.png)
Okinawa Civilian Death Toll: The Four Counts That Do Not Add Up (Cold Files 014)
One in four Okinawans died in the Battle of Okinawa. You will find that sentence in encyclopedias, in guidebooks and in most of the serious journalism written about the island. It is the most repeated fact about the battle, and it is usually made by dividing one number by another that was never built to go underneath it.
Part of the Cold Files series.
The population figure is solid. Okinawa Prefecture counted 590,480 people in February 1944. The death figure usually placed on top of it is around 150,000, and that number is the Okinawan subtotal on a memorial wall at Mabuni which counts deaths from 1931 onward, anywhere on earth, soldiers and civilians together. Put the two in one fraction and you get 25.4 percent.
There are four counting systems for the dead of Okinawa. Each is internally honest. No two of them measure the same thing.
The number that was never a count
The figure cited everywhere as the prefecture's official civilian toll is 94,000. Japan's Ministry of Internal Affairs and Communications reproduces the table: 28,228 Okinawa-origin military and army employees, 65,908 soldiers from other prefectures, 94,000 Okinawan general residents, a Japanese total of 188,136, plus 12,520 American dead, for 200,656.
Read the ministry's note on where it came from. The figures were estimated around 1957 by the Government of the Ryukyu Islands, and the published source is a 1996 volume from the prefecture's welfare department, The Path of Okinawa's Relief Administration.
So the world's standard civilian figure for Okinawa is an estimate made twelve years after the battle, by a government under American administration, by the arm that paid war pensions. It was never a census and never a body count, and the table stands unrevised.
Japan then left the islands out of its own accounting. A written Diet question by the member Teruya Kantoku, 7 September 2015, established that Okinawa was omitted from the 1949 Comprehensive Report on Damage from the Pacific War and excluded from the national war damage surveys run from 1977 to 2013.
What the Americans counted
The United States Army's official history, Okinawa: The Last Battle, published in 1948 by Appleman, Burns, Gugeler and Stevens, gives American battle casualties as 49,151, of which 12,520 were killed or missing. The prefecture's table carries the same 12,520, sourced to US military government records.
For the other side its Appendix C gives 110,071 enemy killed and 7,401 captured, while conceding that "data are not available to permit a complete statistical summary..."
Appleman gives no Okinawan civilian death total. For the dead it offers one sentence: "The bodies of many thousands of other civilians lay scattered in the ditches, in the cane fields, and in the rubble of the villages, or were sealed in caves."
Many thousands. That is the entire civilian death estimate in the United States Army's official history.
The Marine Corps monograph is where the American system admits what it had done. Nichols and Shaw, in Okinawa: Victory in the Pacific, record 107,539 counted enemy dead plus 23,764 estimated sealed in caves or buried by the Japanese, plus 10,755 prisoners, for 142,058. Then this:
"Since this casualty total of 142,058 was 'far above a reasonable estimate of military strength on the island,' Tenth Army intelligence officers concluded that approximately 42,000 civilians had fallen victim to artillery, NGF, and air attacks because of their unfortunate proximity to Japanese combat troops and installations."
Roughly 42,000 civilians were inferred inside the enemy casualty total by subtraction. The total was larger than the garrison could have been, so the remainder had to be civilians. It is the most candid number in the American record, and it is not a civilian death toll. It is a correction to a military one.
The rest of the confusion was manufactured in 1944 and 1945, by conscription. Appleman puts it plainly: "the Japanese Army had conscripted almost all males between the ages of fifteen and forty-five." Between 20,000 and 24,000 Okinawans served in the Boeitai home guard, and Army Ministry ordinances of December 1944 allowed boys aged 14 to 17 into the Tekketsu Kinnotai, the Blood and Iron Imperial Corps, of whom around 1,780 were mobilised and around 890 killed.
Put one of those boys, killed at Shuri at fifteen, into the systems. He is inside Appleman's 110,071 enemy dead. He is inside the prefecture's 28,228 Okinawan military, which means he is explicitly not inside the 94,000 civilians. And he is a name on the wall at Mabuni, cut into the same stone as the civilians, undifferentiated.
