पूर्वकृदन्तप्रकरणम्

Covers SK 2895–3167 (273 sūtras): all present (written in four parts, each coverage-checked by script: 2895–2960, 2961–3027, 3028–3092, 3093–3167; merged coverage re-checked).

Part I covers SK 2895–2960 (66 sūtras): all present (coverage-checked by script; one bulleted line per sūtra). The spine is an Aṣṭādhyāyī run. It covers 3.1.133–3.1.150, the kṛt affixes that take no upapada and close 3.1, and then 3.2.1–3.2.43, the affixes that come with an upapada. One rule is missing from that run: 3.2.27 chandasi vanasanarakṣimathām is Vedic and stands at SK 3408. The SK inserts six rules from books 6 and 8 at the point where the forms first need them:

• 8.3.92 and 8.3.97 (2917–2918: ṣ in -ṣṭha) come after supi sthaḥ;
• 6.3.67 and then 6.3.66 (2942–2943: mum and the short vowel before a khit uttarapada) come after ejeḥ khaś. The SK takes mum first, against AS order;
• 6.4.94 khaci hrasvaḥ (2955) comes after dviṣatparayos tāpeḥ;
• 6.3.69 vācaṃyamapuraṃdarau ca (2957) comes after vāci yamo vrate.

Every AS-ref below is checked against references/ashtadhyayi/sutraani-data.json. All 66 match our source.json, and every SK number matches the reference's own. Two sūtra texts differ from the reference, and in both the reference agrees with the print or BM: 2903 (ours saṃsvratīṇa; reference and BM's pratīka saṃsrvatīṇa) and 2935 (ours lacks ahar; reference and print …bāhvahar…).

pūrvakṛdantaprakaraṇam — outline (p67, SK 2895–3167, all 273 sūtras)

p66 (kṛdantakṛtyaprakaraṇam) closed at 2894 with bhavyageya… (3.4.68), followed by the colophon "iti kṛtyaprakaraṇam". The colophon is in the print, missing from ours, and supplied in brackets.

• Where the prakaraṇa begins. The print heads it "atha kṛdantaprakaraṇam 64" in kaumudī type above SK 2895 (NSP p. 487, scan page-0497). Ours lacks the heading, and the English supplies it in brackets in sk.2895.k1. BM's own opening reads "atha kṛdantaprakaraṇaṃ". Our p67/intro is misplaced Uṇādi boilerplate, not this heading; ignore it. The name pūrvakṛdantaprakaraṇam and the number 67 are ours. The print closes the prakaraṇa after 3167 with "iti pūrvakṛdantaṃ samāptam", which ours also lacks (Part IV).
• TB on the new prakaraṇa does not start in its own block. It opens at the tail of sk.2893.t1, after TB's kṛtya colophon ("ṇvultṛcau. ṇakāro vṛddhyarthaḥ…"), and sk.2895.t1 picks it up mid-stream.
• The first sūtra is ṇvultṛcau (3.1.133). The kṛtya adhikāra of p66 (kṛtyāḥ, SK 2831) ran "up to before ṇvul". From here on the affixes are still kṛt but no longer kṛtya, so tayor eva… (SK 2833) stops confining them to bhāva and karman.
• Where Part I ends. At 2960 meghartibhayeṣu kṛñaḥ (3.2.43), the last khac rule taught with kṛ. The khac run continues into Part II: 2961 kṣemapriyamadre 'ṇ ca (aṇ too, besides khac), 2962 āśite bhuvaḥ…, 2963 saṃjñāyāṃ bhṛtṝvṛji… and 2964 gamaś ca, which Part I's commentators already cite. ḍa begins at 2965.
• Contents of Part I.
• 2895: ṇvul and tṛc, the general agent affixes. The kaumudī runs the aṅga rules of the verb through them: kārakaḥ, kartā … pādahārakaḥ, prakrantā.
• 2896: lyu, ṇini, ac for the nandi-, grahi- and paca-groups: nandanaḥ, grāhī, pacaḥ; the reduplicated carācaraḥ type.
• 2897–2898: ka after igupadha roots and jñā, prī, kṝ, and after ā-finals with an upasarga: kṣipaḥ, prajñaḥ.
• 2899–2901: śa, which brings in the sārvadhātuka stem: pibaḥ, jighraḥ, paśyaḥ; limpaḥ, govindaḥ; dadaḥ.
• 2902–2906: ṇa, optional, invariable or vyavasthita, and ka for the house: jvālaḥ/jvalaḥ, dāyaḥ, dāvaḥ, grāhaḥ/grahaḥ, gṛham.
• 2907–2912: agents with a stated quality (craft, rice/time, repetition, blessing): ṣvun, thakan, ṇyuṭ, vun; nartakaḥ, gāthakaḥ, gāyanaḥ, hāyanaḥ, lavakaḥ, jīvakaḥ.
• 2913–2914: aṇ with an object upapada and the upapada compound: kumbhakāraḥ; ṇa by vārttika (māṃsaśīlā); aṇ against ka (tantuvāyaḥ).
• 2915–2921: ka (and ḍa) with an upapada, and the ṣ of -ṣṭha: godaḥ, brahmajyaḥ, dvipaḥ, ākhūtthaḥ, praṣṭhaḥ, dviṣṭhaḥ, tundaparimṛjaḥ, sarvapradaḥ, gosaṃkhyaḥ.
• 2922–2929: ṭak, and ac after hṛ, grah, arh, ram/jap, śam, śī: sāmagaḥ, aṃśaharaḥ, pūjārhā, stamberamaḥ, śaṃkarā, khaśayaḥ.
• 2930–2933: ṭa after car and sṛ: kurucaraḥ, bhikṣācaraḥ, agresaraḥ, pūrvasaraḥ.
• 2934–2937: ṭa after kṛ: yaśaskarī, divākaraḥ, kiṃkarā (ac), karmakaraḥ; none in śabdakāraḥ.
• 2938–2940: in: stambakariḥ, dṛtihariḥ, phalegrahiḥ, ātmambhariḥ.
• 2941–2952: khaś, with the short vowel and mum before a khit uttarapada: janamejayaḥ, stanandhayaḥ, vahaṃlihaḥ, aruntudaḥ, asūryampaśyāḥ.
• 2953–2960: khac, with khaci hrasvaḥ: priyaṃvadaḥ, vihaṅgaḥ, dviṣantapaḥ, vācaṃyamaḥ, puraṃdaraḥ, sarvaṅkaṣaḥ, abhayaṅkaraḥ.

Threads that run through Part I:

• The kṛt frame, inherited from p66. dhātoḥ (3.1.91, SK 2829) is the heading to the end of adhyāya 3. Under it, kṛd atiṅ (3.1.93, SK 374) makes every affix here a kṛt, and tatropapadaṃ saptamīstham (3.1.92, SK 781) defines the upapada. Now that the kṛtya restriction (SK 2833) has lapsed, kartari kṛt (3.4.67, SK 2832) holds in full. The kaumudī of 2895 opens "these come … in the sense of the agent, by kartari kṛt", so every formation in Part I is agentive unless something says otherwise:
• bahula. Bhaṭṭoji reads SK 2841 as kṛtyalyuṭaḥ alone and splits off bahulam to let any kṛt vary its sense (2895): pādahārakaḥ "one carried by the feet", ṇvul in the karman. TB: the source (ākara) has only kṛto bahulam there.
• ārambhasāmarthya. The part sthaḥ of 2916 would be pointless in the agent sense, so ka comes in bhāva: ākhūtthaḥ.
• What the affix does not denote. A blessing is the speaker's property, not the affix's meaning, "since kṛt affixes are prescribed in the agent sense" (TB 2912). hetu, tācchīlya and ānulomya are dyotya, not vācya, since being denoted would override kartari kṛt (TB 2934).
• Qualifiers of the agent, not upapadas. gehe (2906) and śilpini (2907) qualify the agent (TB). So do vrīhi/kāla (2910), vrīhi/vatsa (2938) and paśu (2939), which name the agent, and the vārttika uttānādiṣu kartṛṣu (2929).
• Upapada and the upapada compound. In a rule of this adhikāra, a seventh-case word (karmaṇi, supi, adhikaraṇe, the listed words) names an upapada (SK 781). upapadam atiṅ (2.2.19, SK 782) then compounds it with the kṛdanta (2913: kumbhaṃ karotīti kumbhakāraḥ).
• TB: kumbha, in the sixth case by kartṛkarmaṇoḥ kṛti (SK 623), is compounded with kāra. BM: kumbhaṃ karoti is the worldly analysis; the technical one is asvapada, made before the sup arises.
• anabhidhāna blocks ādityadarśa (2913).
• The point of a separate upapada affix is often the obligatory compound: sahacara with the pacādi ac is optional (supsupā), bhikṣācara with ṭa is nitya (BM, TB 2931).
• Not every seventh-case word is an upapada: gehe, śilpini (TB), and udi in 2946, which is stated separately to block upapada-hood (TB).
• karmaṇi and supi. The kaumudī at 2918 says that from here on both karmaṇi (2913) and supi (2916) continue, and that with intransitive roots only supi is construed. This governs every 3.2 rule in Part I that names its upapada bare: ram and jap at 2927 (BM: supi only), the 27 words of 2935 (BM: each construed as fits).
• Rules that name another relation: adhikaraṇe (2929, carried into 2930 per BM); kartari (2933, the word pūrva denoting the agent); the word karman (2936, svarūpa, because karmaṇi is already present: BM, TB).
• Following audits C and E, where BM only paraphrases ("…karmaṇy upapade … ity arthaḥ"), the outline says "supplied", not "carried forward per BM".
• The utsarga–apavāda ladder.
• Without an upapada (3.1.133–150). ṇvul/tṛc (2895) is the general rule. Its exceptions are lyu/ṇini/ac by gaṇa (2896), ka (2897–2898), śa (2899–2901) and ṇa (2902–2905), with ka for the house as an exception to ṇa (2906). The qualified agents take ṣvun, thakan, ṇyuṭ, vun (2907–2912).
• With an object upapada (3.2). aṇ (2913) is the general rule; ka for ā-finals (2915) excepts it, and 2914 restores aṇ for hvā/vā/mā "as an exception to ka". The narrower affixes follow: ḍa (vārttika, 2915), ṭak (2922), ac (2923–2929), ṭa (2930–2937), in (2938–2940), khaś (2941–2952), khac (2953–2960).
• vāsarūpa (vā 'sarūpo 'striyām, SK 2830) lets the general affix stand beside a dissimilar exception: kṣepakaḥ, kṣeptā beside kṣipaḥ (2897); kvip beside śa in sāt (2900). Similarly śvapākaḥ takes aṇ "in the other option" beside the pacādi ac (TB 2896).
• Counterexamples fall back to aṇ, with ṇ-vṛddhi visible: tundaparimārjaḥ, sūtragrāhaḥ, bhārahāraḥ, pūrvasāraḥ, karmakāraḥ, śabdakāraḥ, stambakāraḥ, dṛtihāraḥ, vāgyāmaḥ, dviṣatītāpaḥ. In most of these the kaumudī names no affix; the aṇ is the translators' and commentators' identification.
• The feminine decides several apavādas. Some exceptions give the same masculine as aṇ and exist only for the feminine: aṇ (ṇit) would give ṅīp by ṭiḍḍhāṇañ… (SK 470), and the exception gives ṭāp. So ṇa for śīli etc. (māṃsaśīlā, 2913), ac after arh (pūjārhā, 2926), ac with kim etc. (kiṃkarā, 2935); and ac against ṭa in śaṃkarā (2928; the ṅīp/ṭāp reasoning there is the translator's).
• Conversely, ṭ-marked affixes give ṅīp: nadī (nadaṭ, 2896), gāyanī (ṇyuṭ), sāmagī (ṭak), kurucarī (ṭa), yaśaskarī (ṭa), sahacarī (caraṭ). ṣvun gives ṅīṣ: nartakī, rajakī. kiṃkarī is puṃyoga ṅīṣ.
• dheṭ's ṭ: Haradatta allows ṅīp from the khaś-form only (stanandhayī, but dhayā kanyā). BM rejects the restriction (2899); TB reports Vardhamāna, Kṣīrasvāmin and Haradatta for it (2944).
• What the anubandhas do in Part I.
• ṇ (ṇvul, ṇini, ṇa, ṇyuṭ, aṇ) brings vṛddhi: kārakaḥ, grāhī, jvālaḥ, kumbhakāraḥ.
• The l of ṇvul serves the accent (liti). The c of tṛc keeps tṛc apart where only tṛ is named (TB 2895).
• vu → aka, yu → ana by yuvor anākau (7.1.1, SK 1247): kāraka, nartaka, nandana, gāyana.
• k (ka, ṭak) blocks guṇa and triggers ā-loss by āto lopa iṭi ca (6.4.64, SK 2372): kṣipaḥ, godaḥ, sāmagaḥ. It would also trigger saṃprasāraṇa, which is why ḍa replaces ka for jyā and hve (2915).
• ś (śa, khaś) makes the affix sārvadhātuka. That brings śap and the śit substitutes of pāghrā… (7.3.78, SK 2360): pibaḥ, jighraḥ, dhamaḥ, paśyaḥ. It also brings the ṅidvat no-guṇa (vahaṃlihaḥ, 2947). It is TB's proof that ejeḥ is the ṇic stem (2941).
• ḍ (ḍa; ḍit khac) gives ṭi-loss although the stem is not bha: brahmajyaḥ, giriśaḥ, vihaṅgaḥ.
• kh (khaś, khac) gives the short vowel (khity anavyayasya) and then mum (arurdviṣadajantasya mum) on the pūrvapada: śunindhayaḥ, priyaṃvadaḥ. The c of khac specifies it for khaci hrasvaḥ (Vṛtti, reported by TB). Being ārdhadhātuka, khac also allows ṇi-loss (BM 2955).
• The n of in gives an initial-udātta uttarapada (TB 2938).
• Nipātana does much of the work: lavaṇa, viśayī, viṣayī, paribhavī (2896); praṣṭha (2917); the e of agre (2932); phalegrahi, ātmambhari (2940); ugraṃpaśya, iraṃmada, pāṇiṃdhama (2952); vācaṃyama, puraṃdara (2957). ākṛtigaṇa lists: pacādi (2896), mūlavibhujādi (2919); bhvādi itself, for vadha (BM 2895).
• The kaumudī cites by pratīka and lists forms bare. At 2895 most rules are named by opening words only. Derivation stages below (jana-ejaya, śuni-dhaya, dviṣat-tapi …) are BM's or TB's as recorded in the unit notes, or marked as the translator's.
• Where the commentary lies (only the kaumudī is translated, SK-TRANSLATE §25):
• TB printed under a neighbouring unit:
• 2895: tail of sk.2893.t1, then sk.2895.t1.
• 2903–2905: tail of sk.2902.t1.
• 2911, 2912 and the first half of 2913: sk.2902.n1 (coherent there; our sk.2910.t1 has a garbled splice of it).
• 2915's k1 part: tail of sk.2914.t1.
• 2919: tail of sk.2918.t1, then sk.2919.t1, which also carries 2920–2921.
• 2922–2924: sk.2922.t1.
• 2928: opening lemma in sk.2928.t1; the rest, with 2929–2931 and the start of 2934, in sk.2919.n1. 2934 continues as the latter part of sk.2928.t1.
• 2938 and 2940: in sk.2936.t1.
• 2943 and 2944: in sk.2942.t1.
• 2950–2951: tail of sk.2946.t1.
• The 2955 example: tail of sk.2953.t1.
• 2957–2958: sk.2956.t1.
• No TB in print on 2898, 2917, 2925, 2926, 2932, 2933, 2937, 2939, 2947–2949, 2952, 2954, 2959, 2960.
• BM filed elsewhere or missing:
• 2916: second half of sk.2915.b1.
• 2923: inside sk.2922.b1, pratīka lost.
• 2929 and 2934: missing from our BM layer (sk.2929.b1 = BM on 2939; sk.2934.b1 = BM on 2945). Both are read from Wikisource BM.
• 2933: tail of sk.2932.b1. 2936: tail of sk.2935.b1. 2943: opens at the tail of sk.2942.b1. 2954: tail of sk.2953.b1.
• 2955: split between sk.2955.b1 and sk.2954.b1.
• 2960: inside sk.2958.b1, which also has an eye-skip.
• 2959 has no BM anywhere.
• Commentator disagreements, kept neutral in the English:
• 2895: Bhaṭṭoji's yogavibhāga of bahulam, against TB's "only kṛto bahulam in the source";
• 2899/2944: ṅīp from dheṭ;
• 2911: prusṛlvaḥ is a samāhāra pañcamī (BM) or jas by vyatyaya (TB);
• 2918: dviṣṭhaḥ is dvābhyām (BM) or dvayoḥ (TB) tiṣṭhati;
• 2919: vṛddhi in tundaparimṛja (BM one view, TB two);
• 2931: senācaraḥ, "causes the army to go" (BM) or "enters the army" (TB);
• 2935: divākaraḥ;
• 2953: scope of the ḍit khac (BM: viha only; TB: elsewhere too).

Audit corrections the outline follows. Audits B–F cover SK 2908–2972 and found no MUST-FIX. No audit report for 2895–2907 was available when this part was written. Their MINOR items touch only notes and brackets; the outline takes them as follows:

• 2911–2913: TB is not missing. It is at sk.2902.n1 (cite as "TB, printed under sk.2902.n1").
• 2916, 2921, 2924, 2925, 2929, 2931, 2950, 2951, 2953, 2954, 2958: the "anuvṛtti … per BM" labels overstate BM, which only paraphrases. The outline says "supplied".
• 2921: the affix ka is the translator's identification, from the run 2915–2920. No witness names it.
• 2921 and 2916: TB's "Manoramā" is the Prauḍhamanoramā, not BM.
• 2934: TB takes karoti "though it has two anubandhas" (dvyanubandho 'pi), because it is better known; not kṛñ hiṃsāyām.
• 2935: TB rejects the prakriyā-work's optional ac on two grounds: it is against the sources, and it would let aṇ in too. On the Vṛtti's view, kiṃkarī follows even without puṃyoga.
• 2945: khaś goes back to ejeḥ khaś (2941). Only dhmā–dheṭ come from 2944.
• 2953: TB's jñāpana is the answer to an objection ("then put khac in the next sūtra" — "true, but it stands here to show it comes after other roots too").
• 2959 (and 2960): khac is supplied by the translator from the run since 2953. The mum does not prove khac over khaś, since khaś is khit too.
• 2925: Wikisource reads the counterexample as bhārāhāraḥ. 2929: "loke giriśayaḥ" is not in the print, though BM glosses it.
• 2952: the extra "na" in ours reverses the sense and is not in the collation report.

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

• 2895: the heading "atha kṛdantaprakaraṇam".
• 2896: the two missing passages (the three purposes of the pacādi list; jārabharā, śvapacā), and yaṅluk, vṛddhyabhāvo, śvapāko, pāṭūpaṭaḥ, rātricaraḥ.
• 2902 avatānaḥ; 2903 ādantatvāt; 2904 sānubandhakāt; 2907 "tadvati kartari ṣvun".
• 2929: no "loke giriśayaḥ". 2932 agratassaraḥ. 2935 [ahar] in the sūtra.
• 2952 "nipātanāt śyan na". 2959 karīṣaṅkaṣā.

Still open for Ajai (mūla):

• the missing p67 heading (2895);
• the 2896 omissions;
• 2903 saṃsvratīṇa / saṃsrvatīṇa (scan unreadable, class C);
• 2907 kartari;
• 2918 savyāpa (WS savye 'pa) and kuśekuśaṅkvaṅgu (C);
• 2925 bhārahāraḥ / bhārāhāraḥ;
• 2929 loke giriśayaḥ;
• 2935 ahar, and ṅīṣ / ṅīp (C);
• 2952 the double negative;
• 2960 ṛtiṅkaraḥ, in WS and BM but not in ours or the print.

Ordering: SK order throughout; no reordering. Groups follow the affix sequence of AS 3.1.133–3.2.43. The book-6 and book-8 rules sit in the group whose forms they finish.

The four parts. 273 sūtras are too many for one list of groups, so the outline runs in four parts, each with its own lettered groups (I.A …):

• Part I — SK 2895–2960: ṇvul/tṛc and the agent affixes of 3.1.133–150; the upapada affixes 3.2.1–43, from aṇ to khac.
• Parts II–IV — Part II SK 2961–3027 (khac … ḍa after jan; the niṣṭhā and n for its t); Part III SK 3028–3092 (niṣṭhā continued; senses of kta; ṅvanip, atṛn); Part IV SK 3093–3167 (liṭ's participles, śatṛ/śānac, the tācchīlika affixes, ṣṭran, itra).

Part I — SK 2895–2960: the agent affixes of 3.1.133–150, and the upapada affixes of 3.2.1–43 from aṇ to khac

I.A. The general agent affixes ṇvul and tṛc — running example: kṛ + ṇvul → kṛ + vu → kṛ + aka → (vṛddhi) kārakaḥ; kṛ + tṛc → (guṇa) kartṛ → kartā

• 2895 ṇvultṛcau 3.1.133 — ṇvul and tṛc after any dhātu, in the agent sense by kartari kṛt (SK 2832). vu → aka by yuvor anākau (SK 1247). Here kṛ is the dhātu: ṇit aka gives vṛddhi, kārakaḥ; tṛc gives guṇa, kartā. Feminines: kārikā, kartrī. voḍhum arhaḥ voḍhā takes tṛc in the arha sense of arhe kṛtyatṛcaś ca (SK 2822) (BM, TB).
• The kaumudī then runs aṅga rules of the verb through the two affixes, by pratīka:
• ṅit tṛc after kuṭādi (SK 2461), so no guṇa: kuṭitā. ṇvul is ñṇit and so not ṅit: koṭakaḥ. vija iṭ (SK 2536): vijitā.
• hanas to 'ciṇṇaloḥ (SK 2574): ghātakaḥ. āto yuk (SK 2761): dāyakaḥ.
• No vṛddhi after an udātta-upadeśa m-final (SK 2763): śamakaḥ, damakaḥ; but aniṭ yam has it: niyāmakaḥ. janivadhyoś ca (SK 2512): janakaḥ, vadhakaḥ. BM: vadha is a separate bhvādi root, not the substitute for han.
• num: randhakaḥ, jambhakaḥ (SK 2302); radhitā / raddhā (SK 2516); maṅktā, naṃṣṭā / naśitā (SK 2517); rambhakaḥ, rabdhā (SK 2581); lambhakaḥ, labdhā (SK 2582).
• Optional iṭ (SK 2340): eṣitā / eṣṭā, sahitā / soḍhā. daridrā loses its ā (daridritā), but not before ṇvul (daridrāyakaḥ).
• bahula: let SK 2841 be just kṛtyalyuṭaḥ, and split off bahulam (yogavibhāga) so that any kṛt may vary its sense: pādābhyāṃ hriyate pādahārakaḥ, ṇvul in the karman.
• Vārttika: no iṭ for a kṛt in the agent sense after kram in the domain of ātmanepada: prakrantā. Not agent: prakramitavyam; not ātmanepada: saṃkramitā.
• viṣaya means "invariably" (ananyabhāva). So kram, which takes ātmanepada only optionally by anupasargād vā (SK 2716), keeps iṭ: kramitā. "Some" also allow krantā; BM sees dissatisfaction in this view.
• In gamer iṭ… (SK 2401) and na vṛdbhyaś caturbhyaḥ (SK 2348), "parasmaipada" indicates "no taṅ or āna": saṃjigamiṣitā, vivṛtsitā.
• ṇvul after a yaṅ-final: the deleted a is sthānivat, so no vṛddhi: pāpacakaḥ. After yaṅluk nothing stands in the way: pāpācakaḥ (BM, TB).

