← All discoveries
Proto-Elamite computational epigraphy · 2026-04-15

A Proto-Elamite sign M340 substitutes for a 2-sign prefix in a specific household-archive template

Within the 24-tablet `|M153+M342|` household archive at Susa, the sign M340 paradigmatically substitutes for 2-sign prefix spellings `M387 M387` / `M305 M387` / `M122 M387` (read under Kelley-Born-Monroe-Sarkar 2022's Desset-derived values as `na-na` / `M305-na` / `h2-na`) in the line-2 slot of a canonical `[prefix] M054 M388` grain-labor template. On three tablets at matched transaction count 6(N01)/3(N01) — P008022, P008265, P009237 — the substitution preserves the numeric profile exactly, demonstrating semantic equivalence in the accounting slot. This entry originally claimed a phonetic reading M340 = logographic `na-na` under the Desset chain; an adversarial re-review across 30 additional research iterations (iter 23-50) downgraded that claim. The surviving defensible claim is slot-equivalence at the administrative-category level, not phonetic equivalence. Honest score: 5.

Description

CORRECTED ENTRY (2026-04-15, post-iter-50 review). This entry originally framed the M340 substitution as a first empirical phonetic reading of M340 = logographic `na-na`, derived by propagating Kelley-Born-Monroe-Sarkar 2022's Desset-derived `M387 = na` value through the paradigmatic substitution. That strong phonetic framing did NOT survive adversarial review. The defensible claim is weaker: M340 paradigmatically substitutes for 2-sign `[X] na` prefixes at matched transaction counts within one specific household archive, demonstrating SEMANTIC equivalence in an accounting-category slot — not phonetic equivalence at the spelling level. THE CORE STRUCTURAL FACT. Within the `|M153+M342|` household archive (24 tablets at Susa — see the sibling discovery entry 'proto-elamite-three-household-archives-susa' for the broader archive-enumeration context), a 10-tablet sub-cluster follows the canonical template `M157 [prefix] M054 M388 , count(N01)` on line 2, `M288 , count(N01) [± grain capacity]` on line 3, with a subscript line containing `|M153+M342|` or a `|M153+X|` variant. Across these 10 tablets the line-2 prefix slot is filled in four distinct ways: (1) `M387 M387` (two tablets: P008022, P009180); (2) `M305 M387` (one tablet: P008265); (3) `M122 M387` (one tablet: P008033); (4) single-sign `M340` (six tablets: P009190, P009211, P009220, P009237, P009286, P009309). At the matched transaction count of 6(N01) on line 2 + 3(N01) on line 3, three tablets span three of these four prefix writings: P008022 (`M387 M387`, 6/3), P008265 (`M305 M387`, 6/3), P009237 (`M340`, 6/3). The numeric counts are EXACTLY identical across the three prefix writings, demonstrating that the three tablets are recording the same transaction category with three different prefix forms. THE ORIGINAL INFERENCE (which does NOT survive adversarial review). Under the chain M387 = `na` (Kelley-Born-Monroe-Sarkar 2022, propagating Desset 2022's Linear Elamite correspondence), `M387 M387` spells the attested Kelley-wordlist entry `na-na` (3 standalone attestations on P008713, P009127, P008068, plus extensions `na-na-sha` = M387 M387 M377 on P008479 and `na-na-hu-e` = M387 M387 M262 M096 on P234794). The original entry argued that because M340 substitutes paradigmatically for `M387 M387` in the same template slot with the same numeric counts, M340 must be a logographic writing of the same `na-na` morpheme. ADVERSARIAL REVIEW (iter 37 of the 30-iter dig-deep run). An independent agent analyzed the claim by proposing and evaluating strongest alternative readings for M340: (1) M340 as a class determinative (Sumerian-LÚ-like marker for a specific laborer/commodity category); (2) M340 as a specific commodity or worker-subtype logogram; (3) M340 as a ligature of two signs not necessarily M387 M387; (4) M340 as a proper name or toponym; (5) M340 as a temporal/batch marker. The adversarial analysis concluded: paradigmatic substitution in the same slot under the same seals at matched counts forces SEMANTIC equivalence in that slot — it does NOT force PHONETIC equivalence. Multiple logographic systems have cases where a logogram and a syllabic spelling co-occur for the same administrative referent without being phonetic equivalents (Sumerian LÚ vs. specific named professions, GURUŠ vs. `erin₂`, etc.). The most parsimonious alternative, M340 as a class determinative marking 'this is the standard-category laborer entry' for the M153 household's grain-labor genre, is at least as well-supported as the phonetic 'na-na' reading. The unmarked-default frequency (M340 on 6/10 tablets, marked variants on 4/10) is exactly what you'd expect from a determinative, not from a phonetic logogram. THE SURVIVING DEFENSIBLE CLAIM. M340 paradigmatically substitutes for 2-sign `[X] na` prefix writings within the M153+M342 household archive's grain-labor template at matched transaction counts, demonstrating semantic equivalence in the slot. Whether that equivalence is phonetic (logogram for `na-na`), classificatory (determinative for a laborer category), or taxonomic (sibling subtype logogram for a specific worker class) is UNDERDETERMINED by the count-matching evidence alone. Distinguishing these three possibilities would require either (a) M340 appearing in a non-labor template where its reading is independently constrained, or (b) a later-period PE / Elamite descendant sign preserving a phonetic reflex. Neither is currently in the evidence set. PRIOR ART. The M153+M342 compound is catalogued in Dahl's working sign list. M340 is catalogued in Dahl's working sign list with graphic variants M340~b. The compound `M387~ca+M340+M387~ca` (M340 encircled by two M387 variants) is catalogued as a known complex grapheme in Dahl's working sign list. The compound `M157+M340` is also catalogued. Dahl 2005 did NOT propose a phonetic reading for M340; it was an un-read sign in Kelley 2022's syllabary comparison table (neither Dahl 2019 nor Desset 2022 assigns a value). The iter-20/21 claim that M054 functions as a 5-sign ligature for the sequence `M388 M390 M128 M007 M096` (from P008147's long form) is WITHDRAWN — the supposed adjacent-form expansion appears on exactly one tablet corpus-wide (hapax), and that tablet has a different subscript (`|M304+M342|` rather than `|M153+M342|`), meaning P008147 is in a different sub-genre and the alignment at count 6 was cross-sub-genre coincidence rather than a true short-form/long-form identification. Englund 2004's M054 = YOKE/seeder and M388 = male laborer readings stand and underpin the grain-labor interpretation of the template. RELATIONSHIP TO SIBLING ENTRY. The broader archive-enumeration framing of this work is captured in the sibling discovery entry 'proto-elamite-three-household-archives-susa' (same date), which presents the three-archive comparative framework (M327+M342 / M153+M342 / M305+M342) as the strongest quantitative contribution. The M340 substitution is a specific structural finding INSIDE the M153+M342 archive, subordinate to the broader three-archive framing. Both entries share the same scripts and dataset; readers interested in the archive-comparative context should start with the sibling entry.

