The Cups That Vouch for Themselves
Linear Elamite is now read at 97 per cent of its sign occurrences, without a bilingual. More than half of those signs sit on silver vessels that nobody can say came from anywhere.
The sequence that broke it is four signs long.
François Desset had been working through inscriptions engraved on silver beakers held in a private collection in London, and one string of signs kept turning up where a royal name ought to be. The first sign was already known to read ši. The last two were identical to each other. If they were both ha, the whole thing read ši-l-ha-ha — Shilhaha, a ruler of the Sukkalmah dynasty, attested in cuneiform, reigning somewhere around 1950 BCE. Two more names came out of the same texts: Eparti II, his predecessor, and Napirisha, the “Great God.” Between them, those three names fixed nine sign values — e, ha, l, pa, pi₂, ri, ša, ti, u — and from those nine the published decipherment builds outward to 73 values in thirteen documented steps.
That is Linear Elamite: a script used in southern Iran roughly between 2300 and 1880 BCE, first excavated at Susa in 1903, and unread for the following century despite serious attempts by Ferdinand Bork, Carl Frank, Walther Hinz and Piero Meriggi. In 2022, Desset and four colleagues published a claimed near-complete decipherment in Zeitschrift für Assyriologie. As of the project database now maintained at Liège, 1,978 of the corpus’s 2,033 sign occurrences can be read. Five signs remain unread, plus a tail of hapaxes.
There is no Rosetta Stone here
This is the part I did not expect, and it is what makes the case interesting.
Champollion had one text in three scripts. Linear Elamite has nothing of the kind. Asked about it directly, Desset is blunt: “With linear Elamite we do not have bilingual texts stricto sensu. We only have ‘partial bilingualism’.” The monuments from Susa that carry both Linear Elamite and Akkadian cuneiform — the table au lion and its relatives, from the reign of Puzur-Inshushinak around 2150 BCE — share proper names and titles between their two halves, but the two halves are not saying the same thing. They were enough for Bork and Frank to pin down nine signs a century ago. They were not enough to go further.
What Desset had instead was biscriptualism: the same Elamite formulae appearing in cuneiform on one object and in Linear Elamite on another. Not a parallel text, a parallel genre. The foothold comes from dedications being formulaic, the name sitting in a predictable slot, and the language already being known from thousands of later cuneiform tablets.
The reasons to think the readings are right are mostly internal. The values, once fixed, generate grammatical Elamite: verbal forms, precative clauses, titles and epithets that match what is independently attested in Middle and Achaemenid Elamite. They keep working across texts that were not used to derive them. And there is a loud piece of negative evidence: four years on, nobody has published a competing sign list. Manfred Krebernik called the argument “mostly convincing,” Matthew Stolper “clear, coherent and plausible,” Piotr Steinkeller “a major achievement.” Specialists who would enjoy dismantling it have not.
Where the evidence physically sits
The corpus divides three ways. The lowland group from Susa — excavated by French missions from 1903, sitting in the Louvre and the National Museum of Iran — is 18 texts and 533 sign occurrences. The eastern group is eight scraps of between two and eleven signs each. The highland group is 24 inscriptions, every one of them on a silver vessel, and it carries 1,133 sign occurrences: more than half of all the Linear Elamite anybody has ever seen.
The eight beakers that made the script solvable are in that group. Seven came from the Mahboubian collection in London, one from the Schøyen collection, and Desset published them in 2018. None has an archaeological context. Nobody recorded where they came out of the ground, in what kind of building or grave, alongside what else. Desset’s own reconstruction is that they probably come from a Shimashki or Sukkalmah royal cemetery in the Kam-Firuz area of Fars — offered explicitly as a suspicion. The family’s account, as reported in a review of the work, is that the grandfather dug them up in commercial excavations in the 1920s or 1930s and they sat in a safe deposit box for decades.
To their credit, the authors of the decipherment do not paper over this. They write that they “regret that the exact provenance of these artifacts is unknown,” and that whether “the understanding of an unprovenanced text in an undeciphered writing system can validate the genuineness of its physical carrier represents a relevant methodological issue for an epistemological reflection.” That is an academic way of naming a circle, and it is a real one.
The material evidence they lean on is a preliminary metallographic and chemical study of silver samples from the collection, published in the journal Iran in 2018. The publisher would not let me at it, so I am taking its findings as quoted in the decipherment paper: patinas of the kind that form gradually in burial soil, silver alloyed with a little copper, no tin or nickel or zinc, manufacturing traces of the expected kind — all of which “points to ancient artefacts and not to ingenious modern forgeries.”
I believe that about the metal. The thing metallurgy cannot establish is that the engraving is as old as the vessel, and inscriptions added to genuine ancient objects are the standard failure mode in this corner of the market, not an exotic one. What actually reassures me is a different argument: the beakers were in private hands and, in part, in a dealer’s own published catalogue well before anyone could read a line of Linear Elamite. A forger would have had to invent internally consistent text in an unreadable script, put the right kings in the right formulaic positions, and have it parse as grammatical Elamite twenty years later. I do not think anyone did that.
