Book 3 - Pada-kāṇḍa (14): Vṛtti-samuddeśa (On Ccomplex Formation)

3.14.1 unformalized

कुत्साप्रशंसातिशयैः समाप्तार्थं तु युज्यते ।
पदं स्वार्थदयः सर्वे यस्मात् कुत्सादिहेतवः ॥ १ ॥

kutsāpraśaṃsātiśayaiḥ samāptārthaṃ tu yujyate |
padaṃ svārthadayaḥ sarve yasmāt kutsādihetavaḥ || 1 ||

But a word of completed meaning is joined with censure, praise and excess; since all, beginning with its own meaning, are the causes of censure and the rest.

The samuddesa opens on the affixes of attitude. Kutsa-prasamsa-atisayaih: with *censure*, *praise* and *excess* — the taddhitas that add an evaluation: the ka of contempt, the affixes of praise, the tarap and tamap of comparison. Samapta-artham tu yujyate padam: a word of *completed meaning* is *joined* with them — samapta-artha: the word is already whole before the attitude arrives (3.8.17's krtarthatva; 2.4's ekartham). Svarthadayah sarve yasmat kutsadi-hetavah: since *all*, beginning with its *own meaning*, are the *causes* of censure and the rest. So what occasions the contempt is everything about the word's meaning — and the affix does not add a content but registers a stance toward one. The Vrtti-samuddesa's whole subject is what happens when words are joined; it begins with the simplest case, where a completed word takes on an attitude, and even here the question is what the addition adds.

3.14.2 unformalized

देवदत्तादिकुत्सायां वर्तते कुत्सितश्रुतिः ।
कुत्सितस्था तु या कुत्सा तदर्थः को विधीयते ॥ २ ॥

devadattādikutsāyāṃ vartate kutsitaśrutiḥ |
kutsitasthā tu yā kutsā tadarthaḥ ko vidhīyate || 2 ||

The hearing 'censured' operates in the censure of Devadatta and the rest; but that censure which resides in the censured — for that meaning, 'ka' is prescribed.

Two ways of censuring, and they are not the same. Devadattadi-kutsayam vartate kutsita-srutih: the *hearing 'censured'* operates in the *censure of Devadatta* and the rest — one says 'kutsito devadattah', 'Devadatta is contemptible', and the word 'kutsita' names him as censured. Kutsita-stha tu ya kutsa tad-arthah kah vidhiyate: but that *censure* which *resides in the censured* — for *that* meaning, *'ka'* is *prescribed*: 'devadattakah', with the contemptuous suffix, expresses not that he is censured but the censure itself, lodged in him. So: the adjective predicates contempt, the affix carries it. And the difference is the samuddesa's opening lesson about vrtti — what a compressed form does is not what its analysis says.

3.14.3 unformalized

प्रकृष्ट इति शुक्लादिप्रकर्षस्याभिधायकः ।
प्रकृष्टस्य प्रकर्षे तु तरबादिर्विधीयते ॥ ३ ॥

prakṛṣṭa iti śuklādiprakarṣasyābhidhāyakaḥ |
prakṛṣṭasya prakarṣe tu tarabādirvidhīyate || 3 ||

'Prakrsta' is the expressor of the excellence of white and the rest; but in the excellence of the excellent, 'tarap' and the rest is prescribed.

The comparative's structure, and it is doubled. Prakrsta iti sukladi-prakarsasya abhidhayakah: *'prakrsta'* (excellent) is the *expressor* of the *excellence* of white and the rest — the word names the eminence a quality has (3.5.2's prakarsa; 1.64's guna as prakarsa-hetu). Prakrstasya prakarse tu tarab-adir vidhiyate: but in the *excellence of the excellent*, *tarap* and the rest is *prescribed* — 'suklatara', whiter: the comparative marks not excellence but excellence *of what is already excellent*. So the comparative affix operates one level up, on an eminence already predicated. This is the stratification 3.11.19 found in the higher numerals (the countings gathered) and 3.11.3's number numbering itself — and it explains why 'suklatara' is not merely a stronger 'sukla': the affix takes the excellence as its object.

3.14.4 unformalized

कुत्सितत्वेन कुत्स्यो वा न सम्यग् वापि कुत्सितः ।
स्वशब्दाभिहिते केन विशिष्टोऽर्थः प्रतीयते ॥ ४ ॥

kutsitatvena kutsyo vā na samyag vāpi kutsitaḥ |
svaśabdābhihite kena viśiṣṭo'rthaḥ pratīyate || 4 ||

Either what is to be censured by way of being censured, or what is not properly censured — when it is expressed by its own word, by what is a qualified meaning understood?

A dilemma for the contempt-affix. Kutsitatvena kutsyo va: either what is *to be censured* *by way of being censured* — the affix censures what is already censurable, and adds nothing; na samyag vapi kutsitah: or what is *not properly censured* — the affix censures what does not deserve it, and misfires. Sva-sabda-abhihite kena visisto 'rthah pratiyate: when it is expressed by its *own word*, *by what* is a *qualified* meaning understood? If 'kutsita' already says he is contemptible, what does '-ka' qualify? The dilemma is 3.14.2's distinction pressed: if the adjective and the affix both censure, one is idle (2.34's add-or-subtract test: against the sentence's one meaning, addition yields not more meaning but meaninglessness).

3.14.5 unformalized

न च साम्प्रतिकी कुत्सा भेदाभावात् प्रतीयते ।
पूज्यते कुत्सितत्वेन प्रशस्तत्वेन कुत्स्यते ॥ ५ ॥

na ca sāmpratikī kutsā bhedābhāvāt pratīyate |
pūjyate kutsitatvena praśastatvena kutsyate || 5 ||

And a present censure is not understood, from the absence of difference; one is honoured by way of being censured, and censured by way of being praised.

The dilemma of 3.14.4 answered by an observation about irony, and it is one of the finest things in the samuddesa. Na ca samprataki kutsa bheda-abhavat pratiyate: a *present* censure is *not understood*, from the *absence of difference* — the affix does not deliver a censure occurring now, because nothing in the form marks it off. And then the evidence: pujyate kutsitatvena — one is *honoured* *by way of being censured*: the affectionate insult, the friend called a scoundrel; prasastatvena kutsyate — and *censured* *by way of being praised*: the sneer that says 'how admirable'. So the contempt-affix can honour and the praise-affix can wound, and neither carries its attitude reliably. Which answers 3.14.4: the affix does not censure at all; it marks a stance whose direction the occasion supplies (2.315's fourteen determinants; 3.12.5's vivaksita-nibandhana). Irony is not a deviant use — it is what shows that the attitude was never in the form.

3.14.6 unformalized

विशेषणविशेष्यत्वं पदयोरुपजायते ।
व प्रातिपदिकार्थश्च तत्रैवं व्यतिरिच्यते ॥ ६ ॥

viśeṣaṇaviśeṣyatvaṃ padayorupajāyate |
va prātipadikārthaśca tatraivaṃ vyatiricyate || 6 ||

The state of being qualifier and qualified arises for two words; and the nominal stem's meaning is there thus set apart.

The samuddesa turns to its real subject: what happens when two words are joined. Visesana-visesyatvam padayor upajayate: the state of being *qualifier and qualified* *arises* for *two words* — upajayate: arises, comes to be. Neither word is a qualifier by itself; the relation is generated by the joining. Va pratipadika-arthas ca tatra evam vyatiricyate: and the *nominal stem's meaning* is *there* thus *set apart* — the stem's own meaning is separated off from the relational role it acquires. So joining two words produces two things: a qualifier-qualified structure that neither had, and a distinction between what each stem means and what it is doing. This is 3.7.13's doctrine (the karaka-property is not expressed by the noun) at the level of nominal composition, and it opens the whole vrtti-analysis.

3.14.7 unformalized

विशेष्यं स्यादनिर्ज्ञातं निर्ज्ञातार्थो विशेषणम् ।
परार्थत्वेन शेषत्वं सर्वेषामुपकारिणाम् ॥ ७ ॥

viśeṣyaṃ syādanirjñātaṃ nirjñātārtho viśeṣaṇam |
parārthatvena śeṣatvaṃ sarveṣāmupakāriṇām || 7 ||

The qualified would be what is not ascertained; the qualifier is what has an ascertained meaning. From being-for-another, remainder-hood belongs to all that assist.

The two roles defined by epistemic status. Visesyam syad anirjnatam: the *qualified* would be what is *not ascertained* — the head is what one does not yet know: 'gauh' leaves open which cow. Nirjnata-artho visesanam: the *qualifier* is what has an *ascertained* meaning — the modifier is what one knows: 'sukla' brings a settled content to an unsettled one. So the qualifier-qualified relation is not a matter of importance or of grammatical rank but of *what is known*: the known narrows the unknown. And the consequence: para-arthatvena sesatvam sarvesam upakarinam — from *being-for-another*, *remainder-hood* belongs to *all that assist* (2.113's parartha; 3.7.44's sesa). Everything that helps is a remainder — which means the qualifier, for all its determinacy, is the subordinate.

3.14.8 unformalized

विभक्तिभदो नियमाद् गुणगुण्यभिधायिनोः ।
सामानाधिकरणस्य प्रसिध्दिर्द्रव्यशब्दयोः ॥ ८ ॥

vibhaktibhado niyamād guṇaguṇyabhidhāyinoḥ |
sāmānādhikaraṇasya prasidhdirdravyaśabdayoḥ || 8 ||

A difference of case-ending is by rule for the two expressing quality and quality-possessor; the establishing of co-referentiality is for two substance-words.

The morphology of qualification. Vibhakti-bhedo niyamad guna-guny-abhidhayinoh: a *difference of case-ending* is *by rule* for the two expressing *quality* and *quality-possessor* — where one word names a quality as such and the other its bearer, the cases differ: 'suklasya gauh' would be the genitive construction, quality and possessor at different case-levels. Samanadhikaranyasya prasiddhir dravya-sabdayoh: the *establishing of co-referentiality* is for two *substance*-words — samanadhikaranya: same-locus-hood, the grammar's term for two words in the same case referring to one thing: 'suklo gauh', both nominative. So co-reference requires both words to be substance-words, and where one is a pure quality-word the cases split. The rule matters because it means 'sukla' in 'suklo gauh' is not naming a quality at all — it is naming a substance, whitely (3.14.9's example makes this vivid).

3.14.9 unformalized

द्रव्येऽनिर्ज्ञातजातीये कृष्णशब्दः प्रयुज्यते ।
अनिर्ज्ञातगुणे चैव तिलशब्दः प्रवर्तते ॥ ९ ॥

dravye'nirjñātajātīye kṛṣṇaśabdaḥ prayujyate |
anirjñātaguṇe caiva tilaśabdaḥ pravartate || 9 ||

The word 'krsna' is employed with respect to a substance whose universal is unascertained; and the word 'tila' operates with respect to one whose quality is unascertained.

The mechanism of co-reference, and the example is exact. Take 'krsnas tilah', the black sesame. Dravye anirjnata-jatiye krsna-sabdah prayujyate: the word *'krsna'* (black) is employed with respect to a substance *whose universal is unascertained* — 'black' gives you a black *thing* and does not say what kind: the jati is open. Anirjnata-gune ca eva tila-sabdah pravartate: and the word *'tila'* operates with respect to one *whose quality is unascertained* — 'sesame' gives you the kind and leaves the colour open. So each word presents a substance with a gap, and the gaps are complementary: black-something and sesame-of-some-colour. Neither is a qualifier by nature (3.14.6's upajayate: the roles arise in the joining); each fills what the other left open. This is the finest analysis of adjectival co-reference in the ancient world, and it depends on 3.14.7's insight that qualification is the known narrowing the unknown.

3.14.10 unformalized

सामान्यानामसम्बन्धात् तौ विशेषेव्यवस्थितौ ।
रूपाभेदाद् विशेषं तमभिव्यङ्क्तुं न शक्नुतः ॥ १० ॥

sāmānyānāmasambandhāt tau viśeṣevyavasthitau |
rūpābhedād viśeṣaṃ tamabhivyaṅktuṃ na śaknutaḥ || 10 ||

From the non-connection of the generals, those two stand established in the particular; from non-difference of form, they cannot manifest that particular.

Why neither word alone suffices. Samanyanam asambandhat: from the *non-connection of the generals* — blackness-in-general and sesame-hood-in-general are not connected; nothing links the colour-universal to the species-universal. Tau visese vyavasthitau: those two stand established *in the particular* — they meet only in this black sesame seed, not in their generals. Rupa-abhedad visesam tam abhivyanktum na saknutah: from *non-difference of form*, they *cannot manifest* that *particular* — each word's form is the same whether it means the general or this instance, so neither can single out the particular by itself. So the two words meet in a particular that neither can reach alone — and the meeting is what the joining accomplishes. The verse explains why qualification is necessary: generals do not connect, particulars are not namable, and only the crossing of two open generals lands on one thing.

3.14.11 unformalized

तावेवं सन्निपतितौ भेदेन प्रतिपादने ।
अवच्छेदमिवाधाय संशयं व्यपकर्षतः ॥ ११ ॥

tāvevaṃ sannipatitau bhedena pratipādane |
avacchedamivādhāya saṃśayaṃ vyapakarṣataḥ || 11 ||

Those two, thus fallen together, in conveying by way of difference — having as it were deposited a delimitation, draw away the doubt.

The resolution, and its verbs are precise. Tau evam sannipatitau: those two, thus *fallen together* — sannipata: the falling-together, collision, conjunction (3.7.109's sannipata). The two words simply land in one utterance. Bhedena pratipadane: in *conveying by way of difference* — each conveying its own content, differently. Avacchedam iva adhaya samsayam vyapakarsatah: having *as it were deposited* a *delimitation*, they *draw away the doubt*. Two operations: each deposits a limit on the other's openness (avaccheda: delimitation, 2.45's avacchinatti, the relation delimiting the hovering general), and together they *draw away the doubt* — samsayam vyapakarsatah: the ambiguity is dragged off. So qualification is doubt-removal by mutual delimitation, and neither word does anything to the other but stand beside it. The iva is careful: they deposit a delimitation *as it were* — nothing is really added; two openings simply overlap.

3.14.12-14 unformalized

द्रव्यात्मा गुणसंसर्गभेदादाश्रीयते पृथक् ।
जातिस्म्बन्धभेदाच्च द्वितीय इव गुह्यते ॥ १२ ॥
निमित्तैरभिसम्बन्धाद् या निमित्तसरूपता ।
तयैकस्यापि नानात्वं रूपाभेदात् प्रकल्पते ॥ १३ ॥
द्रव्यावस्था तृतीया तु यस्यां संसृज्यते द्वयम् ।
तयोरेवस्थयोर्भेदादाश्रयत्वे नियुज्यते ॥ १४ ॥

dravyātmā guṇasaṃsargabhedādāśrīyate pṛthak |
jātismbandhabhedācca dvitīya iva guhyate || 12 ||
nimittairabhisambandhād yā nimittasarūpatā |
tayaikasyāpi nānātvaṃ rūpābhedāt prakalpate || 13 ||
dravyāvasthā tṛtīyā tu yasyāṃ saṃsṛjyate dvayam |
tayorevasthayorbhedādāśrayatve niyujyate || 14 ||

(12) The substance-self is resorted to severally from a difference of quality-commingling; and from a difference of universal-connection, it is grasped as though a second. (13) That likeness to the occasioning ground which comes from connection with the occasioning grounds — by it, even of one thing, manifoldness is constructed, from non-difference of form. (14) But there is a third substance-state, in which the two are commingled; from the difference of those two states, it is applied to the state of being substratum.

The metaphysics beneath 3.14.9-11's analysis, and it makes one thing into three. (12): dravya-atma guna-samsarga-bhedad asriyate prthak — the *substance-self* is resorted to *severally* from a *difference of quality-commingling*: the sesame taken as black is one resort; jati-sambandha-bhedac ca dvitiya iva grhyate — and from a *difference of universal-connection*, it is grasped *as though a second*: the same seed taken as sesame is another. One thing, two resorts, and the second only *as though*. (13) names the mechanism: nimittaih abhisambandhad ya nimitta-sarupata — the *likeness to the occasioning ground* that comes from connection with the grounds (3.11.13 verbatim) — taya ekasya api nanatvam rupa-abhedat prakalpyate: by it, even of *one* thing, *manifoldness* is *constructed*, from *non-difference of form*. (14) supplies the third: dravya-avastha trtiya tu yasyam samsrjyate dvayam — a *third* substance-state, in which *the two are commingled* — tayor avasthayor bhedad asrayatve niyujyate: from the *difference* of those two states, it is *applied* to the state of being *substratum*. So: the thing as black, the thing as sesame, and the thing as the substratum in which both meet — three states of one seed, and the third is what the co-referential phrase names. The whole apparatus of qualification rests on a threefold construction of one thing by buddhi, which 3.14.15 is about to say outright.

3.14.15 unformalized

बुद्धयैकं भिद्यते भिन्नमेकत्वं चोपगच्छति ।
बुद्धयावस्था विभज्यन्ते सा ह्यर्थस्य विधायिका ॥ १५ ॥

buddhayaikaṃ bhidyate bhinnamekatvaṃ copagacchati |
buddhayāvasthā vibhajyante sā hyarthasya vidhāyikā || 15 ||

By the intellect one is differentiated, and the differentiated attains oneness; by the intellect the states are divided out — for it is the ordainer of the meaning.

The samuddesa's thesis, and it is the Padakanda's. Buddhya ekam bhidyate: by the *intellect*, *one* is *differentiated* — the sesame becomes black-thing and sesame-thing. Bhinnam ekatvam ca upagacchati: and the *differentiated* attains *oneness* — and the two become one seed again, in the phrase. So buddhi both divides the one and unifies the divided, and the whole machinery of 3.14.9-14 is its work. Buddhy-avastha vibhajyante: by the intellect the *states are divided out* — 3.14.14's three states are buddhi's (3.7.104's ekasya buddhy-avasthabhih bhede parikalpite: objecthood, instrument-hood and agency arising when the intellect's states carve one thing). And the last clause is the verse's weight: sa hi arthasya vidhayika — for *it* is the *ordainer of the meaning*. Vidhayika: the ordainer, the one who prescribes (3.11.12's vidhayinim: number ordaining difference among things that have transcended it). The intellect does not find the meaning's structure; it legislates it. Everything the Vrtti-samuddesa will say about compounds and derivatives rests here.

3.14.16 unformalized

व्यपदेशिवदेकस्मिन् बुद्धया नानार्थकल्पना ।
तया कल्पितभेदः सन्नर्थात्मा व्यपदिश्यते ॥ १६ ॥

vyapadeśivadekasmin buddhayā nānārthakalpanā |
tayā kalpitabhedaḥ sannarthātmā vyapadiśyate || 16 ||

As in the 'vyapadesivat' maxim, there is a construction of many meanings by the intellect in one thing; by that, the meaning-self, its difference constructed, is designated.

The doctrine of 3.14.15 given its Paninian precedent. Vyapadesivat: Panini's own metarule (vyapadesivad ekasmin — 'as though there were a designation, in respect of one'), which allows a single item to be treated as though it were two so that a rule requiring two can apply. The grammar has an explicit licence to double one thing. Ekasmin buddhya nanartha-kalpana: in *one* thing, a *construction of many meanings* by the *intellect*. Taya kalpita-bhedah sann artha-atma vyapadisyate: by that, the *meaning-self*, its *difference constructed*, is *designated*. So the sastra's own maxim is the samuddesa's thesis in miniature: buddhi doubles what is single, the doubling is kalpita, and the designation proceeds on the doubled. Bhartrhari is not importing a philosophy into the grammar; he is reading the philosophy the grammar's metarules already presuppose.

3.14.17 unformalized

क्रियाभेदेन दृष्टानामश्मादिनां पुनः पुनः ।
किञ्चिद् दर्शनमन्येन दर्शनेनापदिश्यते ॥ १७ ॥

kriyābhedena dṛṣṭānāmaśmādināṃ punaḥ punaḥ |
kiñcid darśanamanyena darśanenāpadiśyate || 17 ||

Of stones and the like, seen again and again through a difference of action — some one seeing is designated by another seeing.

The doubling of 3.14.16 in perception. Kriya-bhedena drstanam asmadinam punah punah: of *stones* and the like, *seen again and again* through a *difference of action* — the same stone met in different doings: lifted, thrown, stumbled over. Kimcid darsanam anyena darsanena apadisyate: *some one seeing* is *designated by another seeing*. So one identifies a present sighting by reference to a previous one — 'the stone I threw' — and the identification treats two seeings as one object's. The verse supplies the perceptual case of vyapadesivat: buddhi doubles the stone into seen-then and seen-now, and then re-identifies them. And this is what makes qualification possible (3.14.9-11): the two open generals that meet in one particular are two takings of one thing.

3.14.18 unformalized

प्रयोगभेदाद् धातूनां प्रकल्प्य बहुरूपताम् ।
भेदाभेदावुपादाय क्वचिदेकाच्त्वमुच्यते ॥ १८ ॥

prayogabhedād dhātūnāṃ prakalpya bahurūpatām |
bhedābhedāvupādāya kvacidekāctvamucyate || 18 ||

Having constructed the many-formedness of roots from a difference of employment — taking up difference and non-difference, mono-vocalic-ness is somewhere stated.

The doubling in the grammar's own root-list. Prayoga-bhedad dhatunam prakalpya bahu-rupatam: having *constructed* the *many-formedness* of *roots* from a *difference of employment* — one root, employed variously, is treated as several (the Dhatupatha's homonymous entries). Bheda-abhedav upadaya kvacid ekac-tvam ucyate: *taking up difference and non-difference*, *mono-vocalic-ness* is *somewhere stated* — ekac: having one vowel, a Paninian condition for certain rules. So a root is counted as one-vowelled or not according to which of its constructed forms is in view, and the counting takes up both difference and non-difference at once (3.7.179's bhedabheda-abhyupagame na virodho 'sti kascana: in accepting both difference and non-difference there is no contradiction). The grammar doubles its roots and then counts their vowels — and which count applies depends on the doubling.

3.14.19 unformalized

अन्वयव्यतिरेकाभ्यामर्थवान् परिकल्पितः ।
एको धात्वर्थविगमाद् वर्णत्वेनोपचर्यते ॥ १९ ॥

anvayavyatirekābhyāmarthavān parikalpitaḥ |
eko dhātvarthavigamād varṇatvenopacaryate || 19 ||

One thing, constructed as meaningful by agreement and difference, from the departure of the root's meaning is figuratively treated as a mere phoneme.

The limit of the doubling. Anvaya-vyatirekabhyam arthavan parikalpitah: *constructed as meaningful* by *agreement and difference* — the distributional method posits a meaningful element (2.12's anvaya-vyatirekau vyavahara-nibandhanam). Ekah dhatv-artha-vigamad varnatvena upacaryate: *one*, from the *departure of the root's meaning*, is *figuratively treated* as a *mere phoneme*. So an element posited as meaningful, once its root-sense withdraws, is demoted to a bare sound — upacaryate: figuratively (2.253's upacara). This is 2.167's nut and sap (formatives real as units and empty as meanings) given their genesis: they are meaningful posits whose meaning has departed, and the grammar keeps the unit and drops the sense. The doubling of 3.14.18 thus has a terminus: past a point, what analysis isolates is a phoneme.

3.14.20 unformalized

द्रव्यात्मानस्त्रयस्तस्माद् बद्धौ नाना व्यवस्थिताः ।
आश्रयाश्रयिधर्मेणत्ययं पूर्वेभ्य आगमः ॥ २० ॥

dravyātmānastrayastasmād baddhau nānā vyavasthitāḥ |
āśrayāśrayidharmeṇatyayaṃ pūrvebhya āgamaḥ || 20 ||

Therefore three substance-selves stand established as many in the intellect, by the property of substratum and what rests on it — this is the tradition from the earlier ones.

The three states of 3.14.12-14 confirmed on authority. Dravya-atmanas trayas tasmad buddhau nana vyavasthitah: *therefore* *three* substance-selves stand established *as many* *in the intellect* — the thing as qualified, the thing as kind, and the thing as their substratum (3.14.14's third state). Asraya-asrayi-dharmena: by the *property of substratum and what rests on it* — the three are generated by the bearer/borne relation applied to one thing. Ity ayam purvebhya agamah: *this* is the *tradition from the earlier ones* — purvebhya agamah: 3.3.11's very phrase, and 2.485's counsel about waiting on the elders. So the threefold construction is inherited, not invented, and Bhartrhari cites rather than claims it. The Padakanda's boldest analytical apparatus is presented as received.

3.14.21 unformalized

सामानाधिकरण्यं च शब्दयोः कैश्चिदिष्यते ।
विशेषणविशेष्यत्वं संज्ञासंज्ञित्वमेव वा ॥ २१ ॥

sāmānādhikaraṇyaṃ ca śabdayoḥ kaiścidiṣyate |
viśeṣaṇaviśeṣyatvaṃ saṃjñāsaṃjñitvameva vā || 21 ||

And by some, the co-referentiality of two words is accepted; or the state of qualifier and qualified; or else the state of name and named.

Three accounts of what joins two words. Samanadhikaranyam ca sabdayoh kaiscid isyate: by *some*, the *co-referentiality* of two words — samanadhikaranya: same-locus-hood (3.14.8), the two words pointing at one thing. Visesana-visesyatvam: or the state of *qualifier and qualified* (3.14.6-7's relation, arising in the joining). Samjna-samjnitvam eva va: or else the state of *name and named* — the two words as designation and designatum. Three relations for one juxtaposition, and they are not equivalent: co-reference is symmetric, qualification is not, and naming is a third thing again. The verse's catalogue is doing what the samuddesa's catalogues always do — showing that the joining admits several descriptions and that none is privileged (2.179's licensed double description).

3.14.22 unformalized

केषाञ्चिज्जातिगुणयोरेकार्थसमवेतयोः ।
वृत्तिः कृष्णतिलेष्विष्टा शब्दे द्रव्याभिधायिनि ॥ २२ ॥

keṣāñcijjātiguṇayorekārthasamavetayoḥ |
vṛttiḥ kṛṣṇatileṣviṣṭā śabde dravyābhidhāyini || 22 ||

For some, the derivative formation of a universal and a quality inhering in one object is accepted, in 'black sesames', where the word expresses a substance.

The compound's condition, on one account. Kesamcij jati-gunayor eka-artha-samavetayoh: for *some*, of a *universal and a quality* *inhering in one object* — sesame-hood and blackness, both in the seed (eka-artha-samavaya: the Vaisesika's device for relating two qualities through their common bearer, 3.3.13). Vrttih krsna-tilesu ista: the *derivative formation* is accepted in *'black sesames'*. Sabde dravya-abhidhayini: where the *word expresses a substance*. So compounding is licensed when a universal and a quality share a bearer and the word names the bearer — and the condition is precise: both members must reach the same substance, or there is nothing for the compound to be about. This is 3.14.9-10's analysis (the two open generals meeting in a particular) stated as a rule of composition.

3.14.23 unformalized

संस्तु रूपरसादीनामाश्रयो नाभिधीयते ।
द्रव्याभिधानेन विना ततस्ते द्वन्द्वभाविनः ॥ २३ ॥

saṃstu rūparasādīnāmāśrayo nābhidhīyate |
dravyābhidhānena vinā tataste dvandvabhāvinaḥ || 23 ||

But the substratum of form, taste and the rest, being existent, is not expressed; without the expressing of a substance, they are therefore constituents of a dvandva.

The contrasting case. San tu rupa-rasadinam asrayo na abhidhiyate: the *substratum* of *form*, *taste* and the rest, *being existent*, is *not expressed* — the mango's colour and flavour have a bearer, and 'rupa-rasau' does not name it. Dravya-abhidhanena vina tatas te dvandva-bhavinah: *without* the *expressing of a substance*, they are *therefore* *constituents of a dvandva*. So where no substance is named, the two quality-words cannot form a qualifier-qualified compound (3.14.22's condition fails) and fall into a dvandva instead — a mere coordination, 'colour-and-taste'. The rule is elegant: what determines the compound-type is whether a substance is expressed, and the qualities, left without a bearer, can only be listed. This explains why quality-words coordinate and do not qualify each other.

3.14.24 unformalized

द्रव्याभिधायी कृष्णादिराकाङ्क्षावान् प्रवर्तते ।
निमित्तानुविधायित्वात् तत् तिलादौ न विद्यते ॥ २४ ॥

dravyābhidhāyī kṛṣṇādirākāṅkṣāvān pravartate |
nimittānuvidhāyitvāt tat tilādau na vidyate || 24 ||

'Krsna' and the rest, expressing a substance, operates as possessing expectancy; from its conforming to the occasioning ground, that does not exist in 'tila' and the rest.

The asymmetry between the two members of 'krsnas tilah'. Dravya-abhidhayi krsnadir akanksavan pravartate: *'krsna'* and the rest, *expressing a substance*, operates as *possessing expectancy* — 'black' names a substance and *wants* to know which: it is incomplete (3.14.7's visesyam anirjnatam, inverted — here the qualifier is the needy one). Nimitta-anuvidhayitvat: from its *conforming to the occasioning ground* — the colour-word follows whatever bears the colour. Tat tiladau na vidyate: *that* does *not exist* in *'tila'* and the rest — 'sesame' has no such expectancy; it names its kind and is content. So the two words are not symmetric after all: 'krsna' reaches for a bearer and 'tila' does not, and the asymmetry is what makes one the qualifier. This refines 3.14.9-11: the mutual delimitation is real, and the reaching is one-way.

3.14.25 unformalized

एवं जातिमति द्रव्ये प्रत्यासन्ने क्रियां प्रति ।
गुणधर्म गुणाविष्ट द्रव्यं भेदाय कल्पते ॥ २५ ॥

evaṃ jātimati dravye pratyāsanne kriyāṃ prati |
guṇadharma guṇāviṣṭa dravyaṃ bhedāya kalpate || 25 ||

Thus, in a substance possessing a universal, which is near to the action — a substance pervaded by a quality, whose property is the quality, serves for differentiation.

The qualification's office in the sentence. Evam jatimati dravye pratyasanne kriyam prati: in a substance *possessing a universal*, which is *near to the action* — the sesame, ready to figure as a karaka (3.7.94's proximity to the accomplishment). Guna-dharma gunavista dravyam bhedaya kalpate: a *substance pervaded by a quality*, whose *property is the quality*, serves for *differentiation*. So 'krsna' contributes a quality-pervaded substance, and its office is to differentiate the sesame from other sesames — which is what an action needs, since actions take determinate participants (3.7.87's ipsitama: the existing thing not yet attained, taken under the aspect it will have). The verse locates qualification's purpose: it is for the verb, and it hands the verb a narrowed participant.

3.14.26 unformalized

गुणमात्राभिधायित्वं केचिदिच्छन्ति वृत्तिषु ।
अताश्वादिषु सम्बन्धाद् रूढीनामिव रूढिभिः ॥ २६ ॥

guṇamātrābhidhāyitvaṃ kecidicchanti vṛttiṣu |
atāśvādiṣu sambandhād rūḍhīnāmiva rūḍhibhiḥ || 26 ||

Some desire the state of expressing a mere quality in the derivative formations; therefore, in 'asva' and the rest, from connection, it is as of conventional terms by conventional terms.

A rival account of the compound's members. Guna-matra-abhidhayitvam kecid icchanti vrttisu: *some* desire the state of expressing a *mere quality* in the *derivative formations* — in the compound, they say, 'krsna' names blackness itself, not a black substance (against 3.14.8's rule that co-reference requires substance-words). Atah asvadisu sambandhad rudhinam iva rudhibhih: *therefore*, in *'asva'* and the rest, from *connection*, it is as of *conventional terms* by *conventional terms* — the compound's members relate as two rudhi-words do: by convention, not by qualification. So on this view compounds are not qualifications at all; they are conventional pairings (2.37's rudhi-sabdatva acknowledged in the science itself). The account is reported and its cost is visible: it makes every compound an idiom.

3.14.27 unformalized

तिले पूर्वमुपात्ते वा तत्रैव मतुविष्यते ।
स च धर्मः समासेषु गुणस्तस्माद् विशेषणम् ॥ २७ ॥

tile pūrvamupātte vā tatraiva matuviṣyate |
sa ca dharmaḥ samāseṣu guṇastasmād viśeṣaṇam || 27 ||

Or, 'tila' having been taken up first, the 'matup' affix is desired there itself; and that property, in compounds, is the quality — therefore it is the qualifier.

A third account, and it makes the compound a possessive. Tile purvam upatte va: or, *'tila'* having been *taken up first* — the substance-word leads. Tatra eva matub isyate: the *'matup'* affix is desired *there itself* — matup: the possessive suffix ('having'), so that 'krsna-tila' is analysed as 'sesame having blackness'. Sa ca dharmah samasesu gunas tasmad visesanam: and *that property*, in *compounds*, is the *quality* — *therefore* it is the *qualifier*. So the compound conceals a possessive: the substance is the head, the quality is possessed, and the possession makes the quality a qualifier. The account is neat and it explains the asymmetry 3.14.24 found — but it requires an unuttered matup, and 2.219's bahuvrihi already showed the samuddesa's suspicion of members that are not there.

3.14.28 unformalized

अनुस्यूतेव भेदाभ्यामेका प्रख्योपजायते ।
यदा सहविवक्षां तामाहुर्द्वन्द्वैकशेषयोः ॥ २८ ॥

anusyūteva bhedābhyāmekā prakhyopajāyate |
yadā sahavivakṣāṃ tāmāhurdvandvaikaśeṣayoḥ || 28 ||

One presentation arises, as though interwoven by the two differences; when that co-intention is meant, they call it 'dvandva' and 'ekasesa'.

The coordinative compounds, and their mechanism is a weaving. Anusyuta iva bhedabhyam eka prakhya upajayate: *one presentation* arises, *as though interwoven* by the two *differences* — anusyuta: 1.123's very word (all knowledge appears as if interwoven, anuviddham iva, with word), the weaver's term. So two differences are threaded into one showing, and the iva marks it. Yada saha-vivaksam tam ahur dvandva-ekasesayoh: when that *co-intention* is meant, they call it *'dvandva'* and *'ekasesa'*. Saha-vivaksa — co-intention: meaning two things *together*. So the coordinative compounds are what happens when two things are meant at once rather than one through the other — and the difference between qualification and coordination is a difference of vivaksa (3.7.2's kvacit kimcit vivaksyate), not of the things.

3.14.29 unformalized

इतरेतरयोगस्तु भिन्नसङ्घाभिधायिनाम् ।
प्रत्येकं च समृहोऽसौ समूहिषु समाप्यते ॥ २९ ॥

itaretarayogastu bhinnasaṅghābhidhāyinām |
pratyekaṃ ca samṛho'sau samūhiṣu samāpyate || 29 ||

But mutual conjunction belongs to those expressing distinct groups; and that collection is completed severally, in the members of the collection.

The dvandva's structure. Itaretara-yogas tu bhinna-sangha-abhidhayinam: *mutual conjunction* belongs to those expressing *distinct groups* — itaretara-yoga: the grammar's name for the dvandva whose members remain distinct ('rama-krsnau', two men, both there). Pratyekam ca samuho 'sau samuhisu samapyate: and that *collection* is *completed severally*, in the *members* of the collection — pratyekam samapyate: the whole-at-each formula (2.18, 3.8.5). So the dvandva's collection is complete in each member: the pairhood is wholly in Rama and wholly in Krsna, not divided between them. This is why the dvandva takes a dual and not two singulars — and why 2.221-225's analysis of distribution applied to it: the collection is one and its completing is severally.

3.14.30 unformalized

व्यापारसमुदायस्य यथाधिश्रयणादिषु ।
प्रत्येकं जातिवद् वृत्तिस्तथा द्वन्द्वपदेष्वपि ॥ ३० ॥

vyāpārasamudāyasya yathādhiśrayaṇādiṣu |
pratyekaṃ jātivad vṛttistathā dvandvapadeṣvapi || 30 ||

As the derivative formation of an aggregate of operations is severally, like a universal, in the setting-on-the-fire and the rest — so also in the words of a dvandva.

The dvandva's structure paralleled with the action's. Vyapara-samudayasya yatha adhisrayanadisu: as the *aggregate of operations* — in *setting-on-the-fire* and the rest: the cooking's sub-actions (3.8.9's parts of 'paci'). Pratyekam jativad vrttih: the *derivative formation* is *severally*, *like a universal* — the cooking is wholly in the fetching and wholly in the stirring, as cowhood is wholly in each cow (3.1.25's jati manifested in each instance). Tatha dvandva-padesv api: *so also* in the *words of a dvandva*. So the dvandva is to its members what an action is to its sub-actions and a universal to its instances: wholly present in each, not divided among them. The parallel unifies three of the kanda's analyses — and it is the reason 3.14.29's collection completes severally.

3.14.31 unformalized

शौण्डार्धर्चपुरोडशच्छत्रिणोऽत्र निदर्शनम् ।
ते विष्णुमित्रा इति च भिन्नेषु सहचारिषु ॥ ३१ ॥

śauṇḍārdharcapuroḍaśacchatriṇo'tra nidarśanam |
te viṣṇumitrā iti ca bhinneṣu sahacāriṣu || 31 ||

'Saunda', 'ardharca', 'purodasa', 'chatrin' are the illustrations here; and 'they are the Visnumitras' — among distinct accompanying things.

A class of cases where a name spreads to its neighbours. Four specimens: saunda (the drunkard's class); ardharca (the gana of 3.1.103, words whose analysis wavers); purodasa (3.12.22's sacrificial cake, whose name is applied to the grains it will be made from); chatrin (3.12.22's umbrella-bearer, whose designation covers his umbrella-less companions). And a fifth: te visnumitra iti — '*they are the Visnumitras*', a family called by one member's name. Bhinnesu sahacarisu: among *distinct accompanying things* — in every case the things are genuinely distinct and one name covers them. So the samuddesa opens its analysis of composition by assembling the cases where designation already leaks across differences (3.12.21's plaksa-ness attaching to the nyagrodha). Compounding will turn out to be this leak, regularized.

3.14.32 unformalized

अर्थान्तराभिदायित्वं तथार्थान्तरवर्त्तिनाम् ।
याभ्यां चैकमनेकार्थं ताभ्यामेवापरं पदम् ॥ ३२ ॥

arthāntarābhidāyitvaṃ tathārthāntaravarttinām |
yābhyāṃ caikamanekārthaṃ tābhyāmevāparaṃ padam || 32 ||

The state of expressing another meaning belongs likewise to those operating in another meaning; and by which two a single one has many meanings, by those very two another word has them.

The leak of 3.14.31 generalized. Arthantara-abhidayitvam tatha arthantara-vartinam: the state of *expressing another meaning* belongs likewise to those *operating in another meaning* — a word that has strayed into a neighbour's sense goes on expressing that stray sense. Yabhyam ca ekam anekartham tabhyam eva aparam padam: and by *which two* a *single one* has *many meanings*, by *those very two* *another word* has them. So whatever mechanism multiplies one word's senses multiplies another's — the leak is systematic, not idiosyncratic. This is 2.250-251's omni-significance (one word, many meanings, carved by nimittas) applied to composition: the conditions that carve one word carve its neighbour, and the compound is where two such carvings meet.

3.14.33 unformalized

समुदायान्तरत्वाच्च तादृशोऽर्थ न लौकिकः ।
अन्वयव्यतिरेकाभ्यां शास्त्रार्थोऽपि न दृश्यते ॥ ३३ ॥

samudāyāntaratvācca tādṛśo'rtha na laukikaḥ |
anvayavyatirekābhyāṃ śāstrārtho'pi na dṛśyate || 33 ||

And from its being another aggregate, such a meaning is not worldly; nor is a scientific meaning seen by agreement and difference.

The compound's meaning placed nowhere. Samudayantaratvat ca tadrso 'rtha na laukikah: from its being *another aggregate*, such a meaning is *not worldly* — the compound is a new aggregate, and its meaning is not one the world already had (2.208's disputed emergence: from meaningful elements a complex arises with a distinct meaning). Anvaya-vyatirekabhyam sastrartho 'pi na drsyate: nor is a *scientific meaning* seen by *agreement and difference* — the distributional method (2.12's tool) does not find it either: vary the members and the compound's sense does not track. So the compound's meaning is neither in the world's inventory nor recoverable by the analyst's method. It is a genuine novelty, and the samuddesa is honest that its own instrument cannot locate it.

3.14.34 unformalized

दुःखा दुरुपपादा च तस्माद् भाष्येऽप्युदाहृता ।
युगपद्वाचिता सा तु व्यवहारार्थमाश्रिता ॥ ३४ ॥

duḥkhā durupapādā ca tasmād bhāṣye'pyudāhṛtā |
yugapadvācitā sā tu vyavahārārthamāśritā || 34 ||

Therefore it is cited in the Bhasya as difficult and hard to establish; but that simultaneous-expressing-ness is resorted to for the sake of practice.

The difficulty acknowledged on authority. Duhkha durupapada ca tasmad bhasye 'py udahrta: *therefore* it is cited in the *Bhasya* as *difficult* and *hard to establish* — Patanjali himself found it so, and Bhartrhari cites the admission rather than papering over it (2.475's bhasye eva upavarnitah; 3.13.16's Bhasya-citation). Yugapad-vacita sa tu vyavahara-artham asrita: but that *simultaneous-expressing-ness* is *resorted to* *for the sake of practice*. Yugapad-vacita: the compound's expressing its members' senses *at once* — the doctrine that is hard to establish and indispensable. Vyavahara-artham asrita: resorted to for practice's sake. So the account is adopted because the transactions need it, not because it can be defended (3.3.82's beings raised up by construction, as though by a primary word-meaning). The samuddesa's honesty here is exact: this is a working posit.

3.14.35 unformalized

समुदायमुपक्रम्य पदं तस्यां प्रयुज्यते ।
विभागेन समाख्याने ततस्तद् द्वयर्थमुच्यते ॥ ३५ ॥

samudāyamupakramya padaṃ tasyāṃ prayujyate |
vibhāgena samākhyāne tatastad dvayarthamucyate || 35 ||

Having commenced with the aggregate, the word is employed in that; in the designation by division, it is therefore said to have two meanings.

How the two-meaning account arises. Samudayam upakramya padam tasyam prayujyate: having *commenced with the aggregate*, the word is *employed in that* — one starts from the compound as a whole and uses it whole. Vibhagena samakhyane tatah tad dvy-artham ucyate: in the *designation by division*, *therefore*, it is said to have *two meanings*. So the two meanings appear only when one divides — the analysis produces them, and the usage never had them. This is 2.10's apoddhara exactly (words extracted from sentences as suffixes from words) at the compound's grain: the members are what division yields, and 3.14.34's yugapad-vacita is what one must say once one has divided. The samuddesa is tracking its own instrument's artifacts.

3.14.36 unformalized

वीक्येऽपि नियता धर्माः केचिद् वृत्तौ द्वयोस्तथा ।
ते त्वभेदेन सामर्थ्यमात्र एवोपवर्णिताः ॥ ३६ ॥

vīkye'pi niyatā dharmāḥ kecid vṛttau dvayostathā |
te tvabhedena sāmarthyamātra evopavarṇitāḥ || 36 ||

Even in a sentence certain properties are fixed, and likewise in a derivative formation, for the two; but they have been described, without difference, as mere capacity.

Sentence and compound compared. Vakye 'pi niyata dharmah kecid: even in a *sentence* certain *properties are fixed* — the analytic phrase has its own settled behaviour. Vrttau dvayos tatha: and likewise in a *derivative formation*, for the *two* — the compound has its own. So both have fixed properties, and they differ. Te tv abhedena samarthya-matra eva upavarnitah: but they have been described, *without difference*, as *mere capacity* — samarthya-matra: capacity alone, the condition Panini states for compounding (samarthah padavidhih, 2.1.1). So the grammar's single condition — capacity — covers two things with different properties, and does not distinguish them. The verse identifies the gap the samuddesa must fill: what makes capacity issue in a compound rather than a phrase?

3.14.37 unformalized

वृत्तौ विशेषवृत्तित्वाद् भेदे सामान्यवाचिता ।
उपमानसमासादौ श्यामादीनमुदाहृता ॥ ३७ ॥

vṛttau viśeṣavṛttitvād bhede sāmānyavācitā |
upamānasamāsādau śyāmādīnamudāhṛtā || 37 ||

In a derivative formation, from its operating in the particular; in separation, the state of expressing the general — this has been illustrated for 'syama' and the rest, in the comparison-compound and so on.

The samuddesa's first substantive difference between compound and phrase. Vrttau visesa-vrttitvat: in a *derivative formation*, from its *operating in the particular* — inside a compound, a member means a *particular*. Bhede samanya-vacita: in *separation*, the state of expressing the *general* — the same word, standing alone, means the general. And the illustration: upamana-samasadau syamadinam — for *'syama'* (dark) and the rest, in the *comparison-compound*: 'syama' alone means darkness-in-general; in 'megha-syama' (dark as a cloud) it means this particular shade. So composition narrows: the member drops from general to particular by being compounded. This is a real and delicate observation, and it explains 3.14.33's finding — the compound's meaning is not in the world's inventory because the world's inventory has the generals, and the compound has particulars.

3.14.38 unformalized

वृत्तिरन्यपदार्थे या तस्या वाक्येष्वसंभवः ।
चार्थे द्वन्द्वपदानां च भेदे वृत्तिर्न विद्यते ॥ ३८ ॥

vṛttiranyapadārthe yā tasyā vākyeṣvasaṃbhavaḥ |
cārthe dvandvapadānāṃ ca bhede vṛttirna vidyate || 38 ||

That derivative formation which is in another's word-meaning — of that there is no possibility in sentences; and of dvandva-words in the sense of 'ca', there is no formation in separation.

Two compounds that no sentence can render. Vrttir anya-padarthe ya tasya vakyesv asambhavah: the formation which is in *another's word-meaning* — the bahuvrihi, whose meaning is a possessor named by no member (2.219's aprayukta-padas ca artho bahuvrihau: the meaning whose word is not employed) — of *that* there is *no possibility in sentences*: no analytic phrase says what 'citragu' says, because the phrase must name the possessor and the compound does not. Ca-arthe dvandva-padanam ca bhede vrttir na vidyate: and of *dvandva*-words in the sense of *'ca'*, there is no formation *in separation* — the dvandva's togetherness (3.14.28's saha-vivaksa) is lost the moment one says 'and'. So two compound-types have no sentential equivalent at all, and the paraphrase-method (3.7.77's bheda-vakya, disowned) fails for them outright. Composition can say what syntax cannot.

3.14.39 unformalized

भेदे सति निरादीनां क्रान्ताद्यर्थेष्वसंभवः ।
प्राग्वृत्तेर्जातिवाचित्वं न च गौरखरादिषु ॥ ३९ ॥

bhede sati nirādīnāṃ krāntādyartheṣvasaṃbhavaḥ |
prāgvṛtterjātivācitvaṃ na ca gaurakharādiṣu || 39 ||

When there is separation, there is no possibility of 'nir' and the rest in the senses of 'having gone beyond' and so on; and before the formation, there is no expressing of the universal in 'gaura-khara' and the rest.

Two more cases where separation destroys the sense. Bhede sati niradinam krantady-arthesv asambhavah: when there is *separation*, there is *no possibility* of *'nir'* and the rest in the senses of *'having gone beyond'* — the preverb's transcending-sense exists only in composition; 'nir' standing alone means nothing of the kind (3.7.190's obligatory preverbs, whose roots are meaningless apart from them). Prag vrtteh jati-vacitvam na ca gaura-kharadisu: and *before the formation*, there is no *expressing of the universal* in *'gaura-khara'* and the rest — the compound names a kind that its members, uncompounded, do not reach. So the sense arrives *with* the composition and is not in the pieces beforehand. This is 2.36's asvakarna-principle (the compound understood whole, its members idle) given its temporal form: there was no meaning to compound.

3.14.40 unformalized

क्रीडायां जीविकायां च वाक्येनावचनात् तथा ।
न नित्यग्रहणं युक्तं कौटिल्ये यङ्विधौ यथा ॥ ४० ॥

krīḍāyāṃ jīvikāyāṃ ca vākyenāvacanāt tathā |
na nityagrahaṇaṃ yuktaṃ kauṭilye yaṅvidhau yathā || 40 ||

And in 'play' and 'livelihood', from non-expression by a sentence — the taking-up of 'nitya' is not proper, as in the yan-rule for 'crookedness'.

A rule-technical consequence. Kridayam jivikayam ca vakyena avacanat: in *'play'* and *'livelihood'* — Panini's compound-classes for names of games and occupations — from *non-expression by a sentence*: no phrase renders them. Na nitya-grahanam yuktam: the taking-up of *'nitya'* is *not proper* — the grammar need not call these compounds 'obligatory' (nitya-samasa: the class where no analysis exists), because their obligatoriness follows from there being nothing to analyse into. Kautilye yan-vidhau yatha: as in the *yan*-rule for *'crookedness'* — the intensive affix, likewise needing no obligatoriness-marker. So the category 'obligatory compound' is redundant: where no sentence expresses the sense, the compound is obligatory by that fact alone, and stipulating it is otiose (2.443's Panini does not legislate vacuously).

3.14.41 unformalized

निर्धारणादिविषये व्यपेक्षैव यतः स्थिता ।
समासप्रतिषेधानां ततो नास्ति प्रयोजनम् ॥ ४१ ॥

nirdhāraṇādiviṣaye vyapekṣaiva yataḥ sthitā |
samāsapratiṣedhānāṃ tato nāsti prayojanam || 41 ||

Since in the domain of determination and the rest, mutual expectancy alone stands — therefore the prohibitions of compounding have no purpose.

The converse redundancy. Nirdharanadi-visaye vyapeksa eva yatah sthita: since in the domain of *determination* (3.7.147's nirdharana: singling out from a class) and the rest, *mutual expectancy alone* stands — vyapeksa: the mutual reaching of separate words, which is what a *sentence* has (2.4's sakanksa-avayavam). Samasa-pratisedhanam tato na asti prayojanam: *therefore* the *prohibitions of compounding* have *no purpose*. Where only vyapeksa is available, no compound could form, and forbidding it is idle. So the pair 3.14.40-41 finds redundancy at both ends: obligatory-compound rules where nothing else is possible, prohibition-rules where compounding is impossible anyway. And both redundancies point at the same distinction — vyapeksa versus something else — which 3.14.44 will name.

3.14.42 unformalized

विधिभिः प्रतिषेधैश्च भेदाभेदनिदर्शनम् ।
कृतं द्वन्द्वैकवद्भावे सङ्घवृत्त्युपदेशवत् ॥ ४२ ॥

vidhibhiḥ pratiṣedhaiśca bhedābhedanidarśanam |
kṛtaṃ dvandvaikavadbhāve saṅghavṛttyupadeśavat || 42 ||

By injunctions and prohibitions, an exhibition of difference and non-difference has been made — in the dvandva's singular-treatment, as in the teaching of the group-formation.

What the redundant rules of 3.14.40-41 are actually doing. Vidhibhih pratisedhais ca bheda-abheda-nidarsanam krtam: by *injunctions and prohibitions*, an *exhibition of difference and non-difference* has been *made* — the rules that seemed pointless are *exhibiting* something: where compounding is enjoined, non-difference; where prohibited, difference. Dvandva-ekavad-bhave: in the *dvandva's singular-treatment* — the rule making certain dvandvas take a singular ('pani-padam', hand-and-foot as one neuter): non-difference exhibited. Sangha-vrtty-upadesavat: as in the *teaching of the group-formation*. So the rules are nidarsana — exhibitions, teaching-devices (2.13's divisions for the avidvan; 3.7.147's separations for the unlearned). They do not legislate the facts; they display the difference-and-non-difference that composition involves.

3.14.43 unformalized

सामर्थ्यमविशेषोक्तमपि लोकव्यवस्थया ।
वृत्त्यवृत्त्योः प्रयोगज्ञैर्विभक्तं प्रतिपत्तृभिः ॥ ४३ ॥

sāmarthyamaviśeṣoktamapi lokavyavasthayā |
vṛttyavṛttyoḥ prayogajñairvibhaktaṃ pratipattṛbhiḥ || 43 ||

Capacity, though stated without specification, has been divided — by the world's settled arrangement — into formation and non-formation, by comprehenders who know usage.

The gap of 3.14.36 filled, and the filler is usage. Samarthyam avisesa-uktam api: *capacity*, though stated *without specification* — Panini's samarthah padavidhih says only 'capacity', and does not say when capacity yields a compound. Loka-vyavasthaya: by the *world's settled arrangement*. Vrtty-avrttyoh prayoga-jnaih vibhaktam pratipattrbhih: divided into *formation and non-formation*, by *comprehenders* who *know usage* — prayoga-jna: those who know how it is said. So the grammar leaves the distinction unstated and the sistas supply it: what compounds and what does not is settled by usage, and the rule's silence is not a gap but a deference (2.483's the world's practice, followed by the science; 3.13.21's cultivated declaring the gender).

3.14.44 unformalized

अर्थस्य विनिवृत्तत्वाल्लुगादि न विरुध्यते ।
एकार्थीभाव एवातः समासाख्या विधीयते ॥ ४४ ॥

arthasya vinivṛttatvāllugādi na virudhyate |
ekārthībhāva evātaḥ samāsākhyā vidhīyate || 44 ||

From the meaning's having withdrawn, elision and the rest is not contradicted; therefore the designation 'compound' is prescribed in becoming-one-meaning alone.

The samuddesa's central definition, and it is one of the great sentences of Indian grammar. Arthasya vinivrttatvat lug-adi na virudhyate: from the *meaning's having withdrawn*, *elision* and the rest is *not contradicted* — inside a compound, the members' meanings have *withdrawn* (vinivrtta: 2.240's withdrawing meanings; 3.12.27's jahat-svartha, abandoning one's own meaning), and so the grammar may elide their case-endings without violating anything: there is no meaning left for the endings to express. Ekarthibhava eva atah samasa-akhya vidhiyate: *therefore* the designation *'compound'* is prescribed in *ekarthibhava* alone. Ekarthibhava: *becoming-one-meaning* — the compound is not two meanings joined but a becoming-one, and that is what distinguishes it from the sentence's vyapeksa (3.14.41's mutual expectancy). So: sentence = two meanings reaching for each other; compound = two meanings become one. And the morphology follows: endings elide because their meanings are gone.

3.14.45 unformalized

व्यवस्थितविभाषा च सामान्ये कैश्चिदिष्यते ।
तथा वाक्यं व्यपेक्षायां समासोऽन्यत्र शिष्यते ॥ ४५ ॥

vyavasthitavibhāṣā ca sāmānye kaiścidiṣyate |
tathā vākyaṃ vyapekṣāyāṃ samāso'nyatra śiṣyate || 45 ||

And a settled option in the general is accepted by some; thus a sentence in mutual expectancy, a compound elsewhere, is taught.

The distinction of 3.14.44 given its rule-form. Vyavasthita-vibhasa ca samanye kaiscid isyate: a *settled option* in the *general* is accepted by *some* — vyavasthita-vibhasa: the grammar's category for an option that is not free but fixed by domain (some cases take one, some the other, and no case takes either). Tatha vakyam vyapeksayam samaso 'nyatra sisyate: thus a *sentence* in *mutual expectancy*, a *compound* *elsewhere*, is *taught*. So the option is settled: where vyapeksa obtains, a phrase; where ekarthibhava, a compound; and the speaker does not choose. This is why 3.14.40-41's rules were redundant — the domains do not overlap, and the option is settled by which relation obtains. The verse turns 3.14.44's definition into the grammar's operative rule.

3.14.46 unformalized

तुल्यश्रुतित्वात् तत्त्वेऽपि राजादीनामुपाश्रिते ।
वृत्तौ विशेषणाकाङ्क्षाऽगमकत्वान्निवर्तते ॥ ४६ ॥

tulyaśrutitvāt tattve'pi rājādīnāmupāśrite |
vṛttau viśeṣaṇākāṅkṣā'gamakatvānnivartate || 46 ||

From likeness of hearing, even in the reality, when 'rajan' and the rest is resorted to — in a derivative formation, the expectancy for a qualifier withdraws, from its not being a conveyor.

The consequence of ekarthibhava for the members' behaviour. Tulya-srutitvat tattve 'pi rajadinam upasrite: from *likeness of hearing*, even in the *reality*, when *'rajan'* and the rest is *resorted to* — the word 'rajan' sounds the same inside 'rajapurusa' as outside it, and the same thing is meant. Vrttau visesana-akanksa agamakatvat nivartate: in a *derivative formation*, the *expectancy for a qualifier* *withdraws*, from its *not being a conveyor*. So inside the compound, 'rajan' stops wanting a qualifier — it is agamaka, not a conveyor: it no longer conveys anything on its own, so nothing can qualify it. This is 3.14.44's ekarthibhava in its effects: the members' meanings having withdrawn, their expectancies withdraw with them, and the compound admits no internal modification. The rule is real — one cannot insert an adjective into a compound — and here it is derived rather than stipulated.

3.14.47 unformalized

संबन्धिशब्दः सापेक्षो नित्यं सर्वः प्रयुज्यते ।
स्वार्थवत् सा व्यपेक्षास्य वृत्तावपि न हीयते ॥ ४७ ॥

saṃbandhiśabdaḥ sāpekṣo nityaṃ sarvaḥ prayujyate |
svārthavat sā vyapekṣāsya vṛttāvapi na hīyate || 47 ||

A relatum-word, being dependent, is always employed as such, all of it; that mutual expectancy of it, like its own meaning, is not lost even in a derivative formation.

The great exception to 3.14.46, and it is a real one. Sambandhi-sabdah sapekso nityam sarvah prayujyate: a *relatum-word* — one whose very sense is relational: 'father', 'master', 'enemy' — is *always* employed as *dependent*, *all of it*: there is no such thing as a father who is nobody's. Sva-arthavat sa vyapeksa asya vrttav api na hiyate: that *mutual expectancy* of it, *like its own meaning*, is *not lost* even in a *derivative formation*. So where 3.14.46 said the members' expectancies withdraw inside the compound, the relatum-words keep theirs — because their expectancy *is* their meaning: strip 'father' of its reaching-for-a-child and nothing is left to compound. The exception is principled and it marks the limit of ekarthibhava: meanings can withdraw, and a meaning that is nothing but a reaching cannot withdraw without vanishing.

3.14.48 unformalized

समुदायेन संबन्धो येषां गुरुकुलादिना ।
संस्पृश्यावयवांस्ते च युज्यन्ते तद्वता सह ॥ ४८ ॥

samudāyena saṃbandho yeṣāṃ gurukulādinā |
saṃspṛśyāvayavāṃste ca yujyante tadvatā saha || 48 ||

Those whose connection is with an aggregate — with 'guru-kula' and the like — having touched the members, are joined with the possessor of that.

How relatum-words reach through a compound. Samudayena sambandho yesam guru-kuladina: those whose connection is with an *aggregate* — with *'guru-kula'* (the teacher's house, the school) and the like: the compound names a whole. Samsprsya avayavan te ca yujyante tadvata saha: *having touched the members*, they are *joined with the possessor of that*. So a relatum-word connecting to 'guru-kula' touches the members — the teacher, the house — and then joins with what possesses the aggregate. The reaching passes *through* the compound's members without stopping at them. This is why 3.14.47's exception does not break the compound: the relatum-word's expectancy is satisfied by the whole, having grazed the parts. And samsprsya — having touched — is exact: it does not rest there.

3.14.49 unformalized

अबुधान् प्रत्युपायाश्च विचित्राः प्रतिपत्तये ।
शब्दान्तरत्वादत्यन्तभेदो वाक्यसमासयोः ॥ ४९ ॥

abudhān pratyupāyāśca vicitrāḥ pratipattaye |
śabdāntaratvādatyantabhedo vākyasamāsayoḥ || 49 ||

And the means, for the unlearned, are various, for comprehension's sake; from being another word, the difference of sentence and compound is utter.

The samuddesa's methodological reminder, placed where it is needed. Abudhan praty upayas ca vicitrah pratipattaye: the *means*, for the *unlearned*, are *various*, *for comprehension's sake* — the analyses, the paraphrases, the exhibitions of 3.14.42: all upaya, all discardable (2.38's things taken up to be let go; 2.13's divisions for the avidvan). And then the flat statement: sabdantaratvat atyanta-bhedo vakya-samasayoh — from *being another word*, the difference of *sentence and compound* is *utter*. Atyanta-bheda: utterly different. 'Rajnah purusah' and 'rajapurusah' are two words, not one word twice (2.177's rule: different meaning-behaviour, different base; 3.7.77's paraphrase having no contact with the root). So the paraphrase is a ladder and the compound is not the phrase — and everything said by analysing the compound into a phrase is said about the ladder.

3.14.50 unformalized

असमासे समासे च गोरथादिष्वदर्शनात् ।
युक्तादीनां न शास्त्रेण निवृत्त्यनुगमः कृतः ॥ ५० ॥

asamāse samāse ca gorathādiṣvadarśanāt |
yuktādīnāṃ na śāstreṇa nivṛttyanugamaḥ kṛtaḥ || 50 ||

From the non-appearance of 'yukta' and the rest in 'go-ratha' and the like, both in the non-compound and in the compound — no following of withdrawal has been made by the science.

A rule-technical confirmation. Asamase samase ca go-rathadisv adarsanat: from the *non-appearance* of *'yukta'* and the rest in *'go-ratha'* (ox-cart) and the like, *both in the non-compound and in the compound* — the word 'yukta' (yoked) appears in neither: not in 'gobhir yuktah rathah' as a required element, nor in 'go-rathah'. Yuktadinam na sastrena nivrtty-anugamah krtah: *no following of withdrawal* has been *made* by the *science* — the grammar does not track 'yukta' as something that withdrew in compounding, because it was never there. So 'go-ratha' does not mean 'cart yoked with oxen' by suppressing 'yoked'; it means what it means, and the yoking is the hearer's supplement (2.163's penumbra supplied by association). The verse blocks the temptation to read compounds as ellipses.

3.14.51 unformalized

शब्दान्तरत्वाद् युक्तादिः क्वचिद् वाक्ये प्रयुज्यते ।
प्रपर्णप्रपलाशादौ गतशब्दश्च वृत्तिषु ॥ ५१ ॥

śabdāntaratvād yuktādiḥ kvacid vākye prayujyate |
praparṇaprapalāśādau gataśabdaśca vṛttiṣu || 51 ||

From being another word, 'yukta' and the rest is somewhere employed in a sentence; and the word 'gata' in the derivative formations, in 'praparna', 'prapalasa' and so on.

The complement of 3.14.50. Sabdantaratvad yuktadih kvacid vakye prayujyate: from being *another word*, 'yukta' and the rest is *somewhere employed in a sentence* — where one wants the yoking said, one says it, and the sentence is a different word from the compound (3.14.49's atyanta-bheda). Praparna-prapalasadau gata-sabdas ca vrttisu: and the word *'gata'* (gone) in the *derivative formations*, in *'praparna'* (leaf-fallen), *'prapalasa'* and so on — here the compound *does* carry a 'gone' sense that no member names. So: sometimes the sentence has a word the compound lacks (3.14.50), and sometimes the compound has a sense no member supplies (here). Neither is the other's abbreviation. The pair is the samuddesa's evidence for 3.14.49's utter difference — the two forms do not stand in any regular relation of expansion and contraction.

3.14.52 unformalized

विशेषणविशेष्यत्वं कैश्चिदेकस्तथाश्रयः ।
उपाये तत्त्वदर्शित्वादिष्यते वृत्तिवाक्ययोः ॥ ५२ ॥

viśeṣaṇaviśeṣyatvaṃ kaiścidekastathāśrayaḥ |
upāye tattvadarśitvādiṣyate vṛttivākyayoḥ || 52 ||

For some, the state of qualifier and qualified, and likewise one substratum — from their seeing the reality in the means — is accepted for both derivative formation and sentence.

A rival that assimilates the two. Visesana-visesyatvam kaiscid ekas tatha asrayah: for *some*, the state of *qualifier and qualified*, and likewise *one substratum* — the two relations 3.14.21 catalogued. Upaye tattva-darsitvad isyate vrtti-vakyayoh: from their *seeing the reality in the means*, is accepted for *both* derivative formation *and* sentence. Upaye tattva-darsitvat: seeing the reality *in the means* — these theorists take the ladder for the thing (3.14.49's upaya, mistaken for what it reaches). So they hold that compound and phrase have the same structure, differing only in compression. Bhartrhari reports it and 3.14.53 answers: the structure is not shared, because the compound has no parts to bear it.

3.14.53 unformalized

पदं यथैव वृक्षादि विशिष्टेऽर्थे व्यवस्थितम् ।
नीलोत्पलाद्यपि तथा भागाभ्यां वर्तते विना ॥ ५३ ॥

padaṃ yathaiva vṛkṣādi viśiṣṭe'rthe vyavasthitam |
nīlotpalādyapi tathā bhāgābhyāṃ vartate vinā || 53 ||

Just as a word such as 'vrksa' stands established in a qualified meaning — so 'nilotpala' and the rest operates without the two portions.

The answer to 3.14.52, and the parallel is decisive. Padam yatha eva vrksadi visiste 'rthe vyavasthitam: just as a word such as *'vrksa'* (tree) stands established in a *qualified* meaning — 'tree' means a determinate thing, and nobody looks for qualifier and qualified inside it. Nilotpaladi api tatha bhagabhyam vartate vina: so *'nilotpala'* (blue lotus) and the rest operates *without the two portions* — the compound means its thing as simply as 'vrksa' does, and has no more internal structure. So the compound is a *word*, not a compressed phrase: it stands to its meaning as any stem does. This is 2.36's asvakarna (the compound understood whole, its members idle) and 3.14.44's ekarthibhava (becoming-one-meaning) shown in the plainest case — and it refutes 3.14.52 by pointing out that nobody demands parts of 'vrksa'.

3.14.54 unformalized

श्रोत्रियक्षेत्रियादीनां न च वासिष्ठगार्ग्यवत् ।
भेदेन प्रत्ययो लोके तुल्यरूपासमन्वयात् ॥ ५४ ॥

śrotriyakṣetriyādīnāṃ na ca vāsiṣṭhagārgyavat |
bhedena pratyayo loke tulyarūpāsamanvayāt || 54 ||

And of 'srotriya', 'ksetriya' and the rest, the cognition in the world is not by way of difference, as with 'vasistha' and 'gargya' — from the absence of concomitance with a like form.

Evidence from how derivatives are actually understood. Srotriya-ksetriyadinam: of *'srotriya'* (a learned brahmin), *'ksetriya'* (an incurable disease) and the rest — irregular derivatives whose bases are obscure. Na ca vasistha-gargyavat bhedena pratyayo loke: the *cognition* in the *world* is *not by way of difference*, as with *'vasistha'* and *'gargya'* — the patronymics, where everyone hears the base (Vasistha's son) and the derivative. Tulya-rupa-asamanvayat: from the *absence of concomitance* with a *like form* — 'srotriya' has no transparent base to concomit with. So some derivatives are heard as derived and some are not, and the difference is whether a like form recurs elsewhere (2.12's anvaya-vyatireka). Where it does not, the word is simply a word — which is 3.14.53's finding, and here it is confirmed by what hearers actually do.

3.14.55 unformalized

सप्तपर्णादिवद् भेदो न वृत्तौ विद्यते क्वचित् ।
रूढयरूढिविभागोऽपि क्रियते प्रतिपत्तये ॥ ५५ ॥

saptaparṇādivad bhedo na vṛttau vidyate kvacit |
rūḍhayarūḍhivibhāgo'pi kriyate pratipattaye || 55 ||

Like 'saptaparna' and the rest, no difference exists anywhere in a derivative formation; the division into conventional and non-conventional is also made for comprehension's sake.

The samuddesa's flat statement, with the work's favourite example. Sapta-parnadivat: like *'saptaparna'* — the seven-leaved tree that has not seven leaves (2.309's simile: the name indicates without describing, and everyone finds the tree). Bhedo na vrttau vidyate kvacit: *no difference* exists *anywhere* in a *derivative formation* — kvacit: anywhere at all. The compound has no internal division, as saptaparna has no seven. Rudhy-arudhi-vibhago 'pi kriyate pratipattaye: the division into *conventional* and *non-conventional* is *also* made *for comprehension's sake*. So even the grammar's distinction between opaque compounds (rudhi) and transparent ones is a teaching-device: *all* of them are opaque, and the transparent ones merely invite an analysis they do not require (2.37's rudhi-sabdatva acknowledged in the science; 2.36's sarvesu pratyayas tatha: the cognition is like that in all cases).

