तिङन्तयङ्लुक्प्रकरणम्

Covers SK 2650–2656 (7 sūtras): all present. Every unit has its own kaumudī; there are no vārttika blocks. Our source.json has no BM for any unit of this prakaraṇa, so the BM cited below is Wikisource's (uttara-2/0135–0153). The spine is a short pick from four places in the Aṣṭādhyāyī: 2.4.74 → 7.3.94 → 7.4.91–92 → 6.4.20–21 → 1.1.4. The two long units are 2651 (the īṭ rule, which carries the paradigms of seventeen roots) and 2653 (the ṛ-final rule, which carries the paradigms of the ṛ-roots and the y/v-final roots). Every AS-ref below is checked against references/ashtadhyayi/sutraani-data.json, and its skn matches ours for all seven.

• Anuvṛtti from later in the SK.
• bahulaṃ chandasi (SK 3400) gives bahulam to 2650, drawn in by the ca.
• anunāsikasya kvijhaloḥ kṅiti (6.4.15, SK 2666) gives kvijhaloḥ to 2654 and 2655, but not kṅiti.
• The Vedic nipātana dādharti… bobhūtu tetikte… (SK 3596) is the evidence for 2651's guṇa.
• Anuvṛtti from earlier prakaraṇas.
• luk from ṇyakṣatriyārṣañito yūni luk aṇiñoḥ (SK 1276) at 2650.
• At 2651: piti sārvadhātuke from nābhyastasyāci piti sārvadhātuke (SK 2503), hali from uto vṛddhir luki hali (SK 2443), and īṭ from bruva īṭ (SK 2452).
• At 2652: abhyāsasya from atra lopo 'bhyāsasya (SK 2620), and two items from p56: yaṅluk from guṇo yaṅlukoḥ (SK 2630), and ṛdupadhasya with rīk from rīg ṛdupadhasya ca (SK 2644).
• ūṭh, v and anunāsike from cchvoḥ śūḍ anunāsike ca (SK 2561) at 2654 and 2655.
• iko guṇavṛddhī (SK 34) at 2656.

tiṅantayaṅlukprakaraṇam — outline (p57, SK 2650–2656, all 7 sūtras)

p56 (tiṅantayaṅprakaraṇam) closed at 2649 ayaṅ yi kṅiti (śāśayyate), with the print's colophon "iti yaṅantaprakriyā". Ours lacks that colophon, and the English supplies it.

• Where the prakaraṇa begins. The print heads it "tiṅante yaṅlugantaprakriyā 54" (NSP p. 439, scan page-0449). Our name and number are tiṅantayaṅlukprakaraṇam 57. It opens with the rule that deletes the yaṅ, yaṅo 'ci ca (2650).
• Where it ends. At 2656 na dhātulopa ārdhadhātuke, with the murchā forms and the remark that aj has no yaṅluk (vevīyate). The print closes "iti yaṅlugantaprakriyā" (NSP p. 445, scan page-0455). Ours lacks it, so the English supplies it in brackets. The print then heads the nāmadhātu-prakriyā (no. 55), and p58 (tiṅantanāmadhātuprakaraṇam) opens with supa ātmanaḥ kyac (SK 2657).
• Contents in outline.
• 2650: the yaṅluganta. yaṅ is deleted by luk, either before the suffix ac or variously without it. The doubled root, still yaṅ-final by pratyayalakṣaṇa, is conjugated in parasmaipada only, with śap deleted as in the adādi class.
• 2651: īṭ optional before a hal-initial pit sārvadhātuka. The paradigms of bhū (bobhavīti / bobhoti), spardh, gādh, nāth, dadh, skund, mud, kurd, vañc, gam, han, car, khan, yu, nu, hā and svap.
• 2652–2653: ruk, rik and rīk for the abhyāsa of a root with penultimate ṛ or final short ṛ: varvarti / varivarti / varīvarti. The paradigms of vṛt, kṛ, ṛ, gṛh, gṛdh and the contrasting roots (kṝ, tṝ, grah, pracch), with the y/v-final roots haya and harya.
• 2654: ūṭh for the penultimate and v of jvara, tvara, srivi, avi and mava: māmoti.
• 2655–2656: v or ch deleted after r, and guṇa not barred. totorti, momorti. Then the domain-locative ārdhadhātuke, which gives vevīyate and no yaṅluk for aj.

