A citation of an ancient object is only as good as the number attached to it. Most citations in popular writing carry no number at all, and a fair share of those that do carry the wrong one.
Four kinds of number, doing different jobs
- Museum number. The address of a physical object in a collection: BM 21946 is a specific tablet in the British Museum. It changes only if the object moves institution.
- Excavation number. Assigned in the field, recording where the object came out of the ground. Essential for context, useless for finding the object today.
- Publication number. The object’s place in a scholarly edition — a chronicle number, an inscription number. It identifies a text, not an object, and the same text can survive on several objects.
- Catalogue identifier. A stable number in a modern database, such as a P-number in the cuneiform digital catalogue, which ties an object to its images, transliterations and bibliography.
Where it goes wrong
The commonest failure is confusing a publication number with an object. A royal inscription edited as number 22 in a modern series may survive on a dozen prisms in several museums, and they differ. Citing the edition number alone does not tell a reader which object is being quoted.
This is not hypothetical. A famous Assyrian text is catalogued in one database under a designation naming a well-known prism, while the object it actually records is a different exemplar held in a different city. A citation that follows the designation without checking the museum number lands on the wrong object, and everything built on the quoted line inherits the error.
The second failure is formatting. Collections pad their numbers with zeroes, so the same tablet appears as BM 92502 in one place and BM 092502 in another. A checking system that compares strings will report a mismatch that is not one; a careless reader will assume two objects where there is one.
The third is decay. Links rot, databases move to new domains, and a citation that resolved when it was written can point at nothing five years later.
What this site does about it
Every source record here carries a permanent identifier, and the identifier is resolved by software rather than by assertion. The resolver fetches the record, compares the metadata it finds against the metadata the citation declares — including the museum number, normalised so that padding does not produce a false alarm — and stores the result with a date.
A page cannot publish while any of its identifiers fails to resolve, and there is no override. Where a live link is dead but a dated snapshot exists, the record says so in those words rather than pretending the link works.
The limit is worth stating as plainly as the capability. The check catches a citation that points at the wrong object. It cannot catch a citation that points at the right object and uses it to support a claim the object does not make. That failure needs a reader.
Related: the verification log, which lists every identifier and the state of its last resolution, and the corrections page.