Evidence page · names, and what happens to them

What the mouth does to a name.

A name entering a foreign language is reshaped by that language's rules — and the reshaping is regular, predictable, and traceable. Two of the most loaded words in South Asian identity are the results of foreign mouths failing at a local sound.

Sindhu

The flagship case. One river, and every name for the region, the religion and the country descends from it — none of them coined by the people they name.

river → religion

síndhu
Sanskrit · river, the Indus
Indo-Iranian s → h (regular)
Hindu
Old Persian / Avestan
Greek psilosis: initial h dropped
Indós
Greek
Latin territorial -ia
India
Latin → English
Hindu
Persian
Arabic definite article
al-Hind
Arabic
people of → faith of
Hindū · Hindustān · Hindī
Persian → English

“Hindu” and “India” are the same word. One travelled through Persian, one through Greek. The s → h shift is a regular Indo-Iranian sound law, not a mistake: the same law gives Avestan haoma for soma, hapta for sapta, ahura for asura, and Haraxvaitī for Sarasvatī — the Helmand. The name of a religion practised by hundreds of millions is a Persian pronunciation of a river. Sindh and Sindhi keep the original s.

Where síndhu itself came from is a separate question

The chain above is secure from Sanskrit onward — the s→h law, psilosis, and the Persian and Greek branches are not in doubt. The origin of the Sanskrit word is not part of that security.

síndhu is conventionally derived from √syand, "to flow." Dravidian alternatives have been proposed, connecting it to the cintu- / centu- "flow, pour" family. These are minority positions — but the handbooks are not neutral ground. A river in the northwest is precisely where a substrate name would survive, and the Indo-Aryan derivation is not tight enough to foreclose one.

índu — the pressed soma drop, 245 of its 327 Rigvedic occurrences in Book 9 — is marked uncertain by Mayrhofer and has no clear Indo-European cognates. A substrate origin is live. The √ind "to drop" derivation is a traditional gloss, not a demonstrated etymology.

A Sanskrit word with a shaky internal etymology and no Iranian cognate is a candidate loan. Treating the Sanskrit derivation as default and the Dravidian one as speculative reproduces the exact asymmetry this page documents.

Tamiḻ

The second flagship. The sound that defines the language's own name is the one Indo-Aryan cannot produce.

autonym → mispronunciation → language family

tamiḻ
Tamil · ḻ = retroflex approximant
ḻ → ḷ · initial t voiced to d
damiḷa
Prakrit / Pali
ḷ → ḍ
dameḍa · damela
Sinhala Prakrit inscriptions
hypercorrect intrusive r
drāviḍa
Sanskrit
borrowed as a family name
Dravidian
Caldwell, 1856

Tamil distinguishes five lateral and rhotic phonemes — ல ள ழ ர ற — where Sanskrit has two. is not in the Sanskrit inventory at all. mleccha means “speaks indistinctly.” And drāviḍa is Indo-Aryan failing to say ḻ. The accusation and the mispronunciation are the same event, running in both directions. Caldwell took the term from Kumārila Bhaṭṭa's Andhra-Drāviḍa-bhāṣā, c. 650 CE, so that “Tamil” could be left free to mean Tamil.

Meluhha

a name nobody could spell

Meluḫḫa
Sumerian → Akkadian
?
ḫ = velar fricative, the slot for foreign gutturals
mleccha
Sanskrit · “one who speaks badly”
?
Parpola’s proposal
mel-akam
Dravidian · “high country”

ḫḫ → cch is not a sound law. These are best read as independent renderings of one unspellable foreign name, in two writing systems that each lacked the sound. Listed here as the counter-example: not every resemblance is a derivation, and this page is worth nothing if it does not say so.

Names in Greek mouths

The Greek and Prakrit renderings of Indian names are a controlled experiment: we know both endpoints, so the rules are visible.

OriginalRendered asWhat changed
Aléxandros (Greek)Alikasudara Ashokan Prakrit, RE 13ks cluster split; consonant clusters simplified; Prakrit vowel insertion
Aléxandrosal-Iskandar → SikandarArabic reanalysed the first syllable as its own article al-
CandraguptaSandrokottosc → s; -gupta → -kottos
BindusāraAmitrochatesrenders a title, Amitraghāta, “slayer of foes” — not the name
AśokaPiodasses Kandahar Greek edictrenders his epithet Piyadassi, not his name
PāṭaliputraPalibothraretroflex ṭ → l; -putra → -bothra
BharukacchaBarygazaaspirate lost; cch → z
MuciṟiMuzirisTamil ṟ → z; Greek nominative -is added
KāverippūmpaṭṭinamKhaberistruncated to the river element
PāṇḍyaPandionassimilated to an existing Greek name
TāmraparṇīTaprobanecluster simplification, r-loss
Īḻam / SīhaḷaSarandīb · Sīlān → Ceilão → CeylonCONTESTED — see below
Iāones (Ionians)Yauna → Yavana → YonaPersian, then Sanskrit, then Prakrit and Tamil Yavanar

Twice in that table the Greeks recorded a title instead of a name — Bindusāra as Amitrochates, Ashoka as Piodasses. They wrote down what they were told, and what they were told was how the man was styled.

Serendib and Ceylon · unresolved

The handbook derivation takes Arabic Sarandīb from Siṃhaladvīpa, and Ceilão from Arabic SīlānSiṃhala. It is not settled.

