Core Finding: What Carried Over, and What Did Not
At least one of the people who built Taiwan's first naval frogman unit had learned soldiering and the sea in the Japanese Navy on Hainan: the man who commanded it in 1954. Others may have, but have not yet been identified. But the specialized techniques of underwater demolition came from the United States Navy. The Japanese inheritance was people and experience; the American inheritance was doctrine.
One man spans every stage
吳振武: IJN 16th Guard Unit, Sanya (1944–45) → leader of Taiwanese repatriation at Sanya → ROC Marine recruiter (1948) → reported frogman teams (水鬼隊) leader → first commander of the 1954 UDT.
Grade B–C B at the IJN and UDT ends; C for the links betweenEx-IJN volunteers sought as cadre
吳 sought former first- and second-term Japanese naval volunteers as the "backbone" (骨幹) of the rebuilt Marines; 黃金島, a Hainan landing-force veteran whom 吳 is reported to have trained, joined as cadre.
Grade C 黃's cadre service: BSea confidence and small-unit combat
Japanese naval training and years of Hainan counter-guerrilla operations produced men fit for amphibious work. That this pool fed the early frogman (水鬼) teams person by person has not yet been shown.
Grade D (hypothesis)UDT doctrine was American
The 1954 unit trained under a U.S. Navy Troop Training Team; its first instructor completed the U.S. UDT course; a U.S. UDT detachment from Yokosuka trained the first class.
Grade BSources, Evidence Grades, and Terms
The record is thin and uneven, so every claim on this page carries a grade showing how strong its source is. Several common words — "Marines," "SEALs," "PLA" — have to be used carefully because they did not mean in 1940 what they mean now.
Contemporary official record, or an archive's own finding aid summarizing one (JACAR, NIDS, Taiwan National Archives).
Institutional or scholarly secondary work drawing on records (ROC Navy histories, the UDT veterans' association, peer-reviewed studies, state human-rights archives).
Memoir, oral history, biography, or encyclopedia entry not yet matched to a document.
Inference or hypothesis: consistent with the evidence, testable, not yet shown.
Method. An earlier exploratory survey produced a set of candidate claims. Each was then checked against the page or document it cited. Where the check failed or the source was unreachable, the claim is dropped, corrected, or marked as unverified. Three results of that checking deserve mention at the outset: 許昭榮 was recruited by 吳振武 into the Navy's technical ratings (台灣技術員兵), not the Marines; the oral-history sentence sometimes quoted placing 黃金島 and 吳振武 together at Mawei could not be found in the cited interview, whose speaker instead places 吳 at Zuoying; and the published sources disagree on whether 吳 was the UDT's first team commander (隊長) or first battalion commander (大隊長). The page follows the sources where they agree and reports the disagreement where they do not.
"Marines." Imperial Japan had no separate marine corps. The nearest equivalent was the Navy's Special Naval Landing Force (海軍特別陸戦隊, SNLF). On Hainan, the Navy's guard units (警備隊) also fought as naval infantry. A veteran who later said "I was in the Japanese 陸戰隊 (landing force)" may have meant either.
"SEALs." Taiwan's 1954 unit was a UDT — an underwater demolition team. The U.S. Navy's own UDTs were the ancestors of the SEALs, whose first two teams were established in January 1962.[18] "Taiwan's SEALs" is therefore an anachronism for 1954, though the lineage is real.
"PLA." On Hainan in 1939–45 the Communist opponent was the Qiongya (瓊崖) Communist force, later the Qiongya Column (瓊崖縱隊), which Japanese records call Communist forces (共産軍) or Communist "bandits" (共産匪). Veterans speaking decades later naturally called them "the Liberation Army." This page uses the period terms.
Names. 黃圳島 took the name 黃金島 after 1947; both are used, in period. Japanese unit abbreviations follow period usage: 横四特 (Yokosuka 4th SNLF), 佐八特 (Sasebo 8th SNLF), 舞一特 (Maizuru 1st SNLF).
The Japanese occupation of Hainan, the Taiwanese who served in it, the war they fought, and how they came home.
Japanese Forces on Hainan: A Navy-Dominated Occupation
The 1939 invasion was mainly a Navy operation, with a small Army force in support that soon left. For the rest of the war Hainan was run mainly by the Japanese Navy, through five ground formations, each responsible for one part of the island. Knowing which part a veteran served in usually tells you which unit he belonged to.
The Army, briefly
| Formation | Hainan connection | Grade |
|---|---|---|
| Taiwan Mixed Brigade (台湾混成旅団) (Iida Detachment) | Army component of the February 1939 invasion, which the Navy led; later withdrawn.[23] | A / C |
| 48th Division (第48師団) | Formed around the Taiwan Mixed Brigade and Infantry Regiment 47; concentrated on Hainan in spring 1941 after the Fuzhou landing, then went south.[1] JACAR gives organization complete in December 1940; Japanese Wikipedia gives organization on 30 November 1940, completion on 7 February 1941, and a move to Taiwan in August 1941, without mentioning Hainan.[23] | A |
| 23rd Independent Mixed Brigade (独立混成第23旅団) | Formed in Taiwan in January 1943; sent to Hainan from late 1944 in preparation against American operations.[1] | A |
The Navy's permanent structure
The Navy's 4th Base Force became the Hainan Island Base Force and, on 10 April 1941, the Hainan Guard District (海南警備府), the island's principal military and occupation authority. JACAR records that it operated "while receiving human and technical support from the Taiwan Government-General" (台湾総督府から人的・技術的支援を受けつつ).[1] At surrender its five ground formations were:
| Formation | Sector | Lineage and enemy | Taiwanese at handover | Taken over by / concentrated at |
|---|---|---|---|---|
| 15th Guard Unit (第15警備隊) | Haikou (north) | 5th Defense Unit (20 Jan 1939) → 15th Defense → 15th Guard (Nov 1939) | "台湾軍人191名とその他台湾軍属が所属" — 191 Taiwanese servicemen plus Taiwanese civilian employees belonged to the unit | — |
| 16th Guard Unit (第16警備隊) | Sanya (south) | Kure 6th Guard Unit (20 Jan 1939) → 6th Defense → 16th Defense → 16th Guard (Nov 1939); fought Chongqing security forces, guerrillas, and Communist forces | — | Repatriation from 19 Mar 1946; of the unit's own personnel (nationality not stated), 86 kept for technical work and 100 held as war-crimes suspects |
| 横四特 (Yokosuka 4th SNLF) | West | Formed 20 Jan 1939; ~860 men at the invasion; fought security units (保安団), guerrilla units (遊撃隊), Communist forces (共産軍) | 1,825 Taiwanese servicemen and civilian employees (軍人・軍属) in its district, transferred 10 Dec 1945 | New 19th Division; Ledong concentration camp (楽東集中営), 26 Dec 1945 |
| 佐八特 (Sasebo 8th SNLF) | East | Formed 20 Jan 1939; same three enemies | 1,871 in its district, transferred 6 Nov 1945 | 188th Division; Tiandu concentration camp (田独集中営), 6 Dec 1945 |
| 舞一特 (Maizuru 1st SNLF) | Northwest | Formed on Hainan 10 Jun 1940 | 2,221 Taiwanese servicemen, civilian employees, (軍人・軍属・)civilian residents (居留民) (including civilian residents), transferred 25 Nov 1945 | — |
Temporary landing forces. Other SNLFs fought on Hainan for shorter periods: the Kure 6th (about 730 men, invasion of 12–18 February 1939), the Shanghai SNLF's "Sea No. 1 Unit" in the northeast (March–June 1940), and the Maizuru 2nd in the Y4 operation.[2] C (veterans' society list)
The Taiwanese in Hainan's Naval System
Nearly 14,000 Taiwanese worked for the Japanese Navy on Hainan as civilian employees — mostly interpreters and police auxiliaries. About 800 were formal servicemen, mainly in the naval landing forces. Many started as civilians and later passed an exam to become naval volunteers on the island itself.
