Evidence page · comparative, by date
Lay the Aṣṭādhyāyī beside everything else being written about language anywhere in the fifth and fourth centuries BCE. Outside South Asia there is almost nothing. Inside it there are two traditions — and the second is routinely dropped from the comparison.
| Work | Aṣṭādhyāyī, "Eight Chapters" — roughly 4,000 sūtras |
|---|---|
| Date | c. 5th–4th century BCE. Contested at the edges; nobody places it later than the 4th |
| Place | Śalātura, in Gandhāra — he is called Śālāturīya. Near the Kabul–Indus confluence |
| Political context | Gandhāra was an Achaemenid satrapy from c. 518 BCE. The most sophisticated grammatical analysis of the ancient world was produced on a Persian imperial frontier |
| Medium | Almost certainly composed for oral transmission. Its extreme compression is an artefact of being built to be memorised |
Not a description. A generative system.
The sūtras are written in a technical artificial language, distinct from the Sanskrit they describe, with its own conventions — including markers attached to elements purely to control rule application, which are then deleted. A formal notation, built for the purpose.
Fourteen lines that arrange the phonemes so that any natural class needed by the grammar can be named by a two-letter abbreviation — aC for all vowels, haL for all consonants. Kiparsky has shown the ordering is close to optimal for the classes actually required. A compression scheme designed against the rule set that uses it.
A formal precedence system: vipratiṣedhe paraṃ kāryam — where two rules conflict, the later applies. Plus general-versus-exception (utsarga/apavāda), obligatory-versus-optional, and inner-versus-outer conditioning. Meta-rules governing rules.
Semantic roles — agent, patient, instrument, recipient, source, locus — assigned at an intermediate level between meaning and case-ending. This is thematic-role theory, and it is the part modern linguists find least believable as a fourth-century-BCE artefact.
Zero morphemes (lopa), abstract underlying forms, and ordered derivation from underlying to surface.
Bloomfield called it one of the greatest monuments of human intelligence. Its rule format is close enough to Backus–Naur Form that it has been proposed the notation be renamed Panini–Backus Form. The resemblance to generative grammar is structural, not decorative — ordered rules, abstract representations, and a metalanguage.
Every literate society that existed
| Where | What existed | When |
|---|---|---|
| Mesopotamia | Sumerian–Akkadian lexical lists and paradigm tables. Genuinely early and impressive lexicography — but lists, not rules. No generative apparatus | 2nd millennium BCE |
| China | Erya, a thesaurus of glosses. Shuowen Jiezi analysing characters. Lexicography and graphology; syntactic analysis develops far later | c. 3rd c. BCE; 100 CE |
| Greece | Plato's Cratylus — philosophy of naming, not grammar. Aristotle distinguishes noun, verb and connective in passing | c. 360 BCE |
| Greece | Dionysius Thrax, Technē Grammatikē — the first Greek grammar. A short handbook: parts of speech, accidents, definitions. No rule system, no derivation, no metalanguage | c. 100 BCE — two to three centuries after Pāṇini |
| Rome | Varro's De lingua Latina; then Donatus and Priscian, whose handbooks structure European grammar for a thousand years | 1st c. BCE – 6th c. CE |
| Alexandria | Apollonius Dyscolus on syntax — the first sustained Greek syntactic treatment | 2nd c. CE |
| Egypt, Persia, Levant | No comparable grammatical tradition | — |
The nearest thing anywhere is a fifteen-page Greek handbook, two to three centuries later, doing a fraction of the work.
And the gap is not one of quality but of kind. Dionysius Thrax classifies. Pāṇini generates. One tells you what a word is; the other tells you how to build every word in the language from a store of roots and affixes, with the rule interactions specified.
The part usually skipped
Āpiśali · Kāśyapa · Gārgya · Gālava · Cākravarmaṇa · Bhāradvāja · Śākaṭāyana · Śākalya · Senaka · Sphoṭāyana
He cites them where they disagree with him. None of their works survive. Behind the Aṣṭādhyāyī is a grammatical tradition already old enough to have factions.
