Book 3 - Pada-kāṇḍa (12): Upagraha-samuddeśa (On Aspect)

3.12.1 unformalized

य आत्मनेपदाद् भेदः क्वचिदर्थस्य गम्यते ।
अन्यतश्चापि लादेशान्मन्यन्ते तमुपग्रहम् ॥ १ ॥

ya ātmanepadād bhedaḥ kvacidarthasya gamyate |
anyataścāpi lādeśānmanyante tamupagraham || 1 ||

That difference of meaning which is somewhere understood from the atmanepada — and from another substitute of 'la' as well — they consider that to be 'upagraha'.

The samuddesa's subject defined ostensively rather than by essence. Ya atmanepadad bhedah kvacid arthasya gamyate: that *difference of meaning* which is *somewhere* understood from the *atmanepada* — the middle-voice endings, which Sanskrit opposes to the parasmaipada (the 'word for another'). Anyatas ca api la-adesat: and from *another substitute of 'la'* as well — the other personal endings. Manyante tam upagraham: they *consider that* to be *upagraha*. So upagraha is not defined; it is *pointed at*: whatever semantic difference the voice-endings carry, that is it. And kvacit — *somewhere*: not always. The definition's evasiveness is honest, because the Sanskrit middle's meaning is notoriously elusive, and the samuddesa will spend twenty-seven pages showing that it has no single one.

3.12.2 unformalized

क्वचित् साधनमेवासौ क्वचित्तस्य विशेषणम् ।
साधनं तत्र कर्मादि व्यक्तवाचो विशेषणम् ॥ २ ॥

kvacit sādhanamevāsau kvacittasya viśeṣaṇam |
sādhanaṃ tatra karmādi vyaktavāco viśeṣaṇam || 2 ||

Somewhere it is the means itself; somewhere it is the qualifier of that. There the means is the object and the rest; the qualifier is of clear speech.

Upagraha's two roles. Kvacit sadhanam eva asau: *somewhere* it is the *means itself* — the voice-ending marks a karaka: the middle says the agent is also the beneficiary. Kvacit tasya visesanam: *somewhere* it is the *qualifier of that* — it does not mark a karaka but modifies one. Sadhanam tatra karmadi: there the *means* is the *object* and the rest; vyakta-vaco visesanam: the *qualifier* is *of clear speech*. So the voice-ending sometimes *is* a karaka-marker and sometimes merely colours one — two quite different offices in one morphology (3.7.13's assistance understood by the case-endings alone, now with the endings doing two jobs). The samuddesa's problem is set: one ending, two functions, and no rule saying which.

3.12.3 unformalized

क्रिया विषयभेदेन जीविकादिषु भिद्यते ।
लादेशैः स क्रियाभेदो वाक्येष्वपि नियम्यते ॥ ३ ॥

kriyā viṣayabhedena jīvikādiṣu bhidyate |
lādeśaiḥ sa kriyābhedo vākyeṣvapi niyamyate || 3 ||

The action is differentiated by difference of domain, in 'livelihood' and the rest; that difference of action is restricted, by the 'la'-substitutes, even in sentences.

Where the voice does its work. Kriya visaya-bhedena jivikadisu bhidyate: the *action* is *differentiated by difference of domain*, in *'livelihood'* and the rest — the same root means differently in different spheres: 'jiv' in the domain of livelihood is not 'jiv' elsewhere. Ladesaih sa kriya-bhedo vakyesv api niyamyate: that *difference of action* is *restricted*, by the *'la'-substitutes*, *even in sentences*. So the voice-endings restrict which domain-sense of the root is in play — they are disambiguators. This is the domain-relativity that 3.9.32 found for the seasons (one time named by what it is doing) and 2.315's fourteen contextual determinants, now carried by morphology: the ending selects the root's sphere.

3.12.4 unformalized

धात्वर्थस्तद्विशेषश्चाप्युक्तः क्वचिदुपग्रहः ।
धात्वर्थो गन्धनादिः स्याद् व्यतिहारो विशेषणम् ॥ ४ ॥

dhātvarthastadviśeṣaścāpyuktaḥ kvacidupagrahaḥ |
dhātvartho gandhanādiḥ syād vyatihāro viśeṣaṇam || 4 ||

The root's meaning, and also its particular, is somewhere called 'upagraha'; the root's meaning would be 'gandhana' and the rest; reciprocity is the qualifier.

