# The Man in the Iron Mask

Report: https://historical-mysteries.netlify.app/iron-mask/report (a copy is report.html in this folder; its source parts and stylesheet are in report_src/).

## The mystery

On 18 September 1698 the new governor of the Bastille, the royal fortress-prison in Paris, arrived with a prisoner whose face was covered by a mask. The governor, Bénigne de Saint-Mars, had guarded state prisoners for the crown since 1665, in the Alps and then on an island off Cannes. Étienne du Junca, the King's lieutenant at the Bastille, who kept its entry book, wrote that the man's name "is not said". The prisoner died there on 19 November 1703, "always masked with a mask of black velvet", and was buried the next day under a false name, Marchioly, with his age given as forty-five or thereabouts. No paper of the Bastille says who he was.

## Why it is still argued

The writer Voltaire made the mask iron in 1751 and hinted in 1771 at a brother of Louis XIV; Alexandre Dumas made the prisoner the King's twin. Some fifty identifications have been offered since 1745 (the historian Jean-Christian Petitfils counts fifty-two). The evidence is the correspondence between Saint-Mars and the King's war minister from 1669 to 1703, kept while Saint-Mars held his prisoners at Pignerol, a French fortress in the Alps (now Pinerolo, in Italy), at Exilles, a smaller fort nearby, and on the island of Sainte-Marguerite. The letters, first printed in 1801, say where each prisoner was moved and when, never say why this one was held, and are missing at the one point where a name is needed: in January 1687 one of two unnamed prisoners died at Exilles, and the letter that said which is lost.

Most historians who have worked from the letters since Jules Lair, Fouquet's biographer, in 1890 (Georges Mongrédien 1952, who narrowed the choice to two men and would not decide; Petitfils 1970 and 2003; the writer John Noone 1988; the American historian Paul Sonnino 2016; the biographer Josephine Wilkinson 2021) hold that the masked man was a prisoner the war minister called "only a valet", a manservant, arrested in July 1669 and delivered to Pignerol that August. The nineteenth century's answer, from Pierre Roux-Fazillac (1801) to Frantz Funck-Brentano (1894), was Ercole Antonio Mattioli, a diplomat from Mantua seized by the French in 1679, whose name the burial register nearly reproduces. The popular answer, from Voltaire to Marcel Pagnol, is a brother or twin of the King, for whom no document exists. Those who accept the valet still disagree on who he had been (Petitfils: a servant of Henrietta of England, Louis XIV's sister-in-law, during her secret negotiation of 1669 to 1670 with her brother Charles II; Sonnino: a valet of the man who managed Cardinal Mazarin's fortune) and on why he was kept for thirty-four years.

## How the case was judged

Every letter and register entry of the correspondence printed between 1801 and 1953 was read, 180 entries in all, with the variants between editions recorded. Seven dated facts about the man's custody were taken from them, each resting on a named letter or register entry (arrest 1669, service to Fouquet 1675, confinement with Fouquet's other servant 1680, move to Exilles 1681, death of one of the two 1687, entry to the Bastille 1698, death 1703), and twenty candidates were tested against the seven. Only the valet of 1669 passes every documented point. The undocumented point, which of the two Exilles prisoners died in 1687, is argued from two later letters and the subject matter of the whole file, and given a probability. Three links in the chain of custody, weighted 0.99, 0.93 and 0.85, multiply to the headline below.

## Verdict

Evidence label: **Probable**. The evidence clearly favours one answer over all the others, but does not prove it. The percentage moves when the same evidence is weighed differently.

**Who died in the Bastille on 19 November 1703.** The valet whom Captain de Vauroy, an officer of the Dunkirk garrison, delivered to Pignerol in August 1669, entered in the ministers' letters as Eustache Dauger, Danger or d'Angers: **78%**. La Rivière, Fouquet's other valet and the only other man who could have survived the pair held at Exilles in 1687: 14%. Count Ercole Antonio Mattioli, whose name the burial register echoes: 3%. He is excluded by the custody chain of 1681 to 1698 and, on the reading every editor since Lair has given it, by the letter of 10 May 1694 from Barbezieux, the war minister who succeeded his father Louvois in 1691, on "the valet of the prisoner who has died". Someone else: 5%.

