I compared every published transliteration in the CDLI corpus — 135,198 cuneiform texts — against every other, looking for pairs that share long runs of rare sign sequences but are nowhere recorded as related. These are the ten strongest results: tablets in museums thousands of kilometres apart that appear to be pages of the same ancient account books.
When the antiquities trade dispersed the archives of Girsu and Umma around 1900, tablets that had lain in the same basket for four thousand years ended up in Berlin, Paris, Brussels, Copenhagen, London, Moscow, St. Petersburg, New Haven and Cambridge MA — and were published in different series, by different scholars, over seven decades. Text overlap can put the baskets back together. Six of the ten pairs below interlock into a single cluster: the barley-ration lists of the é-munus, the “women's household” of Girsu, issued under Sasa, wife of Urukagina, ca. 2350 BC.
Each entry shows the two publication designations, the holding museums, a sample of the matched sequences (normalized transliteration; ⧸ separates sequences, | marks line breaks), and links to both CDLI records for verification.
VAT 04612 — Vorderasiatisches Museum, Berlin · SM 1904.04.001 — Harvard Museum of the Ancient Near East
VAT 02333 — Vorderasiatisches Museum, Berlin · Erm 14976 — State Hermitage, St. Petersburg
SM 1904.04.012 — Harvard Museum of the Ancient Near East · MRAH O.0649 — Royal Museums of Art & History, Brussels
AO 13320 — Louvre, Paris · Erm 14001 — State Hermitage, St. Petersburg
NMC 05406 — National Museum of Denmark, Copenhagen · BM 026952 — British Museum, London
AM 07 · Erm 14002 — State Hermitage, St. Petersburg
SM 1904.04.011 — Harvard Museum of the Ancient Near East · MRAH O.0647 — Royal Museums of Art & History, Brussels
AO 13320 — Louvre, Paris · SM 1904.07.002 — Harvard Museum of the Ancient Near East
SM 1904.07.002 — Harvard Museum of the Ancient Near East · Erm 14001 — State Hermitage, St. Petersburg
GMII — — Pushkin Museum, Moscow · YBC 03666 — Yale Babylonian Collection
Every transliteration was reduced to a normalized stream of sign tokens and indexed by overlapping four-token sequences. Sequences occurring on more than a dozen tablets — the standard formulae of administrative prose — were discarded; the rest were scored by rarity. Pairs already known to belong together (exemplars of the same composition, entries cross-referenced by CDLI, texts sharing a composite edition) were removed. As a control, the same comparison independently re-found 36 pairs that are cross-referenced in the catalogues, among them a duplicate recorded only in a 1984 collation note (Gomi, Orient 20: MCS 1, 54 // MCS 6, 10) — which is what convinced me the method finds real connections, not coincidences.
For each pair above: the CDLI records on both sides (no cross-references); the BDTNS records of all Ur III tablets involved (no cross-references, no remarks — though BDTNS's dates are what revealed the same-year, nearby-month pattern); Dahl's note on Ur III text duplicates (CDLB 2003:5), which covers the Umma and Drehem gaba-ri copies and none of these; and general literature searches, which turn up no publication citing both members of any pair. What I could not consult are the print commentaries — Selz's AWEL and FAOS 15 above all — so it is possible some of the ED IIIb links are noted there. If they are, I would genuinely like to know; if they are not, these connections appear to be unrecorded.
Shared rare text between two account tablets can mean three things: a true duplicate or ancient copy; two instalments of the same periodic account series, listing the same personnel in consecutive months; or two fragments of one original document. The dated Ur III pairs point clearly to the second — scattered pages of the same ledgers. For the undated ED IIIb cluster the question is open. All three readings would make these pairs worth recording; none of them is recorded today.