The two things upagraha names. Dhatv-arthah tad-visesas ca api uktah kvacid upagrahah: the *root's meaning*, and also *its particular*, is *somewhere called* 'upagraha' — the term covers both the sense and its specification. Dhatv-artho gandhanadih syat: the root's meaning would be *'gandhana'* and the rest — gandhana: divulging, informing; the traditional first item in Panini's list of senses that condition the atmanepada (1.3.24ff.: the middle is used when the root means gandhana, avaksepana, sevana...). Vyatiharo visesanam: *reciprocity* is the *qualifier* — vyatihara: mutual action, another of Panini's conditions (1.3.14: karta-sthe ca). So upagraha names both the root's own sense-condition and the modification laid on it, and the samuddesa is tracking a term that its own tradition uses two ways.

3.12.5 unformalized

क्रियाप्रवृत्तावाख्याता कैश्चित् स्वार्थपरार्थता ।
असती वा सती वापि विवक्षितनिबन्धना ॥ ५ ॥

kriyāpravṛttāvākhyātā kaiścit svārthaparārthatā |
asatī vā satī vāpi vivakṣitanibandhanā || 5 ||

In the proceeding of an action, by some, being-for-oneself and being-for-another is declared — whether unreal or real, grounded in what is intended.

The middle's classical meaning, and its status. Kriya-pravrttau akhyata kaiscit sva-artha-para-arthata: in the *proceeding of an action*, by *some*, *being-for-oneself* and *being-for-another* is declared — the standard account: atmanepada means the fruit accrues to the agent, parasmaipada that it goes elsewhere. Asati va sati vapi vivaksita-nibandhana: *whether unreal or real*, *grounded in what is intended*. And there is the samuddesa's verdict in six syllables: vivaksita-nibandhana — grounded in *vivaksa*. Whether the fruit really comes to the agent is beside the point; what the voice marks is whether the speaker *means* it so. Asati va sati: real or not. So the middle does not report a fact about where benefits land; it records an intention (1.155's criterion; 3.7.2's kvacit kimcit vivaksyate).

3.12.6 unformalized

केषाञ्चित् कर्त्रभिप्राये णिचा सह विकल्पते ।
आत्मनेपदमन्येषां तदर्था प्रकृतिर्यंथा ॥ ६ ॥

keṣāñcit kartrabhiprāye ṇicā saha vikalpate |
ātmanepadamanyeṣāṃ tadarthā prakṛtiryaṃthā || 6 ||

For some, in the agent's intention, it is optional together with the causative affix; for others, the atmanepada is as the base for that purpose.

The rules' disagreement. Kesamcit kartr-abhipraye nica saha vikalpate: for *some*, in the *agent's intention*, it is *optional* together with the *causative affix* — where the agent intends the fruit for himself and a causative is present, the middle is optional. Atmanepadam anyesam tad-artha prakrtir yatha: for *others*, the *atmanepada* is *as the base* for that purpose — the middle is the base-form for that sense, not an option. So two schools: one makes the middle optional in causatives, the other makes it basic. Vikalpate — optional: and 2.85 already recorded that option is the science's last resort. The disagreement is evidence that no fact settles the middle's application — which 3.12.5 predicted, since what it marks is an intention and intentions are not read off situations.

3.12.7 unformalized

क्रीणीष्व वपते धत्ते चिनोति चिनुतेऽपि च ।
आप्तप्रयोगा दृश्यन्ते येषु ण्यर्थोऽभिधीयते ॥ ७ ॥

krīṇīṣva vapate dhatte cinoti cinute'pi ca |
āptaprayogā dṛśyante yeṣu ṇyartho'bhidhīyate || 7 ||

'Krinisva', 'vapate', 'dhatte', 'cinoti' and also 'cinute' — usages of the authoritative are seen, in which the causative's meaning is expressed.

Attested usage against the rules. Apta-prayogah drsyante: *usages of the authoritative* are *seen* — apta: the trustworthy, the sistas whose usage is the criterion (1.27's sadhutva established from the cultivated). And the list: krinisva (buy!), vapate (he sows), dhatte (he places), cinoti and cinute (he gathers, active and middle both). Yesu ny-arthah abhidhiyate: in which the *causative's meaning* is *expressed* — these forms carry a causative sense without causative morphology (3.7.73's antarbhuta-nij-artha: roots with the causative included within). And cinoti / cinute: the same root in both voices, both attested, both authoritative. So the sistas use both, and no rule distinguishes them. The verse is a small dossier of counter-examples, and its force is that usage outruns the rules (2.483's the world's practice, followed by the science).

