स्वरप्रकरणम्

Covers SK 3650–3978 with the 87 phiṭ-sūtras read after SK 3707 (416 sūtras): all present (written in eight parts, each coverage-checked by script: 3650–3701, 3702–phiṭ 46, phiṭ 47–3718, 3719–3770, 3771–3822, 3823–3874, 3875–3926, 3927–3978).

The accent chapter. Part I: the governing paribhāṣā, the vocative, the junction accents of the Tripādī, ekaśruti and the root and word accents of 6.1. Parts II–III: the last 6.1 word accents and Śāntanava's phiṭ-sūtras (the accent of underived stems), then the affix accents (3.1.3–4, 6.1.163ff.). Parts IV–VIII: the compound accents of 6.2 (prakṛtyā pūrvapadam, ādir udāttaḥ, antaḥ, uttarapadādiḥ, the bahuvrīhi and nañ/su rules), then the tiṅanta accents of 8.1 and the svara-saṃcāra that closes the chapter. Accent convention throughout: udātta acute, independent svarita grave, anudātta unmarked; where the 1915 print's signs contradict the rule illustrated, the print is kept and the rule's form given beside it. The commentary layer is the Subodhinī; there is no Bālamanoramā.

svaraprakaraṇam — outline (p71, SK 3650–3978 with phiṭ 1–87, all 416 sūtras)

Part I — SK 3650–3701: the general accents (the governing paribhāṣā, the vocative, the Tripādī junction accents, ekaśruti, sannatara, the āmreḍita), the root accents, and the first word accents (antodātta, penultimate and ādyudātta words, to yato 'nāvaḥ)

Covers SK 3650–3701 (52 sūtras): all present.

• The Aṣṭādhyāyī spine and where the SK moves rules. The SK does not take the accent sūtras in Aṣṭādhyāyī order. It builds three blocks that match the print's three section headings:
• The general accents (3650–3670). These are 6.1.158, 6.1.161, 6.1.222 (cau, the penultimate sūtra of 6.1), 6.1.198, 8.1.19, 8.1.74, 2.1.2, 8.2.4–6 and 8.4.66–67 (just before the closing a a, 8.4.68), then the whole of 1.2.33–40 (3662–3669), then 8.1.3 (3670). 3669 ends by quoting tasya param āmreḍitam (SK 83, 8.1.2) as the frame for 3670.
• The root accents (3671–3679). These are 6.1.162, 6.1.188–190, 6.1.192–195 and 6.1.218. 6.1.191 is held back to 3685, and 6.1.196 goes to SK 3732.
• The word accents (3680–3701). These are 6.1.159–160, 6.1.167, 6.1.180–181, 6.1.191, 6.1.197 and 6.1.199–213. Part II continues with 6.1.214–216 and 6.1.219–221 (SK 3702–3707) before the phiṭ-sūtras.
• Gaps filled later in the SK. 6.1.163–166 (SK 3710–3713), 6.1.168–179 (SK 3714–3725), 6.1.182–187 (SK 3726–3731), 6.1.196 (SK 3732), 6.1.217 (SK 3733) and 6.1.223 (SK 3734) come after the phiṭ-sūtras. So several anuvṛttis in this part are drawn from sūtras the SK teaches later. These are lasārvadhātukam and upadeśa from SK 3730, ādiḥ … anyatarasyām from SK 3731, upottamam from SK 3733, and vibhakti and ṣaṭtricaturbhyaḥ from SK 3714 and 3725.
• AS-refs. All 52 are checked against references/ashtadhyayi/sutraani-data.json by SK number, and all match. The sūtra texts match except 3655. The reference reads vibhāṣitaṃ viśeṣavacane bahuvacanam, while ours and the print read sāmānyavacanaṃ vibhāṣitaṃ viśeṣavacane, which is Pāṇini's wording. The kaumudī itself says the Bhāṣya adds bahuvacanam and moves sāmānyavacanam to SK 413.
• Headings and governing words.
• Print headings (absent from ours): "svaraprakriyā 68" above 3650 (recorded in a note, not bracketed); "atha dhātusvarāḥ 69" above 3671 and "atha pratyayasvarāḥ 70" above 3680 (both supplied in brackets). Ours keeps the closing phrases: "iti sādhāraṇasvarāḥ" ends 3670, and "iti dhātusvarāḥ" ends 3679. The Subodhinī calls 3650–3670 "common" because these accents work within one pada and across two.
• ekaśruti (3662) runs through 3663–3668. subrahmaṇyāyām and svaritasya (3666) run into 3667.
• anta udāttaḥ comes from 6.1.159 (= 3680, which the SK places later). The Subodhinī names it as the source for dhātoḥ (3671). For 3681, 3682, 3683 and 3686 the identification is the translator's.
• pratyayāt pūrvam (3675) runs into 3676 (Subodhinī).
• ādiḥ comes from 3673 for 3685. For 3687–3693 it comes from 3686 (the translator's identification, [Z04]). saṃjñāyām (3692) runs into 3693 (kaumudī), and dvyac (3693) runs into 3701 (Subodhinī).
• The tadanta jñāpaka. The Subodhinī reads the word pūrvam in 3675 as a jñāpaka: in accent rules a seventh-case word means "a form ending in that item". It applies this to SK 3733 and 3679, and expressly not to 3682 (audit H02).
• Accent convention (all of p71). The udātta is marked acute and an independent svarita grave. Anudātta, dependent svarita and pracaya are unmarked, and every line names its udātta syllable.
• Kampa numerals. The numerals 1 and 3 after vī, so and kvā (3659, 3661) are not reproduced; those syllables are marked as independent svarita.
• Examples with no accented copy (3657, 3658 vo 'śvāḥ, 3666, 3667, 3680, 3682, 3690, 3692, 3693, 3696, 3699) are marked only on the syllables the rule names.
• Where the print's signs contradict the rule ([orch], SK 3701), the print's signs are kept and the rule's form is given in brackets: nāvyā́nām [nāvyā̀nām].
• Print readings and audit corrections followed (NSP 1915, scan-confirmed; collation batch10–12; audits H01–H04, with fixes applied in en.yml).
• Mūla after the print.
• 3650 anyatreti.
• 3653 āmantritasyādiḥ.
• 3654 prāguktaṣāṣṭhasya.
• 3655 devīḥ ṣaḷ urvīḥ, devāḥ śaraṇyāḥ.
• 3656 śubhaspatī, teneha, uccair adhīyāna, upāgny adhīyāna (the print misnumbers this sūtra "3556").
• 3658 marutaḥ.
• 3659 te 'vadan, kṣaiprābhinihiteṣu, and the print's anudātta on the a of asyá.
• 3661 pra ya āruḥ (word division, not in the collation).
• 3663 mūrdhā, svādhyāyakāle.
• 3664 uccaistarām.
• 3666: ours and the print cite vibhāṣā chandasi as "3635", read here as SK 3665; dākṣeḥ pitā yajate is supplied; cād antas tena.
• 3668 ekaśrutiḥ.
• 3669 anudāttasyānudāttataraḥ.
• 3672–3673 aniṭy ajādau.
• 3675 mamattu, jāgarṣi tvam.
• 3681: the bhakṣa sentence once; īḷate.
• 3683: saptabhyo jāyamānaḥ, with navatī ca as one example.
• 3686 ñidantasya nidantasya.
• 3687 jyotiṣmataḥ.
• 3694 atasam.
• 3696 niṣṭhā.
• 3700 mahyam.
• 3701 yuñjanty.
• The print's square-bracketed additions in 3666 and 3701 are set in parentheses.
• Ours = print, rendered by sense. 3688 ādyudāttau is read as "udātta". At 3698 the class-C reading at ṣaṣṭi is resolved with the Subodhinī.
• Audit corrections.
• 3654 MUST-FIX: ághnye dévi sárasvati, each ādyudātta. The print marks only aghnye.
• 3682 MUST-FIX: catúraḥ, not caturáḥ.
• 3650: "[you two] guard us"; the Subodhinī's thas → tam.
• 3666: devadattasyá, with the final always udātta.
• 3671: the print marks both go and pā anudātta.
• 3678: kartṛyak is the yak introduced when a sārvadhātuka denoting the agent follows.
• 3685: the Subodhinī reads sarvatrādyudāttasyaiva.
• Where the commentary lies (only the kaumudī is translated, SK-TRANSLATE §25).
• The Subodhinī is the commentary throughout. Some of it is filed under a neighbour:
• 3650: tail of sk.3649.s1, plus sk.3650.s1.
• 3653: tail of sk.3652.s1.
• 3657 and 3658: inside sk.3656.s1.
• 3662–3666: all in sk.3662.s1. For 3663 the rest is our editorial note sk.3659.n1, which is the Subodhinī and not an NSP footnote. There is none on vā nāmadheyasya.
• 3671: tail of sk.3670.s1.
• 3678: tail of sk.3677.s1, plus its own block.
• 3680–3681: in sk.3679.s1.
• 3686: tail of sk.3685.s1. 3688: tail of sk.3687.s1. 3690: tail of sk.3689.s1. 3701: in sk.3700.s1.
• No BM anywhere in Part I. Our layer is empty, and Wikisource uttara-4/0097–0100 prints only the mūla. sk.3655.b1 is misfiled: it is BM on the supād forms under SK 413.
• Where Part I ends. Part I ends at 3701 yato 'nāvaḥ (6.1.213), the ādyudātta of two-vowel yat-final words. Part II opens with īḍavandavṛśaṃsaduhāṃ ṇyataḥ (6.1.214, SK 3702), which continues the same ādyudātta for ṇyat-final words.

I.A. The governing paribhāṣā, and udātta through elision — running example: gopāya + tam → gopāyátaṃ naḥ (udātta on ya, everything else anudātta); deva + ṅīp → (udātta a elided) devī́m (udātta on vī)

• 3650 anudāttaṃ padam ekavarjam 6.1.158 — this is a paribhāṣā for every accent rule (svaravidhi). When a rule gives one vowel of a pada udātta or svarita, all the pada's other vowels are anudātta. gopāyátaṃ naḥ ("[you two] guard us"): once gopāya is a dhātu (SK 2304), the dhātu accent (3671) puts the udātta on ya. gup's own udātta and that of āya are gone.
• Vārttika. An accent taught when another is already present (satiśiṣṭa) is the stronger, except for vikaraṇa accents. In yajñáṃ-yajñam abhí vṛdhé gṛṇītáḥ the later accent of śnā does not win, the ending tas keeps its udātta, and gṛṇītáḥ is antodātta (udātta on taḥ).
• Subodhinī. This is a paribhāṣā, not an adhikāra, because the sūtra is not svarita. "eka" stands for whatever a rule prescribes, so 3688 can leave two udāttas. The rule does not reach the Tripādī, so the svarita that 3660 gives tam survives.
• 3651 anudāttasya ca yatrodāttalopaḥ 6.1.161 — when an udātta is elided, the anudātta after it becomes udātta (udāttanivṛttisvara). deva is antodātta by citaḥ (SK 3710), and ṅīp is anudātta (SK 3709). yasyeti ca (SK 311) elides deva's udātta a, so ṅīp's ī becomes udātta: devī́ṃ vā́cam, udātta on vī.
• 3652 cau 6.1.222 — before añc with its a elided ("c"), the final vowel of the preceding member is udātta. This is an apavāda of 3651. In deva + añc + ṅīp, adri replaces deva's ṭi (SK 418), the a of añc is elided (SK 416) and the vowel is lengthened (SK 417), giving devadrī́cīṃ nayata devayántaḥ. The udātta is on drī, and the following ī is svarita by 3660.
• Vārttika ataddhite. The rule does not apply before a taddhita: dādhīcaḥ, mādhūcaḥ keep the pratyaya accent of aṇ.

I.B. The vocative: ādyudātta, nighāta after a pada, "treated as non-existent", and parāṅgavat — running example: ágna índra váruṇa mítra dévāḥ (each vocative udātta on its first syllable) against imáṃ me gaṅge yamune sarasvati (vocatives after a pada: wholly anudātta)

• 3653 āmantritasya ca 6.1.198 — the first vowel of a vocative (āmantrita, SK 411) is udātta. ágna índra váruṇa mítra dévāḥ has its udāttas on a, i, va, mi, de. This is the ṣāṣṭha rule, and the next sūtra overrides it.
• 3654 āmantritasya ca 8.1.19 — this is the āṣṭamika rule: after a pada, and not at the head of a pāda, the whole vocative is anudātta. The conditions come from the gloss and match SK 402–403.
• imáṃ me gaṅge yamune sarasvati: these vocatives stand after me, so gaṅge, yamune and sarasvati have no udātta.
• At the head of a pāda the ṣāṣṭha rule holds: śútudri stómam, udātta on śu.
• SK 412 (āmantritaṃ pūrvam avidyamānavat) is re-cited. In ágna índra, agne counts as absent, so indra is not "after a pada" and keeps its udātta on i.
• SK 413 (nāmantrite samānādhikaraṇe sāmānyavacanam) is re-cited. When the earlier vocative is the qualificand of a co-referential one, it is not treated as absent: in ágne tejasvin and ágne trātaḥ, tejasvin and trātaḥ undergo nighāta.
• Among synonyms SK 413 does not apply, so each word escapes nighāta: ághnye dévi sárasvati, udātta on a, de, sa (audit H01 MUST-FIX; the print marks only aghnye).
• 3655 sāmānyavacanaṃ vibhāṣitaṃ viśeṣavacane 8.1.74 — Bhaṭṭoji gives the Bhāṣya's recast. A preceding plural vocative is optionally treated as non-existent when a vocative qualifier follows it. In dévīḥ ṣaḷ urvīr urú naḥ kṛṇota, dévīḥ (udātta on de) is taken as present, so ṣaḷ urvīḥ undergoes nighāta. In devāḥ śaraṇyāḥ, nighāta of the second word is optional.
• 3656 subāmantrite parāṅgavat svare 2.1.2 — for accent, a sup-ending word before a vocative counts as part of that vocative.
• drávatpāṇī śúbhaspatī: the genitive śubhaḥ (from śubh with kvip) is absorbed into patī, so the ṣāṣṭhika ādyudātta (3653) falls on śu. dravatpāṇī counts as absent (SK 412), so there is no āṣṭamika nighāta.
• yát te divo duhitar marta bhójanam: divaḥ shares duhitar's nighāta. paraśunā vṛścan is a further example.
• Vārttikas.
• The rule covers only a sixth-case word, or a word expressing a kāraka of the vocative. So it does not apply in ayám agne jaritā́ or eténāgne bráhmaṇā. Alternatively, the restriction comes from carrying samartha (SK 647) forward.
• "Like a member of the preceding (pūrvāṅgavat)" too: in ā́ te pitar marutām and práti tvā duhitar divaḥ, marutām and divaḥ share the nighāta.
• The rule does not cover indeclinables: uccair adhīyāna.
• It does cover the avyayībhāva: in upāgny adhīyāna the compound takes the ādyudātta of 3653.

I.C. The Tripādī junction accents: svarita after a yaṇ or an udātta, the udātta single substitute, and the block before udātta/svarita — running example: agním īḷe (udātta gnim, the following ī svarita); kva + avaram → kvā́varam (merged ā udātta)

• 3657 udāttasvaritayor yaṇaḥ svarito 'nudāttasya 8.2.4 — an anudātta after a yaṇ that replaced an udātta or a svarita becomes svarita.
• abhy abhi hi: the y replaces the udātta final i of the first abhi, so the anudātta a of the repeated abhi becomes svarita.
• khalapvyāśā: the udātta-yaṇ v makes ṅi svarita. That svarita i becomes y, so the ā of āśā is svarita.
• The rule is Tripādī and asiddha, so there is no śeṣanighāta by 3650. The Subodhinī gives the Bhāṣya's yogavibhāga.
• 3658 ekādeśa udāttenodāttaḥ 8.2.5 — a single substitute for an udātta together with another vowel is udātta.
• vas + aśvāḥ → vo 'śvāḥ: anudātta vas plus ādyudātta aśva, merged by pūrvarūpa (SK 86), gives an udātta o.
• kva + avaram → kvā́varaṃ marutaḥ: kva is svarita by tit svaritam (SK 3729) and the a of avara is udātta, so kvā is udātta.
• 3659 svarito vā 'nudātte padādau 8.2.6 — before an anudātta at the start of a pada, that single substitute is optionally svarita. The option has a fixed distribution:
• two short i's give svarita: vi + idam → vī̀daṃ jyótir hṛ́daye (independent svarita vī̀);
• where a long vowel enters, the result is udātta: divi + īyate → asyá ślóko divī́yate (udātta vī);
• the pūrvarūpa of SK 86 gives svarita only: tè 'vadan, sò 'yám ā́gāt.
• The Prātiśākhya line names these three as the praśleṣa, kṣaipra and abhinihita saṃdhis.
• 3660 udāttād anudāttasya svaritaḥ 8.4.66 — in saṃhitā (from SK 3627, Subodhinī), an anudātta after an udātta becomes (dependent) svarita. In agním īḷe the udātta is gnim and ī is svarita. The rule is asiddha, so no śeṣanighāta follows. In tám īśānā́saḥ the udāttas are tam and nā, and ī and saḥ are svarita.
• 3661 nodāttasvaritodayam agārgyakāśyapagālavānām 8.4.67 — the svarita of 3660 does not arise when an udātta or a svarita follows (udaya = para, Subodhinī). Gārgya, Kāśyapa and Gālava hold that it does.
• prá yá ārúḥ: the ā between yá and rú stays anudātta (and is sannatara, 3669).
• vó 'śvāḥ kvā̀ bhī́śavaḥ: the ā of śvāḥ stays anudātta before the independent svarita kvā̀.

I.D. Ekaśruti: calling from afar, the rite, the Veda, and the Subrahmaṇyā formula — running example: āgaccha bho māṇavaka (called from afar: the whole sentence ekaśruti); subrahmaṇya + om → subrahmaṇyó3m (svarita merged o → udātta o3)

• 3662 ekaśruti dūrāt sambuddhau 1.2.33 — when one calls from a distance, the sentence is ekaśruti, an apavāda of traisvarya: āgaccha bho māṇavaka. The Subodhinī takes sambuddhi in its literal sense (so a plural call is covered too). In the example the ṭi is pluta and udātta by dūrād dhūte ca (SK 95). Called from near by, the sentence would keep its three accents.
• 3663 yajñakarmaṇy ajapanyūṅkhasāmasu 1.2.34 — in a sacrificial act a mantra is ekaśruti, except in japa, the nyūṅkhas (the sixteen o-sounds) and the sāmans (chants).
• agnír mūrdhā́ diváḥ kakút keeps its three accents in svādhyāya (udātta gní, dhā́, váḥ, kút) and is monotone in the rite.
• Japa keeps its accent: mámāgne várco vihavéṣv astu, udātta má, vá, vé.
• 3664 uccaistarāṃ vā vaṣaṭkāraḥ 1.2.35 — in the rite, the word vauṣaṭ is optionally uccaistarām ("higher"), or else ekaśruti. By the Subodhinī's first view, brūhipreṣya… (SK 3610) makes its first syllable pluta and udātta, and this rule makes the final udāttatara relative to that. There is no example in ours or the print.
• 3665 vibhāṣā chandasi 1.2.36 — in the Veda, ekaśruti is optional, with a fixed distribution:
• in the saṃhitā there is traisvarya;
• in the brāhmaṇa the Bahvṛcas use ekaśruti;
• other schools follow their own tradition (sampradāya).
• 3666 na subrahmaṇyāyāṃ svaritasya tūdāttaḥ 1.2.37 — in the Subrahmaṇyā nigada, the ekaśruti of 3663 and 3665 does not apply, and an udātta replaces a svarita.
• The rule's example. subrahmaṇya takes yat and is svarita (SK 3729). Its svarita merges with om, and this rule makes the result udātta: subrahmaṇyó3m, udātta pluta o3.
• The print's bracketed addition. It says 3658 would not already give this, because "anudātta" is carried into 3658. The Subodhinī says the opposite at 3668 (recorded).
• Vārttika asāv ity antaḥ. In a first-case word the final is udātta: gārgyó yajate. This sets aside the ñit ādyudātta.
• Vārttika amuṣyety antaḥ. The same for a sixth-case word: dākṣéḥ pitā yajate (supplied from the print).
• Vārttika syāntasyopottamaṃ ca. In a sya-final word the penultimate and the final are both udātta: gārgyásyá pitā yajate, udātta gya and sya.
• Vārttika vā nāmadheyasya. In a name the penultimate is optional: devadattasyá or devadattásyá (audit H02).
• 3667 devabrahmaṇor anudāttaḥ 1.2.38 — in the Subrahmaṇyā, the svarita of deva and brahman becomes anudātta instead of the udātta of 3666. In dévā bráhmāṇa āgacchata the udāttas are dé and brá. vā and hmā, which 3660 would make svarita, are anudātta.

I.E. In saṃhitā: pracaya after a svarita, sannatara before an udātta or svarita, and the anudātta āmreḍita — running example: imám me gaṅge yamune sarasvati (udātta mám; me dependent svarita; every syllable after it ekaśruti); divé-dive (udātta vé; the repeat unaccented)

• 3668 svaritāt saṃhitāyām anudāttānām 1.2.39 — in saṃhitā, the anudāttas after a svarita become ekaśruti (pracaya), whether one, two or many. In imám me gaṅge yamune sarasvati, me is a dependent svarita after mám (3660), and gaṅge yamune sarasvati, all nighāta, are pronounced ekaśruti. The Subodhinī asks why a Tripādī svarita can feed this rule. Its answer: 1.2.32–40 are lifted to the end of the śāstra (kāṇḍotkarṣa).
• 3669 udāttasvaritaparasya sannataraḥ 1.2.40 — an anudātta followed by an udātta or a svarita becomes sannatara (anudāttatara).
• sarasvati śútudri: the final i of sarasvati is sannatara before śú.
• vyàcakṣayat svàḥ: the ya is sannatara before the svarita svàḥ (nyaṅsvarau svaritau, phiṭ 74).
• The unit closes by quoting tasya param āmreḍitam (SK 83) as the frame for 3670.
• 3670 anudāttaṃ ca 8.1.3 — the latter member of a doubled word (āmreḍita) is anudātta. In divé-dive the udātta is the vé of the first word. The repeat has no udātta: its di is a dependent svarita and its ve is ekaśruti. Closing phrase: "iti sādhāraṇasvarāḥ".

I.F. Root accents I: the dhātu is antodātta; the initial of svapādi, hiṃs and reduplicated stems — running example: gopāya (dhātu) → gopāyátam (udātta on the final ya); dā + jhi → (at for jhi) dadati → dádati (initial da udātta)

• 3671 dhātoḥ 6.1.162 — the final vowel of a dhātu is udātta. anta udāttaḥ is carried from 6.1.159 (= 3680), as the Subodhinī says. Examples:
• gopāyátaṃ naḥ: udātta yá, the final of gopāya. taṃ is a dependent svarita and naḥ ekaśruti.
• ási satyáḥ: as + sip, with the s dropped. The udātta is á, the root's own final vowel.
• The heading "[atha dhātusvarāḥ]" is supplied.
• 3672 svapādihiṃsām acy aniṭi 6.1.188 — for svap and the rest (a sub-group running to the end of adādi) and for hiṃs, the initial is optionally udātta before a vowel-initial lasārvadhātuka without iṭ. lasārvadhātukam comes from SK 3730, and ādiḥ … anyatarasyām from SK 3731.
• svápanti, śvásanti, híṃsanti: udātta sva, śva, hiṃ.
• In the other option the form is madhyodātta by the pratyaya accent (udātta on the a of -anti).
• Bhaṭṭoji, following the Vṛtti, limits the rule to kit/ṅit endings. So it does not apply in svapāni, hinasāni (-āni is pit).
• 3673 abhyastānām ādiḥ 6.1.189 — in the same environment, the initial of a reduplicated stem is udātta. ādiḥ is restated so that the rule is obligatory (Subodhinī).
• yé dádati priyā́ vásu: udātta on the first da.
• Being the later rule, it overrides the cit accent of śānac (SK 3710): dádhānā índre.
• 3674 anudātte ca 6.1.190 — the same holds before a lasārvadhātuka that has no udātta (bahuvrīhi, Subodhinī). In dádhāsi rátnaṃ dráviṇaṃ ca dāśúṣe, the ending si is pit and anudātta, so the first da of dadhāsi is udātta.

I.G. Root accents II: the vowel before the suffix; ṇamul, kartṛyak, caṅ — running example: hu + ślu + tip → juhóti (udātta ho, the vowel before the pit ending); cikīrṣa + ṇvul (aka) → cikī́rṣakaḥ (udātta kī)

• 3675 bhīhrībhṛhumadajanadhanadaridrājāgarāṃ pratyayāt pūrvaṃ piti 6.1.192 — for these nine roots as reduplicated stems, before a pit lasārvadhātuka, the vowel before the suffix is udātta. This is an apavāda of 3673. Examples:
• yò 'gnihotráṃ juhóti: udātta ho (yò is an independent svarita by 3659).
• mamáttu naḥ párijmā: udātta on the second ma.
• mātā́ yád vīráṃ dadhánat: udātta dha.
• jāgárṣi tvám: udātta ga.
• The pūrvam jñāpaka. The Subodhinī reads pūrvam as a jñāpaka of the tadanta-saptamī: in accent rules a seventh-case word means a form ending in that item.
• 3676 liti 6.1.193 — before a lit suffix, the vowel before the suffix is udātta (pratyayāt pūrvam carried from 3675). In cikīrṣa + ṇvul (aka) → cikī́rṣakaḥ, the a of san is elided, and the kī right before the lit aka is udātta. The elision does not count as the a (no sthānivattva in accent).
• 3677 ādir ṇamuly anyatarasyām 6.1.194 — before ṇamul, the initial (of reduplicated stems, Bhaṭṭoji's gloss) is optionally udātta.
• lólūyaṃlolūyam: udātta on the first lo. The repetition is anudātta (3670).
• In the other option, the lit accent of 3676 (ṇamul is lit) puts the udātta on the vowel before ṇamul.
• 3678 acaḥ kartṛyaki 6.1.195 — for a root that ends in a vowel in upadeśa, the initial is optionally udātta before the kartṛyak. This is the yak introduced when a sārvadhātuka denoting the agent follows, i.e. the karmakartṛ (audit H03).
• lū́yate kedāraḥ svayam eva ("the field gets reaped by itself"): udātta lū.
• Otherwise yak is udātta: lūyáte.
• 3679 caṅy anyatarasyām 6.1.218 — in a caṅ-final root the penultimate is optionally udātta. upottamam comes from SK 3733, and caṅi means "caṅ-final" by the jñāpaka of 3675.
• mā hi cīkaratām: in cī-ka-ra the upottama is ka, the root's a, giving cīkáratām.
• Otherwise caṅ is udātta: cīkarátām.
• The rule needs three syllables, so it does not apply in mā hi dadhat. Closing phrase: "iti dhātusvarāḥ".

I.H. Word accents I: antodātta words, catur before śas, and the penultimate before jhal-initial endings of numerals — running example: pac + ghañ → (ku, SK 2863) pākáḥ (udātta on the final kaḥ); pañcan + bhis → pañcábhiḥ (udātta on the penultimate ca)

• 3680 karṣātvato ghaño 'nta udāttaḥ 6.1.159 — a ghañ-final word from karṣ (karṣati) or from a root containing ā has its final udātta. This is an apavāda of 3686.
• karṣáḥ and pākáḥ: udātta on the final syllable.
• Since karṣ is cited with śap, tudādi kṛṣ keeps the ādyudātta of 3686: kárṣaḥ.
• The heading "[atha pratyayasvarāḥ 70]" is supplied.
• 3681 uñchādīnāṃ ca 6.1.160 — uñcha and the rest are antodātta:
• yuga is nipātita in the list (ghañ without guṇa) in the senses "age" and "yoke": vaiśvānaráḥ kuśikébhir yugé-yuge, udātta ge. Elsewhere it is yoga, ādyudātta by 3686: yóge-yoge tavástaram.
• bhakṣa is a ghañ-final: gā́vaḥ sómasya prathamásya bhakṣáḥ, udātta kṣaḥ.
• uttama and śaśvattama: úd uttamáṃ varuṇa, śaśvattamám īḷate, udātta maṃ and mam.
• 3682 caturaḥ śasi 6.1.167 — before śas the final vowel of catur is udātta. catur (cat + uran, uṇādi 736) would otherwise be ādyudātta as a nit word.
• catúraḥ kalpayantaḥ: udātta tu, not the ending (audit H03 MUST-FIX).
• Not in cátasraḥ paśya: the r replacing ṛ (aci ra ṛtaḥ, SK 299) is sthānivat for this preceding operation.
• 3683 jhaly upottamam 6.1.180 — in a word ending in a jhal-initial vibhakti after ṣaṭ, tri or catur (from SK 3714 and 3725), the penultimate is udātta. This is an apavāda of SK 3725, which makes the vibhakti udātta.
• adhvaryúbhiḥ pañcábhiḥ, navábhir vā́jair navatī́ ca, saptábhyo jā́yamānaḥ, ā́ daśábhir vivásvataḥ: udātta ca, va, pta, śa.
• Where there is no penultimate, SK 3725 gives the udātta to the vibhakti: ā́ ṣaḍbhír hūyámānaḥ, devaís tribhíḥ.
• Before a non-jhal ending: navānā́ṃ navatīnā́m.
• 3684 vibhāṣā bhāṣāyām 6.1.181 — in ordinary usage, 3683 is optional "in the domain just stated". By the Subodhinī, pañcabhiḥ, saptabhiḥ, tisṛbhiḥ and caturbhiḥ are udātta on the penultimate (ca, pta, …) or else antodātta by SK 3725.

I.I. Word accents II: ādyudātta words — sarva, ñit/nit-final words, pathin/mathin, tavai, and named and listed words — running example: puṃs + ṣyañ (ñit; brāhmaṇādi) → paúṃsya (udātta on the first syllable pau): yásmin víśvāni paúṃsyā

• 3685 sarvasya supi 6.1.191 — before a sup, the first vowel of sarva is udātta (ādiḥ from 3673). sárve nandanti yaśásā: udātta sa. The Subodhinī explains the uñchādi listing (3681): sarva is listed as antodātta only so that sārva takes añ. What is actually heard is always ādyudātta (audit H03).
• 3686 ñnity ādir nityam 6.1.197 — a word ending in an affix with ñ or n as it has its first vowel udātta. ñniti is tadanta by the jñāpaka of 3675.
• puṃs + ṣyañ → paúṃsya: udātta pau.
• cāy + asun (+ nuṭ, uṇādi 639) → cánas: suté dadhiṣva naś cánaḥ, udātta ca.
• 3687 pathimathoḥ sarvanāmasthāne 6.1.199 — before a sarvanāmasthāna, pathin and mathin (ini-final, so antodātta) are ādyudātta.
• ayáṃ pánthāḥ: udātta pa.
• Otherwise: in jyótiṣmataḥ pathó rakṣa, the ṭi is elided in the bha stem, and by 3651 the vibhakti is udātta (tho).
• 3688 antaś ca tavai yugapat 6.1.200 — in a tavai-final word, the first and the last vowels are udātta at once. yugapat overrides the "ekavarjam" of 3650. In dā + tavai → dātavai → (y lost, SK 67) dātavā u, we get harṣase dā́tavā́ u, udātta on dā and vā.
• 3689 kṣayo nivāse 6.1.201 — kṣaya meaning "dwelling" (gha in the locus sense, SK 3296) is ādyudātta: své kṣáye śucivrata, udātta kṣa. The Subodhinī's counter-example is vyādheḥ kṣayaḥ (erac).
• 3690 jayaḥ karaṇam 6.1.202 — jaya meaning an instrument ("one conquers by it", gha by SK 3296) is ādyudātta: jáyo 'śvaḥ, udātta ja. The Subodhinī's counter-example is jayo brāhmaṇānām (bhāve erac).
• 3691 vṛṣādīnāṃ ca 6.1.203 — vṛṣa and the rest, an open list (ākṛtigaṇa), are ādyudātta: vā́jebhir vājínīvatī, índraṃ vā́ṇīḥ, udātta vā. As a ghañ-final, vāja would otherwise be antodātta by 3680.
• 3692 saṃjñāyām upamānam 6.1.204 — a word for the standard of comparison, used as a name, is ādyudātta. In cañcā iva → cáñcā, kan (SK 2051) is elided by lup (SK 2053), and the udātta is ca.
• Because 3686 would already give this through pratyayalakṣaṇa, the rule is a jñāpaka: in accent rules pratyayalakṣaṇa sometimes fails.
• Counter-examples: agnir māṇavakaḥ (not a name); jaitraḥ (not an upamāna).