The civilians who became soldiers after they died
The strangest reclassification happened in peacetime, because of money.
Japan's 1952 relief act for war victims and survivors was applied to Okinawa on 1 April 1953, while the islands were still under American administration. It compensated military personnel and army employees. It did not compensate civilians.
So the Ministry of Health and Welfare sent investigators to Okinawa in 1955 and issued, in July 1957, the Combat Participant Overview Table. It set out twenty categories of circumstance under which a civilian death could be recognised as participation in combat. Municipalities collected declarations, the Ryukyu Government screened them, Tokyo approved them, and payments began in April 1959.
Roughly 55,200 people aged seven and over were certified as combat participants and given quasi-military status, according to the Okinawa Prefectural Archives. The emeritus professor Ishihara Masaie, who has spent a career in these files, puts it at 54,000 of the 94,000 general residents. Either way, more than half of the 94,000 civilians in Japan's official table were simultaneously certified by the Japanese state as participants in the fighting. They are civilians in the death statistics and quasi-soldiers in the pension file, and the same ministry holds both records. After the 1972 reversion the category was extended down to children under six, so infants were posthumously classified as quasi-military personnel. Because that status was military, civilians including those infants were enshrined at Yasukuni, in many cases without their families being asked.
What the screening removed
The deaths this fight is really about happened in caves, and they began before the main landings. American forces took the Kerama Islands on 26 March 1945. On Zamami that day, by the Japanese government's own account, 234 people died, and 53 on Geruma. On Tokashiki on 28 March, at Nishiyama, the same pages record 380 residents dead. The government's own page for Tokashiki describes residents killing themselves and their families with razors, sickles, hand grenades and rope, and says they did so, in its words, "as they had been instructed beforehand."
On 2 April, at Chibichiri Gama in Yomitan, 83 of roughly 140 people sheltering died, about half of them under twelve. At Shimuku Gama, in the same village, within walking distance, about a thousand people sheltered through the same days under the same propaganda. Two men who had lived in Hawaii told them the warnings about what Americans did to prisoners were false, and talked them out of the cave. Essentially everyone lived.
On 30 March 2007 Japan's education ministry announced that seven textbooks from five publishers had been told to alter their accounts of those deaths. The Okinawa Times tracked one sentence through its states. As submitted: "There were residents, who, by the Japanese military, were driven out of shelters or driven into mass suicide." As approved: "There were residents, who, by the Japanese military, were driven out of shelters, or committed suicide."
The army stays in the sentence. It is now attached only to the shelters.
A government written answer to the Diet member Akamine Seiken, issued under Prime Minister Abe Shinzo on 10 July 2007, called the original wording "an expression that risks causing misunderstanding about the actual conditions of the Battle of Okinawa," on the grounds that whether orders were given cannot be definitively determined.
On 29 September 2007 a protest filled Ginowan Seashore Park, and municipal heads from across the prefecture attended. The organising committee announced 116,000. The police released no headcount; Sankei was reported as putting it near 43,000. Every number attached to that crowd came from somebody with a stake in its size.
On 26 December 2007 the ministry accepted corrections from six publishers across eight textbooks. The wording allowed back was "driven into mass suicide by the Japanese military." The word refused was coercion, and the screening opinion itself was never formally withdrawn, which is what the rally had demanded. That is still roughly where the line sits.
What the courts actually found
The courts heard the same question. In August 2005 Umezawa Yutaka, the wartime commander on Zamami, and the brother of the late Tokashiki commander sued the novelist Oe Kenzaburo and the publisher Iwanami Shoten over Oe's 1970 book Okinawa Notes. On 28 March 2008 the Osaka District Court dismissed the claims. Judge Fukami Toshimasa found the Japanese military deeply involved in the mass suicides, resting partly on the distribution of hand grenades, a scarce weapon, to civilians, and on the fact that no mass suicide occurred on Maejima in Tokashiki village, where no Japanese unit was stationed. Then the sentence both sides have quoted ever since: the court said it could not help but hesitate to go so far as to find the suicide order itself. The Osaka High Court affirmed on 31 October 2008 and the Supreme Court closed the case on 21 April 2011.
