असिद्धत्वम् — the tripādī, and why order is the grammar

Why a rule-set needs a way to forget

8.2.1

The Aṣṭādhyāyī is stated as roughly four thousand independent if–then statements, in no announced order, each free to trigger the next. A system like that loops. Rule A changes a word so that rule B's condition is now met; B fires and changes the word so that A's condition is met again; nothing ever settles. Real languages do not behave this way, so the grammar needs a way to say this change may not be used to re-open a question already answered.

पूर्वत्रासिद्धम्The wall at the end of the grammar

— 8.2.1. Four words carry the whole device. पूर्वत्र — "in what precedes"; असिद्धम् — "not accomplished", treated as if it had not happened. From this sūtra to the end of the grammar — the last three pādas, 8.2, 8.3, 8.4, together called the त्रिपादी — every operation is asiddha with respect to everything before 8.2.1. A rule sitting anywhere in 1.1–8.1 must decide whether it applies by looking at the word as the त्रिपादी had not yet touched it, even after the त्रिपादी has, in fact, touched it.

What it does not block: a त्रिपादी rule seeing what an earlier त्रिपादी rule just did. 8.2 feeds 8.3, which feeds 8.4 — that chaining is exactly how the visarga derivations on this series' page 03 work, and nothing here contradicts them. The wall has one direction: later cannot reach back before 8.2.1; forward, within the त्रिपादी, rules see each other exactly as their numbering suggests.

The Kāśikā gives its own examples on this very sūtra. It notes that in अमुष्मै, अमुष्मात्, अमुष्मिन् a certain उत्व-change is asiddha, "so that the स्मा-forms result" — a rule that would have pre-empted them does not get to see the change and act on it. The commentary is quoted in the box further down, in the Kāśikā's own words, because a paraphrase of four thousand years of grammatical argument belongs to whoever wrote it.

Two walls, not one

8.2.1 · 6.4.22

The Aṣṭādhyāyī states this device twice, at two different scales. Once for the whole grammar, and once for a small pocket inside it — and the difference between them is worth being precise about, because they solve different problems.

असिद्धवत्The local pocket

— 6.4.22. From here to 6.4.129 — a much smaller span, inside a single pāda, called आभीयम् after the word आ भात् in the Kāśikā's gloss — the rules are असिद्धवत्, "as if not accomplished", with respect to each other. This is a different shape of problem from 8.2.1: not a one-way wall against everything earlier, but a pocket where a handful of rules that would otherwise feed each other into a loop are made to see only the word as it stood before any of them started.

The Kāśikā's own pair, on this sūtra. एधि, शाधि — a rule needing a stem to still end in a झल् consonant would, if it could see an earlier change in the same pocket, find a vowel instead and fail to fire; असिद्धत्व keeps it looking at the original consonant, so the expected form results rather than a wrong one. आगहि, जही makes the same point the other way — a deletion rule that would otherwise re-fire on its own pocket-mate's output is blocked. The corpus does not decode which rules these two examples name; the point they are quoted for — that a rule inside 6.4.22–129 sees its neighbours as they stood at the door, not as they now stand — needs no further decoding.

The grammar, measured

Both boundaries are exact positions in a strict linear order of all 3,983 sūtras — in the corpus's own count. The bar below is that order, laid out to scale by how many sūtras each of the 32 pādas holds.

ordinary — visible to whatever comes after it, in the usual way आभीयम् 6.4.22–6.4.129 त्रिपादी 8.2.1–8.4.68

A third consequence: where षत्व and णत्व have to live

Retroflexion is not itself असिद्धत्व, but it explains why Pāṇini needs the wall placed where he places it. 8.4.1 turns न् into ण् after र् or ष् — but only समानपदे, "within the same पद". That question — is this still one word, for grammar's purposes — can only be asked once, at the end, after every rule that could still be joining or splitting words has already run. Put णत्व any earlier and "the same पद" would keep changing under it. Because 8.4.1 sits inside the त्रिपादी, "the same पद" is frozen by the time it asks; because it is asiddha for पूर्वत्र rules, its own output cannot be used to reopen whether some earlier compounding rule should have applied differently. 8.3.59 (the ष-त्व of a following स्) sits in exactly the same position for exactly the same reason.