3.14.56-57 unformalized

या सामान्याश्रया संज्ञा विशेषविषया च या ।
बहुलग्रहणान्नास्ति प्रवृत्तिरुभयोस्तयोः ॥ ५६ ॥
सुसूक्ष्मजटकेशादौ समासोऽवयवे यदि ।
स्यात्, स्यात् तन्त्रान्तरङ्गत्वात् बाधकोऽवयवस्वरः ॥ ५७ ॥

yā sāmānyāśrayā saṃjñā viśeṣaviṣayā ca yā |
bahulagrahaṇānnāsti pravṛttirubhayostayoḥ || 56 ||
susūkṣmajaṭakeśādau samāso'vayave yadi |
syāt, syāt tantrāntaraṅgatvāt bādhako'vayavasvaraḥ || 57 ||

(56) That designation which rests on the general, and that whose domain is the particular — from the taking-up of 'bahulam', there is no operation of both of those. (57) If, in 'su-suksma-jata-kesa' and the like, there were a compound in a member — then, from the internality of the single-performance, the member-accent would be the annuller.

Two technical cases. (56): ya samanya-asraya samjna visesa-visaya ca ya — the designation resting on the *general* and that whose domain is the *particular*: two classes of technical term. Bahula-grahanat na asti pravrttir ubhayos tayoh: from the taking-up of *'bahulam'* — Panini's word for 'variously', the marker of a rule that applies irregularly — there is *no operation of both*. Where the grammar says bahulam, it declines to specify, and neither designation governs. (57) turns to accent: su-suksma-jata-kesadau samaso 'vayave yadi syat — if, in *'su-suksma-jata-kesa'* (having fine matted hair) and the like, there were a *compound in a member*: an inner compound inside an outer one. Syat tantra-antarangatvat badhako 'vayava-svarah: then, from the *internality* of the *single-performance*, the *member-accent* would be the *annuller* — the inner compound's accent, being antaranga (settled first: 3.7.182), would override the outer's. So nesting compounds generates accent-conflicts, and antarangatva decides them.

3.14.58 unformalized

समुदायस्य वृत्तौ च नैकदेशो विभाष्यते ।
भेद एव विभाषायां नियतो विषयो यतः ॥ ५८ ॥

samudāyasya vṛttau ca naikadeśo vibhāṣyate |
bheda eva vibhāṣāyāṃ niyato viṣayo yataḥ || 58 ||

And in the derivative formation of an aggregate, a single portion is not optioned; since, in the option, difference alone is the fixed domain.

Why inner compounds cannot be optional. Samudayasya vrttau ca na ekadeso vibhasyate: in the *derivative formation of an aggregate*, a *single portion* is *not optioned* — one cannot make part of a compound optional while the rest stands. Bheda eva vibhasayam niyato visayo yatah: since, in the *option*, *difference alone* is the *fixed domain* — an option requires two separable alternatives, and separability is difference (3.14.45's vyavasthita-vibhasa: the settled option, its domains not overlapping). But a compound is ekarthibhava, becoming-one-meaning (3.14.44), and what has become one has no portions to option. So the impossibility follows from the definition: options need parts, and compounds have none. This is 3.14.55's finding (no difference anywhere in a vrtti) doing rule-technical work.

3.14.59 unformalized

यतश्चाविषयः सोऽस्यास्तस्मान्नास्त्यकृतार्थता ।
अभेदप्रक्रमेऽत्यन्तं भेदानामपसारणात् ॥ ५९ ॥

yataścāviṣayaḥ so'syāstasmānnāstyakṛtārthatā |
abhedaprakrame'tyantaṃ bhedānāmapasāraṇāt || 59 ||

And since that is not its domain, therefore there is no unfulfilled purpose — from the utter removal of the differences, in the commencement of non-difference.

The redundancy-worry answered. One might object: if inner portions cannot be optioned (3.14.58), the rules providing such options are purposeless. Yatas ca avisayah so 'syat tasmat na asti akrtarthata: since *that* is *not its domain*, *therefore* there is *no unfulfilled purpose* — a rule is not idle for failing to apply where it was never meant to. Abheda-prakrame 'tyantam bhedanam apasaranat: from the *utter removal* of the *differences*, in the *commencement of non-difference*. So once ekarthibhava commences, the differences are utterly removed (apasarana: driven off), and the option-rules simply have no territory there — which is not a defect in them. This is 3.14.41's argument (prohibitions of compounding have no purpose where only vyapeksa stands) run in the other direction: rules are not vacuous for having bounded domains.

3.14.60 unformalized

महाकष्टश्रितेत्येवं न स्याद् भेदः पदत्रये ।
वृत्तावयवस्यात्त्वं यस्मान्न प्रतिषिध्यते ॥ ६० ॥

mahākaṣṭaśritetyevaṃ na syād bhedaḥ padatraye |
vṛttāvayavasyāttvaṃ yasmānna pratiṣidhyate || 60 ||

In 'maha-kastha-srita' and the like, there would be no difference among the three words — since the 'a'-substitution of a derivative-formation's member is not prohibited.

A three-word compound and its accent. Maha-kastha-srita ity evam na syad bhedah pada-traye: in *'maha-kastha-srita'* (resorting to great difficulty) and the like, there would be *no difference* among the *three words* — the three would compound into one without internal boundaries. Vrtty-avayavasya attvam yasman na pratisidhyate: since the *'a'-substitution* of a derivative-formation's *member* is *not prohibited* — 'maha' becomes 'maha-' by the rule that substitutes a for the final of 'mahat' in compounds, and nothing blocks it inside a member. So the substitution runs, the boundaries dissolve, and the three-word compound is one word with one accent. The verse is testing 3.14.44's ekarthibhava on a hard case: three members, and still no difference.

3.14.61 unformalized

महारण्यमतीते तु त्रिपदाद् भिद्यते स्वरः ।
यस्मात् तत्रान्तरङ्गत्वाद् बाधकोऽवयवस्वरः ॥ ६१ ॥

mahāraṇyamatīte tu tripadād bhidyate svaraḥ |
yasmāt tatrāntaraṅgatvād bādhako'vayavasvaraḥ || 61 ||

But in 'maharanyam atita', the accent differs from that of a three-word form; since there, from internality, the member-accent is the annuller.

The contrast with 3.14.60. Maharanyam atite tu tri-padad bhidyate svarah: in *'maharanyam atita'* (having passed the great forest), the *accent differs* from that of a *three-word* form. Yasmat tatra antarangatvad badhako 'vayava-svarah: since *there*, from *internality*, the *member-accent* is the *annuller* — 'maharanya' is itself a compound, formed first (antaranga), and its accent is fixed before the outer compounding; so it overrides. So two superficially similar cases behave differently: 3.14.60's three words form one flat compound, and here an inner compound's accent survives into the outer. And the difference is antarangatva — what was formed first. The pair shows that ekarthibhava does not erase everything: what became one *earlier* leaves a mark on what becomes one later.

3.14.62 unformalized

सतिशिष्टबलीयस्त्वात् थाथादिस्वर एव तु ।
द्विपदे तेन युगपत् त्रितयं न समस्यते ॥ ६२ ॥

satiśiṣṭabalīyastvāt thāthādisvara eva tu |
dvipade tena yugapat tritayaṃ na samasyate || 62 ||

But from the greater strength of what is taught while the other is present, it is the 'thathadi'-accent alone; therefore in a two-word form the three are not compounded simultaneously.

The accent-conflict resolved by a metarule. Sati-sista-baliyastvat: from the *greater strength* of *what is taught while the other is present* — the sati-sista principle: a rule taught in the presence of another, and therefore later, prevails (the Paribhasa tradition's own maxim). Thathadi-svara eva tu: it is the *'thathadi'*-accent alone — Panini's accent-rule for the affix class beginning with tha. Dvi-pade tena yugapat tritayam na samasyate: *therefore* in a *two-word* form, the *three* are *not compounded simultaneously*. So the accent-facts prove a derivational fact: because the thathadi-accent wins, the three members cannot have compounded at once; they compounded two at a time. The argument runs from accent back to derivation-order — and it is exactly the sort of thing 3.14.61's antarangatva made possible. The samuddesa is reading composition's history off its phonology.

3.14.63 unformalized

येषामपूज्यमानत्वं परार्थानुगमार्थके ।
विशेषणविशेष्यत्वमपि तेषां न कल्पते ॥ ६३ ॥

yeṣāmapūjyamānatvaṃ parārthānugamārthake |
viśeṣaṇaviśeṣyatvamapi teṣāṃ na kalpate || 63 ||

Those which have the state of not being honoured, in what has for its purpose the following of another's aim — for those, even the state of qualifier and qualified is not conceivable.

The bahuvrihi's members reduced to nothing. Yesam apujyamanatvam parartha-anugamarthake: those which have the state of *not being honoured* — apujyamana: unhonoured, i.e. not principal — in what has for its purpose the *following of another's aim*: the bahuvrihi, whose whole business is to reach a possessor outside itself (3.14.38's anya-padartha). Visesana-visesyatvam api tesam na kalpate: for those, *even* the state of *qualifier and qualified* is *not conceivable*. So in 'citragu' neither 'citra' nor 'go' is principal — both are unhonoured, both point away — and where nothing is principal, nothing can be qualified. This finishes off 3.14.52's rival: the qualifier-qualified structure it wanted to find in every compound is unavailable in the one compound-type that most obviously has two members.

3.14.64 unformalized

विशेषः श्रूयमाणोऽपि प्रधानेषु गुणेषु वा ।
शब्दान्तरत्वाद् वाक्ये तु वृत्तौ नित्यं न विद्यते ॥ ६४ ॥

viśeṣaḥ śrūyamāṇo'pi pradhāneṣu guṇeṣu vā |
śabdāntaratvād vākye tu vṛttau nityaṃ na vidyate || 64 ||

A particular, though heard, among principals or among subordinates — in a sentence, from its being another word; but in a derivative formation it never exists.

Visesah sruyamano 'pi pradhanesu gunesu va: a *particular*, though *heard*, among *principals* or among *subordinates*. Sabdantaratvad vakye: in a *sentence*, from its being *another word* — the phrase has particulars because the phrase is a different word (3.14.49's atyanta-bheda) and its members really do stand apart. Vrttau nityam na vidyate: but in a *derivative formation* it *never* exists — nityam: ever, at all. So the particulars one seems to hear inside a compound belong to the phrase one is silently substituting, not to the compound. This is 3.14.37 sharpened: there the compound's member meant a particular; here the compound has no members to mean anything, and what one hears is one's own paraphrase.

3.14.65 unformalized

विशेषकर्मसम्बन्धे निर्भुक्तेऽपि कृतादिभिः ।
विशेषनिरपेक्षोऽन्यः कृतशब्दः प्रवर्तते ॥ ६५ ॥

viśeṣakarmasambandhe nirbhukte'pi kṛtādibhiḥ |
viśeṣanirapekṣo'nyaḥ kṛtaśabdaḥ pravartate || 65 ||

Even when the connection with a particular object has been consumed by 'krta' and the rest, another word 'krta', independent of the particular, operates.

Visesa-karma-sambandhe nirbhukte 'pi krtadibhih: even when the connection with a *particular object* has been *consumed* — nirbhukta: used up, eaten through — by *'krta'* (done) and the rest. Visesa-nirapekso 'nyah krta-sabdah pravartate: *another* word *'krta'*, *independent* of the particular, *operates*. So in 'krta-purvin' (one who has done it before), the 'krta' that ate its object is not the 'krta' standing in the compound: a second, object-indifferent 'krta' is doing the work. Bhartrhari's move is characteristic — where the analysis produces a contradiction (the object is both consumed and still available), he does not repair the analysis but posits that the analysis was tracking two different words. Which is 3.14.49's principle again, applied inside a single form.

3.14.66 unformalized

अकर्मकत्वे सत्येवं क्तान्तं भावाभिधायि यत् ।
ततः क्रियावता कर्त्रा योगो भवति कर्मणाम् ॥ ६६ ॥

akarmakatve satyevaṃ ktāntaṃ bhāvābhidhāyi yat |
tataḥ kriyāvatā kartrā yogo bhavati karmaṇām || 66 ||

When there is intransitivity, that kta-ending which expresses the bare state — from that arises the objects' connection with an agent possessed of the action.

The mechanism of 3.14.65. Akarmakatve saty evam ktantam bhava-abhidhayi yat: when there is *intransitivity*, that *kta*-ending which expresses the *bhava* — the bare state, the action with no object in view (3.7.115's bhava-vacana; 3.8.44's action as its own object). Tatah kriyavata kartra yogo bhavati karmanam: *from that* arises the *objects'* connection with an *agent* possessed of the *action*. So the sequence is: the kta expresses the state alone; the state implies an agent who has the action; and only then do the objects attach — to the agent, not to the kta. The object is thus reachable without the kta ever taking it, which is exactly what 3.14.65 needed. And the derivation runs through bhava — the actionless action — which is this work's standard solvent for transitivity puzzles.

3.14.67 unformalized

अविग्रहा गतादिस्था यथा ग्रामादिकर्मभिः ।
सम्बध्यते क्रिया तद्वत् कृतपूर्व्यादिषु स्थिता ॥ ६७ ॥

avigrahā gatādisthā yathā grāmādikarmabhiḥ |
sambadhyate kriyā tadvat kṛtapūrvyādiṣu sthitā || 67 ||

As the action residing in 'gata' and the rest, without analysis, is connected with objects such as 'village' — likewise that residing in 'krta-purvin' and the rest.

The precedent. Avigraha gatadi-stha yatha gramadi-karmabhih sambadhyate kriya: as the action residing in *'gata'* (gone) and the rest, *without analysis* — avigraha: unanalysed, the compound taken whole — is *connected* with objects such as *'village'*: 'grama-gata' means gone-to-the-village though nothing in it governs the village. Tadvat krta-purvyadisu sthita: *likewise* that residing in *'krta-purvin'* and the rest. So the connection is made by the whole compound, unanalysed, and no member takes the object. Avigraha is the load-bearing word: the compound relates to the world directly, without passing through its own supposed structure. This is 3.14.53's finding (the compound stands to its meaning as 'vrksa' does) doing the work that the analysis could not.

3.14.68 unformalized

मुण्डिसूत्र्यादयोऽसद्भिर्भागैरनुगता इव ।
विभक्ताः कल्पितात्मनो धातवः कुट्टिचर्चिवत् ॥ ६८ ॥

muṇḍisūtryādayo'sadbhirbhāgairanugatā iva |
vibhaktāḥ kalpitātmano dhātavaḥ kuṭṭicarcivat || 68 ||

'Mundi', 'sutri' and the rest are as though pervaded by parts that are not there; the roots, divided, have imagined selves — like 'kutti' and 'carci'.

One of the work's great verses, and the beginning of its most radical stretch. Mundi-sutryadayo 'sadbhir bhagair anugata iva: *'mundi'* (to shave), *'sutri'* (to thread) and the rest are *as though* pervaded by parts that are *not there* — asadbhir bhagaih: by *unreal* parts; iva: as though. The grammarian analyses 'mundi' into a nominal base plus a denominative affix, and the analysis is a fiction: nothing in the root is that base. Vibhaktah kalpitatmano dhatavah kutti-carcivat: the *roots*, *divided*, have *imagined selves* — kalpita-atman: their very selfhood is a construct — like *'kutti'*, *'carci'*, curadi roots whose supposed derivation is invisible. So division does not find the root's parts; it *invents* them, and the root's identity is an artifact of the inventing. This is 2.10's apoddhara pushed below the word: the extraction goes all the way down and finds constructs at every level.

3.14.69 unformalized

पुत्रीयतौ न पुत्रोऽस्ति विशेषेच्छा तु तादृशी ।
विनैव पुत्रानुगमाद् या पुत्रे व्यवतिष्ठते ॥ ६९ ॥

putrīyatau na putro'sti viśeṣecchā tu tādṛśī |
vinaiva putrānugamād yā putre vyavatiṣṭhate || 69 ||

In 'putriya' there is no son; rather there is a particular desire of such a kind that, without any following of 'son', it stands settled upon a son.

The fiction demonstrated. Putriyatau na putro 'sti: in *'putriya'* — the denominative 'to want a son' — there is *no son*: no putra-element is doing anything in the verb. Visesecchha tu tadrsi: rather there is a *particular desire* of *such a kind*. Vinaiva putra-anugamat ya putre vyavatisthate: which, *without any following of 'son'*, *stands settled upon a son* — the desire is son-directed without containing 'son'. So 'putriya' names a single, unanalysable wanting whose specificity is intrinsic, not composed. This is 3.14.44's ekarthibhava at the root's own level: the meaning has become one, and 'putra + iccha' is the grammarian's reconstruction of something that never was two. The verse is also a small phenomenology — the desire is son-shaped all the way through, and does not consist of a general wanting plus a son.

3.14.70 unformalized

प्राणैर्विना यथा धारिर्जीवतौ प्राणकर्मकः ।
न चात्र धारिर्न प्राणा जीवतिस्तु क्रियान्तरम् ॥ ७० ॥

prāṇairvinā yathā dhārirjīvatau prāṇakarmakaḥ |
na cātra dhārirna prāṇā jīvatistu kriyāntaram || 70 ||

As 'dhr', having the breaths as object, without the breaths, in 'jiv'; and here there is neither 'dhr' nor breaths — 'jiv' is a different action.

The decisive parallel, and the tradition's most famous instance. Pranair vina yatha dharir jivatau prana-karmakah: as *'dhr'* (to hold), having the *breaths* as object, *without* the breaths, in *'jiv'* (to live) — the standard gloss makes 'jivati' = 'pranan dharayati', holds the breaths. Na catra dharir na pranah: and *here* there is *neither* 'dhr' *nor* breaths — the verb contains no holding and no breaths. Jivatis tu kriyantaram: 'jiv' is a *different action*. Living is not breath-holding; it is its own single act, and the gloss is a gloss. So the whole apparatus of derivation-by-paraphrase is put in question by one example everyone accepts: nobody thinks 'jivati' is secretly about pranas, and nobody should think 'putriya' is secretly about a putra. The argument is short and it is fatal.

3.14.71 unformalized

तथा विनेषिपुत्राभ्यां पुत्रीयायां क्रियान्तरम् ।
अन्वाख्यानाय भेदास्तु सदृशाः प्रतिपादकाः ॥ ७१ ॥

tathā vineṣiputrābhyāṃ putrīyāyāṃ kriyāntaram |
anvākhyānāya bhedāstu sadṛśāḥ pratipādakāḥ || 71 ||

Likewise, without 'is' and 'putra', in 'putriya' there is a different action; but the divisions, for the sake of exposition, are similar conveyors.

The parallel closed, and the concession made. Tatha vina isi-putrabhyam putriyayam kriyantaram: likewise, without *'is'* (to desire) and *'putra'*, in *'putriya'* there is a *different action* — as 'jiv' is not dhr-plus-pranas, 'putriya' is not is-plus-putra. Anvakhyanaya bhedas tu sadrsah pratipadakah: *but* the *divisions*, for the sake of *exposition*, are *similar conveyors*. Anvakhyana: the grammarian's proper business — explaining forms already given. Sadrsah pratipadakah: *similar* conveyors — they resemble what they convey, and they do convey it. So Bhartrhari does not abolish the analysis; he relocates it. It is not a discovery of the word's composition but an exposition by resemblance — and it works, which is why it survives its own falsity (3.14.49's upaya; 2.38's what is taken up in order to be let go).

3.14.72-73 unformalized

आक्षेपाच्च प्रयोगेण विषयान्तरवर्तिना ।
सदपीच्छाक्यचः कर्म वाक्य एव प्रयुज्यते ॥ ७२ ॥
प्रसिद्धेन कृतः शब्दो भावगर्हाभिधायिना ।
अभ्यासे तुल्यरूपत्वान्न यङन्तः प्रयुज्यते ॥ ७३ ॥

ākṣepācca prayogeṇa viṣayāntaravartinā |
sadapīcchākyacaḥ karma vākya eva prayujyate || 72 ||
prasiddhena kṛtaḥ śabdo bhāvagarhābhidhāyinā |
abhyāse tulyarūpatvānna yaṅantaḥ prayujyate || 73 ||

(72) And from implication, by a usage operating in another domain, the object of the desiderative kyac, though existing, is employed only in a sentence. (73) The word made by the well-known one expressing censure of the state — in the repetition, from likeness of form, the yan-ending is not employed.

Two rule-technical confirmations. (72): aksepac ca prayogena visayantara-vartina — from *implication*, by a *usage* operating in *another domain*. Sad api iccha-kyacah karma vakya eva prayujyate: the *object* of the desiderative *kyac*, *though existing*, is employed *only in a sentence*. The object exists — 'putriya' is about a son — but it can only be *said* in a phrase, never in the derivative: which is 3.14.69's point in Panini's own currency. (73): prasiddhena krtah sabdo bhava-garha-abhidhayina — the word made by the *well-known* one *expressing censure of the state*. Abhyase tulya-rupatvan na yan-antah prayujyate: in the *repetition*, from *likeness of form*, the *yan*-ending is *not employed* — where the reduplicated form would look identical, the affix is withheld. Both verses show the grammar itself conceding that the analysis and the form come apart.

3.14.74 unformalized

शब्दा यथा विभज्यन्ते भागैरिव विकल्पितैः ।
अन्वाख्येयास्तथा शास्त्रमतिदूरे व्यवस्थितम् ॥ ७४ ॥

śabdā yathā vibhajyante bhāgairiva vikalpitaiḥ |
anvākhyeyāstathā śāstramatidūre vyavasthitam || 74 ||

As words are divided by parts as though imagined, so are they to be expounded; the science stands very far off.

The samuddesa's most famous line, and one of the boldest sentences a grammarian has written about grammar. Sabda yatha vibhajyante bhagair iva vikalpitaih: as *words are divided* by *parts as though imagined* — vikalpita: constructed, optional, made up; iva: as though. Anvakhyeyas tatha: *so* are they *to be expounded* — the exposition follows the imagined division, faithfully. Sastram atidure vyavasthitam: the *science* stands *very far off*. Atidure: at a *great distance*. So Panini's whole apparatus — bases, affixes, substitutions, the entire machinery — is erected on parts that are not there, and the science that erects it is nowhere near the thing (1.1's indivisible sphota; 2.10's apoddhara). This is not scepticism about grammar; it is grammar's self-knowledge. The science works and it does not touch. Bhartrhari says so without flinching and without giving anything up.

3.14.75 unformalized

अर्थस्यानुगमं कञ्चिद् दृष्ट्यैव परिकल्पितम् ।
पदं वाक्ये पदे धातुर्धातौ भागश्च मुण्डिवत् ॥ ७५ ॥

arthasyānugamaṃ kañcid dṛṣṭyaiva parikalpitam |
padaṃ vākye pade dhāturdhātau bhāgaśca muṇḍivat || 75 ||

Some following of the meaning, imagined by the view alone: the word in the sentence, the root in the word, and the part in the root — like 'mundi'.

The distance of 3.14.74 measured at every level. Arthasya anugamam kancid drstya eva parikalpitam: *some* following of the *meaning*, imagined by the *view alone* — drstya eva: by the analytic look and nothing else; parikalpita: fully constructed. And then the three storeys of the fiction: padam vakye — the *word* in the *sentence* (1.73's the sentence alone is real; 2.1's akhandata); pade dhatuh — the *root* in the *word*; dhatau bhagas ca mundivat — the *part* in the *root*, like *'mundi'* (3.14.68's roots with imagined selves). So the same construction repeats at every scale, and at every scale there is only a resemblance of meaning-following, seen by the one who looks for it. The regress is complete and the verse states it flatly: sentence, word, root, sub-root — all the way down, and nothing found.

3.14.76 unformalized

अविप्रयोगः साधुत्वे व्युत्पत्तिरनवस्थिता ।
उपायान् प्रतिपत्तॄणां नाभिमन्येत सत्यतः ॥ ७६ ॥

aviprayogaḥ sādhutve vyutpattiranavasthitā |
upāyān pratipattṝṇāṃ nābhimanyeta satyataḥ || 76 ||

Non-separation is in correctness; derivation is unsettled. One should not regard the means of the comprehenders as true.

The moral, stated in the imperative. Aviprayogah sadhutve: *non-separation* is in *correctness* — what is correct is the undivided form itself; sadhutva attaches to the whole word, not to its parts (1.27's correctness not compositional). Vyutpattir anavasthita: *derivation* is *unsettled* — anavasthita: without a fixed footing, the etymologist's constructions shifting with the etymologist. Upayan pratipattrnam na abhimanyeta satyatah: one should *not* regard the *means* of the *comprehenders* as *true*. Na abhimanyeta satyatah: do not take them for the truth. This is the samuddesa's — perhaps the work's — governing instruction, and it applies retroactively to everything Bhartrhari has himself constructed: the upayas are for comprehension, they are not descriptions, and taking them for descriptions is the one error the whole book is written against.

3.14.77 unformalized

यथैव डित्थे डयतिः पाचके पचतिस्तथा ।
डयतिश्च पचिश्रैव द्वावप्येतावलौकिकौ ॥ ७७ ॥

yathaiva ḍitthe ḍayatiḥ pācake pacatistathā |
ḍayatiśca paciśraiva dvāvapyetāvalaukikau || 77 ||

Just as 'day' in 'dittha', so 'pac' in 'pacaka'; and both 'day' and 'pac' alike are non-worldly.

The argument that gives 3.14.76 its teeth, and it is merciless. Yathaiva ditthe dayatih: just as *'day'* in *'dittha'* — 'dittha' is the grammarians' standard meaningless proper name, the very example of a word with no derivation; and yet one *could* extract a root 'day' from it, and the extraction would be pure fabrication. Pacake pacatis tatha: *so* 'pac' in *'pacaka'* (cook). Dayatis ca pacis caiva dvav apy etav alaukikau: and *both*, 'day' *and* 'pac', are *alaukika* — *not of the world*. So the root everyone accepts and the root everyone would laugh at stand on exactly the same footing: both are products of the same extraction, and neither is out there. The only difference is that one extraction has a tradition behind it. Alaukika is the grammarians' own word for their technical fictions, and Bhartrhari turns it on their most cherished object.

3.14.78 unformalized

प्रकृतिप्रत्ययावूह्यौ पदात् ताभ्यां पदं तथा ।
अनुबन्धस्वरादिभ्यः शिष्टैः शास्त्रं न तान् प्रति ॥ ७८ ॥

prakṛtipratyayāvūhyau padāt tābhyāṃ padaṃ tathā |
anubandhasvarādibhyaḥ śiṣṭaiḥ śāstraṃ na tān prati || 78 ||

Base and affix are to be conjectured from the word, and the word likewise from those two; the science is by the learned, from indicatory letters, accents and the rest — not directed at those.

The circle exhibited, and the science placed. Prakrti-pratyayav uhyau padat: *base* and *affix* are to be *conjectured* from the *word* — uhya: to be surmised, guessed at, reasoned out; they are not given, they are inferred backwards from the form. Tabhyam padam tatha: and the *word* from *those two*, *likewise*. So the circle is closed and admitted: the parts are conjectured from the whole, and the whole is then derived from the parts. Anubandha-svaradibhyah sistaih sastram na tan prati: the *science* is by the *learned*, from *indicatory letters*, *accents* and the rest — the it-markers, the tone-marks, the whole notational apparatus — na tan prati: *not directed at those*. The science is built out of its own devices and is not about them. So the Vrtti-samuddesa's method-stretch ends where it must: grammar is a circle of conjectures made of notation, it stands atidure (3.14.74), it is true of nothing, and it is exactly right.
contested: The final pada admits more than one construal. We take 'na tan prati' as denying that the sastra is *about* its own devices (anubandhas, accents) — the notation is instrumental, not the subject-matter. It can also be read as denying that the sastra is directed at the prakrti and pratyaya of pada a, i.e. that the science does not concern the conjectured parts as realities; on this reading the circle is being disowned rather than the notation. Both readings serve the samuddesa's conclusion; they differ on what 'tan' resumes.

3.14.79 unformalized

शास्त्रदृष्टिस्तु शास्त्रस्य प्राप्तिमात्रेऽप्यनिश्चिते ।
युज्यते प्रत्यवायेन शास्त्रं चक्षुरपश्यताम् ॥ ७९ ॥

śāstradṛṣṭistu śāstrasya prāptimātre'pyaniścite |
yujyate pratyavāyena śāstraṃ cakṣurapaśyatām || 79 ||

But the science's way of seeing, even where the mere obtaining of a rule is undetermined, is joined with harm; the science is an eye for those who do not see.

The counterweight to 3.14.76, and it arrives immediately. Sastra-drstis tu sastrasya praptimatre 'py aniscite: the *science's way of seeing* — sastra-drsti: construing the world by the grammar — even where the *mere obtaining* of a rule is *undetermined*. Yujyate pratyavayena: is *joined with harm* — pratyavaya: fault, damage. So one may not run the sastra's vision past the point where the sastra itself has settled anything. And then the concession, which is also the vindication: sastram caksur apasyatam — the *science* is an *eye* for those who *do not see*. Caksuh: an eye. So 3.14.76's warning does not demote grammar; it locates it. The science is not true, and it is *sight* — for those who lack it there is nothing to see with. The two verses together are the samuddesa's settled position: an eye is not the world, and one does not throw away one's eyes.

3.14.80 unformalized

अर्थान्तराभिधानाच्च पौर्वापर्यं न भिद्यते ।
राजदन्ताहिताग्न्यादिराजाश्वादिषु सर्वथा ॥ ८० ॥

arthāntarābhidhānācca paurvāparyaṃ na bhidyate |
rājadantāhitāgnyādirājāśvādiṣu sarvathā || 80 ||

And from the expressing of another meaning, priority and posteriority are not differentiated, in 'rajadanta', 'ahitagni' and the rest, and in 'rajasva' and the rest, in every way.

Word-order dissolved along with the parts. Arthantara-abhidhanac ca paurvaparyam na bhidyate: from the *expressing of another meaning*, *priority and posteriority* are *not differentiated* — the compound expresses a meaning other than its members', so the order of members carries nothing. Rajadanta-ahitagny-adi-rajasvadisu sarvatha: in *'rajadanta'* (the class where the expected first member goes second), *'ahitagni'* (the bahuvrihi with its participle first), *'rajasva'* and the rest — *sarvatha*: in every way. So the irregular orders are not irregular: order was never semantic to begin with, and Panini's rajadanta-gana simply registers usage. This follows from 3.14.68-78: if the members are not there, their sequence is not there either, and the rules governing sequence are notation (3.14.78's anubandha-svaradi).

3.14.81 unformalized

विनैव प्रत्ययैर्वृत्तौ ये भिन्नार्थाभिधायिनः ।
गर्गादयो लुका तेषां साधुत्वमनुगम्यते ॥ ८१ ॥

vinaiva pratyayairvṛttau ye bhinnārthābhidhāyinaḥ |
gargādayo lukā teṣāṃ sādhutvamanugamyate || 81 ||

Those which, without affixes, express a different meaning in the derivative formation — 'garga' and the rest — their correctness is followed out by means of elision.

How the grammar copes with what it cannot derive. Vinaiva pratyayair vrttau ye bhinnarthabhidhayinah: those which, *without affixes*, express a *different meaning* in the derivative formation — 'gargah' meaning Garga's descendants, with no visible patronymic affix. Gargadayo luka tesam sadhutvam anugamyate: *'garga'* and the rest — their *correctness* is *followed out* by means of *luk*, elision. So the grammar posits an affix and then deletes it, and the form is thereby certified sadhu. Anugamyate is the exact word: the correctness is *followed*, tracked, come after — not conferred. The form was already correct (3.14.76's aviprayogah sadhutve); the luk is how the science catches up with it. This is 3.14.79's eye at work: seeing an affix that is not there is what lets the science see the word at all.

3.14.82 unformalized

सोऽयमित्यभिसंबन्धात् प्रत्ययेन विना यदि ।
बभ्रूवादयः प्रयुज्येरन्नपत्ये नियमो भवेत् ॥ ८२ ॥

so'yamityabhisaṃbandhāt pratyayena vinā yadi |
babhrūvādayaḥ prayujyerannapatye niyamo bhavet || 82 ||

If, from the connection 'this is that', 'babhru' and the rest were employed without an affix, there would be a restriction to offspring.

The cost of taking the elision seriously. So 'yam ity abhisambandhat: from the *connection* 'this is that' — the identification by which a descendant is called by the ancestor's name. Pratyayena vina yadi babhrvadayah prayujyeran: *if* *'babhru'* and the rest were *employed* *without an affix* — i.e. if the naked stem simply meant the descendant, with no elided taddhita behind it. Apatye niyamo bhavet: there would be a *restriction to offspring* — the identification would have to be restricted to offspring, which nothing in 'this is that' warrants: an heir, a pupil, a namesake all satisfy it. So the elided affix earns its keep after all: it is what narrows the identification. The verse concedes the point against itself, which is the samuddesa's habit — the fictions of 3.14.74 are not idle fictions, and where one removes them, real work goes undone.

3.14.83 unformalized

सोऽयमित्यभिसंबन्धे लिङ्गोपव्यञ्जनादृते ।
प्रष्ठादिषु न जायैव नियमेन प्रतीयते ॥ ८३ ॥

so'yamityabhisaṃbandhe liṅgopavyañjanādṛte |
praṣṭhādiṣu na jāyaiva niyamena pratīyate || 83 ||

In the connection 'this is that', apart from an indicatory mark, in 'prastha' and the rest the wife alone is not understood by restriction.

The same argument on a second case. So 'yam ity abhisambandhe linga-upavyanjanad rte: in the connection 'this is that', *apart from* an *indicatory mark* — linga: here the gender-sign, and upavyanjana: the accompanying indication. Prasthadisu na jaya eva niyamena pratiyate: in *'prastha'* and the rest, the *wife alone* is not *understood by restriction* — 'prastha' in the feminine is supposed to yield the headman's wife, but the identification 'this is that' does not by itself pick out a wife rather than a daughter or a servant. So again the bare relation underdetermines, and something else — the gender-mark — must do the restricting. Both 3.14.82 and this verse show the same shape: identification is broad, usage is narrow, and the grammar's machinery is what covers the distance.

3.14.84-85 unformalized

मानमेयाभिसंबन्धविशेषेऽङ्गीकृते तथा ।
प्रस्थादीनामसाधुत्वं तद्धितेन विना भवेत् ॥ ८४ ॥
तद्धितो योगभेदेन वाक्यं वा स्याद् विभाषितम् ।
परिमाणाधिके तत्र प्रथमा शिष्यते पुनः ॥ ८५ ॥

mānameyābhisaṃbandhaviśeṣe'ṅgīkṛte tathā |
prasthādīnāmasādhutvaṃ taddhitena vinā bhavet || 84 ||
taddhito yogabhedena vākyaṃ vā syād vibhāṣitam |
parimāṇādhike tatra prathamā śiṣyate punaḥ || 85 ||

(84) When a particular connection of measure and measured is accepted, the incorrectness of 'prastha' and the rest would follow, without the taddhita. (85) The taddhita, by difference of rule, or else the sentence, would be optional; there, in the case of excess over the measure, the first case-ending is again taught.

Measure-words, and the same lesson a third time. (84): mana-meya-abhisambandha-visese 'ngikrte — when a *particular connection* of *measure and measured* is *accepted*: 'prastha' as a unit, and what is measured by it. Prasthadinam asadhutvam taddhitena vina bhavet: the *incorrectness* of 'prastha' and the rest would follow, *without the taddhita* — without the affix, the measure-word cannot be certified. (85): taddhito yoga-bhedena vakyam va syad vibhasitam — the *taddhita*, by *difference of rule*, or else the *sentence*, would be *optional*. Parimana-adhike tatra prathama sisyate punah: there, in the case of *excess over the measure*, the *first* case-ending is *again* taught — where the quantity exceeds the unit, the nominative returns. So the grammar handles measure by an affix, an option, and a case-rule, and every one of them is machinery (3.14.78).

3.14.86 unformalized

अतिरिक्तस्य साधुत्वे तदेव च निदर्शनम् ।
युज्यतेऽङ्गीकृताधिक्यं तत् सर्वाभिर्विभक्तिभिः ॥ ८६ ॥

atiriktasya sādhutve tadeva ca nidarśanam |
yujyate'ṅgīkṛtādhikyaṃ tat sarvābhirvibhaktibhiḥ || 86 ||

In the correctness of the excess, that very thing is the illustration; that, with its excess accepted, is joined with all the case-endings.

Atiriktasya sadhutve tad eva ca nidarsanam: in the *correctness* of the *excess*, *that very thing* is the *illustration* — the excess-case of 3.14.85 is itself the demonstration. Yujyate 'ngikrtadhikyam tat sarvabhir vibhaktibhih: that, with its *excess accepted*, is *joined with all the case-endings* — once the excess is granted, the form takes every case freely, and no restriction survives. So the measure-word, having been certified by the taddhita, behaves like any nominal: the machinery does its work and then withdraws, leaving an ordinary word. Which is the pattern of the whole stretch — the grammar's constructs are scaffolding, they hold the form up while it is being certified, and the finished word stands without them (3.14.71's upayas, similar and discardable).

3.14.87 unformalized

शुक्लादिषु मतुब्लोपो व्यतिरेकस्य दर्शनात् ।
असाधुत्वनिवृत्त्यर्थं साधवस्ते बिदादिवत् ॥ ८७ ॥

śuklādiṣu matublopo vyatirekasya darśanāt |
asādhutvanivṛttyarthaṃ sādhavaste bidādivat || 87 ||

In 'sukla' and the rest, the elision of 'matup', from the seeing of the difference; for the sake of removing incorrectness they are correct, like 'bida' and the rest.

The construct named for what it is. Sukladisu matub-lopo vyatirekasya darsanat: in *'sukla'* (white) and the rest, the *elision* of *'matup'* — the possessive affix — from the *seeing of the difference*: one sees that 'sukla' behaves as though it meant 'possessing whiteness', so one posits matup and deletes it. Asadhutva-nivrtty-artham sadhavas te bidadivat: *for the sake of removing incorrectness* they are *correct*, like *'bida'* and the rest. Asadhutva-nivrtty-artham is the confession: the elided affix exists in order to *stop the form being ungrammatical* — that is its entire purpose, and it has no other. So the science posits what it needs to certify what usage already gives it (3.14.43's the world's settled arrangement; 3.14.81's anugamyate). Bhartrhari does not object. He only insists one know what one is doing.

3.14.88 unformalized

विशेषणाद् विशेष्येऽर्थे तद्भावाभ्युच्चये सति ।
पुनश्च प्रतिसंहारे वृत्तिमेके प्रचक्षते ॥ ८८ ॥

viśeṣaṇād viśeṣye'rthe tadbhāvābhyuccaye sati |
punaśca pratisaṃhāre vṛttimeke pracakṣate || 88 ||

From the qualifier, in the qualified meaning, when there is an accumulation of that state, and again in the withdrawal — some declare there to be a derivative formation.

A rival account of what composition is, and it is psychological. Visesanad visesye 'rthe tad-bhava-abhyuccaye sati: *from the qualifier*, in the *qualified meaning*, when there is an *accumulation* of *that state* — abhyuccaya: heaping up, the qualifier's character piling into the qualified. Punas ca pratisamhare: and *again* in the *withdrawal* — pratisamhara: the drawing-back, the qualifier retracting once it has deposited its contribution. Vrttim eke pracaksate: *some* declare there to be a *derivative formation*. So on this view a compound is a two-beat mental event: the qualifier pours itself into the qualified and then withdraws, and what is left — qualified, and standing alone — is the compound. This is 3.14.9-11's deposited delimitation given a temporal structure, and 3.14.92-93 will endorse its phenomenology while denying its divisions.

3.14.89-90 unformalized

निमित्ते प्रत्ययः पूर्वो नानुप्राप्तो निमित्तिना ।
निमित्तवति बुद्धेश्च न निमित्तसरूपता ॥ ८९ ॥
संस्कारसहिताद् ज्ञानान्नोपश्लेषः स्मृतेरपि ।
व्यापारे तन्निमित्तानां न ग्राह्यं स्याद् तथाविधम् ॥ ९० ॥

nimitte pratyayaḥ pūrvo nānuprāpto nimittinā |
nimittavati buddheśca na nimittasarūpatā || 89 ||
saṃskārasahitād jñānānnopaśleṣaḥ smṛterapi |
vyāpāre tannimittānāṃ na grāhyaṃ syād tathāvidham || 90 ||

(89) The cognition of the cause, being prior, is not reached by what is caused; and of the cognition in the possessor of the cause, there is no likeness of form with the cause. (90) From cognition accompanied by an impression, there is no contact even from memory; in the operation of those causes, the graspable would not be of that kind.

An epistemological objection to 3.14.88's two-beat account. (89): nimitte pratyayah purvo na anuprapto nimittina — the *cognition of the cause* (the qualifier), being *prior*, is *not reached* by what is *caused*: the earlier cognition is over before the later one exists, so the later cannot get at it to receive a deposit. Nimittavati buddhes ca na nimitta-sarupata: and of the *cognition* in the *cause-possessor* there is *no likeness of form* with the cause — the qualified's cognition does not resemble the qualifier's, so nothing was transferred. (90): samskara-sahitaj jnanan nopaslesah smrter api — even bringing in *impressions* and *memory* gives no *contact* (upaslesa: adhesion). Vyapare tan-nimittanam na grahyam syat tathavidham: in their *operation*, the *graspable* would not be *of that kind*. So the deposit-and-withdraw story has no mechanism: two cognitions cannot touch.

3.14.91 unformalized

अन्तःकरणवृत्तौ च व्यर्था बाह्यार्थकल्पना ।
तस्मादनुपकारो वा ग्राह्यं वा न तथा स्थितम् ॥ ९१ ॥

antaḥkaraṇavṛttau ca vyarthā bāhyārthakalpanā |
tasmādanupakāro vā grāhyaṃ vā na tathā sthitam || 91 ||

And in the operation of the inner organ, the imagining of an external object is purposeless. Therefore either there is no assistance, or the graspable is not so situated.

The objection driven home, and then the dilemma. Antahkarana-vrttau ca vyartha bahyartha-kalpana: in the *operation of the inner organ*, the *imagining of an external object* is *purposeless* — vyartha: pointless. If the qualifier-qualified business happens in the mind, positing an outer object to anchor it adds nothing (2.129's the external is a support one does not need; 3.3.53's meanings that are the intellect's own). Tasmad anupakaro va grahyam va na tatha sthitam: *therefore* either there is *no assistance* — the object contributes nothing — *or* the *graspable* is *not so situated*: it is not out there as supposed. Both horns are unwelcome to 3.14.88's theorists, and Bhartrhari leaves them there. The dilemma clears the ground for 3.14.92's own account, which will keep the phenomenon and drop the transaction.

3.14.92-93 unformalized

अनुस्यूतेव संसृष्टेरर्थे बुद्धिः प्रवर्तते ।
व्याख्यातारो विभज्याथ तां भेदेन प्रचक्षते ॥ ९२ ॥
तदात्मन्यविभक्ते च बुद्ध्यन्तरमुपाश्रिताः ।
विभागमिव मन्यन्ते विशेषणविशेष्ययोः ॥ ९३ ॥

anusyūteva saṃsṛṣṭerarthe buddhiḥ pravartate |
vyākhyātāro vibhajyātha tāṃ bhedena pracakṣate || 92 ||
tadātmanyavibhakte ca buddhyantaramupāśritāḥ |
vibhāgamiva manyante viśeṣaṇaviśeṣyayoḥ || 93 ||

(92) The intellect operates in the meaning as though threaded through, from the conjunction; the expounders, having divided, then declare it by way of difference. (93) And in that self, undivided, having resorted to a second cognition, they think there is as it were a division of qualifier and qualified.

The samuddesa's own account, and it is the work's whole thesis in two verses. (92): anusyuta iva samsrster arthe buddhih pravartate — the *intellect* operates in the *meaning* *as though threaded through*, from the *conjunction*. Anusyuta: sewn through, continuous, one thread running the length of it — the single most beautiful word in the samuddesa, and the exact opposite of composition. The cognition of 'nilotpala' is not blue plus lotus; it is one thread. Vyakhyatharo vibhajya atha tam bhedena pracaksate: the *expounders*, *having divided*, *then* declare it *by way of difference* — the division comes first, from the expositor, and the difference is announced afterwards. (93): tad-atmany avibhakte ca buddhy-antaram upasritah — in that *self*, *undivided*, having *resorted to a second cognition*. Vibhagam iva manyante visesana-visesyayoh: they *think* there is *as it were* a *division* of qualifier and qualified. Iva: as it were. So the division is not in the meaning; it is in the second cognition one brings to it (1.89's the knower's own construction returned to the known). Qualifier and qualified are what a mind that has already divided finds when it looks. This is 1.1's indivisible sphota and 3.14.74's imagined parts and 2.10's apoddhara, all resolved into a single psychological fact: the thread is one, and the expositor cuts it.

3.14.94 unformalized

अबुधान् प्रति वृत्तिं च वर्तयन्तः प्रकल्पिताम् ।
आहुः परार्थवचने त्यागाभ्युच्चयधर्मताम् ॥ ९४ ॥

abudhān prati vṛttiṃ ca vartayantaḥ prakalpitām |
āhuḥ parārthavacane tyāgābhyuccayadharmatām || 94 ||

And carrying on the constructed derivation for the unlearned, they speak of the property of abandonment and of accumulation in the expression of another's meaning.

The tradition's two doctrines placed as upayas. Abudhan prati vrttim ca vartayantah prakalpitam: *carrying on* the *constructed* derivation *for the unlearned* — prakalpita: fabricated (3.14.75's parikalpita); abudhan prati: aimed at those who do not know (3.14.49). Ahuh pararthavacane tyaga-abhyuccaya-dharmatam: they speak of the *property* of *abandonment* and of *accumulation* in the *expression of another's meaning*. Tyaga: the jahat-svartha doctrine — in a bahuvrihi the members abandon their own meanings (3.12.27). Abhyuccaya: the ajahat-svartha doctrine — they retain and heap them up (3.14.88). The tradition's long quarrel between these two, and Bhartrhari's verdict is that both are teaching-devices for the unlearned, spoken over a meaning that was never divided (3.14.92's anusyuta). Neither party is wrong; both are describing their own cut.

3.14.95 unformalized

अन्वयाद् गम्यते सोऽर्थो विरोधी वा निवर्तते ।
द्वयर्थमर्थान्तरे वापि तत्राहुरुपसर्जनम् ॥ ९५ ॥

anvayād gamyate so'rtho virodhī vā nivartate |
dvayarthamarthāntare vāpi tatrāhurupasarjanam || 95 ||

That meaning is understood from agreement, or the contradictory one withdraws; in the two-meaninged, or else in another meaning — there they call it 'subordinate'.

Upasarjana defined by what happens to the loser. Anvayad gamyate so 'rtho: that meaning is understood *from agreement* — the surviving sense, carried by concomitance. Virodhi va nivartate: *or* the *contradictory* one *withdraws* — where two senses conflict, one leaves. Dvyartham arthantare vapi tatrahur upasarjanam: in the *two-meaninged*, or else in *another meaning* — *there* they call it *'upasarjana'*, subordinate. So Panini's technical term for the compound's non-principal member is being derived from three quite different situations: agreement carrying one sense, contradiction expelling one, and the whole thing decamping into a third meaning. The term is one and the phenomena are three — which is what 3.14.96 is about to say in general.

3.14.96 unformalized

उपायमात्रं नानात्वं समूहस्त्वेक एव सः ।
विकल्पाभ्युच्चयाभ्यां वा भेदसंसर्गकल्पना ॥ ९६ ॥

upāyamātraṃ nānātvaṃ samūhastveka eva saḥ |
vikalpābhyuccayābhyāṃ vā bhedasaṃsargakalpanā || 96 ||

Multiplicity is a mere means; that aggregate is one alone. Or, by alternation and accumulation, the construction of difference and of conjunction.

The samuddesa's summary sentence, and it is eight words long. Upaya-matram nanatvam: *multiplicity* is a *mere means* — nanatva: manyness, plurality, the whole apparatus of parts; upaya-matra: a means and *nothing else* (3.14.76's do not take the upayas for true). Samudayas tv eka eva sah: the *aggregate* is *one alone* — eka eva: one, simply. Vikalpa-abhyuccayabhyam va bheda-samsarga-kalpana: *or*, by *alternation* and *accumulation*, the construction of *difference* and of *conjunction* — the two doctrines of 3.14.94 (tyaga and abhyuccaya) recast as two ways of *constructing* what is not there. Kalpana: construction. So the tradition's rival theories of the compound are alternative fabrications of a multiplicity that is a device, laid over an aggregate that is one. This is 3.14.92's anusyuta stated as doctrine.

3.14.97 unformalized

वृत्त वर्तयतामेवमबुधप्रतिपत्तये ।
भिन्नाः सम्बोधनोपायाः पुरुषेष्वनवस्थिताः ॥ ९७ ॥

vṛtta vartayatāmevamabudhapratipattaye |
bhinnāḥ sambodhanopāyāḥ puruṣeṣvanavasthitāḥ || 97 ||

Of those thus carrying on the derivation for the unlearned's comprehension, the means of instruction are diverse, unsettled among persons.

Why the tradition has rival theories at all. Vrttim vartayatam evam abudha-pratipattaye: of those *carrying on* the derivation for the *unlearned's comprehension* (3.14.94's abudhan prati). Bhinnah sambodhana-upayah purusesv anavasthitah: the *means of instruction* are *diverse*, *unsettled among persons* — anavasthita: without fixed footing (3.14.76's vyutpattir anavasthita), and purusesu: varying from teacher to teacher. So the quarrel between jahat-svartha and ajahat-svartha is not a disagreement about the facts: it is two teachers' pedagogy, and pedagogy varies because pupils vary. Nothing in the object adjudicates, because the object is one (3.14.96) and neither doctrine is about it. This is the most irenic thing in the samuddesa and also the most deflationary.

3.14.98 unformalized

वाचिका द्योतिका वापि सङ्खयानां या विभक्तयः ।
तद्रूपेऽवयवे वृत्तौ सङ्खयाभेदो निवर्तते ॥ ९८ ॥

vācikā dyotikā vāpi saṅkhayānāṃ yā vibhaktayaḥ |
tadrūpe'vayave vṛttau saṅkhayābhedo nivartate || 98 ||

Those case-endings which are expressive or indicative of numbers — when the member has that form, in a derivative formation the difference of number withdraws.

Number inside compounds. Vacika dyotika vapi sankhyanam ya vibhaktayah: the case-endings which are *expressive* or *indicative* of *numbers* — the samuddesa's standing distinction (2.190's vacaka against dyotaka: does the ending *say* the number or merely *reveal* it?), and here left open, because it does not matter for the point. Tad-rupe 'vayave vrttau sankhya-bhedo nivartate: when the *member* has *that form*, in a *derivative formation* the *difference of number* *withdraws*. So 'brahmana-kula' does not say how many brahmanas: the number-difference goes out with the case-ending. This is 3.14.44's ekarthibhava applied to number — the meanings having withdrawn, their number withdraws with them (3.14.46's expectancy withdrawing). And it raises the question 3.14.99 must answer: what number, then, does the compound have?

3.14.99 unformalized

अभेदैकत्वसङ्खया वा तत्रान्यैवोपजायते ।
संसर्गरूपं (संसर्गिरूपं) सङ्खयानामविभक्तं तदुच्यते ॥ ९९ ॥

abhedaikatvasaṅkhayā vā tatrānyaivopajāyate |
saṃsargarūpaṃ (saṃsargirūpaṃ) saṅkhayānāmavibhaktaṃ taducyate || 99 ||

Or another number altogether arises there — the number of non-different unity; that is called the unanalysed form of conjunction of the numbers.

The answer, and it is not zero. Abhedaikatva-sankhya va tatra anya eva upajayate: *another* number *altogether* *arises* there — the *number of non-different unity*: abheda-ekatva, a oneness that is not the numeral one but the absence of internal difference. Samsarga-rupam sankhyanam avibhaktam tad ucyate: that is called the *unanalysed form* of the *conjunction* of the *numbers* — avibhakta: undivided (3.14.93's tad-atmany avibhakte). So the compound does not lack number; it has a number of its own, which is the numbers of its members fused past recovery. The members' numbers are in there the way the herbs are in the honey — which is exactly the image 3.14.100 reaches for.
contested: The source records a variant at pada c: samsarga-rupam ('the form which is conjunction') against samsargi-rupam ('the form of what is conjoined'). We translate the former. The variant shifts the emphasis slightly — whether the unanalysed thing is the conjoining or the conjuncts — but on either reading the point stands: what results is undivided.

3.14.100 unformalized

यथौषधिरसाः सर्वे मधुन्याहितशक्तयः ।
अविभागेन वर्तन्ते तां सङ्खयां तादृशीं विदुः ॥ १०० ॥

yathauṣadhirasāḥ sarve madhunyāhitaśaktayaḥ |
avibhāgena vartante tāṃ saṅkhayāṃ tādṛśīṃ viduḥ || 100 ||

As all the herb-juices, their potencies deposited in honey, exist without division — such they know that number to be.

One of the loveliest images in the work, and it does real analytic work. Yatha ausadhi-rasah sarve madhuny ahita-saktayah: as *all* the *herb-juices*, their *potencies deposited* in *honey* — ahita-sakti: their powers *placed into* it, the bees having gathered from a hundred plants. Avibhagena vartante: they exist *without division* — avibhaga: nothing in the honey is this flower's and nothing that flower's, and yet every flower is in it and acting. Tam sankhyam tadrsim viduh: *such* they know that *number* to be. So the compound's abheda-ekatva (3.14.99) is honey: the members' numbers are wholly present, wholly potent, and wholly unrecoverable. The image is better than any argument could be, because it exhibits a case where fusion does not mean loss — the honey has all the herbs' powers and none of their boundaries. Which is what a compound is.

3.14.101 unformalized

भेदानां वा परित्यागात् सङ्खयात्मा स तथाविधः ।
व्यापाराज्जतिभागस्य भेदापोहेन वर्तते ॥ १०१ ॥

bhedānāṃ vā parityāgāt saṅkhayātmā sa tathāvidhaḥ |
vyāpārājjatibhāgasya bhedāpohena vartate || 101 ||

Or, from the abandonment of the differences, that number-self is of such a kind; from the operation of the universal-portion, it operates by the exclusion of differences.

The alternative account of the same fact. Bhedanam va parityagat sankhyatma sa tathavidhah: *or*, from the *abandonment* of the *differences*, that *number-self* is of *such a kind* — not fusion (3.14.100's honey) but renunciation: the differences are simply dropped. Vyaparaj jati-bhagasya bhedapohena vartate: from the *operation* of the *universal-portion*, it operates by the *exclusion of differences* — the jati-element in the meaning does the work, and its way of working is apoha, exclusion. So two accounts stand side by side: the honey, where the differences are all present and indistinguishable, and the universal, where they are excluded. Bhartrhari gives both without choosing — which is 3.14.97's point about upayas being unsettled among persons, and 3.14.96's about multiplicity being a device either way.

3.14.102-103 unformalized

अगृहीतविशेषेण यथा रूपेण रूपवान् ।
प्रख्यायते न शुक्लादिभेदरूपस्तु(रूपन्तु) गृह्यते ॥ १०२ ॥
भेदरूपसमावेशे तथा सत्यविवक्षिते ।
भागः प्रकाशितः कश्चिच्छास्त्रेऽङ्गत्वेन गृह्यते ॥ १०३ ॥

agṛhītaviśeṣeṇa yathā rūpeṇa rūpavān |
prakhyāyate na śuklādibhedarūpastu(rūpantu) gṛhyate || 102 ||
bhedarūpasamāveśe tathā satyavivakṣite |
bhāgaḥ prakāśitaḥ kaścicchāstre'ṅgatvena gṛhyate || 103 ||

(102) As one possessing form is made known by form whose particular is not grasped, and the difference-form 'white' and the rest is not grasped — (103) so, when the entry of the difference-form is present but not intended for expression, some illumined portion is taken in the science as a subsidiary.

How a general can be present without its particulars. (102): agrhita-visesena yatha rupena rupavan prakhyayate — as one *possessing form* is made known by *form* whose *particular is not grasped*: one sees that a thing is coloured before, and without, seeing which colour. Na sukladi-bheda-rupas tu grhyate: the *difference-form* — 'white' and the rest — is *not grasped*. The perception is real, general, and particular-free. (103): bheda-rupa-samavese tatha saty avivaksite — *so*, when the *entry* of the *difference-form* is *present* but *not intended for expression* (avivaksita: the samuddesa's standing device — 3.11's numbers not meant to be said). Bhagah prakasitah kascic chastre 'ngatvena grhyate: some *illumined portion* is taken in the *science* as a *subsidiary*. So the compound's number is like coloured-ness: the differences are in there and unmeant, and what the grammar picks up is the lit portion.

3.14.104 unformalized

सङ्खयासामान्यरूपेण तदा सोंऽशः प्रतीयते ।
अर्थस्यानेकशक्तित्वे शब्दैर्नियतशक्तिभिः ॥ १०४ ॥

saṅkhayāsāmānyarūpeṇa tadā soṃ'śaḥ pratīyate |
arthasyānekaśaktitve śabdairniyataśaktibhiḥ || 104 ||

Then that portion is cognized in the form of number-in-general — the meaning having many powers, by words having fixed powers.

The conclusion of the number-analysis. Sankhya-samanya-rupena tada so 'msah pratiyate: *then* that *portion* is cognized in the form of *number-in-general* — sankhya-samanya: numberhood, not a number. So what a compound carries is the bare fact of quantity with no quantity specified — which is what 3.14.102's coloured-without-a-colour was the model for. And the reason: arthasyaneka-saktitve sabdair niyata-saktibhih — the *meaning* having *many powers*, by *words* having *fixed powers*. The meaning is inexhaustibly potent (3.14.100's honey with all the herbs in it), and words are each fixed to one power. So no word ever takes more than a slice, and the slice a compound takes at the number-place is number-in-general. This is 2.250's omni-significance seen from the other side: the object overflows every word, and each word is narrow by constitution.

3.14.105 unformalized

अव्ययानां च यो धर्मो यश्च भेदवतां क्रमः ।
अभिन्नव्यपदेशार्हमन्तरालं तदेतयोः ॥ १०५ ॥

avyayānāṃ ca yo dharmo yaśca bhedavatāṃ kramaḥ |
abhinnavyapadeśārhamantarālaṃ tadetayoḥ || 105 ||

The property of the indeclinables, and the sequence of those possessing difference — the intermediate between these two deserves the designation 'non-different'.

The compound placed on a scale. Avyayanam ca yo dharmo: the *property* of the *indeclinables* — the avyaya, which has no number at all, admits no differentiation, and is one end of the scale. Yas ca bhedavatam kramah: and the *sequence* of those *possessing difference* — inflected words with their numbers laid out in order: the other end. Abhinna-vyapadesarham antaralam tad etayoh: the *intermediate* between these two *deserves the designation 'non-different'*. Antarala: the space between. So the compound is neither numberless like an indeclinable nor numbered like a noun; it occupies the middle, and abheda (3.14.99's abheda-ekatva) is the name that middle earns. Vyapadesarha — *deserving the designation* — is careful: the name is apt, and it is still a name.

3.14.106 unformalized

अलुकश्चैकवद्भावस्तस्मिन् सति न शिष्यते ।
स च गोषुचरादीनां धर्मोऽस्ति वचनान्तरे ॥ १०६ ॥

alukaścaikavadbhāvastasmin sati na śiṣyate |
sa ca goṣucarādīnāṃ dharmo'sti vacanāntare || 106 ||

And the singular-treatment of the non-elided is not taught when that exists; and that is the property of 'gosucara' and the rest in another number.

A rule made redundant by 3.14.99's doctrine. Alukas ca ekavad-bhavas tasmin sati na sisyate: the *singular-treatment* of the *aluk* — compounds where the case-ending is *not* elided — is *not taught* when *that* (the abheda-ekatva) *exists*: if the compound already has non-different unity, no rule need impose the singular on it. Sa ca gosucaradinam dharmo 'sti vacanantare: and that is the *property* of *'gosucara'* (grazing among cows) and the rest, in *another number*. So the aluk-compounds, which visibly keep their endings, still get their singularity from the doctrine rather than from a rule — and the rule that supplies it is idle (3.14.41's prohibitions with no purpose; 3.14.59's rules with bounded domains). The doctrine of 3.14.99 is paying for itself in rule-economy.

3.14.107 unformalized

जातौ द्विवचनाभावात् तद् वृत्तिषु न विद्यते ।
प्रत्याख्याने तु योगस्य द्रव्ये गोषुचरादयः ॥ १०७ ॥

jātau dvivacanābhāvāt tad vṛttiṣu na vidyate |
pratyākhyāne tu yogasya dravye goṣucarādayaḥ || 107 ||

From the absence of the dual in the universal, that does not exist in derivative formations; but when the rule is rejected, 'gosucara' and the rest are in the substance.

Jatau dvivacana-abhavat: from the *absence of the dual* in the *universal* — a jati has no two: universals do not come in pairs (3.1's the universal is one, everywhere, always). Tad vrttisu na vidyate: *that* — the dual — *does not exist* in *derivative formations*, since the compound's meaning has gone over to the general (3.14.101's jati-bhaga; 3.14.104's number-in-general). Pratyakhyane tu yogasya dravye gosucaradayah: but in the *rejection* of the rule, *'gosucara'* and the rest are in the *substance* — if one throws out the rule, these compounds must denote substances instead, and substances do come in twos. So the choice is exact: keep the rule and the compound means a universal with no dual, or drop it and the compound means a substance that has one. The samuddesa lays out the fork and lets the reader see what the rule is buying.

3.14.108-109 unformalized

आश्रयाद् भेदवत्तायाः सर्वभेदसमन्वयः ।
द्रव्याभिधानपक्षोऽपि जात्याख्यायां न विद्यते ॥ १०८ ॥
सर्वद्रव्यगतिश्चैवमेकशेषश्च नोच्यते ।
प्रत्याख्यातेऽन्यथा सूत्रे भिन्नद्रव्यगतिर्भवेत् ॥ १०९ ॥

āśrayād bhedavattāyāḥ sarvabhedasamanvayaḥ |
dravyābhidhānapakṣo'pi jātyākhyāyāṃ na vidyate || 108 ||
sarvadravyagatiścaivamekaśeṣaśca nocyate |
pratyākhyāte'nyathā sūtre bhinnadravyagatirbhavet || 109 ||

(108) From the substratum's possession of difference, there is concomitance with all differences; the view that substance is expressed also does not obtain in the naming of a universal. (109) And thus the reaching of all substances, and ekasesa, is not stated; if the sutra were rejected otherwise, there would be a reaching of different substances.

The fork of 3.14.107 worked out. (108): asrayad bhedavattayah sarva-bheda-samanvayah — from the *substratum's* possession of *difference*, there is *concomitance with all differences*: if the compound rests on a substance, every difference the substance has comes along. Dravyabhidhana-pakso 'pi jaty-akhyayam na vidyate: the view that *substance is expressed* does *not obtain* in the *naming of a universal* — one cannot have both. (109): sarva-dravya-gatis caivam ekasesas ca nocyate — *thus* the *reaching of all substances*, and *ekasesa* (3.14.30's residue-of-one), is *not stated*: the doctrine makes both rules unnecessary. Pratyakhyate 'nyatha sutre bhinna-dravya-gatir bhavet: if the *sutra* were *rejected otherwise*, there would be a *reaching of different substances* — the wrong result. So the accounting is complete: take the universal, and two rules become idle; take the substance, and the forms come out wrong.

3.14.110 unformalized

वृत्तौ यो युक्तवद्भावो वरणादिषु शिष्यते ।
अभेदैकत्वसङ्खयायां गोदौ तत्र न सिध्यति ॥ ११० ॥

vṛttau yo yuktavadbhāvo varaṇādiṣu śiṣyate |
abhedaikatvasaṅkhayāyāṃ godau tatra na sidhyati || 110 ||

The 'treatment as though joined' which is taught in derivative formation, in 'varana' and the rest — in the number of non-different unity, in 'goda' it does not succeed there.

A difficulty for the doctrine, honestly filed. Vrttau yo yuktavad-bhavo varanadisu sisyate: the *yuktavad-bhava* — Panini's provision that a derivative be treated *as though* it were the analysed phrase, for gender and number — taught in *'varana'* and the rest. Abhedaikatva-sankhyayam godau tatra na sidhyati: in the *number of non-different unity* (3.14.99), in *'goda'* it *does not succeed*. So the rule requires the compound to behave as its phrase does, and the doctrine has just said the compound has a number its phrase never had. In 'goda' the two prescriptions collide and the rule fails. Bhartrhari raises the objection against his own account and states it flatly — which is the samuddesa's manner throughout (3.14.34's the Bhasya found it hard). The following verses will have to answer it.

3.14.111 unformalized

प्राग्वृत्तेर्युक्तवद्भावे षष्ठी भेदाश्रया भवेत् ।
वृत्तौ सङ्खयाविशेषाणां त्यागाद् भेदो निवर्तते ॥ १११ ॥

prāgvṛtteryuktavadbhāve ṣaṣṭhī bhedāśrayā bhavet |
vṛttau saṅkhayāviśeṣāṇāṃ tyāgād bhedo nivartate || 111 ||

Before the derivative formation, under 'treatment as though joined', the genitive would rest on difference; in the formation, from the abandonment of the number-particulars, difference withdraws.

The answer to 3.14.110. Prag vrtter yuktavad-bhave sasthi bhedasraya bhavet: *before* the derivative formation, under *yuktavad-bhava*, the *genitive* would *rest on difference* — the rule looks back to the analysed phrase, where the genitive really did relate two differentiated things. Vrttau sankhya-visesanam tyagad bhedo nivartate: *in* the formation, from the *abandonment* of the *number-particulars*, *difference* *withdraws*. So the two prescriptions do not collide after all: they apply at different moments. Yuktavad-bhava reaches back to a stage where difference obtained; abheda-ekatva describes the stage after. The objection assumed a single time, and there are two — which is 3.14.61's antarangatva argument (what was formed first leaves its mark) in a semantic key.

3.14.112 unformalized

विद्यमानसु सङ्खयासु केचित् सङ्खयान्तरं विदुः ।
अभेदाख्यमुपग्राहि वृत्तौ तच्चोपजायते ॥ ११२ ॥

vidyamānasu saṅkhayāsu kecit saṅkhayāntaraṃ viduḥ |
abhedākhyamupagrāhi vṛttau taccopajāyate || 112 ||

While the numbers exist, some know another number — called 'non-difference', comprehending; and that arises in the derivative formation.

The doctrine restated with its authorities. Vidyamanasu sankhyasu: *while the numbers exist* — the members' numbers are not annihilated; they are there. Kecit sankhyantaram viduh: *some* know *another number* — a further one, over and above them. Abhedakhyam upagrahi: called *'non-difference'*, *comprehending* — upagrahin: taking-in, including; the new number does not replace the old ones but *contains* them. Vrttau tac ca upajayate: and *that* *arises* in the *derivative formation*. So the honey-image of 3.14.100 is now precise: the herb-juices are still there (vidyamanasu), and the honey is a further thing that has taken them in (upagrahi). Nothing is destroyed by composition; something is added that comprehends what it was made of.

3.14.113 unformalized

व्यापारं याति भेदाख्यैस्तत् स्वैरवयवैः क्वचित् ।
आत्मा भेदानपेक्षोऽस्य क्वचिदेति निमित्तताम् ॥ ११३ ॥

vyāpāraṃ yāti bhedākhyaistat svairavayavaiḥ kvacit |
ātmā bhedānapekṣo'sya kvacideti nimittatām || 113 ||

That goes into operation somewhere through its own members, called 'difference'; somewhere its self, independent of the differences, attains causality.

