कर्मकर्तृतिङ्प्रकरणम्

Covers SK 2766–2772 (7 sūtras): all present. Every unit has its own kaumudī. Three units are long: 2766 (the karmavat sūtra with its two vārttikas), 2769 (duhaś ca with the sṛj/yuj and bhūṣā/kirādi/san vārttikas) and 2771 (tapas tapaḥkarmakasyaiva, followed by the snu/nam examples of 2767 and the hetumaṇṇic and Bhāradvājīya vārttikas). Each is outlined under its own number and continued in the groups its vārttikas serve:

• 2766 in A → B;
• 2769 in C → B, D and E;
• 2771 in B → D.

The spine is two Aṣṭādhyāyī runs, interleaved as the print orders them:

• 3.1.87 (2766), 3.1.89 (2767);
• 3.1.62–64 (2768–2770);
• 3.1.88 (2771), 3.1.90 (2772).

Every AS-ref below is checked against references/ashtadhyayi/sutraani-data.json, and all 7 match our source.json.

• Anuvṛtti from p63. The karmakartṛ operations are the ones taught for bhāva and karma:
• yak (sārvadhātuke yak, 3.1.67, SK 2756);
• ātmanepada (bhāvakarmaṇoḥ, 1.3.13, SK 2679);
• ciṇ (ciṇ bhāvakarmaṇoḥ, 3.1.66, SK 2758);
• the ciṇvad-iṭ (6.4.62, SK 2757) (BM).

tapo 'nutāpe ca (3.1.65, SK 2760) bars ciṇ for tap at 2771.

• Anuvṛtti from earlier prakaraṇas:
• kartari ← kartari śap (SK 2167), recast in the first case (2766, 2771; BM, TB);
• cleḥ ← SK 2222; ciṇ / te ← SK 2513; anyatarasyām ← dīpajana… (SK 2328); dhātoḥ ← SK 2629 (2768);
• "l" ← the two-l sūtra-text vyatyayo bahulam / liṅy āśiṣy aṅ (SK 3433–3434; 2766).

karmakartṛtiṅprakaraṇam — outline (p64, SK 2766–2772, all 7 sūtras)

p63 (bhāvakarmatiṅprakaraṇam) closed at 2765 with the dvikarmaka verse and the colophon "iti bhāvakarmaprakriyā", which is supplied in brackets.

• Where the prakaraṇa begins. The print heads it "tiṅante karmakartṛprakriyā 61" (NSP p. 467, scan 0477). In kaumudī type it continues with a whole opening paragraph (scans 0477–0478), which ours lacks: "yadā saukaryātiśayaṃ dyotayituṃ … pacyate odanena | bhidyate kāṣṭhena | kartari tu |". The English supplies it in brackets at the head of sk.2766.k1. Bhaṭṭoji's later "pūrvokte sādhvasir ityādau" refers back to it. The name karmakartṛtiṅprakaraṇam and the number 64 are ours.
• Where it ends. 2772 closes with koṣiṣīṣṭa and raṃkṣīṣṭa. The print then has "iti karmakartṛprakriyā" (NSP p. 470, scan 0480). Ours lacks it, so the English supplies it in brackets. p65 (lakārārthaprakaraṇam) opens with abhijñāvacane lṛṭ (3.2.112, SK 2773).
• Contents in outline.
• Opening and 2766: the object as agent, treated like the object. pacyate odanaḥ, bhidyate kāṣṭham, apāci, abhedi. Only for roots whose action resides in the object, and only for the agent the lakāra expresses.
• 2766 vārttikas, 2771: where karmavattva is barred or newly given. It is barred for roots that stay transitive (anyonyaṃ spṛśataḥ, ajā grāmaṃ nayati). It is restored, bahulam, for duh and pac (udumbaraḥ phalaṃ pacyate). It is enjoined for tap with tapas as object (tapyate tapas tāpasaḥ).
• 2768–2770: ciṇ in the karmakartṛ luṅ. It is optional after vowel-final roots (akāri / akṛta) and after duh (adohi / adugdha / adhukṣata), and barred after rudh (avāruddha).
• 2767, with the 2771 and 2769 vārttikas: ātmanepada without yak or ciṇ.
• The sūtra covers duh, snu and nam: dugdhe, prasnute, namate.
• By vārttika, hetumaṇṇic, śri and brū: kārayate, ucchrayate, brūte.
• By the Bhāradvājīyas, ṇi, śranth, granth and the akarmaka of the taṅ-rule: utpucchayate, granthate, vikurvate.
• By another vārttika, bhūṣā roots, kirādi and san-finals: alaṃkurute, avakirate, cikīrṣate.
• 2769 vārttikas, 2772: śyan in place of yak. sṛj and yuj (sṛjyate srajaṃ bhaktaḥ, yujyate brahmacārī yogam); kuṣ and raj, optionally with parasmaipada (kuṣyati / kuṣyate pādaḥ, rajyati / rajyate vastram).