Purpose

Precise

USE CASE. A Proto-Elamite specialist investigating the M153+M342 household archive (24 tablets at Susa — see sibling entry) now has a characterized sub-cluster of 10 tablets sharing the canonical `[prefix] M054 M388 / M288 / subscript` grain-labor template with a 4-way prefix variant (`M387 M387`, `M305 M387`, `M122 M387`, single-sign M340) and a matched-count 6/3 triple (P008022, P008265, P009237) demonstrating the variants are administratively equivalent. The specialist can now treat these 10 tablets as a uniform dataset for any sub-corpus analysis. The specific downstream utility: (1) future sign-value work on M340 has a concrete slot-equivalence constraint to satisfy — any proposed reading must explain why M340 is administratively interchangeable with two-sign `[X] na` prefixes at matched transaction counts, or must explicitly argue for a semantic distinction not visible in the count data; (2) the same slot-mining method (`paradigmatic_readings.py` + `register_align.py`) can be applied to the other 14 unread content signs in the iter-10 `@top: 1(N34)` classifier cluster to generate further slot-equivalence constraints; (3) the PES0329 (M340-only) and PES0334 (mixed `h2 na` / `na na` / M340) scribal archive identification gives the specialist an administrator-level view of the slot-variant distribution — PES0334 used all three prefix forms across three tablets by the same hand, the strongest within-scribe evidence that the forms are administratively interchangeable. This entry is subordinate to the sibling three-archive entry: readers should cite the sibling for the broader quantitative framework and cite this entry only for the specific M340 slot-equivalence observation within the M153 archive.

For a general reader

This entry is a CORRECTED version of an earlier claim I shipped on the same date. The earlier claim was that the unread Proto-Elamite sign M340 could be read as a logographic writing of the two-sign phonetic sequence `na-na` — because M340 appears in the same template slot where `na na` (the sign M387 written twice) appears on other tablets, and the two tablets happen to record the same transaction count. That claim turned out to be overreach. Under a careful review by a second independent agent I ran, the most honest reading is weaker: M340 is a single sign that substitutes for a two-sign prefix in one specific accounting template, and we can prove the substitution preserves the transaction amount, but we can't prove whether the substitution is PHONETIC (the sign actually spells `na-na`) or CLASSIFICATORY (the sign means 'this is the standard laborer category' and the two-sign prefix means the same thing with a more specific qualifier). The evidence supports the weaker claim but not the stronger one. I'm correcting the entry to match what the evidence actually shows. The core structural fact is real and reproducible: on three specific Proto-Elamite tablets (P008022, P008265, P009237) the accounting template `[prefix] M054 M388 , 6 units / M288 , 3 units` has three different prefix writings — `M387 M387`, `M305 M387`, and single-sign `M340` — but all three record the exact same 6/3 transaction. The sign M340 is administratively interchangeable with those two-sign prefixes in this specific template within this specific household archive. That's a real empirical finding. What I cannot honestly claim is that M340 was PRONOUNCED `na-na`, because the evidence only shows administrative-category equivalence, not phonetic equivalence. A more detailed version of this correction lives in the companion entry 'proto-elamite-three-household-archives-susa' which is the main quantitative finding of this research iteration. Read that one first.