I.B. lyu, ṇini and ac for the three gaṇas — running example: nandi + lyu → (yu → ana) nandanaḥ; grah + ṇini → (vṛddhi) grāhī; pac + ac → pacaḥ

• 2896 nandigrahipacādibhyo lyuṇinyacaḥ 3.1.134 — lyu after the nandi-group, ṇini after the grahi-group, ac after the paca-group, in that order. "ādi" goes with each; nandi is the ṇi-final (BM, TB).
• lyu: nandanaḥ, janārdanaḥ, madhusūdanaḥ, vibhīṣaṇaḥ. With an object word, aṇ would apply; the gaṇa listing gives lyu (TB). lavaṇaḥ has ṇ by nipātana.
• ṇini: grāhī, sthāyī, mantrī. viśayī lacks vṛddhi by nipātana, viṣayī has its ṣ by nipātana too, and paribhāvī / paribhavī has optional vṛddhi.
• ac: pacaḥ. The paca-group is an ākṛtigaṇa. The evidence is ac after kṛ and ghaṭ in SK 3489 and SK 1836, and yaṅluk being taught before ac.
• Its listings serve three ends: attaching anubandhas (nadaṭ → nadī, coraṭ, devaṭ); fullness of statement; and blocking a blocking rule. deva blocks the ka of 2897. jārabharā, śvapacā block the aṇ of 2913; this sentence is in the print but missing from ours. śvapākaḥ (nyaṅkvādi) is aṇ in the other option (TB).
• ac after a yaṅ-final: yaṅ is dropped (yaṅo 'ci ca, SK 2650), and guṇa and vṛddhi are barred (SK 2656): cekriyaḥ, nenyaḥ, loluvaḥ, popuvaḥ, marīmṛjaḥ.
• Vārttikas before ac:
• car, cal, pat and vad may reduplicate, with āk on the abhyāsa. Because the augment is long, there is no shortening and no halādiḥ śeṣaḥ (SK 2179): carācaraḥ, calācalaḥ, patāpataḥ, vadāvadaḥ.
• han also gets gh for the abhyāsa, with kutva in the second member by SK 2430: ghanāghanaḥ.
• pāṭi gets ṇi-luk, ūk and a long abhyāsa: pāṭūpaṭaḥ (print ū).
• Otherwise: caraḥ … hanaḥ, pāṭaḥ. Optional mum by rātreḥ kṛti vibhāṣā (SK 1008): rātriṃcaraḥ / rātricaraḥ.

I.C. ka after an ik-penultimate root, jñā, prī and kṝ, and after ā-finals with an upasarga — running example: kṣip + ka → (k: no guṇa) kṣipaḥ; pra + jñā + ka → (ā-loss) prajñaḥ

• 2897 igupadhajñāprīkiraḥ kaḥ 3.1.135 — ka after a root with an ik penultimate, and after jñā, prī and kṝ (kir = kṛ vikṣepe with i and r). The k blocks guṇa (BM): kṣipaḥ, likhaḥ, budhaḥ, kṛśaḥ. jñaḥ loses its ā; priyaḥ takes iyaṅ; kiraḥ. By vāsarūpa, ṇvul and tṛc also come: kṣepakaḥ, kṣeptā.
• 2898 ātaś copasarge 3.1.136 — ka (supplied, BM) after an ā-final root when an upasarga is upapada. It is the exception to the ṇa of 2903: su + glā (glai) + ka → suglaḥ, prajñaḥ.

I.D. śa: agents built on the sārvadhātuka stem — running example: pā + śa → (ś: sārvadhātuka; piba for pā, SK 2360; śap, pararūpa) pibaḥ

• 2899 pāghrādhmādheṭdṛśaḥ śaḥ 3.1.137 — śa after pā (pāne; not pā rakṣaṇe, which has a luk-vikaraṇa: BM, TB), ghrā, dhmā, dheṭ and dṛś. Being śit, it brings the substitutes of SK 2360: pibaḥ, jighraḥ, dhamaḥ, dhayaḥ, paśyaḥ. 'upasarge' is not carried (TB).
• dhayā kanyā: the ṭ of dheṭ would give ṅīp, as in the khaś-form stanaṃdhayī, but it "is not wanted elsewhere than with khaś", which is Haradatta's view. BM rejects that restriction.
• Vārttika: no śa after ghrā in a name. The proof is Pāṇini's own vyāghrādibhiḥ (SK 735), not vyājighra-.
• 2900 anupasargāl limpavindadhāripārivedyudejicetisātisāhibhyaś ca 3.1.138 — śa (supplied, BM) after these roots without an upasarga.
• limpaḥ, vindaḥ: the citation already has the num of śe mucādīnām (BM).
• dhārayaḥ, pārayaḥ, vedayaḥ, udejayaḥ, cetayaḥ: ṇic stems, then śa, śap, guṇa, ay (BM).
• sāti (sautra, "pleasure") and sāhi: sātayaḥ, sāhayaḥ. By vāsarūpa, kvip after sāti: sāt, the paramātman; sātvantaḥ, the devotees.
• With an upasarga, the affix is ka: pralipaḥ.
• Vārttikas: limp with ni, nilimpāḥ (gods); vind with go etc. in a name, govindaḥ, aravindam (śa against aṇ: BM).
• 2901 dadātidadhātyor vibhāṣā 3.1.139 — śa optionally after dā and dhā: dadaḥ, dadhaḥ (ślu, reduplication, ā-loss: BM, TB). In the other option, the ṇa of 2903 gives dāyaḥ, dhāyaḥ (TB). The kaumudī says "without an upasarga" continues: pradaḥ, pradhaḥ take ka by 2898.

I.E. ṇa, optional, invariable and vyavasthita, and ka for the house — running example: jval + ṇa → (vṛddhi) jvālaḥ / jval + ac → jvalaḥ; dā + ṇa → (yuk) dāyaḥ

• 2902 jvalitikasantebhyo ṇaḥ 3.1.140 — "iti" means ādi (TB). ṇa comes optionally after the roots from jval to kas: vibhāṣā is carried (BM), and ac comes in the other option. It comes only without an upasarga (kaumudī): jvālaḥ / jvalaḥ, cālaḥ / calaḥ; with an upasarga only ac: ujjvalaḥ.
• Vārttika: tan is added, and here neither restriction holds: avatānaḥ (print; ours atānaḥ).
• 2903 śyādvyadhāsrusaṃsrvatīṇavasāvahṛlihaśliṣaśvasaś ca 3.1.141 — ṇa invariably. anupasargāt ceases because 2904 restates it, and vibhāṣā goes with it (BM, TB).
• śyā and avasā are ā-final already; they are named to block the ka of 2898 with an upasarga: avaśyāyaḥ, pratiśyāyaḥ.
• Other forms: dāyaḥ, dhāyaḥ (yuk, SK 2761); vyādhaḥ, āsrāvaḥ, saṃsrāvaḥ, atyāyaḥ, avasāyaḥ, avahāraḥ, lehaḥ, śleṣaḥ, śvāsaḥ.
• 2904 dunyor anupasarge 3.1.142 — ṇa after du and nī without an upasarga. Associated with nīñ, which has an anubandha, du means the anubandha-bearing ṭudu (dunoti): dāvaḥ. du gatau (davati) takes the pacādi ac: davaḥ. Also nāyaḥ; with an upasarga, pradavaḥ, praṇayaḥ.
• 2905 vibhāṣā grahaḥ 3.1.143 — ṇa optionally after grah, as a vyavasthita option: grāhaḥ for an aquatic creature, grahaḥ for a heavenly body.
• The Kāśikā adds bhavateś ca: bhavaḥ (the god; saṃsāra), bhāvāḥ (things). On the Bhāṣya's view, bhāva takes ac from the cur-class bhū "obtain" + ṇic (TB: the addition is not in the Bhāṣya).
• 2906 gehe kaḥ 3.1.144 — ka after grah when the agent is a house. geha qualifies the agent and is not an upapada (TB). It is an exception to 2905 (BM): grah + ka → (saṃprasāraṇa, SK 2412; pūrvarūpa) gṛham. gṛhāḥ "wife" by tātsthya.

I.F. Agents with a stated quality: craft, rice and time, repetition, blessing — running example: nṛt + ṣvun → (vu → aka, guṇa) nartakaḥ → (ṣ: ṅīṣ) nartakī

• 2907 śilpini ṣvun 3.1.145 — ṣvun when the agent has a craft. The print reads "tadvati kartari"; ours lacks kartari. TB: śilpini qualifies the agent, as gehe does.
• Vārttika: only after nṛt, khan, rañj: nartakaḥ, nartakī; khanakaḥ, khanakī. veñ therefore takes ṇa: vāya (BM).
• Vārttika (Bhaṭṭoji's restatement by sense, BM): the n of rañj drops before as, aka, ana: rajakaḥ, rajakī.
• On the Bhāṣya's view, ṣvun is only for nṛt and khan. rañj takes kvun (Uṇādi 190): rajikā. rajakī is then only "the dyer's wife" (puṃyoga).
• 2908 gas thakan 3.1.146 — thakan after gai (śabde, not gāṅ: TB), for a craftsman: gāthakaḥ.
• 2909 ṇyuṭ ca 3.1.147 — ṇyuṭ after gai too: gā + yu → (ana, yuk) gāyanaḥ. Being ṭit, it gives gāyanī. The rule is split from 2908 so that ṇyuṭ alone continues into 2910 (BM, TB).
• 2910 haś ca vrīhikālayoḥ 3.1.148 — ṇyuṭ after hāk and hāṅ when rice or time is the agent: hāyano vrīhiḥ (it leaves the water behind), hāyano varṣam (it leaves things behind, or reaches them).
• 2911 prusṛlvaḥ samabhihāre vun 3.1.149 — vun after pru, sṛ and lū for repetition, which indicates doing it well (sādhukāritva): pravakaḥ, sarakaḥ, lavakaḥ. TB: skill comes from repetition, so one who does it well even once qualifies, one who does it badly often does not.
• 2912 āśiṣi ca 3.1.150 — vun in a blessing: jīvatāt → jīvakaḥ, nandatāt → nandakaḥ. The blessing is the speaker's (a father's) property (TB, BM). Feminine jīvakā: no i, by āśiṣi vunaś ca na (TB). Pāda 3.1 ends here (BM).

I.G. aṇ with an object upapada, and the upapada compound — running example: kumbha (karman) + kṛ + aṇ → (vṛddhi) kumbha + kāra → (upapada compound) kumbhakāraḥ

• 2913 karmaṇy aṇ 3.2.1 — aṇ after a root when an object is upapada, still in the agent sense (BM). The upapada compound follows (SK 782): kumbhaṃ karotīti kumbhakāraḥ. It does not apply to ādityaṃ paśyati (anabhidhāna).
• Vārttikas: ṇa after śīli, kāmi, bhakṣi, ācari, and after īkṣi, kṣami. The masculine is the same as with aṇ, but the feminine takes ṭāp, not ṅīp (BM): māṃsaśīlā, māṃsakāmā, māṃsabhakṣā, kalyāṇācārā, sukhapratīkṣā, bahukṣamā.
• Why gaṅgādhara, bhūdhara, not -dhāra? The object is intended as a mere relation (śeṣa), so there is no object upapada. dhara is formed on its own and compounded as gaṅgāyā dharaḥ (BM, TB).
• 2914 hvāvāmaś ca 3.2.2 — aṇ after hvā (hveñ), vā (veñ) and mā (māṅ, meṅ), as the exception to the ka of 2915: svarga + hvā + aṇ → (yuk) svargahvāyaḥ, tantuvāyaḥ, dhānyamāyaḥ.

I.H. ka (and ḍa) with an upapada, and the ṣ of -ṣṭha — running example: go + dā + ka → (ā-loss, SK 2372) godaḥ; sama + sthā + ka → samasthaḥ; dvi + stha → (ṣ) dviṣṭhaḥ

• 2915 āto 'nupasarge kaḥ 3.2.3 — ka, not aṇ, after an ā-final root without an upasarga when an object is upapada; then āto lopaḥ: godaḥ, pārṣṇitram (trai). With an upasarga, aṇ: gosaṃdāyaḥ.
• Vārttika: wherever ka is taught, ḍa after roots liable to saṃprasāraṇa: brahma + jyā + ḍa → (ṭi-loss by the ḍ) brahmajyaḥ. ka would give brahmajiya (TB).
• "sarvatra" reaches 2898 too: āhvaḥ, prahvaḥ, not āhuva (BM, TB).
• 2916 supi sthaḥ 3.2.4 — the rule is split:
• supi: ka after an ā-final root with any sup as upapada: dvipaḥ, samasthaḥ, viṣamasthaḥ.
• sthaḥ: ka after sthā. Part one already gives the agent sense, so by ārambhasāmarthya this ka is in bhāva: ākhūnām utthānam ākhūtthaḥ. TB derives it: ā-loss, th by udaḥ sthāstambhoḥ… (SK 118), masculine.
• 2917 praṣṭho 'gragāmini 8.3.92 — pra + sthā + ka with ṣ by nipātana (the s follows no iṇ or ku: BM) when going in front is meant: praṣṭhaḥ (an ox). Otherwise prasthaḥ.
• 2918 ambāmbagobhūmisavyāpadvitrikuśekuśaṅkvaṅgumañjipuñjiparamebarhirdivyagnibhyaḥ sthaḥ 8.3.97 — ṣ for the s of the ka-ended stha after these eighteen: dviṣṭhaḥ, triṣṭhaḥ. BM's list adds ambaṣṭhaḥ … savyeṣṭhaḥ, parameṣṭhaḥ, agniṣṭhaḥ. Only the ka-form is meant, not bhūmisthitam (TB).
• The kaumudī adds: from here on karmaṇi and supi both continue, and with intransitive roots supi is construed.
• 2919 tundaśokayoḥ parimṛjāpanudoḥ 3.2.5 — ka after parimṛj and apanud with tunda and śoka as objects. A vārttika limits it to laziness and bringing happiness: tundaparimṛjaḥ (lazy, no vṛddhi), śokāpanudaḥ. Otherwise aṇ: tundaparimārjaḥ, śokāpanodaḥ.
• Vārttika: ka for mūlavibhuja etc., an ākṛtigaṇa: mūlavibhujaḥ (chariot), mahīdhraḥ, kudhraḥ, gilaḥ.
• 2920 pre dājñaḥ 3.2.6 — ka after a dā-form root and jñā with pra, as the exception to aṇ: sarvapradaḥ, pathiprajñaḥ. With another upasarga there is no ka even by 2915: gosaṃpradāyaḥ (TB).
• 2921 sami khyaḥ 3.2.7 — after khyā (khyāñ for cakṣiṅ: BM, TB) with sam: gāḥ saṃcaṣṭe → gosaṃkhyaḥ. ka is the translator's identification (audit C).

I.I. ṭak after gā and pā; ac after hṛ, grah, arh, ram and jap, śam, śī — running example: sāman + gai + ṭak → (ā-loss) sāmagaḥ → (ṭ: ṅīp) sāmagī; aṃśa + hṛ + ac → (guṇa) aṃśaharaḥ (against aṇ in bhārahāraḥ)

• 2922 gāpoṣ ṭak 3.2.8 — ṭak after gai and pā (pibati), without an upasarga. The k serves ā-loss and the ṭ serves ṅīp (TB): sāmagaḥ, sāmagī. With an upasarga, aṇ: sāmasaṃgāyaḥ.
• Vārttika: ṭak after pibati only with surā, śīdhu: surāpī, śīdhupī. Elsewhere ka gives kṣīrapā brāhmaṇī (BM, TB). pā rakṣaṇe gives surāpā.
• 2923 harater anudyamane 'c 3.2.9 — ac after hṛ when it does not mean lifting: aṃśaharaḥ. Lifting takes aṇ: bhārahāraḥ.
• Vārttika: ac (supplied, BM) after grah with śakti, lāṅgala, aṅkuśa, tomara, yaṣṭi, ghaṭa, ghaṭī, dhanus: śaktigrahaḥ, lāṅgalagrahaḥ. ac is not kit, so there is no saṃprasāraṇa (BM).
• Vārttika: with sūtra, in the sense of holding: sūtragrahaḥ; one who merely takes it up gets aṇ: sūtragrāhaḥ.
• TB: ghaṭī beside ghaṭa shows that the liṅgaviśiṣṭa-paribhāṣā is not universal. This is used at 2955.
• 2924 vayasi ca 3.2.10 — ac after hṛ for lifting too, when age is conveyed: kavacaharaḥ kumāraḥ (old enough to bear armour, actually or potentially: TB).
• 2925 āṅi tācchīlye 3.2.11 — ac after ā + hṛ for habit: puṣpāharaḥ. Without habit: bhārahāraḥ (ours = print; WS bhārāhāraḥ, open).
• 2926 arhaḥ 3.2.12 — ac after arh, as the exception to aṇ. The masculine is the same, but the feminine takes ṭāp, not ṅīp (BM): pūjārhā brāhmaṇī.
• 2927 stambakarṇayo ramijapoḥ 3.2.13 — ac (the translator's supply) after ram and jap with stamba, karṇa (supi only, since ram and jap are intransitive: BM). Vārttika: only for an elephant or an informer: stamberamaḥ, karṇejapaḥ, with aluk by SK 972 or SK 966. Otherwise stambe rantā, karṇe japitā (BM, TB).
• 2928 śami dhātoḥ saṃjñāyām 3.2.14 — ac (supplied, BM) after any dhātu with śam, for a name: śambhavaḥ, śaṃvadaḥ.
• The repeated "dhātu" lets ac win even where the ṭa of 2934 would apply by paratva (TB): śaṃkarā, a parivrājikā.
• 2929 adhikaraṇe śeteḥ 3.2.15 — ac after śī with a locus upapada: khe śete → khaśayaḥ. BM on this rule survives only in Wikisource.
• Vārttikas: pārśva etc., which are not loci (TB): pārśvaśayaḥ, pṛṣṭhaśayaḥ, udaraśayaḥ. uttāna etc. as agents: uttānaśayaḥ, avamūrdhaśayaḥ.
• Vārttika: ḍa with giri, in the Veda: giriśaḥ. In the kāvya line "giriśam upacacāra…", giriśa is giri + śa (lomādi, "who has a mountain"). Ours adds "loke giriśayaḥ", which is not in the print.

I.J. ṭa after car and sṛ — running example: kuru (locus) + car + ṭa → kurucaraḥ → (ṭ: ṅīp) kurucarī

• 2930 careṣ ṭaḥ 3.2.16 — ṭa after car with a locus upapada (adhikaraṇe carried: BM): kuruṣu carati → kurucaraḥ, kurucarī. Kuru is a locus, not an object (BM).
• 2931 bhikṣāsenādāyeṣu ca 3.2.17 — ṭa after car with bhikṣā, senā or ādāya, which are not loci (TB): bhikṣācaraḥ, senācaraḥ, ādāyacaraḥ (ādāya is lyabanta).
• sahacarī: caraṭ is read in pacādi. ṭa is prescribed for the nitya compound; sahacara with ac compounds only optionally (BM, TB).
• 2932 puro 'grato 'greṣu sarteḥ 3.2.18 — ṭa (supplied, BM) after sṛ with puras, agratas or agre: purassaraḥ, agratassaraḥ (print), agresaraḥ. The e of agre is by nipātana, against sup-luk. agrasara is by bāhulaka (Haradatta).
• 2933 pūrve kartari 3.2.19 — ṭa after sṛ with pūrva denoting the agent (kartṛ = kartṛvācin, BM): pūrvaḥ sarati → pūrvasaraḥ. As object: pūrvaṃ deśaṃ sarati → pūrvasāraḥ.

I.K. ṭa after kṛ — running example: yaśas + kṛ + ṭa (hetu) → (s for visarga, SK 160) yaśaskara → (ṭ: ṅīp) yaśaskarī vidyā

• 2934 kṛño hetutācchīlyānulomyeṣu 3.2.20 — ṭa after kṛ (object upapada understood: BM, Wikisource) when cause, habit or compliance is conveyed (dyotya). Since they are conveyed, not denoted, kartari kṛt stands (TB). yaśaskarī vidyā (hetu: s by ataḥ kṛkami…, SK 160), śrāddhakaraḥ (habit), vacanakaraḥ (compliance). Otherwise kumbhakāraḥ (TB).
• The kṛ meant is karoti (ḍukṛñ), though it has two anubandhas, because it is better known; not kṛñ hiṃsāyām (TB). Hetu is the ordinary cause, not the technical one (TB).
• 2935 divāvibhāniśāprabhābhāskārāntānantādibahunāndīkimlipilibibalibhaktikartṛcitrakṣetrasaṃkhyājaṅghābāhv[ahar]yattaddhanuraruṣṣu 3.2.21 — ṭa after kṛ with these 27 upapadas, even without hetu etc.
• divākaraḥ, vibhākaraḥ, niśākaraḥ; bhāskaraḥ and ahaskaraḥ (kaskādi s); bahukaraḥ (bahu = abundance, not a numeral); ekakaraḥ, dvikaraḥ; dhanuṣkaraḥ, aruṣkaraḥ (ṣ by SK 159).
• ahar is missing from our sūtra and restored from the print, and the kaumudī's own ahaskaraḥ requires it.
• Vārttika: ac with kim, yad, tad or bahu: kiṃkarā, yatkarā, tatkarā. ac beats ṭa even in hetu etc. because it is later (paratva). So kiṃkarī is puṃyoga ṅīṣ (BM, TB). TB on the Nyāsa, Vṛtti and prakriyā views: see the audit list above.
• 2936 karmaṇi bhṛtau 3.2.22 — ṭa with the word karman itself (svarūpa, because karmaṇi is already present: BM, TB) when wages are in question: karmakaraḥ (a hireling). Otherwise karmakāraḥ.
• 2937 na śabdaślokakalahagāthāvairacāṭusūtramantrapadeṣu 3.2.23 — no ṭa with these nine. The ṭa barred is the one 2934 would give in hetu etc. (BM): śabdakāraḥ and so on, with aṇ (translator).

I.L. in after kṛ, hṛ, grah and bhṛ — running example: stamba + kṛ + in → (guṇa) stambakariḥ vrīhiḥ; ātman + bhṛ + in → (mum by nipātana) ātmambhariḥ

• 2938 stambaśakṛtor in 3.2.24 — in after kṛ with stamba or śakṛt as object (BM). A vārttika requires rice or calf as agent, paired in order (BM): stambakarir vrīhiḥ, śakṛtkarir vatsaḥ. Otherwise stambakāraḥ, śakṛtkāraḥ. Because in is n-marked, the uttarapada is initial-udātta (TB).
• 2939 harater dṛtināthayoḥ paśau 3.2.25 — in (from 2938, the translator's identification) after hṛ with dṛti or nātha when an animal is the agent: dṛtihariḥ, nāthahariḥ (nātha is a nose-rope). Otherwise dṛtihāraḥ, nāthahāraḥ (ours = print; WS -haraḥ).
• 2940 phalegrahir ātmambhariś ca 3.2.26 — nipātana: phalegrahiḥ (the e of the upapada and in after grah), ātmambhariḥ (mum on ātman and in after bhṛ). "ca" adds kukṣimbhariḥ (BM). The Cāndras read ātman, udara and kukṣi, hence Murāri's udarambhari. TB: kukṣimbharāḥ in a current verse is a mistake.

