Chapter 19

Cutting Both Ways: Confirmation Bias and Researcher Bias

Michael Sabom did not believe a word of it. The year was 1976. Raymond Moody’s Life After Life had just appeared, and a social worker named Sarah Kreutziger asked the young cardiologist to give a medical opinion on the strange stories the book was collecting. Sabom went to her seminar fully expecting to demolish the whole thing. He thought near-death experiences were nonsense — a curiosity, maybe, but nothing serious.¹

So he did what doubters do. He set out to refute it.

He started interviewing his own cardiac-arrest patients — not believers Moody had handpicked, but a clinical population from his hospital. He compared what they reported against what they could have known. He kept careful records. He waited for the inconsistencies he was sure he would find.

A year later, Sabom was still looking. Several years later, he wrote one of the most quoted lines in NDE research: “I have searched for such an explanation over the past five years and have not yet found one that is adequate.”²

Sabom never wanted to be an NDE researcher. The evidence drafted him.

That story is a problem for one of the most popular skeptical objections to NDE research: the charge that NDE researchers are biased — that they see what they want to see. Michael Marsh and the philosophers John Martin Fischer and Benjamin Mitchell-Yellin both press this charge hard. We need to take it seriously. Then we need to ask whether the bias mirror reflects the critics, too.

A. The Critic’s Argument

Marsh devotes his third chapter to attacking the integrity of NDE research. NDE researchers, he argues, recruit from populations that are already convinced. They circulate inside IANDS — the International Association for Near-Death Studies — which functions, by his account, more as an advocacy network than a scientific society.³ Their questionnaires are designed by people who already believe. Many of them — Sabom, Ring, Rawlings — are openly religious or sympathetic to spirituality. Their conclusions, Marsh suggests, are baked into their methods.

He spends extended pages on Kenneth Ring. Ring’s Religiousness Index, Marsh argues, smuggles in assumptions. Ring presents NDE-driven shifts in belief as evidence for what he calls a “universalistically-oriented spirituality” — but the inventory itself is loaded toward that conclusion.⁴ Marsh sees the same pattern across the field: researchers who set out to find what they were already inclined to find. He even worries about subtler biases — the way an interviewer’s warmth, or an experiencer’s desire to please, or the cultural script of an IANDS conference can shape what gets remembered and what gets reported.⁵

Fischer and Mitchell-Yellin sharpen the same point philosophically. Their twelfth chapter is titled simply “Confirmation Bias: We Believe What We Want to Believe.” Drawing on Stuart Vyse and Richard Posner, they argue that human beings naturally cling to beliefs that comfort them. We seek out evidence that supports our existing convictions; we ignore or explain away evidence that does not. We form communities of like-minded people who reinforce each other’s views.⁶ What could possibly be more comforting than the belief that death is “only a transition to a place where we would continue to be conscious, engaged in the projects that provide meaning in our lives, and connected to the people we love”?⁷

They pick the neurosurgeon Eben Alexander as their case study. Alexander’s The Map of Heaven, they say, is “one long exercise in confirmation bias.” Alexander does not engage arguments against supernaturalism; he marshals only evidence that supports his view.⁸ They share an anecdote: when one author’s son tried to describe their forthcoming book to an older neighbor, she cut him off, saying, “I will never read your father’s book! I want to believe Eben!”⁹ That, the authors say, is confirmation bias laid bare.

Both critiques converge on the same conclusion. NDE evidence looks strong only because the researchers and the reporters are predisposed to find it strong. Strip the bias away — the argument goes — and the case for survival evaporates.

It is a serious charge. It deserves a serious answer.

B. Weaknesses in the Argument

The trouble is that the bias mirror reflects two ways. And neither Marsh nor Fischer and Mitchell-Yellin look at their own reflection.

Marsh announces his thesis in the introduction to his book — before he has reviewed a single piece of NDE evidence. ECEs, he writes on page xvi, are “brain-state phenomena” produced by “metabolically disturbed brains.”¹⁰ The verdict is delivered before the trial begins. The detailed engagement with veridical evidence does not arrive until chapter 2; the wholesale dismissal of it appears in the front matter. If Sabom’s evangelical Christianity disqualifies his findings, what should we make of an author who tells us his conclusion before reading the data?

