Harry Houdini did not drown in his Water Torture Cell. Not in real life. And not even in the Hollywood movie about him.

He died on Halloween 1926, a hundred years ago, in a hospital bed in Detroit. A punch got the blame. And the only formal finding on that punch came from an insurance company.

Then came the séances. The couple’s secret code had been printed in a book seven months before a medium said it came from Harry.

So we went back to the record of the time. For each part of the story: what did the papers say then, and who told it later?

Albany

Start with the cell, because it did hurt him.

In early October 1926, Houdini opened in Albany, New York. On the opening night, the New York Times later reported, “a piece of apparatus used in his ‘water torture cell’ trick was overturned and struck him on the foot.” “A bone was found to be partly fractured.” His widow’s biographer, two years later, wrote that “Houdini felt one of his ankles snap.”

He was advised to rest. According to the Times, “He declined to cancel his engagements, however, and did not miss a show.”

The Water Torture Cell was not new. As the magician Milbourne Christopher wrote in 1954, “he featured this feat in performances here and abroad for years.”

That foot is the only part the cell plays in the real story of his death.

Montreal

On Tuesday, 19 October, in Montreal, he spoke to students at McGill University about the tricks of fake mediums. Exposing them had become his crusade. In 1924 he had published a pamphlet exposing the Boston medium “Margery.” In February 1926 he took the campaign to a Senate hearing in Washington.

Some time that week, in Montreal, a student hit him in the stomach. That much is in the record. Almost everything else about it comes in more than one version.

The New York Times said a student struck him at a reception after the talk: “One of the students without warning or giving time for Houdini to prepare struck him twice immediately over his appendix.”

The Chicago Tribune put it in his dressing room. A student asked if he would feel a blow, and “before he was aware of the boy’s intention his questioner hit him twice in the pit of the stomach.”

The version you probably know, the couch, the letter, the blows before he was ready, comes from his widow’s authorized biography, Harold Kellock’s Houdini: His Life-Story, published in June 1928. “The student called on Friday morning with a friend. The two sat in Houdini’s dressing-room while he lay on a couch.” Houdini “gave an absent-minded reply which the boy interpreted as consent.” In that telling there were more blows: “At the fourth blow he stopped the young man with a gesture.”

The insurance trade press, in 1927, described a blow “delivered at Houdini’s request by a student of McGill University.”

Which day (the 19th, 21st or 22nd), where, how many blows, and whether he agreed: the papers, the biography and, later, the insurer do not match.

Three students later gave statements. In February 1927 the New York Times reported that “Mrs. Houdini is said to have affidavits from three students of McGill University.” We are not naming them. They were private people, and their names come to us from copies whose originals we could not trace.

Detroit

He went on to Detroit.

On Monday, 25 October, surgeons operated for appendicitis. “His symptoms had mystified his physicians,” the Associated Press reported, “and when they operated they found his appendix far on the left side of the abdomen. Normally it would be at the right.”

Peritonitis followed. There was a second operation on the Friday. A new serum did not work. In the words of the Times: “Like a newly discovered serum, used for the first time in Houdini’s case, it was of no avail.”

On Sunday morning he spoke to his brother. The wire service had him say: “I guess I am all through fighting.” His widow’s biographer had it differently: “I’m tired of fighting, Dash. I guess this thing is going to get me.”

United Press also reported “an incoherent reference to Col. Bob Ingersoll, the famous agnostic.” Kellock disagreed: “Houdini was never delirious,” he wrote, and he called the Ingersoll story a crude story circulated among psychics.

At 1:26 that afternoon, Harry Houdini died in Grace Hospital. It was Sunday, 31 October 1926. Halloween. The papers that week gave the date. None that we read made anything of it being Halloween.

The New York Times put it in the headline: “Magician, Conscious to Last, Loses Fight for Life in Detroit Hospital.” The cause, as the Times gave it: “Death was due to peritonitis, which followed the first operation, that for appendicitis.”

He did not die in the water. He died in a hospital bed, a week after his last show, awake.

The punch in print

The next morning, the punch was news.

On the front page of the New York Times, Monday, 1 November 1926: “Playful Blow Given by Montreal Student as Test Caused Appendix to Break.” His doctors, the papers said, blamed the blow. The Chicago Tribune: “Detroit physicians are of the opinion that it was the blow received at Montreal that caused the ruptured appendix.” Neither paper names the doctors.

Just above it, the Times printed “Death Due to Poisoning.” In 1926, that meant the peritonitis. Nothing more.

