrick Wall published a paper in Science arguing that this is not what the dorsal horn does. Their proposal, which they called gate control theory, was that the dorsal horn contains inhibitory interneurons that sit on the nociceptive transmission line, and that these interneurons are driven by the large fibres, that is, by the fast, myelinated fibres that carry ordinary touch. Activity in the touch fibres therefore closes the gate: it inhibits the transmission of nociceptive signals from the small fibres at that segment, before those signals have gone anywhere.

The consequence is immediate and it is verifiable on the nearest available bruise.

1. You bang your shin

   Nociceptors in the damaged tissue fire. Small A-delta and C fibres carry the signal into the dorsal horn, where they excite the projection neuron that will send it up the spinothalamic tract. Pain.

2. You rub it, without thinking about it

   Rubbing activates the large mechanoreceptor fibres of the same skin segment: Meissner and Pacinian corpuscles above all, since rubbing is exactly the moving, vibrating stimulus they are built for.

3. The large fibres drive the inhibitory interneuron

   That interneuron inhibits the projection neuron. The nociceptive signal is turned down at its very first synapse in the central nervous system, several stages before the brain has any say in it.

4. Less signal ascends, and it hurts less

   The relief is real, it is spinal, and it is mechanical. Nothing about it is psychological, and nothing about it requires the brain's cooperation.

Melzack and Wall's paper matters because of what it made thinkable. Before it, the dominant model was still essentially the one Descartes had drawn in the seventeenth century: a line from the skin to the brain, and pain proportional to what travels along it. Gate control put a modulator on the line. And once there is a modulator on the line, the question of what controls the modulator becomes unavoidable, which is where the next section starts.

The brain turns its own pain down

The gate can be closed from below, by touch fibres. It can also be closed from above, and this is the fact that completes the argument of the page.

Running down from the brain to the dorsal horn is a descending pathway whose function is to suppress nociceptive transmission. Its origin is the periaqueductal grey, a region of the midbrain wrapped around the cerebral aqueduct, which receives input from the hypothalamus, the amygdala, and the prefrontal cortex, that is, from the structures concerned with threat, emotion, and context. The periaqueductal grey projects to the raphe nuclei and to nearby cell groups in the rostral medulla, and those in turn send fibres down the cord to the dorsal horn, where they inhibit the very same projection neurons that gate control described.

The transmitters involved include serotonin and noradrenaline, and, decisively, the body's own opioids: the endorphins, enkephalins, and dynorphins, peptides that bind to the same receptors as morphine. The system is, quite literally, an internal morphine supply with a wire to the spinal cord.

Now count what this single circuit explains.

The existence of a dedicated, opioid-driven, top-down pain-suppression pathway is not a minor detail of pain physiology. It is a structural fact that refutes the transmission model outright. A system whose job was to report damage faithfully would have no reason to build itself a volume control operated from the top, and no reason to wire that control to the amygdala. A system whose job is to produce a protective output appropriate to the situation needs exactly such a control, and it needs it wired to exactly the structures that know what the situation is. The anatomy tells you which kind of system pain is.

Note also that the same descending pathway can, under some conditions, facilitate rather than inhibit nociceptive transmission. The volume control turns both ways. That fact will matter in a moment.

Phantom limbs: the decisive case

After an amputation, the overwhelming majority of people continue to feel the limb that is gone. The phantom has a position, a posture, sometimes an itch, and very often it hurts, sometimes severely and for years. The phantom hand may be felt clenched into a fist so tight that the nails dig into the palm, and the person feels precisely that: nails digging into a palm that does not exist.

Put that beside the transmission model. Pain, on that model, is a signal from damaged tissue. There is no tissue. There is nothing there to send a signal, and no nerve endings, and no injury. The model does not predict phantom limb pain to be rare. It predicts it to be impossible. It is not rare; it is the norm after amputation. This single observation is fatal, and there is no repair available.

On the account this page has been building, it is not merely possible but expected. Pain is produced by the brain, and the brain's representation of the limb does not vanish when the limb does. The cortical territory that served the hand is still there, still connected into the network that produces the experience of a hand. What has changed is the input. And when a piece of cortex loses its input, it does not simply fall silent: it is invaded by its neighbours.

Phantom pain, on the leading account, arises from the mismatch and disorganisation this creates. The motor system sends commands to the limb, the brain expects sensory confirmation, and no confirmation arrives, ever. In a phantom clenched into a fist, the brain issues an unclench command and receives no signal that the hand has opened, and the intractable cramp is what an unresolvable motor-sensory conflict feels like.

The most striking evidence that this is broadly right is therapeutic, and it comes from work by V. S. Ramachandran and colleagues in the 1990s. The mirror box is a box with a vertical mirror in it. The intact hand is placed on one side, the stump on the other, and the patient looks into the mirror so that the reflection of the intact hand appears exactly where the phantom is felt to be. They then move the intact hand, and see a hand moving in the phantom's place. For a proportion of patients, the phantom, at last, moves. And for a proportion of those, a phantom that had been locked in a painful cramp for years unclenches, and the pain eases.

Two honest qualifications, because this page does not oversell. The mirror box does not work for everyone, the size of the effect varies considerably across trials, and the mechanism by which it works is not settled: whether it acts by restoring congruent visual feedback, by reversing cortical reorganisation, by redirecting attention, or by some combination, remains genuinely debated. And the causal link between the extent of cortical remapping and the severity of phantom pain, once thought straightforward, has been complicated by later findings, with some evidence suggesting that persistence of the original limb representation, rather than its erasure, tracks the pain. What is not in doubt is the phenomenon: pain in a limb that is not there, altered by a mirror.

When the alarm itself breaks

Acute pain is useful. It stops you walking on a broken ankle. Chronic pain, defined as pain persisting beyond the normal time for tissue to heal, conventionally three months or more, very often is not useful, and the reason is that in chronic pain the problem has frequently migrated from the tissue into the pain system itself.

Two mechanisms carry most of the weight, and both follow from what has already been established.

Peripheral sensitisation. Damaged and inflamed tissue releases a chemical soup, including prostaglandins, bradykinin, and substance P, which lowers the firing threshold of the nociceptors in that tissue. They become easier to fire. This is why sunburnt skin hurts under a warm shower that would ordinarily feel pleasant: the receptors have been turned up, and a stimulus that was previously below threshold is now above it. This is adaptive, in the short term, because it makes you protect the injured part.

Central sensitisation. The more consequential mechanism. Sustained nociceptive input into the dorsal horn strengthens the synapses there, by mechanisms that overlap substantially with those of long-term potentiation, the very same synaptic strengthening that underlies memory. The dorsal horn neuron becomes more responsive, its receptive field can expand, and the descending pathway, which can facilitate as well as inhibit, may shift towards facilitation. The result is a pain system with its gain turned up, in which normal input produces amplified output.

The clinical signatures of this are named, and they are exactly what the model predicts. Hyperalgesia: a painful stimulus hurts more than it should. Allodynia: a stimulus that is not painful at all, the weight of a bedsheet, the brush of a shirt, is experienced as painful, because a light-touch input is now driving a sensitised nociceptive circuit. And in some conditions, pain that spreads well beyond the original site, because the change is in the central circuit rather than in the tissue.

Common misconceptions

Related topics