Somewhere in Ottawa, in a government building no document we have found bothered to name, there was a dentist's chair. Just a chair -- probably black vinyl, adjustable, the kind that has a headrest and makes a particular sound when you sit down in it, familiar and clinical and somehow always slightly cold. Except this one had a camera on a pulley aimed at your face, and cables running to a box that measured the sweat on your hands, and another attachment monitoring your pulse, and directly in front of you, a screen waiting to show you things.
They called it the fruit machine. Not officially -- officially it was the Special Project, which is the sort of name you give something when you know, even while naming it, that the real name should probably never turn up in a headline. The unofficial name came from an RCMP sergeant whose identity has been absorbed by the same decades that swallowed most of the truth about what happened here. Fruit was the slang for gay men in postwar Ottawa. The machine was the government's instrument for finding them. Together -- the slang, the chair, the camera, the sweat-measuring box -- they constitute Canada's Cold War contribution to the science of human surveillance.
It never worked. Not once, not in a single documented case, did the camera's findings end a career. Keep that in mind. We will come back to it.
The thing about fear is that it makes excellent bureaucratic fuel. It keeps offices lit late. It fills filing cabinets. In Ottawa in the late 1950s the fear was Soviet intelligence, and it was not entirely irrational -- Soviet handlers were, in fact, looking for people to compromise, and a compromised civil servant with access to sensitive material could do genuine damage. The RCMP's Security Service understood this. The Privy Council Office understood it. The Security Panel -- a Cabinet-level body drawing its membership from External Affairs, National Defence, the RCMP, and the PCO -- understood it with particular care, because the Security Panel was where fear got organised into policy.
Fear, once organised, tends to want to expand.
On 12 May 1959, a Privy Council Office official named D.F. Wall sent the Security Panel a memorandum. Its full title was "Security Cases Involving Character Weaknesses, with Special Reference to the Problem of Homosexuality" -- sixteen words of preamble, clearing its throat before saying what it had to say. Wall's argument was compact: gay federal employees were vulnerable to blackmail by Soviet handlers who might discover their secret; blackmail could produce treason; therefore homosexuality was a security risk; therefore gay employees had to be removed. The logic ran clean. Brisk, even. The kind of reasoning that sounds airtight until you notice the load-bearing beam is absent.
The beam: the same documents acknowledged there were no recorded cases of a Canadian homosexual having committed an act of treason under blackmail threat. Not one. The entire policy rested on a premise its own authors knew was unsupported. The Security Panel found this either insufficiently troubling or insufficiently interesting to address, and pressed ahead. That is, honestly, the part of this story most worth sitting with. Not that they were wrong, but that they knew they were working without evidence and kept going.
Cabinet Directive No. 4 had established loyalty screening in 1948. Cabinet Directive No. 29, in 1955, expanded the criteria to include "character weakness" -- a phrase that, in that context and that era, carried a specific meaning that was not alcoholism, not compulsive gambling, not any of the other human frailties that might actually compromise a man's reliability. In 1956 the Directorate of Security and Intelligence created a unit to formalise the work, and here is where the story tips from merely appalling into something almost surreal.
The unit was called Character Weakness.
That was the name. On the paperwork. In the filing system. Character Weakness, a federal government subdivision, operational in Ottawa, staffed by RCMP agents whose job was to find people and put their names in folders.
Agents from Character Weakness watched bars and parks. They photographed people. They cultivated informants inside the civil service -- colleagues who, for whatever combination of fear or ambition or genuine ideological conviction, reported to the RCMP what they had seen or heard or suspected about the person at the next desk. Anyone named in a report got interrogated. The files accumulated. By 1962, more than 800 men and women had been identified across 33 federal departments and agencies as suspected or confirmed homosexuals. The count kept rising. By the late 1960s the RCMP had compiled files on approximately 9,000 people -- the majority of them living in Ottawa, the majority of them citizens who had done nothing except exist in a way the government had decided required documentation and eventual remedy.
Approximately 400 lost their jobs outright through firings and formal discharges. That is the documented floor, not the full accounting. It excludes the careers that ended through forced resignations, denied promotions, or the particular corrosive pressure of knowing you were being watched and eventually concluding you had to leave before the knock came. The actual toll on careers -- on lives -- was higher. The records do not quantify it, presumably because the people maintaining those records were not thinking about the people inside them.
In 1961, someone on the Security Panel had a thought. Surveillance was expensive. Informants were unreliable. What if there were a test?
Frank Robert Wake was the head of the psychology department at Carleton University in Ottawa. The government hired him and sent him to the United States to learn what the Americans were doing -- the FBI, the CIA, and the NSA each ran detection programmes of their own, and Canadian intelligence wanted to understand them. Wake returned with a proposal built partly on research coming out of the University of Chicago, where two psychologists named Eckhard Hess and James Polt had been studying pupil dilation as a measure of interest response. The premise was clean: the pupils expand involuntarily when a person sees something compelling. Wake proposed applying this to the problem at hand. Show federal employees images of increasing explicitness. Photograph their eyes throughout. The camera would tell you what the subject could not control.