Novelty

The M340 paradigmatic slot-equivalence within the `|M153+M342|` archive's grain-labor template is an empirical refinement of Englund 2004's grain-labor framework. The core slot substitution is not visibly preempted by any accessible prior source — M340 has no proposed reading in Dahl 2005, Dahl 2019 (TCL 32) per the Kelley 2022 syllabary comparison table, Desset 2022, or Kelley-Born-Monroe-Sarkar 2022. The slot-equivalence observation itself (that M340 substitutes for 2-sign `[X] na` prefixes at matched counts) is novel relative to the accessible literature. What is NOT novel: the underlying semantic framework (grain-labor template, M054 = YOKE, M388 = male laborer) is all Englund 2004. The M153+M342 compound existence is catalogued in Dahl's working sign list. Afshari & Yousefi Zoshk 2021 is an unverified partial prior art risk on the M153+M342 compound interpretation. The original framing of this entry as a 'first empirical phonetic reading of M340' was an overclaim; the slot-equivalence observation is the defensible contribution. Honest score: 5. A specialist reading this would say 'the slot substitution is a useful structural constraint on future M340 readings; the phonetic reading inference was overreach.' The entry is corrected to state the weaker claim.

How it upholds the rules

1. Not already discovered
The specific M340 slot-equivalence within the M153+M342 archive's grain-labor template is not in any accessible prior source. M340 has no proposed reading in Dahl 2005, Dahl's working sign list, Englund 2004, Desset 2022, or Kelley-Born-Monroe-Sarkar 2022. The sibling entry 'proto-elamite-three-household-archives-susa' carries the broader quantitative contribution of the research iteration; this entry presents the specific structural observation within one archive.
2. Not computer science
Ancient epigraphy on clay tablets. Python is used only as a verifier.
3. Not speculative
The slot substitution, the 10-tablet cluster membership, the 4-way prefix variant, and the 6/3 matched-count triple are all deterministic grep results against the CDLI August 2022 ATF dump. The interpretation of the substitution as semantic equivalence (rather than phonetic equivalence) is the ADVERSARIAL-REVIEW-SURVIVING claim; the stronger phonetic interpretation has been explicitly withdrawn. The iter-20/21 'M054 is a 5-sign ligature' sub-claim has also been withdrawn — the alignment with P008147 was cross-sub-genre coincidental rather than a genuine short-form/long-form identification.

Verification

Reproduction: (1) Clone cdli-gh/data and fetch git-lfs blobs. (2) Extract Proto-Elamite subset via `parse_atf.py`. (3) Grep for tablets matching template `M157 + [M387 M387 | M305 M387 | M122 M387 | M340] + M054 + M388 + M288 + (|M153+M342| or |M153+X| subscript)`. Verify the 10 tablets: P008022, P008033, P008195, P008265, P009190, P009211, P009220, P009237, P009286, P009309 (plus P009180 and P368499 in broader variants). (4) Confirm the 6/3 count-matched triple: P008022 (6 N01 / 3 N01), P008265 (6 N01 / 3 N01), P009237 (6 N01 / 3 N01) — the three tablets span three of the four prefix writings with identical transaction counts. (5) Inspect the PES seal attributions: PES0329 (M340 only, 5-6 tablets), PES0334 (mixed `h2 na` on P008033, `na na` on P009180, M340 on P009286 — three prefix variants within a single administrator). (6) Confirm M340's absence from prior readings: check Kelley 2022's `syllabary_comparison.md` (M340 not in intersection, Dahl-only, or Desset-only lists), `skipgrams_desset.md` (M340 appears only as `M340-ru` with the neighbor M372 = `ru`), `wordlist_desset.md` (no M340 entry). (7) Confirm the adversarial-review alternatives: the slot substitution preserves transaction identity but does not uniquely determine phonetic equivalence vs classificatory equivalence. Multiple interpretations survive; the M340-as-class-determinative reading is at least as well-supported as the M340-as-logographic-`na-na` reading.

Next steps

  • Read the sibling discovery entry 'proto-elamite-three-household-archives-susa' for the broader three-archive quantitative framework that contextualizes this specific slot finding.
  • Distinguishing the three candidate readings for M340 (phonetic `na-na` logogram vs. class determinative vs. subtype logogram) requires finding M340 in a non-labor template where its reading is independently constrained, OR finding a later Elamite descendant sign preserving a phonetic reflex. Neither is currently in the evidence set.
  • Run the same slot-equivalence method on other unread signs in the M153+M342 archive to check whether the pattern generalizes (other signs have matched-count paradigmatic substitutions with Kelley-read neighbors).
  • Check whether PES0334's within-scribe mixed-register usage (three prefix forms across three tablets by one hand) is paralleled by other Proto-Elamite scribes. If yes, the slot substitution is a general scribal convention; if no, it's administrator-specific.

Artifacts

Sources