The headlines were wrong in both directions. No Japanese court has found as historical fact that Umezawa or Akamatsu ordered anyone to die; what the courts decided was whether Oe had reasonable grounds to publish, and they held he did. The screening opinion said no order to any resident can be established; the court only hesitated over two commanders.
But that is not what the screening deleted. It deleted the link between the army and the deaths, and that link the courts found emphatically. The strongest sceptical case, Sono Ayako's 1973 argument that she could find no eyewitness to an order, is an argument about the word ordered. It was never an argument that the army was absent from those caves, and the ministry's edit did not limit itself to the word.
What the monument actually counts
The Cornerstone of Peace at Mabuni was unveiled on 23 June 1995, fifty years after the fighting ended, under Governor Ota Masahide: stone walls in radiating arcs, roughly 2,200 metres of inscribed names.
As of 23 June 2026 it carries 242,659 names. 149,703 from Okinawa Prefecture, 78,365 from other Japanese prefectures, 14,012 Americans, 381 from South Korea, 82 from North Korea, 82 from the United Kingdom, 34 from Taiwan. Ninety-five names were added that year. Three duplicates were removed.
A register that removes names is not a death toll, and the prefecture has never said it is one. Its published policy inscribes everyone who died in the battle between 26 March and 7 September 1945, and then, for Okinawans only, everyone who died anywhere in the world from 18 September 1931, plus war deaths within about a year after the surrender, with survivors of the atomic bombings exempt from the time limit entirely. Three of the Okinawans added in 2026 were hibakusha. Most of the 62 added from other prefectures were crew of the battleship Yamato, sunk at sea on the way.
The 463 Korean names show what the wall is. The Korean memorial tower in the same park commemorates more than 10,000 Koreans brought to Okinawa. To be inscribed, a family must produce official documentation of a death in Okinawa, and Korean conscripts who exist in Korean records only as missing cannot clear it.
So do the arithmetic properly, against the 590,480 counted in February 1944. The prefecture's civilian estimate gives 15.9 percent. Add the 28,228 Okinawan military and you get 122,228, which is 20.7 percent. Only the Cornerstone's 149,703 reaches 25.4 percent, and that figure includes Okinawans who died in China in the 1930s, on the Tsushima Maru, in mainland air raids, and at Hiroshima and Nagasaki.
Kitamura Takeshi's 2019 paper does it another way. Because no death total is fixed, he writes, the rate is said to be one in four or one in five, so he measures Okinawa Island's population instead, February 1944 against December 1945, and finds it fell by 104,663, or 24.3 percent. One in four, reached without a death count at all.
Four systems, four honest answers to four different questions, and a famous ratio that only a fifth method earns.
Sources consulted
Appleman, Burns, Gugeler and Stevens, Okinawa: The Last Battle, Department of the Army, 1948
Nichols and Shaw, Okinawa: Victory in the Pacific, US Marine Corps
Okinawa Prefecture, Cornerstone of Peace inscription totals, 23 June 2026, and the inscription policy
Japan, Ministry of Internal Affairs and Communications, war damage in Okinawa Prefecture
Okinawa Prefecture, The Path of Okinawa's Relief Administration, 1996
Okinawa Prefectural Archives, on the 1957 Combat Participant Overview Table
Ishihara Masaie, Ryukyu Asahi Broadcasting, 8 September 2015
Cabinet written answer to Akamine Seiken, no. 166-474, 10 July 2007
Teruya Kantoku, written question no. 189-408, 7 September 2015
Aniya Masaaki and Steve Rabson, The Asia-Pacific Journal: Japan Focus, 2008
Kitamura Takeshi, Machikaneyama Ronso 53, 2019
More Cold Files: Biafra War Death Toll and No Figure Has Been Accurately Determined.
Next in Cold Files: Srebrenica, July 1995. A United Nations safe area, a Dutch battalion, and a column that walked into the woods.
Cold Files runs alongside The Daily Sitrep. For the morning brief and the long reads as they land, follow Project Leaflet on Telegram: t.me/TheLeaflet
Comments