अव्ययानि — the indeclinables that carry the tone

What अव्यय means, and why it is one class

Four sūtras build the definition outward from a base case, and a fifth says what happens grammatically once a word is in the class. Nothing here is a list memorised for its own sake — each clause exists because some rule downstream needs to be able to say "this word is अव्यय" in two syllables rather than enumerate every case where it matters.

स्वरादि · निपातThe base rule

— 1.1.37: the words of the स्वरादि gaṇa, and the निपातs, get the saṃjñā अव्यय. Two different kinds of word, one rule. स्वरादि is a closed gaṇa — a list, named for its first member स्वर् — of words that simply are avyaya by membership, no further test. निपात is a functional category: a particle that is not a verb, not a nominal, not a prefix, and whose job is to shade meaning rather than name a thing or an action.

The corpus's own स्वरादि gaṇa, exactly as the Gaṇapāṭha lists it — words:

अथ and मा are on this list too. Both also turn up below in the निपात half of the story (चादि, 1.4.57) — the two halves of 1.1.37 are not disjoint vocabularies, they are two different routes into the same saṃjñā, and a word can qualify by either.

निपातःWhere the particles actually come from

1.1.37 hands "निपात" the saṃjñā; it does not say which words are निपात. That work happens in a long run of separate sūtras, 1.4.56 through 1.4.97 — "from here up to before ईश्वरे, these get called निपात." Inside that span sits 1.4.57 , which opens a gaṇa named for its own first word: चादि, "च and the rest."

This is the single richest word-list behind this whole page — words, read straight from the Gaṇapāṭha:

Almost every particle taught below is on this list by name. च, वा, एव, न (as नञ्), मा (as माङ्), किल, अथ, तु, नु, इव and इति are all sitting in it — highlighted above. That is not a coincidence this page is claiming; it is what the Gaṇapāṭha actually says, and it is the strongest piece of corpus support this topic has. The one working particle not on this list, हि, gets an even more specific citation of its own — see below.

1.1.38–41Four more avenues into the same saṃjñā

Taddhita words ending in a partial-case affix

— 1.1.38: a तद्धित-formed word is अव्यय if the affix that made it does not attach across all seven cases (असर्वविभक्तिः). This is the rule behind the whole -तस्, -त्र, -दा, -था family a few sections down. The Kāśikā's own examples:

कृत् words ending in म् or in a diphthong

— 1.1.39: a कृत्-formed word counts too, if it ends in म् or in an एच् sound (ए ऐ ओ औ). Both halves are real — the म्-ending half covers everyday words like the gerund and the infinitive; the एजन्त half is mostly Vedic, and the Kāśikā's example is exactly that:

क्त्वा, तोसुन्, कसुन् — named individually

— 1.1.40. If 1.1.39 already covers म्-final words, why name क्त्वा again? Because at the derivational point where 1.1.39 would need to check, these three affixes have not yet lost the consonant that makes them म्-final — the check would come too early. So Pāṇini just names them outright rather than trust the general rule to catch them in time. Kāśikā:

अव्ययीभाव — the compound that behaves like its own members

— 1.1.41: an अव्ययीभाव compound is itself अव्यय — sensible, since its first member always already is one. Kāśikā's example, a compound of उप "near" with अग्नि "fire":

2.4.82The mechanism — what "unchanging" actually does

— 2.4.82: after an अव्यय, the feminine stem-marker आप् and the case-ending सुप् that would otherwise follow are लुक् — completely elided. Read that again: the rule presupposes the ending is there to begin with. An अव्यय is not a word the grammar forgot to inflect. It is inflected, exactly like any other nominal stem, and then the result of that inflection is wiped out by name.

Why this matters beyond trivia. Because the ending is elided rather than absent, 1.1.62 प्रत्ययलोपे प्रत्ययलक्षणम् — "where an affix is elided, the mark of that affix still counts" — keeps applying downstream. सन्धि, accent, and other rules that care about *which* vibhakti was there can still see it, even though no reader ever will. The elision is orthographic; the grammar's memory of it is not.

The particles that do real work

निपाताः

Thirteen words, none of them inflecting, none of them translatable by a single fixed English word. Each entry below gives one genuine sentence — read the Sanskrit, then the gloss, then the note on where it sits and what governs its placement.

च · वा · अपिJoining, choosing, including

च — and

च never opens the phrase it joins; it follows the word it attaches to, and when it conjoins a list it appears after every member, not once before the whole list the way English "and" does.

