Nakatsumaki

The middle scroll opens with a march. Jimmu leaves Himuka in the south, moves east by sea and land, is beaten back at one landing and turns to approach from another direction, and finally enters the Yamato basin with a giant crow going before him. The kami are still present, still intervening, still sending signs, but the action has moved to a map that a reader in 712 could recognise, and the opponents are chieftains with places and names rather than deities.

That is the scroll's real subject: the change of register. It runs to Ojin, and along the way the material becomes less cosmological and more political. The reign of Sujin brings plague and the establishment of cults to appease it. Yamato Takeru is sent out repeatedly against the edges of the realm, kills by disguise and trickery as often as by arms, is given a sword taken from the serpent's tail in the first scroll, and dies far from home, his spirit leaving as a white bird. The expedition attributed to Empress Jingu carries the claim of authority over territory across the strait.

None of this is annalistic in the way the Chinese histories the court admired were annalistic. There are no reign dates in the Kojiki, no year by year entries. The compilers had those materials, and the Nihon Shoki produced eight years later uses them. The middle scroll instead keeps the shape of a story being told, which is one reason it was long regarded at court as the lesser of the two chronicles.

The middle scroll carries more song than any other part of the Kojiki, and the songs are the oldest layer of Japanese poetry to survive in writing. They stop the narrative rather than decorating it. A general sings before dying; a woman sings to a man who has come or failed to come; a sovereign sings on seeing a landscape or a girl by a river. The prose around them often exists to explain who was singing and why, which suggests that in some cases the song came first and the story was fitted to it.

They also explain the book's strange orthography. Old Japanese had no script of its own, and the compilers had only Chinese characters to write with. For most of the text they could write in a Chinese-influenced style that a trained reader could construe. For the songs they could not, because a song depends on its exact sounds, so the characters are used phonetically, one to a syllable, with meaning set aside. The result is a text that switches between two ways of reading the same signs, and it is the phonetic passages that let modern linguists reconstruct the sound of eighth century Japanese at all.

That preservation was not the purpose. The compilers were recording what had been recited, and a recitation cannot be summarised. The scroll's songs survive because the only honest way to write them down was the difficult way.