A study · England, published in 1894

The census of hallucinations: 17,000 people and the apparitions that coincided with a death

At the end of the 19th century, a group of researchers in London put a question to 17,000 people. They wanted to know whether people had ever seen a living being or an object, felt it touch them, or heard a voice while awake. It had to be something for which they could find no physical cause. Then they counted how often someone saw a known person close to the time that person died. They calculated that such a coincidence happened about 440 times more often than chance would produce. They themselves discounted many cases for exaggeration, and other researchers doubt the calculation.

The study’s question is: when someone sees a loved one at the very moment that person dies far away, can it be chance? We tell it step by step. If you only want to know what it found, go straight to the results.

Step 1

Before the study · The question

An apparition at the moment of a death: chance?

Illustration from The census of hallucinations: 17,000 people and the apparitions that coincided with a death: “An apparition at the moment of a death: chance?”

Some people say they saw or heard a relative or a friend at the very moment that person was dying far away. Later the news of the death arrived.

A coincidence like that is striking. To know whether it is chance, you also have to count the other times, the apparitions that coincided with nothing. If many people see apparitions, some will coincide with a death by pure chance.

In 1886, three researchers had already tried to count coincidences between dreams and deaths in the book “Phantasms of the Living.” This time, the committee studied apparitions seen while awake, and with many more people.

Step 2

The team · As the report tells it

Five researchers from the Society for Psychical Research

Illustration from The census of hallucinations: 17,000 people and the apparitions that coincided with a death: “Five researchers from the Society for Psychical Research”

The Society for Psychical Research was founded in London in 1882 to study methodically the phenomena regarded as paranormal. Henry Sidgwick, a philosopher at the University of Cambridge, was its first president.

The census committee was made up of Henry Sidgwick, his wife Eleanor Sidgwick, the researcher Alice Johnson, and Frederic Myers and Frank Podmore. Myers and Podmore were two of the authors of “Phantasms of the Living.”

Step 3

The survey · As the report tells it

17,000 people answer one question

Many helpers, volunteers who assisted the committee, handed out the same question among their acquaintances. It asked whether people had ever seen a living being or an object, felt it touch them, or heard a voice while sure they were awake. It had to be something for which they could find no physical cause.

17,000 people answered. 2,272 said yes. After removing dreams and cases that did not fit the question, 1,684 people with hallucinations while awake remained. That is about 10 in every 100.

Step 4

The method · As the report tells it

What counts as a coincidence with a death

Illustration from The census of hallucinations: 17,000 people and the apparitions that coincided with a death: “What counts as a coincidence with a death”

The committee called a death-coincidence an apparition of a recognized person who died close to that moment. The person had to die within the 12 hours before or after the apparition. It chose that 12-hour margin for convenience of calculation.

It left out cases in which the witness already knew the person had died. It also left out apparitions seen by children under ten, and those of people who had many hallucinations but did not say how many.

With those rules, 65 coincidences remained, told first-hand by the people who experienced them.

Step 5

The deductions · As the report tells it

From 65 coincidences to 30, out of caution

Illustration from The census of hallucinations: 17,000 people and the apparitions that coincided with a death: “From 65 coincidences to 30, out of caution”

The committee lowered the figure step by step so as not to exaggerate. Some deductions are specific cases. Others are estimates:

3 cases that the helpers had sought out on purpose, because they already knew the story. 62 remained.
22 for exaggeration, an estimate. There were more cases from many years before than recent ones. It should be the other way round, because older events are forgotten more. For the committee, this suggested that over the years some accounts were adjusted until they looked like coincidences. The committee also pointed to another possible reason. Some witnesses may have kept recent cases to themselves because they were still painful. It estimated that about 22 were not real coincidences, without knowing which. It warned that there were too few cases to be sure. Even so, it made the deduction. 40 remained.
8 more, as a precaution, another estimate. The helpers may have mostly asked people they already knew had a story like this. The committee worked it out from the cases the helpers already knew about. 32 remained.
2 doubtful cases, one seen from a distance and one perhaps in bad light. The committee thought it very unlikely that both were among the 32. Even so, it removed them to be on the safe side. 30 remained.

The committee did its calculation with 30 death-coincidences.

