How Okiku came to Himeji Castle’s well

Below the keep of Himeji Castle, in the enclosure called Kami-yamazato, the castle’s Okiku well route leads to an old well. The city’s guide to the castle says that Okiku, the heroine of the ghost story Banshū Sarayashiki, was tortured to death and thrown into it.⁶ The same guide says it was originally called the Tsurubetori well, the well where buckets were drawn.⁷ The Hyōgo Prefectural Museum of History adds that Edo-period documents use that older name, and thinks the new one arrived in the early Taishō years, when the castle opened to visitors.⁷

By 1934 the castle’s own printed guide used the new name, and chose its words with care:⁸

The story it was given is the one Himeji still tells.

Hanano’s five cups

The oldest text of the Himeji story has none of those names. It is an entry in Chikusō yawa, “Chikusō’s evening talks”, a collection of anecdotes from around Himeji and Tatsuno.¹¹ The master of the house is Odagaki Shume-no-suke, a retainer of the Yamana family based at Aoyama. The woman is called Hanano. What goes missing is one of five matched cups made of abalone shell, a gift from the Yamana, and the man who hides it is a young samurai, Kasadera Shin’emon, who wanted her. He tortures her to death; her grudge returns night after night, and the pine she was hung from is called the hanging pine.¹²

The manuscript is in a running hand we could not transcribe in full. We could read the names 小田垣主馬助 and 花野, Kasadera’s name, “鮑貝乃盃”, the abalone-shell cup, and the last line of the entry, which ends on the pine.¹² The museum sums up what separates it from the later story: five cups, not ten plates; a heroine who is not Okiku; and no well.¹³

Both tellings have a pine, and the pine has a history of its own. A book of Buddhist tales printed in 1689 has an entry on the roadside pines of Himeji, which wither, turn red and come back green; in the Enpō era (1673 to 1681) the lord had them fenced. It names no maid.¹⁴ The museum reports that Harima kagami (1762) describes the rainy-season pine as a plain wonder planted by a Himeji retainer, with no word of Okiku, and concludes that the famous pine came first and was drawn into the story later.¹⁴ The 1689 entry doesn’t say when its pines turned red, so we can’t tell whether it is the same tree.

Dated 1577, copied in the 1750s

Chikusō yawa ends with a date and a name:¹¹

We know the text only from the copy made by Amakawa Tomochika of Gochaku, who put it into Banyō manpō chiebukuro, fifty volumes of writings on Harima that he gathered between 1752 and 1760: about 125 of them by the museum’s count, about 130 by the gazetteer’s.¹¹ His own manuscript is in Kobe University Library; Chikusō yawa opens volume 40.¹¹ The museum notes that most of the texts he copied are lost in the original, and the Heibonsha gazetteer of historical place names warns that some of them were careless to begin with and that the copies are full of miscopied and missing characters.¹⁵

So the museum’s claim is conditional: if the colophon is true, Chikusō yawa is the oldest book with the Harima story.¹⁵ Shibuya Takehiro questioned the date in a 2019 article; the editor’s note in the issue of Rekishi to Kōbe that carried it says a tale reworked by many hands can’t be taken on trust just because it states a year.¹⁶ On that basis, it says, recent opinion holds that the Harima story is older than the Edo one.¹⁷ Konita Seiji, writing in 1987 without discussing Chikusō yawa, dated the written Himeji tradition no earlier than the kabuki play of 1720 and Amakawa’s compilation of 1760.¹⁷

In print, one piece at a time

Outside Himeji the pieces turn up one at a time. The table follows the date of each text.

The counting ghost is older in print than any Okiku. Honchō koji innenshū, printed in 1689, tells of a samurai of Matsue in the Shōhō era (1644 to 1648) who pushed a maid down an old well because she broke one of his ten treasured plates:¹

A wise monk ended it by calling out “ten” as soon as she reached nine. The whole passage is transcribed and translated. Konita quotes it as the earliest plate-counting story he found.¹

The oldest Edo plate mansion in a surviving book, according to the museum, is the one at Ushigome in Tōsei chiekagami (1712).¹⁸ It has neither a well nor a counting ghost. The jealous wife of a man named Hattori uses a broken Nanking plate, one of a set of ten, as her excuse to starve and then strangle the maid her husband favoured; the maid’s ghost kills the four men who finished her off, then the wife.¹⁹ The name Aoyama first appears in 1720, in a kabuki play set in Harima, Nishiki sara kumai yakata. It is also the earliest text in which Konita found Kiku and the plates together.²⁰ Only its playbill survives, at Tenri Central Library, and a play listed as Sarayashiki opened that year at the Kaneko-za in Osaka.²⁰

Then came the puppet play. Banshū sarayashiki, by Tamenaga Tarobei and Asada Itchō, opened at the Toyotake-za in Osaka in 1741. Its chief retainer, Aoyama Tetsuzan, is overheard by Okiku plotting to poison his lord Hosokawa Masamoto, hides a plate to blame her and kills her, and her ghost drives him to ruin.²¹ Watanabe Shōgo, in the Nipponica encyclopedia, names this play as the version that made the story most widely known.²¹

