Veḷi

Evidence page · a language with no relatives

The working language of the Persian empire, written for 2,300 years and still partly undeciphered.

Attested for roughly 2,300 years. Written in three scripts, two of which were undeciphered until recently and one of which still is. The working administrative language of the Persian empire — and related to nothing anyone has been able to demonstrate.

The basics

FamilyIsolate. No demonstrated relatives anywhere
WhereSouthwestern Iran — Susa in the Khuzestan lowlands, Anshan (Tal-e Malyan) in the Fars highlands. Kings styled themselves of both
Attestedc. 2600 BCE to c. 360 BCE — a span longer than Latin's. Possibly surviving later as the "Khuzi" that Ibn al-Nadīm mentions in the 10th century CE
TypeAgglutinative, suffixing, verb-final
NotNot Semitic. Not Indo-European. Not Sumerian

Three scripts, and only one of them was ever easy

ScriptDatesStatus
Proto-Elamitec. 3100–2900 BCE STILL UNDECIPHERED. Roughly 1,600 tablets, mostly from Susa. The numerical signs are understood; the rest are not. The second-oldest writing system on earth after proto-cuneiform — and we cannot even be certain it writes Elamite.
Linear Elamitec. 2300–1880 BCE About 40 inscriptions. Undeciphered for a century; Desset and colleagues published a claimed decipherment in 2022, which is significant, recent and not yet settled.
Elamite cuneiformc. 2200–360 BCE Read. Adapted from Akkadian cuneiform. This is the corpus everything we know rests on.
Why this matters for how we read the ancient world

Elamite is the case that proves the pattern. A major civilisation, adjacent to Mesopotamia, trading with the Indus, administering the largest empire on earth — and for most of the last two centuries it was effectively voiceless, because two of its three scripts could not be read.

Not marginal. Just illegible. Which is precisely the position the Indus and Linear A are in now. The Linear Elamite decipherment claim of 2022 is a demonstration that "undeciphered" is a statement about us, not about them.

The Achaemenid archive — and a correction

What Elamite actually administered, and what it did not

The Persepolis Fortification Archive holds roughly 30,000 tablets, dating c. 509–493 BCE, overwhelmingly in Elamite: ration allocations, livestock, work gangs, and travel authorisations for people moving across the empire. Elamite is also one of the three versions of Darius's Behistun inscription, alongside Old Persian and Akkadian.

What the archive does establish is movement. It records rations issued to travellers authorised to cross between satrapies, so the network connecting Ionia to Hindush is documented in ledgers, not inferred. And Darius's Susa foundation charter (DSf) itemises where materials and workers came from — ivory from Ethiopia, Sind and Arachosia; yaka timber from Gandhāra and Carmania — Indian material named in a Persian royal inscription, in Elamite among other versions.

Elamo-Dravidian

The hypothesis, its evidence, and why most Dravidianists reject it

David McAlpin proposed in the 1970s that Elamite and Dravidian descend from a common ancestor, "Proto-Elamo-Dravidian."

What is genuinely similar

  • Agglutinative, suffixing, verb-final — both
  • Noun classification by animacy — Elamite personal vs non-personal; Dravidian rational vs irrational
  • Pronoun shapes McAlpin compared: Elamite u "I", nu "you" against Dravidian *yān/*nān, *nīn
  • Some proposed lexical and verbal-morphology correspondences
  • Geography: Elam sits between Mesopotamia and the Indus, and Brahui survives in Balochistan in between

Why it is not accepted

  • No regular sound correspondences of the kind that established Indo-European. Bhadriraju Krishnamurti, the leading Dravidianist, rejected it on exactly this ground
  • We do not know Elamite's actual phonology. Its cuneiform is borrowed and does not reliably distinguish voiced from voiceless stops — so the phonological comparison is being run on a script, not a sound system
  • Elamite shows no retroflexion, which is Dravidian's most distinctive feature
  • Typological traits — agglutination, SOV, suffixing — are shared by Turkish, Japanese, Dravidian and Sumerian alike. Typology is not genealogy
  • First and second person pronouns in n- and m- recur across unrelated families worldwide
And the genetics cut against the popular version

The hypothesis was revived when South Asian populations were found to carry Iranian-farmer-related ancestry — read as farmers carrying Dravidian eastward.

Narasimhan et al. 2019 undercuts that specific story. The Iranian-related component in Indus Periphery individuals appears to have separated from Ganj Dareh-type Zagros populations before farming began. If so, it did not arrive with agriculture, and the farming-dispersal version of Elamo-Dravidian loses its mechanism.

Honest verdict: not proven, not disproven, and currently short of the evidence its own method requires. It deserves to stay open — the corpus is small, Elamite's phonology is obscured by a borrowed script, and Linear Elamite has only just begun to be read. It does not deserve to be asserted.

And there was no Elamite–Tamil contact under the Persians

The Tamil country was outside the Achaemenid empire entirely — Ashoka still names the Cholas and Pandyas as foreign borderers two centuries later. Any Elamo-Dravidian relationship would belong to deep prehistory, not to the 6th–4th centuries BCE. Conflating the two is a category error worth avoiding.

