संस्कृतव्याकरणम् · 04 of 24 Home
Aṣṭādhyāyī 1.2.29–33 · 6.1 · फिट्सूत्राणि

Pitch, and where
the record is silent

Sanskrit was a pitch-accent language: one syllable of a word was spoken higher than the rest, and which one could change the meaning entirely. The rules for where it falls are fully preserved. The marks themselves, in most of what survives, are not — and a page that taught this without saying so would be teaching you to check something you cannot check.

उदात्तः
=
उच्चैरुदात्तः
high 1.2.29
उदात्तः — high अनुदात्तः — low स्वरितः — the two together

Why it mattered

phonemic

Accent in Sanskrit is not stress and not emphasis. It is a distinguishing feature — two words identical letter for letter can be different words because the high tone sits in a different place. The tradition's own favourite demonstration is a story about a ritual gone wrong.

इन्द्रशत्रुःThe word that killed the wrong personTaittirīya Saṃhitā 2.4.12

Tvaṣṭṛ performed a rite to produce a son who would be the slayer of Indra. The word he wanted was a tatpuruṣa — इन्द्रशत्रुः with the accent on the final member, "one whose enemy-ship is directed at Indra". What he actually uttered had the accent on the first member, making it a bahuvrīhi: "one whose enemy is Indra" — that is, one whom Indra would slay.

The son, Vṛtra, was duly killed by Indra. The accent had inverted the compound.

This is not folklore attached to grammar — it is grammar. 6.2.1 states that in a bahuvrīhi the prior member keeps its own accent, which is exactly the difference the story turns on. Compound accent is a real chapter of the Aṣṭādhyāyī (6.1–6.2), not a curiosity, and it is why a reader who ignores accent is missing a layer of the language that its own grammarians treated as load-bearing.

The three tones, defined

1.2.29–33

Pāṇini defines them in five consecutive sūtras, and they are as brief as anything in the grammar. Read them in order — the third is defined out of the first two.

सूत्रम्textwhat it says
स्वरित is a compound tone, not a third pitch. समाहारः स्वरितः — a combination. And 1.2.32 says the first half-mora of it is udātta: the syllable begins high and falls. That is why svarita behaves in the rules sometimes like udātta and sometimes like anudātta — it is physically both.
अनुदात्तम्Most of a word is low6.1.158 · 8.1.18

— 6.1.158: a pada is anudātta except for one syllable. One high tone per word, and everything else falls away from it. That single sūtra is the whole shape of the system.

— 8.1.18: and a whole word loses its accent when it is not at the start of a sentence, under certain conditions. Accent is not only a property of words; it is a property of words in place.

Where the accent is still doing work

measured

उपदेशेThe Dhātupāṭha marks it, and the grammar reads it1.3.12 · 1.3.72 · 1.3.78

A root is not listed in the Dhātupāṭha as you would write it. It is listed in उपदेश — a citation form carrying markers that no actual word ever shows. Among those markers is the accent, and three sūtras read it to decide something a classical reader meets in every sentence: whether the root takes परस्मैपद or आत्मनेपद endings.

So the accent survives here in the only form that matters to a grammar: as an instruction. Below is what the corpus actually shows when the marker is checked against the pada each root is assigned.

upadeśa markerपरस्मैपदआत्मनेपदउभयपदreading
Read that table honestly, because it is not the tidy result a textbook would print. The svarita marker is nearly deterministic — 68 of 69 roots that carry it are ubhayapada, which is 1.3.72 doing exactly what it says. The anudātta marker is a strong tendency, not a rule: 444 of 704 are ātmanepada, and the rest are not. That is not the corpus being wrong. 1.3.12 says अनुदात्तङित् — anudātta or ṅit — and then some eighty further sūtras (1.3.13–93) assign pada for particular roots and particular senses, overriding the default. The marker is one input among many.

A whole text about one question

फिट्सूत्राणि

फिषोऽन्त उदात्तःThe first sūtra, and the default it setsफिट् 1.1

फिट् is Śāntanava's own term for a प्रातिपदिक — a nominal stem. His first sūtra says the accent falls on its last syllable, and the eighty-six that follow are the exceptions, grouped by meaning: words for the colour red, words for a house, words for a river, words ending in particular affixes.

It is worth browsing even if you never need it, because it shows how a grammarian handles a domain where the general rule covers less than the exceptions do.

What this site cannot show you

honest

Accented text does survive — the Ṛgveda, the Taittirīya and Śatapatha traditions, and the manuscripts and printed editions that preserve their marking (॑ for udātta in some schemes, ॒ for anudātta). But the Aṣṭādhyāyī as it circulates, the Dhātupāṭha's Devanāgarī forms, and the paradigm tables this series is built on are all unaccented. The rules are complete; the worked examples are not.

But the commentary is another matter

This page first said you could not check a single prediction of the accent system against the rest of this site. That was wrong, and page A12 caught it. The sūtras carry no marks and neither do the paradigms — but the commentary carries them in quantity, and it is not scattered evenly:

layerpassagesaccentedshare

So the honest limit is narrower than it first looked, and more useful: the tables cannot show you accent, but the commentary can. If you want to see a rule of this page actually applied to a word, read the Bālamanoramā or the Tattvabodhinī on a sūtra about svara — they mark it because they are arguing about it.

Check yourself

अभ्यासः

Six questions on the system, since the marks are not here to be read.