I.M. khaś, with the short vowel and mum before a khit uttarapada — running example: jana + ej-i + khaś → (śap, guṇa, ay, pararūpa) jana-ejaya → (sup-luk, mum) janamejayaḥ; śunī + dhaya → (hrasva) śuni-dhaya → (mum) śunindhayaḥ

• 2941 ejeḥ khaś 3.2.28 — khaś after eji, i.e. ej + ṇic. The ś-marking, which brings śap, shows that the ṇic stem is meant (TB). An object upapada is understood (BM). The example comes at 2942.
• 2942 arurdviṣadajantasya mum 6.3.67 — mum for arus, dviṣat and an ac-final pūrvapada (not an avyaya) before an uttarapada ending in a khit affix. khiti is carried from 2943, uttarapade from aluk uttarapade (SK 958) (BM). khaś is śit, so śap etc. follow: janān ejayati → janamejayaḥ.
• Vārttika: khaś after aj, dheṭ, tud and jahāti, with vāta, śunī, tila and śardha respectively: vātam ajanti → vātamajāḥ mṛgāḥ.
• 2943 khity anavyayasya 6.3.66 — before a khit uttarapada, a short vowel replaces the final of a non-avyaya pūrvapada (hrasva from SK 999: BM). Shortening comes first, then mum, or the pūrvapada would no longer end in an ac (BM, TB).
• The ī of śunī: śuni-dhaya → śunindhayaḥ; tilantudaḥ.
• śardhaṃjahāḥ māṣāḥ: jahāti carries a causative sense, "they make one release wind".
• 2944 nāsikāstanayor dhmādheṭoḥ 3.2.29 — khaś (supplied, BM). A vārttika sets aside the pairing in order: dheṭ with stana, dhmā (and dheṭ) with nāsikā.
• stanandhayaḥ; feminine stanandhayī by the ṭ of dheṭ.
• nāsikandhamaḥ: dhama by SK 2360, then the short vowel, then mum. Also nāsikandhayaḥ.
• 2945 nāḍīmuṣṭyoś ca 3.2.30 — khaś (from 2941) after dhmā and dheṭ (from 2944) with nāḍī, muṣṭi as objects. Vārttika: the pairing in order is not intended: nāḍindhamaḥ, nāḍindhayaḥ, muṣṭindhamaḥ, muṣṭindhayaḥ. Vārttika: also with ghaṭī, khārī, kharī: ghaṭindhamaḥ, ghaṭindhayaḥ.
• 2946 udi kūle rujivahoḥ 3.2.31 — khaś after ruj and vah preceded by ut, with kūla as object (ruj is transitive). ut is not an upapada (BM, TB): kūlamudrujaḥ, kūlamudvahaḥ.
• 2947 vahābhre lihaḥ 3.2.32 — khaś after lih with vaha (shoulder) or abhra. śap is dropped because lih is adādi. khaś behaves as ṅit, so there is no guṇa of the i: vahaṃlihaḥ (ox), abhraṃlihaḥ (wind).
• 2948 parimāṇe pacaḥ 3.2.33 — khaś after pac with a measure: prasthampacā sthālī, khārimpacaḥ kaṭāhaḥ (khārī shortened, then mum: BM).
• 2949 mitanakhe ca 3.2.34 — and with mita, nakha: mitampacā brāhmaṇī, nakhampacā yavāgūḥ. pac here means heating.
• 2950 vidhvaruṣos tudaḥ 3.2.35 — khaś after tud with vidhu, arus: vidhuntudaḥ (Rāhu). For arus, mum goes after the u, then the conjunct-final s is lost (SK 54): aruntudaḥ.
• 2951 asūryalalāṭayor dṛśitapoḥ 3.2.36 — khaś after dṛś and tap with asūrya, lalāṭa. asūrya is an asamartha compound, since the nañ goes with dṛś (BM): asūryampaśyāḥ (king's wives), lalāṭantapaḥ (the sun).
• 2952 ugrampaśyerammadapāṇindhamāś ca 3.2.37 — three words by nipātana:
• ugraṃpaśyaḥ: ugram is an adverb, and khaś comes after dṛś;
• iraṃmadaḥ (lightning): no śyan, by nipātana. The print reads "śyan na"; ours "na śyan na" reverses the sense;
• pāṇiṃdhamaḥ: a dark road, where hands are sounded.

I.N. khac, and its short vowel — running example: priya + vad + khac → (mum) priyaṃvadaḥ; dviṣat + tāpi + khac → (hrasva, 2955) dviṣat-tapi → (ṇi-loss; mum; t lost) dviṣantapaḥ

• 2953 priyavaśe vadaḥ khac 3.2.38 — khac after vad with priya, vaśa: priyaṃvadaḥ, vaśaṃvadaḥ. Why khac while khaś is current? For 2955 (BM). TB adds that it is placed here to show it comes after other roots too, which answers the objection "then put it in the next rule".
• Vārttika: khac after gam with a sup, for words that are not names (names come by gamaś ca, SK 2964): mitaṅgamaḥ.
• Vārttika: viha for vihāyas, and khac is optionally ḍit: vihaṃgamaḥ / vihaṅgaḥ, bhujaṃgamaḥ / bhujaṃgaḥ. BM ties the ḍit to viha; TB lets it apply elsewhere too.
• 2954 dviṣatparayos tāpeḥ 3.2.39 — khac after tāpi (ṇic of either tap: BM) with dviṣat, para as objects. The forms come at 2955.
• 2955 khaci hrasvaḥ 6.4.94 — before ṇi followed by khac, a short vowel for the penultimate (ṇau from SK 2604, upadhāyāḥ from SK 2364: BM). The ā of tāpi → tapi, then ṇi is lost; dviṣat takes mum (2942) and its conjunct-final t is lost (TB): dviṣantapaḥ, paraṃtapaḥ.
• With khaś, neither the short vowel nor ṇi-loss could happen, since khaś is sārvadhātuka (BM).
• ghaṭa and ghaṭī (2923) show that the liṅgaviśiṣṭa-paribhāṣā is not universal, so dviṣatī does not count as dviṣat: dviṣatīṃ tāpayati → dviṣatītāpaḥ (aṇ, BM).
• 2956 vāci yamo vrate 3.2.40 — khac after yam with vāc, when a vow is meant. The form comes at 2957.
• 2957 vācaṃyamapuraṃdarau ca 6.3.69 — the am-ending of vāc and pur is by nipātana. mum (2942) has no scope here (BM; the reason, that neither ends in a vowel, is the translator's). vācaṃyamaḥ is one under a vow of silence; puraṃdaraḥ is with 2958. Without a vow, aṇ: vāgyāmaḥ.
• 2958 pūḥsarvayor dārisahoḥ 3.2.41 — khac after dāri (dṝ vidāraṇe + ṇic: TB) with pur, and after sah with sarva, in that order (BM).
• puraṃdaraḥ: shortened by 2955, with am by 2957. sarvaṃsahaḥ: sahi is named for words that are not names; names come by SK 2963 (BM, TB).
• The Kāśikā adds dāri with bhaga; "they say" it is got by bāhulaka: bhagaṃdaraḥ.
• 2959 sarvakūlābhrakarīṣeṣu kaṣaḥ 3.2.42 — after kaṣ with sarva, kūla, abhra, karīṣa: sarvaṅkaṣaḥ (a villain), kūlaṅkaṣā (a river), abhraṅkaṣaḥ (the wind), karīṣaṅkaṣā (a whirlwind; print). The affix khac is the translator's supply from the run since 2953. The forms alone cannot tell khac from khaś (audit E).
• 2960 meghartibhayeṣu kṛñaḥ 3.2.43 — khac (supplied, as at 2959) after kṛ with megha, ṛti, bhaya: meghaṅkaraḥ, bhayaṃkaraḥ. "bhaya" includes words ending in it (tadantavidhi: BM, from the Bhāṣya on SK 26): abhayaṅkaraḥ. There is no ṛti example in ours or the print; WS and BM have ṛtiṅkaraḥ.

Part II — SK 2961–3027: the upapada affixes from khac to ḍa after jan, then the niṣṭhā and the n that replaces its t

Covers SK 2961–3027 (67 sūtras): all present.

• AS-refs: all 67 are checked against references/ashtadhyayi/sutraani-data.json, by SK number and sūtra text. So is every SK cross-reference cited below.
• Order: SK order throughout, with no moves. The Aṣṭādhyāyī order runs 3.2.44 → 3.2.101 with interruptions. Bhaṭṭoji leaves out the sūtras he has already taught (3.2.58–60, given in the halanta section as SK 432, 373, 429) and the Vedic ones (3.2.63–67, 71–74 and 88, given later as SK 3409–3419). Where a form needs an aṅga or tripādī rule, he brings it in: 7.2.8, 6.4.41, 8.4.20, 6.4.97, 6.4.40 among the kvip rules, 6.3.68 after khaś, and 1.1.26, 6.4.60, 8.2.42–48, 6.1.24–26, 7.2.15, 8.3.75, 8.3.114 for the niṣṭhā.
• Audit corrections followed (.work/sk-run/p67-audit F–K; all MINOR, applied here even where the en.yml may not yet carry them):
• sūtra-gloss labels. 2964 khac "in a name" is BM's supplement (śeṣa), not an anuvṛtti. 2978 "ader viṭ" is BM's śeṣapūraṇa. 3017 "niṣṭhātasya naḥ" is BM's śeṣapūraṇa, and "from SK 3016" is an inference. 2983: the kaumudī itself supplies "dṛśyate". The han/ḍa/ṭak anuvṛttis at 2966–2968 and 2970 are inferences that the commentators only paraphrase. At 2987 the "supi" anuvṛtti is stated by BM under SK 2988, not 2987.
• readings of the commentators. 2963: BM's reason why rathaṃtara is "mere etymology" is semantic (a sāman cannot be a crosser by chariot); the point that ratha is instrumental is the translator's. 2984: TB's he prāṇ is "an (an prāṇane) + kvip", not "kvip in the sense of breathing". 2998: TB's maxim puts the restriction on the member next to the eva-marked one, and ties are settled by priority (prāthamya) or primacy (prādhānya). 3005: TB's "dṛśigrahaṇa" is the "dṛśyante" of SK 2980.
• senses. 2991: ābhīkṣṇya is "repeatedly", not "habitually" (that is tācchīlya, SK 2988). 3013: "u and k are it" holds of kta as well as of ktavatu. 3021: the sparśa excluded is specifically the cold touch (BM).
• sources and cross-references. 2975: an NSP footnote gives the ṇ of agraṇīḥ by SK 857 (TB's jñāpaka view, against BM). 2977/2980: viṭ, vic and manin etc. come from SK 3413, 3417, 3418. 2978, 3009: the quoted lines are Amarakośa. 2994: TB's last words are at the tail of sk.2993.t1, not lost.
• Print readings followed (NSP 1915, scan-confirmed; ours has slips): 2965 su/dur (not "sudūr"), "uraso lopaś ca" as its own vārttika; 2966 hanter aṇ, karmaṇy upapade; 2972 rājānaṃ; 2973 anāḍhyam āḍhyaṃ, āḍhyaṃkaraṇam; 2974 āḍhyo, ṛtvig-dadhṛk…; 2975 upasarge; 2986 aṅgagat (ours "aṅgat", not in the collation); 2988 brāhmaṇān āmantrayitā, mādhavādibhiḥ; 2989 upapadārthaḥ; 3011 kārakāntareṣu; 3019 ḍīṅaḥ pāṭhasāmarthyāt. Class C for Ajai (ours = print; BM or Wikisource differ): 2986 parītat/sunut (BM purītat/sunat); 2988 sādhukārī (BM sādhudāyī); 2990 sthaṇḍilaśāyī (BM sthaṇḍileśāyī); 3011 "dvijaḥ. brāhmaṇajaḥ" (BM's lemma "dvijo brāhmaṇaḥ").

Where Part II begins. Part I closed at 2960 meghartibhayeṣu kṛñaḥ (3.2.43): khac after kṛñ with megha, ṛti or bhaya as upapada (meghaṅkaraḥ, abhayaṅkaraḥ). Part II opens in the same run of khac after kṛñ (3.2.44 kṣemapriyamadre 'ṇ ca). Throughout, dhātoḥ (SK 2829), kartari kṛt (SK 2832) and tatropapadaṃ saptamīstham (SK 781) stand behind every rule, and the upapada compounds by upapadam atiṅ (SK 782).

Where Part II ends. At 3027 pratistabdhanistabdhau ca (8.3.114), the last of three ṣ-rules that Bhaṭṭoji attaches to niṣṭhā forms. Part III opens at 3028 divo 'vijigīṣāyām (8.2.49), which resumes the n-for-niṣṭhā-t series (dyūnaḥ, but dyūtam).

Contents of Part II

• A. 2961–2964: khac (and aṇ) after kṛñ, bhū, bhṛ … dam and gam: kṣemaṅkaraḥ/kṣemakāraḥ, āśitaṃbhavaḥ, viśvambharaḥ, sutaṃgamaḥ.
• B. 2965–2967: ḍa after gam and han, with ṭi-loss by the ḍ-marker: antagaḥ, uragaḥ, śatruhaḥ, tamopahaḥ.
• C. 2968–2972: han with ṇini and ṭak, and the nipātanas pāṇigha/tāḍagha: kumāraghātī, jāyāghnaḥ, pittaghnam, hastighnaḥ, rājaghaḥ.
• D. 2973–2974: khyun, khiṣṇuc, khukañ in the sense of cvi: āḍhyaṃkaraṇam, āḍhyaṃbhaviṣṇuḥ. The dṛś/spṛś forms of 3.2.58–60 are recalled: sadṛk, tādṛkṣaḥ.
• E. 2975–2979: zero and near-zero affixes with a sup upapada: kvip (mitradviṭ, agraṇīḥ), ṇvi (aṃśabhāk), viṭ (āmāt, kravyāt), kap with gh (kāmadughā).
• F. 2980–2983: manin, kvanip, vanip, vic and kvip after any root; no iṭ before a vaś-initial kṛt; ā for a nasal before viṭ/van: suśarmā, prātaritvā, vijāvā, roṭ, ukhāsrat.
• G. 2984–2987: what happens to the root before kvip: ṇ in he prāṇ, short ā of chādi, m-loss of gam; and ka/kvip after sthā: mitraśīḥ, tanucchat, aṅgagat, śaṃsthaḥ.
• H. 2988–2992: ṇini: tācchīlya, upamāna, vrata, ābhīkṣṇya, man: uṣṇabhojī, uṣṭrakrośī, kṣīrapāyiṇaḥ, darśanīyamānī.
• I. 2993–2994: khaś for thinking of oneself, and am for an ic-final monosyllable: paṇḍitaṃmanyaḥ, kāliṃmanyā, gāṃmanyaḥ, śrimanyam.
• J. 2995–2998: the bhūte heading; ṇini of yaj and han in the past; kvip of han under restriction: somayājī, pitṛvyaghātī, brahmahā.
• K. 2999–3003: kvip of kṛñ, su and ci, and ini of vi-krī, in the past: sukṛt, somasut, agnicit, śyenacit, somavikrayī.
• L. 3004–3006: kvanip after dṛś, yudh and kṛñ: pāradṛśvā, rājayudhvā, sahakṛtvā. The karmaṇi heading ends.
• M. 3007–3011: ḍa after jan with a seventh/fifth-case word, an upasarga, anu, and elsewhere: sarasijam, saṃskārajaḥ, prajā, pumanujā, dvijaḥ, parikhā.
• N. 3012–3013: the name niṣṭhā and the affixes kta/ktavatu in the past: snātam, stutaḥ, kṛtavān.
• O. 3014–3019: n for the niṣṭhā-t: after long kṣī, r/d, saṃyogādi ā, the lvādi, the odit roots: kṣīṇavān, śīrṇaḥ, bhinnaḥ, glānaḥ, lūnaḥ, bhugnaḥ.
• P. 3020–3023: śyaiṅ before the niṣṭhā: saṃprasāraṇa and n by sense: śīnaṃ ghṛtam, śītaṃ jalam, pratiśīnaḥ, abhiśyānam/abhiśīnam.
• Q. 3024–3025: añc and the iṭ-ban for optionally-seṭ roots; the niṣṭhā substitute siddha for ṣatva: samaknaḥ, udaktam, vṛkṇaḥ.
• R. 3026–3027: ṣ withheld: pariskandaḥ, pratistabdhaḥ, nistabdhaḥ.

Threads that run through Part II:

• The headings and how far they reach. bhūte (2995) runs up to, not including, vartamāne laṭ (SK 2151 = 3.2.123), and qualifies the root: "after a root denoting the past" (BM, TB). karmaṇi (2997) runs up to sahe ca (3006), where the kaumudī says it has stopped. Smaller anuvṛttis do much of the work: khac from SK 2953, ḍa from 2965, ṇini from 2988 (kept alive through 2993's "ca" into 2996, TB), kvip from 2998, kvanip from 3004, janer ḍaḥ from 3007, and the niṣṭhā-t → n of 3016 (8.2.42) through 3015–3024. Where the commentators only paraphrase or say "iti śeṣaḥ", this outline says "supplied", not "carried forward" (the audit's recurring correction).
• Zero affixes. kvip, ṇvi, viṭ, vic and kvin leave nothing behind: the it-letters go and the v is lost by ver apṛktasya (SK 375). What is seen is their effect: tuk after a short vowel (hrasvasya piti kṛti tuk, SK 2858: sukṛt, śatrujit, prātaritvā), vṛddhi from ṇ (aṃśabhāk), ā for a nasal (vijāvā), m-loss (aṅgagat), ūṭh (akṣadyūḥ), and pada-final sandhi (mitradviṭ, āmāt).
• The ḍ-marker forces ṭi-loss without bha. In antagaḥ, śatruhaḥ, sarasijam the ṭi goes "by the force of the ḍ-marking" (ḍittva-sāmarthya), because ṭeḥ (SK 316) needs a bha stem (2965, 3007).
• Restriction (niyama) and elaboration (prapañca). If a rule re-teaches what is already available, the commentators read it as a restriction. 2978 (kravyāt: viṭ only, aṇ blocked), 2998 (Kāśikā fourfold vs Bhāṣya twofold), 2999 (threefold), 3000–3001 (fourfold) and 3004 (kvanip only) work this way. A restriction excludes only affixes taught with an upapada, so the niṣṭhā still comes (brahma hatavān, karma kṛtavān: BM, TB). The opposite device is prapañca. satsūdviṣa… (2975) only elaborates kvip ca (2983); BM treats the gam-vārttikas of 2965 the same way, and api in 3011 overrides every condition of the ḍa rules.
• Senses outside the agent. The default sense is the agent (kartari kṛt), but the sūtras name other senses explicitly: karaṇa and bhāva (2962 āśitaṃbhava, 2973 khyun), adhikaraṇa (2965 sugaḥ), karman (3002 śyenacit, with BM's two "karmaṇi"), and bhāva/karman for kta (3013).
• The niṣṭhā-n is tripādī, hence asiddha. Because the n of 8.2.42 ff. is asiddha, kutva comes first (bhugnaḥ, samaknaḥ), saṃprasāraṇa first (jīnaḥ), and ṣatva by vraśca… would apply. A vārttika makes the substitute siddha for ṣatva, svara, affixes and iṭ (vṛkṇaḥ, 3025). In the other direction, a bahiraṅga vṛddhi is asiddha to the niṣṭhā-n rule (kārtiḥ, 3016).
• The kaumudī lists forms bare. The derivation stages cited below (āḍhya-m-karaṇa, śi-ta, sam-ak-na …) are BM's or TB's as recorded in the unit notes, or the standard rules the notes name. Glosses of forms follow the English.

II.A. khac (and aṇ) after kṛñ, bhū, the "name" roots and gam — running example: kṣema + kṛ + khac → kṣema-m-kara → kṣemaṅkaraḥ; kṣema + kṛ + aṇ → kṣemakāraḥ

• 2961 kṣemapriyamadre 'ṇ ca 3.2.44 — with kṣema, priya or madra as upapada, kṛñ takes aṇ, and by "ca" also khac (from 2960). khac is kh-marked, so the upapada takes mum (SK 2942): kṣemaṅkaraḥ; aṇ is ṇit, so the root vowel takes vṛddhi: kṣemakāraḥ. Likewise priyaṅkaraḥ/priyakāraḥ, madraṅkaraḥ/madrakāraḥ.
• Why "aṇ" and not "vā"? With "vā", aṇ would come anyway by karmaṇy aṇ (SK 2913). Naming aṇ keeps out the ṭa of kṛño hetutācchīlyānulomyeṣu (SK 2934) (BM).
• alpārambhāḥ kṣemakarāḥ: when the object is meant as a mere relation (śeṣa), there is no object upapada, so pacādi ac (SK 2896) comes: kṣema-kara.
• 2962 āśite bhuvaḥ karaṇabhāvayoḥ 3.2.45 — with āśita as upapada, bhū takes khac (the kaumudī supplies it) in the sense of the instrument or the bare action. Instrument: āśito bhavaty anena → āśitaṃbhavaḥ odanaḥ ("rice that sates"); bhāva: āśitasya bhavanam → āśitambhavaḥ.
• TB: by vāsarūpa, lyuṭ also comes (āśitabhavanam). ghañ is blocked, since it is of the same form as khac.
• 2963 saṃjñāyāṃ bhṛtṝvṛjidhārisahitapidamaḥ 3.2.46 — in a name, khac comes (BM: supplied, śeṣa) after bhṛ, tṝ, vṛ, ji, dhāri, sah, tap, dam, the upapada being an object.
• viśvaṃ bibharti → viśvambharaḥ (Viṣṇu), viśvaṃbharā (the earth); patiṃvarā kanyā (vṛ); śatruṃjayo hastī (ji); yugaṃdharaḥ parvataḥ (dhāri: BM gives the shortening by khaci hrasvaḥ, SK 2955, and the loss of ṇi); śatruṃsahaḥ, śatruṃtapaḥ, ariṃdamaḥ.
• rathaṃtaram (a sāman): "rathena tarati" is mere etymology, since the meaning of the parts is not carried through. BM's reason is semantic: a sāman cannot cross by chariot. TB says the etymology is kept for accent and avagraha. (The point that ratha is instrumental is the translator's: audit F.)
• dam "quelling": transitive for Mādhava and others; for Haradatta and others it has the causative sense taken in (antarbhāvita-ṇyartha).
• Why "in a name"? TB: kuṭumbaṃbhāraḥ (so the print; an aṇ form). The padaccheda's "vṛji" is two roots, vṛ and ji (audit F; metadata only).
• 2964 gamaś ca 3.2.47 — gam too takes khac in a name (BM: "saṃjñāyāṃ khac" supplied, śeṣa): suta + gam + khac → sutaṃgamaḥ.
• TB: gam is kept out of 2963 so that only gam carries into 2965.

II.B. ḍa after gam and han: the ṭi lost by the ḍ-marker — running example: anta + gam + ḍa → anta-g-a → antagaḥ

• 2965 antātyantādhvadūrapārasarvānanteṣu ḍaḥ 3.2.48 — with anta, atyanta, adhvan, dūra, pāra, sarva or ananta as upapada, gam takes ḍa. "In a name" has ceased. The ṭi am of gam is lost by the force of the ḍ-marking, though the stem is not bha: antaṃ gacchati → antagaḥ, and so on.
• Vārttikas: sarvatra, panna: sarvatragaḥ, pannagaḥ (panna = kta of pad, used as a qualifier of the action). uras with loss of its final (a separate vārttika in the print): urasā gacchati → uragaḥ. su, dur in the sense of a locus (adhikaraṇa): sugaḥ, durgaḥ (print su + dur). anyatrāpi dṛśyate: grāmagaḥ. viha for vihāyas before ḍa: vihagaḥ.
• BM: the vārttikas only elaborate "anyatrāpi dṛśyate". The viha vārttika is needed because vihāyaso vihaḥ belonged to khac (vihaṅgamaḥ). TB: in the object sense, su/dur take khal (SK 3305): sugamaḥ, durgamaḥ.
• 2966 āśiṣi hanaḥ 3.2.49 — han takes ḍa with an object upapada when a wish is understood: śatruṃ vadhyāt → śatru + han + ḍa → śatruhaḥ. Without a wish, aṇ, with t for n (SK 2574): śatrughātaḥ.
• Vārttika: with dāru as upapada, aṇ after āṅ-han and ṭ for the final in a name: dārvāghāṭaḥ. With cāru, optionally: cārvāghāṭaḥ. With sam and an object: varṇasaṅghāṭaḥ / varṇasaṅghātaḥ, padasaṅghāṭaḥ / padasaṅghātaḥ. The "karmaṇi" there stops the carrying-on of "cārau" (BM).
• 2967 ape kleśatamasoḥ 3.2.50 — with kleśa or tamas as upapada, apa-han takes ḍa where there is no wish (BM: anāśīrartham): kleśāpahaḥ putraḥ, tamopahaḥ sūryaḥ.