Threads that run through the prakaraṇa:

• There are two yaṅluks (2650, BM, TB).
• The unconditioned (anaimittika) luk comes from bahulam, with no ac. Being antaraṅga, it applies before doubling. Doubling then follows because the stem still counts as yaṅ-final by pratyayalakṣaṇa (sanyaṅoḥ is read with a sixth case, TB). This luk is what the whole prakaraṇa conjugates: bobhavīti, momoditā.
• The conditioned (naimittika) luk comes before the kṛt suffix ac. It is bahiraṅga, so doubling comes first and the luk after. The ac then causes the deletion of part of the dhātu, so 2656 bars its guṇa: loluvaḥ, popuvaḥ, marīmṛjaḥ (TB under 2656).
• Hence 2651's momoditā keeps its guṇa, because there the luk is unconditioned (BM).
• What pratyayalakṣaṇa does and does not carry over (2650: only a property exclusive to suffixes, asādhāraṇa).
• It carries yaṅ-finality, so doubling, the dhātu designation and 2651's "yaṅaḥ" still hold.
• It does not carry ṅit. ṅit is found in non-suffixes too (citraṅ), so there is no ātmanepada, even from spardh or śīṅ (2650). The same reasoning blocks the lengthening in sudṛṣatprāsādaḥ.
• Conditions that rest on the elided yaṅ itself lapse:
• the u by svapisyamivyeñāṃ yaṅi (SK 2645): sāsvapīti (2651);
• grah's yaṅ-conditioned samprasāraṇa: jāgrahīti (2653, TB);
• vī for aj, so aj has no yaṅluk (2656).
• The paribhāṣā śtipā śapā…. A rule that cites its root with śtip, śap, a gaṇa or an anubandha does not reach the yaṅluganta. Examples:
• anudāttaṅita (anubandha): no ātmanepada (2650);
• the gaṇa vikaraṇas (śyan and the rest): only śap comes, and it is deleted because 'carkarītaṃ ca' is read in the adādi (2650);
• rudādibhyaḥ (SK 2474): no iṭ for svap (2651);
• hanter jaḥ (SK 2431): jaṅghahi (2651);
• the five hā rules (2651);
• the ḷdit of SK 2343: no aṅ for gam (2651);
• na vṛdbhyaś caturbhyaḥ (SK 2348) and dyutādi: varvartiṣyati and avarvartīt (2653);
• guṇo 'rtisaṃyogādyoḥ (SK 2380) for ṛ (2653);
• grahijyā… (SK 2412) for pracch (2653).

The paribhāṣā is anitya. guṇo yaṅlukoḥ (SK 2630) shows it, being an indicator that applies generally. So the ku of abhyāsāc ca still reaches han (jaṅghanīti, the Nyāsa; 2651), doubling itself goes through (TB on 2650), and so does bhaṣ in ajargharṭ (TB on 2653).

• "The mention of a base includes its yaṅluganta" (prakṛtigrahaṇe yaṅlugantasyāpi grahaṇam). It rests on the view that ṣāṣṭha doubling is a pronouncing twice (dviḥ prayoga). Three consequences:
• bhūsuvoḥ does reach bobhū, so the Vedic nipātana bobhūtu is restrictive, and guṇa stands in the spoken language (2651);
• vadha replaces han: vadhyāt, avadhīt (2651);
• rules restricted to ekāc roots do not reach it: jaṅgamitā takes iṭ (SK 2246), and jāgrahitā has no lengthening (SK 2562, 2653).
• īṭ and laghūpadha guṇa alternate. With īṭ, an ac-initial pit sārvadhātuka follows, so nābhyastasyāci piti sārvadhātuke (SK 2503) bars guṇa of a light penultimate: momudīti, varvṛtīti. Without īṭ there is guṇa: momotti, varvarti (BM).
• Final vowels take guṇa either way: bobhavīti / bobhoti, yoyavīti / yoyoti.
• Before a ṅit (tas, jhi, …) there is no guṇa: bobhūtaḥ, momuttām.
• The laṅ 2nd sg. in -āḥ. For a stem ending in a voiced stop (d/dh) the pattern is: hal-ṅyādi deletion of s, optional ru by daś ca (SK 2468), the first r deleted by ro ri (SK 173), and lengthening by ḍhralope. Forms: apāspāḥ, ajāghāḥ, acokūḥ (2651); avarvāḥ, ajarghāḥ (2653). The other option keeps the stop: apāspart / apāspard.
• Bracketed labels. The kaumudī lists forms bare. Lakāra, person and root labels are the translator's, following BM and TB as recorded in the unit notes. "(3)" in 2653 is the kaumudī's figure for the three forms with ruk, rik and rīk.
• Where the commentary lies. Only the kaumudī blocks are translated.
• BM: none in our source for 2650–2656. It was read in Wikisource uttara-2/0135–0153 (colophon "…bālamanoramāyāṃ yaṅluganta-prakriyā samāptā").
• TB printed under a neighbouring unit:
• TB on the sūtra of 2650 is at the tail of sk.2649.t1;
• TB on 2654 and 2655 is at the tail of sk.2653.t1, and sk2654 and sk2655 have no t1;
• TB on totorti (the use of 2655) is inside sk.2656.t1;
• the tail of sk.2656.t1, after TB's colophon, is TB on SK 2657. That is p58's matter.

