Did the creatures of the Shan Hai Jing ever exist?

What the original texts say

The Shan Hai Jing explained in one line: it is a survey, not a story, written in repeating formulas. Fixing that is the first thing to do, because almost every argument about the book starts from the wrong assumption. The Chinese Text Project's edition sets out the eighteen sections: five Classics of Mountains covering the south, west, north, east and centre; four Classics of Lands Beyond the Seas; four Classics of Lands Within the Seas; four Classics of the Great Wilderness; and a final Classic of Lands Within the Seas.

A typical entry names a mountain, says how far and in which direction it lies from the previous one, then lists what grows on it, what metals and stones are in it, and what animals live there — what each looks like, what noise it makes, and what happens to a person who eats it or wears its skin. At the end of each chain of mountains comes the sacrifice the local spirits are owed. That last part matters: the book is as much about ritual obligation as about wildlife.

Penguin's edition, translated by Anne Birrell, describes it as a guided tour of the known world in antiquity, moving outwards from the famous mountains of central China to the lands "beyond the seas", and counts 204 mythical figures inside it, along with early medical prescriptions, omens and sacrifice rites. For dating, the University of California Press puts the compilation between the fourth and first centuries BCE; the Penguin edition gives a wider span, the third century BC to the second century AD. Either way it is a text with a long assembly history and no single author.

How the versions differ

The eighteen-part shape is an editorial decision, not an original one. In his memorial presenting the book to the throne at the end of the Western Han, Liu Xin wrote that the copies he and his colleagues had collated came to thirty-two pian, "now fixed at eighteen". Everything read since descends from that act of tidying.

The catalogues then disagree about what kind of book it is, which is the clearest sign of how unstable its reputation was. The Han History's bibliographic treatise lists thirteen pian and files it under form-and-shape divination. The Sui History lists twenty-three juan with Guo Pu's commentary, under geography. The Old Tang History gives eighteen juan. By the Qing, the imperial catalogue had moved it into fiction. The Qing scholar Bi Yuan proposed a reconstruction: Liu Xiang collated thirteen pian of Mountains and Seas chapters, and Liu Xin added the Great Wilderness chapters to reach eighteen.

Guo Pu, in the Eastern Jin, is the earliest person known to have annotated the whole book. In a 2018 study in the journal of National Taiwan Normal University's Chinese department, Hsieh Hsiu-hui argues that this commentary is precisely what changed the book's standing — it carried the text from being dismissed as "false and reckless" into being read as a classic of broad learning. The pictures came far later still: Richard Strassberg's bestiary reproduces seventy-six woodblock plates from a rare 1597 late-Ming edition, roughly two thousand years younger than the text they illustrate.

For anyone wanting the primary material: the Chinese full text is free at the Chinese Text Project, and there are two standard English translations, Birrell's for Penguin Classics and Strassberg's annotated bestiary for the University of California Press.

How scholars explain it

The argument is two thousand years old. Sima Qian declined to speak about the Kunlun of the Basic Annals of Yu, measuring it against what the envoy Zhang Qian had actually seen; Wang Chong's Lunheng read that refusal as calling the material empty — and elsewhere praised the same book for the sheer breadth of things it had recorded.

The strongest modern version of the sympathetic reading comes from the folklorist Liu Zongdi, who argues the book is a natural history rather than a monster catalogue. Without photographs, an ancient writer describing an unfamiliar animal could only compare its parts to familiar animals, and the comparisons read to us as hybrids. He works through cases: a beast "shaped like a cow, with a snake's tail and wings" is a pangolin; a creature "like a lung, with four eyes and six feet" is a horseshoe crab; a fish "like a chicken with red hair, three tails, six feet and four heads" is a squid, its fins taken for wings and its tentacles for extra heads.

That reading is persuasive in places and cannot be pushed all the way. No identification is universally accepted, and many entries have nothing plausible behind them — they are omens, or gods, or the spirits owed a sacrifice at the end of a mountain chain. The honest summary is that the book mixes observation, hearsay and ritual, and no one has found a rule that sorts them.

One popular claim can be dealt with cleanly: that the creatures are memories of animals from tens of millions of years ago. There is nothing behind it. The last non-avian dinosaurs died around 66 million years ago — argon-argon dating by Paul Renne's team at the Berkeley Geochronology Center, published in Science in 2013, placed the Chicxulub impact at 66.038 million years ago and the extinction at 66.043 million, each to within about eleven thousand years. No human lineage existed to watch, and no fossil or excavation has ever been matched to an entry in the book.

Later retellings

The text kept acquiring layers. Guo Pu also wrote verses in praise of its illustrations, listed in Song book catalogues, so pictures and text have travelled together for a very long time even though what the earliest version looked like is disputed. The 1597 Ming illustrated edition fixed the images most people now picture. The Qing scholar Hao Yixing produced the standard annotated edition, and in the twentieth century Yuan Ke's collated commentary became the working text for Chinese readers.

The strangest retelling is the catalogue drift itself: the same book filed as divination under the Han, as geography under the Sui and Tang, and as fiction by the Qing. Nothing in the text changed. The category that seemed to fit it did.

Verdict

Real or Legend has filed this as disputed. The book is real, old and well documented; the creatures as written did not walk around; and the live scholarly question — how much real observation survives inside those descriptions — is one on which serious researchers genuinely disagree.