In the bulletins from Detroit the week before, we found no student and no punch. He had been taken ill, they said, on the way from Montreal.

And a day after the headline, Montreal denied it. An Associated Press report from Montreal, printed on 2 November, quoted two men from his Montreal engagement: “Harry Houdini did not die of an injury received on Oct. 22 but was so sick when he came here that he was under care of a trained nurse.” A McGill professor “said there was no encounter between the magician and a student.”

So did a punch kill Houdini? Somebody had to decide. The answer did not come from a court.

The insurance verdict

In February 1927, his widow revived the story of the blow. The Times headline: “Mrs. Houdini Revives Report of Fatal Blow.” She was claiming double indemnity: insurance that paid twice if a death was an accident. An insurance trade weekly reported in April that “his widow is now claiming $500,000.” That figure is the family’s.

New York Life investigated. On 17 June 1927, the insurance trade paper The Eastern Underwriter reported the result: “it was found that he had met his death by a blow in the abdomen.” An accidental death. “The company has paid to the widow $50,000. Of this amount $25,000 was accident indemnity.”

That was one policy. Whether his other insurers paid accident money, we did not find.

According to Kellock, “Five of the leading surgeons of Detroit deposed” for the insurer “that death resulted from a gangrenous appendix caused by a blow.” Kellock adds that the student who gave the blows admitted them in a signed statement. We have not seen the depositions.

The only formal finding on the blow that we found came from an insurance company, not from a court.

Some doubted it. Arthur Conan Doyle, a believer in spirits, wrote in 1927, as Kellock quotes him: “The story about a blow from a student is, I believe, apocryphal.” Kellock answered that no American insurance company would pay accident insurance “as a result of death from an ectoplasmic hand.”

In October 1928 a reviewer of Kellock’s book wrote: “The true story of Houdini’s death is told apparently for the first time.” The blow had been on the front page of the New York Times two years before.

And today? Doctors say a punch can cause appendicitis, but rarely. “It is possible that the blows caused Houdini’s appendicitis,” the physician and historian Howard Markel wrote in 2019. The punch may have burst it. He may already have been ill. We can’t know.

A 2006 book suggested that Spiritualists had poisoned him. In 2007 the Associated Press reported that “his great-nephew wants to exhume the magician’s body to determine if enemies poisoned him.” The legal papers were never filed. It remained a theory. He is still in his family plot in Queens.

Home

He went home to New York in a coffin from his own act.

That August, he had stayed sealed in a coffin at the bottom of the Hotel Shelton pool. The Times reported that “exactly 1 hour, 31 minutes and 31 seconds later the casket was lifted to the surface.”

On 2 November the Times said his body was “Coming From Detroit in Bronze Coffin He Used to Remain In an Hour Without Air.” The pool casket, though, was galvanized, and after the test “four tinsmiths were ripping open the six-and-a-half-foot galvanized casket.” Kellock wrote of the coffin from the test: “His body was placed in it for shipment to New York, and later he was buried in it.” Which coffin it really was, the record does not settle.

On 4 November, after a funeral at the Elks’ clubhouse in West 43rd Street, “2,000 HEAR HIM PRAISED,” as the Times put it. He was buried at Machpelah Cemetery, in Cypress Hills, Queens, the Times reported, “with his head pillowed on the letters his mother had written him.”

His own Hollywood

Long before Hollywood made a film about Houdini, Houdini made films. A serial, The Master Mystery. Then features: The Grim Game (1919), Terror Island (1920), The Man from Beyond (1922) and Haldane of the Secret Service (1923).

He sold them the way he sold his escapes. For The Grim Game, the publicity in January 1920 said Houdini himself was on a rope between two planes when they collided and crashed. A theatre ad of February 1920 went further: “The above sum will be paid by Houdini to any person who can prove that the hair-raising airplane chase, collision and crash to earth, as shown in “The Grim Game,” did not actually occur.” The sum was $1,000.

The crash was real. According to his biographer Milbourne Christopher, writing in 1969, the man on the rope was a stunt double.

Variety, reviewing The Man from Beyond in 1922, wrote: “The story is raw melodrama.” Of his swim through the rapids it still said: “It is a true thriller.”

And on Halloween 1922, he faked a séance himself, with a spirit message from Jack London. Then he wrote in his diary, as Kellock quotes it: “I must put on record that it was done wholly by dexterity.”