This is an attribution practice — naming your sources including where you depart from them — and it is worth setting beside everything else on this platform about who credits whom.
| Prātiśākhyas | Phonetic treatises, one per Vedic school, describing sandhi and recitation. Partly pre-Pāṇinian |
| Śikṣā texts | Articulatory phonetics. Place of articulation (sthāna) and manner (prayatna), with velar, palatal, retroflex, dental and labial series — a classification that maps onto modern articulatory phonetics without adjustment |
| Yāska, Nirukta | Etymology and semantics. He too names predecessors and records their disagreements |
| Kātyāyana → Patañjali | The vārttikas and the Mahābhāṣya — a commentarial tradition that argues with the text rather than merely transmitting it |
The form yavanānī, glossed by later commentators as Greek writing, appears in the vārttika literature on Aṣṭādhyāyī 4.1.49 — that is, in Kātyāyana, c. 3rd century BCE, not necessarily in Pāṇini himself. The claim that Pāṇini personally mentions the Greek alphabet is common and goes beyond the evidence.
What holds without it: Śalātura sat in an Achaemenid satrapy where Aramaic was the administrative script, on the road between the Indian plains and Iran. Whoever composed the Aṣṭādhyāyī worked in a multilingual imperial frontier zone where several writing systems were in official use.
Whether writing existed in India in Pāṇini's lifetime is genuinely disputed — Falk argued for Mauryan-period introduction; others push earlier; the Indus script is undeciphered and its relation to Brāhmī unresolved.
What is not disputed is that the Aṣṭādhyāyī is built for memorisation. Its compression, its abbreviation system and its rule ordering are all optimised for oral transmission, and the tradition transmitted it orally for centuries.
The most formally rigorous grammar of the ancient world may have been composed and transmitted with no written text at all. That possibility should change how the phrase "oral culture" is used about anywhere.
Tolkāppiyam and Agattiyam
Surveys of "grammar in the ancient world" reliably cover Mesopotamia, China, Greece, Rome and Alexandria, reach Pāṇini, and stop. South Asia had two grammatical traditions, and the second one is a different kind of object entirely.
| Aṣṭādhyāyī | Tolkāppiyam | |
|---|---|---|
| Size | c. 4,000 sūtras | 1,610 nūṟpā, in three books of nine chapters each |
| Date | c. 5th–4th c. BCE | Heavily contested. Earliest layer often placed c. 300 BCE – 100 CE; almost certainly composite, with later strata. Some Tamil scholarship argues considerably earlier |
| Book 1 | phonology within the rule system | Eḻuttatikāram — letters, articulation, and puṇarcci, the sandhi of word-joining |
| Book 2 | morphology, derivation, syntax | Collatikāram — word classes, case, compounds |
| Book 3 | nothing corresponding | Poruḷatikāram — SUBJECT MATTER. akam and puṟam; the tiṇai landscape system binding place, season, hour and emotion; meypāṭu, the bodily expression of feeling; simile; prosody; convention |
Pāṇini has no poetics, no theory of emotion, and no account of the social and ecological world a sentence is uttered in. The Tolkāppiyam does — and treats them as part of grammar rather than as a separate discipline.
These are not two attempts at the same object. One is a generative engine for producing well-formed words. The other is an integrated theory of language, literature, feeling and landscape. Ranking them against each other measures whichever one you chose the yardstick from — and every European history of linguistics chose Pāṇini's.
The Akattiyam, attributed to Agattiyar, is held by the tradition to be the first Tamil grammar. It is lost — surviving only as scattered sūtras quoted in later commentaries. Tolkāppiyar is traditionally his student.
Exactly as Pāṇini names ten predecessors whose works are gone. Both traditions record that they inherited something, and in both cases the inheritance did not survive.