I.J. Word accents III: niṣṭhā-final words — running example: dā + kta → (dat, SK 3077) dáttaḥ (a name: udātta on the first syllable da), against kṛtám, hṛtám (affix accent: udātta on the final)

• 3693 niṣṭhā ca dvyajanāt 6.1.205 — a name of two vowels ending in a niṣṭhā has its first vowel udātta, provided that vowel is not ā. saṃjñāyām comes from 3692.
• dáttaḥ: udātta da.
• Not cintitaḥ (three vowels) or trātaḥ (ā).
• Not a name: kṛtám, hṛtám.
• 3694 śuṣkadhṛṣṭau 6.1.206 — śuṣka and dhṛṣṭa are ādyudātta even when they are not names. atasáṃ ná śúṣkam: udātta śu (śuṣ + kta, k for t by SK 3030).
• 3695 āśitaḥ kartā 6.1.207 — āśita meaning the agent is ādyudātta: kṛṣánn ít phā́la ā́śitam, udātta ā. The Subodhinī gives two routes: kta in the agent sense by nipātana, or the karman of the causative āśi.
• 3696 rikte vibhāṣā 6.1.208 — rikta is optionally ādyudātta: ríktaḥ, or riktáḥ with the affix accent. As a name it is always ādyudātta by 3693, through pūrvavipratiṣedha.
• 3697 juṣṭārpite ca chandasi 6.1.209 — in the Veda, juṣṭa and arpita are optionally ādyudātta (vā from 3696). There is no example in ours or the print.
• 3698 nityaṃ mantre 6.1.210 — in a mantra they are always ādyudātta. Bhaṭṭoji calls the sūtra dispensable (śakyam akartum):
• júṣṭo dámūnāḥ and ṣáḷara āhur árpitam (udātta jú, ár) already follow from 3697, since the Vedic recitation is fixed;
• arpitā́ḥ ṣaṣṭír ná calācalā́saḥ shows an antodātta arpita (udātta tā́) in a mantra, which the rule would wrongly exclude.

I.K. Word accents IV: yuṣmad/asmad before ṅas and ṅe, and two-vowel yat-final words — running example: asmad + ṅas → (aś; mama; pararūpa) máma (udātta on the first ma, not the ending SK 3658 would give); kam + ṇiṅ + yat (SK 2842) → kā́mya (udātta kā)

• 3699 yuṣmadasmador ṅasi 6.1.211 — before ṅas, the first syllable of yuṣmad and asmad (madik-final, so antodātta) is udātta: nahi ṣas táva no máma, udātta ta and ma.
• 3700 ṅayi ca 6.1.212 — the same before ṅe. The rule is split from 3699 to block a one-to-one pairing (yathāsaṃkhya). Examples: túbhyaṃ hinvānáḥ, máhyaṃ vā́taḥ pavatām, udātta tú and má.
• 3701 yato 'nāvaḥ 6.1.213 — a two-vowel yat-final word (dvyac from 3693) is ādyudātta, except one formed from nau. This is an exception to tit svaritam (SK 3729).
• yuñjánty asya kā́myā: udātta kā.
• "Why 'except nau'?" navatíṃ nāvyā́nām keeps the print's signs. By the rule nāvya keeps the svarita of tit svaritam on vyā, giving [nāvyā̀nām] (the [orch] decision; the discrepancy is listed in the Print Check).
• The Subodhinī takes anāvaḥ as a jñāpaka of svaravidhau vyañjanam avidyamānavat.

Part II — SK 3702 – phiṭ 46: the last Pāṇinian rules on accent within a word (ṇyat words, veṇu and indhāna, tyāga etc., and the matup names in -āvatī, -avatī, -īvatī); then the phiṭ-sūtras from their opening: the stem-final udātta of pāda 1, and the first-syllable heading of pāda 2 up to the heavy-syllable rule and the grain words

Covers SK 3702–3707 and phiṭ 1–46 (52 sūtras): all present.

• The Aṣṭādhyāyī spine and where the SK reorders it.
• 6.1.214–216 and 6.1.219–221 (3702–3707) come in order, continuing Part I's 6.1.213 yato 'nāvaḥ (SK 3701).
• The two rules in between are placed elsewhere: 6.1.217 upottamaṃ riti is SK 3733, and 6.1.218 caṅy anyatarasyām is SK 3679. 6.1.222 cau was taught earlier as SK 3652.
• All six AS-refs were checked against references/ashtadhyayi/sutraani-data.json, by SK number and by sūtra text, and all match. Every SK cross-reference cited below was also checked: 331, 494, 1335, 2009, 2011, 2013, 2019, 2051, 2053, 2469, 2851, 2857, 2863, 2872, 3187, 3658, 3680, 3686, 3701, 3709, 3729, 3730 and 3878.
• The phiṭ-sūtras are not Pāṇini's, so they have no AS-ref. The print places them between SK 3707 and SK 3708 (Ajai, 2026-10-03). They are numbered here by the print's running number, "phiṭ N", where sk1000NN = phiṭ NN.
• source.json gives each phiṭ-sūtra a pāda ref of the form "10.p.n": pāda 1 is phiṭ 1–23, and pāda 2 is phiṭ 24–49. The ref for phiṭ 32 is mis-entered as 10.2.31.
• The headings.
• 3702–3704. "The first syllable is udātta" is Bhaṭṭoji's gloss (ādir udāttaḥ). At 3704 his "vā" is the vibhāṣā of 3703; naming 3703 as its source is the translator's identification.
• 3705–3707. At 3707, "prāgvat" carries forward 3706's antaḥ … udāttaḥ.
• Phiṭ pāda 1. anta udāttaḥ runs on from phiṭ 1 through phiṭ 23. The Subodhinī on phiṭ 23 confirms this ("udātta ity anuvṛtteḥ"); naming phiṭ 1 as the source is the translator's identification. Further carry-overs within the pāda:
• phiṭ 9–10 carry dakṣiṇa from phiṭ 8.
• phiṭ 11 carries chandasi from phiṭ 10 (Subodhinī).
• phiṭ 12 carries kṛṣṇa from phiṭ 11.
• phiṭ 15 carries chandasi and antaḥ from phiṭ 14.
• Phiṭ pāda 2. phiṭ 24 athādiḥ prāk śakaṭeḥ is a heading: the first syllable is udātta in every rule before phiṭ 69 (prāk, audit H07). Exceptions:
• ādiḥ is not construed at phiṭ 42 or 44 (Subodhinī).
• It is not rendered at phiṭ 31, because that rule names the vowel itself.
• phiṭ 40 and 41 carry phiṭ 39's lup-name.
• Print readings followed (NSP 1915, scan pages 0633–0636; collation batches 12–14 and audits H05–H08):
• Supplied in brackets:
• 3702: [vándyaś ca].
• phiṭ 1: the heading [atha phiṭsūtrāṇi; prathamaḥ pādaḥ].
• phiṭ 4: the closing sentence on the antaraṅga ādyudātta.
• phiṭ 6: [nabviṣayasya].
• phiṭ 8: the kośa line.
• phiṭ 23: the pāda-1 colophon.
• phiṭ 24: the heading [dvitīyaḥ pādaḥ].
• phiṭ 29: [óṣṭhāv iva mádhu].
• Readings corrected:
• 3703: indhāno.
• phiṭ 4: āntrébhyas.
• phiṭ 7: the sūtra reads hiṣṭhavatsaratiśatthāntānām (ours: baṃhiṣṭhavatsartiśatthānām).
• phiṭ 13: yáthā piba.
• phiṭ 17: aryasya (ours: āryasya).
• phiṭ 19: nakṣatrāṇām.
• phiṭ 20: āp tad-viṣayāṇām (ours runs it together as āpta-).
• phiṭ 29: chubukād, vihāyāḥ.
• phiṭ 34: anṛttācchīlye, ṛdvarjyam.
• phiṭ 41: āṅgaḥ / aṅgāḥ.
• phiṭ 43: pūrvo, te.
• phiṭ 45: nartuprāṇyākhyāyām.
• phiṭ 46: dhānyānāṃ (ours: dhānyāyāṃ).
• Accents, following the print's signs:
• 3702: u of ukthám is anudātta.
• phiṭ 7: the second baṃhiṣṭha mantra. In abhyū̀rṇvānā́, nā́ is udātta (H05 MUST-FIX, applied).
• phiṭ 26: váne ná vāyáḥ (ours would give vané). The Ṛgveda padapāṭha divides vā | yaḥ.
• phiṭ 30: kṣīrám.
• phiṭ 32: meṣám.
• phiṭ 35: tásya, mā́.
• Ours kept:
• phiṭ 7: saṃvatsaraḥ (print: parivatsaraḥ).
• phiṭ 46: rājamāṣāḥ (print: māṣāḥ).
• phiṭ 30: the bracketed "[kupūrvaṃ kim? dhanuḥ]", which the print lacks. dhanuḥ is no living being; H07 suggests dhenuḥ.
• Class C queries:
• phiṭ 12: vā or vāṃ.
• phiṭ 25: bali or vali.
• phiṭ 44: the print's tittiraḥ and its added khañjarīṭaḥ.
• Where the Subodhinī lies (only the kaumudī is translated, SK-TRANSLATE §25). The Bālamanoramā is absent: our BM layer is empty, and Wikisource uttara-4/0100–0103 is mūla only for the whole range.
• Subodhinī filed under a neighbour:
• 3705–3707: tail of sk.3704.s1, which ends "atha phiṭsūtrāṇi".
• phiṭ 1–4: sk.100001.s1.
• phiṭ 5: runs into the head of sk.100006.s1.
• phiṭ 8 + 10: sk.100008.s1 (10 is garbled in OCR; read from scan 0634).
• phiṭ 11 + 13: sk.100011.s1.
• phiṭ 18–23, the pāda colophon, and phiṭ 26–28: sk.100018.s1.
• phiṭ 30: tail of sk.100029.s1.
• phiṭ 34: tail of sk.100032.s1.
• phiṭ 42 + 44: sk.100042.s1.
• phiṭ 45–49 (and 52–53): sk.100045.s1, which drops a line on phiṭ 49.
• No Subodhinī in the print on phiṭ 9, 12, 24, 25, 31, 33, 37 or 43.
• Audit corrections followed (H05–H08; the only MUST-FIX is phiṭ 7's nā́, applied):
• 3705: strīnāman is "a feminine name", such as a river's, not a woman's name.
• phiṭ 4: the mantra's case is ablative, "from your entrails, from your guts".
• phiṭ 9: the alternative final udātta is phiṭ 1's, carried through. It is not phiṭ 8's, which is limited to sādhu.
• phiṭ 18: the Subodhinī's "ādyudāttaḥ" is a likely slip for antodāttaḥ.
• phiṭ 37: "pītadruḥ saralaḥ" can also be read as one example.
• phiṭ 42: the Subodhinī reports (āhuḥ) the rejection of "the first heavy syllable".
• phiṭ 46: the Subodhinī's "uśīnara-" means uśīra- (phiṭ 67), and the print reads tan na. The Subodhinī on phiṭ 67 turns the point round: there it treats śyāmāka in this sūtra as a jñāpaka.
• Where Part II begins and ends. Part I closed at 3701 yato 'nāvaḥ (6.1.213). Part II opens with 3702 (6.1.214) and runs through the matup names to 3707, then into the phiṭ-sūtras. It stops at phiṭ 46, in the middle of pāda 2. Part III resumes at phiṭ 47 (janapada-).

II.A. ṇyat words, veṇu and indhāna, and tyāga etc. take the udātta on the first syllable — running example: īḍ + ṇyat (ṛhalor ṇyat, SK 2872) → īḍya → ī́ḍyaḥ (udātta ī́)

• 3702 īḍavandavṛśaṃsaduhāṃ ṇyataḥ 6.1.214 — the first syllable of ṇyat-words from īḍ, vand(i), vṛ (= vṛṅ "choose"), śaṃs and duh is udātta (Bhaṭṭoji's gloss ādir udāttaḥ).
• The rule is needed because ṇyat, with two anubandhas, is not covered by the "yat" of yato 'nāvaḥ (SK 3701) (Subodhinī).
• Examples: ī́ḍyo nū́tanair utá; ājúhvāna ī́ḍyo [vándyaś ca]; śréṣṭhaṃ no dhehi vā́ryam; ukthám índrāya śáṃsyam. Each ṇyat form is ādyudātta: ī́ḍya, vándya, vā́rya, śáṃsya (udātta ī́, ván, vā́, śáṃ).
• vā́rya takes ṇyat, not the kyap of SK 2857, because kyap is taught only after vṛñ (Subodhinī).
• 3703 vibhāṣā veṇvindhānayoḥ 6.1.215 — veṇu and indhāna take the udātta on the first syllable optionally.
• veṇu would otherwise be ādyudātta without option, from its nit uṇādi affix. The print's Subodhinī cites the rule as "jṛvṛribhyo nic ca", meaning ajivṛrībhyo nic ca, uṇādi 318.
• indhāna would otherwise be antodātta with cānaś (cit). With śānac it would be madhyodātta: the lasārvadhātuka affix is anudātta (SK 3730), and śnam's a is lost (SK 2469).
• Example: índhāno agním (udātta í: the ādyudātta option).
• 3704 tyāgarāgahāsakuhaśvaṭhakrathānām 6.1.216 — optionally ("vā" from 3703) the first syllable is udātta. The kaumudī gives no examples, so the forms below are the sūtra's own words.
• Three ghañ-forms: tyaj + ghañ → (k for j, SK 2863) tyā́gaḥ; rañj + ghañ → (n lost, SK 3187) rā́gaḥ; hā́saḥ. Otherwise they are antodātta by karṣātvato ghaño 'nta udāttaḥ (SK 3680).
• Three pacādi ac-forms: kúhaḥ, śváṭhaḥ, kráthaḥ. Otherwise they are antodātta by the affix accent (Subodhinī).

II.B. matup names in -āvatī, -avatī and -īvatī — running example: śara + matup (a river name; m → v; ā by śarādīnāṃ ca) → śarāvatī → śarā́vatī (udātta rā́)

• 3705 matoḥ pūrvam āt saṃjñāyāṃ striyām 6.1.219 — in a feminine name (the name of something feminine, such as a river; audit H05), the ā before matup is udātta: udumbarā́vatī, śarā́vatī (udātta rā́).
• The neuter pūrvam refers to the sound-form. Some read pūrvaḥ instead (Subodhinī).
• 3706 anto 'vatyāḥ 6.1.220 — the final vowel of avatī is udātta: vetra + matup + ṅīp → vetravatī́ (udātta tī́).
• The rule is needed because ṅīp, being pit, is anudātta (anudāttau suppitau, SK 3709).
• The v that replaces m is asiddha, but it counts here because the rule depends on it. Also, in an accent rule a consonant counts as absent. So vetravatī ends in avatī (Subodhinī).
• 3707 īvatyāḥ 6.1.221 — "as before": the final vowel of an īvatī-word is udātta too: ahīvatī́, munīvatī́ (udātta tī́).
• The long ī is from śarādīnāṃ ca, and the v comes from saṃjñāyām (Subodhinī).
• In the print the phiṭ-sūtras follow here.

II.C. The phiṭ-sūtras open: the stem's final vowel is udātta, and some meaning-classes keep it — running example: phiṭ uccais → uccaíḥ (udātta caí); pāṭalā → pāṭalā́, not pā́ṭalā by laghāvante (phiṭ 42)

• phiṭ 1 phiṣo 'nta udāttaḥ — "phiṭ" is the earlier teachers' word for the prātipadika (Subodhinī), and its final vowel is udātta: uccaíḥ (udātta caí). The heading [atha phiṭsūtrāṇi; prathamaḥ pādaḥ] is supplied from the print.
• The Subodhinī justifies citing non-Pāṇinian sūtras: the Bhāṣya on ādyudāttaś ca presupposes antodātta and wholly anudātta stems.
• phiṭ 2 pāṭalāpālaṅkāmbāsāgarārthānām — words meaning pāṭalā (a plant), apālaṅka (a tree), ambā ("mother") or sāgara ("ocean") have the udātta on the final vowel.
• Against laghāvante (phiṭ 42), which applies to both synonym sets (Subodhinī): pāṭalā́, phaleruhā́, surūpā́, pākalā́; apālaṅká, vyādhighātá, ārevatá, āragvadhá.
• Against unarvannantānām (phiṭ 32, ṛ-final): mātā́.
• sāgaráḥ, samudráḥ.
• In the feminines, the stem's final udātta and the anudātta ṭāp merge into one udātta ā (translator).
• phiṭ 3 gehārthānām astriyām — words meaning "house", when not feminine, have the udātta on the final vowel: gehám (udātta hám). Without this rule, nabviṣayasya (phiṭ 26) would make it ādyudātta.
• Counter-example: śā́lā (udātta śā́). The exclusion itself shows (jñāpaka) that śālā is ādyudātta.
• phiṭ 4 gudasya ca — guda, when not feminine (astriyām from phiṭ 3), has the final vowel udātta: gudám (udātta dám).
• The feminine is ādyudātta: āntrébhyas te gúdābhyaḥ ("from your entrails, from your guts"; udātta gú). svāṅgaśiṭām (phiṭ 29) applies to guda first, as antaraṅga, and ṭāp comes after (the print's closing sentence, supplied).

II.D. Final udātta set by the word's ending: dh or y before the final, kha, and hiṣṭha, vatsara, ti, śat, tha — running example: chāyā → chāyā́ (udātta yā́), not chā́yā by yāntasyāntyāt pūrvam (phiṭ 62)

• phiṭ 5 dhyapūrvasya strīviṣayasya — in a word that is always feminine (nityastrīliṅga, Subodhinī), a final vowel preceded by dh or y is udātta.
• antardhā́: without this rule, strīviṣayavarṇa- (phiṭ 43) would make it ādyudātta.
• chāyā́, māyā́, jāyā́ (udātta yā́): without it, phiṭ 62 would apply.
• Counter-example for "feminine": bā́hyam (yañ).
• Counter-examples for "restricted": íbhyā (yat, SK 3701) and kṣatríyā (phiṭ 62). These words are also masculine.
• phiṭ 6 khāntasyāśmādeḥ — a kha-final word whose first letter is not ś or m (a-śmādi, Subodhinī) has the udātta on the final vowel.
• nakhám (against phiṭ 29).
• ukhā́ (against phiṭ 31: a pot is a made thing).
• sukhám, duḥkhám (against phiṭ 26).
• Counter-examples: múkham (by phiṭ 29 [or phiṭ 26]); śíkhā, either from the nit uṇādi affix (śīṅaḥ kho nid dhrasvaś ca; our uṇādi 702 reads śīṅo hrasvaś ca) or by phiṭ 29 applied before ṭāp, as antaraṅga.
• phiṭ 7 hiṣṭhavatsaratiśatthāntānām (print) — words ending in hiṣṭha, vatsara, ti, śat or tha have the udātta on the final vowel.
• bahula + iṣṭhan (baṃhi) → baṃhiṣṭhá (udātta ṣṭhá), against the nit ādyudātta (SK 3686). In báṃhiṣṭhair áśvaiḥ… and yád báṃhiṣṭhaṃ… it is ādyudātta by interchange (vyatyaya).
• saṃvatsaráḥ (print: parivatsaraḥ). The kaumudī reports that the accent an indeclinable first member keeps (avyayapūrvapadaprakṛtisvara) is set aside here.
• saptatíḥ, aśītíḥ, catvāriṃśát, against phiṭ 42.
• abhyū̀rṇvānā́ prabhṛthásyāyóḥ: ū̀ is the independent svarita from abhi + ū, nā́ is udātta, and the thá of prabhṛthá is udātta. This tha is already covered by the thāthādi sūtra (SK 3878).

II.E. dakṣiṇa, kṛṣṇa, śukla and gaura: final or first-syllable udātta depending on sense — running example: dakṣiṇa "skilful" → dakṣiṇá (udātta ṇá); in other senses dákṣiṇa (udātta dák)

• phiṭ 8 dakṣiṇasya sādhau — in the sense "skilful" (vīṇāyāṃ dakṣiṇaḥ, Subodhinī), dakṣiṇa has the udātta on the final vowel: dakṣiṇá.
• Denoting relative position, it is a sarvanāman and ādyudātta by phiṭ 29.
• In other senses, its heavy syllable is udātta by phiṭ 42: dákṣiṇa.
• The kośa line is supplied from the print.
• phiṭ 9 svāṅgākhyāyām ādir vā — when dakṣiṇa names a body part, the first and the last syllable are udātta by turns: dakṣiṇo bāhuḥ = dákṣiṇaḥ / dakṣiṇáḥ.
• Why "ākhyā"? For someone facing west, the left hand is the "southern" one. There dakṣiṇa states a position, not the name of a limb.
• phiṭ 10 chandasi ca — in the Veda the same alternation holds even when dakṣiṇa does not name a body part (asvāṅga): dákṣiṇaḥ / dakṣiṇáḥ.
• phiṭ 11 kṛṣṇasyāmṛgākhyā cet — in the Veda (chandasi from phiṭ 10, Subodhinī), kṛṣṇa has the udātta on its final vowel unless it names an animal. This overrides the ādyudātta of varṇānāṃ taṇa- (phiṭ 33).
• kṛṣṇó nonāva vṛṣabháḥ (udātta ṣṇó, bhá); kṛṣṇānāṃ vrīhīṇām.
• As the name of an animal: kṛṣṇo rātryai. Here the word is kṛ́ṣṇaḥ by phiṭ 33 (translator inference).
• phiṭ 12 vā nāmadheyasya — as a proper name, kṛṣṇa takes the final udātta only optionally. The other form is phiṭ 33's ādyudātta. ayáṃ vā kṛṣṇó aśvinā (udātta yáṃ, ṣṇó); kṛṣṇarṣiḥ.
• phiṭ 13 śuklagaurayor ādiḥ — the first syllable of śukla and gaura is udātta.
• "Some" make this invariable. Bhaṭṭoji holds that "vā" carries on, because sáro gauró yáthā piba shows a final udātta (udātta ró).
• The Subodhinī's "eke" carry nāmadheyasya over, and its comment on SK 3932 applies the rule as gaurasyādyantāv udāttau.

II.F. More final-udātta words, by sense or in the Veda; pāda 1 closes — running example: jyeṣṭha "eldest" → jyeṣṭhá (udātta ṣṭhá), against jyéṣṭhaḥ "best" (nit iṣṭhan, SK 3686; udātta jyé)

• phiṭ 14 aṅguṣṭhodakabakavaśānāṃ chandasy antaḥ — in the Veda, aṅguṣṭha, udaka, baka and vaśā have the udātta on the final vowel.
• aṅguṣṭhá: without this rule, svāṅgānām akurvādīnām (phiṭ 52) would make its second syllable udātta. baka is set against phiṭ 30 (Subodhinī).
• vaśā is named as a restriction (niyama): it is antodātta only in the Veda, so in ordinary speech it is ādyudātta, "they say".
• udaka is udakám in the Veda. In bhāṣā it belongs to kardamādi (phiṭ 59) (Subodhinī).
• phiṭ 15 pṛṣṭhasya ca — pṛṣṭha has the udātta on its final vowel in the Veda: pṛṣṭhám. In bhāṣā this is optional (kaumudī): pṛṣṭhám / pṛ́ṣṭham, the second by phiṭ 29 (Subodhinī).
• phiṭ 16 arjunasya tṛṇākhyā cet — arjuna as the name of a grass has the udātta on its final vowel (arjuná). This is an apavāda of unarvannantānām (phiṭ 32). The tree is ādyudātta: arjuno vṛkṣaḥ = árjunaḥ (Subodhinī).
• phiṭ 17 aryasya svāmyākhyā cet (print; ours: āryasya) — arya meaning "master" (a nipātana of SK 2851) has the udātta on its final vowel: aryá. Otherwise ādyudātta would come by phiṭ 62 or by yato 'nāvaḥ (SK 3701).
• phiṭ 18 āśāyā adigākhyā cet — āśā has the udātta on its final vowel when it does not mean "direction".
• By jñāpaka, a synonym of "direction" is ādyudātta: índra ā́śābhyas pári (udātta ā́).
• The Subodhinī's gloss "ādyudāttaḥ" here is a likely slip for antodāttaḥ.
• phiṭ 19 nakṣatrāṇām abviṣayāṇām — names of lunar mansions that take āp have the udātta on the final vowel:
• āśleṣā́ and anurādhā́, against phiṭ 42.
• jyeṣṭhā́, śraviṣṭhā́ and dhaniṣṭhā́, against the nit iṣṭhan (SK 3686).
• phiṭ 20 na kupūrvasya kṛttikākhyā cet — no final udātta for the nakṣatra Kṛttikā, where a ku stands before the final: kṛ́ttikā (udātta kṛ́ by phiṭ 42, Subodhinī).
• "Some" read it as "of words restricted to an āp preceded by ku" (print: āp tad-), so that āryikā and bahulikā are covered too.
• phiṭ 21 ghṛtādīnāṃ ca — an open class (ākṛtigaṇa) with the udātta on the final vowel: ghṛtám mimikṣe (udātta tám). This overrides nabviṣayasya (phiṭ 26) (Subodhinī).
• phiṭ 22 jyeṣṭhakaniṣṭhayor vayasi — in the sense of age, jyeṣṭha and kaniṣṭha have the udātta on the final vowel. This is an apavāda of the nit accent (Subodhinī).
• Examples: jyeṣṭhá āha camasā́, kaniṣṭhá āha catúraḥ (udātta ṣṭhá).
• jya replaces vṛddha by SK 2013, and kan replaces yuvan by SK 2019.
• In the senses "best" and "least" (jya for praśasya, SK 2009 and SK 2011; kan for alpa): jyéṣṭhaḥ, kániṣṭhaḥ, ādyudātta by SK 3686.
• phiṭ 23 bilvatiṣyayoḥ svarito vā — the final vowel of bilva and tiṣya is optionally svarita, otherwise udātta: bilvà / bilvá, tiṣyà / tiṣyá. The kaumudī gives no examples.
• The Subodhinī's "udātta ity anuvṛtteḥ" here is the evidence that udātta runs through pāda 1. The colophon [iti … prathamaḥ pādaḥ] is supplied from the print.

II.G. The ādi-heading: the first syllable is udātta, by gender, number of vowels, numeral, body part or pronoun, ku, and colour — running example: vána (neuter-only, not is-final) → váne (udātta vá, phiṭ 26); páñca (udātta páñ, phiṭ 28)

• phiṭ 24 athādiḥ prāk śakaṭeḥ — a heading, with [dvitīyaḥ pādaḥ] supplied. In the rules before śakaṭiśakaṭyoḥ (phiṭ 69), the first syllable is udātta. Its "udātta" comes from Bhaṭṭoji's glosses and the Subodhinī at phiṭ 23.
• phiṭ 25 hrasvāntasya strīviṣayasya — feminine-only words ending in a short vowel: báliḥ, tánuḥ (udātta bá, tá). The print may read vali "fold" (class C).
• phiṭ 26 nabviṣayasyānisantasya — neuter-only words (nap = napuṃsaka) that do not end in is: váne ná vāyáḥ (vána, udātta vá).
• Counter-example: sarpíḥ (isi, uṇādi 265; udātta pí).
• phiṭ 3, 6 and 21 cite this as the rule they set aside.
• phiṭ 27 tṛṇadhānyānāṃ ca dvyaṣām — grass and grain words with two vowels (aṣ = ac): kúśāḥ, kā́śāḥ, mā́ṣāḥ, tílāḥ.
• Counter-example with many vowels: godhū́māḥ (udātta dhū́ by phiṭ 42, Subodhinī).
• phiṭ 28 nraḥ saṅkhyāyāḥ — numerals ending in n or r: páñca (udātta páñ).
• catvāraḥ is left unmarked: its udātta comes from the augment ām (SK 331), and the Subodhinī's mūla read the example catuṣkapāla instead.
• phiṭ 29 svāṅgaśiṭām adantānām — a-final words for parts of one's own body, and śiṭ words (= sarvanāman):
• kárṇābhyāṃ chúbukād ádhi, [óṣṭhāv iva mádhu] (udātta kár, chú, ó).
• víśvo víhāyāḥ (udātta ví of the pronoun viśva).
• phiṭ 4, 6, 8 and 15 cite this rule.
• phiṭ 30 prāṇināṃ kupūrvam — in words for living beings, the first syllable is udātta when it stands before a ku: kā́kaḥ, vṛ́kaḥ, śúkeṣu.
• Counter-example (not living beings): kṣīrám sarpír mádhūdakám. udaká has its udātta on the final syllable by phiṭ 14.
• [dhanuḥ] is not in the print.
• phiṭ 31 khayy uvarṇaṃ kṛtrimākhyā cet — in the name of something made, a u followed by a khay consonant is udātta: kandúkaḥ (udātta dú). phiṭ 6 applies it to ukhā.
• phiṭ 32 unarvannantānām — words ending in una, ṛ or van:
• váruṇaṃ vo riśā́dasam (udātta vá).
• svásāraṃ tvā kṛṇavai (udātta svá).
• pī́vānaṃ meṣám (udātta pī́).
• phiṭ 2 (mātṛ) and phiṭ 16 (arjuna) cite this rule.
• phiṭ 33 varṇānāṃ taṇatinitāntānām — colour words ending in ta, ṇa, ti, ni or t, one example for each ending in order: étaḥ, háriṇaḥ, śítiḥ, pṛ́śniḥ, hárit. phiṭ 11 overrides it for kṛṣṇa.

II.H. First-syllable udātta for particular words and for lup-names, with the exceptions — running example: cañcā iva → (kan, SK 2051, elided by lup, SK 2053) cáñcā (udātta cáñ: a man like a straw figure)

• phiṭ 34 hrasvāntasya hrasvam anṛttācchīlye — in a word ending in a short vowel, the first short vowel other than ṛ is udātta, when habitual disposition is meant (the translator's construal of tācchīlye): múniḥ (udātta mú).
• anṛt excludes nṛśaṃsaḥ (Subodhinī).
• phiṭ 35 akṣasyādevanasya — akṣa not meaning "gaming": tásya nā́kṣaḥ. The udātta of á is kept in the merged nā́ (SK 3658).
• In the sense of gaming: akṣaír mā́ dīvyaḥ (udātta kṣaí).
• phiṭ 36 ardhasyāsamadyotane — ardha meaning "a part": árdho grāmasya (udātta ár).
• Meaning an equal half: ardhaṃ pippalyāḥ, which is not covered by this rule.
• phiṭ 37 pītadrvarthānām — words meaning pītadru: pī́tadruḥ, sáralaḥ. This may also be read as one example, "pītadru, i.e. sarala".
• phiṭ 38 grāmādīnāṃ ca — grā́maḥ, sómaḥ, yā́maḥ. The Subodhinī supplies ādir udāttaḥ.
• phiṭ 39 lubantasyopameyanāmadheyasya — the name of a thing compared, ending in lup: cáñcā.
• Counter-example: agnir māṇavakaḥ (no lup; Subodhinī).
• A variant reading is sphigantasya: sphik is the earlier grammarians' (prācām) name for lup.
• phiṭ 40 na vṛkṣaparvataviśeṣavyāghrasiṃhamahiṣāṇām — blocks phiṭ 39 for particular trees and mountains and for vyāghra, siṃha and mahiṣa. In tāla iva tālaḥ, merur iva meruḥ, vyāghraḥ, siṃhaḥ and mahiṣaḥ, the first syllable is not udātta.
• The generic words vṛkṣa and parvata still take phiṭ 39 (Subodhinī).
• phiṭ 41 rājaviśeṣasya yamanvā cet — a niyama on phiṭ 39: the name of a particular king takes it only if the name is yamanvā (= vṛddha, SK 1335). ā́ṅgaḥ (udātta ā́); counter-example aṅgāḥ.
• The Subodhinī reports two forms of the niyama without choosing between them: one that "they say" (āhuḥ), and the other the Phiṭ-vṛtti's.

II.I. The heavy-syllable rule and the rules around it — running example: kalyāṇa (final ṇa light; heavy kal and yā) → a heavy syllable is udātta; kukkuṭa → kukkúṭaḥ (light penultimate kú)

• phiṭ 42 laghāv ante dvayoś ca bahvaṣo guruḥ — in a word with many vowels, if the final syllable is light, or the last two are, a heavy syllable is udātta: kalyāṇaḥ, kolāhalaḥ.
• ādiḥ is not construed here, so a heavy syllable that is not initial also takes it, as in vṛṣākapi (SK 494) (Subodhinī).
• The Subodhinī's "by turns" (paryāyeṇa) means either kályāṇaḥ and kolā́halaḥ, one for each word, or either heavy syllable in each word. The English leaves both unmarked.
• This is the rule set aside at phiṭ 2, 7 and 19. It applies at phiṭ 8 (dákṣiṇa), 20 (kṛ́ttikā) and 27 (godhū́ma).
• phiṭ 43 strīviṣayavarṇākṣupūrvāṇām — the first syllable is udātta in:
• feminine-only words: mállikā;
• colour words: śyénī, háriṇī;
• akṣu with something before it: tárakṣuḥ (tara + akṣu).
• phiṭ 5 cites this rule as set aside for antardhā.
• phiṭ 44 śakunīnāṃ ca laghu pūrvam — in words for birds, the light syllable before the final is udātta. ādiḥ is not construed here (Subodhinī). kukkúṭaḥ, tittíriḥ (udātta kú, tí).
• The print adds khañjarīṭaḥ, which is doubtful because its rī is heavy.
• phiṭ 45 nartuprāṇyākhyāyām — in names of seasons and of living beings, the udātta that a rule would otherwise give does not come: vasantaḥ, kṛkalāsaḥ.
• The barred rule is presumably phiṭ 42, which would make san and lā udātta (translator). The resulting accent is not stated.
• phiṭ 46 dhānyānāṃ ca vṛddhakṣāntānām — grain words that are vṛddha and end in k or ṣ have the first syllable udātta: śyā́mākāḥ (udātta śyā́), rā́jamāṣāḥ (udātta rā́; print: māṣāḥ).
• The Subodhinī objects that śyāmāka and māṣa are already covered by uśīra- (phiṭ 67) and by phiṭ 27, so naiṣādakāḥ and rājamāṣāḥ should have been cited; the print reads tan na here.