3.12.8 unformalized

संविधानं पचादीनां क्वचिदर्थः प्रतीयते ।
तन्निमित्ता यथान्यापि क्रियाधिश्रयणादिका ॥ ८ ॥

saṃvidhānaṃ pacādīnāṃ kvacidarthaḥ pratīyate |
tannimittā yathānyāpi kriyādhiśrayaṇādikā || 8 ||

The arranging of 'paci' and the rest is somewhere understood as the meaning; occasioned by that, another action too, such as setting-on-the-fire and the like, is understood.

The middle marking the whole enterprise. Samvidhanam pacadinam kvacid arthah pratiyate: the *arranging* of *'paci'* and the rest is somewhere understood as the *meaning* — 'pacate' can mean not the cooking but the *managing* of a cooking: the householder who has the meal made. Tan-nimitta yatha anya api kriya adhisrayanadika: *occasioned by that*, *another* action too, such as *setting-on-the-fire* and the like, is understood. So the middle can shift the verb from the act to its supervision, and the subordinate acts come along. This is a real and delicate function — Sanskrit's middle does mark a kind of managerial involvement — and it is a third office beside 3.12.2's two. The samuddesa keeps finding new things the ending does.

3.12.9 unformalized

कर्त्रभिप्रायता सूत्रे क्रियाभेदोपलक्षणम् ।
तयाभूता क्रिया या हि तत्कर्ता फलभाग् यतः ॥ ९ ॥

kartrabhiprāyatā sūtre kriyābhedopalakṣaṇam |
tayābhūtā kriyā yā hi tatkartā phalabhāg yataḥ || 9 ||

In the sutra, the agent's-intention is an indication of a difference of action; for that action which is of such a kind — its agent is the one who partakes of the fruit.

The rule's key term reinterpreted. Kartr-abhiprayata sutre kriya-bheda-upalaksanam: in the *sutra*, the *agent's-intention* is an *indication of a difference of action* — Panini's condition (1.3.72: svaritanitah kartr-abhiprayekriya-phale, the middle where the fruit is intended for the agent) does not name a psychological state; it *indicates* a kind of action. Upalaksana: indication, marking-out without describing (3.10.9's man on the horse; 3.3.51's saptaparna). Taya-bhuta kriya ya hi tat-karta phala-bhag yatah: for *that* action which is *of such a kind* — its *agent* is the one who *partakes of the fruit*. So 'agent's intention' names an action-type, and the type is one whose agent gets the fruit. The reinterpretation matters: it moves the middle's condition from the speaker's mind into the action's structure — and 3.12.10 will explain how such a structure is detected.

3.12.10 unformalized

यथोपलक्ष्यते कालस्तारकादर्शनादिभिः ।
तथा फलविशेषेण क्रियाभेदो निदर्श्यते ॥ १० ॥

yathopalakṣyate kālastārakādarśanādibhiḥ |
tathā phalaviśeṣeṇa kriyābhedo nidarśyate || 10 ||

As time is indicated by the appearance of stars and the like — so a difference of action is exhibited by a particular fruit.

How 3.12.9's action-type is known, and the simile is the Kala-samuddesa's. Yatha upalaksyate kalah taraka-darsanadibhih: as *time* is *indicated* by the *appearance of stars* and the like — 3.9.44's very doctrine (the naksatras following time's operation, cihna-matram tu tarakah: the stars are a *mere mark*). Tatha phala-visesena kriya-bhedo nidarsyate: so a *difference of action* is *exhibited* by a *particular fruit*. So one reads the action's type from where its fruit lands, as one reads the hour from the stars — and in both cases the mark is not the thing. The fruit indicates the action-type; it does not constitute it. This is why 3.12.9 could relocate the middle's condition from intention to action-structure without making it a physical fact: the structure is read off a mark, and marks are marks (3.9.45's birdsong and shadow).

3.12.11 unformalized

क्रियाविशेषवचने सामर्थ्यमुपरुध्यते ।
केषाञ्चदन्ये तु कृताः स्वरितेतो ञितस्तथा ॥ ११ ॥

kriyāviśeṣavacane sāmarthyamuparudhyate |
keṣāñcadanye tu kṛtāḥ svariteto ñitastathā || 11 ||

For some, in the expressing of a particular action, the capacity is obstructed; but others are made — those with svarita-marker, and likewise those with n-marker.

