टीकापरम्परा — commentary upon commentary

A stack, not a shelf

परम्परा

Nineteen commentary layers ship with this corpus. Fourteen of them are the actual Sanskrit grammatical tradition — the other five are modern reader aids, and the note at the end of this page says so plainly. Of the fourteen, two — Kātyāyana's vārttikas and Patañjali's Bhāṣya — are A8's and A9's territory. This page is what the tradition built on top of them: twelve layers that do not all address Pāṇini. Most of them address each other.

काशिका → न्यासः A concrete case, before the general claim

The Kāśikā-vṛtti glosses a Pāṇinian sūtra. The Nyāsa — Jinendrabuddhi's काशिकाविवरणपञ्जिका — does not gloss the sūtra. It glosses the Kāśikā: it asks why the Kāśikā's author chose the words he chose, what objection his phrasing forecloses, where his gloss is being economical and where it is silently doing more work than it looks. A reader who opens the Nyāsa expecting a second explanation of Pāṇini's rule will be confused by half of it — because half of it is not about the rule at all. It is about the explanation.

Why this matters for reading, not just for history. A sub-commentary's characteristic move — इति काशिकाकृत्, "so says the Kāśikā's author," naming him and then agreeing, refining, or correcting — only makes sense once a reader knows there is a level below the sentence being read. Section 4 below shows this move happening for real, on one sūtra, in a text from the other tradition entirely.

Two chains, one grammar

वृत्तिः · प्रक्रिया

There are two of these stacks, not one, because there are two ways to walk through the Aṣṭādhyāyī's 3,983 sūtras. Page A7 covers the six routes themselves — Pāṇini's own order against five topical reorderings. What matters here is only what a reordering does to the commentary sitting on top of it: a vṛtti-line text can gloss a sūtra assuming the reader already knows every sūtra that came textually before it, because in that tradition nothing else is possible. A prakriyā-line text cannot assume that — Bhaṭṭoji Dīkṣita moved the sūtras, so a sub-commentary on the Kaumudī is glossing a rule that may now sit beside completely different neighbours than Pāṇini gave it.

वृत्तिशाखाthe vṛtti line — Pāṇini's own order

काशिकाglosses the Aṣṭādhyāyī directly

↓ commented on by

न्यासःon the Kāśikā

पदमञ्जरीon the Kāśikā

प्रक्रियाशाखाthe prakriyā line — reordered topically

सिद्धान्तकौमुदीglosses the Aṣṭādhyāyī, reordered

↓ commented on by

बालमनोरमाon the Kaumudī

तत्त्वबोधिनीon the Kaumudī

प्रौढमनोरमाBhaṭṭoji's own, denser

↓ and abridged as

लघुसिद्धान्तकौमुदीthe Kaumudī, for beginners

↓ commented on by

सुधाon the Laghukaumudī*

सरलाon the Laghukaumudī*

लघुशब्देन्दुशेखरःon the Laghukaumudī

स्वतन्त्रःreorders on its own authority

प्रक्रियासर्वस्वम्Nārāyaṇa Bhaṭṭatiri's own topical rearrangement — a rival to the Kaumudī, not built on it

* The corpus carries no title page or author record for any layer — no field anywhere says "Sudhā comments on the Laghukaumudī." What it carries is where each layer's prose actually falls, and that is measured, not assumed: see §5 for the numbers that put Sudhā, Saralā and the Laghu-śabdenduśekhara here rather than on the full Kaumudī.

Read down the stack

One sūtra, chosen because every one of the twelve layers has text on it — not because it makes an easy story. Each layer below is labelled with what it comments on, not just its name; quoted in full where it is short enough, excerpted with a mark where it is not. English follows only where the translation is mine to stand behind — most of these stay in Sanskrit, on purpose.

आदिरन्त्येन सहेता 1.1.71 — the rule that makes every pratyāhāra work

Already familiar from page 01: this is the sūtra that turns अ इ उ ण् into the pratyāhāra अण् — the first sound, taken together with a later it-marked sound, names everything between them. “The first sound, taken together with an it-marked sound at the end, [names everything] up to that point.” (rendering mine) Twelve commentaries later, that one-line rule has drawn a genuine argument.

Coverage is content

वितरणम्

Only the Kāśikā runs across all 3,983 sūtras without a gap. Every other layer is a record of where the tradition concentrated its attention — and the corpus lets that record be checked, not just eyeballed.

Layer	Comments on	Sūtras	Coverage

Do the two traditions cluster on the same sūtras, or different ones?

Two measurements, not one guess. First, do a vṛtti-line pair and a prakriyā-line pair agree with themselves more than they agree across the divide — Jaccard similarity of their two coverage sets:

Pair	Same line?	Jaccard

Same-line pairs agree more ( average) than cross-line pairs do ( average) — a real but modest effect. The sharper test is what each tradition's gap looks like.

Where the three smallest layers concentrate — checked, not assumed

Sudhā, Saralā and the Laghu-śabdenduśekhara are named after and, on the evidence below, written for the Laghukaumudī specifically — not the full Kaumudī. The measure is lift: how much of a small layer's coverage falls inside a candidate parent's coverage, divided by how much of the whole grammar that parent covers anyway. A lift of 1.0 means no concentration at all — the overlap is exactly what coincidence would give; anything higher is the layer choosing that parent's sūtras on purpose.

Layer	vs. कौमुदी	vs. लघुकौमुदी	vs. बालमनोरमा	vs. तत्त्वबोधिनी

Praudhamanoramā is included as a control: it is not a Laghukaumudī commentary — it is Bhaṭṭoji's own further pass on the full Kaumudī — and its lift toward लघुकौमुदी is the lowest of the four, exactly as it should be if lift is measuring something real.

How many layers does a sūtra attract?

Interactive

Every one of the 3,983 sūtras carries between 2 and 12 of these layers. The bars below are the whole distribution; the table beneath lets you open the extremes — the sūtras the tradition argued over most, and the ones it almost let go.

दर्शय

सूत्रम्	अ.	स्तराः

Click a row above to see which layers cover it, in level order.

This table shows only the two extremes — 399 sūtras with 9 or more layers, 72 with 2 or 3 (the middle, 5-8 layers, is the ordinary case and is already the bulk of the bars above). Every row still links live into the reading view, where the full text of every layer is. Inlining all 3,983 rows here — checked — would have cost roughly 285 KB on its own, more than this whole page's budget.

Check yourself

अभ्यासः

Six questions. Pick an answer to see which layer, or which measurement, decides it.