Why the compound behaves two ways. Vyaparam yati bhedakhyais tat svair avayavaih kvacit: *that* goes into *operation* *somewhere* through its *own members*, called *'difference'* — in some rules the compound acts through its parts, and the parts are live. Atma bhedanapekso 'sya kvacid eti nimittatam: *somewhere* its *self*, *independent of the differences*, attains *causality* — nimittata: the state of being the occasioning cause. So the compound has two modes: through-the-parts and as-a-whole, and which one operates depends on the rule. This is the formal statement of what 3.14.111 discovered — the same compound answers to yuktavad-bhava through its members and to the number-rules through its self — and it saves the doctrine without softening it. The whole is one (3.14.96); it is also, when a rule asks, addressable through parts it does not have.

3.14.114 unformalized

दास्याः पतिरिति व्यक्तो गोदाविति च दृश्यते ।
व्यापारभेदः सङ्खयायास्तस्यादेवं व्यवस्थितः ॥ ११४ ॥

dāsyāḥ patiriti vyakto godāviti ca dṛśyate |
vyāpārabhedaḥ saṅkhayāyāstasyādevaṃ vyavasthitaḥ || 114 ||

'The slave-woman's husband' is manifest, and 'godau' is seen; the difference in the number's operation is therefore settled in this way.

The two modes of 3.14.113 exhibited. Dasyah patir iti vyaktah: *'dasyah patih'* — the slave-woman's husband, the tradition's standing case of a phrase that resists compounding — is *manifest*: here the parts are live and the genitive stands. Godav iti ca drsyate: and *'godau'* — the dual, the very form 3.14.110 said would not come out — *is seen*: it is there in usage, whatever the theory says. Vyapara-bhedah sankhyayah tasmad evam vyavasthitah: the *difference in the number's operation* is *therefore* *settled* in this way. So the doctrine is not being defended a priori: two forms are pointed at, one behaving through parts and one as a whole, and the two-mode account is read off them. Drsyate — *is seen* — is the argument. Usage settles it (3.14.43's loka-vyavastha).

3.14.115 unformalized

द्वयादीनां च द्विपुत्रादौ बाह्यो भेदो निवर्तते ।
विभक्तिवाच्यः स्वार्थत्वान्निमित्तं त्ववतिष्ठते ॥ ११५ ॥

dvayādīnāṃ ca dviputrādau bāhyo bhedo nivartate |
vibhaktivācyaḥ svārthatvānnimittaṃ tvavatiṣṭhate || 115 ||

And of 'dvi' and the rest, in 'dviputra' and the like, the external difference, expressible by the case-ending, withdraws, since it is the word's own meaning; but the cause remains.

Numerals inside compounds. Dvyadinam ca dviputradau bahyo bhedo nivartate: of *'dvi'* and the rest, in *'dviputra'* (two-sonned) and the like, the *external* difference *withdraws* — bahya: outer, the duality that would have been carried by an ending. Vibhakti-vacyah svarthatvat: *expressible by the case-ending*, since it is the word's *own meaning* — 'dvi' *means* two, so it does not also need an ending to say two: what the stem says, the ending need not (3.11's the numeral and the number-ending saying the same thing twice). Nimittam tv avatisthate: but the *cause* *remains* — the duality is still the occasion of the form, even with its outer expression gone. So the number is not lost; it is relocated from the ending into the stem's own sense, and it goes on causing (3.14.113's nimittata).

3.14.116 unformalized

द्वित्वोपसर्जने सङ्घे द्विशब्दस्तत्र वर्तते ।
सोऽयमित्यभिसम्बन्धादुभशब्दे न तत्तथा ॥ ११६ ॥

dvitvopasarjane saṅghe dviśabdastatra vartate |
so'yamityabhisambandhādubhaśabde na tattathā || 116 ||

In an aggregate where duality is subordinate, the word 'dvi' operates there, from the connection 'this is that'; in the word 'ubha' it is not so.

Dvitva-upasarjane sanghe dvi-sabdas tatra vartate: in an *aggregate* where *duality* is *subordinate*, the word *'dvi'* *operates* there — where twoness is the subordinate element (3.14.95's upasarjana), 'dvi' can stand. So 'yam ity abhisambandhat: from the *connection* 'this is that' — the identification that lets a numeral name a group. Ubha-sabde na tat tatha: in the word *'ubha'* (both) it is *not so*. So 'ubha' will not do what 'dvi' does: it cannot be subordinated, cannot enter the identification. The difference is that 'dvi' counts and 'ubha' *totalizes* — it says not two but *the* two, both of them, and a totality has nothing above it to be subordinate to. The following verses work out what follows from this.

3.14.117 unformalized

उभयस्तस्य तुल्यार्थो वृत्तौ नित्यं प्रयुज्यते ।
सूत्रेऽपि नित्यग्रहणं तदर्थमभिधीयते ॥ ११७ ॥

ubhayastasya tulyārtho vṛttau nityaṃ prayujyate |
sūtre'pi nityagrahaṇaṃ tadarthamabhidhīyate || 117 ||

'Ubhaya', equal in meaning to it, is always employed in the derivative formation; in the sutra too, the taking-up of 'nitya' is stated for that purpose.

Ubhayas tasya tulyartho vrttau nityam prayujyate: *'ubhaya'* — *equal in meaning* to 'ubha' — is *always* employed in the *derivative formation*: where a compound is wanted, 'ubhaya' goes in and 'ubha' does not. Sutre 'pi nitya-grahanam tad-artham abhidhiyate: in the *sutra* too, the taking-up of *'nitya'* is stated *for that purpose*. So Panini's word 'nitya' in the relevant rule exists precisely to force this substitution — and here, unlike 3.14.40's redundant nitya, the word earns its place. The pair is instructive: the same technical device is idle in one rule and load-bearing in another, and only inspection tells which. Bhartrhari has no general verdict on the grammar's economy; he checks case by case.

3.14.118 unformalized

आपि के चापरार्थत्वान्नाभेद उपजायते ।
उभे इति ततः स्वार्थे भेदवृत्तिः प्रयुज्यते ॥ ११८ ॥

āpi ke cāparārthatvānnābheda upajāyate |
ubhe iti tataḥ svārthe bhedavṛttiḥ prayujyate || 118 ||

And in 'ap' and 'ka', from their serving another's purpose, non-difference does not arise; therefore 'ubhe' — the difference-usage is employed, in its own meaning.

Api ke ca aparartatvat na abheda upajayate: in *'ap'* and *'ka'* — the affixes — from their *serving another's purpose*, *non-difference* does *not arise*: these affixes exist for someone else's sake and so cannot bring about the fusion of 3.14.99. Ubhe iti tatah svarthe bheda-vrttih prayujyate: *therefore* *'ubhe'* — the *difference-usage* is employed, *in its own meaning*. So where the affix cannot fuse, the form keeps its difference and stands in its own sense: 'ubhe', the dual, with its twoness intact. This confirms 3.14.116 — 'ubha' resists subordination and therefore resists fusion, and its dual survives everywhere.
contested: Pada a is compressed and its reading is not certain. We take 'api ke ca' as naming the affixes ap (tap) and ka, whose 'serving another's purpose' (apararthatva) then explains why abheda fails. The line can also be construed with 'api' as the particle and 'ke' otherwise resolved; the verse's conclusion — that 'ubhe' keeps its difference — is unaffected.

3.14.119 unformalized

स्त्रीत्वाभिधानपक्षेऽपि गुणभावविपर्ययः ।
स्वभवादपरार्थत्वात् तत्र भेदो न हीयते ॥ ११९ ॥

strītvābhidhānapakṣe'pi guṇabhāvaviparyayaḥ |
svabhavādaparārthatvāt tatra bhedo na hīyate || 119 ||

Even on the view that femininity is expressed, there is a reversal of subordination; from its own nature, from its serving another's purpose, difference is not lost there.

The same result on the rival account. Stritva-abhidhana-pakse 'pi: even on the view that *femininity is expressed* — that the feminine affix says femininity rather than merely revealing it (3.13's vacaka/dyotaka quarrel about gender). Guna-bhava-viparyayah: there is a *reversal of subordination* — what was subordinate becomes principal and vice versa. Svabhavad apararthatvat tatra bhedo na hiyate: from its *own nature*, from its *serving another's purpose*, *difference* is *not lost* there. So whichever way one decides the gender-question, 'ubhe' keeps its difference — the affix serves another and cannot fuse. Bhartrhari's habit again: show that the result survives the disputed premise, so the dispute need not be settled to get on (3.14.98's vacika va dyotika vapi, left open for the same reason).

3.14.120 unformalized

तस्माद् द्विवचनाट्टापश्चोभयोऽन्यत्र दृश्यते ।
प्रत्ययं तयपं हित्वा नास्त्युत्तरपदे पुनः ॥ १२० ॥

tasmād dvivacanāṭṭāpaścobhayo'nyatra dṛśyate |
pratyayaṃ tayapaṃ hitvā nāstyuttarapade punaḥ || 120 ||

Therefore, from the dual and from 'tap', 'ubhaya' is seen elsewhere; leaving aside the affix 'tayap', it does not exist again in the second member.

Tasmad dvivacanat tapas ca ubhayo 'nyatra drsyate: *therefore*, from the *dual* and from *'tap'*, *'ubhaya'* is *seen elsewhere* — the form turns up outside the compound-context, where the dual and the feminine affix operate normally. Pratyayam tayapam hitva na asty uttarapade punah: *leaving aside* the affix *'tayap'*, it does *not exist* *again* in the *second member*. So 'ubhaya' is derived by tayap and, once derived, cannot recur as the latter member of a compound. The restriction is exactly what 3.14.116-119 predicted: a totalizer resists every subordinate position, and the second member of a compound is a subordinate position. The rule falls out of the semantics rather than being stipulated — which is the pattern the whole samuddesa has been demonstrating.

3.14.121 unformalized

प्राप्तिः प्रगृह्यसंज्ञाया न स्यात् प्रत्ययलक्षणात् ।
कुमार्यगारे न ह्यस्ति समासो वचनान्तरे ॥ १२१ ॥

prāptiḥ pragṛhyasaṃjñāyā na syāt pratyayalakṣaṇāt |
kumāryagāre na hyasti samāso vacanāntare || 121 ||

The obtaining of the term 'pragrhya' would not arise from the affix-indication; for in 'kumaryagara' there is no compound in another number.

Praptih pragrhya-samjnaya na syat pratyaya-laksanat: the *obtaining* of the term *'pragrhya'* — the designation for a final vowel exempt from sandhi, which duals characteristically get — would *not arise* from the *affix-indication*: pratyaya-laksana, the principle that a rule conditioned by an affix still applies once the affix is elided. Kumaryagare na hy asti samaso vacanantare: for in *'kumaryagara'* (the girl's chamber) there is *no compound* in *another number*. So the pragrhya-designation cannot be got at through an elided dual, because the compound in question is not in that number at all: there is no dual there to be exempted, elided or otherwise. This is 3.14.98's number-difference withdrawing, cashed out in a sandhi-rule — the compound's number is not the members' number, and the rules that reach for the members' number find nothing.

3.14.122 unformalized

एकद्वयोर्यञादीनां विभाषा लुङ् न कल्पते ।
यौष्माकस्तावकश्चेति भेदाभावान्न सिध्यति ॥ १२२ ॥

ekadvayoryañādīnāṃ vibhāṣā luṅ na kalpate |
yauṣmākastāvakaśceti bhedābhāvānna sidhyati || 122 ||

For one and two, the optional elision of 'yañ' and the rest is not conceivable; 'yausmaka' and 'tavaka' do not succeed, from the absence of difference.

Eka-dvayor yanadinam vibhasa lun na kalpate: for *one* and *two*, the *optional elision* of *'yan'* and the rest is *not conceivable* — the patronymic affix's optional luk requires a plurality to operate on, and one and two do not supply it. Yausmakas tavakas ceti bhedabhavan na sidhyati: *'yausmaka'* and *'tavaka'* — the derivatives from 'yusmad' and 'tvad' — do *not succeed*, from the *absence of difference*. So two rule-failures from the same cause: where the number has fused (3.14.99's abheda-ekatva), the rules that need a difference to work on have nothing. Bhartrhari is cataloguing the doctrine's costs as carefully as its benefits (3.14.106's benefit, this verse's cost), and he does not net them out.

3.14.123 unformalized

दृष्टो गार्ग्यतरे भेदस्तथा गर्गतरा इति ।
युष्मत्पिता त्वत्पितेति तथादेशौ व्यवस्थितौ ॥ १२३ ॥

dṛṣṭo gārgyatare bhedastathā gargatarā iti |
yuṣmatpitā tvatpiteti tathādeśau vyavasthitau || 123 ||

Difference is seen in 'gargyatara', and likewise in 'gargatarah'; 'yusmat-pita' and 'tvat-pita' — thus the two substitutes are settled.

The counter-evidence, admitted. Drsto gargyatare bhedah: *difference* is *seen* in *'gargyatara'* — the comparative, which requires the difference 3.14.122 said was absent. Tatha gargatara iti: and *likewise* in *'gargatarah'*. Yusmat-pita tvat-piteti tatha adesau vyavasthitau: *'yusmat-pita'* and *'tvat-pita'* — thus the *two substitutes* are *settled*: the plural-based and singular-based substitutes both occur, each in its own domain. Drstah is again the operative word (3.14.114's drsyate): the forms exist and the theory must accommodate them. So difference does survive in some derivatives, and 3.14.124 must say how — which it does, by relocating the number into the base.

3.14.124 unformalized

उपाधिभूता या सङ्खया प्रकृतौ समवस्थिता ।
आदेशैः संज्ञया वापि विभक्त्याव्यज्यते विना ॥ १२४ ॥

upādhibhūtā yā saṅkhayā prakṛtau samavasthitā |
ādeśaiḥ saṃjñayā vāpi vibhaktyāvyajyate vinā || 124 ||

That number which, being an adjunct, is settled in the base, is manifested by substitutes, or by a technical term, even without a case-ending.

The resolution of 3.14.122-123. Upadhi-bhuta ya sankhya prakrtau samavasthita: that *number* which, being an *upadhi* — an adjunct, a qualifying condition rather than a constituent — is *settled in the base*. Adesaih samjnaya vapi vibhaktya vyajyate vina: is *manifested* by *substitutes*, or by a *technical term*, *even without a case-ending*. Vyajyate: is *manifested*, revealed — not said (3.14.98's dyotika, and the word is chosen). So the number that 3.14.98 said withdraws from the compound has not gone anywhere: it has become an adjunct of the base, where no ending expresses it and the substitutes betray it. This is why 3.14.123's forms come out: 'tvat-pita' shows its singular through its substitute, having no ending to show it with. Number in a compound is not spoken; it is given away.

3.14.125 unformalized

शौर्पिके मासजाते च परिमाणं स्वभावतः ।
उपाधिभूतामाश्रित्य सङ्ख्यां भेदेन वर्तते ॥ १२५ ॥

śaurpike māsajāte ca parimāṇaṃ svabhāvataḥ |
upādhibhūtāmāśritya saṅkhyāṃ bhedena vartate || 125 ||

In 'saurpika' and in 'masa-jata', the measure, by its own nature, resting on the number which is an adjunct, operates by way of difference.

The upadhi-number of 3.14.124 at work. Saurpike masa-jate ca parimanam svabhavatah: in *'saurpika'* (measured by a surpa) and *'masa-jata'* (a month old), the *measure*, by its *own nature*. Upadhi-bhutam asritya sankhyam bhedena vartate: *resting on* the *number* which is an *adjunct*, operates *by way of difference*. So measure-words carry their difference through the adjunct-number in the base, not through any ending: 'masa-jata' says one month and no ending says it. Svabhavatah is doing work — the measure is *by its own nature* quantitative, so it needs no help to reach the number that is sitting in its base. This finishes the account 3.14.84-86 began, and the two now agree: measure is certified by machinery and carried by the base.

3.14.126 unformalized

वयस्विनि परिच्छेदः क्रीते चापि न गम्यते ।
इष्टो भेदादृते तत्र परिमाणमनर्थकम् ॥ १२६ ॥

vayasvini paricchedaḥ krīte cāpi na gamyate |
iṣṭo bhedādṛte tatra parimāṇamanarthakam || 126 ||

In 'vayasvin' and in 'krita', the delimitation is not understood; the sense intended, apart from difference — there the measure is purposeless.

The contrast case. Vayasvini paricchedah krite capi na gamyate: in *'vayasvin'* (aged, possessing age) and in *'krita'* (bought), the *delimitation* is *not understood* — pariccheda: the cutting-off, the definite quantity. 'Vayasvin' says one has age, not how much; 'krita' says bought, not for how much. Isto bhedad rte tatra parimanam anarthakam: the sense *intended*, *apart from difference* — there the *measure* is *purposeless*. Anarthaka: without purpose. So these words carry no adjunct-number at all (3.14.124), and reaching for one gets nothing. The pair with 3.14.125 is the point: some derivatives have quantity in their base and some have only quality, and no rule can tell which — one has to know the word. Which is 3.14.43 once more: the world's settled arrangement, and the science following after.

3.14.127 unformalized

भिन्नस्याभेदवचनात् प्रस्थादिभ्यः शसो विधिः ।
तद्धर्मत्वादभेदात् तु घटादिभ्यो न दृश्यते ॥ १२७ ॥

bhinnasyābhedavacanāt prasthādibhyaḥ śaso vidhiḥ |
taddharmatvādabhedāt tu ghaṭādibhyo na dṛśyate || 127 ||

From the expression of non-difference of what is different, the rule of 'sas' after 'prastha' and the rest; but from non-difference, because that is its property, it is not seen after 'ghata' and the rest.

Bhinnasya abheda-vacanat prasthadibhyah saso vidhih: from the *expression of non-difference* of what is *different*, the rule of *'sas'* after *'prastha'* and the rest — the distributive affix, giving 'prastha-sah', prastha by prastha: many prasthas said as one distributive sweep. Tad-dharmatvad abhedat tu ghatadibhyo na drsyate: but from *non-difference*, *because that is its property*, it is *not seen* after *'ghata'* and the rest. So the distributive attaches to measure-words and not to pots: a measure is intrinsically a repeatable unit — abheda is *its property* — and a pot is not. This is 3.14.125-126's finding again (quantity is in some bases and not others), now settling which stems take an affix. The grammar's gana is not arbitrary; it tracks which words have a unit built into them.

3.14.128 unformalized

श्रूयते वचनं यत्र भावस्तत्र विशिष्यते ।
निवर्तते यद् वचनं तस्य भोवो न विद्यते ॥ १२८ ॥

śrūyate vacanaṃ yatra bhāvastatra viśiṣyate |
nivartate yad vacanaṃ tasya bhovo na vidyate || 128 ||

Where the number-ending is heard, there the state is specified; that ending which withdraws — of that, there is no being.

A methodological principle, stated with force. Sruyate vacanam yatra bhavas tatra visisyate: *where* the *number-ending* is *heard*, *there* the *state* is *specified* — an audible ending does specifying work. Nivartate yad vacanam tasya bhavo na vidyate: that ending which *withdraws* — of *that*, there is *no being*. Bhavo na vidyate: it has no existence at all. So Bhartrhari refuses the halfway house: an elided ending is not a quiet ending, it is *not there*. This is stronger than 3.14.87's asadhutva-nivrtty-artham (where the elided matup was allowed to exist in order to certify) and stands in tension with it — the samuddesa is exploring both, and the tension is the subject-matter (3.14.113's two modes). What is not heard does not specify.

3.14.129 unformalized

कार्यं सत्ताश्रयं शास्त्रदप्रवृत्तिरदर्शनम् ।
वाक्ये दृष्टं यदत्यन्तमभावस्तस्य वृत्तिषु ॥ १२९ ॥

kāryaṃ sattāśrayaṃ śāstradapravṛttiradarśanam |
vākye dṛṣṭaṃ yadatyantamabhāvastasya vṛttiṣu || 129 ||

The operation rests on existence; from the science, non-application is non-appearance. What is seen utterly in the sentence — of that there is absence in derivative formations.

The consequence of 3.14.128. Karyam satta-asrayam: the *operation* rests on *existence* — a rule cannot act on what is not there. Sastrad apravrttir adarsanam: from the *science*, *non-application* is *non-appearance* — the rule's failing to apply *is* the thing's not showing up; there is no further fact. Vakye drstam yad atyantam abhavas tasya vrttisu: what is *seen utterly* in the *sentence* — of that there is *absence* in *derivative formations*. So the sentence's visible articulation is *absent* — not hidden, not suppressed — in the compound. This is the samuddesa's core claim (3.14.49's atyanta-bheda) stated in the vocabulary of rule-application: nothing withdrew into the compound, because nothing was there to withdraw.

3.14.130 unformalized

संज्ञाविषयभेदार्थं प्रसक्तादर्शनं स्मृतम् ।
क्षूयमाणं तु वचनं विशिष्टमुपलभ्यते ॥ १३० ॥

saṃjñāviṣayabhedārthaṃ prasaktādarśanaṃ smṛtam |
kṣūyamāṇaṃ tu vacanaṃ viśiṣṭamupalabhyate || 130 ||

For the sake of differentiating the domain of the technical term, the non-appearance of what was applicable is taught; but the ending which is heard is apprehended as specified.

Samjna-visaya-bhedartham prasakta-adarsanam smrtam: for the sake of *differentiating the domain* of the *technical term*, the *non-appearance* of what was *applicable* is *taught* — prasakta: what would have applied; adarsana: its not showing. The grammar deliberately teaches that an applicable thing fails to appear, and the teaching's purpose is to carve the term's domain. So the elision is a demarcation-device, not a report (3.14.42's rules as nidarsana). Sruyamanam tu vacanam visistam upalabhyate: but the ending which *is heard* is apprehended as *specified*. The audible one does real work; the elided one draws a boundary. Both are the grammar's; only one is the word's.
contested: Pada c is transmitted in our source as 'ksuyamanam', which yields no sense here; we read 'sruyamanam' ('being heard'), which the contrast with 3.14.128's 'sruyate' and with pada d's 'upalabhyate' requires. We note the source reading rather than silently normalising it.

3.14.131 unformalized

अभावो वा लुको यत्र रूपवान् वा विधीयते ।
व्यभिचारान्निमित्तस्य तत्रासाधुः प्रसज्यते ॥ १३१ ॥

abhāvo vā luko yatra rūpavān vā vidhīyate |
vyabhicārānnimittasya tatrāsādhuḥ prasajyate || 131 ||

Where absence, or elision, or a substitute possessing form is prescribed — there, from the deviation of the cause, incorrectness follows.

The cost of the whole apparatus. Abhavo va luko yatra rupavan va vidhiyate: where *absence*, or *luk*, or a *substitute possessing form* is *prescribed* — the grammar's three ways of handling what is not audibly there. Vyabhicaran nimittasya tatra asadhuh prasajyate: *there*, from the *deviation* of the *cause*, *incorrectness* *follows*. Vyabhicara: straying, failure of invariable concomitance — the occasioning cause does not hold steady across the cases. So wherever the grammar posits an absence, the conditioning breaks down somewhere and a bad form is generated. This is 3.14.76's vyutpattir anavasthita made concrete: the derivational machinery leaks, and it leaks precisely at the points where it posits what is not there (3.14.74's imagined parts).

3.14.132 unformalized

भेदः संख्या विशेषोवाव्याख्यातो वृत्तिवाक्ययोः ।
सर्वत्रैव विशेषस्तु नावश्यं तादृशो भवेत् ॥ १३२ ॥

bhedaḥ saṃkhyā viśeṣovāvyākhyāto vṛttivākyayoḥ |
sarvatraiva viśeṣastu nāvaśyaṃ tādṛśo bhavet || 132 ||

Difference, number, or particular has been expounded for the derivative formation and the sentence; but the particular need not everywhere be of that kind.

A caution against over-generalizing what has been established. Bhedah samkhya viseso va vyakhyato vrtti-vakyayoh: *difference*, *number*, or *particular* has been *expounded* for the *derivative formation* and the *sentence* — the samuddesa's results so far. Sarvatra eva visesas tu na avasyam tadrso bhavet: but the *particular* need *not everywhere* be *of that kind*. Na avasyam: not necessarily. So the analysis was of certain cases and does not automatically extend. Bhartrhari is guarding his own conclusions against the very error he charged others with (3.14.76's taking the upayas for true): a result reached on some forms is an upaya for those forms, and carrying it everywhere would make it a doctrine. The samuddesa polices itself by the standard it set.

3.14.133 unformalized

जातेश्च भेदहेतुत्वान्न लिङ्गेन विशेष्यते ।
प्रधानं मृगदुग्धादौ गार्गीपुत्रे न स क्रमः ॥ १३३ ॥

jāteśca bhedahetutvānna liṅgena viśeṣyate |
pradhānaṃ mṛgadugdhādau gārgīputre na sa kramaḥ || 133 ||

And from the universal's being the cause of difference, the principal is not qualified by gender in 'mrga-dugdha' and the rest; in 'gargi-putra' that sequence does not hold.

Jates ca bheda-hetutvat: from the *universal's* being the *cause of difference* — the jati is what differentiates, so it does the qualifying. Na lingena visesyate pradhanam mrga-dugdhadau: the *principal* is *not* qualified by *gender* in *'mrga-dugdha'* (deer's milk) and the rest — the milk is not made feminine by the doe: gender does not reach across into the principal. Gargi-putre na sa kramah: in *'gargi-putra'* (Gargi's son) that *sequence* does *not hold* — here the feminine of the first member is doing something, and the sequence reverses. So two compounds, alike in form, differ in whether gender penetrates. This is 3.14.132's caution being exercised at once: the particular is not everywhere of the same kind, and here is a pair to prove it.

3.14.134 unformalized

अभेदे लिङ्गसंख्याभ्यां योगाच्छुक्लं पटाइति ।
प्रसक्ते शास्त्रमारब्धं सिद्धये लिङ्गसंख्ययोः ॥ १३४ ॥

abhede liṅgasaṃkhyābhyāṃ yogācchuklaṃ paṭāiti |
prasakte śāstramārabdhaṃ siddhaye liṅgasaṃkhyayoḥ || 134 ||

In non-difference, from connection with gender and number, 'suklam patah' being applicable — the science is undertaken for the establishing of gender and number.

Why the grammar has gender-and-number rules at all. Abhede linga-samkhyabhyam yogat: in *non-difference*, from *connection* with *gender and number*. Suklam pata iti prasakte: *'suklam patah'* — the ungrammatical mixture, a neuter singular adjective with a masculine plural noun — being *applicable*: without a rule, that is what would come out, since abheda leaves the qualifier with no reason to agree. Sastram arabdham siddhaye linga-samkhyayoh: the *science* is *undertaken* for the *establishing* of *gender and number*. So agreement is not a fact of the meaning — the meaning is undivided and has no agreement in it (3.14.92's anusyuta) — and the grammar must impose it. This is a striking admission: concord is the sastra's contribution, not the world's.

3.14.135 unformalized

परार्थं शेषभावं यो वृत्तिषु प्रतिपद्यते ।
गुणो विशेषणत्वेन स सूत्रे व्यपदिश्यते ॥ १३५ ॥

parārthaṃ śeṣabhāvaṃ yo vṛttiṣu pratipadyate |
guṇo viśeṣaṇatvena sa sūtre vyapadiśyate || 135 ||

That quality which, for another's sake, attains the state of remainder in derivative formations — that is designated in the sutra by way of being a qualifier.

Pararthaam sesa-bhavam yo vrttisu pratipadyate: that *quality* which, *for another's sake*, attains the *state of remainder* in *derivative formations* — sesa: what is left over, the residue after the principal has taken what it needs. Guno visesanatvena sa sutre vyapadisyate: that is *designated* in the *sutra* by way of being a *qualifier*. Vyapadisyate: is *designated* — the samuddesa's careful verb for a name that is apt without being a description (3.14.105's vyapadesarha). So Panini's 'visesana' names whatever is left over serving another's purpose, and the naming is a convenience. The qualifier is not a thing in the meaning; it is the grammar's label for a residue — which is why 3.14.134's concord had to be imposed on it.

3.14.136 unformalized

शब्दान्तरत्वाद् वाक्येषु विशेषा यद्यपि श्रुताः ।
वृत्तेरभिन्नरूपत्वात् तेषु वृत्तिर्न विद्यते ॥ १३६ ॥

śabdāntaratvād vākyeṣu viśeṣā yadyapi śrutāḥ |
vṛtterabhinnarūpatvāt teṣu vṛttirna vidyate || 136 ||

Although particulars are heard in sentences, from their being other words — from the derivative formation's non-different form, in them there is no derivative formation.

Sabdantaratvad vakyesu visesa yady api srutah: although *particulars* are *heard* in *sentences*, from their being *other words* — the phrase has audible particulars, because the phrase is a different word (3.14.49's atyanta-bheda; 3.14.64). Vrtter abhinna-rupatvat tesu vrttir na vidyate: from the *derivative formation's* *non-different form*, in *them* there is *no* derivative formation. So one cannot find the compound inside the phrase: the phrase's particulars are not the compound's members, and there is no compounding going on in a phrase at all. This closes the door 3.14.52's rival wanted open. The compound is not a phrase that has undergone anything; the two are different words, and the particulars one hears belong to one of them only.

3.14.137-138 unformalized

रूपाच्च शब्दसंस्कारः सामान्यविषयो यतः ।
तस्मात् तदाश्रयं लिङ्गं वचनं च प्रसज्यते ॥ १३७ ॥
सलिङ्गं च ससङ्ख्यं च ततो द्रव्याभिधायिना ।
सम्बध्यते पदं तत्र तयोर्भिन्ना श्रुतिर्भवेत् ॥ १३८ ॥

rūpācca śabdasaṃskāraḥ sāmānyaviṣayo yataḥ |
tasmāt tadāśrayaṃ liṅgaṃ vacanaṃ ca prasajyate || 137 ||
saliṅgaṃ ca sasaṅkhyaṃ ca tato dravyābhidhāyinā |
sambadhyate padaṃ tatra tayorbhinnā śrutirbhavet || 138 ||

(137) And since the word's formation, from its form, has the general for its domain — therefore gender and number resting on that follow. (138) Then the word, with gender and with number, is connected with what expresses the substance; there, the hearing of those two would be separate.

How gender and number get into a word at all. (137): rupac ca sabda-samskarah samanya-visayo yatah — the *word's formation*, from its *form*, has the *general* for its *domain*: the grammar operates on shapes, and shapes are general (3.14.78's science made of notation). Tasmat tad-asrayam lingam vacanam ca prasajyate: *therefore* *gender* and *number* resting on *that* *follow* — they attach to the general form. (138): salingam ca sasamkhyam ca tato dravyabhidhayina sambadhyate padam — *then* the *word*, with gender and number already on it, is *connected* with what *expresses the substance*. Tatra tayor bhinna srutir bhavet: *there*, the *hearing* of those two would be *separate*. So gender and number arrive with the form, before the substance is in view, and one would then hear them twice — which is the difficulty 3.14.139 answers.

3.14.139 unformalized

भाविनो बहिरङ्गस्य वचनादाश्रयस्य ये ।
लिङ्गसङ्ख्ये गुणानां ते सूत्रेण प्रतिपादिते ॥ १३९ ॥

bhāvino bahiraṅgasya vacanādāśrayasya ye |
liṅgasaṅkhye guṇānāṃ te sūtreṇa pratipādite || 139 ||

The gender and number of the qualities, which come from the statement of the future, external substratum — those are taught by the sutra.

The answer. Bhavino bahirangasya vacanad asrayasya ye lingasankhye gunanam: the *gender and number* of the *qualities*, which come from the *statement* of the *substratum* — and the substratum is *bhavin*, *future* (not yet in view when the form was made), and *bahiranga*, *external* (settled later: 3.7.182's antaranga/bahiranga). Te sutrena pratipadite: *those* are *taught* by the *sutra*. So the double hearing of 3.14.138 does not arise: the qualifier's gender and number are not its own at all — they are borrowed from a substratum that comes afterwards and from outside, and the borrowing is what the rule prescribes. This is 3.14.134's admission completed: concord is imposed, and it is imposed by making the qualifier answer to something not yet there.

3.14.140 unformalized

विशेषवृत्तेरपि च रूपाभेदादलक्षितः ।
यस्माद् विशेषस्तेनात्र भेदकार्यं न कल्पते ॥ १४० ॥

viśeṣavṛtterapi ca rūpābhedādalakṣitaḥ |
yasmād viśeṣastenātra bhedakāryaṃ na kalpate || 140 ||

And even though it operates in the particular, since the particular is unmarked, from the non-difference of form — therefore here the difference-operation is not conceivable.

Visesa-vrtter api ca rupabhedad alaksitah yasmad visesah: even though it *operates in the particular* (3.14.37's visesa-vrttitva), the *particular* is *unmarked* — alaksita: uncharacterized, bearing no sign — from the *non-difference of form*: the form does not show which particular it has got. Tena atra bheda-karyam na kalpate: *therefore* here the *difference-operation* is *not conceivable*. So a rule that needs a difference to operate on finds none, because the difference — though present — has left no mark on the shape, and the grammar reads shapes (3.14.137's samanya-visaya). This is the deepest version of the samuddesa's recurring finding: the science's blindness to particulars is not an oversight but a consequence of its being a science of forms.

3.14.141 unformalized

विशेष एव सामान्यं विशेषाद् भिद्यते यतः ।
अभेदो हि विशेषाणामाश्रितो विनिवर्तकः ॥ १४१ ॥

viśeṣa eva sāmānyaṃ viśeṣād bhidyate yataḥ |
abhedo hi viśeṣāṇāmāśrito vinivartakaḥ || 141 ||

The general is the particular itself, since it is differentiated from the particular; for the non-difference of the particulars, when resorted to, is exclusive.

One of the boldest lines in the kanda. Visesa eva samanyam: the *general* *is* the *particular itself* — eva: nothing else. Visesad bhidyate yatah: *since* it is *differentiated from* the particular — and here is the paradox stated on purpose: the general is the particular, and it is what the particular is marked off from. The general is not a further thing; it is the particular seen as excluding others. Abhedo hi visesanam asrito vinivartakah: for the *non-difference* of the *particulars*, when *resorted to*, is *exclusive* — vinivartaka: what turns others away. So sameness *is* exclusion: to take several particulars as non-different just is to turn away everything else. There is no positive general over and above the turning-away. This is very close to what the Buddhists would build apoha out of, arrived at from the other side — Bhartrhari's samanya is real, and its reality consists in excluding.

3.14.142 unformalized

यद् यदाश्रीयते तत् तदन्यस्य विनिवर्तकम् ।
भेदाभेदविभागस्तु सामान्ये न निरूप्यते ॥ १४२ ॥

yad yadāśrīyate tat tadanyasya vinivartakam |
bhedābhedavibhāgastu sāmānye na nirūpyate || 142 ||

Whatever is resorted to, that excludes something else; but the division into difference and non-difference is not determined in the general.

The principle generalized, and then withheld. Yad yad asriyate tat tad anyasya vinivartakam: *whatever* is *resorted to*, *that* *excludes* something *else* — every resort is an exclusion, without exception. To take up anything at all is to turn away what it is not (3.14.101's bhedapohena vartate; 2.129's the mind's fixing on one thing being its exclusion of the rest). Bhedabheda-vibhagas tu samanye na nirupyate: *but* the *division* into *difference and non-difference* is *not determined* in the *general*. Na nirupyate: is not ascertained, cannot be settled. So the question 'is the general the same as its particulars or different?' has no answer — not because we lack information, but because the general is constituted by exclusion and exclusion is not one of the two. Bhartrhari states his own doctrine and then declines to metaphysicalize it: the general excludes, and asking whether it is one with what it excludes is asking the wrong thing.

3.14.143 unformalized

अपोद्धारश्च सामान्यमिति तस्योरकारिणः ।
निमित्तावस्थमेवातस्तत् स्वधर्मेण गृह्यते ॥ १४३ ॥

apoddhāraśca sāmānyamiti tasyorakāriṇaḥ |
nimittāvasthamevātastat svadharmeṇa gṛhyate || 143 ||

And the general is an extraction — so it stands only in the condition of a cause; therefore it is grasped by its own property.

The general's status settled by naming its origin. Apoddharas ca samanyam: the *general* is an *apoddhara* — an *extraction*, the work's technical term for what the analyst lifts out of an undivided whole (2.10's words extracted from sentences as suffixes from words; 3.14.75's three storeys of extraction). So the samanya is not found; it is pulled out. Nimittavastham eva atah: *therefore* it stands *only in the condition of a cause* — nimitta-avastha: it exists in the state of being an occasioning condition, and in no other state. Tat sva-dharmena grhyate: it is *grasped by its own property*. So the general is real exactly as far as it causes, and no further — which is 3.14.142's refusal to say whether it is one with its particulars, now given its ground. An extraction that occasions has nothing left over to be same or different.
contested: Pada b in our source reads 'tasyorakarinah', which we cannot construe and take to be corrupt. We translate the verse from padas a, c and d, whose sense is secure, and leave b unrendered rather than emend on conjecture. Something like 'tasya upakarinah' ('of that which assists') would fit the syntax, but we do not press it.

3.14.144 unformalized

अनिर्धारितधर्मत्वाद् भेदा एव विकल्पिताः ।
निमित्तैर्व्यपदिश्यन्ते सामान्याख्याविशेषिताः ॥ १४४ ॥

anirdhāritadharmatvād bhedā eva vikalpitāḥ |
nimittairvyapadiśyante sāmānyākhyāviśeṣitāḥ || 144 ||

From the non-ascertainment of their properties, the differences alone, imagined, are designated by their causes, qualified by the name of the general.

What is actually going on when one uses a general term. Anirdharita-dharmatvad: from the *non-ascertainment* of their *properties* — one does not have the particulars' properties settled; one never does. Bheda eva vikalpitah: the *differences alone*, *imagined* — vikalpita: constructed (3.14.74's bhagair iva vikalpitaih). Nimittair vyapadisyante samanyakhya-visesitah: are *designated* by their *causes*, *qualified* by the *name of the general*. So there are only particulars, ill-known and constructed, and the general's name is a qualification laid over them by way of their occasioning conditions. Nothing general is being referred to. This is 3.14.143's apoddhara cashed out: the extraction leaves the particulars where they were and adds a name.

3.14.145 unformalized

यदा तु व्यपदिश्येते लिङ्गसंख्ये स्वभावतः ।
प्रयोगेष्वेव साधुत्वं वाक्ये प्रक्रम्यते तदा ॥ १४५ ॥

yadā tu vyapadiśyete liṅgasaṃkhye svabhāvataḥ |
prayogeṣveva sādhutvaṃ vākye prakramyate tadā || 145 ||

But when gender and number are designated by their own nature, then correctness is commenced in the sentence, in usages alone.

Yada tu vyapadisyete linga-samkhye svabhavatah: *when* gender and number are designated *by their own nature* — not imposed by rule (3.14.134's sastram arabdham) but belonging to the words themselves. Prayogesv eva sadhutvam vakye prakramyate tada: *then* *correctness* is *commenced* in the *sentence*, in *usages* alone. Prayogesv eva: in usages *and nothing else*. So on this alternative, concord is not the grammar's imposition: it is a fact of how words are used, and sadhutva starts from usage and stays there. The verse offers the rival to 3.14.134-139's account, and it is the more deflationary of the two — there is no mechanism, only what the sistas say (3.14.43's loka-vyavastha).

3.14.146 unformalized

तत्र प्रयोगोऽनियतो गुणानामाश्रयैः सह ।
सामान्यं यत् तदत्यन्तं तत्रैव समवस्थितम् ॥ १४६ ॥

tatra prayogo'niyato guṇānāmāśrayaiḥ saha |
sāmānyaṃ yat tadatyantaṃ tatraiva samavasthitam || 146 ||

There, the usage of the qualities with their substrata is unfixed; that general which is utter is settled there alone.

Tatra prayogo 'niyato gunanam asrayaih saha: *there*, the *usage* of the *qualities* with their *substrata* is *unfixed* — aniyata: not settled, varying: which quality-word goes with which substratum is not determined by anything. Samanyam yat tad atyantam tatraiva samavasthitam: that *general* which is *utter* is *settled* *there alone* — atyanta: complete, unqualified; tatra eva: in *that* — in the usage itself, and nowhere else. So on 3.14.145's view the general has no seat but the practice: the usage is unfixed, and the general is wholly resident in it. This is the most radically usage-based moment in the samuddesa, and Bhartrhari lets it stand alongside the rule-based account without adjudicating (3.14.97's upayas unsettled among persons).

3.14.147 unformalized

न गोत्वं शाबलेयस्य गौरिति व्यपदिश्यते ।
शुक्लत्वं बाहुलेयस्य शुक्ल इत्यपदिश्यते ॥ १४७ ॥

na gotvaṃ śābaleyasya gauriti vyapadiśyate |
śuklatvaṃ bāhuleyasya śukla ityapadiśyate || 147 ||

The cowness of Sabaleya is not designated as 'cow'; the whiteness of Bahuleya is designated as 'white'.

An asymmetry between universals and qualities, exhibited in two words. Na gotvam sabaleyasya gaur iti vyapadisyate: the *cowness* of *Sabaleya* — the dappled ox, the tradition's stock individual — is *not* designated as *'cow'*: 'gauh' names the animal, never the cowness. Suklatvam bahuleyasya sukla ity apadisyate: the *whiteness* of *Bahuleya* *is* designated as *'white'* — 'sukla' does reach the quality itself. So jati-words point past the universal to the individual, and guna-words do not: the two behave oppositely, and the grammar treats them alike. This is why 3.14.133's gender did not penetrate 'mrga-dugdha' and did penetrate 'gargi-putra'; and it is 3.14.132's warning made good — the particular is not everywhere of the same kind.

3.14.148 unformalized

व्यतिरेके च सत्येवं मतुपः श्रवणं भवेत् ।
लुगन्वाख्यायते तस्माद् रसादिभ्यश्च नास्ति सः ॥ १४८ ॥

vyatireke ca satyevaṃ matupaḥ śravaṇaṃ bhavet |
luganvākhyāyate tasmād rasādibhyaśca nāsti saḥ || 148 ||

And when there is thus a difference, the hearing of 'matup' would occur; therefore elision is expounded, and after 'rasa' and the rest it does not exist.

Vyatireke ca saty evam matupah sravanam bhavet: when there is thus a *difference* — between the quality and its possessor — the *hearing* of *'matup'*, the possessive affix, *would occur*: if 'sukla' means whiteness and the cloth *has* it, the affix should be audible. Lug anvakhyayate tasmat: *therefore* *luk* is *expounded* — the grammar posits the affix and deletes it (3.14.87's matub-lopa, for the same reason). Rasadibhyas ca na asti sah: and after *'rasa'* (taste) and the rest, *it does not exist* — no matup, elided or otherwise. So the device is applied to some quality-words and withheld from others, and the withholding needs its own explanation. 3.14.149 gives it.

3.14.149 unformalized

यत्सोऽयमिति संबन्धाद्रूपाभेदेन वर्तते ।
शुक्लादिवत् ततो लोपस्तद्रसादौ न विद्यते ॥ १४९ ॥

yatso'yamiti saṃbandhādrūpābhedena vartate |
śuklādivat tato lopastadrasādau na vidyate || 149 ||

Since, from the connection 'this is that', it operates by non-difference of form, like 'sukla' and the rest — therefore the elision does not exist in 'rasa' and the rest.

Yat so 'yam iti sambandhad rupabhedena vartate: since, from the *connection* 'this is that', it operates by *non-difference of form* — the identification runs directly, with no formal gap between quality and possessor. Sukladivat: *like* 'sukla' and the rest. Tato lopas tad rasadau na vidyate: *therefore* the *elision* does *not exist* in *'rasa'* and the rest. So the argument is: where the identification is immediate, no affix was ever needed, and so none can be elided (3.14.128's what withdraws has no being). 'Rasa' just identifies with its possessor; 'sukla' has to be got there by a suppressed matup. The grammar's uneven treatment tracks a real unevenness in how the words work — which is 3.14.147's asymmetry again, one level down.

3.14.150 unformalized

आवेशो लिङ्गसंख्याभ्यां क्वचिन्मञ्चादिवत् स्थितः ।
सोऽयमित्यभिसंबन्धे स प्रष्ठादौ न विद्यते ॥ १५० ॥

āveśo liṅgasaṃkhyābhyāṃ kvacinmañcādivat sthitaḥ |
so'yamityabhisaṃbandhe sa praṣṭhādau na vidyate || 150 ||

The entry by gender and number stands somewhere, like 'the platforms' and the rest; in the connection 'this is that', it does not exist in 'prastha' and the rest.

Avesah linga-samkhyabhyam kvacin mancadivat sthitah: the *entry* by *gender and number* stands *somewhere*, *like 'manca'* and the rest — 'manca hasanti', *the platforms laugh*: the platforms take the gender and number of the people on them, and the metonymy is complete enough to govern agreement. Avesa: entering, taking possession. So 'this is that' can be strong enough that one thing wears another's gender. So 'yam ity abhisambandhe sa prasthadau na vidyate: in the connection 'this is that', *it does not exist* in *'prastha'* and the rest. But not always: the measure-words do not take on what they measure. So the identification comes in strengths, and the grammar has to know which strength a word commands.

3.14.151 unformalized

लिङ्गं लिङ्गपरित्यागे सूत्रे प्रत्ययशासनम् ।
सोऽयमित्यभिसंबन्धात् पुंशब्दे स्त्र्यभिधायिनी ॥ १५१ ॥

liṅgaṃ liṅgaparityāge sūtre pratyayaśāsanam |
so'yamityabhisaṃbandhāt puṃśabde stryabhidhāyinī || 151 ||

The gender, in the abandonment of gender, is in the sutra the teaching of the affix; from the connection 'this is that', in a masculine word, the one expressing the feminine.

Lingam linga-parityage sutre pratyaya-sasanam: the *gender*, in the *abandonment of gender*, is in the *sutra* the *teaching of the affix* — where a word has dropped its gender, what the rule supplies is not a gender but an affix: the grammar deals in morphology and lets the gender follow. So 'yam ity abhisambandhat pum-sabde stry-abhidhayini: from the connection 'this is that', in a *masculine word*, the one *expressing the feminine*. So a feminine affix lands on a masculine stem, carried by the identification. This is 3.14.150's avesa in a specific case, and it shows how the grammar handles what it cannot state: it prescribes a suffix and lets the semantics arrive by identification.

3.14.152 unformalized

आश्रये लिङ्गसंख्याभ्यामाश्रितं व्यपदिश्यते ।
विशेषणानां चाजातेरिति शास्त्रव्यवस्थया ॥ १५२ ॥

āśraye liṅgasaṃkhyābhyāmāśritaṃ vyapadiśyate |
viśeṣaṇānāṃ cājāteriti śāstravyavasthayā || 152 ||

In the substratum, what is resorted to is designated by gender and number; and by the settled arrangement of the science, 'of qualifiers, and not of a universal'.

Asraye linga-samkhyabhyam asritam vyapadisyate: in the *substratum*, what is *resorted to* is *designated* by *gender and number* — the qualifier takes its gender and number from where it rests (3.14.139's borrowed concord). Visesananam ca ajater iti sastra-vyavasthaya: and by the *settled arrangement* of the *science*, *'of qualifiers, and not of a universal'* — the rule's own words, restricting the provision to qualifiers and excluding universals. So the borrowing is not universal: a jati-word does not take its substratum's gender, and the sutra says so. This is 3.14.147's asymmetry finding its way into Panini's wording — the grammar had noticed that universals and qualities behave differently, and had written the difference into a rule.

3.14.153 unformalized

निमित्तानुवाधायित्वाद् ये धर्मा भेदहेतुषु ।
त आश्रयेऽपि विद्यन्त इति बुद्धिर्निवर्त्यते ॥ १५३ ॥

nimittānuvādhāyitvād ye dharmā bhedahetuṣu |
ta āśraye'pi vidyanta iti buddhirnivartyate || 153 ||

The cognition that 'those properties which are in the difference-causes, from their conforming to the cause, exist also in the substratum' — is set aside.

An error blocked. Nimitta-anuvidhayitvad ye dharma bheda-hetusu: *those properties* which are in the *difference-causes*, from their *conforming to the cause* — anuvidhayin: following after, taking their shape from. Ta asraye 'pi vidyanta iti buddhir nivartyate: the *cognition* that they *exist also in the substratum* — is *set aside*. Nivartyate: is turned back. So one must not infer that because a qualifier's properties follow its cause, those properties are also in the thing qualified. The properties belong to the conditions, and the substratum does not acquire them by being conditioned. This guards 3.14.152's restriction from the obvious slide, and it is a small instance of the samuddesa's constant discipline: what the analysis puts somewhere for its own reasons must not be allowed to migrate into the object.

3.14.154 unformalized

आख्यायते च शास्त्रेण लोकरूढा स्वभावतः ।
निमित्ततुल्या गोदादौ प्रवृत्तिर्लिङ्गसंख्ययोः ॥ १५४ ॥

ākhyāyate ca śāstreṇa lokarūḍhā svabhāvataḥ |
nimittatulyā godādau pravṛttirliṅgasaṃkhyayoḥ || 154 ||

And the operation of gender and number in 'goda' and the rest — established in the world by its own nature, similar to the cause — is stated by the science.

The relation of science and usage, stated in one line. Akhyayate ca sastrena: is *stated* by the *science* — akhyayate: is *reported*. Loka-rudha svabhavatah: *established in the world* *by its own nature* — rudha: settled, grown into place, the word for what convention has made unanalysable (3.14.55's rudhi). Nimitta-tulya godadau pravrttir linga-samkhyayoh: the *operation* of gender and number in *'goda'* and the rest, *similar to the cause*. So the order is: the world settles it by its own nature, the operation resembles the causes, and the science *reports*. The grammar is not the source; it is downstream (3.14.81's anugamyate, correctness *followed out*). This is 3.14.145-146's usage-account and 3.14.134's rule-account reconciled — the rules are true reports of what usage settled.

3.14.155 unformalized

हरीतक्यादिषु व्यक्तिः सङ्खया खलतिकादिषु ।
मनुष्यलुब्विशेषाणामभिधेयाश्रयं द्वयम् ॥ १५५ ॥

harītakyādiṣu vyaktiḥ saṅkhayā khalatikādiṣu |
manuṣyalubviśeṣāṇāmabhidheyāśrayaṃ dvayam || 155 ||

In 'haritaki' and the rest, the individual; in 'khalatika' and the rest, number; of the particulars of 'manusya' and of the luk, the pair rests on what is expressed.

Three cases sorted. Haritakyadisu vyaktih: in *'haritaki'* (the myrobalan) and the rest, the *individual* — the word reaches the individual fruit. Sankhya khalatikadisu: in *'khalatika'* and the rest, *number*. Manusya-lub-visesanam abhidheyasrayam dvayam: of the *particulars* of *'manusya'* and of the *luk*, the *pair* — gender and number — *rests on what is expressed*. Abhidheyasraya: resting on the expressed, i.e. taking their character from the meaning rather than the form. So three word-classes and three different anchors: individual, number, and the meaning itself. The catalogue is unglamorous and it is the samuddesa's method — 3.14.132 said the particular is not everywhere the same, and this is what checking looks like.

3.14.156 unformalized

जातिप्रयोगे जात्या चेत् सम्बन्धमुपगच्छति ।
विशेषणं ततो धर्माञ् जातेस्तत् प्रतिपद्यते ॥ १५६ ॥

jātiprayoge jātyā cet sambandhamupagacchati |
viśeṣaṇaṃ tato dharmāñ jātestat pratipadyate || 156 ||

If, in the use of the universal, the qualifier comes into connection with the universal — then it acquires the properties of the universal.

Jati-prayoge jatya cet sambandham upagacchati visesanam: *if*, in the *use of the universal*, the *qualifier* comes into *connection* with the *universal*. Tato dharman jates tat pratipadyate: *then* it *acquires* the *properties* of the *universal*. So a qualifier attached to a jati-word takes on the jati's characteristics — including, presumably, the absence of number that 3.14.107 established (jatau dvivacanabhavat). The verse is conditional throughout: *if* the connection is with the universal. Because it need not be — the qualifier might attach to the individual instead, and then everything changes. 3.14.157 takes the other branch.

3.14.157 unformalized

लुबन्ते सन्निपतितं जातेरन्यद् विशेषणम् ।
लुबन्तस्य प्रधानत्वात् तद्धर्मैर्व्यपदिश्यते ॥ १५७ ॥

lubante sannipatitaṃ jāteranyad viśeṣaṇam |
lubantasya pradhānatvāt taddharmairvyapadiśyate || 157 ||

A qualifier other than the universal, having fallen upon a form ending in luk — from that form's being principal, it is designated by its properties.

The other branch. Lubante sannipatitam jater anyad visesanam: a qualifier *other than the universal*, having *fallen upon* a form ending in *luk* — sannipatita: alighted, come to rest on. Lubantasya pradhanatvat tad-dharmair vyapadisyate: from *that* form's being *principal*, it is *designated by its properties*. So where the qualifier lands on a luk-ending word rather than on the universal, it takes that word's properties instead — because the luk-ending word is principal there. The pair 3.14.156-157 is a clean statement of the samuddesa's general finding: a qualifier has no properties of its own; it takes whichever ones the thing it rests on has (3.14.139's borrowed concord, 3.14.152's asraye). What varies is only what it lands on.

3.14.158 unformalized

अबहुर्वीहिद्वन्द्वस्त्र्यतिशयेषु ये ।
भेदा भाष्यानुसारेण वाच्यास्ते लिङ्गसंङ्खययोः ॥ १५८ ॥

abahurvīhidvandvastryatiśayeṣu ye |
bhedā bhāṣyānusāreṇa vācyāste liṅgasaṃṅkhayayoḥ || 158 ||

Those differences of gender and number which occur in the non-bahuvrihi, the dvandva, the feminine and the superlative — are to be stated in accordance with the Bhasya.

The gender-and-number stretch closed by deferral. Abahurvrihi-dvandva-stry-atisayesu ye bheda: those *differences* which are in the *non-bahuvrihi*, the *dvandva*, the *feminine* and the *superlative* — four remaining classes, each with its own gender-and-number irregularities. Bhasyanusarena vacyas te linga-samkhyayoh: are *to be stated* in *accordance with the Bhasya*. Bhasyanusarena: following Patanjali. So Bhartrhari declines to redo what the Bhasya has done and points the reader there (3.14.34's citing the Bhasya's admission; 2.475). The gesture is characteristic: the karikas are not a complete grammar and do not pretend to be — they are a philosophy of one, and where the grammar has already spoken well, Bhartrhari says so and moves on.
contested: Pada a is metrically short in our source (reading 'abahurvihi-dvandva-stry-atisayesu ye'), and the compound's first member is transmitted as 'abahurvihi' rather than the expected 'abahuvrihi'. We render the sense the four-item list requires. The list's membership, not the metre, carries the verse.

3.14.159 unformalized

यदि षष्ठीद्वितीयान्तान्निकृष्टात् तरबादयः ।
न्यक्कारिणी स्युरुत्कृष्टे प्रकृतेः स्यात् विलिङ्गता ॥ १५९ ॥

yadi ṣaṣṭhīdvitīyāntānnikṛṣṭāt tarabādayaḥ |
nyakkāriṇī syurutkṛṣṭe prakṛteḥ syāt viliṅgatā || 159 ||

If 'tarap' and the rest came after the inferior thing ending in genitive or accusative, then, in the case of the superior which does the humbling, there would be a gender-discrepancy of the base.

The comparative's derivation, and a difficulty. Yadi sasthi-dvitiyantan nikrstat tarabadayah: *if* *'tarap'* and the rest — the comparative affixes — came after the *inferior* thing ending in *genitive* or *accusative*: i.e. if 'X-tara' were derived from the one X surpasses. Nyakkarini syur utkrste prakrteh syad vilingata: in the case of the *superior* which does the *humbling*, there would be a *gender-discrepancy* of the *base* — vilingata: the base would end up with the wrong gender, since the surpassed and the surpassing need not share one. So the derivation would make the comparative's gender depend on whoever is being outdone, which is plainly wrong. The next verses spell out how wrong.

3.14.160-161 unformalized

काल्यां कालाद् द्वितीयान्तात् काले काल्यास्तरब् भवेत् ।
न्यक्कर्तरि तथा गार्ग्ये गर्गेभ्यः प्रत्ययो भवेत् ॥ १६० ॥
न्यक्कर्तृषु च गर्गेषु गार्ग्यात् स्यात् तच्च नेष्यते ।
कुमार्याः स्वार्थिके ङीप् स्यात् प्रकृत्यर्थो हि नाधिकः ॥ १६१ ॥

kālyāṃ kālād dvitīyāntāt kāle kālyāstarab bhavet |
nyakkartari tathā gārgye gargebhyaḥ pratyayo bhavet || 160 ||
nyakkartṛṣu ca gargeṣu gārgyāt syāt tacca neṣyate |
kumāryāḥ svārthike ṅīp syāt prakṛtyartho hi nādhikaḥ || 161 ||

(160) In 'kali', 'tarap' would come from 'kala' ending in the accusative; in 'kala', from 'kali'. Likewise, when Gargya does the humbling, the affix would come from the Gargas. (161) And when the Gargas do the humbling, it would come from Gargya — and that is not desired. For 'kumari', in the svarthika, 'nip' would apply; for the base-meaning is not in excess.

The absurdity of 3.14.159 spelled out on four forms. (160): kalyam kalad dvitiyantat kale kalyas tarab bhavet — in the feminine *'kali'*, 'tarap' would come from masculine *'kala'*; and in *'kala'*, from feminine *'kali'*: each would take the other's gender, and the derivation would deliver 'kalitara' where 'kalatara' is wanted. Nyakkartari tatha gargye gargebhyah pratyayo bhavet: likewise where *Gargya* humbles, the affix would come from the *Gargas* — the plural. (161): nyakkartrsu ca gargesu gargyat syat tac ca nesyate — and conversely; *and that is not desired*. Kumaryah svarthike nip syat prakrtyartho hi nadhikah: for *'kumari'*, in the *svarthika*, *'nip'* would apply, for the *base-meaning* is *not in excess* — the affix adds nothing, so nothing licenses the change. Four forms, four wrong results.

3.14.162 unformalized

षठ्यन्तादधिके तस्माद् गुणे स्वाश्रयवर्तिनि ।
उत्कृष्टसमवेतायां क्रियायां वा विधीयते ॥ १६२ ॥

ṣaṭhyantādadhike tasmād guṇe svāśrayavartini |
utkṛṣṭasamavetāyāṃ kriyāyāṃ vā vidhīyate || 162 ||

Therefore it is prescribed after a genitive-ending, when the quality residing in its own substratum is in excess — or in an action inhering in the superior.

The correct derivation. Sasthyantad adhike tasmad gune svasraya-vartini: *therefore* after a *genitive*-ending, when the *quality* residing in its *own substratum* is *in excess* — the comparative is built on the *quality's* excess, and the quality stays in its own seat: svasraya-vartin. Utkrsta-samavetayam kriyayam va vidhiyate: *or* in an *action* *inhering* in the *superior*. So 'tarap' attaches where the excess is — in the quality, or in the action — and not to whoever is surpassed. The gender-discrepancy of 3.14.159-161 never arises, because the base was never the inferior party. The whole objection came from putting the affix in the wrong place, and the repair is to locate the excess correctly.

3.14.163 unformalized

उपात्तं च प्रकृत्यर्थो द्रव्यमेवाश्रयस्तयोः ।
सोऽयमित्यभिसंबन्धादभेदेन प्रतीयते ॥ १६३ ॥

upāttaṃ ca prakṛtyartho dravyamevāśrayastayoḥ |
so'yamityabhisaṃbandhādabhedena pratīyate || 163 ||

And what is taken up — the base-meaning — the substance alone is the substratum of those two; from the connection 'this is that', it is cognized by non-difference.

Upattam ca prakrtyartho dravyam eva asrayas tayoh: what is *taken up*, the *base-meaning* — the *substance alone* is the *substratum* of those *two* (gender and number). Dravyam eva: the substance and nothing else; qualities and universals do not hold gender or number in their own right. So 'yam ity abhisambandhad abhedena pratiyate: from the connection 'this is that', it is *cognized by non-difference*. So the substance is where gender and number live, and everything else that seems to have them has borrowed them by identification (3.14.150's avesa). This is the resolution of the whole gender-and-number stretch: there is one bearer, and the rest is 'this is that'.

3.14.164 unformalized

रूपाभेदाच्च तद् द्रव्यमाकाङ्क्षावत् प्रतीयते ।
विशेषैर्भिन्नरूपैस्तदाश्रयैरिव युज्यते ॥ १६४ ॥

rūpābhedācca tad dravyamākāṅkṣāvat pratīyate |
viśeṣairbhinnarūpaistadāśrayairiva yujyate || 164 ||

And from non-difference of form, that substance is cognized as having expectancy; it is joined with particulars of different forms, as though with substrata.

Rupabhedac ca tad dravyam akanksavat pratiyate: from *non-difference of form*, that *substance* is cognized as *having expectancy* — akanksavat: with a gap in it, reaching (2.4's sakanksa). Because its form shows no differences, the substance presents itself as incomplete and wanting qualification. Visesair bhinna-rupais tad asrayair iva yujyate: it is *joined* with *particulars* of *different forms*, *as though* with *substrata*. Iva: as though. So the substance takes on particulars as if they were its supports — when in fact it is the support and they are the borrowers (3.14.163). The relation inverts in the cognition, and iva marks the inversion. This is 3.14.9-11's two open generals drawing away each other's doubt, now stated for substance and quality.

3.14.165 unformalized

भिन्नरूपेषु यल्लिङ्गं विशेषेषु व्यवस्थितम् ।
संख्या च ताभ्यां द्रव्यात्मा सोऽभिन्नो व्यपदिश्यते ॥ १६५ ॥

bhinnarūpeṣu yalliṅgaṃ viśeṣeṣu vyavasthitam |
saṃkhyā ca tābhyāṃ dravyātmā so'bhinno vyapadiśyate || 165 ||

That gender which is settled in the particulars of different forms, and number — by those two the substance-self, undivided, is designated.

Bhinna-rupesu yal lingam visesesu vyavasthitam samkhya ca: the *gender* *settled* in the *particulars* of *different forms*, *and number* — the two live in the particulars. Tabhyam dravyatma so 'bhinno vyapadisyate: by *those two* the *substance-self*, *undivided*, is *designated*. Abhinnah: undivided (3.14.99's abheda-ekatva; 3.14.93's tad-atmany avibhakte). So the traffic runs both ways: the substance lends its gender and number to the particulars (3.14.163), and the particulars are what designate the substance by means of them. And the substance so designated is undivided — it does not acquire the particulars' divisions by being pointed at through them. Vyapadisyate again: designated, not described (3.14.135).

3.14.166 unformalized

आश्रयः समवायी च निमित्तं लिङ्गसंख्ययोः ।
कर्तृस्थभावकः शेतिरतो भाष्य उदाहृतः ॥ १६६ ॥

āśrayaḥ samavāyī ca nimittaṃ liṅgasaṃkhyayoḥ |
kartṛsthabhāvakaḥ śetirato bhāṣya udāhṛtaḥ || 166 ||

The substratum, and what inheres, are the cause of gender and number; therefore 'si', whose state resides in the agent, is cited in the Bhasya.

Asrayah samavayi ca nimittam linga-samkhyayoh: the *substratum*, and what *inheres*, are the *cause* of gender and number — two causes, not one: where the thing rests, and what rests in it. Kartr-stha-bhavakah setir ato bhasya udahrtah: *therefore* *'si'* (to lie down), whose *state resides in the agent*, is *cited in the Bhasya*. So Patanjali's example is chosen because 'si' is the clean case: its action does not go out to an object but stays in the one who lies (3.7.115's kartr-stha-bhava; 3.8's intransitive action). With nothing but the agent in view, one can see that the agent's gender and number are doing all the work. Bhartrhari is reading the Bhasya's choice of example as an argument — which it was.

3.14.167 unformalized

निमित्तमाश्रयत्वेन गृह्येत यदि साधनम् ।
कर्मापदिष्टयोः प्राप्तिस्तत्र स्याल्लिङ्गसंख्ययोः ॥ १६७ ॥

nimittamāśrayatvena gṛhyeta yadi sādhanam |
karmāpadiṣṭayoḥ prāptistatra syālliṅgasaṃkhyayoḥ || 167 ||

If the cause were grasped as the substratum — the means — then there would arise, in that case, the gender and number stated for the object.

The reason 3.14.166 needed an intransitive example. Nimittam asrayatvena grhyeta yadi sadhanam: *if* the *cause* were grasped as the *substratum* — namely the *means*, the sadhana. Karma-apadistayoh praptis tatra syal linga-samkhyayoh: then there would arise, in that case, the *gender and number* stated for the *object*. So if any karaka could serve as the seat, a transitive verb would let the object's gender and number in, and the form would come out wrong. Hence 'si': no object, no leak. This is a small, exact piece of reasoning about why an example is the example it is, and it shows the samuddesa's respect for the Bhasya's craft (3.14.158's bhasyanusarena).

3.14.168 unformalized

शास्त्रे निर्मित्तभावेन समुदायादपोद्घृतः ।
स्त्र्यर्थस्तस्येच्छया योगः प्रकृत्या प्रत्ययेन वा ॥ १६८ ॥

śāstre nirmittabhāvena samudāyādapodghṛtaḥ |
stryarthastasyecchayā yogaḥ prakṛtyā pratyayena vā || 168 ||

In the science, the feminine-meaning is extracted from the aggregate by way of being a cause; its connection, at one's pleasure, is with the base or with the affix.

The feminine begins, and it begins by being named an extraction. Sastre nimitta-bhavena samudayad apoddhrtah stry-arthah: in the *science*, the *feminine-meaning* is *extracted* from the *aggregate* by way of being a *cause* — apoddhrta: pulled out (3.14.143's apoddharas ca samanyam, in the same words). So femininity is not found in the word; it is lifted out of an undivided whole because the grammar needs a condition to hang rules on. Tasya icchaya yogah prakrtya pratyayena va: its *connection*, *at one's pleasure*, is with the *base* or with the *affix*. Icchaya: as one wishes. So having extracted it, one may put it wherever one likes — in the stem or in the suffix — and nothing decides. That is what being an extraction means (3.14.96's upaya-matram nanatvam).

3.14.169 unformalized

स्त्रीशब्दो गुणशब्दत्वात् तुल्यधर्मा सितादिभिः ।
गुणमात्रे प्रयुज्यते संस्त्यानवति वाश्रये ॥ १६९ ॥

strīśabdo guṇaśabdatvāt tulyadharmā sitādibhiḥ |
guṇamātre prayujyate saṃstyānavati vāśraye || 169 ||

The word 'stri', being a quality-word, has properties similar to 'sita' and the rest; it is employed for the quality alone, or for the substratum possessing condensation.

Stri-sabdo guna-sabdatvat tulya-dharma sitadibhih: the word *'stri'*, from being a *quality-word*, has *properties similar* to *'sita'* (white) and the rest — it behaves like a colour-term, not like a jati-term (3.14.147's asymmetry: guna-words reach the quality, jati-words do not). Guna-matre prayujyate samstyanavati va asraye: it is employed for the *quality alone*, *or* for the *substratum* possessing *samstyana*. Samstyana: *condensation*, coagulation — from the root sty~ai, and the grammarians' etymology of 'stri'. So the word is ambiguous exactly as 'sukla' is: it may name the quality or the thing that has it. Everything in the following verses turns on which.

3.14.170 unformalized

स्त्र्यर्थः संस्त्यानवद् द्रव्यं प्रकृत्यर्थश्च यद्यसौ ।
द्रव्योपलक्षणार्थत्वं संस्त्यानस्य तथा सति ॥ १७० ॥

stryarthaḥ saṃstyānavad dravyaṃ prakṛtyarthaśca yadyasau |
dravyopalakṣaṇārthatvaṃ saṃstyānasya tathā sati || 170 ||

If the feminine-meaning is the substance possessing condensation, and that is the base-meaning — then condensation has the purpose of indicating the substance.

