Chapter 14

Phantom Limbs and the “Illusory Body” Argument

In 1872, an American doctor named Silas Weir Mitchell coined a strange new phrase. He had been treating Civil War veterans — men whose arms and legs had been blown off at Gettysburg, at Antietam, at the Wilderness. Many of them complained that they could still feel the missing limb. A foot that wasn’t there itched. A hand long buried in some surgeon’s pail still seemed to clench into a fist. Mitchell called these strange sensations phantom limbs.¹

A century and a half later, the science of phantom limbs has filled volumes. We now know that the brain holds an inner map of the body in the parietal cortex — an internal picture of arms and legs, fingers and toes, that doesn’t simply dissolve when a limb is lost.² Even children born without arms can sometimes feel phantom fingers. One eleven-year-old girl, born without forearms or hands, used her phantom fingers to count out arithmetic problems.³ The brain, it turns out, can build a body part out of nothing.

Michael Marsh sees a parallel. If a brain can manufacture a phantom finger, why not a phantom body? Why not the whole floating self that NDE experiencers describe hovering above the operating table? In Marsh’s reading, the out-of-body experience is just one more illusion in a long list of body-image deceptions the brain is known to perform. Phantom limb, phantom torso, phantom self — same machinery, different scale.⁴

It’s a clever argument. It’s also — I want to argue in this chapter — the wrong argument. Phantom limbs prove something fascinating about the brain. They do not prove what Marsh needs them to prove. The leap from a felt-but-imaginary finger to a perceiving-but-imaginary self is a much bigger jump than it looks.

A. Marsh’s Phantom-Body Analogy

Marsh lays out the case carefully. Chapter 5 of Out-of-Body and Near-Death Experiences is essentially a tour through the neurology of body-image. He walks the reader through phantom pain in amputees, through the strange way paraplegics (people paralyzed below a spinal injury) still feel a body that ought to be invisible to them, through the eerie experiments where a vibrating tendon makes someone’s nose feel a foot long. The point of all this, Marsh tells us up front in the preface, is to soften us up for a single conclusion: if the cortex can build an arm out of nothing, or a torso for a man whose neck is broken, then the cortex can build an entire phantom body and place it floating above the bed.⁵

The argument is meant to dissolve the OBE the way a chemist dissolves salt. Once you see the brain doing something like an OBE in another setting — building body parts out of pure cortical signal — the OBE looks less like a window into another reality and more like one more well-known cortical trick.

Marsh is at his strongest when he describes the ordinary illusions of the healthy brain. A normal subject in a lab can be made to feel his nose stretch outward by 30 centimeters when an experimenter vibrates his bicep tendon while he touches his face. Astronauts in zero gravity can feel upside-down with no way to tell which way is up. A clever mirror trick can make a phantom hand uncramp.⁶ All of this happens in brains that are awake, alert, and not in any way oxygen-starved or dying. Marsh stresses this point, and he’s right to. Body-image is constantly being constructed by the brain. We feel where our limbs are because the parietal cortex tells us so. When the parietal cortex is fed bad information, we feel where our limbs aren’t.⁷

From here, Marsh widens the lens. He notes that some people with brain injuries or epilepsy report a doppelgänger — an autoscopic or heautoscopic experience — in which they seem to see their own body from outside themselves. He notes that disturbances of the temporo-parietal junction can produce brief sensations of floating outside the body, a topic we covered in Chapter 13. He notes that the brain’s inner body-map is so flexible that under the right perturbation it can place the felt self in a different location entirely. Phantom limbs are not exotic. They are the tip of an iceberg.

And so — here is the key move — if all of that is true, why should the OBE component of an NDE be any different? Marsh’s conclusion at the end of chapter 5 is sweeping. The brain, he reasons, has already shown it can rebuild a missing arm and even a torso for a man whose spine has been severed; it takes only a small further leap of imagination, he suggests, to conceive of that same brain producing an entire person at a distant location, with consciousness traveling along.⁸

That sentence is doing a lot of work. Notice how the action verbs travel. Rebuild a missing limb. Rebuild a torso. Rebuild the entire person, complete with consciousness, somewhere across the room. Each step is presented as a small additional act of imagination beyond the one before. Whether the steps actually shrink that small — whether the imagination really is little — is the question this chapter is built to answer.

B. Where the Analogy Strains

Marsh is a careful neurologist. He is not a careful logician at this particular point. Three problems sit just under the surface of his phantom-body argument, and once we see them, the analogy starts to look much less impressive than it first appears.