रामः च सीता च वनं गच्छतः।

Rama and Sītā — both — go to the forest. (lit. "Rama-and Sītā-and the-forest go-both")

वा — or

Same second-position habit as च, one disjunct at a time.

रामः वा लक्ष्मणः वा वनं गच्छतु।

Let either Rama or Lakṣmaṇa go to the forest.

वा has a second job, inside the grammar itself. — 1.1.44: when Pāṇini's own sūtras use न or वा, the two together get the technical name विभाषा, "optionality." The Kāśikā glosses this line as "इतिकरणोऽर्थनिर्देशार्थः" — "the word इति here is there to point at the meaning being discussed." Pāṇini is not shy about using ordinary particles to talk about his own grammar.

अपि — even / also / and — position decides

— 1.4.96 names अपि individually and lists four senses at once: सम्भावना "even" (a possibility being raised), अन्ववसर्ग "go ahead" (permission), गर्हा "even — how could you" (reproach), and समुच्चय "also" (addition). Which one a reader hears is not marked by anything except where अपि stands and what the sentence is doing.

रामः अपि गच्छति।

Rama, too, goes. (समुच्चय — inclusion)

बालः अपि इदं जानाति।

Even a child knows this. (सम्भावना — a low bar, cleared)

अपि keeps a second citizenship. The same word is also an उपसर्ग under 1.4.58 — the same syllables, a different job, depending on whether it sits next to a verb root (उपसर्ग) or a full clause (निपात).

इव · खलु · किल · नुComparing, and coloring the tone

इव — like, as if

Enclitic: it follows the thing being compared, never precedes it.

चन्द्रः इव मुखम्।

A face like the moon.

खलु — indeed, verily

A mild, softening emphasis — often inviting the listener's agreement rather than just asserting.

सत्यम् खलु वदति।

He does indeed speak the truth.

किल — it is said, apparently

Reportative: the speaker is passing on a claim, not vouching for it.

सः पण्डितः किल।

He's a scholar, apparently — so I hear.

नु — then? indeed?

Deliberative, mostly in questions expressing doubt or wonder.

किं नु करोमि?

What, then, am I to do?

एव · न · मा · हि · तु · अथAsserting, denying, reasoning, contrasting

एव — the emphatic that changes what is asserted

एव does not mean any one English word; it narrows whatever it follows to the exclusion of alternatives. Because where it lands changes what gets narrowed, it gets the interactive demo below rather than a single fixed example here.

रामः एव गच्छति।

It is Rama — and no one else — who goes.

न — not

सः न गच्छति।

He does not go.

मा — don't (a different mechanism from न)

— 3.3.175: मा, for prohibition, takes लुङ् (the aorist stem) rather than the imperative. And — 6.4.74: specifically in construction with मा, the augment (अट्) that लुङ् normally carries is blocked.

मा भूत् एवम्।

Let this not happen. (भू + लुङ् is ordinarily अभूत् — मा strips the आ- and leaves भूत्)

न and मा are not interchangeable negators. न simply denies; it never touches the verb's own shape. मा actively reaches into the verb's tense-formation and removes a piece of it. Mistaking one for the other is one of the most common learner errors with negation.

हि — for, because (never sentence-initial)

— 8.1.34, on accent: हि, when attached to a finite verb, keeps its prominence rather than going toneless like most निपातs in a sentence. Kāśikā's own examples show exactly where it sits:

सः न आगच्छति, व्याधितः हि अस्ति।

He isn't coming — he's ill, you see. (हि follows व्याधितः, the first word of its own clause, never starts one)

तु — but, however

रामः गच्छति, लक्ष्मणः तु तिष्ठति।

Rama goes, but Lakṣmaṇa stays.

अथ — now, then (opening a new matter)

अथ marks a transition — the classic use is opening a whole work: this site's own Yogasūtra reader begins अथ योगानुशासनम्, "now, the teaching of yoga [begins]," and the Brahmasūtra opens the same way, अथातो ब्रह्मजिज्ञासा.

अथ रामायणस्य कथा आरभ्यते।

Now the story of the Rāmāyaṇa begins.

इति

the quotation marker

Sanskrit has no word for "that" in the English subordinating sense — no exact match for "he said that he was going," "the fact that he was ill." One particle does the entire job instead, by closing off whatever came before it as the thing being reported, named, or reasoned from. Arguably no single word carries more of the syntax of Sanskrit prose.

इतिFour jobs, one particle

1 · Closing quoted speech

रामः "अहं गच्छामि" इति अवदत्।

Rama said, "I am going." (lit. "...'I go' — thus — he said")

2 · Closing quoted thought

सः "अहं जेष्यामि" इति अचिन्तयत्।

He thought, "I will win."