Threads that run through the prakaraṇa:

• Two kinds of non-agent as agent (the opening). When the agent's activity is not meant to be expressed, in order to convey ease (saukaryātiśaya), other kārakas become agents, being independent in their own activity.
• An instrument or locus as agent. It is an ordinary agent, so the lakāra is in the agent: sādhv asiś chinatti, kāṣṭhāni pacanti, sthālī pacati. 2766 does not apply, which is why the sūtra says "karmaṇā".
• The object as agent. The root mostly (prāyeṇa) becomes intransitive. Single-object roots like bhid and chid do; dvikarmaka roots stay transitive (TB). So the lakāra may be in bhāva, the impersonal pacyate odanena, or in the agent, which is 2766's domain.
• Kāryātideśa, not śāstrātideśa (2766). The two give the same result. But the extension of rules exists only for the sake of the operations, so the extension of operations is the principal view and is adopted (BM, TB). This decides p62's 2745: the later 2766 blocks the parasmaipada of anuparābhyāṃ kṛñaḥ. Audit MINOR: the English bracket "[not an extension of the designation karman]" names the wrong contrast. The contrast is with śāstrātideśa.
• The "vat" guards against a saṃjñā reading. As a saṃjñā, the root would stay transitive, and impersonal pacyate odanena would be impossible (BM, TB).
• The action must reside in the object (karmastha-kriyā). The test is whether the object shows a difference produced by the action (kriyākṛta viśeṣa): cooked against uncooked rice, split against unsplit wood. A village gone to looks no different, so gam is agent-residing (kartṛstha) and takes no karmavattva: gacchati grāmaḥ, ārohati hastī, adhigacchati śāstrārthaḥ.
• kṛ means utpādana. Coming into being resides in the object, so kāriṣyate ghaṭaḥ. If kṛ meant effort (yatna) it would be agent-residing (TB: against the Naiyāyikas).
• anuvyavasyamāne 'rthe. It is agent-residing, so there is no yak. śyan follows and o is deleted (otaḥ śyani, 7.3.71, SK 2510). The -āna is cānaś (tācchīlyavayovacanaśaktiṣu cānaś, 3.2.129, SK 3109), not an ātmanepada ending.
• Only the agent denoted by the lakāra is karmavat. "l" is carried forward from the two-l sūtra-text (SK 3433–3434). The mechanism is disputed: BM says a second l was lost by halo yamāṃ yami lopaḥ; TB says the anusvāra became l by parasavarṇa. The English names neither. Two consequences:
• In a bhāva-lakāra the agent does not take the second case.
• No kṛtya, kta or khalartha in the karmakartṛ, only in bhāva: bhettavyaṃ kusūlena. BM, via the Śabdenduśekhara, limits the kta exclusion to the niṣṭhā-kta (SK 3013) (sino grāsaḥ svayam eva). TB says kta should be dropped from the list. The English keeps Bhaṭṭoji.
• "svayam" in "pacyate odanaḥ svayam eva". It means "by itself as the means" (ātmanā karaṇena), not as agent; otherwise the lakāra would be in the karma (BM, TB, after Kaiyaṭa). Hence "of itself" in the glosses.
• Commentator disagreements, recorded neutrally:
• 2766 k3: which example truly illustrates sakarmakāṇāṃ pratiṣedhaḥ. BM says anyonyaṃ spṛśataḥ, and that ajā grāmaṃ nayati is not a true case. TB says the reverse.
• 2766: the motive for "kiṃca" (BM: the "dhātoḥ"-anuvṛtti objection, which TB in the print rejects; TB: proximity). Audit MINOR.
• 2769: whether pac is really dvikarmaka. Both say duhipacyoḥ then becomes a fresh injunction of karmavattva for pac.
• 2771: eva. The Bhāṣya makes it a niyama (BM), and "vṛttyanusāriṇaḥ" hints at a conflict.
• Where the commentary lies. Only the kaumudī blocks are translated.
• TB on the opening paragraph is in sk.2765.t1. Our OCR skips its first part: anuṣaṅga, prāyeṇa, mathnāti sāgaro 'mṛtam, svayaṃ pradugdhe…, and TB's dissent that the primary object can also be meant as agent (ajā grāmaṃ nayati). See scan 0477.
• BM on 2768 is at the end of sk.2767.b1; 2768 has no b1.
• TB on 2771 begins at the tail of sk.2770.t1.
• TB on 2772, the colophon and TB on p65's opening are in sk.2771.t1; 2772 has no t1.
• TB on namate daṇḍaḥ is truncated in ours. The print continues "uddeśānurodhāc ca karmasthakriyatvam asti…" (scan 0480).