Hollywood’s Houdini

Hollywood had been trying to film his life since the 1930s. Paramount’s first project, dated 1935 by the AFI Catalog, was sold on in 1944: “Vanguard Films has purchased “Houdini” from Paramount,” the Film Daily reported.

In September 1951 Paramount announced it again. Houdini “will be biographed by Paramount, with George Pal as producer,” Variety reported. “Picture will be based on a book by Harold Kellock”: that same 1928 biography.

“Everything in the film is based on fact,” Pal said in December 1952. “We took certain dramatic license on the time element, but biographically, our film is pretty close to the facts.”

Tony Curtis played Houdini. Janet Leigh, his real wife, played Bess. George Marshall directed.

Here is the thing. Even Hollywood didn’t drown him.

Read the reviews from May 1953. Variety: “In the film plot, Houdini comes to his death as result of a ruptured appendix, fatal attack occurring while he is doing a water torture act suspended head first in a tank.” Harrison’s Reports: “He suffers an appendicitis attack while in the tank, and when Janet screams the glass tank is broken to free him. He regains consciousness, but soon dies.” Motion Picture Herald has him “die of peritonitis during his first performance of a spectacular and newly-conceived escape.”

The film made the cell a new escape, never tried in public. The real Houdini had been doing it for years. And the punch, by later accounts, never made it into the finished film. Wild About Houdini, John Cox’s Houdini blog, reports that an early script summary of 1952 included it.

The reviewers knew what kind of film it was. Variety called it a “reel, not real, biopic.” Motion Picture Herald was kinder: Pal “here carefully reconstructs the well remembered facts about the great showman.”

The film also gave him a Halloween superstition. We have not seen the film, and we do not describe it here. The next January, in Variety, Milbourne Christopher quoted the collector Sidney Radner, who said he owned the real water cell. Radner called the film’s version of the cell ridiculous, and objected to “the fictional Hallowe’en superstition” and to “the manner of Houdini’s death involving a failure to do the upside-down, water cell escape.” Christopher closed the article with a thought: if Houdini could come back from the grave, “the Paramount Studios would have had a wrathful visit from the man.”

How he “drowned”

So where did the drowning come from?

Not from the 1953 reviews. None that we read says he drowned. They name the appendix, or peritonitis.

But by the 1970s, magicians’ own ads said the cell had killed him. Advertisements of 1976, shown on Wild About Houdini, said he died in the cell. Another magician’s poster, undated, called it “The Death Defying Act That Ultimately Took the Life of The Great Houdini.” Cox has written that the drowning belief “was born of the 1953 film Houdini.” On the reviews of the time, the film put his death in the tank. The drowning came later, in the retelling.

Even the film reference books slip. The AFI Catalog’s note on the film says he died of a ruptured appendix “in Montreal on 31 Oct 1926.” He died in Detroit.

The code

The other Halloween legend is the séance.

Harry and Bess had a secret code, from their old mind-reading act. Kellock printed “their code words for numbers”: ten words for ten numbers. Pray meant one. Be quick meant ten.

They made a pact: if he died, she would try to reach him. In 1928, as she described it to Kellock, the pact was a private one. “Each Sunday at a fixed hour she was to take her favorite photograph of him” and sit with it for half an hour. “There has been no result.”

In January 1929, a medium named Arthur Ford gave her a message in the code, beginning with the word Rosabelle. Decoded, it read: Rosabelle, believe. On 9 January she signed a statement that it “IS THE CORRECT MESSAGE PRE-ARRANGED BETWEEN MR. HOUDINI AND MYSELF.”

Here is the catch. The ten code words were already in print, in her own authorized biography. Its copyright page reads “Published, June, 1928”: seven months earlier. Even the medium’s own defender, Francis Fast, admitted in 1929 that the words “were published in a book by Harold Kellock.”

To be fair, the book did not print the first word, Rosabelle.

The next day, 10 January, a New York tabloid, the Evening Graphic, called it a hoax. That is the account of its reporter, Rea Jaure, who wrote in April that the paper “alone announced the communication from Houdini to Ford as a gigantic hoax.” We have not seen the Graphic’s page. She also claimed that Bess had rehearsed the séance with her the day before. That is her claim. Bess and Ford both denied it. Bess wrote, in a letter to Walter Winchell printed by Fast: “I was no party to any fraud.”

A year later, in March 1930, a wire story quoted her: “several great spiritualistic advocates declare that I succeeded, but it is untrue.” “There is only a void.”

The roof

Then came the night the legend remembers.

