At a glanceExisting decipherment · partial (77% of the cipher)
- The documents
- About 850 American messages, 29 January–28 September 1932, copied in Tokyo as typed English clear text under the American heading. About 70% are US Navy traffic, 25% State Department, 3% War Department; the rest are ship-movement bulletins.
- Where
- Sixteen bundles in the records of the Navy Minister’s secretariat (Kobun bikan, Showa 7), National Institute for Defense Studies, Tokyo, online through JACAR, Ref. C05022467800 to C05022469300, 796 images. The images are all rights reserved and not reproduced here.
- The cipher
- American, but not in the file: no cipher groups were copied, only the recovered words, with the missing ones left blank or reduced to their first letters, and notes such as “what follows: code not deciphered”.
- How it was read
- It was not read here. The Japanese Navy’s 1932 readings were transcribed (bundles 1–8) and set against the State Department text printed in FRUS, word by word.
- What it shows
- Tokyo recovered about 77% of the words of 53 State Department telegrams (45–95% by telegram). The paragraphs left unread are the most sensitive ones. The wrong words are alphabetical neighbours of the true ones, which fits a one-part code.
- Still open
- The confidential passages; which code it was (the Gray code is a compatible hypothesis, not a proof); a full transcription of bundles 9–16.
- Credit
- Brice Dudout (Europe chiffrée) found the file, inventoried it, transcribed and compared it, and designed the one-part-code test, working with Claude agents; he sent it on 29 October 2026. American texts: FRUS (US government, public domain).
01 The documents
Each copy is typed in English under the copied American heading (sender, addressee, number, word count, date, Tokyo time), with a handwritten Japanese title and distribution stamps, among them the round seal of Commander Nakasugi Hisajō, head of the Naval General Staff’s cryptanalytic annex according to Kotani. About 70% of the lines are US Navy traffic (Asiatic Fleet, Cavite, Yangtze Patrol, 4th Marines), 25% State Department (Peiping legation, Shanghai and Nanking consulates, Tokyo embassy) and 3% War Department. Most State Department telegrams from China went over Navy radio and were intercepted in Japan, about forty at Hiratsuka; cables of the Tokyo embassy carry Japanese telegraph marks and probably came from the telegraph administration.
Brice Dudout inventoried all sixteen bundles (about 850 messages), transcribed bundles 1–8 image by image, and compared 53 State Department telegrams with FRUS 1932 vol. III–IV and Japan 1931–1941 vol. I. The working notes are in French.
02 The code
The American code is not seen: the file holds the Japanese result, not the cipher groups, and no copy of the State Department’s own file with its groups was available. What the file allows is a test of the code’s nature from the errors in the Japanese reading.
- A one-part code. Wrong readings share the initial of the true word in 63 of 137 cases (46%) against 6.3% when the FRUS words are shuffled (10,000 draws); truncations have the right initial in 95 of 119 (80%) against 6.1%; p < 1e-4 for both and in each of the four batches. Wrong words are alphabetical neighbours (functioning becomes “fundamental”, admission “admonition”, million “military”): median alphabetical distance 0.10 against 0.30 for controls, replicated on the held-out batch d. The protocol was written before the calculation (
stats_initials.py). - Which code. Compatible with the Gray code, the State Department’s ordinary one-part code since 1918, used without superencipherment; Kotani credits Nakasugi’s office with solving it. This is a hypothesis: neither groups nor a Gray copy are available.
- The Japanese code book. Of 1,972 distinct FRUS words, 1,107 are always read right, 357 sometimes, 270 only replaced by another word, 181 never read and 57 only omitted (
reconstructed_book.tsv). The errors spread over the whole alphabet.
03 What Tokyo read
Word for word, the Japanese copy renders on average about 77% of the FRUS text, from 45% to 95% by telegram. The four batches were measured somewhat differently (sense recovered, identical words, manual count), so the figure is an order of magnitude. Structure, figures, dates and common names are nearly always right; losses fall on small, rare or abstract words and unusual proper names. Wrong values persist for months: believe is rendered “calculate” from March to September.
