25 · 11 · 39 · 4 · 4 · 10 · 3 · 54 · 50 · 19 · 1 · 18 · 1 · 5 · 9 · 58 · 15 · 1
4 · 17 · 1 · 42 · 32 · 77 · 23 · 75 · 6 · 3 · 18 · 20 · 36 · 8 · 21
5.3.27.38.32.14.21.8.66.8.70.39.5.9.12.18.2.3.56.5.1.7.3.2.13.19.3.25.9.3.16.6

London · book cipher on the author's own printed book · 1652–53 · octastich read

Thomas Urquhart’s Cyphral Octastich and Distich

The Jewel (1652), final leaf · eight lines and a “Decagram”, 285 numbers to 201 · Logopandecteision (1653), “The Cyphral Distich”, 64 numbers to 70 · Schmeh’s Top 50, no. 28

The octastich is read in full from the EEBO-TCP text of The Jewel and a fresh transcription of its 285 numbers from the 1983 edition, without using the plaintext published by Vals AI on 31 August 2026; 238 of 284 numbers are first-occurrence hits against 0.43 for both controls. The distich rule does not reproduce.

Daniel Bourdeau · solved

Summary. Urquhart printed a verse under the octastich saying the key was “gather’d out of my Exskybalorum” and that he loved “an aphaeresified treason”: the cipher takes word-initials from the book it is printed in. The k-th number, counting straight through the poem, is a word index into physical page k of the 1652 Jewel, and Urquhart took the first word on the page that began with the letter he wanted. That habit makes the rule testable without any plaintext: whether number n on page k is the first word of its own initial is a fact about the page. The public transcription (Schmeh, 2015) turned out to be short by thirteen numbers; the 1983 Jack and Lyall edition, photographed on the HCPortal record, gives 285. Decoded one number per page against the TCP text, 238 of 284 are first-occurrence hits, against 0.43 for the same numbers shuffled and 0.44 for random pages, and the poem reads: Great Lord, mantaine that regal familie, / Whereof King Charls the second is the head, / And grant that he may beare the supreme sweigh / Where English, Scots and Irsh are borne and bred; / And […] this usurp’d authoritie / Reigne in his royal predecessors stead: / Let him be our sole Cesar, Artur, Hector, / Our Emperour, King, Monarch and Protector. / Amen, so be it. Vals AI published this rule and plaintext on 31 August 2026 as a Claude Fable 5.1 result; here it is reached from the text and the numbers alone, and their two caveats are traced to defects of the digitised copy. The distich claim, i-th number into the i-th Proquiritation, gives 34 of 64 first-occurrence hits, chance for texts that short, and no English. Status: octastich read, ten letters of line 5 excepted; distich open.

01 The two poems

The distich stands under the heading “THE CYPHRAL DISTICH” after the thirty-two Proquiritations that close Logopandecteision, followed by six lines promising that “an honest skilful man” will find in it “his own heart’s wishes, and the Author’s minde”. Reticuli Labs showed on 1 September 2026, and the 1834 Maitland Club reprint confirms, that this page is in the 1834 Works (p. 417) and not in the British Library’s 1653 copy. The octastich stands on the last leaf of The Jewel (1652), a leaf missing from every digitised copy; its public transcription is Klaus Schmeh’s, from a scan sent him in 2014, and the HCPortal record holds a scan with the verse beneath it, which a reader quoted in 2019:

To this Octastick if you will subjoyn / A Decagram of this same stuff of mine, / All gather’d out of my Exskybalorum, / You’ll find a Rule by which, with great decorum, / You may most comfortably regulate / Your actions, thoughts and speeches; and know that / I love an aphaeresified treason / Better then any prosthesized reason.

Aphaeresis is the loss of a word’s first letter. Urquhart says, in his way, that the cipher is made of initials taken out of his own book.

02 A test that never sees the plaintext

A book cipher of this kind is normally judged by whether the letters it yields make sense, which invites the objection that a determined reader can make sense of anything. Urquhart’s habit removes that objection. If he scanned each page from the top and stopped at the first word beginning with the letter he needed, then for the right page the number he wrote is the index of the first word of its own initial. That is a property of the number and the page alone. For a wrong page, or a wrong rule, it holds at the base rate, which in this book is about 43 % for a random page and index because small indices are common and the first few words of any page are first occurrences by construction.