The first problem is the most basic. A phantom limb is a felt sensation of presence — a tingling, a clenching, a pain. It is not a perception of an outside event. The amputee with a phantom hand cannot use that phantom hand to read the next morning’s newspaper. He cannot send it through the wall to count the people in the next room. The phantom is a private feeling, locked inside the patient’s nervous system. It tells him about the presence of an arm. It does not tell him about the world.⁹

NDE out-of-body perceptions are the opposite of this. The whole reason researchers find them interesting is that they reach outward. Patients describe specific tools on a specific cart on a specific date, and someone walks over and finds the tools right where they said. That is not body-image at all. That is world-image. The brain’s ability to manufacture phantom parts of the self says nothing about its ability to manufacture phantom parts of the world. Those are two different jobs in two different cortical systems.

The second problem is the size of the leap. Marsh asks us to grant the cortex an enormous new power on the strength of a much smaller demonstration. We have good evidence that the brain holds an inner schema of where the limbs are. We do not have evidence that the brain can spontaneously construct an entire vantage point twelve feet above the operating table, complete with novel visual content the patient has never seen, accurate enough to be matched against later inspection of the room. Phantom limbs are old, well-worn cortical engrams reasserting themselves after their input is cut. NDE perceptions are new — specific, unanticipated content that the patient never previously encoded. The two are not on a sliding scale. They are different categories of cortical work, and the second has no demonstrated mechanism in the literature.¹⁰

The third problem is one Marsh almost stumbles into himself. Early in chapter 5 he reminds us that phantom limbs occur in brains that are “hardly considered to be hypoxic, and certainly not in any sense agonal.”¹¹ Translation: phantom limbs appear in healthy, well-oxygenated brains, not in dying ones. But the OBEs Marsh wants to explain are reported during cardiac arrest, when the brain is anything but healthy. So if we follow the analogy strictly, the phantom-limb mechanism predicts that NDEs should not occur in patients whose brains have just gone into shutdown — which is exactly the population in which they are most often reported. The analogy works against the conclusion Marsh wants to draw from it.

C. Phantom Limbs Cannot See

Now to the heart of the matter. The phantom-limb analogy is meant to dissolve the strangeness of the OBE. The way to test whether it succeeds is simple: take an NDE that contains real, externally verified information about the world — the kind of report no inner body-map could possibly generate — and ask whether the phantom-limb story can absorb it.

Consider a man rushed into a hospital in the Netherlands in 1979. He had been found in a coma, blue, no pulse, no breathing. The paramedics had begun resuscitation in the ambulance and continued in the emergency room. A male nurse, identified in the literature only as TG, took over. To clear the patient’s airway, TG removed the man’s upper denture and laid it on a wooden pull-out shelf attached to a metal cart that held the resuscitation drugs. The cart had been built in-house by the hospital’s own maintenance shop. There was nothing standard about it. The patient remained pulseless for over an hour. He was eventually stabilized, transferred to intensive care, and kept in a medically induced coma for several days.¹²

About a week later, the patient saw TG walk onto the cardiac ward and pointed at him in surprise. “You! You know where my dentures are.” TG didn’t. The patient described, with notable accuracy, the cart with bottles on top, the sliding wooden shelf underneath, and the resuscitation team bent over his unconscious body. He said he had watched the whole thing from above. He said he had been afraid the team would give up too soon, and described their reluctant conversation about whether to continue. The case was reported by Pim van Lommel in The Lancet in 2001 and was later corroborated in detail by independent researchers who interviewed the medical staff.¹³

Now turn this case against the phantom-limb hypothesis. What would Marsh’s mechanism need to do to explain it? The cortex would have to (a) construct a phantom self at the ceiling, (b) furnish that phantom self with a vantage point from which a custom-built metal cart with a particular wooden shelf was visible, (c) generate the specific information that the patient’s dentures had been placed on that shelf, (d) reproduce the medical team’s actual conversation about whether to abandon resuscitation, and (e) do all of this in a brain that, at the relevant moment, had no measurable cardiac output and no measurable cortical activity. Phantom limbs do not even attempt the first of these tasks, much less the rest of them.

Or take the case of Maria, a Spanish-speaking migrant worker admitted to Harborview Medical Center in Seattle in 1977 after a heart attack.¹⁴ Three days into her stay she had a second arrest. After resuscitation she described, in halting English to social worker Kimberly Clark Sharp, the layout of the resuscitation room and the printer paper that had spilled on the floor. Then she said something stranger. She had drifted out of the building and noticed a dark blue tennis shoe sitting on a third-floor exterior window ledge on the north side of the hospital. She described the shoe’s wear pattern over the little toe and the way one lace had been tucked under the heel. Sharp went floor by floor along the north wing, pressing her face against each window. When she found a window where a shoe was sitting on the ledge, she had to put her face against the glass to see it — and even then she could not see the worn fabric or the tucked lace from the inside. The details Maria reported were visible only from outside the building.