The p62–66 audit (.work/sk-run/p62-66-audit/c-sk2766-2776.md) found no MUST-FIX here. A fixer is applying the MINOR items in parallel. This outline already follows the audit:

• 2766: "kāryātideśa" is contrasted with śāstrātideśa (above).
• 2766: "That very action now resides in the agent; it is not similar to it." The objection is that identity rules out similarity, since similarity presupposes difference (TB, BM). Not "[merely] similar".
• 2766: TB in the print rejects the "dhātoḥ" objection that BM gives as Bhaṭṭoji's motive for "kiṃca".
• 2771: kārayate = "[Devadatta] makes [of himself, unprompted]" (BM), rather than "gets himself to make".
• 2772: BM's encoded "varuām" = vasram. In that encoding the garbled "ruā" regularly stands for sr. So ours, BM, Wikisource and probably the print all read vasram. The English keeps vastram ("cloth") with a note.

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

• the opening paragraph and the colophon;
• 2766: utpādana (ours "rū-"), pratiṣedha (ours "-tha");
• 2769: yagapavādaś ca, kirādi;
• 2771: tapaḥkarmakasyeti, uttapati, śrathnīte / grathnīte (ours "śranthīte granthīte");
• 2767: "ciṇ tu vikalpeneṣyate" is the print's reading (BM calls it a "kvacit pāṭhaḥ").
• The "(vārtikam)" tags at 2766 k3/k5, 2769 k3/k5/k7 and 2771 k3/k5 are editorial. 2766 k5 is the tag alone, rendered "—".

Still open for Ajai (mūla):

• 2766: "viklatti" (print = ours). viklitti is meant, as BM and Wikisource read.
• 2769: "lugvā…" is cross-referred as SK 365 (print = ours). It is SK 2365 (lug vā duhadihalihaguhām ātmanepade dantye, 7.3.73).
• 2772: vasram / vastram (above).
• Every mūla correction listed under Print readings stands in ours unedited (source.json has not been changed).

Ordering: the groups follow outcome, not strict SK order.

• 2767 (A 3.1.89), placed by the print right after the duh/pac vārttika, is outlined in D with the other yak/ciṇ prohibitions.
• 2771 (A 3.1.88) is outlined in B, and its snu/nam examples and vārttikas in D.
• 2769's vārttikas go to B (the udumbara example), D (bhūṣā/kirādi/san) and E (sṛj/yuj).
• The SK numbers within each group stay in order.

