Historical mysteries · Read from the letters, 1669 to 1703

Who Was the Man in the Iron Mask?

The letters that passed between the prisoner's jailer and the King's ministers from 1669 to 1703, the Bastille's own registers and the burial entry, read in full and tested against twenty proposed candidates.

The mystery

On Thursday 18 September 1698, at three in the afternoon, a litter came through the gate of the Bastille, the royal fortress-prison in Paris. In it sat the prison's new governor, Bénigne de Saint-Mars, who had guarded state prisoners for the French crown since 1665, first in the Alps and then on an island off the coast of Provence, and beside him a prisoner whose face was covered by a mask. Étienne du Junca, the King's lieutenant at the Bastille, who kept the book in which every arrival was entered, wrote that Saint-Mars had brought "an old prisoner he had at Pignerol, whom he keeps always masked, whose name is not said". The man waited in a tower room until dark. At nine that night du Junca himself led him to the third chamber of another tower, furnished some days before on Saint-Mars's orders, where a sergeant who had come with the governor would bring his food.

He was there for five years. On Sunday 18 November 1703 he felt unwell coming out of mass. He died the next evening at about ten, du Junca wrote, "without having had a great illness", still "the unknown prisoner, always masked with a mask of black velvet". He was buried the following afternoon in the cemetery of Saint-Paul, the prison's parish. The burial register gives a name du Junca called "as unknown" as the man, Marchioly, and an age, "forty-five years or thereabouts", signed for by two officers of the Bastille who had never seen his face. Forty livres were paid for the burial, and no paper of the Bastille gives his name. The one letter about him from the minister in charge of the prison calls him "your prisoner from Provence".

Why it is still argued

The state never said who the man was or why he was held, and writers supplied answers. In 1751 the writer Voltaire described the mask as having a chin-piece with steel springs, and later accounts called it iron; in 1771 he let it be understood that the prisoner had been an elder brother of Louis XIV. Alexandre Dumas made him the King's twin in a novel of the 1840s. Since 1745 some fifty identifications have been offered, and the historian Jean-Christian Petitfils counts fifty-two: a royal brother, a diplomat from Mantua, the Duke of Monmouth (Charles II's illegitimate son), the playwright Molière, the musketeer d'Artagnan, the disgraced finance minister Nicolas Fouquet.

The evidence comes from the years before the Bastille, and it is incomplete at one point. Saint-Mars had held the man since 1669: at Pignerol, a French fortress in the Alps (now Pinerolo, in Italy) where he guarded state prisoners until 1681; at Exilles, a smaller fort in the same mountains, until 1687; and on the island of Sainte-Marguerite, off Cannes, until 1698. All that time he wrote about his prisoners to Louvois, the King's war minister in Paris, who had charge of the state prisons; the minister wrote back, and hundreds of the letters were kept. They were first printed in 1801 by Pierre Roux-Fazillac, a former member of the revolutionary Convention, and every serious argument since has been made from them. The letters say where each prisoner was moved and when, but never why this one was held, and they fail at the one point where a name is needed: in January 1687 one of two unnamed prisoners died at Exilles, the other went on to the island and then to the Bastille, and the letter that said which of the two had died is lost.

Three answers have been given. Most historians who have worked from the letters since 1890 hold that the masked man was a prisoner whom Louvois called "only a valet", a manservant, arrested in July 1669 on the Channel coast and delivered to Pignerol that August under orders that no one should see him or hear him speak. Jules Lair, the biographer of Fouquet, argued this in 1890. Georges Mongrédien, a historian of the seventeenth century, narrowed the choice in 1952 to two men, this valet or the diplomat from Mantua, and would not decide between them. Petitfils (1970 and 2003), the writer John Noone (1988), the American historian Paul Sonnino (2016) and the biographer Josephine Wilkinson (2021) decide for the valet. The nineteenth century's answer, given by Roux-Fazillac and defended last in 1894 by Frantz Funck-Brentano, who catalogued the Bastille's papers, was that diplomat, Ercole Antonio Mattioli, whom the French had seized in 1679 for double-dealing over the sale of a fortress; the name on the burial register is three letters from the spelling Saint-Mars used for his. The popular answer, from Voltaire to the film-maker Marcel Pagnol, is a brother or twin of the King, for whom no document exists. Others have proposed a disgraced nobleman who bore the valet's name (Harry Thompson, 1987), d'Artagnan (Roger Macdonald, 2005) and Fouquet himself, his recorded death in 1680 being a fiction (Jean-Marc Dabos, 2025).

Those who accept the valet are divided by two further questions, which the letters do not answer. The first is who he had been. Petitfils makes him a servant of Henrietta of England, Louis XIV's sister-in-law, known at court as Madame, who in 1669 and 1670 carried the King's secret negotiation with her brother Charles II of England. Sonnino makes him a valet of the man who managed the fortune of Cardinal Mazarin, the King's chief minister until his death in 1661. The second is why he was kept for thirty-four years, and masked. Either he knew something that stayed dangerous, or, as Petitfils and Bernard Caire, who examined the original letters in the 1980s, have both argued, the secret was of small importance and the appearance of mystery was Saint-Mars's own doing.

How the case was judged here

This report takes the letters as the evidence and reads all of them. Every letter and register entry of the Pignerol, Exilles, Sainte-Marguerite and Bastille correspondence printed between 1801 and 1953 was read, one hundred and eighty entries in all, and where the editions disagree on a word both readings are recorded. From them seven dated facts about the man's custody were taken, each resting on a named letter or register entry: the arrest and delivery of 1669, the permission of 1675 to serve Fouquet, the confinement with Fouquet's other servant in 1680, the move to Exilles in 1681, the death of one of the two in 1687, the entry into the Bastille in 1698 and the death of 1703. Twenty candidates were tested against the seven. A candidate documented alive and free, in another prison, or dead at any of those dates fails.

Only the valet of 1669 passes every documented point, and one point is not fully documented: no letter says which of the two Exilles prisoners died in January 1687. The other was La Rivière, who had come to Pignerol with Fouquet as a free servant in 1665. From two later letters and the subject matter of the whole file this report argues that the survivor was the valet of 1669, and puts that at 85 per cent. Two other steps in the chain of custody carry their own weights, 99 and 93 per cent. Multiplied, they give 78 per cent that the man buried as Marchioly was the valet arrested in 1669, 14 per cent that he was La Rivière, 3 per cent Mattioli, and 5 per cent someone the letters never name. The weights are judgements about the documents rather than measurements, and the report shows how the headline moves when they change. On the two questions the letters do not answer, the report reaches no name for the valet's master, and finds that no minister ever ordered a mask: the mask was the jailer's practice from 1687, approved in general terms by the King in 1698, and the length of the detention follows the practice of the King's prisons, which had no procedure for releasing a man held for what he knew and in this file used none.

What this report covers

The panel below states the three answers with their probabilities. The report then sets out the method and the seven fixed points; follows the letters in order through the arrest of 1669, the eleven years at Pignerol, the years at Exilles, the missing link of January 1687, the island and the Bastille; tests the twenty candidates in a table; takes up who the valet was and why he was held; and closes with the strongest cases against the verdict, what would overturn it, what in it is new and what comes from earlier historians, what could not be read, and the sources.

Verdict

1. The prisoner who died in the Bastille on 19 November 1703

• 78%The valet delivered to Pignerol in August 1669 by Captain de Vauroy, an officer of the Dunkirk garrison, and entered in the ministers' letters as Eustache Dauger, Danger or d'Angers. His true name and his master are not in the record.
• 14%La Rivière, Nicolas Fouquet's valet, the only other man who could have been the survivor of the two Exilles prisoners in 1687.
• 3%Count Ercole Antonio Mattioli, whose name the burial register echoes; excluded by the custody chain of 1681 to 1698 and, on the reading every editor of the letters since Lair (1890) has given it, by a letter of 10 May 1694.
• 5%Someone else: a prisoner brought from Pignerol in 1694, or a man the printed record never names.

Figure 1. The four outcomes, weighted as in the text. Made for this report from the probability arithmetic below (three link weights, 0.99 × 0.93 × 0.85).

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. Which of the two Exilles prisoners died in January 1687 is decided by inference rather than by any document, so the headline is the product of three judged weights and is printed with its sensitivity below.

2. Who the valet was before July 1669

Not identified. No document names his master, his birthplace or his employment. The two current hypotheses, Petitfils's servant of Madame carrying letters in the negotiation that produced the secret treaty of Dover in 1670, and Sonnino's valet of Antoine-Hercule Picon, the manager of Mazarin's fortune, are both inferences from the same silence. Petitfils's one concrete name, a Gaspard Eustache baptised at Calais on 27 December 1643, is no more than a matching surname in a parish register; Sonnino writes that he found no Auger or Dauger among the Picons' servants. Against each other the two hypotheses are roughly even, and together they are less likely than an employment the record does not point to at all.