The 78% is the product of three judgements, each argued from named letters: that the Bastille prisoner of 1698 was the man Saint-Mars brought from Sainte-Marguerite (put at 99%); that the island prisoner was the man brought from Exilles in 1687 (93%, against the thesis of Stanislas Brugnon, an archivist who reconstructed the island's prison register, that the island prisoner died in 1693); and that the survivor of Exilles was the 1669 valet rather than La Rivière (85%, on Barbezieux's "depuis vingt ans" of 13 August 1691, his "ce qu'a fait vostre ancien prisonnier" of 17 November 1697, and the subject matter of the whole file). Change one weight and the headline moves: read the 1691 "twenty years" as a loose round number (60%) and the valet falls to about 55% with La Rivière at about 37%; give Brugnon even odds and the valet falls to about 42%. All of these figures are conditional on the weights chosen. They are judgements about the documents rather than measurements. The script code/probability.py prints the headline and the sensitivity table (data/probability_sensitivity.md).

**Who the valet was before July 1669.** Nobody knows. Petitfils makes him a servant of Madame, as Henrietta of England was known at court, carrying letters in the negotiation that produced the secret treaty of Dover in 1670 (about 25%); Sonnino makes him a valet of Antoine-Hercule Picon, who managed Mazarin's fortune (about 20%); an employment the surviving papers do not point to at all takes the rest (about 55%). Both hypotheses are inferences from the same silence, and both authors say as much in their own terms. Petitfils writes that the prisoner's life in prison "nous prouve que son affaire, quelle qu'elle ait pu être, n'était pas vraiment de la plus haute importance"; Sonnino writes that he found no Auger or Dauger among the Picons' servants. These three figures are judgements about unwritten history.

**Why he was held for thirty-four years.** The letters say the secret was something he had been employed at or had done before 1669 ("ce à quoi il a été employé auparavant que d'être à Pignerol", 1678; "ce qu'a fait vostre ancien prisonnier", 1697), and that every order had one object: that he never speak of it, above all to Lauzun, a courtier held beside Fouquet at Pignerol from 1671 to 1681 whose release was expected. No minister ordered a mask. The mask was Saint-Mars's practice from 1687, and in August 1698 the King approved, in general terms, whatever precautions Saint-Mars had proposed for the journey to Paris. The thirty-four years are the ordinary working of Louis XIV's state prisons, which had no procedure for releasing a man held for what he knew and in this file used none, under a minister (Barbezieux) who inherited the precautions and a jailer whose posts and pay depended on continuing to hold him.

This is a reading of the documents with a stated probability. The mystery is not solved: which of the two Exilles prisoners died in January 1687 is decided by inference, and no letter says.

## Where this verdict stands among historians

**Agrees with** the position most historians who have worked from the letters since 1890 have held, which is that the masked man was this valet. The question is still argued. Jules Lair (1890) first identified the masked man with the valet of 1669. Georges Mongrédien (1952) eliminated every Pignerol prisoner but Mattioli and the valet, showed that La Rivière died at Exilles in 1687, and declined to choose ("Si oui, c'est Dauger, si non, c'est Matthioli", p. 231). Petitfils (1970, 2003), John Noone (1988), Sonnino (2016) and Josephine Wilkinson (2021) decide for the valet. Against or aside: Funck-Brentano (1894) held for Mattioli, the writer Harry Thompson (1987) for Eustache Dauger de Cavoye, a disgraced nobleman who bore the valet's name, the writer Roger Macdonald (2005) for d'Artagnan, and Sarah Madry, in an article of 2016, argues that the word "valet" in the 1669 letter cannot be relied on. The report reaches the valet at a lower confidence than his proponents state. The Calais arrest is Brugnon's finding (1987); the reading of the name as Danger or d'Angers is Lair's and Bernard Caire's, who examined the original letters in the 1980s; the deleted words "ce qu'il a vu" in the ministry's draft of 23 December 1678 are Caire's find; the reading of the mask as the jailer's device is Noone's and Petitfils's.

## What is new

Nothing here is new evidence. No document unknown to the printed editions was found; everything below is a new reading of old evidence.