A. The object as agent, treated like the object: yak, taṅ, ciṇ and the ciṇvad-iṭ — running example: odanaṃ pacati → (odana meant as agent) pac + laṭ (kartari) → (karmavat: yak, taṅ) pacyate odanaḥ; luṅ: cli → ciṇ → apāci

• 2766 karmavat karmaṇā tulyakriyaḥ 3.1.87 — an agent (kartari, from SK 2167, recast in the first case) denoted by the lakāra, whose action is like the action residing in the object, is treated like the object. This is a kāryātideśa.
• The work in pacyate odanaḥ:
• odana, the former karma, is now the agent;
• softening (viklitti) is the action that resided in it as object;
• pac stays a pac-action, and is now agent-located.
• karmavattva brings:
• yak (SK 2756): pac-ya;
• taṅ (SK 2679): -te.
• The agent is expressed by the tiṅ, so it is in the first case: pacyate odanaḥ ("the rice cooks [by itself]"), bhidyate kāṣṭham.
• luṅ with ciṇ (SK 2758): apāci, abhedi. With the ciṇvad-iṭ (SK 2757): kāriṣyate ghaṭaḥ.
• "Similar" though identical? The softening is the same action, now in the agent, and not similar to it. Answer: the states of being object and being agent differ, so there is a conditioned (aupādhika) difference in the action co-located with them. On that the usage "similar" rests.
• "karmaṇā". It excludes an instrument or locus as agent (sādhv asiś chinatti, sthālī pacati: no yak). "kiṃca" adds that it also excludes agent-residing roots: gacchati grāmaḥ, ārohati hastī, adhigacchati śāstrārthaḥ smarati śraddadhāti ca.
• For kṛ, the l-condition, bhettavyaṃ kusūlena and anuvyavasyamāne, see the threads above.

B. Where karmavattva is barred, restored or newly given — running example: udumbaraṃ phalaṃ pacati (kālaḥ) → (udumbara as agent; pac still transitive through phala) → duhipacyoḥ: yak, taṅ → udumbaraḥ phalaṃ pacyate; tapas tapati → tapyate tapas tāpasaḥ

• 2766 vārttika (k2) sakarmakāṇāṃ pratiṣedho vaktavyaḥ (print pratiṣedho; ours "pratiṣetho") — no karmavattva for roots that remain transitive when an object becomes agent, as dvikarmaka roots do (BM).
• anyonyaṃ spṛśataḥ: each is agent and object of the touching; without this there would be spṛśyete (BM).
• ajā grāmaṃ nayati: the goat, the primary object, as agent, still transitive through grāma (TB).
• The commentators dispute which example is apt; see the threads above.
• 2766 vārttika (k4) duhipacyor bahulaṃ sakarmakayor iti vācyam — for duh and pac, though they stay transitive, karmavattva comes variously. It is a pratiprasava of k2.
• The pac example is given at 2769 k1: udumbaraḥ phalaṃ pacyate. udumbara, the gauṇa karma of "kālaḥ udumbaraṃ phalaṃ pacati", is meant as agent. phala keeps pac transitive, so k2 would bar karmavattva and k4 restores it: yak and taṅ (BM, TB).
• BM and TB: even if pac is not truly dvikarmaka, k4 is then a fresh injunction (apūrvavidhi) of karmavattva for pac.
• 2771 tapas tapaḥkarmakasyaiva 3.1.88 — the agent of tap, only when tap has tapas as its object, is treated like the object. tapaḥ is a sixth case; kartari and karmavat are carried forward (BM).
• This is a vidhi. In tap = "earn", the accomplishing resides in the agent (the ascetic), so 2766 would not apply (BM).
• The Vṛtti-followers call eva superfluous; BM says that in the Bhāṣya it is a niyama.
• In tapyate tapas tāpasaḥ ("the ascetic practises / earns austerity"):
• tāpasa is the agent;
• tapas is the object;
• karmavattva → yak, taṅ.
• luṅ: ciṇ is barred by tapo 'nutāpe ca (SK 2760), so sic: atapta [tapas tāpasaḥ].
• Why "tapaḥkarmaka"? uttapati suvarṇaṃ suvarṇakāraḥ: no karmavattva.

