In May of 1863, the Surgeon General of the Union army, William Hammond, ordered the conversion of a four-hundred-bed wooden hospital on Turner's Lane in north Philadelphia into the first hospital in the United States devoted exclusively to diseases of the nervous system. He put a thirty-four-year-old physician in charge. The physician's name was Silas Weir Mitchell. Over the next two years Mitchell and his colleagues George Morehouse and William Keen would treat several thousand soldiers with nerve injuries from gunshot wounds, mortar fragments, severed limbs and the slow attritional infections of the field hospital. Approximately sixty thousand amputations were performed during the four years of the American Civil War, the largest concentrated source of nerve-injury patients in the country's history to that point. A substantial fraction of those amputees ended up at Turner's Lane.

What Mitchell saw was that many of his patients reported pain. They reported pain in limbs that no longer existed. A man whose right foot had been amputated three months before would describe the soles of that foot as burning, the toes as cramped, the heel as resting on a sharp pebble. A man whose arm had been removed at the shoulder would feel the ghost of his fingers, sometimes spread, sometimes balled into a fist that he could not unclench. The pain was not occasional. For some it was constant. For some it was unbearable. Mitchell wrote that the most reliable thing about it was that nothing the medicine of 1864 could offer would touch it.

He decided to take the patients seriously. In 1864 he and his two colleagues published a hundred-and-sixty-page book, Gunshot Wounds and Other Injuries of Nerves, describing the cases. In 1872 Mitchell published a longer book under his own name, Injuries of Nerves and Their Consequences, which became the founding text of American neurology. He coined two terms in those books that are still in clinical use today. Causalgia, from the Greek for burning pain, for the chronic, often unbearable nerve pain after a partial nerve injury. And phantom limb for the sensation of a limb that had been amputated.

The implication of phantom limb pain, which Mitchell understood and which became the most important inheritance he gave to American neurology, was that pain could not be a simple signal carried up a wire from the body. There was no wire. The pain was being generated somewhere above the cut. It was being generated inside the central nervous system. The body part that hurt did not need to exist for the hurt to be real.

Mitchell knew this in 1864. He could not explain it.

For the next century the question of how the central nervous system decided to maintain a pain signal stayed approximately where Mitchell had left it. In 1965 a Canadian psychologist named Ronald Melzack and an English physiologist named Patrick Wall published a paper in Science arguing that pain was modulated, not reflexive, that the spinal cord contained a "gate" that could amplify or quiet incoming signals. The mechanism of the gate was a guess. In 1973, three laboratories in three countries independently discovered that the brain has its own receptors for opium-derived molecules, implying it produced its own opioids. The endorphins were named in 1975. The clinical results of all this were inconclusive enough that for the next forty years American medicine treated chronic pain mostly with stronger and stronger doses of synthetic opioids, with the public-health consequences that have since followed.

Mitchell's question was never quite answered. Why does the pain stay on, in a leg that is not there?

*

In April of this year, the Journal of Neuroscience published a paper from a team at the University of Colorado Boulder titled "Caudal Granular Insular Cortex to Somatosensory Cortex I: A Critical Pathway for the Transition of Acute to Chronic Pain." It identifies, in the rat brain, a small, previously obscure region of deep cerebral cortex called the caudal granular insular cortex, the CGIC, that is both necessary and sufficient for the maintenance of chronic pain after an injury has healed. The CGIC sends instructions to the somatosensory cortex, which sends them down to the spinal cord, which keeps the pain signal running. When the team silenced the CGIC, chronic pain that had already been established in the animals melted away. Acute pain, the kind that protects an animal from immediate harm, was unaffected. The body could still detect a hot stove. The brain had stopped insisting on remembering it.

The paper does not yet describe the human equivalent of the CGIC. The clinical work to translate the rodent finding into a non-opioid pain therapy has not begun. The same has been said about every other promising chronic-pain finding of the last sixty years, and they have all been harder than they looked. But the structural lesson Mitchell took from his Civil War amputees was correct, and the new paper is the first detailed picture of the structure he inferred. There is a circuit in the brain that decides to keep the pain on. The leg does not have to be there. The arm does not have to be there. The brain can run the program by itself.

Mitchell's most famous patient was a fictional one. In the summer of 1866 he published a short story in the Atlantic Monthly called "The Case of George Dedlow." Dedlow is a young Union surgeon who is wounded at Chickamauga and ends up at a hospital in Philadelphia having lost both arms and both legs. He sits, in the final scene, in his wheelchair at a séance, and feels his legs walking back into the room. The story was published anonymously, and the Atlantic's readers wrote in by the thousand asking for Dedlow's address so they could send him donations and prayer letters. Mitchell did not correct them. He did not say that Dedlow was a composite of his amputee patients, or that the legs walking back into the room were the medical mystery of the war.

A hundred and sixty years later, in a laboratory in Boulder, the legs are still walking. We have just figured out which part of the brain is moving them.