How Patañjali
वदति — argues

Sanskrit grammar has one text with three authors, a century or more apart, and a rule for what to do when they disagree.

पाणिनिः states the 3,983 sūtras of the Aṣṭādhyāyī. कात्यायनः, generations later, writes वार्त्तिकानि — terse notes flagging where a sūtra over-generates, under-generates, or could be shorter. पतञ्जलिः, later still, writes the महाभाष्यम् — not a line-by-line gloss of either, but a discussion about both: is the vārttika's objection real? Does Pāṇini's rule survive it as stated, or does it need the fix? Tradition sums up the whole arrangement in four words: यथोत्तरं मुनीनां प्रामाण्यम् — "authority increases with lateness." Where the three disagree, the later voice is followed — not because it is cleverer, but because it has already had to answer the earlier one.

That makes the Bhāṣya something narrower and more useful than "commentary." It is Patañjali arguing — with an imagined objector, with Kātyāyana's vārttikas, sometimes with a straw version of his own last sentence — and the arguing has a shape: a question is raised, an objection follows (पूर्वपक्षः, "the prior side"), a reply answers it (सिद्धान्तः, "the settled view"), often more than once before the discussion moves on. Nothing on this page labels which sentence is which. What this page does instead is point at six phrases Patañjali uses, over and over, to make exactly those moves — because he does the labelling himself, in his own words, and a reader only has to learn to notice them.

These are not a scholar's invented taxonomy. They are Patañjali's own recurring vocabulary, counted directly out of the corpus's text of the Bhāṣya — every occurrence tokenised on word boundaries, not on a raw substring, because "न वा" as a bare substring also turns up as the tail of one word running into the head of the next (…अर्थेन वाक्येन…, …कीलेन वा…) and would inflate the count with pairs Patañjali never wrote as a phrase.

Kātyāyana's vārttikas are proposals, not the last word — Patañjali tests each one the same way he tests Pāṇini.

The seven densest passages measured for this page run 8,000 to 34,000 characters — honest to excerpt from, dishonest to call "unpacked." is complete at a few hundred characters, and it still shows two different markers doing two different jobs, plus a fresh objection that carries no marker at all — which is itself the point: not every turn in an argument is signalled.

The full passage, for reference — every step above, in order, markers highlighted the same way as the reader further down the page:

Pick a passage — an opening excerpt of one of the eight densest, or the vārttika passage above — and its markers light up. Turn a marker off in the legend to see the same text without that highlight; step through the densest passages in order with the walker below the picker.

of the 1,702 commented sūtras carry none of the six markers. That is not a gap in the counting — it is most of the Bhāṣya, and it should look like nothing in particular, because it is nothing in particular: continuous exposition, examples, cross-references, without a turn to signal.

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