And Agattiyar has been claimed. In Sanskrit tradition, Agastya is a Vedic ṛṣi who crosses the Vindhyas to civilise the south. In Tamil tradition he founds Tamil grammar and the Siddhar medical lineage. Whether the figure is a Tamil culture hero later given a Vedic pedigree, a genuine northern import, or a conflation of two people, is contested — and the "Agastya brought civilisation south" version is itself a Sanskritising move, of exactly the kind catalogued elsewhere on this platform.
Tolkāppiyam is also said to follow the Aindra grammatical school — another lost tradition, argued by Burnell to be pre-Pāṇinian. Whether the Tamil grammar is modelled on Sanskrit grammar or developed in parallel is genuinely disputed. The third book is the strongest argument for independence: nothing in the Sanskrit tradition would have produced it.
Very little, and the little is interesting
| Śālāturīya | "the man from Śalātura" — the name is a place, in Gandhāra |
| Dākṣīputra | Patañjali refers to him as "son of Dākṣī" — by his mother's name |
| everything else | Legend. The biographical stories come from sources a thousand years later and are worthless as evidence |
His own text is the better source for his world: it cites the usages of northerners and easterners, names peoples and places of the northwest, and describes the spoken language, not only the scriptural one.
The frontier keeps producing the decisive work
A speaker deep inside a monolingual region has no occasion to think of their language as a system. A speaker at a boundary does — constantly, and practically.
Gandhāra sat between Indo-Aryan, Iranian, and Dardic speech, on the roads to Bactria and Iran, inside an Achaemenid satrapy where Aramaic was the administrative script and record-keeping was routinely multilingual. Taxila was a teaching centre that both Buddhist and Brahmanical sources describe students travelling to.
And it was far from the Kuru-Pañcāla ritual heartland — the same distance that mattered for every other case on this platform.
Yājñavalkya the breakaway, the teaching-kings, the Buddha in the east, Pāṇini in the northwest. The decisive work keeps arriving from the edge.
Run the test that is run everywhere else on this platform
No other Indo-European tradition produced anything remotely comparable. Greek grammar arrives late and thin. Latin follows Greek. Iranian produced nothing of the kind, despite sharing the Vedic ritual inheritance.
So by the same test used for karma, mokṣa and dhyāna: grammatical science is not Indo-European inheritance. It is South Asian.
The proximate engine is the Vedic recitation apparatus — padapāṭha, which breaks continuous recited text into isolated words, and the permuted recitations (krama, jaṭā, ghana) that recombine them. Preserving a text by memory to that standard requires analysing speech into segments and reassembling it. That is a phonological analysis engine, built for a preservation task.
And notice what it had to describe. The Śikṣā classification handles a sound system with a full retroflex series — which is not Indo-European and did not arrive with the language. The analytic apparatus was built to handle a phonology the incoming language did not bring with it.
Tolkāppiyam performs the same operation for Tamil, with its own five-way lateral and rhotic inventory. Two traditions, on the same subcontinent, doing something no other branch of the family did anywhere.
bhāṣā, prākṛta, and who was allowed to speak what
Pāṇini distinguishes chandas, the Vedic, from bhāṣā, the spoken language — and describes both. Beyond that lay the Prakrits, and they were not treated as equals.
| Prakrit | Where | Feature |
|---|---|---|
| Māgadhī | the eastern Gangetic plain | l for r, ś for s. Mocked in the Śatapatha, used by Ashoka, assigned to servants in drama — see the arc below |
| Ardhamāgadhī | east | the language of the Jain canon |
| Śaurasenī | Mathura region | the prose Prakrit of drama |
| Mahārāṣṭrī | Deccan | the preferred Prakrit for lyric verse |
| Gāndhārī | the northwest | written in Kharoṣṭhī — the language of the oldest surviving South Asian manuscripts |
| Paiśācī | uncertain | "the language of the piśācas" — named for demons |
Dramatic convention, codified in the Nāṭyaśāstra and observed for centuries, assigns speech by status: Sanskrit for high-status men. Śaurasenī for women and for respectable characters of lower standing. Māgadhī for servants and low-status characters.