The pitch, as proposals went, had a certain appeal to it. Objective. Scientific. Involuntary. The kind of methodology a Security Panel could endorse without embarrassment.
The apparatus was built by the RCMP's Identification Branch. The Defence Research Board, operating under the Department of National Defence, provided funding. The subject sat in the dentist's chair facing a screen. The slides cycled through: neutral images first, then increasingly explicit photographs of men and women, then a word-association exercise using contemporary homosexual slang. The camera on the pulley photographed the subject's pupils throughout. Perspiration sensors and a pulse monitor ran simultaneously, building what was intended to be a physiological portrait from which orientation could be read.
The cover story for subjects was that the device measured stress responses.
That cover lasted until word circulated through the federal workforce, which it did, as word does. After that, finding people willing to sit in the chair became essentially impossible. Men who understood what the chair was actually measuring declined to be measured. And even setting that aside: the methodology carried flaws that the funding bodies appear never to have raised. The brightness of individual slides varied, and pupil dilation responds to luminance as readily as to content. The camera's off-angle position -- it could not go directly in front without blocking the view -- reduced measurement accuracy. The differences being sought were sub-millimetre shifts in pupil diameter, beyond the reliable resolution of 1960s photographic equipment. The fundamental premise, that gay and straight subjects would produce categorically distinct physiological responses to the same visual material, had no validated scientific basis.
By the mid-1960s the project had produced no usable data, established no reliable distinctions between subjects, and identified nobody. The Defence Research Board cut funding in 1967. Wake filed his final project reports in May 1969. The machine was dismantled. No physical remnant of it survives. Wake himself died in 1993.
Carleton University's official position, released in 2016, is that the institution has no internal documents confirming Wake's engagement in this research and bears no responsibility for activities conducted outside the scope of his employment. A Carleton professor who has written extensively on Wake and the purge confirmed that no evidence exists of the university funding the work -- that money came from the federal government. The university has not apologised for Wake's role.
Now. Here is the thing I mentioned we would come back to.
The fruit machine never worked. It identified no one, produced no actionable findings, ended no careers through its own output. And the purge it was built to accelerate ran for thirty more years after the machine was abandoned. Understand what that means. The machine was an attempt to make existing work more efficient. When the attempt failed, the work continued by the methods that had been producing results all along.
Surveillance. Informants. Interrogation. Polygraph. When the RCMP determined enough evidence existed, suspects were brought in and questioned at length about their lives, their relationships, the names of people they knew. Some were coerced into providing those names. Admitting to homosexuality, when the interrogation moved in that direction, ended the career in the same room where the admission was made.
John Wendell Holmes, an assistant under-secretary at External Affairs, was interrogated at RCMP headquarters in November 1959. He resigned the following April. David Moffat Johnson left External Affairs in 1960; the circumstances of his departure have been examined in the scholarship and more recent academic work suggests the picture is not as clear as earlier accounts indicated. John Watkins -- a former Canadian ambassador to Moscow, a man who had given decades to the public service -- died of a heart attack on 12 October 1964 in Montreal, during an RCMP interrogation. The RCMP's presence in that room was concealed at the time.
Senior men. Experienced, credentialled, with long careers built on genuine public service. Also men who had been classified, on the basis of a policy its own architects acknowledged had no evidentiary foundation, as inherent security risks.
The Canadian Armed Forces maintained a formal exclusion under a military directive called CFAO 19-20, in force until 1992, when a serving captain took the military to Federal Court and won. The RCMP's security-clearance framework mirrored it through the same period. Same-sex sexual activity between consenting adults had been partially decriminalised when Bill C-150 came into force on 27 June 1969 -- the same year Wake was filing his last reports on the machine -- but the purge of public servants and military personnel ran for more than two decades after that. The legal class period for the eventual settlement ends 20 June 1996, which tells you people were still being formally affected on that date.
For roughly twenty years after the machine was abandoned, almost none of what I have described above was public knowledge. The memos, the files, the interrogation records, the documents in which the Security Panel's own authors acknowledged the absence of supporting evidence for their policy -- all of it sat in classified federal archives, quiet.
In April 1992, a Canadian Press Access to Information request produced a substantial release of RCMP and Security Panel records. The numbers that emerged -- 9,000 files, approximately 400 careers formally destroyed, a machine built by a government-funded university researcher to detect gay civil servants by measuring their pupils -- landed in Canadian newsrooms. Some considerable coverage followed. The Prime Minister condemned the historical practices.
Then, largely, it receded. No parliamentary inquiry. No compensation. No formal reckoning beyond the condemnation itself, which is not the same thing as a reckoning. And the 1992 release was incomplete in any case -- records had been withheld, others destroyed, the full inventory of those filing cabinets still not entirely surfaced.
Twenty-five years after that.
A class-action lawsuit filed in late 2016 -- brought by three former military members, each discharged because of their sexual orientation -- had placed the question on a formal legal footing. Advocates had worked the political track for years. And on 28 November 2017, the Prime Minister stood in the House of Commons and apologised.