II.C. han with ṇini and ṭak; the nipātanas pāṇigha, tāḍagha — running example: jāyā + han + ṭak → jāyā-ghn-a → jāyāghnaḥ

• 2968 kumāraśīrṣayor ṇiniḥ 3.2.51 — with kumāra or śīrṣa as object, han takes ṇini: kumāraghātī. śīrṣa for śiras is by nipātana: śīrṣaghātī.
• TB: the rule lets ṇini come without the senses of tācchīlya, āvaśyaka or ādhamarṇya (cf. SK 2988), and secures śīrṣa.
• 2969 lakṣaṇe jāyāpatyoṣ ṭak 3.2.52 — with jāyā or pati as object, han takes ṭak when the agent bears a mark. lakṣaṇa = lakṣaṇavat, by arśaādi ac (SK 1933). The penultimate a of han is lost (SK 2363, BM) and h becomes gh before n (SK 358): jāyāghno nā. ṭ-marked, so the feminine takes ṅīp: patighnī strī. The "killing" is figurative (gauṇa): the mark only presages it (TB).
• 2970 amanuṣyakartṛke ca 3.2.53 — ṭak also when the agent is non-human: jāyāghnas tilakālakaḥ, patighnī pāṇirekhā, pittaghnaṃ ghṛtam. Counter-example: ākhughātaḥ śūdraḥ.
• pralambaghnaḥ, śatrughnaḥ, kṛtaghnaḥ (human agents) take ka by mūlavibhujādi (the vārttika under SK 2919). coraghāto hastī takes aṇ by bahulaka (SK 2841). TB: ṭak and aṇ count as the same form, so vāsarūpa cannot give aṇ.
• 2971 śaktau hastikapāṭayoḥ 3.2.54 — with hastin or kapāṭa as upapada, ṭak when capability is meant, for a human agent: hastighno nā (able, by his own strength, to kill an elephant: TB), kapāṭaghnaś coraḥ. kavāṭa is a variant (BM's lemma reads it). TB counter-example: hastighātaḥ (killing by poison).
• 2972 pāṇighatāḍaghau śilpini 3.2.55 — for an artisan, nipātana with pāṇi or tāḍa as upapada: ṭak, loss of the ṭi, gh for h: pāṇighaḥ, tāḍaghaḥ. Otherwise: pāṇighātaḥ, tāḍaghātaḥ.
• Vārttika: rājagha is added to the nipātana: rājānaṃ hanti → rājaghaḥ (print rājānaṃ).

II.D. Making or becoming what one was not: khyun, khiṣṇuc, khukañ — running example: anāḍhyam āḍhyaṃ kurvanty anena → āḍhya + kṛ + khyun → āḍhya-m-kar-ana → āḍhyaṃkaraṇam

• 2973 āḍhyasubhagasthūlapalitanagnāndhapriyeṣu cvyartheṣv acvau kṛñaḥ karaṇe khyun 3.2.56 — with āḍhya … priya as object, in the cvi sense (becoming what was not) but not ending in cvi, kṛñ takes khyun in the instrument sense. BM: yu → ana, mum from the kh, and khyun is an exception to lyuṭ: āḍhyaṃkaraṇam.
• Why "acvau"? āḍhyīkurvanty anena: no khyun. The Kāśikā says that by the force of the prohibition not even lyuṭ comes. On the Bhāṣya's view lyuṭ does come, and "acvau" serves the next rule.
• 2974 kartari bhuvaḥ khiṣṇuckhukañau 3.2.57 — under the same conditions, bhū takes khiṣṇuc and khukañ in the agent sense. "acvau" runs on (TB). anāḍhya āḍhyo bhavati → āḍhyaṃbhaviṣṇuḥ; with khukañ the ñ gives vṛddhi: āḍhyaṃbhāvukaḥ.
• The kaumudī then recalls 3.2.58–60, taught in the halanta section. spṛśo 'nudake kvin (SK 432): ghṛtaspṛk; "karmaṇi" has stopped, so also mantraspṛk. ṛtvig… (SK 373). tyadādiṣu dṛśo 'nālocane kañ ca (SK 429).
• Vārttika: samāna and anya also take kvin/kañ after dṛś: sadṛk, sadṛśaḥ; anyādṛk, anyādṛśaḥ (BM: in the karmakartṛ sense). Vārttika: ksa too: tādṛkṣaḥ, sadṛkṣaḥ, anyādṛkṣaḥ.

II.E. Zero and near-zero affixes with a sup upapada: kvip, ṇvi, viṭ, kap — running example: mitra + dviṣ + kvip → mitra-dviṣ → mitradviṭ

• 2975 satsūdviṣadruhaduhayujavidabhidacchidajinīrājām upasarge 'pi kvip 3.2.61 — sad, sū, dviṣ, druh, duh, yuj, vid, bhid, chid, ji, nī, rāj take kvip with a sup upapada, with or without an upasarga. kvip vanishes entirely; ṣ is pada-final: mitradviṭ / pradviṭ. Likewise dyusat / upaniṣat (ṣ by sadir aprateḥ), aṇḍasūḥ/prasūḥ, mitradhruk/pradhruk, godhuk/pradhuk, aśvayuk/prayuk, vedavit/nivit.
• BM: "upasarge 'pi" is a jñāpaka that elsewhere "sup" excludes an upasarga. TB: sū is the adādi sūte, and every vid except vidḷ lābhe is included.
• Vārttika: after agra and grāma, ṇ for the n of nī: agraṇīḥ, grāmaṇīḥ. An NSP footnote gives the ṇ by pūrvapadāt saṃjñāyām agaḥ (SK 857), agreeing with TB (jñāpita) against BM (a new statement) (audit G). TB: with aṇ there is no ṇ: grāmanāyaḥ.
• 2976 bhajo ṇviḥ 3.2.62 — bhaj takes ṇvi with a sup or upasarga upapada. ṇvi vanishes; the ṇ gives vṛddhi of the penultimate (BM): aṃśa + bhaj + ṇvi → aṃśabhāj → aṃśabhāk, prabhāk.
• 2977 ado 'nanne 3.2.68 — ad takes viṭ with a sup upapada other than anna. viṭ comes from the Vedic janasanakhanakramagamo viṭ (A 3.2.67 = SK 3413; BM, TB). āmam atti → āmāt, sasyāt. With anna: annādaḥ (aṇ, SK 2913).
• 2978 kravye ca 3.2.69 — with kravya too, viṭ (BM: "ader viṭ" is śeṣapūraṇa). The form would come from 2977 anyway, so the rule is to block aṇ: kravyāt only.
• "kravyādo 'srapa āśaraḥ" (the Amarakośa list of rākṣasa names: audit G) has aṇ with pakvamāṃsa as upapada, and pṛṣodarādi replaces it with kravya: kravyādaḥ.
• 2979 duhaḥ kab ghaś ca 3.2.70 — duh takes kap with a sup upapada, and gh replaces its final: kāmaṃ dogdhi → kāma-dugh-a → kāmadughā (dhenuḥ). kab is kap, voiced before gh.

II.F. manin, kvanip, vanip, vic and kvip after any root — running example: su + śṝ + manin → su-śar-man → suśarmā; vi + jan + vanip → vi-jā-van → vijāvā

• 2980 anyebhyo 'pi dṛśyante 3.2.75 — "in the Veda" has stopped. manin, kvanip, vanip (from āto manin-kvanib-vanipaś ca, A 3.2.74 = SK 3418) and vic (from vij upe chandasi, A 3.2.73 = SK 3417) are seen after other roots too (TB). The examples come in 2981–2982.
• 2981 neḍ vaśi kṛti 7.2.8 — no iṭ for a kṛt affix beginning with vaś (BM: tadādi-vidhi). śṝ + manin, guṇa, no iṭ: suśarmā. iṇ + kvanip, with tuk (SK 2858, TB): prātaritvā.
• 2982 viḍvanor anunāsikasyāt 6.4.41 — before viṭ and van, ā replaces a final nasal: vi + jan + vanip → vi-ja-ā-van → vijāvā (savarṇadīrgha). oṇṛ: ā for ṇ, av for o: avāvā.
• vic: roṭ, reṭ (ruṣ, riṣ with guṇa; v lost by ver apṛktasya), sugaṇ (gaṇ; BM, Wikisource).
• 2983 kvip ca 3.2.76 — kvip too is seen after any root, with or without an upapada, in the Veda and in ordinary usage. The kaumudī supplies "dṛśyate" (audit G). 2975 is only an elaboration of this rule. Examples: ukhāsrat (sraṃs: n lost, s → d/t), parṇadhvat, vāhabhraṭ (bhraṃś; the Vṛtti reads vahābhraṭ, per Haradatta: TB).

II.G. What kvip does to the root, and ka/kvip after sthā — running example: aṅga + gam + kvip → aṅga-ga → aṅga-ga-t → aṅgagat; he prāṇ; śam + sthā + ka → śaṃsthaḥ

• 2984 antaḥ 8.4.20 — the pada-final n of an becomes ṇ after a cause in an upasarga. aniteḥ (SK 2478), raṣābhyāṃ no ṇaḥ (SK 235) and upasargāt (SK 2287) carry on; antaḥ is nominative for genitive (BM via Wikisource; our BM layer lacks it). pra + an + kvip: n-loss is barred in the vocative by na ṅisaṃbuddhyoḥ (SK 352), so he prāṇ. TB reads this as "an prāṇane + kvip" (audit G).
• Further kvip forms follow. śāsa id aṅhaloḥ (SK 2486) and lengthening (SK 433) give mitraśīḥ. Vārttika: i for the penultimate of āśās before kvi gives āśīḥ. With i and u for ṝ: gīḥ, pūḥ.
• 2985 ismantrankviṣu ca 6.4.97 — before is, man, tran, kvi, a short vowel replaces the ā of chādi. chādeḥ comes from SK 3297 and hrasvaḥ from SK 2955 (BM). tanuṃ chādayati → tanucchat. With is, man, tran (BM): chadiḥ, chadma, chatram.
• The kaumudī then runs other kvip formations. Lengthening (SK 2666) and m → n (SK 341): pratān, praśān. ūṭh by cchvoḥ (SK 2561): akṣadyūḥ. ūṭh by jvaratvara… (SK 2654): jūḥ, tūḥ, srūḥ, ūḥ. vṛddhi (SK 73): janauḥ. Also mūḥ, sumūḥ (sumvau by oḥ supi, SK 281: TB). rāl lopaḥ (SK 2655): mūḥ (mūrchā), dhūḥ (dhurvī).
• 2986 gamaḥ kvau 6.4.40 — gam loses its nasal before kvi (anunāsikalopa supplied from SK 2428, BM), then tuk: aṅgagat (print; ours "aṅgat").
• Vārttika: "of gam and the rest": parītat, saṃyat, sunut (print; BM purītat, sunat: Class C). Vārttika: ū replaces the penultimate, and the nasal goes: agre gacchati → agregūḥ, agrebhrūḥ.
• 2987 sthaḥ ka ca 3.2.77 — sthā with a sup upapada (with or without an upasarga) takes ka, and by "ca" kvip. śaṃsthaḥ (ka, ā lost); śaṃsthāḥ (kvip). The rule exists to block the ac of śami dhātoḥ (SK 2928). BM states the "supi" anuvṛtti under SK 2988.
• There is no ī (SK 2462) in śaṃsthāḥ. BM explains this from the Bhāṣya's usage and pṛṣodarādi; TB says sthānivattva is barred for an operation on a letter.

II.H. ṇini: habit, comparison, vow, repetition, and man — running example: uṣṇa + bhuj + ṇini → uṣṇa-bhoj-in → uṣṇabhojī

• 2988 supy ajātau ṇinis tācchīlye 3.2.78 — with a sup upapada not denoting a genus, a root takes ṇini for habitual disposition: uṣṇabhojī, śītabhojī. Genus: brāhmaṇān āmantrayitā (tṛn; print reading). Not habit: uṣṇaṃ bhuṅkte kadācit.
• The Vṛtti allows ṇini only without an upasarga and so needs a vārttika for ut/prati-āṅ-sṛ; Haradatta and Mādhava follow it. Bhaṭṭoji rejects this as against the Bhāṣya: ṇini with an upasarga is established (upajīvin, anuyāyin, visārin, anujīvin).
• Vārttikas for ṇini without habit: in the sense of sādhukārin, sādhukārī (BM/Wikisource sādhudāyī: Class C); with brahman, vad: brahmavādī.
• 2989 kartary upamāne 3.2.79 — ṇini when the upapada's agent is the standard of comparison for the affix's agent: uṣṭra iva krośati → uṣṭrakrośī, dhvāṅkṣarāvī. This gives ṇini without habit and with a genus word. Counter-examples: apūpān iva bhakṣayati māṣān (not the agent); uṣṭraḥ krośati (no comparison).
• 2990 vrate 3.2.80 — ṇini with a sup upapada when a vow is meant: sthaṇḍilaśāyī (print; BM sthaṇḍileśāyī, aluk: Class C). Our BM layer lacks this block; Wikisource has it.
• 2991 bahulam ābhīkṣṇye 3.2.81 — variously, ṇini for repetition: kṣīrapāyiṇa uśīnarāḥ ("the Uśīnaras repeatedly drink milk"; not "habitually": audit H). BM: this lets a genus word (kṣīra) be the upapada.
• 2992 manaḥ 3.2.82 — ṇini after man. Because "bahula" runs on, only the divādi manyate is meant (BM, TB): darśanīyamānī, "one who thinks (another) handsome".

II.I. Thinking of oneself: khaś, and am for an ic-final monosyllable — running example: paṇḍita + man + śyan + khaś → paṇḍita-m-manya → paṇḍitaṃmanyaḥ; go + am + manyaḥ → gāṃmanyaḥ

• 2993 ātmamāne khaś ca 3.2.83 — when manyate means thinking with oneself as object, it takes khaś, and by "ca" ṇini. khaś is ś-marked, so it is sārvadhātuka and takes śyan; it is kh-marked, so mum (SK 2942) (BM): paṇḍitaṃmanyaḥ, paṇḍitamānī.
• With a feminine, khity anavyayasya (SK 2943) shortens and blocks puṃvadbhāva by paratva: kāliṃmanyā. An indeclinable is left alone: divāmanyā.
• TB: "ca" keeps ṇini, not khaś, running into 2996.
• 2994 ica ekāco 'm pratyayavac ca 6.3.68 — an ic-final monosyllabic first member takes am before a khit uttarapada, and that am counts as the case-affix am. khiti comes from SK 2943 and uttarapade from SK 958 (BM). It is an apavāda of mum.
• go + am: ā by auto 'mśasoḥ (SK 285) gives gāṃmanyaḥ. With strī, iyaṅ optionally (SK 302): striyaṃmanyaḥ / strīṃmanyaḥ. With nṛ, guṇa (SK 275, BM): naraṃmanyaḥ. Also bhuvaṃmanyaḥ.
• śrimanyaṃ kulam: on the Bhāṣya's authority śrī is shortened as a neuter (SK 318), with neither mum nor am. TB's closing words are filed at the tail of sk.2993.t1 (audit H).

II.J. The bhūte heading; ṇini in the past; kvip of han under restriction — running example: brahman + han + kvip → brahma-han → brahmahā

• 2995 bhūte 3.2.84 — heading: "after a root denoting the past", up to but not including vartamāne laṭ (SK 2151) (BM, TB).
• 2996 karaṇe yajaḥ 3.2.85 — with an instrument upapada, yaj in the past takes ṇini in the agent sense. ṇini runs on as svarita, not khaś (TB). somena iṣṭavān → somayājī, agniṣṭomayājī.
• 2997 karmaṇi hanaḥ 3.2.86 — with an object upapada, han in the past takes ṇini: pitṛvyaghātī (t by SK 2574, gh by SK 358: BM). "karmaṇi" is a heading up to sahe ca (3006).
• TB: the Kāśikā requires "kutsita" here; the kaumudī says nothing of it.
• 2998 brahmabhrūṇavṛtreṣu kvip 3.2.87 — with brahman, bhrūṇa or vṛtra as object, han in the past takes kvip: brahmahā, bhrūṇahā, vṛtrahā.
• kvip ca (2983) already gives it, so the rule is a restriction. The Kāśikā reads it fourfold: only with brahman etc., only after han, only in the past, only kvip. The Bhāṣya reads it twofold: only with brahman etc., only kvip.
• BM: only an affix taught with an upapada is excluded, so the niṣṭhā still comes (brahma hatavān). TB's niyama labels follow the maxim that the restriction falls next to the eva-member (audit H).

II.K. kvip of kṛñ, su, ci, and ini of vi-krī — running example: su + kṛ + kvip → su-kṛ-t → sukṛt

• 2999 sukarmapāpamantrapuṇyeṣu kṛñaḥ 3.2.89 — with su, karman, pāpa, mantra, puṇya as upapada, kṛñ in the past takes kvip, with tuk after the short ṛ (SK 2858): sukṛt, karmakṛt, pāpakṛt, mantrakṛt, puṇyakṛt.
• The Kāśikā's threefold restriction works as follows. "Only kvip": no aṇ in karma kṛtavān. "Only after kṛñ": mantrādhyāyaḥ, not kvip. "Only in the past": no kvip for makes/will make. There is no "only with su etc.", so śāstrakṛt, bhāṣyakṛt also stand.
• 3000 some suñaḥ 3.2.90 — with soma as object, su in the past takes kvip: somasut. The restriction is fourfold here and in the next rule. TB, from the scan: somasāva/agnicāya excluded; surāsāva, iṣṭakācāya, somakrāya with aṇ.
• 3001 agnau ceḥ 3.2.91 — with agni as object, ci in the past: agnicit (one who has piled the fire-altar).
• 3002 karmaṇy agnyākhyāyām 3.2.92 — with an object upapada, ci takes kvip in the object sense when the whole names a fire-site. BM: one "karmaṇi" supplies the upapada and the other the affix-sense: śyena iva citaḥ → śyenacit.
• 3003 karmaṇīni vikriyaḥ 3.2.93 — with an object upapada, vi-krī in the past takes ini: vi-kre-in → vikrayin, giving somavikrayī, ghṛtavikrayī.
• Vārttika: "kutsita" is to be added: only a contemptible object (selling soma is forbidden). TB: dhānyavikrāyaḥ (aṇ).

II.L. kvanip after dṛś, yudh and kṛñ; karmaṇi ends — running example: pāraṃ dṛṣṭavān → pāra + dṛś + kvanip → pāradṛśvā

• 3004 dṛśeḥ kvanip 3.2.94 — dṛś takes kvanip; "object" and "past" continue: pāradṛśvā.
• BM, TB: kvanip is already available (SK 3418, 2980), so the rule restricts. It excludes manin and vanip and the upapada affixes, but not the niṣṭhā. TB: other times are not excluded (viśvadṛśvan of the present is no fault).
• 3005 rājani yudhikṛñaḥ 3.2.95 — with rājan as object, yudh and kṛñ take kvanip. yudh has the causative sense taken in: rājānaṃ yodhitavān → rājayudhvā; rājakṛtvā (tuk).
• TB (Śabdakaustubha): 3005–3006 are a restriction like 3004. Since reaching these forms through the "dṛśyante" of SK 2980 is a last resort, the three sūtras may equally be called prescriptive (audit I).
• 3006 sahe ca 3.2.96 — with saha as upapada, the same kvanip. "karmaṇi" has stopped: sahayudhvā, sahakṛtvā.

II.M. ḍa after jan — running example: sarasi + jan + ḍa → sarasi-j-a → sarasijam

• 3007 saptamyāṃ janer ḍaḥ 3.2.97 — with a seventh-case upapada, jan in the past takes ḍa. The ṭi goes by ḍittva. The seventh-case ending stays by SK 966 or 972 (BM): sarasijam. mandurāyāṃ jātaḥ → mandurajaḥ, with the shortening by ṅyāpoḥ… (SK 1001).
• 3008 pañcamyām ajātau 3.2.98 — with a fifth-case upapada that is not a genus word: saṃskārajaḥ, adṛṣṭajaḥ. janer ḍaḥ runs on, and BM supplies "bhūtārthāt".
• 3009 upasarge ca saṃjñāyām 3.2.99 — with an upasarga, when the whole is a name: pra-jan-ḍa → prajā. The kaumudī quotes "prajā syāt santatau jane", the Amarakośa nānārtha line (audit I).
• 3010 anau karmaṇi 3.2.100 — anu-jan with an object upapada: pumāṃsam anurudhya jātā → pumanujā. jan is transitive here because it means birth following upon (anurodha) (BM, TB).
• 3011 anyeṣv api dṛśyate 3.2.101 — ḍa after jan is seen with other upapadas too: ajaḥ, dvijaḥ, brāhmaṇajaḥ (print; BM's lemma "dvijo brāhmaṇaḥ": Class C).
• api allows every condition of the ḍa rules to lapse (sarvopādhivyabhicāra), even the root and the kāraka: paritaḥ khātā → parikhā (khan, object sense).

II.N. The niṣṭhā: name and affixes — running example: snā + kta → snātam (bhāva); stu + kta → stutaḥ (karman); kṛ + ktavatu → kṛtavān (agent)

• 3012 ktaktavatū niṣṭhā 1.1.26 — kta and ktavatu are each called niṣṭhā (BM). TB answers the charge of mutual dependence: the name is "one that will be" (bhāvinī saṃjñā).
• 3013 niṣṭhā 3.2.102 — a niṣṭhā comes after a root denoting the past. kta takes bhāva and karman by tayor eva… (SK 2833); ktavatu takes the agent by kartari kṛt (SK 2832). u and k are it-markers (the k of kta too: audit J). Examples: snātaṃ mayā, stutas tvayā viṣṇuḥ, viṣṇur viśvaṃ kṛtavān.
• BM: by laḥ karmaṇi… (SK 2152), intransitive roots take kta in bhāva and transitive roots take it in karman.