Part III — phiṭ 47 – SK 3718: the end of the phiṭ-sūtras' second pāda (ādyudātta); the third pāda (second syllable udātta, first-or-second options, fixed places); the fourth pāda (syllables in turn, svarita words, anudātta words, particles and upasargas, doublings); then the "remaining accent rules" for affixes, cit and kit finals, and udātta vibhaktis

Covers phiṭ 47–87 and SK 3708–3718 (52 sūtras): all present.

• The spine. This part has two sources, and only the second is the Aṣṭādhyāyī.
• phiṭ 47–87 (sk100047–sk100087) are Śāntanava's phiṭ-sūtras. The print sets them between SK 3707 and 3708. They have no AS-refs. Our source.json numbers them with a pseudo-ref "10.pāda.n". sk100059 carries "10.4.59", but phiṭ 59 belongs to the third pāda, so the ref should be 10.3.59 (metadata only).
• Pādas. phiṭ 47–49 close the second pāda; 50–68 are the third; 69–87 are the fourth. The print's colophons follow phiṭ 49, 68 and 87, and its headings stand before 50 and 69. 87's colophon is "iti śāntanavācāryapraṇīteṣu phiṭsūtreṣu turīyaḥ pādaḥ". Ours lacks all of these, so each is supplied in brackets.
• SK 3708–3718 go back to the Aṣṭādhyāyī under the print's heading "atha svaraprakaraṇaśeṣāḥ 72", which is supplied in brackets before 3708. The run is 3.1.3–3.1.4 (3708–3709), the affix-accent rules after pratyayaḥ and paraś ca (3.1.1–2 = SK 180–181). Then come 6.1.163–6.1.172 (3710–3718), except 6.1.167 caturaḥ śasi, which was taught earlier as SK 3682. 6.1.158–162 are SK 3650, 3680, 3681, 3651 and 3671, and 6.1.173 opens Part IV at SK 3719.
• AS-refs. All 11 were checked against references/ashtadhyayi/sutraani-data.json by SK number and by sūtra text, and all match. So were the cross-references cited below: SK 180, 1009, 2140, 2146, 2147, 3650, 3652, 3657, 3660, 3668, 3671, 3680, 3683, 3686, 3701, 3729, 3734, 3873.
• The headings and how far they run (adhikāra = a governing heading; anuvṛtti = a word carried forward from an earlier sūtra).
• udātta. The word goes back to phiṭ 1 (phiṣo 'nta udāttaḥ).
• ādiḥ ("the first syllable"). It runs from phiṭ 24 (athādiḥ prāk śakaṭeḥ) to just before phiṭ 69, where 69 itself marks the limit. It governs 47–49, 67 and 68, and 66 restates it.
• dvitīyam ("the second syllable") is the adhikāra of phiṭ 50. It runs to just before phiṭ 66 (prāg īṣāt). tryacām ("of three-vowel words") is the adhikāra of phiṭ 51. It runs to just before phiṭ 57 (prāṅ makarāt).
• The fourth pāda. Bhaṭṭoji's glosses repeat the "each syllable in turn" of phiṭ 69 at 70–72. At 73 the translator supplies it, flagged as a supplement. Other words run on as follows:
• svarita from 74 into 75–77, and antaḥ from 76 into 77.
• ādyudāttāḥ from 80 into 81 (Subodhinī).
• anudāttaḥ from 84 into 85 (Subodhinī).
• The "udātta" of 82 is the translator's supplement.
• The SK run. pratyayaḥ (SK 180) governs 3708–3709. "The final is udātta" runs on into 3710–3718. Naming karṣātvato ghaño 'nta udāttaḥ (SK 3680, 6.1.159) as its source is the translator's identification. ekācaḥ and tṛtīyādir vibhaktiḥ run on from 3714: 3715 and 3717 take both (Subodhinī), 3716 takes "vibhakti" and "from the third onward", and 3718 takes "vibhakti". antodāttāt runs from 3715 into 3717 (Bhaṭṭoji) and 3718 (Subodhinī). asarvanāmasthānam runs from 3716 into 3717–3718. Wherever a sutra_en names the source rule of a carried word and no commentator does, the identification is the translator's (audits H10–H12). This outline marks those cases "supplied".
• Accent convention and the print's signs. Accent is written in IAST: udātta acute (á), independent svarita grave (à), anudātta unmarked, and dependent svaritas left unmarked.
• Where the marks come from. Up to phiṭ 73, and in most of the SK run, neither our text nor the print marks the examples. There the outline marks only the syllable the rule names, as the en.yml does. In 74, 78, 85, 87 and 3710–3717 the marks are decoded from the print's signs.
• Where the print's signs contradict the rule, the [orch] SK 3701 policy is followed: the print's signs stay in the block, and a note gives the rule's form.
• phiṭ 78 símasmai is printed with udātta on sí, against the rule.
• phiṭ 78 utá tvaḥ páśyan follows the print. Ours marks paśyán.
• phiṭ 85 táṃ nemím ṛbhávo yathā follows the print. Ours would make thā udātta.
• Print readings followed (NSP 1915, scans page-0636–0639; collation batch13–15; audits H09–H12):
• The print's reading is followed: phiṭ 50 hayādeḥ (ours halādeḥ); phiṭ 68 aṣāḍhā (ours āṣāḍhā); phiṭ 74 nyaṅṅ uttānaḥ, vyacakṣayat svaḥ; phiṭ 78 tvaḥ (ours tvaṃ); phiṭ 82 putra; phiṭ 87 nityādi (ours ity ādi).
• Ours or Wikisource is kept where the print lacks or misprints:
• phiṭ 47: the print lacks the sentence "[stated because phiṭ 62 would apply]", so it is bracketed.
• phiṭ 56: the print has yute; ours yuta is kept.
• phiṭ 62: the print lacks talayaḥ, valayaḥ.
• phiṭ 64: the print's purūvasām is a misprint.
• phiṭ 72 dhūprabhṛtīn (C).
• phiṭ 75–77: the print lacks most of the examples, and its sūtra for 76 is a different one (tilyaśikyakāśmarya… rāja-).
• phiṭ 77 tato bilva (print nano vilvasya).
• phiṭ 85: the print's padānte is a misprint.
• phiṭ 86. The print sets parasyānta udāttaḥ inside the sūtra. Ours divides it as gloss and is kept (class M).
• SK run. 3716 bāhū́n (ours bāhūnū, C). In 3717 the plain-copy slips (babhū, mabhyāṃ, mahe, upatatvāgne) are corrected from the accented copy and the print. 3717 apā́ṃ is an accent query (C).
• Where the Subodhinī and BM lie (only the kaumudī is translated, SK-TRANSLATE §25):
• BM. There is none anywhere in this part. Our BM layer is empty, and Wikisource uttara-4/0103–0105 prints only the mūla.
• Subodhinī filed under a neighbour:
• phiṭ 47, 48, 49, 52 and 53 under sk.100045.s1, with its own "iti dvitīyaḥ pādaḥ" after 49.
• phiṭ 57 at the tail of sk.100056.s1; phiṭ 63 at the tail of sk.100062.s1.
• phiṭ 81's tail at the end of sk.100068.s1.
• phiṭ 87, SK 3708 and SK 3709 in sk.100086.s1.
• SK 3711–3714 in sk.3710.s1.
• SK 3716 at the tail of sk.3715.s1. The head of SK 3717's commentary is at the tail of sk.3716.s1, and the rest is sk.3717.s1.
• Missing from our layer. The Subodhinī on phiṭ 78, 79 and the head of 81 stands in the last line of NSP p. 627 and is missing here. The en.yml quotes it from the print. It is a commentary-OCR fix candidate.
• No Subodhinī in print on phiṭ 50, 51, 59, 69–77, 80, 82, 83 and 84.
• Audit corrections (H09–H12, applied to en.yml):
• One MUST-FIX, phiṭ 87: the form is práprāyám, with udātta on both prá and yám.
• MINOR:
• phiṭ 48: the alternatives are marked.
• phiṭ 57: the Subodhinī reads peṣṭā.
• phiṭ 59: the gaṇa attribution of the kardamādi examples is hedged.
• phiṭ 60, 61, 81 and 85: the source rules of the anuvṛtti are owned as the translator's.
• phiṭ 63: the kecit reading is rendered literally.
• phiṭ 82: the gloss is "with your patrons, O son".
• SK 3710: the takát súte signs. SK 3716: aniditām. SK 3718: nānvanaḥ is the print's own reading.

III.A. Initial udātta, the close of the second pāda: country names, hay-initial la-finals, two-vowel ik-finals — running example: kekaya (janapada, vowel-final) → (ādiḥ udātta) kékayaḥ; kṛṣi (two vowels, ik-final) → kṛ́ṣiḥ

• phiṭ 47 janapadaśabdānām aṣāntānām — in a country name ending in an aṣ (the phiṭ name for ac), the first syllable is udātta (ādiḥ, phiṭ 24). kékayaḥ: the udātta is ké. Under phiṭ 62 it would fall on the penultimate ka.
• Ours adds "[stated because yāntasyāntyāt pūrvam (phiṭ 62) would apply]", which the print lacks.
• The Subodhinī objects that phiṭ 62, the later rule, ought to prevail (paratvāt), so aṅgāḥ, vaṅgāḥ are the right examples.
• phiṭ 48 hayādīnām asaṃyuktalāntānām antaḥ pūrvaṃ vā — hay = hal. In a consonant-initial word ending in a la that does not follow a conjunct, the first syllable is udātta, or optionally the syllable before the final. This is the Subodhinī's construal; the wording could also mean "the final, or the one before it". pálalam / palálam; śálalam / śalálam.
• Counter-examples: ekalaḥ (vowel-initial) and mallaḥ (the l follows a conjunct, ll).
• Some carry janapadānām forward (pañcālāḥ, kośalāḥ; Subodhinī).
• phiṭ 49 igantānāṃ ca dvyaṣām — in a two-vowel word ending in an ik, the first syllable is udātta. kṛ́ṣiḥ: the udātta is kṛ́. The Subodhinī reports that "some" carry janapadānām here too (kuravaḥ, cedayaḥ) and treat kṛṣi as a counter-example, because it is antodātta in akṣair mā dīvyaḥ kṛṣim it kṛṣasva. [End of the second pāda.]

III.B. The second syllable udātta in three-vowel words: the dvitīyam and tryac headings — running example: lalāṭa (svāṅga, three vowels, l-initial) → (dvitīyam udātta) lalā́ṭam; pādapa (heavy-initial, pa-final) → pādápaḥ

• phiṭ 50 atha dvitīyaṃ prāg īṣāt — adhikāra: "the second [syllable is udātta]" holds until just before īṣāntasya hayādeḥ (phiṭ 66). It supplies the lā́ of lalā́ṭam. The heading "[tṛtīyaḥ pādaḥ]" is supplied. The kaumudī quotes 66 as hayādeḥ, following the print.
• phiṭ 51 tryacāṃ prāṅ makarāt — adhikāra: "of three-vowel words" holds until just before makaravarūḍha… (phiṭ 57). la-lā-ṭa has three vowels, so it qualifies.
• phiṭ 52 svāṅgānām akurvādīnām — in a three-vowel body-part word that does not begin with a k-class consonant (ku), r or v, the second syllable is udātta. lalā́ṭam: the udātta is lā́. The counter-examples follow their own rules (Subodhinī): kapólaḥ by phiṭ 42; rásanā and vádanam by phiṭ 29.
• phiṭ 53 mādīnāṃ ca — in a three-vowel word beginning with m, the second syllable is udātta. maláyaḥ, makáraḥ: the udātta falls on la and ka.
• The Subodhinī calls these poor examples: makara has phiṭ 57 and malaya has phiṭ 62. For malaya, phiṭ 62 gives the same syllable.
• It would cite mahendra instead.
• phiṭ 54 śādīnāṃ śākānām — in a three-vowel pot-herb name beginning with ś, the second syllable is udātta. śītányā, śatápuṣpā: the udātta is tá in both.
• śatapuṣpā has four syllables, although the rule stands under the tryac heading. Neither commentator remarks on this.
• The Subodhinī reports "some" who read sādīnām and cite sarṣapā.
• phiṭ 55 pāntānāṃ gurvādīnām — in a three-vowel word ending in pa and beginning with a heavy syllable, the second syllable is udātta. pādápaḥ, ātápaḥ: the udātta is dá / tá (the second syllable).
• anūpám has a light first syllable, and nīpám has two syllables, so the rule does not reach them.
• Their accent is final, by anor apradhānakanīyasī (SK 3923) and ner anidhāne (SK 3926) (Subodhinī).
• phiṭ 56 yutānyaṇyantānām — in a three-vowel word ending in yuta, ani or aṇi, the second syllable is udātta. ayútam, dhamániḥ, vipáṇiḥ: the udātta falls on yú, má and pá. The print reads yute; ours yuta is kept.

III.C. The first or the second syllable: named words, the Veda, the kardamādi — running example: makara → (ādir vā; else dvitīyam) mákaraḥ / makáraḥ; sugandhitejana → súgandhitejanāḥ / sugándhitejanāḥ / sugandhitéjanāḥ

• phiṭ 57 makaravarūḍhapārevatavitastekṣvārjidrākṣākalomākāṣṭhāpeṣṭhākāśīnām ādir vā — for these twelve words the first syllable is udātta, or, under the dvitīyam heading still in force, the second. The tryac heading ends here, which is why the two-vowel ikṣu and umā are allowed in. mákaraḥ / makáraḥ; várūḍhaḥ / varū́ḍhaḥ. The Subodhinī reads pārāvata and peṣṭā. It says kāśīnānām would be expected and calls the missing nuṭ a sage's licence (ārṣa).
• phiṭ 58 chandasi ca — in the Veda, words other than makara and the rest also take the udātta on the first or the second syllable, as usage shows (lakṣyānusārāt). ādir vā is supplied from 57. The Subodhinī's examples are kāśyapaḥ, kardamaḥ, kulaṭā, udakam, gāndhāriḥ, given unmarked; the last four may belong to phiṭ 59.
• phiṭ 59 kardamādīnāṃ ca — for kardama and the rest, the first or the second syllable is udātta (Bhaṭṭoji's gloss). The rule is not limited to the Veda: the Subodhinī on udaka says "bhāṣāyām". There is no example. The print reads ādir dvitīyaṃ vodāttam.
• phiṭ 60 sugandhitejanasya te vā — in sugandhitejana the syllable te is udātta, or else the first or the second. The three take the udātta in turn: súgandhitejanāḥ / sugándhitejanāḥ / sugandhitéjanāḥ (su, gan, te). The Subodhinī says the "first" is the first syllable of the whole word.
• phiṭ 61 napaḥ phalāntānām — in a neuter (nap) word ending in phala, the first or the second syllable is udātta. rā́jādanaphalam / rājā́danaphalam. The Subodhinī's counter-example dāsīphalo vṛkṣaḥ is a masculine bahuvrīhi.

III.D. Fixed places: the penultimate, nā and a light syllable; the dvitīyam heading closes at īṣā; named lists with initial udātta — running example: kulāya (ya-final) → (antyāt pūrvam) kulā́yaḥ; halīṣā (hay-initial, īṣā-final) → (ādir vā) hálīṣā; mahiṣī (wife) → máhiṣī

• phiṭ 62 yāntasyāntyāt pūrvam — in a word ending in the syllable ya, the syllable before the last is udātta. kulā́yaḥ, taláyaḥ, valáyaḥ: the udātta falls on lā́ and lá. The print gives only kulāyaḥ.
• phiṭ 63 thāntasya ca nālaghunī — in a word ending in tha, the syllable nā and a light syllable are udātta. sanā́thā sabhā: the udātta is marked on nā́ alone.
• The rule does not say where the light syllable stands.
• The Subodhinī reports a kecit reading, āntasya: "of a word ending in ā and of the word nā, the light syllables before the last are udātta". Its examples are nā́nā, dívā, múdhā, each with a single udātta.
• phiṭ 64 śiśumārodumbarabalīvardoṣṭrārapurūravasāṃ ca — in śiśumāra, udumbara, balīvarda, uṣṭrāra and purūravas, the syllable before the last is udātta (from 62, by Bhaṭṭoji's gloss), or the second. There are no examples. The Subodhinī gives the variant śiṃśumāra. The print's purūvasām is a misprint.
• phiṭ 65 sāṅkāśyakāmpilyanāsikyadārvāghāṭānām — in these four names the second syllable is optionally udātta. There are no examples. The Subodhinī notes how they are formed: sāṅkāśya and kāmpilya take ṇya by vuñchaṇ- (SK 1292), and dārvāghāṭa takes aṇ (vārttika under SK 2966). Neither commentator states the accent in the other case.
• phiṭ 66 īṣāntasya hayāder ādir vā — this is the limit of the dvitīyam heading. In a consonant-initial word ending in īṣā, the first syllable is optionally udātta. hálīṣā, lā́ṅgalīṣā: the udātta falls on há and lā́. The print's halīpā is the ṣ/p letter shape.
• phiṭ 67 uśīradāśerakapālapalālaśaivālaśyāmākaśārīraśarāvahṛdayahiraṇyāraṇyāpatyadevarāṇām — in these thirteen words the first syllable is udātta (ādiḥ, phiṭ 24). The division into thirteen is the translator's. There are no examples.
• Subodhinī: the rule names hṛdaya to stop the second-syllable udātta of phiṭ 52, and most of the others to stop phiṭ 42.
• It calls the naming of śyāmāka (already covered by phiṭ 46) and devara (already covered by phiṭ 42) questionable (cintya).
• phiṭ 68 mahiṣyaṣāḍhayor jāyeṣṭakākhyā cet — mahiṣī as a wife and aṣāḍhā as a brick (respectively, per the Subodhinī) take the udātta on the first syllable. máhiṣī jāyā; áṣāḍhā upadadhāti. The print reads aṣāḍhā; ours has āṣāḍhā. [End of the third pāda.]

III.E. Each syllable in turn: the ādiḥ heading ends — running example: śakaṭi → (akṣaram akṣaraṃ paryāyeṇa) śákaṭiḥ / śakáṭiḥ / śakaṭíḥ

• phiṭ 69 śakaṭiśakaṭyor akṣaram akṣaraṃ paryāyeṇa — this is the limit of ādiḥ (phiṭ 24). In śakaṭi and śakaṭī, each syllable in turn is udātta. śákaṭiḥ / śakáṭiḥ / śakaṭíḥ; śákaṭī / śakáṭī / śakaṭī́. The heading "[caturthaḥ pādaḥ]" is supplied.
• phiṭ 70 goṣṭhajasya brāhmaṇanāmadheyasya — when goṣṭhaja is the name of a brāhmaṇa, each syllable in turn is udātta: góṣṭhajaḥ / goṣṭhájaḥ / goṣṭhajáḥ.
• Elsewhere: goṣṭhajáḥ paśuḥ ("a beast born in a cow-pen"). It is antodātta by the kṛduttarapadaprakṛtisvara, under which a final member ending in a kṛt affix keeps its own accent.
• Identifying that accent with SK 3873 is the translator's presumption.
• phiṭ 71 pārāvatasyopottamavarjam — in pārāvata every syllable except the penultimate (va) is udātta in turn: pā́rāvataḥ / pārā́vataḥ / pārāvatáḥ.
• phiṭ 72 dhūmrajānumuñjakeśakālavālasthālīpākānām adhūjalasthānām — in these four words the syllables other than dhū, ja, la, sthā (one per word) are udātta in turn. For example sthālī́pākaḥ / sthālīpā́kaḥ / sthālīpākáḥ, and dhūmrajānu with mra, jā or nu. The text does not say which la of kālavāla is left out. The pairing of word with excluded syllable is implied, not stated.
• phiṭ 73 kapikeśaharikeśayoś chandasi — kapikeśaḥ, harikeśaḥ, in the Veda. Neither Bhaṭṭoji nor any commentator says what the rule does. The translator supplies the "in turn" of 69, so no accented forms are given.

III.F. Independent svarita, fixed by name — running example: nyaṅ → (svarita) nyàṅ; tila + yat → tilya → (antaḥ svaritaḥ, not the ādyudātta of SK 3701) tilyàm

• phiṭ 74 nyaṅsvarau svaritau — nyaṅ and svar are svarita. nyàṅṅ uttānáḥ; vyàcakṣayat svàḥ. The other marks follow the print's signs. The print reads nyaṅṅ, with the ṅ doubled (ṅamuṭ), and vyacakṣayat.
• phiṭ 75 nyarbudavyalkaśayor ādiḥ — in nyarbuda and vyalkaśa the first syllable is svarita (from 74). nyàrbudāya svā́hā (svā́hā is ādyudātta by phiṭ 80); pākadūrvā vyàlkaśā. The print has no examples.
• phiṭ 76 tilyaśikyamatyakarṣmaryadhānyakanyārājanyamanuṣyāṇām antaḥ — in these words the final syllable is svarita. The rule applies where yato 'nāvaḥ (SK 3701) would put the udātta on the first syllable of a two-vowel yat-word. tilyàm, śikyàm, matyàm, kārṣmaryàm, dhānyàm asi, rājanyò vajrasya rūpam, pratīcīṃ manuṣyā̀ḥ. The print's sūtra is tilyaśikyakāśmarya…rāja-, and its only example is tilyam.
• phiṭ 77 bilvabhakṣyavīryāṇi chandasi — in the Veda, bilva, bhakṣya and vīrya take svarita on the final syllable. tato bilvà ud atiṣṭhat. There is no example for bhakṣya. The example balaṃ virujavīryam is left unmarked because its word division is uncertain.

III.G. Anudātta words, particles and upasargas — running example: svāhā (nipāta) → (ādyudātta) svā́hā; eva → (antaḥ) evá; yathā at the end of a pāda → anudātta: táṃ nemím ṛbhávo yathā

• phiṭ 78 tvattvasamasimety anuccāni — tvat, tva, sama, sima are "not high", that is, anudātta (Subodhinī, print only). Examples, with the marks decoded from the print's signs:
• starī́r u tvat.
• utá tvaḥ páśyan. Ours has tvaṃ and paśyán.
• nábhantām anyaké same, with sama anudātta after the udātta ké.
• símasmai, as both witnesses print it. Its udātta on sí goes against the rule and is recorded, not corrected.
• phiṭ 79 simasyātharvaṇe 'nta udāttaḥ — in the Atharvaveda, sima has its final syllable udātta. The restriction is prāyikam (it states what usually holds), or it means "sima as seen there". So the accent also holds in the Ṛgveda: vāsas tanute simásmai, with the udātta on má.
• phiṭ 80 nipātā ādyudāttāḥ — particles take the udātta on the first syllable. svā́hā: the udātta is svā́.
• phiṭ 81 upasargāś cābhivarjam — upasargas, except abhi, are also ādyudātta (Subodhinī; carried from 80). The kaumudī is empty, only a daṇḍa. In this part the rule is visible in úpa (SK 3717) and the prá of práprāyám (phiṭ 87).
• phiṭ 82 evādīnām antaḥ — eva and the rest take the udātta on the final syllable. "udātta" is the translator's supplement. evá, evám, nūnám, sahá, as in sahá te putra sūribhiḥ ("with your patrons, O son"). The variant reading is evamādīnām. Bhaṭṭoji calls the earlier authorities' ādyudātta saha (at SK 1009) questionable (cintyam).
• phiṭ 83 vācādīnām ubhāv udāttau — in vācā and the rest both syllables are udātta. "Both" sets aside anudāttaṃ padam ekavarjam (SK 3650). There is no example.
• phiṭ 84 cādayo 'nudāttāḥ — ca and the rest are anudātta; Bhaṭṭoji says only "spaṣṭam". Presumably these are the cādi of 1.4.57, set apart from phiṭ 80 (translator's inference).
• phiṭ 85 yatheti pādānte — yathā at the end of a metrical pāda is anudātta (from 84). táṃ nemím ṛbhávo yathā, with yathā wholly low, following the print's signs. The counter-example is yáthā no áditiḥ kárat: away from the pāda's end, yáthā is ādyudātta as a nipāta (phiṭ 80).

III.H. Doubled words, and the phiṭ-sūtras close — running example: paṭu + paṭu (prakāra) → (final of the latter member udātta) paṭupaṭúḥ; dive + dive (nitya) → (latter member wholly anudātta) divé-dive

• phiṭ 86 prakārādidviruktau — in a doubling for likeness (prakāre guṇavacanasya, SK 2147) and the like, the final of the latter member is udātta. paṭupaṭúḥ ("somewhat clever"): the udātta is ṭú.
• The Subodhinī objects that SK 2146 treats the doubling as a karmadhāraya, so samāsasya (SK 3734) would already make it antodātta.
• It replies that the phiṭ-sūtras are older than Pāṇini. The print puts parasyānta udāttaḥ inside the sūtra.
• phiṭ 87 śeṣaṃ sarvam anudāttam — in other doublings ("nitya and the rest", the print's reading; nityavīpsayoḥ, SK 2140), the latter member is wholly anudātta. práprāyám: the second pra is low, and prá and yám are udātta (audit H12 MUST-FIX). divé-dive: ve is udātta and the second dive is low. [End of the fourth pāda and of the phiṭ-sūtras.]

III.I. The affix's own accent: ādyudātta pratyaya; sup and pit anudātta — running example: agni (affix ni, uṇādi 490) → (pratyaya ādyudātta) agníḥ; kṛ + tavya → kartávyam; yajña + sya (sup) → (anudātta) yajñásya

• 3708 ādyudāttaś ca 3.1.3 — an affix (pratyayaḥ, SK 180) takes the udātta on its first vowel ("and only that", eva). agníḥ: the affix is ni, with the udātta on ní. kartávyam: the affix is tavya, with the udātta on tá. tavyat, being tit, would be svarita by SK 3729 (Subodhinī). The heading "[atha svaraprakaraṇaśeṣāḥ]" is supplied.
• 3709 anudāttau suppitau 3.1.4 — an exception (apavāda) to 3708: sup endings and pit affixes are anudātta. yajñásya: the ending sya is low. ná yó yúcchati: the root is udātta by dhātoḥ (SK 3671), the low śap becomes a dependent svarita by SK 3660, and the low tip becomes pracaya by SK 3668.

III.J. The final udātta of cit and kit: base + affix, the cit taddhita, the kit taddhita — running example: anya + akac (cit) → (final of the whole udātta) anyaké; agni + ḍhak (kit) → āgneyáḥ

• 3710 citaḥ 6.1.163 — the final vowel of what has a cit affix is udātta. The vārttika saprakṛteḥ applies this to the whole base-plus-affix, for bahuc and akac. Examples:
• nábhantām anyaké same, with the udātta on ké of the akac form (SK 2026, Subodhinī).
• yaké sárasvatīm ánu.
• takát súte: kát is udātta, and the signs make su udātta as well.
• Naming SK 3680 as the source of "the final is udātta" is the translator's identification.
• 3711 taddhitasya 6.1.164 — the final of a cit taddhita is udātta. 3710 already covers this, but the rule is taught to block the ñit accent (SK 3686, translator's identification). kauñjāyanā́ḥ: cphañ (SK 1099) gives the cit affix, whose final is nā́. Without this rule the ñit accent would win as the later rule (Subodhinī).
• 3712 kitaḥ 6.1.165 — the final of a kit taddhita is udātta. yad āgneyáḥ: agni + ḍhak (SK 1236), with the udātta on yá.

III.K. The udātta vibhakti: after tisṛ, after monosyllables, after añc, after ūṭh, idam, pad and the rest, after aṣṭā — running example: vāc + ṭā → (monosyllable before su; third case on) vācā́; aṣṭā + bhis → aṣṭābhír

• 3713 tisṛbhyo jasaḥ 6.1.166 — after tisṛ, the final of jas is udātta. tisró dyā́vaḥ savitúḥ: the udātta is ró. When r replaces ṛ (SK 299), jas would otherwise become svarita by SK 3657 (Subodhinī).
• 3714 sāv ekācas tṛtīyādir vibhaktiḥ 6.1.168 — "sau" is the seventh-case plural ending su. After a stem that is monosyllabic before su, a vibhakti from the third case onward is udātta. vācā́ virūpaḥ: the udātta is cā́, on the ending. Counter-examples:
• rā́jño nú te: rājñ- is monosyllabic only through elision of its a, not before su.
• vidhaté rā́jani tvé: rājan is not monosyllabic.
• ná dadarśa vā́cam: the second-case ending is not covered.
• 3715 antodāttād uttarapadād anyatarasyām anityasamāse 6.1.169 — after an antodātta, monosyllabic (from 3714) final member, the third-case-onward vibhakti is optionally udātta. This holds only outside a compound prescribed under the nitya adhikāra. paramavācā́ / paramavā́cā. The Subodhinī's counter-examples are agnicit, avācā and rājadṛṣadā.
• 3716 añceś chandasy asarvanāmasthānam 6.1.170 — in the Veda, a vibhakti after añc that is not a sarvanāmasthāna is udātta.
• índro dadhīcáḥ: the udātta is cáḥ. This overrides cau (SK 3652), which would have made the prior member antodātta.
• "asarvanāmasthāna" rather than "from the third case" brings in śas: pratīcó bāhū́n.
• 3717 ūḍidampadādyappumraidyubhyaḥ 6.1.171 — after ūṭh, idam, the six pad-words (SK 228), ap, pum, rai and div, a non-sarvanāmasthāna vibhakti is udātta. ekācaḥ and antodāttāt run on. Examples by stem:
• ūṭh. praṣṭhauháḥ, praṣṭhauhā́. By the vārttika ūṭhy upadhāgrahaṇam, ūṭh counts only as the penultimate sound, so the rule does not apply in akṣadyúvā.
• idam. ebhír nṛ́bhir. Not in anvādeśa (re-mention), where aś is anudātta by SK 350: mā́bhyāṃ gā́ ánu.
• The pad-words. padbhyā́m, dadbhír, māsí-māsi, hṛdá ā́.
• ap, pum, rai. apā́ṃ phénena, puṃsáḥ, rāyā́ vayám, rāyó dhartā́.
• div. úpa tvāgne divé-dive.
• 3718 aṣṭano dīrghāt 6.1.172 — after aṣṭan in its long-ā form, a vibhakti from śas onward is udātta. aṣṭābhír daśabhiḥ: the udātta is bhír.
• The rule is an exception to jhaly upottamam (SK 3683), which still governs short aṣṭabhiḥ.
• The Subodhinī reads "long" as a jñāpaka (a wording that shows something not stated outright) that the ā of SK 371 is optional.

Part IV — SK 3719–3770: the udātta vibhakti, nadī, matup and nām after special stems and its prohibitions; tit, the lasārvadhātuka, sic, thal and rit to the close of the pratyayasvaras; samāsasya; then 6.2.1–36: the prior member that keeps its own accent in the bahuvrīhi, tatpuruṣa, karmadhāraya, dvigu, avyayībhāva and dvandva

Covers SK 3719–3770 (52 sūtras): all present.