Step 6

The calculation · As the report tells it

About 1,300 apparitions of living people

Illustration from The census of hallucinations: 17,000 people and the apparitions that coincided with a death: “About 1,300 apparitions of living people”

To compare, it was necessary to know how often someone sees a known person who does not die. The survey contained 381 apparitions of recognized living people. Those of people who had many hallucinations but did not say how many were removed. 350 remained.

Many of these apparitions are forgotten over time. The committee compared what people remembered from the last year with earlier years, and calculated that only 1 in 4 is remembered. So it multiplied them by 4: 350 times 4 is 1,400. Then it removed those of children. It got about 1,300.

With death-coincidences it did the opposite. It assumed that none is forgotten, because they leave a much deeper mark.

Step 7

The result · As the report tells it

About 440 times more than chance would produce

Illustration from The census of hallucinations: 17,000 people and the apparitions that coincided with a death: “About 440 times more than chance would produce”

The committee used the death rate for England and Wales between 1881 and 1890. With those figures, out of every 19,000 people, one dies on a particular day. That is the probability that any given person dies on a given day, and it is what chance would produce.

The survey had 30 death-coincidences against about 1,300 apparitions of living people. That is 1 in 43. Chance would produce 1 in 19,000, and 19,000 divided by 43 is about 440. So there were about 440 times more coincidences than chance would produce.

The committee had left out doubtful apparitions, such as those seen from far away or in bad light. If they are added, the total of apparitions is larger, the proportion of coincidences falls, and the figure becomes about 292 times.

Step 8

In the report · As the committee replies

Three objections the committee raised itself

Illustration from The census of hallucinations: 17,000 people and the apparitions that coincided with a death: “Three objections the committee raised itself”

According to the committee, anyone who did not believe in a paranormal explanation would have to argue one of these three ideas. This is how it answered each one:

That the coincidences are misreported. The difference from chance is very large. For the committee, almost all the accounts would have to be misreported at once for the result to be chance. It printed 31 of the 62 cases for the reader to judge.
That the helpers sought them out. In 26 of the 62 cases, the helpers say they knew nothing of the experience before asking the question.
That the witness, worried about a sick person, imagined seeing them. Of the 62 cases, in at least 19 the witness did not know the person was ill. In another 18 they knew the person was ill, but not that they were in danger. In another 20, the witness knew the illness was serious, or it cannot be ruled out that they knew. In the remaining 5, the witness was anxious for other reasons or the case is unclear.

The committee also spoke in person with 21 of the 62 witnesses. Edmund Gurney, another researcher of the society who was not on the committee, spoke with 3 more. They found mistakes in 2 accounts.

Step 9

1894 · As the committee concludes

For the committee, it is not chance

The committee concludes that these coincidences are not due to chance. It writes that, if its figures are accepted, nobody with basic knowledge of probability could say it is chance.

For the committee, the explanation is telepathy, communication from mind to mind. It believes that whoever dies somehow conveys their image to a person who loves them.

Let’s see what can and can’t be known from these data.

What is known?

What the census of hallucinations found

What it found

About 10 in every 100 people said they had had a hallucination while awake. That was 1,684 out of 17,000.
There were 65 accounts of an apparition that coincided with a death. The committee lowered the figure to 30 for its calculation. It estimated how many might be exaggerated or deliberately sought out, without saying which ones.
According to its calculation, that is about 440 times more than chance would produce. Counting doubtful apparitions, about 292.
There were more old cases than recent ones, more than expected. The committee read this as a sign that memory exaggerates over the years.

What it couldn’t find out

Whether each coincidence happened as told. Almost everything is told by the witness themselves, often years later. 24 of the 62 witnesses were interviewed in person, 21 by the committee and 3 by Edmund Gurney.
Whether the time was adjusted after the death became known. A witness might remember the apparition as closer to the death than it was. The committee deducted about 22 coincidences for this reason. It could not check whether more errors remained among the 30 it used.
How much it matters that many of the people seen were ill. A seriously ill person is more likely to die on a given day than an average person. But the calculation uses the average death rate of England and Wales. In 20 of the 62 cases, the witness knew the illness was serious, or it cannot be ruled out that they knew.
How many coincidences are forgotten. The committee assumed none, and could not check it.

How each side interprets it

What the committee concludes

Henry Sidgwick and the committee of the Society for Psychical Research, in 1894. They believe the coincidences between apparitions and deaths are not due to chance, and that the explanation is telepathy.