II.O. n for the t of the niṣṭhā — running example: bhid + kta → bhid-ta → bhin-na → bhinnaḥ; kṣi + ktavatu → kṣī-tavat → kṣīṇavān

• 3014 niṣṭhāyām aṇyadarthe 6.4.60 — before a niṣṭhā outside the senses of ṇyat (bhāva and karman), so in the agent, kṣi gets a long vowel. kṣiyaḥ comes from SK 3338 and dīrghaḥ from SK 3544 (BM): kṣi + tavat → kṣī-tavat.
• 3015 kṣiyo dīrghāt 8.2.46 — after long kṣī, n for the niṣṭhā-t, then ṇ after ṣ: kṣīṇavān. In bhāva/karman there is no lengthening and so no n: kṣitaḥ kāmo mayā (antarbhāvita-ṇyartha, BM, TB).
• The kaumudī also recalls that śryukaḥ kiti (SK 2381) bars iṭ: śritaḥ, bhūtaḥ, kṣutaḥ. Vārttika: ūrṇu is treated like ṇu, so it counts as single-vowelled and takes no iṭ: ūrṇutaḥ. Also nutaḥ, vṛtaḥ.
• 3016 radābhyāṃ niṣṭhāto naḥ pūrvasya ca daḥ 8.2.42 — after r or d, n replaces the niṣṭhā-t, and the preceding d becomes n too: bhinnaḥ, chinnaḥ.
• śṝ: i by ṝta id dhātoḥ (SK 2390), r follows, then n and ṇatva: śīrṇaḥ.
• It does not apply in kārtiḥ (kṛta + iñ), because the vṛddhi is bahiraṅga and so asiddha. TB: r/d must immediately precede the t, so caritam and uditam keep their t (iṭ intervenes).
• 3017 saṃyogāder āto dhātor yaṇvataḥ 8.2.43 — after the ā of a root that begins with a conjunct and contains a yaṇ: drāṇaḥ, styānaḥ, glānaḥ, mlānaḥ (glai → glā by SK 2370: BM). BM supplies "niṣṭhātasya naḥ" as śeṣapūraṇa (audit J).
• 3018 lvādibhyaḥ 8.2.44 — after the twenty-one lvādi (the pvādi of the kryādi class without pūñ): lūnaḥ. jyā takes saṃprasāraṇa first (SK 2412; it is siddha and nitya, TB), then lengthening (SK 2559) and n by this rule: jīnaḥ.
• Vārttika: du, gu take n and are lengthened: dūnaḥ, gūnaḥ. ṭudu is excluded, which is why Māgha has "dutayā".
• Vārttika: pūñ in the sense of perishing: pūnā yavāḥ, but pūtam.
• Vārttika: si(nu) whose object-agent is the mouthful: sino grāsaḥ. Counter-examples: sitā pāśena sūkarī, sito grāso devadattena.
• 3019 oditaś ca 8.2.45 — after an o-marked root. With bhujo, the n is asiddha, so kutva comes first: bhugnaḥ. With ṭuośvi, saṃprasāraṇa and lengthening: ucchūnaḥ. ohāk: prahīṇaḥ.
• svādaya oditaḥ: sūnaḥ, dūnaḥ. ḍīṅ takes no iṭ, because its reading among the odit roots would otherwise be pointless (print pāṭhasāmarthyāt): uḍḍīnaḥ.

II.P. śyaiṅ before the niṣṭhā: saṃprasāraṇa and n by sense — running example: śyai + kta → śyā-ta → śi-ta → śī-ta → śīnam / śītam

• 3020 dravamūrtisparśayoḥ śyaḥ 6.1.24 — when it means a liquid hardening (kāṭhinya) or touch, śyā (from śyaiṅ) takes saṃprasāraṇa before a niṣṭhā (from SK 1003, BM): śyā-ta → śi-ta.
• 3021 śyo 'sparśe 8.2.47 — after śyā, n for the niṣṭhā-t except in touch. "Touch" means the particular touch called cold (BM, audit J). Lengthening by halaḥ (SK 2559): śīnaṃ ghṛtam. Touch: śītaṃ jalam.
• Outside both senses there is no saṃprasāraṇa: saṃśyāno vṛścikaḥ, with n by SK 3017 (TB).
• 3022 prateś ca 6.1.25 — after prati, saṃprasāraṇa even outside those senses (BM): pratiśīnaḥ.
• 3023 vibhāṣābhyavapūrvasya 6.1.26 — after abhi/ava, optionally: abhiśyānam / abhiśīnaṃ ghṛtam, avaśyānaḥ / avaśīnaḥ vṛścikaḥ. It is a fixed option (vyavasthita), so samavaśyānaḥ only.

II.Q. añc and the iṭ-ban for optionally-seṭ roots; the niṣṭhā substitute siddha for ṣatva — running example: sam + añc + kta → sam-ac-ta → sam-ak-ta → samak-na → samaknaḥ

• 3024 añco 'napādāne 8.2.48 — after añc, n for the niṣṭhā-t unless an apādāna is construed. The examples are printed under 3025: samaknaḥ, but udaktam udakaṃ kūpāt.
• 3025 yasya vibhāṣā 7.2.15 — a root that takes iṭ optionally anywhere takes no iṭ before the niṣṭhā. niṣṭhāyām comes from SK 3039 and na iṭ from SK 2981 (BM).
• añc is optional before ktvā by udito vā (SK 3328), so: sam-añc-ta, n lost (SK 415), c → k (the n being asiddha), no iṭ, n by 3024: samaknaḥ. BM reads yasya as yasmāt; TB takes it as a genitive with the ārdhadhātuka.
• Since the n is asiddha, vraśca… (SK 294) would give ṣ. Vārttika: the niṣṭhā substitute is siddha for ṣatva, svara, affixes and iṭ, so vraśc + ta → vṛkṇaḥ, vṛkṇavān.

II.R. ṣ withheld: pariskanda, pratistabdha, nistabdha — running example: pari + skand → pariskandaḥ (eastern), pariṣkandaḥ / pariskandaḥ (elsewhere)

• 3026 pariskandaḥ prācyabharateṣu 8.3.75 — among the eastern Bharatas, pariskanda is a nipātana without ṣ. pareś ca (SK 2399) would give ṣ optionally, and elsewhere both forms stand. BM: pari-skand with the niṣṭhā-t lost; TB: with ac, or the niṣṭhā-t elided as the Kāśikā says.
• The kaumudī closes by setting up the next rule: ṣ would be due by stambheḥ (SK 2272).
• 3027 pratistabdhanistabdhau ca 8.3.114 — in pratistabdha and nistabdha, no ṣatva: prati + stanbh + kta → pratistabdhaḥ, nistabdhaḥ. There is no BM or TB on this sūtra.

Part III (SK 3028–3092) — the niṣṭhā continued: t-substitutes, iṭ, the seṭ niṣṭhā, nipātanas, root changes, the senses of kta, ṅvanip and atṛn

Covers SK 3028–3092 (65 sūtras): all present.

• AS-refs: all 65 are checked against references/ashtadhyayi/sutraani-data.json, by SK number (skn) and sūtra text.
• Order: SK order throughout, with no reordering.
• Sources: each unit's en.yml and source.json, with the commentary notes where BM or TB settles a reading, plus .work/sk-run/p67-notes.md (T13–T18). The finished audits that cover this part are K (3025–3037) and L (3038–3050). No audit report yet exists for 3051–3092.

Where Part III begins / ends

• Begins. Part II defined the niṣṭhā (ktaktavatū niṣṭhā 1.1.26, SK 3012) and prescribed it in the past (niṣṭhā 3.2.102, SK 3013). It then opened the tripādī run of t-substitutes with radābhyāṃ niṣṭhāto naḥ pūrvasya ca daḥ (8.2.42, SK 3016) and gave the general iṭ-ban yasya vibhāṣā (7.2.15, SK 3025). Part III picks up that 8.2 run at 3028 divo 'vijigīṣāyām (8.2.49).
• Ends. At 3092 jīryater atṛn (3.2.104), still under bhūte (3.2.84, SK 2995). Part IV opens with chandasi liṭ (3.2.105, SK 3093), followed by kānac and kvasu (3094–3095).

Contents of Part III

• A. 3028–3034: more substitutes for the niṣṭhā-t: n (dyūnaḥ, nirvāṇaḥ), k (śuṣkaḥ), v (pakvaḥ), m (kṣāmaḥ, prastīmaḥ).
• B. 3035–3037: phulla by nipātana, and praphulta with an upasarga (āditaś ca, ti ca).
• C. 3038–3040: optional n (nunnaḥ/nuttaḥ), no iṭ after śvi and the īdit roots, and roots that refuse n (dhyātaḥ, mattaḥ).
• D. 3041–3044: fixed forms vitta, bhitta, ṛṇa; sphī for sphāy.
• E. 3045–3050: iṭ restored to the niṣṭhā (and ktvā) against yasya vibhāṣā.
• F. 3051–3056: a seṭ niṣṭhā that is not kit (śayitaḥ, pavitaḥ), and kta in ādikarman.
• G. 3057: ṇi lost before a seṭ niṣṭhā (bhāvitaḥ), with a catalogue of where the niṣṭhā keeps or lacks iṭ.
• H. 3058–3065: aniṭ niṣṭhā forms fixed for particular senses (kṣubdha … abhyarṇa).
• I. 3066–3071: the niṣṭhā from ṇic-stems (vṛtta, śṛta, dānta …), and the options of iṭ (ruṣṭa/ruṣita, hṛṣṭa/hṛṣita, apacita).
• J. 3072–3073: pī for pyāy, and a short vowel for hlād.
• K. 3074–3081: root changes before a kit t-initial affix: i, hi, dath, t, the long upasarga, jagdhi, and long kṣī.
• L. 3082–3085: ṣ for s in niṣṇāta, pratiṣṇāta, kapiṣṭhala, viṣṭhala.
• M. 3086–3090: which sense the kta expresses: agent, locus, present, neuter bhāva.
• N. 3091–3092: other past participles: ṅvanip (sutvā, yajvā) and atṛn (jaran).

Threads that run through Part III:

• Niṣṭhā means kta and ktavatu (SK 3012). The kaumudī mostly cites the kta form in the masculine nominative. It adds the ktavatu form where a rule turns on it (phullavān 3035, śayitavān 3052, marṣitavān 3055, jīrṇavān 3092). A niṣṭhā is kit, so by default it blocks guṇa. It denotes the past and, by tayor eva kṛtyaktakhalarthāḥ (SK 2833), bhāva and karman. Groups F and M widen this.
• Three carried headings.
• (1) 'niṣṭhātaḥ naḥ'. It runs from SK 3016 through 8.2.60 (3028–3043): the niṣṭhā-t becomes n, or k, v or m, or is kept (3040), or is fixed by nipātana (3041–3043). The kaumudī often supplies the words itself. Several units make the SK 3016 attribution explicitly the translator's.
• (2) 'na iṭ' + 'niṣṭhāyām'. 'na iṭ' comes from neḍ vaśi kṛti (7.2.8, SK 2981) and 'niṣṭhāyām' from śvīdito niṣṭhāyām (7.2.14, SK 3039). Together they govern 7.2.15–30 (3025, 3036, 3054, 3058–3071) (BM). The SK teaches 7.2.14 at 3039, after 7.2.16 at 3036, so BM's anuvṛtti reaches back in SK order.
• (3) The iṭ re-prescriptions 7.2.47–54 (3045–3050), each answering yasya vibhāṣā.
• yasya vibhāṣā (SK 3025) is the pivot of iṭ. A root that is optionally seṭ anywhere takes no iṭ before the niṣṭhā. Its optional iṭ may come by udito vā (SK 3328), tīṣasaha… (SK 2340), svaratisūti… (SK 2279), niraḥ kuṣaḥ (SK 2560) or vibhāṣā gamahana… (SK 3099). This gives vittam 3041, gūḍhaḥ/vataḥ/tataḥ 3057 and iṣṭaḥ 3089.
• 3045–3050 override it.
• 3057 shows it is not obligatory (patitaḥ) and holds only for ekāc roots (daridritaḥ).
• 3059 and 3069–3070 restrict it or turn it into an option.
• kit-hood. 1.2.19–22 (3051, 3052, 3055, 3056) deny kit-hood to a seṭ niṣṭhā, so it takes guṇa: śayitaḥ, pavitaḥ, marṣitaḥ, dyotitam. BM takes 'na seṭ' from na ktvā seṭ (SK 3322) and 'kit' from asaṃyogāl liṭ kit (SK 2242). Group K's rules all need a "t-initial kit" affix, carried from dyatisyati… (3074).
• Anuvṛtti from later in the SK. Bhaṭṭoji gathers every rule that shapes a niṣṭhā form from AS 8.2, 6.1, 7.2, 1.2, 3.4, 6.4, 7.4, 6.3, 2.4, 8.3, 3.2 and 3.3. So BM draws on sūtras the SK teaches later:
• 'niṣṭhāyām' from SK 3044 into 3033;
• 'ktvāniṣṭhayoḥ' from SK 3049 into 3046–3048;
• 'niṣṭhā' from SK 3052 into 3051;
• 'na seṭ' from SK 3322 into 3051–3056.
• Tense headings. bhūte (SK 2995) governs the 3.2 rules, including 3091–3092. BM carries vartamāne from SK 2151 by maṇḍūkapluti to 3088–3089 (3087's closing line). 3090 holds for time in general (kālasāmānye).
• 3086 is cited throughout. Many earlier examples are kta in the agent by gatyarthākarmaka…: nirvāṇaḥ 3029, kṣāmaḥ 3032, rucitam 3056, hṛṣṭaṃ loma 3070, kṣīṇāyuḥ 3081.
• Commentator disagreements are recorded, not resolved.
• 3028: dyūnaḥ is stuta (BM) or kṣīṇa (TB).
• 3029: wind in nirvāṇo dīpo vātena is a karaṇa (BM) or a hetu (TB).
• 3040: khyā is superfluous (BM) or khyā prakathane (TB).
• 3041: the nipātana is framed against SK 3038 (BM) or against all the vid roots (TB).
• 3051: why ktvā alone continues: BM gives the contrast with a bare 'pūṅaś ca', TB the anukṛṣṭa 'ca'.
• 3062: kṛcchra 'denotes' the cause of pain (BM) or reaches it by lakṣaṇā (TB).
• 3067: the milk/oblation restriction comes from a vārttika (BM) or from the vyavasthita vibhāṣā of SK 3023 (TB).
• Derivation steps. The kaumudī usually gives forms bare. Steps cited below (rule names, intermediate strings) are BM's or TB's as recorded in the unit notes. Where a step is the translator's inference, it is marked.

Audit corrections followed (K and L; the en.yml files may not carry them yet):

• 3034: 'niṣṭhātasya maḥ' is a śeṣa per BM. BM names no anuvṛtti source (SK 3016/3032 is the translator's).
• 3037: TB on the vārttika says the statement excludes the forms utphulta, saṃphulta. The print's 'utphulla saṃphulla' there is a slip.
• 3041: TB says vid sattāyām and vid vicāraṇe are not used at all in the senses bhoga and pratyaya, and vid jñāne must take iṭ. Hence 'vindateḥ' (vid lābhe). (Our TB OCR's 'varjitaṃ' is the print's 'vṛttir na'.)
• 3044: TB derives sphītikāma from the ṇic-stem of 'sphītam ācaṣṭe' by the uṇādi rule aca iḥ.
• 3045: the iṭ is obligatory. Restating 'iṭ' stops the anuvṛtti of 'vā' from SK 2279 (BM), as at 3046.
• 3048: only BM attests 'vimohane'. TB's lemma 'lubho—' does not decide it.

Print readings followed (NSP 1915, scan-confirmed; textual notes in the units):

• 3029 vātaś cet;
• 3035 ñiphalā, kṣīva (also prakṣīvitaḥ 3037);
• 3037 carphalor ataḥ (ours 'carphalopata'; not in the collation);
• 3041 niṣṭhāntasya, vinnaḥ;
• 3057 the eye-skip '[nipātanād iṭ. patitaḥ. se 'sici (SK 2506) iti]', and daridritaḥ;
• 3058 mliṣṭam avispaṣṭam;
• 3060 'ca nipātyate' (ours 'na', which reverses the sense; escalated);
• 3062 niṣṭhāyā, syāt;
• 3066 vartitā;
• 3068 nipātyante;
• 3069 ādittve;
• 3070 hṛṣṭo;
• 3078 anudattam;
• 3087 the eye-skip '[ca. ramāpatir idaṃ yātaḥ. tenedaṃ yātam. bhujeḥ karmaṇi]', plus yathāprāptam, āsitam, asti;
• 3082 niṣṇātaḥ śāstreṣu as one example.
• At 3048 the print's 'lubho 'vimohane' (with avagraha) is a misprint, and ours is kept.

Still open for Ajai (mūla, class C):

• 3068 jñapitaḥ (ours = print = TB) vs jñāpitaḥ (Wikisource; BM's lemma);
• 3072 pīnaṃ sukham (ours = print) vs mukham (svāṅga requires it);
• 3086 glānaḥ (ours = print = TB) vs mlānaḥ (BM, Wikisource);
• 3087 BM presupposes 'pakṣe' (āser akarmakatvāt);
• 3089 ākṛṣṭaḥ (ours) vs ākruṣṭaḥ (TB's root list; the English follows TB);
• 3060–3061 b/v (dṛḍha/paribṛḍha);
• 3065 nāsannaṃ vs āsannaṃ;
• 3090 klība/klīva.

III.A. More substitutes for the niṣṭhā-t: n, k, v, m — running example: div + kta → dyū + ta → dyūnaḥ; śuṣ + ta → śuṣkaḥ; pac + ta → pakvaḥ; kṣai → kṣāmaḥ; pra-styā → prastīmaḥ

• 3028 divo 'vijigīṣāyām 8.2.49 — after div, n replaces the niṣṭhā-t, except when the sense is desire to conquer.
• div + ta: the v → ūṭh by cchvoḥ śūḍ anunāsike ca (SK 2561), giving dyū + ta. The t after dyū → n: dyūnaḥ.
• BM glosses dyūnaḥ 'praised', TB 'exhausted'. Both are recorded.
• With vijigīṣā (dice are thrown to win): dyūtam "gambling".
• 3029 nirvāṇo 'vāte 8.2.50 — after vā preceded by nir, n replaces the niṣṭhā-t, provided wind is not the agent.
• nir + vā + ta → nirvā + na → nirvāṇaḥ (ṇ after the r): nirvāṇo 'gniḥ "the fire is extinguished", nirvāṇo muniḥ "the sage has come to rest". The kta is agentive by SK 3086.
• Wind as agent: nirvāto vātaḥ.
• In nirvāṇo dīpo vātena the wind is karaṇa (BM) or hetu (TB), not agent, so the n stands.
• 3030 śuṣaḥ kaḥ 8.2.51 — after śuṣ, k replaces the niṣṭhā-t ('niṣṭhātaḥ … ity eva'): śuṣ + ta → śuṣ + ka → śuṣkaḥ "dried".
• 3031 paco vaḥ 8.2.52 — after pac, v. pac + ta → pac + va. The v is asiddha, so coḥ kuḥ (SK 378) still sees c before the jhal t, and the c becomes k: pakvaḥ "cooked" (BM; the rule identification is the translator's).
• 3032 kṣāyo maḥ 8.2.53 — after kṣai, m. kṣai → kṣā by ādeca upadeśe 'śiti (SK 2370). Then kṣā + ta → kṣā + ma → kṣāmaḥ. BM gives it two senses, 'wasted' and 'made to waste', so it is left unglossed.
• 3033 styaḥ prapūrvasya 6.1.23 — saṃprasāraṇa for styā (styai with ā made) after pra, before a niṣṭhā. BM: 'saṃprasāraṇam' from SK 1003, 'niṣṭhāyām' from SK 3044.
• pra + styā + ta → pra + sti-a + ta; pūrvarūpa gives sti, and halaḥ (SK 2559) lengthens: prastī + ta.
• There is no example here. The forms come under the next sūtra.
• 3034 prastyo 'nyatarasyām 8.2.54 — after pra-styā, m optionally replaces the niṣṭhā-t. 'niṣṭhātasya maḥ' is a śeṣa per BM (audit K).
• prastī + ta → prastīmaḥ "congealed" / prastītaḥ.
• The alternative gets no n: after saṃprasāraṇa, prastī- has neither a yaṇ nor a final ā for SK 3017 (TB).
• Without pra: styānaḥ, with n by SK 3017 (translator's reading, in line with TB's saṃstyānaḥ).

III.B. phulla by nipātana, and praphulta with an upasarga — running example: ñiphalā + kta → phal + la → phullaḥ; pra + phal + ta → praphultaḥ

• 3035 anupasargāt phullakṣīvakṛśollāghāḥ 8.2.55 — without an upasarga, phulla, kṣīva, kṛśa and ullāgha are nipātanas.
• phulla. Only the l for the niṣṭhā-t is nipātana: ñiphalā viśaraṇe + ta → phal + la. The u comes by 3037 and the absence of iṭ by 3036 (BM): phullaḥ. The nipātana reaches the kta inside ktavatu: phullavān.
• The other three. Only the loss of the t of kta (not of ktavatu) is nipātana. Because that loss is asiddha, the iṭ that would follow is also blocked by nipātana: kṣīv + ta → kṣīvaḥ "intoxicated", kṛśaḥ "thin", ullāghaḥ "free of disease". The senses are restricted by the nipātana (BM).
• 3036 āditaś ca 7.2.16 — after an ādit root, no iṭ comes to a niṣṭhā ('na iṭ' from SK 2981, 'niṣṭhāyām' from SK 3039, per BM).
• ñiphalā is ādit, so phal + ta stays aniṭ, both in phullaḥ and in praphultaḥ.
• The kaumudī gives no example of its own.
• 3037 ti ca 7.4.89 — u replaces the a of car and phal before a t-initial kit affix. BM: car/phal from SK 2636, ata ut from SK 2637.
• With pra there is no nipātana, so the t stays: pra + phal + ta → pra + phul + ta → praphultaḥ. With pra the other three take iṭ: prakṣīvitaḥ, prakṛśitaḥ, prollāghitaḥ.
• BM gets 'kiti' from SK 2456 by maṇḍūkapluti, then calls that anuvṛtti "nirmūlā niṣphalā ca".
• praphullam (in lodhradrumaṃ sānumataḥ praphullam) is phull + the pacādi ac (SK 2896). So 3035 serves "to exclude phulta and the like". TB: that wording is illustrative; the rule also secures phullavān and the initial accent (SK 3693).
• Vārttika: utphulla and saṃphulla are to be added. TB: this excludes utphulta, saṃphulta (audit K).

III.C. Optional n, no iṭ after śvi and the īdit, and roots that refuse n — running example: nud + kta → nunnaḥ / nuttaḥ; undī → unnaḥ / uttaḥ; but dhyai → dhyātaḥ, mad → mattaḥ

• 3038 nudavidondatrāghrāhrībhyo 'nyatarasyām 8.2.56 — after nud, vid, und, trā, ghrā and hrī, n optionally replaces the niṣṭhā-t. For hrī the n was not due; for the others it was obligatory (SK 3016/3017). So this is an option in both cases (BM, TB).
• nud + ta: the t → n, and the preceding d → n (pūrvasya ca daḥ), giving nunnaḥ; otherwise nuttaḥ.
• vid means only the rudh-class vid vicāraṇe, by sāhacarya with und: vinnaḥ / vittaḥ. vid jñāne (seṭ) gives viditaḥ; vid sattāyām gives vinnaḥ (obligatory).
• The kaumudī breaks off at "undī" (udāharaṇasūcanam, BM). Its examples resume under 3039.
• 3039 śvīdito niṣṭhāyām 7.2.14 — after śvi and an īdit root, no iṭ comes to a niṣṭhā ('na iṭ' from SK 2981, per BM).
• undī + ta → und + ta, with no iṭ. The n is optional by 3038, and the root's own n is lost by aniditām (SK 415; TB): unnaḥ / uttaḥ.
• Then 3038's examples: trāṇaḥ, trātaḥ; ghrāṇaḥ, ghrātaḥ; hrīṇaḥ, hrītaḥ.
• The śvi form (śūnaḥ) comes at 3057.
• 3040 na dhyākhyāpṝmūrchimadām 8.2.57 — after dhyā, khyā, pṝ, mūrch and mad, the niṣṭhā-t does not become n.
• dhyai → dhyā + ta → dhyātaḥ: the n due by SK 3017 does not come. Likewise khyātaḥ. BM thinks khyā is superfluous; TB takes it as khyā prakathane.
• pṝ + ta → pūrtaḥ: no iṭ by śryukaḥ kiti (SK 2381), and ut by SK 2494.
• mūrch + ta: rāl lopaḥ (SK 2655) drops the ch, and āditaś ca blocks iṭ: mūrtaḥ.
• mad + ta → mattaḥ: īdit, so no iṭ.