• The Aṣṭādhyāyī spine and where the SK moves rules. The part runs 6.1.173 → 6.1.223, taking only the rules not already placed, then 6.2.1–36 unbroken:
• 6.1.173–179 (3719–3725), then 6.1.182–187 (3726–3731). 6.1.180–181 (jhaly upottamam, vibhāṣā bhāṣāyām) were taught as SK 3683–3684, among the vibhakti and dhātu accents.
• 6.1.196 (3732). 6.1.188–195 are SK 3672–3678 and 3685; 6.1.197–216 are SK 3686–3704.
• 6.1.217 (3733) closes the affix accents. 6.1.218 is SK 3679, 6.1.219–221 are SK 3705–3707, 6.1.222 (cau) is SK 3652. 6.1.223 samāsasya (3734) opens the compound accents.
• 6.2.1–6.2.36 (3735–3770) in strict order, nothing pulled forward or deferred.
• AS-refs. All 52 were checked against references/ashtadhyayi/sutraani-data.json by SK number, and all match. So were the cross-referenced rules cited below (SK 3657 = 8.2.4, 3658 = 8.2.5, 3673, 3676, 3677, 3686, 3701, 3710, 3714–3717, 3783, 3873, 3878, 3906, 808, 809, 2009, 2011, 2019).
• Headings, adhikāras and how far they run. Our text lacks two print headings; both are supplied in brackets (collation class M): the colophon iti pratyayasvarāḥ after 3733, and the heading samāsasvarāḥ before 3734.
• For the vibhakti rules: antodāttāt (SK 3715) runs through 3719–3724 and stops at 3725 (Subodhinī). The śas-onward vibhakti (SK 3716) runs into 3719–3721, and the vibhakti (SK 3714) into 3725 and 3727.
• In the matup and nām rules: hrasva and matup (3722) carry into 3723, the latter turned locative (Subodhinī), and nām (3723) into 3724.
• In the prohibitions: na (3726) carries into 3727–3728, ādiḥ (3725) into 3727 (Subodhinī), and jhal (3727) into 3728. anyatarasyām in 3732 comes from SK 3677 (Subodhinī).
• In 6.2: prakṛtyā pūrvapadam (3735) is the adhikāra of all of 6.2 to 6.2.63, so it covers the whole of 3735–3770 and runs on into Part V. tatpuruṣe (3736) runs through 3757. karmadhāraye (3759) runs into 3760–3762, kumāraḥ (3760) into 3761–3762, and ādiḥ (3761) into 3762. dvigau (3763) runs through 3765 and anyatarasyām (3764) into 3765. dvandve (3768) runs into 3769–3770.
• The content of each of these is in Bhaṭṭoji's glosses. Where a sutra_en names the source rule, that is the translator's identification, owned in a note (audits H13, H16).
• Print readings followed (NSP 1915, scans 0639–0647; collation batch15–17):
• 3720: kṛṇvaté savitré (the print's signs). 3721 sét pṛ́śniḥ. 3722: svaravidhau vyañjanam avidyamānavat (ours drops m a, which reverses the sense), and reśabdāc ca labelled as a vārttika. 3725 ṣaḍbhíḥ (C).
• 3726 dyumnám (C). 3727 upa tvāgne. 3728 nṛbhiḥ. 3729 kva nūnam. 3730 anhviṅoḥ (example hnute), vṛtrāṇy, katīha.
• 3736 bhojyoṣṇam. 3737 lohitakalmāṣaḥ. 3742 ṅīṣ (C). 3750 and 3763: the examples our text doubles are given once. 3753 aiśvaryavācini and vākpatiḥ. 3768 andhakavṛṣṇayaḥ and rājanyagrahaṇam.
• The "(vārttikam)" tags in our text are editorial and are dropped (3730, 3736, 3746, 3763). In 3763, tam adhīṣṭo… and dvigor yap are Aṣṭādhyāyī sūtras (5.1.80 = SK 1744, 5.1.82 = SK 1746), not vārttikas.
• Accent policy ([orch]): the print's signs are kept in the block. A syllable is marked only where the accented copy, Bhaṭṭoji or the Subodhinī fixes it. Where the translator derives it from a named affix, that is flagged below as the translator's step. Counter-examples are left unmarked.
• Where the Subodhinī and BM lie (only the kaumudī is translated, SK-TRANSLATE §25):
• BM. None. Our BM layer is empty for all 52 units, and Wikisource uttara-4/0105–0109 prints only the mūla.
• Subodhinī filed under a neighbour:
• 3719: the tail of sk.3718.s1. 3721–3722: in sk.3720.s1. 3726: the tail of sk.3725.s1.
• 3731–3736: in sk.3730.s1. 3737: in sk.3737.s1 plus sk.3730.n1, which is the Subodhinī run-on, not an NSP footnote.
• 3744–3745: in sk.3743.s1. 3749: the tail of sk.3748.s1. 3753, 3754 and the head of 3755: in sk.3752.s1. 3757: in sk.3756.s1. 3769: the tail of sk.3768.s1.
• A missing band. The whole Subodhinī band of NSP p. 636 (scan 0646) is missing from our layer: the rest of 3759, all of 3760–3766, and the head of 3767. The translators read it from the scan. The rest of 3767 is in sk.3758.s1.
• No Subodhinī in print on 3724, 3728 or 3751.
• Audit corrections followed (H13–H16, applied to en.yml):
• H13 (MUST-FIX): in 3726, śunaḥśepa is vanaspatyādi, so both members keep their accent (śúnaś cic chépam). In 3730 the form is indhé rā́jā.
• H14 (MUST-FIX): in 3733, riti is a tadanta locative: the udātta is the penultimate syllable of the word ending in the rit affix. In 3732 the rule likewise applies in a word ending in thal with iṭ. 3738 arítra- and 3741 mū́tra- are now marked. In 3740 the bracketed "[an act]" is dropped.
• H15: in 3746 the dvigu denotes the measure.
• H16: the Subodhinī's gaṇa had paṭu (3758, a variant). 3766 sāṅkāśyá-, 3768 śvāphalká- and 3769 dvā́daśa are now marked (translator's steps). In 3767 pari means "except", not "all round".
• Where Part IV ends. It ends at 3770 ācāryopasarjanaś cāntevāsī (6.2.36). Part V opens with kārtakaujapādayaś ca (6.2.37, SK 3771), still under prakṛtyā pūrvapadam in the dvandva.

IV.A. The nadī and the śas-onward vibhakti made udātta after śatṛ and after a yaṇ — running example: jñā + śnā + śatṛ → jānat (antodātta) + ṅīp → jānatī́ (udātta tī); netṛ (tṛc, antodātta) + ṅīp → (ṛ → r) netrī́ (udātta trī)

Accent: in this group the following element (the nadī ī or the vibhakti) takes the udātta that the stem-final would otherwise keep. In 3721, where that udātta is denied, the vibhakti is an independent svarita.

• 3719 śatur anumo nadyajādī 6.1.173 — after an antodātta stem in śatṛ without num, a nadī and a vowel-initial vibhakti from śas on are udātta.
• The stem: jñā → jā (SK 2511), and the ā of nā is lost (SK 2483), giving jānat. Then jānat + ṅīp → ácchā rávaṃ prathamā́ jānatī́ (udātta on the nadī, tī). With the vibhakti: kṛṇvaté (udātta té).
• Not after a stem that is not antodātta: dádhatī, where the first syllable is udātta by SK 3673.
• Not where there is num: tudántī. The śatṛ after the a of śa is anudātta (SK 3730), and the udātta is the ekādeśa (SK 3658). The Subodhinī takes "anumaḥ" as a jñāpaka that the ekādeśa is not asiddha here.
• 3720 udāttayaṇo halpūrvāt 6.1.174 — after a yaṇ that replaces an udātta vowel and follows a consonant, the nadī and the śas-onward vibhakti are udātta. SK 3657 would otherwise make them svarita.
• tṛc stems are antodātta by citaḥ (SK 3710). So codayitṛ + ṅīp → codayitrī́ sūnṛ́tānām, eṣā́ netrī́, and savitṛ + ṅe → ṛtáṃ devā́ya kṛṇvaté savitré: the udātta is on trī́, trī́, tré.
• The Subodhinī would add n to the rule: vākpatnī.
• 3721 noṅdhātvoḥ 6.1.175 — not after the yaṇ of ūṅ or of a dhātu. Only the śas-onward vibhakti is meant; the Subodhinī says no nadī can follow these.
• The vibhakti, denied the udātta, becomes svarita by SK 3657: brahmabandhū + ṭā → brahmabandhvā̀ (independent svarita on vā̀).
• subhū (kvip; antodātta by SK 3873) + ṅe → sét pṛ́śniḥ subhvè.

IV.B. Udātta matup and nām; the consonant-initial vibhakti after ṣaṭ, tri and catur — running example: abdi (antodātta) + matup → abdimā́n (udātta mā); ṣaṣ + bhis → ṣaḍbhír (udātta bhír)

• 3722 hrasvanuḍbhyāṃ matup 6.1.176 — after an antodātta stem ending in a short vowel, and after nuṭ, matup is udātta.
• abdi (ki, SK 3271) and udani (ghṛtādi, phiṭ 21) are both short-final antodātta stems: yó abdimā́m̐ udanimā́m̐ íyarti (udātta mā́m̐).
• With nuṭ: akṣan + matup → (n lost; nuṭ) akṣaṇvántaḥ kárṇavantaḥ (udātta ván).
• Not after a stem that is not antodātta: mā́ tvā vidad íṣumān (iṣu is ādyudātta).
• The maxim svaravidhau vyañjanam avidyamānavat does not hold here: marútvān, niyútvān, where matup stays anudātta.
• Vārttika reśabdāc ca: rayi → re leaves no short vowel, yet matup is still udātta: revā́m̐ íd revátaḥ.
• 3723 nām anyatarasyām 6.1.177 — after such a stem whose vowel is short before matup, nām is optionally udātta.
• sumati (bahuvrīhi, antodātta by SK 3906) + nām → cétantī sumatīnā́m (udātta nā́m).
• The Subodhinī: "before matup" lets a formerly short vowel count (agnīnām), and leaves tisṛṇām to SK 3725.
• 3724 ṅyāś chandasi bahulam 6.1.178 — in the Veda, after ṅī, nām is udātta variously; Bhaṭṭoji glosses "vā". devasenā́nām abhibhañjatīnā́m (udātta nā́m), against jáyantīnāṃ marúto yantu, where nām is anudātta. No Subodhinī.
• 3725 ṣaṭtricaturbhyo halādiḥ 6.1.179 — after a ṣaṭ numeral, and after tri and catur, a consonant-initial vibhakti is udātta.
• antodāttāt has stopped here. Otherwise ādyudātta pañcan, navan and catur (phiṭ 28) would be missed (Subodhinī).
• ṣaṣ + bhis → ā́ ṣaḍbhír hūyámānaḥ; tri + bhis → tribhíṣ ṭváṃ deva (udātta bhír, bhíṣ).
• Not a vowel-initial vibhakti: catasraḥ (Subodhinī; cf. SK 3682).

IV.C. Where the vibhakti stays anudātta: go, śvan, avarṇa nominatives, div, nṛ — running example: go + ām → gávām (udātta ga; SK 3714 would give gavā́m); div + bhis → dyúbhiḥ, against divé

• 3726 na gośvansāvavarṇarāḍaṅkruṅkṛdbhyaḥ 6.1.182 — after go, śvan, a word with avarṇa in the nominative singular (sau), rāṭ, aṅ, kruṅ and kṛt, the vibhakti udātta taught before does not apply. The Subodhinī: this means the "ṣāṣṭhika" rules from SK 3714 on, and so it also blocks SK 3722 in vīravattamam.
• gávāṃ śatā́, góbhyo gātúm: the udātta stays on go.
• śúnaś cic chépam: a vanaspatyādi compound, so both members keep their accent, śú and ché (H13).
• "sau" is the nominative singular, so tad and kim (sa, ka) count: tébhyo dyumnám, téṣāṃ pāhi śrudhī́ hávam (udātta te).
• rāṭ, aṅ (añc in kvip, only where the n stays: prāñcā), kruṅ and kṛt have no example in the kaumudī (Subodhinī).
• 3727 divo jhal 6.1.183 — after div, a vibhakti beginning with a jhal is not udātta. "ādiḥ" comes from SK 3725 (Subodhinī).
• dyúbhir aktúbhiḥ (udātta dyú).
• A vowel-initial ending is not covered: úpa tvāgne divé-dive, where the vibhakti dé is udātta by SK 3717.
• 3728 nṛ cānyatarasyām 6.1.184 — after nṛ, the same prohibition holds optionally. In nṛ́bhir yemānáḥ it applies: udātta nṛ́, and bhis is a dependent svarita. No Subodhinī.

IV.D. The last affix accents: tit, the lasārvadhātuka, sic, thal with iṭ, rit — running example: kim + at (tit) → (kva for kim) kvà (svarita); vṛdh + śap + śānac (muk) → várdhamānam (the ending stays anudātta)

• 3729 tit svaritam 6.1.185 — an affix with t as it is svarita. kim + at (SK 1959), with kva for kim (SK 1960): kvà nūnám. kvà is an independent svarita.
• 3730 tāsyanudātteṅidadupadeśāl lasārvadhātukam anudāttam ahnviṅoḥ 6.1.186 — after tāsi, an anudāttet root, a ṅit in upadeśa, or what ends in a in upadeśa, a lasārvadhātuka is anudātta, except after hnu and iṅ. It is an apavāda to the affix accent.
• Examples: kartā, kartārau, kartāraḥ (after tās); ya āste (anudāttet); abhícaṣṭe ánṛtebhiḥ (ṅit); púrubhujā canasyátam (śap counts as a in upadeśa).
• It also overrides the cit accent, being later: várdhamānaṃ své dáme. Muk does not intervene (Subodhinī).
• Counter-examples: abhí vṛdhé gṛṇītáḥ; ható (han is not a-final in upadeśa, so tas keeps its udātta); katīha pacamānāḥ (cānaś, not a la); śiśye (ārdhadhātuka); hnute; yad adhīte.
• Vārttika: not after vid, indh or khid: indhé rā́jā. The te keeps the affix udātta (H13).
• 3731 ādiḥ sico 'nyatarasyām 6.1.187 — in a form ending in sic, the first syllable is optionally udātta. yā + luṅ (sic, iṭ, sak) → yāsiṣṭáṃ vartír aśvinā. The print marks the other option, with udātta ṣṭáṃ; the Subodhinī says an ādyudātta example must be sought elsewhere.
• 3732 thali ca seṭīḍanto vā 6.1.196 — in a word ending in thal with iṭ, optionally (anyatarasyām, from SK 3677) the iṭ or the final is udātta, or, by "vā" (Subodhinī), the first. Failing all three, liti (SK 3676) gives the udātta to the vowel before the affix. So all four syllables are udātta in turn: lúluvitha / lulúvitha / luluvítha / luluvithá.
• 3733 upottamaṃ riti 6.1.217 — in a word ending in an affix with r as it, the penultimate syllable is udātta. riti is a tadanta locative (Subodhinī's jñāpaka at SK 3675; H14). ā + hu + anīyar → yad āhavanī́ye (udātta nī). [iti pratyayasvarāḥ.]

IV.E. The compound's own accent, and the bahuvrīhi prior member — running example: yajña + śrī (ṣaṣṭhī-tatpuruṣa) → yajñaśrī + am → yajñaśríyam (final udātta śri); citrá + śravas (bahuvrīhi) → citráśravas- (prior member keeps trá)

• 3734 samāsasya 6.1.223 — [samāsasvarāḥ.] The final of a compound is udātta. yajñaśríyam: the iyaṅ that replaces ī takes its udātta by nearest likeness (Subodhinī). This is the default that every rule from 3735 on overrides.
• 3735 bahuvrīhau prakṛtyā pūrvapadam 6.2.1 — in a bahuvrīhi, a prior member that contains an udātta or svarita keeps its own accent. "udātta-svarita" is carried forward (Subodhinī), and the rule is the adhikāra of 6.2.
• citra is antodātta (phiṭ 1): satyáś citráśravastamaḥ (udātta trá).
• A wholly anudātta prior member leaves only the 3734 accent: sama (anudātta, phiṭ 78) + pāda → samapādáḥ (udātta dáḥ).

IV.F. The tatpuruṣa prior member keeps its accent: the seven classes, and the first named final members — running example: kiri (antodātta) + kāṇa → kiríkāṇaḥ (udātta rí); gamana (lyuṭ, ādyudātta by liti) + cira → gámanaciram (udātta gá)

Accent: in each form only the prior member's preserved udātta is marked, and the rest of the compound is anudātta. A counter-example falls back on samāsasya (3734) and is left unmarked unless the text says otherwise.

• 3736 tatpuruṣe tulyārthatṛtīyāsaptamyupamānāvyayadvitīyākṛtyāḥ 6.2.2 — in a tatpuruṣa, seven kinds of prior member keep their accent. "tatpuruṣe" runs through 3757.
• A word meaning equal: túlyaśvetaḥ (yat, SK 3701).
• A 3rd-case word: kiríkāṇaḥ.
• A 7th-case word: patayán mandayátsakham.
• An upamāna: śastrī́śyāmā (ṅīṣ).
• A 2nd-case word: muhūrtásukham.
• A kṛtya: bhojyoṣṇam (left unmarked).
• An avyaya, which a vārttika limits to nañ, ku and the nipātas: ayajño vā eṣaḥ. Not snātvākālakaḥ, which takes the compound-final accent.
• 3737 varṇo varṇeṣv anete 6.2.3 — a colour word before colour words other than eta.
• lóhitakalmāṣaḥ: itan is nit, so the udātta is on lo.
• kṛṣṇasāraṅgaḥ is left unmarked: kṛṣṇa is antodātta in the Veda (phiṭ 11) and ādyudātta in ordinary speech (phiṭ 33).
• Counter-examples: paramakṛṣṇaḥ, kṛṣṇatilāḥ, kṛṣṇaitaḥ.
• 3738 gādhalavaṇayoḥ pramāṇe 6.2.4 — before gādha or lavaṇa, in a compound of measure (pramāṇa = the delimitation of extent).
• arítragādham udakam (itra, madhyodātta, so udātta ri).
• gólavaṇam (go, from ḍo, antodātta).
• Counter-example: paramagādham.
• 3739 dāyādyaṃ dāyāde 6.2.5 — before dāyāda, a prior member naming what is inherited. dhánadāyādaḥ: dhana takes kyu and is ādyudātta, so the udātta is on dhá. Counter-example: paramadāyādáḥ (compound-final).
• 3740 pratibandhi cirakṛcchrayoḥ 6.2.6 — before cira or kṛcchra, a prior member naming an obstruction. gámanaciram, vyāháraṇakṛcchram: both are lyuṭ forms and take the liti accent, so the udātta is on gá and há. Counter-example: mūtrakṛcchram.
• 3741 pade 'padeśe 6.2.7 — before pada meaning pretext.
• mū́trapadena prasthitaḥ: mūtra is ghañ-final, so ādyudātta by SK 3686 (the translator's step).
• uccārápadena: ghañ, antodātta by SK 3878, so udātta rá.
• Counter-example: viṣṇupadam.
• 3742 nivāte vātatrāṇe 6.2.8 — before nivāta meaning a shelter from the wind.
• kuṭī́nivātam (gaurādi ṅīṣ, so udātta ṭī).
• kuḍyánivātam (ḍyak, so udātta dyá).
• Counter-example: rājanivāte vasati, where nivāta means "side".
• 3743 śārade 'nārtave 6.2.9 — before śārada meaning new, not "of the season". rájjuśāradam udakam: rajju is formed by sṛjer asum ca (uṇādi 15) and is nit, so the udātta is on rá. Counter-example: uttamaśāradam.
• 3744 adhvaryukaṣāyayor jātau 6.2.10 — before adhvaryu or kaṣāya, in the name of a class.
• kaṭhá (pacādi ac) + adhvaryu → kaṭhā́dhvaryuḥ: the udātta ekādeśa comes by SK 3658 (the translator's step).
• dauvārikákaṣāyam (ṭhak, so udātta ká).
• Counter-example: paramādhvaryuḥ.
• 3745 sadṛśapratirūpayoḥ sādṛśye 6.2.11 — before sadṛśa or pratirūpa, in a compound of likeness. pitṛ́sadṛśaḥ: pitṛ takes tṛc, which is cit, so the udātta is on tṛ́. Counter-example: paramasadṛśaḥ, where the sense is honouring. There is no pratirūpa example.

IV.G. More named final members in the tatpuruṣa: a measure-dvigu, vāṇija, mātra and the rest, sukha/priya, svāmin, pati, nedīyas, pūrva, savidha — running example: gṛha (ka, antodātta) + pati → gṛhápatiḥ (udātta ha); bhuvana (kyun, ādyudātta) + pati → bhúvanapatiḥ (optional)

• 3746 dvigau pramāṇe 6.2.12 — before a dvigu that denotes a measure (H15).
• prā́cyasaptasamaḥ: prācya takes yat and is ādyudātta by SK 3701, so the udātta is on prā. The mātrac is elided by the vārttika pramāṇe lo dvigor nityam.
• Counter-examples: vrīhiprasthaḥ (not a dvigu); paramasaptasamam (a samāhāra-dvigu, not a measure).
• 3747 gantavyapaṇyaṃ vāṇije 6.2.13 — before vāṇija, a prior member naming the destination or the merchandise. madrávāṇijaḥ (rak, so udātta drá); góvāṇijaḥ. Counter-example: paramavāṇijaḥ.
• 3748 mātropajñopakramacchāye napuṃsake 6.2.14 — before mātra, upajña, upakrama or chāyā, in a neuter compound.
• bhikṣā́mātram (the a of SK 3280, then ṭāp).
• pā́ṇinyupajñam (iñ).
• nandópakramam (pacādi ac, merged into an udātta o; the translator's step).
• íṣucchāyam (nit).
• Counter-example: kuḍyacchāyā (feminine).
• 3749 sukhapriyayor hite 6.2.15 — before sukha or priya, meaning good for. gámanapriyam, gámanasukham (litsvara, so udātta gá). Counter-example: paramasukham.
• 3750 prītau ca 6.2.16 — the same when affection is meant. brāhmaṇásukhaṃ pāyasam, chātrápriyo 'nadhyāyaḥ: the affix accent puts the udātta on ṇá and trá. Counter-example: rājasukham. The Subodhinī: the rule is split from 3749 to stop yathāsaṃkhya.
• 3751 svaṃ svāmini 6.2.17 — before svāmin, a prior member naming property. gósvāmī (udātta go). Counter-example: paramasvāmī. No Subodhinī.
• 3752 patyāv aiśvarye 6.2.18 — before pati meaning lordship. dámūnā gṛhápatir dáme (gṛha takes ka, SK 2906, so udātta ha).
• 3753 na bhūvākciddidhiṣu 6.2.19 — not for bhū, vāc, cit, didhiṣū: bhūpatiḥ, vākpatiḥ, citpatiḥ, didhiṣūpatiḥ. The antodātta of the prior member is withheld. The resulting accent is not stated, so the forms are left unmarked.
• 3754 vā bhuvanam 6.2.20 — bhuvana keeps its accent only optionally. Where it does: bhúvanapatiḥ (kyun, nit, so udātta bhú). The other option, bhuvanapatiḥ, is not stated.
• 3755 āśaṅkābādhanedīyassu saṃbhāvane 6.2.21 — before āśaṅka, ābādha or nedīyas, of supposition. gámanāśaṅkam asti, gámanābādham, gámananedīyaḥ. Counter-example: paramanedīyaḥ.
• 3756 pūrve bhūtapūrve 6.2.22 — before pūrva meaning "formerly". āḍhyápūrvaḥ: āḍhya takes ka and is antodātta by SK 3878, so the udātta is on dhyá. Counter-example: paramapūrvaḥ.
• 3757 savidhasanīḍasamaryādasaveśasadeśeṣu sāmīpye 6.2.23 — before savidha and the other four, meaning nearness.
• madrásavidham, madrásaveśam, madrásadeśam.
• kāśmī́rasamaryādam (phiṭ 42).
• gāndhārasanīḍam is left unmarked. The Subodhinī's text had gāndhāri-.
• Counter-examples: samaryādaṃ kṣetram, caitrasamaryādam.

IV.H. Before quality words, and in the karmadhāraya: vispaṣṭādi; the bhāva noun before śra, jya, kan, avama and pāpa; kumāra — running example: gamana (lyuṭ, bhāva) + śreṣṭha → gámanaśreṣṭham (udātta gá); kumāra (antodātta) + śramaṇā → kumāráśramaṇā, then + pratyenas → kúmārapratyenāḥ (udātta moves to kú)

• 3758 vispaṣṭādīni guṇavacaneṣu 6.2.24 — vispaṣṭa and the rest, before words for a quality, keep their accent.
• víspaṣṭakaṭukam: the gati vi is udātta by SK 3783.
• The compound is a supsupā compound, not a karmadhāraya (Subodhinī).
• Counter-examples: paramalavaṇam, vispaṣṭabrāhmaṇaḥ.
• The gaṇa: vispaṣṭa, vicitra, vyakta, saṃpanna, paṇḍita, kuśala, capala, nipuṇa. The Subodhinī's list also had paṭu.
• 3759 śrajyā'vamakanpāpavatsu bhāve karmadhāraye 6.2.25 — in a karmadhāraya, a prior member naming the action itself keeps its accent before a final member containing śra, jya or kan (SK 2009, 2011, 2019), before avama, or before one containing pāpa.
• gámanaśreṣṭham, gámanajyāyaḥ, gámanāvamam, gámanakaniṣṭham, gámanapāpiṣṭham.
• Counter-examples: gamanaśobhanam; gamanaśreyaḥ (gamana as instrument, SK 3293); a ṣaṣṭhī-samāsa.
• 3760 kumāraś ca 6.2.26 — kumāra also keeps its accent in a karmadhāraya. It is pacādi ac and cit, so antodātta: kumāráśramaṇā (udātta rá). The Subodhinī: only the pratipadokta compound of SK 752 is meant, though "some" take every karmadhāraya.
• 3761 ādiḥ pratyenasi 6.2.27 — before pratyenas, the first syllable of kumāra is udātta instead: kúmārapratyenāḥ (udātta kú).
• 3762 pūgeṣv anyatarasyām 6.2.28 — before a word for a guild, the same holds optionally: kúmāracātakāḥ, kúmārajīmūtāḥ. In the other option 3760 applies: kumārácātakāḥ.

IV.I. The dvigu prior member — running example: pañcan (ādyudātta, phiṭ 28) + kapāla → páñcakapālaḥ (udātta pá); bahú + aratni → bahvàratniḥ (independent svarita)

• 3763 igantakālakapālabhagālaśarāveṣu dvigau 6.2.29 — in a dvigu, the prior member keeps its accent before an ik-final word, a word for time, kapāla, bhagāla or śarāva.
• Before an ik-final word: páñcāratniḥ.
• Before a word for time: dáśamāsyaḥ, páñcamāsyaḥ, with yap by SK 1746 inside the adhikāra of SK 1744. dáśa is marked by the translator, under phiṭ 28.
• Before the three vessel words: páñcakapālaḥ, páñcabhagālaḥ, páñcaśarāvaḥ.
• Counter-examples: pañcāśvaḥ, paramāratniḥ.
• 3764 bahv anyatarasyām 6.2.30 — bahu (antodātta) keeps its accent only optionally.
• bahvàratniḥ: u → v, and the following a becomes an independent svarita by SK 3657.
• bahúmāsyaḥ, bahúkapālaḥ (udātta hú).
• In the other option the compound is end-udātta (Subodhinī).
• 3765 diṣṭivitastyoś ca 6.2.31 — before diṣṭi or vitasti, in a dvigu, optionally: páñcadiṣṭiḥ, páñcavitastiḥ.

IV.J. Seventh-case words, upasargas in the avyayībhāva, and dvandvas — running example: bhrāṣṭra (ṣṭran, nit) + pakva → bhrā́ṣṭrapakvaḥ (udātta bhrā); śini (ādyudātta) + vāsudeva → śínivāsudevāḥ (udātta śí)

• 3766 saptamī siddhaśuṣkapakvabandheṣv akālāt 6.2.32 — a 7th-case word that does not denote time keeps its accent before siddha, śuṣka, pakva or bandha.
• sāṅkāśyásiddhaḥ: ṇya gives udātta śyá (the translator's step, H16).
• ātapaśuṣkaḥ is left unmarked.
• bhrā́ṣṭrapakvaḥ; cakrábandhaḥ (cakra, from ka, is antodātta).
• Counter-example: pūrvāhṇasiddhaḥ.
• The rule re-enjoins the 3736 accent that the kṛt (thāthādi, SK 3878) accent had blocked.
• 3767 paripratyupāpā varjyamānāhorātrāvayaveṣu 6.2.33 — pari, prati, upa and apa, which are ādyudātta, keep their accent before a word naming what is excluded, or a part of the day or night.
• The rule serves the avyayībhāva: for the bahuvrīhi and tatpuruṣa, 3735–3736 already give the result.
• páritrigartaṃ vṛṣṭo devaḥ ("except over Trigarta"), ápatrigartam.
• prátipūrvāhṇam, úpapūrvāhṇam, úpapūrvarātram.
• Only apa and pari have the excluding sense. Counter-example: pratyagni.
• 3768 rājanyabahuvacanadvandve 'ndhakavṛṣṇiṣu 6.2.34 — in a dvandva of plural rājanya words naming Andhakas and Vṛṣṇis, the prior member keeps its accent.
• śvāphalkácaitrakāḥ: aṇ gives udātta lká (the translator's step).
• śínivāsudevāḥ (udātta śí).
• Counter-examples: dvaipyabhaimāyanāḥ (not of anointed lineage); saṃkarṣaṇavāsudevau (dual); vṛṣṇikumārāḥ (not a dvandva); kurupañcālāḥ.
• 3769 saṃkhyā 6.2.35 — a numeral prior member in a dvandva keeps its accent.
• dvā́daśa: dvi is udātta and takes ā by SK 808 (the translator's step).
• tráyodaśa: the trayas of SK 809 is nipātita as ādyudātta.
• The Subodhinī's text had ekādaśa and "antodātta".
• 3770 ācāryopasarjanaś cāntevāsī 6.2.36 — in a dvandva of pupil words that have a teacher as upasarjana, the prior member keeps its accent.
• pāṇinī́yarauḍhīyāḥ: cha makes them madhyodātta, so the udātta is on nī.
• The qualifier attaches to the whole dvandva: hence not pāṇinīyadevadattau.
• Counter-examples: chāndasavaiyākaraṇāḥ (no teacher); āpiśalapāṇinīye śāstre (not pupils).

Part V — SK 3771–3822 (6.2.37–6.2.88): the prior member keeps its own accent (named words, case-words, the gati, the optional set), then the ādir udāttaḥ heading and the compounds whose prior member is made ādyudātta

Covers SK 3771–3822 (52 sūtras): all present.