| Measure | Figure | Source | Grade |
|---|---|---|---|
| Taiwanese military personnel on Hainan at war's end | 817 | 鍾淑敏、沈昱廷、陳柏棕 (2024), from the Yasukuni registers[3] | B |
| Taiwanese civilian military employees (軍屬) | 13,953 | ||
| Combined — about 70% of all Taiwanese on Hainan | 14,770 | ||
| Taiwanese naval volunteer system announced / in force | 11 May / 1 Aug 1943 | 陳柏棕 (2012)[4] | B |
| First cohort (1,000) enters training | 1 Oct 1943 | ||
| Volunteer exam held at Beili (北黎), Hainan | 600-odd sat; 100 passed | Civil Historical Materials Digital Platform, 1937–1949 (民間史料數位平台1937–1949)[5] | C |
The same study states that the Taiwanese servicemen on Hainan "were mainly naval landing forces, because the landing forces, to make up for insufficient manpower, conscripted large numbers of Taiwanese, trained them, and distributed them among the various detachments" (軍人主要是海軍陸戰隊……大量徵用臺灣人加以訓練，然後分派到各個分遣隊); the civilian employees were mostly interpreters and police or police auxiliaries (巡查、巡查補).[3] Chen Po-tsung's study of the naval volunteers adds that the volunteers on Hainan were men "originally dispatched there as civilian employees" (原先派遣至當地擔任軍屬者).[4] The pathway was therefore not only Taiwan → training → Hainan, but also Taiwan → Hainan as a civilian → local examination → naval volunteer.
Named individuals
| Person | Hainan service | After 1945 | Grade |
|---|---|---|---|
| 張子涇 (b. 1921, Fengyuan) | Naval interpreter, Sasebo 8th SNLF (佐八特), Oct 1941–Aug 1945 | Leader of the Taiwanese at the Jiaji (嘉積) camp, Sep 1945–Jun 1946; returned by sailing junk, July 1946 Memoir published 2017.[6] | C |
| 黃圳島 / 黃金島 | Navy civilian employee from March 1942; engine training at Sanya; passed the volunteer exam at Beili; trained at Yokosuka Naval Barracks (横須賀海兵団); practical combat training in the Navy 7th Mobile Unit (海軍第七機動部隊); served in the Yokosuka 4th landing force | Interned; sailed from Haikou May 1946, reached Taiwan 14 June 1946. See §07.[7] | B |
| Note: Chinese Wikipedia dates 黃圳島's entry into the naval volunteers to 1942, which is impossible, since the system began in 1943; the sequence above, from the National Human Rights Museum, is followed. | |||
| 洪燦榮 | Hainan Navy Motoyoshi Unit (海南海軍本吉部隊) | — | B |
| 陳順和 | First-term volunteer sent to Hainan; unit given as the Wenchang company of the Haikou 15th Guard (文昌中隊) — row reading uncertain | — | C |
The War They Fought
For six years the Japanese naval units on Hainan fought guerrillas — Nationalist, Communist, and local. That war included the killing of villagers during sweeps. This account includes both the combat and the civilian deaths, and treats Japanese casualty numbers with caution.
The opponents. JACAR's entries for the 16th Guard Unit, Yokosuka 4th SNLF (横四特), and Sasebo 8th SNLF (佐八特) each describe the same three opponents: Chongqing's security units (重慶軍の保安団), guerrilla formations (遊撃隊), and Communist forces (共産軍).[1] A The Communist movement on Hainan, grew in the later war years: named the Qiongya Column (瓊崖縱隊) in autumn 1944, it is reported to have reached about 7,000 men and to hold nearly three-fifths of the island by 1945 — the years in which the Taiwanese volunteer cohorts served.[24] C
The operations. The island-wide sweeps were numbered Y1, Y2, and onward; landing-force platoons were regularly combined into temporary punitive columns (討伐隊) for them. A surviving series of Yokosuka 4th SNLF (横四特) battle reports (戦闘詳報) runs from December 1943 to 1945, with sections recording claimed results, Japanese losses, ammunition expended, and number of sweeps.[8] A (finding aid)
Civilian deaths. The Japanese record cannot be read as a count of combatants. The economist and historian 斉藤日出治 (Saitō Hideharu) has examined the Hainan Guard District's war diaries against local testimony and argues that massacres of residents were carried out alongside the Y operations.[9] In a later project he matched villagers' accounts of a 1945 killing at Longwei New Village (龍衛新村) in western Hainan against Yokosuka 4th SNLF (横四特)'s own battle report No. 5 of 15 April 1945, "The Battle of Longwei New Village, 2 March 1945."[10] B
Surrender, Internment, and the Long Way Home
After August 1945 the Taiwanese on Hainan were separated from the Japanese, disarmed, held in camps, and left waiting for ships — in many cases for most of a year. Where a man was held after the war does not prove where he served during it.
| Date | Event | Grade |
|---|---|---|
| 6 Nov 1945 | 1,871 Taiwanese in the Sasebo 8th SNLF (佐八特) district transferred to Chinese control; Sasebo 8th SNLF (佐八特) concentrated at Tiandu (田独), within Sanya (today's Jiyang District), by 6 Dec.[1] | A |
| 25 Nov 1945 | 2,221 Taiwanese (including civilian residents) in the Maizuru 1st SNLF (舞一特) district transferred.[1] | A |
| 10 Dec 1945 | 1,825 Taiwanese in the Yokosuka 4th SNLF (横四特) district transferred; Yokosuka 4th SNLF (横四特) concentrated at Ledong (楽東) by 26 Dec.[1] | A |
| 1945–46 | 吳振武 chosen as head of repatriation (遣送部長) for Taiwanese servicemen in the Sanya area.[11] | C |
| 19 Mar 1946 | Repatriation of the 16th Guard Unit begins.[1] | A |
| 2 May 1946 | The Sanya port commander cables that 5,658 Taiwanese soldiers and 1,386 Taiwanese civilians — 7,044 in all — have waited long for ships in hardship, and asks for transport.[12] | A |
| May–Jun 1946 | 黃圳島 sails from Haikou; lands at Xuejia, Tainan, on 14 June.[7] | B |
| Jul 1946 | 張子涇 returns to Taiwan by sailing junk.[6] | C |
| Aug 1947 | A file on repatriating Taiwanese from Singapore contains a name list of Taiwanese still remaining on Hainan (海南島殘留台籍民人名表), whose indexed localities include Erlin Township (二林鎮).[13] | A (finding aid) |
How one officer and his network carried Hainan experience into the rebuilt ROC Marines and the first frogman unit.
吳振武: The Man Who Spans Every Stage
吳振武 is the only person so far who can be followed through every link of the chain, from the Japanese Navy on Hainan to command of Taiwan's first frogman unit. Most of his biography comes from secondary sources, so parts of it still need documents behind them.