The technical apparatus. Kriya-visesa-vacane samarthyam uparudhyate kesamcid: for *some*, in the *expressing of a particular action*, the *capacity* is *obstructed* — where the root already expresses the special action, the voice-ending's capacity is blocked (it has nothing left to do). Anye tu krtah svarita-itah nitas tatha: but *others are made* — those with the *svarita*-marker and those with the *n*-marker: Panini's anubandhas, the diacritic markers on roots in the Dhatupatha that determine which voice a root takes (1.3.72's svaritanitah). So the roots are pre-sorted by invisible markers, and the sorting decides the voice. The verse introduces the samuddesa's last topic: whether these markers encode a meaning or merely a class.

3.12.12 unformalized

अनुबन्धश्च सिद्धेऽर्थे स्मृत्यर्थमनुषज्यते ।
तुल्यार्थेष्वपि चावश्यं न सर्वेष्वेकधर्मता ॥ १२ ॥

anubandhaśca siddhe'rthe smṛtyarthamanuṣajyate |
tulyārtheṣvapi cāvaśyaṃ na sarveṣvekadharmatā || 12 ||

And a marker is attached, for the sake of remembrance, in a meaning already established; and even among things of like meaning, there is not necessarily one property for all.

The markers' status, and it is deflationary. Anubandhas ca siddhe 'rthe smrty-artham anusajyate: a *marker* is *attached*, *for the sake of remembrance*, in a *meaning already established* — the anubandha does not create the root's behaviour; it is a mnemonic for behaviour the usage already has (2.13's divisions for the unlearned; 3.11.6's world explaining by another's property for comprehension's sake). Tulya-arthesv api ca avasyam na sarvesv eka-dharmata: and even among things of *like meaning*, there is *not necessarily* *one property* for *all*. So roots that mean alike need not behave alike, and the markers record the behaviour rather than deriving it from the meaning. This is the samuddesa's central negative result: the voice-assignment is not a function of the sense.

3.12.13 unformalized

दृशीक्ष्योः सदृशेऽप्यर्थे नाभेदः प्रतिपूर्वयोः ।
ण्यर्थोपादायिनस्तस्मान्न तुल्यार्थाः पचादिभिः ॥ १३ ॥

dṛśīkṣyoḥ sadṛśe'pyarthe nābhedaḥ pratipūrvayoḥ |
ṇyarthopādāyinastasmānna tulyārthāḥ pacādibhiḥ || 13 ||

In 'drs' and 'iks', though the meaning is similar, there is no non-difference when preceded by 'prati'; therefore those which take up the causative's meaning are not of like meaning with 'paci' and the rest.

The evidence for 3.12.12. Drsi-iksyoh sadrse 'py arthe: in *'drs'* (to see) and *'iks'* (to look), *though the meaning is similar* — two roots, near-synonyms. Na abhedah prati-purvayoh: there is *no non-difference* when *preceded by 'prati'* — pratidrs and pratiks behave differently, though drs and iks are alike. Ny-artha-upadayinah tasman na tulya-arthah pacadibhih: *therefore* those which *take up the causative's meaning* are *not of like meaning* with 'paci' and the rest. So near-synonymous roots diverge under the same preverb, and roots with built-in causative sense are not equivalent to plain roots. The verse is fine distributional evidence (2.12's anvaya-vyatireka) that meaning does not predict behaviour — which is why the markers of 3.12.11 are needed.

3.12.14 unformalized

उम्भ्यर्थे वर्तमानस्य करोतेर्भिन्नधर्मणः ।
ण्यर्थोपादायिता तस्मान्नियताः शब्दशक्तयः ॥ १४ ॥

umbhyarthe vartamānasya karoterbhinnadharmaṇaḥ |
ṇyarthopādāyitā tasmānniyatāḥ śabdaśaktayaḥ || 14 ||

Of 'kr', standing in the meaning of 'umbh', whose properties are different — the taking-up of the causative's meaning; therefore the capacities of words are fixed.