3. 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 the object of the orders was that he should 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 ever ordered a mask. The mask is first attested in a rumour of September 1687, and its use at the Bastille was Saint-Mars's practice, approved in general terms by the King in August 1698. The long detention is the ordinary working of Louis XIV's state prisons, which released nobody whose secret was still thought live, continued under a minister (Barbezieux, who succeeded his father Louvois in 1691) who inherited the precautions, and a jailer whose posts and pay depended on continuing to hold him.

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) reduced the choice to Mattioli or the valet, showed that La Rivière died at Exilles in 1687, and declined to choose; Petitfils (1970, 2003), Noone (1988), Sonnino (2016) and Wilkinson (2021) decide for the valet, while Funck-Brentano (1894) held for Mattioli, Thompson (1987) for Eustache Dauger de Cavoye, the disgraced nobleman who bore the valet's name, Macdonald (2005) for d'Artagnan, and Sarah Madry, in a 2016 article in the Cahiers du dix-septième, a journal of seventeenth-century studies, argues that the word "valet" itself cannot be relied on. This report reaches the valet at a lower confidence than his proponents state, because one link in the chain, the death of one of two unnamed prisoners at Exilles in January 1687, is decided by inference and not by any document.

What moves the headline. The 78% is the product of three links: that the Bastille prisoner of 1698 was the man Saint-Mars brought from Sainte-Marguerite (put at 99%), that the island prisoner of 1698 was the prisoner brought from Exilles in 1687 (put at 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 1687 survivor was the 1669 valet rather than La Rivière (put at 85%, on the strength of Barbezieux's "twenty years" of 1691 and "what he has done" of 1697). Read the 1691 letter 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%. The 78% and every figure derived from it are conditional on those three weights, which are judgements about the documents rather than measurements; the arithmetic is published with the working files for this case. The candidate table and the sensitivity are below.

The question and the seven fixed points

The beginning and the end of this prisoner's custody are documented, and the years between them are disputed. The end is the Bastille: du Junca recorded on 18 September 1698 that the new governor Saint-Mars had arrived with "an old prisoner he had at Pignerol, whom he keeps always masked, whose name is not said", and on 19 November 1703 that "the unknown prisoner, always masked with a mask of black velvet" had died. The parish of Saint-Paul buried him the next day under the name Marchioly, aged "forty-five years or thereabouts". The beginning is a letter of 19 July 1669 in which Louvois tells Saint-Mars to prepare a dungeon at Pignerol for "the man named Eustache d'Auger", adding that "as he is only a valet" he needs no great furniture. The question is whether the same man stands at both ends, and, if so, who he was and why the state kept him.

The method is the one Jules Lair used in 1890 and Georges Mongrédien in 1952, applied to the whole printed record rather than to extracts. Every letter of the Pignerol, Exilles, Sainte-Marguerite and Bastille correspondence that the nineteenth- and twentieth-century editions print was read. The editions, each the work of a historian who printed part of the file, are Roux-Fazillac (1801), Joseph Delort (1825 and 1829), François Ravaisson's Archives de la Bastille (vols. III and VII to XI, 1866 to 1880), Théodore Iung (1873), Jules Loiseleur (1882), Marius Topin (1883), Lair (1890), Funck-Brentano (1894 and 1901), Émile Laloy (1913), and Mongrédien's article of 1953; the report cites them below by surname, volume and page. One hundred and eighty entries were transcribed with the archive reference each edition gives, and the variants between editions were tabulated (the transcriptions and the tables are published with this report as working files). From that record seven fixed points were taken, each resting on a named document:

1. (a) Arrested in the Dunkirk or Calais area in late July 1669 and delivered to Saint-Mars at Pignerol by Captain de Vauroy between 20 and 24 August 1669; "only a valet" (Louvois, 19 July 1669; the King's orders of 28 July; Saint-Mars, 21 August).
2. (b) Allowed in 1675 to serve Nicolas Fouquet as a valet, and never Lauzun (Louvois, 30 January and 11 March 1675).
3. (c) Shut up with Fouquet's valet La Rivière after Fouquet's death in March 1680, the two becoming "the two prisoners of the lower tower" (Louvois, 8 April 1680 and 12 May 1681).
4. (d) Taken by Saint-Mars to Exilles in October 1681 while Mattioli stayed at Pignerol (Louvois, 9 June 1681; Saint-Mars to the abbé d'Estrades, the French ambassador at Turin, 25 June 1681).
5. (e) The survivor of the two after one died in January 1687, carried to the île Sainte-Marguerite in a covered chair, arriving 30 April 1687 (Louvois, 13 January 1687; Saint-Mars, 20 January and 3 May 1687).
6. (f) Brought to the Bastille on 18 September 1698 as "an old prisoner he had at Pignerol" (Barbezieux, June to August 1698; du Junca's register).
7. (g) Died 19 November 1703 and buried as "Marchioly, 45 ans ou environ" (du Junca; Saint-Paul register).

A candidate who is documented alive and free, or in another prison, or dead, at the date of any fixed point fails that point. The age on the burial register is treated as a weak test, because the men who supplied it did not know the prisoner and the same entry carries a false name. Points (a) to (d) and (f) to (g) are documented as events. Point (e) is documented as an event but not as an identity: no surviving letter says which of the two Exilles prisoners died. Point (e) is therefore argued separately below.

Two cautions on the record. First, no original manuscript was consulted. The letters survive in three places: the Archives nationales series K 120 (Saint-Mars's own file of the letters he received, to 1681), the war-office registers now at the Service historique de la Défense, the French army's historical archive at Vincennes (series A1, which the editions cite as "P. x, v. y, Dépôt de la guerre", page and volume), and the Bastille's own registers of entries and exits, now in the Arsenal library in Paris (mss. 5133 and 5134). All were read through the editions, with each edition's page and reference recorded. Six documents were also seen in the photographs of the originals published on masque-de-fer.com by Jean-Marc Dabos, who runs the site. Second, the editions disagree on words that matter. Where a reading changes the sense, both readings are given, and the report says which is the more literal transcription.

1669: the arrest

The earliest document about this prisoner is a letter of 19 July 1669 from Louvois, secretary of state for war, written from Saint-Germain, the King's residence outside Paris, to Bénigne d'Auvergne de Saint-Mars, the captain who had guarded Fouquet at Pignerol since 1665.

Nine days later the King signed three orders, to Captain de Vauroy, sergeant-major of Dunkirk, to the marquis de Piennes, governor of Pignerol, and to Saint-Mars. The war-office minutes of all three leave the name blank ("le nommé ….", Iung 1873, pp. 187–188; vol. 234, fols 271, 272 and 274 per Laloy). The order to Vauroy gives the only stated ground: "estant mal satisfait de la conduite du nommé …., et voulant m'assurer de sa personne", "being ill satisfied with the conduct of the man named …., and wishing to make sure of his person". Vauroy was to seize him, conduct him himself to Pignerol, and come back to report. Saint-Mars acknowledged delivery in a letter the editions date 21 August: "M. de Vauroy a remis entre mes mains le nommé Eustache d'Auger", adding that he told the man in Vauroy's presence that if he spoke to him or anyone of anything but his needs "je lui mettrois mon épée dans le ventre" (Roux-Fazillac 1801, pièces; Laloy, p. 104). Louvois's reply of 10 September acknowledges letters "des 24 et dernier du mois passé", so the delivery fell between 20 and 24 August.

The same reply allows the prisoner a prayer book, mass on Sundays and feast days "without being in the same place" as Fouquet's chaplain, and confession "three or four times a year if he wishes it, and no more" (Delort 1829, I, p. 157). On 25 September the prisoner "sent to you last" is unwell and may be bled without waiting for orders (I, p. 158). On 17 December, after Fouquet's valet Champagne was caught passing letters, Louvois orders half-circle grilles and close hurdles on the windows of "the one who has just arrived" so that he can see nothing but the sky, "to avoid similar accidents" (I, pp. 159–161). On 26 March 1670 Louvois has heard that a man called Honneste or one of Fouquet's valets spoke to "the prisoner brought to you by the major of Dunkirk" and asked whether he had nothing of consequence to say, and the prisoner answered that they should leave him in peace; Saint-Mars is to inspect the cell and make it so that the prisoner "ne puisse voir ni estre veu de personne, et ne puisse parler à qui que ce soit" (I, pp. 167–168). Saint-Mars, 12 April 1670: curious people ask news of "mon prisonnier" and why so many retrenchments, and he tells them "des contes jaunes", tall tales (I, p. 169).