| Stage | Record | Source | Grade |
|---|---|---|---|
| Birth and education | Born 1918 at Touqianxizai, Pingtung (屏東頭前溪仔); studied physical education at the Tokyo Higher Normal School (東京高等師範學校, today's University of Tsukuba). | zh.wikipedia[22] | C |
| Commission | Trained in the 3rd class of naval reserve students, line branch (第三期海軍兵科預備学生), and was commissioned ensign (少尉) — the reserve-officer route for college graduates, not the volunteer system. | zh.wikipedia[22] | C |
| IJN, Hainan | "In autumn 1944 吳振武 was transferred to Hainan, to the 16th Guard Unit directly under the Hainan Guard District, as a lieutenant (junior grade) battalion commander (大隊長)" (1944年秋……第十六警備隊，擔任中尉大隊長); soon promoted to lieutenant (大尉). | National Library biography[11] | B |
| Trainer of 黃圳島 | "吳振武 was the senior who guided 黃圳島 when he received practical combat training in the Navy 7th Mobile Unit (海軍第七機動部隊)" (即黃圳島在海軍第七機動部隊接受實戰訓練的指導前輩). | Taiwan Dadi Cultural and Educational Foundation (台灣大地文教基金會) biography[14] | C |
| Repatriation | Chosen head of repatriation for Taiwanese servicemen in the Sanya area. | zh.wikipedia, citing published biography[11] | C |
| 228 Incident, Taichung (Mar 1947) | A physical-education teacher at Taichung Normal School, he was asked by the local crisis committee to head a security-maintenance corps (治安維持總隊長); he refused to lend the school's rifles to 謝雪紅, was shot in the thigh, and could not take up the post. The 27 Brigade (二七部隊) article records the wounding on 6 March and the crisis committee's account that he shot himself by accident while cleaning a gun; other accounts differ. He remained on the wanted list. | zh.wikipedia (吳振武; 二七部隊)[22][25] | C |
| Possible second record | A National Human Rights Museum entry for a 吳振武 born in 1918, of Taichung, records service on the Taichung crisis committee, command of the 27 Brigade's student squad, arrest, and about three months' detention under torture. It does not mention naval service and conflicts with the account below; whether it is the same man is unconfirmed. | National Human Rights Museum[26] | B (identity unconfirmed) |
| Into the ROC Navy | Under the protection of Navy commander 桂永清, he went to Nanjing on 15 March 1947 and served at Navy Headquarters as a lieutenant commander (少校); then trained at Mawei, served at Xiamen and elsewhere, and later reported at Zuoying. By July 1947 he was writing, as a lieutenant commander staff officer (少校參謀), to recruit former Japanese naval volunteers. | zh.wikipedia[22]; National Human Rights Museum (許昭榮)[27] | C; July letter B |
| Marine recruiter | "Personally selected one hundred Taiwanese officers and men" (親自遴選一百名臺灣籍官兵) for the Marines; sought "first- and second-term former Japanese naval volunteers as the backbone of the Marine Corps" (招募第一、二期的前日本海軍志願兵作為海軍陸戰隊骨幹). | zh.wikipedia, citing 李碧鏘 and 許昭榮's autobiography[11] | C |
| frogman teams (水鬼隊) | "During 胡宗南's command of the Southeast Anti-Communist National Salvation Army, 吳振武 repeatedly led frogman teams (水鬼隊) on important intelligence and demolition missions along the Fujian coast." | zh.wikipedia[11] | C |
| UDT command | First team commander (隊長) of the Navy Underwater Demolition Team (海軍水中爆破隊), formed 1 March 1954; a commander (中校) at the time; Wikipedia gives 1954–1960. | UDT veterans' association; 黃種雄's account[15][16] | B |
| Detention and release | After 桂永清's death (12 August 1954), the Navy Headquarters political department had him arrested on a charge of taking part in the 228 Incident and sent to the Taipei Garrison Command; after about six months of interrogation he was released, found not guilty. The year, and its overlap with his UDT command, are not stated. | zh.wikipedia[22] | C |
| Later life | Moved to Navy sports and recreation posts; retired as a commander (中校); taught physical education at Kaohsiung Medical College and Kaohsiung Normal College; a record-setting thrower (discus 1947, hammer 1951); died 1 July 1999. | zh.wikipedia[22] | C |
The Rebuilt Marines and the Ex-IJN Cadre
The ROC Marine Corps was rebuilt in 1947. 吳振武, who had joined the ROC Navy that March, was writing to former Japanese naval volunteers by July, and was recruiting at Zuoying by late 1948, seeking men the Japanese Navy had already trained. One of them was 黃金島, a Hainan landing-force veteran whom 吳 is said to have trained during the war. He served as a Marine cadre and was arrested in 1952, by then at the Armored Forces School.
| Date | Event | Source | Grade |
|---|---|---|---|
| 16 Sep 1947 | "Rebuilt the Marine battalion at Mawei, Fujian, with six companies under it" (重建海軍陸戰隊大隊於福建馬尾，下轄六個中隊). | ROC Navy[17] | A |
| 1948 | The Marine battalion is merged with the Navy guard regiment into the 1st, 2nd, and 3rd Marine Regiments (海軍陸戰隊第1、2、3團). | zh.wikipedia (中華民國海軍陸戰隊)[29] | C |
| Late autumn 1948 | 吳振武 recruits at Zuoying; 黃圳島, wanted by the authorities after the 228 Incident, joins him under the name 黃金島. | Taiwan Dadi Cultural and Educational Foundation (台灣大地文教基金會)[14] | C |
| — | 黃金島 serves "as a cadre at the Zuoying Third Naval Base of the Marine Corps" (到海軍陸戰隊左營海軍第三基地擔任幹部). | National Human Rights Museum[7] | B |
| Jul–Oct 1947 | 許昭榮, a second-term Japanese naval volunteer (1943), receives in July a letter from 吳振武, whom he did not know, then a lieutenant commander staff officer (少校參謀), urging him to enlist; in October he enters the Navy Technical Ratings Battalion (海軍技術員兵大隊) — not the Marines. He later served on the mainland coast until 1950, and in 1958 was court-martialed and sentenced to ten years. | National Human Rights Museum (許昭榮)[27]; zh.wikipedia[11] | B |
| June 1952 | 黃金島 arrested at the Armored Forces School; sentence later changed from 12 years to life; released 14 July 1975, under the special amnesty (特赦) of that year, after 23 years, 1 month, and 13 days. | National Human Rights Museum[7] | B |
The 228 context. 黃圳島 fought on the side of the armed resistance in central Taiwan during the 228 Incident of 1947: the National Human Rights Museum records that he led an independent security unit and was appointed commander of the guard by 鍾逸人, and Wikipedia describes him as commanding in the front line at the Wuniulan battle.[7] His later service in the ROC Marines, under the recruiter who had trained him in the Japanese Navy, and his arrest in 1952 belong to the same record. They are reported here because they bear on the lineage, not to take a side in how the period is remembered.
A corrected quotation. A sentence sometimes attributed to 許昭榮 — that 黃金島 "also came back from Hainan and went with 吳振武 to Mawei to organize the Marines" — does not appear in the 2006 oral-history interview to which it is cited. In that interview 許 is the interviewer; the interviewee, 陳淡而, says instead that 吳 "went into the Navy, was sent to Zuoying, to help them organize the Marines" (吳振武就進去海軍，派到左營，要幫他們組陸戰隊).[19]
The recruiter's own position. The officer who recruited Taiwanese veterans into the rebuilt Marines had himself been on a 228 wanted list in March 1947, and after 1954 was detained and cleared on the same charge (§06). How he passed from a wanted list to a naval staff appointment within weeks — Wikipedia credits the protection of Navy commander 桂永清 — is a question for the archives.
Frogmen (水鬼, "Water Ghosts") Before the Name: The Undocumented Years
Before 1954 there were already swimmer teams raiding the mainland coast, and 吳振武 is reported to have led some of them. Almost nothing about who these men were has been published. This is where former Japanese naval men are most likely to be found — and where the evidence is weakest.
What is reported. The only published statement located for this page is the Wikipedia line that 吳振武 "repeatedly led frogman teams (水鬼隊) on important intelligence and demolition missions along the Fujian coast" during 胡宗南's command of the Southeast Anti-Communist National Salvation Army.[11] C The attribution conflicts with 胡宗南's own Wikipedia article, which gives his command (1951–53, under the alias 秦東昌) as the Jiangsu–Zhejiang Anti-Communist National Salvation Army (江浙反共救國軍) on the Dachen islands off Zhejiang; the Fujian-coast guerrilla forces were under 胡璉 at Kinmen, and the only "Southeast" body on record, 黃炳炎's Southeast People's Anti-Communist National Salvation Army (東南人民反共救國軍), was merged into those forces in 1950.[30] Either the commander or the coast is probably wrong. If the account is confirmed from operational records, it would place clandestine swimmer operations under an ex-IJN officer before the American UDT curriculum arrived.
The obvious question. Where would such a commander have found swimmers, boat handlers, and men used to small-unit fighting? His documented recruiting network — former Japanese naval volunteers sought as Marine cadre — is a plausible source. D No roster of these early teams has been located.