But notice the shape of that. It is an argument from how hard the forgery would have been, and it is the strongest argument available precisely because the evidence that would settle it — the context — was destroyed before anyone wrote it down.
The beakers made the script readable, and the readability is now the best argument that the beakers are real.
The fight is not where the coverage put it
The most substantial published criticism is Jacob Dahl’s, in Akkadica in 2023, and it is aimed at a claim most readers probably merged with the decipherment: that Linear Elamite descends from proto-Elamite, the older and still-unread script from the same region.
Dahl’s objections are specific. The graphical comparisons in the 2022 paper cite proto-Elamite sign numbers from superseded versions of his own sign list — one of them, M483, was retired in 2005 — and seven of the compared signs do not exist in the current proto-Elamite corpus at all. Proto-Elamite has well over 1,500 glyphs, so matching a hundred-glyph script against it will throw up close resemblances by chance; Desset concedes this in principle. Five centuries separate the two corpora with nothing in between, and writing systems do not sit still for five centuries — cuneiform is transformed over the same span. Worse for the descent story, the Linear Elamite signs that resemble proto-Elamite ones cluster among the highest-frequency proto-Elamite signs, which is what you would expect if a scribe copied the signs he happened to find on some dug-up tablets. Dahl’s alternative is that late third-millennium scribes invented an archaising script as an act of cultural differentiation from Mesopotamia, reaching for old tablets the way Neo-Assyrian scribes later produced deliberately venerable-looking inscriptions.
None of that touches the sign values. A script can be phonetically solved and genealogically misdescribed at the same time, and I think that is roughly where things stand.
There is a second, quieter disagreement about what kind of system it is. The team says purely phonographic — an alpha-syllabary of vowels, consonants and CV signs, with no logograms — which is what licenses the headline that this is possibly the oldest purely phonographic writing ever found. Michael Mäder’s analysis puts it nearer 70% phonographic, with the rest doing logographic work. The older reference literature describes it as primarily syllabic with a few logograms. The superlative rides entirely on the strongest version of the claim.
And then there is the strangest feature of the whole business. When Smithsonian Magazine reported scholarly disagreement over the work, Gianni Marchesi — one of the five co-authors — replied that as far as he and his colleagues were aware it “has only received positive reactions,” and that the journalist had manufactured the controversy. Dahl, in print, says the sign comparisons would make his predecessors turn in their graves. Both of those can be true, because they are about different claims, and the fact that the participants cannot agree on whether there is a controversy tells you how cleanly the two questions came apart.
What would actually settle it
One inscription. Excavated in context, photographed before cleaning, and read with the published sign list before the reading is published. That is the experiment, it is cheap, and it is the only thing that would convert a coherent internal system into a tested one. The eastern finds that did come out of regular excavation are two to eleven signs long, too short to test anything, and some unprovenanced brick tablets from Jiroft are under suspicion as forgeries.
In the meantime the corpus is creeping upward: 40 texts in 2021, 45 now; the unread residue has gone from 3.7% to 2.7%. What I keep turning over is the inversion in it. The philology here is tighter than the archaeology. The field’s confidence is highest exactly where its physical chain of custody is weakest, and that is not a scandal, it is just what happens when more than half of a writing system’s surviving record reaches scholarship through the antiquities market rather than out of a trench.
One last thing I liked. Asked whether computers, statistics or machine learning had contributed, Desset’s answer was “All hogwash!” — what mattered, he said, was knowledge of the language and the period, some luck, and perseverance. Given how much of the current excitement about undeciphered scripts assumes the next one will fall to a model, it is worth noticing that this one fell to a man recognising a doubled sign at the end of a king’s name.
Sources
- The Decipherment of Linear Elamite Writing — Desset, Tabibzadeh, Kervran, Basello & Marchesi, Zeitschrift für Assyriologie 112 (2022), 11–60
- Proto-Elamite and Linear Elamite, a Misunderstood Relationship? — Jacob L. Dahl, Akkadica 144/2 (2023), 107–126
- Hatamti–Linear Elamite Database, "About" — University of Liège
- Breaking the Code: Ancient Iran's Linear Elamite Script Deciphered — François Desset, ANE Today
- On the Decipherment of Linear Elamite Writing: François Desset interviewed by Robert M. Kerr — The Postil Magazine
- Have Scholars Finally Deciphered Linear Elamite, a Mysterious Ancient Script? — Smithsonian Magazine
- Correcting Misconceptions from a Smithsonian Magazine Article on Linear Elamite Research — kavehfarrokh.com
- Beyond Mesopotamia: Linear Elamite Deciphered — Tom Stevenson, London Review of Books
- Elam iv. Linear Elamite — Encyclopaedia Iranica
- Linear Elamite (corpus groups and sign counts) — Wikipedia