Three points in this first year bear on the identification. The name is written out on the two originals (19 July and 21 August) and blank on the three minutes, which is why every editor from Roux-Fazillac onward has read the name differently: d'Auger (Roux-Fazillac, Delort, Laloy), Dauger (Iung, Lair's transcription, Mongrédien, Sonnino), Danger (Lair's judgement, Petitfils, and the handwriting analysis that Bernard Caire, who examined the original letters, presented in 1987), and d'Angers in the original of 8 April 1680, which is legible in Jean-Marc Dabos's photograph. The place of arrest is nowhere stated; "Dunkirk" is an inference from Vauroy's post, and "Calais" rests on a reimbursement order for Vauroy's journey that the archivist Stanislas Brugnon found among the papers of Colbert, the King's finance minister (Bibliothèque nationale, Mélanges Colbert, vol. 282, fol. 276 v°), and produced at a conference on the Iron Mask held at Cannes in 1987: 3,000 livres (a state prisoner's keep at Pignerol was set at four livres a day) "pour avoir été lui quatrième de ladite ville de Dunkerque en celle de Calais, et de ladite ville en celle de Pignerol, lui cinquième", four men from Dunkirk to Calais and five from Calais to Pignerol, the fifth being the prisoner (as quoted by Petitfils 2003, p. 244 n. 11). Petitfils, Noone, the historian Michel Vergé-Franceschi and Sonnino accept it; the document itself was not consulted for this report, and Petitfils's transcription is the only one in print. And the precautions are severe but not unique. On 10 March 1674 Louvois sent Saint-Mars another prisoner, "who, though obscure, is still a man of consequence", to be treated "in the same way as the prisoner M. de Vauroy brought you", and Le Tellier, Louvois's father and his predecessor as war minister, writing on 23 March, said why: he was "un fripon insigne qui, en matière fort grave, a abusé des gens considérables" (Ravaisson, III, pp. 176–177). The ministry stated a reason for the prisoner of March 1674, and no surviving letter states one for Eustache.

1670 to 1680: a valet among valets

For eleven years the man is one prisoner among several in Saint-Mars's charge at Pignerol, and the letters treat him both as a servant and as a man whose past must not be spoken of. In August 1671 he is ill, and Saint-Mars writes that "il me donne assez d'occupation pour lui souhaiter la santé" (Laloy, pp. 109–110). On 20 February 1672 Saint-Mars, unable to find servants willing to be shut up with his prisoners, proposes "ce prisonnier qui est dans la tour, et que vous m'avez envoyé par M. le major de Dunkerque" as a valet for Lauzun: "Je ne pense pas qu'il dît à M. de Lauzun d'où il sort, après que je lui aurais défendu; je suis sûr qu'il ne lui dirait pas aussi aucune nouvelle, ni ne me demanderait point de sortir de sa vie, comme font tous les autres" (Ravaisson, III, pp. 117–119). No reply is printed, but the answer is known from what followed. On 30 December 1673: "Pour le prisonnier de la tour, que M. de Vauroy m'a amené, il ne dit rien, il vit content comme un homme tout à fait résigné à la volonté de Dieu et du Roi" (Ravaisson, III, pp. 174–175).

Fouquet's valet Champagne had died "at Notre-Dame de septembre" (1674 by Lair's dating, 1675 by Laloy's), and Eustache took his place beside the older valet La Rivière. In December 1678 the King prepared to soften Fouquet's prison, and before doing so put one question to him.

Fouquet's answer is lost. On 17 February 1679 Louvois told him that "Sa Majesté veut bien se remettre à vous de la conduite qu'il faudra tenir à l'égard d'Eustache Danger" (Ravaisson, III, pp. 209–210), and the memoir (the written instructions) of 20 January 1679 that opened Fouquet's and Lauzun's apartments to each other and to visitors carries two paragraphs about "le nommé Eustache": whenever Lauzun or any stranger comes into Fouquet's room, Saint-Mars is to withdraw Eustache and not put him back "until there is no one there but him and his old valet"; when Fouquet walks in the citadel Eustache stays in the room, and follows only when Fouquet walks alone with his old valet (Delort 1829, I, pp. 280–285). On 15 February 1679: "Le Roy se remet à vous de régler avec monsieur Foucquet, comme vous jugerez à propos, ce qui regardera la seureté de la personne du nommé Eustache Danger, vous recommandant sur tout de faire en sorte qu'il ne parle à personne en particulier" (I, pp. 286–288). On 13 September 1679 Louvois asks for "des nouvelles de la santé d'Eustache Dangers" (Ravaisson, III, p. 212; the final s is visible in the photograph of K 120 no. 277).

Fouquet died on 23 March 1680. Saint-Mars reported the death, a hole he had found between Fouquet's and Lauzun's rooms, and his judgement that Lauzun knew most of the important things Fouquet had known "et que le nommé la Rivière ne les ignore pas". Louvois's reply of 8 April 1680 is the document on which the later chain of custody depends.

One letter of this spring has been used to argue that the chain of custody breaks here. On 12 March 1680 Louvois told Saint-Mars that the King did not intend "que vous payez à monsieur Foucquet les gages de celuy de ses valets qui est mort" (Delort 1829, I, p. 316; K 120 no. 299, photographed on masque-de-fer.com). Jean-Marc Dabos reads this as the death of Eustache and builds on it a thesis that the masked man was Fouquet himself, his death of 23 March being a fiction. The rest of the file tells against that reading, though the alternative reading is itself an inference. Fouquet's valets had been deprived of wages since January 1670 for their infidelity, and La Rivière asked for his arrears in September 1679 (Iung, p. 196); the most economical reading is that Saint-Mars had asked whether Fouquet should be paid the arrears of the valet who had died, Champagne, and that Louvois said no. The letter of Saint-Mars that Louvois was answering is not in the editions, so that reading is not documented; what is documented is that four weeks later Louvois names both Eustache and La Rivière as living men to be shut up, orders Fouquet's furniture disposed of, releases the body to Fouquet's son (8 and 9 April), and the Visitation Sainte-Marie convent in Paris entered his burial in its register on 28 March 1681 (Lair, II, pp. 473–474).

1681 to 1687: the two blackbirds

Lauzun was released in April 1681 and Saint-Mars asked for another post. The King gave him Exilles, a fort in the Alps between Pignerol and Briançon, and decided which prisoners would go with him.

On 9 June the two are to leave Pignerol "dans une litière" under the escort of Saint-Mars's company, the rest of the prisoners staying under the lieutenant Villebois at two écus (six livres) a day "for the food of these three prisoners", and Mattioli's clothes are to be carried to Exilles "to be able to return them to him if ever His Majesty ordered him set free" (Delort 1825, p. 269). Delort read that last clause as proof that Mattioli went to Exilles. Saint-Mars's own letter to the abbé d'Estrades of 25 June 1681 says the opposite: "j'aurai en garde deux merles que j'ai ici, lesquels n'ont point d'autres noms que messieurs de la tour d'en bas; Matthioli restera ici avec deux autres prisonniers", "I shall have in my keeping two blackbirds I have here, who have no other names than the gentlemen of the lower tower; Mattioli will stay here with two other prisoners" (Ravaisson, III, pp. 212–213, from the d'Estrades papers). Louvois then delayed the move so that it would not look contrived: the repairs are to be slowed "so as to have a pretext not to leave until about the first days of October", "observant de vous conduire de manière qu'il ne paroisse point d'affectation" (22 July 1681; Delort 1825, p. 274). The reason for the delay was Catinat, the officer who was hidden at Pignerol under a false name while France negotiated the secret purchase of Casale, a fortress in northern Italy, from the Duke of Mantua. No letter gives the day of departure; the first letter from Exilles is of 4 December 1681.

At Exilles the pair were held in the manner Saint-Mars described on 12 July 1681 before leaving: they would not leave their room to hear mass, a lieutenant would sleep above them, two sentries would watch the tower day and night, and the old curé of Exilles could confess them but was forbidden "on the King's behalf to know who these prisoners are, nor their names, nor what they have been" (Roux-Fazillac, p. 110; Iung, pp. 398–399). On 14 December 1681 Louvois allowed them clothes, "mais il faut que les habits durent trois ou quatre ans à ces sortes de gens-là", "for that sort of people" (Topin, p. 332). On 2 March 1682 the King wanted the prisoners "que l'on nommoit à Pignerol de la Tour d'en bas" to have no dealings with anyone, not even the garrison (Iung, p. 400). On 16 April 1684 Louvois asked Saint-Mars for news of "vos prisonniers" and "ce que vous savez de la naissance du nommé La Rivière et de l'aventure par laquelle il fut mis au service de feu M. Foucquet" (Iung, p. 402): La Rivière was at Exilles, and the minister did not know who he was. On 30 January 1685 what the prisoners had told Saint-Mars was "d'aucune conséquence" (Iung, p. 403). On 7 June 1685 one of them wanted to make his will (Iung, p. 403). On 23 December 1685: "Mes prisonniers sont toujours malades et dans les remèdes. Du reste, ils sont dans une grande quiétude" (Roux-Fazillac, p. 113). On 9 January 1686 Louvois complained that Saint-Mars should have named "quel est celuy de vos prisonniers qui est devenu hydropique", and on 3 November 1686 allowed the dropsical one to be confessed "lorsque vous verrez apparence d'une prochaine mort. Jusque-là il ne faut pas que lui ou son camarade aient aucune communication" (Iung, p. 404).