Why the record is thin. Clandestine coastal operations in the early 1950s were among the least documented activities of the period, and many records remain unpublished or unindexed in ROC military archives. There may also be a cultural reason. The Koyanagi interviews — some 4,000 pages in 44 volumes, recorded with 47 senior Imperial Navy officers between 1956 and 1961 — were kept confidential until 2010, one sign of how guarded that navy's record remained.[20] Taiwanese veterans of that navy, who after 1945 served a different state and lived through 228 and the White Terror, had further reasons to say little. 張子涇's Hainan memoir appeared only in 2017.
The 1954 UDT: American Doctrine, Taiwanese Cadre
The formal unit began on 1 March 1954; its first members had only four weeks of initial training, and 黃種雄 counts 22 of them. A Taiwanese officer who had just finished the complete American course came back as instructor, and an American UDT team from Japan trained the first complete class. Besides 吳振武, six early men are named in 黃種雄's account; their earlier lives are not known.
| Item | Record | Grade |
|---|---|---|
| Founding | "Established at Zuoying on 1 March 1954; the initial members received only four weeks of training" (1954年3月1日……於左營成立，初期隊員僅受4周訓練). The unit already existed when the U.S. Troop Training Team (T.T.T.) instruction began.[15] | B |
| Instructor | 黃種雄: "On 4 January 1954 I began the twelve-week UDT training" (民國四十三年一月四日我開始了為期十二週的UDT訓練); returned as instructor.[16] | C |
| U.S. detachment | "The U.S. Navy UDT sent two officers and three petty officers from Yokosuka naval base in Japan to Zuoying" (由日本橫須賀海軍基地派來左營二位軍官、三位士官).[16] | C |
| U.S. units | A 2024 commentary in UpMedia (上報) states that the U.S. Troop Training Team began instruction on 1 March 1954 and that U.S. Navy UDT-5 assisted from May, and names one U.S. instructor, Quartermaster Second Class Dennison.[31] | C |
| Leadership | "The commander was Commander 吳振武, the deputy Lieutenant 尤世坤, both outstanding naval athletes. There were twenty-two members" (隊長是吳振武中校，副隊長是尤世坤上尉，兩位都是海軍的運動健將。隊員有二十二人).[16] | C |
| Named members | Deputy 尤世坤; assistants 張文斌, 王樹俊, 楊振儀, 朱和璞; and 吳肇嶼, named among sixteen trainees (「吳肇嶼等十六位」) of the first complete class. Whether the assistants were among the 22 is not stated.[16] | C |
What the roster does and does not show. No published biography located for this page links 尤世坤, 張文斌, 王樹俊, 楊振儀, 朱和璞, or 吳肇嶼 to Japanese naval service. That absence is weak evidence either way: pre-1954 biographies of these men are simply not exposed in searchable sources. The note that the commander and his deputy were "outstanding athletes" suggests that the unit also recruited on swimming and athletic ability, independently of any Japanese background. The first cadre most likely drew on several pools at once: ex-IJN Taiwanese, Navy personnel generally, and strong swimmers. D
What is proven, what is plausible, how to tell the Hainan units apart, one family's oral history, and the author's own case.
Evidence Ledger
This table lists every link in the argument with its current status, so a reader can see at a glance which parts rest on documents and which are still hypotheses.
| Proposition | Status | Grade |
|---|---|---|
| Hainan was a Navy-dominated occupation with five sectoral ground formations | Proven | A |
| Hundreds of Taiwanese served as naval servicemen on Hainan, mainly in landing forces | Proven | B |
| 191 Taiwanese servicemen belonged to the 15th Guard Unit at war's end | Proven | A |
| A named Taiwanese served in Sasebo 8th SNLF (佐八特) (張子涇) | Reported (memoir) | C |
| A named Taiwanese served in Yokosuka 4th SNLF (横四特) (黃圳島) | Proven at biography level | B |
| 吳振武 commanded in the 16th Guard Unit | Reported (National Library biography) | B |
| 吳 was commissioned through the naval reserve-student course | Reported (Wikipedia) | C |
| 吳 was on a 228 wanted list and later detained and cleared | Reported (Wikipedia) | C |
| 吳 trained 黃圳島 during the war | Reported | C |
| ROC Marines rebuilt at Mawei, 16 Sep 1947 | Proven | A |
| 吳 recruited ex-IJN naval volunteers as Marine cadre | Reported (autobiography via Wikipedia); his July 1947 letter to 許昭榮 documented | C / B |
| 黃金島 served as Marine cadre | Proven at biography level | B |
| 許昭榮 joined the Marines | Incorrect — technical ratings | — |
| 吳 led frogman teams (水鬼隊) before 1954 | Reported, single source | C |
| 吳 was the 1954 UDT's first commander | Proven (title disputed: team commander, 隊長, vs battalion commander, 大隊長) | B |
| UDT doctrine came from the U.S. Navy | Proven | B |
| Ex-IJN enlisted men were among the early frogmen (水鬼) | Plausible, untested | D |
| Any Hainan veteran besides 吳振武 (for example, one of the many Taiwanese in the Sasebo 8th SNLF (佐八特) district) became an early frogman (水鬼) or UDT founder | Plausible, no named case | D |
| Japanese techniques formed the basis of ROC UDT doctrine | Not supported | — |
Unit Fingerprints: Telling the Hainan Formations Apart
Family memories rarely include a unit's formal name, but they often include a place. Each Hainan formation had its own territory, so a single remembered place name can point to the right unit — as long as wartime postings are kept separate from post-war camps.
| If the veteran remembered… | Points to | Basis |
|---|---|---|
| Haikou (海口); Wenchang (文昌) | 15th Guard Unit (第15警備隊) | Haikou (海口): sector (JACAR) A; Wenchang (文昌): single uncertain appendix reading C |
| Sanya (三亞), Yulin (榆林), Yaxian (崖縣) as a wartime posting | 16th Guard Unit (第16警備隊) | Sector (JACAR) A |
| Beili (北黎), Basuo (八所), Shilu (石碌) (iron mine), Changjiang (昌江), Baisha (白沙), Huangliu (黃流), Gan'en (感恩); Baoping (保平), in today's Yazhou District of Sanya, where the boundary with the 16th Guard's Yaxian area is unclear | Yokosuka 4th SNLF (横四特) | Western sector (JACAR) A; place names from JACAR item summaries C |
| 楽東 (Ledong) camp after surrender | Yokosuka 4th SNLF (横四特) | Concentration site (JACAR) A |
| 嘉積 (Jiaji), Bo'ao (博鰲), Wanning (萬寧), Tanmen (潭門), Lingshui (陵水), Wanquan River (萬泉河); Tengqiao (藤橋), in today's Haitang District of Sanya on the Lingshui border, a weaker marker | Sasebo 8th SNLF (佐八特) | Eastern sector (JACAR) A; place names from item summaries C |
| 田独 (Tiandu) camp after surrender | Leans Sasebo 8th SNLF (佐八特) | Concentration site (JACAR) A — but Tiandu (田独) lies within Sanya |
| Nada (那大) or the northwest | Maizuru 1st SNLF (舞一特) | Northwestern sector (JACAR) A |
Clues that do not discriminate
- "Japanese Navy 陸戰隊 (landing force)," hard sea swimming, fighting Communists in the hills, internment, and eventual return fit every formation.
- "Yokosuka" remembered from training usually means the Yokosuka naval barracks (横須賀海兵団), where volunteers trained, not the Yokosuka 4th landing force; "Sasebo" likewise.
- The swimmer's trick of letting a long loincloth (六尺褌) trail behind to make a shark think its prey is larger is described as an old fishermen's method in 石原慎太郎's story 「生き残りの水兵」 ("The Surviving Sailor," 2001);[21] an ethnographic source has not been located D. Whatever its origin, it marks seafaring culture, not any one unit.