What Elam actually was

Not a province of Mesopotamia

c. 3100 BCEProto-Elamite writing at Susa — independent of Mesopotamian cuneiform, and among the earliest writing anywhere
3rd mill.Trade across the Iranian plateau — Jiroft, Shahdad, Tepe Yahya, chlorite vessels moving to both Mesopotamia and the Indus. Some scholars group these as a "Trans-Elamite" world
c. 1250 BCEChogha Zanbil — one of the best-preserved ziggurats in existence, built by Untash-Napirisha
c. 1158 BCEShutruk-Nahhunte sacks Babylon and carries the Code of Hammurabi stele back to Susa — which is why it was excavated in Iran, not Iraq
646 BCEAshurbanipal destroys Susa and boasts of it in Assyrian annals
c. 550–330Elamite becomes the bookkeeping language of the Achaemenid empire

Its pantheon was its own — Napirisha, Inshushinak lord of Susa, Humban, and prominent goddesses including Pinikir and Kiririsha. Elamite succession sometimes ran brother to brother and gave unusual weight to the sister's son, which has been read as a matrilineal element — argued, not settled.

The Code of Hammurabi was found in Iran because Elam took it as loot.

Did the Greeks get their content through Persia, or from further east?

Undeterminable, and worth saying so

The Achaemenid empire connected Ionia to Gandhāra for two centuries with a shared administration. That establishes a channel. It does not establish a direction.

What can be said: the empire itself was an administrative rather than a literary culture — the Old Persian corpus is small and almost entirely royal inscription. It moved people, goods, scripts and officials at scale, and produced little philosophy of its own. The pipe is well documented. What flowed through it is not.

Anyone claiming to know which specific ideas travelled which way, on current evidence, is going beyond it — including in the direction this platform would prefer.

Linear Elamite, read in 2022

A decipherment, and a negative result that matters here

In 2022 François Desset and colleagues published a proposed decipherment of Linear Elamite — the second-millennium-BCE script distinct from the later Elamite cuneiform. The work rests on a group of inscribed silver vessels and on bilingual anchoring through royal names.

It is contested. The corpus is small, several of the key objects are unprovenanced, and the sign values are not universally accepted. Treat it as a serious proposal under review, not a settled reading.

And the result relevant to this platform is negative. The texts read so far are royal and dedicatory. No Meluhha. No reference to the Indus region, its people, its goods or its trade.

A script from the civilisation that sat directly between Mesopotamia and the Indus has been partly read, and it does not mention the Indus.

That is worth stating carefully in both directions. It is not evidence that contact did not happen — the Mesopotamian record documents Meluhha extensively, and Elam sat on the route. What it shows is that the surviving Linear Elamite corpus is not the kind of text that records trade: it is royal inscription, and royal inscriptions name kings and gods.

The absence is a fact about genre and corpus size, not about history. The Mesopotamian Meluhha references survive because Mesopotamia produced administrative archives by the tens of thousands. Elam's surviving Linear corpus is a few dozen objects.

Elamite and Dravidian — what the case actually consists of

FeatureElamiteDravidianWeight
Word orderSOVSOVWEAK — most of Eurasia is SOV, including Indo-Aryan
AgglutinationYes, suffixingYes, suffixingWEAK — a common type, not an inheritance
No grammatical genderCorrectCorrect in the nominal systemMODERATE — a real shared absence
Animate / inanimate classYesYesMODERATE
Proposed lexical correspondencesMcAlpin assembled a set of proposed cognates with stated sound correspondencesTHE CRUX
Regular sound lawsProposed by McAlpin; not accepted as demonstrated by the wider fieldUNRESOLVED
Why it stays unresolved rather than refuted

David McAlpin proposed Elamo-Dravidian in the 1970s and 80s with a full apparatus — proposed cognate sets, proposed regular correspondences, and a reconstruction. That is the correct form for such a proposal, and it distinguishes his work from resemblance-hunting.

It has not been accepted. The objections are real: the Elamite corpus is limited and its phonology partly inferred from a cuneiform script not designed for it; the proposed cognate sets are small; and the time depth required strains what the comparative method can demonstrate.

But "not accepted" here means something specific. Testing Elamo-Dravidian requires competence in Elamite and in Dravidian comparative linguistics. The number of scholars holding both is very small, and the proposal has been evaluated by few of them.

A hypothesis that requires two rare specialisms to assess will be under-assessed, and under-assessment looks identical to rejection in a citation count.

The defensible statement: Elamo-Dravidian is a serious proposal with a stated method that has not been demonstrated and has not been refuted. Anyone who tells you it is established, or that it is discredited, is going beyond the state of the evidence.

And Sumerian is a separate question with a worse answer

Sumerian is an isolate. Proposed links to Dravidian, to Elamite, to Indo-European and to a dozen other families have all been made and none is accepted.

The typological similarities are real — Sumerian is agglutinating and SOV, like Elamite and Dravidian, and unlike Akkadian which surrounded it. But typology is not descent, and the same features occur across unrelated families worldwide.

What is documented and not disputed is contact: Meluhha goods, Meluhha personnel, a Meluhha village and an interpreter of Meluhhan, all in Mesopotamian records. Contact is attested; relatedness is not. These are different claims and conflating them damages the first.