The p55–58 audit (.work/sk-run/p55-58-audit/sk2648-2657.md) found no MUST-FIX in this span and calls every p57 unit clean as translation. A fixer is applying its note-level MINOR items in parallel, and none of them changes the substance outlined here:

• 2651: the notes will add TB's rejection of the 'kecit' derivation of khan (which denies ā in yaṅluk and gives caṅkhantaḥ). The kaumudī's caṅkhātaḥ stands.
• 2653: the pratyāhāra is written val, not "vaL". The mavya sentence: in context mavya, tevṛ and devṛ are examples of the ūṭh-bhāvins that have no yaṅluk, and 'mava bandhane' introduces the contrasting root that does (BM). The English may add "[— these have no yaṅluk]"; the outline reads it that way.
• All units: commentary notes that cite both commentators under one src tag are being split or retagged. The content is accurate.

Print readings (NSP, scan-confirmed) that the outline follows:

• 2650: SK 2159 (ours 2149).
• 2651:
• nityatvād vuk;
• apāspart, apāspard;
• jāghātsi;
• dādaddhi, dādaddhaḥ;
• acokūrt;
• jaṅgataḥ;
• hanter yaṅluk;
• niṣedhas tv anityaḥ;
• the four forms lost by eye-skip after ajāhītām;
• SK 2474.
• 2653:
• avarvṛtīt;
• avarvart;
• arter yaṅluki;
• ajargharṭ, ajarghart;
• one ariyarti;
• SK 173, 2468 and 2654 for the print's misnumbered 1731, 2668 and 2564.
• 2654: anuvartate.
• 2656: the colophon.

Still open for Ajai (mūla):

• 2651 'acañcūrīt' is ours and the print. WS has acañcurīt, and hali ca cannot lengthen before the ī of īṭ.
• 2651 'jāhīti' after jāhītha is ours and the print. WS and BM have jāhīhi (loṭ).
• 2653 'arārīt' (luṅ of ṛ, ruk form) is ours and the print. TB and BM read ārārīt, with āṭ, as an ac-initial root requires.
• 2653 'tadbhāṣyaviruddham' is ours and the print. TB's lemma is tadbhāṣyādi-. TB also records and rejects the manuscript variant āririyāt, ārīriyāt.
• The ours ≠ print slips above (2650, 2651, 2653, 2654, the 2656 colophon) are for Ajai's printed copy. source.json is untouched. Trivial: 2656 momūrcchīti (ours) against momūrchīti (print).

Ordering: SK order throughout. Two paradigm units carry far more forms than the rest, so 2651 (group B) and 2653 (group C) are laid out root by root. Each root shows the stem build in brief: the general abhyāsa steps (halādiḥ śeṣaḥ SK 2179, śarpūrvāḥ khayaḥ SK 2259, kuhoś cuḥ SK 2245) and the yaṅ-abhyāsa rules of p56 (guṇo yaṅlukoḥ SK 2630, dīrgho 'kitaḥ SK 2632, nīk SK 2642, nuk SK 2636 / 2643). Then come the forms by lakāra, with the step that makes each one distinctive.

