Henry of Navarre to Ségur

01 The documents

Gallica serves the volume (ark:/12148/btv1b10035574w) over IIIF; the canvases are unlabelled, and the cipher pages sit at canvases 154 (f. 143), 257 (f. 233), 263 (f. 239), 325 (f. 288v, a slip attached to the letter of 1 April 1586) and 359 (f. 321, the key letter with interlinear decipherment). The 1585–86 letters are secretarial copies in one chancery hand: clear French with cipher passages of a few words to fourteen lines, the figures dotted apart, and a handful of tokens written in letters (pi, na, gne, Ra, H, L) or as a small square. Together: 461 figures over 102 distinct values, 25 letter-glyphs.

f.233  X 59 96 118 80 38 52 62 29 9 83 105 34 107 28 35 115 85 30 44 6 45 13 | 9 60 46 48 100 47 31 45 40 5 68 22 110 105 51 53 [s'il est possible]
       4 45 23 20 80 68 9 59 111 86 38 33 45 94 38 112 30 62 29 110 29 115 85 30 43 52 69 24 105 79 80 45 13 15 25 6 34 83 40 18 104 30 12 15 52 79 69 24
       38 15 84 39 9 20 14 38 [le plustost que faire se pourra ...]
f.288v gne 113 22 40 13 29 14 48 43 105 85 93 23 90 65 89 122 24 38 13 45 65 122 40 na 11 15 98 24 41 e 23 90 41 59 49 pe pu 13 46 10 106 105 ...

The 1583 letter at f. 143 turned out to be to Ségur, from a correspondent at Beaupréau, and its clear text describes the system: “vous y avez toutes choses tellement dispersées en lettres, syllabes, doubles, nulles et tout le reste qu’il est impossible de le descouvrir … Les noms sont en lettres et syllabes.” Letters, syllables, doubles, nulls, and names written with letter-signs: that is exactly what the 1585–86 key proves to be.

02 Solving

1. Two ranges. Values 1–53 carry 311 tokens, 54–123 carry 146, and only three exceed 123. Counting the upper range by residue mod 5 gives 60 · 30 · 22 · 20 · 14 for one alignment and near-flat counts for the other four: blocks of five with one dominant position, which is what an alphabetical syllabary (a e i o u, e commonest) produces. The structure was fixed before a single letter was known.
2. A structured annealer. Each number below the table is a free homophone (22 letters or a null, nulls capped); each block of five is one consonant with the vowel given by position; letter-glyphs are skipped; and the clear French around every cipher passage is kept as fixed context for a 16th-century French 5-gram model (Catherine de Médicis’ letters, Teulet, Labanoff; 9.4 M letters). Moves reassign or swap letter symbols and block consonants. Five of twelve seeds converged on one key.
3. Controls. Passages of Catherine de Médicis enciphered with random keys of the same shape, the same token counts and the same interleaving of clear text. With the table start guessed at 64 the method read one control at 90 % and one at 47 %; with the true structure (start 54, fourteen consonants) it reads a control at 96.6 % of letter tokens and 12 of 14 blocks, missing only x and f at one token each.
4. The blind key was already alphabetical. Without being told so, the annealer put a at 4–6, b at 7, c at 9–10, d at 11–12, e at 13–15, g at 18, h at 20–21, i at 22–24, l at 26, m at 27–28, n at 29–30, o at 31, p at 34–35, r at 38–39, s at 40–42, t at 43–44 and u at 45–47. Against the final key it had 86 % of the letter tokens and 77 % of the syllables right even with the table start wrong by two blocks. The text read capituler et conclure promptement avec ceux que vous adviserez … le duc Casimir … faire la levée … la faire marcher … deux mille reistres … dix mil escus … Vivarès ou Daulphiné at once.
5. Reading fixed the rest. et conclure = 52 62 29 9 83 105, secours = 110 62 46 39 42 and le duc Casimir = 80 68 9 59 111 86 38 force 62 = co, 59 = ca, 65 = de; bien = 56 14 29 gives 56 = bi: the table begins at 54 with b c d f g l m n p q r s t v. que vous and qu’il fix 99–103 = qua que qui …; Clervant fixes 119 = va; nece□aire makes the square ss, mi2e makes 2 = ll, lev1 makes 1 = ée; 52 = et, 53 a null. Every digit the French disputed was re-read at twice the working zoom; two were my errors (90 for 40 in les deux, 63 for 65 in de Clervant), the others stand as written and are noted below.

03 The key

Doubles: 1 = ée, 2 = ll, □ = ss, 3 unused here. Word-signs in letters: pi = est, Ra = faictes, qu = vous, na = affaires (or nouvelles), gne = Je at the head of f. 288v; gna, pe, pu, e, X, gla, H, L, Ne, qua and the numbers 180, 900 unread. The letter counts (three homophones for a e h i o u, two for the other consonants, one for g x z) are the design of the volume’s other Navarre key at f. 321 (Tomokiyo: a 12–14, e 20–23, i 30–32, syllables 63–142, nomenclature from 162), re-numbered.