One word, two stems, and a wall between them

the argument

Two nouns, the same case and number, from the corpus's own declension tables. The two outcomes differ for a single reason, and running the derivation with the wall removed produces a form that is not attested anywhere.

वृक्षैः · राजभिःतृतीया बहुवचन, side by side

— 7.1.9: after an अ-ending अङ्ग, the instrumental-plural ending भिस् is replaced by ऐस्. वृक्ष ends in अ, so वृक्ष + भिस् becomes वृक्षैः, no complications. राजन् ends in न् — 7.1.9 has nothing to fire on, and by every ordinary reading of "apply the rules whose conditions are met", it simply does not apply.

Except that राजन्'s final न् does not survive to the end of the derivation. 8.2.7 deletes the final न् of a प्रातिपदिक — and 8.2.7 is inside the त्रिपादी (its own governing अधिकार stack, in the corpus's sūtra record, lists 82001 पूर्वत्रासिद्धम् by name). After it fires, the stem in front of भिस् does end in अ: राज + भिस्. If 7.1.9 could see that, it would fire — 27 sūtras too late by the numbering, but the corpus places no clock on a rule beyond its number. It does not fire, because 8.2.7's effect is asiddha for anything outside the त्रिपादी, 7.1.9 included. भिस् reaches the end of the word unchanged, becomes भिः by the same रुत्व–विसर्ग pair the visarga page walked through, and the attested form is राजभिः.

The counterfactual step is not a citation. Nothing in the corpus computes "what if 8.2.1 did not hold" — that branch is reasoned out here, on the page, to show what the wall is actually stopping. Its source is P. Kiparsky, "On the Architecture of Pāṇini's Grammar" (Stanford), which gives this exact pair — वृक्षैः against राजभिः, never *राजैः — as the standard argument for why न्-लोप has to sit in the last three pādas.

The same wall, forward-facing — from page 03

The two chains below are the reference page's own cited derivations, re-shown here for what they prove about ordering rather than about visarga. Every step after 8.2.66 is a त्रिपादी rule watching the previous त्रिपादी rule's output — which is exactly what असिद्धत्व permits. The two chains start identically and diverge only because a different rule is waiting at the far end.

What this means for reading a derivation

practice

Every prakriyā player in this series — page 03's sandhi chains, the लकार and कृदन्त derivations on pages 11–18 — steps through rules in a fixed sequence and never lets a later step change an earlier one's answer. That sequence is not house style. It is asiddhatva, followed exactly.

A rule is checked once, against the word as it stands at that point in the derivation — not against some later state the derivation has not reached yet, and not against an earlier state it has already left.
Once the derivation crosses into 8.2.1, no step can send it back out to reconsider a rule from 1.1–8.1. The two visarga chains above could not have run यण् or गुण again after रुत्व just because रुत्व happened to produce something that looked like a vowel.
Inside 6.4.22–6.4.129, the same discipline applies at a smaller scale: a rule in that pocket sees its neighbours as they stood before entering the pocket, not as they now stand.

None of this is a rule you apply by hand while translating a sentence. It is why the step order on every other page in this series is not arbitrary — and why a table that lists outcomes without the rule order behind them is, as page 03 put it, "the shadow of something simpler."

All 32 pādas, and where the walls fall

Reference

पाद	सूत्राणि	परिसरः	स्थितिः

Where this page's numbers and forms come from. The tripādī and ābhīya boundaries, the sūtra count in every pāda and the canonical order of all 3,983 sūtras are architecture(), read straight from core/architecture.json — the same file the reference site uses to draw its own adhikāra map. वृक्ष and राजन्'s declensions are paradigm() cells, cross-checked against the prakriyā engine already. The sūtra texts are sutra() lookups that would fail the build if any reference here were wrong. What is editorial, and marked as such above: the choice of वृक्ष/राजन् as the demonstration (Kiparsky's), the counterfactual step, and this page's summary of the Kāśikā's own examples on 8.2.1 and 6.4.22.

Check yourself

अभ्यासः

Six questions. Each comes down to asking which side of a wall a rule is standing on.