An Oral-History Case: The Erlin Veteran
A spoken account from Erlin in Changhua — given by the veteran himself about twenty years ago and preserved in family memory — describes a Taiwanese man who served in a Japanese naval landing force on Hainan, led a platoon, fought Communist guerrillas, survived internment, and came home. Every part of the story is historically plausible. None of it yet identifies his unit, and one detail puts limits on when he could have served.
| Remembered detail | How it fits the record |
|---|---|
| From 二林 (Erlin), Changhua | An indexed 1947 name list of Taiwanese remaining on Hainan includes Erlin Township (二林鎮) among its localities (§05). The link to any individual is not yet made. |
| Japanese naval landing force (陸戰隊) | Consistent with the finding that Taiwanese servicemen on Hainan were mainly in the landing forces (§03). |
| Hard open-sea swimming; the shark and the long cloth | Consistent with naval training and seafaring lore; not unit-specific (§11). |
| Platoon leader (小隊長; a post held in the U.S. Marine Corps by a lieutenant (O-1/O-2), and in a U.S. Navy SEAL platoon by a Navy lieutenant (O-3), the same grade as an Army or Marine Corps captain), mostly in southern Hainan | A platoon command narrows little by itself. "Southern Hainan" could mean Sanya (16th Guard), the southwest around Huangliu (黃流) and Ledong (楽東), pointing to the Yokosuka 4th SNLF (横四特); or the southeast around Lingshui (陵水) and Tengqiao (藤橋, on the Sanya–Lingshui border), pointing to the Sasebo 8th SNLF (佐八特). |
| Many close-range engagements with Communist forces; opponents usually appeared in groups | All three southern formations fought Communist forces; the Communist presence grew in 1943–45. The family memory describes repeated short-distance, group-on-group fighting rather than sniper-type engagements. |
| Internment, then return to Taiwan | Typical: thousands waited at Sanya into mid-1946 (§05). |
The Author's Case: Lineage and Counterintelligence
The author is the grandson of the Erlin veteran in §12. This section asks whether that ancestry — a Taiwanese who rose from the lowest rank to platoon leader (in U.S. terms, a lieutenant's post — at the grade of a captain in a Navy SEAL platoon) in a Japanese naval landing force and fought the Qiongya (瓊崖) Communist forces — bears on whether his son or grandson could be trusted or recruited by the PRC, and whether suspecting them on that basis would be reasonable. The estimates are the author's own analytic judgments; subsection D sets out how each was derived, and subsection E compares them with an average Taiwanese adult, so that readers can check them.
A. The grandfather's record: oral-history anchor and comparative model
The strongest evidence for the grandfather's personal combat total is his own testimony, given directly to family members about twenty years ago. When specifically asked how many Communist fighters he himself had personally killed, he answered, “I personally killed so many Communist fighters that I lost count.” On another occasion, when asked to estimate rather than answer generally, he recalled approximately 45–55 or more. His remembered engagements were predominantly at short distance, against Communist adversaries who commonly appeared in groups, and the family remembers combat occurring with very high frequency during roughly twenty-two months of wartime service.
He began at the lowest enlisted rank and rose to platoon leader (小隊長, equivalent to an ROC 排長, platoon leader; a post normally held in the U.S. Marine Corps by a lieutenant (O-1/O-2), while a U.S. Navy SEAL platoon may be led by a Navy lieutenant (O-3), pay-grade equivalent to an Army or Marine Corps captain). That rise most likely required service from before 1943, reaching Hainan first as a civilian employee or police auxiliary (probability about 63 in 100; service of any kind before 1943, about 75 in 100; see D1).
The page therefore no longer derives his personal total from island-wide casualty figures. The oral-history estimate is the anchor; the statistical model below asks the narrower question: is a remembered total of at least roughly 45–55 operationally plausible for about twenty-two months of frequent close-range group combat?
| Proposition | Evidence status |
|---|---|
| Repeated direct personal killing in close-range combat | Strong family oral-history evidence |
| Approximate remembered personal total | About 45–55 or more Communist fighters — family oral-history datum, not a wartime tally |
| Exact total | Unknown |
| Under 20 | Difficult to reconcile with the veteran's own recalled estimate |
| 60–100 | Plausible, but inferential above the remembered anchor |
| 100+ | Possible, not demonstrated |
Comparative model: how unusual is 45–55?
No comprehensive U.S. database records lifetime personal enemy kills across the Army, Marines, Navy, Reserve components, National Guard, Coast Guard, or federal/state/local law enforcement. Published individual totals are strongly selected toward roles such as snipers and aviators; infantry and special-operations records usually report results at patrol, platoon, or unit level. Law-enforcement national use-of-force data are incident-based rather than career kill ledgers, and participation is voluntary.[37] Accordingly, the distributions below are reference-class models, not measured historical frequencies.
Three empirical anchors constrain the model without turning it into a false census. Pew found that about 29 percent of U.S. veterans reported combat experience,[35] so even before asking about killing an opponent, combat exposure is a minority experience. For one unusually well-documented Vietnam SEAL period, June 1969–April 1970, SEALs in the Mekong Delta conducted 1,192 missions resulting collectively in 581 Viet Cong killed and 193 captured; that calibrates mission-level tempo and lethality but not individual attribution.[36] The FBI's national law-enforcement collection began nationwide in 2019 and records qualifying incidents, not lifetime named-officer totals.[37]
| Personal direct kills | General combat-exposed | Heavy infantry / recon | High-tempo raider / SOF |
|---|---|---|---|
| 0 | 35–55% | 15–30% | 5–15% |
| 1–4 | 25–35% | 20–30% | 10–20% |
| 5–9 | 8–15% | 15–20% | 10–15% |
| 10–19 | 4–8% | 12–18% | 15–20% |
| 20–39 | 1–4% | 8–15% | 18–25% |
| 40–59 | <2% | 4–8% | 10–18% |
| 60–99 | <1% | 2–5% | 5–12% |
| 100+ | Very rare | <2% | 1–5% |
What the comparison can support. A remembered total of at least 45–55 would be extraordinary against all uniformed military and law-enforcement personnel, because zero would dominate that pooled population. Against combat-exposed personnel it belongs in a small upper tail; against heavily exposed infantry or reconnaissance personnel it remains high; only within a deliberately selected high-tempo raider/SOF reference class does the 40–59 range become relatively common. Because no career-level U.S. dataset exists, this page does not assign a literal percentile or a historical rank.
Hainan operational plausibility check
Roughly twenty-two months is about ninety-five weeks. Family memory describes combat as occurring almost every week, but "combat" may have meant patrol or sweep activity rather than an actual firefight every time. To avoid taking the phrase literally, the model brackets substantive armed contact at roughly 40–60 percent of weeks, or about 38–57 contacts. Because both sides usually encountered one another in groups at short distance, personal kills should accumulate in bursts rather than at a constant one-per-contact rate.
| Illustrative personal result in a contact | Number of contacts | Contribution |
|---|---|---|
| 0 | 18 | 0 |
| 1 | 14 | 14 |
| 2 | 8 | 16 |
| 3 | 4 | 12 |
| 4 | 2 | 8 |
| Total | 46 substantive contacts | 50 personal kills |
Who the opponents were, and in what numbers. According to the family, the grandfather never encountered Western forces; every enemy he fought was Chinese. On the ground the same was true of both units in question: neither the Yokosuka 4th SNLF (横四特) nor the Sasebo 8th SNLF (佐八特) fought Western ground forces on Hainan. Their opponents were Nationalist (Chongqing) security and guerrilla units, local armed groups, and — increasingly, and in the later war above all — the Qiongya Communist forces, whose successor formation was later incorporated into the People's Liberation Army.
| Measure | Figure |
|---|---|
| Qiongya Communist column at its peak, 1945 | About 7,000 men, holding nearly three-fifths of the island[24] |
| Communist personnel dead over the whole occupation (the survey's category) | 6,133, as reported from the Hainan provincial survey of wartime losses (not yet verified against the report)[32] |
| Of these, killed by Japanese naval ground forces over the whole occupation | About 3,000–4,300 (estimate: 50–70% of 6,133) |
| Of these, killed by naval ground forces from mid-1942 to the surrender | About 1,700–3,200 (estimate) |
| All dead recorded by Japanese forces island-wide, February 1939–July 1942, all categories | 24,387 (secondary figure; see §04) |
| The same, extended to the surrender | Low-to-mid 30,000s (extrapolation, not a record) |
Attrition. The two figures measure different things: the 6,133 are cumulative deaths over six and a half years, while about 7,000 is the column's strength at its 1945 peak. Since at least some 13,000 people served in the column, its deaths amount to as much as about 47 percent of all who served — a high attrition rate for a guerrilla force, much of it inflicted by the Japanese naval ground forces in which the grandfather served. For the question in this section, the point is scale: this is the record a PRC political review would associate with his units. The proportion rests on the 6,133 figure and awaits its verification.