Fischer and Mitchell-Yellin display the same one-way mirror. They press confirmation bias hard against believers. They never once turn the lens on themselves. The very tendencies they describe — tenacious clinging to existing beliefs, seeking out information that accords with those beliefs, forming communities of like-minded people — apply just as well to academic philosophers committed to physicalism. Chapter 25 takes up the unstated physicalist assumption that runs through almost all NDE skepticism; here we just need to note that the argument is asymmetric. Confirmation bias is presented as a universal human problem — and then deployed in only one direction.

A second weakness is that the bias charge is, in form, ad hominem. It tries to discredit a conclusion by impeaching the people who reached it. That is a perfectly legitimate move when motives can be shown to have distorted the data. It is not legitimate as a substitute for engaging the data. Both Marsh and Fischer and Mitchell-Yellin spend significant pages on motivations and surprisingly few on the strongest verified cases. That is the wrong order of business. Even a researcher with terrible priors can stumble onto a true result. The way to test the result is to test the data, not to interview the researcher’s pastor.

A third weakness is the selection of targets. Fischer and Mitchell-Yellin choose Eben Alexander as their textbook example. They may well be right about The Map of Heaven; that book reads like marketing more than science. But the rigorous prospective studies — van Lommel’s Lancet paper, Parnia’s AWARE trials, Janice Holden’s veridical-perception analysis in the field’s standard handbook, Bruce Greyson’s longitudinal work, the verified cases compiled in The Self Does Not Die — barely appear in their critique.¹¹ A skeptic who picks the weakest popular book as the representative of an entire research field has not refuted the field. He has refuted that book.

C. The Pro-NDE Response

Three lines of evidence dismantle the bias objection.

1. Many leading NDE researchers began as skeptics.

Confirmation bias is supposed to draw researchers toward what they already believe. The pattern in this field runs the opposite direction. Some of the most rigorous NDE investigators began as committed materialists or skeptics. The evidence dragged them, sometimes against their will, to revise their views.

Pim van Lommel is the clearest case. The Dutch cardiologist conducted what is still the most ambitious prospective NDE study in the literature, published in The Lancet in 2001.¹² Before the study, van Lommel was a thoroughgoing materialist. He has said it plainly: “That death is the end used to be my own belief.”¹³ When his first cardiac patient described an out-of-body experience, van Lommel filed it away and forgot about it for years. Only after several similar reports did he finally take the question seriously.

Sabom we have already met. He set out, in his own words, to refute Moody. He tested the claims for a full year before he began to believe people were genuinely leaving their bodies.¹⁴ Maurice Rawlings, the cardiologist who would later write Beyond Death’s Door, said before his first NDE encounter that he “had always thought of death as painless extinction” and viewed religion as “all hocus-pocus.”¹⁵ The Welsh nurse-researcher Penny Sartori dismissed her first NDE case as “wishful thinking” before changing her mind only after years of careful study in an intensive-care unit.¹⁶

Bruce Greyson, the University of Virginia psychiatrist who developed the most widely used NDE measurement scale, says it directly: “Most near-death researchers did not go into their investigations with a belief in mind-body separation, but came to that hypothesis based on what their research found.”¹⁷

That is not the signature of confirmation bias. It is closer to its opposite. Confirmation bias predicts that researchers will find what they expected to find. What we actually see in NDE research is investigators who came to conclusions they had previously dismissed as nonsense. If commitment-protection were really driving the field, the field should be filled with believers reporting belief. Instead it is filled with skeptics reporting that the evidence has changed their minds.

2. Prospective hospital studies are designed to control for bias.

The bias charge would have force if the field still relied on the methods of the 1970s — collecting voluntary letters from people who had already told their stories at IANDS gatherings. It does not. The serious work of the past quarter century has been prospective: identifying cardiac-arrest patients in advance, interviewing them as soon as they regain consciousness, comparing their reports against control groups, and verifying anything they say about events during the arrest with hospital records and staff.

Van Lommel’s Lancet study is the model. Three hundred forty-four resuscitated cardiac patients across ten Dutch hospitals.¹⁸ Every patient had the same diagnosis. Every patient was interviewed within five days of resuscitation. Every patient was re-interviewed at two years and again at eight years to test whether memories drifted over time. NDE-reporting patients were compared against non-NDE patients with identical clinical profiles. The whole architecture of the study was built to detect the kinds of distortions a biased researcher could introduce. Van Lommel did not get to choose his subjects. He had to take whoever survived.