• The Aṣṭādhyāyī spine. The run is 6.2.37–6.2.88, taken one-to-one and in order (SK 3771 = 6.2.37 … SK 3822 = 6.2.88). The SK moves nothing here: there are no gaps, nothing pulled forward and no rules from other books. All 52 AS-refs are checked against references/ashtadhyayi/sutraani-data.json, by SK number, and all match. So do all 49 SK cross-references cited below, checked by script against the same file (from the samāsa rules 690–749 and the kṛt rules 2913–3288 to the svara rules 3652–3885).
• The headings and what runs on.
• bahuvrīhau prakṛtyā pūrvapadam (6.2.1, SK 3735) gives "the prior member keeps its own accent" from 3771 through 3797. 3771 also takes dvandve from SK 3768. Naming these source rules in the sutra_en frames is the translator's identification, and it is owned in a note in every unit.
• Chains that run on:
• caturthī (3777) runs into 3778–3779.
• kte (3779) runs into 3780–3783.
• karmaṇi (3782) runs into 3783.
• gatir anantaraḥ (3783) runs into 3784–3787.
• aniganto 'ñcatau vapratyaye (3786) gives the añc environment to 3787.
• anyatarasyām (3788) is the "optionally" of 3788–3797. Bhaṭṭoji writes only "vā" from 3794 on; tying that to 3788 is the translator's identification.
• karmadhāraye (3791) runs into 3792–3793. brāhmaṇakumārayoḥ (3792) runs into 3793, and rājā (3793) into 3794. śilpini (3796) runs into 3797.
• ādir udāttaḥ (6.2.64, SK 3798) is a new adhikāra: the first vowel of the prior member is udātta. The Subodhinī says pūrvapadam is turned into the sixth case by sense here. Bhaṭṭoji does not say how far the heading runs. It governs every rule left in this part and carries on into Part VI as far as antaḥ (6.2.92, SK 3826).
• 3800 is obligatory: the "vā" of 3788 stops at 3797.
• vibhāṣā (3801) carries into 3802 (Subodhinī).
• Chains under ādir udāttaḥ:
• ake (3807) runs into 3808.
• aṇi (3809) runs into 3810–3811, and akṛñaḥ (3810) into 3811.
• ṇini (3813) runs into 3814.
• je (3816) runs into 3817, which puts antyāt pūrvam in place of ādiḥ.
• anivasantaḥ (3818) runs into 3819 on some views only.
• prasthe (3821) runs into 3822.
• Print readings followed (NSP 1915, scan-confirmed; collation batch17–19 and audits H17–H20):
• Mūla. 3771 sāvarṇimāṇḍūkeyau. 3772 mahadvrīhiḥ (ours mahahvrīhiḥ). 3776 taitilakadrūpaṇyakambala (ours has a stray visarga) and upratyayāntaḥ kuruḥ as one sentence. 3777 neuter dāru. 3784 gatiḥ. 3787 nyaṅṅ uttānaḥ. 3803 śāstrāṇi. 3808 prāgdeśavācinām. 3810 ced aṇ. 3821 indraprasthaḥ as one word.
• Words the print has and ours lacks, supplied in brackets: 3781 ahīnavācini; 3799 saptamīsamāsaḥ. kāranāmni ca 968 (ours skips by eye); 3822 śoṇāprasthaḥ.
• 3819 dākṣihradaḥ (ours and Wikisource -hṛdaḥ) is not in the collation.
• Accent sign. In 3786 the print has the svarita sign over vṛ́ṣṇyāni.
• Ours kept: 3775 gauḥ (print gau); 3806 gobiḍāla (the print may read goviḍāla; class C). 3805 samāśa[śā]layaḥ is supplied against both ours and the print. The print misnumbers 3813 as "4813".
• Print Check queries, ours = print: 3776 "kṛtyāḥ 2831" (the accent meant is the kṛtya member of SK 3736); 3808 "paryāye ṇvul" (SK 3288 prescribes ṇvuc). In 3776 our source tags the prose after each vārttika "(vārttikam)" where the print has only its vārttika sign; the tag is not rendered.
• The commentary (only the kaumudī is translated, SK-TRANSLATE §25):
• Subodhinī only. There is no Bālamanoramā anywhere in the range. Our layer is empty, and Wikisource uttara-4/0109–0112 is mūla only.
• Subodhinī filed under a neighbour:
• The head of 3772 is in sk.3771.s1, and the head of 3776 is in sk.3775.s1.
• 3778, 3779 and 3780 are in sk.3777.s1.
• 3783 is the tail of sk.3782.s1, and the head of 3784 is the tail of sk.3783.s1.
• 3789 is the tail of sk.3788.s1, and 3791 is the second half of sk.3790.s1.
• 3798 and the head of 3799 are in sk.3797.s1.
• 3800 is the tail of sk.3799.s1, and 3802 is in sk.3801.s1.
• 3807 is the tail of sk.3806.s1, and 3808 the tail of sk.3807.s1.
• 3810 is the tail of sk.3809.s1.
• 3816–3817 are in sk.3815.s1.
• sk.3822.s1 also carries 3823–3824.
• Where the Subodhinī differs from the kaumudī (recorded in the notes, not followed):
• 3771: ṛṣyaṇ in the apatya sense.
• 3775: it reads gosādī with ṇini.
• 3786: Haradatta's paratvāt = iṣṭatvāt.
• 3818: it takes nivasanta as the dwelling-place.
• Accent marking and the audits.
• Where the marks come from. The text itself gives accented copies only for the mantras in 3776, 3784–3787. Every other mark is placed on the syllable that Bhaṭṭoji or the Subodhinī names, or that follows from a stated ground (aṇ/rak → affix vowel by SK 3708; ñit/nit → ādyudātta by SK 3686; cit → antodātta by SK 3710; ṇyat → svarita by SK 3729). The text states no accent for the counter-examples, and they are left unmarked unless the Subodhinī states it.
• The audits. H17–H20 found no MUST-FIX. Their MINORs are applied:
• accents marked in 3771, 3783 and 3788–3793;
• the accent statements in 3794–3796 bracketed;
• 3795's thāthādi note corrected (a kāraka prior member also triggers SK 3878);
• 3800 construed as "in [a compound] denoting one engaged";
• 3814 gardabhoccārī confirmed as the print's reading.
• Where Part V ends. At 3822 mālādīnāṃ ca (6.2.88), ādyudātta of vṛddha prior members before prastha. Part VI opens with amahan navaṃ nagare 'nudīcām (6.2.89, SK 3823), still under ādir udāttaḥ.

V.A. Listed dvandvas and named prior members keep their own accent: mahat, kṣullaka, uṣṭra, go, kuru and dāsībhārādi — running example: mahat (antodātta) + vrīhi → mahā́vrīhiḥ (udātta on hā, the rest anudātta)

• 3771 kārtakaujapādayaś ca 6.2.37 — in the dvandvas of the kārtakaujapādi list (dvandve from SK 3768), the prior member keeps its own accent. kārta ends in aṇ, whose vowel carries the udātta (SK 3708): kārtákaujapau (udātta tá). sāvarṇi ends in iñ and is ādyudātta (SK 3686): sā́varṇimāṇḍūkeyau (udātta sā). The rest of each dvandva is anudātta. The list itself is not reproduced. The Subodhinī derives kārta and kaujapa with ṛṣyaṇ (offspring), against Bhaṭṭoji's "idam" sense.
• 3772 mahān vrīhyaparāhṇagṛṣṭīṣvāsajābālabhārabhāratahailihilarauravapravṛddheṣu 6.2.38 — mahat (antodātta by nipātana, uṇādi 241) keeps its accent before the ten listed words. mahā́vrīhiḥ, mahā́parāhṇaḥ, mahā́gṛṣṭiḥ, mahā́hailihilaḥ: the udātta is on hā (ā by SK 807).
• Before iṣvāsa, udātta ā + i → e. That gives mahéṣvāsaḥ (udātta by SK 3658) or, per the Subodhinī, mahèṣvāsaḥ (svarita by SK 3659).
• The rule holds only in the compound taught specifically (pratipadokta) by sanmahat… (SK 740). Not in the ṣaṣṭhī mahato vrīhiḥ → mahadvrīhiḥ.
• Subodhinī: pravṛddha is not redundant, because SK 3780 takes only the śreṇyādi compound.
• 3773 kṣullakaś ca vaiśvadeve 6.2.39 — kṣullaka, and by "ca" mahat too, keeps its accent before vaiśvadeva. kṣulla is formed from kṣudh + lā + ka. Then the taddhita ka of ajñāte (SK 2028) makes kṣullaka antodātta: kṣullakávaiśvadevam (udātta ká). With mahat: mahā́vaiśvadevam (hā).
• 3774 uṣṭraḥ sādivāmyoḥ 6.2.40 — uṣṭra (uṣ + ṣṭran, nit, so ādyudātta) keeps its accent before sādin and vāmin: úṣṭrasādī, úṣṭravāmī (udātta ú).
• 3775 gauḥ sādasādisārathiṣu 6.2.41 — go keeps its accent before sāda, sādi and sārathi. The monosyllable carries the udātta: gósādaḥ, gósādiḥ, gósārathiḥ (udātta gó). Subodhinī: for sāda and sādin this is an apavāda of the kṛt accent, for sārathi of the compound accent.
• 3776 kurugārhapatariktagurvasūtajaratyaślīladṛḍharūpāpārevaḍavātaitilakadrūpaṇyakambalo dāsībhārāṇāṃ ca 6.2.42 — in seven listed compounds and the open list (ākṛtigaṇa) dāsībhārādi, the prior member keeps its accent:
• kurúgārhapatam: kuru is u-final, so the udātta is on rú. By vārttika, vṛji too: vṛ́jigārhapatam.
• ríktaguruḥ (by rikte vibhāṣā, SK 3696).
• ásūtajaratī and áślīladṛḍharūpā: nañ-compounds, so ādyudātta.
• pārévaḍavā: pāra is antodātta by ghṛtādi (phiṭ 21), and the seventh-case e is udātta by SK 3658. This is a compound in the sense of iva, with aluk by nipātana.
• taitilákadrūḥ (aṇ).
• paṇya is yat-final and ādyudātta by SK 3701. By vārttika, only as a name: páṇyakambalaḥ. Otherwise the compound takes the compound-final udātta, paṇyakambaláḥ, because the kṛtya accent of SK 3736 is only for the pratipadokta compound (SK 749).
• dāsī́bhāraḥ, deváhūtiḥ. Any tatpuruṣa that should keep the prior member's accent but has no rule of its own belongs to this list.
• Mantra: sá rāyé sá púraṃdhyām. pura is ādyudātta (phiṭ 26), so the udātta is on pú. The aluk is Vedic.

V.B. A case-word prior member keeps its accent: the fourth case before tadartha, artha and kta; a non-niṣṭhā in a karmadhāraya; the second and third cases before kta — running example: yūpāya dāru → yū́padāru (yūpa nit: udātta yū); kaṣṭaṃ śritaḥ → kaṣṭáśritaḥ (udātta tá, against thāthādi's final udātta)

• 3777 caturthī tadarthe 6.2.43 — a fourth-case prior member keeps its accent when the following member is what is for its purpose. yūpa (pa by uṇādi 307, nit, so ādyudātta) + dāru: yū́padāru (udātta yū; dāru anudātta). The print reads neuter dāru.
• 3778 arthe 6.2.44 — likewise before artha. devá (pacādi ac) + artha: the a + a merger is udātta by SK 3658, giving devā́rtham (udātta dā).
• 3779 kte ca 6.2.45 — likewise before a kta-final word. gobhyo hitam → góhitam (udātta gó; fourth case by SK 631).
• 3780 karmadhāraye 'niṣṭhā 6.2.46 — in a karmadhāraya before a kta-final word, a prior member that is not a niṣṭhā keeps its accent. Only the śreṇyādi compound (SK 738) is meant (Subodhinī).
• śréṇikṛtāḥ: śreṇi is ādyudātta (ni is nit), so the udātta is on śré.
• pūgákṛtāḥ: pūga is antodātta, so the udātta is on gá.
• Counter-examples: śreṇyā kṛtam → śreṇikṛtam (not a karmadhāraya); kṛtākṛtam (a niṣṭhā prior member, SK 739).
• 3781 ahīne dvitīyā 6.2.47 — before a kta-final word, in a compound that denotes what is not abandoned, a second-case prior member keeps its accent. This is the apavāda of thātha- (SK 3878).
• kaṣṭáśritaḥ: kaṣṭa is antodātta, so the udātta is on tá.
• grā́magataḥ: grāma is nit (man), so the udātta is on grā.
• Counter-example under ahīne: kāntārātītáḥ, with thāthādi's final udātta.
• Vārttika anupasarge: sukhaprāptáḥ.
• 3782 tṛtīyā karmaṇi 6.2.48 — before a kta-final word in the object sense, a third-case prior member keeps its accent.
• tvayā ūtāḥ → tvótāsaḥ: tva + ūta, merged by SK 3658, so the udātta is on tvó. tā is the dependent svarita.
• rudráhataḥ: rudra takes rak, so the udātta is on drá.
• mahārājáhataḥ: ṭac makes it antodātta, so the udātta is on já.
• Counter-example: rathena yātaḥ → rathayātaḥ, where the kta is agentive because yā is a root of motion.

V.C. The gati keeps its own accent: before kta, before a t-initial nit kṛt, before tavai, and before añc + v — running example: puras (antodātta) + hita → puróhitam (udātta ro); parā + añc (kvin) → párāñcaḥ (udātta pa)

• 3783 gatir anantaraḥ 6.2.49 — before a kta in the karman sense, a gati immediately adjacent to the root keeps its accent. This is the apavāda of thātha- (SK 3878).
• puras takes asi (SK 1975), and the affix accent makes it antodātta: puróhitam (udātta ró).
• Under anantara: in abhyúddhṛtaḥ, abhi is not adjacent to the root, so ud keeps its own udātta (ú).
• With a kāraka prior member, only the thāthādi accent applies, since it is satiśiṣṭa: dūrādāgatáḥ (final udātta).
• 3784 tādau ca niti kṛty atau 6.2.50 — before a kṛt affix that begins with t, is nit and is not tu, the adjacent gati keeps its accent. This is the apavāda of the kṛt accent (SK 3873).
• pra + bhūti (ktin): ágne rāyó nṛ́tamasya prábhūtau (udātta prá).
• sáṅgatiṃ góḥ (udātta sá).
• Counter-examples: prajalpākaḥ (ṣākan, SK 3135, not t-initial); prakartā (tṛc, not nit); āgantuḥ (tun).
• 3785 tavai cāntaś ca yugapat 6.2.51 — before a tavai-final word, the gati keeps its accent and the end of the tavai-word is udātta, together. The Subodhinī gets "udātta" from prakṛtyā; "yugapat" excludes taking the two in turn.
• anu is ādyudātta (phiṭ 81): ánvetavā́ u. The udāttas are á of anu and vā (vai → vā before u).
• 3786 aniganto 'ñcatau vapratyaye 6.2.52 — a gati not ending in ik keeps its accent before añc + v (kvin, SK 373).
• yé párāñcas tān: parā is ādyudātta, so the udātta is on pá.
• Under anigantaḥ: pratyáñco yantu (prati ends in ik). The kṛt accent puts the udātta on tyáñ.
• "Being later" than the cu accent (SK 3652), this rule alone applies in jahí vṛ́ṣṇyāni kṛṇuhī́ párācaḥ. The udātta is on pá, not on rā by the cu accent.
• Under vapratyaye: udañcanam.
• 3787 nyadhī ca 6.2.53 — ni and adhi do the same, although they end in ik.
• The udātta i of ni → y, and the a of añc becomes svarita by SK 3657. So nyàṅṅ uttānáḥ has nyàṅ as an independent svarita.
• adhyaṅ.

V.D. The optional prior-member accent (anyatarasyām, 3788–3797) — running example: īṣat (antodātta) + kaḍāra → īṣátkaḍāraḥ (udātta ṣat) or, by the compound accent, īṣatkaḍāráḥ; rājan (ādyudātta) + brāhmaṇa → rā́jabrāhmaṇaḥ (udātta rā)

• 3788 īṣad anyatarasyām 6.2.54 — īṣat (antodātta by phiṭ, per the Subodhinī) optionally keeps its accent: īṣátkaḍāraḥ (udātta ṣát) or īṣatkaḍāráḥ. With a kṛt final member only the kṛt accent applies: īṣadbhedaḥ (khal).
• 3789 hiraṇyaparimāṇaṃ dhane 6.2.55 — a prior member that is a quantity of gold optionally keeps its accent before dhana: dvisuvarṇam → dvisuvarṇadhanam. The text names no udātta syllable. There is no tatpuruṣa heading, so this rule reaches a bahuvrīhi too. Being later, it keeps the accent optional there against SK 3735.
• Counter-examples: prasthadhanam (not gold); kāñcanadhanam (not a quantity); niṣkamālā (not dhana).
• 3790 prathamo 'ciropasampattau 6.2.56 — prathama, meaning "new", optionally keeps its accent. It is prath + amac, and cit gives antodātta (translator's inference): prathamávaiyākaraṇaḥ, "one just beginning grammar" (udātta má). Counter-example: prathamo vaiyākaraṇaḥ, "the first grammarian"; the Subodhinī says that compound is always antodātta.
• 3791 katarakatamau karmadhāraye 6.2.57 — katara and katama in a karmadhāraya. ḍatarac is cit, so katara is antodātta: katarákaṭhaḥ (udātta rá). karmadhāraye is there for the next rule, since the pratipadokta compound (SK 742) already is one.
• 3792 āryo brāhmaṇakumārayoḥ 6.2.58 — ārya before brāhmaṇa or kumāra, in a karmadhāraya. ṇyat is tit, so ārya is antasvarita (SK 3729): āryàkumāraḥ, āryàbrāhmaṇaḥ. There is no udātta: yà is an independent svarita and the rest is anudātta. Counter-examples: paramabrāhmaṇaḥ, āryakṣatriyaḥ.
• 3793 rājā ca 6.2.59 — rājan likewise. It is kanin, nit, so ādyudātta: rā́jabrāhmaṇaḥ, rā́jakumāraḥ (udātta rā). The rule is split off so that the next rule takes rājan alone.
• 3794 ṣaṣṭhī pratyenasi 6.2.60 — rājan as a sixth-case word before pratyenas: rā́japratyenāḥ (udātta rā). It does not apply to the karmadhāraya rājā cāsau pratyenāś ca.
• 3795 kte nityārthe 6.2.61 — before a kta-final word, in a compound meaning "constantly". nitya is tyap, so ādyudātta: nítyaprahasitaḥ (udātta ní). This is a dvitīyā compound by kālāḥ (SK 690). Without the option, thāthādi (SK 3878) gives nityaprahasitáḥ. Counter-example: muhūrtaprahasitaḥ (thāthādi).
• 3796 grāmaḥ śilpini 6.2.62 — the word grāma before a word meaning an artisan. grāma is ādyudātta: grā́manāpitaḥ (udātta grā). Counter-examples: paramanāpitaḥ, grāmarathyā.
• 3797 rājā ca praśaṃsāyām 6.2.63 — rājan before an artisan-word, when praise is meant: rā́janāpitaḥ, rā́jakulālaḥ (udātta rā). Counter-examples: rājanāpitaḥ, the same compound with no praise; rājahastī.

V.E. ādir udāttaḥ: the prior member is ādyudātta by what the compound denotes (a customary due, one engaged, a superintendent, a wretched artisan, contempt) — running example: mukuṭe kārṣāpaṇam → múkuṭekārṣāpaṇam (udātta mú); go + ballava → góballavaḥ (udātta gó)

• 3798 ādir udāttaḥ 6.2.64 — adhikāra: the first vowel of the prior member is udātta. The Subodhinī turns pūrvapadam into the sixth case by sense. This heading replaces "keeps its own accent" for the rest of this part.
• 3799 saptamīhāriṇau dharmye 'haraṇe 6.2.65 — a seventh-case word or a hārin (taker) as prior member is ādyudātta before a word for a customary due (dharmya), but not before haraṇa.
• Seventh case: múkuṭekārṣāpaṇam, háledvipadikā (the compound by SK 721, aluk by SK 968).
• hārin: yā́jñikāśvaḥ, vaíyākaraṇahastī.
• Counter-examples under dharmye: stamberamaḥ, karmakaravarddhitakaḥ.
• Under aharaṇe: vāḍavaharaṇam, where the Subodhinī gives final udātta by SK 3884. That exclusion is a jñāpaka: the hārin accent beats even the later kṛt accent, so vā́ḍavahāryam.
• 3800 yukte ca 6.2.66 — in a compound denoting one engaged (yukta, from yuja samādhau): góballavaḥ, cowherd (udātta gó). The rule is obligatory.
• 3801 vibhāṣādhyakṣe 6.2.67 — optionally before adhyakṣa: gávādhyakṣaḥ (udātta gá). It is an option against 3800, since adhyakṣa also denotes one engaged.
• 3802 pāpaṃ ca śilpini 6.2.68 — pāpa before an artisan-word, optionally per the Subodhinī: pā́panāpitaḥ, "a wretched barber" (udātta pā). This holds only in the pratipadokta karmadhāraya of pāpāṇake kutsitaiḥ (SK 733), not in a ṣaṣṭhī compound.
• 3803 gotrāntevāsimāṇavabrāhmaṇeṣu kṣepe 6.2.69 — before a word meaning a descendant or a pupil, or before māṇava or brāhmaṇa, when contempt is meant:
• descendant: bhā́ryāsauśrutaḥ;
• pupil: kúmārīdākṣāḥ, ódanapāṇinīyāḥ;
• māṇava: bhíkṣāmāṇavaḥ;
• brāhmaṇa: bháyabrāhmaṇaḥ.
• The udātta is on the first syllable each time (bhā, kú, ó, bhí, bhá).
• Counter-examples: dāsīśrotriyaḥ; paramabrāhmaṇaḥ.

V.F. ādir udāttaḥ by the following member's class: an ingredient, food, a comparison, an aka-noun of livelihood or of an eastern game — running example: madhu + maireya → mádhumaireyaḥ (udātta má); danta + lekhaka → dántalekhakaḥ (udātta dán)

• 3804 aṅgāni maireye 6.2.70 — a word for an ingredient before maireya (a liquor): mádhumaireyaḥ (udātta má). Counter-examples: paramamaireyaḥ; puṣpāsavaḥ.
• 3805 bhaktākhyās tadartheṣu 6.2.71 — a name of a food before a word for what is meant for it: bhíkṣākaṃsaḥ, bhā́jīkaṃsaḥ (udātta bhí, bhā; ṣaṣṭhī compounds per the Subodhinī).
• Counter-example: samāśa[śā]layaḥ, since samāśa is the bare act of eating.
• Counter-example: the bahuvrīhi bhikṣā́priyaḥ, where the prior member is antodātta (udātta kṣā).
• 3806 gobiḍālasiṃhasaindhaveṣūpamāne 6.2.72 — before go, biḍāla, siṃha, saindhava used as the standard of comparison: dhānyaṃ gaur iva → dhā́nyagavaḥ (vyāghrādi, SK 735). Here the ṭac of SK 729 loses its cit accent. Also góbiḍālaḥ, tṛ́ṇasiṃhaḥ, sáktusaindhavaḥ. Counter-example: paramasiṃhaḥ.
• 3807 ake jīvikārthe 6.2.73 — before an aka-final member, in a compound denoting a livelihood (SK 711): dántalekhakaḥ (ṇvul; udātta dán). Counter-examples: ramaṇīyakartā (not aka); ikṣubhakṣikāṃ me dhārayasi (not a livelihood).
• 3808 prācāṃ krīḍāyām 6.2.74 — before an aka-final member, in the name of a game of the eastern people. Here prācām is a sixth-case word (Subodhinī), not "in the view of the easterners". úddālakapuṣpabhañjikā (ṇvul by SK 3286; udātta ú).
• Counter-examples: jīvaputrapracāyikā (a northern game); tava puṣpapracāyikā (a turn). For the second the print has ṇvul where SK 3288 would give ṇvuc.

V.G. ādir udāttaḥ before aṇ, pāla and ṇini — running example: chatra + dhṛ + aṇ → chátradhāraḥ (udātta chá); tantu + vā + aṇ → tántuvāyaḥ (udātta tán); puṣpa + hārin → púṣpahārī (udātta pú)

• 3809 aṇi niyukte 6.2.75 — before an aṇ-final member (karmaṇy aṇ, SK 2913), in a compound denoting one appointed: chátradhāraḥ (udātta chá). Counter-example: kāṇḍalāvaḥ. Subodhinī: niyukta comes from yujir yoge; the yukta of 3800 comes from yuja samādhau.
• 3810 śilpini cākṛñaḥ 6.2.76 — the same before aṇ in a compound denoting an artisan, unless the aṇ is after kṛñ: tántuvāyaḥ (aṇ by SK 2914; udātta tán). Counter-examples: kāṇḍalāvaḥ; kumbhakāraḥ.
• 3811 saṃjñāyāṃ ca 6.2.77 — the same in a name, with akṛñaḥ still holding: tántuvāyo nāma kṛmiḥ (udātta tán). Counter-example: rathakāro nāma brāhmaṇaḥ.
• 3812 gotantiyavaṃ pāle 6.2.78 — go, tanti, yava before pāla, where the compound does not mean "appointed": gópālaḥ, tántipālaḥ, yávapālaḥ (udātta gó, tán, yá). Counter-examples: vatsapālaḥ; gorakṣaḥ.
• 3813 ṇini 6.2.79 — before a ṇini-final member: púṣpahārī (udātta pú). The print misnumbers this sūtra 4813.
• 3814 upamānaṃ śabdārthaprakṛtāv eva 6.2.80 — a prior member that is a standard of comparison gets 3813's accent only when the base of the ṇini means sound. úṣṭrakrośī, dhvā́ṅkṣarāvī (ṇini by SK 2989; udātta ú, dhvā).
• Counter-example: vṛkavañcī (vañc does not mean sound).
• Counter-example: gardabhoccārī. ud + car means sound, but the root car alone does not.
• 3815 yuktārohyādayaś ca 6.2.81 — the listed compounds yuktārohin and the rest: yúktārohī, ā́gatayodhī, kṣī́rahotā (udātta yú, ā, kṣī). The list restricts both members. For kṣīrahotā this rule overrides manktin- (SK 3885).

V.H. ādir udāttaḥ before ja, grāma, ghoṣādi, śālā and prastha (and the penultimate vowel of a bahvac before ja) — running example: kuṭī + ja → kúṭījaḥ (udātta kú); indra + prastha → índraprasthaḥ (udātta í)

• 3816 dīrghakāśatuṣabhrāṣṭravaṭaṃ je 6.2.82 — a prior member ending in a long vowel, and kāśa, tuṣa, bhrāṣṭra, vaṭa, before ja. kuṭī + ja (ḍa, SK 3007): kúṭījaḥ. Also kā́śajaḥ, túṣajaḥ, bhrā́ṣṭrajaḥ, váṭajaḥ (udātta kú, kā, tú, bhrā, vá).
• 3817 antyāt pūrvaṃ bahvacaḥ 6.2.83 — before ja, a many-vowelled prior member has its penultimate vowel udātta, in place of the first.
• upasárajaḥ (udātta sá).
• āmalákījaḥ (udātta lá). Here this rule, being later, overrides 3816 for the long-ī-final āmalakī (Subodhinī).
• Counter-example: dagdhajāni (dagdha has only two vowels).
• 3818 grāme 'nivasantaḥ 6.2.84 — before grāma, unless the prior member denotes the dwellers: mállagrāmaḥ (grāma = a band; udātta má), dévagrāmaḥ (udātta dé). Counter-example: dākṣigrāmaḥ, the Dākṣis' dwelling.
• 3819 ghoṣādiṣu ca 6.2.85 — before ghoṣa and the rest: dā́kṣighoṣaḥ, dā́kṣikaṭaḥ, dā́kṣihradaḥ (udātta dā). The print reads hrada. On some views anivasantaḥ continues (Subodhinī).
• 3820 chātryādayaḥ śālāyām 6.2.86 — chātri and the rest before śālā: chā́triśālā, vyā́ḍiśālā (udātta chā, vyā). When the compound is neuter, this rule still applies by precedence of the earlier rule (pūrvavipratiṣedha) over SK 3857, which would make śāla ādyudātta: chā́triśālam.
• 3821 prasthe 'vṛddham akarkyādīnām 6.2.87 — before prastha, a prior member that is not vṛddha, karkī and the rest excepted: índraprasthaḥ (udātta í). Counter-examples: dākṣiprasthaḥ (vṛddha); karkīprasthaḥ, makarīprasthaḥ.
• 3822 mālādīnāṃ ca 6.2.88 — mālā and the rest before prastha. The rule is for vṛddha words: mā́lāprasthaḥ, [śóṇāprasthaḥ] (śoṇa is vṛddha by SK 1338; udātta mā, śó).

Part VI — SK 3823–3874: the last ādyudātta pūrvapadas; the pūrvapada made antodātta (antaḥ) in tatpuruṣa and bahuvrīhi; the uttarapada made ādyudātta (uttarapadādiḥ) in bahuvrīhi, avyayībhāva, dvigu and tatpuruṣa; the prakṛtyā rules from bhagāla to vanaspati

Covers SK 3823–3874 (52 sūtras): all present.

• The Aṣṭādhyāyī spine. The run is 6.2.89–6.2.140, taken in strict order. No rule is moved, and none is pulled in from another book. Bhaṭṭoji adds two vārttikas: divodāsa etc. in the Veda (3825) and the precedence of the avyaya accent (3864).
• AS-refs. All 52 are checked against references/ashtadhyayi/sutraani-data.json, by SK number and by sūtra text, and all match. So does every SK cross-reference cited below.
• The headings, and how far they run. Bhaṭṭoji names each limit at 3826, 3840, 3845 and 3871. The tatpuruṣa carry-over is the Subodhinī's.
• pūrvapadam. It comes from bahuvrīhau prakṛtyā pūrvapadam (SK 3735, 6.2.1) and governs 3823–3844. The first-case word naming the member is read in the sense of the sixth ("the final of X"; Subodhinī at 3798, 3826, 3846).
• ādir udāttaḥ (SK 3798, Part V) makes the pūrvapada's first syllable udātta. It closes with 3825.
• antaḥ (3826) makes the pūrvapada's final udātta, "up to uttarapadādiḥ". It covers 3826–3844.
• bahuvrīhau (3840) runs "up to the avyayībhāva", so it covers 3840–3854. 3855 names the avyayībhāva.
• uttarapadādiḥ (3845) makes the uttarapada's first syllable udātta. uttarapada runs to the end of the pāda. ādi runs up to prakṛtyā bhagālam (3871), so it covers 3845–3870. At 3859, ādi is restated to move it onto the pūrvapada.
• tatpuruṣe. The Subodhinī says that tatpuruṣe of 3857 is stated "for what follows" and continues "even across intervening rules" (at 3857 and 3873). It governs 3857–3873, though not 3872, which is bahuvrīhi.
• prakṛtyā (3871) runs up to antaḥ (SK 3877), so it covers 3871–3874 and continues into Part VII.
• Smaller carry-overs:
• amahannavam 3823→3824; arme 3824→3825; pure 3833→3834–3835.
• dikśabdāḥ 3837→3838–3839; saṃjñāyām 3840→3841; udarāśveṣuṣu 3841→3842.
• karṇa 3846→3847; saṃjñaupamyayoḥ 3847→3848–3849.
• soḥ 3851→3852–3854; chandasi 3853→3854.
• napuṃsake 3857→3858–3859; akarmadhāraye 3864→3865; putraḥ 3866→3867; aprāṇiṣaṣṭhyāḥ 3868→3869.
• The en.yml marks the source-rule identifications that are the translator's.
• Accent. udātta is acute, independent svarita is grave, and anudātta and dependent svarita are unmarked. Each line names the syllable the rule makes udātta.
• Unaccented examples. Most examples have no accented copy. Only the syllable the rule affects is marked, and counter-examples are left unmarked unless Bhaṭṭoji or the Subodhinī states their accent.
• Accented copies. The Vedic quotations at 3825, 3840, 3850–3854, 3873 and 3874 have accented copies, and their signs are transcribed.
• Single substitutes. A merged vowel takes udātta by ekādeśa udāttenodāttaḥ (SK 3658). At the head of a pada it is optionally svarita by SK 3659, as in guḍòdakam (3830). A yaṇ before an anudātta makes it svarita by SK 3657, as in haryàśvaḥ (3841).
• Left unresolved. At 3850, the signs on amṛtam (ours = print) put no udātta on mṛ, though the rule requires it. At 3871, kumbhīpālam cannot be madhyodātta.
• Print readings and audit corrections followed. These are NSP 1915 readings, scan-confirmed (collation batch19–21; audits H21–H24).
• Corrections after the print:
• 3828 antodāttam; 3831 atirātraḥ; 3837 antodāttā bhavanty eṣu.
• 3851: nañsubhyām, ity asyāpavādaḥ, suvahmā, supeśasas karati, paratvāt, susrotaskaḥ.
• 3852: suprátūrtim anehásam. 3853: yad ādyudāttaṃ and suráthā̃.
• 3858 dākṣikanthā; 3859 cihaṇakantham; 3866 māhiṣaputraḥ and pumbhyaḥ.
• 3864: the vārttika ends at pūrvavipratiṣedhena, and kucelam and kurājyam are Bhaṭṭoji's examples (M).
• Not in the collation. 3852 sā́mrājyāya: the print has a svarita stroke where ours has the bar. 3854 suvī́reṇa and gómataḥ.
• Ours = print, kept or queried:
• 3830 guhodakam is read as guḍodakam, since the print's own Subodhinī and Wikisource have guḍodakam (class C, a shared misprint).
• 3834 tathā (WS yathā). 3835 mṛdeḥ (the Subodhinī implies mṛdu).
• 3837: the heading janapada; pūrvayātam may be a doublet.
• 3846 śaṅkukarṇaḥ keeps its short u. 3851: the bar under mā in animānáḥ is as printed.
• 3860 nāgara- (Subodhinī nagara-). 3863 -sthalam (Subodhinī -sthalī).
• 3871 kumbhīpālam (WS kumbhīkapālam).
• 3873: amsvaraḥ is read as the accent of ām. prapacatideśyārtham is read with the Subodhinī's "and the like".
• Audit sense-points:
• 3825 (H21, MUST-FIX). The saṃghāta-vigṛhīta gloss covers madra alone and madrāśma. It does not cover aśman alone.
• 3829. "not maidenhood itself". 3835. nehāsti is the print's own lemma.
• 3844. śuṣka is ādyudātta by SK 3694, not by a phiṭ-sūtra.
• 3851 (H23, MUST-FIX). With kap, the syllable before kap is udātta (sukarmákaḥ), not su.
• 3852 (H23, MUST-FIX). sā́mrājyāya.
• 3874. nṝ + ap by ṝdor ap (SK 3232).
• Where the commentary lies. Only the kaumudī is translated (SK-TRANSLATE §25). The commentary is the Subodhinī, filed as follows (NSP pp. 643–649).
• Under a neighbour:
• 3823–3824: sk.3822.s1, which is Part V's unit.
• 3827: tail of sk.3826.s1, plus sk.3827.s1. 3828: tail of sk.3827.s1.
• 3829: sk.3829.s1, plus the head of sk.3825.n1. That editorial note is no NSP footnote. It is the Subodhinī run-on printed below the rules on p. 644, and it also carries 3830, 3831, 3832 and the lemma of 3833.
• 3833's gloss: end of sk.3829.s1. 3835: tail of sk.3834.s1, plus sk.3835.s1. 3837: tail of sk.3835.s1, plus sk.3837.s1.
• 3841–3843: inside sk.3840.s1. 3851's head: tail of sk.3850.s1. 3853: all in sk.3851.s1.
• 3861: tail of sk.3860.s1. The 3864 vārttika: tail of sk.3863.s1. 3866–3868: sk.3865.s1.
• 3872: tail of sk.3871.s1, plus sk.3872.s1. 3873: sk.3872.s1 from "gatikā—", plus sk.3873.s1. 3874: tail of sk.3873.s1, plus sk.3874.s1.
• No Subodhinī on 3836, 3849 or 3852, or on the sūtra of 3864 (print confirms).
• No BM anywhere in Part VI. Our layer is empty, and Wikisource uttara-4/0112–0116 prints only the mūla. sk.3861.b1 is misfiled: it is BM on SK 91.
• Where Part VI ends. Part VI ends at 3874 ubhe vanaspatyādiṣu yugapat (6.2.140), where both members keep their accent. Part VII opens with devatādvandve ca (6.2.141, SK 3875), still under prakṛtyā.

