01 The letter

The copy is headed “Lettre de M. Bayard a M. le Mal de Montmorency”, dated “5 Janvier 1526”, with the source reference “V. 41. f. 4” in the margin. It is signed “Gilbert Bayard”, “A Tollede le ve jo. de Janvier”. It reports Bayard’s talks with the viceroy Lannoy, with Lallemant and with Don Hugo de Moncada on the four points still open before the treaty of Madrid: the hostages, Fuenterrabía, and the sovereignty of the Constable of Bourbon.
The cipher comes where Bayard tells how the Imperial side had asked whether he had spoken to any Italian ambassador: “…et qu’il y avoit neuf moys que je n’avoye salué celuy de Ferrare a Calatheu [cipher] comme j’espere vous dire dimenche.”
02 The cipher
There are 56 signs of 29 kinds (measured on signs.txt): Greek-like letters (Φ, θ, ξ, φ, γ, ω, ∂), figures (7), Latin letters (n, o, R, b, x, k, e), triangles filled and open, a barred I, a circled m, and an m with a superscript m. With 29 kinds in 56 signs there must be homophones, nulls or code signs. Several of the shapes (7, R, n, θ, x, Δ, ε, ω, and the “ny” and K signs that are nulls in Raince’s key) belong to the pool used by French ciphers of 1525–26. The copyist drew them by imitation, so small differences between shapes may not be meaningful.
03 What was tried
- Print. Champollion-Figeac, Captivité du roi François Ier (Paris 1847), no. CCXVIII, pp. 462–463, prints the letter from “copie de Bréquigny, vol. 91”. His text has “…que j’avoye salué celui de Ferrare. A ce soir…”, with no mention of a cipher.
- Calvimont’s key (same volume, f. 67, key recovered by Norbert Biermann). Its signs were matched by shape (7 = m, 8/∂ = a, 5 = e, 6/b = t, barred I = d, ω = l) and fixed in a solver (simulated annealing). The result scores worse than with no values fixed and gives no French.
- Raince’s 1526 key (Tomokiyo, as re-measured for the Raince write-up), with the nulls removed. It gives −3.5 per sign under the 1530s French model, against −1.4 for the free solution, and the output is gibberish. The shapes are shared, but the values are not.
- Ciphertext only. Two homophonic annealing runs against a 5-gram model of 1530s French despatches: one free, and one allowing homophones only for common letters, with 24 restarts. The best restarts share no text. At 56 signs the problem is underdetermined.
04 What would finish it
- The original letter (“V. 41 f. 4”, probably a Béthune or Brienne volume). Its signs would be clean, and a decipherment may be written on it.
- Another cipher letter of Bayard’s from the 1525–26 embassy in Spain, or a key for Montmorency’s correspondence with him.
05 Sources
- BnF, Clairambault 325, f. 84r–v (Gallica); DECODE R2285.
- A. Champollion-Figeac, Captivité du roi François Ier, Paris 1847, no. CCXVIII (archive.org).
- S. Tomokiyo, “French Ciphers during the Reign of Francis I”, cryptiana, section “BnF Clair.325”, including Biermann’s Calvimont key and the Raince 1526 key.
Transcription, solver scripts, notes and profile: targets/bayard1526/.