The samuddesa's methodological conclusion, and it is the Padakanda's. Umbhy-arthe vartamanasya karoteh bhinna-dharmanah: of *'kr'* (to do), *standing in the meaning of 'umbh'* (to fill), *whose properties are different* — the same root in a borrowed sense behaves by other rules. Ny-artha-upadayita: the *taking-up of the causative's meaning*. Tasman niyatah sabda-saktayah: *therefore* the *capacities of words* are *fixed*. Niyatah — fixed: not derivable, not predictable from meaning, simply so (3.3.29's yogyata: the word-meaning fitness, beginningless, as the eye's for colour; 2.97's siddha niyata, established and fixed). So the samuddesa's answer to what the middle means is: nothing derivable. The voice-capacities are fixed as sense-capacities are fixed, and one learns them rather than deducing them. Which is why 3.12.1 could only point at upagraha and not define it.

3.12.15 unformalized

तथा ह्यनुप्रयोगस्य करोतेरात्मनेपदे ।
पूर्ववद्ग्रहणं प्राप्ते स्वरितं समुपस्थितम् ॥ १५ ॥

tathā hyanuprayogasya karoterātmanepade |
pūrvavadgrahaṇaṃ prāpte svaritaṃ samupasthitam || 15 ||

For thus, in the atmanepada of 'kr' which is a subsequent employment — the taking-up 'as before' being obtained, the svarita presented itself.

A technical case where a marker is invoked to save a rule. Anuprayogasya karoteh atmanepade: in the *atmanepada* of *'kr'* which is a *subsequent employment* — the periphrastic perfect (edham cakre), where 'kr' is added after the reduplicated stem and must take the same voice as the main root. Purvavad-grahanam prapte: the taking-up *'as before'* being *obtained* — Panini's rule that the auxiliary follows the main root's voice (1.3.63: aam pratyayavat krno 'nuprayogasya). Svaritam samupasthitam: the *svarita* *presented itself* — the marker (3.12.11's svarita-it) is what makes the rule applicable. So the anubandha does real work in the derivation, and 3.12.12's account of it as a mere mnemonic is under pressure: a mnemonic does not condition a rule. The verse is honest about the tension it creates.

3.12.16 unformalized

एकत्वेऽपि क्रियाख्याते साधनाश्रयसंख्यया ।
भिद्यते न तु लिङ्गाख्यो भेदस्तत्र तदाश्रितः ॥ १६ ॥

ekatve'pi kriyākhyāte sādhanāśrayasaṃkhyayā |
bhidyate na tu liṅgākhyo bhedastatra tadāśritaḥ || 16 ||

Even where the action stated is one, it is differentiated by the number of the means' substratum; but the difference called 'gender' is not there dependent on that.

Number and gender contrasted in their reach into the verb. Ekatve 'pi kriya-akhyate sadhana-asraya-samkhyaya bhidyate: even where the *action stated is one*, it is *differentiated* by the *number* of the *means' substratum* — the verb agrees in number with its agent: one cooking, and 'pacati' or 'pacanti' according to how many cook. Na tu linga-akhyo bhedas tatra tad-asritah: but the difference called *'gender'* is *not there dependent on that* — Sanskrit verbs do not agree in gender. So number reaches the verb and gender does not, though both are properties of the same nouns. The asymmetry is a fact about the language and it prepares the Linga-samuddesa: gender is a nominal affair, sealed off from the verb, which is why it can be (as 3.13 will argue) a property of words rather than of things.

3.12.17 unformalized

तस्मादवस्थितेऽप्यर्थे कस्यचित् प्रतिबध्यते ।
शब्दस्य शक्तिः स त्वेष शास्त्रेऽन्वख्यायते विधिः ॥ १७ ॥

tasmādavasthite'pyarthe kasyacit pratibadhyate |
śabdasya śaktiḥ sa tveṣa śāstre'nvakhyāyate vidhiḥ || 17 ||

Therefore, even where the meaning stands settled, the capacity of some word is obstructed; and that very thing is what is expounded in the science as a rule.

The samuddesa's central finding, stated as a general law. Tasmad avasthite 'py arthe kasyacit pratibadhyate sabdasya saktih: *therefore*, even where the *meaning stands settled*, the *capacity* of *some word* is *obstructed* — the meaning is there, the word could express it, and its capacity is blocked (3.9.24's jara obstructing a capacity while its contraries are released). Sa tv esa sastre 'nvakhyayate vidhih: and *that very thing* is what is *expounded in the science* as a *rule*. So the grammar's rules *are* these obstructions: a rule is a place where a word's capacity is blocked though the meaning would license it. This is a striking account of what a grammatical rule is — not a description of how meaning maps to form, but a record of where the mapping is prevented. And it follows from 3.12.13-14's finding: capacities are fixed, not derivable, so the rules record blockages rather than deriving expressions.