3 · Marking a title or name

रामायणम् इति काव्यम् प्रसिद्धम्।

The poem called the Rāmāyaṇa is famous. (lit. "'Rāmāyaṇa' — thus [called] — poem, famous")

4 · Closing a reason — subordination without a subordinator

सः व्याधितः इति न आगच्छत्।

He did not come, because he was ill. (lit. "he [was] ill — thus [being the case] — did-not-come")

Same mechanism throughout. In every one of the four, इति closes off a stretch of language — words someone said, words someone thought, a name, a reason — and hands it to the verb as a single unit. English needs four different devices (quotation marks, "that," "called," "because") for what Sanskrit does with one particle in one position: right after the thing being closed.

Where this section's citation comes from. इति is on the चादि gaṇa list above (1.4.57) — its निपात status is exactly as attested as च's or एव's. What is not corpus-sourced is the four-way breakdown of its uses: that classification is this page's own, illustrated with authored sentences, since the corpus has no sentence bank to draw genuine examples of इति-in-context from.

Move the particle, change what's asserted

Interactive

One three-word sentence — रामः वनं गच्छति, "Rama goes to the forest." Nothing about the words changes. Only where एव or अपि lands changes, and that alone is enough to change what the sentence is actually claiming. Watch the English shift with each click.

particle

placed after

The adverbial families — a system, not a list

1.1.38 · 5.3

This is the taddhita half of 1.1.38 laid out in full. Four affixes each attach to the same small set of pronoun-and-quantifier stems, in the senses "from," "at/in," "at the time of," and "in the manner of" — and once it's a grid, the pattern is obvious even though no single sūtra states it as a grid.

stem	-तस् · 5.3.7 from	-त्र · 5.3.10 at, in	-दा · 5.3.15 at the time of	-था · 5.3.23 in the manner of

किम् breaks the pattern in two places. Its "where" cell has two attested answers — क्व, from the किम्-specific 5.3.12 (which the Kāśikā calls an अपवाद, an override, of the general त्रल्), and कुत्र from the general 5.3.10 itself — the Kāśikā notes "केचिदिच्छन्ति। कुत्र।", "some [grammarians] want कुत्र [too]." Both are real Sanskrit; grammarians disagreed on which the rule intends. And its "manner" cell, कथम्, does not use -था at all — a separate affix थमुँ (5.3.25 ) replaces it specifically after किम्.

इदम् does not follow the pattern anywhere. Every one of its cells above is its own separate suppletive rule — इह "here" (5.3.11, overriding त्रल्), एतर्हि "now" (5.3.16, overriding दा in the time sense specifically), इत्थम् "this way" (5.3.24, sharing थमुँ with किम्). None of them is इदम् plus a family affix. If a first course presents this grid as fully regular, इदम् is the row that shows it is not.

One more family belongs here in spirit, though it is not part of 1.1.38's -तस्/-त्र/-दा/-था group: 5.1.115 — the affix वति, "same as, like." The Kāśikā's own examples are on ordinary nouns, not pronouns:

Applied to the pronoun stems above, the same affix regularly gives तद्वत्, यद्वत्, सर्ववत् — "like that," "like which," "like all." These three are this page's own extension of the rule to match the shape of the grid above, not the Kāśikā's chosen illustration, and are marked as such rather than presented as quoted.

Where this grid comes from. Every filled cell is typed directly out of the Kāśikā passage cited beside it — none of it is generated. The corpus's paradigm database (shabda/) covers declinable nominal stems only; none of these words are declinable, so none of them are in it, and C.has() returns false for every one. This is exactly the gap the brief for this page warned about, stated rather than papered over.

Full reference

Coverage

The 21 kṛt affixes the engine itself marks अव्यय

Affix	इत्-markers	Group	Stated at

Every particle taught above, and its citation

Particle	Sense	Attested

How to read the two tables above. The kṛt table is pulled whole from core/vidyut.json — every row the prakriyā engine itself flags avyaya: true, no filtering. One entry, कसे, carries no group and no citing sūtra in the underlying data; the table shows that gap as a gap rather than inventing a value. The particle table's "attested" column reports exactly what sections 1 and 2 of this page found: a place in the चादि gaṇa (1.4.57), an individual sūtra, or — for none of the fourteen, as it turns out — neither. That last possibility existed going in; the page reports what was actually found, not what was expected.

Check yourself

अभ्यासः

Six questions. Pick an answer to see which sūtra decides it.