C. ciṇ in the karmakartṛ luṅ: optional, or barred — running example: kaṭaḥ svayam eva, kṛ + luṅ ta → cli → ciṇ (optional) → akāri / (sic, deleted) akṛta

• 2768 acaḥ karmakartari 3.1.62 — after a vowel-final root, in the karmakartṛ, ciṇ optionally replaces cli when ta follows. cleḥ comes from SK 2222, ciṇ / te from SK 2513 and anyatarasyām from SK 2328. dhātoḥ (SK 2629) is qualified by acaḥ, with tadantavidhi (BM, TB).
• In akāri, the ṛ of kṛ is the ac-final, and ta in the karmakartṛ is the trigger. With ciṇ: vṛddhi, r, and ta deleted by ciṇo luk (SK 2329): akāri.
• Without ciṇ: sic, deleted by hrasvād aṅgāt (8.2.27, SK 2369): akṛta ("[the mat] got made [of itself]").
• Counter-examples, where ciṇ is obligatory (TB): abhedi kāṣṭham (not ac-final); akāri ghaṭaḥ kulālena (not karmakartṛ).
• 2769 duhaś ca 3.1.63 — after duh too, the same: optional ciṇ in the karmakartṛ before ta. Everything in 2768 except acaḥ carries over (BM).
• adohi (ciṇ, laghūpadha-guṇa). Otherwise ksa (śala igupadhād aniṭaḥ ksaḥ, 3.1.45, SK 2336), and then optional luk of ksa by lug vā… (7.3.73, SK 2365; ours "365"): adugdha / adhukṣata.
• Outside the karmakartṛ, ciṇ is obligatory: adohi gaur gopena (TB).
• (k1's udumbara example → B; k2–k5 → E; k6–k7 → D.)
• 2770 na rudhaḥ 3.1.64 — after rudh, in the karmakartṛ, no ciṇ. karmakartari comes from 2768 and ciṇ from SK 2513 (BM, TB).
• avāruddha gauḥ ("the cow got confined [of herself]"): sic, no ciṇ.
• Outside the karmakartṛ: avārodhi gaur gopena.

D. ātmanepada without yak or ciṇ — running example: gāṃ payo dogdhi → (gauḥ, gauṇa karma, as agent) duh + laṭ → (no yak) śap, luk → gauḥ payo dugdhe