3.12.18 unformalized

यस्यार्थस्य प्रसिद्ध्यर्थमारभ्यन्ते पचादयः ।
तत् प्रधानं फलं तेषां न लाभादि प्रयोजनम् ॥ १८ ॥

yasyārthasya prasiddhyarthamārabhyante pacādayaḥ |
tat pradhānaṃ phalaṃ teṣāṃ na lābhādi prayojanam || 18 ||

That meaning for the accomplishing of which 'paci' and the rest are undertaken — that is their principal fruit; profit and the like is not their purpose.

The verb's fruit distinguished from the agent's motive. Yasya arthasya prasiddhy-artham arabhyante pacadayah: that *meaning* for the *accomplishing* of which *'paci'* and the rest are *undertaken*. Tat pradhanam phalam tesam: *that* is their *principal fruit* — the cooking's fruit is the cooked, and nothing else. Na labhadi prayojanam: *profit* and the like is *not* their *purpose* — the cook may cook for money, and the money is not the cooking's fruit. So the verb's phala is internal to the action and the agent's motives are external — which matters for the middle, since 3.12.9 made its condition 'the agent partakes of the fruit'. If the fruit is the cooked rice and not the wages, then the cook who cooks for hire still partakes of the fruit, and the middle should apply. The distinction is being drawn to control the next verses' hard case.

3.12.19 unformalized

यत्रौभौ स्वामिदासौ तु प्रारभेते सहक्रियाम् ।
युगपद् धर्मभेदेन धातुस्तत्र न वर्तते ॥ १९ ॥

yatraubhau svāmidāsau tu prārabhete sahakriyām |
yugapad dharmabhedena dhātustatra na vartate || 19 ||

But where both master and servant undertake a joint action — the root does not there operate simultaneously, by a difference of properties.

The hard case. Yatra ubhau svami-dasau tu prarabhete saha-kriyam: where *both* *master and servant* *undertake* a *joint action* — both are cooking, in different senses: the master by arranging (3.12.8's samvidhana), the servant by stirring. Yugapad dharma-bhedena dhatus tatra na vartate: the *root* does *not there operate* *simultaneously*, by a *difference of properties*. One verb cannot carry both relations at once (3.7.144's cloud, where contradictory properties cannot be intended simultaneously). So 'they cook' cannot mean master-arranging and servant-stirring in one breath. The case is well chosen: it is exactly where the middle's fruit-condition (3.12.9) becomes undecidable, since both partake and differently. And 3.12.20 gives the resolution.

3.12.20 unformalized

यत्र प्रतिविधानार्थः पचिस्तत्रात्मनेपदम् ।
परस्मैपदमन्यत्र संस्कराद्यभिधायिनि ॥ २० ॥

yatra pratividhānārthaḥ pacistatrātmanepadam |
parasmaipadamanyatra saṃskarādyabhidhāyini || 20 ||

Where 'paci' has the meaning of arranging, there the atmanepada; the parasmaipada elsewhere, when it expresses the refinement and the rest.

The resolution of 3.12.19, and it splits the verb. Yatra prati-vidhana-arthah pacih: where *'paci'* has the meaning of *arranging* — the master's sense (3.12.8's samvidhana). Tatra atmanepadam: *there* the *atmanepada*. Parasmaipadam anyatra samskaradyabhidhayini: the *parasmaipada* *elsewhere*, when it expresses the *refinement* and the rest — the servant's sense, the actual working of the food. So the voice sorts the two senses: the manager takes the middle, the labourer the active. And this is a real fact about Sanskrit — the middle does carry a shade of self-involved supervision — given here its precise statement. The verse resolves 3.12.19 by denying that one verb must serve both: two senses, two voices, and the joint action described twice.

3.12.21 unformalized

संविधातुश्च सान्निध्यात् दासे धर्मोऽनुषज्यते ।
प्लक्षशब्दस्य सान्निध्यान्न्यग्रोधे प्लक्षता यथा ॥ २१ ॥

saṃvidhātuśca sānnidhyāt dāse dharmo'nuṣajyate |
plakṣaśabdasya sānnidhyānnyagrodhe plakṣatā yathā || 21 ||

And from the proximity of the arranger, the property is attached to the servant — as, from the proximity of the word 'plaksa', plaksa-ness attaches to the nyagrodha.