The first horn of 3.14.169's ambiguity. Stry-arthah samstyanavad dravyam prakrtyarthas ca yady asau: *if* the *feminine-meaning* is the *substance* possessing *samstyana*, and *that* is the *base-meaning*. Dravya-upalaksanarthatvam samstyanasya tatha sati: then *samstyana* has the *purpose of indicating the substance* — upalaksana: an indicator, a mark that identifies without characterizing (3.14.55's saptaparna, whose seven leaves identify a tree that has none). So on this horn, condensation is not what 'stri' means; it is only how the word finds its object. The femininity is the substance, and the coagulation is a pointer. Bhartrhari sets out the consequence without comment and turns, in the next verses, to what it costs.

3.14.171 unformalized

संस्त्यानेन क्वचिद् द्रव्यं दृष्टं यदुपलक्षितम् ।
अनङ्गीकृतसंस्त्यानात् तद्वृत्तेः प्रत्ययो भवेत् ॥ १७१ ॥

saṃstyānena kvacid dravyaṃ dṛṣṭaṃ yadupalakṣitam |
anaṅgīkṛtasaṃstyānāt tadvṛtteḥ pratyayo bhavet || 171 ||

Somewhere a substance is seen indicated by condensation; from the derivative formation of that, condensation not being accepted, the affix would come.

Samstyanena kvacid dravyam drstam yad upalaksitam: *somewhere* a *substance* is *seen* *indicated* by *samstyana* — the indicator does work, in some cases. Anangikrta-samstyanat tad-vrtteh pratyayo bhavet: from the *derivative formation* of *that*, *samstyana* not being *accepted*, the *affix* would come. So where one derives from such a form without granting the condensation, the feminine affix attaches anyway — and it attaches for the wrong reason, since what licensed it has been denied. The verse is registering a leak in 3.14.170's account: an indicator that is not part of the meaning cannot be relied on to condition affixation, and yet the affixes come. 3.14.172 gives the case.

3.14.172 unformalized

भूतादयः षडाख्याश्च संस्त्यानेनोपलक्षिते ।
ब्राह्मण्यादौ यदा वृत्तास्तेभ्यः स्युः प्रत्ययास्तदा ॥ १७२ ॥

bhūtādayaḥ ṣaḍākhyāśca saṃstyānenopalakṣite |
brāhmaṇyādau yadā vṛttāstebhyaḥ syuḥ pratyayāstadā || 172 ||

And when 'bhuta' and the rest, the six designations, operate on 'brahmani' and the like, indicated by condensation — then the affixes would come from them.

Bhutadayah sad akhyas ca: *'bhuta'* and the rest, the *six designations* — a fixed set of terms in the grammar. Samstyanena upalaksite brahmanyadau yada vrttah: when they *operate on* *'brahmani'* and the like, *indicated by samstyana* (3.14.170's upalaksana). Tebhyah syuh pratyayas tada: *then* the *affixes* would come *from them*. So the feminine affix would be conditioned by the six designations rather than by the femininity — the indicator has been allowed to do the licensing, and the derivation has slipped its moorings. This is 3.14.131's vyabhicara in a specific instance: posit a condition that is only an indicator, and it will stray. The problem 3.14.171 announced now has a name and a form.

3.14.173 unformalized

तद्वन्तो हि प्रधानत्वात् प्रत्ययानां प्रयोजकाः ।
सामानाधिकरण्येऽपि तस्माट्टाबादिसम्भवः ॥ १७३ ॥

tadvanto hi pradhānatvāt pratyayānāṃ prayojakāḥ |
sāmānādhikaraṇye'pi tasmāṭṭābādisambhavaḥ || 173 ||

For those possessing that, being principal, are the occasioners of the affixes; therefore even in co-reference, 'tap' and the rest is possible.

The repair. Tadvanto hi pradhanatvat pratyayanam prayojakah: those *possessing that* — the substances that have the samstyana, not the samstyana itself — being *principal*, are the *occasioners* of the *affixes*. Prayojaka: what sets them in motion. So the affix is conditioned by the possessor, not by the indicator, and 3.14.172's slippage is blocked: the six designations were never the licensers, because they were not principal. Samanadhikaranye 'pi tasmat tabadi-sambhavah: *therefore* *even in co-reference*, *'tap'* and the rest is *possible* — even where two words share a referent (3.14.9-11's krsnas tilah), the feminine affix can attach, because the possessor is there to license it. The principal does the work; the mark only points.

3.14.174 unformalized

गुणमात्राभिधायित्वं स्त्रीशब्दे वर्ण्यते यदा ।
प्रकृत्यर्थश्च संस्त्यानं स्वार्थिकाः प्रत्ययास्तदा ॥ १७४ ॥

guṇamātrābhidhāyitvaṃ strīśabde varṇyate yadā |
prakṛtyarthaśca saṃstyānaṃ svārthikāḥ pratyayāstadā || 174 ||

When the word 'stri' is described as expressing the quality alone, and the base-meaning is condensation — then the affixes are svarthika.

The second horn of 3.14.169. Guna-matra-abhidhayitvam stri-sabde varnyate yada: *when* the word *'stri'* is described as expressing the *quality alone* — not the substance that has it. Prakrtyartham ca samstyanam: and the *base-meaning* is *samstyana* — the condensation itself is what the stem says. Svarthikah pratyayas tada: *then* the *affixes* are *svarthika* — affixes that add nothing to the base's meaning, mere formal furniture. So on this horn the feminine suffixes are semantically empty: the stem already means femininity, and the suffix only marks what is there. The two horns of 3.14.169 thus give two quite different jobs to the same affixes — which is 3.14.168's point exactly: the extraction may be lodged in the base or in the affix, *icchaya*, as one pleases, and the grammar comes out either way.

3.14.175 unformalized

संस्त्याने केवले वृत्तिः प्रकृतीनां न विद्यते ।
तदाविष्टे ततो द्रव्ये गृह्यन्ते समवस्थिताः ॥ १७५ ॥

saṃstyāne kevale vṛttiḥ prakṛtīnāṃ na vidyate |
tadāviṣṭe tato dravye gṛhyante samavasthitāḥ || 175 ||

In mere condensation, the operation of the bases does not exist; therefore they are grasped as settled in the substance pervaded by it.

The cost of 3.14.174's horn. Samstyane kevale vrttir prakrtinam na vidyate: in *mere* condensation, the *operation* of the *bases* does *not exist* — kevala: bare, by itself. If 'stri' names only the quality, the stems have nothing to work on: one cannot decline a coagulation. Tad-aviste tato dravye grhyante samavasthitah: *therefore* they are *grasped* as *settled* in the *substance* *pervaded* by it — avista: entered, permeated. So the bases must after all be lodged in the substance, and the second horn collapses back into the first (3.14.170). Which is 3.14.163's finding once more: dravyam eva asrayah — the substance alone is the seat, and every account that tries to do without it comes back.

3.14.176 unformalized

उपकारि च संस्त्यानं येषु शब्देष्वपेक्षितम् ।
तेभ्यष्टाबादयस्तच्च भूतादिष्वविवक्षितम् ॥ १७६ ॥

upakāri ca saṃstyānaṃ yeṣu śabdeṣvapekṣitam |
tebhyaṣṭābādayastacca bhūtādiṣvavivakṣitam || 176 ||

And in those words in which condensation is required as assisting, from them come 'tap' and the rest; and that is not intended for expression in 'bhuta' and the rest.

Upakari ca samstyanam yesu sabdesv apeksitam: in those words in which *condensation* is *required* as *assisting* — upakarin: helping, contributing; apeksita: needed. Tebhyas tabadayah: *from them* come *'tap'* and the rest — the feminine affixes attach where the condensation is a live contributor. Tac ca bhutadisv avivaksitam: and that is *not intended for expression* in *'bhuta'* and the rest — avivaksita: not meant to be said (3.14.103's standing device). So the same element is contributory in one class of words and unmeant in another, and the affixation follows the contribution. This settles 3.14.171-172's leak on principle rather than by patch: the affix goes where the condensation is doing something.

3.14.177 unformalized

संस्त्यानं प्रत्ययस्यार्थः शुद्धमाश्रीयते यदा ।
तदा द्विवचनानेकप्रत्ययत्वं न सिध्यति ॥ १७७ ॥

saṃstyānaṃ pratyayasyārthaḥ śuddhamāśrīyate yadā |
tadā dvivacanānekapratyayatvaṃ na sidhyati || 177 ||

When pure condensation is resorted to as the affix's meaning — then the dual and the multiplicity of affixes do not succeed.

The second horn's second cost. Samstyanam pratyayasya arthah suddham asriyate yada: when *pure* condensation is resorted to as the *affix's* meaning — the svarthika account of 3.14.174, taken strictly. Tada dvivacana-aneka-pratyayatvam na sidhyati: *then* the *dual* and the *multiplicity of affixes* do *not succeed*. So one cannot get 'two women': a quality has no dual (3.14.107's jatau dvivacanabhavat, and a bare guna is no better off), and there is nothing for a second affix to attach to. The horn fails on morphology as 3.14.175 showed it failing on the bases. Bhartrhari is not refuting the account so much as walking it to its end and letting it stop.

3.14.178 unformalized

जातिश्चेत् स्त्रीत्वमेवासौ भेदोऽन्यत्राविवक्षितः ।
यस्माद् भिन्नैरपि द्रव्यैस्तदेकं सद् विशिष्यते ॥ १७८ ॥

jātiścet strītvamevāsau bhedo'nyatrāvivakṣitaḥ |
yasmād bhinnairapi dravyaistadekaṃ sad viśiṣyate || 178 ||

If femininity is a universal, then it is that alone; difference is elsewhere unintended, since, though the substances differ, that, being one, is specified.

A third possibility. Jatis cet stritvam eva asau: *if* femininity is a *universal*, then it is *that alone* — a single jati, striitva, present wherever a woman is. Bhedo 'nyatra avivaksitah: *difference* is *elsewhere* *unintended*. Yasmad bhinnair api dravyais tad ekam sad visisyate: *since*, though the *substances differ*, *that*, being *one*, is *specified* — the many women specify the one femininity. So this account explains what 3.14.177's account could not: the differences are in the substances, the femininity is one, and the plural is theirs, not its. It is the strongest of the three and the samuddesa tests it too — 3.14.179 immediately.

3.14.179 unformalized

मात्राणां हि तिरोभावे परिमाणं न विद्यते ।
कुमार्य इति तेन स्यात् कुमार्यां भेदसम्भवात् ॥ १७९ ॥

mātrāṇāṃ hi tirobhāve parimāṇaṃ na vidyate |
kumārya iti tena syāt kumāryāṃ bhedasambhavāt || 179 ||

For in the disappearance of the measures, the measure does not exist; therefore 'kumaryah' would come about, from the possibility of difference in 'kumari'.

The test of 3.14.178. Matranam hi tirobhave parimanam na vidyate: in the *disappearance* of the *measures*, the *measure* does *not exist* — where the units go, the quantity goes: no units, no measure (3.14.127's the unit is the measure-word's own property). Kumarya iti tena syat kumaryam bheda-sambhavat: *therefore* *'kumaryah'* — the plural — *would come about*, from the *possibility of difference* in *'kumari'*. So on the universal-account, the plural has to come from somewhere, and the only place left is the individual girls; the femininity contributes nothing to it. Which is exactly what 3.14.178 wanted — and the verse notes the consequence rather than objecting. The account survives the test.

3.14.180-181 unformalized

जातिसङ्खयासमाहारैर्याथैव सहचारिणि ।
द्रव्ये क्रियाः प्रवर्तन्ते एकात्मत्वेऽप्यपेक्षिते ॥ १८० ॥
मूर्तिभ्यो मूर्तिधर्माणां तथाऽभेदस्य दर्शनात् ।
सामानाधिकरण्यं च क्रियायोगश्च कल्पते ॥ १८१ ॥

jātisaṅkhayāsamāhārairyāthaiva sahacāriṇi |
dravye kriyāḥ pravartante ekātmatve'pyapekṣite || 180 ||
mūrtibhyo mūrtidharmāṇāṃ tathā'bhedasya darśanāt |
sāmānādhikaraṇyaṃ ca kriyāyogaśca kalpate || 181 ||

(180) Just as actions operate in an accompanying substance by means of universal, number and aggregate, even where oneness of self is required — (181) so, from the seeing of the non-difference of the properties of forms from the forms, both co-reference and connection with action are conceivable.

The analogy that licenses the whole account. (180): jati-sankhya-samaharair yathaiva sahacarini dravye kriyah pravartante — as *actions* operate in an *accompanying substance* by means of *universal*, *number* and *aggregate*: three routes by which an action reaches a thing. Ekatmatve 'py apeksite: *even where* *oneness of self* is *required* — even where the thing must be one, the three routes work. (181): murtibhyo murti-dharmanam tatha 'bhedasya darsanat — from the *seeing* of the *non-difference* of the *properties of forms* from the *forms*: the shape's properties are not other than the shape. Samanadhikaranyam ca kriya-yogas ca kalpate: *co-reference* and *connection with action* are *conceivable*. So because a form's properties are not separate from it, both agreement and predication go through — without any of it requiring real division (3.14.92's anusyuta).

3.14.182 unformalized

सामानाधिकरण्ये तु मतुब्लोपादपेक्षिते ।
लुक् तद्धितलुकीति स्याल्लुक् तत्राप्युपलक्षणम् ॥ १८२ ॥

sāmānādhikaraṇye tu matublopādapekṣite |
luk taddhitalukīti syālluk tatrāpyupalakṣaṇam || 182 ||

But in co-reference, when the elision of 'matup' is required, 'luk taddhita-luki' would apply; there too the luk is an indicator.

Samanadhikaranye tu matub-lopad apeksite: in *co-reference*, when the *elision of matup* is *required* (3.14.148's device). Luk taddhita-lukiti syat: *'luk taddhita-luki'* would apply — Panini's rule chaining one elision to another. Luk tatrapy upalaksanam: *there too* the *luk* is an *indicator* — upalaksana: a mark, not a constituent (3.14.170's upalaksanartha). So the elision, having been posited to certify the form, turns out to be functioning as an indicator, and the rule that chains to it is chaining to a pointer. Tatrapi — *there too* — is the point: this is the same thing happening again, and by now one should recognize it. The grammar's absences do indicative work throughout, and Bhartrhari's contribution is to say so each time.

3.14.183 unformalized

केषाञ्चित् त्यक्तभेदेषु द्रव्येष्वेव विधीयते ।
संस्त्यानवत्सु टाबादिरभेदेन समन्वयात् ॥ १८३ ॥

keṣāñcit tyaktabhedeṣu dravyeṣveva vidhīyate |
saṃstyānavatsu ṭābādirabhedena samanvayāt || 183 ||

For some, 'tap' and the rest is prescribed only in substances possessing condensation whose differences are abandoned, from concomitance by non-difference.

Kesancit tyakta-bhedesu dravyesv eva vidhiyate samstyanavatsu tabadih: for *some*, *'tap'* and the rest is prescribed *only* in *substances* possessing *condensation* whose *differences are abandoned* — tyakta-bheda: the differences given up (3.14.101's bhedanam parityagat). Abhedena samanvayat: from *concomitance* by *non-difference*. So a fourth position: the feminine affix requires substances that have shed their differences, and it attaches by the concomitance of what is left. This keeps 3.14.163's dravya as the seat while getting the unity that 3.14.178's universal supplied. The samuddesa now has four accounts of the feminine on the table and has refuted none — which is 3.14.97 again: the upayas are unsettled among persons, because they are upayas.

3.14.184 unformalized

सामान्यभूतो द्रव्यात्मा परिच्छिन्नपरिग्रहः ।
क्रियाभिर्युज्यते भेदैर्भागश्चावतिष्ठते ॥ १८४ ॥

sāmānyabhūto dravyātmā paricchinnaparigrahaḥ |
kriyābhiryujyate bhedairbhāgaścāvatiṣṭhate || 184 ||

The substance-self, having become general, its grasp delimited, is joined with actions, with the differences; and the portion remains.

Samanya-bhuto dravyatma: the *substance-self*, having *become general* — samanya-bhuta: not a universal, but a substance gone general (3.14.141's visesa eva samanyam: the general *is* the particular). Paricchinna-parigrahah: its *grasp delimited* — pariccheda: the cutting-off (3.14.126). Kriyabhir yujyate bhedaih: is *joined* with *actions*, with the *differences*. Bhagas ca avatisthate: and the *portion* *remains* — avatisthate: stands, persists. So the generalized substance takes on actions and differences and something is left over that does not: the bhaga persists through. This is the samuddesa's picture of how a thing can be one and still be caught in the traffic of predication — the joining happens at the edges, and a portion stands out of it.

3.14.185 unformalized

शुक्लादिष्वाश्रयद्रव्यं प्राधान्येनाभिधीयते ।
स्त्रीत्वं तु प्रत्ययार्थ्तवादभिधाविषयो यतः ॥ १८५ ॥

śuklādiṣvāśrayadravyaṃ prādhānyenābhidhīyate |
strītvaṃ tu pratyayārthtavādabhidhāviṣayo yataḥ || 185 ||

In 'sukla' and the rest, the substratum-substance is expressed as principal; but femininity, since it is the affix's meaning, is the domain of expression.

The disanalogy that has been threatening since 3.14.169. Sukladisv asraya-dravyam pradhanyena abhidhiyate: in *'sukla'* and the rest, the *substratum-substance* is *expressed as principal* — 'sukla' says the white *thing*, and the whiteness rides along. Stritvam tu pratyayarthatvad abhidha-visayo yatah: *but* femininity, since it is the *affix's meaning*, is the *domain of expression* — it is what gets said. So the parallel with the colour-words breaks: in 'sukla' the substance is principal and the quality subordinate; in the feminine the quality is what the affix expresses, and the relation inverts. 3.14.169 proposed the analogy; here it is withdrawn where it fails. This is 3.14.132's discipline being applied to the samuddesa's own working assumption.

3.14.186 unformalized

सोऽयमित्यभिसम्बन्धादाश्रयं प्रतिपद्यते ।
स्त्रीत्वं स्वभावसिद्धो वा गुणभावविपर्ययः ॥ १८६ ॥

so'yamityabhisambandhādāśrayaṃ pratipadyate |
strītvaṃ svabhāvasiddho vā guṇabhāvaviparyayaḥ || 186 ||

From the connection 'this is that', femininity attains a substratum; or else the reversal of subordination is established by nature.

Two ways out of 3.14.185's inversion. So 'yam ity abhisambandhad asrayam pratipadyate stritvam: from the connection 'this is that', *femininity* *attains a substratum* — the identification hands it a seat it does not have of its own (3.14.150's avesa). Or: svabhava-siddho va guna-bhava-viparyayah: the *reversal of subordination* is *established by nature* — svabhava-siddha: settled by the thing's own nature; the inversion is simply how it is, and needs no mechanism. So either the identification supplies what is missing, or nothing is missing and the inversion is a fact. Bhartrhari offers both and endorses neither. By this point in the samuddesa that is not evasion but method (3.14.119's showing the result survives the disputed premise).

3.14.187 unformalized

साकाङ्क्षत्वाद् गुणत्वेन सामान्यं वोपदिश्यते ।
व्यक्तीनामात्मधर्मोऽसावेकप्रख्यानिबन्धनः ॥ १८७ ॥

sākāṅkṣatvād guṇatvena sāmānyaṃ vopadiśyate |
vyaktīnāmātmadharmo'sāvekaprakhyānibandhanaḥ || 187 ||

Or the general is taught as a quality, from its having expectancy; that is a property of the self of the individuals, having a single presentation for its basis.

A fifth account, and the most interesting. Sakanksatvad gunatvena samanyam va upadisyate: *or* the *general* is *taught as a quality*, from its *having expectancy* — sakanksa: reaching, incomplete (3.14.164's akanksavat). The general is classed as a guna because, like a quality, it cannot stand alone. Vyaktinam atma-dharmo 'sau: *that* is a *property of the self* of the *individuals* — not a thing over them but their own nature. Eka-prakhya-nibandhanah: having a *single presentation* for its *basis* — eka-prakhya: one appearing, one showing-forth. So the general is: the individuals' own property, grounded in the fact that they present alike. Not a universal, not a name, not an exclusion merely — a single presentation, and the property is that. This is as close as the work comes to a positive account of samanya, and it is built out of appearing (3.14.141's the general is the particular itself).

3.14.188-189 unformalized

एवम्भूता च सावस्था भागभेदपरिग्रहे ।
कृते बुद्ध्यैव भेदानामाश्रयत्वेऽवकल्पिते ॥ १८८ ॥
निष्कृष्टेष्वपि भेदेषु व्यक्तिरूपाश्रये ततः ।
लिङ्गप्रत्यवमर्शेन लिङ्गसंख्ये प्रपद्यते ॥ १८९ ॥

evambhūtā ca sāvasthā bhāgabhedaparigrahe |
kṛte buddhyaiva bhedānāmāśrayatve'vakalpite || 188 ||
niṣkṛṣṭeṣvapi bhedeṣu vyaktirūpāśraye tataḥ |
liṅgapratyavamarśena liṅgasaṃkhye prapadyate || 189 ||

(188) And that state is of such a kind, when the grasping of the portion-difference has been made by the intellect alone, the substratum-hood of the differences being conceived — (189) then, even when the differences are extracted, in the substratum having individual-form, by the reflective reference to gender it attains gender and number.

The whole gender-and-number analysis referred to the intellect. (188): evambhuta ca sa avastha bhaga-bheda-parigrahe krte buddhya eva — that *state* is of *such a kind*, when the *grasping* of the *portion-difference* has been *made by the intellect alone*. Buddhya eva: by the intellect and by nothing else — the divisions were never in the object (3.14.15's buddhya ekam bhidyate; 3.14.93's buddhy-antaram upasritah). Bhedanam asrayatve 'vakalpite: the *substratum-hood* of the *differences* being *conceived* — avakalpita: constructed. (189): niskrstesv api bhedesu vyakti-rupasraye tatah — *even when* the *differences* are *extracted* (niskrsta: pulled out, as in apoddhara), then, in the *substratum* having *individual-form*. Linga-pratyavamarsena linga-samkhye prapadyate: by the *reflective reference* to *gender*, it *attains* gender and number. Pratyavamarsa: the mind's turning back on what it laid down. So gender and number arrive by the intellect referring back to its own extraction — which is 3.14.74's imagined parts, and the whole samuddesa, in one word.

3.14.190 unformalized

अन्तरेण चशब्दस्य प्रयोगं द्वन्द्वभाविनाम् ।
अविशिष्टार्थवृत्तित्वं रूपाभेदात् प्रतीयते ॥ १९० ॥

antareṇa caśabdasya prayogaṃ dvandvabhāvinām |
aviśiṣṭārthavṛttitvaṃ rūpābhedāt pratīyate || 190 ||

Without the employment of the word 'ca', the operating in an unspecified meaning of those which are to be a dvandva is cognized, from non-difference of form.

The dvandva returns, and the samuddesa's method is by now familiar. Antarena ca-sabdasya prayogam: *without* the *employment* of the word *'ca'* — no 'and' is spoken. Dvandva-bhavinam avisistartha-vrttitvam: of those which are *to be* a *dvandva* — bhavin: future, not yet compounded — the *operating in an unspecified meaning*: before the compounding, the members mean indeterminately. Rupabhedat pratiyate: is *cognized*, from *non-difference of form*. So the members that will make a dvandva already operate unspecifiedly, and one knows it because their form shows no difference. The togetherness of 3.14.38 (the dvandva's saha-vivaksa, lost the moment one says 'ca') is thus present before the compound exists — which is why no 'ca' is needed to make it.

3.14.191 unformalized

विकल्पवति वा वृत्तिर्निवर्त्येऽर्थे समुच्चिते ।
तेषामज्ञातशक्तीनां द्योतकेन नियम्यते ॥ १९१ ॥

vikalpavati vā vṛttirnivartye'rthe samuccite |
teṣāmajñātaśaktīnāṃ dyotakena niyamyate || 191 ||

Or the derivative formation, in a meaning that has alternation, or is to be excluded, or is accumulated — of those whose powers are unknown, it is restricted by the indicator.

What 'ca' is for. Vikalpavati va vrttir nivartye 'rthe samuccite: the formation, in a meaning that has *alternation*, or is *to be excluded*, or is *accumulated* — three of the particle's senses, and no form shows which. Tesam ajnata-saktinam dyotakena niyamyate: of those whose *powers are unknown*, it is *restricted* by the *indicator* — ajnata-sakti: the words' capacities are not known from their shapes; the dyotaka narrows them. So the particle does not add a meaning; it *restricts* a range that was already there and unreadable. This is the dyotaka doctrine at its clearest (2.190's vacaka against dyotaka): 'ca' reveals which of the members' latent powers is in play, and reveals nothing that was not latent.

3.14.192 unformalized

वृत्तौ विशिष्टरूपत्वाच्चशब्दो विनिवर्तते ।
अर्थमेदेऽपि सारूप्यात् तच्चार्थेनापदिश्यते ॥ १९२ ॥

vṛttau viśiṣṭarūpatvāccaśabdo vinivartate |
arthamede'pi sārūpyāt taccārthenāpadiśyate || 192 ||

In the derivative formation, from its specified form, the word 'ca' withdraws; even where the meaning differs, from likeness of form that is designated by the meaning.

Why the dvandva has no 'and'. Vrttau visista-rupatvat ca-sabdo vinivartate: in the *derivative formation*, from its *specified form*, the word *'ca'* *withdraws* — the compound's own shape already specifies what 'ca' would have restricted, so the restrictor is redundant (3.14.191's niyamyate). Artha-bhede 'pi sarupyat tac ca arthena apadisyate: even where the *meaning differs*, from *likeness of form*, that is *designated* by the *meaning*. So the compound does the particle's work by being the shape it is, and the particle drops out — not elided (3.14.128's what withdraws has no being) but unnecessary. 'Rama-krsnau' needs no 'and' because its form is already the restriction.

3.14.193 unformalized

चस्य चासत्त्वभूतोऽर्थः स एवाश्रीयते यदि ।
तद्धर्मत्वं ततो द्वन्द्वे चादिष्वर्थकृतं हि तत् ॥ १९३ ॥

casya cāsattvabhūto'rthaḥ sa evāśrīyate yadi |
taddharmatvaṃ tato dvandve cādiṣvarthakṛtaṃ hi tat || 193 ||

And if the non-substantial meaning of 'ca' is itself resorted to — then in the dvandva there would be the property of that; for in 'ca' and the rest, that is made by the meaning.

Casya ca asattva-bhuto 'rthah sa eva asriyate yadi: *if* the *non-substantial* meaning of *'ca'* — asattva-bhuta: not a thing, the standard characterization of what particles mean — is *itself* resorted to as the dvandva's meaning. Tad-dharmatvam tato dvandve: *then* in the *dvandva* there would be *the property of that* — the compound would inherit ca's non-substantiality, and so would have no gender and no number, since only substances have those (3.14.163's dravyam eva asrayah). Cadisv artha-krtam hi tat: for in *'ca'* and the rest, *that* is *made by the meaning*. So making the dvandva mean what 'ca' means would make it an indeclinable. The consequence is unacceptable and the following verses look for a way round.

3.14.194 unformalized

चार्थः शक्तेः क्वचिद् भेदात् कथञ्चित् समवस्थितः ।
द्योतकाश्चादयस्तस्य वक्ता द्वन्द्वस्तु तद्वताम् ॥ १९४ ॥

cārthaḥ śakteḥ kvacid bhedāt kathañcit samavasthitaḥ |
dyotakāścādayastasya vaktā dvandvastu tadvatām || 194 ||

The meaning of 'ca', from a difference of power somewhere, is somehow settled; 'ca' and the rest are its indicators, but the dvandva is the expresser of those which possess it.

The way round 3.14.193, and it is precise. Carthah sakteh kvacid bhedat kathancit samavasthitah: the *meaning of 'ca'*, from a *difference of power* somewhere, is *somehow* settled — kathancit: somehow; Bhartrhari does not pretend to have located it. Dyotakas cadayas tasya: *'ca'* and the rest are its *indicators*. Vakta dvandvas tu tadvatam: *but* the *dvandva* is the *expresser* of *those which possess it* — tadvat: the *possessors* of ca's meaning, i.e. the things conjoined, not the conjoining. So the distinction that saves the account: 'ca' indicates the togetherness; the dvandva *says the things that have it*. The compound therefore means substances after all, and takes gender and number (3.14.193's difficulty dissolved). Dyotaka against vacaka, doing real work.

3.14.195 unformalized

विकल्पाद्यभिघेयस्य चार्थस्यान्यपदार्थता ।
द्योतकत्वान्न कल्पेत तस्मात् सदुपलक्ष्यते ॥ १९५ ॥

vikalpādyabhigheyasya cārthasyānyapadārthatā |
dyotakatvānna kalpeta tasmāt sadupalakṣyate || 195 ||

The state of being another's word-meaning, belonging to ca's meaning — which is to be expressed as alternation and so on — would not be conceivable, since it is an indicator; therefore, existing, it is indicated.

Vikalpady-abhidheyasya carthasya anya-padarthata: the state of being *another's word-meaning* (3.14.38's bahuvrihi-relation), belonging to *ca's meaning* — which is *to be expressed* as *alternation* and so on. Dyotakatvan na kalpeta: would *not be conceivable*, since it is an *indicator*: an indicator cannot be somebody else's word-meaning, because it is not a meaning at all. Tasmat sad upalaksyate: *therefore*, *existing*, it is *indicated*. Sat: it *is* there — Bhartrhari does not make ca's sense unreal; he makes it unsayable. It exists and is pointed at, and no word says it. This is the exact boundary the dyotaka doctrine draws, and 3.14.194's dvandva lives on the other side of it.

3.14.196 unformalized

तत्र स्वाभाविकं लिङ्गं शब्दधर्मे व्यपेक्षिते ।
शब्दः कश्चित् तमेवार्थं कथञ्चित् प्रतिपद्यते ॥ १९६ ॥

tatra svābhāvikaṃ liṅgaṃ śabdadharme vyapekṣite |
śabdaḥ kaścit tamevārthaṃ kathañcit pratipadyate || 196 ||

There, the gender is natural, when a word-property is required; some word somehow attains that very meaning.

Tatra svabhavikam lingam sabda-dharme vyapeksite: *there*, the *gender* is *natural* — svabhavika: belonging to the thing's own nature (3.14.186's svabhava-siddha) — when a *word-property* is *required*. Sabdah kascit tam eva artham kathancit pratipadyate: *some* word *somehow* attains *that very* meaning. Kascit... kathancit: *some* word, *somehow* — the vagueness is not carelessness; it is the claim. Which word reaches which meaning, and by what route, is not derivable: it happens, and the grammar reports it (3.14.154's akhyayate ca sastrena). So gender is natural, the reaching is contingent, and the science stands at its distance (3.14.74's atidure). The verse's diffidence is its content.

3.14.197 unformalized

शब्दादर्थाः प्रतायन्ते स भेदानां विधायकः ।
अनुमानं विवक्षायाः शब्दादन्यन्न विद्यते ॥ १९७ ॥

śabdādarthāḥ pratāyante sa bhedānāṃ vidhāyakaḥ |
anumānaṃ vivakṣāyāḥ śabdādanyanna vidyate || 197 ||

Meanings are spread out from the word; it is the ordainer of the differences. Other than the word, there is no inference of the intention to speak.

One of the summit verses of the work. Sabdad arthah pratayante: *meanings* are *spread out* from the *word* — pratayante, from pra-tan: woven forth, extended like a fabric from its warp. Not caused by, not indicated by: *spread out from*. Sa bhedanam vidhayakah: *it* is the *ordainer* of the *differences* — vidhayaka: the one who ordains, lays down (3.14.15's sa hi arthasya vidhayika, of the intellect; the same word, now of the word). All the divisions this samuddesa has been unmaking — parts, members, genders, numbers — are the word's own ordinances. Anumanam vivaksayah sabdad anyan na vidyate: there is *no inference* of the *intention to speak* *other than the word*. So one cannot get behind the word to the speaker's meaning: the word is the only evidence there is of what was meant, and there is no independent access to vivaksa. This is the whole doctrine — 1.123's no cognition without word; 2.32's the word is the measure — and it is why 3.14.196 could only say kathancit. The word ordains, and nothing stands behind it to be consulted.

3.14.198-199 unformalized

समुच्चितः स्याद् द्वन्द्वार्थो गुणभूतसमुच्चयः ।
समुच्चयो वापि भवेद् गुणभूतसमुच्चितः ॥ १९८ ॥
समुच्चितस्य प्राधान्ये लिङ्गसंख्ये स्वभावतः ।
समुच्चयस्य प्राधान्ये शास्त्रं स्यात् प्रतिपादकम् ॥ १९९ ॥

samuccitaḥ syād dvandvārtho guṇabhūtasamuccayaḥ |
samuccayo vāpi bhaved guṇabhūtasamuccitaḥ || 198 ||
samuccitasya prādhānye liṅgasaṃkhye svabhāvataḥ |
samuccayasya prādhānye śāstraṃ syāt pratipādakam || 199 ||

(198) The dvandva's meaning would be the accumulated, with the accumulation subordinate; or else the accumulation, with the accumulated subordinate. (199) Where the accumulated is principal, gender and number are by nature; where the accumulation is principal, the science would be the teacher.

The fork, cleanly drawn. (198): samuccitah syad dvandvartho guna-bhuta-samuccayah — the dvandva means the *accumulated things*, with the *accumulation* subordinate. Samuccayo vapi bhaved guna-bhuta-samuccitah: *or* the *accumulation*, with the *things* subordinate. (199): samuccitasya pradhanye linga-samkhye svabhavatah — where the *things* are principal, gender and number are *by nature*: substances have them, and nothing need be said (3.14.163). Samuccayasya pradhanye sastram syat pratipadakam: where the *accumulation* is principal, the *science* would be the *teacher* — the accumulation is no substance, so the grammar must supply what the meaning cannot (3.14.134's sastram arabdham). So the fork is exactly the fork of 3.14.145 against 3.14.134: does concord come from the world or from the rules? And the answer depends on what a dvandva is about.

3.14.200 unformalized

समुच्चयवतोऽर्थस्य प्राधान्येऽप्यपरे विदुः ।
निमित्तानुविधायित्वादसिद्धिं लिङ्गसंख्ययोः ॥ २०० ॥

samuccayavato'rthasya prādhānye'pyapare viduḥ |
nimittānuvidhāyitvādasiddhiṃ liṅgasaṃkhyayoḥ || 200 ||

Others know that, even where the meaning possessing accumulation is principal, gender and number are not established — from their conforming to the cause.

An objection to the first horn. Samuccayavato 'rthasya pradhanye 'py apare viduh: *others* know that *even where* the *meaning possessing accumulation* is *principal* — i.e. even on 3.14.198's first branch, where the things are what is meant. Nimitta-anuvidhayitvad asiddhim linga-samkhyayoh: gender and number are *not established*, from their *conforming to the cause* — anuvidhayin: following after (3.14.153's word). If gender and number must follow their occasioning cause, and the cause is the accumulation, then they cannot come 'by nature' from the substances after all. So even the natural horn needs the science. The objection is good and 3.14.201-202 must meet it.

3.14.201 unformalized

समुच्चयो निमित्तं चेत् स्यान्निमित्तानुवर्तनम् ।
अन्वयव्यतिरेकाभ्यां चार्थो द्वन्द्वनिबन्धनः ॥ २०१ ॥

samuccayo nimittaṃ cet syānnimittānuvartanam |
anvayavyatirekābhyāṃ cārtho dvandvanibandhanaḥ || 201 ||

If the accumulation is the cause, there would be conformity to the cause; by agreement and difference, ca's meaning is the basis of the dvandva.

The objection's premise examined. Samuccayo nimittam cet syan nimitta-anuvartanam: *if* the *accumulation* is the *cause*, there would be *conformity to the cause* — the conditional is granted, and its antecedent is now the question. Anvaya-vyatirekabhyam cartho dvandva-nibandhanah: by *agreement and difference* — the distributional test (2.12's method) — *ca's meaning* is the *basis* of the *dvandva*: where the togetherness is, the compound is; where it is not, not. So the test says the accumulation *is* the basis, and 3.14.200's antecedent holds. Bhartrhari is running the objection's own argument to its strongest and then, in 3.14.202, showing that the conclusion still does not follow.

3.14.202 unformalized

समुच्चितनिमित्तत्वे चार्थस्यापगमेऽपि वा ।
स्वभावसिद्धे द्वन्द्वस्य लिङ्गसंख्ये व्यवस्थिते ॥ २०२ ॥

samuccitanimittatve cārthasyāpagame'pi vā |
svabhāvasiddhe dvandvasya liṅgasaṃkhye vyavasthite || 202 ||

Whether the accumulated is the cause, or even where ca's meaning has departed — the dvandva's gender and number stand settled, established by nature.

The answer. Samuccita-nimittatve carthasya apagame 'pi va: *whether* the *accumulated* is the *cause*, *or even* where *ca's meaning* has *departed* — both branches, including the one where the togetherness is simply gone. Svabhava-siddhe dvandvasya linga-samkhye vyavasthite: the dvandva's gender and number *stand settled*, *established by nature*. So 3.14.200's inference fails not because its premise is false but because the conclusion does not depend on it: however the causation is arranged, the compound's gender and number are there. Svabhava-siddha and vyavasthita together: settled by nature, and standing. This is 3.14.119's move once more — show the result survives every branch, and the branching stops mattering.

3.14.203 unformalized

पदान्तरस्थस्यार्थस्य द्योतकत्वान्न युज्यते ।
निपातो लिङ्गसङ्खयाभ्यां द्वन्द्वस्त्वर्थस्य वाचकः ॥ २०३ ॥

padāntarasthasyārthasya dyotakatvānna yujyate |
nipāto liṅgasaṅkhayābhyāṃ dvandvastvarthasya vācakaḥ || 203 ||

The particle, being an indicator of a meaning residing in another word, is not joined with gender and number; but the dvandva is the expresser of the meaning.

The ground of 3.14.202, stated plainly. Padantara-sthasya arthasya dyotakatvan na yujyate nipato linga-sankhyabhyam: the *particle*, being an *indicator* of a meaning *residing in another word*, is *not joined* with *gender and number*. Padantara-stha: the meaning it indicates is not its own — it sits in someone else's word (3.14.191's restricting the members' latent powers). So a particle can have no gender: it has no meaning of its own to bear one. Dvandvas tv arthasya vacakah: *but* the *dvandva* is the *expresser* of the *meaning* — vacaka, not dyotaka. So the compound is categorially unlike the particle whose work it does, and 3.14.193's threatened contagion cannot occur: the dvandva says things, and things have gender and number (3.14.194).

3.14.204 unformalized

निमित्तानुविधाने च द्रव्यधर्मानपेक्षणात् ।
गुणप्रधानभावेन क्रियायोगो न कल्पते ॥ २०४ ॥

nimittānuvidhāne ca dravyadharmānapekṣaṇāt |
guṇapradhānabhāvena kriyāyogo na kalpate || 204 ||

And in conformity to the cause, from the non-requiring of the substance's properties, connection with action by way of subordinate and principal is not conceivable.

The cost of 3.14.200's position, now charged. Nimitta-anuvidhane ca dravya-dharma-anapeksanat: in *conformity to the cause*, from the *non-requiring* of the *substance's properties* — if gender and number merely follow the accumulation, the substance's own properties are not needed and drop out of view. Guna-pradhana-bhavena kriya-yogo na kalpate: *connection with action* by way of *subordinate and principal* is *not conceivable*. So the compound could not be predicated of: an action needs a substance with properties to attach to, and this account has dispensed with them. The objection of 3.14.200, followed out, costs the dvandva its ability to appear in a sentence — which 3.14.205 makes fatal.

3.14.205 unformalized

यस्य नास्ति क्रियायोगः स्वतन्त्रोऽसौ न विद्यते ।
अर्थो द्वन्द्वस्य तत्र स्यादुपादानमनर्थकम् ॥ २०५ ॥

yasya nāsti kriyāyogaḥ svatantro'sau na vidyate |
artho dvandvasya tatra syādupādānamanarthakam || 205 ||

That which has no connection with action does not exist independently; the dvandva's meaning would there be — and the taking-up is purposeless.

The fatal step. Yasya nasti kriya-yogah svatantro 'sau na vidyate: that which has *no connection with action* does *not exist independently* — a thing that cannot enter into action has no standing of its own; being is being caught up in action (3.8's the whole world is a weave of actions; 3.7.3's the karakas as what is capable). Artho dvandvasya tatra syad upadanam anarthakam: the *dvandva's meaning* would *there* be — and the *taking-up* is *purposeless*: anarthaka (3.14.126's word). So 3.14.200's account leaves the dvandva with a meaning that cannot act and therefore is not, and the grammar's taking it up is pointless. The refutation is complete, and its principle is worth marking: for this work, to be is to be able to be predicated.

3.14.206 unformalized

समुच्चयवतोऽर्थस्य वाचको नानुवर्तते ।
निमित्तमपि चास्यार्थः स्वधर्मैर्युज्यते ततः ॥ २०६ ॥

samuccayavato'rthasya vācako nānuvartate |
nimittamapi cāsyārthaḥ svadharmairyujyate tataḥ || 206 ||

The expresser of the meaning possessing accumulation does not conform; and its meaning, being also the cause, is therefore joined with its own properties.

The settled position. Samuccayavato 'rthasya vacako na anuvartate: the *expresser* of the meaning *possessing accumulation* does *not conform* — na anuvartate: the dvandva, being a vacaka (3.14.203), does not follow after its cause the way an indicator would. Nimittam api ca asya arthah sva-dharmair yujyate tatah: and its *meaning*, being *also the cause*, is *therefore joined with its own properties*. So the compound's meaning is both what it says and what conditions it, and it therefore carries its own gender and number rather than borrowing them — which is 3.14.202's svabhava-siddha, now derived rather than asserted. The dvandva-analysis closes: 'ca' indicates and has nothing; the dvandva says and has everything.

3.14.207 unformalized

बाह्यो नास्त्याश्रयो द्वन्द्वे विशेषौ तत्र हि श्रुतौ ।
समुच्चयस्तदाधारस्तद्धर्मैर्व्यपदिश्यते ॥ २०७ ॥

bāhyo nāstyāśrayo dvandve viśeṣau tatra hi śrutau |
samuccayastadādhārastaddharmairvyapadiśyate || 207 ||

In a dvandva there is no external substratum, for the two particulars are heard there; the accumulation, having them for its support, is designated by their properties.

Why the dvandva is not a bahuvrihi. Bahyo nasty asrayo dvandve: in a *dvandva* there is *no external substratum* — nothing outside it that it is about (3.14.38's anya-padartha, which the bahuvrihi has and this does not). Visesau tatra hi srutau: for the *two particulars* are *heard* *there* — the dvandva says its own members, audibly. Samuccayas tad-adharas tad-dharmair vyapadisyate: the *accumulation*, having *them* for its *support*, is *designated by their properties*. So the accumulation rests on the members and takes its gender and number from them — no outside seat is needed, because the members are right there. This grounds 3.14.206's svabhava-siddha: the dvandva has its own properties because it has its own members.

3.14.208 unformalized

यो वावयवभेदाभ्यां भेदवद्भ्यामिवान्वितः ।
एकः समूहो धर्मान् स भागयोः प्रतिपद्यते ॥ २०८ ॥

yo vāvayavabhedābhyāṃ bhedavadbhyāmivānvitaḥ |
ekaḥ samūho dharmān sa bhāgayoḥ pratipadyate || 208 ||

Or that one aggregate which is followed, as it were, by the two member-differences which possess difference — acquires the properties of the two portions.

The same result by the other route. Yo va avayava-bhedabhyam bhedavadbhyam iva anvitah: that which is *followed*, *as it were*, by the *two member-differences* which *possess difference* — iva again (3.14.68's asadbhir bhagair anugata iva, in nearly the same words): the aggregate is only *as though* accompanied by its members. Ekah samuho dharman sa bhagayoh pratipadyate: *one* *aggregate*, it *acquires* the *properties* of the *two portions*. Ekah samuhah: one aggregate (3.14.96's samudayas tv eka eva sah). So the compound is one and takes two members' properties without having two members — the iva carries the whole weight. This is the samuddesa's settled shape: unity in fact, duality as device.

3.14.209 unformalized

एकश्च द्वयात्मकोऽर्थोऽसौ भेदाभेदसमन्वितः ।
यौ भेदावाश्रितस्तत्स्थे लिङ्गसङ्खये प्रपद्यते ॥ २०९ ॥

ekaśca dvayātmako'rtho'sau bhedābhedasamanvitaḥ |
yau bhedāvāśritastatsthe liṅgasaṅkhaye prapadyate || 209 ||

And that meaning is one and of two selves, endowed with difference and non-difference; resting on the two differences, it attains the gender and number residing in them.

Ekas ca dvayatmako 'rtho 'sau: that *meaning* is *one* and *of two selves* — the paradox stated without apology. Bhedabheda-samanvitah: *endowed with difference and non-difference* — bhedabheda, the doctrine that will bear Bhartrhari's name in the later tradition, here in its native habitat: a compound's meaning. Yau bhedav asritas tat-sthe linga-sankhye prapadyate: *resting on* the *two differences*, it *attains* the *gender and number* *residing in them*. So the dvandva is genuinely both — one meaning with two selves — and it draws its concord from the side of itself that is two. Note that 3.14.142 refused to settle whether the general is same or different from its particulars; here the refusal has become a positive characterization. What cannot be resolved into either is what a compound is.

3.14.210 unformalized

यथा स्वशब्दाभिहिते चैत्रार्थे न प्रयुज्यते ।
चैत्रशब्दो बहुर्वीहावप्रयोगस्तथा भवेत् ॥ २१० ॥

yathā svaśabdābhihite caitrārthe na prayujyate |
caitraśabdo bahurvīhāvaprayogastathā bhavet || 210 ||

As the word 'Caitra' is not employed when the Caitra-meaning is expressed by its own word — so there would be non-employment in the bahuvrihi.

The bahuvrihi's problem, opened. Yatha sva-sabdabhihite caitrarthe na prayujyate caitra-sabdah: as the word *'Caitra'* is *not employed* when the Caitra-*meaning* is already *expressed by its own word* — one does not say a thing twice. Bahuvrihav aprayogas tatha bhavet: *so* there would be *non-employment* in the *bahuvrihi*. So if 'citragu' expresses its possessor, the possessor's own name should become unusable alongside it — one could not say 'citragur devadattah'. But one can. So either the compound does not express the possessor, or the principle is wrong. The following verses take the first branch.

3.14.211 unformalized

यथा गौरिति शुक्लादेरभिधानं न विद्यते ।
एवं यस्याभिसम्बन्धो गोभिस्तावत् प्रतीयते ॥ २११ ॥

yathā gauriti śuklāderabhidhānaṃ na vidyate |
evaṃ yasyābhisambandho gobhistāvat pratīyate || 211 ||

As by 'gauh' there is no expression of 'sukla' and the rest — so, of whomever there is a connection, that much is cognized by 'go'.

The first branch. Yatha gaur iti sukhlader abhidhanam na vidyate: as by *'gauh'* there is *no expression* of *'sukla'* and the rest — 'cow' does not say white, though every cow has some colour. Evam yasya abhisambandho gobhis tavat pratiyate: *so*, of *whomever* there is a *connection*, *that much* is cognized by *'go'*. Tavat: *that much* and no more. So 'citragu' conveys only that someone stands connected to variegated cows — it does not name him, any more than 'gauh' names a colour. The possessor is reached and not expressed, and 3.14.210's difficulty dissolves: the compound never said Caitra, so Caitra's name remains free.

3.14.212 unformalized

सम्बन्धी नियतो रूढश्चित्राणां न च विद्यते ।
गवां यथा वज्रपाणिस्त्र्यक्षो वापि व्यवस्थितः ॥ २१२ ॥

sambandhī niyato rūḍhaścitrāṇāṃ na ca vidyate |
gavāṃ yathā vajrapāṇistryakṣo vāpi vyavasthitaḥ || 212 ||

And there is no fixed, conventionally established relatum of the variegated cows — as 'vajrapani' or 'tryaksa' is settled.

The proof of 3.14.211. Sambandhi niyato rudhas citranam na ca vidyate: there is *no fixed*, *conventionally established* relatum of the *variegated* ones — rudha: grown into place by convention (3.14.55). Nobody in particular is *the* owner of dappled cows. Gavam yatha vajrapanis tryakso vapi vyavasthitah: as *'vajrapani'* — thunderbolt-in-hand, which means Indra and nobody else — *or* *'tryaksa'* — three-eyed, which means Siva — is *settled*. So some bahuvrihis have hardened into names and pick out one being by convention; 'citragu' has not. And that is the evidence: if the compound *expressed* its possessor, every bahuvrihi would be a proper name. Only the ones convention has frozen are, and the freezing is extraneous to the form.

3.14.213 unformalized

शब्दान्तरत्वाद् वाक्येषु विशेषा यद्यपि श्रुताः ।
वृत्तिशब्दोऽन्य एवायं सामान्यस्याभिधायकः ॥ २१३ ॥

śabdāntaratvād vākyeṣu viśeṣā yadyapi śrutāḥ |
vṛttiśabdo'nya evāyaṃ sāmānyasyābhidhāyakaḥ || 213 ||

Although particulars are heard in sentences, from their being other words — this derivation-word is quite another, an expresser of the general.

Sabdantaratvad vakyesu visesa yady api srutah: although *particulars* are *heard* in *sentences*, from their being *other words* — the analytic phrase 'citra gavo yasya' does name its parts. Vrtti-sabdo 'nya eva ayam: this *derivation-word* is *quite another* — anya eva: simply a different word (3.14.49's atyanta-bheda, and 3.14.136 in nearly these words). Samanyasya abhidhayakah: an *expresser of the general*. So the compound says a general — someone-or-other-with-dappled-cows — where the phrase says particulars. This is why 3.14.212 found no fixed relatum: the compound was never in the business of picking one out. It expresses a general and leaves the individual to be found.

3.14.214 unformalized

अगोरचित्रगोश्चैव रूपभेदान्निवर्तकः ।
एवं चित्रगुविशेषाणां रूपाभेदात्तु वाचकः ॥ २१४ ॥

agoracitragoścaiva rūpabhedānnivartakaḥ |
evaṃ citraguviśeṣāṇāṃ rūpābhedāttu vācakaḥ || 214 ||

It excludes the non-cow and the non-variegated-cowed alike, from difference of form; and thus, from non-difference of form, it is the expresser of the citragu-particulars.

How a general expression reaches particulars, and the mechanism is exclusion. Agor acitragos ca eva rupa-bhedan nivartakah: it *excludes* the *non-cow* and the *non-citragu*, from *difference of form* — nivartaka: what turns away (3.14.141's vinivartakah; 3.14.142's whatever is resorted to excludes something else). Where the form differs, the word turns it away. Evam citragu-visesanam rupabhedat tu vacakah: *thus*, from *non-difference of form*, it is the *expresser* of the *citragu-particulars*. So the same fact — form-difference — does both jobs: where it obtains the word excludes, where it fails the word expresses. The general is nothing but this two-sided working of form, and every particular that survives the exclusion is expressed. Bhartrhari's samanya is an exclusion that leaves particulars standing (3.14.187's eka-prakhya).

3.14.215 unformalized

यथा चित्रगुरित्येतत्प्रयुक्ते न प्रयुज्यते ।
एवं यदि स्यात् सामान्यं तस्य न स्यात् प्रतिश्रुतिः ॥ २१५ ॥

yathā citragurityetatprayukte na prayujyate |
evaṃ yadi syāt sāmānyaṃ tasya na syāt pratiśrutiḥ || 215 ||

As this 'citragu' is not employed when [the thing] has been employed — so, if it were a general, there would be no re-hearing of it.

Yatha citragur ity etat prayukte na prayujyate: as this *'citragu'* is *not employed* when [the thing] has been *employed*. Evam yadi syat samanyam tasya na syat pratisrutih: *so*, *if* it were a *general*, there would be *no re-hearing* of it — pratisruti: hearing again, echo. So the objection: on the general-account, the word should not bear repetition — one general once said is said. But 'citragu' can be said again of the same man, and heard again. Hence, allegedly, it is no general. The verse is compressed and it is an objection, not a doctrine; 3.14.216 answers by showing that generals and particulars exclude each other in both directions.
contested: Pada b, 'prayukte na prayujyate', has no expressed subject; we supply 'the thing' from 3.14.210, whose construction it echoes. Alternatively the locative absolute may resume 'citragu' itself (the word not being employed once employed). The objection's force — that a general should not tolerate repetition — is the same on either supply.

3.14.216 unformalized

सर्वादयो विशेषास्तु प्रदेशानां निवर्तकाः ।
यथा प्रदेशाः सामान्यप्रदेशान्तरबाधकाः ॥ २१६ ॥

sarvādayo viśeṣāstu pradeśānāṃ nivartakāḥ |
yathā pradeśāḥ sāmānyapradeśāntarabādhakāḥ || 216 ||

But 'sarva' and the rest, the particulars, are excluders of the specific instances — as the specific instances are annullers of the general and of other instances.

The answer to 3.14.215, and it makes exclusion mutual. Sarvadayo visesas tu pradesanam nivartakah: *'sarva'* (all) and the rest, the *particulars*, are *excluders* of the *specific instances* — pradesa: a particular case, an instance. So the general terms turn away the instances. Yatha pradesah samanya-pradesantara-badhakah: *as* the *instances* are *annullers* of the *general* and of *other instances* — badhaka: what sublates, the strong word. So the traffic runs both ways: generals exclude instances, and instances annul both the general and each other. Nothing is privileged; every term turns away whatever it is not (3.14.142's yad yad asriyate). And so 3.14.215's difficulty fails: a general is not consumed by being said, because its being said is an exclusion, and exclusions repeat.

3.14.217 unformalized

विभक्त्यर्थाभिधानाद् वा षष्ठी नानुप्रयुज्यते ।
द्रव्यस्यानभिधानात्तु तच्छब्दोऽनुप्रयुज्यते ॥ २१७ ॥

vibhaktyarthābhidhānād vā ṣaṣṭhī nānuprayujyate |
dravyasyānabhidhānāttu tacchabdo'nuprayujyate || 217 ||

Or, from the case-meaning's being expressed, the genitive is not employed afterwards; but from the substance's not being expressed, the word 'tad' is employed afterwards.

Two facts about how bahuvrihis are used. Vibhakty-arthabhidhanad va sasthi na anuprayujyate: from the *case-meaning's* being *expressed*, the *genitive* is *not employed* afterwards — the compound has already said the possessive relation, so no 'whose' need follow. Dravyasya anabhidhanat tu tac-chabdo 'nuprayujyate: *but* from the *substance's* *not being expressed*, the word *'tad'* *is* employed afterwards. So the compound says the relation and does not say the possessor — and usage reflects both halves precisely: no genitive (the relation is in), a pronoun (the substance is out). This is 3.14.211's tavat confirmed from the outside: what a word expresses can be read off what one must add to it.

3.14.218 unformalized

सामानाधिकरण्यं चेन्मतुब्लोपात् प्रकल्पते ।
मतुपोऽपि तदर्थत्वादनवस्था प्रसज्यते ॥ २१८ ॥

sāmānādhikaraṇyaṃ cenmatublopāt prakalpate |
matupo'pi tadarthatvādanavasthā prasajyate || 218 ||

If co-reference is conceivable from the elision of 'matup' — since 'matup' also has that meaning, an infinite regress follows.

The elision-device, finally broken. Samanadhikaranyam cen matub-lopat prakalpate: *if* *co-reference* is conceivable from the *elision of matup* — the device of 3.14.87, 3.14.148, 3.14.182. Matupo 'pi tad-arthatvad anavastha prasajyate: since *matup* *also* has *that* meaning, an *infinite regress* follows — anavastha: no standing, the endless recession. Because if possession must be carried by a matup, then the matup's own possession needs another matup, and so without end. So the device that has been quietly doing the work all through this stretch is now shown to be viciously regressive. Bhartrhari has let it run and run and then pulled it — which is 3.14.76's vyutpattir anavasthita made literal: the derivation has no footing, and here is the proof.

3.14.219 unformalized

सम्बन्धस्य च सम्बन्धी सम्बन्धोऽन्यः प्रसज्यते ।
विभक्त्यर्थप्रधाने च क्रियायोगो न कल्पते ॥ २१९ ॥

sambandhasya ca sambandhī sambandho'nyaḥ prasajyate |
vibhaktyarthapradhāne ca kriyāyogo na kalpate || 219 ||

And of the relation, another relation as relatum follows; and where the case-meaning is principal, connection with action is not conceivable.

A second regress and a second failure. Sambandhasya ca sambandhi sambandho 'nyah prasajyate: of the *relation*, *another relation* as *relatum* *follows* — the relation needs relating to its terms, and that relating needs relating (3.3's the sambandha-samuddesa's central difficulty: relation cannot be a relatum without regress). Vibhakty-artha-pradhane ca kriya-yogo na kalpate: and where the *case-meaning* is *principal*, *connection with action* is *not conceivable*. So the second failure is the fatal one of 3.14.205: a meaning that cannot enter into action does not independently exist. Making the relation principal in a bahuvrihi thus both regresses and de-realizes. The account is finished.

3.14.220 unformalized

विभक्त्यर्थप्रधानत्वात् ततस्तत्रेति न क्रिया ।
दृश्यादिः कर्मकर्त्रादिनिमित्तत्वाय कल्पते ॥ २२० ॥

vibhaktyarthapradhānatvāt tatastatreti na kriyā |
dṛśyādiḥ karmakartrādinimittatvāya kalpate || 220 ||

From the case-meaning's being principal, therefore 'there' — there is no action; 'drs' and the rest is conceivable as the cause of object, agent and the rest.

Vibhakty-artha-pradhanatvat tatas tatreti na kriya: from the *case-meaning's* being *principal*, *therefore* 'tatra' — 'there' — there is *no action*: a locative standing on its own has no verb to be locative to, and a bare relation is such a locative. Drsyadih karma-kartradi-nimittatvaya kalpate: *'drs'* (to see) and the rest is *conceivable* as the *cause* of *object*, *agent* and the rest — it is the *verb* that generates the karakas (3.7.1's the sadhanas are what the action needs; 3.8.2's the action is the principal). So relations cannot be principal because karakas are made by verbs, and a compound with a relation at its centre has no verb in it. The refutation is now complete on principle: relation is not a thing one can put in the middle of anything.

3.14.221 unformalized

अन्तर्भवेच्च सम्बन्धः प्राधान्याभिहितः कथम् ।
स प्रातिपदिकार्थश्च तथाभूतः कथं भवेत् ॥ २२१ ॥

antarbhavecca sambandhaḥ prādhānyābhihitaḥ katham |
sa prātipadikārthaśca tathābhūtaḥ kathaṃ bhavet || 221 ||

And how could a relation be included, being expressed as principal? And how could that, being such, be the meaning of a nominal base?

Two rhetorical questions, and they close the case. Antarbhavec ca sambandhah pradhanyabhihitah katham: *how* could a *relation* be *included* — antarbhu: contained within something — while being *expressed as principal*? What is principal is not contained; the two are contraries. Sa pratipadikarthas ca tatha-bhutah katham bhavet: and *how* could *that*, being *such*, be the meaning of a *nominal base*? A pratipadika means a thing; a relation is not a thing (3.3's relation as neither substance nor quality). So a bahuvrihi cannot have its relation as its meaning, because a nominal stem cannot mean a relation and nothing that is principal can be contained. Katham... katham: the questions have no answers, and are not meant to.

3.14.222 unformalized

असम्भवात्तु सम्बन्धे सम्बन्धसहचारिणि ।
जातिसङ्ख्यासमाहारकार्याणामिव सम्भवः ॥ २२२ ॥

asambhavāttu sambandhe sambandhasahacāriṇi |
jātisaṅkhyāsamāhārakāryāṇāmiva sambhavaḥ || 222 ||

But from the impossibility in the relation, there is possibility in what accompanies the relation — as of the operations of universal, number and aggregate.

The resolution. Asambhavat tu sambandhe: from the *impossibility* in the *relation* — 3.14.219-221's demonstration that nothing can be predicated of a relation. Sambandha-sahacarini sambhavah: in what *accompanies* the relation, there is *possibility* — sahacarin: the accompanier (3.14.31's bhinnesu sahacarisu, where the samuddesa began). Jati-sankhya-samahara-karyanam iva: *as* of the *operations* of *universal*, *number* and *aggregate* (3.14.180's three routes). So the grammar's operations do not fall on the relation, which cannot bear them; they fall on what goes along with it, exactly as universal, number and aggregate reach a substance by accompanying it. The bahuvrihi means neither its members nor its relation nor its possessor: it means what accompanies the relation, and the possessor is found by that.

3.14.223 unformalized

सोऽयमित्यभिसम्बन्धाद् विशिष्टाश्रयवाचिनाम् ।
शुक्लादिवल्लिङ्गसङ्ख्ये शास्त्रारम्भाद् भविष्यतः ॥ २२३ ॥

so'yamityabhisambandhād viśiṣṭāśrayavācinām |
śuklādivalliṅgasaṅkhye śāstrārambhād bhaviṣyataḥ || 223 ||

From the connection 'this is that', of those expressing a specified substratum — as with 'sukla' and the rest, gender and number will come about from the undertaking of the science.

So 'yam ity abhisambandhad visistasraya-vacinam: from the connection 'this is that', of those *expressing a specified substratum* — the bahuvrihis, which reach a substratum without naming it (3.14.211's tavat). Sukladivat linga-sankhye sastrarambhad bhavisyatah: as with *'sukla'* and the rest, *gender and number* will come about from the *undertaking of the science*. Sastrarambhat: because the grammar takes it up (3.14.134's sastram arabdham, in the same idiom). So the bahuvrihi's concord is not natural (against 3.14.202's dvandva, which had it by svabhava): it must be legislated, because there is no substance in the compound to bear it. The two compound-types are thus opposite in exactly the place one would expect.

3.14.224 unformalized

भेदेन तु विवक्षायां सामान्ये वा विवक्षिते ।
सलिङ्गस्य ससङ्ख्यस्य पदार्थस्यागतिर्भवेत् ॥ २२४ ॥

bhedena tu vivakṣāyāṃ sāmānye vā vivakṣite |
saliṅgasya sasaṅkhyasya padārthasyāgatirbhavet || 224 ||

But when the intention to express is by way of difference, or when the general is intended — there would be no reaching of a word-meaning possessing gender and number.

Bhedena tu vivaksayam samanye va vivaksite: when the *intention to express* is *by way of difference*, *or* when the *general* is *intended* — vivaksa, the will to say (3.14.197's the word is the only inference of it). Salingasya sasamkhyasya padarthasya agatir bhavet: there would be *no reaching* — agati: no access, no way through — of a *word-meaning* possessing *gender and number*. So on either intention, differentiating or generalizing, one cannot get to a gendered and numbered meaning: the differentiated has no unity to bear them, the general has no substance. This is why 3.14.223 had to call in the sastra: neither vivaksa delivers concord, and something must.

3.14.225 unformalized

साधुत्वं न विभक्त्यर्थमात्रे वृत्त्स्य दृश्यते ।
कृत्स्नार्थवृत्तेः साधुत्वमित्यर्थग्रहणं कृतम् ॥ २२५ ॥

sādhutvaṃ na vibhaktyarthamātre vṛttsya dṛśyate |
kṛtsnārthavṛtteḥ sādhutvamityarthagrahaṇaṃ kṛtam || 225 ||

Correctness is not seen of what operates in the case-meaning alone; 'correctness belongs to what operates in the whole meaning' — thus the taking-up of the word 'artha' is made.

A word in Panini's rule explained. Sadhutvam na vibhakty-artha-matre vrttasya drsyate: *correctness* is *not seen* of what operates in the *case-meaning alone* — a form that reaches only the relation is not sadhu (3.14.219-221's demonstration that a relation cannot be the meaning). Krtsnartha-vrtteh sadhutvam ity artha-grahanam krtam: 'correctness belongs to what operates in the *whole meaning*' — thus the *taking-up of 'artha'* is *made*. So the word 'artha' in the sutra is there to require that the form reach the *whole* meaning, relation and relata together. Bhartrhari is reading a single word of Panini's as carrying the doctrine of 3.14.222 — the operations fall on what accompanies the relation, and the rule says so by saying 'artha'.

3.14.226 unformalized

सोऽयमित्यभिसम्बन्धाद् द्रव्यवृत्तिरयं यदा ।
सलिङ्गस्य ससङ्ख्यस्य तदा साधुत्वमुच्यते ॥ २२६ ॥

so'yamityabhisambandhād dravyavṛttirayaṃ yadā |
saliṅgasya sasaṅkhyasya tadā sādhutvamucyate || 226 ||

When, from the connection 'this is that', this operates in the substance — then correctness is stated of what possesses gender and number.

So 'yam ity abhisambandhad dravya-vrttir ayam yada: when, from the connection 'this is that', this *operates in the substance* — the compound reaches a substance by identification, not by naming (3.14.211). Salingasya sasamkhyasya tada sadhutvam ucyate: *then* *correctness* is stated of what *possesses gender and number*. So the order is: identification first, substance reached, concord acquired, correctness pronounced. Nothing here is in the compound's meaning; all of it is in what the compound gets to. This completes 3.14.225: 'artha' in the rule means the whole reach of the word, and the reach ends in a substance (3.14.163's dravyam eva asrayah, for the last time in this stretch).

3.14.227 unformalized

अन्तर्भूतविभक्त्यर्थे षष्ठी न श्रूयते यथा ।
तथाऽश्रुतिः प्रसज्यते लिङ्गसङ्ख्याभिधायिनाम् ॥ २२७ ॥

antarbhūtavibhaktyarthe ṣaṣṭhī na śrūyate yathā |
tathā'śrutiḥ prasajyate liṅgasaṅkhyābhidhāyinām || 227 ||

As the genitive is not heard when the case-meaning is included — so the non-hearing of the expressers of gender and number would follow.

An objection to 3.14.226. Antarbhuta-vibhakty-arthe sasthi na sruyate yatha: as the *genitive* is *not heard* when the *case-meaning* is *included* — 3.14.217's fact: the compound has swallowed the relation, so no 'whose' follows. Tatha asrutih prasajyate linga-sankhyabhidhayinam: *so* the *non-hearing* of the *expressers* of *gender and number* would *follow*. So if inclusion silences the genitive, inclusion should silence the gender-and-number affixes too — and then no bahuvrihi could inflect. The objection is exact and it uses the samuddesa's own finding against it. 3.14.228 states the consequence in its full severity.

3.14.228 unformalized

साधर्म्यमव्ययेन स्याद् बहुर्वीहेस्तथा सति ।
लिङ्गसङ्ख्यानिमित्तस्य संस्कारस्यापवर्तनात् ॥ २२८ ॥

sādharmyamavyayena syād bahurvīhestathā sati |
liṅgasaṅkhyānimittasya saṃskārasyāpavartanāt || 228 ||

The bahuvrihi would then have similarity with an indeclinable, from the removal of the formation which is the cause of gender and number.

The consequence, and it is the one Bhartrhari has been avoiding since 3.14.193. Sadharmyam avyayena syad bahuvriheh tatha sati: the *bahuvrihi* would then have *similarity* with an *indeclinable*. Linga-sankhya-nimittasya samskarasya apavartanat: from the *removal* of the *formation* which is the *cause* of gender and number — apavartana: the turning away, the withdrawal of the very derivational process that would have supplied concord. So 3.14.227's objection makes the bahuvrihi an avyaya — uninflectable, like 'ca' (3.14.193's threatened contagion, arriving from the other direction). The result is plainly false: 'citragum pasya' is good Sanskrit. So something in the objection is wrong, and 3.14.229 finds it.

3.14.229 unformalized

प्रयुक्तेन च सम्बन्धाच्चैत्रादिश्रवणं भवेत् ।
विना विभक्त्या सम्बन्धो विभक्त्या विद्यते विना ॥ २२९ ॥

prayuktena ca sambandhāccaitrādiśravaṇaṃ bhavet |
vinā vibhaktyā sambandho vibhaktyā vidyate vinā || 229 ||

And from connection with what is employed, the hearing of 'Caitra' and the rest would occur; the relation exists without a case-ending — it exists without a case-ending.