In a single play, you can hear a person's assigned place in the order from the language the author put in their mouth. A woman of the highest rank speaks Prakrit; her husband speaks Sanskrit. The hierarchy is not described in the text — it is the medium of the text.
And one Prakrit is named after demons. Set that beside mleccha, barbaros, Vaidehaka and Draviḍa: the same operation, applied to a dialect.
One dialect, four verdicts, in this order
| When | Status | Evidence |
|---|---|---|
| c. 900–700 BCE | The first speech ever called mleccha | Śatapatha Brāhmaṇa 3.2.1.24 mocks the asuras for saying he 'lavo instead of he 'arayo — l for r, the defining eastern feature. The word mleccha enters the language pointing east |
| c. 500–300 BCE | The language of the region that changed everything | Magadha produces the Buddha, Mahāvīra and Makkhali Gosāla. Ardhamāgadhī is the language of the Jain canon. The doctrines of rebirth and liberation are credited by the Upaniṣads themselves to kings of the east |
| c. 260 BCE | The chancellery language of the largest empire South Asia had seen | The eastern recensions of Ashoka's edicts show the Māgadhī features — lāja for rāja. The Mauryan state issued its moral and administrative proclamations in it, on stone, from Kandahar to Karnataka |
| c. 200 BCE – 200 CE onward | Assigned to servants and buffoons | Sanskrit dramatic convention gives Māgadhī and its sub-dialects to low-status characters. In the Mṛcchakaṭika, the villain-buffoon Śakāra speaks a Māgadhī-derived dialect whose eastern features are the joke |
Mocked as barbarism, then adopted by the Buddha's homeland, then used to govern an empire, then written into plays as the speech of servants.
Māgadhī was not low-status when Ashoka used it. The dramatic convention that ranks it postdates the Mauryan empire. So this is not a language that was always marginal and stayed so — it is a language that governed the subcontinent and was subsequently assigned to servants by a literary establishment writing in Sanskrit.
And the ranking is a fiction. Real people did not speak by dramatic rule. What the convention records is not how anyone talked but how a Sanskrit-writing elite wished the order of speech to be understood — and that wish was durable enough to outlast the empire it demoted.
Pali is not Māgadhī. The Theravāda tradition calls its language Māgadhabhāsā, but linguists identify Pali as a western or central Prakrit. The Buddha most likely spoke an eastern dialect; the canon that survives is not in it.
The Upaniṣads are in Sanskrit, not Māgadhī. The eastern claim established earlier on this platform is about whose doctrines they are — kings of Videha and Kāśī, and a region the Śatapatha marks as beyond the Sadānīrā — not about the language the texts are written in.
What holds without qualification: Ardhamāgadhī is the Jain canonical language; the eastern Ashokan recensions carry the Māgadhī features; the Śatapatha mocks those same features centuries earlier; and the drama assigns them to servants centuries later.
The Vinaya records two monks of brahmin background asking permission to put the teaching into chandas — the elevated Vedic register. The request is refused, with the instruction that each is to learn it sakāya niruttiyā, in his own speech.
A foundational refusal of the prestige language, at the origin of the tradition — and the reason Buddhist texts exist in Gāndhārī, Pali, Chinese and Tibetan rather than in one sacred tongue whose access could be restricted.
Pāṇini describes bhāṣā, the spoken language of his time, alongside chandas, the Vedic. He records regional variation and cites usages he attributes to northerners and easterners. He is documenting a living language.
What his successors built was different. saṃskṛta means "refined"; prākṛta means "natural, of the source." Grammatical correctness became a marker of standing, and access to the grammatical tradition was socially restricted, as Vedic study was.
The tool was descriptive. The institution built on it was not. That distinction matters when the Aṣṭādhyāyī is invoked as evidence of a civilisation's genius — the genius is real, and the use made of it is a separate question with a separate answer.
And note where he came from: Gandhāra, the frontier — not the Gangetic Brahmanical heartland. As with the teaching-kings of the Upaniṣads and Yājñavalkya the breakaway, the decisive intellectual work keeps arriving from the edge.