How the master's property leaks onto the servant. Samvidhatus ca sannidhyat dase dharmo 'nusajyate: from the *proximity of the arranger*, the *property* is *attached* to the *servant* — the servant, working beside the master, takes on the master's arranging-property, and so the middle can be used of him too. And the simile: plaksa-sabdasya sannidhyan nyagrodhe plaksata yatha — as, from the *proximity of the word 'plaksa'*, *plaksa-ness* attaches to the *nyagrodha*: two fig-trees named together, and the first's name colours the second (the grammarians' standing example of ekasesa-adjacent contamination). So properties leak by proximity, in trees and in households alike. This is 2.299's lamp lighting what it was not brought for, and 3.11.13's likeness to the occasioning ground: what stands near takes on what it does not have.

3.12.22 unformalized

पुरोडाशाभिधानं च धानादिषु यथा स्थितम् ।
छत्रिणा चाभिसंबन्धाच्छत्रिशब्दाभिधेयता ॥ २२ ॥

puroḍāśābhidhānaṃ ca dhānādiṣu yathā sthitam |
chatriṇā cābhisaṃbandhācchatriśabdābhidheyatā || 22 ||

And the designation 'purodasa' is as it stands in respect of the grains and the rest; and from connection with an umbrella-bearer, there is the state of being expressible by the word 'umbrella-bearer'.

Two more cases of leaking designation. Purodasa-abhidhanam ca dhanadisu yatha sthitam: the designation *'purodasa'* (the sacrificial cake) is as it *stands* in respect of the *grains* and the rest — the name of the finished cake is applied to the grains it will be made from (3.7.7's janyamana karman: the object that does not exist until the action completes, held in place by buddhi). Chatrina ca abhisambandhat chatri-sabda-abhidheyata: and from *connection with an umbrella-bearer*, there is the state of being expressible by the word *'umbrella-bearer'* — the classic chatrino gacchanti: 'the umbrella-bearers go', said of a group in which one has an umbrella. So the name leaks from the future thing to the present stuff, and from one man to his company. Both are 3.12.21's mechanism, and both are ordinary speech.

3.12.23 unformalized

अर्थात् प्रतीतमन्योन्यं पारार्थ्यमविवक्षितम् ।
इत्ययं शेषविषयः कैश्चिदत्रानुवर्ण्यते ॥ २३ ॥

arthāt pratītamanyonyaṃ pārārthyamavivakṣitam |
ityayaṃ śeṣaviṣayaḥ kaiścidatrānuvarṇyate || 23 ||

Mutual being-for-another, understood from the meaning, is unintended; thus this is described by some as the domain of the 'remainder'.

The relation of master and servant placed grammatically. Arthat pratitam anyonyam para-arthyam avivaksitam: *mutual being-for-another*, *understood from the meaning*, is *unintended* — each works for the other (the servant for the master's meal, the master for the servant's wage), and the mutuality is not what anyone means to say. Avivaksita: unintended (1.136's canon; 3.1.51's forced categories). Ity ayam sesa-visayah kaiscid atra anuvarnyate: thus *this* is described by *some* as the *domain of the 'remainder'* — the sesa, the genitive's territory, where relations stand that no karaka covers (3.7.44's sasthi sese). So the master-servant relation is a sesa-relation: real, understood, unintended, and expressed by the genitive if at all. It is not a karaka and it does not condition the voice.

3.12.24 unformalized

अथ प्रतिविधाता यो हलैः कृषति पञ्चभिः ।
भाष्ये नोहाहृतं कस्मात् प्राप्तं तत्रात्मनेपदम् ॥ २४ ॥

atha pratividhātā yo halaiḥ kṛṣati pañcabhiḥ |
bhāṣye nohāhṛtaṃ kasmāt prāptaṃ tatrātmanepadam || 24 ||

Now, that arranger who ploughs with five ploughs — why has it not been cited in the Bhasya? The atmanepada is obtained there.