III.D. Fixed forms in restricted senses — vitta, bhitta, ṛṇa — and sphī — running example: vid (lābhe) + kta → vittam (dhanam) ~ vinnaḥ

• 3041 vitto bhogapratyayayoḥ 8.2.58 — vitta is a nipātana of the niṣṭhā form of vindati (vid lābhe) in the senses 'enjoyed' and 'known'.
• vittaṃ dhanam "wealth"; vittaḥ puruṣaḥ "a well-known man". In other senses: vinnaḥ "obtained".
• There is no iṭ because vid lābhe is veṭ before kvasu (SK 3099), so yasya vibhāṣā (SK 3025) applies.
• BM: it fixes the absence of 3038's n. TB: vid jñāne must take iṭ, and vid sattāyām and vid vicāraṇe are not used in these senses, hence 'vindateḥ' (audit L).
• 3042 bhittaṃ śakalam 8.2.59 — bhitta, with no n for the niṣṭhā-t of bhid, meaning 'a fragment'.
• bhid + ta → bhittam. Otherwise bhinnam "split" (SK 3016).
• BM: bhitta is rūḍha and a synonym of śakala.
• 3043 ṛṇam ādhamarṇye 8.2.60 — n for the t of kta after ṛ, in a debtor's transaction.
• ṛ + ta → ṛṇam; otherwise ṛtam "true".
• TB: ādhamarṇya, by lakṣaṇā, is the transaction itself, which also accounts for uttamarṇa (SK 573).
• 3044 sphāyaḥ sphī niṣṭhāyām 6.1.22 — sphī replaces sphāy(ī) before a niṣṭhā: sphāy + ta → sphī + ta → sphītaḥ.
• TB derives sphītikāma by aca iḥ (audit L).
• Its 'niṣṭhāyām' is the one BM carries back to 3033 and 3072.

III.E. iṭ restored to the niṣṭhā (and ktvā) against yasya vibhāṣā — running example: nir + kuṣ + kta → niṣ + kuṣ + i + ta → niṣkuṣitaḥ

• 3045 iṇ niṣṭhāyām 7.2.47 — after nir-kuṣ ('niraḥ kuṣaḥ', SK 2560, carried), iṭ comes to a niṣṭhā, obligatorily (audit L).
• nir-kuṣ is optionally seṭ before tṛc etc., so yasya vibhāṣā would bar iṭ. This renewed prescription (punarvidhi) overrides it: nir + kuṣ + i + ta → niṣkuṣitaḥ.
• 3046 vasatikṣudhor iṭ 7.2.52 — after vas (vasati) and kṣudh, iṭ comes obligatorily to ktvā and a niṣṭhā ('ktvāniṣṭhayoḥ' from SK 3049, per BM). This overrides ekāca upadeśe 'nudāttāt (SK 2246).
• vas + i + ta: saṃprasāraṇa (SK 2409) and ṣ (SK 2410) give uṣitaḥ.
• Likewise kṣudhitaḥ. TB rejects the Dhātuvṛtti's itac derivation.
• 3047 añceḥ pūjāyām 7.2.53 — after añc 'to honour', iṭ comes to ktvā and a niṣṭhā. Otherwise udito vā (SK 3328) and yasya vibhāṣā would interfere.
• añc + i + ta → añcitaḥ: the n is kept by nāñceḥ pūjāyām (SK 424).
• In the sense of going: aktaḥ, with the n lost by SK 415.
• 3048 lubho vimohane 7.2.54 — after the tud-class lubh 'to bewilder', iṭ comes obligatorily to ktvā and a niṣṭhā. This overrides tīṣasaha… (SK 2340) and yasya vibhāṣā.
• lubh + i + ta → lubhitaḥ "bewildered".
• In the sense of greed (the div-class root): lubdhaḥ "greedy" (agentive by SK 3089).
• 3049 kliśaḥ ktvāniṣṭhayoḥ 7.2.50 — after kliś, iṭ is optional before ktvā and a niṣṭhā ('vā' from SK 2279, per BM).
• For kliśa upatāpe the iṭ would be obligatory (SK 2184).
• For kliśū vibādhane the option before ktvā already exists (ūdit), but before a niṣṭhā yasya vibhāṣā would bar iṭ.
• Result: kliśitaḥ / kliṣṭaḥ.
• 3050 pūṅaś ca 7.2.51 — after pūṅ too, iṭ is optional before ktvā and a niṣṭhā. BM: 'iṭ' from 3045, 'vā' from 2279, 'ktvāniṣṭhayoḥ' from 3049.
• It overrides śryukaḥ kiti (SK 2381). The forms come at 3051.

III.F. A seṭ niṣṭhā that is not kit, and kta at the start of an action — running example: śī + kta → śī + i + ta (not kit) → śe + i + ta → śayitaḥ

• 3051 pūṅaḥ ktvā ca 1.2.22 — after pūṅ, a seṭ ktvā and niṣṭhā are not kit. BM: 'na seṭ' from SK 3322, 'kit' from SK 2242, 'niṣṭhā' from SK 3052.
• With iṭ (3050): pū + i + ta takes guṇa and av: pavitaḥ. Without iṭ, kit-hood stays: pūtaḥ.
• ktvā is named here only so that it alone continues into nopadhāt… (SK 3324).
• 3052 niṣṭhā śīṅsvidimidikṣvididhṛṣaḥ 1.2.19 — after śīṅ, svid, mid, kṣvid and dhṛṣ, a seṭ niṣṭhā is not kit.
• śī + i + ta → śe + i + ta → śayitaḥ, śayitavān.
• The ṅ of śīṅ excludes yaṅluk. There the niṣṭhā stays kit, with no guṇa and y by er anekācaḥ (TB): śeśyitaḥ, śeśyitavān.
• Vārttika: a niṣṭhā is to be stated in ādikarman. Once the first moments are past the action is begun but not yet past, so the past-tense niṣṭhā would not reach it (BM, TB).
• 3053 ādikarmaṇi ktaḥ kartari ca 3.4.71 — the kta of ādikarman denotes the agent, and by 'ca' also bhāva and karman (BM: from SK 2833; TB gives the purpose of 'ca' by takrakauṇḍinya).
• No example here. The kaumudī reprises the rule under 3080: prakṛtaḥ kaṭaṃ saḥ "he has begun the mat"; prakṛtaḥ kaṭas tena.
• 3054 vibhāṣā bhāvādikarmaṇoḥ 7.2.17 — in bhāva and ādikarman, iṭ to a niṣṭhā after an ādit root is optional. BM: 'āditaḥ' from 3036, 'niṣṭhā' from 3039, 'na' from 2981.
• pra + svid + i + ta is not kit by 3052, so it takes guṇa: prasveditaś caitraḥ "Caitra has begun to sweat". In bhāva: prasveditaṃ tena.
• In 3052, svid is the bhvādi ñiṣvidā, by association with the ñīt roots. The divādi svid gives only sviditaḥ (kit).
• mid and kṣvid are both divādi and bhvādi: prameditaḥ, prakṣveditaḥ, pradharṣitaḥ (and the -vān forms). TB rejects Haradatta's narrowing of kṣvid.
• Why "seṭ" (in 3052)? prasvinnaḥ (aniṭ, kit, n).
• 3055 mṛṣas titikṣāyām 1.2.20 — after mṛṣ 'to forbear', a seṭ niṣṭhā is not kit.
• mṛṣ + i + ta → marṣitaḥ, marṣitavān.
• In another sense the niṣṭhā stays kit: apamṛṣitaṃ vākyam "an indistinct utterance". TB: this shows that roots have many meanings.
• 3056 udupadhād bhāvādikarmaṇor anyatarasyām 1.2.21 — after a root with penultimate u, in bhāva and ādikarman, a seṭ niṣṭhā is optionally not kit.
• dyut + i + ta → dyutitam / dyotitam; muditam / moditaṃ sādhunā. In ādikarman: pradyutitaḥ / pradyotitaḥ, pramuditaḥ / pramoditaḥ sādhuḥ.
• Counter-examples: viditam (no penultimate u); rucitaṃ kārṣāpaṇam (not bhāva or ādikarman); kruṣṭam (not seṭ).
• Vārttika: only for roots with the vikaraṇa śap. Not here: gudhyati → gudhitam.

III.G. ṇi lost before a seṭ niṣṭhā, and where the niṣṭhā keeps or lacks iṭ — running example: bhū + ṇic → bhāvi + i + ta → bhāv + ita → bhāvitaḥ

• 3057 niṣṭhāyāṃ seṭi 6.4.52 — before a seṭ niṣṭhā, ṇi is deleted. BM: 'ṇeḥ' from SK 2313 and 'lopaḥ' from SK 2372. ṇer aniṭi could not apply, since the affix is seṭ.
• bhāvi + i + ta → bhāvitaḥ, bhāvitavān.
• The kaumudī then runs through further niṣṭhā forms:
• No iṭ by śvīditaḥ (SK 3039). śvi + ta takes saṃprasāraṇa: śu → śū by halaḥ, and n by oditaś ca (SK 3019), giving śūnaḥ. Also dīptaḥ (dīpī).
• No iṭ by yasya vibhāṣā (veṭ roots). guhū: gūḍhaḥ (ḍh, dh, ṣṭutva, ḍh-loss, lengthening). vanu: vataḥ; tanu: tataḥ (n lost by anudāttopadeśa…, SK 2428).
• iṭ despite yasya vibhāṣā. pat is veṭ before san, so yasya vibhāṣā should bar iṭ.
• But iṭ is given by nipātana in dvitīyā śrita… (SK 686): patitaḥ (print).
• Alternatively: kṛt, cṛt and nṛt are already veṭ by se 'sici (SK 2506), so their ī-marking would be pointless. It therefore shows that yasya vibhāṣā is not obligatory. Hence also dhāvitam in ibharājadhiyā.
• Only for one-vowelled roots. yasya vibhāṣā is limited to ekāc roots, so the polysyllabic daridrā keeps its iṭ: daridritaḥ (print; ours 'daridrataḥ').

III.H. Aniṭ niṣṭhā forms fixed for particular senses — running example: kṣubh + kta → kṣubh + ta → kṣubdhaḥ (mantha) ~ kṣubhitam

• 3058 kṣubdhasvāntadhvāntalagnamliṣṭaviribdhaphāṇṭabāḍhāni manthamanastamaḥsaktāvispaṣṭasvarānāyāsabhṛśeṣu 7.2.18 — eight aniṭ forms are fixed as whole words (samudāyena) for eight senses, respectively.
• kṣubdhaḥ, for mantha (barley-meal in liquid, or the churning-stick): dh by jhaṣas tathoḥ (SK 2280).
• svāntam "mind" and dhvāntam "darkness": lengthened by SK 2666.
• lagnam "attached": its n is also a nipātana.
• mliṣṭam "indistinct" and viribdhaḥ "a tone": the i of mlecch and rebh is also a nipātana.
• phāṇṭam, a decoction made without effort. Mādhava says it is the stage just before fresh butter.
• bāḍham "intensely".
• In other senses: kṣubhitam, svanitam, dhvanitam, lagitam, mlecchitam, virebhitam, phaṇitam, bāhitam. But kṣubdho rājā stands, because the āgama-śāstra is not obligatory.
• 3059 dhṛṣiśasī vaiyātye 7.2.19 — dhṛṣ and śas are aniṭ in the niṣṭhā only in the sense of impudence.
• The rule is restrictive: dhṛṣ (ādit, 3036) and śas (udit, by yasya vibhāṣā) would be aniṭ anyway.
• dhṛṣṭaḥ, viśastaḥ "ill-behaved"; elsewhere dharṣitaḥ, viśasitaḥ.
• Vṛtti: dhṛṣ is not used in the sense of impudence in bhāva or ādikarman. Haradatta calls dhṛṣ's ā-marking "to be considered". Mādhava gives an option in bhāva and ādikarman outside impudence: dhṛṣṭam / dharṣitam, pradhṛṣṭaḥ / pradharṣitaḥ.
• 3060 dṛḍhaḥ sthūlabalayoḥ 7.2.20 — dṛḍha is a nipātana meaning 'stout' and 'strong' (bala = balavat, by arśaādi ac).
• From dṛh and dṛhi + kta, the nipātana gives: no iṭ, ḍh for the t, loss of h, and for the idit dṛhi loss of the n: dṛḍhaḥ.
• Otherwise: dṛhitaḥ, dṛṃhitaḥ.
• Why the h-loss must be nipātana: a deletion taught in adhyāya 7 is not asiddha, so draḍhimā and paridraḍhayya come out right (BM, TB).
• 3061 prabhau parivṛḍhaḥ 7.2.21 — parivṛḍha means 'master', with the same nipātana ("prāgvat"): pari + vṛh + ta → parivṛḍhaḥ. Otherwise: parivṛhitaḥ, parivṛṃhitaḥ.
• 3062 kṛcchragahanayoḥ kaṣaḥ 7.2.22 — kaṣ is aniṭ in the niṣṭhā meaning 'painful' or 'impenetrable'.
• kaṣ + ta → kaṣṭaḥ: kaṣṭo mohaḥ (a cause of pain); kaṣṭaṃ śāstram (hard to penetrate).
• Otherwise: kaṣitam.
• 3063 ghuṣir aviśabdane 7.2.23 — ghuṣ is aniṭ in the niṣṭhā, except in the sense of making known by words.
• ghuṣṭā rajjuḥ "a rope made" or "stretched".
• In that sense: ghuṣitaṃ vākyam. TB: the exclusion itself shows that the curādi ṇic of that root is optional.
• 3064 ardeḥ saṃnivibhyaḥ 7.2.24 — ard after sam, ni or vi is aniṭ.
• sam + ard + ta: n replaces the t and the preceding d (SK 3016), then ṇatva: samarṇaḥ, nyarṇaḥ, vyarṇaḥ. (BM only in Wikisource.)
• Otherwise: arditaḥ.
• 3065 abheś cāvidūrye 7.2.25 — ard after abhi is aniṭ when the sense is 'not far'.
• abhyarṇam "not very far", or "not near" (TB keeps the two apart).
• Otherwise: abhyarditam.

III.I. The niṣṭhā from ṇic-stems, and the options of iṭ — running example: dami + kta → dāntaḥ / damitaḥ

• 3066 ṇer adhyayane vṛttam 7.2.26 — from vṛt + ṇic, in the sense 'studied', the nipātana gives kta without iṭ and with luk of ṇic.
• varti + ta → vṛt + ta → vṛttaṃ chandaś chātreṇa "the metre has been learned by the student".
• It is luk, not lopa, so there is no guṇa (BM, TB).
• Otherwise: vartitā rajjuḥ "a twisted rope".
• 3067 śṛtaṃ pāke 6.1.27 — before kta, śṛ replaces śrā and its causative śrapi, when milk or an oblation is cooked.
• śrā / śrapi + ta → śṛ + ta → śṛtaṃ kṣīram, either "softened of itself" or "cooked".
• Other things: śrāṇam, śrapitam.
• Where the restriction comes from: TB (the vyavasthita vibhāṣā of SK 3023) and BM (a vārttika) differ.
• 3068 vā dāntaśāntapūrṇadastaspaṣṭacchannajñaptāḥ 7.2.27 — from ṇic-stems ('ṇeḥ' from 3066, per BM), seven forms are optionally fixed.
• dami + ta → dāntaḥ (no iṭ; lengthened by SK 2666), śāntaḥ, pūrṇaḥ, dastaḥ, spaṣṭaḥ, channaḥ (n by SK 3016), jñaptaḥ.
• In the other alternative: damitaḥ, śamitaḥ, pūritaḥ, dāsitaḥ, spāśitaḥ, chāditaḥ, jñapitaḥ.
• 3069 ruṣyamatvarasaṃghuṣāsvanām 7.2.28 — after ruṣ, am, tvar, saṃ-ghuṣ and ā-svan, the niṣṭhā optionally takes iṭ. 'vā' continues from 3068 and 'ṇeḥ' stops (BM).
• ruṣṭaḥ / ruṣitaḥ; āntaḥ / amitaḥ.
• tvar + ta: ūṭh (SK 2654), then n (SK 3016) and ṇ: tūrṇaḥ / tvaritaḥ. tvar's ā-marking "has little purpose".
• saṃghuṣṭaḥ / saṃghuṣitaḥ; āsvāntaḥ / āsvanitaḥ.
• 3070 hṛṣer lomasu 7.2.29 — after hṛṣ, iṭ is optional when the hair is the agent: hṛṣṭaṃ / hṛṣitaṃ loma "the hair has bristled".
• Vārttika: also in the senses of astonishment and being thwarted: hṛṣṭo / hṛṣito maitraḥ.
• Elsewhere: hṛṣu alīke is aniṭ (udit), and hṛṣa tuṣṭau is seṭ.
• 3071 apacitaś ca 7.2.30 — apacita is an optional nipātana of cāy.
• apa + cāy + ta → apacitaḥ "honoured", or apacāyitaḥ.
• BM quotes the Bhāṣya's ci for the causative ciñ.

III.J. pī for pyāy, and a short vowel for hlād — running example: pyāy + kta → pī + na → pīnaḥ; pra + hlād + ta → prahlannaḥ

• 3072 pyāyaḥ pī 6.1.28 — pī optionally replaces pyāy before a niṣṭhā. BM: 'vibhāṣā' from SK 3023, 'niṣṭhāyām' from SK 3044.
• The option has a fixed domain (vyavasthita). For a body part it is obligatory: pīnaṃ sukham (print; see open items). Elsewhere: pyānaḥ / pīnaḥ svedaḥ.
• The n is by oditaś ca. In pyāna, iṭ is barred by 3039 and the y is deleted (TB).
• Not with an upasarga: prapyānaḥ. But with āṅ, for andhu and ūdhas: āpīno 'ndhuḥ "a well that has filled", āpīnam ūdhaḥ "a swollen udder".
• 3073 hlādo niṣṭhāyām 6.4.95 — a short vowel replaces the vowel of hlād before a niṣṭhā ('hrasvaḥ' from SK 2955, per BM).
• pra + hlād + ta → pra + hlad + ta. There is no iṭ (īdit, 3039), and n replaces both the t and the preceding d (SK 3016): prahlannaḥ "delighted".

III.K. Root changes before a kit t-initial affix: i, hi, dath, t, the long upasarga, jagdhi, long kṣī — running example: sthā + kta → sthi + ta → sthitaḥ; dā → dath + ta → dattaḥ; pra + dā + ta → pra + t + ta → prattaḥ

• 3074 dyatisyatimāsthām it ti kiti 7.4.40 — i replaces the final of do, so, mā and sthā before a t-initial kit affix. It is an exception to ītva (SK 2462) and to dad (SK 3077).
• do + ta → di + ta → ditaḥ; sitaḥ; mitaḥ (mā, māṅ and meṅ are all included); sthitaḥ.
• 3075 śācchor anyatarasyām 7.4.41 — the same i, optionally, for śo and cho.
• śitaḥ / śātaḥ "sharpened"; chitaḥ / chātaḥ "cut".
• The option is vyavasthita: for śo it is obligatory in respect of a vow: saṃśitaṃ vratam, saṃśito brāhmaṇaḥ.
• 3076 dadhāter hiḥ 7.4.42 — hi replaces dhā (dadhāti) before a kit t ('ti kiti' from 3074, per BM).
• abhi + dhā + ta → abhi + hi + ta → abhihitam "said"; nihitam "deposited".
• Naming the root with śtip excludes dheṭ (TB).
• 3077 do dad ghoḥ 7.4.46 — dath replaces the ghu dā before a kit t. Cartva then applies by khari ca (SK 121).
• dā + ta → dath + ta → dattaḥ. For the non-ghu dāp: dātaḥ.
• Alternatively the substitute ends in t. Then dasti does not lengthen in vidattam, since dasti needs a t-initial uttarapada.
• Or it ends in d or dh. The niṣṭhā-n or the dh that would then follow is prevented by the sannipāta paribhāṣā.
• 3078 aca upasargāt taḥ 7.4.47 — after a vowel-final upasarga, t replaces the vowel of the ghu dā. 'acaḥ' does double duty (BM, TB).
• pra + dā + ta → pra + d-t + ta, then cartva: prattaḥ, avattaḥ.
• Bhāṣya verse: avadattam, vidattam, pradattam (in ādikarman), and also sudattam, anudattam, nidattam.
• By 'ca' the form otherwise obtained (with dad) also holds.
• 3079 dasti 6.3.124 — an ik-final upasarga is lengthened before an uttarapada that begins with the t that replaced dā. BM, TB: 'ikaḥ' from SK 1045, 'upasargasya' from SK 1044, 'dīrghaḥ' from SK 174, 'uttarapade' from SK 958.
• ni + dā + ta → ni + t + ta → nīttam; sūttam (BM: doubtful).
• Cartva counts as siddha here, because this rule depends on the t it produces (āśrayāt).
• The kaumudī then lists more niṣṭhā forms. With ītva (SK 2462): dhītam (dheṭ), gītam, pītam. With ātva (SK 2504): jātam, sātam, khātam.
• 3080 ado jagdhir lyap ti kiti 2.4.36 — jagdhi replaces ad before lyap and before a kit t. Its i is only for pronunciation.
• ad + ta → jagdh + ta; dh replaces the t (SK 2280), and the first dh is optionally dropped by jharo jhari (SK 71): jagdhaḥ "eaten".
• Here the kaumudī reprises 3053: prakṛtaḥ kaṭaṃ saḥ / kaṭas tena. Then pra + kṣi + ta: lengthened by SK 3014, with n/ṇ by SK 3015: prakṣīṇaḥ saḥ "he has begun to waste away".
• 3081 vā''krośadainyayoḥ 6.4.61 — kṣi is optionally lengthened before a niṣṭhā (not in the ṇyat sense, from SK 3014, per BM) in imprecation and in wretchedness.
• kṣīṇāyur bhava "may your life waste away", or kṣitāyuḥ.
• kṣīṇo 'yaṃ tapasvī "this ascetic is wasted", or kṣitaḥ.

III.L. ṣ for s in niṣṇāta, pratiṣṇāta, kapiṣṭhala, viṣṭhala — running example: ni + snā + ta → ni-ṣnā-ta → niṣṇātaḥ

• 3082 ninadībhyāṃ snāteḥ kauśale 8.3.89 — after ni and nadī, ṣ replaces the s of snā when skill is meant. BM: 'saḥ' from SK 335; apadāntasya mūrdhanyaḥ (SK 210) is the governing rule.
• ni + snā + ta → niṣṇātaḥ śāstreṣu "expert in the śāstras". The ṇ comes by ṣṭutva (SK 113, left unstated).
• nadyāṃ snāti → nadī + snā + ka (by supi, SK 2916, split off as a rule of its own) → nadīṣṇaḥ "skilled at river-bathing".
• 3083 sūtraṃ pratiṣṇātam 8.3.90 — after prati, ṣ when the word qualifies sūtra: pratiṣṇātaṃ sūtram "clean". Otherwise: pratisnātam.
• 3084 kapiṣṭhalo gotre 8.3.91 — kapiṣṭhala when it names a gotra (the father of Kāpiṣṭhali). Otherwise: kapisthalam "the place of monkeys".
• 3085 vikuśamiparibhyaḥ sthalam 8.3.96 — after vi, ku, śami and pari, ṣ replaces the s of sthala, and th becomes ṭh by ṣṭutva: viṣṭhalam, kuṣṭhalam, śamiṣṭhalam, pariṣṭhalam.

