Mysteries Department
Sign inRequest access
Proto-cuneiform: the archaic texts of UrukResearch sessionsSession 3
Research session 3 · Proto-cuneiform: the archaic texts of Uruk

The archaic names benchmarked against real Sumerian names

Scored by the same procedure, Early Dynastic Sumerian names carry the stock Sumerian name elements four to six times more than chance, and the Uruk III names barely above it. The EN excess is an archive-level difference, and the EN-first pattern is not a name template.

Session 32 Nov 2026Testing3 runs7 evidence1 figure
Session 2 · The archaic names tested against Englund's Sumerian name elementsSession 4 · Dependents-only benchmark: the names gap is not a class effect

Going in the sweep before the work

What changed since Session 22 Nov 2026, 16:37 → 31 Oct 2026, 15:21
0datasets added
0sources added
0news stories matched
0duplicates or retired records
Top next step going in: Dependents-only benchmark
Ranked 1 of 20 · could decide · good odds · ~4 h
STEP

In brief

  1. Real Sumerian names carry the stock Sumerian name elements at odds ratios of 4.5–5.8 from about 2800 BC on; the Uruk III names reach 1.35. The archaic names are measurably not Sumerian-pattern names.
  2. The jump falls between Uruk III and Archaic Ur, the same interval in which the first phonetic name spellings appear.
  3. The EN excess is a difference between archives, not a few odd tablets, and the EN-first pattern is not a name template: the same names are transliterated in both orders.
What this changes: Names question: H-NAMES-SUMERIAN fails its kill criterion and the new answer H-NAMES-NOTSUMERIAN is supported. Central question: new evidence for H-UNKNOWN and against H-SUMERIAN, each limited by the dependents caveat.

What movedNo question changed standing

Key question
What language, if any, do the proto-cuneiform texts record?
+2 evidence, 1 for, 1 against1 next step addedtaken up again
The language is unknown (Englund's agnosticism) · 4 for, 1 againstPossible
Proto-cuneiform records Sumerian · 3 for, 6 againstPossible
Openunchanged
What language, if any, do the proto-cuneiform texts record?this question
Do the roughly 440 personal names in the archaic texts follow Sumerian, Semitic or some other naming pattern?
+5 evidence, 2 for, 3 against1 answer proposed4 next steps done, 1 addedtaken up again
None of these. · 1 for, 0 againstUntested → Possible
The names lack the Sumerian name profile · proposed this session · 5 for, 0 againstPossible
The names favour Sumerian name elements · 1 for, 3 againstPossible
and 1 more answer on the question's page
Openunchanged

What's next3 steps set · 4 done · ranked

1
Officials' names
Could decideFair odds~10 h
On: What language, if any, do the proto-cuneiform texts record?
28score
Details

Test name-like entries attached to officials and institutions in Uruk III, to separate the scribes' own naming culture from the listed workers'.

Taken up
Restrict the Early Dynastic benchmark to names of dependent workers (ED IIIb Lagash ration lists: igi-nu-du8, geme2, dumu) to check that the gap with the archaic names is not a social-class effect.
On: Do the roughly 440 personal names in the archaic texts follow Sumerian, Semitic or some other naming pattern?Set this sessionTaken up in Session 4
Taken up
Mark personal names in the 394 Archaic Ur tablets in sign-name transliteration and repeat the element test directly, removing the reading-to-sign mapping step.
On: When and how did phonetic spelling begin in Mesopotamian writing?Set this sessionTaken up in Session 7
Done this session
Score the Early Dynastic IIIa Sumerian personal names (Oracc ePSD2 PN lemmata for ED IIIa admin texts, or the Fara name lists) with the same element-share procedure, mapping readings to sign names with OGSL, to learn what share a genuine Sumerian name corpus gives. Needs Oracc access from a machine that can reach oracc.museum.upenn.edu.
On: Do the roughly 440 personal names in the archaic texts follow Sumerian, Semitic or some other naming pattern?Set in an earlier session
Done this session
Fit a per-tablet random-effects model of EN share in names to see whether the unprovenanced excess comes from one or two Schøyen tablets or is spread across the group.
On: Do the roughly 440 personal names in the archaic texts follow Sumerian, Semitic or some other naming pattern?Set in an earlier session
Done this session
Score a tagged Early Dynastic IIIa Sumerian name list (Oracc ePSD2 PN lemmata, or the Fara name lists) with the same procedure, to learn what element share a genuine Sumerian name corpus gives.
On: Do the roughly 440 personal names in the archaic texts follow Sumerian, Semitic or some other naming pattern?
Done this session
Fit a per-tablet random-effects model of EN share in names, to see whether the enrichment comes from one or two Schøyen tablets or is spread across the unprovenanced group.
On: Do the roughly 440 personal names in the archaic texts follow Sumerian, Semitic or some other naming pattern?

Next steps live on their questions, where anyone can take one up. Ranked by how likely each is to move a key question, for the time it takes.

