Dossiers / Shanghai · Peiping · Tokyo · Washington · 1932

Dossier · Japanese intelligence and American codes

1932: the Japanese Navy reads the State Department

From January to September 1932, through the Shanghai crisis and the League of Nations inquiry in Manchuria, a small office of the Japanese Naval General Staff circulated, under the heading “米特情 極秘” (“special intelligence, America, top secret”), the clear text of intercepted American telegrams. By setting these copies against the text the State Department itself published, we can measure, telegram by telegram, what Tokyo read, and what it did not.

Dossier by Brice Dudout · reading and comparison with Claude · documents: JACAR (Military History Research Center, NIDS, Tokyo) · American texts: Foreign Relations of the United States · version of 28 October 2026

1932: the Shanghai crisis and Manchuria

In September 1931 Japan’s Kwantung Army occupied Manchuria. On 28 January 1932 the fighting spread to Shanghai, where the US Asiatic Fleet, the Yangtze Patrol and the 4th Marines protected the International Settlement. Secretary of State Henry L. Stimson refused to recognise Japan’s conquests; the Lytton Commission investigated for the League of Nations from February to September; Japan recognised Manchukuo on 15 September. Throughout those eight months, the telegrams between the American posts in China, the Tokyo embassy and Washington were one of the threads of the crisis.

On the Japanese side, Herbert Yardley’s revelation in 1931 that the Americans had read Japanese telegrams during the Washington Conference of 1921–1922 came as a shock. The Navy already had a service of its own: according to the historian Ken Kotani, in 1929 it created, within the intelligence branch of the Naval General Staff (軍令部), a cryptanalytic “annex” (別室) headed by Commander 中杉久治郎 (Nakasugi Hisajirō), with a handful of officers and typists. The round seal “中杉” appears on many of the documents studied here.

The file

The documents are sixteen bundles filed among the records of the Navy Minister’s secretariat for 1932 (公文備考 昭和7年, “事件付属（米特情 極秘 号外 外）” (1)–(16)), now available online from the Japan Center for Asian Historical Records (JACAR), ref. C05022467800 to C05022469300. They run to 796 images and about 850 messages, from 29 January to 28 September 1932. The Japanese numbered them 32 to 548 in the “米特情 極秘” series, plus an unnumbered series of about 330 ship-movement bulletins.

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. No cipher groups are copied: there is only the recovered clear text, missing words left blank (“____”) or reduced to their first letters (“co....”), and working notes such as “以下未解読暗号” (“what follows: code not deciphered”) or “不明ノ個所多キモ参考ノタメ” (“many obscure passages; circulated for information”).

Of the 828 lines of our inventory, about 70 % come from the US Navy (Asiatic Fleet, Cavite, Yangtze, Marines), 25 % from the State Department and 3 % from the War Department. The Navy dominates from February to April, during the fighting at Shanghai; from July almost nothing remains but the State Department, around the Lytton Commission and Manchukuo.

How the telegrams reached Tokyo

• Radio intercept. Most State Department telegrams from China went over Navy radio (Cavite station, the cruiser Houston, the Peiping legation radio). The Japanese copy the American preamble, including the serial number that FRUS also prints. Most copies are noted as received in Tokyo; about forty at Hiratsuka, three at the Navy Communications School.
• The Japanese telegraph network. Cables of the Tokyo embassy, the military and naval attachés and the Nagasaki consulate carry a commercial preamble and Japanese marks (“D.C.”, “G.H.”, “G.N.”) found only on this channel. These copies very probably came from the telegraph administration, not from intercept.
• The same day. The Japanese copy of Shanghai telegram no. 121 (27 February) is timed 21:29 Tokyo; Washington logged its receipt at 9:36 a.m., that is 23:36 in Tokyo (FRUS 1932, III, doc. 491).

What Tokyo read

We found in FRUS the printed text of 53 State Department telegrams copied by the Japanese, from February to September 1932. Word for word, the Japanese copy renders on average about 77 % of the text, from 45 % to 95 % depending on the telegram. The four batches were not measured in quite the same way (sense recovered, identical words, manual count), so the rate is an order of magnitude. Structure, figures, dates and common proper names are nearly always right; the losses fall on small words, rare or abstract words, and unusual proper names, spelled out in pieces.

The same errors recur for nine months: believe is rendered “calculate” from March to September, his “trusty”, question “realization”. The service worked with a single reconstructed code book, whose wrong values it corrected only slowly.

The confidential passages are not read. In six telegrams the unread part begins at a paragraph break and covers exactly the most sensitive matter. For instance:

• Forbes, Tokyo, 1 February: the only paragraph that FRUS paraphrases, as the Department did for its confidential codes, is a blank in the Japanese copy; the rest is read at about 60 %;
• Johnson, Peiping, 12 July: the copy notes “(28 words Enciphered)” and “(251 words Enciphered)”, that is, the source he names and his personal judgements;
• Grew, Tokyo, 2 September: the forty-five unread words are the Vice Minister for Foreign Affairs’ confidence;
• Johnson, 7 September, on the Lytton report: only the first, factual paragraph is read; the note “以下未解読暗号” covers the nine paragraphs of political analysis.