III.M. Which sense the kta expresses: agent, locus, present, neuter bhāva — running example: gam + kta (kartari) → gaṅgāṃ gataḥ; ās + kta (adhikaraṇe) → mukundasyāsitam idam

• 3086 gatyarthākarmakaśliṣaśīṅsthā''savasajanaruhajīryatibhyaś ca 3.4.72 — kta comes in the agent after roots meaning going, after intransitive roots, and after śliṣ, śīṅ, sthā, ās, vas, jan, ruh and jṝ. It also comes in bhāva and karman. BM: 'kartari' from SK 3053, 'bhāve karmaṇi' from SK 2152.
• Examples: gaṅgāṃ gataḥ (nasal lost by SK 2428), gaṅgāṃ prāptaḥ, glānaḥ saḥ, lakṣmīm āśliṣṭo hariḥ, śeṣam adhiśayitaḥ, vaikuṇṭham adhiṣṭhitaḥ (i by 3074), śivam upāsitaḥ, haridinam upoṣitaḥ (iṭ by 3046), rāmam anujātaḥ, garuḍam ārūḍhaḥ, viśvam anujīrṇaḥ.
• śīṅ, sthā, ās and vas are named because with upasargas they take objects (SK 542, 544), so 'akarmaka' would not cover them (BM).
• In the other alternative: prāptā gaṅgā yena.
• 3087 kto 'dhikaraṇe ca dhrauvyagatipratyavasānārthebhyaḥ 3.4.76 — kta comes in the sense of the locus after roots meaning remaining still, going and eating. By 'ca' it also comes as otherwise obtained.
• Verse: mukundasyāsitam idam (the seat), idaṃ yātaṃ ramāpateḥ (the path), bhuktam etad anantasya (the place of eating). In each the agent is in the sixth case (SK 626).
• ās is intransitive, so it also takes the agent and bhāva: āsito mukundaḥ, āsitaṃ tena.
• Roots of going take the agent and the object: [ramāpatir idaṃ yātaḥ, tenedaṃ yātam] (supplied from the print).
• bhuj takes the object: anantenedaṃ bhuktam. bhuktā brāhmaṇāḥ is bhukta + the matvarthīya ac.
• What follows is under vartamāne (BM: by maṇḍūkapluti from SK 2151).
• 3088 ñītaḥ ktaḥ 3.2.187 — kta comes in the present after a ñīt root.
• ñikṣvidā → kṣviṇṇaḥ (no iṭ by 3036).
• ñiindhī → iddhaḥ (no iṭ by 3039; the nasal is lost).
• 3089 matibuddhipūjārthebhyaś ca 3.2.188 — kta comes in the present after roots meaning desire, understanding and honouring. It is in bhāva and karman, by SK 2833.
• mati means desire here, since buddhi is listed separately.
• rājñāṃ mataḥ, iṣṭaḥ "desired by kings". The sixth case is by SK 625. iṣṭaḥ is aniṭ by yasya vibhāṣā.
• buddhaḥ, viditaḥ; pūjitaḥ, arcitaḥ.
• 'ca' gathers in śīlitaḥ, rakṣitaḥ, kṣāntaḥ, ākruṣṭaḥ, juṣṭaḥ and so on (a Bhāṣya verse, per BM).
• 3090 napuṃsake bhāve ktaḥ 3.3.114 — kta comes in bhāva, in the neuter, in time in general: jalpitam "talk", śayitam "lying down", hasitam "laughter".
• BM: only after intransitive roots. In gataṃ haṃsasya the object is not intended.

III.N. Other past participles: ṅvanip and atṛn — running example: su + ṅvanip → su + t + van → sutvā; jṝ + atṛn → jar + at → jaran (beside jīrṇaḥ)

• 3091 suyajor ṅvanip 3.2.103 — ṅvanip comes in the past after su (sunoti) and yaj: sutvā, sutvānau; yajvā, yajvānau.
• The ṅ blocks guṇa. The p is for the accent and for tuk (TB).
• Being ṅit and not kit, it causes no saṃprasāraṇa in yajvā (SK 2409; TB).
• 3092 jīryater atṛn 3.2.104 — atṛn comes in the past ('bhūte' ity eva) after jṝ.
• jṝ + at, with guṇa and with num because the affix is ugit (SK 361): jaran, jarantau, jarantaḥ. There is no lengthening, because the loss of the conjunct-final t is asiddha (TB).
• By the vāsarūpa principle (SK 2830) the niṣṭhā also comes: jīrṇaḥ, jīrṇavān.

Part IV (SK 3093–3167): liṭ's participles, śatṛ/śānac and sat, the tācchīlika affixes, ṣṭran and itra

*Covers SK 3093–3167 (75 sūtras): all present, each on its own line (coverage checked by script). SK 3093 and 3094 have no kaumudī of their own (ours and the print): Bhaṭṭoji explains them together with 3095 under sk.3095.k1, and BM does the same (sk.3093.b1 and sk.3094.b1 are pratīka only). Every AS-ref is checked against references/ashtadhyayi/sutraani-data.json; all 75 match. One sūtra text differs: at 3119 the reference reads glājisthaś ca ksnuḥ, which is BM's reading (k = cartva of g); ours and the print have gsnuḥ.*

• Where Part IV begins. Part III ended at 3092 jīryater atṛn (3.2.104): jaran, still under bhūte (3.2.84, SK 2995). Part IV opens at 3093 chandasi liṭ (3.2.105), with the same past-tense heading, and then runs the AS in order through 3.2.105–3.2.186. 3.2.110–3.2.123 are not here (they are the lakāra rules, e.g. luṅ SK 2218 and vartamāne laṭ SK 2151). Bhaṭṭoji inserts seven rules from other books at the point where they are first needed:
• 7.2.67 (3096) after kvasuś ca, and 7.2.68 (3099) after the bhāṣā kvasu rules: iṭ for vasu;
• 7.2.82–83 (3101, 3104) among the śatṛ/śānac rules: muk, and ī for āna after ās;
• 7.1.36 (3105): vasu for the śatṛ of vid;
• 3.3.14 (3107) after tau sat: sat for lṛṭ;
• 7.2.9 (3163) after the ṣṭran rule: no iṭ before tra etc.
• Where Part IV ends. At 3167 kartari carṣidevatayoḥ (3.2.186), the last itra rule (pavitram). This also ends the pūrvakṛdanta. The 1915 print closes "iti pūrvakṛdantaṃ samāptam ‖" after 3167 (NSP p. 516, scan page-0526). Ours lacks the colophon, so the English supplies it in brackets: "[Here the pūrvakṛdanta is concluded.]" TB's own colophon "iti pūrvakṛdantam" stands in sk.3161.t1, followed by the start of TB's Uṇādi. sk.3167.b1 ends with BM's colophon and a stray BM rudhādi opening. p68 (uṇādiprakaraṇam) follows, opening at sk3168 (na raparasṛpi…, 8.3.110) and the uṇādi sūtras.

Contents of Part IV

• IV.A. 3093–3095, 3097–3098: liṭ in the Veda and its participles kānac and kvasu; kvasu in the bhāṣā. tasthivāṃsam, niṣeduṣīm, śuśruvān, upeyivān, anāśvān, anūcānaḥ.
• IV.B. 3096, 3099: the iṭ of vasu. ādivān, dadivān, jakṣivān; jagmivān/jaganvān; dadṛśivān/dadṛśvān.
• IV.C. 3100–3104: śatṛ and śānac for laṭ. pacantam, pacamānam, san brāhmaṇaḥ, he pacan, śayānā bhuñjate, āsīnaḥ.
• IV.D. 3105–3107: vasu for vid's śatṛ, the name sat, sat for lṛṭ. vidvān, viduṣī; kariṣyantam/kariṣyamāṇam.
• IV.E. 3108–3113: independent present affixes. śānan (yajamānaḥ), cānaś (bhogaṃ bhuñjānaḥ), and śatṛ for iṅ, dhāri, dviṣ, suñ, arh (adhīyan, dviṣan, arhan).
• IV.F. 3114–3120: the tācchīlika heading; tṛn, iṣṇuc, gsnu, knu. kartā kaṭān, alaṃkariṣṇuḥ, jiṣṇuḥ, sthāsnuḥ, gṛdhnuḥ.
• IV.G. 3121–3127: ghinuṇ and vuñ. śamī, pramādī, unmādī, saṃparkī, pralāpī; nindakaḥ, ādevakaḥ.
• IV.H. 3128–3133: yuc and its limits. calanaḥ, vartanaḥ, caṅkramaṇaḥ, krodhanaḥ; knūyitā, sūditā; kamrā/kamanā.
• IV.I. 3134–3138: ukañ, ṣākan, ini, āluc. pātukaḥ, jalpākaḥ/varākī, prajavī, jayī, spṛhayāluḥ, śayāluḥ.
• IV.J. 3139–3144: ru, kmarac, ghurac, kurac, kvarap, gatvara. dāruḥ, ghasmaraḥ, bhaṅguraḥ, viduraḥ, naśvaraḥ, gatvarī.
• IV.K. 3145–3147: ūka and ra. jāgarūkaḥ, yāyajūkaḥ, namraḥ, ajasram.
• IV.L. 3148–3151: u, and ki/kin treated like liṭ. cikīrṣuḥ, bhikṣuḥ, vinduḥ, icchuḥ, sumnayuḥ; papiḥ, jagmiḥ, cakriḥ, nemiḥ.
• IV.M. 3152–3156: najiṅ, āru, kru/klukan, varac. svapnak, tṛṣṇak, vandāruḥ, bhīruḥ, īśvaraḥ, yāyāvaraḥ.
• IV.N. 3157–3160: kvip, which closes the tācchīlika range; kvip after other roots; bhū with kvip or ḍu. vibhrāṭ, vāk, didyut, juhūḥ, dhīḥ, pratibhūḥ, prabhuḥ, śambhuḥ.
• IV.O. 3161–3167: ṣṭran and itra, for the object, the instrument and the agent, with the iṭ-bar. dhātrī, dātram, śastram, potram, khanitram, pavitram.

Ordering. SK order throughout except one move: 3096 (vasv ekājādghasām) is outlined in IV.B with 3099, the other iṭ rule for vasu. In the SK it stands between kvasuś ca (3095) and bhāṣāyāṃ sadavasaśruvaḥ (3097).

Threads that run through Part IV:

• The headings.
• bhūte (3.2.84, SK 2995) governs 3093–3099. In the Veda it is the past "in general" (bhūtasāmānye): the anadyatana/parokṣa conditions of parokṣe liṭ (SK 2171) are not intended (BM).
• From 3100 the affixes replace or accompany the laṭ of vartamāne laṭ (SK 2151). 3108 supplies "vartamāne" from it (BM).
• ā kves tacchīla… (3114) governs 3115 up to and including the kvip of 3157 (āṅ as abhividhi, BM/TB). BM and the en.yml still read the tacchīla senses into 3158.
• kartari kṛt (3.4.67, SK 2832) gives the agent sense by default. 3161 (karmaṇi), 3162–3166 (karaṇe) and 3167 (karaṇa for a ṛṣi) override it.
• Substitute of l, or independent affix.
• Substitutes of l (lādeśa): kvasu, kānac (3094–3095); śatṛ, śānac (3100); vasu (3105); sat for lṛṭ (3107). Because the vāsarūpa option does not hold for lādeśas, the "vā" of 3094 has to be stated (BM on 3097).
• Independent affixes: śānan (3108), cānaś (3109), and the śatṛ of 3110–3113 (BM, TB). So śānan comes only in the agent. cānaś, not being ātmanepada, also follows parasmaipadin han (śatrūn nighnānaḥ).
• śit, so sārvadhātuka. śatṛ and śānac take śap (pacantam, 3100). āna after an a-final aṅga takes muk (pacamāna, yajamāna, pavamāna: 3101, 3108). cānaś does not follow a, so it has no muk (bhuñjānaḥ).
• What the it-markers do in the tācchīlika affixes. The kaumudī lists forms bare; BM and TB name the markers.
• k/g/ṅ-it, so no guṇa: gsnu (jiṣṇuḥ; a g read into kṅiti ca, SK 2217), knu, kurac, kvarap, kru, klukan, najiṅ.
• gh, so kutva by cajoḥ ku ghiṇṇyatoḥ (SK 2863): saṃparkī, bhaṅguraḥ.
• ñ/ṇ, so vṛddhi: ukañ (pātukaḥ), vuñ (ākrośakaḥ), ghinuṇ (pramādī).
• p, so tuk: kvarap (itvaraḥ), kvip (grāvastut).
• ṣ, so ṅīṣ: ṣākan (varākī). ṣṭran gives dhātrī, but daṃṣṭrā is a disputed exception (3163).
• u-it: num/ṅīp for śatṛ and vasu (pacantam, viduṣī). For ghinuṇ, on the Bhāṣya's view, it gives optional shortening by ugitaś ca (śaminitarā / śaminītarā, 3121).
• vāsarūpa in the tācchīlika domain (vā 'sarūpo 'striyām, SK 2830). Each step rests on a jñāpaka.
• 3126: the statement of vuñ, though ṇvul (SK 2895) would give the same forms, shows that the general affixes tṛc etc. do not come in the tacchīla senses.
• 3130: naming pad, though 3129 already gives it yuc, shows that the tācchīlika affixes do not alternate with one another. Hence no tṛn after alaṃkṛñ.
• 3133: naming dīp, though ra (3147) already blocks yuc, shows that this bar is only generally valid (prāyika). Hence kamrā/kamanā, kamprā/kampanā; unmādī beside unmadiṣṇuḥ (3121); gantā beside gāmukaḥ (TB, 3134).
• 3158: the vārttika kvib vaci… is stated so that tṛn does not block (BM).
• Upasargas. In this range the tācchīlika affixes generally do not come after a root with an upasarga (BM on 3114, from the Bhāṣya). The exceptions are named in the sūtras: alaṃkṛñ, nirākṛñ … (3116); saṃpṛc … (3122); vi- (3123); apa-/vi- (3124); pra- (3125); an upasarga with devi and kruś (3127).
• Veda and bhāṣā.
• Vedic: 3093–3095 (but "poets use them widely", 3095), 3117–3118 and 3150–3151.
• "chandasi" ends at 3119 (BM: by explanation).
• The two Vedic sūtras 3150–3151 are cited to set out the bhāṣā vārttika on ki/kin (BM, TB).
• The kaumudī lists forms bare. The derivation steps below (reduplication, saṃprasāraṇa, ṣatva, guṇa, kutva …) are BM's and TB's, as recorded in the unit notes. The commentary filing is irregular in this stretch:
• TB on 3093–3097 is in sk.3091.t1; on 3108–3109 in sk.3105.t1; on 3118–3120 in sk.3117.t1; on 3131–3137 in sk.3130.t1; on 3163–3167 in sk.3161.t1.
• sk.3135.b1 is BM on 3166, and BM on 3135 survives only in Wikisource.

Audit status. The finished audit reports in .work/sk-run/p67-audit/ (A–L) stop at SK 3050, so no audit correction touches Part IV. This part follows the en.yml and the T19–T25 entries in .work/sk-run/p67-notes.md.

Print readings (NSP, scan-confirmed) that the English follows, with textual notes:

• 3101 māṅy (ours māḍy);
• 3109 kavacaṃ bibhrāṇaḥ (ours bibhrāṇam);
• 3116 utpatiṣṇuḥ; 3118 aprabhaviṣṇu vaiṣṇavam (no visarga);
• 3121 ugitaś ca, [śamī] supplied, jhalantānām;
• 3122 abhyāghātī (ours abhyaghātī; a new slip, not in the collation);
• 3126 vāsarūpa-; 3130 laṣapatapada (ours -pata; new);
• 3146/3151 yaṅanta- (ours yaṅnta-);
• 3149 iṣeś chatvam, vinduḥ;
• 3156 ajādeśo, sthānivattvam;
• 3160 vibhur vyāpakaḥ (ours vibhūḥ);
• 3161 dhāñaś (ours ghāñaś).

Readings where ours is kept, the print being in error:

• 3099: the print repeats jaganvān for jaghanvān.
• 3123 and 3124: kaṣ- and vilāṣī stand where the print shows p; this is the ṣ/p limit of its type.

Still open for Ajai (mūla):

• 3101 k3: ours and the print both read "etadvacanasāmarthyāl luk", which makes no sense. BM requires laṭ, and the English supplies [laṭ].
• 3101 k1: ours = print = TB lemma "aṅgasyātaḥ". Wikisource and BM's lemma presuppose "adantasyāṅgasya".
• 3115: kartā kaṭān (ours = print = TB) vs kaṭam (BM lemma, Wikisource).
• 3116: the full list of 12 forms vs "alaṃkariṣṇur ityādi" (BM/TB lemmas, Wikisource).
• 3138: tandrāluḥ, with no śraddhāluḥ (ours = print) vs BM's lemma "śraddhālur iti".
• 3143: sūtra naśji (ours) vs naśaji (print) vs naśiji (BM, Wikisource).
• 3145: sūtra jāgurūkaḥ (ours; both commentaries need "jāguḥ") vs the print's jāgarūkaḥ.
• 3156: "ato lopaḥ 2380" in ours and the print. It should be SK 2308; the English gives 2308.
• 3158: vārttika kib (print) vs kvib (ours).
• 3165: dhavitram (ours = print kaumudī) vs dhuvitram (TB's argument, Wikisource).
• 3167: the missing colophon, supplied in brackets.

IV.A. liṭ in the Veda and its participles kānac and kvasu; kvasu in the bhāṣā — running example: ni + sad + liṭ (bhāṣā, past) → kvasu → (reduplication; e + abhyāsa-loss) ni-sed-vas → (ṅīp; vas → us; ṣ) niṣeduṣīm

• 3093 chandasi liṭ 3.2.105 — in the Veda, liṭ comes in the past in general (bhūtasāmānye). "bhūte" comes from SK 2995. The anadyatana/parokṣa conditions of SK 2171 do not apply (BM). TB: this is not already covered by chandasi luṅlaṅliṭaḥ (SK 3423), which needs a dhātusaṃbandha. No kaumudī of its own. This is the liṭ that 3094–3095 replace.
• 3094 liṭaḥ kānaj vā 3.2.106 — kānac optionally replaces that liṭ (in the Veda, in the past). "liṭ" is restated so that any liṭ is taken, including that of parokṣe liṭ (BM). TB: kānac's k-marker serves tistirāṇa, blocking the guṇa of ṛcchatyṝtām (SK 2383). In cakrāṇa kit-ness already follows from asaṃyogāl liṭ kit (SK 2242).
• 3095 kvasuś ca 3.2.107 — kvasu also replaces it. It is split off from 3094 so that kvasu alone carries into 3097 (BM). Kaumudī: kvasu and kānac are Vedic only (trimunimatam), "but poets use them widely":
• sthā + kvasu → tasthivāṃsam: lengthening by atvasantasya (SK 425), num because ugit;
• adhi + gam + kvasu → adhijagmuṣaḥ: upadhā-loss by SK 2363; saṃprasāraṇa by vasoḥ saṃprasāraṇam (SK 435) before śas (BM).
• 3097 bhāṣāyāṃ sadavasaśruvaḥ 3.2.108 — in the bhāṣā, after sad, vas, śru, liṭ comes optionally in the past in general, and kvasu necessarily replaces it. "vā" goes with liṭ; in the other option luṅ comes (BM).
• ni + sad → ni-sed-vas (e and abhyāsa-loss by ata ekahalmadhye…), then ṅīp (ugit), saṃprasāraṇa of vas, and ṣ by sadir aprateḥ (SK 2271): niṣeduṣīm. There is no iṭ, since no val-initial affix remains (TB); masculine niṣedivān.
• adhi + vas → adhyūṣuṣaḥ: yajādi saṃprasāraṇa, SK 2409. TB reads it before ṅas, BM before śas.
• śru → śuśruvān.
• 3098 upeyivān anāśvān anūcānaś ca 3.2.109 — three nipātanas:
• upa + iṇ: liṭ even in the bhāṣā, kvasu, and iṭ → upeyivān. This iṭ is the val-conditioned one, which 3096 had denied and this rule re-allows (BM). It does not come in upeyuṣī, where saṃprasāraṇa removes the v. "upa" is not a condition (vyākhyāna): īyivān, samīyivān.
• nañ + aś (aśnāti) + kvasu, without iṭ → anāśuṣaḥ (Bhāravi).
• anu + vac + kānac (agent; saṃprasāraṇa) → anūcānaḥ, "one who has completed the recitation of the Veda".

IV.B. The iṭ of vasu — running example: ad + kvasu → (reduplication, halādiśeṣa, lengthening, savarṇadīrgha) ād-vas → (still ekāc) ād-i-vas → ādivān; gam → jagmivān / jaganvān

• 3096 vasv ekājādghasām 7.2.67 — vasu takes iṭ only after roots that are ekāc even after reduplication, after ā-final roots, and after ghas. BM: neḍ vaśi kṛti (SK 2981) is overridden by the kṛādi-niyama, and this sūtra then restricts (niyama) the iṭ so restored.
• ekāc: ādivān (ad), ārivān (ṛ, guṇa by SK 2383).
• ā-final: dadivān. dad is not ekāc, and the ā is lost only before the iṭ.
• ghas (for ad by SK 2424): jakṣivān.
• Counter: babhūvān.
• SK order: this rule stands between 3095 and 3097; it is moved here.
• 3099 vibhāṣā gamahanavidaviśām 7.2.68 — after gam, han, vid, viś, iṭ is optional: jagmivān / jaganvān, jaghnivān / jaghanvān, vividivān / vividvān, viviśivān / viviśvān.
• With iṭ, the upadhā is lost by SK 2363 (TB).
• vid here is vindati's (sāhacarya with viś). vetti gives only vividvān, since SK 2981 bars iṭ.
• Vārttika dṛśeś ca: dadṛśivān / dadṛśvān. (The print repeats jaganvān for jaghanvān; this is its slip.)

IV.C. śatṛ and śānac for laṭ: the present participles — running example: pac + laṭ → śatṛ → pac + śap + at → pacat (num) → pacantaṃ caitraṃ paśya; → śānac → paca + āna → (muk) pacamānam

• 3100 laṭaḥ śatṛśānacāv aprathamāsamānādhikaraṇe 3.2.124 — śatṛ and śānac replace laṭ (the laṭ of vartamāne laṭ, SK 2151, by "anantarasya": BM) when it is co-referent with a non-first-case word.
• The it-letters are ś, ṛ and ś, c. Being śit they are sārvadhātuka (SK 2166), so śap comes; num because ugit: pacantaṃ caitraṃ paśya.
• pacantam = "the substratum of the activity of cooking" (BM, TB).
• 3101 āne muk 7.2.82 — when āna follows, an a-final aṅga takes muk ("ataḥ" from ato yeyaḥ, SK 2212). Being kit, it goes at the end: paca + m + āna → pacamānaṃ caitraṃ paśya.
• Kaumudī: laṭ is restated in 3100 to prescribe more (adhikavidhāna). So śatṛ/śānac sometimes come even with a first-case word: as + śatṛ → san brāhmaṇaḥ, with a lost by śnasor allopaḥ (BM).
• Vārttika māṅy ākrośe: mā jīvan yaḥ…, "he is not living who…". māṅi luṅ (SK 2219) would give luṅ, and [laṭ] comes by force of the statement (open: ours and the print read "luk").
• 3102 sambodhane ca 3.2.125 — also in address: he pacan, he pacamāna. BM: since 3100's restated laṭ already covers the first case, this is elaboration (prapañcārtha).
• 3103 lakṣaṇahetvoḥ kriyāyāḥ 3.2.126 — when the root-meaning is a mark (lakṣaṇa) or cause (hetu) of another action:
• lakṣaṇa: śayānā bhuñjate yavanāḥ (lying down marks the eaters);
• hetu as result: arjayan vasati, "he dwells in order to earn";
• hetu as producing cause: hariṃ paśyan mucyate.
• Across padas, kṛtyacaḥ (SK 2835) gives ṇ: prapīyamāṇaḥ somaḥ.
• 3104 īdāsaḥ 7.2.83 — after ās, ī replaces the first sound of āna (ādeḥ parasya, SK 44): ās + āna → ās + īna → āsīnaḥ.