VI.A. The last ādyudātta pūrvapadas: before nagara and arma — running example: brahma + nagaram → bráhmanagaram (udātta on bra); gupta + armam → gúptārmam (udātta on gu)

• 3823 amahannavaṃ nagare 'nudīcām 6.2.89 — before nagara, a pūrvapada other than mahat or nava is ādyudātta (ādir udāttaḥ, SK 3798), unless the town is a northern one: bráhmanagaram, udātta on bra.
• "Other than" applies to each word separately (Subodhinī).
• Counter-examples: mahat and nava (mahānagaram, navanagaram); a northern town (kārtikanagaram).
• 3824 arme cāvarṇaṃ dvyac tryac 6.2.90 — before arma, a pūrvapada of two or three vowels ending in a/ā is ādyudātta, still excluding mahat and nava (from 3823): gúptārmam, kúkkuṭārmam, udātta on gu, ku.
• Not a/ā-final: bṛhadarmam. Four vowels: kapiñjalārmam.
• Excluded by name: mahārmam (mahat → mahā by SK 807 is a-final, so it would be within reach, Subodhinī) and navārmam.
• 3825 na bhūtādhikasaṃjīvamadrāśmakajjalam 6.2.91 — before arma (from 3824), bhūta, adhika, saṃjīva, madra, aśman, kajjala are not ādyudātta: bhūtārmam, adhikārmam, saṃjīvārmam, kajjalārmam.
• madra and aśman are named both singly and as the compound (saṃghāta-vigṛhīta): madrārmam, aśmārmam, madrāśmārmam.
• Subodhinī: the rule really covers madra alone and madrāśma, which is a-final by SK 796. aśman is not a-final, so aśmārmam only follows the Vṛtti (audit H21).
• Vārttika: in the Veda, divodāsa etc. are added to the ādyudātta section: dívodāsāya dāśúṣe, udātta on di. The sixth-case ending of divas is not elided (aluk).

VI.B. The pūrvapada made antodātta (antaḥ): sarva, and before giri, nikāya, kumārī, udaka, a dvigu and sabhā — running example: sarva + śvetaḥ → sarváśvetaḥ (udātta on va, the final of sarva; the rest anudātta)

• 3826 antaḥ 6.2.92 — adhikāra: from here on, the final of the pūrvapada is udātta, up to uttarapadādiḥ (3845). It is the frame for groups B–E.
• 3827 sarvaṃ guṇakārtsnye 6.2.93 — sarva is antodātta when it denotes the totality of a quality: sarváśvetaḥ "white all over" (a karmadhāraya by SK 726) and sarvámahān, udātta on va.
• Not sarva: paramaśvetaḥ, even though here too the whiteness pervades its substrate.
• Not a quality: sarvasauvarṇaḥ, where the totality is of the product.
• Not the totality of the quality: sarvaśvetaḥ = sarveṣāṃ śvetataraḥ "whiter than all", where the totality is of the possessors.
• 3828 saṃjñāyāṃ girinikāyayoḥ 6.2.94 — in a name, before giri or nikāya: añjanā́giriḥ (the ā is by SK 1038; udātta on nā) and mauṇḍínikāyaḥ (udātta on ḍi).
• Not names: paramagiriḥ, brāhmaṇanikāyaḥ.
• 3829 kumāryāṃ vayasi 6.2.95 — before kumārī, when an age is conveyed: vṛddhákumārī "an old maiden", udātta on dha.
• kumārī here means only "not joined to a man", so it can stand in apposition with vṛddha. The age that counts is the one the other word names, not maidenhood itself.
• Counter-example: paramakumārī.
• 3830 udake 'kevale 6.2.96 — before udaka, when the compound means a mixture (akevala = miśra): guḍa + udakam.
• The final ḍa is udātta and the u of udaka is anudātta, so their single substitute o is udātta (SK 3658): guḍódakam.
• Or the o is svarita by SK 3659: guḍòdakam.
• Not a mixture: śītodakam.
• 3831 dvigau kratau 6.2.97 — before a dvigu, in the name of a sacrifice: garga + trirātraḥ → gargátrirātraḥ "the Gargas' three-night rite", udātta on ga, the final of garga.
• Not a dvigu: atirātraḥ. Not a rite: bilvasaptarātraḥ.
• 3832 sabhāyāṃ napuṃsake 6.2.98 — before sabhā, in a neuter compound (neuter by SK 827): gopālásabham (udātta on la) and strī́sabham (udātta on strī).
• Not sabhā: brāhmaṇasenam. Not neuter: rājasabhā.
• Only the neuter specifically taught for sabhā (pratipadokta, SK 826–827) counts. So the rule does not apply to the bahuvrīhi in ramaṇīyasabhaṃ brāhmaṇakulam.

VI.C. Antodātta before pura and bila — running example: devadatta + puram → devadattápuram (udātta on ta, the final of devadatta)

• 3833 pure prācām 6.2.99 — before pura, in the eastern country. The Subodhinī reads prācām as a region, not as the eastern grammarians' view. devadattápuram, nāndī́puram, udātta on ta, dī.
• Not eastern: śivapuram.
• 3834 ariṣṭagauḍapūrve ca 6.2.100 — before pura (from 3833), in a compound beginning with ariṣṭa or gauḍa: ariṣṭápuram, gauḍápuram, udātta on ṣṭa, ḍa.
• "pūrva" makes the condition a bahuvrīhi, "having them first". So the rule also reaches longer prior members, and their final is udātta: ariṣṭāśritápuram, gauḍabhṛtyápuram.
• 3835 na hāstinaphalakamārdeyāḥ 6.2.101 — before pura, hāstina, phalaka, mārdeya are not antodātta, which blocks 3833: hāstinapuram, phalakapuram, mārdeyapuram. No accent is stated for them.
• mārdeya is "the offspring of Mṛdi", with ḍhak because the word is in the śubhrādi list (SK 1126).
• 3836 kusūlakūpakumbhaśālaṃ bile 6.2.102 — before bila, kusūla, kūpa, kumbha, śāla are antodātta: kusūlábilam "the opening of a granary", kūpábilam, kumbhábilam, śālábilam, udātta on la, pa, bha, la.
• Not these words: sarpabilam. Not bila: kusūlasvāmī.

VI.D. Direction words as prior member — running example: pūrva + pañcālāḥ → pūrvápañcālāḥ (udātta on va, the final of pūrva)

• 3837 dikśabdā grāmajanapadākhyānacānarāṭeṣu 6.2.103 — a direction word is antodātta before a word for a village, a country or a narrative, or before the word cānarāṭa. For cānarāṭa the form itself is meant; for the others, the meaning (Subodhinī).
• Village names (SK 727): pūrvéṣukāmaśamī, where a + i → e is udātta, or svarita by SK 3659; and aparákṛṣṇamṛttikā.
• Country: pūrvápañcālāḥ (karmadhāraya by SK 737).
• Narrative: pūrváyātam, pūrváyāyātam (yāyāta is a work on Yayāti).
• cānarāṭa: pūrvácānarāṭam.
• "śabda" lets in a direction word used of time, as pūrva is in pūrvayāyātam.
• 3838 ācāryopasarjanaś cāntevāsini 6.2.104 — direction words (from 3837) are also antodātta before a word for a pupil in which the teacher is subordinate: pūrvápāṇinīyāḥ "the earlier Pāṇinīyas", udātta on va.
• No teacher in the word: pūrvāntevāsī. Not a pupil: pūrvapāṇinīyaṃ śāstram.
• 3839 uttarapadavṛddhau sarvaṃ ca 6.2.105 — before an uttarapada whose vṛddhi is taught under the heading uttarapadasya (SK 1396), sarva and the direction words are antodātta: sarvápāñcālakaḥ, aparápāñcālakaḥ, udātta on va, ra.
• The affix is vuñ by SK 1349, and the later member's vṛddhi is by SK 1398–1399.
• The heading is named so as to exclude sarvabhāsaḥ and sarvakārakaḥ, whose vṛddhi comes from SK 254.

VI.E. In a bahuvrīhi the prior member is antodātta — running example: viśva + karman → viśvákarmā (udātta on śva, the final of viśva, replacing its own ādyudātta)

• 3840 bahuvrīhau viśvaṃ saṃjñāyām 6.2.106 — in a bahuvrīhi that is a name, viśva is antodātta: viśvákarmā, viśvádevaḥ, and Vedic ā́ viśvádevaṃ sátpatim.
• This is an apavāda to the pūrvapada's own accent (SK 3735), which would keep víśva ādyudātta, viśva having nit kvan.
• Not a bahuvrīhi: viśvadevāḥ (karmadhāraya). Not a name: víśvadevaḥ.
• The bahuvrīhi adhikāra opens here.
• 3841 udarāśveṣuṣu 6.2.107 — in a bahuvrīhi that is a name (from 3840), before udara, aśva, iṣu.
• vṛkódaraḥ and mahéṣuḥ: the merged vowel is udātta.
• hari + aśvaḥ: the udātta i becomes y and the following a is svarita (SK 3657), giving haryàśvaḥ.
• Subodhinī: maheṣuḥ is antodātta anyway; the true example is suvarṇapuṅkheṣuḥ.
• 3842 kṣepe 6.2.108 — the same before udara, aśva and iṣu when censure is meant, with no name required: ghaṭódaraḥ "pot-bellied", kaṭukā́śvaḥ, calācaléṣuḥ.
• In anudara both rules apply. By vipratiṣedha the later nañsubhyām (SK 3906) wins and the compound is antodātta: anudaráḥ.
• 3843 nadī bandhuni 6.2.109 — before bandhu, a prior member ending in a nadī vowel (SK 266) is antodātta: gārgī́bandhuḥ, udātta on gī. This overrides the ñit ādyudātta of gā́rgī (SK 3686).
• Not nadī-final: bráhmabandhuḥ (brahman is ādyudātta by its nit affix manin).
• Not bandhu: gā́rgīpriyaḥ (the pūrvapada keeps its own accent).
• 3844 niṣṭhopasargapūrvam anyatarasyām 6.2.110 — a prior member ending in a niṣṭhā and preceded by an upasarga is optionally antodātta: pradhautápādaḥ, udātta on ta.
• In the other option it keeps the gati accent: prádhautapādaḥ (Subodhinī).
• Not a niṣṭhā: prasevakamukhaḥ. No upasarga: śuṣkamukhaḥ (śuṣka is ādyudātta by SK 3694).

VI.F. The uttarapada's first syllable (uttarapadādiḥ): karṇa, other body-parts and śṛṅga in a bahuvrīhi — running example: śukla + karṇaḥ → śuklakárṇaḥ (udātta on kar, the first syllable of karṇa)

• 3845 uttarapadādiḥ 6.2.111 — adhikāra: the first syllable of the uttarapada is udātta. uttarapada runs to the end of the pāda; ādi runs up to prakṛtyā bhagālam (3871).
• The target moves from the pūrvapada's final to the uttarapada's first syllable.
• 3846 karṇo varṇalakṣaṇāt 6.2.112 — in a bahuvrīhi, karṇa after a word for a colour or a mark is ādyudātta: śuklakárṇaḥ "white-eared" and śaṅkukárṇaḥ "with a peg-mark on the ear".
• The mark is a cattle brand (Subodhinī).
• By SK 1036 one would expect śaṅkū-; the short u is kept as transmitted.
• Not karṇa: śvetapādaḥ. Not after a colour or mark: śobhanakarṇaḥ (pūrvapada accent).
• 3847 saṃjñaupamyayoś ca 6.2.113 — karṇa is also ādyudātta in a name or a comparison: maṇikárṇaḥ (name) and gokárṇaḥ "cow-eared", udātta on kar.
• 3848 kaṇṭhapṛṣṭhagrīvājaṅghaṃ ca 6.2.114 — kaṇṭha, pṛṣṭha, grīvā, jaṅghā are ādyudātta in a name or a comparison (from 3847). The udātta is on kaṇ, pṛ, grī, jaṅ.
• Names: śitikáṇṭhaḥ, kāṇḍapṛ́ṣṭhaḥ, sugrī́vaḥ, nāḍījáṅghaḥ.
• Comparisons: kharakáṇṭhaḥ, gopṛ́ṣṭhaḥ, aśvagrī́vaḥ, gojáṅghaḥ.
• Subodhinī: in sugrīvaḥ, SK 3906 ought to prevail. This is recorded, and the example is kept.
• 3849 śṛṅgam avasthāyāṃ ca 6.2.115 — śṛṅga is ādyudātta when a stage of life is meant, that is, an animal's age marked by the growth of its horns. It is also ādyudātta in a name or a comparison. The udātta is on śṛṅ.
• Stage of life: udgataśṛ́ṅgaḥ, dvyaṅgulaśṛ́ṅgaḥ.
• Name: ṛṣyaśṛ́ṅgaḥ. Comparison: meṣaśṛ́ṅgaḥ.
• Not a stage of life: sthūlaśṛṅgaḥ.

VI.G. After nañ and su in a bahuvrīhi: exceptions to nañsubhyām (SK 3906) — running example: na me jarā → a + jara → ajáram (udātta on ja); su + karman → sukármāṇaḥ (udātta on kar)

• 3850 naño jaramaramitramṛtāḥ 6.2.116 — after nañ, jara, mara, mitra, mṛta are ādyudātta, an apavāda to SK 3906: ajáram, amáram, amítram ardaya, śrávo devéṣv amṛ́tam, udātta on ja, ma, mi, mṛ.
• The accented copy, which agrees with the print, puts no udātta on mṛ. This is unresolved.
• Not after nañ: brāhmaṇamitraḥ. Another word: aśatruḥ, which is antodātta by SK 3906.
• 3851 sor manasī alomoṣasī 6.2.117 — after su, an uttarapada ending in man or as, other than loman and uṣas, is ādyudātta. The udātta is on kar, vah, man, var, pe:
• sukármāṇaḥ suyújaḥ; animānáḥ suváhmā; sumánāḥ suvárcāḥ; supéśasas karati.
• Not after su: kṛtakarmā. Not man- or as-final: surā́jā (rājan has kanin). Excluded by name: sulomā, sūṣāḥ.
• With kap, kapi pūrvam (SK 3907) prevails as the later rule, and the syllable before kap is udātta: sukarmákaḥ, susrotáskaḥ.
• 3852 kratvādayaś ca 6.2.118 — kratu and the rest are also ādyudātta after su: sā́mrājyāya sukrátuḥ, suprátīkaḥ, suhávyaḥ, suprátūrtim anehásam, udātta on kra, pra, ha, pra.
• 3853 ādyudāttaṃ dvyac chandasi 6.2.119 — in the Veda, after su, an uttarapada that is already ādyudātta and has two vowels stays ādyudātta: ádhā sváśvāḥ (aśva has nit kvan) and suráthā̃ ātithigvé (ratha has nit kthan), udātta on a, ra.
• Not ādyudātta: yā́ subāhúḥ. bāhu is antodātta, so SK 3906 gives the final udātta.
• Three vowels: sugúr asat suhiraṇyáḥ.
• 3854 vīravīryau ca 6.2.120 — vīra and vīrya are also ādyudātta after su in the Veda: suvī́reṇa rayíṇā, suvī́ryasya gómataḥ, udātta on vī.
• Naming vīrya is a jñāpaka that yato 'nāvaḥ (SK 3701) does not make yat-final vīrya ādyudātta. If it did, 3853 would already cover the form.

VI.H. The uttarapada ādyudātta in an avyayībhāva and a dvigu — running example: upa + kūlam → upakū́lam (udātta on kū)

• 3855 kūlatīratūlamūlaśālā'kṣasamam avyayībhāve 6.2.121 — in an avyayībhāva, kūla, tīra, tūla, mūla, śālā, akṣa, sama are ādyudātta: upakū́lam, upatī́ram, upatū́lam, upamū́lam, upaśā́lam.
• In upā́kṣam the merged ā is udātta.
• In suṣámam, niḥṣámam the ṣ is asiddha, so the word still counts as sama. These two forms are avyayībhāvas by the tiṣṭhadgu list.
• Not these words: upakumbham. Not an avyayībhāva: paramakūlam.
• 3856 kaṃsamanthaśūrpapāyyakāṇḍaṃ dvigau 6.2.122 — in a dvigu, kaṃsa, mantha, śūrpa, pāyya, kāṇḍa are ādyudātta: dvikáṃsaḥ "bought for two kaṃsas", dvimánthaḥ, dviśū́rpaḥ, dvipā́yyam, dvikā́ṇḍam. The udātta is on kaṃ, man, śūr, pāy, kāṇ.
• In each the taddhita affix is elided by luk (Subodhinī).
• Not a dvigu: paramakaṃsaḥ.

VI.I. Neuter tatpuruṣas in śālā and kanthā, and the cihaṇādi pūrvapadas — running example: brāhmaṇa + śālā → (neuter tatpuruṣa) brāhmaṇaśā́lam (udātta on śā)

• 3857 tatpuruṣe śālāyāṃ napuṃsake 6.2.123 — in a neuter tatpuruṣa ending in śālā, śālā is ādyudātta: brāhmaṇaśā́lam (neuter by SK 828).
• Not a tatpuruṣa: dṛḍhaśālaṃ brāhmaṇakulam (bahuvrīhi). Not śālā: brāhmaṇasenam. Not neuter: brāhmaṇaśālā.
• The Subodhinī says tatpuruṣe is stated here for the rules that follow (see above).
• 3858 kanthā ca 6.2.124 — kanthā is also ādyudātta in a neuter tatpuruṣa: sauśamikántham, āhvarakántham, udātta on kan. These are place names, neuter by saṃjñāyāṃ kanthośīnareṣu.
• Not neuter: dākṣikanthā.
• 3859 ādiś cihaṇādīnām 6.2.125 — in a neuter tatpuruṣa ending in kanthā, the first syllable of the pūrvapada cihaṇa etc. is udātta: cíhaṇakantham, mándurakantham, udātta on ci, man.
• ādi is restated to move the udātta from the uttarapada to the pūrvapada.

VI.J. Tatpuruṣa uttarapadas made ādyudātta under a condition — running example: putra + celam → putracélam "a son no better than a rag" (udātta on ce)

• 3860 celakheṭakaṭukakāṇḍaṃ garhāyām 6.2.126 — when contempt is expressed, cela, kheṭa, kaṭuka, kāṇḍa are ādyudātta: putracélam, nāgarakhéṭam, dadhikáṭukam, prajākā́ṇḍam, udātta on ce, khe, ka, kā.
• The compounds are vyāghrādi compounds (SK 735).
• No contempt: paramacelam.
• 3861 cīram upamānam 6.2.127 — cīra is ādyudātta when it is the standard of comparison: vastracī́ram "a garment like a rag" and kambalacī́ram, udātta on cī.
• Not a standard of comparison: paramacīram.
• 3862 palalasūpaśākaṃ miśre 6.2.128 — palala, sūpa, śāka are ādyudātta when a mixture is expressed (compound by SK 697): ghṛtapálalam, ghṛtasū́paḥ, ghṛtaśā́kam, udātta on pa, sū, śā.
• Not a mixture: paramapalalam.
• 3863 kūlasūdasthalakarṣāḥ saṃjñāyām 6.2.129 — kūla, sūda, sthala, karṣa are ādyudātta in a name, here names of villages: dākṣikū́lam, śāṇḍisū́dam, dāṇḍāyanasthálam, dākṣikárṣaḥ. The udātta is on kū, sū, stha, ka.
• Not a name: paramakūlam.
• 3864 akarmadhāraye rājyam 6.2.130 — rājya is ādyudātta in a tatpuruṣa that is not a karmadhāraya: brāhmaṇarā́jyam, udātta on rā.
• A karmadhāraya: paramarājyam.
• Vārttika: the avyaya accent (SK 3736) prevails over the accent of cela, rājya etc. by pūrvavipratiṣedha. So in kucelam, kurājyam the ku keeps its own accent, and the uttarapada is not made ādyudātta.
• 3865 vargyādayaś ca 6.2.131 — vargya and the rest are also ādyudātta in a non-karmadhāraya tatpuruṣa: arjunavárgyaḥ, vāsudevapákṣyaḥ, udātta on va, pa.
• A karmadhāraya: paramavargyaḥ.
• vargādi is a sub-list (antargaṇa) inside digādi, whose words take yat (SK 1429).
• 3866 putraḥ puṃbhyaḥ 6.2.132 — putra after words for males is ādyudātta: dāśakipútraḥ, māhiṣapútraḥ, udātta on pu.
• Not putra: kaunaṭimātulaḥ. Not after a male: dākṣīputraḥ.
• 3867 nācāryarājartviksaṃyuktajñātyākhyebhyaḥ 6.2.133 — putra (from 3866) is not ādyudātta after names of a teacher, a king, a priest, an in-law or a kinsman. "Names" takes in their synonyms and particular kinds:
• ācāryaputraḥ, upādhyāyaputraḥ, śākaṭāyanaputraḥ;
• rājaputraḥ, īśvaraputraḥ, nandaputraḥ;
• ṛtvikputraḥ, yājakaputraḥ, hotuḥputraḥ (aluk by SK 981);
• śyālaputraḥ; jñātiputraḥ, bhrātuṣputraḥ (ṣ by SK 144).
• The rule only cancels 3866. The resulting accent is not stated.
• 3868 cūrṇādīny aprāṇiṣaṣṭhyāḥ 6.2.134 — cūrṇa and the rest are ādyudātta after a sixth-case word naming something not living: mudgacū́rṇam "mudga-bean flour", udātta on cū.
• A living being: matsyacūrṇam.
• 3869 ṣaṭ ca kāṇḍādīni 6.2.135 — the six words of 3860–3863 (kāṇḍa, cīra, palala, sūpa, śāka, kūla) are also ādyudātta after such a word, without their earlier conditions. The udātta is on kā, cī, pa, sū, śā, kū:
• darbhakā́ṇḍam, darbhacī́ram, tilapálalam, mudgasū́paḥ, mūlakaśā́kam, nadīkū́lam.
• Not one of the six: rājasūdaḥ. A living being: dattakāṇḍam.
• 3870 kuṇḍaṃ vanam 6.2.136 — kuṇḍa, used figuratively of a forest, is ādyudātta: darbhakúṇḍam "a darbha thicket", udātta on ku.
• Its literal sense: mṛtkuṇḍam.
• This is the last rule of the ādi adhikāra.

VI.K. prakṛtyā: the uttarapada, or both members, keeps its own accent — running example: kumbhī + bhagālam → kumbhībhagā́lam (bhagāla keeps its madhyodātta, udātta on gā; kumbhī loses its accent)

• 3871 prakṛtyā bhagālam 6.2.137 — in a tatpuruṣa, an uttarapada meaning bhagāla keeps its own accent. The Subodhinī takes bhagālam by its meaning, not its form, on the authority of the commentarial explanation (vyākhyāna) alone.
• kumbhībhagā́lam, kumbhīnadā́lam are madhyodātta (by phiṭ 42).
• kumbhīpālam is printed, but it cannot be madhyodātta; WS has kumbhīkapālam.
• prakṛtyā is an adhikāra up to antaḥ (SK 3877).
• 3872 śiter nityābahvajbahuvrīhāv abhasat 6.2.138 — in a bahuvrīhi after śiti, an uttarapada that never has many vowels, other than bhasat, keeps its own accent. This is an apavāda to SK 3735.
• śitipā́daḥ: pāda is ādyudātta as vṛṣādi (SK 3691).
• śityáṃsaḥ: aṃsa is ādyudātta by its nit affix.
• Not after śiti: darśanī́yapādaḥ, which follows SK 3735, with darśanī́ya by SK 3733.
• bhasat: śítibhasat, where śiti keeps its ādyudātta (phiṭ 33).
• 3873 gatikārakopapadāt kṛt 6.2.139 — in a tatpuruṣa (carried from 3857; 3872's bahuvrīhi is not svarita, so it does not continue), a kṛt-final uttarapada after a gati, a kāraka or an upapada keeps its own kṛt accent:
• After a gati: prakā́rakaḥ, praháraṇam.
• After a kāraka: śóṇā dhṛṣṇū́ nṛvā́hasā, idhmapravráścanaḥ.
• After an upapada: uccaiḥkā́ram (SK 3381), īṣatkáraḥ.
• The pūrvapada loses its accent.
• Not after these: devasya kārakaḥ. Its sixth case is a residual relation (śeṣa), not the object (SK 709).
• Some say "kṛt" is only for clarity. In prapacatitarām, the later-taught (satiśiṣṭa) accent of ām prevails. Others say "kṛt" serves to exclude prapacatideśya and the like, which then take the avyaya accent (SK 3736).
• 3874 ubhe vanaspatyādiṣu yugapat 6.2.140 — in vanaspati and the rest, both members at once keep their own accent, so each compound has two udāttas:
• vánaspátiṃ vána ā́.
• bṛ́haspátiṃ yáḥ: bṛhat is ādyudātta by nipātana.
• harṣayā śácīpátim: śacī has ṅīn as a śārṅgaravādi word (SK 527).
• tánūnápād ucyate.
• nárāśáṃsaṃ vājínam: the long ā is by nipātana.
• śúnaḥśépam.

Part VII — SK 3875–3926: the deity dvandva and its limit; the antaḥ heading, under which uttarapadas become antodātta after a gati, kāraka or upapada, after a seventh- or third-case word, and after nañ; then in the bahuvrīhi (prathama/pūraṇa, stana, mitra/ajina, antara, mukha, kta, nañ/su with kap); then after upasargas and particles, from vana to ner anidhāne

Covers SK 3875–3926 (52 sūtras): all present.

• The Aṣṭādhyāyī spine. The run is 6.2.141–6.2.192, unbroken and in AS order. The SK moves nothing in or out of this part.
• Non-sūtra material. There are two vārttikas: ṛṣipratiṣedho mitre (3899) and "ater dhātulope iti vaktavyam" (3925). There is one gaṇasūtra in the cārvādi list, rājāhnoś chandasi (3894).
• AS-refs. All 52 were checked against references/ashtadhyayi/sutraani-data.json by SK number and by sūtra text, and all match.
• Cross-references. Those cited below also match: SK 3651 (6.1.161), 3657 (8.2.4), 3689 (6.1.201), 3691 (6.1.203), 3701 (6.1.213), 3710 (6.1.163), 3734 (6.1.223), 3735 (6.2.1), 3736 (6.2.2), 3767 (6.2.33), 3782 (6.2.48), 3783 (6.2.49), 3784 (6.2.50), 3786 (6.2.52), 3840 (6.2.106), 3844 (6.2.110), 3845 (6.2.111), 3857 (6.2.123), 3871 (6.2.137), 3873 (6.2.139), 3874 (6.2.140), 3231 (3.3.56), 3232 (3.3.57), 3291 (3.3.116), 692 (2.1.30), 693 (2.1.31), 697 (2.1.35), 703 (2.2.9), 717 (2.1.40), 922 (6.3.26), 953 (5.4.85), 956 (5.4.71), 1049 (6.3.130), 1050 (8.4.5), 510 (4.1.54), 555 (1.4.94), 833 (5.4.153), 2017 (6.4.158), 2974 (3.2.57), 3116 (3.2.136), 3134 (3.2.154), 2988 (3.2.78), 3855 (6.2.121). The phiṭ references also match: phiṭ 26, 29, 38, 81.
• The headings, and how far they run.
• prakṛtyā (SK 3871). Bhaṭṭoji says it runs "up to antaḥ (SK 3877)", so 3875–3876 are its last two rules. ubhe … yugapat (SK 3874) carries into 3875.
• antaḥ (3877). This heading supplies "[the final syllable is] udātta" (Bhaṭṭoji: anta udāttaḥ) to every later rule in the part, and it runs on into Part VIII. The bracketed "uttarapada" that goes with it is the translator's supplement, taken from Bhaṭṭoji's gloss uttarapadam antodāttam at 3881. Our sūtra record labels 3877 "uttarapadāntodāttasvaraprakaraṇam", but the print has no heading there.
• The tatpuruṣa rules.
• tatpuruṣe comes from SK 3857. The Subodhinī says so at 3873, and it is named at 3882, 3885 (Bhaṭṭoji), 3886 and 3895.
• gatikārakopapadāt (SK 3873) carries into 3878 ("ity eva") and 3880. At 3882 the Subodhinī calls this the "tritayādhikāra", the heading of all three.
• Other carry-overs: kārakāt 3882 → 3884–3885; saṃjñāyām 3880 → 3882; tṛtīyāyāḥ 3887 → 3888.
• nañaḥ 3889 → 3890–3895. Bhaṭṭoji confirms guṇapratiṣedhe "ity eva" at 3891.
• ackau 3891 → 3892; ākrośe 3892 → 3893.
• bahuvrīhau (3896). Bhaṭṭoji says it runs up to, but not including, vanaṃ samāse (3912), so it governs 3896–3911. Within it:
• mukhaṃ svāṅgam 3901 → 3902–3903; jātikālasukhādibhyaḥ 3904 → 3905.
• nañsubhyām 3906 → 3907–3908; kapi 3907 → 3908; bahoḥ 3909 → 3910.
• From 3912, any compound. samāse (3912) ends the bahuvrīhi heading (Subodhinī: bahuvrīhāv ity asya nivṛttyartham; Bhaṭṭoji: samāsamātre). It carries into 3913–3914. Other carry-overs: upasargāt 3911 → 3912, 3914; vanam 3912 → 3913; mukham 3919 → 3920; apāt 3920 → 3921 (Subodhinī: "uttarārthaḥ"); anoḥ 3923 → 3924.
• Print readings followed (NSP 1915 pp. 649–656, scans 0659–0666; collation batches 21–23; audits H25–H28):
• Ours corrected to the print.
• Letters: 3875 ubhe (ours uṅe); 3876 śukrāmanthinau; 3878 prabhedaḥ, puruṣṭutaḥ; 3881 pravṛddhaḥ (C); 3887 svarūpagrahaṇam; 3888 sarpirmiśrāḥ, sopasargagrahaṇasya; 3889 kārṇaveṣṭakikam, guṇapratiṣedhe.
• More letters: 3896 bahūny and plural idaṃprathamāḥ; 3910 bahuguṇā / bahuguṇo; 3921 apakukṣiḥ; 3924 anvādiṣṭaḥ; 3925 atipadā gāyatrī.
• Accent: in 3875 the print has a svarita on ru of índrāváruṇau, where ours has an anudātta. This is not in the collation (audit H25). The udātta syllables are the same on either marking.
• Kept as transmitted, rendered by sense or queried for the Print Check.
• 3876: ours = print anudāttāv, rendered "begins with an anudātta" (V).
• 3881: prayutaḥ (Wikisource prayukta).
• 3883: Bhaṭṭoji's itthambhūte, rendered by the sūtra's sense (V).
• 3889: dvigārdabharathikaḥ (Wikisource and the Subodhinī's lemma have vi-).
• 3894: anāhaḥ (Wikisource anahaḥ).
• 3908: antyād eva pūrvapadam, rendered "the syllable before the last".
• 3919: siddhe bahuvrīhyartham, read as siddhe 'bahuvrīhy-.
• 3925: atikāraḥ (Wikisource and the Subodhinī read atikārakaḥ).
• The editorial "(vārtikam)" tags in 3899 and 3925 are dropped.
• Where the commentary lies (only the kaumudī is translated, SK-TRANSLATE §25):
• BM. There is no BM anywhere in this part. Our b-layer is empty, and Wikisource uttara-4/0116–0120 prints only the mūla.
• Subodhinī GAP. The whole Subodhinī on 3876–3881 and the head of 3882 (NSP p. 650, scan page-0660) is missing from our s-blocks. sk.3875.s1 jumps from agniṣṭoma to the tail on 3882. The en.yml notes for these units were read from the scan.
• Subodhinī filed under a neighbour:
• 3875: the tail of sk.3874.s1. 3882: the tail of sk.3875.s1.
• 3884: the tail of sk.3883.s1 and the head of sk.3884.s1. 3885: begins at the tail of sk.3884.s1.
• 3890: the tail of sk.3889.s1. 3891: in sk.3890.s1. 3896: the tail of sk.3895.s1. The forms of 3897: under sk.3898.s1.
• 3903: the tail of sk.3902.s1. 3905: the tail of sk.3904.s1.
• 3907 and 3908: in sk.3906.s1. The rest of 3908, and 3909: in sk.3907.s1. 3911: in sk.3910.s1.
• 3917–3919: in sk.3916.s1. 3922–3924: in sk.3921.s1.
• The tail of sk.3926.s1 belongs to 3927.
• No Subodhinī in print on 3877 or 3914.
• Audit corrections followed (H25–H28, now in the en.yml):
• MUST-FIX. 3894: the two anubandhas of khiṣṇuc are kh and c.
• MINOR.
• 3875: the Subodhinī's "rag-antaḥ" for indra is printed as such, and is taken as a slip for ran (uṇādi 186), which gives índra-.
• 3876: the scope of "what was stated before" is SK 3875 only.
• Uṇādis are cited by number throughout.
• 3882: tatpuruṣe and ktāntayoḥ are supplied from the Subodhinī.
• 3888: the Subodhinī's "apare" take sandhi as the perception of a separate identity.
• 3901–3903: the antaḥ and bahuvrīhau sources are cited.
• 3912: the mantra is read tásyédime.
• 3914: samāse is sourced from the Subodhinī on 3912.
• 3915 and 3917: upasargas are ādyudātta by phiṭ 81 (upasargāś cābhivarjam), not by SK 3767.
• 3926: the Subodhinī's "nirmūlam" is reported as a print oddity.
• Accent convention. udātta is written acute and the independent svarita grave; anudātta is unmarked. Only the syllable the rule makes udātta is marked, and counter-examples are left unmarked unless the commentary states their accent. Where Part VII ends: at 3926 ner anidhāne (6.2.192). Part VIII opens with prater aṃśvādayas tatpuruṣe (6.2.193, SK 3927), still under antaḥ.