The reply. Prayuktena ca sambandhac caitradi-sravanam bhavet: from *connection* with what is *employed*, the *hearing* of *'Caitra'* and the rest would *occur* — the possessor's name does turn up alongside the compound, and that is what a bahuvrihi is for (3.14.210-211: the compound reaches him without saying him, so his name stays sayable). Vina vibhaktya sambandho vibhaktya vidyate vina: the *relation* exists *without a case-ending*. So the disanalogy 3.14.227 missed: a relation can subsist unexpressed, and gender cannot. The genitive is silent because the relation survives its silence; the gender-affixes are audible because the gender does not. Inclusion silences only what can survive being silent.
contested: Pada d as transmitted ('vibhaktya vidyate vina') repeats pada c's content with the words reordered, which may be emphatic or may be a transmission fault. We render the sense once and note the duplication rather than reconstruct a different fourth pada.

3.14.230 unformalized

अभिधानेऽपि सङ्खयायाः संख्यात्वं न निवर्तते ।
षष्ठ्यर्थस्याभिधाने तु स्यात् प्रातिपदिकार्थता ॥ २३० ॥

abhidhāne'pi saṅkhayāyāḥ saṃkhyātvaṃ na nivartate |
ṣaṣṭhyarthasyābhidhāne tu syāt prātipadikārthatā || 230 ||

Even in the expression of number, numberhood does not withdraw; but in the expression of the genitive's meaning, there would be the state of being a nominal-base-meaning.

Abhidhane 'pi sankhyayah sankhyatvam na nivartate: *even in the expression* of *number*, *numberhood* does *not withdraw* — say a number and it is still a number; expressing it does not consume it. Sasthy-arthasya abhidhane tu syat pratipadikarthata: *but* in the *expression* of the *genitive's meaning*, there would be the state of being a *nominal-base-meaning*. And that is impossible — 3.14.221 asked how a relation could be a pratipadika's meaning, and the question had no answer. So the asymmetry of 3.14.229 is now grounded: number survives expression, relation cannot be expressed at all. The two are different in kind, and the rules treat them differently because they are.

3.14.231 unformalized

अनुप्रयोगसिद्ध्यर्थं न विभक्त्यर्थकल्पना ।
वस्त्वन्तरमुपक्षिप्तमिति केचित् प्रचक्षते ॥ २३१ ॥

anuprayogasiddhyarthaṃ na vibhaktyarthakalpanā |
vastvantaramupakṣiptamiti kecit pracakṣate || 231 ||

The imagining of the case-meaning is not for the sake of establishing the subsequent employment; another thing has been hinted at — so some declare.

Anuprayoga-siddhy-artham na vibhakty-artha-kalpana: the *imagining* of the *case-meaning* is *not* for the sake of *establishing* the *subsequent employment* — kalpana: construction (3.14.96's bheda-samsarga-kalpana). One does not posit the case-meaning merely to license the 'tad' that follows (3.14.217). Vastv-antaram upaksiptam iti kecit pracaksate: *another thing* has been *hinted at* — so *some* declare. Upaksipta: thrown out, alluded to, gestured toward. So on this view the posited case-meaning is pointing at something else altogether, and the anuprayoga is not the reason for it. Bhartrhari reports without endorsing (kecit), and 3.14.232 sets out what the something else might be.

3.14.232 unformalized

सम्बन्धिभिर्विशिष्टानां सम्बन्धाना निमित्तता ।
सम्बन्धैर्वा विशिष्टानां तद्वतां स्यान्निमित्तता ॥ २३२ ॥

sambandhibhirviśiṣṭānāṃ sambandhānā nimittatā |
sambandhairvā viśiṣṭānāṃ tadvatāṃ syānnimittatā || 232 ||

The causality belongs to the relations qualified by the relata; or the causality would belong to the possessors of those, qualified by the relations.

The fork that organizes the rest of the stretch. Sambandhibhir visistanam sambandhanam nimittata: the *causality* belongs to the *relations* *qualified by the relata* — relation principal, terms qualifying. Sambandhair va visistanam tadvatam syan nimittata: *or* the causality would belong to the *possessors of those*, *qualified by the relations* — terms principal, relation qualifying. So: which is the bahuvrihi's occasioning cause, the relation-as-qualified or the possessor-as-qualified? Everything 3.14.219-221 established says the second must win — a relation cannot be principal. But Bhartrhari puts both, because the tradition holds both, and 3.14.233-234 name the holders.

3.14.233 unformalized

केचित् संयोगिनो दण्डाद् विषाणात् समवायिनः ।
तद्वति प्रत्ययानाहुर्बहुर्वीहिं तथैव च ॥ २३३ ॥

kecit saṃyogino daṇḍād viṣāṇāt samavāyinaḥ |
tadvati pratyayānāhurbahurvīhiṃ tathaiva ca || 233 ||

Some say the affixes come, in the possessor of that, from 'danda' for the conjoined and from 'visana' for the inherent; and the bahuvrihi likewise.

The second horn's partisans. Kecit samyogino dandad visanat samavayinah: *some* say — from *'danda'* (staff) for the *conjoined*, giving 'dandin', the staff-bearer whose staff is merely in contact with him; and from *'visana'* (horn) for the *inherent*, giving 'visanin', the horned one whose horns inhere in him. Two relations, samyoga and samavaya, the Vaisesika pair. Tadvati pratyayan ahuh: they say the *affixes* come *in the possessor of that*. Bahuvrihim tathaiva ca: and the *bahuvrihi* *likewise*. So the possessive affixes attach to the possessor, and the compound works the same way: possessor principal, relation qualifying — 3.14.232's second horn, with two well-known relations to instance it.

3.14.234 unformalized

भिन्नं सम्बन्धिभेदेन सम्बन्धमपरे विदुः ।
निमित्तं सविभक्त्यर्थः समासेनाभिधीयते ॥ २३४ ॥

bhinnaṃ sambandhibhedena sambandhamapare viduḥ |
nimittaṃ savibhaktyarthaḥ samāsenābhidhīyate || 234 ||

Others know the relation, differentiated by the difference of the relata, as the cause; what bears the case-meaning is expressed by the compound.

The first horn's partisans. Bhinnam sambandhi-bhedena sambandham apare viduh: *others* know the *relation*, *differentiated by the difference of the relata*, as the *cause* — the relation is principal, and it is many because its terms are many: staff-bearing and horn-having are different relations. Nimittam savibhakty-arthah samasena abhidhiyate: what *bears the case-meaning* is *expressed by the compound*. So on this view the compound says the relation-with-its-terms, and that is what conditions the affixes. It is 3.14.232's first horn, and it has 3.14.219-221 against it. Bhartrhari sets both parties out and lets the earlier argument do the deciding — he does not repeat it.

3.14.235 unformalized

प्रधानमन्यार्थतया भिन्नं स्वैरुपसर्जनैः ।
निमित्तमभिधेयं वा सर्वपश्चादपेक्ष्यते ॥ २३५ ॥

pradhānamanyārthatayā bhinnaṃ svairupasarjanaiḥ |
nimittamabhidheyaṃ vā sarvapaścādapekṣyate || 235 ||

The principal, differentiated by its own subordinates through being for another's purpose, is required last of all — whether as cause or as what is expressed.

Pradhanam anyarthataya bhinnam svair upasarjanaih: the *principal*, *differentiated* by its *own subordinates* through *being for another's purpose* — the bahuvrihi's members, which serve an outside possessor (3.14.63's apujyamana: unhonoured, none of them principal). Nimittam abhidheyam va sarva-pascad apeksyate: is *required* *last of all* — sarva-pascat: after everything — whether as *cause* or as what is *expressed*. So the possessor comes last: the members are said, the relation is included, and only at the end is the one they were all for required. This is the bahuvrihi's peculiar temporal shape, and it explains 3.14.139's bhavin, bahiranga substratum — the thing that governs the compound's concord arrives after the compound is finished.

3.14.236 unformalized

स्वामिनि व्यतिरेकश्च वाक्ये यद्यपि दृश्यते ।
प्राधान्य एव तस्येष्टो बहुर्वीहिर्विवक्षिते ॥ २३६ ॥

svāmini vyatirekaśca vākye yadyapi dṛśyate |
prādhānya eva tasyeṣṭo bahurvīhirvivakṣite || 236 ||

And although difference is seen in the owner in the sentence — the bahuvrihi is desired only when his principality is intended.

Svamini vyatirekas ca vakye yady api drsyate: although *difference* is *seen* in the *owner* in the *sentence* — in 'citra gavo yasya', the owner stands apart, distinct, named by a relative pronoun. Pradhanya eva tasya isto bahuvrihir vivaksite: the *bahuvrihi* is *desired* *only* when *his principality* is *intended*. So the phrase and the compound differ in what they make principal: the phrase foregrounds the cows and appends the owner; the compound is chosen precisely when the owner is to be principal. This is 3.14.49's atyanta-bheda given its motive — one does not compound to abbreviate, one compounds to shift what is principal. Vivaksa decides (3.14.224), and the form follows.

3.14.237 unformalized

गवां विशेषणत्वेन यदा तद्वान् प्रवर्तते ।
अस्यैता इति तत्रार्थे बहुर्वीहिर्न विद्यते ॥ २३७ ॥

gavāṃ viśeṣaṇatvena yadā tadvān pravartate |
asyaitā iti tatrārthe bahurvīhirna vidyate || 237 ||

When the possessor operates as the qualifier of the cows — in that meaning, 'these are his', there is no bahuvrihi.

The test of 3.14.236, negative half. Gavam visesanatvena yada tadvan pravartate: when the *possessor* operates as the *qualifier* of the *cows* — the owner qualifying the herd rather than the reverse. Asya eta iti tatra arthe bahuvrihir na vidyate: in *that* meaning — *'these are his'* — there is *no bahuvrihi*. So where the cows are principal and the owner merely tells which cows, the compound is unavailable: one must use the phrase. The bahuvrihi is not a way of saying 'his cows'; it is a way of saying 'the man with the cows', and the difference is which way the qualification runs. This is as sharp a statement of the compound's semantics as the samuddesa gives.

3.14.238 unformalized

यदा प्रत्यवमर्शस्तु तासां स्वामी गवामिति ।
गोभिस्तदाभिसम्बन्धो निमित्तत्वाय कल्पते ॥ २३८ ॥

yadā pratyavamarśastu tāsāṃ svāmī gavāmiti |
gobhistadābhisambandho nimittatvāya kalpate || 238 ||

But when the reflective reference is 'he is the owner of those cows' — then the connection with the cows is conceivable as the cause.

The positive half. Yada pratyavamarsas tu tasam svami gavam iti: when the *reflective reference* is *'he is the owner of those cows'* — pratyavamarsa: the mind's turning back on what it has laid down (3.14.189's linga-pratyavamarsa, in the same sense). Gobhis tada abhisambandho nimittatvaya kalpate: *then* the *connection with the cows* is *conceivable as the cause*. So when the mind's backward glance lands on the *owner*, the cow-connection becomes the occasioning condition and the bahuvrihi is licensed. The whole determination is thus in the pratyavamarsa — in which way the cognition turns — and not in the world, where a man and his cows stand in one relation regardless (3.14.188's buddhya eva: by the intellect alone). Which is where this samuddesa has said everything is.

3.14.239 unformalized

अपेक्षमाणः सम्बन्धं रूढित्वस्य निवृत्तये ।
निमित्तानुविधायित्वात् तद्धर्मार्थः प्रसज्यते ॥ २३९ ॥

apekṣamāṇaḥ sambandhaṃ rūḍhitvasya nivṛttaye |
nimittānuvidhāyitvāt taddharmārthaḥ prasajyate || 239 ||

Requiring the relation, for the removal of conventionality — from conforming to the cause, the meaning possessing that property follows.

Apeksamanah sambandham rudhitvasya nivrttaye: *requiring* the *relation*, for the *removal of conventionality* — rudhitva: the state of having hardened into a name (3.14.212's vajrapani and tryaksa, frozen by convention). A compound that keeps requiring its relation cannot congeal into a proper name: the requirement holds it open. Nimitta-anuvidhayitvat tad-dharmarthah prasajyate: from *conforming to the cause*, the meaning *possessing that property* *follows*. So the live relation both prevents rudhi and delivers the properties. This explains why most bahuvrihis are not names: they go on requiring, and requiring keeps them general (3.14.213's samanyasya abhidhayakah).

3.14.240 unformalized

नानाचित्रा इति यथा निमित्तमनुरुध्यते (वर्तते) ।
नानाभूतेऽपि वृत्तः सन् बहुर्वीहिस्तथा भवेत् ॥ २४० ॥

nānācitrā iti yathā nimittamanurudhyate (vartate) |
nānābhūte'pi vṛttaḥ san bahurvīhistathā bhavet || 240 ||

As 'nana-citrah' conforms to its cause — so the bahuvrihi, though operating in what has become manifold, would be.

Nana-citra iti yatha nimittam anurudhyate: as *'nana-citrah'* — the manifoldly variegated — *conforms* to its *cause*: the plural is there because the variegation is many, and the form follows the fact. Nana-bhute 'pi vrttah san bahuvrihis tatha bhavet: *so* the *bahuvrihi*, though *operating* in what has *become manifold*, would be. So the compound may reach a plurality and still take its concord from its cause rather than from the plurality — as 'nana-citrah' does. The point is that manifoldness in the object does not force manifoldness in the form: the form answers to what conditions it (3.14.153's the properties belong to the conditions and do not migrate).
contested: The source records a variant in pada b: 'anurudhyate' ('conforms to') against 'vartate' ('operates in'). We translate the former, which the argument requires — the verse is about conformity to the nimitta, as pada b of 3.14.239 has just established. On the variant the sense would be that 'nana-citrah' merely operates in its cause, which is weaker but not incompatible.

3.14.241 unformalized

सम्बन्धिनि निमित्ते तु द्रव्यधर्मो न हीयते ।
लिङ्गाभावो हि लिङ्गस्य विरोधित्वेन वर्तते ॥ २४१ ॥

sambandhini nimitte tu dravyadharmo na hīyate |
liṅgābhāvo hi liṅgasya virodhitvena vartate || 241 ||

But when the relatum is the cause, the substance's property is not lost; for the absence of gender operates as the contradictory of gender.

Sambandhini nimitte tu dravya-dharmo na hiyate: when the *relatum* is the *cause* — 3.14.232's second horn, the one that survives — the *substance's property* is *not lost*. Lingabhavo hi lingasya virodhitvena vartate: for the *absence of gender* operates as the *contradictory* of gender. So the choice is stark: either the cause is a substance and gender comes with it, or the cause is a relation and gender is *contradicted* — not merely absent but excluded. Virodhitvena: as a contrary. This is why 3.14.228's avyaya-consequence was so threatening: relation does not merely fail to supply gender, it repels it. Hence the relatum must be the cause, and 3.14.232's fork is settled.

3.14.242 unformalized

संख्यावांल्लिङ्गश्चार्थोऽभिन्नधर्मा निमित्ततः ।
आसन्न एव द्रव्यत्वात् तद्धर्मैर्न विरुध्यते ॥ २४२ ॥

saṃkhyāvāṃlliṅgaścārtho'bhinnadharmā nimittataḥ |
āsanna eva dravyatvāt taddharmairna virudhyate || 242 ||

And the meaning possessing number and gender, of undivided properties, from its cause — being near, from substantiality, is not contradicted by its own properties.

Samkhyavaml lingas ca artho 'bhinna-dharma nimittatah: the meaning possessing *number* and *gender*, of *undivided properties*, from its *cause* — abhinna-dharma: its properties not split off from it (3.14.181's murti-dharmas not other than the murti). Asanna eva dravyatvat tad-dharmair na virudhyate: *being near*, from *substantiality*, it is *not contradicted* by *its properties*. Asanna: close at hand, adjacent. So because the meaning is substance-like it is *near* its gender and number, and no contradiction arises — where 3.14.241's relation was far and repellent. The bahuvrihi's meaning is thus a near-substance: not a substance, near enough to hold concord. That is as precise as this can be made.

3.14.243 unformalized

विभक्त्यर्थेन चाविष्टं शुद्धं चेति द्विधा स्थितम् ।
द्रव्यं शुद्धस्य यो धर्मः स न स्यादन्यधर्मणः ॥ २४३ ॥

vibhaktyarthena cāviṣṭaṃ śuddhaṃ ceti dvidhā sthitam |
dravyaṃ śuddhasya yo dharmaḥ sa na syādanyadharmaṇaḥ || 243 ||

Substance stands in two ways: pervaded by the case-meaning, and pure. That property which belongs to the pure would not belong to what has another property.

A distinction with consequences. Vibhakty-arthena ca avistam suddham ceti dvidha sthitam dravyam: *substance* stands *in two ways* — *pervaded by the case-meaning* (avista: entered into, 3.14.175's word), and *pure*. Suddhasya yo dharmah sa na syad anya-dharmanah: that *property* which belongs to the *pure* would *not* belong to what has *another property*. So the two substances have different properties and one cannot argue from the one to the other: the substance caught up in a relation is not the substance simpliciter, and what holds of the second need not hold of the first. This is the ground of 3.14.242's asanna: the bahuvrihi's meaning is the pervaded substance, near the pure one and not it.

3.14.244 unformalized

द्रव्यमात्रस्य निर्देशे भेदोऽयमविवक्षितः ।
ग्रन्थे पूर्वत्र भेदस्तु द्वितीयेऽनुप्रदर्शितः ॥ २४४ ॥

dravyamātrasya nirdeśe bhedo'yamavivakṣitaḥ |
granthe pūrvatra bhedastu dvitīye'nupradarśitaḥ || 244 ||

In the indication of mere substance, this difference is not intended; in the earlier part of the work, the difference; but in the second it has been shown.

A rare internal cross-reference. Dravya-matrasya nirdese bhedo 'yam avivaksitah: in the *indication* of *mere substance*, *this difference* — 3.14.243's pure against pervaded — is *not intended*: where one says simply 'substance', the distinction is not being drawn (3.14.103's avivaksita). Granthe purvatra bhedas tu dvitiye 'nupradarsitah: *in the earlier part of the work*, the *difference*; but in the *second* it has been *shown*. So Bhartrhari is pointing his reader back — the Dravya-samuddesa (3.2) and the Dravyalaksana-samuddesa (3.4) treated substance one way, and the second kanda another. Granthe purvatra: *earlier in the book*. The Vakyapadiya knows itself as a single structure and says so here.

3.14.245 unformalized

द्रव्यस्य ग्रहणं चात्र लिङ्गसंख्याविशेषणम् ।
द्रव्याश्रितत्वं हि तयोस्ततोऽन्यस्य न सिध्यतः ॥ २४५ ॥

dravyasya grahaṇaṃ cātra liṅgasaṃkhyāviśeṣaṇam |
dravyāśritatvaṃ hi tayostato'nyasya na sidhyataḥ || 245 ||

And here the taking-up of the word 'substance' is the qualifier of gender and number; for those two rest on substance, and do not succeed for anything else.

Dravyasya grahanam ca atra linga-samkhya-visesanam: the *taking-up* of the word *'substance'* — its appearance in the rule — is the *qualifier* of *gender and number*: the word is there to specify which gender and number are meant, namely a substance's. Dravyasritatvam hi tayoh: for those *two* *rest on substance*. Tato 'nyasya na sidhyatah: and for *anything else* they do *not succeed*. So the samuddesa's long gender-and-number enquiry ends where 3.14.163 said it would: dravyam eva asrayah, and every other candidate — quality, universal, relation, accumulation — has been tried and has failed. The word 'dravya' in the sutra is Panini registering the same finding.

3.14.246 unformalized

सम्बन्धिभिन्नसम्बन्धपरिच्छिन्ने प्रवर्तते ।
समासो द्रव्यसामान्ये विशिष्टार्थानुपातिनि ॥ २४६ ॥

sambandhibhinnasambandhaparicchinne pravartate |
samāso dravyasāmānye viśiṣṭārthānupātini || 246 ||

The compound operates in the substance-general, delimited by relations differentiated by their relata, conforming to the specified meaning.

The bahuvrihi's meaning, finally stated in full. Sambandhi-bhinna-sambandha-paricchinne: *delimited* by *relations* *differentiated by the relata* (3.14.234's bhinnam sambandhi-bhedena, now used affirmatively). Pravartate samaso dravya-samanye: the *compound* operates in the *substance-general* — dravya-samanya: not a substance and not a universal, but substance-in-general, the near-substance of 3.14.242. Visistartha-anupatini: *conforming to the specified meaning* — anupatin: running after, following on. So: 'citragu' means substance-in-general, cut down by the cow-relation, running toward whatever specified thing satisfies it. Every word of this has been earned across forty verses, and none of it is spare.

3.14.247 unformalized

द्रव्यधर्मानतिक्रान्तो भेदधर्मेष्ववस्थितः (ष्वशक्तितः) ।
भविष्यदाश्रयापेक्षे लिङ्गसङ्ख्ये प्रपद्यते ॥ २४७ ॥

dravyadharmānatikrānto bhedadharmeṣvavasthitaḥ (ṣvaśaktitaḥ) |
bhaviṣyadāśrayāpekṣe liṅgasaṅkhye prapadyate || 247 ||

Not transgressing the substance's property, settled in the difference-properties, it attains gender and number which depend on a future substratum.

Dravya-dharma-anatikrantah: *not transgressing* the substance's property — it stays within what a substance may do, and so may hold concord (3.14.245). Bheda-dharmesv avasthitah: *settled in* the *difference-properties*. Bhavisyad-asraya-apekse linga-sankhye prapadyate: it *attains* *gender and number* which *depend* on a *future substratum*. Bhavisyat: the substratum is *yet to come* — 3.14.139's bhavin and bahiranga, and 3.14.235's sarva-pascat: the possessor arrives last. So the bahuvrihi's concord is owed to something that does not yet exist when the compound is formed, and the compound holds the place open. That is the whole peculiarity of the form, and here it is stated exactly.
contested: Pada b is transmitted with a variant: 'bheda-dharmesv avasthitah' ('settled in the difference-properties') against 'bheda-dharmesv asaktitah' ('from incapacity in the difference-properties'). We translate the former. The variant would reverse the pada's force — the compound failing at the differences rather than resting in them — and would fit 3.14.241's repulsion-argument; but 'avasthitah' better suits pada a's 'anatikrantah', which describes staying within bounds rather than falling short.

3.14.248 unformalized

शास्त्रप्रवृत्तिभेदेऽपि लौकिकार्थो न भिद्यते ।
नञ्समासे यतस्तत्र त्रयः पक्षा विचारिताः ॥ २४८ ॥

śāstrapravṛttibhede'pi laukikārtho na bhidyate |
nañsamāse yatastatra trayaḥ pakṣā vicāritāḥ || 248 ||

Even where the science's operation differs, the worldly meaning is not differentiated; since, in the negative compound, three positions have been considered there.

The negative compound begins, and it begins with the samuddesa's constant refrain. Sastra-pravrtti-bhede 'pi laukiko 'rtho na bhidyate: *even where* the *science's operation* *differs*, the *worldly meaning* is *not differentiated* — however many ways the grammar derives it, 'abrahmana' means what it means (3.14.74's the science stands very far off; 3.14.97's the upayas unsettled among persons). Nan-samase yatas tatra trayah paksa vicaritah: since, in the *nan-compound*, *three positions* have been *considered* there. So three rival accounts are on record and the word is indifferent to all of them. This is the frame for everything that follows: the analyses will multiply and nothing in the language will move.

3.14.249 unformalized

शब्दान्तरेऽपि चैकत्वमाश्रित्यैव विचारणा ।
अब्राह्मणादिषु नञः प्रयोगो न हि विद्यते ॥ २४९ ॥

śabdāntare'pi caikatvamāśrityaiva vicāraṇā |
abrāhmaṇādiṣu nañaḥ prayogo na hi vidyate || 249 ||

Even where it is another word, the consideration proceeds only by resorting to oneness; for in 'abrahmana' and the rest, the employment of 'nañ' does not exist.

Sabdantare 'pi ca ekatvam asritya eva vicarana: *even where* it is *another word*, the *consideration* proceeds *only* by *resorting to oneness* — the analysis treats the compound and the phrase as one thing in order to discuss them, though 3.14.49 established they are utterly different words. The unity is the analyst's, assumed to make the discussion possible. Abrahmanadisu nanah prayogo na hi vidyate: for in *'abrahmana'* and the rest, the *employment* of *'nañ'* does *not exist* — the technical particle 'nañ' is a grammarian's object; what one says is 'a-'. So the three positions of 3.14.248 are being debated about a thing that is not in the word: nañ is notation (3.14.78's anubandha-svaradi).

3.14.250 unformalized

प्राक् समासात् पदार्थानां निवृत्तिर्द्योत्यते नञा ।
स्वभावतो निवृत्तानां रूपाभेदादलक्षिता ॥ २५० ॥

prāk samāsāt padārthānāṃ nivṛttirdyotyate nañā |
svabhāvato nivṛttānāṃ rūpābhedādalakṣitā || 250 ||

Before the compound, the exclusion of the word-meanings is indicated by 'nañ'; of those excluded by their own nature, unmarked from non-difference of form.

The first position, and it is the dyotaka one. Prak samasat padarthanam nivrttir dyotyate nana: *before* the compound, the *exclusion* of the *word-meanings* is *indicated* by *'nañ'* — dyotyate: indicated, not expressed (3.14.191's the particle restricts what was latent). Svabhavato nivrttanam rupabhedad alaksita: of those *excluded* *by their own nature*, *unmarked* from *non-difference of form*. So the exclusion was already there — everything excludes what it is not, by its own nature (3.14.142's yad yad asriyate tat tad anyasya vinivartakam) — and it went unmarked because the form did not show it (3.14.140's alaksitah). 'Nañ' does not negate; it marks an exclusion that every word was already performing. This is a beautiful position and it follows from everything the samuddesa has said about generals.

3.14.251 unformalized

ब्राह्मणादिस्थया वाक्येष्वाख्यातपदवाच्यया ।
क्रियया यस्य सम्बन्धो वृत्तिस्तस्य न विद्यते ॥ २५१ ॥

brāhmaṇādisthayā vākyeṣvākhyātapadavācyayā |
kriyayā yasya sambandho vṛttistasya na vidyate || 251 ||

That which has connection with the action residing in 'brahmana' and the rest, expressed by the verb-word in sentences — of that there is no derivative formation.

Brahmanadi-sthaya vakyesv akhyata-pada-vacyaya kriyaya yasya sambandhah: that which has *connection* with the *action* residing in *'brahmana'* and the rest, *expressed by the verb-word* in *sentences* — the being-a-brahmin that the phrase 'na brahmanah' denies by its verb. Vrttis tasya na vidyate: of *that* there is *no derivative formation*. So the sentence's negation attaches to an action said by a verb, and no compound can do that: compounds have no verbs (3.14.220's vibhakty-artha-pradhanatvat... na kriya). Hence 'abrahmana' is not the compressed 'na brahmano bhavati' — it cannot be, because what the phrase negates is not available to be compounded.

3.14.252 unformalized

पाचकादिपदस्था चेन्नञा सम्बन्ध्यते क्रिया ।
तत्र सत्तानुपादानात् त्रिपक्षी नोपपद्यते ॥ २५२ ॥

pācakādipadasthā cennañā sambandhyate kriyā |
tatra sattānupādānāt tripakṣī nopapadyate || 252 ||

If the action residing in 'pacaka' and the rest is connected with 'nañ' — there, from the non-taking-up of existence, the three-fold position is not established.

Pacakadi-pada-stha cen nana sambandhyate kriya: *if* the *action* residing in *'pacaka'* (cook) and the rest is *connected with 'nañ'* — 'apacaka', the non-cook: here the base does contain an action, cooking, so perhaps the negation reaches it. Tatra satta-anupadanat tri-paksi na upapadyate: *there*, from the *non-taking-up of existence*, the *three-fold position* is *not established*. So the three accounts of 3.14.248 all assumed that nañ negates an *existence*, and 'apacaka' negates a cooking without touching any existence — the cook still is. So the trio does not apply here, and its generality is broken. Bhartrhari is dismantling the framework of the debate rather than joining it.

3.14.253 unformalized

सत्तयैवाभिसम्बन्धो यदि सर्वत्र कल्प्यते ।
असन्निति समासेऽस्मिन् सत्तान्या परिकल्प्यताम् ॥ २५३ ॥

sattayaivābhisambandho yadi sarvatra kalpyate |
asanniti samāse'smin sattānyā parikalpyatām || 253 ||

If connection with existence alone is imagined everywhere — then in this compound 'asat', let another existence be imagined.

The reductio, and it is very fine. Sattaya eva abhisambandho yadi sarvatra kalpyate: *if* connection with *existence alone* is *imagined* *everywhere* — kalpyate: imagined, constructed (3.14.231's kalpana). Asann iti samase 'smin satta anya parikalpyatam: then in *this* compound *'asat'* — the non-existent — let *another existence* be *imagined*. Parikalpyatam: let it be imagined — the imperative is sardonic. So if every nañ-compound negates an existence, then 'asat' negates one too, and one must find an existence for it to negate, and then another to be negated in that, and so on. The account destroys itself on its own crucial word. The argument is three lines long and it is complete.

3.14.254 unformalized

क्त्वान्ते च तुमुनन्ते च नञ्समासे न दृश्यते ।
विशेषणविशेष्यत्वं नञा (नञः) सत्ताभिधायिनां ॥ २५४ ॥

ktvānte ca tumunante ca nañsamāse na dṛśyate |
viśeṣaṇaviśeṣyatvaṃ nañā (nañaḥ) sattābhidhāyināṃ || 254 ||

In the ktva-ending and in the tumun-ending, in the negative compound, the qualifier-qualified relation of 'nañ' with those expressing existence is not seen.

Two more counter-cases. Ktvante ca tumun-ante ca nan-samase na drsyate: in the *ktva*-ending (the gerund: 'akrtva', not having done) and the *tumun*-ending (the infinitive), in the *nan-compound*, it is *not seen*. Visesana-visesyatvam nanah satta-abhidhayinam: the *qualifier-qualified* relation of *'nañ'* with *those expressing existence*. So in 'akrtva' the nañ stands in no qualifier-qualified relation to anything existence-expressing — there is no existence in a gerund to qualify. Na drsyate: it is *not seen* (3.14.114's drsyate: the argument is observation). So the framework fails again, on forms everyone uses. Between 3.14.252 and this verse the existence-account has been refuted from four directions, and the samuddesa has not yet said what it thinks nañ does.

3.14.255 unformalized

क्रियायाः साधनाधारसामान्ये नञ् व्यवस्थितः ।
ततो विशिष्टैराधारैर्युज्यते ब्राह्मणादिभिः ॥ २५५ ॥

kriyāyāḥ sādhanādhārasāmānye nañ vyavasthitaḥ |
tato viśiṣṭairādhārairyujyate brāhmaṇādibhiḥ || 255 ||

'Nañ' is settled in the general of the substratum which is the action's means; then it is joined with specified substrata — 'brahmana' and the rest.

The samuddesa's own account of the negative, at last. Kriyayah sadhana-adhara-samanye nan vyavasthitah: *'nañ'* is *settled* in the *general* of the *substratum* which is the *action's means* — not in an existence (3.14.252-254's refuted framework), but in the general of a *seat*. Tato visistair adharair yujyate brahmanadibhih: *then* it is *joined* with *specified substrata* — *'brahmana'* and the rest. So the order is: nañ occupies a general first, and is *afterwards* fitted to a specific base. It does not negate the brahmin; it settles in a general seat and then that seat turns out to be brahmin-shaped. This is why 'abrahmana' means a ksatriya and not simply anything-not-a-brahmin: the general was a seat, and seats are furnished.

3.14.256 unformalized

वृत्तौ यथा गताद्यर्थमुपादाय निरादयः ।
युज्यन्ते साधनाधारैर्नञ्समासेऽपि स क्रमः ॥ २५६ ॥

vṛttau yathā gatādyarthamupādāya nirādayaḥ |
yujyante sādhanādhārairnañsamāse'pi sa kramaḥ || 256 ||

As, in the derivative formation, 'nir' and the rest, having taken up the meaning 'gone' and so on, are joined with the substrata which are means — in the negative compound too, that is the sequence.

The precedent for 3.14.255. Vrttau yatha gatady-artham upadaya niradayah: as in the *derivative formation*, *'nir'* and the rest, having *taken up* the meaning *'gone'* and so on — 3.14.39's preverbs, whose transcending-sense exists only in composition. Yujyante sadhana-adharaih: are *joined* with the *substrata* which are *means*. Nan-samase 'pi sa kramah: in the *nan-compound* too, *that* is the *sequence*. So nañ works exactly as 'nir' does: it acquires a sense in the composing and is then fitted to a base. Neither is a negation or a preposition operating on something already there. Sa kramah — *that* sequence: the ordering is the whole of the explanation.

3.14.257 unformalized

तत्रासति नञो वृत्तेर्ब्राह्मणक्षत्रियादिभिः ।
विशेषणविशेष्यत्वं कल्प्यते कुब्जखञ्जवत् ॥ २५७ ॥

tatrāsati naño vṛtterbrāhmaṇakṣatriyādibhiḥ |
viśeṣaṇaviśeṣyatvaṃ kalpyate kubjakhañjavat || 257 ||

There, in 'asat', from the derivation of 'nañ', the qualifier-qualified relation with 'brahmana', 'ksatriya' and the rest is imagined — as with 'hunchbacked' and 'lame'.

Tatra asati nano vrtteh: *there*, in *'asat'*, from the *derivation* of *'nañ'*. Brahmana-ksatriyadibhir visesana-visesyatvam kalpyate: the *qualifier-qualified* relation with *'brahmana'*, *'ksatriya'* and the rest is *imagined* — kalpyate: constructed, not found. Kubja-khanjavat: as with *'kubja'* (hunchbacked) and *'khanja'* (lame) — two attributes of one man, neither qualifying the other, held together only by co-reference (3.14.9-11's krsnas tilah). So the relation between the negation and the base is like that between a hunchback's two defects: they meet in a subject and not in each other. The negative compound has no internal qualification, only a shared seat.
contested: Pada b's syntax is strained: 'asati' and 'vrtteh' do not construe smoothly together. We read 'asati' as locative of 'asat' (picking up 3.14.253's example) with 'vrtteh' ablative. Read instead as a locative absolute — 'the derivation of nañ not existing' — the verse would say that the qualifier-qualified relation is imagined precisely where nañ has no derivation, which also suits 'kalpyate'. The comparison to kubja-khanja stands on either construal.

3.14.258 unformalized

कामचारे च सत्येवमसतः स्यात् प्रधानता ।
गुणत्वमितरेषां च तेषां वा स्यात् प्रधानत् ॥ २५८ ॥

kāmacāre ca satyevamasataḥ syāt pradhānatā |
guṇatvamitareṣāṃ ca teṣāṃ vā syāt pradhānat || 258 ||

And where there is thus free play, 'asat' would be principal and the others subordinate; or they would be principal.

The cost of 3.14.257's imagined relation. Kama-care ca saty evam: where there is thus *free play* — kama-cara: doing as one likes, the state of a question nothing settles (3.14.168's icchaya: at one's pleasure). Asatah syat pradhanata gunatvam itaresam ca: *'asat'* would be *principal* and the *others* *subordinate*. Tesam va syat pradhanata: *or* *they* would be principal. So if the relation is merely imagined, nothing decides which side is principal, and both assignments are available. This is exactly what one should expect of a construct (3.14.96's upaya-matram nanatvam), and it is a defect only if one thought the construct was a description.

3.14.259 unformalized

प्राधान्येनाश्रिताः पूर्वं श्रुतेः सामान्यवृत्तयः ।
विशेष एव प्राकान्ता ब्राह्मणक्षत्रियादयः ॥ २५९ ॥

prādhānyenāśritāḥ pūrvaṃ śruteḥ sāmānyavṛttayaḥ |
viśeṣa eva prākāntā brāhmaṇakṣatriyādayaḥ || 259 ||

'Brahmana', 'ksatriya' and the rest — resorted to as principal at first, operating in the general because of what is heard — are commenced upon precisely in the particular.

Pradhanyena asritah purvam: resorted to as *principal* *at first* — the base-word comes first and comes as principal. Sruteh samanya-vrttayah: *operating in the general* because of *what is heard* — the audible word 'brahmana' means the general, brahmin-hood at large. Visesa eva prakranta brahmana-ksatriyadayah: they are *commenced upon* *precisely in the particular* — prakranta: begun, set out from. So the movement is general-to-particular within a single word: 'brahmana' sounds the general and the discourse begins from a particular. This is 3.14.37's visesa-vrttitva (in composition the member means the particular) shown as a movement rather than a state, and it is what 3.14.255's 'then' was describing.

3.14.260 unformalized

यथा गौरादिभिस्तेषामवच्छेदो विधीयते ।
असताप्यनभिव्यक्तं तादात्म्यं व्यज्यते तथा ॥ २६० ॥

yathā gaurādibhisteṣāmavacchedo vidhīyate |
asatāpyanabhivyaktaṃ tādātmyaṃ vyajyate tathā || 260 ||

As their delimitation is effected by 'gaura' and the rest — so, by 'asat' too, an unmanifested identity is manifested.

The positive characterization of what a negative does. Yatha gauradibhis tesam avacchedo vidhiyate: as their *delimitation* is effected by *'gaura'* (fair) and the rest — an ordinary adjective cuts a class down. Asata api anabhivyaktam tadatmyam vyajyate tatha: *so*, by *'asat'* too, an *unmanifested identity* is *manifested*. Tadatmya: identity, being-that-self; anabhivyakta: unmanifested; vyajyate: is made manifest. So the negative does what the adjective does — it delimits — and what it delimits into view is an *identity* that was there unmanifested. 'Abrahmana' does not remove a brahmin; it manifests a ksatriya-identity that was in the field unshown. This is 3.14.250's position perfected: exclusion is how things get shown, and nañ shows by excluding.

3.14.261 unformalized

यथा सत्ताभिधानाय सन्नर्थः परिकल्प्यते ।
तथा सत्ताभिधानाय निरुपाख्यो विकल्प्यते ॥ २६१ ॥

yathā sattābhidhānāya sannarthaḥ parikalpyate |
tathā sattābhidhānāya nirupākhyo vikalpyate || 261 ||

As the existent meaning is imagined for the expression of existence — so the indefinable is constructed for the expression of non-existence.

The parity argument. Yatha satta-abhidhanaya sann arthah parikalpyate: as the *existent meaning* is *imagined* for the *expression of existence* — even to say 'it is', one constructs an existent for the saying to be about. Tatha satta-abhidhanaya nirupakhyo vikalpyate: so the *indefinable* — nirupakhya: that of which no account can be given, the technical term for the utterly unreal — is *constructed* for the expression of non-existence. So both existence and non-existence require a posit, and the posit for non-existence is a thing defined as having no definition. Neither side is innocent. This is 3.14.253's reductio turned into a general observation: reference to being and to non-being are equally constructions (3.14.197's the word ordains the differences).
contested: Pada c reads 'satta-abhidhanaya' in our source, duplicating pada a. The parallel ('yatha... tatha') plainly requires 'asatta-abhidhanaya' — for the expression of *non*-existence — which is what we translate; otherwise the verse says the same thing twice and the nirupakhya has no work. We note the transmitted reading rather than pass over the emendation.

3.14.262 unformalized

क्षत्रियादौ पदं कृत्वा बुद्धिः सत्तान्तराश्रया ।
जात्या भिन्नां ततः सत्तां प्रसक्तामपकर्षति ॥ २६२ ॥

kṣatriyādau padaṃ kṛtvā buddhiḥ sattāntarāśrayā |
jātyā bhinnāṃ tataḥ sattāṃ prasaktāmapakarṣati || 262 ||

The intellect, having made a footing in 'ksatriya' and the rest, resting on another existence, then draws away the applicable existence differentiated by the universal.

The negative described as a mental movement — and this is the samuddesa's real answer. Ksatriyadau padam krtva buddhih: the *intellect*, having *made a footing* — padam krtva: having taken a step, planted itself — in *'ksatriya'* and the rest. Satta-antarasraya: *resting on another existence*. Jatya bhinnam tatah sattam prasaktam apakarsati: *then* *draws away* the *applicable* *existence* *differentiated by the universal*. Apakarsati: pulls off, drags away (3.14.11's apakarsati, of the two generals drawing away each other's doubt). So: the mind plants itself on a ksatriya, stands on that existence, and from there *pulls away* the brahmin-existence that was pressing. The negation is a shift of footing followed by a withdrawal — done by the buddhi, in two moves, and there is no negative fact anywhere in it.

3.14.263 unformalized

अभाव इति भावस्य प्रतिषेधे विवक्षिते ।
सोपाख्यत्वमनाश्रित्य प्रतिषेधो न कल्पते ॥ २६३ ॥

abhāva iti bhāvasya pratiṣedhe vivakṣite |
sopākhyatvamanāśritya pratiṣedho na kalpate || 263 ||

When the negation of being is intended as 'non-being' — without resorting to definability, negation is not conceivable.

Abhava iti bhavasya pratisedhe vivaksite: when the *negation of being* is *intended* as *'non-being'* — the hardest case, where what is denied is being itself. Sopakhyatvam anasritya pratisedho na kalpate: *without resorting to definability* — sopakhya: that of which an account *can* be given, the opposite of 3.14.261's nirupakhya — *negation* is *not conceivable*. So even to deny being, one must have something definable to plant one's foot on (3.14.262's padam krtva). A negation of the utterly indefinable is not a negation, because there is nowhere to stand. This is why 3.14.261's posit was unavoidable: the mind's withdrawal requires a footing, and a footing must be describable.

3.14.264 unformalized

अनेकधर्मवचनैः शब्दैः सङ्घाभिधायिभिः ।
एकदेशेषु वर्तन्ते तुल्यरूपाः स्वभावतः ॥ २६४ ॥

anekadharmavacanaiḥ śabdaiḥ saṅghābhidhāyibhiḥ |
ekadeśeṣu vartante tulyarūpāḥ svabhāvataḥ || 264 ||

By words expressing many properties, expressing an aggregate — they operate in the parts, similar in form by their own nature.

Aneka-dharma-vacanaih sabdaih sangha-abhidhayibhih: by words *expressing many properties*, *expressing an aggregate* — words like 'brahmana', which name a bundle: birth, study, conduct, and the rest. Ekadesesu vartante tulya-rupah svabhavatah: they *operate in the parts*, *similar in form* by their *own nature*. So an aggregate-word applies to its parts as readily as to the whole, and the form does not distinguish the two applications: they are tulya-rupa, alike in shape, by nature. This is the premise of the paradox that 3.14.265 is about to spring, and it is not a trick — it is a fact about every word that names a bundle.

3.14.265 unformalized

यथैकदेशकरणात् कृत इत्यभिधीयते ।
अकृतश्चेति सङ्घातः स एवाब्राह्मणे क्रमः ॥ २६५ ॥

yathaikadeśakaraṇāt kṛta ityabhidhīyate |
akṛtaśceti saṅghātaḥ sa evābrāhmaṇe kramaḥ || 265 ||

As, from the doing of one part, an aggregate is spoken of as 'done' and also as 'not done' — that same sequence is in 'abrahmana'.

The paradox, and it is superb. Yatha ekadesa-karanat krta ity abhidhiyate akrtas ceti sanghatah: as, from the *doing of one part*, an *aggregate* is spoken of as *'done'* *and* as *'not done'* — a half-built house is built and unbuilt, and both are true, because 3.14.264's words apply to parts and wholes alike. Sa eva abrahmane kramah: *that same* sequence is in *'abrahmana'*. So a man with some of the brahmin-properties and not others is a brahmana and an abrahmana at once. This is not a puzzle to be dissolved; it is what aggregate-words do, and 3.14.266 draws the consequence without flinching.

3.14.266 unformalized

ब्राह्मणोऽब्राह्मणस्तस्मादुपन्यासात् प्रसज्यते ।
अकृते वा कृतासङ्गादविशिष्टं कृताकृतात् ॥ २६६ ॥

brāhmaṇo'brāhmaṇastasmādupanyāsāt prasajyate |
akṛte vā kṛtāsaṅgādaviśiṣṭaṃ kṛtākṛtāt || 266 ||

From that presentation, a brahmana would be a non-brahmana; or, in the undone, from the attachment of the done, it is not different from the done-and-undone.

The consequence. Brahmano 'brahmanas tasmad upanyasat prasajyate: from *that presentation*, a *brahmana* would be a *non-brahmana* — the contradiction, stated as such. Akrte va krtasangad avisistam krtakrtat: *or*, in the *undone*, from the *attachment* of the *done*, it is *not different* from the *done-and-undone*. So either the brahmin is a non-brahmin, or the undone thing is indistinguishable from the half-done. Both are unacceptable, and both follow from 3.14.264's premise, which is true. So something must be added — and 3.14.267 adds it: the grammar's own decision.

3.14.267 unformalized

अमुख्यसम्भवे तत्र मुख्यस्य विनिवृत्तये ।
शास्त्रान्वाख्यानसमये नञ् प्रयुक्तो विशेषकः ॥ २६७ ॥

amukhyasambhave tatra mukhyasya vinivṛttaye |
śāstrānvākhyānasamaye nañ prayukto viśeṣakaḥ || 267 ||

There, when a non-primary is possible, 'nañ' — employed at the time of the science's exposition — is a differentiator, for the exclusion of the primary.

The resolution, and it is frankly conventionalist. Amukhya-sambhave tatra mukhyasya vinivrttaye: *there*, when a *non-primary* is *possible*, for the *exclusion of the primary* — where a word might be taken in its full or its partial sense, the full sense must be got out of the way. Sastra-anvakhyana-samaye nan prayukto visesakah: *'nañ'*, *employed* at the *time of the science's exposition*, is a *differentiator*. Anvakhyana-samaye: at the moment of the grammarian's *exposition* (3.14.71's anvakhyanaya bhedah). So nañ is the grammar's device for forcing the non-primary reading, deployed when the science is expounding. It resolves 3.14.266's contradiction by decision, not by discovery — which is what a science that stands atidure (3.14.74) has to do.

3.14.268 unformalized

पदार्थानुपघातेन दृश्यते च विशेषणम् ।
अथ जातिमतोऽर्थस्य कश्चिद् धर्मो निवर्तितः ॥ २६८ ॥

padārthānupaghātena dṛśyate ca viśeṣaṇam |
atha jātimato'rthasya kaścid dharmo nivartitaḥ || 268 ||

And a qualifier is seen not to damage the word-meaning. Now, some property of the meaning possessing a universal has been excluded.

Two steps toward a new difficulty. Padartha-anupaghatena drsyate ca visesanam: a *qualifier* is *seen* *without damaging* the *word-meaning* — anupaghata: no injury: adding 'blue' to 'lotus' does not destroy lotus-hood, and qualifiers generally leave their bases intact (3.14.260's delimitation, which cuts down without cutting up). Atha jatimato 'rthasya kascid dharmo nivartitah: *now*, *some property* of the meaning *possessing a universal* has been *excluded*. So nañ, if it is a qualifier (3.14.257), must likewise leave the base intact while excluding one of its properties. And that raises the question of 3.14.269: which property, and does it matter?

3.14.269 unformalized

अवश्यं ब्राह्मणे कश्चित् क्वचिद् धर्मो न विद्यते ।
विशेषावचनात् तत्र नञः श्रुतिरनर्थिका ॥ २६९ ॥

avaśyaṃ brāhmaṇe kaścit kvacid dharmo na vidyate |
viśeṣāvacanāt tatra nañaḥ śrutiranarthikā || 269 ||

Necessarily, some property somewhere does not exist in a brahmana; there, from the non-statement of the particular, the hearing of 'nañ' is purposeless.

The difficulty, and it is sharp. Avasyam brahmane kascit kvacid dharmo na vidyate: *necessarily* — avasyam — *some* property *somewhere* does *not exist* in a *brahmana*: no one has every property; every brahmin lacks something. Visesa-avacanat tatra nanah srutir anarthika: *there*, from the *non-statement of the particular*, the *hearing of 'nañ'* is *purposeless* — anarthika (3.14.126, 3.14.205's word). So if nañ merely excludes some property or other, it says nothing: every brahmin already fails some property, and 'abrahmana' would be true of all of them. Unless nañ says *which* property — and it does not. The account of 3.14.268 collapses into vacuity.

3.14.270 unformalized

अविशिष्टस्य पर्यायो नञ्विशिष्टः प्रसज्यते ।
अन्वाख्यानाद्धि साधुत्वमेवम्भूते प्रतीयते ॥ २७० ॥

aviśiṣṭasya paryāyo nañviśiṣṭaḥ prasajyate |
anvākhyānāddhi sādhutvamevambhūte pratīyate || 270 ||

The nañ-qualified would become a synonym of the unqualified; for correctness in such a case is cognized from the exposition.

The consequence of 3.14.269. Avisistasya paryayo nan-visistah prasajyate: the *nañ-qualified* would become a *synonym* of the *unqualified* — 'abrahmana' and 'brahmana' would mean the same, since every brahmin lacks something. Paryaya: a synonym — the reductio's final form, and it is worse than 3.14.266's contradiction. Anvakhyanad dhi sadhutvam evambhute pratiyate: for *correctness* in *such a case* is cognized from the *exposition* — anvakhyana: the grammarian's exposition (3.14.267's samaya). So what saves the word from synonymy is not anything in it: it is the science's exposition, which decides how it is to be taken. The negative compound is legible because the grammarians settled how to read it — and Bhartrhari, having demonstrated that nothing else could settle it, says so.

3.14.271 unformalized

पदार्थानुपघातेन यद्यप्यत्र विशेषणम् ।
उपचारसतोऽर्थस्य सावस्था द्योत्यते नञा ॥ २७१ ॥

padārthānupaghātena yadyapyatra viśeṣaṇam |
upacārasato'rthasya sāvasthā dyotyate nañā || 271 ||

Although here the qualifier does not damage the word-meaning — that state of the meaning which exists by figurative transfer is indicated by 'nañ'.

The answer to 3.14.268-270. Padartha-anupaghatena yady apy atra visesanam: *although* here the *qualifier* does *not damage* the *word-meaning* (3.14.268's principle, conceded). Upacara-sato 'rthasya sa avastha dyotyate nana: that *state* of the meaning which *exists by figurative transfer* is *indicated* by *'nañ'*. Upacara-sat: existing by upacara — by transferred usage, the mode of being that belongs to what is only figuratively so (3.3.82's beings raised up by construction). Dyotyate: *indicated*, not expressed (3.14.250's dyotyate nana). So nañ points at the base's *figurative* condition — 'abrahmana' shows brahmin-hood being held in a transferred way — and 3.14.269's vacuity is avoided: nañ does not exclude some property or other, it marks a mode of holding.

3.14.272 unformalized

विशेष्येषु यथाभूतः पदार्थः समवस्थितः ।
तथाभूते तथाभावो गम्यते भेदहेतुभिः ॥ २७२ ॥

viśeṣyeṣu yathābhūtaḥ padārthaḥ samavasthitaḥ |
tathābhūte tathābhāvo gamyate bhedahetubhiḥ || 272 ||

As the word-meaning stands settled in the qualified things — in such a one, the being-such is understood through the difference-causes.

Visesyesu yathabhutah padarthah samavasthitah: as the *word-meaning* stands *settled* in the *qualified things* — however it happens to sit in its instances. Tathabhute tathabhavo gamyate bheda-hetubhih: in *such a one*, the *being-such* is *understood* through the *difference-causes*. So the mode in which the meaning sits — fully, partially, figuratively — is read off the differentiating conditions, and nothing else reveals it. This grounds 3.14.271: nañ can indicate a figurative holding because the mode of holding is what the bheda-hetus disclose anyway. The negative is one such cause among others, doing what they all do.

3.14.273 unformalized

निवृत्तेऽवयवस्तस्मिन् पदार्थे वर्तते कथम् ।
नानिमित्ता हि शब्दस्य प्रवृत्तिरुपपद्यते ॥ २७३ ॥

nivṛtte'vayavastasmin padārthe vartate katham |
nānimittā hi śabdasya pravṛttirupapadyate || 273 ||

When that word-meaning is excluded, how does the member operate? For a word's operation without a cause is not possible.

The objection that will not go away. Nivrtte 'vayavas tasmin padarthe vartate katham: when that *word-meaning* is *excluded*, *how* does the *member* operate? If 'abrahmana' has excluded brahmin-hood, what is the 'brahmana' in it doing? Na animitta hi sabdasya pravrttir upapadyate: for a *word's operation* *without a cause* is *not possible*. Animitta: causeless. So the base cannot just sit there: a word operates only when something occasions it, and its occasion has just been denied. This is the sharpest form of the negative compound's puzzle, and 3.14.274-275 test the obvious answers.

3.14.274 unformalized

आराच्छब्दवदेकस्य विरुद्धेऽर्थे स्वभावतः ।
शब्दस्य वृत्तिर्यद्यस्ति नञः श्रुतिरनर्थिका ॥ २७४ ॥

ārācchabdavadekasya viruddhe'rthe svabhāvataḥ |
śabdasya vṛttiryadyasti nañaḥ śrutiranarthikā || 274 ||

If, like the word 'arat', a single word by its own nature operates in a contradictory meaning — then the hearing of 'nañ' is purposeless.

The first answer, and its cost. Arac-chabdavad ekasya viruddhe 'rthe svabhavatah sabdasya vrttir yady asti: *if*, like the word *'arat'* — which means both *near* and *far*, Sanskrit's most famous auto-antonym — a *single* word *by its own nature* operates in a *contradictory* meaning. Nanah srutir anarthika: then the *hearing of 'nañ'* is *purposeless*. So one might say that 'brahmana' natively means both brahmin and non-brahmin, as 'arat' means both near and far — and then the base needs no cause from outside (3.14.273 answered). But then nañ is doing nothing, since the base already had the sense. The answer buys the base's operation at the price of the negative.

3.14.275 unformalized

अथ स्वभावो वचनादन्वाख्येयत्वमर्हति ।
तद्वाच्यमप्रसिद्धत्वान्नञार्थो विनिवर्त्यते ॥ २७५ ॥

atha svabhāvo vacanādanvākhyeyatvamarhati |
tadvācyamaprasiddhatvānnañārtho vinivartyate || 275 ||

Now, the nature deserves to be expounded from the statement; that is to be expressed, from its not being well-known; the meaning is excluded by 'nañ'.

The second answer. Atha svabhavo vacanad anvakhyeyatvam arhati: *now*, the *nature* *deserves to be expounded* from the *statement* — anvakhyeya: to be expounded (3.14.74's anvakhyeyas tatha). The word's dual nature is not self-evident and must be taught. Tad vacyam aprasiddhatvat: *that* is *to be expressed*, from its *not being well-known* — aprasiddha: not established in usage, unlike 'arat', whose double sense everyone knows. Nana artho vinivartyate: the *meaning* is *excluded* *by 'nañ'*. So: 'brahmana' does not natively mean non-brahmin (which is why 3.14.274's parallel fails), and nañ is what excludes — it has work precisely because the base's contrary sense is not established. The negative earns its keep from the base's not being 'arat'.

3.14.276 unformalized

यद्यप्युभयवृत्तित्वं प्रधानं तु प्रतीयते ।
प्रस्थानं गम्यते शुद्धे तदर्थेऽपि न तिष्ठतौ ॥ २७६ ॥

yadyapyubhayavṛttitvaṃ pradhānaṃ tu pratīyate |
prasthānaṃ gamyate śuddhe tadarthe'pi na tiṣṭhatau || 276 ||

Although there is operation in both, the principal is cognized; setting-out is understood in the pure root, and even in that meaning, not in 'tisthati'.

A parallel from the preverbs. Yady apy ubhaya-vrttitvam pradhanam tu pratiyate: *although* there is *operation in both*, the *principal* is *cognized* — where a form could go two ways, one of them is what is understood. Prasthanam gamyate suddhe tad-arthe 'pi na tisthatau: *setting-out* is understood in the *pure* root, and *even in that meaning*, *not in 'tisthati'*. So 'pra-stha' means setting out, and the sense belongs to the compounded root — while 'tisthati', the same root uncompounded, means standing still and never means departure. The preverb's sense is not in the root and does not stay in it (3.14.39's niradinam krantady-arthesv asambhavah; 3.14.256's sa kramah). Nañ is the same: what it contributes is not recoverable from the base, and does not survive decomposition.

3.14.277 unformalized

किमर्थमतथाभूते सति मुख्यार्थसम्भवे ।
भेदे ब्राह्मणशब्दस्य वृत्तिरभ्युपगम्यते ॥ २७७ ॥

kimarthamatathābhūte sati mukhyārthasambhave |
bhede brāhmaṇaśabdasya vṛttirabhyupagamyate || 277 ||

For what purpose, when the primary meaning is possible in what is not such, is the operation of the word 'brahmana' accepted in difference?

Kim artham atathabhute sati mukhyartha-sambhave: *for what purpose*, when the *primary meaning* is *possible* in what is *not such* — the ksatriya is not a brahmin, and yet the word's primary sense somehow reaches him. Bhede brahmana-sabdasya vrttir abhyupagamyate: is the *operation* of the word *'brahmana'* *accepted* *in difference*? So the question is put bluntly: why let the word run into the very thing it is not? Kim artham: to what end? The samuddesa has been circling this and here demands the point of the whole construction. 3.14.278 answers: because that is how one is told anything.

3.14.278 unformalized

अयं पदार्थ एतस्मिन् क्षत्रियादौ न विद्यते ।
इति तद्वचनः शब्दः प्रत्ययाय प्रयुज्यते ॥ २७८ ॥

ayaṃ padārtha etasmin kṣatriyādau na vidyate |
iti tadvacanaḥ śabdaḥ pratyayāya prayujyate || 278 ||

'This word-meaning does not exist in this ksatriya and the like' — thus the word expressing that is employed, for the sake of cognition.

The answer, and it is the practical one. Ayam padartha etasmin ksatriyadau na vidyate: *'this word-meaning does not exist in this ksatriya'* and the like — the content of what one wants conveyed. Iti tad-vacanah sabdah pratyayaya prayujyate: *thus* the word *expressing that* is *employed*, *for the sake of cognition* — pratyayaya: so that one may know. So 'brahmana' runs into the ksatriya (3.14.277's question) because that is the only way to say that brahmin-hood is *not* there: one must bring the brahmin-hood into view to remove it (3.14.262's padam krtva, planting the foot in order to pull away; 3.14.263's sopakhyatvam, needing something definable to deny). The word goes where it does not belong because negation has no other route.

3.14.279 unformalized

बुद्धेर्विषयतां प्राप्ते शब्दादर्थे प्रतीयते ।
प्रवृत्तिर्वा निवृत्तिर्वा श्रुत्या ह्यर्थोऽनुषज्यते ॥ २७९ ॥

buddherviṣayatāṃ prāpte śabdādarthe pratīyate |
pravṛttirvā nivṛttirvā śrutyā hyartho'nuṣajyate || 279 ||

When the meaning, from the word, has attained the state of being the intellect's object — either operation or exclusion is cognized; for the meaning is attached by the hearing.

The general principle behind 3.14.278. Buddher visayatam prapte sabdad arthe pratiyate: when the *meaning*, *from the word*, has attained the state of being the *intellect's object* — the word first makes the meaning an object of thought; that is what a word does (3.14.197's sabdad arthah pratayante). Pravrttir va nivrttir va: *either* *operation* *or* *exclusion* is cognized — only then, and equally either way. Srutya hy artho 'nusajyate: for the *meaning* is *attached* by the *hearing* — anusajyate: is annexed, comes along with. So affirmation and denial are alike downstream of the word's presenting: one cannot deny what one has not first been given, and the giving is the hearing. This is the whole doctrine of negation in one line, and it is the doctrine of 1.123 (no cognition without word) applied to absence.

3.14.280 unformalized

असम्यगुपदेशाद् वा निमित्तात् संशयस्य वा ।
शब्दप्रवृत्तिर्न त्वस्ति लोष्टादिषु विपर्ययम् ॥ २८० ॥

asamyagupadeśād vā nimittāt saṃśayasya vā |
śabdapravṛttirna tvasti loṣṭādiṣu viparyayam || 280 ||

Or from incorrect teaching, or from a cause of doubt — a word does operate; but there is no reversal in the case of 'clod' and the rest.

The limit of 3.14.279's principle. Asamyag-upadesad va nimittat samsayasya va sabda-pravrttih: from *incorrect teaching*, or from a *cause of doubt*, a *word* does *operate* — words are applied wrongly, out of bad instruction or genuine uncertainty, and the misapplication is real. Na tv asti lostadisu viparyayam: *but* there is *no reversal* in *'losta'* (a clod of earth) and the rest. So the errors have a range: one may call a ksatriya a brahmin by ignorance or doubt, and nobody calls a clod one. The word's straying is bounded by something — the resemblance that made the doubt possible. This is 3.14.214's rupa-bheda: where the form differs enough, the word turns it away, and no mistake can reach it.

3.14.281 unformalized

अनेकस्मादस इति प्राधान्ये सति सिध्यति ।
सापेक्षत्वं प्रधानानामेवं युक्तं त्वतल्विधौ ॥ २८१ ॥

anekasmādasa iti prādhānye sati sidhyati |
sāpekṣatvaṃ pradhānānāmevaṃ yuktaṃ tvatalvidhau || 281 ||

From the many, 'as' — it succeeds where there is principality; the dependence of principals is thus proper in the 'atal'-rule.

A rule-technical case, and its reading is not easy. Anekasmad asa iti pradhanye sati sidhyati: from the *many*, *'as'* — it *succeeds* where there is *principality*: the form comes out when the relevant element is principal, and not otherwise. Sapeksatvam pradhananam evam yuktam tv atal-vidhau: the *dependence* of *principals* is *thus proper* in the *'atal'*-rule. Sapeksatva: mutual dependence — and the point is that even principals may depend, which the atal-rule requires. So the samuddesa notes that principality does not entail independence: a principal can still be sapeksa (3.14.47's relatum-words, whose expectancy is their meaning). The nañ-compound will need this, since its base is principal and dependent at once.
contested: Padas a-b are highly compressed and we render them with reserve. 'Anekasmad asa iti' appears to cite a rule-fragment; without the rule's identification the construction of 'asa' (whether the root 'as', a form of it, or a pratyaya) cannot be fixed. The verse's usable content — that principals may nonetheless be sapeksa, and that the atal-rule requires this — rests on padas c-d, which are clear.

3.14.282 unformalized

एकस्य च प्रधानत्वात् तद्विशेषणसन्निधौ ।
प्रधानधर्माप्यावृत्तिरतो न वचनान्तरम् ॥ २८२ ॥

ekasya ca pradhānatvāt tadviśeṣaṇasannidhau |
pradhānadharmāpyāvṛttirato na vacanāntaram || 282 ||

And from one's being principal, in the presence of its qualifier — the recurrence even of the principal's property; therefore there is no other statement.

Ekasya ca pradhanatvat tad-visesana-sannidhau: from *one's* being *principal*, in the *presence* of *its qualifier* — sannidhi: proximity, the qualifier standing by. Pradhana-dharmapy avrttih: the *recurrence* *even* of the *principal's property* — avrtti: coming round again, repetition. Ato na vacanantaram: *therefore* there is *no other statement* — no further rule is needed. So when a principal has its qualifier alongside, the principal's own property recurs, and the grammar need not say so separately. This is rule-economy of the kind 3.14.106 found (a rule made idle by a doctrine), and it prepares the number-analysis of 3.14.283-286: the principal's property recurring is exactly what will happen to number in 'advi'.

3.14.283 unformalized

प्रधानमत्र भेद्यत्वादेकार्थो विकृतो नञा ।
हित्वा स्वधर्मान् वर्तन्ते द्व्यादयोऽप्येकतां गताः ॥ २८३ ॥

pradhānamatra bhedyatvādekārtho vikṛto nañā |
hitvā svadharmān vartante dvyādayo'pyekatāṃ gatāḥ || 283 ||

Here the principal, from being what is differentiated, is the one-meaning modified by 'nañ'; and 'dvi' and the rest too, having gone to oneness, operate having abandoned their own properties.

Pradhanam atra bhedyatvad ekartho vikrto nana: *here* the *principal*, from being *what is differentiated* — bhedya: the thing to be cut into, not the cutter — is the *one-meaning* *modified* by *'nañ'*. Ekartha: 3.14.44's ekarthibhava, becoming-one-meaning; vikrta: altered. So the negative compound is one meaning that nañ has modified, not two meanings in relation. Hitva sva-dharman vartante dvyadayo 'py ekatam gatah: and *'dvi'* and the rest too, having *gone to oneness*, operate having *abandoned their own properties*. So the numerals in a nañ-compound drop their numbers — 'advi' has no two in it — because they have gone into the one-meaning (3.14.101's bhedanam parityagat; 3.14.115's the numeral's outer difference withdrawing).

3.14.284 unformalized

ब्राह्मणत्वं यथापन्ना नञ्युक्ताः क्षत्रियादयः ।
द्वित्वादिषु तथैकत्वं नञ्योगादुपचर्यते ॥ २८४ ॥

brāhmaṇatvaṃ yathāpannā nañyuktāḥ kṣatriyādayaḥ |
dvitvādiṣu tathaikatvaṃ nañyogādupacaryate || 284 ||

As 'ksatriya' and the rest, joined with 'nañ', have attained brahmana-hood — so in 'dvitva' and the rest, oneness is figuratively applied, from the connection with 'nañ'.

The parallel that organizes the numeral-case. Brahmanatvam yatha apanna nan-yuktah ksatriyadayah: as *'ksatriya'* and the rest, *joined with 'nañ'*, have *attained brahmana-hood* — the striking formulation: in 'abrahmana' the ksatriya has *acquired* brahmin-hood, figuratively, in order to have it denied (3.14.271's upacara-sat; 3.14.278's the word must reach him to remove it). Dvitvadisu tatha ekatvam nan-yogad upacaryate: *so* in *'dvitva'* and the rest, *oneness* is *figuratively applied* from the *connection with 'nañ'* — upacaryate: is transferred. So the negated numeral acquires a figurative oneness exactly as the negated ksatriya acquires a figurative brahmin-hood. Nañ confers before it removes, both times.

3.14.285 unformalized

एकत्वयोगमासज्य स धर्मः प्रतिषिध्यते ।
द्व्यादिभ्यस्तेषु तच्छब्दो वर्तते ब्राह्मणादिवत् ॥ २८५ ॥

ekatvayogamāsajya sa dharmaḥ pratiṣidhyate |
dvyādibhyasteṣu tacchabdo vartate brāhmaṇādivat || 285 ||

Having attached the connection with oneness, that property is negated; from 'dvi' and the rest, in those the word 'tad' operates, as with 'brahmana' and the rest.

Ekatva-yogam asajya sa dharmah pratisidhyate: having *attached* the *connection with oneness*, *that property* is *negated* — asajya: having fastened it on. So the sequence is explicit: attach, then deny. The oneness is put there in order to be taken away, which is what 3.14.284's upacara amounted to. Dvyadibhyas tesu tac-chabdo vartate brahmanadivat: from *'dvi'* and the rest, in *those* the word *'tad'* operates, as with *'brahmana'* and the rest — the pronoun follows, as it did at 3.14.217, because the substance is not expressed. So the numeral-negation behaves in every respect like the caste-negation, down to the pronoun. The parallel is complete.

3.14.286 unformalized

आविष्टसङ्ख्यो वाक्येऽसौ यथा द्वयादौ प्रयुज्यते ।
वृत्तौ तस्य प्रधानत्वात् सा सङ्ख्या न निवर्तते ॥ २८६ ॥

āviṣṭasaṅkhyo vākye'sau yathā dvayādau prayujyate |
vṛttau tasya pradhānatvāt sā saṅkhyā na nivartate || 286 ||

As, in a sentence, that which is pervaded by number is employed in 'dvi' and the rest — in the derivative formation, from its being principal, that number does not withdraw.

The stretch's conclusion, and it inverts 3.14.283. Avista-sankhyo vakye 'sau yatha dvyadau prayujyate: as, *in a sentence*, that *pervaded by number* — avista: entered into (3.14.243's word) — is employed in *'dvi'* and the rest. Vrttau tasya pradhanatvat sa sankhya na nivartate: in the *derivative formation*, from *its being principal*, that *number* does *not withdraw*. So where the numeral is *principal*, its number survives the composition — against 3.14.283, where the numerals abandoned their properties on going into the one-meaning. The two verses are not in conflict: they describe the numeral subordinate and the numeral principal, and 3.14.282 has already established that the principal's property recurs. What a compound does to its members depends on which of them is principal — which is where this samuddesa began (3.14.95's upasarjana).