An objection from the Bhasya's silence. Atha prati-vidhata yo halaih krsati pancabhih: the *arranger* who *ploughs* with *five ploughs* — a man who manages five ploughmen: an arranger by 3.12.8's criterion, and 'krsati' is active. Bhasye na udahrtam kasmat: *why has it not been cited* in the *Bhasya*? Praptam tatra atmanepadam: the *atmanepada* is *obtained* there — by 3.12.20's rule, the arranging sense should take the middle, and Patanjali does not list this case. So either the rule is wrong or the Bhasya is incomplete. The objection is from authority's silence, and it is a real move in this tradition: what the Bhasya does not say, it means not to say (2.475's bhasye eva upavarnitah, where the citation was positive). Bhartrhari takes the silence seriously enough to answer it.

3.12.25 unformalized

प्रतीतत्वात् तदर्थस्य शेषत्वं यदि कल्प्यते ।
न स्यात् प्राप्तविभाषासौ स्वरितेतां निवर्तिका ॥ २५ ॥

pratītatvāt tadarthasya śeṣatvaṃ yadi kalpyate |
na syāt prāptavibhāṣāsau svaritetāṃ nivartikā || 25 ||

If, from the meaning's being understood, remainder-hood is constructed — that option, being already obtained, would not be the excluder of those having a svarita-marker.

The first answer to 3.12.24, and it is rejected. Pratitatvat tad-arthasya sesatvam yadi kalpyate: if, from the *meaning's being understood*, *remainder-hood* is *constructed* — if one says the ploughing-arrangement is a sesa-relation (3.12.23) and so does not condition the voice. Na syat prapta-vibhasa asau svarita-itam nivartika: that *option*, being *already obtained*, would *not be* the *excluder* of those having a *svarita-marker*. The reasoning: if the sense were sesa, no option would arise at all, and the rule providing an option for svarita-marked roots would have nothing to exclude. Prapta-vibhasa: an option on what is already obtained — the technical category. So the sesa-answer proves too much: it removes the option the grammar provides, and grammars do not provide vacuous options (2.443's Panini does not legislate vacuously).

3.12.26 unformalized

शुद्धे तु संविधानार्थे कैश्चिदत्रेष्यते कृषिः ।
तद्धर्मा यजिरित्येवं न स्यात्तत्रात्मनेपदम् ॥ २६ ॥

śuddhe tu saṃvidhānārthe kaiścidatreṣyate kṛṣiḥ |
taddharmā yajirityevaṃ na syāttatrātmanepadam || 26 ||

But in the pure meaning of arranging, 'krs' is accepted here by some; the atmanepada would not be there, as with 'yaji' whose property is that.

The second answer. Suddhe tu samvidhana-arthe kaiscid atra isyate krsih: in the *pure* meaning of *arranging*, *'krs'* is accepted here by *some* — the ploughing taken as pure management, no working. Tad-dharma yajir iti evam na syat tatra atmanepadam: the *atmanepada* would *not be* there, as with *'yaji'* whose *property is that* — 'yaj' (to sacrifice) is a root whose sense is inherently managerial (the sacrificer arranges; the priests act) and which nonetheless takes the middle by its marker, not by its sense. So if 'krs' in the pure arranging sense behaved like 'yaj', it would take the middle — and it does not. Therefore the arranging sense alone does not confer the middle, and 3.12.20's rule needs a further condition. The answer is offered and the verse's structure leaves it standing as a report.

3.12.27 unformalized

अत्र तूपपदेनायमर्थभेदः प्रतीयते ।
प्राप्ते विभाषा क्रियते तस्मान्नास्त्यात्मनेपदम् ॥ २७ ॥

atra tūpapadenāyamarthabhedaḥ pratīyate |
prāpte vibhāṣā kriyate tasmānnāstyātmanepadam || 27 ||

But here this difference of meaning is understood by the upapada; the option is made on what is already obtained — therefore there is no atmanepada.

The samuddesa's resolution of 3.12.24, and it is grammatical rather than semantic. Atra tu upapadena ayam artha-bhedah pratiyate: *here* this *difference of meaning* is understood by the *upapada* — the accompanying word ('with five ploughs') carries the arranging-sense, not the root. Prapte vibhasa kriyate: the *option* is made on what is *already obtained*. Tasman na asti atmanepadam: *therefore* there is *no atmanepada*. So the ploughman's case takes the active because the arranging-sense comes from the upapada and not from 'krs' itself — and the voice follows the root's own capacity (3.12.14's niyatah sabda-saktayah: the capacities of words are fixed), not the sentence's sense. This is the samuddesa's thesis holding under pressure: the voice is a property of the word, fixed, not derivable from what the sentence means. The Bhasya's silence is thus explained: there was no case to cite.