This case is a textbook test of Marsh’s analogy and the analogy fails. There is no body-image story under which a sick patient’s parietal cortex spontaneously generates accurate information about a small worn-out shoe sitting on a third-floor exterior ledge that no one in the room had ever mentioned. A phantom hand is one thing. A phantom hand that knows where the shoes are is something else entirely.

Or consider a smaller, quieter case. Linda Morris and Kathleen Knafl, two nurses with Ph.D.s, interviewed nineteen colleagues about their experiences with critically ill patients. One nurse described a woman who had had an OBE during cardiac arrest. The patient mentioned, almost as an aside, that there was a penny on top of one of the cabinets in the resuscitation room — a place where you would have to stand on something to see. Another nurse climbed up. The penny was there.¹⁵

What links these cases is not their drama. It is the simple fact that they all contain information the patient could not have acquired from inside her body. A phantom limb is a closed loop. It runs entirely on what the cortex already had. These NDE perceptions are open loops. They contain content that has to come from somewhere outside the patient’s usual sensory range, and they verify back accurately to the world. Janice Holden’s exhaustive review of veridical NDE reports in the Handbook of Near-Death Experiences found that 92 percent of accurate elements in such cases were entirely correct — with another 6 percent partly correct.¹⁶ No closed cortical mechanism — phantom-body or otherwise — predicts those numbers.

There is a deeper conceptual confusion at the bottom of Marsh’s argument that is worth naming. He runs together two very different cortical accomplishments under the single word “construct.” The brain constructs the felt sense of having a body. Granted. But the brain also perceives the surrounding world, which is not the same operation at all and uses different cortical territory. Body-schema research, including the work Marsh cites from Catherine Reed and Martha Farah, deals with the first kind of construction.¹⁷ Veridical NDE evidence challenges the brain to do the second kind. Treating the two as scalable versions of the same trick is the move that lets Marsh’s analogy seem stronger than it is. Once you separate them, the analogy has nothing to say about the cases that matter most.

Some readers will already be reaching for a deeper move — what philosophers call the filter or transmission model of consciousness, in which the brain is more like a receiver than a generator. We engage that model fully in Chapter 23. For now, the narrower point will hold the chapter together. Even if you are agnostic about filter theories, even if you bring no metaphysical commitments to the table at all, the phantom-limb argument cannot bridge the gap between “the brain can build a felt arm” and “the brain can build an accurate three-dimensional view of a real room from outside the body.” That bridge has not been built, and the phantom limb is not its foundation.

D. Two Counter-Objections

A skeptic could push back on this in two main ways. The first objection takes the form of a concession and a shrug. Maybe the brain doesn’t literally see the room, the skeptic might say. Maybe it confabulates a scene from memory and prior knowledge of hospital rooms, and we just happen to remember the cases where the confabulation got lucky.

This is reasonable in the abstract and crumbles against the data. Confabulated content is generic. It runs on stereotypes, fills in plausible details, and stumbles on specifics. The cases above are all specific and unanticipated. The wear pattern on a particular tennis shoe. A wooden shelf on a hand-built metal cart. A penny atop a cabinet that no one had mentioned. None of this is the stuff of stereotyped guessing. And as we will see in Chapter 18, when memory researchers actually compare NDE memories with confabulated ones using objective scales, the NDE memories pattern as more like real perceptual memories, not less.¹⁸

The second objection appeals to recent virtual-reality experiments. Henrik Ehrsson’s rubber-hand illusion and the body-swap experiments by Olaf Blanke and Bigna Lenggenhager show that healthy subjects, given the right visual and tactile manipulations, can be made to feel a fake body as their own.¹⁹ If the brain can be tricked into that illusion, why is the OBE so impossible?

Because those experiments do not produce what NDEs produce. They produce a brief sense of relocation in healthy subjects who are watching a manipulated scene with their own eyes. Take away the cameras and the touch, and the illusion collapses in seconds. No one has ever come out of a rubber-hand illusion knowing the location of a worn shoe on an exterior ledge. The experiments demonstrate that the brain’s body-map is flexible, which no one denies. They do not demonstrate that a flatlining brain can manufacture accurate, externally verifiable perceptions of distant rooms. The gap between a slipped sense of self in a bright lab and an unconscious patient correctly describing her own resuscitation is enormous, and no experiment has begun to close it.

The phantom-limb chapter of Marsh’s book is one of his most engaging. It is also one of his most quietly disappointing, because the wonder of phantom limbs is real, and the leap to phantom selves is not. We are left, at the end of the analogy, exactly where we started: with a body of NDE evidence in which patients accurately report things they could not have known, and with a skeptical literature still searching for a brain mechanism that can do the same. Chapter 15 turns next to a different proposal — that NDEs are simply dreams — and asks whether that explanation fares any better.