The scale a PRC reviewer would see. Political review assigns categories; it does not count one man's kills. The naval landing forces and guard units were the main Japanese ground force on Hainan for most of the occupation, so a large share of the Communist column's dead fell to them. For a PRC reviewer, the grandfather served in the forces that killed several thousand of the fighters the PRC claims as part of its own lineage — a record that Mao-era review classed with collaboration (漢奸, collaborator) and "historical counter-revolution" (歷史反革命) — and it is that record, measured at the scale of his units and the occupation rather than his personal count, that political review would weigh.
B. The counterintelligence questions
| Question | Probability (out of 100) |
|---|---|
| The son or grandson is accepted into CCP membership, the PLA, or a PRC post of political trust (as ROC citizens) | < 1 |
| The same, supposing they went through PRC political review (政審) inside the PRC system with this family history | 2–5 |
| Approached by PRC intelligence or United Front (統戰, united-front work) outreach during a career, as an ordinary private citizen | 1–6 |
| The same, for someone with national-security access or a public research profile on the PRC | 20–40 |
| Effect of the lineage on the chance of being approached | About ×1 (range ×0.8–×1.2): no real effect |
| If one of them volunteered to the PRC: rejected or treated as a provocation | 55 (against 40 for an identical volunteer without this family history) |
| If one of them volunteered: used under tight control and tested as a possible planted double agent | 40 (against 50) |
| If one of them volunteered: trusted as an asset | 5 (against 10) |
| If he volunteered: used only under tight control and tested as a possible planted double agent | 40 (against 50) |
| If he volunteered: rejected, or treated as a provocation | 55 (against 40) |
| The son works for the PRC, judged on the evidence | About 0.02–0.11 (at most about 1 in 1,000) |
| The grandson (the author) works for the PRC, judged on the evidence | About 0.005–0.08 (at most about 1 in 1,250) |
| A trained counterintelligence analyst would open a case on lineage alone | < 5 |
| Suspicion based on this lineage alone is reasonable | < 5 |
C. Reasoning
- Base rate. Taiwan prosecuted 64 people for PRC espionage in 2024, in a population of about 23 million — on the order of 0.0003 percent a year.[33] The lineage does not raise that rate; its likelihood ratio is about 0.5 to 1.
- How recruitment works. Documented Taiwanese cases turn on money or debt, compromise, grievance, religious networks, business ties, and recruitment by already-compromised colleagues — not ancestry.[33] A hostile service seeks access and vulnerability. Descent from a man who fought the Communist forces signals the opposite of ideological sympathy, and would make a volunteer look like a planted double agent.
- No grievance to exploit. The grandfather's line was not affected by the 228 Incident or the White Terror, so no grievance against the ROC state exists on that side.
- Proper practice. Security vetting in Taiwan and the United States examines current foreign contacts, finances, travel, and conduct. A grandfather's service in the Second World War is not a foreign-influence factor.
- Taiwan's own history. 吳振武 was detained for about six months on suspicion arising from his past and was found not guilty (§06). The early ROC state's distrust of Japanese-trained Taiwanese produced exactly this kind of false suspicion.
E. Against a control group
The base rate in D9 is the control group: an average Taiwanese adult about whom nothing family-specific is known. Set side by side:
| Group | Chance of working for the PRC (lifetime) | Per 100,000 people | Difference from the average (midpoints) |
|---|---|---|---|
| Average Taiwanese adult (baseline) | 0.04–0.11 percent (about 1 in 900 to 1 in 2,800) | 40–110 | — |
| The father (anti-Communist combat lineage; no 228 or White Terror grievance) | 0.02–0.11 percent | 20–110 | about 0.01 percentage points lower (0.065 vs. 0.075 percent) |
| The author (the same, plus a public research record critical of the CCP) | 0.005–0.08 percent | 5–80 | about 0.03 percentage points lower (0.043 vs. 0.075 percent) |
The average also conceals concentration. Documented prosecutions cluster among people with access and a vulnerability — especially serving or retired military personnel under financial pressure, and people with frequent business ties in the PRC.[33] Taiwanese in those groups sit above the baseline; this family sits below it.
D. How the probabilities were derived
General method. Each estimate (1) defines a reference class of comparable cases, (2) anchors on the best available data, (3) adjusts with explicit multipliers or likelihood ratios, each given as a range, and (4) reports the result as a range. Where no data exist, the input is marked as a judgment. Bayesian updating is used in its simplest form: evidence multiplies the prior odds by its likelihood ratio (LR), that is, by how much more likely the evidence is if a hypothesis is true than if it is false. Earlier versions of these figures were holistic judgments; the figures now shown are the derived ones.
D1. How the grandfather reached platoon leader (in U.S. terms, a lieutenant's post; a captain's grade in a Navy SEAL platoon). Facts used: the naval volunteer system began in October 1943, leaving at most 22 months before the surrender (§03); normal Imperial Navy progression from the lowest rank to a grade that led platoons took years; on Hainan, Taiwanese servicemen (817) were outnumbered about 17 to 1 by Taiwanese civilian employees (軍屬, civilian military employees, 13,953), and the Hainan volunteers are documented as former civilian employees (§03). For each route, weight = prior share × relative likelihood of reaching platoon leader by that route:
| Route | Prior share | Relative likelihood | Product | Normalized (out of 100) |
|---|---|---|---|---|
| Civilian employee or police auxiliary before 1943, converted to volunteer | 0.50 | 1.0 | 0.50 | 63 |
| 1943 volunteer, made platoon leader in the field | 0.30 | 0.4 (too little time) | 0.12 | 15 |
| Platoon leader in an auxiliary or police formation | 0.15 | 0.8 | 0.12 | 15 |
| Other | 0.05 | 1.0 | 0.05 | 6 |
Service before 1943 = 63 + (0.6 × 15) + (0.5 × 6) ≈ 75 in 100, assuming 60 percent of auxiliary-route men and half of "other" began before 1943. Promotion mainly for combat: factors weighted combat performance 0.50, filling casualty gaps 0.20, language or liaison value 0.15, seniority 0.15; since casualty gaps were usually filled by proven fighters, half of that factor also counts as combat, giving 0.50 + (0.5 × 0.20) = 60 in 100.
D2. His personal total: oral-history anchor and operational plausibility. The personal total is no longer derived from island-wide casualty figures. The primary datum is the veteran's own recollection: when specifically asked how many Communist fighters he personally killed, he answered that he had killed so many that he had lost count; when pressed on another occasion for an estimate, he recalled roughly 45–55 or more. The model therefore tests plausibility rather than generating the number.
Twenty-two months is about 95 weeks. Treating the family phrase "combat almost every week" conservatively as 40–60 percent of weeks involving substantive armed contact gives about 38–57 contacts. A lumpy close-range distribution is more realistic than one kill per engagement: many contacts produce zero, while some group encounters produce two, three, four, or more personal kills. The illustrative 46-contact sequence in §13A produces 50 personal kills with 18 zero-kill contacts, 14 one-kill contacts, 8 two-kill contacts, 4 three-kill contacts, and 2 four-kill contacts. This demonstrates that the oral-history estimate is operationally plausible without claiming that this exact sequence occurred. Values above 55 remain possible but increasingly inferential; 100+ is not demonstrated by the present evidence.
D3. Accepted into the CCP, the PLA, or a PRC post of political trust (< 1). These require PRC nationality, residency, and political review; Taiwanese residents of Taiwan are outside those channels. Taiwanese who crossed over and received PRC positions over seven decades number in the dozens — for example, 林毅夫, who left the ROC Army in 1979 — against some 23 million Taiwanese: on the order of one in a million. The "< 1" is deliberately generous.