It looks as though these telegrams mixed two systems: the ordinary code, largely read, and an enciphered confidential code for sensitive matter. This is a hypothesis, which only the American file copies, with their groups, could test.

Which code?

The cryptanalyst saw only code groups, not letters. If his wrong readings nevertheless begin with the same letter as the true word, the order of the groups must have followed the alphabetical order of the words: an unknown group lying between two known groups falls between two dictionary words. He then gives its beginning (“co....”), or takes a neighbour: functioning becomes “fundamental”, admission “admonition”, million “military”. This is the signature of a one-part, alphabetical code being reconstructed.

The gap is highly significant (p < 10⁻⁴) and appears in each of the four batches. The measurement protocol was written before the calculation. The result is compatible with the Gray code, the State Department’s ordinary code since 1918, a one-part code used without superencipherment; Kotani indeed credits Nakasugi’s office with solving the Gray. It does not prove that the code was the Gray: other one-part codes existed, and neither the groups nor a copy of the Gray are available to check.

And the US Navy?

The Asiatic Fleet traffic copied by the Japanese was mostly sent in clear: ship movements broadcast from Cavite, Yangtze Patrol reports, press summaries. It is rendered almost without gaps, with only the errors of Morse reception.

Sixteen dispatches from late January to 12 March 1932 are the exception. Under the US Navy’s Communication Instructions 1929, the first group of a dispatch gives the originating office number and the day of the month, and office number “00” is reserved for code dispatches. All sixteen carry a “00” group, followed by a five-letter group of a distinctive form, which the Japanese copy as it stands before giving the rest in clear. That group is very probably an indicator of the naval code. If this reading is right, the Japanese service was also reading coded US Navy dispatches as early as February 1932. The American originals would be needed to confirm it.

Where the Japanese copy corrects FRUS

The Japanese copy is a second, independent witness of the same groups as received. Three gaps are common to both texts: the group had been garbled in transmission. Elsewhere, the Japanese reader sometimes reads better than Washington:

The reverse also happens: the negative in “have not yet entered Harbin” disappears, Chinese becomes “Japanese”. The Japanese reader of 1932 could thus be misled on specific points.

What these documents tell us

That the Japanese Navy read American traffic before the war is not new. Kotani writes in one sentence that Nakasugi’s office read, during the Shanghai Incident, the telegrams of the Shanghai consulate general to the State Department, and credits it with solving the Gray code; according to the historian Moriyama Atsushi, in 1941 Japan read most State Department telegrams, except the strip cipher. We have found no publication that uses these sixteen bundles or compares their texts with FRUS.

What this dossier adds, for the first time as far as we know: a Japanese corpus of about 850 dated American intercepts from 1932; 53 telegrams compared with the text published by the United States; a measure of Japanese success (about three words in four); proof, telegram by telegram, that the confidential passages escaped the Japanese; a statistical argument for a one-part code; a few corrections to the FRUS text; and the indication that coded naval dispatches were read as early as February 1932.

What remains uncertain

• No cipher groups are copied in the documents: the identification of the code (Gray) remains a compatible hypothesis, not a proof. The file copies held at the US National Archives (RG 59) could settle it.
• The reading rates were measured somewhat differently from batch to batch; “about 77 %” is an order of magnitude.
• The meaning of the Japanese marks (A.C., A.G., A.O., D.C., G.H.) is not established.
• What we say of Nakasugi’s office rests on Kotani, read through a review, and on the Japanese Wikipedia; we have not yet read Kotani’s pages directly, nor Moriyama’s articles of 2004 and 2007.
• The reading of coded naval dispatches rests on the form of the headings; it remains to be confirmed on the Asiatic Fleet originals.

References

1. Archive

   Japan Center for Asian Historical Records (JACAR), ref. C05022467800 to C05022469300: 「公文備考 昭和7年 T 事件附属」, “事件付属（米特情 極秘 号外 外）” (1)–(16), 海軍大臣官房記録 (Military History Research Center, NIDS). Copy of Grew no. 223: C05022469300, stamped images 0766–0767; Johnson no. 1078: same item, image 0777. Images not reproduced (all rights reserved).

2. Edition

   Foreign Relations of the United States, 1932, vol. III, The Far Eastern Crisis, and vol. IV, The Far East; Japan, 1931–1941, vol. I (Office of the Historian). Examples: IV, doc. 238 (Grew no. 223); IV, doc. 247 (Johnson no. 1078); III, doc. 491; III, doc. 288; III, doc. 605; IV, doc. 3; Japan 1931–1941, I, doc. 126 (Forbes no. 33).

3. Manual

   Navy Department, Communication Instructions 1929, Part IV, “Procedure and Examples”, arts. 3016–3023 (navy-radio.com).

4. Studies

   小谷賢 (Kotani Ken), 『日本軍のインテリジェンス』, Tokyo, Kōdansha, 2007, ch. 3; English ed. Japanese Intelligence in World War II, Oxford, Osprey, 2009. 森山優 (Moriyama Atsushi), “戦前期における日本の暗号解読能力に関する基礎研究”, 国際関係・比較文化研究, 3-1, 2004. Japanese Wikipedia, “中杉久治郎” (accessed 28 October 2026).