Sam Parnia’s AWARE studies went further still. Parnia’s team installed hidden visual targets — pictures placed face-up on shelves where they could only be seen from above — in resuscitation rooms across multiple hospitals.¹⁹ The targets were chosen specifically so that any veridical claim could be tested against an objective record. If a patient said, “I saw a picture of an elephant on the shelf,” researchers could check whether the elephant image was the one assigned to that room. The design is the opposite of confirmation bias: it builds in conditions under which the supernaturalist hypothesis could be falsified. A field driven by motivated reasoning does not invent tests it could fail.

Or take Janice Miner Holden’s review of veridical perception cases, published in the Handbook of Near-Death Experiences — the standard academic reference in the field. Holden examined ninety-three published cases of out-of-body perception during NDEs that included verifiable details. Ninety-two percent were entirely accurate. Six percent contained some small error. Only one percent were wholly mistaken.²⁰ A field driven by confirmation bias does not produce that pattern. A field driven by confirmation bias produces a wash of unverified anecdotes — not a 92 percent accuracy rate against external checks.

These methodological choices were not made in spite of the bias objection. They were made because of it. Serious NDE researchers have spent decades trying to build studies that bias cannot easily distort.

3. Veridical perception is an objective check that bias cannot dissolve.

This last point is the most important. Confirmation bias can shape the interpretation of subjective reports. It cannot conjure objective facts.

Either the dentures were on the crash cart or they were not. Either the surgical instrument matched the description or it did not. Either the cardiologist who left the room had the hair color the patient described or she had a different one. These are public facts that can be checked. They are not what the researcher hoped to find; they are what the world is. (Chapter 18 handles the related question of whether memories of those facts are reliable; here we are simply noting that the facts themselves are external to anyone’s priors.)

Bias can lead a researcher to ask leading questions. It can lead an interviewee to embellish. But it cannot make a congenitally blind woman accurately describe the pattern on a tie she has never seen with functioning eyes. It cannot make a cardiac patient correctly identify a unique surgical instrument that was first introduced after he lost consciousness. The veridical cases compiled in The Self Does Not Die — over one hundred independently verified accounts — are an empirical anchor that no amount of researcher bias can pull free.²¹

Fischer and Mitchell-Yellin, to their credit, half-recognize this. They acknowledge that veridical claims, if accurate, would matter. Their response is to redirect: such cases must be rare, must be misreported, must be unverifiable.²² But that is a very different argument from “researchers see what they want to see.” It is an argument about the data themselves — and it is the argument the rest of this book engages, chapter by chapter.

D. Counter-Objections

Two replies remain.

“Belief in an afterlife is uniquely comforting; physicalism offers no comparable inducement.”

Fischer and Mitchell-Yellin lean on this point heavily. Terror management theory, they say, predicts that people will reach for survival narratives because they fear death.²³

Three things weigh against the argument. First, it commits the genetic fallacy. Even if a belief offers comfort, that has no bearing on whether it is true. Many true beliefs are comforting; many false beliefs are also comforting; comfort is not a reliable guide to either. Second, materialism offers comforts of its own — no future judgment, no moral accountability after death, no questions about ultimate destiny. Some people find that prospect deeply settling. The “we believe what we want to believe” sword cuts the materialist as cleanly as the believer. Third, the empirical pattern is mixed. Distressing NDEs exist. Nancy Evans Bush has documented them carefully.²⁴ If experiencers were producing the experiences they wanted, distressing NDEs should be vanishingly rare. They are not. By some estimates they make up between ten and twenty percent of NDEs.²⁵ And many of the most evidential cases come from atheists and skeptics who were emphatically not seeking spiritual reassurance.

“Pro-NDE research is funded by the Templeton Foundation, which has its own theological agenda.”

The Templeton Foundation does fund research in this area. It also funded Fischer and Mitchell-Yellin’s Near-Death Experiences, as the authors openly acknowledge in their preface.²⁶ A foundation that supports both sides of a debate is not a partisan distorter; it is a patron of inquiry. Funding can compromise research only when it skews methods. Van Lommel’s Lancet paper, Parnia’s AWARE protocols, Greyson’s NDE Scale — all stand on their methodological merits, regardless of where their authors found their grants.

The bias objection is real. Every researcher carries commitments — including the ones who write skeptical books. The right response is not to declare inquiry hopeless. It is to design studies that bias cannot easily distort, to test claims against objective records, and to follow the evidence even when it leads somewhere our priors did not expect. That is what the best NDE researchers have done. It is why their work survives. The next chapter takes up a related but distinct question: whether the content of NDEs is just what we would expect a culture to invent.