The story says she held a séance every Halloween for ten years. In 1928, the pact was Sundays. The every-Halloween line appears in October 1936, in the publicity for the last one. Variety: “Mrs. Houdini has tried each Hallowe’en for last nine years without success.” That is the earliest we found. We found no report of a Halloween séance by Bess from 1927 to 1935. Cox writes that “it does not appear Bess held her seances on Halloween at all.” That is not proof that none was held. The story was repeated in 1943, in an International News Service interview with her: she “every year on the anniversary of his death would hold a seance.”

In the Billboard of 31 October 1936, Bess placed a notice, “In Memory,” to the man “Who went away October 31, 1926.”

That night, ten years to the night, on the roof of the Knickerbocker Hotel in Hollywood, about two hundred people waited for him, “200 of Houdini’s friends,” as the wire put it. Her adviser, Edward Saint, led it: “The substance of the seance was delivered by Edward Saint.” A small electric light burned over Houdini’s photograph. According to Billboard, it was one of sixteen Houdini séances held that night in the United States, England, Australia and British Columbia, at her special request.

It was meant to go out coast to coast, on a radio “hook-up.” We could not find a network that carried it.

At the end, Saint asked her for her decision. The wire reports had her say: “My last hope is gone. I shall never try again.” A fuller report, from Winnipeg, reprinted in December: “‘He has not come,’ she said. ‘I turn out the light.’” A longer goodbye, known from a record set sold a few years later and from books since 1996, was not in any 1936 newspaper we read.

The storm that broke over the roof? One man remembered it, twenty years later. William Larsen Sr., writing in 1956 as quoted by Wild About Houdini, admitted the papers never mentioned it: “Probably it meant nothing for the newspapers failed to mention it.” No 1936 report we read mentions rain.

And her most famous line, “Ten years is long enough to wait for any man”? She did not say it on the roof. She said it in an International News Service interview, datelined Hollywood, 5 February 1943: “‘Ten years is long enough to wait for any man,’ she said today.” We found it in no earlier source. Within days, she was dead.

Magicians revived the Halloween séance after the war. Cox reports that a séance of 1948 is “generally seen as the genesis of what is today known as” the Official Houdini Séance.

Even last year’s official séance press release, in 2025, says the 1936 séance was “nationally broadcast” and puts the “ten years” line on the roof.

What the record shows

Houdini did not drown. He died of peritonitis in a hospital bed in Detroit, on Halloween 1926, a week after his last show, awake. The Water Torture Cell’s only part in that story is a partly fractured bone in his foot, in Albany.

The punch happened, in some version. Its day, its place, the number of blows and whether he agreed are told differently by the papers of 1926, his widow’s biography and the insurer. Whether it killed him, nobody can prove. The only formal finding on it came from an insurance company.

Hollywood didn’t drown him either. The 1953 film moved his death into the tank, by appendix. The drowning came later.

The code proves less than it seemed. The ten code words were in print seven months before Arthur Ford’s message. The first word, Rosabelle, was not.

The every-Halloween séances before 1936, and “ten years is long enough” on the roof, are later layers.

This Halloween, the centenary, séances are announced inside his old house in New York.

We’ll leave the light on.

Evidence limits

This account checks the story of Houdini’s death, and the legends that grew from it, against the record of the time. It does not judge Houdini, Bess Houdini, the students or anyone’s motives, and it assigns no cause of death beyond what the doctors and the insurer said in 1926 and 1927. “Earliest we found” and “we found no” mean in the sources we could search: the Internet Archive’s newspapers, magazines and books (including its full-text search), the Media History Digital Library’s trade papers, Papers Past (New Zealand) and Trove (Australia). Chronicling America (Library of Congress) could not be searched during the research; two of its pages are used for pictures only. That is not proof that an earlier telling does not exist. Not all papers could be searched. The Montreal and Albany papers of 1926, the New York Evening Graphic of January 1929, and most United States papers of 1 and 2 November 1936 were out of reach. The Detroit bulletins were searched, not read page by page. The radio search covered the trade press and the Midwest edition of Radio Guide for 31 October 1936. The New York Times of 1926 and 1927 was read off the page images. Kellock’s biography was read in an Internet Archive scan of a 1930 printing whose copyright page reads “Published, June, 1928”; that “Rosabelle” is not in it rests on a full-text search of that scan. Conan Doyle’s words are quoted as Kellock printed them; the Strand page was not seen. The stunt double in The Grim Game rests on Christopher’s 1969 biography, read in an unofficial upload, and is not named here. The January 1920 Grim Game publicity was seen as a full-text search snippet. The 1976 advertisements, the undated poster, Larsen’s 1956 memory and the 2025 press release were seen only as Wild About Houdini shows them. The students’ affidavits are known to us only from web transcriptions whose originals’ holder is not named; the students are not named here. Bess Houdini’s letter to Winchell is quoted as Fast printed it. Rea Jaure’s claims are given as her claims, with the denials. The poisoning theory is reported as a theory. We have not seen the 1953 film; it is described only in the words of its reviewers. The video shows no frames or clips from the 1953 film; publicity stills from it appear only as printed in a 1953 trade advertisement, for commentary.