• 2767 na duhasnunamāṃ yakciṇau 3.1.89 — for duh, snu and nam in the karmakartṛ, no yak and no ciṇ. karmakartari comes from the co-utterance with 2766, or by a frog's leap (maṇḍūkapluti) from 2768 (BM).
• duh: this rule bars only yak. ciṇ stays optional by 2769 (the kaumudī; BM: "kvacit pāṭhaḥ").
• In gauḥ payo dugdhe, gauḥ is the gauṇa karma as agent. yak is barred, so śap comes and is deleted by adiprabhṛtibhyaḥ śapaḥ (2.4.72, SK 2423). The ātmanepada is by karmavattva (SK 2679).
• payo gāṃ dugdhe is not the example, by "gauṇe karmaṇi duhyādeḥ" (2765) (BM, TB).
• snu (given at 2771 k1): prasnute [gauḥ] ("lets her milk flow"). snu means eagerness for the flow of milk, residing in the cow; vatso gāṃ prasnauti has the causative sense incorporated (TB). luṅ: no ciṇ, but the ciṇvad-iṭ still optionally: prāsnāviṣṭa / prāsnoṣṭa.
• nam (2771 k1): namate daṇḍaḥ ("the stick bends [of itself]"), with the causative sense incorporated; anaṃsta. TB in the print: karmasthakriyatva holds through what is intended (uddeśa), the bending.
• 2771 vārttika (k2, introduced by Bhaṭṭoji's "yakciṇoḥ pratiṣedhe") hetumaṇṇiśribrūñām upasaṃkhyānam — add hetumaṇṇic-finals, śri and brū to the prohibition.
• kārayate ("[Devadatta] makes [of himself, unprompted]"; BM); acīkarata, with caṅ by ṇiśri… (3.1.48, SK 2312).
• ucchrayate daṇḍaḥ; udaśiśriyata.
• The ciṇvad-iṭ still comes, since ṇi's and śri's vowels are in upadeśa (BM): kāriṣyate, ucchrāyiṣyate.
• brūte kathā ("the story tells [itself]"); avocata.
• 2771 vārttika (k4, the Bhāradvājīyas) ṇiśranthigranthibrūñātmanepadākarmakāṇām upasaṃkhyānam — add ṇi (any ṇi, so ṇiṅ too), śranth, granth, brū, and roots made intransitive in the taṅ-rule.
• ṇiṅ: puccham udasyati → utpucchayate gauḥ. With the causative sense incorporated, utpucchayate gām; with the cow again as agent, utpucchayate gauḥ, udapupucchata. yak and ciṇ are barred, so śap and caṅ come.
• śranth / granth: named for their ṇic-less option (ādhṛṣīya): granthati grantham → granthate granthaḥ, agranthiṣṭa; śranthate, aśranthiṣṭa. The kryādi roots take śnā: śrathnīte, grathnīte svayam eva.
• Intransitive in the taṅ-rule: vikurvate saindhavāḥ ("the horses prance"), taṅ by veḥ śabdakarmaṇaḥ (1.3.34, SK 2707) and akarmakāc ca (SK 2708). With the prompting dropped, as karmakartṛ: vyakāriṣṭa… / vyakṛta, vyakṛṣātām, vyakṛṣata.
• 2769 vārttika (k6) bhūṣākarmakirādisanāṃ cānyatrātmanepadāt — for roots meaning adorning, the kirādi (a subgroup of tudādi) and san-finals, every karma-operation except the ātmanepada is barred: no yak, no ciṇ, no ciṇvad-iṭ (k7).
• alaṃkurute kanyā, alamakṛta;
• avakirate hastī, avākīrṣṭa;
• girate, agīrṣṭa;
• ādriyate, ādṛta;
• cikīrṣate kaṭaḥ, acikīrṣiṣṭa. Desire resides in the agent, but the karmasthatā is reckoned with respect to kṛ's action.

E. śyan in place of yak — running example: kuṣṇāti pādaṃ devadattaḥ → (pāda as agent) kuṣ + śyan + parasmaipada → kuṣyati pādaḥ, or yak + taṅ → kuṣyate

• 2769 vārttika (k2) sṛjiyujyoḥ śyaṃs tu — for sṛj and yuj used transitively, the agent is variously treated like the object, but with śyan in exception to yak (k3). Being n-marked, śyan gives initial acute accent (BM, TB).
• vārttika (k4) sṛjeḥ śraddhopapanne kartary eveti vācyam: for sṛj, only with an agent endowed with faith.
• In sṛjyate srajaṃ bhaktaḥ ("the devotee makes a garland [with faith]"), the lakāra is in the main agent. karmavattva gives taṅ, with śyan for yak: asarji.
• yujyate brahmacārī yogam: the brahmacārin, the object of "yoga unites him", is now agent (BM).
• 2772 kuṣirajoḥ prācāṃ śyan parasmaipadaṃ ca 3.1.90 — for kuṣ and raj in the karmakartṛ, optionally (prācām), there is not yak but śyan, together with parasmaipada. karmakartari comes by a frog's leap from 2768, and na and yak from 2767 (BM, TB). The parasmaipada is an apavāda of the ātmanepada.
• In kuṣyati pādaḥ svayam eva ("the foot gets pulled of itself"), pāda, the former object, is agent, and śyan stands in yak's place: kuṣ-ya-ti. In the other option, yak and taṅ: kuṣyate.
• rajyati / rajyate vastram (ours vasram).
• Only where yak would be. Elsewhere there is neither śyan nor parasmaipada, which is enjoined jointly with śyan: koṣiṣīṣṭa, raṃkṣīṣṭa (ātmanepada).
• Then [iti karmakartṛprakriyā].