1. The whole printed record from 1669 to 1703 (Roux-Fazillac, Delort, Ravaisson, Iung, Topin, Loiseleur, Lair, Funck-Brentano, Laloy, Mongrédien's 1953 article) is transcribed into 180 dated entries with each edition's archive reference, and the variants between editions are tabulated (notes/A, notes/B), so the chain of custody can be checked line by line.
2. The weak link is treated as a weak link. The survivor of Exilles carries a probability and a published sensitivity instead of being stated as a fact, and twenty candidates are tested against seven fixed points by a script (code/fixed_points.py), so anyone who disagrees with a cell can change the file and rerun it.
3. Three small readings not found in the works consulted (several of which could be read only in fragments): the ministers' unbroken use of "ancien prisonnier" from January 1696 to the orders of 1698 as the answer to Brugnon's thesis of a death in 1693; the letter of 12 March 1680 on a dead valet's wages read as Champagne's wage arrears rather than as the death of Eustache (against Jean-Marc Dabos's thesis that the masked man was Fouquet), an inference from the surrounding letters and not a documented exchange; and the variant "ce qu'a fait" (Topin) against "ce que fait" (Iung) in the letter of 17 November 1697, which the original (Service historique de la Défense, A1 vol. 1392, p. 171) would settle.

## What was not read

- Petitfils 2003 and Sonnino 2016, the two books that defend the current identifications, were not read in full: about 1,050 and 1,000 search-service fragments of the lending copies, plus their reviews (Genet, Venturino, Ruff, Parrott) and interviews in full. Mongrédien 1952 was read in 3,439 page-numbered Gallica snippets plus his 1953 article in full. Noone, Wilkinson, Thompson and Macdonald were read in fragments only; Petitfils 1970 through four contemporary reviews.
- Not obtainable: the Cannes 1987 proceedings (Brugnon, Caire), the Pinerolo 1974 proceedings, Brugnon's 1992 reconstruction of the island register (a typescript at the Arsenal), Caire's articles of 1988, 1997 and 1998, Sonnino's 1992 article on Claude Imbert, the reviews of Sonnino in the Journal of Modern History, the Catholic Historical Review and the American Historical Review, Duvivier 1932, and Bazeries and Burgaud 1893 (known through Batiffol's review).
- No original manuscript was consulted. Six letters were checked against the photographs on masque-de-fer.com. Vauroy's reimbursement order (Mélanges Colbert 282, fol. 276 v°), the minute of 23 December 1678 with its deleted words, and the Saint-Lazare order of 17 August 1678 are known only through the authors who report them.
- The Bibliothèque nationale's Gallica site refused automated access throughout; Google Books, HathiTrust, Cairn, JSTOR and several university presses likewise.

## Files

- report.html: the report (self-contained HTML). report_src/: its four source parts and stylesheet.
- data/fixed_points.json: the seven fixed points (a) to (g) with their documents. data/candidates.json: twenty candidates scored against them. data/fixed_points_table.md: the generated table. data/custody_chronology.csv: who was where, 1669 to 1703. data/key_documents.csv: 46 load-bearing letters and register entries with edition, page and archive reference. data/probability_sensitivity.md: the headline arithmetic.
- code/fixed_points.py regenerates the candidate table (`python3 code/fixed_points.py md`). code/probability.py prints the headline and its sensitivity (`python3 code/probability.py`).
- notes/A_pignerol_1669-1681.md (100 entries, roster, 44 discrepancies); notes/B_exilles_marguerite_bastille_1681-1703.md; notes/C_candidates.md; notes/D_historiography_and_secrecy.md (the legend 1687 to 1790; scholarship 1801 to 2021; the secrecy theses tested against 26 letters); notes/E_1669_context_and_dauger_identity.md; notes/F_modern_works_access.md (Petitfils, Sonnino, Mongrédien, Caire, Brugnon, Noone, Wilkinson, Thompson, Macdonald: what text could be obtained, how, and the authors' theses in their own words); notes/G_open_access_articles.md; notes/H_masque-de-fer.com_transcriptions.md.
- sources/: OCR texts, fragment harvests, reviews, and the facsimile images from masque-de-fer.com.