3.14.287 unformalized

प्रतिषेध्यो यथाभूतस्तथाभूतोऽनुषज्यते ।
वचनान्तरयोगे हि न सोऽर्थः प्रतिषिध्यते ॥ २८७ ॥

pratiṣedhyo yathābhūtastathābhūto'nuṣajyate |
vacanāntarayoge hi na so'rthaḥ pratiṣidhyate || 287 ||

The thing to be negated is annexed exactly as it is; for in connection with another number, that meaning is not negated.

Pratisedhyo yathabhutas tathabhuto 'nusajyate: the *thing to be negated* is *annexed* *exactly as it is* — yathabhutas tathabhutah: however it is, so it comes along (3.14.279's srutya hy artho 'nusajyate, the same verb). One cannot deny a thing under a description it does not have. Vacanantara-yoge hi na so 'rthah pratisidhyate: for in connection with *another number*, that *meaning* is *not negated* — change the number and the denial misses. So negation is exact: it strikes the thing as presented and nothing else, which is why 3.14.286's number mattered. Deny 'two' and you have not denied 'twos'.

3.14.288 unformalized

अशुक्ल इति कृष्णादिर्यथार्थः सम्प्रतीयते ।
सङ्ख्यान्तरं तथानेक इत्यत्राप्यभिधीयते ॥ २८८ ॥

aśukla iti kṛṣṇādiryathārthaḥ sampratīyate |
saṅkhyāntaraṃ tathāneka ityatrāpyabhidhīyate || 288 ||

As by 'non-white' the meaning 'black' and the like is cognized — so in 'not-one' too, another number is expressed.

The paryudasa principle, in two examples. Asukla iti krsnadir yatha arthah sampratiyate: as by *'asukla'* (non-white) the meaning *'black'* and the like is *cognized* — one does not hear 'not white' and think of virtue or Tuesday: one thinks of another colour. The negation lands inside the genus. Sankhyantaram tatha aneka ity atrapi abhidhiyate: *so* in *'aneka'* (not-one) too, *another number* is *expressed*. So 'aneka' means some other number, not the absence of number. This is paryudasa — the negation that selects a sibling rather than annihilating — and Bhartrhari establishes it by pointing at what one actually understands (3.14.260's asata api tadatmyam vyajyate: the negative manifests an identity).

3.14.289 unformalized

क्रियाप्रसङ्गात् सर्वेषु कर्मस्वङ्गीकृतेषु च ।
एकस्मिन् प्रतिषिद्धेऽपि प्राप्तमन्यत् प्रतीयते ॥ २८९ ॥

kriyāprasaṅgāt sarveṣu karmasvaṅgīkṛteṣu ca |
ekasmin pratiṣiddhe'pi prāptamanyat pratīyate || 289 ||

And from the action's applicability, all the objects being accepted — even when one is negated, another, being applicable, is cognized.

Why paryudasa works. Kriya-prasangat sarvesu karmasv angikrtesu ca: from the *action's applicability*, *all the objects* being *accepted* — the verb has already brought the whole field of its possible objects into view (3.14.291 will spell this out). Ekasmin pratisiddhe 'pi praptam anyat pratiyate: *even when one* is *negated*, *another*, being *applicable*, is *cognized*. So the reason 'asukla' delivers black is that the colour-field was already open when white was denied: deny one member of an open field and the field's remainder presents itself. Paryudasa is not a special kind of negation; it is ordinary negation performed inside a field that the verb had already spread out.

3.14.290 unformalized

क्रिया श्रुतिश्च प्रक्रान्ते प्रसज्यप्रतिषेधेन ।
पर्युदासे तु नियतं सङ्ख्येयान्तरमुच्यते ॥ २९० ॥

kriyā śrutiśca prakrānte prasajyapratiṣedhena |
paryudāse tu niyataṃ saṅkhyeyāntaramucyate || 290 ||

The action and what is heard are within what has been commenced, by prasajya-negation; but in paryudasa, a determinate other countable is stated.

The two negations distinguished, and this is the tradition's most consequential distinction. Kriya srutis ca prakrante prasajya-pratisedhena: the *action* and *what is heard* are within what has been *commenced*, by *prasajya*-negation — prasajya-pratisedha: the negation that denies the predication itself ('he does not cook'), and it stays inside what was begun: the verb is what is denied. Paryudase tu niyatam sankhyeyantaram ucyate: *but* in *paryudasa*, a *determinate other* countable is *stated* — niyatam: fixed. So prasajya denies the act; paryudasa asserts a sibling. The first says no, the second says another. And Bhartrhari's contribution here is 3.14.289's ground: paryudasa can name a sibling only because the action had already laid out the family.

3.14.291 unformalized

धात्वर्थः कर्मविषयो व्यपदिष्टः स्वसाधनैः ।
अर्थात् सर्वाणि कर्माणि प्रागाक्षिप्यावतिष्ठते ॥ २९१ ॥

dhātvarthaḥ karmaviṣayo vyapadiṣṭaḥ svasādhanaiḥ |
arthāt sarvāṇi karmāṇi prāgākṣipyāvatiṣṭhate || 291 ||

The root-meaning, having objects for its domain, designated by its own means — by implication, having first implied all objects, stands.

The ground of 3.14.289, stated in full. Dhatv-arthah karma-visayo vyapadistah sva-sadhanaih: the *root-meaning*, having *objects for its domain*, *designated by its own means* — the verb comes with its karakas (3.7.1's the action requires its sadhanas). Arthat sarvani karmani prag aksipya avatisthate: *by implication*, having *first implied* *all objects*, it *stands*. Aksipya: having drawn in, implicated; prak: beforehand. So the verb, before any object is specified, has already implicated *all* its possible objects and stands over them. That standing field is what a paryudasa-negation carves. This is a fine piece of semantics: the verb's transitivity is a prior claim on a whole domain, and particular objects are selections from it.

3.14.292-293 unformalized

निर्ज्ञातसाधनाधारे यत्राख्याते प्रयुज्यते ।
अनेक इति पश्चाच्च तिष्ठतीत्यनुषज्यते ॥ २९२ ॥
साध्यत्वात् तत्र सिद्धेन क्रिया द्रव्येण लक्ष्यते ।
प्रागेवाङ्गीकृतं द्रव्यमतः पूर्वेण भिद्यते ॥ २९३ ॥

nirjñātasādhanādhāre yatrākhyāte prayujyate |
aneka iti paścācca tiṣṭhatītyanuṣajyate || 292 ||
sādhyatvāt tatra siddhena kriyā dravyeṇa lakṣyate |
prāgevāṅgīkṛtaṃ dravyamataḥ pūrveṇa bhidyate || 293 ||

(292) Where 'aneka' is employed in a verb whose means and substratum are known, and afterwards 'tisthati' is annexed — (293) there, since the action is to be accomplished, it is indicated by an accomplished substance; the substance was accepted already beforehand, and therefore differs from the earlier one.

(292): nirjnata-sadhana-adhare yatra akhyate prayujyate aneka iti — where *'aneka'* is employed in a *verb* whose *means and substratum* are *known*. Pascac ca tisthatity anusajyate: and *afterwards* *'tisthati'* is *annexed* — 'aneke tisthanti', not-one stand. (293): sadhyatvat tatra siddhena kriya dravyena laksyate — since the *action* is *to be accomplished* (sadhya: not yet done), it is *indicated* by an *accomplished* substance (siddha: already there). The standing contrast of this work (3.8's sadhya-kriya against the siddha karakas; 3.1's the accomplished indicating the accomplishable). Prag eva angikrtam dravyam atah purvena bhidyate: the *substance* was *accepted already beforehand*, and *therefore differs from the earlier* one. So the substance that indicates the action is not the one the verb had implicated at 3.14.291: it was accepted earlier and is a different thing. The negation's 'other' is thus genuinely other.

3.14.294 unformalized

सङ्ख्यैव प्रतिषेधेन सङ्ख्यान्तरमपेक्षते ।
वाक्येऽपि तेन नैकत्वमात्रमेव निवर्त्यते ॥ २९४ ॥

saṅkhyaiva pratiṣedhena saṅkhyāntaramapekṣate |
vākye'pi tena naikatvamātrameva nivartyate || 294 ||

Number itself, by the negation, requires another number; therefore even in a sentence, not merely oneness is excluded.

Sankhya eva pratisedhena sankhyantaram apeksate: *number itself*, by the *negation*, *requires* *another number* — apeksate: needs, reaches for. The negation of a number is not satisfied by absence; it demands a substitute from the same family (3.14.288's asukla giving black). Vakye 'pi tena na ekatva-matram eva nivartyate: *therefore* even *in a sentence*, *not merely oneness* is *excluded*. So 'na ekah' does not simply remove the one and leave a blank: it puts another number there. And the point of 'vakye 'pi' — *even in a sentence* — is that this is not a compound-effect: paryudasa is how negation works in the language at large, and the compound only makes it visible.

3.14.295 unformalized

स्नेहान्तरादवच्छेदा यस्तथासत्तेः प्रतीयते ।
तैलेन भोजने प्राप्ते न त्वन्यदुपसेचनम् ॥ २९५ ॥

snehāntarādavacchedā yastathāsatteḥ pratīyate |
tailena bhojane prāpte na tvanyadupasecanam || 295 ||

That which is cognized from the delimitation by another oil, from such proximity — when eating with oil is applicable, there is no other seasoning.

Paryudasa's reach measured, and it is short. Sneha-antarad avacchedad yas tatha asatteh pratiyate: what is cognized from the *delimitation* by *another oil*, from such *proximity* — asatti: nearness, the adjacency that decides which sibling the negation lands on. Tailena bhojane prapte na tv anyad upasecanam: when *eating with oil* is *applicable*, there is *no other seasoning* — deny the sesame oil and one gets a different *oil*, not ghee, not salt. So the negation does not roam its genus freely; it takes the nearest neighbour, and asatti fixes which. This is the exact converse of 3.14.280's lostadisu viparyayam (the word cannot stray as far as a clod): both verses bound the drift, one for error and one for negation, and both bound it by nearness.

3.14.296 unformalized

एकार्थे वर्तमानाभ्यामसता ब्राह्मणेन च ।
यदा जात्यन्तरं बाह्यं क्षत्रियाद्यपदिश्यते ॥ २९६ ॥

ekārthe vartamānābhyāmasatā brāhmaṇena ca |
yadā jātyantaraṃ bāhyaṃ kṣatriyādyapadiśyate || 296 ||

When, by 'asat' and 'brahmana', both operating in one meaning, an external other class — ksatriya and the rest — is designated.

Ekarthe vartamanabhyam asata brahmanena ca: by *'asat'* and *'brahmana'*, both *operating in one meaning* — ekarthe: in one meaning, the co-reference of 3.14.257's kubja-khanja. Yada jaty-antaram bahyam ksatriyady apadisyate: when an *external other class* — *ksatriya* and the rest — is *designated*. Bahyam: *external*: the class designated is outside both words' own senses. So two co-referring words, neither of which means ksatriya, together designate a ksatriya. The mechanism is not composition but co-reference plus paryudasa: they meet in one meaning and the meaning they meet in is the neighbour (3.14.295's asatti).

3.14.297 unformalized

श्यामेव शस्त्री कन्येति यथान्यद् व्यपदिश्यते ।
असन् ब्राह्मण इत्याभ्यां तथान्यं क्षत्रियादयः ॥ २९७ ॥

śyāmeva śastrī kanyeti yathānyad vyapadiśyate |
asan brāhmaṇa ityābhyāṃ tathānyaṃ kṣatriyādayaḥ || 297 ||

As by 'the dark one', 'the knife-bearing one', 'the girl', something else is designated — so by 'asat' and 'brahmana', ksatriya and the rest designate something else.

The parallel for 3.14.296, and it is well chosen. Syama eva sastri kanya iti yatha anyad vyapadisyate: as by *'syama'* (dark), *'sastri'* (knife-bearing), *'kanya'* (girl), *something else* is *designated* — three words, none of which means this particular girl, converging on her. Each contributes a delimitation; the referent is what survives all three (3.14.9-11's two open generals drawing away each other's doubt). Asan brahmana ity abhyam tatha anyam ksatriyadayah: *so* by *'asat'* and *'brahmana'*, *ksatriya* and the rest designate *something else*. So the negative compound is a convergence of delimitations, exactly like a string of adjectives — and the negative is one adjective among them. This is 3.14.260's finding (nañ delimits as 'gaura' does) given its clearest illustration.

3.14.298 unformalized

असास्नो गौरिति यथा गवयो व्यपदिश्यते ।
जात्यन्तरं न गोरेव सास्नाऽभावः प्रतीयते ॥ २९८ ॥

asāsno gauriti yathā gavayo vyapadiśyate |
jātyantaraṃ na goreva sāsnā'bhāvaḥ pratīyate || 298 ||

As by 'a dewlapless cow' the gavaya is designated — another class; the absence of a dewlap in a cow itself is not cognized.

The example that clinches paryudasa. Asasno gaur iti yatha gavayo vyapadisyate jaty-antaram: as by *'asasnah gauh'* — a *cow without a dewlap* — the *gavaya* is *designated*, *another class*: the gavaya (the nilgai, the wild ox) is what one thinks of, and it is a different species. Na gor eva sasna-abhavah pratiyate: the *absence of a dewlap* *in a cow itself* is *not cognized* — one does not picture a mutilated cow. So the negation of a defining property does not produce a defective member of the class: it produces the neighbouring class. This is the strongest possible case for paryudasa, because prasajya would give exactly the wrong result, and everyone's understanding refuses it.

3.14.299 unformalized

तुल्यरूपं यथाख्यातं कण्टकै(कङ्कटै)र्भेदहेतुभिः ।
खदिरं जातिभेदेन खर्जूरात्(बर्बुरात्) प्रतिपद्यते ॥ २९९ ॥

tulyarūpaṃ yathākhyātaṃ kaṇṭakai(kaṅkaṭai)rbhedahetubhiḥ |
khadiraṃ jātibhedena kharjūrāt(barburāt) pratipadyate || 299 ||

As what has been declared to be of similar form is apprehended, by thorns as the difference-causes, as khadira, distinct in class from kharjura.

Tulya-rupam yatha akhyatam kantakair bheda-hetubhih: as what has been *declared* to be of *similar form* — two trees that look alike — by *thorns* as the *difference-causes*. Khadiram jati-bhedena kharjurat pratipadyate: one *apprehends* *khadira* (the acacia) as *distinct in class* from *kharjura* (the date-palm). So a single differentiating mark sorts two similar things into two classes, and one apprehends the class, not the mark. This is 3.14.298's gavaya again in a botanical key: the differentia does not modify the genus, it delivers a different species. The thorns are 3.14.295's asatti made visible — the nearest neighbour is the one the mark separates you from.
contested: The source records variants at both padas b and d: 'kantakaih' ('thorns') against 'kankataih', and 'kharjurat' against 'barburat'. We translate the first of each. 'Barbura' (a thorny acacia) would in fact suit an argument from thorns better than the date-palm does, and may be the sounder reading; we flag it rather than choose against the primary text.

3.14.300 unformalized

अविद्यमानब्राह्मण्यो यादृशो ब्राह्मणो भवेत् ।
अङ्गीकृतोपमानेन तथान्यार्थोभिऽधीयते ॥ ३०० ॥

avidyamānabrāhmaṇyo yādṛśo brāhmaṇo bhavet |
aṅgīkṛtopamānena tathānyārthobhi'dhīyate || 300 ||

As a brahmana in whom brahmana-hood is absent would be — so, by an accepted comparison, another meaning is expressed.

Avidyamana-brahmanyo yadrso brahmano bhavet: as a *brahmana* in whom *brahmana-hood is absent* would be — the contradictory figure, and everyone knows whom it means: the man who is one by birth and not by conduct. Angikrta-upamanena tatha anyartho 'bhidhiyate: so, by an *accepted comparison*, *another meaning* is *expressed*. Angikrta-upamana: a comparison that has been *granted*, taken as established. So the negative compound works as a conceded simile does: 'abrahmana' says *the sort of thing a brahmin without brahmin-hood would be*, and the hearer supplies the rest. The negation is a figure, and figures work by being accepted (3.14.271's upacara-sat: existing by transfer).

3.14.301 unformalized

अवृष्टयो यथावर्षा नीहाराभ्रसमावृताः ।
तद्रूपत्वात् स हेमन्त इत्यभिन्नः प्रतीयते ॥ ३०१ ॥

avṛṣṭayo yathāvarṣā nīhārābhrasamāvṛtāḥ |
tadrūpatvāt sa hemanta ityabhinnaḥ pratīyate || 301 ||

As rainless days, though not raining, covered with mist and cloud — from having that form, 'that is winter' — it is cognized as non-different.

A lovely example, and it makes the point better than argument. Avrstayo yatha avarsa nihara-abhra-samavrtah: as *rainless* days, *not raining*, *covered with mist and cloud* — days in the cold season that look exactly like the rains: overcast, dim, wet-seeming, and dry. Tad-rupatvat sa hemanta ity abhinnah pratiyate: *from having that form*, *'that is winter'* — it is cognized as *non-different*. So one says of such a day that it is winter *and* one perceives it as the rains, because the form is the same. Tad-rupatvat: from having that form (3.14.214's rupabhedat). The negation is exactly this: the form is there, the thing is not, and the cognition takes what it sees. 'Abrahmana' is an overcast day in the cold season.

3.14.302 unformalized

अपरे ब्राह्मणादीनां सर्वेषां जातिवाचिनाम् ।
द्रव्यस्यान्यपदार्थत्वे नञा योगं प्रचक्षते ॥ ३०२ ॥

apare brāhmaṇādīnāṃ sarveṣāṃ jātivācinām |
dravyasyānyapadārthatve nañā yogaṃ pracakṣate || 302 ||

Others declare the connection with 'nañ' of all the universal-expressing words like 'brahmana' — where the substance is another's word-meaning.

A further position recorded. Apare brahmanadinam sarvesam jati-vacinam: *others*, of *all* the *universal-expressing* words like *'brahmana'*. Dravyasya anya-padarthatve nana yogam pracaksate: *declare* the *connection with 'nañ'* — where the *substance* is *another's word-meaning*: i.e. they treat every nañ-compound on jati-words as a bahuvrihi, whose substance lies outside it (3.14.38's anya-padartha). So 'abrahmana' would be, structurally, 'one for whom brahmin-hood is absent' — a possessive compound, not a determinative. This is a real proposal and it explains 3.14.300's 'the sort of thing a brahmin without brahmin-hood would be' rather neatly. Bhartrhari records it as apare and, as usual, lets the reader weigh it (3.14.97: the upayas are unsettled among persons).

3.14.303 unformalized

न चैवंविषयः कश्चिद् बहुर्वीहिः प्रकल्पते ।
अगुरश्व इति व्याप्तिर्नञ्समासेन यस्य न ॥ ३०३ ॥

na caivaṃviṣayaḥ kaścid bahurvīhiḥ prakalpate |
aguraśva iti vyāptirnañsamāsena yasya na || 303 ||

And no bahuvrihi of such a domain is conceivable; 'a horse is cowless' — of which there is no pervasion by the negative compound.

The refutation of 3.14.302. Na ca evam-visayah kascid bahuvrihih prakalpate: *no* bahuvrihi of *such a domain* is *conceivable*. Agur asva iti vyaptir nan-samasena yasya na: *'agur asvah'* — *a horse is cowless* — of which there is *no pervasion* by the *nan-compound*. Vyapti: pervasion, the co-extension a real bahuvrihi would give. If 'abrahmana' were possessive, then 'agu' would cover every cowless thing — horses, stones, gods — and it does not: 'agu' means a man without cattle, not a horse. So the possessive reading over-generates catastrophically, and the test is one word long. The bahuvrihi-account dies where 3.14.298's gavaya lives: paryudasa picks the neighbour, and possession picks the world.

3.14.304 unformalized

द्वन्द्वैकदेशिनोरुक्ता परवल्लिङ्गता यतः ।
अवर्षासु ततोऽसिद्धिरिष्टयोर्लिङ्गसंख्ययो ॥ ३०४ ॥

dvandvaikadeśinoruktā paravalliṅgatā yataḥ |
avarṣāsu tato'siddhiriṣṭayorliṅgasaṃkhyayo || 304 ||

Since for the dvandva and for the ekadesin the gender of the latter member is stated — therefore in 'avarsah' the desired gender and number are not established.

A second failure. Dvandva-ekadesinor ukta paravat-lingata yatah: for the *dvandva* and the *ekadesin* the *gender of the latter* is *stated* — Panini's paravat-lingata: the compound takes its final member's gender. Avarsasu tato 'siddhir istayor linga-samkhyayoh: *therefore*, in *'avarsah'* — the non-rains, 3.14.301's overcast winter days — the *desired* gender and number are *not established*. So the rule would give 'avarsa' the gender of 'varsa', and what one wants is 'hemanta', a masculine singular. The forms do not come out. Between this and 3.14.303 the alternative accounts are exhausted, and the samuddesa returns to its own.

3.14.305 unformalized

विशेषणं ब्राह्मणादि क्रियासंबन्धिनोऽसतः ।
यदा विषयभिन्नं तत् तदाऽसत्त्व प्रतीयते ॥ ३०५ ॥

viśeṣaṇaṃ brāhmaṇādi kriyāsaṃbandhino'sataḥ |
yadā viṣayabhinnaṃ tat tadā'sattva pratīyate || 305 ||

When 'brahmana' and the rest, the qualifier of the 'asat' connected with the action, is differentiated by its domain — then non-being is cognized.

Visesanam brahmanadi kriya-sambandhino 'satah: *'brahmana'* and the rest, the *qualifier* of the *'asat'* *connected with the action* — the arrangement 3.14.255 proposed: nañ settles in a general seat, and the base specifies it. Yada visaya-bhinnam tat tada asattva pratiyate: when *that* is *differentiated by its domain*, *then* *non-being* is *cognized*. So non-being is not given and then qualified; it is *produced* by the qualifier's cutting the domain. 'Asat' alone conveys nothing determinate — it is 3.14.261's nirupakhya — and 'brahmana' is what makes it a determinate absence. The negative needs its base more than the base needs it.

3.14.306 unformalized

ब्राह्मणत्वेन चासत्त्वादुच्यतेऽसत्तदन्यथा ।
असदित्यपि सत्त्वेन सतः सत्ता निवर्त्यते ॥ ३०६ ॥

brāhmaṇatvena cāsattvāducyate'sattadanyathā |
asadityapi sattvena sataḥ sattā nivartyate || 306 ||

And from non-being as a brahmana, it is called 'asat' — otherwise; and by 'asat' too, the existence of the existent is excluded by way of being.

Brahmanatvena ca asattvad ucyate 'sat tad anyatha: from *non-being* *as a brahmana*, it is *called 'asat'* — *otherwise*: the ksatriya is 'asat' only under a description, and in himself he is perfectly sat. Anyatha: in another way, i.e. he is otherwise than the word says. Asad ity api sattvena satah satta nivartyate: and by *'asat'* too, the *existence* of the *existent* is *excluded* *by way of being* — sattvena: through being. So even the flattest negation works through a positive: to exclude a thing's existence one must have it *as being* first, and the exclusion runs along the being. This is 3.14.263's sopakhyatva (one needs something definable to deny) and 3.14.279's the meaning is annexed by the hearing — nothing is denied except what has first been given.

3.14.307 unformalized

सामान्यद्रव्यवृत्तित्वान्निमित्तानुविधायिनः ।
अयोगो लिङ्गसंख्याभ्यां स्याद् वा सामान्यधर्मता ॥ ३०७ ॥

sāmānyadravyavṛttitvānnimittānuvidhāyinaḥ |
ayogo liṅgasaṃkhyābhyāṃ syād vā sāmānyadharmatā || 307 ||

From its operating in the general substance, of what conforms to its cause — there would be no connection with gender and number; or there would be the property of the general.

Samanya-dravya-vrttitvat nimitta-anuvidhayinah: from its operating in the *general substance*, of what *conforms to its cause* — dravya-samanya, 3.14.246's substance-in-general. Ayogo linga-samkhyabhyam syad va samanya-dharmata: there would be *no connection* with *gender and number*; *or* there would be the *property of the general*. So two unwelcome results: either the compound gets no concord at all, or it gets the general's own gender and number — which, since a general has none (3.14.107's jatau dvivacanabhavat), is the first result under another name. The difficulty is real and 3.14.310 will meet it by letting the specified substrata supply what the general cannot.

3.14.308 unformalized

प्रागसत्त्वाभिधायित्वं समासे द्रव्यवाचिता ।
निमित्तानुविधानं च न सर्वत्र स्वभावतः ॥ ३०८ ॥

prāgasattvābhidhāyitvaṃ samāse dravyavācitā |
nimittānuvidhānaṃ ca na sarvatra svabhāvataḥ || 308 ||

Before, the expressing of non-substance; in the compound, the expressing of substance; and conformity to the cause is not everywhere, by nature.

Prag asattva-abhidhayitvam: *before* — uncompounded — the expressing of *non-substance*: 'nañ' by itself says no thing (3.14.193's asattva-bhuta, the particle's mode of meaning). Samase dravya-vacita: *in the compound*, the *expressing of substance* — compounded, it says a thing. So nañ changes category by being compounded, exactly as 'nir' does (3.14.256's sa kramah). Nimitta-anuvidhanam ca na sarvatra svabhavatah: and *conformity to the cause* is *not everywhere*, *by nature* — the premise of 3.14.307's difficulty does not hold universally. So the difficulty is answered by denying its generality: things do not always follow their causes, and this is one of the places they do not.

3.14.309 unformalized

निमित्तानुविधाने च क्रियायोगो न कल्पते ।
तथाचाव्यपदेश्यत्वादुपादानमनर्थकम् ॥ ३०९ ॥

nimittānuvidhāne ca kriyāyogo na kalpate |
tathācāvyapadeśyatvādupādānamanarthakam || 309 ||

And in conformity to the cause, connection with action is not conceivable; and thus, from being indesignable, the taking-up is purposeless.

The decisive argument against 3.14.307's premise, and it is the samuddesa's standing one. Nimitta-anuvidhane ca kriya-yogo na kalpate: in *conformity to the cause*, *connection with action* is *not conceivable* (3.14.204, in the same words). Tatha ca avyapadesyatvad upadanam anarthakam: and thus, from being *indesignable* — avyapadesya: not even nameable (3.14.135's vyapadisyate, negated) — the *taking-up* is *purposeless*. So if the compound merely followed its cause, it could not be predicated of, and a thing that cannot be predicated of does not independently exist (3.14.205's yasya nasti kriya-yogah svatantro 'sau na vidyate). Not merely unusable: unnameable. Anarthaka closes it.

3.14.310 unformalized

असत्सामान्यवृत्तिर्वा विशेषैः क्षत्रियादिभिः ।
प्रयुक्तैराश्रयैर्भिन्नो याति तल्लिङ्गसङ्ख्यताम् ॥ ३१० ॥

asatsāmānyavṛttirvā viśeṣaiḥ kṣatriyādibhiḥ |
prayuktairāśrayairbhinno yāti talliṅgasaṅkhyatām || 310 ||

Or 'asat', operating in the general, differentiated by the particulars — 'ksatriya' and the rest, the substrata employed — goes to their gender and number.

The resolution. Asat-samanya-vrttir va: *'asat'*, operating in the *general*. Visesaih ksatriyadibhih prayuktair asrayair bhinnah: *differentiated* by the *particulars* — *'ksatriya'* and the rest, the *substrata* *employed*. Yati tal-linga-sankhyatam: it *goes to* *their* gender and number. So the general has no concord of its own (3.14.307's worry, granted) and takes the specified substrata's — which are employed, present, and gendered. Yati: it *goes to* them. This is 3.14.152's borrowing once more, and it is the samuddesa's uniform answer wherever concord is wanted and no substance is at hand: find the substance the word reaches, and take its gender.

3.14.311 unformalized

प्रागाश्रयो हि भेदाय प्रधानेऽभ्यन्तरीकृतः ।
पुनः प्रत्यवमर्शेन विभक्त इव दृश्यते ॥ ३११ ॥

prāgāśrayo hi bhedāya pradhāne'bhyantarīkṛtaḥ |
punaḥ pratyavamarśena vibhakta iva dṛśyate || 311 ||

For the substratum, previously internalized into the principal for the sake of difference, is again seen as though divided, by the reflective reference.

One of the most exact psychological sentences in the samuddesa. Prag asrayo hi bhedaya pradhane 'bhyantarikrtah: the *substratum*, *previously* *internalized* into the *principal* *for the sake of difference* — abhyantarikrta: made inner, taken inside. The seat was swallowed by the principal, and swallowed *in order to* differentiate it. Punah pratyavamarsena vibhakta iva drsyate: *again*, by the *reflective reference*, it is *seen* *as though divided* — pratyavamarsa (3.14.238's word): the mind turning back on what it laid down; iva: as though. So: the mind takes the seat inside, then looks back, and sees it standing out again — and the standing-out is an appearance. This is where the compound's members come from: not from the compound, but from the backward glance (3.14.188's buddhya eva; 3.14.93's buddhy-antaram upasritah).

3.14.312 unformalized

समासे श्रूयते स्वार्थो येन तद्वास्तदाश्रयः ।
द्रव्यं तु लिङ्गसङ्ख्यावदसताभ्यन्तरीकृतम् ॥ ३१२ ॥

samāse śrūyate svārtho yena tadvāstadāśrayaḥ |
dravyaṃ tu liṅgasaṅkhyāvadasatābhyantarīkṛtam || 312 ||

In the compound its own meaning is heard, by which the possessor of that has that for substratum; but the substance possessing gender and number is internalized by 'asat'.

Samase sruyate svartho yena tadvas tad-asrayah: in the compound *its own meaning* is *heard*, by which the *possessor of that* has *that* for *substratum* — what is audible in 'abrahmana' is brahmin-hood, and it is what the possessor rests on. Dravyam tu linga-sankhyavad asata abhyantarikrtam: *but* the *substance* possessing *gender and number* is *internalized* by *'asat'* — abhyantarikrta again (3.14.311). So the audible member gives the substratum and the negative swallows the substance — and the swallowed substance is what carries the concord that 3.14.310 said the compound goes to. The division of labour is complete: the base is heard, the nañ digests, and the gender comes out of the digestion.

3.14.313 unformalized

एकार्थविषयौ शब्दौ तस्मिन्नन्यार्थवर्तिनौ ।
असतैव तु भेदानां सर्वेषामुपसंग्रहः ॥ ३१३ ॥

ekārthaviṣayau śabdau tasminnanyārthavartinau |
asataiva tu bhedānāṃ sarveṣāmupasaṃgrahaḥ || 313 ||

The two words, having one meaning for their domain, operate there in another's meaning; but by 'asat' alone there is the comprehension of all the differences.

Ekartha-visayau sabdau tasminn anyartha-vartinau: the *two words*, having *one meaning* for their *domain* — 'asat' and 'brahmana', co-referring (3.14.296's ekarthe vartamanabhyam) — *operate* there *in another's meaning*: both point past themselves to the ksatriya. Asata eva tu bhedanam sarvesam upasamgrahah: *but* by *'asat'* *alone* there is the *comprehension* of *all the differences* — upasamgraha: the gathering-in of the whole field. So the base narrows and the negative *gathers*: 'asat' holds open everything that is not a brahmin, and 'brahmana' is what tells you which opening you are looking through. This is the exact inverse of the usual picture, in which the negative subtracts.

3.14.314 unformalized

ते क्षत्रियादिभिर्वाच्या वाच्या वा सर्वनामभिः ।
यान्तीवान्यपदार्थत्वं नञो रूपाविकल्पनात् ॥ ३१४ ॥

te kṣatriyādibhirvācyā vācyā vā sarvanāmabhiḥ |
yāntīvānyapadārthatvaṃ naño rūpāvikalpanāt || 314 ||

They are to be expressed by 'ksatriya' and the rest, or by pronouns; they go, as it were, to the state of being another's word-meaning, from the non-construction of nañ's form.

Te ksatriyadibhir vacya vacya va sarvanamabhih: *they* — the differences gathered at 3.14.313 — are *to be expressed* by *'ksatriya'* and the rest, *or* by *pronouns*: one names the sibling or one says 'that one' (3.14.217's tac-chabdo 'nuprayujyate). Yanti iva anya-padarthatvam nano rupa-avikalpanat: they go, *as it were*, to the state of *being another's word-meaning*, from the *non-construction* of *nañ's form* — iva: as though. So because nañ contributes no form of its own, the compound *looks like* a bahuvrihi — which is why 3.14.302's theorists thought it was one. Bhartrhari grants the resemblance and marks it with iva. The appearance is real and the identity is not (3.14.303's agur asvah).

3.14.315 unformalized

विशेषस्याप्रयोगे तु लिङ्गसङ्ख्ये न सिध्यतः ।
अवर्षादिषु दोषश्च हेमन्तोऽन्याश्रयो यतः ॥ ३१५ ॥

viśeṣasyāprayoge tu liṅgasaṅkhye na sidhyataḥ |
avarṣādiṣu doṣaśca hemanto'nyāśrayo yataḥ || 315 ||

But when the particular is not employed, gender and number do not succeed; and there is a fault in 'avarsah' and the rest, since 'hemanta' has another substratum.

The account's honest remainder. Visesasya aprayoge tu linga-sankhye na sidhyatah: when the *particular* is *not employed*, gender and number do *not succeed* — 3.14.310's solution requires the substratum to be *prayukta*, actually said; where it is not, nothing supplies concord. Avarsadisu dosas ca hemanto 'nyasrayo yatah: and there is a *fault* in *'avarsah'* and the rest, since *'hemanta'* has *another substratum*. So 3.14.304's difficulty returns and is not solved: 'avarsah' should give 'hemanta', and hemanta rests elsewhere. Dosah: a *fault* — Bhartrhari names it in his own account and does not repair it. That is the samuddesa's manner, and it has been consistent since 3.14.34.

3.14.316 unformalized

आकृतिः सर्वशब्दानां यदा वाच्या प्रतीयते ।
एकत्वादेकशब्दत्वं न्याय्यं तस्यां च वर्ण्यते ॥ ३१६ ॥

ākṛtiḥ sarvaśabdānāṃ yadā vācyā pratīyate |
ekatvādekaśabdatvaṃ nyāyyaṃ tasyāṃ ca varṇyate || 316 ||

When the form is cognized as what all words express — then, from its oneness, singular-wordhood is described as proper in it.

A last position, and it is the akrti-vadin's. Akrtih sarva-sabdanam yada vacya pratiyate: when the *akrti* — the form, the shape, the configuration — is cognized as what *all words* *express*: the doctrine that every word means a form (the Vajapyayana position, against Vyadi's dravya; 3.1's opening quarrel). Ekatvad eka-sabdatvam nyayyam tasyam ca varnyate: from its *oneness*, *singular-wordhood* is described as *proper* in it. So if words mean forms, and a form is one, every word should be singular. Nyayyam: proper, following the reasoning. The consequence is stated neutrally, and 3.14.317-318 draw out what it costs.

3.14.317 unformalized

आविष्टलिङ्गता तस्यां स्याद् ग्राम्यपशुसङ्घवत् ।
द्रव्येभेदेऽपि चैकत्वात् तत्रैकवचनं भवेत् ॥ ३१७ ॥

āviṣṭaliṅgatā tasyāṃ syād grāmyapaśusaṅghavat |
dravyebhede'pi caikatvāt tatraikavacanaṃ bhavet || 317 ||

In it there would be pervaded-gender, as with the aggregate of domestic animals; and even where the substances differ, from oneness the singular would obtain there.

Avista-lingata tasyam syad gramya-pasu-sanghavat: in it there would be *pervaded-gender* — avista-linga: gender entered into it from outside (3.14.150's avesa) — as with the *aggregate of domestic animals*: 'gramya-pasu', a collective that takes one gender though it covers cows, goats and sheep. Dravya-bhede 'pi ca ekatvat tatra eka-vacanam bhavet: and *even where the substances differ*, from *oneness* the *singular* would obtain. So on the akrti-account every word behaves like a collective noun: one form, one gender, one number, whatever the multiplicity underneath. That is a coherent position and it is very nearly 3.14.99's abheda-ekatva — which is why Bhartrhari can state it without hostility.

3.14.318 unformalized

आश्रयाणां हि लिङ्गैः सा नियतैरेव युज्यते ।
तथा च युक्तवद्भावे प्रतिषेधो निरर्थकः ॥ ३१८ ॥

āśrayāṇāṃ hi liṅgaiḥ sā niyataireva yujyate |
tathā ca yuktavadbhāve pratiṣedho nirarthakaḥ || 318 ||

For it is joined only with the fixed genders of the substrata; and thus, in 'treatment as though joined', the prohibition is purposeless.

The consequence, and it is a rule-economy. Asrayanam hi lingaih sa niyatair eva yujyate: for *it* — the akrti — is joined *only* with the *fixed genders* of the *substrata*: the form has no gender of its own and takes the seats', which are settled (3.14.310's yati tal-linga-sankhyatam, on the rival account). Tatha ca yuktavad-bhave pratisedho nirarthakah: and *thus*, in *yuktavad-bhava*, the *prohibition* is *purposeless* — nirarthaka. So the akrti-doctrine makes one of Panini's prohibitions idle: if concord always comes from the fixed seats, no rule need forbid it coming from elsewhere. This is 3.14.106's pattern (a doctrine paying for itself in rule-economy) offered to the opponent's account — Bhartrhari finds the rival's advantage for it, and does not pretend the finding is his.

3.14.319 unformalized

सर्वत्राविष्टलिङ्गत्वं लोकलिङ्गपरिग्रहे ।
विरोधित्वात् प्रसज्येत नाश्रितं तच्च लौकिकम् ॥ ३१९ ॥

sarvatrāviṣṭaliṅgatvaṃ lokaliṅgaparigrahe |
virodhitvāt prasajyeta nāśritaṃ tacca laukikam || 319 ||

If the worldly gender were taken up, pervaded-gender everywhere would follow, from contradiction; and that worldly gender is not resorted to.

Sarvatra avista-lingatvam loka-linga-parigrahe: *pervaded-gender everywhere*, if the *worldly gender* is *taken up* — loka-linga: gender as the world knows it, sexual difference. Virodhitvat prasajyeta: would *follow*, from *contradiction*. Na asritam tac ca laukikam: and that *worldly* gender is *not resorted to*. So grammatical gender is not the world's gender, and the science does not use the world's. This is the premise the whole gender-analysis has been resting on and here it is stated: the linga of the sastra is a technical category, and looking for sex in it is a category error (3.14.137's the word's formation has the general for its domain). 3.14.320 proves it in one line.

3.14.320 unformalized

सामान्यमाकृतिर्भावो जातिरित्यत्र लौकिकम् ।
लिङ्गं न सम्भवत्येव तेनान्यत् परिगृह्यते ॥ ३२० ॥

sāmānyamākṛtirbhāvo jātirityatra laukikam |
liṅgaṃ na sambhavatyeva tenānyat parigṛhyate || 320 ||

In 'samanya', 'akrti', 'bhava', 'jati' — here the worldly gender is not possible at all; therefore another is taken up.

The proof, and it is beautiful because it is drawn from the grammarians' own vocabulary. Samanyam akrtir bhavo jatir ity atra: in *'samanya'* (neuter), *'akrti'* (feminine), *'bhava'* (masculine), *'jati'* (feminine) — four words for one thing, the universal, in three genders. Laukikam lingam na sambhavaty eva: here the *worldly* gender is *not possible at all* — na sambhavaty eva: not even possible; a universal has no sex, and yet these four have three genders among them. Tena anyat parigrhyate: *therefore* *another* is *taken up*. So gender cannot come from the meaning, because four synonyms differ in it. The argument is one verse long and it settles the matter — and the examples are the terms of the debate itself.

3.14.321 unformalized

प्रवृत्तिरिति सामान्यं लक्षणं तस्य कथ्यते ।
आविर्भावस्तिरोभावः स्थितिश्चेत्यथ भिद्यते ॥ ३२१ ॥

pravṛttiriti sāmānyaṃ lakṣaṇaṃ tasya kathyate |
āvirbhāvastirobhāvaḥ sthitiścetyatha bhidyate || 321 ||

'Pravrtti' — the general characteristic of it is stated; then it is divided as manifestation, disappearance, and persistence.

The other gender, and it is not what one expects. Pravrttir iti samanyam laksanam tasya kathyate: *'pravrtti'* — the *general characteristic* of *it* is *stated*: what the akrti is, generally, is *operation* — pravrtti: going forth, activity, the thing's being under way. Avirbhavas tirobhavah sthitis ca ity atha bhidyate: *then* it is *divided* as *manifestation*, *disappearance*, and *persistence*. So the universal is not a static form but a threefold activity: coming-into-view, going-out-of-view, and staying. This is the sat-bhava-vikara tradition (the six modifications) compressed to three, and it is the samuddesa's most metaphysical moment: what a word means is a rhythm.

3.14.322 unformalized

प्रवृत्तिमन्तः सर्वेऽर्थास्तिसुभिश्च प्रवृत्तिभिः ।
सततं न वियुज्यन्ते वाचश्चैवात्र सम्भवः ॥ ३२२ ॥

pravṛttimantaḥ sarve'rthāstisubhiśca pravṛttibhiḥ |
satataṃ na viyujyante vācaścaivātra sambhavaḥ || 322 ||

All things possess operation, and are never disjoined from the three operations; and here is the possibility of speech.

Pravrttimantah sarve 'rthah: *all* things *possess operation* — nothing is inert; everything is under way. Tisrbhis ca pravrttibhih satatam na viyujyante: they are *never* *disjoined* from the *three* operations — satatam: constantly, at every moment: everything is at all times manifesting, vanishing, and persisting at once. Vacas caiva atra sambhavah: and *here* is the *possibility of speech*. So language is possible *because* of this: a word can catch a thing only where the thing shows a manifestation to be caught, and the ceaseless three-beat rhythm is what offers words their purchase. Atra sambhavah — *here* is where speech can happen. Without the pravrttis there would be nothing for a word to be about.

3.14.323 unformalized

यश्चाप्रवृत्तिधर्मार्थश्चितिरूपेण गृह्यते ।
अनुयातीव सोऽन्येषां प्रवृत्तीविष्वगाश्रयाः ॥ ३२३ ॥

yaścāpravṛttidharmārthaścitirūpeṇa gṛhyate |
anuyātīva so'nyeṣāṃ pravṛttīviṣvagāśrayāḥ || 323 ||

And that thing whose property is non-operation, which is grasped in the form of consciousness — it follows, as it were, the operations of others, which rest everywhere.

The exception to 3.14.322, and it is the self. Yas ca apravrtti-dharmarthah: that *thing* whose *property* is *non-operation* — apravrtti: it does not go forth; it does not manifest, vanish or persist, because those are changes and it does not change. Citi-rupena grhyate: which is *grasped in the form of consciousness* — citi: awareness (1.1's the changeless Word-Brahman; 1.4's the one that appears as many). Anuyati iva so 'nyesam pravrttir visvag-asrayah: it *follows*, *as it were*, the *operations of others*, which *rest everywhere*. Iva: as though. So consciousness does not move and appears to move with everything that does — as the moon appears to run through the clouds. Visvag-asrayah: resting on all sides. The changeless is the only thing outside the rhythm, and it looks like the busiest thing there is.

3.14.324 unformalized

तेनास्य चितिरूपं च चितिकालश्च भिद्यते ।
तस्य स्वरूपभेदस्तु न कश्चिदपि विद्यते ॥ ३२४ ॥

tenāsya citirūpaṃ ca citikālaśca bhidyate |
tasya svarūpabhedastu na kaścidapi vidyate || 324 ||

Therefore its consciousness-form and its consciousness-time are differentiated; but of it there is no difference of its own form at all.

Tena asya citi-rupam ca citi-kalas ca bhidyate: *therefore* its *consciousness-form* and its *consciousness-time* are *differentiated* — the awareness appears as this awareness and that, at this moment and that: it acquires shape and date from what it seems to follow (3.14.323's anuyati iva). Tasya sva-rupa-bhedas tu na kascid api vidyate: *but* of it there is *no* difference of *its own form* *at all* — na kascid api: not the least. So all the differentiation is in the apparent following and none in the thing: consciousness is one and is dated and shaped by borrowing. This is 1.4's ekam eva yad amnatam bhinnam saktivyapasrayat, and the samuddesa has arrived at it from the grammar of gender.

3.14.325 unformalized

अचेतनेषु संक्रान्तं चैतन्यमिव दृश्यते ।
प्रतिबिम्बकधर्मेण यत्तच्छब्दनिबन्धनम् ॥ ३२५ ॥

acetaneṣu saṃkrāntaṃ caitanyamiva dṛśyate |
pratibimbakadharmeṇa yattacchabdanibandhanam || 325 ||

Consciousness is seen as though transferred into the unconscious, by the property of reflection; and that is the basis of the word.

One of the great images of the work. Acetanesu samkrantam caitanyam iva drsyate: *consciousness* is *seen* *as though* *transferred* into the *unconscious* — samkranta: gone across, migrated; iva: as though. Pratibimbaka-dharmena: by the *property of reflection* — as a face is in a mirror without going there. So awareness appears to have got into stones and pots, and it has not: what one sees is a reflection, and reflections do not move. Yat tac-chabda-nibandhanam: and *that* is the *basis of the word* — sabda-nibandhana: what the word rests on. So words are possible because consciousness appears in the insentient: a thing can be *meant* only by seeming to have some awareness lodged in it. This is 1.132's the word-principle in all things, and it grounds 3.14.322's vacas ca eva atra sambhavah — speech is possible where the reflection falls.

3.14.326 unformalized

अवस्था तादृशी नास्ति या लिङ्गेन न युज्यते ।
क्वचित्तु शब्दसंस्कारो लिङ्गस्यानाश्रये सति ॥ ३२६ ॥

avasthā tādṛśī nāsti yā liṅgena na yujyate |
kvacittu śabdasaṃskāro liṅgasyānāśraye sati || 326 ||

There is no such state as is not joined with gender; but somewhere there is the word's formation, when there is no substratum of gender.

Avastha tadrsi nasti ya lingena na yujyate: there is *no* such *state* as is *not joined with gender* — every condition of a thing has one of the three (3.14.321's manifestation, disappearance, persistence: and this is what the linga now turns out to be). Kvacit tu sabda-samskaro lingasya anasraye sati: *but* *somewhere* there is the *word's formation* when there is *no substratum* of gender. So gender is universal in the states and there are still words with no seat for it — the indeclinables (3.14.105's avyayanam yo dharmah). The two facts stand side by side: everything has gender, and some words have nowhere to put it. Bhartrhari does not reconcile them here.

3.14.327 unformalized

कृत्तद्धिताभिधेयानां भावानां न विरुध्यते ।
शस्त्रे लिङ्गं गुणावस्था तथा चाकृतिरिष्यते ॥ ३२७ ॥

kṛttaddhitābhidheyānāṃ bhāvānāṃ na virudhyate |
śastre liṅgaṃ guṇāvasthā tathā cākṛtiriṣyate || 327 ||

Of the states expressed by krt and taddhita, gender is not contradicted; in the science, gender is the condition of a quality, and thus the akrti is accepted.

Krt-taddhita-abhidheyanam bhavanam na virudhyate: of the *states* expressed by *krt* and *taddhita* — the deverbal and denominal abstracts, 'pakti' and 'sauklya' — *gender* is *not contradicted*: these take gender readily, though they name no substances. Sastre lingam gunavastha: in the *science*, *gender* is the *condition of a quality* — guna-avastha: a quality's state, which is 3.14.321's pravrtti under another name. Tatha ca akrtir isyate: and *thus* the *akrti* is *accepted*. So the grammarians' linga is a quality's condition, not a sex — which is what 3.14.319-320 argued and what the krt-taddhita abstracts confirm: they have gender because they have conditions, not because they have bodies.

3.14.328 unformalized

लिङ्गं प्रति न भेदोऽस्ति द्रव्यपक्षेऽपि कश्चन ।
तस्मात् सप्त विकल्पा ये सैवात्राविष्टलिङ्गता ॥ ३२८ ॥

liṅgaṃ prati na bhedo'sti dravyapakṣe'pi kaścana |
tasmāt sapta vikalpā ye saivātrāviṣṭaliṅgatā || 328 ||

As regards gender there is no difference at all, even on the substance-view; therefore, whatever the seven alternatives may be, that same pervaded-gender obtains here.

Lingam prati na bhedo 'sti dravya-pakse 'pi kascana: as regards *gender* there is *no difference* *at all*, *even on the substance-view* — the akrti-vadin and the dravya-vadin (3.14.316's ancient quarrel) come out the same on gender. Tasmat sapta vikalpa ye sa eva atra avista-lingata: *therefore*, whatever the *seven alternatives* may be, that same *pervaded-gender* obtains here. Sapta vikalpah: the seven positions the tradition had enumerated. So all seven agree, and the great dispute makes no difference to the case at hand. This is 3.14.119 and 3.14.202's move for the last time, and its scope is now maximal: the result survives every position anyone holds, and so the positions need not be adjudicated at all.

3.14.329 unformalized

वचने नियमः शास्त्राद् द्रव्यस्याभ्युपगम्यते ।
यतस्तदाकृतौ शास्त्रमन्यथैव समर्थ्यते ॥ ३२९ ॥

vacane niyamaḥ śāstrād dravyasyābhyupagamyate |
yatastadākṛtau śāstramanyathaiva samarthyate || 329 ||

The restriction on the substance's number is accepted from the science; since the science regarding that akrti is justified quite otherwise.

Vacane niyamah sastrad dravyasya abhyupagamyate: the *restriction* on *number* of the *substance* is *accepted from the science* — the grammar, not the world, settles how many a substance-word says (3.14.134's sastram arabdham). Yatas tad-akrtau sastram anyatha eva samarthyate: since the *science* regarding *that akrti* is *justified quite otherwise* — anyatha eva: on a wholly different footing. So the two objects are handled by two disjoint justifications: what makes the rules right for substances does not make them right for forms. The verse is registering that the sastra is not a single system with one ground — it is a set of accommodations, each answering to what it accommodates (3.14.74's science standing far off).

3.14.330 unformalized

वर्तते यो बहुष्वर्थो भेदे तस्य विवक्षिते ।
स्वाश्रयैर्व्युपदिष्टस्य शास्त्रे वचनमुच्यते ॥ ३३० ॥

vartate yo bahuṣvartho bhede tasya vivakṣite |
svāśrayairvyupadiṣṭasya śāstre vacanamucyate || 330 ||

That meaning which operates in many — when its difference is intended, of it as designated by its own substrata, number is stated in the science.

Vartate yo bahusv arthah: that *meaning* which operates in *many* — the general, present in its instances. Bhede tasya vivaksite: when its *difference* is *intended* — vivaksa again: the will to say the difference (3.14.224). Sva-asrayair vyupadistasya sastre vacanam ucyate: of it as *designated by its own substrata*, *number* is *stated* in the *science*. So the sequence: a general in many; a wish to say the manyness; the general designated through its seats; and then the grammar supplies a plural. The number is at the end of a chain that begins in an intention, and no step of it is in the object. This is 3.14.188's buddhya eva in the number's own case.

3.14.331 unformalized

यदा त्वाश्रयभेदेन भेद एव प्रतीयते ।
आकृतेर्द्रव्यपक्षेण तदा भेदो न विद्यते ॥ ३३१ ॥

yadā tvāśrayabhedena bheda eva pratīyate |
ākṛterdravyapakṣeṇa tadā bhedo na vidyate || 331 ||

But when difference itself is cognized through difference of substratum — then, of the akrti, on the substance-view, there is no difference.

Yada tu asraya-bhedena bheda eva pratiyate: when *difference itself* is cognized through *difference of substratum* — the difference one grasps is the seats', not the form's. Akrter dravya-paksena tada bhedo na vidyate: *then*, of the *akrti*, on the *substance-view*, there is *no difference*. So on the dravya-account the form stays undivided while its seats multiply, and the plural one hears reports the seats. This is the pair to 3.14.330: there the science supplied a number to an intended difference; here the difference belongs to the substrata and the form takes none. The two positions differ only in where they lodge the multiplicity, and 3.14.328 has already said it will not matter.

3.14.332 unformalized

अभेदे त्वेकशब्दत्वाच्छास्त्राच्च वचने सति ।
एकशेषो न वक्तव्यो वचनानां त सम्भवः ॥ ३३२ ॥

abhede tvekaśabdatvācchāstrācca vacane sati |
ekaśeṣo na vaktavyo vacanānāṃ ta sambhavaḥ || 332 ||

But in non-difference, from singular-wordhood and from the science's number — ekasesa need not be stated; there is, however, the possibility of the numbers.

Abhede tv eka-sabdatvat sastrac ca vacane sati: in *non-difference*, from *singular-wordhood* (3.14.316's consequence) and from the *science's* number. Ekaseso na vaktavyah: *ekasesa* need *not* be *stated* — the rule that leaves one of several identical words standing becomes unnecessary, since on this account there was only ever one word (3.14.30; 3.14.109). Vacananam tu sambhavah: there is, *however*, the *possibility* of the *numbers* — they remain available. So the akrti-doctrine retires a rule and keeps the plurals: a real economy, and Bhartrhari credits it (3.14.318's nirarthakah pratisedhah, the same generosity). He is auditing the rival position's books and finding them in order.

3.14.333 unformalized

ननु चानभिधेयत्वे द्रव्यस्य तदपाश्रयः ।
आकृतेरुपकारोऽयं द्रव्याभावान्न कल्पते ॥ ३३३ ॥

nanu cānabhidheyatve dravyasya tadapāśrayaḥ |
ākṛterupakāro'yaṃ dravyābhāvānna kalpate || 333 ||

But if substance is not expressible, this assistance to the akrti, resting on that, is not conceivable — from the absence of substance.

The objection. Nanu ca anabhidheyatve dravyasya: *but if* *substance* is *not expressible* — on the strict akrti-view, words mean forms and substances are never said. Tad-apasrayah akrter upakaro 'yam dravyabhavan na kalpate: this *assistance* to the *akrti*, *resting on that*, is *not conceivable*, from the *absence of substance*. Upakara: the help the substrata were giving — 3.14.330-331 had the seats supplying number and difference. But if the seats are inexpressible they cannot help. So the akrti-account has been quietly living off substances it officially does not have. The objection is exact and 3.14.334 answers it exactly.

3.14.334 unformalized

व्यपदेशोऽभिधेयेन न शास्त्रे कश्चिदाश्रितः ।
द्रव्यं नाम पदार्थो यो न च स प्रतिषिध्यते ॥ ३३४ ॥

vyapadeśo'bhidheyena na śāstre kaścidāśritaḥ |
dravyaṃ nāma padārtho yo na ca sa pratiṣidhyate || 334 ||

No designation in the science rests on what is expressed; and that word-meaning which is called substance — it is not denied.

The answer, and it is two moves. Vyapadeso 'bhidheyena na sastre kascid asritah: *no* designation in the *science* rests on *what is expressed* — the grammar's technical terms do not attach to meanings; they attach to forms and notation (3.14.78's anubandha-svaradibhyah... na tan prati). So the objection's premise — that inexpressibility disables assistance — misreads how the science works: it never used expressibility. Dravyam nama padartho yo na ca sa pratisidhyate: and that *word-meaning* which is *called substance* — it is *not denied*. So the akrti-vadin never said there are no substances; he said words do not express them. The substances are there, helping, unsaid. The distinction between what is and what is expressed is the whole of the reply — and it is the distinction this work is built on.

3.14.335 unformalized

गुणभावोऽभिधेयत्वं प्रति द्रव्यस्य नाश्रितः ।
उपकारी गुणः शेषः पदार्थ इति कल्पना ॥ ३३५ ॥

guṇabhāvo'bhidheyatvaṃ prati dravyasya nāśritaḥ |
upakārī guṇaḥ śeṣaḥ padārtha iti kalpanā || 335 ||

The substance's subordination is not resorted to with regard to expressibility; the construction is: 'the assisting quality is the remainder — that is the word-meaning'.

Guna-bhavo 'bhidheyatvam prati dravyasya na asritah: the *substance's* *subordination* is *not resorted to* with regard to *expressibility* — the substance is not subordinate *qua* expressible, since on this view it is not expressed at all (3.14.334). Upakari gunah sesah padartha iti kalpana: the *construction* is — the *assisting quality* is the *remainder*, and *that* is the *word-meaning*. Kalpana: a construction (3.14.96's bheda-samsarga-kalpana). So subordination is a relation among *constructed* elements, not a fact about what is or is not said. This separates two things the debate had run together: being unexpressed and being subordinate. The substance is the first and not the second.

3.14.336 unformalized

द्रव्ये न गुणभावोऽस्ति विना द्रव्याभिधायिताम् ।
आकृतौ वा प्रधानत्वमत एवं समर्थ्यते ॥ ३३६ ॥

dravye na guṇabhāvo'sti vinā dravyābhidhāyitām |
ākṛtau vā pradhānatvamata evaṃ samarthyate || 336 ||

There is no subordination in the substance without substance-expression; or else the akrti's principality — therefore it is justified thus.

Dravye na guna-bhavo 'sti vina dravyabhidhayitam: there is *no subordination* in the *substance* *without* *substance-expression* — one cannot make a thing subordinate without saying it: subordination is a relation between things said (3.14.335's kalpana). Akrtau va pradhanatvam ata evam samarthyate: *or else* the *akrti's* *principality* — *therefore* it is *justified* thus. So the choice: either the substance is expressed (and can then be subordinated) or the akrti is principal (and the substance is simply outside the account). The akrti-vadin takes the second, and the verse concedes it is coherent: samarthyate, it is justified. The account is consistent, and its consistency is bought by putting substances outside the semantics altogether.

3.14.337 unformalized

कैश्चिद् गुणप्रधानत्वं नामाख्यातवदिष्यते ।
न वृत्तिवत् परार्थस्य गुणभावस्तु वर्ण्यते ॥ ३३७ ॥

kaiścid guṇapradhānatvaṃ nāmākhyātavadiṣyate |
na vṛttivat parārthasya guṇabhāvastu varṇyate || 337 ||

Some accept subordinate-and-principal as with noun and verb; but the subordination of what serves another's purpose is not described as in a derivative formation.

Kaiscid guna-pradhanatvam nama-akhyatavad isyate: *some* accept *subordinate-and-principal* as with *noun and verb* — the sentence's own hierarchy: the verb principal, the nominals subordinate (3.8.2's the action is the principal; 2.326's the verb as the sentence's core). Na vrttivat pararthasya guna-bhavas tu varnyate: but the *subordination* of what *serves another's purpose* is *not* described as in a *derivative formation*. So the sentence's subordination is not the compound's: in a sentence the subordinate members are still there, saying things; in a compound they are not (3.14.53's bhagabhyam vartate vina). The distinction matters because the akrti-debate keeps borrowing sentence-relations to describe word-internal ones, and this verse marks the borrowing as illicit.

3.14.338 unformalized

गुणभूतस्य नानात्वादाकृतेरेकशब्दता ।
सिद्धो वचनभेदश्च द्रव्यभेदसमन्वयात् ॥ ३३८ ॥

guṇabhūtasya nānātvādākṛterekaśabdatā |
siddho vacanabhedaśca dravyabhedasamanvayāt || 338 ||

From the manifoldness of what is subordinate, the akrti has singular-wordhood; and the difference of number is established from concomitance with the difference of substances.

Guna-bhutasya nanatvad akrter eka-sabdata: from the *manifoldness* of what is *subordinate*, the *akrti* has *singular-wordhood* — and the logic is exactly inverted from what one expects: because the *subordinate* is many, the *principal* is one-worded. The manyness is all lodged below, and the form above is single. Siddho vacana-bhedas ca dravya-bheda-samanvayat: and the *difference of number* is *established* from *concomitance* with the *difference of substances*. So the plural one hears is the substances' and the word is single: the account gets both facts, and gets them from the same division of labour. This is 3.14.331 stated as a doctrine rather than a hypothetical.

3.14.339 unformalized

साधनं गुणभावेन क्रियाया भेदकं यथा ।
आख्यातेष्वेकशब्दाया जातेर्द्रव्यं तथोच्यते ॥ ३३९ ॥

sādhanaṃ guṇabhāvena kriyāyā bhedakaṃ yathā |
ākhyāteṣvekaśabdāyā jāterdravyaṃ tathocyate || 339 ||

As the means, by subordination, is the differentiator of the action in verbs — so the substance is said to be that of the universal, which has a single word.

The analogy that makes 3.14.338 intelligible. Sadhanam guna-bhavena kriyaya bhedakam yatha akhyatesu: as the *means*, *by subordination*, is the *differentiator* of the *action* in *verbs* — 'pacati' is one action-word and its karakas make it this cooking and that (3.7's the sadhanas; 3.8.44's the action one and differentiated by its means). Eka-sabdayah jater dravyam tatha ucyate: *so* the *substance* is said to be that of the *universal*, which has a *single word*. So the universal stands to its substances as the verb to its karakas: one word above, the differentiating many below. This is the samuddesa's most economical statement of the akrti-position, and it makes it look, for the first time, obviously right.

3.14.340-341 unformalized

एकत्वे तुल्यरूत्वाच्छब्दानां प्रतिपादने ।
निमित्तात् तद्वतोऽर्थस्य विशिष्टग्रहणे सति ॥ ३४० ॥
सोऽयमित्यभिसम्बन्धादाश्रयैराकृतेः सह ।
प्रवृत्तौ भिन्नशब्दायां लिङ्गसंख्ये प्रसिध्यतः ॥ ३४१ ॥

ekatve tulyarūtvācchabdānāṃ pratipādane |
nimittāt tadvato'rthasya viśiṣṭagrahaṇe sati || 340 ||
so'yamityabhisambandhādāśrayairākṛteḥ saha |
pravṛttau bhinnaśabdāyāṃ liṅgasaṃkhye prasidhyataḥ || 341 ||

(340) In oneness, from likeness of form in the conveying of words; from the cause, when there is the specified grasping of the meaning possessing that — (341) from the connection 'this is that', in the akrti's operating together with the substrata, which has a differentiated word — gender and number are established.

The mechanism, in two verses. (340): ekatve tulya-rupatvac chabdanam pratipadane — in *oneness*, from *likeness of form* in the *conveying* of *words*: the form is one and the words that convey it are alike. Nimittat tadvato 'rthasya visista-grahane sati: *from the cause*, when there is the *specified grasping* of the *meaning possessing that*. (341): so 'yam ity abhisambandhad asrayair akrteh saha — from the connection *'this is that'*, of the *akrti* *together with* the *substrata*. Pravrttau bhinna-sabdayam linga-samkhye prasidhyatah: in the *operation*, which has a *differentiated word*, *gender and number* are *established*. So: the form is one, it identifies with its seats, and in *that joint operation* — where the word does differentiate — concord arises. Neither the form alone nor the seats alone would give it; the identification is what does.

3.14.342 unformalized

प्राक् च जात्यभिसंबन्धात् सर्वनामाभिधेयता ।
वस्तूपलक्षणाः सत्त्वे प्रयुज्यन्ते त्यदादयः ॥ ३४२ ॥

prāk ca jātyabhisaṃbandhāt sarvanāmābhidheyatā |
vastūpalakṣaṇāḥ sattve prayujyante tyadādayaḥ || 342 ||

And before connection with the universal, there is expressibility by pronouns; 'tyad' and the rest, indicating things, are employed for substance.

Prak ca jaty-abhisambandhat sarvanama-abhidheyata: *before* connection with the *universal*, there is *expressibility by pronouns* — before a thing is brought under any kind, it can still be pointed at: 'that'. Vastupalaksanah sattve prayujyante tyad-adayah: *'tyad'* and the rest, *indicating things* — upalaksana: indicating without characterizing (3.14.170) — are *employed for substance*. So the pronoun is the language's access to a substance as such, prior to every universal: it indicates and says nothing. This answers a real question the akrti-account raises — if words mean forms, how is a bare substance ever reached? — and the answer is: by pronouns, which do not mean forms. It is why 3.14.217 and 3.14.314 kept finding pronouns wherever a substance was unexpressed.

3.14.343 unformalized

पाकौ पाका इति यथा मेदकः कैश्चिदाश्रयः ।
इष्यते चानुपादानो धर्मोऽसौ गुणवाचिनाम् ॥ ३४३ ॥

pākau pākā iti yathā medakaḥ kaiścidāśrayaḥ |
iṣyate cānupādāno dharmo'sau guṇavācinām || 343 ||

As in 'the two cookings', 'the cookings', the substratum is accepted by some as the differentiator; and that property of quality-words is without any taking-up.

Pakau paka iti yatha bhedakah kaiscid asrayah isyate: as in *'pakau'*, *'pakah'* — the two cookings, the cookings — the *substratum* is *accepted by some* as the *differentiator*: the abstract noun pluralizes because its seats are several, not because cooking is (3.14.339's sadhanam bhedakam). Ca anupadano dharmo 'sau guna-vacinam: and that *property* of *quality-words* is *without any taking-up* — anupadana: the differentiating seats are never mentioned, and they differentiate anyway. So 'pakah' says several cookings without saying whose, and the plural is owed to seats that the word does not touch. This is the akrti-account's cleanest evidence: the number is demonstrably not in what is said.

3.14.344 unformalized

आश्रयस्यानुपादाने केवलं लभते यदि ।
आधारधर्मान् सामान्यं पुरस्तात् तद्विचारितम् ॥ ३४४ ॥

āśrayasyānupādāne kevalaṃ labhate yadi |
ādhāradharmān sāmānyaṃ purastāt tadvicāritam || 344 ||

If, without any taking-up of the substratum, the general alone obtains the substratum's properties — that has been considered earlier.

Asrayasya anupadane kevalam labhate yadi adhara-dharman samanyam: *if*, without any *taking-up* of the *substratum*, the *general* *alone* obtains the *substratum's properties* — the position 3.14.343 just described, stated as a question. Purastat tad vicaritam: *that* has been *considered earlier* — purastat: before, in the earlier part of the work. So Bhartrhari points back rather than repeat himself (3.14.244's granthe purvatra, the same gesture) — the Jati-samuddesa (3.1) took up whether a universal can wear its instances' properties, and the reader is sent there. The Vakyapadiya is again treating itself as one continuous argument, and the cross-reference is the only answer this verse gives.

3.14.345 unformalized

जातौ पूर्वं प्रवृत्तानां शब्दानां जातिवाचिनाम् ।
अशब्दवाच्यात् सम्बन्धाद् व्यक्तिरप्युपजायते ॥ ३४५ ॥

jātau pūrvaṃ pravṛttānāṃ śabdānāṃ jātivācinām |
aśabdavācyāt sambandhād vyaktirapyupajāyate || 345 ||

Of universal-expressing words, previously operating in the universal — from a relation not expressed by any word, the individual also arises.

How a jati-word gets to an individual. Jatau purvam pravrttanam sabdanam jati-vacinam: of *universal-expressing* words, *previously operating* in the *universal* — the word starts at the kind, always. Asabda-vacyat sambandhad vyaktir apy upajayate: from a *relation* *not expressed by any word*, the *individual* *also* *arises*. Asabda-vacya: not sayable by a word — the relation by which 'cow' arrives at this cow is itself unsaid, and unsayable (3.3's the sambandha-samuddesa: relation cannot be expressed). Upajayate: *arises*. So the individual is not reached by any further word: it comes up out of an unsayable relation, and the word never left the universal. This is why 3.14.342 needed pronouns — and why 'cow' can pick out this cow without meaning her.

3.14.346 unformalized

सोऽयमित्यभिसम्बन्धाज्जातिधर्मोपचर्यते ।
द्रव्यं तदाश्रयो भेदो जातेश्चाभ्युपगम्यते ॥ ३४६ ॥

so'yamityabhisambandhājjātidharmopacaryate |
dravyaṃ tadāśrayo bhedo jāteścābhyupagamyate || 346 ||

From the connection 'this is that', the universal's property is figuratively applied; and the substance, resting on that, is accepted as the universal's difference.