The correspondence is incomplete at this point. Saint-Mars's letter of 5 January 1687 announcing a death is lost. Louvois's reply of 13 January survives: "J'ai reçu votre lettre du 5 de ce mois, par laquelle j'apprends la mort d'un de vos prisonniers" (Iung, p. 405; P. 231, v. 779). From Saint-Mars's next letter, 20 January 1687, the prisoner is referred to in the singular, "mon prisonnier", and in the singular in every later letter.

The weak link: who died at Exilles

Everything after January 1687 is documented for one man. Whether that man is the valet of 1669 or Fouquet's valet La Rivière is not stated in any surviving letter, and the reader should know that every author who names the survivor is inferring. Lair (1890, II, p. 479) wrote that the dropsical prisoner who died on 4 January 1687 was La Rivière and that Saint-Mars was left with "Eustache Danger"; Funck-Brentano and Petitfils follow him; Iung, who thought Eustache had stayed at Pignerol, made the dead man the Jacobin, a Dominican friar held at Pignerol since 1674; Roux-Fazillac, who had no letter between December 1685 and January 1687, asked "est-ce Matthioly ou le moine?" and left it open. The date of death is also an inference: 4 January is deduced from the letter of the 5th, and Funck-Brentano's "end of December 1686" is a mis-citation of Lair.

Three later letters bear on the identity. Jules Lair drew the conclusion first (1890, t. II, pp. 485–486): the masked man was "un des deux Messieurs de la Tour d'en bas", La Rivière died at Exilles in January 1687, and the survivor was Danger. Georges Mongrédien restated it from the later letters (Le Masque de fer, 1952, pp. 196–197 and 222: "nous avons démontré qu'il est mort à Exiles en 1687"), and Mongrédien himself noted (p. 230) that Lair's whole identification presupposes that the 1703 prisoner never left Saint-Mars's custody, which no letter states in so many words before 1698. This report reaches the same conclusion by the same route, and assigns it a probability.

A new minister, writing from his father's registers, identifies the one prisoner who requires special precautions by the length of his custody. The valet of 1669 had been in Saint-Mars's hands for twenty-two years. La Rivière had come to Pignerol with Fouquet in January 1665, twenty-six years before, as a free servant, and had been a prisoner in his own right for eleven years, since April 1680. "Vingt ans" is a round number in either case, but twenty-two rounds to twenty, and a clerk working from a file that opens with the dispatches of July 1669 would arrive at that figure. Editors who preferred other candidates read the number to fit them: Iung made it eighteen years for a prisoner of 1673, Topin seventeen for the Jacobin of 1674, and Paul Lacroix, who held for Fouquet (1836), twenty-seven.

The third piece of evidence is what the letters of the whole file are about. From 1669 to 1680 the minister's letters are about Eustache: his cell, his windows, his health, whether he has been spoken to, whether he has spoken, to whom he may be lent, what Fouquet knows of his past. La Rivière appears in them as Fouquet's servant, to have his wages stopped, to be enquired after in 1684 because the minister did not know where he came from, and to be shut up in 1680 because he might have heard what Eustache and Fouquet knew. Nothing in the record makes La Rivière a man who had "done" something. After 1687 the state builds a prison for the survivor at a cost of 6,926 livres (16 March and 17 August 1687), refuses to name him in writing, moves him to the Bastille "so that he be seen and known by no one", and keeps him there for five years. That treatment fits the prisoner the ministry had been protecting since 1669 rather than La Rivière, who had been shut up only because he might have heard.

Three points can be argued for La Rivière, and this report judges each of them weak. The prisoner who wanted to make a will in June 1685 is more likely a servant with wages owed than "un valet" arrested with nothing. The state had treated the two as one unit since 1680, so the special prison of 1687 proves only that the survivor was one of the pair and does not say which. And Barbezieux, a minister of twenty-three, may have used "vingt ans" loosely. Taking these together, this report puts the probability that the survivor of January 1687 was the valet of 1669 at 85 per cent and the probability that he was La Rivière at 15 per cent. Mongrédien treated the point as settled and Petitfils does not discuss it as a difficulty; that is the one place where this report is less confident than Lair, Mongrédien and Petitfils.

1687 to 1698: Sainte-Marguerite

Louvois gave Saint-Mars the government of the îles de Lérins, the two islands off Cannes of which Sainte-Marguerite is the larger, on 8 January 1687, before he knew of the death, and the letter still speaks of "les prisonniers" in the plural (Iung, pp. 404–405). Saint-Mars's letter of 20 January is the first to speak of one man, and it proposes the covered chair from which the first rumour of a mask followed.

Louvois left the manner of transfer to him, "pourvu que vous en puissiez répondre" (end of January 1687; Iung, p. 406). On 16 March the King paid 5,026 livres for a prison and lodging on the island "pour rendre sûre la personne de votre prisonnier", and Saint-Mars was to leave immediately after Easter; on 23 March, recovering from a fever on the island, he promised to bring his prisoner "sans que personne le voie ni lui puisse parler" and not to let him hear mass until he was in the new cell with its adjoining chapel (Roux-Fazillac, p. 115). The journey took twelve days.

On 8 January 1688 Saint-Mars reported that he had put "mon prisonnier quy est toujours valtudinaire à son ordinaire" in one of the two new cells, "grandes, belles et claires", and added: "Dans toute sette province l'on dit que le mien est monsieur de Baufort, et dautres dissent que cest le fis de feu Cronvel" (Iung, pp. 410–411, from the original then in private hands). The duc de Beaufort, a grandson of Henri IV lost in the fighting at Candia, in Crete, in June 1669, and a son of Oliver Cromwell were the guesses of Provence in 1688; the guess of a royal brother does not appear in the letters. He also sent an expense account for the prisoner without detail, "pour que personne par qui il passe puisse pénétrer autre chose que ce quils croyent".

After Louvois's death the letters thin out and change hands. Barbezieux's letter of 13 August 1691 on the "prisoner of twenty years" is quoted above. In January 1694 the lieutenant La Prade, who had kept the Pignerol donjon (the keep, where the state prisoners were held) since 1681, reported that "le plus ancien est mort et qu'il n'en sait pas le nom", and Barbezieux asked Saint-Mars to send the name in cipher (Iung, p. 90; Topin, p. 350 n.): the ministry could not itself say which prisoner had died. In March 1694, with Pignerol about to be handed back to the Duke of Savoy, the three remaining state prisoners of the donjon were sent one by one to the islands, Barbezieux writing on 26 February "Je ne vous en mande pas le nom, persuadé que vous le sçavez" and on 20 March "Vous sçavez qu'ils sont de plus de conséquence, au moins un, que ceux qui sont présentement aux îles" (Topin, pp. 340, 350). The "one" is Mattioli. On the reading of the next letter that every editor since Lair has given it, he died within a month.

On 6 January 1696 Saint-Mars described his daily routine for Barbezieux, and used for the first time in his own hand the phrase the ministers would use for the next two years: the first lieutenant to arrive "prend les clefs de la prison de mon ensien prisonnier par ou l'on commence, il ouvre les trois portes et entre dans la chambre du prisonnier quy luy remet honnestement les plats et assiettes"; the second lieutenant inspects everything that goes in or out "et voir s'il n'y a rien d'écrit sur les vaisselles"; the bed, the grilles, the privy and "fort souvent" the man himself are searched, "après luy avoir demandé fort sivilement s'il na pas besoin d'autre chose" (Iung, pp. 415–417). Barbezieux's letter of 17 November 1697 on "ce qu'a fait vostre ancien prisonnier" is quoted above. Then, in 1698, Saint-Mars was offered the Bastille.