Confidential passages. In six telegrams the unread part begins at a paragraph break and covers the most sensitive matter: Forbes, Tokyo, 1 February; Johnson, Peiping, 12 July (“28 words Enciphered”, “251 words Enciphered”); Grew, Tokyo, 2 September, no. 223 (45 words on the Vice Minister’s confidence); Johnson, 7 September, no. 1078, on the Lytton report, where only the first paragraph is read and the nine paragraphs of political analysis are marked as not deciphered. One hypothesis is an ordinary code largely read and an enciphered confidential code for sensitive matter.
US Navy. The Asiatic Fleet traffic was mostly in clear. Sixteen dispatches of late January to 12 March 1932 carry an office-number group “00” followed by a five-letter group, very probably the naval code indicator, so coded Navy dispatches were probably read as early as February 1932 (to be confirmed on the originals).
Where the Japanese copy corrects FRUS. Cunningham, Shanghai, 9 February: FRUS “Rear Admiral Shimottowa [Shiozawa?]”, Japanese copy “Shiozawa”; Shanghai no. 222, 6 May: “about 24,000 men of the 3d Division” against FRUS “about 2400 men of the 14th division” (probably right). The reverse also occurs.
04 What remains uncertain
- Confidential passages (for example Grew no. 223, last 45 words; Johnson no. 1078, paragraphs 2–10): not read in 1932, and no groups were copied, so nothing can be read from the file; the American text is known from FRUS.
- The code: needs the State Department file copies (NARA RG 59, decimal files in
frus_pairs.md) with their groups. - Bundles 9–16 are inventoried, with the comparison against FRUS in the gaps files, but not transcribed in full; that needs the JACAR images, which cannot be redistributed.
- Rates are measured differently from batch to batch; the Japanese marks A.C., A.G., A.O., D.C. and G.H. are not explained; Kotani’s pages and Moriyama’s articles are known through a review and Japanese Wikipedia, not read directly.
Full texts
The transcription and the decryption of every document on this page, so that the reading can be used and checked. Short files are shown here; all are in the repository.
米特情 copies, bundles (1)-(8), Jan-Apr 1932 read in part
US Navy Asiatic Fleet traffic and State Department telegrams on the Shanghai fighting, the League of Nations (Lytton) inquiry and Manchukuo, as read in Tokyo.
Builds coherent sentences, but under the read bar of 95% of words read as sense; the holes are marked in the reading.
Transcription: none of the cipher. No cipher groups survive: the Japanese copied only the clear text they had recovered.
Decryption: C05022467800.md
Decryption: C05022467900.md
Decryption: C05022468000.md
Decryption: C05022468100.md
Decryption: C05022468200.md
Decryption: C05022468300.md
Decryption: C05022468400.md
Decryption: C05022468500.md
05 Sources
- JACAR (jacar.go.jp), Ref. C05022467800 to C05022469300; copy of Grew no. 223: C05022469300, images 0766–0767; Johnson no. 1078: same item, image 0777. Images not reproduced.
- Foreign Relations of the United States, 1932, vol. III (The Far Eastern Crisis) and vol. IV; Japan, 1931–1941, vol. I (Office of the Historian).
- Navy Department, Communication Instructions 1929, Part IV, arts. 3016–3023.
- Kotani Ken, Nihongun no interijensu (Kōdansha, 2007), ch. 3; English edition Japanese Intelligence in World War II (Osprey, 2009). Moriyama Atsushi, article in Kokusai kankei hikaku bunka kenkyu 3-1 (2004).
- Brice Dudout, his write-up (site currently offline).
Folder: targets/tokyo1932/ (notes, the inventory, transcriptions of bundles 1–8, the FRUS comparison and the one-part-code test, as contributed by Brice Dudout).