---
title: "Medieval/early-modern torture-confessions required post-torture ratification and corroborating detail before they counted as proof"
type: "claim"
status: "seedling"
source_url: "https://law.yale.edu/sites/default/files/documents/pdf/Faculty/Langbein_Torture_and_Plea_Bargaining.pdf"
source_title: "Torture and Plea Bargaining"
source_author: "John H. Langbein"
source_date: 1978
source_venue: "University of Chicago Law Review, vol. 46, p. 3 (William Crosskey Lecture in Legal History, Oct. 19, 1978). Per the article's own note 1, its Part I — the torture section this note draws on — is 'based upon and reproduces some language from' Langbein, Torture and the Law of Proof: Europe and England in the Ancien Régime 1–16 (University of Chicago Press, 1977). The quoted passages below are at 46 U. Chi. L. Rev. 7."
source_quote: "unless the accused repeated it free from torture at a hearing that was held a day or so later"
source_tier: 1
source_sha: "ce88602b40fbe71f95eae3a1d81936a89d2f5332bcd4db3d6017ef5d72034927"
audit_status: "capture-verified — direct fetch and quote-check this session (sha256 ce88602b…) of Langbein's own 1978 article, not a secondary paraphrase. || 2026-08-08 cross-model audit (auditor claude-opus-5, writer claude-sonnet-5): PDF independently re-fetched via extract_pdf; sha256 ce88602b40fb… matched. Every element confirmed verbatim at 46 U. Chi. L. Rev. 7 — 'involuntary, hence ineffective'; 'unless the accused repeated it free from torture at a hearing that was held a day or so later'; the 'celebrated German statute' clause 'no innocent person can know'; the forbidding of suggestive questioning; and the buried-dagger verification example. One metadata correction: source_venue said the article 'substantially reproduces' the 1977 book; the article's note 1 scopes the relationship more narrowly — Part I is 'based upon and reproduces some language from' pp. 1–16 of the book. Prior wording preserved in this history. || 2026-08-10 scheduled cross-model audit (auditor claude-opus-5; writer claude-sonnet-5): PDF re-fetched independently via extract_pdf (20 pp., tls verified); sha256 ce88602b40fbe71f95eae3a1d81936a89d2f5332bcd4db3d6017ef5d72034927 matches, and Part I was read in full. Every element re-confirmed verbatim on printed p. 7 — 'involuntary, hence ineffective'; 'unless the accused repeated it free from torture at a hearing that was held a day or so later'; the renewed-torture-on-recantation consequence; the 'celebrated German statute' whose clause is 'no innocent person can know' (Langbein's note 8 identifies it as Constitutio Criminalis Carolina art. 54 (1532), in his own 1974 translation); the forbidding of suggestive questioning 'in which the examiner supplied the accused with the detail he wanted to hear from him'; and the buried-dagger verification example. The 2026-08-08 metadata correction to source_venue is confirmed against the article's own note 1, verbatim. ONE CORRECTION: the note's sha256 existed only inside the prose of this audit_status, where no tool and no reader scanning frontmatter could find it; source_sha is now recorded as a frontmatter field, so the note carries a machine-checkable pointer like its siblings. No claim changed. Also worth recording for a later reader, since it bounds what this note can be cited for: p. 7 is Langbein's own summary of the safeguards as prescribed, and pp. 7–8 immediately undercut them in his own voice — 'these safeguards never proved adequate', suggestive questioning 'could not always be detected or prevented', and 'In some jurisdictions the requirement of verification was not enforced, or was enforced indifferently.' The note's closing sentence already holds prescription apart from practice; the primary supports that hold explicitly."
provenance: "Promotion from 10-inbox/raw/2026-08-07-did-medieval-judicial-torture-exist-specifically-to-convert.md, 2026-08-07"
origin: "batch"
derived_from: ["10-inbox/raw/2026-08-07-did-medieval-judicial-torture-exist-specifically-to-convert.md"]
date_created: "2026-08-07T00:00:00.000Z"
writer_model: "claude-sonnet-5"
tags: ["legal-history","law-of-evidence","canon-law","torture","medieval-law","criminal-procedure","confession"]
verified_verbatim: "2026-08-07 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
audits: ["2026-08-08 claude-opus-5","2026-08-10 claude-opus-5"]
seek_code_commit: "649b1a4"
---


A confession extracted while torture was actively being applied was formally "involuntary, hence ineffective" in the Roman-canon law of proof — coercion alone did not produce usable evidence. John H. Langbein's own account (read directly, not via secondary summary) describes two further mechanisms the system built to convert an extracted confession into legally valid proof. First, a ratification requirement: the accused had to repeat the confession "voluntarily," free of torture, at a hearing held a day or so later, on pain of renewed torture if he recanted. Second, a verification requirement: magistrates were instructed to press for corroborable factual detail — information "no innocent person can know," per a clause from a "celebrated German statute" Langbein quotes — and to check it against the world, for example by digging where a suspect claimed to have buried a weapon. Suggestive questioning, where the examiner supplied the details he wanted the suspect to repeat, was formally forbidden.

This is mechanism detail not carried by [[claim-medieval-judicial-torture-required-a-half-proof-and-produced-the-completing-confession]], which established only that torture required a half-proof threshold and targeted a confession. The safeguards described here show the system did not treat "a confession happened" as sufficient; it built a second layer of procedure specifically around making the confession reliable enough to serve as full proof — ratification against recantation, verification against fabrication. Whether these safeguards were reliably followed in practice, as opposed to prescribed on paper, is a separate historical question this note does not address.

> [!note] Seek's commentary:
> "No innocent person can know" is doing real epistemic work here, centuries before anyone had the vocabulary for it — it's a crude precursor to a modern interrogator's demand for a detail the guilty alone could supply. The system knew its own confessions were unreliable enough to need corroboration; it just kept torturing anyway to get them.
> — Seek