Conspiracy theories: what the record proves, and what it doesn't

A conspiracy is a legal fact: two or more people agreeing in secret to do something wrong, and prosecutors charge it daily. A conspiracy theory explains an event by a secret plot, offered before or instead of the evidence that would establish one. The difference is not the subject matter but whether the claim came out of the documents or arrived ahead of them.

You will see it repeated that the CIA coined conspiracy theory in 1967 to discredit Warren Commission critics. Dispatch 1035-960 is real and did coach stations on countering those critics, but the phrase appears decades earlier, including in Karl Popper's 1945 writing on the conspiracy theory of society. The document exists; the origin story attached to it does not.

The CIA approved MK-Ultra in April 1953 under Allen Dulles to research interrogation and behavior control, including LSD given to people who had not consented. Director Richard Helms ordered the files destroyed in 1973. It surfaced anyway: the Church Committee in 1975, then a 1977 FOIA request that turned up a misfiled cache of about 20,000 financial records.

From 1932 the US Public Health Service tracked 399 Black men with syphilis and 201 without it in Macon County, Alabama, telling them they were being treated for bad blood. Penicillin became standard around 1947 and was deliberately withheld. Employee Peter Buxtun went to the press after internal complaints went nowhere; the Associated Press story ran on 25 July 1972 and the study ended within months.

In November 1975 the intelligence chiefs of Argentina, Bolivia, Chile, Paraguay and Uruguay met in Santiago and agreed to hunt each other's exiles across borders; Brazil joined later. On 22 December 1992 lawyer Martin Almada and judge Jose Agustin Fernandez found roughly 700,000 pages in a police station in Lambare, Paraguay. Those files became evidence in trials in Argentina, Chile and Italy, including the 2016 Buenos Aires verdict convicting former officers of running the network. The secret held about 17 years, from that meeting to that police station.

COINTELPRO was one FBI division targeting a defined list of groups; Condor coordinated a handful of intelligence services against named exiles. Neither required silence from hundreds of thousands of people across rival countries and generations. The secrets that held were narrow, bureaucratic and dull, and they still leaked within 10 to 20 years.

Not one was cracked by someone spotting an oddity in a photograph. Tuskegee ended because an employee walked files to a reporter, COINTELPRO because activists stole documents from a field office, Condor because a judge opened a cabinet in Paraguay.

The hardest problem for the hoax claim is not the flag or the shadows, it is the retroreflectors. Apollo 11, 14 and 15 left mirror arrays that observatories including the Cote d'Azur and Apache Point still bounce lasers off to measure the Moon's distance to within centimeters. Add 382 kilograms of returned samples distributed to labs worldwide, and Lunar Reconnaissance Orbiter photographs from 2009 to 2011 showing descent stages and footpaths.

Contrails form when hot, humid engine exhaust meets air below roughly minus 40 degrees Celsius. Whether they vanish in seconds or spread into cirrus for hours depends on the humidity of that layer, which is why two aircraft at different altitudes leave different trails. In 2016 a survey in Environmental Research Letters asked 77 atmospheric chemists and geochemists to review the evidence proponents offered; 76 reported seeing none that required a secret program.

Building 7 fell at 5:20 pm on 11 September 2001 without being hit by a plane, the strongest surface case for demolition. NIST's 2008 report attributes it to fires burning nearly seven hours, unfought because the water mains were severed, causing thermal expansion that failed a critical column connection. The counterclaim requires days of access to place charges in an occupied federal building, explosives surviving hours of fire, and no crew member ever talking.

This is a documentary story, not a medical argument. Andrew Wakefield's 1998 Lancet paper described 12 children; the journal retracted it on 2 February 2010 and Britain's General Medical Council struck him off the register on 24 May 2010 after finding dishonesty. Brian Deer's BMJ investigation of January 2011 documented that the paper's data did not match the children's hospital records, and that Wakefield had been paid through a lawyer preparing litigation against manufacturers.

Psychologists call it proportionality bias. In a 2007 study, Patrick Leman and Marco Cinnirella gave people the same fictional assassination account with one detail changed: in one version the president died, in the other he survived. Readers of the fatal version were markedly more likely to suspect a wider plot. Same evidence, different explanation demanded.

Jennifer Whitson and Adam Galinsky reported in Science in 2008 that people primed to feel a loss of control saw more images in visual noise and perceived more conspiracies in ambiguous stories. Mistaking a shadow for a predator costs less than the reverse, so the system is tuned to overfire. Job loss, illness and upheaval are when it overfires most.

Believing what the mainstream rejects can supply what a job or a congregation used to: a group, a role, the sense of being one of the few who see clearly. Belief tracks distrust in institutions far more tightly than education, and communities with a documented history of being lied to by health authorities or police score higher on suspicion. That is not irrationality. It is memory.

No subscription or press pass required, and all of these hold primary material rather than commentary about it.

The Warren Commission concluded in 1964 that Oswald acted alone. The House Select Committee on Assassinations concluded in 1979 that a conspiracy was probable, largely on acoustic analysis of a police dictabelt recording the National Academy of Sciences later disputed. Two official findings, in direct disagreement, one resting on contested audio.

The Pentagon's All-domain Anomaly Resolution Office published a historical review in March 2024 reporting no evidence that any US program recovered extraterrestrial technology. The same report described something stranger: officials using cover stories about alien technology to protect classified aircraft programs, meaning some rumors were seeded by the government for unrelated reasons.

Diplomats and intelligence officers in Havana, Guangzhou and elsewhere reported sudden severe symptoms from 2016. A 2023 US intelligence community assessment judged it very unlikely that a foreign adversary caused them, and NIH studies published in 2024 found no distinct brain imaging signature versus controls. Affected personnel and some members of Congress dispute both. The symptoms were documented; the cause is not settled.

In Guatemala, the belief that Washington organized the 1954 overthrow of Jacobo Arbenz was dismissed for years and is now confirmed by declassified files on PBSUCCESS. A population demonstrably lied to about one thing discounts the next official statement, which is a rational response to a track record even when it produces wrong answers in specific cases.

Recommendation systems optimize for watch time and shares, and outrage produces those signals better than a careful correction does. That is a design consequence rather than an editorial plot, and it means a claim reaches millions before anyone opens the underlying document.

Much high-volume conspiracy content is monetized through supplements, survival gear and paid livestreams. Alex Jones sold nutritional products alongside the Sandy Hook claims that produced defamation judgments near $1.44 billion in Connecticut and $49 million in Texas in 2022. Money does not make a claim false, but a claim that reliably sells something has a second reason to exist.

Sander van der Linden and Jon Roozenbeek at Cambridge tested inoculation: showing people manipulation techniques before they meet them. Their game Bad News, where players practice building fake news, measurably lowered how credible participants later rated misleading content, and a 2022 collaboration with Google Jigsaw replicated the effect using short videos run as YouTube ads. Separately, Thomas Wood and Ethan Porter tested corrections across 52 contentious issues with over 10,000 participants and found no backfire in the overwhelming majority.

Four of the six below point at something that genuinely happened at least once. That is why the other two travel so easily: they borrow their neighbors' credibility.