How a Decipherment Becomes Evidence

Copiale: the ordinary letters were the surprise

The Copiale manuscript combines familiar Roman letters with unfamiliar symbols. Knight, Megyesi and Schaefer first tested the familiar letters as the message-bearing alphabet. That hypothesis failed. Comparing neighboring contexts helped reveal groups of homophones among the unfamiliar signs; German language hypotheses and human interpretation guided the search.

1. Source: preserve diacritics and distinguish symbol types before counting them. A familiar-looking letter need not play its familiar role.
2. Competing hypothesis: treat ordinary Roman letters as ciphertext, then as separators. The successful reading uses unaccented Roman letters as word boundaries. Calling them meaningless nulls loses that function.
3. Mixed key: the system includes multi-letter values such as CH, SCH and ST, logograms, and a colon that doubles the preceding consonant. A one-letter substitution cannot represent the whole manuscript.
4. Evidence: test mappings across occurrences and read the resulting German in context. Some material remained unresolved in the 2011 paper; a successful passage does not certify every glyph.

Try it: in the workbench, enter an original exercise g1 g2 g3 g1 with g1→A, g2→CH and g3→space. Compare that typed key with forcing every sign to one letter. This exercise is invented; it is not a transcription from Copiale.

Source: Knight, Megyesi and Schaefer, “The Copiale Cipher” (2011), especially the transcription and decipherment discussion, PDF pp. 2–7. The paper and its manuscript figures supply the historical evidence.

Mary Stuart: preserve the dot, then test the story

The 57-letter collection published in 2023 contains more than 150,000 transcribed cipher symbols and about 50,000 plaintext words. Its 219 graphical symbol types include distinctions that cannot be compressed into a 26-letter alphabet.

1. Source: record the archive, shelfmark and folio. The collection spans BnF Français 2988, Français 20506, Cinq Cent de Colbert 470 and Français 3158. Keep letter identity separate from a modern filename.
2. Reading alternatives: diacritics distinguish symbols. If a mark might be a dot or damage, record both readings and its image region. Do not silently merge the two types to make a solver work.
3. Language hypothesis: early Italian expectations gave way to French. A modern-English score is not evidence against a French reading.
4. Search and interpretation: simulated annealing recovered letter homophones. Names, word codes, repeat and erase controls, and other nomenclature required further manual work. Preserve the literal reading before normalizing spelling or translating.
5. Independent corroboration: surviving plaintext versions and historical correspondence help test the interpretation. They provide evidence beyond the language score used to select a candidate.

Try it: transcribe an invented pair as g17 g17-dot, save a span with both alternatives, and record why one is preferred. Revise the source and inspect the earlier annotation: its original location is retained. This tests provenance, not recovery of a historical letter.

Source: Lasry, Biermann and Tomokiyo, “Deciphering Mary Stuart’s lost letters from 1578–1584” (2023), §3.2, §4.2, §5.1 and Appendix A; PDF pp. 9–13, 16–26, 32 and 96–98. Figure 2 shows letter F38 and the importance of the source image.

Z340: replay the first 153 symbols

This is an authentic source-backed replay using the published key and the first nine rows of the November 1969 cipher. The grid uses case-sensitive ASCII symbol IDs from David Oranchak’s repository, not facsimiles of Zodiac’s handwriting. It retains all 63 symbol types.

Start at the upper left, move down one row and right two columns, wrapping independently at both edges of this 9 × 17 section. Compare that route with reading rows in order. The key stays fixed.

Loading the published first-section fixture…

Do not repair the result invisibly. The diplomatic output includes FAN, BRINGO, BECAASE and PARADLCE. A normalized interpretation reads FUN, BRINGS, BECAUSE and PARADISE. Those are editorial corrections, not different outputs from the key.

This route does not solve all 340 symbols. In the second nine-row section, the published reconstruction skips LIFE IS and corrects a shifted segment on row 15. The final two rows include reversed words. The eventual placement of LIFE IS is subjective. The physical procedure used to create the cipher remains uncertain.

The solvers submitted their report to the FBI on 5 December 2020; the public announcement and FBI confirmation followed on 11 December. The replay verifies a published solution mechanically. It does not reproduce the original search or turn a plausible sentence into independent historical proof.

Sources: Oranchak, Blake and Van Eycke, “The Solution of the Zodiac Killer’s 340-Character Cipher”, §5.4–5.5, figures 43–44 and 53–54; solver’s ciphertext and decoded-grid transcription. Download the exact replay fixture and provenance.

What the comparison teaches

Copiale challenges the assumed role of a symbol. Mary Stuart challenges transcription, language and key structure. Z340 challenges reading order and makes exceptions visible. Each case needs a different test. The workbench preserves those hypotheses and their evidence instead of treating readable output as a verdict.

Build an evidence record · Use Cipher Detective · Explore cryptanalysis techniques