The EEBO-TCP transcription of The Jewel was parsed into word lists per page. The 1652 printer numbered pages 34–35, 38–39, 42–43 and 46–47 twice and skipped 60 and 110, so pages are counted physically from p. 1, and three duplicated page breaks that the TCP carries for re-shot images are removed. Schmeh’s 272-number transcription, the only one in circulation since 2015, omits the run-over lines of stanzas 2 and 6, one figure, and misreads one; a dynamic-programme alignment that may skip pages still placed it at 231 first-occurrence hits in 267 and read five lines. The photographs of the 1983 edition on the HCPortal record give the full 285, which then decode straight, one number per page.

RunFirst-occurrence rate
The 285 numbers, one per physical page0.838 (238/284)
Same numbers shuffled, 200 trialsmean 0.434, max 0.491
Numbers in order, pages randomly permuted, 200 trialsmean 0.436, max 0.509
Schmeh’s 272 numbers, aligned with page skips0.865 (231/267), vs 0.686 shuffled and re-aligned
Distich, i-th number into i-th Proquiritation0.53 (34/64), no English
Distich, shuffled, against pages of The Jewel, 100 trialsmean 0.50, max 0.62

Line 1 alone makes the case: 28 of its 33 numbers are exactly the first word of the needed initial on physical page k, and the other five are one word off (25 for 26, 18 for 17, 23 for 22, 75 for 74) or a TCP variant, which is the size of difference one expects between a modern tokenisation and a 1652 page.

03 The reading

Capitals are first-occurrence hits, lower case a word a few places off in the TCP tokenisation, _ the one page (physical 158) whose text the TCP does not carry. The right-hand column is the reading.

wREATLORDMApTAInETHATREGtaFAMILIE          Great Lord, mantaine that regal familie
WtEREOFKINGCHARLStHaStCONtISTaEHiAD        Whereof King Charls the second is the head
ANDGRAnTTHatHEDAYBEARETHEFUPaEMESgEIGS     And grant that he may beare the supreme sweigh
WHERDENGLIsHSCoTSANDIOSoAREBORNEANDBRED    Where English, Scots and Irsh are borne and bred
AaDCOnERTHTO_aIttSUaPDApTHORIEIE           And [conerthto?] this usurp’d authoritie
REIGtEINMISnOtALPREDECESSORSSifAt          Reigne in his royal predecessors stead
SETHpLBEOURSOLtCESAtAmtaRHECBOR            Let him be our sole Cesar, Artur, Hector
OeREMPoRPUwJINGMONARCHANDPROTECTOR         Our Emperour, King, Monarch and Protector
AMsNSOwLtt                                 Amen, so be it

The reading agrees with Vals’ published plaintext throughout, including the nine letters of line 5 that neither decoder can read (conerthto, pages 149–157, eight of them exact first-occurrence hits) followed by the one letter on the TCP’s near-empty page 158. Their “one page used twice from position 159” is that page. Sweigh is Scots swey, “sway”; mantaine, Charls, Cesar and Artur are period spellings. A royalist prayer for Charles II, written in London in March 1652 by a prisoner of Worcester, hidden in a book dedicated to the vindication of Scotland: entirely in character.

04 The distich does not follow

The same test applied to the distich under Vals’ Proquiritation rule, using Reticuli’s tokenised Proquiritations, yields 34 first-occurrence hits in 64 and the string NTMAMNDN_OHTISEABRPTTSGSNTAATVTD, not O God uphold King Charls the second. Texts of thirty to a hundred words give high base rates, so 34 of 64 is chance. Page-index variants on either book, with either numbering, with the second line restarting at page 1, all sit inside the shuffled control. Reticuli’s ten infeasible positions stand. The distich is a different construction, or the same construction on a text not yet tried, and it remains unread.

05 What remains

06 Method and reproduction

In urquhart/: ct.py holds both transcriptions of the octastich; pages.py parses the TCP XML of A95749 into pages under both numberings; decode_1983_full.tsv gives every position with page, word, letter and first-occurrence flag; align.py is the alignment used on the defective transcription; the controls are in the notes. The TCP text comes from the Text Creation Partnership’s GitHub mirror; the 1983 page photographs from the HCPortal record, whose API answers at api.hcportal.eu.

07 Sources