04 Reading

f. 233, October 1585. Clear text in roman, cipher in bold, unread signs in braces; an English rendering follows each letter, cipher passages again in bold.

Seulement je vous prieray, si de ça me l’avez fait, {X} capituler et conclure promptement avec ceux que vous adviserez, s’il est possible avec le duc Casimir ou par son consentement, et faire la levée la plus grande et la faire marcher le plustost que faire se pourra. Car pour que quelque prompt secours que est necessaire, j’escry à Mr de Clervant de faire la levée de deux mille reistres selon qu’il est obligé {gla} la contracte avec les Eglises[?] … en l’an 1579[?], pour le moings quelque partie s’il ne peult le tout, et s’il n’a le moyen de ce faire, qu’il s’aide des dix mil escus de {180}, ainsy que ledit Sr de Clervant vous dira … que je soye bien tost adverty … du chemin qu’il tiendra en ce cas, evitant le danger … chemins le Vivarès ou Daulphiné pourroient estre propres, en nous advertissant de bonne heure.

Monsieur de Ségur. I am troubled at having had no news of you since your departure from England; I have written to you four times … Only I will ask you, if you have not already done so from over there, {X} to treat and conclude promptly with those you shall judge fit, if it is possible with Duke Casimir or by his consent, and to make the largest levy [you can] and have it march as soon as may be. For since some prompt succour is necessary, I am writing to M. de Clervant to make the levy of two thousand reiters which he is bound to {gla} [under] the contract with the Churches[?] … in the year 1579[?], at least some part if he cannot [raise] the whole, and if he has not the means to do it, let him use the ten thousand écus from {180}, as the said Sieur de Clervant will tell you, to which I pray you lend a hand, and that I may soon be advised of everything … of the road he will take in that case, avoiding the danger …; the Vivarais or Dauphiné roads might be suitable, giving us early warning. I pray you send my news to my brother Monsieur the Prince wherever he may be … Your affectionate master and perfect friend, Henry.

f. 239, Montauban, early 1586.

Seulement je vous diray de {900} n’espargner aucun moyen {gne} {gna} secourir en toute diligence, et remuez toutes pieces pour ung si necessaire effect, obligez-y tout. … [postscript, another hand, signed D.:] puis {e} on n’a eu de vou[s] {na}. Monsieur, je suis perpetuellement vostre serviteur. Nous esperons que Dieu benira vos desseins. Drak a faict une tres grande prise. C’est le premier homme de marine et des plus vaillans qui soit au monde.

Only I will tell you {900} to spare no means {gne} {gna} to succour [us] with all diligence, and to set every piece in motion for so necessary an effect; commit everything to it. And for the rest, make very certain account of me … [postscript, signed D.:] since {e} we have had no {news} from you. Sir, I am perpetually your servant. We hope God will bless your designs. Drake has made a very great prize. He is the first seaman, and among the most valiant, in the world. I kiss your hands most humbly as your most humble and most faithful servant. D.

f. 288v, slip with the letter of 1 April 1586.

{Je} suis en extreme peine de n’avoir eu de vos {affaires} depuis {e}; je ne sçay {pe} {pu} e[s]crire, ne à qui, ne en quel estat sont nos {affaires}. Je vous prie me mander de vos nouvelles, et faictes s’il vous est possible la plus grande levée qui ayt esté faite. Les deux mille reistres que Monsieur de Clervant est obligé fournir nous seroient tres necessaires, et mesmes en {H} {L}. Je vous prie en communiquer avec luy. Encores que nos ennemys n’ayent faict aucun effort contre nous et qu’il y ayt en eux plus de mauvaise volonté que d’execution, si est ce que l’assistance et secours {qua} sont tres necessaires. Nos places sont bien fortifiées. {Ne} avons envoyé … en Valen[ce?], dont long temps a nous n’avons eu nouvelles.

{I} am in extreme anxiety at having had no [word] of your {affairs} since {e}; I do not know {where} {to you} to write, nor to whom, nor in what state our {affairs} stand. I pray you send me your news, and make, if it is possible for you, the greatest levy that was ever made. The two thousand reiters that Monsieur de Clervant is bound to furnish would be very necessary to us, and even more so in {H} {L}. I pray you confer with him about it. Although our enemies have made no effort against us, and there is in them more ill will than execution, yet assistance and succour {qua} are very necessary. Our places are well fortified. {We} have sent … to Valen[ce?], from whom we have long had no news. Make, I pray you, always very sure account of the friendship and good will of your most affectionate master and perfect friend, Henry.

• 4 45 23 20 (avec) reads a-v-i-h, and 20 recurs where c is wanted in contra-h-te and Valen-h, while marcher, chemin and Daulphiné need 20 = h after 9/10 = c. The digits were re-read and stand; the encipherer’s slip of one line in the key table (10 → 20) is the simplest account.
• e-v-c-ri-re for escrire (46 for an s), nu-i-ne for peine (93 for 95), e-m-li-se-s for Eglises (28 for 18) and pie-be-s for pieces (66 for 60/61) are each one adjacent-value error, mine or the clerk’s.
• The interlinear 14 45 39 65 9 80 38 119 29 44 above line 5 of f. 288v has its caret after 111 in line 6: mon-si-[eur de Clervant] est obligé.

