Presences and voices reported at World War II monuments in southern Okinawa
A caller who gives their name as [quote withheld] tells Art Bell that while serving in the Marines in Okinawa, Japan, the caller and a Japanese girlfriend visited the southern area with its World War II monuments at sunset. There they felt presences around them, and the caller heard voices they could not understand. The caller says their girlfriend ran back to the car in tears and never told them what she had heard, seen, or felt. Bell responds by saying he had lived on Okinawa himself and calls it “a haunted island.”
This clip comes from the source recording. AI chose where it starts and ends, so it may start or end a little early or late. The word-for-word transcript is not published. You can hear the source recording on the Internet Archive (file “Ghost to Ghost 1996.mp3”, starting at 1:52:37). Is this you, or someone you know? Ask for a correction or removal.
The story as told
The caller, who gives their name as [quote withheld] says they experienced something while the caller [quote withheld] in Okinawa, Japan. Bell interjects that he knows Okinawa and lived there. The caller says they had a Japanese girlfriend and that they were visiting [quote withheld] The caller says they were there watching [quote withheld] The caller apologizes for being nervous, then reports that [quote withheld] and that [quote withheld] The caller says they looked to their Japanese girlfriend and she [quote withheld] According to the caller, [quote withheld] and would not talk about it. The caller says it is [quote withheld] Bell replies that he also lived there, at Kadena Air Force Base and then Naha, and says there were [quote withheld] He concludes [quote withheld] The caller offers no explanation of their own for what they and their girlfriend experienced.
Reported events, in order
- 0:07
The caller says they experienced something while serving in the Marines in Okinawa, Japan.
- 0:15
The caller says they had a Japanese girlfriend at the time.
- 0:19
They were visiting the southern area of Okinawa where the World War II monuments are.
- 0:29
They were there watching the sun go down along the horizon.
- 0:40
The caller reports feeling the presence of people all around them.
- 0:47
The caller reports hearing voices in their heads, like people talking, which they could not understand.
- 0:58
The caller says their girlfriend ran back to the car with tears rolling in her eyes.
- 1:08
The caller says she never told them what she heard, saw, or felt and would not discuss it.
- 1:21
Bell says he lived on Okinawa at Kadena and Naha, describes violent deaths in battles there, and calls it a haunted island.
Claims
Each claim is labeled as something the person telling the story noticed (observation), what they made of it (interpretation), something others told them (hearsay), or something the host said (host statement).
- observation· 0:40
The caller reports feeling the presence of people all around them at the monument site.
- observation· 0:47
The caller reports hearing voices in their heads, like people talking, that they could not understand.
- observation· 0:58
The caller reports their girlfriend running back to the car with tears in her eyes.
- hearsay· 1:08
The caller says their girlfriend never told them what she heard, saw, or felt.
- host statement· 1:21
Bell says he lived on Okinawa at Kadena and Naha and that many people died violent deaths in battles there.
- host statement· 1:41
Bell says that Okinawa is a haunted island.
Scores, with reasons
An AI model gave these scores using a fixed set of scoring rules (scoring-v1). They are editorial judgments and rough guides, not the odds that anything paranormal happened. How scoring works.
A brief but truly firsthand story with some concrete detail of setting (Okinawa, the southern monument area, sunset, the Marines, a girlfriend present). It is consistent and has no contradictions, but it has no date or precise location, and ordinary explanations go unaddressed.
Evidence for and against
- Firsthand story: the caller places themself at the scene.
- Concrete details of setting: Marines, Okinawa, the southern monument area, watching the sun go down.
- The sequence is consistent, and the answers follow the story without shifting details.
- No date, unit, exact site, or other specifics that could be checked are given.
- The central claim (“presence of people,” [quote withheld]) is vague and rests on sensation alone.
- Ordinary explanations such as distant conversation or ambient sound at an exposed outdoor site fit and are not addressed by the caller.
- The labels showing who is speaking switch between the host and the caller within single sentences, so who said some lines is uncertain.
- The story is under two minutes long and ends without development.