He named the fruit machine explicitly: an absurd device, a failed technology funded with the intention of using it against Canadians. He described the broader campaign as nothing short of a witch-hunt. He acknowledged that the blackmail rationale -- the entire premise of that 1959 memo -- had never been supported by evidence, and that the government had known this. He said: we were wrong. We apologise. I am sorry. We are sorry.
The apology came with a $145-million settlement announced the same day. The Federal Court approved the Final Settlement Agreement on 22 June 2018. Bill C-66, the Expungement of Historically Unjust Convictions Act, received Royal Assent on 21 June 2018, creating a mechanism for people to have their criminal convictions for consensual same-sex activity erased from the record entirely.
The settlement divided into four parts: $110 million for individual compensation to eligible class members -- current and former federal public servants, Canadian Armed Forces members, and RCMP officers harmed between 1 December 1955 and 20 June 1996; $15 million for class-counsel legal fees; $5 million for external administration costs; and between $15 million and $25 million for reconciliation and memorialisation administered by the LGBT Purge Fund. Individual awards ranged from a minimum of $5,000 to a maximum of $150,000; most claimants received between $5,000 and $50,000, calibrated to documented harm. When the Federal Court discharged the administrator on 7 January 2022, the final count stood at 719 claimants: 629 military, 78 federal public servants, 12 RCMP.
The LGBT Purge Fund received $23,713,413 from the settlement. It is legally required to spend those funds on reconciliation and commemoration. It must wind down permanently by 30 June 2027. Among the things it is building is a national monument in Ottawa.
Here is what the machine's failure actually shows us.
The purge never needed it. The device was an attempt to make the identification work faster, cleaner, less dependent on witnesses and confessions and human fallibility. When that attempt failed, the identification work continued for thirty years by the methods that had been functioning all along. The machine was an ambition. The purge was the practice. And the practice required nothing from the machine except, perhaps, the argument that the work was scientific.
What the machine reveals is the logic underneath. Not a rogue operation. Not a handful of bad actors working without authorisation. A Cabinet-sanctioned, memo-documented, unit-named, budget-funded apparatus for removing a category of person from government service -- built on a premise its architects admitted had no evidentiary support, sustained through six decades and multiple governments, and never subjected to a single parliamentary debate because no single Order in Council ever launched it. It ran on directives and memoranda and the Security Panel's quiet authority, paperwork moving through proper channels toward outcomes that destroyed real lives, the whole apparatus so thoroughly embedded in the bureaucratic structure that it required no announcement and generated no visible rupture. It was just the way things worked.
Nine thousand files. Four hundred careers, at minimum. An untallied further toll in resignations, demotions, and lives abandoned before the knock came. An ambassador dead in an interrogation room. A class-action settlement that took sixty years to materialise, by which point many of the people it was intended for were already gone.
And a machine that never worked, sitting in its dentist's chair in a building nobody named, camera on a pulley, waiting for someone to sit down in it -- until the money ran out and they took it apart, and nobody thought what remained was worth keeping.
The machine is the emblem. The purge is the story. The emblem survives because it is easier to picture -- strange, specific, almost darkly funny in its failure -- than the thirty years of ordinary methodical damage that surrounded it. But the emblem points somewhere. Straight at the memo. Straight at the logic. Straight at the nine thousand files and what it meant to be inside them.
BEHIND THE STORY
The research question that anchored this piece was simple and turned out to be important: what, exactly, can be verified about the fruit machine? Popular accounts of it are broadly accurate and frequently wrong in the specifics, most commonly by conflating the machine -- a failed research project that never deployed operationally, that ended no career through its own output -- with the larger purge it was built to assist, which ran for decades and demonstrably did. The piece was written to preserve that distinction throughout.
Primary materials examined included declassified Security Panel memoranda made available through federal Access to Information requests, most substantially the April 1992 document release; official government apology transcripts and backgrounders published on federal websites; Federal Court settlement documents and their supporting financial disclosures; and published materials from the not-for-profit trust established under the settlement to administer reconciliation funds. Academic literature on the subject spans a peer-reviewed 1994 history journal article, a 1995 labour history journal article, and a 2010 monograph based directly on the Security Panel records. A 2025 peer-reviewed international relations journal article re-examined one of the key individual External Affairs cases and complicated the account given in earlier scholarship; the piece reflects that complication without resolving what that scholarship has not resolved.
The description of the device's scientific basis -- specifically the University of Chicago pupillometry work the lead researcher cited in his written proposal to the Security Panel -- rests on the 1995 academic article and a secondary scholarly blog written by a credentialled psychologist, both of which draw on the Security Panel documents rather than the documents directly. This is a limitation of the sourcing chain and is noted here accordingly.
Editorial decisions of note: the piece does not speculate about what occurred during the funded period between 1963 and 1967 because the accessible record does not support specifics. The 400-career figure is presented as a floor rather than a count because the documented research makes clear it excludes a range of harm. The 9,000 figure refers to files compiled, not jobs lost; any conflation of those two numbers misrepresents what is known. The circumstances surrounding one External Affairs departure remain contested in current scholarship and are written to reflect that uncertainty.
AI was used in the research and drafting process. All factual claims were cross-referenced against primary or peer-reviewed secondary sources before publication.