Brugnon's objection: a death on the island in 1693

The one modern author who breaks the chain here is Stanislas Brugnon, the archivist whose Cannes paper of 1987 and privately printed Reconstitution du registre d'écrou des prisonniers de l'île Sainte-Marguerite (1992) supplied Petitfils with much of his Pignerol chronology. Brugnon held that the prisoner of 1687 died on the island in 1693 and never went to the Bastille. His evidence, as quoted by the Cannes public library, is the abbé Papon's Histoire générale de la Provence (1778): an officer of eighty-two at the Fort Royal, the fortress on Sainte-Marguerite, remembered that his father, "l'homme de confiance de Saint-Mars", had "été prendre le mort à l'heure de minuit dans la prison et l'avait porté sur ses épaules dans le lieu de la sépulture". Brugnon named the father as Jacques Souchon, a cadet of Saint-Mars's company, and dated the death to 1693.

The objection fails against the ministerial usage. Saint-Mars's "mon ensien prisonnier" of January 1696, Barbezieux's "vostre ancien prisonnier" of November 1697 and the three letters of summer 1698 are one continuous term for one man, and a man who had died in 1693 could not be "your old prisoner" in 1697; Barbezieux's letter of 13 August 1691 had already fixed which prisoner the precautions concerned. Papon's midnight burial, if it records an actual event, fits Mattioli, who on that reading died on the island in April 1694 and whose burial no letter records. Petitfils, who used Brugnon's papers through his widow, did not follow him on this point. The probability that the 1698 Bastille prisoner was the man carried from Exilles in 1687 is put here at 93 per cent, the residue covering Brugnon's reading and the possibility that one of the Pignerol prisoners of 1694 was the "ancien prisonnier qu'il avait à Pignerol" of du Junca's entry.

1698 to 1703: the Bastille

Saint-Mars left the islands at the end of August 1698 with his nephew Formanoir, the sergeant Rosarges and the turnkey Antoine Ru, went up the Rhône, and, according to the Gazette d'Amsterdam, a French-language newspaper printed in Holland, of 3 October, "a fait mettre un prisonnier qu'il avait amené avec lui, et il en a laissé un autre à Pierre-Cise, en passant à Lyon" (Iung, p. 422), Pierre-Cise being the state prison above Lyon. The second prisoner is the subject of Bernard Caire's article "Chi era il prigioniero di Pierre-Scise?" (1998), which Petitfils cites for a reimbursement order showing that another prisoner travelled with him; neither article was consulted. The entry register of the Bastille, kept by du Junca, records the arrival.

On 3 November 1698 Pontchartrain, who had charge of the Bastille as secretary of state for the King's household, allowed "votre prisonnier de Provence" to confess and take communion as often as Saint-Mars judged fit (Topin, p. 356; Iung, p. 423). Iung inferred that the mask was for the walk across the courtyard to mass, which is what the Bastille tradition later said. On 30 April 1701 du Junca put a new prisoner "en compagnie avec Tirmont, dans la seconde chambre de la tour de la Bretaudière, avec l'ancien prisonnier, tous les deux bien renfermés" (Ravaisson, X, p. 369): the old prisoner had moved down a floor and, on Ravaisson's reading, shared the tower but not the room. The Bastille of 1698 to 1703 had between twenty and forty prisoners, and du Junca's register shows the practice of false names at burial for others besides this one (the suicide Dupressoir-Louvart was buried in June 1702 "sous le nom de Pierre Massuque").

The Saint-Paul register was destroyed in the fires of the Commune in 1871; a facsimile published in Topin's fifth edition (1878) reads, in Funck-Brentano's transcription (1894, p. 296): "Le 19e, Marchioly, âgé de quarante-cinq ans ou environ, est décédé dans la bastille, duquel le corps a esté inhumé dans le cimetière de st Paul, sa paroisse, le 20me du présent, en présence de monsieur Rosage, majeur de la bastille, et de Mr Reglhe, chirurgien majeur de la bastille, qui ont signés. Rosarges, Reilhe." Earlier readers from Griffet to Lair had read the name as Marchialy or Marchiali. Two later readers corrected them: the critic Paul de Saint-Victor first read Marchioly, and Pierre Bertrand connected it to Saint-Mars's habitual spelling "Marthioly" for Mattioli. The age was supplied by Bastille officers who had known the prisoner for five years without seeing his face. Everything else about the death, the burning of the furniture, the scraping of the walls, the drugs in the coffin, comes from Griffet in 1769 reporting the garrison's memory, from the major Chevalier's sheet printed in 1789, and from a letter of 1768 by Formanoir de Palteau, Saint-Mars's great-nephew; no document of 1703 records any of it.

Every candidate against the fixed points

Twenty candidates have been proposed since 1745. The table tests each against the seven fixed points; the names in brackets after each candidate are the authors who proposed him, with the year. "Fail" means a document places the man elsewhere, free, or dead at that date; "open" means the documents do not decide; "n/a" means the point does not apply. A short script generates the table from a candidate file published with this case, so anyone who disagrees with a cell can change the file and rerun it.

Table 1. Twenty candidates against the seven fixed points. Made for this report from the candidate file and the documents cited in each row.

Only the valet of 1669 passes every documented point. Two candidates deserve a word because they were once the consensus or still have an argument.

Mattioli was the identification of the war archives themselves in the nineteenth century, from Roux-Fazillac in 1801 to Funck-Brentano's article of 1894, and it rests on the burial name, which is strong, and on a misreading by Delort of the letter of 9 June 1681. The strong point is the burial name: a clerk who had to write something wrote a name three letters from "Marthioly", the spelling Saint-Mars used for Mattioli, and Funck-Brentano called that discovery the "victoire décisive". The misreading is Delort's use of the sentence about Mattioli's clothes in the letter of 9 June 1681 to send him to Exilles, which Saint-Mars's own letter of 25 June contradicts. Against the name stand the documented facts that Mattioli was arrested in May 1679, held openly under his own name in the Catinat letters, beaten with a stick and denied a physician (Louvois, 15 and 22 May 1679), left at Pignerol in 1681, moved to the islands only in March 1694, and dead there with his valet before 29 April 1694. Why the Bastille chose his name for the register is not known. The likeliest answer is the one Lair gave: the name was in the Pignerol papers Saint-Mars carried with him, and a Mantuan diplomat was as good a false identity as any. The probability that the masked man was Mattioli is put at 3 per cent, a weight that reflects the burial name set against six failed fixed points.

The royal twin is the popular candidate and the only one that cannot be tested against a document, because no document records him. He fails on the only ground available: the man of 1669 was "un valet" whose furniture was to be cheap, whom Saint-Mars offered as a servant to Lauzun in 1672, who served Fouquet from 1675 to 1680, whose bed was sold for thirteen écus in 1687, and whose clothes had to last three or four years "à ces sortes de gens-là". Sarah Madry (2016) has argued that "valet" in the letter of 19 July 1669 is a pun on the jack of spades and on Louvois's rival Louis d'Auger de Cavoye, and that Louvois knew nothing of the prisoner; the argument does not reach the six later letters that treat the man as a servant.

Who the valet was

The record says little about the man before Pignerol. He was entered as Eustache d'Auger or Danger or d'Angers; the ministers used the Christian name alone from 1678 onward ("le nommé Eustache"), which Laloy took to mean a nickname given at his entry into prison and Bernard Caire, after examining the handwriting of the originals, took to mean that Eustache was the surname and "d'Angers" a place of origin. He was a valet, or at least treated as one for thirty-four years, and was arrested by the major of Dunkirk, probably at Calais. He was a Catholic who heard mass and confessed, and was ill in 1669, 1671 and 1687 and "always valetudinary" in 1688. He could serve, and Fouquet found him acceptable for five years. In 1680 he did something with drugs that Louvois found hard to believe possible in a cell. Saint-Mars wrote in 1673 that he said nothing and "lived content". He died after a day's illness in 1703.

Every proposed identification of his master argues from the dates of events near the arrest. The arrest fell in the weeks after the execution of the Huguenot conspirator Roux de Marsilly (22 June 1669), the recall of the abbé Pregnani, an Italian astrologer whom Louis XIV had sent to Charles II's court, from his mission (May to July), and the first exchanges of the secret negotiation between Louis XIV and Charles II that Madame, Henrietta of England, carried and that produced the secret treaty of Dover of June 1670, in which Charles II promised to join France against the Dutch and to declare himself a Catholic. Vauroy was posted at Dunkirk, and the road from Calais is the route from England. The Scottish writer Andrew Lang (1903) made him Marsilly's valet Martin, and the report of Colbert de Croissy, the French ambassador in London, of 1 July 1669 that Martin refused to come to France, followed by the order of Lionne, the foreign minister, of 13 July to drop the matter, dispose of that. Arthur Barnes, a Catholic priest (1908), made him Pregnani, who died in Rome in 1678 or 1679. Maurice Duvivier, in a book of 1932, made him Eustache Dauger de Cavoye, a disgraced nobleman of the right name, and Mongrédien published in 1953 the two letters Cavoye wrote in 1678 and about 1679 from Saint-Lazare, a house of correction in Paris where he had been held since 1668. Théodore Iung (1873) connected him to the poisoners, ten years before the Chambre ardente, the special court that tried the poisoning cases of 1679 to 1682. Petitfils's first book (1970) made him a valet of Colbert hired by Louvois for a poisoning in December 1668, hidden from the King and from Colbert; the King's own signature on the order of 28 July 1669 and Colbert's payment of Vauroy's journey tell against it, and Petitfils had dropped it by 1991. Émile Laloy (1913, pp. 199–200), after abandoning Pregnani, kept the man himself, "à peu près sûrement un prêtre", and put the secret "plutôt dans 'ce à quoi il avait été employé' que dans sa personnalité même"; that is the first statement of the reading followed in this report.

