Why does a missing limb hurt?

In a survey of 914 people with limb loss, 79.9 per cent reported phantom pain in the four weeks before they were asked. Pain in the residual limb itself was reported separately, by 67.7 per cent. A 2020 meta-analysis of studies from many countries put phantom pain lower, at 64 per cent, and lower again in developing countries.

It happens to people who never had the limb. Among 125 people with missing limbs, phantoms were reported by at least a fifth of those born limb-deficient, and a fifth of those phantoms hurt. A 2023 study of 99 adults born without a limb found phantom pain uncommon among them.

The older assumption was that the trouble lay in the stump. The evidence gathered since points mainly to the central nervous system, and a limb that was never there is a case the stump account cannot reach.

The periphery has not dropped out. In 2014 a study of 31 amputees found that a local anaesthetic delivered to the spinal nerve roots extinguished phantom pain, quickly and reversibly, and its authors argue that the pain starts there.

The cortex holds a map of the body, and on it the face sits beside the hand. In monkeys, when input from an arm is cut, the hand territory starts to answer to touch on the face. In human arm amputees, studies at a single point in time found the lip representation shifted towards the former hand area.

The size of that shift correlates with the amount of phantom pain reported, and in the study that established it, on thirteen arm amputees measured neuromagnetically, the correlation was 0.93. It has been replicated, and it is the central evidence for the standard account.

The shift itself is now disputed. A 2025 study scanned three people before and up to five years after arm amputation and found their hand and lip maps unchanged. Its authors argue that single-time comparisons, and analyses that ignored the phantom hand’s own activity, had produced the appearance of reorganisation.

A face touched in a particular place can be felt on a phantom hand. In one patient, touching the lower face produced sensations localised to individual fingers. Ramachandran and colleagues read this as the face taking over the hand’s territory; in the 1995 study, whether a patient felt it was unrelated to how far the map had shifted.

The correlation is robust and the direction is not settled. Pain can itself alter the map, which would make the reorganisation a consequence.

A 2013 study argues close to the opposite. Imaging 18 amputees with and without phantom pain, it found the representation of the missing hand preserved rather than overwritten, and the more phantom pain people had, the more of its structure and activity their former hand area had kept.

In 2020 the authors of the two accounts reviewed the field together. On the studies where a treatment changes both the brain and the pain, their verdict is that the two move together and that nothing in those studies says which way the causation runs.

Placing a mirror so that the intact limb appears where the missing one was gives the brain the sight of a limb in the empty place, which is meant to restore its representation and so quiet the pain. It is the treatment the standard account predicts, and patients do report relief.

The trials that compare it against a placebo do not establish it. A 2023 review found five of them, and one reported a significant difference from the control group. The reviewers put that down to small samples and low methodological quality rather than to a settled negative, and they note that looser meta-analyses reach the opposite conclusion.

So the treatment is unproven rather than refuted, and it cannot at present be used as an argument for the account that inspired it.