Tamil Īḻam is attested in Tamil-Brahmi and in Sangam literature, and serious scholarship argues it is independent of — and possibly prior to — Sīhaḷa, with Sīlān derivable from it directly. Burrow and Emeneau derived Īḻam from Sīhaḷa; others reverse it. The direction is genuinely open.

And Siṃhala's own "lion" etymology is a nirvacana — a Sanskrit gloss with a lion-ancestor story attached to a name that may never have meant lion. Deriving everything from the Sanskrit form and calling it settled is the error this page exists to expose.

The regular laws behind all of it

Indo-Iranian s → h. Sanskrit soma / sapta / asura / sindhu / sarasvatī → Avestan haoma / hapta / ahura / hindu / haraxvaitī. Utterly regular, and it produces the word Hindu.

Eastern r → l. Māgadhī has l where western Indo-Aryan has r. Ashokan rāja appears as lāja in eastern recensions of the same edict. This is the dialect the Śatapatha mocks as he ’lavo for he ’arayo — the first recorded use of mleccha.

v ~ b merger. Eastern Indo-Aryan merges them: Bengali says biśwa for viśva. Sanskrit Vaṅga gives both Banga and Bengal.

Prakrit cluster simplification. Sanskrit clusters assimilate to geminates: sapta → satta, hasta → hattha, sarpa → sappa, dharma → dhamma. This is why Ashoka's word is dhamma, not dharma.

Tamil has no phonemic aspirates or voiced stops. Sanskrit borrowings flatten: bhāgavata → pākavata. Voicing is allophonic — which is exactly why tamiḻ comes out of Prakrit as damiḷa with a d.

Greek psilosis. Ionic Greek dropped initial h. Persian Hindu arrives as Indós — and the missing h is why English has both Hindu and India from one river.

And the traffic ran the other way

European vocabulary for the goods that moved west is South Asian, and mostly Dravidian at the root.

EnglishViaFrom
riceGreek óryza → Latin → Old FrenchTamil arisi
gingerGreek zingíberis → LatinPali siṅgivera ← Tamil iñci-vēr, “ginger root”
pepperGreek péperi → LatinSanskrit pippalī, long pepper
sugarArabic sukkar → ItalianSanskrit śarkarā, “grit, gravel”
camphorArabic kāfūrSanskrit karpūra
berylGreek bḗryllosPrakrit veruliya ← Sanskrit vaiḍūrya, from a place-name
mangoPortuguese mangaTamil/Malayalam māṅkāy
catamarandirectTamil kaṭṭumaram, “tied wood”

Sanskrit's own word for pepper is yavanapriya“beloved of the Greeks.” A commodity named in Sanskrit after the foreigners who wanted it, whose own word for it was borrowed from India.

Trace a name far enough and you find out who was doing the naming.

How to use this page without abusing it

Resemblance is not derivation. Sound change is a claim about regular correspondences, not about words that look alike. Three traps worth naming, all of which circulate widely:

yoni and Yona are unrelated — one is a Sanskrit word from a root meaning to join or contain, the other is the Ionians via Persian. Nisha and Pārthiban are from Sanskrit niśā, night, and Pārtha, son of Pṛthā — not from Nisa or Parthia. Vyāghra has nothing to do with a modern pharmaceutical trade name.

The Sindhu and Tamiḻ chains hold because they run on regular, independently attested laws and are documented at every step. Meluhha is on this page precisely because it doesn't.

Manu 3.9 — the names you may not marry

A naming rule that works as a boundary rule

The prohibition, in the Sanskrit

na ṛkṣa-vṛkṣa-nadī-nāmnīṃ na antya-parvata-nāmikām
na pakṣy-ahi-preṣya-nāmnīṃ na ca bhīṣaṇa-nāmikām

"He should not marry a girl named after a constellation, a tree or a river; nor one named after a low-caste person or a mountain; nor one named after a bird, a snake or a servant; nor one with a frightening name."

Eight categories of name, and every one of them is a naming convention.

Forbidden name-typeWhere such names are ordinary
ṛkṣa — star, constellationTamil mīṉ names: aṟumīṉ, ēḻumīṉ, vaṭamīṉ. Naming after asterisms is standard Dravidian practice
vṛkṣa — treeTree names are ubiquitous in Ādivāsi and Dravidian naming, and in clan totems
nadī — riverRiver names as personal names
parvata — mountainHill and mountain names — the kuṟiñci landscape
pakṣin — bird · ahi — snakeTotemic naming. Bird and snake clan names are widespread among Ādivāsi communities and in Nāga traditions
antya — last-born, low-casteExplicit
preṣya — servantExplicit

The rule bans marrying a woman named for a star, a tree, a river, a mountain, a bird or a snake — which is a description of how the people outside the system named their daughters.

Read as a naming rule it is arbitrary. Read as a boundary rule it is precise. The listed categories are not random: they are totemic and topographic naming, the conventions of clan and landscape rather than of lineage and gotra. A single verse converts onomastics into an endogamy filter.

And 3.10 gives the positive: marry one with a saumya name — auspicious, mild, Sanskritic.

And 3.5 supplies the machinery

asapiṇḍā ca yā mātur asagotrā ca yā pituḥ

"One who is not sapiṇḍa of his mother and not of the same gotra as his father." Both terms occur zero times in the Rigveda, the Atharvaveda, the Taittirīya Saṃhitā and the Śatapatha. The rule arrives with the text that states it.