VII.A. The last of prakṛtyā: both members of a deity dvandva keep their accent, unless the second begins with an anudātta — running example: índra + váruṇa → índrāváruṇau (two udāttas, í and vá); against indra + agní → indrāgníbhyām (one udātta, on gní)

• 3875 devatādvandve ca 6.2.141 — in a dvandva of deity names, both members keep their original accent at once (ubhe yugapat from SK 3874, prakṛtyā from SK 3871). indra is ādyudātta (ran, uṇādi 186) and varuṇa is ādyudātta (unan, uṇādi 333), and the first member takes ānaṅ by SK 922. The result is ā́ yá índrāváruṇau, with udātta on í and vá. In índrābṛ́haspátī vayám, bṛhaspati, a vanaspatyādi word, adds two udāttas of its own, so the compound has three: í, bṛ́, pá. Counter-examples: plakṣanyagrodhau (the trees are not deities), agniṣṭomaḥ (not a dvandva).
• 3876 nottarapade 'nudāttādāv apṛthivīrudrapūṣamanthiṣu 6.2.142 — the double accent of 3875 does not hold when the uttarapada begins with an anudātta, except for pṛthivī, rudra, pūṣan and manthin. agni is antodātta (ni, uṇādi 490), so indrāgníbhyāṃ káṃ vṛ́ṣaṇaḥ has a single udātta, on gní.
• The excepted words keep both accents. dyā́vāpṛthivī́ janáyan: dyā́ (nipātana) and vī́. sómārudráu: rudra ends in rak (uṇādi 179), so the udāttas are só and drá. índrāpūṣáṇau: pūṣan is antodātta by nipātana (uṇādi 157), so í and ṣá. śukrāmanthínau: thí.
• Why the sūtra says "uttarapada". Without it, the seventh-case anudāttādau would qualify the dvandva itself and would wrongly reach candrāsūryau. anudāttādau is what divides the domains of the injunction and the prohibition.
• The Subodhinī reads the pṛthivī exception as a jñāpaka for svaravidhau vyañjanam avidyamānavat.

VII.B. The antaḥ heading: an uttarapada in tha, atha, ghañ, kta, ac, ap, itra or ka after a gati, kāraka or upapada becomes antodātta, and so does a kta-form after su, after an upamāna, in a name, and in pravṛddha and the rest — running example: pra + bhid + ghañ → prabhedáḥ (udātta on the final dá, where the kṛt accent of SK 3873 would have kept ñit bhéda ādyudātta)

• 3877 antaḥ 6.2.143 — an adhikāra: "the final syllable [of the uttarapada is udātta]". The kaumudī says only that it is a heading. It ends the prakṛtyā section (SK 3871) and governs everything that follows. There is no Subodhinī on it.
• 3878 thāthaghañktājabitrakāṇām 6.2.144 — after a gati, kāraka or upapada (ity eva, from SK 3873), an uttarapada ending in tha, atha, ghañ, kta, ac, ap, itra, ka is antodātta. The examples and their udātta syllables:
• tha: prabhṛthásyāyóḥ (thá; kthan by uṇādi 159–160, which would have been ādyudātta as nit).
• atha: āvasatháḥ (thá; uṇādi 396).
• ghañ: prabhedáḥ (dá).
• kta: dhartā́ vajrī́ puruṣṭutáḥ (tá). It is analysed with the seventh case, puruṣu stutaḥ, so that tṛtīyā karmaṇi (SK 3782) cannot come in.
• ac: prakṣayáḥ (yá; erac, SK 3231). It is not already antodātta, because kṣaya "dwelling" is ādyudātta by SK 3689.
• ap: pralaváḥ (vá; SK 3232).
• itra: pralavitrám (trá).
• ka: govṛṣáḥ (ṣá; mūlavibhujādi; it would otherwise be ādyudātta by SK 3691).
• Not after a gati: sustutaṃ bhavatā. Here su is a karmapravacanīya (SK 555), so the rule does not apply.
• 3879 sūpamānāt ktaḥ 6.2.145 — after su or an upamāna, a kta-final uttarapada is antodātta. ṛtásya yónau sukṛtásya: in sukṛtá, tá is udātta, overriding gatir anantaraḥ (SK 3783). śaśaplutáḥ, "leapt like a hare": tá, overriding SK 3782.
• 3880 saṃjñāyām anācitādīnām 6.2.146 — in a name, a kta-final uttarapada after a gati, kāraka or upapada is antodātta, except ācita and the rest. upahūtáḥ śākalyaḥ, parijagdháḥ kauṇḍinyaḥ: udātta on tá and dhá, overriding SK 3783. Excepted: ācitam, āsthāpitam.
• 3881 pravṛddhādīnāṃ ca 6.2.147 — the kta-final uttarapada of pravṛddha and the rest is antodātta: pravṛddháḥ (ddhá), prayutáḥ (tá). The rule is for words that are not names (asaṃjñārtha), since in a name 3880 already gives the result. The list is an ākṛtigaṇa.

VII.C. Antodātta after a kāraka or after a case-form: datta and śruta in a blessing, itthambhūta-kṛta, ana, man, ktin and their like, puṇya, ūna and kalaha, miśra — running example: deva + datta (blessing name) → devadattáḥ (udātta ttá); māṣa + ūna → māṣonám (udātta nám)

• 3882 kārakād dattaśrutayor evāśiṣi 6.2.148 — a niyama on 3880. In a tatpuruṣa, in a name, in a blessing, only datta and śruta (kta-forms) after a kāraka are antodātta. devadattáḥ, viṣṇuśrutáḥ: udātta on the final syllable.
• "after a kāraka": saṃbhūtó rāmāyaṇaḥ. sam is a gati, so this escapes the niyama and is antodātta by 3880 (the translator's inference, supported by the tritayādhikāra).
• "datta and śruta": devapālitaḥ takes tṛtīyā karmaṇi (SK 3782), not 3880.
• "eva": it restricts the kāraka, not datta and śruta. So saṃśrutáḥ, with no kāraka, is still antodātta.
• "in a blessing": devakhātā takes SK 3782, and so does the conch-name devadattaḥ in anāhato nadati devadattaḥ.
• 3883 itthambhūtena kṛtam iti ca 6.2.149 — in the sense "done by one in such a state", the kta-final member is antodātta. It is an apavāda to SK 3782. suptapralapitám, pramattagītám: udātta on tám. Here kṛ means action in general, not "bringing into being", so babbling also counts as "done".
• 3884 ano bhāvakarmavacanaḥ 6.2.150 — after a kāraka (from 3882), a member in ana (lyuṭ) that denotes the action or the object is antodātta. It is an apavāda to SK 3873.
• payaḥpānáṃ sukham (nám): bhāva lyuṭ, an upapada-samāsa by SK 3291.
• rājabhojanā́ḥ śālayaḥ (nā́): karman lyuṭ.
• Counter-examples: hastādāyaḥ (not ana); dantadhāvanam (instrument lyuṭ); nidarśanam (not after a kāraka).
• 3885 manktinvyākhyānaśayanāsanasthānayājakādikrītāḥ 6.2.151 — after a kāraka, in a tatpuruṣa, these members are antodātta: man, ktin, vyākhyāna, śayana, āsana, sthāna, yājaka and the rest, and krīta. It is an apavāda to the kṛt accent; for krīta, the Subodhinī says it overrides SK 3782 instead.
• The examples, each with its final syllable udātta: rathavartmá, pāṇinikṛtíḥ, chandovyākhyānám, rājaśayanám, rājāsanám, aśvasthānám, brāhmaṇayājakáḥ (yājakādi = the list of SK 703), gokrītáḥ.
• Not after a kāraka: prabhūtau saṅgatim, which takes SK 3784.
• 3886 saptamyāḥ puṇyam 6.2.152 — puṇya after a seventh-case word is antodātta. adhyayanapuṇyám (yám), compounded by the yogavibhāga saptamī (SK 717). Without this rule, tatpuruṣe tulyārtha- (SK 3736) would apply. Counter-example: vedapuṇyam (third case).
• 3887 ūnārthakalahaṃ tṛtīyāyāḥ 6.2.153 — after a third-case word, members meaning "deficient", and kalaha, are antodātta. This is an exception to the tṛtīyā-pūrvapada accent. māṣonám, māṣavikalám, vākkalaháḥ (há). The compounds are by SK 693.
• "Some" (kecit) read artha as the word itself: dhānyārtháḥ. On their view, ūna stands for its meaning.
• tṛtīyāyāḥ is said for clarity.
• 3888 miśraṃ cānupasargam asandhau 6.2.154 — after a third-case word, miśra without an upasarga is antodātta, except where sandhi (an alliance by pact) is meant. tilamiśrā́ḥ, sarpirmiśrā́ḥ (śrā́).
• Counter-examples: guḍadhānāḥ; tilasaṃmiśrāḥ; brāhmaṇamiśro rājā.
• The exclusion anupasargam is a jñāpaka: where SK 693 mentions miśra, miśra with an upasarga is also taken.

VII.D. After nañ in a tatpuruṣa: the uttarapada becomes antodātta, in place of nañ's own á- — running example: na kārṇaveṣṭakikam → akārṇaveṣṭakikám (udātta kám); na pacaḥ (unable) → apacáḥ (udātta cáḥ)

• 3889 naño guṇapratiṣedhe sampādyarhahitālamarthās taddhitāḥ 6.2.155 — after nañ that negates a quality, members in taddhitas of the senses sampādin, arha, hita, alam are antodātta. The quality is the taddhita's pravṛttinimitta, here sampāditva and the like.
• akārṇaveṣṭakikám (ṭhañ), "a face not suited to earrings"; acchaidikáḥ (ṭhak); avatsīyáḥ (cha); asāntāpikáḥ (ṭhañ).
• Counter-examples: dvigārdabharathikaḥ (no nañ; queried for vi-); agārdabharathikaḥ ("other than", not negation); apāṇinīyaḥ (not sampādin etc.); avoḍhā (tṛc, not a taddhita).
• 3890 yayatoś cātadarthe 6.2.156 — after such a nañ, members in the taddhitas ya, yat, when not "for that", are antodātta. apāśyā́ (yā́), adantyám (yám).
• Counter-examples: apādyam (tadartha); adeyam (kṛt yat); adantyam "other than dental"; avāmadevyam (ḍya, excluded by tadanubandhakagrahaṇe nātadanubandhakasya).
• 3891 ackāv aśaktau 6.2.157 — after nañ, a member in ac or ka is antodātta when inability is meant, guṇapratiṣedhe still holding. apacáḥ "unable to cook", avilikháḥ: udātta on the final.
• Counter-examples: apaco dīkṣitaḥ (barred by śāstra, not unable); anyo 'yaṃ pacād apacaḥ ("other than a cook").
• 3892 ākrośe ca 6.2.158 — the same members are also antodātta in reviling: apacó jālmaḥ (có), avikṣipáḥ (páḥ).
• 3893 saṃjñāyām 6.2.159 — in a name, with abuse meant, the uttarapada after nañ is antodātta: adevadattáḥ (ttá), "a non-Devadatta".
• 3894 kṛtyokeṣṇuccārvādayaś ca 6.2.160 — after nañ, members in a kṛtya, ukañ or iṣṇuc, and the cārvādi words, are antodātta.
• Examples: akartavyáḥ; anāgāmukáḥ (SK 3134); analaṅkariṣṇúḥ (SK 3116); acārúḥ, asādhúḥ.
• khiṣṇuc is included under iṣṇuc, despite its anubandhas kh and c, because it is taught with initial i: anāḍhyambhaviṣṇúḥ (SK 2974).
• Gaṇasūtra rājāhnoś chandasi: arājā́, anāháḥ in the Veda only. In bhāṣā, nañ keeps its own accent (árājā, the translator's reading).
• 3895 vibhāṣā tṛnnannatīkṣṇaśuciṣu 6.2.161 — after nañ, optionally: tṛn-forms, anna, tīkṣṇa, śuci. akartā́, anannám, atīkṣṇám, aśucí: udātta on the final syllable. In the other option, the avyaya nañ keeps its accent (á-, SK 3736).

VII.E. The bahuvrīhi heading opens: prathama or an ordinal after idam, etad, tad; stana after a numeral; mitra and ajina in names; antara "screen" — running example: idaṃ prathamaṃ yasya → idaṃprathamáḥ (udātta má), and with kap idaṃprathamákaḥ (still má)

• 3896 bahuvrīhāv idametattadbhyaḥ prathamapūraṇayoḥ kriyāgaṇane 6.2.162 — in a bahuvrīhi, after idam, etad, tad, prathama and an ordinal are antodātta when acts are counted. idaṃprathamáḥ, etaddvitīyáḥ, tatpañcamáḥ: udātta on the final syllable.
• Counter-examples.
• anena prathamaḥ: a tatpuruṣa by yogavibhāga of SK 692.
• yatprathamaḥ.
• tadbahuḥ: not prathama or an ordinal.
• idaṃprathamāḥ "of which this one is first, the leading one": a counting of things, not acts.
• idaṃprathamāḥ = idaṃpradhānāḥ: prathama here means "chief", so nothing is counted.
• With kap. The uttarapada is the operand (kāryin), so the udātta stays on prathama's last syllable before kap: idaṃprathamákaḥ.
• The adhikāra. bahuvrīhau runs to just before SK 3912.
• 3897 saṃkhyāyāḥ stanaḥ 6.2.163 — stana after a numeral is antodātta in a bahuvrīhi. dvistanī́ (nī́): the udātta of the elided a passes to ṅīṣ by SK 3651. catuḥstanā́ (nā́).
• Counter-examples: darśanīyastanā (ritsvara of the first member); dviśirāḥ.
• The Subodhinī disagrees about catuḥstanā (printed under 3898): catur keeps its own accent, giving cátuḥstanā, so the form takes no ṅīṣ. The English follows Bhaṭṭoji.
• 3898 vibhāṣā chandasi 6.2.164 — in the Veda, 3897 is optional: dvistanāṃ karoti. The kaumudī does not say which option this shows, so it is left unmarked.
• 3899 saṃjñāyāṃ mitrājinayoḥ 6.2.165 — in a name, mitra and ajina are antodātta. devamitráḥ (trá), kṛṣṇājinám (nám). Counter-example: priyamitraḥ.
• Vārttika ṛṣipratiṣedho mitre: this accent is barred when a seer is named. In viśvāmitra ṛṣiḥ, bahuvrīhau viśvaṃ saṃjñāyām (SK 3840) takes effect instead, giving viśvā́mitraḥ (vā́), with the long ā by SK 1049 (Subodhinī).
• 3900 vyavāyino 'ntaram 6.2.166 — antara after a word for what intervenes (a screen) is antodātta. vastrāntaráḥ (ráḥ), "he for whom a cloth is the screen". Counter-example: ātmāntaraḥ, "of another nature".

VII.F. mukha and kta-forms in the bahuvrīhi — running example: gaura + mukha → gauramukháḥ (udātta kháḥ), against go + mukha → gómukhaḥ (first member's own udātta gó, mukha unaccented)

• 3901 mukhaṃ svāṅgam 6.2.167 — mukha denoting a body part (svāṅga, in its technical definition) is antodātta: gauramukháḥ. Counter-example: dīrghamukhā śālā, where mukha means the front of a hall.
• 3902 nāvyayadikśabdagomahatsthūlamuṣṭipṛthuvatsebhyaḥ 6.2.168 — not after an avyaya, a direction word, or go, mahat, sthūla, muṣṭi, pṛthu, vatsa. Here the first member keeps its own accent (SK 3735). Each first member is antodātta, as the Subodhinī says.
• The forms, with the udātta on the first member: uccaírmukhaḥ (caí), prā́ṅmukhaḥ (prā́ṅ, pra udātta by SK 3786 plus the ekādeśa), gómukhaḥ, mahā́mukhaḥ, sthūlámukhaḥ, muṣṭímukhaḥ, pṛthúmukhaḥ, vatsámukhaḥ.
• For go, muṣṭi and vatsa as upamāna, this rule also sets aside the option of 3903, by pūrvavipratiṣedha.
• 3903 niṣṭhopamānād anyatarasyām 6.2.169 — after a niṣṭhā-form or an upamāna, mukha is optionally antodātta. prakṣālitamukháḥ (kháḥ); siṃhamukháḥ.
• The other forms of prakṣālitamukhaḥ: prakṣālitá- by SK 3844, or the gati accent on prá- kept through pūrvapadaprakṛtisvara.
• 3904 jātikālasukhādibhyo 'nācchādanāt kto 'kṛtamitapratipannāḥ 6.2.170 — a kta-member after a word for a genus, a time, or sukha and the rest is antodātta, but not after a word for a covering, and not kṛta, mita or pratipanna. sāraṅgajagdháḥ (dháḥ), māsajātáḥ, sukhajātáḥ, duḥkhajātáḥ (táḥ).
• Counter-examples: putrajātaḥ; vastracchannaḥ; kuṇḍakṛtaḥ, kuṇḍamitaḥ, kuṇḍapratipannaḥ. kuṇḍa is ādyudātta by phiṭ 26.
• Jñāpaka. In such compounds the niṣṭhā-form stands second (paranipāta), against SK 899.
• 3905 vā jāte 6.2.171 — jāta after these words is optionally antodātta: dantajātáḥ, māsajātáḥ. In the other option the first member keeps its accent: dántajātaḥ (phiṭ 29), mā́sajātaḥ (phiṭ 38).

VII.G. After nañ and su in the bahuvrīhi, and the udātta before kap; bahu as nañ; guṇa and the rest — running example: na vrīhiḥ yasya → avrīhíḥ (udātta hí); with kap avrī́hikaḥ (udātta vrī́, the syllable before the last of vrīhi)

• 3906 nañsubhyām 6.2.172 — after nañ and su, the uttarapada is antodātta: avrīhíḥ (hí), sumāṣáḥ (ṣáḥ).
• The Subodhinī: if samāsāntas belong to the samāsa, the samāsa is the operand, with 3907 as the jñāpaka. If they belong to the uttarapada, the a of anṛcaḥ (SK 940) is also udātta.
• 3907 kapi pūrvam 6.2.173 — when kap follows, the syllable before kap is udātta. abrahmabandhū́kaḥ (dhū́), sukumārī́kaḥ (rī́); kap is by SK 833. The examples end in long vowels only, because 3908 takes the short ones.
• 3908 hrasvānte 'ntyāt pūrvam 6.2.174 — if the member (reckoned before kap) ends in a short vowel, the syllable before the last is udātta. avrī́hikaḥ (vrī́), sumā́ṣakaḥ (mā́).
• The renewed pūrvam is a niyama: for a short-final member, only that syllable is udātta, not the one before kap. Hence ajñakáḥ takes 3906 and is antodātta on the kap-form.
• 3909 bahor nañvad uttarapadabhūmni 6.2.175 — after bahu, when it expresses the plurality of what the uttarapada denotes, the accent is as after nañ. bahuvrī́hikaḥ (vrī́), bahumítrakaḥ (mí), both by 3908. Counter-example: bahumānaḥ, where bahu is antodātta (uṇādi 29) and keeps its own accent.
• 3910 na guṇādayo 'vayavāḥ 6.2.176 — after bahu, guṇa and the rest, when they denote parts, are not antodātta: bahuguṇā rajjuḥ, bahvakṣaraṃ padam, bahvadhyāyaḥ. The list is an ākṛtigaṇa.
• Counter-example: bahuguṇo dvijaḥ, where the guṇas are qualities.
• Only the negation is stated, so no accent is marked.

VII.H. After an upasarga: a permanent body part (the last bahuvrīhi rule), then in any compound vana, antar-vana and anta; not after ni or vi; pari, pra, nir — running example: pra + pṛṣṭha → prapṛṣṭháḥ (udātta ṣṭháḥ); pari + anta → paryantáḥ (udātta táḥ), against ni + anta → nyàntaḥ (independent svarita yà)

• 3911 upasargāt svāṅgaṃ dhruvam aparśu 6.2.177 — in a bahuvrīhi, after an upasarga (that is, pra and the rest), a permanent body part other than parśu is antodātta: prapṛṣṭháḥ, pralalāṭáḥ.
• Counter-examples: darśanīyapṛṣṭhaḥ; praśākho vṛkṣaḥ (not a body part); udbāhuḥ (not permanent); viparśuḥ.
• 3912 vanaṃ samāse 6.2.178 — in any compound, vana after an upasarga is antodātta. samāse ends the bahuvrīhi heading. tásyédime pravaṇé: udātta on ṇé, with ṇ by SK 1050.
• 3913 antaḥ 6.2.179 — vana after antar (a non-upasarga use) is antodātta: antarvaṇáḥ deśaḥ (ṇáḥ). The fifth case of antar is dropped as a chāndasa luk.
• 3914 antaś ca 6.2.180 — in any compound, the word anta "end" after an upasarga is antodātta: paryantáḥ, samantáḥ (táḥ). There is no Subodhinī in print.
• 3915 na nivibhyām 6.2.181 — not after ni, vi. The first member keeps its own accent: SK 3735 in a bahuvrīhi, SK 3736 in a tatpuruṣa. ni and vi are ādyudātta by phiṭ 81. After yaṇ, the following anudātta a becomes independent svarita by SK 3657: nyàntaḥ, vyàntaḥ.
• 3916 parer abhitobhāvi maṇḍalam 6.2.182 — after pari, a word for what is on both sides (kūla and the like) and maṇḍala are antodātta: parikūlám, parimaṇḍalám (lám). This sets aside the first member's accent, SK 3767 and SK 3855.
• 3917 prād asvāṅgaṃ saṃjñāyām 6.2.183 — after pra, a non-body-part member is antodātta in a name: pragṛhám (hám). Counter-example: prapadam, which the Subodhinī gives as prápadam (pra ādyudātta by phiṭ 81).
• 3918 nirudakādīni ca 6.2.184 — nirudaka and the rest are antodātta: nirudakám, nirupalám (kám, lám). The Subodhinī analyses them as a bahuvrīhi or a prādi compound. As an avyayībhāva, SK 3734 would already give the accent.

VII.I. After abhi, apa, adhi, anu, ati and ni: mukha and the apa-list, uparistha, the non-principal, puruṣa, the non-kṛt member, and ni outside "concealment" — running example: apa + mukha → apamukhám (udātta khám); anu + jyeṣṭha (prior member principal) → anujyeṣṭháḥ (udātta ṣṭháḥ)

• 3919 abher mukham 6.2.185 — after abhi, mukha is antodātta: abhimukhám. SK 3911 would already give this. The rule is for non-bahuvrīhi compounds, for an impermanent mukha, and for a mukha that is not a body part: abhimukhā́ śālā (khā́). SK 3901 does not apply after an avyaya (SK 3902).
• 3920 apāc ca 6.2.186 — after apa, mukha is antodātta: apamukhám. This sets aside pūrvapadaprakṛtisvara and SK 3767. The yogavibhāga is there so that only apāt carries into 3921.
• 3921 sphigapūtavīṇāñjo'dhvakukṣisīranāmanāma ca 6.2.187 — after apa, sphiga, pūta, vīṇā, añjas, adhvan, kukṣi, words for the plough, and nāman are antodātta.
• The examples, each with its final syllable udātta: apasphigám, apapūtám, apavīṇám, apāñjáḥ, apādhvā́, apakukṣíḥ, apasīrám, apahalám, apanāmá.
• adhvan is named for the case where upasargād adhvanaḥ (SK 953) does not apply. This is the jñāpaka that samāsāntas are anitya; with ac, citaḥ (SK 3710) would already give the result.
• sphiga, pūta and kukṣi are named for non-bahuvrīhi, impermanent and non-body-part uses (against SK 3911).
• 3922 adher uparistham 6.2.188 — after adhi, a word for what stands above is antodātta: adhidantáḥ (táḥ), "a tooth grown over a tooth". Counter-example: adhikaraṇam, which the Subodhinī gives as adhikáraṇam (SK 3873).
• 3923 anor apradhānakanīyasī 6.2.189 — after anu, a non-principal member, and kanīyas, are antodātta.
• anujyeṣṭháḥ ("following the eldest"; a prādi compound with the prior member principal).
• anukanīyā́n (yā́n). kanīyas is named for its principal use.
• Counter-example: anujyeṣṭhaḥ, "the eldest, who follows".
• 3924 puruṣaś cānvādiṣṭaḥ 6.2.190 — after anu, puruṣa meaning "mentioned afterwards" is antodātta: anupuruṣáḥ (ṣáḥ). Counter-example: anupuruṣaḥ, "the man who follows".
• 3925 ater akṛtpade 6.2.191 — after ati, a non-kṛdanta member, and pada, are antodātta: atyaṅkuśáḥ nāgaḥ (śáḥ), atipadā́ gāyatrī (dā́). Counter-example: atikāraḥ (print; WS atikārakaḥ).
• Vārttika ater dhātulope: the condition should be that a verbal root is elided, not "non-kṛt". Then the rule does not apply in atigārgyaḥ "an excellent Gārgya", and it does apply in atikārukáḥ (káḥ), "having surpassed the artisan". This cures both avyāpti and ativyāpti (Subodhinī).
• 3926 ner anidhāne 6.2.192 — after ni, the uttarapada is antodātta when the sense is not concealment (nidhāna = aprakāśatā): nimūlám, nyakṣám (lám, kṣám). Counter-example: nidaṇḍaḥ, "a stick put away". The Subodhinī's gloss of nimūlam, ending "nirmūlam", is a print oddity and is not followed.

Part VIII — SK 3927–3978: the last compound accents (antaḥ after prati, upa, su and ud; dvi/tri and saktha in the bahuvrīhi; parādi in the Veda); then the tiṅanta accents, from the nighāta of a verb after a non-verb through the particles and conditions that cancel it or restore it, to the gati before a gati or an accented verb; and Bhaṭṭoji's closing svara-saṃcāra on agním īḷe

Covers SK 3927–3978 (52 sūtras): all present.

• The Aṣṭādhyāyī spine. The run has two blocks, each in AS order.
• 6.2.193–199 (3927–3933) closes 6.2 and with it the compound accents. Bhaṭṭoji ends 3933 with "[here end] the accents of compounds (samāsasvarāḥ)".
• 8.1.27–71 (3934–3978) is the only stretch of 8.1 that the SK takes here. The rest of 8.1 is elsewhere: 8.1.1 and 8.1.4–15 with doubling (SK 2139–2150); 8.1.2 at SK 83; 8.1.3 at 3670; 8.1.6 at 3599; 8.1.16–18, 8.1.20–26 and 8.1.72–73 with the yuṣmad/asmad substitutes (SK 401–413); 8.1.19 at 3654; 8.1.74 at 3655.
• Non-sūtra material.
• Vārttikas: the śloka parādiś ca parāntaś ca pūrvāntaś cāpi dṛśyate … (3933); malopaś ca vaktavyaḥ (3974); gatigrahaṇe upasargagrahaṇam iṣyate (3975); kriyākutsana iti vācyam, pūti as cit, and vā bahvartham (3976).
• The svara-saṃcāra. Inside 3978 comes the "svarasaṃcāraprakāraḥ" section, the worked accent analysis of the first ṛc, and then the SK's closing verse.
• AS-refs. All 52 were checked against references/ashtadhyayi/sutraani-data.json by SK number and by sūtra text, and all match.
• Cross-references. The ones cited below also match: SK 20 (1.4.57), 78 (6.1.94), 95 (8.2.84), 194 (6.1.107), 402 (8.1.17), 403 (8.1.18), 408 (8.1.24), 418 (6.3.92), 555 (1.4.94), 626 (2.3.68), 707 (2.2.13), 761 (2.2.18), 762 (1.4.61), 768 (1.4.67), 788 (5.4.91), 852 (5.4.113), 854 (5.4.115), 879 (5.4.140), 880 (5.4.141), 1975 (5.3.39), 2140 (8.1.4), 2163 (1.4.106), 2778 (3.2.118), 2780 (3.2.120), 2783 (3.3.4), 2789 (3.3.131), 2856 (3.1.108), 3076 (7.4.42), 3231 (3.3.56), 3268 (3.3.90), 3623 (8.2.104), 3651 (6.1.161), 3653 (6.1.198), 3654 (8.1.19), 3658 (8.2.5), 3659 (8.2.6), 3660 (8.4.66), 3668 (1.2.39), 3669 (1.2.40), 3671 (6.1.162), 3689 (6.1.201), 3708 (3.1.3), 3709 (3.1.4), 3710 (6.1.163), 3730 (6.1.186), 3734 (6.1.223), 3736 (6.2.2), 3783 (6.2.49), 3873 (6.2.139), 3878 (6.2.144).
• The headings, and how far they run.
• antaḥ (SK 3877) still supplies "the final syllable is udātta" for 3927–3932.
• tatpuruṣe, named at 3927, carries into 3928–3930. Bhaṭṭoji asks "tatpuruṣe kim?" at 3928 and 3930.
• vibhāṣā (3930) carries into 3931–3933 (Subodhinī at 3933). bahuvrīhau belongs to 3931 alone.
• At 3933, parādiḥ replaces antaḥ: there it is the first syllable of saktha (from 3932).
• tiṅantasvarāḥ (3934–3978). This heading is the print's "tiṅantasvarāḥ 73", which ours lacks; it is supplied in brackets at 3934.
• The frame. Throughout, padāt (SK 402) and anudāttaṃ sarvam apādādau (SK 403) are the adhikāras. tiṅ comes from 3935. na comes from 3936; the Subodhinī says "the prohibitions begin here". yuktam comes from 3937. These source identifications are the translator's, from Bhaṭṭoji's "tiṅantaṃ nānudāttam".
• Local carries:
• aprātilomye 3940 → 3941 (Subodhinī); hi 3941 → 3942.
• yāvadyathābhyām 3943 → 3944–3945; the pūjāyām of 3944 → 3945.
• A fresh pūjāyām at 3946 → 3947. aho 3947 → 3948, whose vibhāṣā → 3949.
• All of 3951 → 3952. apūrvam 3954 → 3955–3956. āho utāho 3956 → 3957.
• All of 3958 except lṛṭ → 3959; 3959 → 3960 → 3961.
• agateḥ 3964 → 3965 only, drawn in by "ca" and not continuing into 3966 (Subodhinī).
• prathamā 3966 → 3967–3970 and 3972. vibhāṣā 3970 → 3971. chandasi 3971 → 3972.
• The terms of 3974 → 3975 → 3976. gatiḥ 3977 → 3978.
• Where padāt stops. The Subodhinī says padāt has ceased at 3976.
• Three kinds of rule. 3936–3973 prohibit the nighāta of 3935: the verb keeps its own udātta. Three exceptions inside that run:
• 3944–3945 cancel the prohibition of 3943, so the verb is anudātta after all.
• 3953 and 3963 are niyamas.
• 3948, 3949, 3952, 3957, 3960–3961, 3970 and 3971 are options.
• 3974–3978 impose anudātta.
• Print readings followed (NSP 1915 pp. 656–664, scans 0666–0674; collation batches 23–24; audits H29–H32).
• Ours corrected to the print:
• 3930 utpucchaḥ; 3931 pareṣv eṣv, with the svarita on ra of saptáraśmim; 3939 taṃ vayam; 3945 prapacati.
• 3946 svadhām, with the bar under ri (eriré); 3955 ciduttaram; 3958 bhokṣyante.
• 3959: the dropped sentence "āgaccha deva grāmaṃ paśyatv evaṃ rāmaḥ. sarvagrahaṇāt tv iha syād eva", supplied in brackets.
• 3961 "(SK 3937)" (ours 3967); 3966 sātim; 3968 upādhyāyam; 3969 four slips; 3972 svādv atti; 3973 yadadryaṅ; 3977 haribhir yāhi.
• 3978 phiṣaḥ, antodātta, "tiṅṅ atiṅaḥ 3935", samāsāntodātte, anudāttatve. Also supplied in 3978: the heading "svarasaṃcāraprakāraḥ 74", the verse number and the colophon.
• Queried for the Print Check:
• 3932: ṣacaś against pacaś (C), and akrāntāt against ākrāntāt (B).
• 3941: ruhaṃtam read as ruhatam (C).
• 3948: the transmitted pūjāyām, where the Subodhinī's gloss presupposes apūjāyām.
• 3950: ākarṣīḥ read as akārṣīḥ (V).
• 3959: paśyāvaḥ read as paśyāva.
• 3965: pareṣu, quoted by the Subodhinī but absent from the mūla.
• 3966: nighātārtham read as a-nighāta (avagraha).
• Presentation: 3933, the vārttika śloka set as one in the print (M); 3976, kriyākutsana iti vācyam set as a vārttika (M).
• Where the commentary lies (only the kaumudī is translated, SK-TRANSLATE §25).
• BM. There is no BM anywhere in this part. Our layer is empty, and Wikisource uttara-4/0120–0124 prints only the mūla.
• Subodhinī filed under a neighbour:
• 3927: the tail of sk.3926.s1. 3932, 3933, 3934 and 3935: all in sk.3931.s1, with the colophon "iti svarasubodhinyāṃ samāsasvaraprakaraṇam" after 3933.
• 3937: the tail of sk.3936.s1. 3946: the tail of sk.3945.s1. 3949: the tail of sk.3948.s1, plus its own s1. 3953: the tail of sk.3952.s1, plus its own s1.
• 3960: the tail of sk.3959.s1. 3962 and the first part of 3963: in sk.3961.s1. The rest of 3963 and the head of 3964: in sk.3963.s1.
• 3966 and 3967: in sk.3965.s1. 3970: the tail of sk.3969.s1, plus its own s1. 3974: the tail of sk.3973.s1, plus its own s1. 3977: the tail of sk.3976.s1.
• No Subodhinī in print on 3928 or 3956. For 3956 the only remark is the jñāpaka in sk.3954.s1.
• Lines missing from our s-blocks: the last line of p. 660, under sk.3959.s1 (on 3960); the last line of p. 663 and the closing colophon, under sk.3978.s1.
• Audit corrections followed (H29–H32, now in the en.yml).
• MUST-FIX:
• 3949: the two options are vídyotate / vidyótate, not a wholly anudātta vidyotate.
• 3954: the counter-example is kaṭáṃ jā́tu kariṣyasi, where jātu keeps its own udātta.
• 3963: in yád aṅgiraḥ, the first syllable of aṅgiraḥ is the svarita.
• MINOR, accent: the protected verb is now marked and named in 3942, 3947, 3950, 3951, 3953, 3959, 3960, 3961, 3962, 3963, 3964 and 3965.
• MINOR, wording: 3931's gloss uses nominatives; 3934–3935 now carry apādādau and present SK 403 as an adhikāra; 3948 adds the Subodhinī's "[other than praise]" in sutra_en; 3959–3960 use "not wholly different".
• Accent convention. udātta is acute and the independent svarita grave; anudātta, dependent svarita and pracaya are unmarked. In the tiṅanta rules only the protected verb, and whatever the commentary states, is marked. An unmarked neighbouring word means "not discussed", not "anudātta" (translator notes, accepted by the audits).
• Where Part VIII ends. The part ends at 3978 tiṅi codāttavati (8.1.71), the svara-saṃcāra and the close of the vaidika-svara-prakriyā. That is also the end of p71.