So 'yam ity abhisambandhaj jati-dharmopacaryate: from the connection 'this is that', the *universal's property* is *figuratively applied* — upacaryate: transferred (3.14.284's word). The individual wears cowhood by transfer, not by containing it. Dravyam tad-asrayo bhedo jates ca abhyupagamyate: and the *substance*, *resting on that*, is *accepted* as the *universal's* *difference*. So the substance is what differentiates the universal (3.14.339) and it does so while wearing the universal's property figuratively. The relation is thus doubly non-literal: the universal is in the individual by transfer, and the individual is the universal's difference by acceptance (abhyupagamyate). Nothing here is a fact about cows.

3.14.347 unformalized

मञ्चशब्दो यथाऽऽधेयं मञ्चेष्वेव व्यवस्थितः ।
तत्त्वेनाह तथा जातिशब्दो द्रव्येषु वर्तते ॥ ३४७ ॥

mañcaśabdo yathā''dheyaṃ mañceṣveva vyavasthitaḥ |
tattvenāha tathā jātiśabdo dravyeṣu vartate || 347 ||

As the word 'platform', settled in the platforms alone, states what rests on them as its reality — so the universal-word operates in substances.

The image that makes 3.14.345-346 concrete, and it is the samuddesa's favourite. Manca-sabdo yatha adheyam mancesv eva vyavasthitah tattvena aha: as the word *'manca'*, *settled in the platforms alone*, states what *rests on them* — the people — as its *reality*. 'Manca hasanti', the platforms laugh (3.14.150). The word never leaves the platforms and what it says is the people, and it says them *tattvena*, as the real thing. Tatha jati-sabdo dravyesu vartate: *so* the *universal-word* *operates in substances*. So 'cow' is settled in cowhood and states the individual cow as its reality — the word stays where it is and its meaning is elsewhere. The whole account of reference in this samuddesa is in that image.

3.14.348 unformalized

तत्र जातिपदार्थत्वं तथैवाभ्युपगम्यते ।
जातिरुत्सङ्ख्या तु द्रव्यात्मन्यनुषज्यते ॥ ३४८ ॥

tatra jātipadārthatvaṃ tathaivābhyupagamyate |
jātirutsaṅkhyā tu dravyātmanyanuṣajyate || 348 ||

There, the universal's being the word-meaning is accepted just so; but the universal — and number — is annexed to the substance-self.

Tatra jati-padarthatvam tathaiva abhyupagamyate: *there*, the *universal's* being the *word-meaning* is *accepted just so* — the akrti-position is granted in full: what the word means is the kind. Jatir ut sankhya tu dravyatmany anusajyate: but the *universal* — *and number* — is *annexed* to the *substance-self*. Anusajyate: is attached, comes along (3.14.279, 3.14.287's word). So the universal is the meaning and it *lands* on the substance, bringing number with it. The two claims coexist: the word means the kind, and the kind attaches to the thing. This is 3.14.347's manca exactly — settled in the platforms, landing on the people.
contested: Pada c reads 'jatir utsankhya tu' in our source, which we cannot construe; we read 'ut' as the particle 'uta' ('and', 'or rather') with 'sankhya' as a second subject, giving 'the universal — and number — is annexed'. An alternative is to take an unattested compound 'ut-sankhya'; we do not press it. The verse's force rests on 'dravyatmany anusajyate', which is secure.

3.14.349 unformalized

अस्येदमिति वा यत्र सोऽयमित्यपि वा श्रुतिः ।
वर्तते परधर्मेण तदन्यदभिधीयते ॥ ३४९ ॥

asyedamiti vā yatra so'yamityapi vā śrutiḥ |
vartate paradharmeṇa tadanyadabhidhīyate || 349 ||

Where the hearing is 'this is of this', or 'this is that' — it operates by another's property; that other is what is expressed.

Asyedam iti va yatra so 'yam ity api va srutih: where the *hearing* is *'this is of this'* — the possessive identification — *or* *'this is that'* — the plain one: the two formulas that have run through this entire samuddesa. Vartate para-dharmena: it *operates* by *another's property* — para-dharma: a property that is not its own. Tad anyad abhidhiyate: *that other* is what is *expressed*. So wherever one of these identifications is in play, the word is working by a borrowed property and what it expresses is the other thing. Bhartrhari is naming, at the end, the device he has used at almost every turn: 'so 'yam' is how a word says what it does not mean.

3.14.350 unformalized

यत् प्रधानं न तस्यास्ति स्वरूपमनिरूपणात् ।
गुणस्य चात्मना द्रव्यं तद्भावेनोपलक्ष्यते ॥ ३५० ॥

yat pradhānaṃ na tasyāsti svarūpamanirūpaṇāt |
guṇasya cātmanā dravyaṃ tadbhāvenopalakṣyate || 350 ||

That which is principal — of it there is no own-form, from non-determination; and the substance is indicated by the quality's self, by way of being that.

A deep and strange verse. Yat pradhanam na tasyasti svarupam anirupanat: that which is *principal* — of *it* there is *no own-form*, from *non-determination*: anirupana: it is not determined, not characterized. What is principal has no nature of its own, precisely because nothing determines it — determination comes from the subordinate, and the principal is what receives it. Gunasya ca atmana dravyam tad-bhavena upalaksyate: and the *substance* is *indicated* by the *quality's self*, *by way of being that*. Upalaksyate: is *indicated* (3.14.170's upalaksana: marked, not described). So the substance — the principal — has no form and is pointed at through a quality that it is said to *be*. The principal is empty and is shown by its subordinates: which is 3.14.164's substance cognized as having expectancy, and 3.2's dravya as the undetermined that all determination is of.

3.14.351 unformalized

गुणस्य भेदकाले तु प्राधान्यमुपजायते ।
संसर्गश्रुतिरर्थेषु साक्षादेव न वर्तते ॥ ३५१ ॥

guṇasya bhedakāle tu prādhānyamupajāyate |
saṃsargaśrutirartheṣu sākṣādeva na vartate || 351 ||

But at the time of differentiating, the quality attains principality; the hearing of conjunction does not operate directly in the meanings.

Gunasya bheda-kale tu pradhanyam upajayate: at the *time of differentiating*, the *quality* attains *principality* — bheda-kale: at the moment of the cutting. So the hierarchy of 3.14.350 inverts momentarily: while the subordinate is doing its differentiating work, it is the principal — because the principal has no form of its own and is passive to the cut. Samsarga-srutir arthesu saksad eva na vartate: the *hearing of conjunction* does *not operate* *directly* in the *meanings* — saksat: immediately. So the joining one hears is never straight onto the meanings: it goes through the differentiation, and the meanings are reached only after the cut has been made. Principality is a function of the moment, not a standing rank.

3.14.352 unformalized

जातौ वृत्तो यदा द्रव्ये स शब्दो वर्तते पुनः ।
जातेरेव पदार्थत्वं न तदाभ्युपगम्यते ॥ ३५२ ॥

jātau vṛtto yadā dravye sa śabdo vartate punaḥ |
jātereva padārthatvaṃ na tadābhyupagamyate || 352 ||

When that word, having operated in the universal, operates again in the substance — then the universal's alone being the word-meaning is not accepted.

Jatau vrtto yada dravye sa sabdo vartate punah: when that word, having *operated in the universal*, operates *again* in the *substance* — the two-step of 3.14.345-347: 'cow' settles on cowhood and lands on the cow. Jater eva padarthatvam na tada abhyupagamyate: *then* the universal's *alone* being the word-meaning is *not accepted*. So the akrti-account cannot survive its own second step: if the word goes on to the substance, the universal is not the *only* meaning. Punah — *again* — is what breaks it. The position was coherent (3.14.336's samarthyate) so long as the word stayed put; it does not stay put, and 3.14.353 says why that is not a refutation.

3.14.353 unformalized

प्रवृत्तानां पुनर्वृत्तिरेकत्वेनोपवर्ण्यते ।
प्रतिपत्तेरुपायेषु न तत्त्वमनुगम्यते ॥ ३५३ ॥

pravṛttānāṃ punarvṛttirekatvenopavarṇyate |
pratipatterupāyeṣu na tattvamanugamyate || 353 ||

The re-operation of what has operated is described as by oneness; in the means of comprehension, reality is not followed.

Pravrttanam punar vrttir ekatvena upavarnyate: the *re-operation* of what has *operated* is *described* as by *oneness* — the word's second step (3.14.352's punah) is accounted for by saying it never left: one operation, described twice. Pratipatter upayesu na tattvam anugamyate: *in the means of comprehension*, *reality* is *not followed* — upaya: means, devices (3.14.76's do not take the upayas for true; 3.14.96's upaya-matram nanatvam). So the two-step is a device and the doctrine that reconciles it is another device, and neither tracks how things are. This is the samuddesa's governing sentence applied to its own analysis of the akrti-question — and it means 3.14.352 was never a refutation, because there was nothing there to refute.

3.14.354 unformalized

अपृथक्।छब्दवाच्यस्य जातिराश्रीयते यदा ।
द्रव्यस्य सति संस्पर्शे तदा जातिपदार्थता ॥ ३५४ ॥

apṛthak|chabdavācyasya jātirāśrīyate yadā |
dravyasya sati saṃsparśe tadā jātipadārthatā || 354 ||

When the universal of what is expressed by a non-separate word is resorted to, there being contact with substance — then the universal is the word-meaning.

The first of three permutations. Aprthak-sabda-vacyasya jatir asriyate yada: when the *universal* of what is expressed by a *non-separate word* is *resorted to* — aprthak-sabda-vacya: said by a word that is not separate from it; the universal and its word are not two. Dravyasya sati samsparse: there being *contact* with *substance* — samsparsa: touch. The substance is grazed, not meant (3.14.48's samsprsya avayavan, the same verb). Tada jati-padarthata: *then* the *universal* is the *word-meaning*. So: substance touched, universal meant — the akrti-position, stated as one option among several rather than as the truth.

3.14.355 unformalized

द्रव्यस्य सति संस्पर्शे द्रव्यमाश्रीयते यदा ।
वाच्यं तेनैव शब्देन तदा द्रव्यपदार्थता ॥ ३५५ ॥

dravyasya sati saṃsparśe dravyamāśrīyate yadā |
vācyaṃ tenaiva śabdena tadā dravyapadārthatā || 355 ||

When, there being contact with substance, the substance is resorted to as expressible by that very word — then the substance is the word-meaning.

The second permutation, and the words are almost identical. Dravyasya sati samsparse dravyam asriyate yada vacyam tena eva sabdena: when, there being *contact* with substance, the *substance* is *resorted to* as *expressible* by *that very word*. Tada dravya-padartha: *then* the *substance* is the *word-meaning*. So the same touch, the same word, and now the substance is meant — the Vyadi-position (3.1's dravya-paksa). Bhartrhari sets 3.14.354 and this side by side with only one word changed, and the effect is deliberate: the great quarrel of the grammarians turns on which of two things one *resorts to* (asriyate), and nothing in the situation decides. Both descriptions fit the same fact.

3.14.356 unformalized

अपृथक्च्छब्दवाच्यापि भेदमात्रे प्रवर्तते ।
यदा सम्बन्धवज्जातिः सापि द्रव्यपदार्थता ॥ ३५६ ॥

apṛthakcchabdavācyāpi bhedamātre pravartate |
yadā sambandhavajjātiḥ sāpi dravyapadārthatā || 356 ||

When the universal, though expressed by a non-separate word, operates in mere difference, as possessing relation — that too is the substance's being the word-meaning.

The third permutation, a hybrid. Aprthak-sabda-vacya api bheda-matre pravartate yada sambandhavat jatih: when the *universal*, *though* expressed by a *non-separate word*, operates *in mere difference*, as *possessing relation*. Sa api dravya-padartha: *that too* is the *substance's* being the word-meaning. So a universal that has gone over into differentiation is functionally a substance-account: once the jati is operating in bheda-matra — mere difference — it is doing the substance's work, and the label 'universal' has stopped mattering. This is the samuddesa dissolving the debate rather than settling it: the positions are not three theories but three descriptions, and one of them turns into another when you push it.

3.14.357 unformalized

अत्यन्तभिन्नयोरेव जातिद्रव्याभिधायिनोः ।
अवाच्यस्योलपकारित्व आश्रिते तूभयार्थता ॥ ३५७ ॥

atyantabhinnayoreva jātidravyābhidhāyinoḥ |
avācyasyolapakāritva āśrite tūbhayārthatā || 357 ||

Of the two utterly different expressers of universal and substance — when the assisting of the inexpressible is resorted to, there is both-meaning-ness.

The fourth position. Atyanta-bhinnayor eva jati-dravyabhidhayinoh: of the *two utterly different* expressers of *universal* and *substance* — atyanta-bhinna: the two are held wholly apart, no hybrid (against 3.14.356). Avacyasya upakaritva asrite tu ubhayarthata: when the *assisting* of the *inexpressible* is *resorted to*, there is *both-meaning-ness*. Avacya: the inexpressible — the relation, which no word says (3.14.345's asabda-vacya). So: the two words stay utterly distinct, an unsayable relation assists, and the result means *both*. Ubhayarthata: the jati-sabda-vada's classical compromise. And it works only because something inexpressible is doing the joining — which has been true of every account in this samuddesa.

3.14.358 unformalized

आश्रिते त्वाश्रयकृतं भेदमभ्युपगच्छता ।
पुनश्चाप्येकशब्दत्वं जातिशब्देऽनुवर्णितम् ॥ ३५८ ॥

āśrite tvāśrayakṛtaṃ bhedamabhyupagacchatā |
punaścāpyekaśabdatvaṃ jātiśabde'nuvarṇitam || 358 ||

But by one who accepts the difference made by the substratum in what is resorted to — singular-wordhood in the universal-word has again been described.

Asrite tv asraya-krtam bhedam abhyupagacchata: by one *accepting* the *difference* *made by the substratum* in what is *resorted to* — the position of 3.14.338: the manyness is the seats'. Punas ca api eka-sabdatvam jati-sabde 'nuvarnitam: *again* *singular-wordhood* in the *universal-word* has been *described*. Punah, anuvarnitam: again, described-after — Bhartrhari marks that he is repeating himself. So the survey closes by returning to where it started, and the return is noted as a return. Four positions, each internally sound, none refuted, and the last is the first. This is what 3.14.353 predicted: in the means of comprehension, reality is not followed, and so the means do not converge.

3.14.359 unformalized

अनिर्ज्ञातस्य निर्ज्ञानं येन तन्मानमुच्यते ।
प्रस्थादि तेन मेयात्मा साकल्येनावधार्यते ॥ ३५९ ॥

anirjñātasya nirjñānaṃ yena tanmānamucyate |
prasthādi tena meyātmā sākalyenāvadhāryate || 359 ||

That by which there is knowledge of the unknown is called 'measure'; 'prastha' and the rest — by it the measured's self is determined in its entirety.

The comparison-compound begins, and it begins by defining its neighbour. Anirjnatasya nirjnanam yena tan manam ucyate: that by which there is *knowledge* of the *unknown* is called *'mana'* — measure. Prasthadi tena meyatma sakalyena avadharyate: *'prastha'* and the rest — *by it* the *measured's self* is *determined* *in its entirety*. Sakalyena: completely, without remainder. So a measure gives total knowledge of what it measures: say 'a prastha of rice' and nothing about the quantity is left over. This is the benchmark against which 3.14.360 will define comparison — and comparison will fall short of it, deliberately.

3.14.360 unformalized

अनिर्ज्ञातं प्रसिद्धेन येन तद्धर्म गम्यते ।
साकल्येनापरिज्ञानादुपमानं तदुच्यते ॥ ३६० ॥

anirjñātaṃ prasiddhena yena taddharma gamyate |
sākalyenāparijñānādupamānaṃ taducyate || 360 ||

That by which the unknown, having that property, is understood through the well-known — from the knowledge being incomplete, that is called 'comparison'.

The definition, and the contrast with 3.14.359 is the whole of it. Anirjnatam prasiddhena yena tad-dharma gamyate: that by which the *unknown*, *having that property*, is understood through the *well-known* — prasiddha: what is established, familiar. One knows the cloud; one learns the complexion. Sakalyena aparijnanad upamanam tad ucyate: *from the knowledge being incomplete*, *that* is called *'upamana'*. Sakalyena aparijnanat: precisely because it is *not* complete. So comparison is defined by its shortfall: a measure exhausts and a comparison does not, and that is not a defect in comparison but its identity. 'Megha-syama' does not tell you the shade; if it did it would be a measure. This is one of the finest short definitions in the work.

3.14.361 unformalized

द्वयोः समानो यो धर्म उरमानोपमेययोः ।
समास उपमानानां शब्दैस्तदभिधायिभिः ॥ ३६१ ॥

dvayoḥ samāno yo dharma uramānopameyayoḥ |
samāsa upamānānāṃ śabdaistadabhidhāyibhiḥ || 361 ||

That property common to the two — the comparans and the comparandum — the compound of comparanda is with the words expressing that.

Dvayoh samano yo dharma upamana-upameyayoh: that *property* *common* to the *two* — the *upamana* (that with which one compares: the cloud) and the *upameya* (that which is compared: the complexion). Samasa upamananam sabdais tad-abhidhayibhih: the *compound of comparanda* is with the *words expressing that* — the common property's words. So 'megha-syama' compounds 'megha' with the word that says the shared property, 'syama'. The rule is stated flatly and it raises at once the question 3.14.362-365 will worry: is the darkness in the cloud and the darkness in the girl one property or two? The whole comparison depends on the answer, and there is no good answer.

3.14.362 unformalized

आधारभेदाद् भेदो यः श्यामत्वे सोऽविवक्षितः ।
गुणोऽसावाश्रितैकत्वो भिन्नाधारः प्रतीयते ॥ ३६२ ॥

ādhārabhedād bhedo yaḥ śyāmatve so'vivakṣitaḥ |
guṇo'sāvāśritaikatvo bhinnādhāraḥ pratīyate || 362 ||

That difference in darkness which comes from difference of substratum is not intended; that quality, having resorted to oneness, is cognized as having different substrata.

Adhara-bhedad bhedo yah syamatve so 'vivaksitah: the *difference* in *darkness* which comes from *difference of substratum* is *not intended* — avivaksita: not meant to be said (3.14.103, 3.14.176's device). The cloud's dark and the girl's dark do differ, and the comparison does not mean the difference. Guno 'sav asrita-ekatvah bhinnadharah pratiyate: that *quality*, *having resorted to oneness*, is cognized as *having different substrata*. Asrita-ekatva: it has *taken up* oneness — the oneness is resorted to, not found. So the comparison works by an intended oneness laid over an actual difference: the darkness is made one so that the two may be compared. The device is exactly 3.14.99's abheda-ekatva, doing rhetorical work.

3.14.363 unformalized

गुणयोर्नियतो भेदः गुणजातेस्तथैकता ।
एकत्वेऽत्यन्तभेदे वा नोरमानस्य सम्भवः ॥ ३६३ ॥

guṇayorniyato bhedaḥ guṇajātestathaikatā |
ekatve'tyantabhede vā noramānasya sambhavaḥ || 363 ||

The difference of the two qualities is fixed; likewise the oneness of the quality-universal. In oneness, or in utter difference, comparison is not possible.

The dilemma, stated with perfect symmetry. Gunayor niyato bhedah: the *difference* of the *two qualities* is *fixed* — the two darknesses are two, and no wishing makes them one. Guna-jates tatha ekata: *likewise* the *oneness* of the *quality-universal* — and darkness-in-general is one, equally fixed. Ekatve 'tyanta-bhede va na upamanasya sambhavah: *in oneness*, *or* in *utter difference*, *comparison* is *not possible*. So: if the property is one, there is nothing to compare (identity is not likeness); if the properties are two, there is no ground of comparison. And both are true at once, of the same darkness. Comparison requires a thing that is neither one nor two — which is 3.14.209's ekas ca dvayatmakah, and here it is a condition of a figure of speech.

3.14.364 unformalized

जातिमात्रव्यपेक्षायामुपमार्थो न कश्चन ।
श्यामत्वमेकं गुणयोरुभयोरपि वर्तते ॥ ३६४ ॥

jātimātravyapekṣāyāmupamārtho na kaścana |
śyāmatvamekaṃ guṇayorubhayorapi vartate || 364 ||

When only the universal is required, there is no purpose of comparison; the one darkness operates in both the qualities.

The first horn pressed. Jati-matra-vyapeksayam upamartho na kascana: when *only* the *universal* is *required*, there is *no purpose* of *comparison* — na kascana: none whatever. Syamatvam ekam gunayor ubhayor api vartate: the *one* darkness *operates* in *both* the qualities. So if the comparison is to rest on darkness-in-general, it says nothing: the cloud is dark and she is dark, and 'dark like a cloud' has told you only that she is dark — which 'dark' had already said. The comparison collapses into a tautology. So the shared property cannot be the universal, and 3.14.365 shows it cannot be the particulars either.

3.14.365 unformalized

येनैव हेतुना श्यामा शस्त्री तत्र प्रवर्तते (प्रतीयते) ।
स हेतुर्देवदत्तायाः प्रत्यये न विशिष्यते ॥ ३६५ ॥

yenaiva hetunā śyāmā śastrī tatra pravartate (pratīyate) |
sa heturdevadattāyāḥ pratyaye na viśiṣyate || 365 ||

The very cause by which 'dark' operates in the knife there — that cause is not distinguished in the cognition of Devadatta.

The second horn. Yena eva hetuna syama sastri tatra pravartate: the *very cause* by which *'dark'* operates in the *knife* there. Sa hetur devadattayah pratyaye na visisyate: *that* cause is *not distinguished* in the *cognition* of *Devadatta* — the dark girl of 3.14.297. Na visisyate: it is not marked off, not told apart. So the reason 'dark' applies to the knife and the reason it applies to her are indistinguishable in the cognition — one cannot get at the particular darknesses to compare them, because the cognition does not separate them. So the comparison cannot rest on the particulars either: they are inaccessible. Between 3.14.364 and this verse, both horns of 3.14.363 are closed, and upamana works anyway.

3.14.366 unformalized

आश्रयाद् यो गुणे भेदो जातेर्या चाविशिष्टता ।
ताभ्यामुभाभ्यां द्रव्यात्मा सव्यापारः प्रतीयते ॥ ३६६ ॥

āśrayād yo guṇe bhedo jāteryā cāviśiṣṭatā |
tābhyāmubhābhyāṃ dravyātmā savyāpāraḥ pratīyate || 366 ||

That difference in the quality which comes from the substratum, and that non-specificity of the universal — by both of these together, the substance-self is cognized as engaged in operation.

The resolution, and it is characteristic: take both horns. Asrayad yo gune bhedo: the *difference* in the *quality* from the *substratum* — 3.14.362's actual difference, 3.14.363's niyato bhedah. Jater ya ca avisistata: and the *non-specificity* of the *universal* — 3.14.364's one darkness in both. Tabhyam ubhabhyam dravyatma sa-vyaparah pratiyate: by *both of these* *together*, the *substance-self* is *cognized* as *engaged in operation*. Sa-vyapara: with its activity, under way (3.14.321's pravrtti). So the comparison does not need the property to be either one or two: the difference and the non-specificity work *together*, and what they deliver is a substance in action. Neither horn alone gives a comparison; both at once do. This is bhedabheda (3.14.209) doing the work no single answer could.

3.14.367 unformalized

सोऽयमेकत्वनानात्वे व्यवहारः समाश्रितः ।
भेदाभेदविमर्शेन व्यतिकीर्णेन वर्तते ॥ ३६७ ॥

so'yamekatvanānātve vyavahāraḥ samāśritaḥ |
bhedābhedavimarśena vyatikīrṇena vartate || 367 ||

That transaction, resorted to in oneness and in manifoldness, operates by an intermingled reflection of difference and non-difference.

The general form of 3.14.366's answer. So 'yam ekatva-nanatve vyavaharah samasritah: that *transaction* — vyavahara: dealing, practice, the whole business of speaking and understanding — *resorted to* in *oneness and in manifoldness*: it stands with a foot in each. Bhedabheda-vimarsena vyatikirnena vartate: it *operates* by a *reflection* of *difference and non-difference* that is *vyatikirna* — *intermingled*, shot through one another, poured together. So the practice does not choose between one and many: it runs on a reflection in which both are mixed past separating. Vyatikirna is the word to keep — not a synthesis, not a compromise, but an interpenetration that no analysis will part. This is the samuddesa's last word on the topic, and it is a description of how speech works, not a metaphysics.

3.14.368 unformalized

श्यामेत्येवाभिधीयते जातिमात्रे विवक्षिते ।
शस्त्र्यादीनामुपादाने तत्र नास्ति प्रयोजनम् ॥ ३६८ ॥

śyāmetyevābhidhīyate jātimātre vivakṣite |
śastryādīnāmupādāne tatra nāsti prayojanam || 368 ||

When only the universal is intended, one simply says 'dark'; there is no purpose in taking up 'knife' and the rest.

Syama ity eva abhidhiyate jati-matre vivaksite: when *only* the *universal* is *intended*, one simply says *'dark'* — ity eva: just that. Sastry-adinam upadane tatra nasti prayojanam: there is *no purpose* in *taking up* *'knife'* and the rest. So if darkness-in-general were the point, the comparison would be idle: one would say 'dark' and stop. The fact that speakers reach for the knife proves they want something the universal does not give. This is 3.14.364's tautology-argument again, now used constructively: the existence of comparison is evidence that there is something below the universal to be gestured at.

3.14.369 unformalized

अशब्दवाच्यो यो भेदः श्याममात्रे न वर्तते ।
श्यामेषु केषुचिद् वृत्तिर्यस्यसोऽत्र व्यपेक्ष्यते ॥ ३६९ ॥

aśabdavācyo yo bhedaḥ śyāmamātre na vartate |
śyāmeṣu keṣucid vṛttiryasyaso'tra vyapekṣyate || 369 ||

That difference not expressible by a word does not operate in mere darkness; the one whose operation is in some dark things — that is what is required here.

Asabda-vacyo yo bhedah syama-matre na vartate: that *difference* *not expressible by a word* — asabda-vacya (3.14.345's word for the unsayable relation) — does *not operate* in *mere darkness*: the shade that makes this dark a cloud's dark is not in darkness-at-large. Syamesu kesucid vrttir yasya so 'tra vyapeksyate: the one whose *operation* is in *some* dark things — *that* is what is *required* here. Kesucit: in *some* — not all and not one. So what the comparison needs is a middle thing: narrower than darkness, wider than this darkness, and unsayable. It has no word, and 'megha-syama' is how one points at it. That is what a simile is for.

3.14.370 unformalized

श्यामेषु केषुचित् किञ्चित् किञ्चित् सर्वत्र वर्तते ।
सामान्यं कश्चिदेकस्मिन् श्यामे भेदो व्यवस्थितः ॥ ३७० ॥

śyāmeṣu keṣucit kiñcit kiñcit sarvatra vartate |
sāmānyaṃ kaścidekasmin śyāme bhedo vyavasthitaḥ || 370 ||

Some general operates in some dark things; some operates everywhere; some difference is settled in a single dark thing.

The scale, laid out in one verse, and it is a genuine discovery. Syamesu kesucit kincit: *some* general in *some* dark things — a middle-level kind: cloud-dark, sword-dark. Kincit sarvatra vartate samanyam: *some* general operates *everywhere* — darkness as such. Kascid ekasmin syame bhedo vyavasthitah: *some* *difference* is *settled* in a *single* dark thing — this shade, here, and nowhere else. So generality is not two-valued: between the universal and the individual there is a graded range of intermediate kinds, and language has words for almost none of them. The comparison-compound is the device for reaching the middle rungs (3.14.369's kesucit). This is a very fine piece of semantics and it is stated in twenty syllables.

3.14.371 unformalized

तथाहि सति सौरभ्ये भेदो जात्युत्पलादिषु ।
गन्धानां सति भेदे तु सादृश्यमुपलभ्यते ॥ ३७१ ॥

tathāhi sati saurabhye bhedo jātyutpalādiṣu |
gandhānāṃ sati bhede tu sādṛśyamupalabhyate || 371 ||

So indeed: fragrance being present, there is a difference of scents among jati-flowers, lotuses and the rest; but the difference being present, a similarity is apprehended.

3.14.370's scale confirmed from another sense. Tatha hi sati saurabhye bhedo jaty-utpaladisu gandhanam: *so indeed* — fragrance being *present*, there is a *difference of scents* among *jati*-flowers, *lotuses* and the rest: they are all fragrant and no two smell alike. Sati bhede tu sadrsyam upalabhyate: *but* the difference being *present*, a *similarity* is *apprehended*. So the differences are real and one perceives, across them, a resemblance — not the bare universal 'fragrance' but a family likeness among these scents. The example is well chosen: nobody can say what jasmine and lotus have in common, everybody smells it, and no word names it (3.14.369's asabda-vacya). The middle rung is where the nose lives.

3.14.372 unformalized

गुणानामाश्रयाद् भेदः स्वतो वार्यनुगम्यते ।
अनिर्देश्याद् विशेषाद् वा सङ्कराद्वा गुणान्तरैः ॥ ३७२ ॥

guṇānāmāśrayād bhedaḥ svato vāryanugamyate |
anirdeśyād viśeṣād vā saṅkarādvā guṇāntaraiḥ || 372 ||

The difference of qualities is followed out from the substratum, or from itself, or from an indescribable particular, or from mixture with other qualities.

Four sources for the differences of 3.14.371. Gunanam asrayad bhedah: the *difference of qualities* from the *substratum* — the cloud's dark differs because clouds do. Svato va: *or from itself* — the quality is intrinsically this one. Anirdesyad visesad va: *or* from an *indescribable particular* — anirdesya: not statable; the shade that is simply this shade and has no account. Sankarad va gunantaraih: *or* from *mixture* with *other qualities* — the dark is what it is because of what it is blended with. So the sources are catalogued and not ranked, and one of them is confessedly indescribable. This is the samuddesa being honest about where its analysis stops: some differences have no reason, and one names the fact rather than the reason.

3.14.373 unformalized

उपमानं प्रसिद्धत्वात् सर्वत्र व्यतिरिच्यते ।
उपमेयत्वमाधिक्ये साम्ये वा न निवर्तते ॥ ३७३ ॥

upamānaṃ prasiddhatvāt sarvatra vyatiricyate |
upameyatvamādhikye sāmye vā na nivartate || 373 ||

The comparans, from its being well-known, exceeds everywhere; the state of being compared does not withdraw in excess or in equality.

Upamanam prasiddhatvat sarvatra vyatiricyate: the *comparans*, from its being *well-known*, *exceeds* *everywhere* — prasiddha (3.14.360's word): the cloud is famously dark, and its darkness always outdoes what is compared to it. Vyatiricyate: it exceeds, stands over. Upameyatvam adhikye samye va na nivartate: the *state of being compared* does *not withdraw* in *excess* or in *equality*. So even where the compared thing matches or exceeds the standard, it remains the compared thing: 'moon-faced' does not stop being a simile if the face is lovelier than the moon. The direction of comparison is fixed by fame, not by degree. This is why 3.14.359's mana and 3.14.360's upamana differ in kind: measures track quantity and comparisons track renown.

3.14.374 unformalized

अन्यैस्तु मानं जात्यादि भेद्यस्यार्थस्य वर्ण्यते ।
अनिर्ज्ञातस्वरूपो हि ज्ञेयोऽर्थस्तेन मीयते ॥ ३७४ ॥

anyaistu mānaṃ jātyādi bhedyasyārthasya varṇyate |
anirjñātasvarūpo hi jñeyo'rthastena mīyate || 374 ||

By others, the universal and the rest is described as the measure of the meaning to be differentiated; for the knowable meaning, whose own form is unknown, is measured by it.

Anyais tu manam jaty-adi bhedyasya arthasya varnyate: by *others*, the *universal* and the rest is described as the *measure* of the meaning *to be differentiated* — the jati as a mana (3.14.359), measuring its instances. Anirjnata-svarupo hi jneyo 'rtho tena miyate: for the *knowable* meaning, whose *own form* is *unknown*, is *measured by it*. So the individual, whose nature is not known in itself (3.14.350's anirupana: the principal has no own-form), is measured by the kind it falls under. This is a real position: universals are the measuring-instruments of particulars. And it makes 3.14.360's upamana a defective mana — which the following verses will resist.

3.14.375-376 unformalized

मितस्तु स्वेन मानेन प्रसिद्धो यो गुणाश्रयः ।
ऐश्रयान्तरमानाय स्वधर्मेण प्रवर्तते ॥ ३७५ ॥
रूपान्तरेण संस्पर्शो रूपान्तरवतां सताम् ।
भिन्नेन यस्य भेद्यानामुपमानं तदुच्यते ॥ ३७६ ॥

mitastu svena mānena prasiddho yo guṇāśrayaḥ |
aiśrayāntaramānāya svadharmeṇa pravartate || 375 ||
rūpāntareṇa saṃsparśo rūpāntaravatāṃ satām |
bhinnena yasya bhedyānāmupamānaṃ taducyate || 376 ||

(375) But the quality-substratum which, measured by its own measure, is well-known — operates by its own property for the measuring of another substratum. (376) The contact, by a different form, of things possessing different forms, of things to be differentiated — that is called comparison.

(375): mitas tu svena manena prasiddho yo gunasrayah — the *quality-substratum* which, *measured by its own measure*, is *well-known*: the cloud has been measured already, by whatever measures clouds, and is famous. Asrayantara-manaya sva-dharmena pravartate: it *operates by its own property* for the *measuring of another substratum*. So the comparans is a thing that has finished being known and is now lent out as an instrument. (376): rupantarena samsparso rupantaravatam satam bhinnena yasya bhedyanam — the *contact*, by a *different form*, of *things possessing different forms*, of things *to be differentiated*. Upamanam tad ucyate: *that* is called *comparison*. So: two things of different forms, touched by way of a form that is other than both — that is a simile. Samsparsa (3.14.354's touch) is the word: comparison grazes, and 3.14.359's mana grips.

3.14.377 unformalized

धर्मः समानः श्यामादिरुपमानोपमेययोः ।
आश्रीयमाणप्राधान्यो धर्मेणान्येन भिद्यते ॥ ३७७ ॥

dharmaḥ samānaḥ śyāmādirupamānopameyayoḥ |
āśrīyamāṇaprādhānyo dharmeṇānyena bhidyate || 377 ||

The common property 'dark' and the rest, of the comparans and the comparandum, whose principality is resorted to, is differentiated by another property.

Dharmah samanah syamadir upamana-upameyayoh: the *common property* — *'dark'* and the rest — of *comparans and comparandum*. Asriyamana-pradhanyah: whose *principality* is *resorted to* — asriyamana: being taken up; the property's principality is conferred by the taking, not owned (3.14.351's the quality attains principality at the moment of differentiating). Dharmena anyena bhidyate: is *differentiated* by *another property*. So the common property is made principal and then cut by a further property — the middle rung of 3.14.370, reached by a second cut. Two operations, and the simile is their product: elevate the shared quality, then differentiate it. Neither alone gives a comparison.

3.14.378 unformalized

शस्त्रीकुमार्योः सदृशः श्याम इत्येवमाश्रिते ।
व्यपदेश्यमनेनेति निमित्तं गुणयोः स्थितम् ॥ ३७८ ॥

śastrīkumāryoḥ sadṛśaḥ śyāma ityevamāśrite |
vyapadeśyamaneneti nimittaṃ guṇayoḥ sthitam || 378 ||

When it is resorted to thus — 'the dark of the knife and of the girl is similar' — the cause 'it is to be designated by this' stands in the two qualities.

Sastri-kumaryoh sadrsah syama ity evam asrite: when it is *resorted to* thus — *'the dark of the knife and of the girl is similar'*: the comparison spelled out as a statement of resemblance. Vyapadesyam anena iti nimittam gunayoh sthitam: the *cause* — *'it is to be designated by this'* — *stands* in the *two qualities*. So what the two darknesses share is not a property but a *designability*: each is such that the other may name it. The nimitta is a fitness-to-be-designated, lodged in the qualities themselves. This is exact and it sidesteps 3.14.363's dilemma: the shared thing need not be one property or two, because it is not a property — it is a licence to name.

3.14.379 unformalized

यदा निमित्तैस्तद्वन्तो गच्छन्तीव तदात्मताम् ।
भेदाश्रयं तदाख्यानमुपमानोपमेययोः ॥ ३७९ ॥

yadā nimittaistadvanto gacchantīva tadātmatām |
bhedāśrayaṃ tadākhyānamupamānopameyayoḥ || 379 ||

When the possessors of those, through the causes, go as it were to identity with them — that statement of comparans and comparandum rests on difference.

Yada nimittais tadvanto gacchanti iva tad-atmatam: when the *possessors of those* — the knife and the girl — through the *causes*, go *as it were* to *identity* with them: iva again. The girl becomes, as it were, the knife's darkness. Bhedasrayam tad akhyanam upamana-upameyayoh: *that statement* of comparans and comparandum *rests on difference*. So the identity is figurative and the statement is grounded in difference: the two must stay apart for the going-toward to mean anything (3.14.363's ekatve na upamanasya sambhavah). The iva is not decoration — it is what keeps the simile from collapsing into an identification. Which is what 3.14.380 is about.

3.14.380 unformalized

तत्त्वासङ्गविवक्षायां येषु भेदो निवर्तते ।
लुप्तोपमानि तान्याहुस्तद्धर्मेण समाश्रयात् ॥ ३८० ॥

tattvāsaṅgavivakṣāyāṃ yeṣu bhedo nivartate |
luptopamāni tānyāhustaddharmeṇa samāśrayāt || 380 ||

Those in which the difference withdraws, when identity-attachment is intended to be expressed — those they call 'elliptical comparisons', from the resort by way of that property.

The figure named. Tattva-asanga-vivaksayam yesu bhedo nivartate: those in which the *difference withdraws*, when *identity-attachment* is *intended to be expressed* — tattva-asanga: clinging to the very thing, the wish to say she *is* that darkness and not merely like it. Lupta-upamani tany ahuh: those they call *'lupta-upama'* — the *elided comparison*: 'sastri-syama', knife-dark, with the 'iva' gone. Tad-dharmena samasrayat: from the *resort* by way of *that property*. So dropping the 'iva' is not abbreviation: it answers to a different intention. Where 3.14.379's iva held the two apart, here the speaker wants them together and the marker of difference goes. The compound is chosen because it says the identity the phrase could not (3.14.236's the bahuvrihi is desired when his principality is intended).

3.14.381 unformalized

शस्त्र्यां प्रसिद्धं श्यामत्वं मानं सा तेन मीयते ।
अन्याश्यामातु तद्रूपा (त्वतद्रूपा) तेनात्यन्तं न मीयते ॥ ३८१ ॥

śastryāṃ prasiddhaṃ śyāmatvaṃ mānaṃ sā tena mīyate |
anyāśyāmātu tadrūpā (tvatadrūpā) tenātyantaṃ na mīyate || 381 ||

The darkness well-known in the knife is the measure; she is measured by it. But another dark thing, having that form, is not measured by it completely.

Sastryam prasiddham syamatvam manam sa tena miyate: the *darkness* *well-known* in the *knife* is the *measure*; *she* is *measured by it* — 3.14.374's proposal applied: the famous dark measures the unfamous. Anya syama tu tad-rupa tena atyantam na miyate: but *another* dark thing, *having that form*, is *not* measured by it *completely* — atyantam: utterly, without remainder (3.14.359's sakalyena). So even where the measuring works, it does not exhaust: another dark thing of the same form is left partly unmeasured. And that remainder is precisely 3.14.360's definition of comparison — the shortfall is not a failure of the measure but the nature of the thing.
contested: Pada c is transmitted with a variant: 'tad-rupa' ('having that form') against 'atad-rupa' ('not having that form'). We translate the former, on which the verse says that even a like-formed thing escapes complete measurement — which is the stronger point and matches 3.14.360's definition. On the variant the claim is the weaker one that unlike things are not measured at all.

3.14.382 unformalized

शस्त्रीं स्वेन गुणेनातो मिमानामाश्रयमान्तरम् ।
असमाप्तगुणं सिद्धेरुपमानं प्रचक्षते ॥ ३८२ ॥

śastrīṃ svena guṇenāto mimānāmāśrayamāntaram |
asamāptaguṇaṃ siddherupamānaṃ pracakṣate || 382 ||

Therefore they call the comparans that other substratum which measures the knife by its own quality — whose quality is incomplete as regards establishment.

The definition closed. Sastrim svena gunena atah mimanam asrayam antaram: that *other substratum* which *measures* the *knife* by its *own quality*. Asamapta-gunam siddheh upamanam pracaksate: they call the *comparans* that whose *quality* is *incomplete* as regards *establishment* — asamapta-guna: its quality not finished, not brought to completion. So the comparans is defined by its incompleteness: it measures, and its measuring never finishes (3.14.381's atyantam na miyate). A mana completes and an upamana does not, and the samuddesa has now said this three times in different words (3.14.360, 3.14.381, here) because it is the whole distinction. A simile is a measure that has agreed not to finish.

3.14.383 unformalized

उपमेये स्थितो धर्मः श्रुतोऽन्यत्रानुमीयते ।
श्रुतोऽथवोपमानस्थ उपमेयेऽनुमीयते ॥ ३८३ ॥

upameye sthito dharmaḥ śruto'nyatrānumīyate |
śruto'thavopamānastha upameye'numīyate || 383 ||

The property residing in the comparandum, being heard, is inferred elsewhere; or, heard as residing in the comparans, it is inferred in the comparandum.

Upameye sthito dharmah sruto 'nyatra anumiyate: the *property* residing in the *comparandum*, being *heard*, is *inferred* *elsewhere*. Sruto 'thava upamana-stha upameye 'numiyate: *or*, *heard* as residing in the *comparans*, it is *inferred* in the *comparandum*. So the simile runs in either direction: what is said of one term is inferred of the other, and which term gets the saying is a matter of arrangement. Anumiyate: *inferred* — and the word matters. A comparison does not state the shared property twice; it states it once and leaves the other occurrence to be worked out. That is the mechanism of 3.14.360's incompleteness.

3.14.384 unformalized

अधीयते ब्राह्मणवत् क्षत्रिया दृश्यते ।
उपमेयस्य भिन्नत्वाद् वचनं क्षत्रियाश्रयम् ॥ ३८४ ॥

adhīyate brāhmaṇavat kṣatriyā dṛśyate |
upameyasya bhinnatvād vacanaṃ kṣatriyāśrayam || 384 ||

'The ksatriyas study like brahmins' is seen; from the comparandum's difference, the number rests on the ksatriyas.

Adhiyate brahmanavat ksatriya drsyate: *'the ksatriyas study like brahmins'* is *seen* — 'brahmanavat', with the comparison-suffix. Upameyasya bhinnatvad vacanam ksatriyasrayam: from the *comparandum's difference*, the *number* rests on the *ksatriyas*. So the verb's plural answers to the ksatriyas and not to the brahmins, though the brahmins are what is named: the compared thing carries the concord, and the standard is inert. This confirms 3.14.379's bhedasrayam: the comparison rests on difference, and the difference shows up in the grammar. The standard lends its fame (3.14.373's prasiddhatvat) and nothing else.
contested: Pada a is metrically short in our source ('adhiyate brahmanavat ksatriya drsyate'), a syllable or more having apparently dropped. The construction and the point — that the verb agrees with the ksatriyas — are unaffected.

3.14.385 unformalized

साधारणं ब्रुवन् धर्मं क्वचिदेव व्यवस्थितम् ।
सामान्यवचनः शब्द इति सूत्रेऽपदिश्यते ॥ ३८५ ॥

sādhāraṇaṃ bruvan dharmaṃ kvacideva vyavasthitam |
sāmānyavacanaḥ śabda iti sūtre'padiśyate || 385 ||

Stating a common property that is settled only somewhere — 'a word expressing the general' is what the sutra says.

Sadharanam bruvan dharmam kvacid eva vyavasthitam: *stating* a *common property* *settled only somewhere* — kvacid eva: in some places only, 3.14.370's middle rung, the general that is neither everywhere nor in one. Samanya-vacanah sabda iti sutre 'padisyate: *'a word expressing the general'* is *stated in the sutra*. So Panini's phrase 'samanya-vacana' names exactly the thing 3.14.369-370 had to construct: a word whose general is a middle one. The grammar had the category and Bhartrhari has supplied its semantics. This is the pattern of the whole samuddesa — the rules were right and unexplained, and the explanation is the work.

3.14.386 unformalized

नाभेदेन न भेदेन गुणो द्विष्ठोऽभिधीयते ।
भिन्नयोर्धर्मयोरेकः श्रूयतेऽन्यः प्रतीयते ॥ ३८६ ॥

nābhedena na bhedena guṇo dviṣṭho'bhidhīyate |
bhinnayordharmayorekaḥ śrūyate'nyaḥ pratīyate || 386 ||

A quality residing in two is expressed neither by non-difference nor by difference; of the two different properties, one is heard, the other is cognized.

The resolution of 3.14.363's dilemma, and it is exact. Na abhedena na bhedena guno dvistho 'bhidhiyate: a *quality residing in two* is expressed *neither by non-difference nor by difference* — both horns declined, in one line. Bhinnayor dharmayor ekah sruyate 'nyah pratiyate: of the *two different properties*, *one* is *heard*, the *other* is *cognized*. So the properties are two — that much is conceded to the difference-horn — and the comparison does not express both: it says one and lets the other be understood (3.14.383's anumiyate). The oneness is never asserted, so the identity-horn is never taken. The simile's whole trick is that it says half and the hearer supplies half, and the halves are different properties.

3.14.387 unformalized

नात्यन्ताय मिमीते यत् सामान्ये समवस्थितम् ।
सादृश्यादुपमेयार्थः समीपे परिकल्प्यते ॥ ३८७ ॥

nātyantāya mimīte yat sāmānye samavasthitam |
sādṛśyādupameyārthaḥ samīpe parikalpyate || 387 ||

That which, settled in the general, does not measure completely — from similarity, the comparandum's meaning is imagined in proximity.

Na atyantaya mimite yat samanye samavasthitam: that which, *settled in the general*, does *not* measure *completely* — 3.14.381's atyantam na miyate, now as a definition. Sadrsyad upameyarthah samipe parikalpyate: from *similarity*, the *comparandum's meaning* is *imagined* *in proximity* — samipe: *near by*; parikalpyate: constructed (3.14.75's word). So the comparison does not place the compared thing *at* the standard but *near* it, and the placing is a construction. Samipa is the whole of what a simile achieves: not identity, not measure, but nearness — and the nearness is imagined.

3.14.388 unformalized

मानं प्रति समीपं वा सादृश्येन प्रतीयते ।
परिच्छेदाद्धि सादृश्यमिह मानोपमानयोः ॥ ३८८ ॥

mānaṃ prati samīpaṃ vā sādṛśyena pratīyate |
paricchedāddhi sādṛśyamiha mānopamānayoḥ || 388 ||

Or the proximity to the measure is cognized by similarity; for from delimitation there is here a similarity of measure and comparison.

Manam prati samipam va sadrsyena pratiyate: the *proximity* to the *measure* is cognized by *similarity* — one knows how near one has come to a full measure by how like it the case is. Paricchedad dhi sadrsyam iha mana-upamanayoh: for from *delimitation* there is *here* a *similarity* of *measure and comparison* — pariccheda: the cutting-off (3.14.126's word). So measure and comparison are themselves alike, because both delimit — and the samuddesa's whole distinction between them (3.14.359-360's sakalyena and its absence) is a difference of degree within one operation. The verse quietly softens its own dichotomy: a mana is an upamana that finished.

3.14.389 unformalized

एकजातिव्यपेक्षायां तदेवेत्यवसीयते ।
भेदस्यैव व्यपेक्षायामन्यदेवेति गम्यते ॥ ३८९ ॥

ekajātivyapekṣāyāṃ tadevetyavasīyate |
bhedasyaiva vyapekṣāyāmanyadeveti gamyate || 389 ||

When one universal is required, 'it is that very thing' is determined; when only the difference is required, 'it is something else' is understood.

Eka-jati-vyapeksayam tad eveti avasiyate: when *one universal* is *required*, *'it is that very thing'* is *determined*. Bhedasya eva vyapeksayam anyad eveti gamyate: when *only the difference* is required, *'it is something else'* is understood. So identity and otherness are not found but *required*: which one obtains depends on what the cognition is reaching for (vyapeksa: the reaching). The same two things are the same thing and different things according to what one wants of them. This is 3.14.367's vyatikirna — the intermingled reflection — stated as a rule of thumb: identity and difference are two ways of needing.

3.14.390 unformalized

कर्मत्वं करणत्वं च भेदेनैवाश्रितं यतः ।
अत्यन्तैकत्वविषयान्न स्यात् तेनात्र संशयः ॥ ३९० ॥

karmatvaṃ karaṇatvaṃ ca bhedenaivāśritaṃ yataḥ |
atyantaikatvaviṣayānna syāt tenātra saṃśayaḥ || 390 ||

Since object-hood and instrument-hood are resorted to only by way of difference — from the domain of utter oneness there would be no doubt here.

Karmatvam karanatvam ca bhedena eva asritam yatah: since *object-hood* and *instrument-hood* are *resorted to* *only by way of difference* — the karakas require distinct terms: one does not cut a thing with itself (3.7's the sadhanas as distinguished capacities). Atyanta-ekatva-visayan na syat tena atra samsayah: from the *domain of utter oneness*, there would be *no doubt* here. So where oneness is complete, no karaka-question can even arise: doubt about what is object and what is instrument presupposes two. The verse is a small proof that 3.14.389's 'requiring the difference' is not optional in some places — the karakas need it constitutively.

3.14.391 unformalized

भेदेन तुल्यरूपत्वाच्छालींस्तानिति दृश्यते ।
जात्यभेदात् स एवायमिति भिन्नोऽभिधीयते ॥ ३९१ ॥

bhedena tulyarūpatvācchālīṃstāniti dṛśyate |
jātyabhedāt sa evāyamiti bhinno'bhidhīyate || 391 ||

From likeness of form by way of difference, 'those rice-grains' is seen; from non-difference of universal, 'this is that very one' — the different is expressed.

Bhedena tulya-rupatvac chalims tani iti drsyate: from *likeness of form* by way of *difference*, *'those rice-grains'* is *seen* — the plural: many grains, alike, distinct. Jaty-abhedat sa eva ayam iti bhinno 'bhidhiyate: from *non-difference of universal*, *'this is that very one'* — the *different* is *expressed*. So of the same grains one says 'those, plural, alike' and 'this is the very same' — and the second says a *different* thing while asserting identity. Bhinno 'bhidhiyate: what is expressed is the different one, under the form of identity. This is 3.14.389 exhibited: the grains oblige either requirement, and the language has both idioms ready.

3.14.392 unformalized

कथं ह्यवयवोऽन्यस्य स्यादन्य इत् चोच्यते ।
अत्यन्तभेदे नानात्वं यत्र तत्त्वं न विद्यते ॥ ३९२ ॥

kathaṃ hyavayavo'nyasya syādanya it cocyate |
atyantabhede nānātvaṃ yatra tattvaṃ na vidyate || 392 ||

For how could a member of another be, and yet be called 'other'? In utter difference there is manifoldness, where identity does not exist.

Katham hy avayavo 'nyasya syad anya iti ca ucyate: *how* could a *member of another* be, and yet be *called 'other'*? — a part belongs to its whole and is other than it, and both cannot be flatly true. Atyanta-bhede nanatvam yatra tattvam na vidyate: in *utter difference* there is *manifoldness*, *where* *identity* does *not exist*. So if the parts were utterly other, there would be many and no whole at all — which is not what one finds. The question is left standing as a question (katham hi), and its point is that partaking will not reduce to identity or to difference. This is 3.14.367's vyatikirna again, met in mereology.

3.14.393 unformalized

अभेदस्य विवक्षायामेकत्वं सङ्घसङ्घिनोः ।
सङ्घिनोर्न त्वभेदोऽस्ति तथान्यत्वमुदाहृतम् ॥ ३९३ ॥

abhedasya vivakṣāyāmekatvaṃ saṅghasaṅghinoḥ |
saṅghinorna tvabhedo'sti tathānyatvamudāhṛtam || 393 ||

When non-difference is intended, there is oneness of the aggregate and the aggregated; but there is no non-difference of the aggregated ones among themselves — thus otherness has been illustrated.

The answer to 3.14.392. Abhedasya vivaksayam ekatvam sangha-sanghinoh: when *non-difference* is *intended*, there is *oneness* of the *aggregate* and the *aggregated* — the forest and its trees are one, when one means them so. Sanghinor na tv abhedo 'sti: but there is *no* non-difference of the *aggregated ones* *among themselves* — the trees are not one another. Tatha anyatvam udahrtam: *thus* *otherness* has been *illustrated*. So the part is one with the whole and other than its fellows, and 3.14.392's paradox dissolves once one asks *other than what*. The whole absorbs, the siblings repel, and vivaksa selects which relation is in view (3.14.389).

3.14.394 unformalized

तत्राभिन्नव्यपेक्षायामुपमार्थो न विद्यते ।
यो हि गोरिति विज्ञाने हेतुः सोऽस्ति गवान्तरे ॥ ३९४ ॥

tatrābhinnavyapekṣāyāmupamārtho na vidyate |
yo hi goriti vijñāne hetuḥ so'sti gavāntare || 394 ||

There, when the non-different is required, there is no purpose of comparison; for the cause which is in the cognition 'cow' exists in another cow too.

Tatra abhinna-vyapeksayam upamartho na vidyate: when the *non-different* is *required*, there is *no purpose* of *comparison* — 3.14.363's first horn, and 3.14.364's tautology. Yo hi gor iti vijnane hetuh so 'sti gavantare: for the *cause* which is in the cognition *'cow'* exists in *another cow* too. So one does not compare a cow to a cow: whatever makes this one a cow makes that one a cow, and there is nothing to say. Comparison needs the terms to differ, and where the universal is what is required, they do not. The verse re-establishes 3.14.379's bhedasrayam from the other side.

3.14.395 unformalized

व्यावृत्तानां विशेषाणां व्यापारे तु विवक्षिते ।
न कश्चिदुपकारोऽस्ति बुद्धेर्बुद्ध्यन्तरं प्रति ॥ ३९५ ॥

vyāvṛttānāṃ viśeṣāṇāṃ vyāpāre tu vivakṣite |
na kaścidupakāro'sti buddherbuddhyantaraṃ prati || 395 ||

But when the operation of the excluded particulars is intended — there is no assistance at all of one cognition toward another.

The other horn, and it is worse. Vyavrttanam visesanam vyapare tu vivaksite: when the *operation* of the *excluded* particulars is *intended* — vyavrtta: turned away, mutually excluding (3.14.141's vinivartaka). Na kascid upakaro 'sti buddher buddhy-antaram prati: there is *no assistance at all* of *one cognition* toward *another*. Upakara: help (3.14.333's word). So where the particulars exclude one another, the cognition of one contributes nothing to the cognition of the other, and again no comparison is possible. Between 3.14.394 and this verse both horns are closed for the second time — and comparison happens. So neither pure identity nor pure difference is ever what is required: 3.14.367's vyatikirna is.

3.14.396 unformalized

किञ्चिद् यत्रास्ति सामान्यं यदि भेदाश्च केचन ।
गोत्वं गोष्वस्ति सामान्यं भेदाश्च शबलादयः ॥ ३९६ ॥

kiñcid yatrāsti sāmānyaṃ yadi bhedāśca kecana |
gotvaṃ goṣvasti sāmānyaṃ bhedāśca śabalādayaḥ || 396 ||

If somewhere there is some general and there are some differences — cowness is the general in cows, and the differences are 'dappled' and the rest.

Kincid yatra asti samanyam yadi bhedas ca kecana: *if* *somewhere* there is *some* general and *some* differences — the condition comparison needs, per 3.14.394-395: neither general alone nor differences alone. Gotvam gosv asti samanyam bhedas ca sabaladayah: *cowness* is the *general* in *cows*, and the *differences* are *'sabala'* (dappled) and the rest. So the cow-case has both: one cowness, many coats. And that is the structure a comparison requires — a shared kind carrying real differences. The verse is a specification, not an example: it says what must be present for a simile to have work to do.

3.14.397 unformalized

सामान्यं श्यामताद्येव तद्धि साधारणं द्वयोः ।
तदेव सिद्ध्यसिद्धिभ्यां भेद इत्यपदिश्ते ॥ ३९७ ॥

sāmānyaṃ śyāmatādyeva taddhi sādhāraṇaṃ dvayoḥ |
tadeva siddhyasiddhibhyāṃ bheda ityapadiśte || 397 ||

The general is just darkness and the rest; for that is common to the two. That very thing, by establishment and non-establishment, is designated as the difference.

Samanyam syamatady eva tad dhi sadharanam dvayoh: the *general* is *just darkness* and the rest; for *that* is *common* to the *two* — the shared property, and nothing more elevated. Tad eva siddhy-asiddhibhyam bheda ity apadisyate: *that very thing*, by *establishment* and *non-establishment*, is *designated* as the *difference*. Siddhi and asiddhi: being made out and not being made out — the darkness is established in the knife (prasiddha: 3.14.373) and not established in the girl. So the same darkness is the general *and* the difference, according to whether it is made out or not. That is a remarkable claim: the difference between the two darks is not a further property but the difference between being famous and being obscure.

3.14.398 unformalized

श्यामत्वमेव सामान्यमन्येषामुभयोः स्थितम् ।
सम्पूर्णत्वात् तदन्यस्माद् विशेष इति गम्यते ॥ ३९८ ॥

śyāmatvameva sāmānyamanyeṣāmubhayoḥ sthitam |
sampūrṇatvāt tadanyasmād viśeṣa iti gamyate || 398 ||

For others, darkness itself is the general, standing in both; from its completeness, it is understood as the particular as distinct from that other.

Syamatvam eva samanyam anyesam ubhayoh sthitam: for *others*, *darkness itself* is the *general*, *standing in both*. Sampurnatvat tad anyasmad visesa iti gamyate: from its *completeness*, it is understood as the *particular*, as distinct *from that other* — sampurna: full, complete: the darkness that is fully realized in the knife is thereby a *particular*, marked off from the darkness that is not. So the same doctrine as 3.14.397 in another vocabulary: the general becomes a particular by being complete. Not by acquiring a differentia — by being fully there. This is as unusual an account of the general-particular relation as the tradition offers, and the comparison-compound is where Bhartrhari found it.

3.14.399 unformalized

आकृतौ वापि सामान्ये क्वचिदेव व्यवस्थिताः ।
श्यामादयोऽवसीयन्ते विशेषास्त इहाश्रिताः ॥ ३९९ ॥

ākṛtau vāpi sāmānye kvacideva vyavasthitāḥ |
śyāmādayo'vasīyante viśeṣāsta ihāśritāḥ || 399 ||

Or, settled only somewhere in the form, in the general — 'dark' and the rest are determined; those particulars are what are resorted to here.

Akrtau vapi samanye kvacid eva vyavasthitah: settled *only somewhere* in the *akrti*, in the *general* — kvacid eva again (3.14.370, 3.14.385): the middle rung. Syamadayo 'vasiyante visesas ta iha asritah: *'dark'* and the rest are *determined*; *those particulars* are what are *resorted to* here. So on this account 'dark' is already a particular relative to the form it sits in — and the comparison resorts to *it*, not to darkness at large. The vocabulary shifts (akrti for jati) and the finding is 3.14.370's: what a simile reaches for lives between the universal and the individual, and the language has only compounds to reach it with.

3.14.400 unformalized

जातेरभेदे भेदे वा सादृश्यं तत् प्रचक्षते ।
कश्चित् तदाचिदर्थात्मा तथाभूतोऽपदिश्यते ॥ ४०० ॥

jāterabhede bhede vā sādṛśyaṃ tat pracakṣate |
kaścit tadācidarthātmā tathābhūto'padiśyate || 400 ||

In the non-difference or the difference of the universal, they declare that similarity; some meaning-self, being such, is designated.

Jater abhede bhede va sadrsyam tat pracaksate: in the *non-difference* *or* the *difference* of the *universal*, they *declare* that *similarity* — abhede bhede va: either way, and the va is doing real work: the theorists differ and the similarity is claimed on both accounts. Kascit... arthatma tathabhuto 'padisyate: *some* *meaning-self*, being *such*, is *designated*. So resemblance is asserted from both sides of the great division, and what is designated is a meaning-self of some sort — kascit: some, indeterminate. Bhartrhari is registering that sadrsya survives every account of the universal because it is not built out of universals at all.
contested: Pada c is transmitted as 'kascit tadacid arthatma', in which 'tadacid' will not construe. We translate from the remainder ('kascid arthatma tathabhuto 'padisyate'), whose sense is clear, and leave the corrupt element unrendered rather than emend on conjecture.

3.14.401 unformalized

यत्रार्थे प्रत्ययाभेदो न कदाचिद् विकल्पते ।
अविद्यमानभेदत्वात् स एक इति गम्यते ॥ ४०१ ॥

yatrārthe pratyayābhedo na kadācid vikalpate |
avidyamānabhedatvāt sa eka iti gamyate || 401 ||

That meaning in which the non-difference of cognition never varies — from the non-existence of difference, it is understood as one.

Yatra arthe pratyaya-abhedo na kadacid vikalpate: that *meaning* in which the *non-difference of cognition* *never* *varies* — na kadacit: at no time; vikalpate: wavers, admits an alternative. Avidyamana-bhedatvat sa eka iti gamyate: from the *non-existence of difference*, it is *understood as one*. So oneness is a fact about cognition's steadiness, not about the object: what one never cognizes as two, one calls one. The criterion is epistemic through and through (3.14.188's buddhya eva), and it is the honest form of what identity comes to in this work: the absence of a wavering.

3.14.402 unformalized

योऽर्थ आश्रितनानात्वः स एवेत्यपदिश्यते ।
व्यापारं जातिभागस्य तत्रापि प्रतिजानते ॥ ४०२ ॥

yo'rtha āśritanānātvaḥ sa evetyapadiśyate |
vyāpāraṃ jātibhāgasya tatrāpi pratijānate || 402 ||

That meaning which has resorted to manifoldness is designated as 'that very one'; there too they assert the operation of the universal-portion.

Yo 'rtha asrita-nanatvah sa evety apadisyate: that *meaning* which has *resorted to manifoldness* is *designated* as *'that very one'* — the thing has taken up manyness and is still called the same (3.14.391's sa eva ayam of the rice-grains). Vyaparam jati-bhagasya tatrapi pratijanate: *there too* they *assert* the *operation* of the *universal-portion*. So even where manyness has been resorted to, a jati-portion is at work holding the 'same' together. Bhaga: portion — not a universal but a *part* of the meaning that does universal-work. The samuddesa's characteristic move: not two entities but one meaning with sides.

3.14.403 unformalized

जातिभागाश्रया प्रख्या तत्राभिन्ना प्रवर्तते ।
व्यक्तिभागाश्रया बुद्धिस्तत्र भेदेन जायते ॥ ४०३ ॥

jātibhāgāśrayā prakhyā tatrābhinnā pravartate |
vyaktibhāgāśrayā buddhistatra bhedena jāyate || 403 ||

The presentation resting on the universal-portion operates there as undivided; the cognition resting on the individual-portion arises there by difference.

The two portions of 3.14.402, each with its own cognition. Jati-bhagasraya prakhya tatra abhinna pravartate: the *presentation* resting on the *universal-portion* operates there as *undivided* — prakhya: the showing-forth (3.14.187's eka-prakhya-nibandhanah, of the general). Vyakti-bhagasraya buddhis tatra bhedena jayate: the *cognition* resting on the *individual-portion* *arises* there *by difference*. So one object, two portions, two cognitions — one showing it whole and one bringing it up divided. Both are of the same thing at the same time, which is why 3.14.367's reflection is vyatikirna, intermingled: the two cognitions are not stages, they are simultaneous and they overlap.

3.14.404-405 unformalized

अन्यत्र वर्तमानं सद् भेदाभेदसमन्वितम् ।
निमित्तं पुनरन्यत्र नानात्वेनेव गृह्यते ॥ ४०४ ॥
आधारेषु पदन्यासं कृत्वोपैति तदाश्रयम् ।
स सादृश्यस्य विषय इत्यन्यैरपदिश्यते ॥ ४०५ ॥

anyatra vartamānaṃ sad bhedābhedasamanvitam |
nimittaṃ punaranyatra nānātveneva gṛhyate || 404 ||
ādhāreṣu padanyāsaṃ kṛtvopaiti tadāśrayam |
sa sādṛśyasya viṣaya ityanyairapadiśyate || 405 ||

(404) The cause, existing in one place, endowed with difference and non-difference, is again grasped elsewhere as though by manifoldness. (405) Having planted a foot in the substrata, it approaches what rests on them; that, others declare, is the domain of similarity.

(404): anyatra vartamanam sad bhedabheda-samanvitam — the *cause*, existing *in one place*, endowed with *difference and non-difference* (3.14.209's bhedabheda). Nimittam punar anyatra nanatvena iva grhyate: is *again* grasped *elsewhere* *as though* by *manifoldness* — iva. (405): adharesu pada-nyasam krtva upaiti tad-asrayam — having *planted a foot* in the *substrata*, it *approaches* what *rests on them*. Pada-nyasam krtva: *having taken a step* — the same image as 3.14.262's padam krtva, where the intellect planted itself on a ksatriya in order to pull the brahmin away. Sa sadrsyasya visaya ity anyair apadisyate: *that*, *others* declare, is the *domain of similarity*. So resemblance lives where the cause, having stepped onto the seats, reaches past them. Similarity is a movement, not a property — and this samuddesa has said the same of negation.

3.14.406 unformalized

परापेक्षे यथा भावे कारणाख्या प्रवर्तते ।
तथान्याधिगमापेक्षमुपमानं प्रचक्षते ॥ ४०६ ॥

parāpekṣe yathā bhāve kāraṇākhyā pravartate |
tathānyādhigamāpekṣamupamānaṃ pracakṣate || 406 ||

As the designation 'cause' operates in a state that depends on another — so, they declare, comparison depends on the knowing of another.

Para-apekse yatha bhave karanakhya pravartate: as the designation *'cause'* operates in a *state* that *depends on another* — nothing is a cause in itself; it is a cause of something, and the title is relational through and through (3.3's relation; 3.7's the karakas as capacities conferred by the action). Tatha anya-adhigama-apeksam upamanam pracaksate: *so* they declare *comparison* to depend on the *knowing of another* — anya-adhigama: the getting-to-know of the other. So upamana is a relational title like 'cause': nothing is a comparans in itself, only for a knowing that needs it. This is why 3.14.373's fame decided direction — a standard is a standard for someone.

3.14.407 unformalized

गुरुशिष्यपितापुत्रक्रियाकालादयो यथा ।
व्यवहारास्तथौपम्यमप्यपेक्षानिबन्धनम् ॥ ४०७ ॥

guruśiṣyapitāputrakriyākālādayo yathā |
vyavahārāstathaupamyamapyapekṣānibandhanam || 407 ||

As the transactions 'teacher and pupil', 'father and son', 'action and time' and the rest — so comparison too is founded on dependence.

The list that makes 3.14.406 plain, and it is well chosen. Guru-sisya-pita-putra-kriya-kaladayo yatha vyavaharah: as the *transactions* — *teacher-pupil*, *father-son*, *action-time* and the rest: correlatives, each of which is nothing without its mate (3.14.47's sambandhi-sabdah sapekso nityam sarvah prayujyate: a relatum-word is *wholly* dependent). Tatha aupamyam apy apeksa-nibandhanam: *so* *comparison* too is *founded on dependence* — apeksa-nibandhana: its ground is the reaching. So a comparans is to its comparandum as a father to a son: not a thing with a property but a term in a reaching. And the last pair — kriya-kala, action and time — quietly recalls the Kala-samuddesa, where time was shown to be nothing but this kind of dependence.

3.14.408 unformalized

श्यामत्वमुपमाने चेद् वृत्तं वृत्तौ प्रयुज्यते ।
उपमेयं समासेन बाह्यं तत्राभिधीयते ॥ ४०८ ॥

śyāmatvamupamāne ced vṛttaṃ vṛttau prayujyate |
upameyaṃ samāsena bāhyaṃ tatrābhidhīyate || 408 ||

If the darkness which operated in the comparans is employed in the derivative formation — then the comparandum, being external, is expressed by the compound.