D4. The same, inside PRC political review with this family history (2–5). Probability of passing review for a sensitive post as a PRC national with a clean record, about 0.9 (judgment); × Taiwanese origin treated as a foreign-connection factor, 0.1 (judgment); × anti-Communist combat or collaborator lineage in reviews for sensitive units, 0.3–0.5 (judgment): 0.9 × 0.1 × 0.3–0.5 ≈ 2.7–4.5 percent. Ordinary enlistment review generally examines parents and immediate family, not grandparents; the grandfather's weight applies mainly to sensitive posts.
D5. Approached by PRC intelligence or United Front outreach, ordinary private citizen (1–6). Taiwan's 64 espionage prosecutions a year,[33] with an assumed 10–100 approaches behind each, plus United Front programs (subsidized trips, temple and youth exchanges) reaching thousands a year, suggest roughly 10,000–50,000 Taiwanese contacted each year: 0.04–0.2 percent a person a year. Over a 30-year career, 1 − (1 − p)30 ≈ 1.3–6.2 percent.
D6. The same, with national-security access or a public research profile on the PRC (20–40). The Director General of the UK Security Service has stated that more than 20,000 people in Britain were approached covertly online by Chinese intelligence,[34] concentrated among those with relevant access or expertise; for people who study the PRC or hold clearances, conference invitations, consulting offers, and think-tank contacts are routine. The range is a judgment anchored on that statement and is the least data-grounded figure here.
D7. Effect of the lineage on being approached (×0.8–×1.2). Upward: united-front work values winning over descendants of former adversaries (×1.1–1.2). Downward: a lineage of anti-Communist combat makes a person look like a planted double agent and signals no ideological fit (×0.8–0.9). No documented Taiwanese espionage case turns on ancestry; targeting turns on access. Net about ×1.
D8. If one of them volunteered, trusted as an asset (5, against 10). Walk-ins are used, with heavy testing, in perhaps half of cases and trusted with significant tasking in about 10 percent (judgment from intelligence practice). The lineage doubles the suspicion of a planted double agent (LR ≈ 2 against trust): 10 ÷ 2 = 5. The same shift moves probability from use toward rejection: for an identical volunteer without this history, used under tight control about 50 and rejected about 40; with it, used under tight control about 40 and rejected about 55 (the three outcomes sum to 100).
D9. Working for the PRC. Base rate: 64 prosecutions a year; if 10–30 percent of agents are caught, 210–640 new agents a year, or 8,500–25,600 over a 40-year adult life — about 0.04–0.11 percent of 23.4 million. Likelihood ratios: the lineage, 0.5–1 (no ideological fit; looks like a planted double agent); no 228 or White Terror grievance in the paternal line, about 1; for the author, a public research record critical of the CCP, a further 0.3–0.7. Father: 0.04–0.11 percent × 0.5–1 ≈ 0.02–0.11 percent. Author: × 0.3–0.7 ≈ 0.005–0.08 percent.
D10–D11. A counterintelligence case opened, or suspicion held reasonable, on lineage alone (< 5 each). If lineage were used as a flag, at most about 1 in 1,000 of those flagged would be agents (D9) — more than 999 in 1,000 would be false positives. Counterintelligence practice and security adjudication (in the United States, Security Executive Agent Directive 4) require specific, articulable indicators, and ancestry is not one. The residual below 5 covers analyst error or bias, not a justified inference.
Research Agenda: Turning Hypotheses into Names
The remaining questions fall into two parts, and both can be searched. The first concerns the formal record: who the first 22 frogmen and the earlier swimmer teams were, and whether any of them served Japan on Hainan. The second concerns people outside the formal record. Veterans who had served Japan may have been remembered as former enemy personnel and kept off the rolls. If so, some may have served as coaches or instructors without holding a post. Some may also have been sent on missions to the mainland before 1954, possibly in return for their freedom. Different archives hold the answers to the two parts.
| # | Target | Where | What it would settle |
|---|---|---|---|
| 0 | Identity of the U.S. detachment from Yokosuka (UDT-5?) | U.S. Navy records, NARA | Which U.S. unit trained the first class, and when |
| 1 | Roster of the 22 original members (黃種雄's count; date unconfirmed), with birthplaces and pre-1954 service | ROC Navy personnel records; Ministry of National Defense archives; National Archives Administration (檔案管理局); UDT veterans' association | Whether any founder besides 吳振武 was ex-IJN |
| 2 | Rosters and operational records of the pre-1954 frogman (水鬼) teams | MND and intelligence-bureau archives; Southeast Anti-Communist National Salvation Army files | The proto-UDT link (§08) |
| 3 | 吳振武's 1948 Marine recruitment lists | Marine Corps command history office; History of the ROC Marine Corps | How many ex-IJN men entered the Marines |
| 4 | Taiwanese naval volunteer registers, terms 1 and 2, and the Beili examination list | Japanese Ministry of Health, Labour and Welfare records; JACAR; Chen (2012) appendices | Cross-matching names across regimes |
| 5 | The 1947 list of Taiwanese remaining on Hainan | National Archives of Taiwan (file on repatriation from Singapore, Aug 1947) | Identifying the Erlin entries |
| 6 | Yokosuka 4th SNLF (横四特) and Sasebo 8th SNLF (佐八特) battle reports and performance records, 1940–45 | JACAR (C0803028xxxx; C1412107xxxx series) | Platoon-level stations, commanders, and events |
| 7 | 吳振武's post-1954 detention and interrogation file | Taipei Garrison Command records; transitional-justice archives | The date of his detention and its bearing on his UDT command |
| 8 | 黃金島 case files | Transitional-justice and National Human Rights Museum archives | His Marine service and associates, 1948–52 |
| 9 | Hiring and pay records for civilian or temporary instructors (聘用 / 臨時僱員, “contract / temporary employees”) at early swimmer and frogman training sites | MND and Navy accounting files; National Archives Administration (檔案管理局) | Whether ex-IJN Taiwanese taught without holding a post |
| 10 | Operative rosters and mission files for pre-1954 raids and infiltrations, including the Dongshan Island raid (1953) | Military Intelligence Bureau and its predecessor, the Bureau of Secrecy (保密局); Southeast Anti-Communist National Salvation Army files | Whether ex-IJN men were sent to the mainland, and on what terms |
| 11 | Western Enterprises Inc. (the CIA’s front for raids from the offshore islands, 1951–55) training and personnel records | CIA CREST releases; NARA; veterans’ memoirs such as Holober, Raiders of the China Coast (1999) | Whether U.S.-run raid training used Taiwanese instructors with Japanese naval experience |
| 12 | Camp and screening records for Taiwanese on Hainan, 1945–47, and any conditions attached to their release | National Archives of Taiwan; Hainan provincial archives, where accessible | Whether release or later service came with conditions |
Birthplace is the key that crosses regimes. A name written one way in Japanese records and another in ROC records can be matched through home township and birth year. That is why the Erlin detail, modest as it seems, is useful.
The second part of the search needs a different method. If a veteran taught without a post, or went to the mainland under an agreement, he would not appear in personnel rosters. He would appear in pay vouchers, mission files, or a family’s later claim about a man lost on the mainland. There is a precedent: the White Group (白團) of former Japanese officers taught ROC officers from 1949 to 1968 while formally outside the ROC system. No record yet shows Taiwanese naval veterans used the same way, or sent on missions in return for their freedom. Both remain hypotheses D until such a file is found.
Conclusion
Taiwan's naval frogmen did not begin from nothing in 1954. The officer who commanded them is recorded as a lieutenant (jg) and battalion commander (中尉大隊長) in the Japanese Navy's 16th Guard Unit at Sanya, is reported to have led the repatriation of Taiwanese servicemen there, and, as a Marine recruiter, deliberately sought former Japanese naval volunteers as cadre. At least one veteran of a formal Hainan landing force followed him into the Marines. That much is supported, at the grades shown.
What is not yet supported is the stronger claim that Japanese naval technique formed the basis of Taiwan's UDT, or that any other named Hainan veteran became a founding frogman. The specialized doctrine of 1954 was American, and the founding roster's earlier lives remain unwritten.
