What did not survive the check

The most famous experiment against psychiatric diagnosis

David Rosenhan's 1973 paper in Science is the most-cited evidence against psychiatric diagnosis. It cannot be relied on as evidence — though it has never been retracted.

Documents located in Rosenhan's private papers by the journalist Susannah Cahalan, and reported in the peer-reviewed literature by the sociologist Andrew Scull, establish that: Rosenhan's own admission record at Haverford State Hospital in February 1969 contains far more symptom reporting than the three words he published — hallucinations lasting months, thought insertion and suicidal thoughts; and that a ninth participant, Harry Lando, whose experience contradicted the published findings, was excluded.

And the precision we owe: "documented serious doubts" is not the same as "proven fraudulent". Cahalan's book (2019) is a secondary source, a trade non-fiction work. Robert Spitzer had already made a methodological critique in 1975.

Why it is here and not in the main narrative: because the same chain of evidence — journalistic research in private papers — is rejected elsewhere on this page. A standard that changes with the conclusion is not a standard.

[54] Rosenhan DL, Science 1973;179(4070):250-258 · [55] Cahalan S, The Great Pretender, 2019 — secondary · [56] Scull A, History of Psychiatry · [57] Spitzer RL, 1975

And the rest, with the reason each failed

"The pharmaceutical industry created the DSM." Ties are documented, with numbers per panel. Causation is not. DSM-III grew out of the academically originated Feighner criteria of 1972.

"A psychiatric seeding trial is documented." The canonical documented seeding trial is gabapentin's STEPS — an epilepsy trial, not a psychiatric one. The general claim that "psychiatry is full of seeding trials" does not verify and is not published.

"A court found Study 329 fraudulent." It did not. There is a 2026 Expression of Concern, which is a holding notice.

"X% of the psychiatric literature is ghostwritten." No study measures this. Healy and Cattell's "23 of 85 possibly originated" is not a measurement.

"Hundreds of thousands of victims of Soviet punitive psychiatry." The documented cases are orders of magnitude fewer. The large numbers have no primary source we could locate.

"Greece has the highest involuntary admission rate in Europe." Refuted by the very paper cited for it: below the median of 22 countries. See Part C.

"Freeman performed 3,439 lobotomies; his youngest patient was four." The first comes from a trade biography drawing on Freeman's own records; the second found no primary source.

And two that concern this page itself.

The peak number of residents at Leros. No single-date census was found in any document. Other figures do exist, and each is a different kind of thing: statutory bed establishment from the government gazette (2,650 in 1965, 2,750 in 1981), an approximation by the authors of the 2009 account (around 2,400 by the late 1960s), and cumulative register entries (5,070 by 31 July 2009). None of these is "the peak number", and we do not present any of them as one.

A correction of our own. We had written here that the figures 4,000 and 2,569 are not independent of each other. Having obtained the full 2009 account, that does not hold: 2,569 is the sum of 33 documented group transfers between 26 July 1964 and 15 February 1975, while the 4,000 in the 2016 paper describes referrals over a longer period, up to 1982, and cites a peak year of "900 in 1965" that does not appear in the 2009 account at all (whose largest single transfer is 358 people). They measure different things over different periods. Where the 2016 paper got its own figure remains unknown — its reference list was not obtained.

43 claims checked: 16 confirmed, 17 partial, 8 rejected, 2 unverifiable

From the newer material: one number that is not published

"16,500 sterilisations without consent in Japan." The number of sterilisations performed without consent under the Eugenic Protection Law is given variously as 16,000 and 16,500, in sources that all trace back to the same ministry data. The ministry's own release was not located, so which figure is correct — and whether the difference is rounding or a different definition — could not be determined. The total of 25,000 holds; its subdivision does not, and it is therefore not used in Part J.

That the chemical restraint of children in US psychiatric facilities rose by 141%

It circulates on campaign websites as a finding of the American Academy of Pediatrics. The paper exists (Pediatrics 2024) and the number is in its own abstract — but it counts days with restraint, which rose 141% while patient days rose 138%. The overall rate per patient day did not change significantly (RR 1.2; p=0.47); it rose in four diagnoses and fell in two. And “in psychiatric facilities” is the wrong setting: the study excluded psychiatric units and measured paediatric wards. The number is true; the claim carrying it is not. What stands from the same paper — which children are restrained — is in Part B.

[74] Masserano B et al. Pediatrics 2024;153(1):e2023062784