The picture, plainly
What we actually have
A retelling, not a citation. What follows is the story as the most widely circulated popular accounts tell it — folklore encyclopedias, translations, and the versions that have travelled furthest. It is synthesized rather than keyed line by line to the dated primary texts, and popular versions drift from what those texts actually say. The cited, source-by-source record begins further down this page, and where the two disagree, the record wins.
A child-sized figure covered in hair, squatting at the edge of a river, eating a fish. Its teeth are the detail: not fangs but file-teeth — a row of narrow, serrated, rasping teeth, drawn deliberately.
That's the gangikozō, from Sekien's 1784 volume.
And it looks exactly like a feral kappa, which is precisely the thing you should not conclude.
One picture, no fieldwork
Being blunt about it: the gangikozō appears nowhere else. Not in classical literature, not in any regional folk tradition, not in a village account or a collected tale. It exists in Sekien's collection and nowhere before it, and modern scholarship classifies it as art-only.
Scholars go further and specifically warn against filing it with the kappa. Same river, same crouch, same approximate size — and no shared pedigree whatsoever. The resemblance is a resemblance, not a relationship.
That warning is worth taking seriously, because the pull to tidy this material into families is enormous and it is how bad taxonomy happens.
Watch it grow a biography
Now the useful part, because this creature is a clean specimen of something that happened to dozens of others.
Sekien, 1784, gives you: a hairy child at a river, with file-teeth. That's the attested content. All of it.
Then the twentieth century arrives. Mizuki's version reads the gangi in the name as a different word — the stepped wooden pier structures on a riverbank — and so renders it "gear-toothed child." And it acquires, in the process:
- bobbed hair - webbed hands and feet - a fish-eating method, head-first - and a folk remedy: a fishmonger who escaped one by throwing it the cheapest large fish he had
None of that is in Sekien. The webbing, the haircut, the diet, and above all that charming fishmonger anecdote — which reads exactly like a village tale — are post-1950 embroidery with no Edo source behind them.
Then the encyclopedias inherit the whole bundle, undifferentiated, and print it as the creature's folklore.
So a single plate with one attested feature became, within two centuries, a creature with anatomy, behaviour, diet, and a traditional method of escape. And the only way anyone can tell is that somebody kept track of which century each piece arrived in.
The fishmonger story is genuinely good, incidentally. It deserves to be enjoyed. It just needs its date on it.
Field guide
At a glance
- Behavior
- Squats hair-covered at the water's edge, file-like teeth bared, eating fish — per Sekien's one plate, which is the entire record.
- Powers
- The teeth are the sole attested detail.
- Weaknesses
- The fishmonger-bribe story (throw it the cheapest large fish) is Mizuki-era embroidery with no Edo source.
- Oddities
- Scholars explicitly warn against filing it as a kappa: same river, same posture, no shared pedigree.
- A clean accretion specimen — Sekien drew teeth; Mizuki supplied diet, hairstyle, anatomy, and a folk-remedy; the encyclopedias inherited the bundle.
Overview
A hair-covered child-figure squatting at the water's edge, file-like teeth bared, eating fish — Sekien, Konjaku Hyakki Shūi (1784) [sekien-1784]; [ja-gangikozo]. Also written 岸崖小僧 ("cliff-bank child"). It looks like a kappa's feral cousin — and that resemblance is precisely what the scholarship tells you not to trust.
The record
- The provenance verdict: gangikozō "appears nowhere else in classical literature or folk traditions beyond Sekien's collection" — modern sources classify it art-only, and scholars explicitly caution against filing it into kappa lore [ja-gangikozo]. One picture, zero fieldwork, full stop.
- Mizuki's layer: the gangi pun read as 雁木 — wooden gear/stepped-pier structures — so "gear-toothed child"; his account adds bobbed hair, webbed hands and feet, fish eaten head-first, and the fishmonger who escaped one by throwing it the cheapest large fish [ja-gangikozo]. Note for the Mizuki-invention audit: every one of those details is post-Sekien embroidery — the bribe-anecdote has no Edo source.
The reading
A clean specimen for the authenticity-spectrum's methods sidebar: this is how a one-plate invention accretes a biography. Sekien draws teeth; Mizuki supplies diet, hairstyle, anatomy, and a folk-remedy; the encyclopedias inherit the package; and two centuries later the site's job is to un-bundle it in public. The kappa cross-reference earns its place only as a warning label — same river, same posture, no shared pedigree. File the fishmonger story under "modern accretion, charming, uncited."
Two-track labeling
The image
Citations
The record
- [ja-gangikozo]「岸涯小僧」ja.wikipedia via WebFetch (retrieved 2026-07-26), Tier C carrying Mizuki.modern reference Japanese Wikipedia's article on this subject, as retrieved in 2026 — often carries regional detail the English article lacks.All-source record
- [sekien-1784]per 02-HISTORICAL-CATALOGS.md.Edo, 1784 Sekien's fourth and final catalog (Gazu Hyakki Tsurezure Bukuro) — the tsukumogami volume: household objects come alive.All-source record
Where the evidence thins
Open questions
- Mizuki-invention audit line item (which encyclopedia first printed the fishmonger anecdote?).