The rest of the volume, f. 333 and f. 366 (added later on 16 September 2026)

All 440 canvases of the volume were fetched and read on contact sheets. Cipher occurs on six leaves only: the four above, Tomokiyo’s ff. 321 and 333, and one leaf not on his list, f. 366, an autograph of Navarre to Ségur from the autumn of 1586 (on a fait avancer l’armée et les forces contre moy … d’user de toute la diligence que vous pourrez pour effectuer ce dont je vous ay chargé) with three short cipher lines at its foot. Its low figures and a letter-written sign gu put it in the ff. 233 key; the last line reads escri, the first two (ri-e-c-h-a-v-f-fe-n, re-de-du-pi-n gu la-que) look like names in the king’s own spelling and are left open. There is no further secretary letter in this key in the volume, so the word-signs stay unglossed. Nine other letters of Navarre in the volume are in clear.

f. 333 (La Rochelle, 10 July 1586, a news-sheet in the third person signed Henry) was transcribed and read with Tomokiyo’s f. 321 table: aussi tost que les reistres marcheront il montera à cheval; […] et son frere sont de bonne volonté; il fault avancer la levée et la faire marcher le plustost qu’on pourra. The reading adds 24 = f, 34 = l, 36 = m and 61 = et to his table, and shows that key to be the ff. 233 key shifted by eight in the letters (a 12–14 … u 53–55, et 60–62) with h and j rows added to the syllabary. The free annealer on f. 333 alone (58 letter tokens) did not converge, and neither did it on a matched control of that size; but an annealer that builds in the chancery design itself (homophones in alphabetical order, so the letter key is a vector of homophone counts) re-derives the f. 333 letter key blind, every seed agreeing, with 93 % of the letter tokens right against Tomokiyo’s table (90 % on a matched alphabetical control), the syllable rows being the weak part at that size; the key rests on the f. 321 interlinear decipherment and on the French it yields here. f. 321 itself has every figure struck through by the decipherer and was not re-transcribed.

05 Context

Jacques de Ségur-Pardaillan was Navarre’s agent in England, Denmark and the Empire from 1583, and Claude-Antoine de Vienne, sieur de Clervant, his colleague for the German levy; Duke John Casimir of the Palatinate was the prince through whom Huguenot armies of reiters had been raised since 1567. Navarre’s October 1585 letter follows the bull of 9 September excommunicating him and the summer campaign that cost him most of his Guyenne places; the “contract with the Churches” and Clervant’s two thousand reiters refer to the standing obligation under which the Huguenot assemblies financed the German horse, and the ten thousand écus from [180] is most naturally the English subsidy (an inference, not a reading). The February and April 1586 letters belong to the long wait for the army that was finally assembled in 1587 under the baron of Dohna and destroyed at Auneau. The postscript’s Drak is Drake’s West Indies voyage: Santo Domingo fell on 1 January 1586, and the news reached France in February.

06 Method and files

1. Images from Gallica IIIF (segur/fetch_gallica.py); transcription at full resolution with 2× re-reads (ct_233.txt, ct_239.txt, ct_288.txt, ct_143.txt).
2. Mod-5 residue test and structured annealer with clear-text context (segur/solve.py).
3. Matched controls with random keys of the same shape (segur/control.py, eval_control.py).
4. Key file and renderer (key_v4.json, decode.py); edited reading with token notes (reading.md).
5. Sweep of the volume (fetch_sweep.py, mksheets.py); f. 333 and f. 366 transcriptions (ct_333.txt, ct_366.txt); Tomokiyo’s table extended (key_321.json); design-constrained annealer and controls (solve_mono.py, control5.txt); word-sign likelihood ranking (gloss_check.py, gloss_check.txt).

Checked: transcription, structure test, three controls, convergence over seeds, alphabetical order of the recovered key, word-level French of every passage. Not checked: Tomokiyo’s f. 321 table figure by figure against the struck-through leaf; the word-signs; the historical glosses; the second Ségur volume (500 Colbert 402, not digitised); BnF fr. 17820 is an eighteenth-century fair copy in clear. The sender attribution and the readings marked [?] should be verified before they are repeated.

07 Sources

• BnF, Cinq cents de Colbert 401, Négociation de M. de Segur, baron de Pardaillan, pour le roi de Navarre avec les Princes protestans d’Allemagne, 1584–1588, I, Gallica.
• S. Tomokiyo, “Henry III’s Figure Ciphers” and “Different Version” in French Ciphers during the Reign of Henry III; “Unsolved Historical Ciphers” (cryptiana).
• Paul Destray’s and La Ferrière’s editions used for the French language model are listed in ducroc/NOTES.md.

Analysis scripts, transcriptions and the key are in the repository (segur/). Manuscript rights: Bibliothèque nationale de France.