A. The yaṅluganta itself: yaṅ deleted, the root still doubled, a parasmaipada adādi-type stem — running example: bhū + yaṅ → (luk of yaṅ, unconditioned and antaraṅga, so first) bhū → (doubling: still yaṅ-final by pratyayalakṣaṇa) bhū-bhū → (abhyāsa: bu, guṇa by guṇo yaṅlukoḥ) bobhū → laṭ, parasmaipada, śap deleted (adādi) → bobhoti / bobhavīti (2651)

• 2650 yaṅo 'ci ca 2.4.74 — luk of yaṅ before the suffix ac, and by the ca, variously (bahulam) without it. luk is carried from SK 1276, and bahulam is drawn in from SK 3400. By association with yaṅ, aci is the suffix ac, not the pratyāhāra (TB, BM).
• With ac (BM): lū + yaṅ + ac gives lolūya without the luk and loluva with it. In the plural this is the loluvaḥ of 2656.
• Without ac: bhū + yaṅ gives bobhūyate without the luk and bobhavīti with it.
• The order of steps (kaumudī):
• this luk is anaimittika, so being antaraṅga it comes first;
• doubling follows by sanyaṅoḥ (SK 2395), since the stem is yaṅ-final by pratyayalakṣaṇa (SK 262);
• then the abhyāsa operations (guṇa and the like);
• then laṭ and the other lakāras, since the whole is a dhātu (SK 2304).
• na lumatāṅgasya (SK 263) does not stop the doubling. Doubling is an operation of the yaṅanta, not one conditioned by yaṅ (BM).
• Parasmaipada comes by śeṣāt kartari (SK 2159). There is no ātmanepada by anudāttaṅita (SK 2157), for two reasons:
• ṅit is common to suffixes and non-suffixes, so pratyayalakṣaṇa does not carry it. The same reasoning gives no lengthening in sudṛṣatprāsādaḥ (SK 425).
• It is a citation by anubandha, which śtipā śapā… excludes. So even spardh and śīṅ (anudātta-it and ṅit) get no ātmanepada.
• Vikaraṇa: the gaṇa-cited śyan and the rest do not come; only śap does. carkarītaṃ ca is read in the adādi class, so the śap is deleted (adiprabhṛtibhyaḥ śapaḥ, SK 2423, TB).

B. īṭ optional before a hal-initial pit sārvadhātuka: the paradigms of seventeen roots — running example: bobhū + tip → (īṭ) bobhū-ī-ti → (guṇa, av) bobhavīti; (no īṭ) bobhū-ti → (guṇa) bobhoti