VIII.A. The last antaḥ rules in the tatpuruṣa: after prati, upa and su, and the optional utpuccha — running example: prati + aṃśu (pratigato 'ṃśuḥ) → pratyaṃśúḥ (udātta on the final śú)

• 3927 prater aṃśvādayas tatpuruṣe 6.2.193 — after prati, in a tatpuruṣa, aṃśu and the rest are antodātta.
• Examples: pratyaṃśúḥ (śú), pratijanáḥ (ná), pratirājā́ (jā́, the final of pratirājan).
• No ṭac in pratirājā, because samāsāntas are non-obligatory. The Subodhinī reasons from the list itself. ṭac by SK 788 is cit, so it would already make the compound antodātta, and listing rājan would be pointless. So the listing is the jñāpaka.
• 3928 upād dvyajajinam agaurādayaḥ 6.2.194 — after upa, in a tatpuruṣa, an uttarapada of two vowels, and ajina, are antodātta, except gaura and the rest.
• Examples: upadeváḥ (vá), upendráḥ (drá), upājinám (nám).
• Counter-examples: upagauraḥ and upataiṣaḥ (gaurādi); upasomaḥ (a bahuvrīhi, "one whom soma has reached").
• There is no Subodhinī on this sūtra.
• 3929 sor avakṣepaṇe 6.2.195 — after su, when censure is meant, the uttarapada is antodātta: supratyavasitáḥ (tá).
• su itself means praise (it is a karmapravacanīya, SK 555). The sentence is censure because it is said out of spite: to someone sitting at ease in a misfortune.
• The thāthādi rule (SK 3878) is not in question here (Subodhinī). The reason, that su is not a gati, is the translator's.
• Counter-examples: kubrāhmaṇaḥ (not su); suvṛṣaṇam (no censure).
• 3930 vibhāṣotpucche 6.2.196 — in the tatpuruṣa utpuccha ("gone out from the tail"), puccha after ud is optionally antodātta: utpuccháḥ (chá).
• Why the option holds both ways. Take the sense "one who raises the tail": that is utpucchayati + ac by erac (SK 3231), and there the thāthādi accent (SK 3878) would make the final udātta obligatory, so this rule makes it optional. In the first derivation, the avyaya prior member's accent (SK 3736) would have blocked the final udātta altogether (Subodhinī).
• Counter-example: the bahuvrīhi utpucchaḥ, "one by whom the tail is raised". The print reads utpucchaḥ.

VIII.B. dvi and tri in the bahuvrīhi, saktha, and the Vedic parādi; the compound accents close — running example: dvi + pād (pāda, a lost, SK 879) → dvipā́t (udātta on the final pā́): dvipā́c cátuṣpāc caráthāya

• 3931 dvitribhyāṃ pāddanmūrdhasu bahuvrīhau 6.2.197 — in a bahuvrīhi, after dvi and tri, before pād, dat or mūrdhan, the compound is optionally antodātta. vibhāṣā comes from 3930.
• Examples: dvipā́t; tripā́d ūrdhváḥ (pā́); dvidán (dán; dat by SK 880); trimūrdhā́naṃ saptáraśmim (dhā́).
• mūrdhan without its samāsānta. The sūtra says mūrdhan, not mūrdha, so the final udātta comes even without a samāsānta. This is the jñāpaka that samāsāntas are non-obligatory.
• With the samāsānta. When ṣa is added (SK 854), the rule still operates on the bahuvrīhi, of which ṣa is a part, so the final udātta remains possible: dvimūrdháḥ, trimūrdháḥ (dháḥ).
• Counter-examples: kalyāṇamūrdhā (not dvi or tri; kalyāṇa keeps its own accent, by phiṭ 42 per the Subodhinī); dvimūrdhā (a tatpuruṣa, "the head of two").
• 3932 sakthaṃ cākrāntāt 6.2.198 — saktha (the stem with ṣac, SK 852) is optionally antodātta, provided it does not follow a word ending in kra.
• Examples: gaurasaktháḥ, ślakṣṇasaktháḥ (tháḥ). In the other option, gaura keeps its own accent, with first and last syllables udātta (phiṭ 13), so the word has three accents (Subodhinī).
• Counter-example: cakrasaktháḥ. Here the final is udātta obligatorily, because ṣac is cit (SK 3710). So this rule is a prāptavibhāṣā on the cit accent.
• Print queries: ṣacaś (C); the pratīka akrāntāt (B).
• 3933 parādiś chandasi bahulam 6.2.199 — in the Veda, the first syllable of the following member (saktha, from 3932) is variously udātta: ajisáktham (sák).
• The vārttika śloka. parādi, parānta and pūrvānta are seen, and so is pūrvādi. So there is interchange (vyatyaya), and the sūtra says bahulam. Bhaṭṭoji's examples:
• parādi: tuvijātā́ urukṣáyā (kṣá, the first syllable of kṣaya).
• parānta: ní yéna muṣṭihatyáyā (tyá); yás tricakráḥ (krá).
• pūrvānta: viśvā́yuḥ (śvā́, the final of viśva merged into ā).
• The Subodhinī's doubts. It argues that urukṣaya is really a bāhulaka case. Its true parādi example is codayitrī sūnṛtānām.
• This ends the samāsasvarāḥ.

VIII.C. The tiṅanta nighāta: gotrādi after a verb, the verb after a non-verb, and the first exception, luṭ — running example: agní + am → agním (udātta gní) + īḷe → agním īḷe (īḷe after a non-tiṅ pada: wholly anudātta)

• 3934 tiṅo gotrādīni kutsanābhīkṣṇyayoḥ 8.1.27 — the print heads this sūtra "tiṅantasvarāḥ". After a pada ending in tiṅ, gotra and the rest are wholly anudātta in contempt and repetition. This runs under padāt (SK 402) and anudāttaṃ sarvam apādādau (SK 403).
• Examples: pacati gotram ("he torments his family", contempt); pacatipacati gotram (repetition; the doubling is by SK 2140). Likewise pravacana, prahasana, prakathana, pratyāyana.
• The two senses qualify the list itself (pāṭhaviśeṣaṇa). So wherever gotrādi is named later, as in 3964, only these senses count. The Subodhinī gets this by splitting the rule: the second half acts as a paribhāṣā.
• Counter-examples: pacati pāpam (pāpa is not in the list); khanati gotraṃ sametya kūpam (neither sense).
• 3935 tiṅṅ atiṅaḥ 8.1.28 — a tiṅanta after a pada not ending in tiṅ is wholly anudātta (nighāta, "nihanyate").
• agním īḷe: the udātta is on gní; īḷe is nighāta, and its ī is only a dependent svarita.
• This is the rule that 3936–3973 suspend.
• The Bhāṣya rejects the word atiṅ (Subodhinī).
• 3936 na luṭ 8.1.29 — the prohibitions begin. A luṭ form is not nighāta: śvaḥ kartā́, śvaḥ kartā́rau (udātta tā́ in both).
• The dual. In kartārau the ending after tās is anudātta, so tās keeps the udātta.
• The singular. In kartā, ṭi is lost before ḍā, and ḍā becomes udātta by the accent that follows the loss of an udātta. Identifying that accent with SK 3651 is the translator's.

VIII.D. Particles that keep the verb accented: yad…yatra, naha, satya, aṅga, hi, yāvat/yathā; and the two cancellations in praise — running example: yád agne syā́m aháṃ tvám (syā́m keeps its udātta syā́ after the particle yad; agne, an āmantrita after a pada, anudātta)

• 3937 nipātair yadyadihantakuvinnecceccaṇkaccidyatrayuktam 8.1.30 — a tiṅanta connected with the particles yad, yadi, hanta, kuvit, net, cet, caṇ, kaccit, yatra is not anudātta.
• Examples: yád agne syā́m; yúvā yádī kṛtháḥ (thá); kuvíd aṅgá ā́san (ā́); ácittibhiś cakṛmā́ kác cit (mā́); putrā́so yátra pitáro bhávanti (bhá).
• Why "nipātaiḥ". The word is the jñāpaka that neither pratipadokta nor sāhacarya decides anything in this section. That is how the derived yāvat (with vatup) is also taken at 3943 (Subodhinī).
• yukta means a direct connection.
• 3938 naha pratyārambhe 8.1.31 — with naha, in a pratyārambha (a beginning joined with a prohibition): naha bhokṣyáse (kṣyá, the suffix sya; "you shall not eat at all!").
• Counter-example: naha vai tásmiṃl loké dákṣiṇam icchanti. Here icchanti is nighāta by 3935.
• 3939 satyaṃ praśne 8.1.32 — with satya, in a question: satyaṃ bhokṣyáse? (kṣyá).
• Counter-example: satyám íd vā́ u táṃ vayám índraṃ stavāma. stavāma is nighāta. Print: taṃ vayam.
• 3940 aṅgāprātilomye 8.1.33 — with aṅga, when there is no contrariness: aṅga kurú (rú, the vikaraṇa u; a permission).
• Counter-example: aṅga kūjasi vṛṣala, idānīṃ jñāsyasi jālma. This is said in indignation, and kūjasi is nighāta.
• 3941 hi ca 8.1.34 — with hi, again with no contrariness (Subodhinī): ā́ hí ṣmā yā́ti (yā́); ā́ hí ruhátam (há). Print query: ruhaṃtam (C).
• 3942 chandasy anekam api sākāṅkṣam 8.1.35 — in the Veda, with hi, even more than one tiṅanta that are in mutual expectancy are kept accented.
• anṛtaṃ hi matto vádati, pāpmā cainaṃ yunā́ti: vá is the root accent and nā́ is that of śnā. Here hi = "since".
• By api, a single tiṅanta is also covered (Subodhinī).
• 3943 yāvadyathābhyām 8.1.36 — with yāvat or yathā: yáthā cit káṇvam ā́vatam (ā́).
• The ending is a third case, so the rule also applies when the particle follows the verb: pacati yāvat (Subodhinī).
• 3944 pūjāyāṃ nānantaram 8.1.37 — in praise, a tiṅanta immediately after yāvat or yathā is anudātta: the na cancels 3943. yāvat pacati śobhanam, yathā pacati śobhanam ("how well he cooks!"): pacati is wholly anudātta.
• Counter-examples: yāvad bhuṅkte (no praise); yāvad devadattaḥ pacati śobhanam (not immediately after). In both, 3943 keeps the verb's accent.
• 3945 upasargavyapetaṃ ca 8.1.38 — the same holds when an upasarga intervenes: yāvat prápacati śobhanam. pra keeps its own udātta, prá, and pacati is anudātta.
• "Immediately" still holds for the verb with its upasarga: yāvad devadattaḥ prapacati śobhanam. Print: prapacati.

VIII.E. Praise, haste and permission: tu/paśya/paśyata/aha, aho (fixed, then optional), purā, nanu — running example: aho devadattaḥ pácati śobhanam (with aho, in praise: pacati keeps its root udātta pá)

• 3946 tupaśyapaśyatāhaiḥ pūjāyām 8.1.39 — a tiṅanta connected with tu, paśya, paśyata or aha is not anudātta, in praise.
• ā́d áha svadhā́m ánu púnar garbhatvám eriré: eriré keeps its final udātta, ré.
• Why pūjā is stated again (Subodhinī). If the pūjā of 3944 carried on, it would bring its cancelling force into the domain of na luṭ (3936) and the following rules. Stated afresh, it is a plain prohibition of nighāta, and it is not limited to "immediately after".
• 3947 aho ca 8.1.40 — with aho, in praise: aho devadattaḥ pácati śobhanam (pá).
• It is a separate rule so that the option of 3948 holds for aho alone, not for tu and the rest.
• 3948 śeṣe vibhāṣā 8.1.41 — in the remaining case, a verb connected with aho is optionally anudātta.
• The Subodhinī glosses the remaining case as "other than praise".
• aho kaṭaṃ kariṣyati: either wholly anudātta, or kariṣyáti (ṣyá, the suffix accent).
• Print Check. The kaumudī transmits pūjāyām. The Subodhinī reads the example as envy, not praise, which presupposes apūjāyām.
• 3949 purā ca parīpsāyām 8.1.42 — with purā, when haste is meant, optionally.
• adhīṣva māṇavaka purā vidyotate vidyut ("study, boy: the lightning will soon flash"). The laṭ is future in sense by SK 2783.
• The two options (audit MUST-FIX):
• vídyotate: dyotate is wholly anudātta, and vi keeps its own udātta (phiṭ 81).
• vidyótate: the root udātta is on dyó, and vi becomes anudātta by 3978.
• Counter-example: na tena sma purādhīyate ("long past"; laṭ by SK 2778).
• 3950 nanv ity anujñaiṣaṇāyām 8.1.43 — with nanu, when permission is requested: nanu gácchāmi bhoḥ (gá; "permit me, as I am going").
• iti marks nanu as a single particle.
• Counter-example, a reply to a question: akārṣīḥ kaṭaṃ tvam? nanu karomi. The laṭ there is past in sense by SK 2780. The print has ākarṣīḥ (V, Print Check).

VIII.F. Questions and indefinites: kim of the action and its omission, ehi manye in jest, jātu, kiṃvṛtta + cit, āho/utāho — running example: kiṃ dvijaḥ pácaty ā́hosvid gacchati (the first verb keeps its root udātta pá; ā́hosvit ādyudātta)

• 3951 kiṃ kriyāpraśne 'nupasargam apratiṣiddham 8.1.44 — with kim asking about the action, a verb without an upasarga and not negated is kept accented: pácati.
• Who keeps the accent. The Subodhinī reports a dispute over the second verb, gacchati. Haradatta: it is nighāta. Others: both verbs are protected. The outline leaves gacchati unmarked.
• Counter-examples:
• kiṃ bhaktaṃ pacaty apūpān vā (a question about a means, not the action).
• kiṃ paṭhati (censure, not a question).
• kiṃ prapacati uta prakaroti (upasarga).
• kiṃ dvijo na pacati (negated).
• 3952 lope vibhāṣā 8.1.45 — when kim is not used (lopa means non-use; no rule elides it), the whole of 3951 holds optionally.
• devadattaḥ pacaty āhosvit paṭhati: pacati is nighāta, or pácati (pá).
• 3953 ehimanye prahāse lṛṭ 8.1.46 — a lṛṭ form connected with ehi manye is kept accented in jest.
• ehi mánye bhaktaṃ bhokṣyáse (kṣyá; mánye ādyudātta). The persons are exchanged by SK 2163: "you think: I shall eat".
• A niyama. ehi is the loṭ of a verb of motion, so 3958 would already apply. This rule restricts it to jest.
• Counter-example: ehi manyase odanaṃ bhokṣye, where bhokṣye is nighāta. The Bhāṣya's counter-example is ehi manye rathena yāsyasi.
• 3954 jātv apūrvam 8.1.47 — with jātu that has nothing before it: jā́tu bhokṣyáse (jā́, a particle; kṣyá).
• Counter-example (audit MUST-FIX): kaṭáṃ jā́tu kariṣyasi. jātu keeps its own udātta, and kariṣyasi is nighāta.
• apūrvam qualifies jātu, not the verb. The jñāpaka is anantaram in 3956.
• 3955 kiṃvṛttaṃ ca ciduttaram 8.1.48 — with a form of kim (a vibhakti-final, or one with ḍatara or ḍatama) that is followed by cit, and with nothing before it (from 3954).
• Examples: kaścid bhuṅkte, kataraś cit, katamaś cit. bhuṅkte keeps its own accent, which is not stated.
• Counter-examples: ko bhuṅkte (no cit); rāmaḥ kiṃcit paṭhati (something before it). Print: ciduttaram.
• 3956 āho utāho cānantaram 8.1.49 — with āho or utāho, with nothing before them, when the verb immediately follows: āho bhuṅkte, utāho bhuṅkte.
• Counter-example: deva āho bhuṅkte. Here bhuṅkte is nighāta by 3935.
• There is no Subodhinī on this sūtra.
• 3957 śeṣe vibhāṣā 8.1.50 — in the remaining case, where the verb is separated from āho or utāho, it is optionally anudātta.
• āho devaḥ pacati: pacati is nighāta, or pácati (pá).

VIII.G. A verb connected with the loṭ of a verb of motion: lṛṭ, then loṭ, with an upasarga optionally, and with hanta — running example: āgaccha deva grāmaṃ drakṣyásy enam (lṛṭ after the motion loṭ āgaccha: drakṣyasi keeps its suffix udātta ṣyá)

• 3958 gatyarthaloṭā lṛṇ na cet kārakaṃ sarvānyat 8.1.51 — a lṛṭ form connected with the loṭ of a verb of motion is not anudātta, provided its kāraka is not wholly different from the loṭ's.
• Examples: drakṣyási (ṣyá); uhyantāṃ devadattena śālayo rāmeṇa bhokṣyante. In the second, the object (the rice) is the same, though the agents differ.
• Counter-examples:
• paca … bhokṣyase (not a verb of motion).
• āgaccheḥ … (liṅ, not loṭ).
• … paśyasi … (not lṛṭ).
• pitā te odanaṃ bhokṣyate (a different individual agent).
• sarva. Because of sarva, the rule still applies in tvaṃ cāhaṃ ca drakṣyāvaḥ, where the lṛṭ expresses the loṭ's kāraka and another besides.
• Print: bhokṣyante.
• 3959 loṭ ca 8.1.52 — a loṭ form too is kept accented: āgaccha deva grāmaṃ páśya (pá).
• Counter-examples: paca devaudanaṃ bhuṅkṣvainam (not a verb of motion); … paśyasi (not loṭ); [… paśyatv evaṃ rāmaḥ] (a different agent, so nighāta). The last is supplied from the print.
• Because of sarva: tvaṃ cāhaṃ ca paśyāva. The print has paśyāvaḥ; this is a Print Check item.
• Why a separate rule (yogavibhāga). It lets the option of 3960 apply to loṭ only.
• 3960 vibhāṣitaṃ sopasargam anuttamam 8.1.53 — such a loṭ, with an upasarga and not in the first person, is accented only optionally (a prāptavibhāṣā).
• āgaccha deva grāmaṃ praviśa. The two options:
• praviśá (śá): pra becomes anudātta by 3978.
• práviśa: viśa is anudātta, and pra keeps its own udātta.
• Counter-examples: paśya (no upasarga) and āgacchāni … praviśāni (first person). In both, 3959 applies invariably.
• 3961 hanta ca 8.1.54 — with hanta, the same option holds: hanta praviśa gives praviśá / práviśa.
• This is a prāptavibhāṣā against 3937, whose list includes hanta.
• Counter-examples: hanta kuru (no upasarga, so 3937 prohibits nighāta outright); hanta prabhuñjāvahai (first person).
• Print: "(SK 3937)".

VIII.H. Address after ām, and words standing after the verb: yad/hi/tu in the Veda, cana, cit, iva, gotrādi, taddhita, āmreḍita, cādi — running example: devaḥ pácati cana (cana follows; pacati, not after a gati, keeps its root udātta pá)

• 3962 āma ekāntaram āmantritam anantike 8.1.55 — an āmantrita after ām, with one pada between, is not anudātta when the addressee is not near.
• ām pacasi dévadatta3: dé, the ṣāṣṭha ādyudātta of SK 3653. Without this rule it would be nighāta by SK 3654. The pluta is by SK 95.
• Counter-examples: ām prapacasi devadatta3 (two padas between); ām pacati devadattaḥ (not an āmantrita); ām pacasi devadatta (near).
• The Subodhinī gives two views of anantika, "far" or "neither far nor near".
• 3963 yaddhituparaṃ chandasi 8.1.56 — in the Veda, a tiṅanta followed by yad, hi or tu is not anudātta.
• Examples:
• udásṛjo yád aṅgiraḥ: dá is the augment. The first syllable of aṅgiraḥ is svarita (audit MUST-FIX).
• uśánti hí (śá).
• ākhyāsyā́mi tu te (syā́; ā is anudātta by 3978).
• A niyama. 3937, 3941 and 3946 would already give the result. The rule restricts it: these particles must stand after the verb, and others (yadi, hanta) do not count.
• So jāye svā rohāvaihi: rohāva, connected with the motion-loṭ ehi that follows it, is nighāta, and 3959 is withdrawn.
• 3964 canacidivagotrāditaddhitāmreḍiteṣv agateḥ 8.1.57 — when cana, cit, iva, a gotrādi word, a taddhita or an āmreḍita follows, a tiṅanta not preceded by a gati is not anudātta.
• Examples: devaḥ pácati cana, … pácati cit, pácatīva, pácati gotram, pácatikalpam, pácatipacati.
• Only an anudātta taddhita is meant (kalpap is pit). In pacatideśyaḥ the taddhita's own accent prevails (Subodhinī).
• Counter-example: devaḥ prapacati cana (after a gati).
• 3965 cādiṣu ca 8.1.58 — when ca, vā, ha, aha or eva follows, with agateḥ carried by "ca": devaḥ pácati ca khā́dati ca (pá, khā́).
• This is the list of SK 408, not the cādi of SK 20 (Subodhinī).
• Counter-example: devaḥ prapacati ca prakhādati ca. Only the first verb is saved, by 3966; the second is nighāta.
• Print Check: pareṣu.

VIII.I. The first verb (prathamā) with ca/vā, ha, aha and elided particles; vai/vāva, eka/anya; and always after yad — running example: itó vā sātím ī́mahe (with vā: the first tiṅanta īmahe keeps its udātta ī́)

• 3966 cavāyoge prathamā 8.1.59 — in connection with ca or vā, the first tiṅ-vibhakti is not anudātta.
• Examples: gāś ca cārayati vīṇāṃ vā vādayati; itó vā sātím ī́mahe (ī́).
• "First" in the mantra. īmahe counts as first relative to the verb understood in the next two clauses, divo vā pārthivād adhi and indraṃ maho vā rajasaḥ (Subodhinī).
• Why yoge. It brings in ca or vā standing before the verb as well, and that connection too saves the verb from nighāta. Print Check: the avagraha in nighātārtham.
• Why "first". Later verbs are not covered.
• agateḥ does not continue here. Print: sātim.
• 3967 heti kṣiyāyām 8.1.60 — with ha, in a breach of proper conduct: svayaṃ ha rathena yāti3 upādhyāyaṃ padātiṃ gamayati3.
• yāti, the first verb, is kept accented; gamayati is not first. The pluta is by SK 3623.
• 3968 aheti viniyoge ca 8.1.61 — with aha, in a commission for several purposes, and in kṣiyā.
• Examples: tvam aha grāmaṃ gaccha; svayam aha rathena yāti3 upādhyāyaṃ padātiṃ nayati.
• 3969 cāhalopa evety avadhāraṇam 8.1.62 — when ca or aha is not used, and eva is restrictive, the first verb is not anudātta.
• ca-lopa, same agent: deva eva grāmaṃ gacchatu, deva evāraṇyaṃ gacchatu.
• aha-lopa, different agents: deva eva grāmaṃ gacchatu, rāma evāraṇyaṃ gacchatu.
• Counter-example: deva kveva bhokṣyase. Here eva expresses impossibility, not restriction.
• In both examples, the first gacchatu keeps its accent.
• 3970 cādilope vibhāṣā 8.1.63 — when ca, vā, ha, aha or eva is not used, the first verb is accented optionally.
• índra vā́jeṣu no 'va: ava is anudātta here. The Subodhinī says this shows only the nighāta option.
• śuklā vrīhayo bhavanti, śvetā gā ājyāya duhanti.
• vrīhibhir yajeta, yavair yajeta (vā-lopa): the first verb is yájeta (yá) in one option, and the second is nihata.
• 3971 vaivāveti ca cchandasi 8.1.64 — in the Veda, with vai or vāva, the first verb is optionally accented: ahar vai devānām āsīt, ayaṃ vāva hasta āsīt. There is no accented copy, so nothing is marked.
• 3972 ekānyābhyāṃ samarthābhyām 8.1.65 — in the Veda, with eka and anya when they mean the same ("the one … the other"), the first verb is not anudātta.
• Examples: ajām ekāṃ jinvati, ajām ekāṃ rakṣati; táyor anyáḥ píppalaṃ svādv átti (á of átti).
• Counter-example: eko devān upātiṣṭhat, where eka is a number.
• Print: svādv atti.
• 3973 yadvṛttān nityam 8.1.66 — after a pada containing yad, a tiṅanta is always accented.
• Examples: yo bhuṅkte; yadadryaṅ vāyur vāti. In the second, the operation holds even though a word intervenes; the jñāpaka for this is in 3974.
• Print: yadadryaṅ (adri for the ṭi, SK 418).

VIII.J. Anudātta imposed: the praised word, the verb with its gati, the censured action, the gati before a gati or an accented verb; then the svara-saṃcāra — running example: yat prapácati (the verb keeps pá after yat, so the gati pra goes anudātta) against yat kāṣṭhāṃ prapacati (in praise: pra and pacati both anudātta)

• 3974 pūjanāt pūjitam anudāttaṃ kāṣṭhādibhyaḥ 8.1.67 — after kāṣṭha and the rest, words expressing praise, the word for what is praised is anudātta. pūjanāt is taken as plural.
• kāṣṭhādhyāpakaḥ ("a wonderful teacher"): adhyāpaka is anudātta.
• Vārttika malopaś ca: dāruṇādhyāpakaḥ, ajñātādhyāpakaḥ.
• Scope. This is an apavāda to the compound's final udātta. On the Vṛtti's view it is wanted only in a compound, not in dāruṇam adhyāpakaḥ. The Subodhinī reports that the Bhāṣya, Kaiyaṭa and Haradatta want it outside compounds.
• Why pūjitam is stated. It secures the immediately following word. That makes it the jñāpaka that immediacy is not relied on in this section even with a fifth case; that jñāpaka is used at 3973.
• 3975 sagatir api tiṅ 8.1.68 — a tiṅanta, together with its gati or without one, is anudātta as the thing praised.
• yat kāṣṭhāṃ prapacati: pra and pacati are both anudātta. This is where 3937 would otherwise have blocked 3935.
• Vārttika gatigrahaṇe upasargagrahaṇam: only an upasarga counts as the gati here. So in yat kāṣṭhāṃ śuklīkaroti, śuklī (a gati by SK 762) is not made anudātta.
• 3976 kutsane ca supy agotrādau 8.1.69 — in censure, before a subanta other than a gotrādi word, a tiṅanta (with or without a gati) is anudātta.
• Examples: pacati pūti, prapacati pūti, pacati mithyā.
• Counter-examples: prapacati śobhanam (no censure); pacati kliśnāti (no subanta follows); pacati gotram (a gotrādi word).
• Vārttika kriyākutsana iti vācyam: the censure must be of the action, not of the agent: pacati; pūtir devadattaḥ.
• Vārttika on pūti: pūti has c as its anubandha, so pūtí is antodātta (tí, SK 3710).
• Vārttika vā bahvartham: a verb denoting many is optionally anudātta: pacanti pūti.
• padāt has ceased (Subodhinī).
• 3977 gatir gatau 8.1.70 — a gati followed by a gati is anudātta.
• abhyúddharati: abhi is anudātta. ud keeps its udātta, ú (phiṭ 81), and harati is nighāta.
• Counter-example to "gati", as transmitted: dattaḥ pacati.
• Counter-example to "gatau": ā́ mandraír indra háribhir yāhí mayū́raromabhiḥ. ā is followed by mandraiḥ, which is not a gati, so ā keeps ā́.
• Print: haribhir yāhi.
• 3978 tiṅi codāttavati 8.1.71 — a gati before a tiṅanta that has an udātta is anudātta. This is an apavāda to phiṭ 80.
• yat prapácati: pacati keeps pá by 3937, so pra is anudātta.
• Counter-example: prápacati. pacati is nighāta by 3935, so pra keeps prá.
• Why tiṅ. It makes the whole verb, not the bare root, the measure of "having an udātta". So the rule also applies when the udātta is on the affix, as in yat prakaróti (Subodhinī). This closes the tiṅantasvarāḥ.
• The svara-saṃcāra (heading "svarasaṃcāraprakāraḥ 74", supplied from the print) works through agním īḷe puróhitaṃ yajñásya devám ṛtvíjam hótāraṃ ratnadhā́tamam:
• agni is antodātta. Mādhava derives this by phiṭ 1; in fact it comes from the ghṛtādi list, or from the affix ni.
• agním. am is anudātta as a sup. The single substitute (SK 194) is udātta by SK 3658: agním (gní).
• īḷe is nighāta (3935). In saṃhitā its ī becomes svarita (SK 3660), and ḷe becomes ekaśruti/pracaya (SK 3668).
• puras + hita. puras ends in asi (SK 1975) and hita in dhā → hi (SK 3076). The gati compound (SK 768, 761) passes through the compound-final udātta, then SK 3736, SK 3873 and thāthādi, each overriding the last. Finally gatir anantaraḥ (SK 3783) gives puróhitam (ró). The u of puras is anudāttatara (SK 3669).
• The other words:
• yajñásya (ñá): naṅ (SK 3268), with the sup made svarita.
• devám (vá).
• ṛtvíjam (tví).
• hótāram (hó): the nit affix tṛn.
• ratnadhā́tamam (dhā́): ratna is ādyudātta by phiṭ 26, and tamap is pit, so anudātta.
• The close. The section ends with "iti svaraprakaraṇam", then the closing verse to Bhavānī and Viśvanātha, then the supplied colophon of the vaidika-svara-prakriyā.