अनुवृत्तिः — the words a sūtra does not say

Terse on the page, long in force

Orientation

The Aṣṭādhyāyī has 3,983 sūtras and covers the whole grammar of a classical language in roughly the length of a long short story. That is only possible because most of what a rule needs is not written in it — it is already in force from something said earlier, and stays in force until it is told to stop.

आद्गुणःTwo syllables, read as four words

Page 03 used this rule without pausing on it: देव + इन्द्रः → देवेन्द्रः. Read cold, आद्गुणः ("of आ, गुणः") does not even say what follows the आ, or that the two words must be in the same breath group. Nothing about it is wrong — it is simply not finished, and was never meant to be read alone.

What the page shows and what a reader must supply

Put back together: is what आद्गुणः actually says. Two words on the page, four needed to apply it — and every one of the other three is recoverable, because the corpus records exactly where each was last stated.

This is not an editorial gloss. The added words are not this page's paraphrase of what the rule "really means" — they are the corpus's own अनुवृत्तिः and अधिकारः fields, annotated by hand against the traditional commentaries for 3,489 of the 3,983 sūtras, 9,266 edges in all. Nothing below is inferred from context; every arrow points at a specific earlier sūtra.

Which way it runs, and where it stops

6.1.77 →

Anuvṛtti only runs forward — a word is carried out of the sūtra that states it and into the ones after, never backward into what came before. And it does not run forever: at some point the topic changes enough that the word no longer applies, and it lapses.

The tradition has names for the two ways this shows up. मण्डूकप्लुति — "the frog's leap" — is a word that skips over one or more sūtras that do not need it and reattaches further down, exactly as a frog lands past the ground in between. सिंहावलोकन — "the lion's glance back" — is the coarser habit of a whole topic looking back over everything it has passed, the way a lion checks the ground it has already covered before moving on. Both are the commentators' vocabulary for patterns in the text, not something this corpus encodes — what it encodes is only the edge itself, one word to the sūtra it last came from. The page below is real enough to earn the first name.

अचि, leaping across six.one.seventy-seven to six.one.one-twenty-four

— 6.1.77 — states अचि once. Twenty-eight later sūtras cite it directly as their source for that word, spread across forty-eight sūtra-slots from to . It is not every sūtra in that stretch — the chart marks which ones, and the gap in the middle is fourteen sūtras long.

अनुवृत्तिः is not अधिकारः

The confusion worth killing

Both put a word into a sūtra that does not state it, and both leave a reader who does not know the difference reading the wrong sentence. They are not the same mechanism, and page A3 covers the second in full — this is only the contrast.

अनिदितां हल उपधायाः क्ङितिOne sūtra, both routes, side by side

6.4.24 is four words on the page. Reading it needs seven — and the three missing ones arrive by two different routes:

	अनुवृत्तिः	अधिकारः
what it is	a word, carried forward one sūtra's decision at a time	a heading, declared once and binding over a stated range
how far it reaches	only as far as the next sūtra keeps citing it — it can lapse early, or leap	every single sūtra between its start and its own stated end, no exceptions
in this example	नलोपः, from — one hop	अङ्गस्य () and असिद्धवदत्राभात् ()
measured reach	11 sūtras cite नलोपः from 6.4.23 directly

The test that actually works. Do not ask whether a word "feels like" a heading. Ask whether it is stated once and never re-earned — that is an adhikāra, and its reach is written on the record as a hard end-point (govEnd). If instead every sūtra that carries the word has its own citation back to a specific predecessor, and some sūtras in between opt out, that is anuvṛtti.

The graph is acyclic, and that is a fact about the text

Measured

Every from edge in the corpus points at a strictly earlier sūtra. Nothing inherits from something that has not been said yet — which is exactly why the Aṣṭādhyāyī can be read once, forward, and why a program can compute what page A2 shows without ever getting stuck in a loop.

The longest chain the corpus contains

Following, at each sūtra, whichever of its own inherited words reaches back the furthest — and repeating — the deepest run found anywhere in the 3,983 is sūtras long, entirely inside one pāda:

Nothing to inherit

sūtras carry no recorded anuvṛtti at all — most of them the first sūtra of a new topic, where there is nothing yet to carry forward. The grammar's own opening is the clearest case: and each define a term from nothing, and the very next sūtra, , inherits both of them at once — the first anuvṛtti edges in the text.

Walk the ladder

Interactive

Pick a sūtra. Read what it inherits and where each word was last stated — then click a source to make it the focus, and keep climbing. Switch direction to see the reverse: what later sūtras take from this one.

आरम्भः

दिशा

The most-inherited words in the grammar

Reference

Word	Times inherited	First stated at

Where this page's numbers come from. Every count is computed from sutra_rec()'s own अनुवृत्तिः, अधिकारः and usedBy fields across all 3,983 sūtras at build time — walked fresh each build, not typed once and left to drift. The ladder above ships a curated slice of that graph, chosen to let a reader climb three real chains end to end rather than the whole 9,266-edge graph; anything it points outside that slice is still shown with its real text, just not climbable inside the widget.

Check yourself

अभ्यासः

Six questions, each turning on the same distinction: what is written, what is carried in, and by which route.