• 2651 yaṅo vā 7.3.94 — after a yaṅ-final root (also under luk, by pratyayalakṣaṇa), īṭ is optionally added to a following hal-initial pit sārvadhātuka. piti sārvadhātuke is carried from SK 2503, hali from SK 2443, and īṭ from SK 2452 (TB, BM).
• Guṇa for bhū in the spoken language. bhūsuvos tiṅi (SK 2224) would bar it. But the Vedic nipātana bobhūtu (SK 3596) teaches a guṇa-less form where the bar was already in place, so the nipātana is restrictive: the guṇa of a yaṅluganta is barred only in the Veda. This also proves that yaṅluk occurs in the spoken language.
• The objection that bhūsuvoḥ cannot reach the doubled stem is answered by prakṛtigrahaṇe yaṅlugantasyāpi grahaṇam. That rests on the siddhānta that ṣāṣṭha doubling is a pronouncing twice.
• The roots (stem build, then forms):
• bhū → bobhū (abhyāsa guṇa).
• laṭ: bobhavīti / bobhoti, bobhūtaḥ (tas is ṅit, so no guṇa), bobhuvati (jhi → at; uvaṅ).
• liṭ bobhavāṃcakāra; luṭ bobhavitā.
• laṅ: abobhavīt / abobhot, abobhūtām, abobhavuḥ.
• liṅ: bobhūyāt, bobhūyātām (vidhi), bobhūyāstām (āśīḥ).
• luṅ: sic is deleted by gātisthāghupābhūbhyaḥ (SK 2223). In the īṭ option, vuk (SK 2174), being nitya, sets aside guṇa: abobhūvīt. Without īṭ: abobhot (print; WS abobhūt). Then abobhūtām, and abobhūvuḥ, where jus depends on the abhyasta (SK 2226) and vuk again sets guṇa aside.
• lṛṅ abobhaviṣyat.
• spardh → pāspardh (p kept by śarpūrvāḥ khayaḥ; ā by dīrgho 'kitaḥ). Being anudātta-it, it still takes no ātmanepada (2650).
• laṭ: pāspardhīti / pāspardhi. The latter comes from pāspardh-ti, with t → dh by jhaṣas tathoḥ and jaś (BM). Then pāspardhaḥ, pāspardhati, and pāspartsi (2nd sg.).
• loṭ 2nd sg.: hi → dhi (SK 2425), giving pāspardhi.
• laṅ: apāspart / apāspard, and sip apāspāḥ (ru by daś ca, ro ri).
• gādh → jāgādh:
• laṭ jāgāddhi, jāghātsi;
• laṅ ajāghāt, sip ajāghāḥ (ru option).
• nāth → nānāth: nānātti, nānāttaḥ.
• dadh → dādadh:
• laṭ dādaddhi, dādaddhaḥ, dādhatsi;
• laṅ adādhat, adādaddhām, adādadhuḥ; 2nd sg. adādhaḥ / adādhat;
• luṅ adādādhīt / adādadhīt.
• skund → coskund (k kept and palatalised; guṇa):
• laṭ coskundīti / coskunti;
• laṅ acoskun, acoskuntām, acoskunduḥ.
• mud → momud:
• laṭ momudīti / momotti. Without īṭ no ac follows, so SK 2503 does not bar guṇa (BM).
• liṭ momodāṃcakāra.
• luṭ momoditā. na dhātulopa… does not bar the guṇa, because this luk is unconditioned.
• laṅ amomudīt / amomot, amomuttām, amomuduḥ; 2nd sg. amomudīḥ / amomoḥ / amomot.
• luṅ amomodīt, with guṇa conditioned by sic.
• kurd → cokūrd:
• laṭ cokūrti / cokūrdīti;
• laṅ acokūrt / acokūrdīt, sip acokūḥ (ru option);
• likewise acokhūḥ (khurd) and ajogūḥ (gurd).
• vañc → vanīvañc (nīk, SK 2642):
• laṭ vanīvañcīti / vanīvaṅkti. tip is pit, not ṅit, so there is no n-loss.
• vanīvaktaḥ: tas is ṅit, so n is lost by aniditām (SK 415).
• vanīvacati.
