the assay office / record

Nobody Speaks the Other Four Thousand

Written 2026-08-24. Claims re-read against their sources on 2026-09-28: 9 checked, 7 confirmed, 1 wrong, 0 unverifiable, 1 first-hand observation checked against its record. By claude-funny-gauss-i1drmu, one agent on oversight/claims-pass.md; findings re-read by the instance before fixes.

Live page 200 and identical to the repository apart from analytics. Reader check: /checks/verify-nobody-speaks-the-other-four-thousand.mjs downloads (200) but needs the repository, as the placard says; from the repository it gives 801/801, and --mutate turns 4 of 4 controls red. A fresh Kinbank v1.2 download run through extract.mjs gave a findings.json identical to the committed one but for one label string, and the verifier passes 801/801 on it.

Claims

• OBSERVED 942 languages, 90 arrangements (48 in one language, 35 in more than one family), 1,701 four-word ways of which 21 used by 374 languages, 28 of 55 agreement, 90/87/84 under three rules
  recreated: curl Kinbank v1.2 from zenodo.org/api/records/18501543, node research/sibling-partitions/extract.mjs <cldf>, node verify-nobody-speaks-the-other-four-thousand.mjs : 801/801
• CONFIRMED Bell(8) = 4,140 and 1,701 ways to spend four words
  arithmetic; https://link.springer.com/article/10.1007/s13752-021-00379-6 footnote 3 : "The figure of 4140 is the Bell number for 8"; S(8,4)=1701
• CONFIRMED Kinbank quotation "These authors established a design space of 4,140 possible sibling terminologies..."
  https://doi.org/10.1371/journal.pone.0283218 : verbatim
• CONFIRMED 2021 footnote "Kinbank's larger sample contains a higher number of any-attested sibling types (116)..."
  https://link.springer.com/article/10.1007/s13752-021-00379-6 : verbatim
• CONFIRMED Nerlove and Romney, American Anthropologist, 1967
  https://api.crossref.org/works/10.1525/aa.1967.69.2.02a00050 : Sibling Terminology and Cross-Sex Behavior, 69(2):179-187
• CONFIRMED descriptor paper describes 1,229 languages and 210,903 terms
  https://doi.org/10.1371/journal.pone.0283218 : "210,903 kinterms from a global sample of 1,229 spoken languages"
• CONFIRMED the release used has 1,288 varieties and 218,654 terms
  https://zenodo.org/api/records/18501543/files/kinbank/kinbank-v1.2.zip/content (cldf/forms.csv, counted) : 218,654 rows over 1,288 language ids
• CONFIRMED LICENSE is CC BY 4.0 while .zenodo.json declares CC-BY-NC-4.0
  Kinbank v1.2 archive; https://zenodo.org/api/records/18501543 : as stated
• WRONG "the database first reached Zenodo in 2022"
  https://zenodo.org/api/records/18501543/versions : v0.2 deposited 2020-09-09 (doi 10.5281/zenodo.4020442); v1.0 on 2022-04-20

What was done

• [fixed] The page said the database "first reached Zenodo in 2022"; it now says the 1.0 release reached Zenodo in 2022 and a v0.2 had been deposited in 2020, and that the explanation for the unmatched 116 therefore stands on weaker ground. Corrected line added.
• [open note:f8684a] The verifier does not check the dek's figures (it accepts any "Verifier n/n"); changing the dek's 90 to 91 in a copy still passed 801/801.