What others object

G. N. M. Tyrrell, a researcher of apparitions, in 1943. He wrote that he could not attach any importance to the final figure of the calculation. This view is quoted by the philosopher Keith Augustine.
Donald West, a researcher of the same society, in 1948. He repeated the question by post to 1,519 people. Some people believed their hallucination warned of something real. West could not confirm any of those. He wrote that his study had too few cases to judge the 1894 calculation.
Keith Augustine, a philosopher and critic of the idea of life after death, in 2022. He writes that the figure of 440 is often quoted, but that the reasoning behind it is dubious.

What can be said from the data

According to the committee’s calculation, chance does not explain the 30 coincidences. For the committee, almost all the accounts would have to be wrong at once. The risk lies in memory. A witness might remember the apparition as closer to the death than it was. The committee estimated that about 22 of the 62 reported coincidences were not real ones, and deducted them. It could not know whether more errors remained among the 30.

So the census does not show that a dying person can appear to someone else. Nor does it show that these accounts are false.

What would settle it

It would take what the committee itself could not have:

writing down the apparition, with its time, before the news of the death arrives;
checking the time of death in a document from the period;
and repeating the survey with many people, ideally soon after each apparition.

From 17,000 answers to 30 coincidences

The committee lowered the figure from 65 coincidences to 30. It deducted 35, more than half.

Even so, according to its calculation, far more remained than chance would produce.

The figures of the census of hallucinations
Item	Figure
People who answered	17,000
People with hallucinations while awake, not counting dreams	1,684
Death-coincidences reported by the witness	65
Coincidences the committee used in its calculation, after its deductions	30
Apparitions of living people, estimated	1,300
Times more than chance, according to the committee	440

What these figures don’t tell

They are accounts. Each coincidence is told by the person who experienced it, often years later.
The forgetting figures are estimates. The committee calculated that 1 in 4 apparitions is remembered, and so multiplied by 4. And it assumed that no death-coincidence is forgotten.

Frequently asked questions

What was the 1894 census of hallucinations?

It was a survey of 17,000 people by the Society for Psychical Research in London. It asked whether people had ever seen a living being or an object, felt it touch them, or heard a voice while awake. It had to be something for which they could find no physical cause. The committee counted how often someone saw a known person close to the time that person died.

Is it chance to see a loved one just as they die far away?

The committee of the 1894 census calculated that such coincidences happened about 440 times more often than chance would produce. Other researchers doubt the calculation, because it relies on accounts sometimes told years later. The committee already deducted cases in case the witness remembered the apparition as closer to the death than it was. It could not know whether it deducted enough.

What is a crisis apparition?

It is when someone sees, hears or feels a known person who at that moment is in danger or dying far away. The 1894 census counted as a coincidence an apparition that happened within 12 hours before or after the death.

Where this page comes from

The study

Henry Sidgwick, Alice Johnson, Frederic Myers, Frank Podmore and Eleanor Sidgwick (1894), “Report on the Census of Hallucinations,” Proceedings of the Society for Psychical Research 10:25-422, public domain. The calculation is in chapters XII and XIIIRead for free

Assessments and criticism

Donald West (1948), “A Mass-Observation questionnaire on hallucinations,” Journal of the Society for Psychical Research 34:187-196Read for free
Keith Augustine (2022), “How not to do survival research,” Journal of Scientific Exploration 36(2):366-398Read for free

To learn more

Dreams that warned of a death: the 24 cases in “Phantasms of the Living” (Gurney, Myers, and Podmore, 1886)Read
337 encounters with the dead in Iceland: the Haraldsson study (2009)Read
“My name is Margret”: a shared-death experience in IcelandRead

Our archive

A reading of chapters XII and XIII, the count in chapter II and the 31 printed cases, checked on October 23, 2026

More studies

AWARE: the pictures hidden on the shelves
Fifteen hospitals put up pictures that could only be seen from the ceiling, to check whether anyone saw them during a cardiac arrest.
AWARE II: a random picture and the brain on the screen
Above the patient was a tablet with a random picture. Headphones said the names of three fruits. And the doctors measured the brain during resuscitation.
Van Lommel: 344 cardiac arrest survivors in ten hospitals
Ten hospitals asked every cardiac arrest survivor what they remembered. They saw them again after 2 and 8 years.