The two hypotheses now in the field are these.

• Jean-Christian Petitfils, Le Masque de fer: entre histoire et légende (Perrin, 2003; revised 2004). Eustache Danger was "un humble valet affecté au service de Madame", who had handled or carried secret papers concerning Charles II's conversion and the alliance against the Dutch, talked too much, and was seized at Calais by Vauroy with three archers (as summarised by his reviewers Stéphane Genet, writing in 2003 for H-Net, an online network of scholarly reviews, and Diego Venturino, in the journal Dix-huitième siècle in 2005, and in his own chapter "Du danger du secret au secret de Danger"). Petitfils reads the name as the surname Eustache with a place of origin, and offers one man from the Calais registers as a possible match: Gaspard Eustache, son of Jehan Eustache and Jossine Cléché, born 20 December 1643 and baptised on the 27th at Notre-Dame de Calais (Archives départementales du Pas-de-Calais, 5 MI 193 R28; Petitfils 2003, p. 167). Nothing but the surname and the town connects that child to the prisoner, and a birth in 1643 gives sixty at death against the register's "forty-five or thereabouts". Petitfils also sets the limit of his own thesis: 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". The mask, in Petitfils's account, was "une mystification montée de toute pièce par le geôlier" and "ne fut jamais demandé par les ministres". Genet raised three objections which Petitfils's book, in the parts that could be read, does not remove: why forbid the man's name if he was nobody; why keep him after 1672, when the Dover secret had lost its value; and why he was so docile.
• Paul Sonnino, The Search for the Man in the Iron Mask: A Historical Detective Story (Rowman and Littlefield, 2016). Eustache Dauger was a valet of Antoine-Hercule Picon, the man who managed Mazarin's fortune, and "the most trusted of those men", to whom Picon had entrusted "some of the most sensitive details of Mazarin's peculations" (the historian Julius Ruff's review for H-France, an online review journal, 2016). The secret was that Mazarin had built part of his fortune by cheating Henrietta Maria, Charles I's French-born queen, and Charles I himself, with the connivance of Louis XIV; the trigger was that Louis was courting Charles II in 1669; the offence was that Dauger "blabbed at the wrong time", perhaps in drink. Sonnino had spent a decade on the candidacy of Claude Imbert, an agent of the cardinal de Retz, Mazarin's old enemy, until he found Imbert alive and free in 1675, 1679 and 1681; the Picon thesis dates from 2003, when Sonnino found Picon's marriage contract of 30 December 1662 (Archives nationales, notarial archive, Minutier central, XVI/477), and grew out of his archival work on the three versions of Mazarin's will (French Historical Studies, 2014). Sonnino states that Dauger was "neither the valet of Nicolas nor of Basil" Fouquet, and that he "found no trace of individuals with the name of Auger or Dauger who might have provided the Picons with a suitable valet". The historian David Parrott's review (French History, 2017) says the fair thing: "Dauger's status as one of Picon's valets rests on a balance of probabilities rather than written evidence."

Both theses explain the timing with the same events and the place with the same road, and both are consistent with the one substantive clue the letters give, that the secret was an employment. They differ on whether the secret would have expired. Madame's negotiation had become public policy by 1672, whereas Mazarin's money, on Sonnino's account, implicated the King himself. That is Sonnino's structural advantage and Petitfils's weakness, and Petitfils meets it by saying that the prisoner was hidden "non parce qu'il était connu, mais parce qu'il était inconnu et, au fond, sans importance depuis les années 1680" (Venturino's summary). The letter of 12 May 1681, in which the King moves the two valets as the prisoners "of consequence" and leaves Mattioli behind, is hard to square with a prisoner who no longer mattered. No document names a valet of Picon or a servant of Madame. This report gives the Madame hypothesis about 25 per cent and the Picon hypothesis about 20 per cent, and leaves more than half the probability with an employment the surviving papers do not point to at all. These two figures are judgements about events for which no document survives.

Why thirty-four years

That question contains three: what the state was protecting, why the man was never released, and where the mask came from. The letters describe the kind of secret being protected but never state its content. They also show how a man once imprisoned remained so, and where the mask originated.

What was being protected

The orders of 1669 forbid the prisoner's communication with others and say nothing about his being seen. The dungeon's windows were to give on no approachable place and its doors were to keep the sentries from hearing; Saint-Mars alone was to carry his food; he was threatened with death for speaking of anything but his needs. In March 1670 the precaution is repeated as "ne puisse voir ni estre veu de personne, et ne puisse parler à qui que ce soit". In 1675 he may serve Fouquet, who would never leave, and never Lauzun, who would. In 1678 the King wants to know whether he has told La Rivière "ce à quoi il a été employé auparavant que d'être à Pignerol". In 1679 he is to be withdrawn whenever anyone but Fouquet and the old valet is in the room, and is to "speak to no one in private". In 1680 La Rivière, who "n'ignore pas" Fouquet's secrets, goes into the same room under the same fiction of release. In 1681 the reason for their arrest is not to be written even for the minister. In 1697 Saint-Mars is to explain to no one "ce qu'a fait vostre ancien prisonnier". The consistent object across twenty-eight years is that a valet's account of his own past employment should reach nobody, and in particular nobody who would return to court.

The file also rules three things out. His face was of no concern to anyone until Saint-Mars made it one in 1687. He was treated as a valet throughout, in furniture, clothing and wages. And the ministry did not treat his regime as unique. Louvois sent the prisoner of March 1674 to be treated "de la même façon que le prisonnier que M. de Vauroy vous a amené", and Le Tellier called that one "un fripon insigne qui, en matière fort grave, a abusé des gens considérables". The ministry applied the same regime to at least one other prisoner held for what he knew, and stated the grounds in his case. In a file of more than a hundred letters that gives grounds for other prisoners, no minister gives grounds for Eustache.

Why he was never released

Louis XIV's state prisons had no procedure for releasing a man held for what he knew, and this file shows none being used. Louvois's own letter of 17 December 1669 states the principle: if Saint-Mars removed La Rivière from Fouquet's service, he was to keep him seven or eight months first, "afin que, s'il avoit pris des mesures pour porter des nouvelles de son maître, elles soient si vieilles en ce temps-là qu'elles ne puissent en rien préjudicier". The length of the delay was matched to how long the news would stay useful. Of the state prisoners in Saint-Mars's Pignerol donjon, Fouquet died there in 1680, the Jacobin about 1694, Dubreuil, a spy held since 1676, on the island about 1697, and Mattioli, on the reading given above, on the island in 1694; Lauzun was released because a duchess, the King's cousin Mademoiselle de Montpensier, bought his freedom by settling estates on the duc du Maine, the King's illegitimate son. After Louvois died in 1691 the orders continued unchanged. Barbezieux's first letter asks for the same precautions as before; in 1694 the ministry does not know the name of the prisoner who has died at Pignerol and asks Saint-Mars for it in cipher; in 1697 the standing order is to continue. Whether Barbezieux or Pontchartrain ever knew what the man had done, the letters do not say. And the man's jailer had every reason to keep him. Saint-Mars was paid 500 livres a month at Pignerol and kept that pay at Exilles "moyennant quoi vos appointemens seront aussi forts que ceux des gouverneurs des grandes places de Flandre" (12 May 1681); he received a gratification of 15,000 livres in 1679 and 2,000 écus for his furniture in 1698; his free company, the soldiers raised and paid to guard his prisoners, and his lieutenants went where his prisoner went; and in January 1688 he was advertising his new cells to the minister "pour quand vous auriez des prisonniers à vouloir mettre en toute seureté". This does not require the secret to have remained dangerous, only that no one with authority ever ruled that it was not, and the file records no such ruling. Mongrédien (1952, pp. 194, 200 and 229) added that the reasons for keeping him multiplied inside the prison: after 1675 he had heard Fouquet's confidences, in 1680 he had handled the "drogues", and Lauzun had been told he was gone. Petitfils (1991, 2003) and Caire (1991) both concluded that the original secret "n'est pas aussi important qu'on l'a imaginé" and that Saint-Mars inflated it; Genet's question, why keep a man whose secret had lost its value, is answered by the practice of the prisons rather than by the content of the secret.