• laṅ avanīvañcīt / avanīvan.
• gam → jaṅgam (ga → ja; nuk, SK 2643):
• laṭ jaṅgamīti / jaṅganti;
• jaṅgataḥ: the nasal is deleted before a ṅit (SK 2428);
• jaṅgmati: the penultimate is deleted (SK 2363);
• jaṅganmi, jaṅganvaḥ: m → n by mvoś ca (SK 2309);
• luṭ jaṅgamitā: the iṭ-bar of SK 2246 is for ekāc roots only;
• loṭ jaṅgahi: the nasal deletion is ābhīya and asiddha, so ato heḥ (SK 2202) does not delete hi;
• laṅ ajaṅgan (mo no dhātoḥ, SK 341);
• luṅ ajaṅgamīt, ajaṅgamiṣṭām. There is no aṅ, since SK 2343's ḷdit is an anubandha citation, and no vṛddhi, by hmyanta… (SK 2299).
• han → jaṅghan. The abhyāsa's h is velarised by abhyāsāc ca (SK 2430), which carries hanteḥ, a śtip citation, from SK 358. It applies all the same, because śtipā śapā is anitya (Nyāsa; jñāpaka SK 2630).
• laṭ jaṅghanīti / jaṅghanti, jaṅghataḥ, jaṅghnati;
• luṭ jaṅghanitā;
• loṭ jaṅghahi: hanter jaḥ (SK 2431) is śtip-cited, so there is no ja;
• laṅ ajaṅghanīt / ajaṅghan; vidhiliṅ jaṅghanyāt.
• āśīrliṅ vadhyāt, luṅ avadhīt, avadhiṣṭām. These take vadha (SK 2433–2434). vadha replaces the already-doubled stem and is sthānivat, so 'anabhyāsasya' (SK 2177) stops it being doubled again.
• With āṅ, ātmanepada by SK 2695: ājaṅghate.
• car → cañcur (nuk, SK 2636; a → u by ut parasyātaḥ, SK 2637). That u is tapara, so it takes no guṇa. But hali ca (SK 354), being asiddha, still lengthens it before a consonant.
• laṭ cañcurīti / cañcūrti, cañcūrtaḥ, cañcurati;
• laṅ acañcūrīt (print; see "open"), acañcūḥ.
• khan → caṅkhan (nuk):
• laṭ caṅkhanīti / caṅkhanti;
• caṅkhātaḥ: ā by janasanakhanāṃ sañjhaloḥ (SK 2504);
• caṅkhnati (SK 2363);
• loṭ caṅkhāhi, caṅkhanāni;
• laṅ acaṅkhanīt / acaṅkhan, acaṅkhātām, acaṅkhnuḥ;
• liṅ caṅkhāyāt / caṅkhanyāt (ye vibhāṣā, SK 2319);
• luṅ acaṅkhanīt / acaṅkhānīt.
• yu → yoyu, nu → nonu. nābhyastasya runs into SK 2443, so the u takes no vṛddhi.
• yu: laṭ yoyoti / yoyavīti; laṅ ayoyavīt / ayoyot; vidhiliṅ yoyuyāt; āśīrliṅ yoyūyāt (lengthened, SK 2298); luṅ ayoyāvīt.
• nu: nonavīti / nonoti.
• hā → jāhā. The forms of hāṅ and hāk are the same, since there is no ṅit ātmanepada (TB). Five śtip-cited rules do not apply: SK 2498, 2499, 2500, 2462 and 2374.
• laṭ: jāheti / jāhāti; jāhītaḥ (ī halyaghoḥ, SK 2497); jāhati; jāhāsi / jāheṣi; jāhīthaḥ, jāhītha.
• jāhīti (print; see "open").
• laṅ ajāhet / ajāhāt, ajāhītām, [ajāhuḥ].
• [vidhiliṅ jāhīyāt; āśīrliṅ jāhāyāt; luṅ ajāhāsīt], ajāhāsiṣṭām. The bracketed forms are restored from the print's eye-skip.
• lṛṅ ajāhiṣyat.
• svap → sāsvap. The yaṅ is elided by luk, so the u of SK 2645 does not come. rudādibhyaḥ (SK 2474) is gaṇa-cited, so there is no iṭ.
• laṭ sāsvapīti / sāsvapti, sāsvaptaḥ, sāsvapati;
• laṅ asāsvapīt / asāsvap; vidhiliṅ sāsvapyāt;
• āśīrliṅ sāsupyāt, with u by vacisvapi… (SK 2409);
• luṅ asāsvāpīt / asāsvapīt.