IV.D. vasu for vid's śatṛ, the name sat, and sat for lṛṭ — running example: kṛ + lṛṭ → sya + śatṛ / śānac → kariṣya-t / kariṣya-m-āṇa → kariṣyantaṃ kariṣyamāṇaṃ paśya; vid + śatṛ → vasu → vidvān

• 3105 videḥ śatur vasuḥ 7.1.36 — after vid (vetti only, BM), vasu optionally replaces śatṛ. "anyatarasyām" comes from SK 2197.
• vidan (no substitute) / vidvān; feminine viduṣī (ṅīp, ugit; saṃprasāraṇa by SK 435; ṣ).
• TB: the explicit u-marker lets vasoḥ saṃprasāraṇam cover kvasu too.
• 3106 tau sat 3.2.127 — śatṛ and śānac are named sat (saṃjñā). The name feeds 3107.
• 3107 lṛṭaḥ sad vā 3.3.14 — sat optionally replaces lṛṭ (BM: only the lṛṭ of SK 2193). It is a fixed option (vyavasthita-vibhāṣā): sat is obligatory in the environments of 3100–3103.
• with a non-first case: kariṣyantaṃ kariṣyamāṇaṃ paśya;
• before an affix: kāriṣyataḥ;
• before an uttarapada: kariṣyadbhaktiḥ (BM: karmadhāraya; TB's parallel under 3101 is tatpuruṣa);
• in address: he kariṣyan;
• hetu: arjayiṣyan vasati.
• Sometimes it comes even with a first-case word: kariṣyati iti kariṣyan.

IV.E. Independent present affixes: śānan, cānaś, and śatṛ in restricted senses — running example: yaj + śānan → yaja + āna → (muk) yajamānaḥ; adhi + iṅ + śatṛ → adhīyan

• 3108 pūṅyajoḥ śānan 3.2.128 — after pūṅ, yaj, śānan comes in the present ("vartamāne" from SK 2151). It is not a lādeśa, so it comes only in the agent (BM). TB: the ś gives sārvadhātuka status, the n the accent. pū → pavamānaḥ, yaj → yajamānaḥ, both with śap and muk.
• 3109 tācchīlyavayovacanaśaktiṣu cānaś 3.2.129 — cānaś comes when disposition, age or capacity is meant:
• bhogaṃ bhuñjānaḥ (disposition), kavacaṃ bibhrāṇaḥ (age; print), śatrūn nighnānaḥ (capacity).
• Not a lādeśa, so not ātmanepada: it follows parasmaipadin han. No muk, since it does not follow a (BM, TB).
• 3110 iṅdhāryoḥ śatr akṛcchriṇi 3.2.130 — after iṅ and dhāri, śatṛ comes when the action is easy for the agent: adhīyan, dhārayan. Counter: kṛcchreṇādhīte, dhārayati. TB: the rule is needed because iṅ is ātmanepadin.
• 3111 dviṣo 'mitre 3.2.131 — śatṛ after dviṣ for an enemy: dviṣan.
• 3112 suño yajñasaṃyoge 3.2.132 — śatṛ after su(ñ) for pressing connected with a sacrifice: sarve sunvantaḥ sarve yajamānāḥ satriṇaḥ. BM: "saṃyoga" takes in the master of the sacrifice; the plural is possible only in a sattra.
• 3113 arhaḥ praśaṃsāyām 3.2.133 — śatṛ after arh in praise: arhan. TB's counter: arhati cauro vadham.

IV.F. The tācchīlika heading; tṛn, iṣṇuc, gsnu, knu — running example: kṛ + tṛn → (guṇa) kar-tṛ → kartā (kaṭān); alam + kṛ + iṣṇuc → alaṃkariṣṇuḥ; ji + gsnu → (g-it: no guṇa) jiṣṇuḥ

• 3114 ā kves tacchīlataddharmatatsādhukāriṣu 3.2.134 — heading: up to and including the kvip of 3157, the affixes come in agents whose habit (tacchīla), duty (taddharma) or skill (tatsādhukārin) is that action. BM: tacchīla acts by nature, without regard to a result; taddharman acts from a sense of duty.
• 3115 tṛn 3.2.135 — tṛn: kartā kaṭān. The object stays in the second case because in na lokāvyaya… (SK 627) "tṛn" is a pratyāhāra (BM, TB). BM's taddharma example: muṇḍayitāraḥ śrāviṣṭhāyanāḥ.
• 3116 alaṃkṛñnirākṛñ…sahacara iṣṇuc 3.2.136 — iṣṇuc after twelve bases: alaṃkariṣṇuḥ, nirākariṣṇuḥ, prajaniṣṇuḥ, utpatiṣṇuḥ, utpaciṣṇuḥ, unmadiṣṇuḥ, rociṣṇuḥ, apatrapiṣṇuḥ, vartiṣṇuḥ, vardhiṣṇuḥ, sahiṣṇuḥ, cariṣṇuḥ. kṛñ = ḍukṛñ karaṇe only (TB).
• 3117 ṇeś chandasi 3.2.137 — in the Veda, iṣṇuc after a ṇi-final root: pṛ + ṇi + iṣṇuc → (ay for ṇi, SK 2311) pārayiṣṇavaḥ.
• 3118 bhuvaś ca 3.2.138 — in the Veda, iṣṇuc after non-causative bhū: bhaviṣṇuḥ.
• aprabhaviṣṇu vaiṣṇavam outside the Veda: "poets are unrestrained".
• The Vṛtti takes "ca" to gather in bhrājiṣṇuḥ and kṣayiṣṇuḥ; this "is not seen in the Bhāṣya".
• 3119 glājisthaś ca gsnuḥ 3.2.139 — "chandasi" ceases. gsnu after glā, ji, sthā, and by "ca" bhū:
• g-it, not k-it, so sthāsnuḥ gets no ī by SK 2462.
• A g is read into kṅiti ca (BM), so there is no guṇa: jiṣṇuḥ; also glāsnuḥ.
• A g is read into śryukaḥ kiti (SK 2381), so there is no iṭ: bhūṣṇuḥ.
• Vārttika daṃśeś chandasy upasaṃkhyānam: daṅkṣṇavaḥ paśavaḥ. The nasal stays because gsnu is not kit/ṅit (TB).
• 3120 trasigṛdhidhṛṣikṣipeḥ knuḥ 3.2.140 — knu: trasnuḥ (no iṭ, SK 2981), gṛdhnuḥ, dhṛṣṇuḥ (k-it: no guṇa), kṣipnuḥ.

IV.G. ghinuṇ and vuñ — running example: śam + ghinuṇ → śam-in → (upadhā-vṛddhi barred, SK 2763) śamī; nind + vuñ → nind-aka → nindakaḥ

• 3121 śamity aṣṭābhyo ghinuṇ 3.2.141 — ghinuṇ after the eight divādi roots śam etc. ("iti" = "ādi").
• Its u is an it-marker (Bhāṣya), not merely for pronunciation (Kāśikā). So in śaminitarā / śaminītarā the shortening is optional by ugitaś ca (SK 987). num in śaminau is avoided by drawing "jhal" back from SK 314.
• Vṛddhi is barred by nodāttopadeśasya… (SK 2763): śamī, tamī, damī, śramī, bhramī, kṣamī, klamī. But pramādī, since mad is not m-final (BM).
• ut + mad: iṣṇuc by 3116, and also ghinuṇ by vāsarūpa: unmādī. The tācchīlika bar on vāsarūpa is only prāyika.
• 3122 saṃpṛcānurudhāṅyamāṅyasa…abhyāhanaś ca 3.2.142 — ghinuṇ after 27 bases.
• saṃparkī (k by cajoḥ ku ghiṇṇyatoḥ, SK 2863), anurodhī, āyāmī, āyāsī, parisārī, saṃsargī, paridevī (devṛ, not the causative of dīv), saṃjvārī, … tyāgī, rāgī, bhāgī, aticārī, apacārī, āmoṣī.
• abhyāghātī: t by SK 2574, gh by SK 358. This is the print's reading; ours has abhyaghātī.
• 3123 vau kaṣalasakatthasrambhaḥ 3.2.143 — with vi: vikāṣī, vilāsī, vikatthī, visrambhī.
• 3124 ape ca laṣaḥ 3.2.144 — laṣ with apa, and by "ca" with vi: apalāṣī, vilāṣī.
• 3125 pre lapasṛdrumathavadavasaḥ 3.2.145 — with pra: pralāpī, prasārī, pradrāvī, pramāthī, pravādī, pravāsī.
• 3126 nindahiṃsakliśakhādavināśaparikṣipapariraṭaparivādivyābhāṣāsūyo vuñ 3.2.146 — vuñ after ten bases (the first case stands for the fifth): nindakaḥ, hiṃsakaḥ, etc.
• vināśa and parivādi are causatives; asūy is a kaṇḍvādi (BM).
• ṇvul (SK 2895) would already give these forms. So stating vuñ is a jñāpaka: in the tacchīla senses, tṛc and the other general affixes do not come by vāsarūpa (SK 2830).
• 3127 devikruśoś copasarge 3.2.147 — vuñ after devi (ṇi-final div) and kruś, with an upasarga: ādevakaḥ, ākrośakaḥ. Without one, tṛn: devayitā, kroṣṭā.

IV.H. yuc and its limits — running example: cal + yuc → (yu → ana, SK 1247) cal-ana → calanaḥ; vṛt (anudātteta) + yuc → vartanaḥ

• 3128 calanaśabdārthād akarmakād yuc 3.2.148 — yuc after an intransitive root meaning motion or sound: calanaḥ, copanaḥ, kampanaḥ; śabdanaḥ, ravaṇaḥ. Counter (transitive): paṭhitā vidyām (tṛn).
• 3129 anudāttetaś ca halādeḥ 3.2.149 — yuc after an intransitive anudātteta, hal-initial root: vartanaḥ, vardhanaḥ.
• Counters: bhavitā (not anudātteta), edhitā (not hal-initial), vasitā vastram (transitive).
• BM: the word ādi lets the rule cover jugupsanaḥ.
• 3130 jucaṅkramyadandramyasṛgṛdhijvalaśucalaṣapatapadaḥ 3.2.150 — yuc after ten bases. ju is sautra ("going, speed"): javanaḥ, caṅkramaṇaḥ (y-loss by SK 2631, a-loss by SK 2308), dandramaṇaḥ, saraṇaḥ.
• pad already has yuc by 3129. Naming it here keeps it from being blocked by the ukañ of 3134. Jñāpaka: the tācchīlika affixes do not alternate with one another by vāsarūpa. Hence no tṛn after alaṃkṛñ (only iṣṇuc).
• 3131 krudhamaṇḍārthebhyaś ca 3.2.151 — yuc after roots meaning anger or adorning: krodhanaḥ, roṣaṇaḥ; maṇḍanaḥ, bhūṣaṇaḥ.
• 3132 na yaḥ 3.2.152 — no yuc after a y-final root, against 3129: knūyitā, kṣmāyitā (tṛn). TB: "na yaḥ" is not nayati.
• 3133 sūdadīpadīkṣaś ca 3.2.153 — no yuc after sūd, dīp, dīkṣ: sūditā, dīpitā, dīkṣitā.
• The ra of 3147 already blocks yuc after dīp, yet dīp is named. This shows the mutual bar is prāyika: kamrā / kamanā yuvatiḥ, kamprā / kampanā śākhā.
• madhusūdanaḥ is by the nandi list (lyu, SK 2896).

IV.I. ukañ, ṣākan, ini, āluc — running example: pat + ukañ → (vṛddhi) pāt-uka → pātukaḥ; jalp + ṣākan → jalpākaḥ (fem. varākī); ji + ini → (guṇa, ay) jayī

• 3134 laṣapatapadasthābhūvṛṣahanakamagamaśṝbhya ukañ 3.2.154 — ukañ: lāṣukaḥ, pātukaḥ, and so on.
• BM's "and so on": pādukaḥ; sthāyukaḥ (yuk, SK 2761); bhāvukaḥ, varṣukaḥ; ghātukaḥ (SK 2574, 358).
• TB: gantā beside gāmukaḥ (prāyika).
• 3135 jalpabhikṣakuṭṭaluṇṭavṛṅaḥ ṣākan 3.2.155 — ṣākan: jalpākaḥ, bhikṣākaḥ, kuṭṭākaḥ, luṇṭākaḥ, varākaḥ. The ṣ gives ṅīṣ (SK 498): varākī. TB reports the view that ṣākan is taught where yuc, iṣṇuc or ki/kin would otherwise come.
• 3136 prajor iniḥ 3.2.156 — ini after pra + ju (sautra): prajavī, prajavinau, prajavinaḥ.
• 3137 jidṛkṣiviśrīṇvamāvyathābhyamaparibhūprasūbhyaś ca 3.2.157 — ini after ten more bases: jayī, darī, kṣayī, viśrayī, atyayī, vamī, avyathī, abhyamī, paribhavī, prasavī.
• In avyatha, nañ is compounded with the root by nipātana (BM).
• TB: praju should have been read here and 3136 dropped.
• 3138 spṛhigṛhipatidayinidrātandrāśraddhābhya āluc 3.2.158 — āluc:
• The three curādi adanta roots: spṛhayāluḥ, gṛhayāluḥ, patayāluḥ (ay for ṇi, SK 2311). The lost a is sthānivat, so there is no guṇa or vṛddhi (BM, TB).
• dayāluḥ, nidrāluḥ; tad + drā → tandrāluḥ, with tad n-final by nipātana.
• Vārttika śīṅo vācyaḥ: śayāluḥ.
• śraddhāluḥ is absent from ours and the print (BM's lemma presupposes it).

IV.J. ru, kmarac, ghurac, kurac, kvarap, and gatvara — running example: naś + kvarap → naś-vara → naśvaraḥ; iṇ + kvarap → (k-it: no guṇa; p-it: tuk) i-t-vara → itvaraḥ

• 3139 dādheṭsiśadasado ruḥ 3.2.159 — ru: dāruḥ, dhāruḥ, seruḥ, śadruḥ, sadruḥ. TB: dā = ḍudāñ, do, deṅ only.
• 3140 sṛghasyadaḥ kmarac 3.2.160 — kmarac: sṛmaraḥ, ghasmaraḥ, admaraḥ. ghas is a separate base, not ad's substitute (BM).
• 3141 bhañjabhāsamido ghurac 3.2.161 — ghurac: bhañj → (kutva, SK 2863) bhaṅguraḥ, bhāsuraḥ, meduraḥ.
• 3142 vidibhidicchideḥ kurac 3.2.162 — kurac (k-it: no guṇa): viduraḥ (vid "to know"), bhiduram, chiduram.
• 3143 iṇnaśjisartibhyaḥ kvarap 3.2.163 — kvarap: itvaraḥ (tuk, SK 2858), itvarī (ṅīp, SK 470), naśvaraḥ, jitvaraḥ, sṛtvaraḥ.
• 3144 gatvaraś ca 3.2.164 — gam + kvarap, with the nasal lost by nipātana (no jhal follows): gatvaraḥ, gatvarī.

IV.K. ūka and ra — running example: jāgṛ + ūka → (guṇa + r) jāgar-ūka → jāgarūkaḥ; yaj + yaṅ → yāyaj(ya) + ūka → yāyajūkaḥ

• 3145 jāgurūkaḥ 3.2.165 — ūka after jāgṛ ("jāguḥ" is its fifth case): jāgarūkaḥ. The form is not a nipātana, or ūka would not carry into 3146 (BM, TB).
• 3146 yajajapadaśāṃ yaṅaḥ 3.2.166 — ūka after yaṅ-final yaj, jap, daś.
• "daś" names daṃś by its n-loss that is still to come.
• yāyajūkaḥ (a lost by SK 2308, y by SK 2631, abhyāsa lengthened by SK 2632); jañjapūkaḥ, dandaśūkaḥ (nuk, SK 2638).
• 3147 namikampismyajasakamahiṃsadīpo raḥ 3.2.167 — ra: namraḥ, kampraḥ, smeraḥ. nañ + jas → ajasram, "continuous". Then kamraḥ, hiṃsraḥ, dīpraḥ.

IV.L. u, and ki/kin treated like liṭ — running example: kṛ + san → cikīrṣa + u → (a lost) cikīrṣuḥ; pā + ki (liḍvat: reduplication) → pa-pā-i → (ā lost) papiḥ

• 3148 sanāśaṃsabhikṣa uḥ 3.2.168 — u after a san-final base, after āśaṃs and after bhikṣ: cikīrṣuḥ, āśaṃsuḥ, bhikṣuḥ. "san" means the affix, not the roots ṣaṇ; the proof is that vijigīṣu is read in the garga list (BM, TB).
• 3149 vindur icchuḥ 3.2.169 — u with nipātana: num for vid (vetti), giving vinduḥ; ch for the ṣ of iṣ, giving icchuḥ.
• 3150 kyāc chandasi 3.2.170 — in the Veda, u after a kya-final base: sumna + kyac → sumnaya + u → sumnayuḥ. No ī, because na cchandasy aputrasya (SK 3588) blocks SK 2658 (BM).
• 3151 ādṛgamahanajanaḥ kikinau liṭ ca 3.2.171 — in the Veda, ki and kin after ā-final and ṛ-final roots and after gam, han, jan. They are treated like liṭ (liḍvat: reduplication etc.):
• papiḥ somam, dadir gāḥ, babhrir vajram (second case, by SK 627), jagmir yuvā, jaghnir vṛtram, jajñiḥ.
• Vārttika bhāṣāyāṃ dhāñkṛsṛgamijaninamibhyaḥ: dadhiḥ, cakriḥ, sasriḥ, jagmiḥ, jajñiḥ, nemiḥ.
• Vārttika sāsahivāvahicācalipāpatīnām upasaṃkhyānam: ki/kin after yaṅ-final sah etc., and no nīk for pat.

IV.M. najiṅ, āru, kru/klukan, varac — running example: sthā + varac → sthāvaraḥ; yā + yaṅ → yāyāya + vara → (a lost; y lost) yāyāvaraḥ

• 3152 svapitṛṣor najiṅ 3.2.172 — najiṅ (ṅ-it: no guṇa): svapnak; tṛṣṇak, tṛṣṇajau, tṛṣṇajaḥ. "dhṛṣeś ca" is from the Kāśikā etc., not the Bhāṣya: dhṛṣṇak.
• 3153 śṝvandyor āruḥ 3.2.173 — āru: śarāruḥ, vandāruḥ.
• 3154 bhiyaḥ kruklukanau 3.2.174 — kru, klukan after bhī: bhīruḥ, bhīlukaḥ. Vārttika krukann api vācyaḥ: bhīrukaḥ (so the Amarakośa, TB).
• 3155 stheśabhāsapisakaso varac 3.2.175 — varac: sthāvaraḥ, īśvaraḥ (no iṭ, SK 2981), bhāsvaraḥ, pesvaraḥ, kasvaraḥ.
• 3156 yaś ca yaṅaḥ 3.2.176 — varac after yaṅ-final yā.
• The a of yaṅ is lost by ato lopaḥ (SK 2308; ours and the print say "2380").
• That loss would be sthānivat by acaḥ parasmin (SK 50). na padānta… (SK 51) denies this for the y-deletion, so y goes by lopo vyor vali (SK 873).
• "vare luptaṃ na sthānivat" keeps the ā of yā from being lost: yāyāvaraḥ.

IV.N. kvip, which closes the tācchīlika range; kvip after other roots; bhū with kvip or ḍu — running example: vi + bhrāj + kvip → (ṣ, SK 294; jaś, car) vibhrāṭ; vac + kvip (lengthening, no saṃprasāraṇa) → vāk; pra + bhū + ḍu → (ṭi lost) prabhuḥ

• 3157 bhrājabhāsadhurvidyutorjipṝjugrāvastuvaḥ kvip 3.2.177 — kvip, the limit of ā kveḥ (3114):
• vibhrāṭ; bhāḥ, bhāsau; dhūḥ, dhurau (v lost by SK 2655, lengthening by SK 433); vidyut; ūrk; pūḥ, purau.
• "dṛśi" is drawn back from 3158, so ju gets a long vowel: jūḥ, juvau, juvaḥ.
• grāvan is compounded with stu by nipātana: grāvastut (tuk).
• 3158 anyebhyo 'pi dṛśyate 3.2.178 — kvip after other roots too: chit, bhit. "dṛśi" also brings in other operations (lengthening, no saṃprasāraṇa, reduplication, shortening). Four vārttikas:
• kvib vacipracchyāyatastukaṭaprujuśrīṇāṃ dīrgho 'saṃprasāraṇaṃ ca: vāk, prāṭ, āyatastūḥ, kaṭaprūḥ, (jūḥ), śrīḥ. These are already Uṇādi forms; they are restated so that tṛn does not block them.
• dyutigamijuhotīnāṃ dve ca: didyut, jagat. The abhyāsa name comes from "dṛśi".
• juhoter dīrghaś ca: juhūḥ. dṝ "to fear" takes shortening: dadṛt.
• dhyāyateḥ saṃprasāraṇaṃ ca: dhīḥ.
• 3159 bhuvaḥ saṃjñāntarayoḥ 3.2.179 — kvip after bhū only (niyama) in a name or for one standing between:
• mitrabhūḥ (a name);
• pratibhūḥ, the surety between creditor and debtor. Not a name, since names last as long as their bearer (BM, TB).
• 3160 viprasambhyo ḍv asaṃjñāyām 3.2.180 — ḍu after bhū with vi, pra, sam, not in a name. The ṭi is lost by force of the ḍ (BM).
• vibhuḥ (print; ours vibhūḥ), prabhuḥ, sambhuḥ "begetter". In a name, kvip by 3159: vibhūḥ.
• Vārttika mitadrvādibhya upasaṃkhyānam: mitadruḥ, śatadruḥ, śambhuḥ, where bhavati has the ṇic sense within it.

IV.O. ṣṭran and itra: affixes of the object, the instrument and the agent, with the iṭ-bar — running example: dā + ṣṭran (karaṇa) → dā-tra → dātram; śas + ṣṭran → (no iṭ, 3163) śastram; pū + itra → (guṇa, av) pavitram

• 3161 dhaḥ karmaṇi ṣṭran 3.2.181 — ṣṭran after dhā (dheṭ and dhāñ: print) in the sense of the object (karmaṇi is not an upapada, TB). The ṭ is ṣṭutva of t and goes with the ṣ: dhā + tra → dhātrī, "mother, āmalakī, earth, foster-mother".
• 3162 dāmnīśasayuyujastutudasisicamihapatadaśanahaḥ karaṇe 3.2.182 — ṣṭran in the sense of the instrument after 13 roots. "dām" is dāp with p → m (SK 116); "daś" is daṃś. dānti anena → dātram, netram.
• 3163 titutratathasisusarakaseṣu ca 7.2.9 — no iṭ before the ten kṛt affixes ti, tu, tra, ta, tha, si, su, sara, ka, sa ("na iṭ", "kṛti" from SK 2981). Inserted here out of AS order, because iṭ was suspected for śas and pat (BM).
• The ṣṭran forms of 3162: śastram, yotram, yoktram, stotram, totram, setram, sektram, meḍhram, patram, daṃṣṭrā, naddhrī.
• daṃṣṭrā has no ṅīṣ (anitya, TB; or ajādi ṭāp).
• 3164 halasūkarayoḥ puvaḥ 3.2.183 — ṣṭran after pūṅ/pūñ for an instrument that is part of a plough or a hog: potram, its fore-part/snout. No iṭ, by 3163 (BM).
• 3165 artilūdhūsūkhanasahacara itraḥ 3.2.184 — itra (karaṇa) after ṛ, lū, dhū, sū, khan, sah, car: aritram, lavitram, dhavitram (TB, Wikisource dhuvitram), savitram, khanitram, sahitram, caritram.
• 3166 puvaḥ saṃjñāyām 3.2.185 — itra after pū in a name (karaṇe continues): pavitram, the strainer of the ājya and the ring worn on the ring-finger.
• 3167 kartari carṣidevatayoḥ 3.2.186 — itra after pū: in the instrument sense for a ṛṣi (= a mantra) and in the agent sense for a deity. This is the Kāśikā's respective pairing; the Śabdakaustubha rejects it (TB).
• pūyate 'nena → pavitram (the Pāvamānī hymns, BM);
• agniḥ pavitraṃ sa mā punātu, "Agni is the purifier".
• [iti pūrvakṛdantaṃ samāptam] is supplied in brackets from the print. p68 follows.