Shigaraki plates at Kiri no baba

The oldest Himeji plate-counting story we could read in full is in Seiban kaidan jikki, compiled by Haruna Tadanari of Sayo and printed in 1754. Its narrator writes in the Kan’en era (1748 to 1751).³ It names nobody. In the days of the Kodera, a man of rank entertained the house elders with Shigaraki plates for ten guests; a maid broke one and hid the halves under the box. When the master had her count them, she stopped at nine, and he cut her down before she finished explaining.³ Then:

No priest could stop it. The family moved and soon died out, and the house at Kiri no baba, the narrator says, still stood empty “even now”.³ There is no well in it. The story is transcribed and translated from a copy of the 1754 print in the Himeji City Museum of Literature. Konita lists the book among the plate-mansion texts printed before 1758.³

Aoyama Shuzen’s Nanking plates

Four years later came the Edo version. Sarayashiki bengiroku, which Konita calls the first work of the Banchō Sarayashiki and a strong influence on the storytellers who came after, is attributed to the storyteller Baba Bunkō and dated 1758.²² In it the bannerman Aoyama Shuzen owns ten Nanking plates; on the second day of the first month of Jōō 2 (1653) his maid Kiku breaks one, and he cuts off her right middle finger.²² She throws herself into the well of the house, her ghost counts plates, and a monk, Ryōyo, lays her to rest by saying “ten”.²²

Okamoto Kidō’s play Banchō sarayashiki, first staged at the Hongō-za in Tokyo in February 1916, dropped the ghost. His Okiku loves the bannerman Aoyama Harima and breaks one of his family’s plates on purpose, to test his love; when he learns why, he kills her for doubting him.²³

How the well moved into the castle

In 1899 Yanai Masao’s guide to Himeji said that Okiku’s shrine stood inside the grounds of the Jūnisho shrine, where worshippers offered plates with their prayers.²⁴ He had little patience with the story itself:²

He summarised it anyway, from a thirteen-volume novel called Himeji sarayashiki whose author he did not know.² In that version Danshirō takes Okiku to his house west of the castle, ties her to a pine in the garden, and after her death her body is thrown into the deep well in his garden; the counting is heard in that house.² The museum’s account of the Jitsuroku behind Himeji-jō shi agrees that Danshirō tortured her at his house near the castle’s Kuruma gate, but has him take her into the castle and throw her into a well while Tetsuzan watches.² The museum dates the Jitsuroku to the late Edo period; Konita reports a claim that the original was written by a Himeji retainer around 1748, and that the copy once kept at the Jūnisho shrine was lost in the war.⁹

Some time before 1921 the Edo historian Mitamura Engyo paid to see the treasures of the Okiku shrine, and heard the shrine’s own account. In it Tetsuzan cut Okiku down into a well inside the castle.⁴ The shrine showed him the bucket from the well she sank in and nine small red Raku dishes shaped like chrysanthemums. He noted, with a pun on sara, “plate”, that the story he had just been told had no plates in it at all.⁴ Konita found that Mitamura’s account matches the shrine’s printed leaflet.⁴ By 1934 the castle guide was pointing visitors to the Okiku well in Kami-yamazato.⁸

Insects with their hands tied

In 1795 people made much of an outbreak of “Okiku insects”. Yanai, citing old records, says they were believed to come out of the well each year on the anniversary of her death.²⁵ Fujisawa Morihiko (1918) places the outbreak in the Himeji castle town; Mitamura (1921), following an account by the writer Shokusanjin, places it in the disused well of the Kida house at Amagasaki, which had its own Okiku, thrown into a well in 1696 over a needle found in her master’s rice.²⁵²⁶ Mitamura concluded that the Okiku insect became widely known only after 1795.²⁵

The insect is the pupa of a swallowtail butterfly, Byasa alcinous. Takakura Tadahiro, in the Heibonsha encyclopedia, explains the name: its shape looked like a woman tied to a tree with her hands behind her back.²⁷ The museum says the pupae were once sold at summer temple fairs in Himeji.²⁷

Two stories joined

Plate-mansion stories are told in many places. Watanabe gives examples from Kōchi, the Gotō islands and Fukuoka, and the museum reports that Itō Atsushi (2002) found 48 sites between Iwate and Kagoshima.²⁸

Konita’s reading of the early texts is that two separate tales were joined: one about a maid named Kiku abused by her master, and one about a ghost counting plates. He thought they were probably joined in Harima, because until 1758 only the Harima versions have Kiku break a plate.²⁹ The museum describes the same gap from the Himeji side: Chikusō yawa lacks Okiku and the ten plates, and something had to add them.²⁹

The step nobody has found

Neither account can say who took that step. The museum asks when the name Okiku and the ten plates were joined to the Chikusō yawa plot, and answers that the process is not clear.⁵ Konita could not decide whether the joining was the work of the author of the 1720 kabuki play or of an older Harima Jitsuroku.⁵ Between Hanano’s abalone cup, dated by a colophon to 1577, and a Kiku with plates on the kabuki stage in 1720, no text we know of connects the two, and of the 1720 play only the playbill is left.²⁰⁵