Where the mask came from

No minister ordered a mask. The word does not occur in any of the hundred and eighty letters and register entries read for this report from 1669 to 1703 until du Junca's entry of 18 September 1698, where it is Saint-Mars's practice ("lequel il fait tenir touiours masqué"). The documented sequence runs as follows. January 1687: Saint-Mars proposes a chair covered in oilcloth so that no one can see the prisoner, and Louvois leaves the means to him. May 1687: "personne au monde ne l'a vu … chacun cherche à deviner qui peut être mon prisonnier". September 1687: the bishop of Agde's newsletter reports a steel mask in a sedan chair and orders to kill the prisoner if he says his name. January 1688: Provence says he is Beaufort or Cromwell's son. July 1698: Barbezieux, in the King's name, tells Saint-Mars to take his precautions "pour empescher qu'il ne soit veu ny conneu de personne", and in August the King approves the precautions Saint-Mars listed in a letter now lost. September 1698: the prisoner enters the Bastille masked and is kept so, in black velvet, for the walk to mass. 1751: Voltaire turns the velvet into a mask "dont la mentonnière avait des ressorts d'acier". Petitfils and Noone both concluded that the mask was Saint-Mars's device, and this report confirms it with one qualification: the King approved in August 1698 whatever precautions Saint-Mars had proposed for the journey, so the mask, if it was among them, had the King's general sanction from then on. Saint-Mars had written in 1687 that his precautions made everyone try to guess who the prisoner was, and the guesses of 1688 were a duke and Cromwell's son. A velvet mask on a prisoner no one in Paris had seen invited the same guessing.

The strongest counter-cases

• Mattioli (Funck-Brentano 1894): the burial name against six documented failures, above all Barbezieux's letter of 10 May 1694. Put at 3 per cent.
• La Rivière as the survivor of 1687: undocumented either way; against him the "twenty years" of 1691, the "what he has done" of 1697 and the subject matter of the file. Put at 15 per cent within the chain, 14 per cent overall.
• Brugnon's death of 1693: Papon's hearsay of 1778 against the ministers' continuous "ancien prisonnier" of 1696 to 1698. Put at about 5 per cent together with the other ways the 1698 prisoner could be a different Pignerol prisoner.
• The royal brother (Voltaire to Marcel Pagnol): no document, and the valet's furniture, service, bed and clothes against it. No separate share of the probability is assigned to it beyond the 5 per cent residue.
• Fouquet (Dabos 2025): the 12 March 1680 wages letter read as a death, against the letter of 8 April 1680 that names both valets alive, the release of the body, and the Visitation burial entry of 1681.

What would overturn this

• Saint-Mars's letter of 5 January 1687, or any Exilles account, naming the prisoner who died. If it names Eustache, the Bastille prisoner was La Rivière and the first two verdict lines change places.
• A burial or garrison record of Sainte-Marguerite between 1687 and 1698 recording the death of the prisoner of the new cell, which is what Brugnon's reconstruction of the island register claims to establish. It was not available here.
• The original of Barbezieux's letter of 17 November 1697 (Service historique de la Défense, A1 vol. 1392, p. 171) reading "ce que fait" rather than "ce qu'a fait", which would remove one of the three supports for the 1669 identification of the survivor.
• Any document naming the master or employment of "Eustache" in 1669: Vauroy's reimbursement order in the Mélanges Colbert as Brugnon described it, Fouquet's sealed reply of January 1679, or the Calais notarial records Sonnino cites. Such a document would settle the second question and might change the third.
• The minute of Louvois's letter of 19 July 1669 read in a hand other than the scribe's, or the name shown to have been added later, which would bear on whether the name was real.

What is new here, and what is not

The identification is not new. Jules Lair established it in 1890 from the K 120 letters ("Il était Français, catholique, valet de profession. Tout porte à croire qu'il s'appelait réellement Eustache Danger", t. II, p. 454), Loiseleur had already shown in 1882 that the masked man of 1687 came from Exilles and could not be Mattioli, Mongrédien in 1952 eliminated every Pignerol prisoner but Mattioli and the valet and showed that La Rivière died in 1687, while declining to choose between the two ("Si oui, c'est Dauger, si non, c'est Matthioli", p. 231), Petitfils has defended the valet since 1970, and Noone (1988), Sonnino (2016) and Wilkinson (2021) start from him. The reading of the name as Danger or d'Angers is Lair's and Caire's; the arrest at Calais is Brugnon's (1987); the steel-mask report of 1687 was printed by Azéma (1963) and brought into the debate by Brugnon; the deleted "ce qu'il a vu" in the minute of 23 December 1678 is Caire's find (1987, 1991). The reading of the mask as the jailer's device is Noone's and Petitfils's, and the judgement that the secret was small is Caire's and Petitfils's. The refutations of Martin, Cavoye and Bulonde belong to Laloy (1913), Mongrédien (1953) and Batiffol (1894). The observation that the secret was an employment rather than an identity is Laloy's (1913, p. 200) and Caire's.

What this report adds is smaller and of three kinds. First, the whole printed record for 1669 to 1703 has been transcribed into one hundred and eighty dated entries with each edition's archive reference and the variants between editions tabulated, so that the chain of custody can be checked line by line rather than trusted; the entries, the candidate file, the script that produces the table and the script that produces the headline and its sensitivity are published with this case. Second, the 1687 identification is reported as an inference with a probability rather than as an established fact. Lair and Petitfils state the identity of the 1687 survivor as a fact, and Mongrédien wrote that he had demonstrated it; here it carries a probability and a published sensitivity, and the Bastille identification is reported at 78 per cent rather than as certain. Nothing in this report is new evidence: no document unknown to the printed editions was found, and the three points that follow are new readings of old evidence. Third, three small points that were not found in the works read, several of which were read only in fragments: the ministers' continuous use of "ancien prisonnier" from Saint-Mars's letter of 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 the dead valet's wages read as Champagne's wage arrears rather than as a death in 1680, an inference from the surrounding letters rather than a documented exchange; and the variant "ce qu'a fait" against "ce que fait" in the letter of 17 November 1697 flagged as a reading that bears on the argument and that the original would settle. The comparison of the 1669 orders with the 1674 order to treat another prisoner "in the same way" is in Laloy and Ravaisson; its use here, to show that the ministry could state a reason and did not, is this report's.

On the second and third questions the answer is that nobody knows. This report has not identified the valet either. It has measured how far the two current identifications rest on documents, and neither rests on a document that names the prisoner.

What was not read

• Petitfils (2003 and 2004) and Sonnino (2016), the two books that defend the current identifications, were not read in full. That is the largest gap in this report. Petitfils was reached through about 1,050 search-service fragments of the lending copy (130 characters each, from some 400 queries), his reviewers Genet (2003) and Venturino (2005) in full, a 2005 interview and a third-party summary; Sonnino through about 1,000 fragments, the reviews by Ruff (2016) and Parrott (2017, by summary) and by Publishers Weekly, and three interviews. The reviews of Sonnino in the Journal of Modern History, the Catholic Historical Review and the American Historical Review, and his 1992 article on Claude Imbert, were not obtainable. The quotations from both books above are therefore fragments, exact as far as they go; the objections raised may be answered in pages not seen. The two authors' own statements of their theses, in their words, are collected in the working notes published with this case.
• Mongrédien's Le Masque de fer (1952) was read in 3,439 snippets from Gallica, the Bibliothèque nationale's digital library, each carrying the printed page number, plus its full table of contents, and his 1953 article in full; his conclusion (pp. 221–231) is quoted from those snippets. Noone (1994), Wilkinson (2021), Thompson (1987) and Macdonald (2005) were read in search fragments only. Petitfils's 1970 book is known only from four contemporary reviews. Duvivier (1932) was not read; Bazeries and Burgaud (1893) were read through Batiffol's review and Laloy.
• The proceedings of the Cannes colloquium of 1987 (Brugnon, Caire) and of Pinerolo 1974 were not obtainable; the proceedings of a second conference, at Pinerolo in 1991 (Caire, Petitfils, Noone), were read in fragments. Brugnon's reconstruction of the island register (1992, a typescript at the Arsenal) and Caire's articles of 1988, 1997 and 1998 were not obtainable. Vauroy's reimbursement order, the deleted words in the minute of 23 December 1678 and the Saint-Lazare order of 17 August 1678 are known only through those authors' reporters.
• No original manuscript was consulted. The war-office registers (Service historique de la Défense, series A1), the K 120 file and the Arsenal registers were read through the editions; six letters were checked against photographs. Iung, Topin, Lair and Funck-Brentano differ on dates, folios and words, and where they differ the report says so rather than choosing on authority.
• The Bibliothèque nationale's Gallica site refused automated access throughout, which is why Mongrédien 1952 was read in snippets and the newsletter of 4 September 1687 only in Dabos's photograph.
• The probabilities are the author's judgements about the weight of the documents. The arithmetic that produces the headline is stated so that a reader who weighs the 1691 letter differently can redo it.

