The Ones That Were Real

What this is

Seventeen programmes. Each was alleged before it was proven, denied while it ran, and eventually documented — by a committee report, a court order, a leaked file, a declassified memo. The chart above gives two dates for each one: when it operated, and when the record became public.

The gap between those dates runs to a median of 28 years. The longest is 49. Nine of the seventeen were still going when they surfaced.

Every row names a primary document rather than somebody's account of one, and every row closes with what that document does not establish. That last part is not throat-clearing. It is most of the reason to build the page at all.

The honest version of this list is less useful than the dishonest one

A list of vindicated conspiracy theories is the most-used move in this subject, and it almost always arrives doing one job: they called Tuskegee a conspiracy theory too. What you are meant to conclude is that a claim resembling a proven case gets some credit for the resemblance.

It doesn't, and the chart shows why. Seventeen is a numerator. The denominator — everything claimed over the same period that didn't pan out — is not seventeen, or seventy, or seven hundred. This archive alone holds more than 1,300 entries. A list of hits with no account of the misses tells you the category isn't empty. That's worth knowing. It is also the entire extent of what it's worth.

The second thing it won't do is get you from a conspiracy to the conspiracy. Nearly every case here is narrower than what was said about it at the time. MKULTRA was real, and what survives of it is an accounting file, because Richard Helms ordered the research records destroyed in January 1973. So it proves the programme existed and was funded. It does not tell you what the experiments found. Operation Northwoods is a proposal that was submitted and never carried out. The zinc cadmium sulfide tests over St Louis happened; the harm from them was never shown. And the one that gets overstated most reliably is the CIA's relationship with the press, where the Church Committee's finding of roughly fifty journalists is solid and the codename everybody attaches to it appears nowhere in the record.

A proven case is a proven case. It is not a licence to round up the next one.

The mechanism is the interesting part

Sort the chart by how each programme surfaced. The distribution is the finding.

Four came out through a congressional committee. Four through a reporter. Three through routine declassification, three through a leak, one through a FOIA request, one through a court. And one — COINTELPRO, the most consequential domestic programme on the list — because eight people broke into an FBI office in Media, Pennsylvania in March 1971 and posted the files to the press.

Sit with that, because it cuts against the frame these cases usually get enlisted to support. Every one of them became known through an institution. A Senate committee. A federal court. A newspaper. An archive with a declassification schedule. A statute that lets a citizen demand a file. The machinery that is supposed to be captured is the machinery that produced every document on this page.

The corollary is less comfortable. Not one of them surfaced because the agency responsible decided to say so. The CIA took until 2013 to state plainly that it had organised the 1953 coup in Iran: sixty years after the fact, and thirteen years after its own history of the operation had leaked. The NSA study concluding that the Gulf of Tonkin attack of 4 August 1964 never happened was written in 2001 by the agency's own historian and sat classified until 2005, and the thing that finally forced it out was a reporter drawing the parallel to Iraq.

Secrecy has a half-life and it is long

Disclosure doesn't arrive on a schedule, and it doesn't arrive early.

The ones that surfaced fastest had an outside witness. Iran-Contra broke in a Lebanese magazine five years in. The Business Plot came apart inside a year because the general it was pitched to went to Congress about it. The ones that ran longest were the ones where everybody involved shared an interest in saying nothing and no outsider was in the room: a hospital ward, a weapons range, a line in a research budget.

Tuskegee is the outlier that explains the others. It ran forty years and was never secret. It was published in medical journals throughout, discussed at conferences, read by physicians who did nothing about it. Concealment was never required. All that was required was that nobody with standing objected. That is a different failure from a cover-up, and on the evidence here it is the more durable one.

What would make this page better

Three things, none of them available.

A denominator. The comparison that would mean something is between the claims that proved out and the claims made alongside them that didn't, and nobody has assembled the second list because there's no principled way to bound it.

A base rate for whatever is running now. If the median programme here stayed quiet for 28 years, the arithmetic puts things begun in the late 1990s about due. That's a reason to keep reading documents. It is not a reason to believe any particular claim about what they'll say.

And the files that no longer exist. Helms gave the order seven months before MKULTRA formally ended. The twenty thousand pages that survived did so because somebody had filed them with the budget paperwork instead. Everything on this page is what happened to be kept.

Reading them yourself

Most of this is online and free. Committee reports are at govinfo.gov, and the Church Committee volumes are mirrored by the Assassination Archives and Research Center. The National Security Archive at George Washington University holds the Iran coup history and the Northwoods memorandum among much else, with the provenance of each release stated on the page. The declassified-document archive on this site indexes NARA, GovInfo and the CIA's CREST collection and matches them against the files here.

If you want something that isn't public yet, the guide to filing a FOIA request covers how, what it costs, and what to do when they say no. One of the seventeen above came out that way.