Sources

1920 to 1929

  • The Star (Oneonta, N.Y.), 4 Feb 1920, p. 2, Oneonta Theatre advertisement for The Grim Game
  • Atlanta Constitution, 6 Jan 1920, cinema column on The Grim Game (Internet Archive full-text search)
  • Variety, 7 Apr 1922, p. 41, review of The Man from Beyond (Internet Archive)
  • Houdini, HOUDINI Exposes the tricks used by the Boston Medium “Margery” (Adams Press, 1924), title page (Internet Archive)
  • The New York Times, 6 Aug 1926, p. 32, the Hotel Shelton pool test (Internet Archive)
  • St. Louis Post-Dispatch, 26 Oct 1926, p. 1, “Houdini Operated On, Is in Grave Condition” (AP, Detroit) (Internet Archive)
  • The New York Times, 1 Nov 1926, pp. 1 and 6, “Harry Houdini Dies After Operations” (Internet Archive)
  • Chicago Daily Tribune, 1 Nov 1926, p. 1, “Houdini, Noted Stage Figure, Dies in Detroit”
  • Rushville Daily Republican (Indiana), 1 Nov 1926, p. 1, “Houdini’s Last Act Is Loosed By Death” (United Press, Detroit) (Internet Archive)
  • St. Louis Post-Dispatch, 1 Nov 1926, p. 1
  • The New York Times, 2 Nov 1926, p. 27, “Injury in Montreal Is Denied” (AP, Montreal) and the bronze coffin (Internet Archive)
  • The Washington Times, 5 Nov 1926, Fay King cartoon
  • The New York Times, 5 Nov 1926, p. 21, “Houdini Eulogized in Impressive Rites” (Internet Archive)
  • The New York Times, 26 Feb 1927, p. 3, “Mrs. Houdini Revives Report of Fatal Blow” (Internet Archive)
  • The National Underwriter (Life Insurance Edition), 1 Apr 1927, p. 13, “Houdini’s Beneficiary Wants Double Indemnity” (Internet Archive)
  • The Eastern Underwriter, 17 Jun 1927, p. 15, “Houdini Insurance Settled” (Internet Archive)
  • The New Republic, 17 Oct 1928, p. 248, “A Great Magician,” review of Kellock (Internet Archive)
  • Rea Jaure, “Rev. Ford faked Houdini message,” Science and Invention, Apr 1929, pp. 1151, 1203, 1209 (Internet Archive)
  • Greymouth Evening Star, 23 Feb 1929, p. 12 (US syndicated story, San Francisco)

1930 to 1943

1944 to 1954

Books

Articles, posts and catalogues since 2007

Pictures, pages and film in the video

Photographs, posters and lantern slides, Library of Congress, Prints and Photographs Division (McManus-Young Collection unless noted):

Other photographs:

Silent film clips (no sound), Internet Archive:

Web pages shown as still screenshots, captured on 21 Oct 2026, for commentary: the AFI Catalog entry for Houdini (1953); Wild About Houdini, 23 Mar 2012, and the 2025 Official Houdini Seance press release as printed there; the AP report of 23 Mar 2007, via Fox News.

The video’s narration uses a synthetic channel voice. No AI-generated pictures are used. Newspaper, magazine and book pages are shown for commentary. No frames or clips from the 1953 film are used, and the silent film clips carry no sound.

Corrections

Hollywood Case File separates documented claims from reported accounts. If you know of a newspaper before 1953 that said Houdini drowned in the Water Torture Cell, a report of one of Bess Houdini’s Halloween séances from 1927 to 1935, or a primary record that changes or clarifies this account, use the site’s corrections and contact route, with a date and a publication. A published correction should identify what changed, why it changed, the date, and the supporting evidence.