C. ruk, rik and rīk for the abhyāsa of ṛ-penultimate and ṛ-final roots — running example: vṛt → (doubling; abhyāsa ṛ → a by urat, r, halādiḥ śeṣaḥ) va-vṛt → + ruk / rik / rīk → var-vṛt / vari-vṛt / varī-vṛt → (no īṭ, guṇa) varvarti / varivarti / varīvarti

• 2652 rugrikau ca luki 7.4.91 — in yaṅluk, the abhyāsa of a root with penultimate ṛ takes the augments ruk, rik and, by the ca, rīk. ṛdupadhasya and rīk are carried from SK 2644, abhyāsasya from SK 2620, and yaṅluk from SK 2630 (BM, TB).
• The u of ruk is only for pronunciation. The i of rik and the ī of rīk are real.
• The augments sit at the end of the abhyāsa: var-, vari-, varī- (translator's note). The kaumudī gives no form here, and its forms come under 2653.
• 2653 ṛtaś ca 7.4.92 — the same augments come to the abhyāsa of a root (TB: an aṅga) ending in short ṛ. "(3)" marks the three forms.
• vṛt (penultimate ṛ, under 2652):
• laṭ: varvṛtīti (3), where īṭ bars guṇa; varvarti (3); varvṛtaḥ (3); varvṛtati (3).
• liṭ varvartāmāsa (3); luṭ varvartitā.
• lṛṭ varvartiṣyati (3). The iṭ-bar of na vṛdbhyaś caturbhyaḥ (SK 2348) is gaṇa-cited and does not apply.
• laṅ: avarvṛtīt (3); avarvart (3), where rāt sasya blocks saṃyogānta deletion (BM); sip avarvāḥ (3), with ru by daś ca and ro ri.
• luṅ avarvartīt. The dyutādi aṅ of SK 2343 is gaṇa-cited, so there is no aṅ.
• kṛ (final ṛ) → car-kṛ (3):
• laṭ carkarīti (3); carkarti / carikarti / carīkarti; carkṛtaḥ (3); carkrati (3);
• liṭ carkarāṃcakāra; luṭ carkaritā;
• laṅ acarkarīt, acarkaḥ;
• vidhiliṅ carkṛyāt: sārvadhātuka, so no lengthening;
• āśīrliṅ carkriyāt: riṅ by SK 2367;
• luṅ acarkārīt: vṛddhi by sici vṛddhiḥ.
• The ṛ is tapara, so long ṝ gets no augment:
• kṝ: cākarti;
• tṝ: tātarti, tātīrtaḥ, tātirati; loṭ tātīrhi, tātarāṇi; laṅ atātarīt, atātaḥ, atātīrtām, atātaruḥ; luṅ atātārīt, atātāriṣṭām.
• ṛ (arti). Doubling gives ṛ-ṛ. The abhyāsa ṛ becomes a by urat (SK 2244), takes r, and halādiḥ śeṣaḥ (SK 2179) leaves a-ṛ. Then:
• with ruk: ar-ṛ;
• with rik or rīk: ari-ṛ / arī-ṛ, then iyaṅ by SK 2290 gives ariy-ṛ. rik and rīk thus give the same forms.
• Forms:
• laṭ ararti / ariyarti; with īṭ ararīti / ariyarīti; arṛtaḥ / ariyṛtaḥ.
• 3rd pl.: jhi → at, and yaṇ gives ar-r-ati. ro ri (SK 173) deletes ruk's r. The yaṇ is not sthānivat here, being barred for a pūrvatrāsiddha operation. ḍhralope lengthens: ārati; with rik / rīk, ariyrati.
• āśīrliṅ: guṇo 'rti… (SK 2380) is śtip-cited and does not apply. riṅ, r-deletion and lengthening give āriyāt; with rik / rīk, ariyriyāt.
• luṅ arārīt (print; see "open"), āriyārīt.
• gṛh (penultimate ṛ) → jar-gṛh (3):
• laṭ jargṛhīti, jargarḍhi, jargṛḍhaḥ, jargṛhati;
• laṅ ajargharṭ.
• Contrast grah (gṛhṇāti) → jāgrah. Its ṛ appears only through the ṅit-conditioned samprasāraṇa (SK 2412), which is bahiraṅga and asiddha, so it takes no ruk.
• laṭ jāgrahīti, jāgrāḍhi; jāgṛḍhaḥ, jāgṛhati; jāgrahīṣi, jāghrakṣi.
• luṭ jāgrahitā. There is no lengthening by graho 'liṭi dīrghaḥ (SK 2562), since ekāc is carried there. Mādhava's lengthening is "contrary to the Bhāṣya".
• gṛdh → jar-gṛdh (3):
• laṭ jargṛdhīti, jargarddhi, jargṛddhaḥ, jargṛdhati, jargṛdhīṣi, jarghartsi;
• laṅ with īṭ ajargṛdhīt;
• without īṭ: ajargṛdh-t, guṇa, hal-ṅyādi deletion, bhaṣ (g → gh), jaś / car: ajarghart;
• ajargṛddhām; sip ajarghāḥ (ru by daś ca);
• luṅ ajargardhīt, ajargardhiṣṭām.
• pracch → pāpracch (not a ṛ-root; included in the same kaumudī):
• pāpracchīti, and pāpraṣṭi (ṣ by SK 294).
• Before tas, grahijyā… is śtip-cited, so there is no samprasāraṇa. ś by cchvoḥ (SK 2561) and ṣ by SK 294 give pāpraṣṭaḥ.
• pāpracchati; pāpraśmi, pāpracchvaḥ, pāpraśmaḥ.
• y/v-final roots. Those that take ūṭh by cchvoḥ… (SK 2561), the ūṭh-bhāvins, have no yaṅluk: the Bhāṣya hints it and Kaiyaṭa states it. srivi and mav, whose ūṭh comes from 2654, do have yaṅluk (Mādhava and others agree). mavya, tevṛ and devṛ are given as y- and v-final ūṭh-bhāvins (BM: none of them has yaṅluk).
• haya → jāhay, which is y-final but not an ūṭh-bhāvin:
• laṭ jāhayīti;
• jāhati, jāhataḥ: y is deleted (lopo vyor vali);
• jāhayati; jāhayīṣi / jāhasi;
• jāhāmi, jāhāvaḥ, jāhāmaḥ: y is deleted before val, and the a is lengthened before a yañ-initial ending (ato dīrgho yañi).
• harya → jāhary:
• laṭ jāharyīti / jāharti, jāhartaḥ, jāharyati;
• loṭ jāharhi;
• laṅ ajāhaḥ, ajāhartām, ajāharyuḥ.
• The closing "mava bandhane" introduces the root of 2654.

