storiesMost of them are for children

The claim is in this repository, in this project's handwriting.

`data/vocab/against.json` lists twenty-nine harms a charm can be said to answer. Under the key `child-harm`, labelled "Harm to a child", the gloss read: "the largest single class of amulet in the world". No source was attached to it. It is an assertion made by the author of a controlled vocabulary, in a file nobody reads, and it then governs how every record in the atlas gets coded.

That is the worst place for an unsourced claim to sit, because it is upstream of the evidence. So this record counted — and the gloss has since been changed on the strength of what the count found. The last section says what it now reads.

Counted across the 286 records in `data/nodes/`, the finished set.

200 records carry an `against[]` block at all. Between them they hold 736 claim rows — one row per claim, never per charm, so a charm said to stop three things appears three times.

The ranking of those rows:

bad luck, unspecified — 90
illness — 83
the evil eye — 67
ghosts and hungry spirits — 50
demons and night-hags — 43
harm to a house or a field — 40
harm to a child — 39
another person's sorcery — 39
infertility — 30
danger to the dead — 26

Harm to a child is joint seventh, tied with sorcery. Add the separate key for death in childbirth, which carries 18 more rows, and the two together come to 57 rows across 50 distinct records — which puts the combined class fourth, behind unspecified bad luck, behind illness, and behind the evil eye.

Of the 109 amulet records, 22 name harm to a child or death in childbirth. Eighteen records list `cradle` among the places the object is worn: azabache, the Bes amulet, the cimaruta, the dreamcatcher, the evil-eye bracelet, the figa, the hamsa, the kamea, the nazar, the ojo de venado, the Taweret amulet, wood as a material, and the sign records for the eye, the fig fist and the lion, plus three whole traditions — Egyptian amulet craft, European folk apotropaic practice, and Latin American folk Catholic practice.

So: on this atlas's own coding, protection of infants is among the largest classes and is not the largest. The gloss overstated what the records underneath it said.

Inference — this count was first run at 234 records, when the combined class sat third, ahead of the evil eye. Fifty-two records later it sits fourth and the evil eye has passed it. The claim did not get truer or falser; the sample got bigger. That is worth printing, because a count taken once and quoted forever is how an unsourced gloss gets made in the first place.

The rows are not a uniform thing. They come from four different kinds of evidence and it matters which.

From ritual texts. Mesopotamian incantations name Lamashtu's victims as the unborn and the newly born, and the plaque carries the order "you must not approach (this) person." Ritual texts from Uruk prescribe a bronze Pazuzu necklace against miscarriage, and householders set his image in dwellings for infants. A Greek amulet naming the demon Abyzou orders her off by name — "Flee, flee, Abyzou, [from] Sisinios and Sisinnia; the voracious dog dwells here." — and the Testament of Solomon has the same demon saying she walks the world looking for babies.

From household practice recorded by outsiders. Varro records the Roman custom of hanging a phallic charm at a baby's neck. Roman families gave a boy a bulla nine days after birth; Isidore of Seville defines the lunula as a woman's ornament — "Lunulae are female ornaments in the likeness of the moon, little hanging gold bullae." Jewish families in Alsace, southern Germany and Switzerland set coral on a boy's neckband for the brit milah to draw the look away, into the early twentieth century. Frances Densmore's Ojibwe informants told her in 1929 that the netted hoop caught "any harm that might be in the air as a spider's web catches and holds whatever comes in contact with it".

From Egypt's own material. Ivory apotropaic wands, called birth tusks, became common in the Middle Kingdom, around 2040 to 1782 BCE, and were used to protect expectant mothers and children; the cowroid amulet, imitating a cowrie shell, went into a woman's girdle for the same purpose. Bes and Taweret are household gods of exactly this job, invoked in rituals performed at home rather than in the state temples.

From parents alive now. Edward Tabor, writing in Hektoen International in the autumn of 2023, describes the azabache handed to Puerto Rican, Cuban and Dominican infants within days of birth: a jet fist about 8 mm long or a faceted polyhedron 3 to 4 mm across, with a small red bead on the chain. 29% of the wearers he surveyed had the fist, 71% the polyhedron; 40% of the babies in his clinic who wore one also wore a Virgin Mary medallion. What parents named it against was specific: diarrhoea, respiratory infections, deformities, and above all sudden infant death. One mother's answer to him is the whole subject in eight words: "Of course it works. He's never been sick before this."

Inference — five separate reasons the number above is not the world's number.

The harms are this project's keys, not the carriers' words. Every row exists because somebody here read an account and decided which of twenty-nine keys it fell under. `bad-luck`, the largest bucket at 90 rows, is a residual — glossed "Bad luck, unspecified" — and it is where a claim goes when a source does not say what the charm is for. A charm hung over a cradle and described in its source only as lucky lands in `bad-luck`, not in `child-harm`. The coding rule pushes the count in one direction and there is no way to measure by how much.

One charm, many rows. The ranking counts claims, not objects: an amulet with a long entry generates more rows than one with a short entry, regardless of how many people carry either.

The corpus is one language wide. These records were built from English-language encyclopedia articles, museum pages and a handful of journal and news items. A tradition with thin English coverage contributes few rows whatever its practice looks like on the ground.

Survival is not frequency. Durable amulets, buried in graves that were later excavated, are over-represented against the cloth, dough, herb and thread charms beside them — and charms for children are disproportionately cheap and organic.

And a count of records is not a count of people. Nothing here weighs the nazar, worn by tens of millions, against the tupilaq, which is one row like any other.

What the count does show is narrower and still worth having: across four continents and five thousand years, in traditions with no contact between them, the class of harm that recurs with almost no gaps is harm to a small child. Mesopotamia names a demon for it. Egypt keeps two household gods for it. Rome puts something on the neck nine days after birth. Asturias cuts a jet fist for it. Puerto Rico pins one inside a shirt now.

On the evidence this atlas holds, the sentence "most of them are for children" is not supported and the sentence "more of them are for children than for almost anything else" is.

That is a real difference, and the record prints the difference rather than quietly rewriting the file and moving on. The gloss was softened. `child-harm` in `data/vocab/against.json` now reads: "harm to an infant or a small child; common across the atlas, though not the largest class in it — see /story/most-of-them-are-for-children/". The superlative is gone and the vocabulary now points at the count instead of standing in front of it. A reader who meets the key on the /against/ page is reading a description, not a finding.

One last observation, offered as reasoning and not as fact. Inference — the charms in this class are unusual in that the wearer never chooses them. A bulla goes on at nine days, an azabache within the week, a nazar before the child can see it. Every other class in this atlas is worn by somebody who decided to wear it. This one is put on by somebody else, on behalf of a person who cannot yet be asked — which is why the claimant named in the record is almost always a parent, and almost never the wearer.