- 0:07
- 0:19
- 0:47
What the model inferred (not said on air): The believability score is my own judgment combining source proximity (firsthand, 20 of 25), moderate specificity (9 of 20), consistency (15 of 20), weak resistance to ordinary explanations (6 of 20), and partial calibration (9 of 15). The caller does not claim to know what the presences or voices were.
The story rests on the caller’s word plus a claimed second witness who never speaks on air and never said what she experienced. No documentation, recording, or record that could be checked is presented.
Evidence for and against
- A first name and city are given, which could in principle be checked.
- A second person is said to have been present and to have reacted.
- The second witness is not named, does not speak on air, and is said never to have described her experience.
- No photograph, recording, date, unit record, or other physical evidence is mentioned.
- No independent source is cited for any part of the story.
- Whether the girlfriend’s reaction had any connection to the reported voices is not stated.
- Problems with the labels showing who is speaking make it unclear which lines are the caller’s and which are the host’s.
- 0:58
- 1:08
What the model inferred (not said on air): A claimed but silent second witness places this in the 11 to 25 range (“other witnesses claimed”). Nothing in the transcript reaches the 26 to 40 range, which requires claimed physical evidence or a second witness speaking on air.
The picture of unseen presences all around at a war memorial at dusk, voices they could not understand, and a companion fleeing in tears without ever explaining why is truly unsettling. But there is no threat, contact, harm, or escalation, and the story is very brief.
Evidence for and against
- Unseen presences felt all around them.
- Voices heard in the head that the caller could not understand.
- A companion runs to the car in tears and refuses ever to discuss it.
- The host ends by calling the island haunted after describing violent wartime deaths.
- No malice, physical contact, injury, or pursuit is reported.
- The encounter is a single short episode with no build-up.
- The caller’s manner is not treated as evidence of anything here.
- The girlfriend’s unstated experience could add to or take away from the effect, but it is unknown.
- The labels in the transcript are unreliable, so exactly how the caller’s and host’s lines fit together is unclear.
- 0:40
- 1:08
What the model inferred (not said on air): My rating reflects the combination of an unseen crowd of presences at a battlefield memorial, an unexplained tearful flight by the second witness, and the absence of any direct threat.
The frame, a battlefield or war memorial where presences and voices are sensed, is a familiar ghost-story pattern. But the combination of voices heard “in the head,” a Japanese girlfriend’s unexplained tears, and her lifelong refusal to describe what she experienced is a distinctive detail.
Evidence for and against
- Unusual detail: voices described as heard in the head rather than as sound from outside.
- Unusual detail: a second witness whose experience is permanently unavailable.
- The setting is a particular kind of site (World War II monuments in southern Okinawa) rather than a house.
- Feeling presences and hearing voices at a battlefield or memorial are stock motifs.
- No rare event or unusual combination of events beyond the above.
- The part of the score compared with the rest of the archive is provisional in this pilot. Only my own judgment of the content is reliable here.
- 0:47
- 0:19
What the model inferred (not said on air): My score sits just below the [quote withheld] midpoint because the frame is common but the voices in the head and the silent second witness are not.
Scoring confidence 0.55 · analysis confidence 0.60 · checked against outside sources: no · evidence available: The caller’s firsthand story; Claimed second witness who does not speak on air; The host’s own statements about having lived on Okinawa. The originality score blends the model's score (48) with how rare the story's details are across the archive. This method is still provisional.
Ordinary explanations to consider
These are possibilities only. None is a finding about what happened.
- Distant conversation by other people at or near the monument site was mistaken for voices “in the head,” possibly because it was in a language the caller did not understand. · The caller says the voices were [quote withheld] while the caller was outdoors at a public monument area.
- Expectation and suggestion at a war memorial, combined with low light at sunset, produced a felt sense of presence and unclear impressions of sound. · The caller says they were at [quote withheld] and watching the sun go down.
- Wind, surf, or other outdoor sound at dusk was taken for voices. · The story places the events outdoors at dusk along the horizon, with no source described for the sound.
What stands out
The voices are described as heard “in our head” rather than as sound from outside. The second witness’s experience is left permanently unknown, because the caller says she refused to discuss it. The site is a particular kind of place, World War II monuments in southern Okinawa at sunset, rather than a house or building.
Related stories
Matched by shared recurring details, categories, and wording.