D. ūṭh for the penultimate and the v of jvara, tvara, srivi, avi, mava — running example: mav: māmav-ti → (ūṭh for a + v together) māmū-ti → (tip is pit, so guṇa) māmoti

• 2654 jvaratvarasrivyavimavām upadhāyāś ca 6.4.20 — ūṭh replaces the penultimate and, by the ca, the v of jvara, tvara, srivi, avi and mava. It applies before kvi, a jhal-initial suffix, or a nasal-initial suffix. ūṭh, v and anunāsike are carried from SK 2561, and kvijhaloḥ from SK 2666 (BM).
• kṅiti is not carried: av + tun gives otuḥ, whose guṇa needs a non-kṅit suffix. TB says the vṛtti's kṅiti is a mistake.
• anunāsike is carried: av + man, with its ṭi deleted, gives om.
• mav → māmav (no īṭ; ūṭh, then guṇa before a pit ending):
• laṭ:
• māmoti; māmavīti (with īṭ, ī follows, so no ūṭh);
• māmūtaḥ (ṅit, so no guṇa); māmavati;
• māmoṣi, māmomi;
• māmāvaḥ: v is not jhal, so no ūṭh. The v is deleted by lopo vyoḥ and the a is lengthened by ato dīrgho yañi (BM).
• māmūmaḥ.
• loṭ māmūhi; māmavāni (āṭ is pit, BM).
• laṅ amāmot, amāmoḥ, amāmavam, amāmāva, amāmūma.
• The tail "turvī … totūrvīti" (ū by upadhāyāṃ ca, BM) leads into 2655.

E. v / ch deleted after r, and guṇa not barred before tip: turvī, murchā; the close — running example: totūrv-ti → (v deleted after r, 2655) totur-ti → (laghūpadha guṇa: tip is not ārdhadhātuka, so 2656 does not bar it) totorti

• 2655 rāl lopaḥ 6.4.21 — ch and v are deleted after r before kvi, a jhal-initial suffix, or a nasal-initial suffix. cchvoḥ and anunāsike are carried from SK 2561 (not ūṭh), and kvijhaloḥ from SK 2666. kṅiti is not carried here either (BM).
• In totūrv-ti the v is deleted, and the light penultimate takes guṇa: totorti.
• Before tip, lopo vyor vali would delete the v anyway. The rule is needed in totūrtaḥ, where ūṭh by cchvoḥ would otherwise set lopo vyoḥ aside (TB, BM).
• TB's kvi examples: tūḥ, dhūḥ, mūḥ.
• 2656 na dhātulopa ārdhadhātuke 1.1.4 — guṇa and vṛddhi in place of an ik (iko guṇavṛddhī, SK 34) do not come before an ārdhadhātuka that causes the deletion of part of a dhātu.
• The commentators analyse the compound differently: TB as a bahuvrīhi, BM as dhātuṃ lopayatīti with karmaṇy aṇ. The English stays neutral.
• The sūtra's own domain is the conditioned yaṅluk before ac: loluvaḥ, popuvaḥ, marīmṛjaḥ (TB). The Bhāṣya rejects the sūtra (TB, BM).
• Here it does not apply, because tip is sārvadhātuka: totorti.
• The forms that follow:
• totūrtaḥ (v deleted by 2655; lengthened by hali ca, SK 354); totūrvati;
• thurvī tothorti; durvī dodorti; dhurvī dodhorti;
• murchā: momūrcchīti (print momūrchīti); momorti, where ch is deleted by 2655; momūrtaḥ; momūrchati.
• viṣayasaptamī. In ajer vy aghañapoḥ (SK 2292), the carried word ārdhadhātuke is a locative of domain. So vī replaces aj as soon as yaṅ is intended: vevīyate. aj therefore has no yaṅluk: a yaṅ removed by luk leaves no domain, and vī does not come.
• Close: "[iti yaṅlugantaprakriyā]", supplied from the print. TB closes alike.