What was done

  1. Oracc reached. Oracc's server leaves out its intermediate certificate (InCommon RSA Server CA 2). Adding that certificate, from the address published in Oracc's own certificate, to the trust bundle let the downloads verify normally. The lemmatized Early Dynastic administrative corpora of ePSD2 (ED I–II, IIIa and IIIb, CC0) and the OGSL sign list were downloaded.

  2. Benchmark (next step 1). Every lemmatized personal name was mapped from readings to sign names through OGSL and scored with the same element test as the archaic names. Englund's element list, a length-matched null and a text-cluster bootstrap were used throughout [@englund2009smell] [@andersson2012kingship]. Sumerian names contain an element in 40% (ED I–II), 47% (ED IIIa) and 53% (ED IIIb) of cases, with odds ratios of 5.8, 4.9 and 4.5. The archaic names reach 12.4% and an odds ratio of 1.35.

Figure 1

Real Sumerian names carry the stock name elements four to six times more than chance; the Uruk III names barely do

Token odds ratio of Englund's Sumerian name elements in names against each corpus's other words, with 95% bootstrap intervals; Oracc ePSD2 and CDLI

02468odds ratio (names vs other words)Uruk III names, c. 3100 BC (540)Archaic Ur, ED I–II, c. 2800 BC (326)ED IIIa, c. 2550 BC (11,799)ED IIIb, c. 2400 BC (31,083)
02468odds ratio (names vs other words)Uruk III na…Archaic Ur,…ED IIIa, c.…ED IIIb, c.…
odds ratio95% low95% high
An odds ratio of 1 means names use the elements no more than ordinary words. The jump falls between Uruk III and Archaic Ur, where the first phonetic name spellings appear.
Drawn by the site

  1. Per-tablet heterogeneity (next step 3). EN-name rates vary only modestly across the 41 tablets (chi-square 51.2 on 40 df, p = 0.11). The excess belongs to the unprovenanced Schøyen and Brussels group as a whole, at 16–33% on most tablets, against 0–7% on the excavated Jemdet Nasr tablets.

  2. EN order (next step 2). Instead of collating 80 names on photographs, the transliterations were checked for internal consistency. Of 58 recurring multi-sign names, 22 are transliterated in more than one order, including 5 of the 10 that contain EN. The same name is not written in a fixed order, so EN-first is a tendency, not a template [@born2021tribute]. A pilot look at the MS 3035 line drawing was not recorded as evidence, because the glyph identifications were not secure.

  3. Answers. H-NAMES-SUMERIAN fails its own kill criterion. A new answer, H-NAMES-NOTSUMERIAN, states Englund's point as a comparison with real Sumerian names. It was framed after the benchmark, and is labelled so.

Runs 3
RunMeasuredResultAgainstDecision
Sumerian benchmark: the same element test on Early Dynastic names (Oracc ePSD2) · 1 figureShare of names containing at least one of Englund's elements against a length-matched null; token odds ratio with a text-cluster bootstrapEach corpus's own non-proper-noun lemmatized words (ED) or non-name administrative entries (Uruk III)The Sumerian name corpora are enriched three- to four-and-a-half-fold in every period, and the archaic names are not. This meets the kill criterion of H-NAMES-SUMERIAN: the archaic names do not have a Sumerian onomastic profile.
Per-tablet heterogeneity of EN in the archaic namesPearson chi-square of EN-name counts against a common rate; method-of-moments between-tablet SD; leave-one-tablet-out ratesA common EN rate across all 41 name tabletsThe EN excess is spread across the unprovenanced Schøyen and Brussels tablets and absent from the excavated Jemdet Nasr ones. It is an archive-level difference, not the product of one or two tablets.
Order consistency of repeated archaic namesShare of recurring multi-sign names (same sign set, two or more attestations) transliterated in more than one orderRepeated names are not written in a fixed order, and half the repeated EN names occur in both orders. The EN-first tendency is not a name template, so session 2's speculative EN-first evidence is withdrawn as support. Image collation is not needed to reach this conclusion.
Answers proposed 1
ANS
Answers taken up again 2
QST
QST
Evidence found 7
EVD
EVD
EVD
EVD
EVD
EVD
EVD
Used: methods 2
MTH
MTH
Used: datasets 1
DAT
Consulted: sources 2
SRC
SRC
Produced: datasets 1
DAT

Cite it

Mysteries Department. (2026). Session 3: The archaic names benchmarked against real Sumerian names [Research session SES0DFVRK8PYQFX]. https://mysteriesdept.com/mysteries/proto-cuneiform/sessions/SES0DFVRK8PYQFX/session-3-the-archaic-names-benchmarked-against-real
Private beta Downloads are for beta members

We're in a private beta while we get things right. Everything here is free to read; downloading datasets, sources and code is open to members for now.

Apply to the betaSign in
It takes a couple of minutes. We'll reply by email.