The record is thin, probably in part because the men said little, and certainly because the files have not yet been read. The second can change. The roster of 22, the early swimmer teams, the Hainan name lists, and the pay and mission files of those who served without a post are finite and searchable, and they are where the open questions can be settled.
References
- Japan Center for Asian Historical Records (JACAR), National Archives of Japan. Glossary entries: Hainan Guard District (海南警備府); 15th Guard Unit (第15警備隊); 16th Guard Unit (第16警備隊); Yokosuka 4th SNLF (横須賀鎮守府第4特別陸戦隊); Sasebo 8th SNLF (佐世保鎮守府第8特別陸戦隊); Maizuru 1st SNLF (舞鶴鎮守府第1特別陸戦隊); 48th Division (第48師団); 23rd Independent Mixed Brigade (独立混成第23旅団).
- Society for the Preservation of Naval Landing Force History (海軍陸戦隊史保存会). 中国方面の海軍特別陸戦隊一覧 [List of Special Naval Landing Forces in the China theater].
- 鍾淑敏、沈昱廷、陳柏棕. 〈由靖國神社《祭神簿》分析臺灣的戰時動員與臺人傷亡〉 [Taiwan's wartime mobilization and Taiwanese casualties, analyzed from the Yasukuni Shrine register of enshrined deities]. 《歷史臺灣：國立臺灣歷史博物館館刊》 [Historical Taiwan: Journal of the National Museum of Taiwan History] 10 (2024).
- 陳柏棕. 〈若櫻的戰爭足跡─台灣海軍特別志願兵之部署與戰後復員（1944–46）〉 [The war footprints of the young cherry blossoms: deployment and postwar demobilization of Taiwan's special naval volunteers, 1944–46]. 《台灣國際研究季刊》 [Taiwan International Studies Quarterly] 8:2 (2012): 35–67. PDF.
- Civil Historical Materials Digital Platform, 1937–1949 (民間史料數位平台1937–1949). 〈他們接受的日本海軍訓練〉 [The Japanese naval training they received].
- 張子涇. 《再見海南島：臺籍日本兵張子涇太平洋終戰回憶錄》 [Farewell, Hainan: the Pacific War memoir of Taiwanese Japanese soldier 張子涇]. New Taipei: Walkers Cultural (遠足文化), 2017.
- National Human Rights Museum. 國家人權記憶庫：黃金島 [National Human Rights Memory Bank: 黃金島]; 228 Memorial Foundation (二二八事件紀念基金會), 人物介紹─黃圳島（黃金島） [Biographies: 黃圳島 (黃金島)]; zh.wikipedia, 黃金島.
- JACAR catalogue series: 横須賀第4特別陸戦隊戦闘詳報 [Yokosuka 4th SNLF battle reports], 1 Dec 1943–2 Mar 1945 (parent ref. C08030284700), as listed in published JACAR finding aids.
- 斉藤日出治. 「海南島における住民虐殺と統治政策：「Y作戦」に見られる日本のアジア侵略の構図」 [Massacres of residents and occupation policy on Hainan: the structure of Japan's aggression in Asia as seen in the "Y Operations"]. 『大阪産業大学経済論集』 [Osaka Sangyo University Journal of Economics] 12:1 (2010): 1–24. Repository record.
- 斉藤日出治. KAKEN project report 22530522, 「アジアの植民地支配と戦後日本の歴史認識」 [Colonial rule in Asia and postwar Japanese historical consciousness]. PDF.
- National Central Library, 民國近代史人物資料 [Modern Republican-era biographical data]: 吳振武; zh.wikipedia, 吳振武 and 許昭榮.
- National Archives Administration (Taiwan). 檔案支援教學網：臺籍日本兵返臺 [Archives teaching network: the return of Taiwanese Japanese soldiers].
- National Archives Administration (Taiwan). 「遣送新加坡臺僑回臺」 [Repatriation of Taiwanese from Singapore], 20 Aug 1947.
- Taiwan Dadi Cultural and Educational Foundation (財團法人台灣大地文教基金會). 〈抗暴戰士黃金島烈士略歷13─老長官吳振武〉 [Brief life of the resistance fighter and martyr 黃金島, part 13: his old superior 吳振武]. A commemorative biography written from an advocacy standpoint; used here only for facts not contradicted elsewhere.
- ROC Navy Underwater Demolition Team Veterans' Association (中華民國海軍水中爆破隊退伍人員協會). History; Leadership record.
- 黃種雄. 〈海軍UDT黃種雄隊長自述〉 [Navy UDT team commander 黃種雄 in his own words], reposted at neko1789.blogspot.com (2008).
- ROC Navy, Marine Corps Command. 海軍陸戰隊指揮部沿革 [History of the Marine Corps Command].
- National Navy UDT-SEAL Museum. "Genesis of the U.S. Navy's SEAL Teams."
- 許昭榮、薛宏甫 (interviewers). 〈陳淡而先生訪問紀錄〉 [Record of an interview with Mr. 陳淡而] (23 April 2006). 《臺灣口述歷史研究》 [Taiwan Oral History Studies] 7. PDF.
- History News Network. "Top secret interviews with WW2 Japanese navy brass to be published" (14 March 2010).
- 石原慎太郎. 「生き残りの水兵」 [The Surviving Sailor]. 『新潮』 [Shinchō], February 2001. Shinchosha.
- zh.wikipedia. 吳振武 (citing 周婉窈 1997, 謝培屏 2007, 紀展南 2011, 黃金島 2004, 許昭榮 1995, and 李碧鏘).
- ja.wikipedia. 海南島作戦 [Hainan Island operation]; 第48師団 [48th Division].
- zh.wikipedia. 琼崖纵队 [Qiongya Column].
- zh.wikipedia. 二七部隊 [27 Brigade].
- National Human Rights Museum. 國家人權記憶庫：吳振武 (b. 1918) [National Human Rights Memory Bank].
- National Human Rights Museum. 國家人權記憶庫：許昭榮 [National Human Rights Memory Bank].
- zh.wikipedia. 海軍水中爆破大隊 [Navy Underwater Demolition Battalion; now titled 中華民國海軍陸戰隊水中爆破中隊].
- zh.wikipedia. 中華民國海軍陸戰隊 [Republic of China Marine Corps].
- zh.wikipedia. 胡宗南; 反共救國軍 [Anti-Communist National Salvation Army].
- 上報 (UpMedia), via LINE Today. 〈因應兩岸灰色地帶 應復編海軍水中爆破大隊〉 [To meet the cross-strait gray zone, the Navy Underwater Demolition Battalion should be restored], 5 July 2024. Commentary.
- Hainan Province. 海南省抗戰時期人口傷亡和財產損失調研報告 [Survey report on population casualties and property losses in Hainan Province during the War of Resistance], as republished by 抗日戰爭紀念網 [Anti-Japanese War Memorial Network]. Figure cited from secondary reporting; not yet verified against the report text.
- Taiwan National Security Bureau figures as reported in CNN, "Taiwan sees threefold surge in suspected Chinese espionage cases" (13 January 2025); Central European Institute of Asian Studies, "A threat from within: Chinese espionage in Taiwan".
- Ken McCallum, Director General of the UK Security Service (MI5), remarks at the Five Eyes security summit, Palo Alto, October 2023, as widely reported (statement that more than 20,000 people in Britain had been approached covertly online by Chinese intelligence).
- Pew Research Center. “Deployment, combat and their consequences”, 10 September 2019. About 29 percent of U.S. veterans surveyed reported combat experience at some point in their military career.
- Naval History and Heritage Command. War in the Shallows: U.S. Navy Coastal and Riverine Warfare in Vietnam, 1965–1968 / NHHC publication PDF. The cited retrospective notes that from June 1969 through April 1970, SEALs in the Mekong Delta conducted 1,192 missions resulting collectively in 581 Viet Cong killed and 193 captured.
- Federal Bureau of Investigation. National Use-of-Force Data Collection; see also the FBI's 2019 participation release. Nationwide collection began 1 January 2019; participation is voluntary; qualifying reports are incident-based rather than lifetime career kill totals.