Sources

Manuscript series (read through the editions below)

• Archives nationales, K 120 (cartons des rois; Saint-Mars's file of letters received, 1669–1681) and K 129. Photographs of K 120 nos 132, 275, 277, 299 and 302 at masque-de-fer.com (Jean-Marc Dabos), harvested 14 September 2026.
• Service historique de la Défense, Vincennes, série A1 (the "Dépôt de la guerre" registers cited by Iung and Topin as "P. x, v. y"), vols 234, 618, 684, 779–783, 1034, 1242–1247, 1392, 1431–1432.
• Bibliothèque de l'Arsenal, mss. 5133 and 5134 (du Junca's registers of entries and exits).
• Bibliothèque nationale, papiers d'Estrades (Saint-Mars to d'Estrades, 1681, printed by Ravaisson III).
• Bibliothèque Sainte-Geneviève, ms. 1477, fol. 396 v° (manuscript newsletter of 4 September 1687 attributed to Louis Fouquet, bishop of Agde; photograph at masque-de-fer.com; printed by X. Azéma, Un prélat janséniste, Louis Foucquet, évêque et comte d'Agde, Paris, Vrin, 1963, p. 150, not seen).
• Bibliothèque nationale, Mélanges Colbert, vol. 282, fol. 276 v° (reimbursement of Captain de Vauroy's journey, 1669; found by S. Brugnon; quoted by Petitfils 2003, p. 244 n. 11; not seen).
• Register of burials of Saint-Paul, 20 November 1703 (destroyed 1871; facsimile in Topin, 5th ed., 1878, and Revue encyclopédique, 1 April 1894).

Editions of the correspondence

• Roux-Fazillac, Recherches historiques et critiques sur l'homme au masque de fer (Paris, an IX/1801). Bibliothèque municipale de Lyon, Numelyo, BML_00GOO0100137001102795544.
• J. Delort, Histoire de l'homme au masque de fer (Paris, 1825); Histoire de la détention des philosophes et des gens de lettres à la Bastille et à Vincennes, t. I (Paris, 1829). archive.org/details/histoiredeladte05delogoog.
• F. Ravaisson, Archives de la Bastille, t. III (Paris, 1866), t. VII–XI (1874–1880). archive.org/details/archivesdelabast03rava, archivesdelabast07rava, archivesbastille10rava.
• Th. Iung, La Vérité sur le Masque de fer (Paris, 1873). archive.org/details/lavritsurlemasq00jung.
• J. Loiseleur, Trois énigmes historiques (Paris, 1882). archive.org/details/troisnigmeshist00loisgoog.
• M. Topin, L'Homme au masque de fer, 6th ed. (Paris, 1883); English translation (London, 1870).
• J. Lair, Nicolas Foucquet, t. II (Paris, 1890). archive.org/details/nicolasfoucquetp02lairuoft.
• F. Funck-Brentano, "L'homme au masque de velours noir, dit le Masque de fer", Revue historique 56 (1894), pp. 253–303; Légendes et archives de la Bastille (Paris, 1901). archive.org/details/lgendesetarchi00funcuoft.
• E. Laloy, Le Masque de fer: Jacques Stuart de la Cloche, l'abbé Prignani, Roux de Marsilly (Paris, 1913). archive.org/details/masqueJacques00lalo.
• H. Griffet, Traité des différentes sortes de preuves qui servent à établir la vérité de l'histoire (Liège, 1769; ed. 1770). archive.org/details/traitedesdiffere00grif.
• G. A. Ellis, The True History of the State Prisoner commonly called the Iron Mask (London, 1826). archive.org/details/truehistorystat00elligoog.
• Journal du marquis de Dangeau, t. III (Paris, 1854). archive.org/details/journaldumarquis03dang.

Scholarship

• L. Batiffol, review of Burgaud and Bazeries, Bibliothèque de l'École des chartes 55 (1894), pp. 359–360. archive.org/details/bibliothquedel55sociuoft; Persée.
• A. Lang, The Valet's Tragedy and Other Studies (London, 1903). Project Gutenberg 2073.
• A. S. Barnes, The Man of the Mask (London, 1908). archive.org/details/manofmaskstudyin00barnuoft.
• M. Duvivier, Le Masque de fer (Paris, 1932). Not read.
• G. Mongrédien, Le Masque de fer (Paris, Hachette, 1952, 256 pp.), Gallica ark:/12148/bpt6k3357084j, read in 3,439 page-numbered snippets and the table of contents; "Deux documents inédits sur le Masque de fer", XVIIe siècle 17–18 (1953), pp. 55–58. archive.org/details/xviie-siecle_1953_17-18.
• S. Brugnon, "L'identité de l'homme au masque de fer", pp. 27–38, and B. Caire, "Eustache et son secret", pp. 41–49, in Il y a trois siècles, le Masque de fer. Actes du colloque international, Cannes, 12–13 septembre 1987 (Cannes, 1989; 2nd ed. 1996). Not obtained; known through Petitfils, Noone, Sonnino, Wilkinson and Madry.
• B. Caire, "Le secret d'Eustache Danger", pp. 176–207, and J.-C. Petitfils, "La double légende et le secret d'État", pp. 210–237, in La Maschera di Ferro e il suo tempo. Atti del convegno internazionale di studio, Pinerolo, 13–15 settembre 1991 (Pinerolo, Alzani, 1992). archive.org lending item pinerololamasche0000unse; search fragments only.
• S. Brugnon, Reconstitution du registre d'écrou des prisonniers de l'île Sainte-Marguerite à compter du 30 avril 1687 et jusqu'au 19 mars de l'année 1704 (typescript, 1992), Bibliothèque de l'Arsenal, fol. Z 2753. Not obtained.
• J. Noone, The Man Behind the Iron Mask (Stroud, Alan Sutton, 1988; rev. ed. 1994). archive.org lending item manbehindironmas0000noon; search fragments only.
• J.-C. Petitfils, L'Homme au masque de fer (Paris, Perrin, 1970). Not read; known through the reviews in Le Figaro littéraire (27 April 1970), Historia 289 (December 1970) and the Bulletin critique du livre français 296 (1970).
• J.-C. Petitfils, Le Masque de fer: entre histoire et légende (Paris, Perrin, 2003, 310 pp.; coll. tempus, 2004). archive.org lending items lemasquedeferent0000peti and lemasquedeferent0000peti_h1r0; about 1,050 search fragments; reviews by S. Genet (H-Net, October 2003) and D. Venturino (Dix-huitième siècle 37, 2005, pp. 645–646, Persée).
• P. Sonnino, "On the Trail of the Iron Mask: The Candidacy of Claude Imbert", Proceedings of the Western Society for French History 19 (1992), not obtained; "The Three Testaments of Cardinal Mazarin", French Historical Studies 37 (2014), pp. 421–436, doi:10.1215/00161071-2689643, abstract only; The Search for the Man in the Iron Mask: A Historical Detective Story (Lanham, Rowman and Littlefield, 2016, xiv + 269 pp.). archive.org lending item searchformaninir0000sonn; about 1,000 search fragments; reviews by J. R. Ruff (H-France Review 16, 2016, no. 258, h-france.net/vol16reviews/vol16no258Ruff.pdf), D. Parrott (French History 31, 2017, p. 122, doi:10.1093/fh/crw073, by summary) and Publishers Weekly (30 November 2015); interviews in The UCSB Current (27 April 2016) and Live Science (2016).
• S. Madry, "The Valet: The Marquis de Louvois's Invited Guest", Cahiers du dix-septième 17 (2016), pp. 1–27. Read in full.
• J. Wilkinson, The Man in the Iron Mask: The True Story of Europe's Most Famous Prisoner (New York, Pegasus, and Stroud, Amberley, 2021). archive.org lending item 9781643137421; publisher's pages and search fragments only.
• H. Thompson, The Man in the Iron Mask (London, 1987), and R. Macdonald, The Man in the Iron Mask (London, 2005). Search fragments only.
• J.-M. Dabos, masque-de-fer.com (2024–2025), 52 pages harvested, with photographs of the originals listed above.

Working files published with this case: the transcriptions and notes, the fixed points and the candidate file, the two scripts (fixed_points.py, which produces the candidate table, and probability.py, which produces the headline and its sensitivity), and the source texts.