The rule-technical consequences begin. Syamatvam upamane ced vrttam vrttau prayujyate: if the *darkness* which *operated in the comparans* is *employed* in the *derivative formation* — the shared property enters the compound from the standard's side. Upameyam samasena bahyam tatra abhidhiyate: then the *comparandum*, being *external*, is *expressed by the compound*. Bahya: outside (3.14.207's bahyo nasty asrayo dvandve). So 'sastri-syama' would have its meaning outside itself — which makes it structurally a bahuvrihi (3.14.38's anya-padartha), with all that follows for accent and gender. The following verses test whether that is tolerable.

3.14.409 unformalized

टाबन्त एव चैत्रादौ श्यामाशब्दस्तथा भवेत् ।
सूत्रे च प्रथमाभावान्न श्यामाद्युपसर्जनम् ॥ ४०९ ॥

ṭābanta eva caitrādau śyāmāśabdastathā bhavet |
sūtre ca prathamābhāvānna śyāmādyupasarjanam || 409 ||

The word 'syama', ending in 'tap', would thus occur in 'Caitra' and the like; and from the absence of the nominative in the sutra, 'syama' and the rest is not the subordinate.

Tab-anta eva caitradau syama-sabdas tatha bhavet: the word *'syama'*, ending in *'tap'* — the feminine — would thus occur in *'Caitra'* and the like: a masculine name would take a feminine-marked member, which is wrong. Sutre ca prathama-abhavan na syamady-upasarjanam: and from the *absence of the nominative* in the *sutra*, *'syama'* and the rest is *not* the *subordinate* — Panini's definition makes upasarjana what stands in the nominative in the rule, and 'syama' does not. So 3.14.408's reading fails twice: it produces a bad form and it misidentifies the subordinate. The tests are purely formal and they are decisive.

3.14.410 unformalized

अथवैकविभक्तित्वाद् गुणत्वाद् वोपसर्जनम् ।
नैव तित्तिरिकल्माष्यामिष्टः स्त्रीप्रत्ययो भवेत् ॥ ४१० ॥

athavaikavibhaktitvād guṇatvād vopasarjanam |
naiva tittirikalmāṣyāmiṣṭaḥ strīpratyayo bhavet || 410 ||

Or, from having a single case-ending, or from being a quality, it is the subordinate; the desired feminine affix would not occur in 'tittiri-kalmasi'.

Athava eka-vibhaktitvad gunatvad va upasarjanam: *or*, from having a *single case-ending*, or from *being a quality*, it *is* the subordinate — two alternative criteria that would make 'syama' subordinate after all, against 3.14.409. Na eva tittiri-kalmasyam istah stri-pratyayo bhavet: the *desired feminine affix* would *not* occur in *'tittiri-kalmasi'* — partridge-speckled, of a cow. So the alternative criteria buy the subordination and lose the feminine. Every route through the rules costs something, and Bhartrhari is walking each one to where it breaks rather than declaring a winner. This is 3.14.267's finding in advance: the science settles such things by decision.

3.14.411 unformalized

सति शिष्टबलीयस्त्वाद् बाह्ये ङीषि च सत्यपि ।
उपमानस्वरो न स्यात् तस्मात् स्त्र्यन्तः समस्यते ॥ ४११ ॥

sati śiṣṭabalīyastvād bāhye ṅīṣi ca satyapi |
upamānasvaro na syāt tasmāt stryantaḥ samasyate || 411 ||

From the greater strength of what is taught while the other is present, and even where the external 'nis' is there — the comparans-accent would not occur; therefore the feminine-ending is compounded.

Sati sista-baliyastvat: from the *greater strength* of what is *taught while the other is present* — the sati-sista maxim (3.14.62, where it settled another accent-conflict). Bahye nisi ca saty api: even where the *external 'nis'* — the feminine affix — is *there*. Upamana-svaro na syat: the *comparans-accent* would *not* occur. Tasmat stry-antah samasyate: *therefore* the *feminine-ending* is *compounded*. So the accent-facts force the conclusion: since the comparans-accent does not appear, the feminine must have been inside the compound when it formed. The argument runs from phonology back to structure — exactly as at 3.14.62, and it is the samuddesa's most reliable instrument.

3.14.412 unformalized

गुणे न चोपमानस्थे सापेक्षत्वं प्रकल्पते ।
प्राधान्यस्य (प्रधानस्य) तथा न स्याद् व्याघ्रादौ लिङ्गदर्शनम् ॥ ४१२ ॥

guṇe na copamānasthe sāpekṣatvaṃ prakalpate |
prādhānyasya (pradhānasya) tathā na syād vyāghrādau liṅgadarśanam || 412 ||

And dependence is not conceivable in a quality residing in the comparans; thus the seeing of gender in 'tiger' and the rest would not belong to the principal.

Gune na ca upamana-sthe sapeksatvam prakalpate: *dependence* is *not conceivable* in a *quality residing in the comparans* — the standard's quality is complete in itself and reaches for nothing (3.14.373's it exceeds everywhere). Pradhanyasya tatha na syad vyaghradau linga-darsanam: thus the *seeing of gender* in *'vyaghra'* (tiger) and the rest would *not* belong to the *principal*. So in 'purusa-vyaghra', man-tiger, the gender one sees would not be the principal's — which is wrong. The difficulty comes from denying dependence to the comparans's quality, and it shows the denial is too strong.
contested: Pada c is transmitted with a variant: 'pradhanyasya' ('of the principality') against 'pradhanasya' ('of the principal'). We translate the latter's sense, which the argument about where the gender is seen requires. The variant is unlikely to be more than a scribal alternation.

3.14.413 unformalized

तस्मात् सति गुणत्वेऽपि प्राधान्यं विग्रहान्तरे ।
नैवञ्जतीयकं शास्त्रे सम्भवत्युपसर्जनम् ॥ ४१३ ॥

tasmāt sati guṇatve'pi prādhānyaṃ vigrahāntare |
naivañjatīyakaṃ śāstre sambhavatyupasarjanam || 413 ||

Therefore, even though it is a quality, there is principality in another analysis; a subordinate of this kind is not possible in the science.

Tasmat sati gunatve 'pi pradhanyam vigrahantare: *therefore*, *even though* it is a *quality*, there is *principality* *in another analysis* — vigraha-antara: a different resolution of the compound. So the same element is quality and principal, depending on which analysis one runs (3.14.351's the quality attains principality at the moment of differentiating). Na evanjatiyakam sastre sambhavaty upasarjanam: a *subordinate of this kind* is *not possible* in the *science*. So the grammar has no category for what this element is — a principal quality — and the analyses multiply because none of them fits. This is the honest end of the comparison-compound's rule-technical treatment: the thing is real and the science has no box for it (3.14.74's atidure).

3.14.414 unformalized

उपमेयात्मनि श्यामे वर्तमेनोऽभिधीयते ।
उपमानेष्वनिर्दिष्टः सामर्थ्यात् सम्प्रतीयते ॥ ४१४ ॥

upameyātmani śyāme vartameno'bhidhīyate |
upamāneṣvanirdiṣṭaḥ sāmarthyāt sampratīyate || 414 ||

Existing in the comparandum's self, 'dark' is expressed; not stated in the comparanda, it is understood from capacity.

Upameyatmani syame vartamano 'bhidhiyate: existing in the *comparandum's self*, *'dark'* is *expressed* — the property is said of the compared thing, which is where it was doubtful (3.14.397's asiddhi). Upamanesv anirdistah samarthyat sampratiyate: *not stated* in the *comparanda*, it is *understood from capacity* — samarthya: the words' own power to convey without saying (3.14.36's samarthya-matra; 2.13's what capacity carries). So the simile states the property of the obscure term and leaves it to be understood of the famous one — which is 3.14.386's exact division of labour (one is heard, the other is cognized), now assigned: the heard one belongs to the comparandum. The compound says what needed saying and trusts capacity with the rest.

3.14.415 unformalized

द्रव्यमात्रेऽपि निर्दिष्टे चन्द्रवक्त्रेऽनुगम्यते ।
विशिष्ट एव चन्द्रस्थो गुणो नोपप्लवादयः ॥ ४१५ ॥

dravyamātre'pi nirdiṣṭe candravaktre'nugamyate |
viśiṣṭa eva candrastho guṇo nopaplavādayaḥ || 415 ||

Even when only the substance is indicated, in 'moon-face' a specific quality residing in the moon is what follows — not eclipses and the rest.

A lovely observation, and it is decisive. Dravya-matre 'pi nirdiste candra-vaktre: even when *only the substance* is *indicated* — 'candra' names the moon, a thing, and no property is said. Anugamyate visista eva candra-stho gunah: a *specific quality* residing in the *moon* is what *follows* — visista eva: a *particular* one: the beauty, the coolness, the fullness. Na upaplavadayah: *not* *eclipses* and the rest. So 'moon-faced' does not mean her face is periodically swallowed, or spotted, or waning — though all of that is as much the moon's as its beauty is. The substance is named and one specific quality comes along and the rest, though equally present, does not. This is why the comparison-compound cannot be analysed: nothing in it selects, and the selection happens anyway.

3.14.416 unformalized

भेदभावनयैतच्च समासेऽप्युपवर्ण्यते ।
विशिष्टगुणभिन्नेऽर्थे पदमन्यत् प्रयुज्यते ॥ ४१६ ॥

bhedabhāvanayaitacca samāse'pyupavarṇyate |
viśiṣṭaguṇabhinne'rthe padamanyat prayujyate || 416 ||

And this is described, in the compound too, by the cultivation of difference; another word is employed in a meaning differentiated by the specific quality.

Bheda-bhavanaya etac ca samase 'py upavarnyate: this is *described*, in the *compound* too, by the *cultivation of difference* — bhava-na: the bringing-into-being by attention, the mind's dwelling on something until it stands out (the term is the yogins', and Bhartrhari uses it for what an analysis does). Visista-guna-bhinne 'rthe padam anyat prayujyate: *another word* is *employed* in a meaning *differentiated by the specific quality*. So the selection of 3.14.415 is described by *cultivating* a difference that the compound does not contain — the analyst dwells on the beauty until it separates from the eclipses. This is the exact status of every analysis in the samuddesa: a bhavana, not a finding (3.14.353's in the means of comprehension, reality is not followed).

3.14.417 unformalized

यदि भिन्नाधिकरणो वचनादनुगम्यते ।
मृगीव चपलेत्यत्र पुंवद्भावो न सिध्यति ॥ ४१७ ॥

yadi bhinnādhikaraṇo vacanādanugamyate |
mṛgīva capaletyatra puṃvadbhāvo na sidhyati || 417 ||

If a different locus is followed from the statement — then in 'fickle as a doe' the treatment-as-masculine does not succeed.

Yadi bhinnadhikarano vacanad anugamyate: if a *different locus* is *followed* from the *statement* — if the comparans and comparandum are taken to sit in different substrata. Mrgi iva capala ity atra pumvad-bhavo na sidhyati: then in *'mrgi iva capala'* — fickle as a doe — the *pumvad-bhava* does *not succeed*: Panini's provision that a feminine first member in a compound be treated as masculine. So the difference-of-locus reading blocks the rule and the form comes out wrong. The comparison's terms must share a locus for the grammar to work — which is what 3.14.379's identity-as-it-were was, seen from the rules.

3.14.418 unformalized

अस्त्रीपूर्वपदत्वात् तु पुंवद्भावो भविष्यति ।
तथैव मृगदुग्धादो न चेत् स्त्र्यर्थो विवक्ष्यते ॥ ४१८ ॥

astrīpūrvapadatvāt tu puṃvadbhāvo bhaviṣyati |
tathaiva mṛgadugdhādo na cet stryartho vivakṣyate || 418 ||

But from the first member's not being feminine, the treatment-as-masculine will occur; likewise in 'deer-milk' and the rest, if the feminine meaning is not intended.

Astri-purva-padatvat tu pumvad-bhavo bhavisyati: from the *first member's* *not being feminine*, the *pumvad-bhava* *will* occur — the rule applies where the member is not feminine to begin with, so nothing needs converting. Tathaiva mrga-dugdhadau na cet stryartho vivaksyate: *likewise* in *'mrga-dugdha'* and the rest, *if* the *feminine meaning* is *not intended* — 3.14.133's deer-milk, where the doe's gender did not penetrate. So the rule is saved by noting that the member was never feminine in the relevant sense: vivaksa decides (na cet... vivaksyate), and where femininity is not meant, there is no femininity to treat as masculine. The repair is semantic and the rule is formal, which is the samuddesa's usual arrangement.

3.14.419 unformalized

शस्त्रीव शस्त्री श्यामेति देवदत्तैव कथ्यते ।
तस्यामेवोभयं तस्मादुच्यते शास्त्रविग्रहे ॥ ४१९ ॥

śastrīva śastrī śyāmeti devadattaiva kathyate |
tasyāmevobhayaṃ tasmāducyate śāstravigrahe || 419 ||

'Dark like a knife' — Devadatta alone is spoken of; therefore both are stated of her alone, in the science's analysis.

Sastri iva sastri syama iti devadatta eva kathyate: *'dark like a knife'* — *Devadatta* *alone* is *spoken of*: whatever the analysis produces, the sentence is about the girl and about nobody else. Tasyam eva ubhayam tasmad ucyate sastra-vigrahe: *therefore* *both* are stated of *her alone*, in the *science's analysis* — sastra-vigraha: the grammar's resolution of the compound. So the vigraha must put both the darkness and the knife-likeness in her, because that is where the utterance put them. The analysis is disciplined by what the sentence is about — which is the only discipline an analysis has (3.14.71's anvakhyanaya, and 3.14.353).

3.14.420 unformalized

पुंवद्भावस्य सिद्ध्यर्थं पक्षे स्त्रीप्रत्ययस्य च ।
बह्वपेक्ष्यमतस्तस्यामुभयप्रतिपादनम् ॥ ४२० ॥

puṃvadbhāvasya siddhyarthaṃ pakṣe strīpratyayasya ca |
bahvapekṣyamatastasyāmubhayapratipādanam || 420 ||

For the establishing of the treatment-as-masculine, and, on the alternative, of the feminine affix — much has to be taken into account; therefore both are conveyed of her.

Pumvad-bhavasya siddhy-artham pakse stri-pratyayasya ca: for the *establishing* of the *pumvad-bhava*, and, *on the alternative*, of the *feminine affix* — the two rules of 3.14.417-418, pulling opposite ways. Bahv apeksyam atah: *much* has to be *taken into account* — bahu apeksyam: a great deal must be considered. Tasyam ubhaya-pratipadanam: *therefore* *both* are *conveyed* of *her*. So the grammar's two provisions are reconciled by locating both in the girl (3.14.419), and Bhartrhari concedes the reconciliation is laborious. Bahv apeksyam is a rare admission of the machinery's cost, and it stands where the machinery is at its worst.

3.14.421 unformalized

प्राप्तिः प्रगृह्यसंज्ञाया न स्यात् प्रत्ययलक्षणात् ।
कुमार्यगारे न ह्यस्ति समासो वचनान्तरे ॥ १२१ ॥

prāptiḥ pragṛhyasaṃjñāyā na syāt pratyayalakṣaṇāt |
kumāryagāre na hyasti samāso vacanāntare || 121 ||

The obtaining of the term 'pragrhya' would not arise from the affix-indication; for in 'kumaryagara' there is no compound in another number.

This verse is identical to 3.14.121, and in our source it carries the number 121 rather than 421 — the repetition is the transmission's, not the reader's. Its argument belongs with 3.14.121's context (number-difference withdrawing in a compound, so that pratyaya-laksana cannot reach an elided dual), and it has no evident work to do here among the comparison-compounds. We print it where the source prints it and note the doubling rather than silently delete: a duplicated karika is evidence about the text's history, and suppressing it would destroy that evidence.
contested: The source repeats this verse verbatim from 3.14.121 and mis-numbers it '121'. Whether the doubling is scribal (an eye-skip that copied a leaf's worth of material twice) or reflects a genuine variant recension, we cannot determine from a single witness. Readers with access to the critical editions should check whether the verse is attested here at all.

3.14.422 unformalized

अथ श्यामेव शस्त्रीवं श्यामेत्येवं प्रयुज्यते ।
शस्त्री यथेयं श्यामेति तावदेव प्रतीयते ॥ ४२२ ॥

atha śyāmeva śastrīvaṃ śyāmetyevaṃ prayujyate |
śastrī yatheyaṃ śyāmeti tāvadeva pratīyate || 422 ||

Now, 'dark as a knife' is thus employed; 'this is dark, as a knife is' — only that much is cognized.

Atha syama iva sastri syama ity evam prayujyate: *'dark as a knife'* is thus *employed*. Sastri yatha iyam syama iti tavad eva pratiyate: *'this is dark, as a knife is'* — *only that much* is *cognized*. Tavad eva: just that and no more (3.14.211's tavat, of the bahuvrihi's reach). So the whole content of the simile is: she is dark, in the way a knife is. No shade is specified, no property is shared in the strong sense, nothing is measured (3.14.360's aparijnana). The verse states the simile's meaning in its own words and the statement is disappointingly small — which is the point. That smallness is what the last sixty verses have been accounting for.

3.14.423 unformalized

उपलक्षणमात्रार्था गुणस्यास्य यदि श्रुतिः ।
पृथग् द्वयोः श्रुतोऽप्येष नेष्टस्यार्थस्य वाचकः ॥ ४२३ ॥

upalakṣaṇamātrārthā guṇasyāsya yadi śrutiḥ |
pṛthag dvayoḥ śruto'pyeṣa neṣṭasyārthasya vācakaḥ || 423 ||

If the hearing of this quality has mere indication for its purpose — then, heard separately of the two, it is not the expresser of the desired meaning.

Upalaksana-matrartha gunasya asya yadi srutih: if the *hearing* of this *quality* has *mere indication* for its purpose — upalaksana: a pointer, not a description (3.14.170, 3.14.182). Prthag dvayoh sruto 'py esa na istasya arthasya vacakah: *heard separately* of the *two*, it is *not the expresser* of the *desired meaning*. So if 'dark' merely indicates, then saying it of the knife and of the girl separately gives two indications and no comparison: the desired meaning — the likeness — is not expressed by either. The indicator-reading cannot get the simile's content. So the quality must be doing more than pointing, and 3.14.424 says what.

3.14.424 unformalized

उपमेयं तु यद् वाच्यं तस्य चेत् प्रतिपादने ।
सव्यापारा गुणास्तत्र सर्वस्योक्तिः सकृच्छ्रुतौ ॥ ४२४ ॥

upameyaṃ tu yad vācyaṃ tasya cet pratipādane |
savyāpārā guṇāstatra sarvasyoktiḥ sakṛcchrutau || 424 ||

But if, in conveying the comparandum which is to be expressed, the qualities are engaged in operation there — then in a single hearing there is the statement of all.

Upameyam tu yad vacyam tasya cet pratipadane sa-vyapara gunas tatra: if, in *conveying* the *comparandum* which is *to be expressed*, the *qualities* are *engaged in operation* there — sa-vyapara: with their activity, at work (3.14.366's word). Sarvasya uktih sakrc-chrutau: then in a *single hearing* there is the *statement of all*. Sakrt: once. So one saying does the whole work — the darkness, the likeness, the girl — because the qualities are active rather than merely indicative. This answers 3.14.423: the quality is not a pointer, it is an operation, and an operation said once reaches everything it operates on. Sakrc-chrutau is the whole reply.

3.14.425 unformalized

प्रकाराधारभेदेन विशेषे समवस्थितः ।
शब्दान्तराभिसम्बन्धे सामान्यवचनः कथम् ॥ ४२५ ॥

prakārādhārabhedena viśeṣe samavasthitaḥ |
śabdāntarābhisambandhe sāmānyavacanaḥ katham || 425 ||

Settled in the particular through difference of mode and of substratum — in connection with another word, how is it 'expressing the general'?

Prakara-adhara-bhedena visese samavasthitah: *settled in the particular* through difference of *mode* and of *substratum* — prakara: the manner in which it is dark; adhara: what it is dark in. Both differ, so the quality is a particular. Sabdantara-abhisambandhe samanya-vacanah katham: in *connection with another word*, *how* is it *'expressing the general'* — Panini's samanya-vacana (3.14.385)? So the objection: the rule calls this word a general-expresser and everything just said makes it a particular. Katham: how? The question is left open here and 3.14.426-427 answer it by listing the ways the tradition has tried.

3.14.426-427 unformalized

सादृश्यमात्रं सामान्यं द्विष्ठं कैश्चित् प्रतीयते ।
गुणो भेदेऽप्यभेदेन द्विवृत्तिर्वा विवक्षितः ॥ ४२६ ॥
व्यापारो जातिभागस्य द्रव्ययोर्वाभिधित्सितः ।
रूपात् सामान्यवाचित्वं प्राग्वा वृत्तेरुदाहृतम् ॥ ४२७ ॥

sādṛśyamātraṃ sāmānyaṃ dviṣṭhaṃ kaiścit pratīyate |
guṇo bhede'pyabhedena dvivṛttirvā vivakṣitaḥ || 426 ||
vyāpāro jātibhāgasya dravyayorvābhidhitsitaḥ |
rūpāt sāmānyavācitvaṃ prāgvā vṛtterudāhṛtam || 427 ||

(426) By some, mere similarity is cognized as the general, residing in two; or the quality, though in difference, is intended as operating in two by non-difference. (427) Or the operation of the universal-portion, of the two substances, is what is meant to be expressed; or the expressing of the general from the form, prior to the derivation, has been illustrated.

Four answers to 3.14.425, and Bhartrhari lists them without choosing. (426): sadrsya-matram samanyam dvistham kaiscit pratiyate — by *some*, *mere similarity* is the *general*, *residing in two*: resemblance itself is the universal. Guno bhede 'py abhedena dvi-vrttir va vivaksitah: *or* the *quality*, though *in difference*, is *intended* as *operating in two* by *non-difference* (3.14.362's asrita-ekatva). (427): vyaparo jati-bhagasya dravyayor va abhidhitsitah — *or* the *operation of the universal-portion*, of the *two substances* (3.14.402-403). Rupat samanya-vacitvam prag va vrtter udahrtam: *or* the *expressing of the general* *from the form*, *prior to the derivation* (3.14.137's samanya-visaya). So: similarity, intended oneness, the jati-portion, and the form — four ways to make the particular a general-expresser, all available, none refuted. This is the samuddesa's method at full extension (3.14.97).

3.14.428 unformalized

व्याघ्रशब्दो यदा शौर्यात् पुरुषार्थोऽवतिष्ठते ।
तदाधिकरणाभेदात् समासस्यास्ति सम्भवः ॥ ४२८ ॥

vyāghraśabdo yadā śauryāt puruṣārtho'vatiṣṭhate |
tadādhikaraṇābhedāt samāsasyāsti sambhavaḥ || 428 ||

When the word 'tiger', by reason of valour, comes to stand for the man-meaning — then, from non-difference of locus, the compound is possible.

Vyaghra-sabdo yada sauryat purusartho 'vatisthate: when the word *'tiger'*, by reason of *valour*, comes to *stand for* the *man*-meaning — the word has migrated: 'vyaghra' now means the man (3.14.347's the manca stating the people as its reality). Tada adhikarana-abhedat samasasya asti sambhavah: *then*, from *non-difference of locus*, the *compound* is *possible*. So 'purusa-vyaghra' can compound because both words have come to sit in one place — the man. This is 3.14.417's requirement satisfied: shared locus, and the rules run. The migration is what makes the compound available.

3.14.429 unformalized

शूरशब्दप्रयोगे तु व्याघ्रशब्दो मृगे स्थितः ।
भिन्नेऽधिकरणे वृत्तेस्तत्र नैवास्ति सम्भवः ॥ ४२९ ॥

śūraśabdaprayoge tu vyāghraśabdo mṛge sthitaḥ |
bhinne'dhikaraṇe vṛttestatra naivāsti sambhavaḥ || 429 ||

But when the word 'brave' is employed, the word 'tiger' stands in the animal; the locus being different, there is no possibility of derivation there.

The contrast, and it is exact. Sura-sabda-prayoge tu vyaghra-sabdo mrge sthitah: when the word *'brave'* is *employed*, the word *'tiger'* *stands in the animal* — say 'brave as a tiger' and the tiger stays a tiger: the explicit quality-word holds it in place and blocks the migration of 3.14.428. Bhinne 'dhikarane vrtteh tatra na eva asti sambhavah: the *locus* being *different*, there is *no possibility* of derivation there. So 'sura-vyaghra' will not compound, and 'purusa-vyaghra' will — and the difference is whether the shared quality is said. Saying it keeps the terms apart; leaving it unsaid lets them fuse. This is the finest thing in the comparison-stretch: the simile compounds only when it does not explain itself.

3.14.430 unformalized

उष्ट्रावयवतुल्येषु मुखेषूष्ट्रश्रुतिर्यथा ।
वर्तते गृहतुल्ये च प्रासादे मधुराश्रुतिः ॥ ४३० ॥

uṣṭrāvayavatulyeṣu mukheṣūṣṭraśrutiryathā |
vartate gṛhatulye ca prāsāde madhurāśrutiḥ || 430 ||

As the word 'camel' operates in faces resembling a camel's part; and the word 'Madhura' in a palace resembling a house.

Two migrations, of the sort 3.14.428 licensed. Ustra-avayava-tulyesu mukhesu ustra-sruti: the word *'camel'* operates in *faces* resembling a *camel's part* — one calls a man's face a camel, meaning the muzzle: the word travels from a part to a whole that is like the part. Vartate grha-tulye ca prasade madhura-srutih: and the word *'Madhura'* in a *palace* resembling a *house* — the city's name for a grand building. So words move to whatever resembles what they mean, and the resemblance need not be with the whole: a part will carry the word across (3.14.31's chatrin, whose designation covered his umbrella-less fellows — the samuddesa's opening case). The Vrtti-samuddesa began with names leaking across differences, and it is still tracking the leak four hundred verses later.

3.14.431 unformalized

यथाध्ययनयोः साम्यमध्येत्रोरपदिश्यते ।
तथा क्रियागतैर्धर्मैरुच्यन्ते साधनाश्रयाः ॥ ४३१ ॥

yathādhyayanayoḥ sāmyamadhyetrorapadiśyate |
tathā kriyāgatairdharmairucyante sādhanāśrayāḥ || 431 ||

As the similarity of two studyings is designated of the two students — so the substrata of the means are spoken of by properties belonging to the action.

The comparison of actions begins. Yatha adhyayanayoh samyam adhyetror apadisyate: as the *similarity* of *two studyings* is *designated* *of the two students* — the likeness is between the studyings and one says it of the *studiers*: 'he studies like a brahmin' (3.14.384). Tatha kriya-gatair dharmair ucyante sadhana-asrayah: *so* the *substrata of the means* are *spoken of* by *properties belonging to the action*. So the action's properties are predicated of the persons — the resemblance lives in the studying and is asserted of the student. Which is 3.14.383's inference-across, now running from action to agent.

3.14.432 unformalized

इवार्थे यच्च वचनं पूर्वसूत्रे च यो विधिः ।
क्रियाशब्दश्रुतौ भेदो न कश्चित् विद्यते तयोः ॥ ४३२ ॥

ivārthe yacca vacanaṃ pūrvasūtre ca yo vidhiḥ |
kriyāśabdaśrutau bhedo na kaścit vidyate tayoḥ || 432 ||

The statement in the sense of 'iva', and the rule in the previous sutra — in the hearing of an action-word there is no difference at all between those two.

Ivarthe yac ca vacanam purva-sutre ca yo vidhih: the *statement* in the sense of *'iva'*, and the *rule* in the *previous sutra* — two of Panini's provisions for comparison. Kriya-sabda-srutau bhedo na kascit vidyate tayoh: in the *hearing* of an *action-word* there is *no difference at all* between those two — na kascit: none whatever. So where the compared thing is an action, the two rules coincide and one of them is idle. This is the samuddesa's standing audit (3.14.40-41, 3.14.106, 3.14.318): rules are checked for redundancy case by case, and the redundancy is reported without recommendation.

3.14.433 unformalized

वत्यन्तावयवे वाक्ये यदौपम्यं प्रतीयते ।
तत्प्रत्ययविधौ सूत्रे निर्देशोऽयं विचार्यते ॥ ४३३ ॥

vatyantāvayave vākye yadaupamyaṃ pratīyate |
tatpratyayavidhau sūtre nirdeśo'yaṃ vicāryate || 433 ||

The comparison cognized in a sentence having a 'vati'-ending member — in the sutra prescribing that affix, this indication is what is considered.

Vaty-antavayave vakye yad aupamyam pratiyate: the *comparison* cognized in a *sentence* having a *vati*-ending member — 'brahmanavat', in the manner of a brahmin: the affix that makes an adverb of likeness. Tat-pratyaya-vidhau sutre nirdeso 'yam vicaryate: in the *sutra* prescribing that affix, *this indication* is what is *considered*. So the next stretch is an examination of the vati-rule's wording: what exactly does the sutra pick out when it prescribes the comparison-affix? The question sounds technical and it is about what an action is.

3.14.434 unformalized

क्रियेत्युपाधिः प्राथम्यात् प्रकृत्यर्थस्य यद्यपि ।
न प्रातिपदिकं तत्र क्रियावाच्युपपद्यते ॥ ४३४ ॥

kriyetyupādhiḥ prāthamyāt prakṛtyarthasya yadyapi |
na prātipadikaṃ tatra kriyāvācyupapadyate || 434 ||

Although 'action' is an adjunct of the base-meaning by primacy — a nominal base expressing action is not possible there.

Kriyety upadhih prathamyat prakrty-arthasya yady api: although *'action'* is an *upadhi* — an adjunct, a qualifying condition (3.14.124's word) — of the *base-meaning*, *by primacy*. Na pratipadikam tatra kriya-vacy upapadyate: a *nominal base* *expressing action* is *not possible* there. So the rule requires an action and requires a nominal stem, and a nominal stem cannot express an action: verbs do that. The rule asks for something that cannot exist. This is the difficulty the following verses work over, and it is not verbal — it goes to the categorial gulf between nama and akhyata that 3.14.337 was already marking.

3.14.435 unformalized

सत्त्वृत्तस्य शेषे वा तृतीया साधनेऽपि वा ।
तिङामसत्त्वाचित्वादुभयं तन्न विद्यते ॥ ४३५ ॥

sattvṛttasya śeṣe vā tṛtīyā sādhane'pi vā |
tiṅāmasattvācitvādubhayaṃ tanna vidyate || 435 ||

Or the third case, of what operates in a substance, in the remainder or in the means — from the finite verbs' expressing non-substance, neither of those exists.

Sattva-vrttasya sese va trtiya sadhane 'pi va: the *third case* — the instrumental — of what *operates in a substance*, in the *remainder* or in the *means*: two of the instrumental's uses, both requiring a substance. Tinam asattva-vacitvad ubhayam tan na vidyate: from the *finite verbs'* expressing *non-substance*, *neither* of those *exists*. Asattva-vacitva: expressing what is not a thing — the standing characterization of what verbs and particles mean (3.14.193). So a verb cannot take the instrumental in either use, because both need a thing and verbs do not supply one. The categorial gulf of 3.14.434 is now producing consequences in the case-system.

3.14.436 unformalized

पाकादस्तृतीयान्ताः सत्त्वधर्मसमन्वयात् ।
न क्रियेत्यपदिश्यन्ते कृत्वोर्थप्रत्यये यथा ॥ ४३६ ॥

pākādastṛtīyāntāḥ sattvadharmasamanvayāt |
na kriyetyapadiśyante kṛtvorthapratyaye yathā || 436 ||

'Paka' and the rest, ending in the third case, are not designated as 'action', from their concomitance with substance-properties — as in the affix in the sense of 'krtvas'.

Pakadas trtiyantah sattva-dharma-samanvayat na kriyety apadisyante: *'paka'* and the rest, ending in the *third* case, are *not* designated as *'action'*, from their *concomitance with substance-properties* — 'pakena', by cooking: the noun takes a case, and taking a case is a substance's behaviour (3.14.435). So the deverbal noun is not an action for the rule's purposes, however much cooking it names. Krtvo-'rtha-pratyaye yatha: as in the *affix* in the sense of *'krtvas'* — the frequency-suffix, where the same finding holds. So the rule's 'action' cannot mean deverbal nouns either: they have gone over to the substance side by taking endings.

3.14.437 unformalized

ये चाव्ययकृतः केचित् क्रियाधर्मसमर्न्विताः ।
तेषामसत्त्वाचित्वं तिङ्न्तैर्न विशिष्यते ॥ ४३७ ॥

ye cāvyayakṛtaḥ kecit kriyādharmasamarnvitāḥ |
teṣāmasattvācitvaṃ tiṅntairna viśiṣyate || 437 ||

And those indeclinable krt-forms which are endowed with action-properties — their expressing of non-substance is not distinguished from that of the finite verbs.

Ye ca avyaya-krtah kecit kriya-dharma-samanvitah: those *indeclinable krt-forms* — gerunds and infinitives — which are *endowed with action-properties*: these do carry the action, and they take no endings. Tesam asattva-vacitvam tinantair na visisyate: their *expressing of non-substance* is *not distinguished* from that of the *finite verbs*. So the avyaya-krts are on the verb's side of the gulf: they mean actions and they mean no thing, exactly as tin-forms do. Which closes the last door — the rule wanted a nominal stem expressing an action, and everything expressing an action is either a verb or verb-like, and everything nominal has become a substance.

3.14.438 unformalized

कृत्वसुज्विषया यापि शयितव्यादिषु क्रिया ।
उपमानोपमेयत्व तत्राप्यन्तमसम्भवि ॥ ४३८ ॥

kṛtvasujviṣayā yāpi śayitavyādiṣu kriyā |
upamānopameyatva tatrāpyantamasambhavi || 438 ||

And that action which is the domain of 'krtvasuc', in 'sayitavya' and the rest — there too the comparans-comparandum relation is utterly impossible.

Krtvasuj-visaya yapi sayitavyadisu kriya: the *action* which is the *domain of krtvasuc*, in *'sayitavya'* (to be slept) and the rest — the gerundives, which do name actions. Upamana-upameyatva tatrapy atyantam asambhavi: *there too* the *comparans-comparandum* relation is *utterly impossible*. So even the gerundive, which looks like a nominal expressing an action, cannot enter a comparison. The catalogue is complete: verb, deverbal noun, indeclinable krt, gerundive — none of them will serve, and 3.14.434's rule still asks for one.
contested: Pada d is transmitted as 'tatrapy antam asambhavi', in which 'antam' does not construe. We read 'atyantam' ('utterly'), which requires only the loss of a syllable in transmission and gives the emphatic sense the argument wants. Pada c's 'upamanopameyatva' is likewise metrically short for the expected 'upamanopameyatvam'.

3.14.439 unformalized

न केवलौ द्रव्यगुणौ तद्वान् वाप्युपमीयते ।
शयितव्यादिभिस्तेषु नोपमार्थोऽस्ति कश्चन ॥ ४३९ ॥

na kevalau dravyaguṇau tadvān vāpyupamīyate |
śayitavyādibhisteṣu nopamārtho'sti kaścana || 439 ||

Neither the bare substance and quality, nor their possessor, is compared; by 'sayitavya' and the rest there is no purpose of comparison in them.

Na kevalau dravya-gunau tadvan vapy upamiyate: neither the *bare* *substance and quality*, *nor* their *possessor*, is *compared* — the three candidates for what a comparison relates, all declined. Sayitavyadibhis tesu na upamartho 'sti kascana: by *'sayitavya'* and the rest there is *no purpose of comparison* in them — na kascana: none at all (3.14.364's phrase). So gerundives compare nothing to nothing: neither the substance, nor the quality, nor the qualified thing is in play. The failure is total and it is not a failure of the words — it is that actions do not stand still to be compared. 3.14.441-442 will say why.

3.14.440 unformalized

उपमानोपमेयत्वे द्रव्ये चानुक्तधर्मिणि ।
निमित्तत्वेन गम्यन्ते रूढयोगाः क्रियागुणाः ॥ ४४० ॥

upamānopameyatve dravye cānuktadharmiṇi |
nimittatvena gamyante rūḍhayogāḥ kriyāguṇāḥ || 440 ||

In the comparans-comparandum relation, and in a substance whose property is unstated — the conventionally connected actions and qualities are understood as the cause.

Upamana-upameyatve dravye ca anukta-dharmini: in the comparans-comparandum relation, and in a *substance* whose *property* is *unstated* — anukta-dharmin: nothing said about what it is like (3.14.415's candra-vaktra, where no property was named). Nimittatvena gamyante rudha-yogah kriya-gunah: the *conventionally connected* *actions and qualities* are *understood* as the *cause*. Rudha-yoga: the connection that convention has settled (3.14.55's rudhi). So when nothing is said, what supplies the tertium is convention: the moon's beauty and not its eclipses, because that is what the moon is conventionally connected with. This is the answer 3.14.415 needed, and it took twenty-five verses to reach it: the selection is made by rudhi.

3.14.441 unformalized

होतव्यसदृशो होतेत्यत्राप्यर्थो न विद्यते ।
विरोधात् क्रियया तस्मात् क्रियावान् नोपमीयते ॥ ४४१ ॥

hotavyasadṛśo hotetyatrāpyartho na vidyate |
virodhāt kriyayā tasmāt kriyāvān nopamīyate || 441 ||

'The hotr is like what is to be offered' — here too there is no meaning; from contradiction with the action, therefore, an action-possessor is not compared.

Hotavya-sadrso hota ity atrapy artho na vidyate: *'the hotr is like what is to be offered'* — here too there is *no meaning*: the priest is not like the oblation, and the sentence is not merely false but empty. Virodhat kriyaya: from *contradiction with the action* — the offerer and the offered are opposed *within* the action, and the action's own structure forbids their likeness. Tasmat kriyavan na upamiyate: *therefore* an *action-possessor* is *not compared*. So the karakas of one action cannot be compared to each other: the action that relates them is what keeps them unlike (3.7's the sadhanas as distinguished capacities; 3.14.390's karmatvam karanatvam ca bhedena eva asritam).

3.14.442 unformalized

क्रिया समानजातीय तद्भावान्नोपमीयते ।
जातिभेदेऽपि पाकेन (वा केन) भिन्नाः पाकादयः क्रियाः ॥ ४४२ ॥

kriyā samānajātīya tadbhāvānnopamīyate |
jātibhede'pi pākena (vā kena) bhinnāḥ pākādayaḥ kriyāḥ || 442 ||

An action of the same class is not compared, from its being that; and even where the class differs, the actions 'cooking' and the rest are differentiated by 'cooking'.

The dilemma for action-comparison, and it is 3.14.363's in a new key. Kriya samana-jatiya tad-bhavan na upamiyate: an action of the *same class* is *not compared*, *from its being that* — identity again: cooking is not like cooking, it *is* cooking (3.14.394's the cause in the cognition 'cow' exists in another cow). Jati-bhede 'pi pakena bhinnah pakadayah kriyah: and *even where the class differs*, the actions *'paka'* and the rest are *differentiated by 'paka'* — where the classes differ, the differentiator is the action itself, and there is no neutral ground. So: same class, nothing to compare; different class, nothing shared. Actions admit no comparison at all — which is why 3.14.438-439 failed.
contested: The source records a variant at pada d: 'pakena' ('by cooking') against 'va kena' ('or by what?'). We translate the former. On the variant the pada becomes a rhetorical question — 'by what, then, are the actions differentiated?' — which is also apt; the dilemma stands either way.

3.14.443 unformalized

आधारभेदाद् भिन्नायामुपमानस्य सम्भवः ।
अध्येतव्येन विप्राणां तुल्यमध्ययनं विशाम् ॥ ४४३ ॥

ādhārabhedād bhinnāyāmupamānasya sambhavaḥ |
adhyetavyena viprāṇāṃ tulyamadhyayanaṃ viśām || 443 ||

Where it is differentiated by difference of substratum, comparison is possible: 'the vaisyas' study is like the brahmins' studying'.

The exception, and it is the one that saves the whole apparatus. Adhara-bhedad bhinnayam upamanasya sambhavah: where the action is *differentiated by difference of substratum*, *comparison* is *possible*. Adhyetavyena vipranam tulyam adhyayanam visam: *'the study of the vaisyas is like the brahmins' studying'*. So one action, two seats, and the seats' difference is what lets the two studyings be compared. 3.14.442's dilemma assumed the action alone; bring in the substrata and the middle ground appears — the studyings are one in kind and two in seat, which is exactly the structure 3.14.396 said a comparison needs. This is why 3.14.431 opened by locating the resemblance in the students.

3.14.444 unformalized

अर्थात् प्रकरणाद् वापि यत्रोपेक्ष्यं प्रतीयते ।
सामर्थ्यादनपेक्षस्य तस्य वृत्तिः प्रसज्यते ॥ ४४४ ॥

arthāt prakaraṇād vāpi yatropekṣyaṃ pratīyate |
sāmarthyādanapekṣasya tasya vṛttiḥ prasajyate || 444 ||

Where what is to be expected is cognized from the meaning or from the context — the derivation of that, being independent, follows from capacity.

Arthat prakaranad vapi yatra upeksyam pratiyate: where what is *to be expected* is cognized from the *meaning* or from the *context* — prakarana: the situation of utterance, one of the meaning-fixing conditions (2.316's list). Samarthyad anapeksasya tasya vrttih prasajyate: the *derivation* of *that*, being *independent*, *follows* from *capacity* — samarthya (3.14.414's word). So where sense or context has already supplied what the form would have reached for, the form no longer reaches: it is anapeksa, independent, and capacity carries the rest. Context makes a word self-sufficient, and self-sufficiency licenses compounding.

3.14.445 unformalized

तैलापाकेन तुल्ये च धृतपाके विवक्षिते ।
क्रियावदपि कार्याणां दर्शनात् प्रत्ययो भवेत् ॥ ४४५ ॥

tailāpākena tulye ca dhṛtapāke vivakṣite |
kriyāvadapi kāryāṇāṃ darśanāt pratyayo bhavet || 445 ||

And when a ghee-cooking similar to an oil-cooking is intended — from the seeing of operations even as with an action, the affix would occur.

3.14.443's exception given a second instance. Taila-apakena tulye ca ghrta-pake vivaksite: when a *ghee-cooking* *similar* to an *oil-cooking* is *intended* — two cookings, different substrata, comparable (3.14.443's structure exactly). Kriyavad api karyanam darsanat pratyayo bhavet: from the *seeing of operations* *even as with an action*, the *affix* would *occur*. So the rules apply to the compared actions as they would to any action, and the vati-affix comes — which is what 3.14.434's difficulty seemed to forbid. The resolution: the rule's 'action' means an action *seated*, and seated actions behave like nominals because their seats do.

3.14.446 unformalized

अतिङ्ग्रहणमेवं तु समासस्य निवर्तकम् ।
गमनं कारकस्येति ण्वुल्यन्यस्मिन्न संभवेत् ॥ ४४६ ॥

atiṅgrahaṇamevaṃ tu samāsasya nivartakam |
gamanaṃ kārakasyeti ṇvulyanyasminna saṃbhavet || 446 ||

Thus the taking-up of 'atin' is what excludes the compound; 'the going of the agent' — 'nvul' would not be possible in another.

Atin-grahanam evam tu samasasya nivartakam: *thus* the taking-up of *'atin'* — Panini's 'not a finite verb', in the relevant rule — is what *excludes* the *compound*. So the sutra's word 'atin' does the work 3.14.434-437 identified as needed: it keeps finite verbs out, since they are the one thing that cannot enter. Gamanam karakasyeti nvul anyasmin na sambhavet: *'the going of the agent'* — *'nvul'* would *not* be possible in *another*. So the affix is confined to its proper base and cannot migrate. Bhartrhari's habit again: a single word of Panini's turns out to encode a result the karikas take a dozen verses to establish (3.14.225's artha-grahanam krtam).

3.14.447 unformalized

सर्वस्य परिहारार्थं समुदायत्वमाश्रितम् ।
शुद्धायाः संभवान्नस्यात् क्रियाया ब्राह्मणादिषु ॥ ४४७ ॥

sarvasya parihārārthaṃ samudāyatvamāśritam |
śuddhāyāḥ saṃbhavānnasyāt kriyāyā brāhmaṇādiṣu || 447 ||

For the avoidance of all of it, aggregate-hood is resorted to; from the possibility of the pure action, it would not obtain in 'brahmana' and the rest.

Sarvasya pariharartham samudayatvam asritam: for the *avoidance* of *all of it* — every difficulty of 3.14.434-442 — *aggregate-hood* is *resorted to*: take the compound as an aggregate and the questions about which member expresses the action lapse (3.14.96's samudayas tv eka eva sah). Suddhayah sambhavan na syat kriyaya brahmanadisu: from the *possibility* of the *pure* action, it would *not* obtain in *'brahmana'* and the rest. So the escape works only because a pure, seatless action is not available: if it were, the aggregate-reading would be unnecessary. The device is a last resort and the verse says so — parihara: an evasion, and Bhartrhari calls it one.

3.14.448 unformalized

उपमानविवक्षायां स्वधर्मश्च निवर्तते ।
क्रियाया न श्रुताद् यस्मादुपमानं समाप्यते ॥ ४४८ ॥

upamānavivakṣāyāṃ svadharmaśca nivartate |
kriyāyā na śrutād yasmādupamānaṃ samāpyate || 448 ||

When the comparans is intended to be expressed, its own property withdraws; since comparison is not completed from the heard action.

Upamana-vivaksayam sva-dharmas ca nivartate: when the *comparans* is *intended* to be expressed, its *own property* *withdraws* — the standard's proper nature drops away when it is put to work as a standard (3.14.375's it operates by its own property for the measuring of another: the property is lent out and the thing behind it goes quiet). Kriyaya na srutad yasmad upamanam samapyate: since *comparison* is *not completed* from the *heard* action. Samapyate: is *finished* (3.14.382's asamapta-guna). So the audible action does not finish the comparison — nothing does (3.14.360) — and meanwhile the standard's own nature has withdrawn. The comparans gives everything and keeps nothing.

3.14.449 unformalized

तृतीयाऽप्याश्रितो भेदो धर्मः साधारणो द्वयोः ।
व्यापारवान् न कृत्स्नस्य साम्यं कृत्स्नेन विद्यते ॥ ४४९ ॥

tṛtīyā'pyāśrito bhedo dharmaḥ sādhāraṇo dvayoḥ |
vyāpāravān na kṛtsnasya sāmyaṃ kṛtsnena vidyate || 449 ||

The third case too — the difference resorted to, the property common to the two, engaged in operation; there is no similarity of the whole with the whole.

Trtiya apy asrito bhedah dharmah sadharano dvayoh vyaparavan: the *third case* too — the *difference resorted to*, the *property common to the two*, *engaged in operation* (3.14.424's sa-vyapara). Na krtsnasya samyam krtsnena vidyate: there is *no similarity* of the *whole* with the *whole*. So a comparison never matches its terms entire: only a property, only under an aspect, only partly. Krtsnena krtsnasya: whole with whole — and if there were such a similarity it would be identity, and there would be no comparison (3.14.363). Every simile is partial by constitution, and the partiality is not a shortfall to be regretted.

3.14.450 unformalized

द्रव्ये वापि क्रियायां वा निमित्तात् तत् प्रकल्पते ।
क्रियाणां विद्यमानत्वाद् वृत्तिर्न स्याद् गवादिषु ॥ ४५० ॥

dravye vāpi kriyāyāṃ vā nimittāt tat prakalpate |
kriyāṇāṃ vidyamānatvād vṛttirna syād gavādiṣu || 450 ||

In a substance or in an action, that is conceivable from the cause; from the existence of actions, there would be no derivative formation in 'go' and the rest.

Dravye vapi kriyayam va nimittat tat prakalpate: in a *substance* *or* in an *action*, *that* is conceivable *from the cause* — the comparison can be lodged either place, so long as an occasioning cause is found. Kriyanam vidyamanatvad vrttir na syad gavadisu: from the *existence of actions*, there would be *no derivative formation* in *'go'* and the rest. So if any action's mere existence licensed the comparison-affix, then 'go' — which has plenty of actions about it — would take it, and it does not. The cause must be more than the action's being there. This is 3.14.440's rudha-yoga argument in advance: what licenses is convention, not availability.

3.14.451 unformalized

अभावात् केवलायास्तु तद्वानर्थः प्रतीयते ।
प्रधानासम्भवे युक्ता लक्षणार्था क्रियाश्रुतिः ॥ ४५१ ॥

abhāvāt kevalāyāstu tadvānarthaḥ pratīyate |
pradhānāsambhave yuktā lakṣaṇārthā kriyāśrutiḥ || 451 ||

But from the absence of the pure action, the meaning possessing it is what is cognized; when the principal is impossible, the hearing of the action, having indication for its purpose, is proper.

Abhavat kevalayas tu tadvan arthah pratiyate: from the *absence* of the *pure* action, the *meaning possessing it* is what is cognized — there is no seatless action to be got at, so what one reaches is the possessor (3.14.447's the pure action is not available). Pradhana-asambhave yukta laksanartha kriya-srutih: when the *principal* is *impossible*, the *hearing of the action*, having *indication* for its purpose, is *proper*. Laksanartha: for the sake of indicating — the action-word becomes a pointer (3.14.170's upalaksana). So where the action cannot be principal, its word demotes itself to an indicator of the possessor — and that demotion is *yukta*, proper. The rule's 'kriya' means: use the action to point at whoever has it.

3.14.452 unformalized

क्रियाधर्मान्तरेषु सापेक्षाः क्रियाशब्दाः क्रियान्तरे ।
उपकाराय गृह्यन्ते यथैव ब्राह्मणादयः ॥ ४५२ ॥

kriyādharmāntareṣu sāpekṣāḥ kriyāśabdāḥ kriyāntare |
upakārāya gṛhyante yathaiva brāhmaṇādayaḥ || 452 ||

Action-words, dependent, in other action-properties, in another action — are taken up for assistance, just as 'brahmana' and the rest are.

Kriya-dharmantaresu sapeksah kriya-sabdah kriyantare: *action-words*, *dependent* — sapeksa: reaching (3.14.47's the relatum-word is wholly dependent) — in *other action-properties*, in *another action*. Upakaraya grhyante yathaiva brahmanadayah: are *taken up for assistance*, *just as* *'brahmana'* and the rest are. Upakara: help (3.14.333, 3.14.395). So an action-word in a comparison is an auxiliary, exactly as the base of a nan-compound was: it is brought in to help, and it is not what is meant. The parallel with 'brahmana' is the samuddesa recognizing its own pattern — the negative compound and the comparison-compound do the same thing with their bases.
contested: Padas a-b are metrically long in our source ('kriya-dharmantaresu sapeksah kriya-sabdah kriyantare'), apparently by dittography of a 'kriya'-element. The construction is not in doubt.

3.14.453 unformalized

यथा प्रकर्षः सर्वत्र निमित्तान्तरहेतुकः ।
द्रव्यवद् गुणशब्देऽपि स निमित्तमपेक्षते ॥ ४५३ ॥

yathā prakarṣaḥ sarvatra nimittāntarahetukaḥ |
dravyavad guṇaśabde'pi sa nimittamapekṣate || 453 ||

As excellence everywhere is caused by some other cause — so, as in a substance, in a quality-word too it requires a cause.

Yatha prakarsah sarvatra nimittantara-hetukah: as *excellence* *everywhere* is caused by *some other cause* — prakarsa: superiority, the comparative's content: nothing is superior in itself, only by reason of something else (3.14.162's the excess is in the quality, in its own seat). Dravyavad guna-sabde 'pi sa nimittam apeksate: so, *as in a substance*, in a *quality-word* too it *requires a cause*. So the comparative needs its ground wherever it attaches, substance or quality alike. The verse ties the comparison-compound back to 3.14.159-162's comparatives: both are about excess, and excess is always owed.

3.14.454 unformalized

यो य उच्चार्यते शब्दः स स्वरूपनिबन्धनः ।
यथा तथोपमानेषु व्यपेक्षा न निवर्तते ॥ ४५४ ॥

yo ya uccāryate śabdaḥ sa svarūpanibandhanaḥ |
yathā tathopamāneṣu vyapekṣā na nivartate || 454 ||

Whatever word is uttered, it is grounded in its own form; and since it is so, in the comparanda the mutual expectancy does not withdraw.

Yo ya uccaryate sabdah sa sva-rupa-nibandhanah: *whatever* word is *uttered*, it is *grounded in its own form* — sva-rupa-nibandhana: its own shape is what it rests on (3.14.137's sabda-samskarah rupat; and 1.68's the word first means itself). Yatha tatha upamanesu vyapeksa na nivartate: *since it is so*, in the *comparanda* the *mutual expectancy* does *not withdraw*. So because each word stands on its own form, the reaching between them survives compounding — against 3.14.46, where expectancy withdrew inside a compound. The comparison-compound is thus like the relatum-words of 3.14.47: its members go on reaching, because what they reach with is their own form and composition does not take that away.

3.14.455 unformalized

कियावृत्तेस्तृतीयान्तस्यैतच्चासंगते सति ।
प्रसिद्धन्यायकरणो भाष्ये युजिरुदाहृतः ॥ ४५५ ॥

kiyāvṛttestṛtīyāntasyaitaccāsaṃgate sati |
prasiddhanyāyakaraṇo bhāṣye yujirudāhṛtaḥ || 455 ||

And this, of what ends in the third case operating in an action, there being non-connection — 'yuj', having the well-known reasoning for its instrument, is cited in the Bhasya.

Kriya-vrttes trtiyantasya etac ca asamgate sati: this, of what ends in the *third case* operating in an *action*, there being *non-connection*. Prasiddha-nyaya-karano bhasye yujir udahrtah: *'yuj'* — to join — *having the well-known reasoning for its instrument*, is *cited in the Bhasya*. So Patanjali's example is 'nyayena yuktam', joined by reasoning: a verb of joining with an instrumental that is a nyaya. Bhartrhari cites the Bhasya rather than argue (3.14.34, 3.14.158, 3.14.166's habit), and the example is about to be used against itself in 3.14.456.

3.14.456 unformalized

अन्तर्भूते तु करणे प्रयोगो न पुनर्भवेत् ।
न्यायेनायुक्तमित्यत्र जीवतौ प्राणकर्मवत् ॥ ४५६ ॥

antarbhūte tu karaṇe prayogo na punarbhavet |
nyāyenāyuktamityatra jīvatau prāṇakarmavat || 456 ||

But when the instrument is included, the employment would not occur again — as in 'unjoined by reasoning', like the breath-object in 'lives'.

The samuddesa's own great example, returning four hundred verses later. Antarbhute tu karane prayogo na punar bhavet: when the *instrument* is *included*, the *employment* would *not* occur *again* — if 'yukta' already contains its instrument, one cannot say the instrument over. Nyayena ayuktam ity atra: in *'nyayena ayuktam'* — unjoined by reasoning: and here the instrument *is* said again. Jivatau prana-karmavat: *like* the *breath-object* in *'jivati'* — 3.14.70's decisive case: 'jivati' is glossed 'pranan dharayati' and contains neither breaths nor holding, so nothing prevents saying them. Exactly so here: 'ayukta' does not contain its instrument, whatever the gloss says, and 'nyayena' is free to be said. The parallel is Bhartrhari's, it is exact, and it shows the karikas as a single argument.

3.14.457 unformalized

शास्त्राभ्यासाय भेदोऽयमयुक्तमिति वर्ण्यते ।
अशोभनमसम्बद्धमिति रूढिर्व्यवस्थिता ॥ ४५७ ॥

śāstrābhyāsāya bhedo'yamayuktamiti varṇyate |
aśobhanamasambaddhamiti rūḍhirvyavasthitā || 457 ||

This distinction is described as 'ayuktam' for the practice of the science; the convention 'improper', 'unconnected' is what is settled.

Sastrabhyasaya bhedo 'yam ayuktam iti varnyate: this *distinction* is described as *'ayuktam'* for the *practice of the science* — sastra-abhyasa: the exercise of grammar, its drill. The analysis into 'a-yukta' is a teaching-exercise. Asobhanam asambaddham iti rudhir vyavasthita: the *convention* — *'improper'*, *'unconnected'* — is what is *settled*. Rudhi: what usage has grown into place (3.14.55, 3.14.440). So 'ayukta' means improper by convention, and the derivation from 'yuj' is drill. This is 3.14.267's conclusion for the nan-compound arriving for this word: what fixes the sense is settled usage, and the science's analysis is exercise.

3.14.458 unformalized

विविभक्तिः प्रकृत्यर्थं प्रत्युपाधिः कथं भवेत् ।
विभक्तिपरिणामे च प्रकल्प्यं विषयान्तरम् ॥ ४५८ ॥

vivibhaktiḥ prakṛtyarthaṃ pratyupādhiḥ kathaṃ bhavet |
vibhaktipariṇāme ca prakalpyaṃ viṣayāntaram || 458 ||

How could a case-ending be an adjunct with regard to the base-meaning? And in the transformation of the case-ending, another domain must be imagined.

Vibhaktih prakrty-artham praty upadhih katham bhavet: *how* could a *case-ending* be an *upadhi* — an adjunct, a conditioning qualifier (3.14.124) — with regard to the *base-meaning*? The ending is outside the stem; it cannot condition what the stem means. Vibhakti-parinama ca prakalpyam visayantaram: and in the *transformation* of the case-ending, *another domain* must be *imagined* — prakalpya: to be constructed. So a rule that makes the ending qualify the base needs both an impossible conditioning and an invented domain to change into. Katham: how — the question is not answered, and the following verses show what one has to build to answer it.

3.14.459-460 unformalized

विभक्त्यन्तरयोगो हि यस्य तद्विषयान्तरे ।
विभक्त्यन्तरसम्बन्धः सामर्थ्यादनुमीयते ॥ ४५९ ॥
सारूप्यात् तु तदेवेदमिति तत्रोपचर्यते ।
शब्दान्तरे विभक्त्या तु युक्तं शास्त्रे तदश्रुतम् ॥ ४६० ॥

vibhaktyantarayogo hi yasya tadviṣayāntare |
vibhaktyantarasambandhaḥ sāmarthyādanumīyate || 459 ||
sārūpyāt tu tadevedamiti tatropacaryate |
śabdāntare vibhaktyā tu yuktaṃ śāstre tadaśrutam || 460 ||

(459) For that which has connection with another case-ending — in another domain of it, connection with another case-ending is inferred from capacity. (460) But from likeness of form, 'this is that very one' is there figuratively applied; whereas in another word, joined with a case-ending — that is not heard in the science.

(459): vibhakty-antara-yogo hi yasya tad-visayantare vibhakty-antara-sambandhah samarthyad anumiyate — for what has *connection with another case-ending*, in *another domain* of it, a *connection with another case-ending* is *inferred from capacity*. So the invented domain of 3.14.458 gets its ending by inference, not by rule. (460): sarupyat tu tad eva idam iti tatra upacaryate — but from *likeness of form*, *'this is that very one'* is there *figuratively applied* (3.14.391's sa eva ayam; 3.14.284's upacaryate). Sabdantare vibhaktya tu yuktam sastre tad asrutam: but in *another word*, joined with a *case-ending* — *that is not heard* in the *science*. Asrutam: not heard, i.e. Panini nowhere says it. So the whole construction rests on an inference, a figure, and something the grammar never states. Bhartrhari lays the scaffolding bare and leaves it standing.

3.14.461 unformalized

प्रकृतिश्चेत् तृतीयान्ता तेनेत्यस्मात् प्रतीयते ।
क्रियेति प्रथमान्ता सा कथं भवितुमर्हति ॥ ४६१ ॥

prakṛtiścet tṛtīyāntā tenetyasmāt pratīyate |
kriyeti prathamāntā sā kathaṃ bhavitumarhati || 461 ||

If the base ends in the third case, it is cognized from 'tena'; how can that, being 'kriya' ending in the first case, be proper?

Prakrtis cet trtiyanta tena ity asmat pratiyate: *if* the *base* ends in the *third* case, it is cognized from *'tena'* — the sutra's word 'tena' shows an instrumental is wanted. Kriyeti prathamanta sa katham bhavitum arhati: *how* can *that*, being *'kriya'* ending in the *first* case, be *proper*? — the sutra's other word, 'kriya', is a nominative. So the rule says 'tena' and 'kriya' and the two cases will not agree: one demands an instrumental base and the other supplies a nominative. Katham... arhati: how can it be right? The sutra appears to contradict itself, and 3.14.462 presses the point.

3.14.462 unformalized

क्रिययेति तृतीया च प्रयोगे कस्य कल्प्यताम् ।
तेनेत्यस्य हि सम्बन्धः सूत्रस्थेन न विद्यते ॥ ४६२ ॥

kriyayeti tṛtīyā ca prayoge kasya kalpyatām |
tenetyasya hi sambandhaḥ sūtrasthena na vidyate || 462 ||

And in the employment 'kriyaya', the third case — for whom is it to be imagined? For the connection of 'tena' with what stands in the sutra does not exist.

Kriyaya iti trtiya ca prayoge kasya kalpyatam: in the employment *'kriyaya'* — the instrumental of 'kriya', which the rule seems to need — the *third case*: *for whom* is it to be *imagined*? Kalpyatam: let it be imagined — the imperative is 3.14.253's sardonic one. Tena ity asya hi sambandhah sutra-sthena na vidyate: for the *connection* of *'tena'* with *what stands in the sutra* does *not exist*. So there is nothing in the rule for 'tena' to connect with, and the instrumental one must supply has no owner. The sutra as it stands does not construe. Bhartrhari has been reading Panini generously for four hundred verses; here he reports that a rule will not parse, and does not soften it.

3.14.463 unformalized

सोपस्कारेषु सूत्रेषु वाक्ये शेषः समर्थ्यते ।
तेन यत् तत् तृतीयान्तं क्रिया चेत् सेति गम्यते ॥ ४६३ ॥

sopaskāreṣu sūtreṣu vākye śeṣaḥ samarthyate |
tena yat tat tṛtīyāntaṃ kriyā cet seti gamyate || 463 ||

In sutras requiring supplementation, the remainder is justified in the sentence; therefore, of what ends in the third case — if it is an action, 'that one' is understood.

The answer to 3.14.461-462, and it is a concession dressed as a solution. Sopaskaresu sutresu vakye sesah samarthyate: in sutras *requiring supplementation* — sopaskara: needing to be filled out — the *remainder* is *justified* *in the sentence*: the missing words are supplied from the running context of the Astadhyayi. Tena yat tat trtiyantam kriya cet sa iti gamyate: *therefore*, of what ends in the *third* case — *if* it is an *action*, *'that one'* is *understood*. So the rule construes once one supplies a pronoun that Panini did not write. 3.14.462 said the connection does not exist; here it is manufactured. Sopaskara is the technical name for the manufacture, and the samuddesa uses it without pretending it is discovery.

3.14.464 unformalized

उपाधेः कस्यचिद् वाक्ये प्रयोग उपलभ्यते ।
प्रतीयमानधर्मान् यो न कदाचित् प्रयुज्यते ॥ ४६४ ॥

upādheḥ kasyacid vākye prayoga upalabhyate |
pratīyamānadharmān yo na kadācit prayujyate || 464 ||

The employment of some adjunct is found in a sentence; that which has the cognized properties is never employed.

Upadheh kasyacid vakye prayoga upalabhyate: the *employment of some adjunct* is *found* in a *sentence* — some upadhis do get said. Pratiyamana-dharma yo na kadacit prayujyate: that which has the *cognized properties* is *never* employed — na kadacit: at no time. So there are two kinds of conditioning element: the ones that show up in the sentence, and the ones that carry what is actually understood and never show up at all. The second kind is the interesting one, and 3.14.465 gives the example that makes it undeniable.
contested: Pada c is transmitted as 'pratiyamana-dharman' (accusative plural), which will not construe with the relative 'yah' and the singular verb. We read a nominative singular 'pratiyamana-dharma' ('that which has the cognized properties'), a minimal correction.

3.14.465 unformalized

नीलमुत्पलमित्यत्र न विशेष्ये न भेदके ।
कश्चित् तद्धर्मवचनो वाक्ये शब्दः प्रयुज्यते ॥ ४६५ ॥

nīlamutpalamityatra na viśeṣye na bhedake |
kaścit taddharmavacano vākye śabdaḥ prayujyate || 465 ||

In 'blue lotus', neither in the qualified nor in the differentiator is any word expressing that property employed in the sentence.

The example, and it is perfectly chosen. Nilam utpalam ity atra na visesye na bhedake: in *'nilam utpalam'* — the blue lotus — *neither* in the *qualified* *nor* in the *differentiator*. Kascit tad-dharma-vacano vakye sabdah prayujyate: is *any* word *expressing that property* *employed* in the *sentence*. So: what makes 'nila' qualify 'utpala' rather than the reverse, what makes them co-refer at all — the relation itself — is said by neither word and by no third word. The sentence has two words and conveys three things. This is 3.14.9-11's krsnas tilah and 3.3's the unsayable relation, and here it is used to show that the sentence's most important element is always unemployed.

3.14.466 unformalized

अत्यन्तानुगमात् तत्र न सूत्रे न च विग्रहे ।
विभक्तिपरिणामेन किञ्चिदस्ति प्रयोजनम् ॥ ४६६ ॥

atyantānugamāt tatra na sūtre na ca vigrahe |
vibhaktipariṇāmena kiñcidasti prayojanam || 466 ||

From the utter concomitance there, there is no purpose at all — neither in the sutra nor in the analysis — in a transformation of the case-ending.

Atyanta-anugamat tatra: from the *utter concomitance* there — atyanta-anugama: the two go together completely; blue and lotus are simply co-present in the cognition. Na sutre na ca vigrahe vibhakti-parinamena kincid asti prayojanam: there is *no purpose at all* — *neither* in the *sutra* *nor* in the *analysis* — in a *transformation of the case-ending* (3.14.458's device). So where the concomitance is total, the elaborate machinery of case-transformation has nothing to do: nothing needs converting, because nothing was ever apart. This retires 3.14.458-460's whole construction for the cases that matter most — which is where 3.14.465 was heading.

3.14.467 unformalized

तृतीयान्तं क्रियेत्येतद् विग्रहे न प्रयुज्यते ।
यथा दण्डः प्रहरणं क्रीडायामिति दृश्यते ॥ ४६७ ॥

tṛtīyāntaṃ kriyetyetad vigrahe na prayujyate |
yathā daṇḍaḥ praharaṇaṃ krīḍāyāmiti dṛśyate || 467 ||

The third-case-ending 'kriya' is not employed in the analysis; as 'the staff is the weapon in the play' is seen.

Trtiyantam kriyety etad vigrahe na prayujyate: the *third-case* form of *'kriya'* is *not employed* in the *analysis* — the instrumental 3.14.462 demanded never appears in any actual vigraha. Yatha dandah praharanam kridayam iti drsyate: as *'the staff is the weapon in the play'* is *seen* — the analysis of 'danda-praharanam kridayam', a compound in Panini's play-and-livelihood class (3.14.40). And the analysis has no instrumental in it. So the rule's supposed requirement is contradicted by the analyses the grammar itself gives. Drsyate: it is *seen* — the argument is observation of practice, as always (3.14.114).

3.14.468 unformalized

घविधौ घश्च संज्ञायामिति सूत्र उदाहृतम् ।
उपादानं प्रयोगेषु तस्यात्यन्तं न विद्यते ॥ ४६८ ॥

ghavidhau ghaśca saṃjñāyāmiti sūtra udāhṛtam |
upādānaṃ prayogeṣu tasyātyantaṃ na vidyate || 468 ||

In the rule for 'gha', 'and gha, in a name' is cited in the sutra; its taking-up is never found in usages.

Gha-vidhau ghas ca samjnayam iti sutra udahrtam: in the *rule for 'gha'*, *'ghas ca samjnayam'* is *cited* in the *sutra*. Upadanam prayogesu tasya atyantam na vidyate: its *taking-up* in *usages* is *never* found — atyantam na vidyate: not at all, ever. So the rule mentions a condition that no actual usage exhibits. This pairs with 3.14.467: there the analysis lacked what the rule demanded; here the usage lacks what the rule mentions. The grammar's apparatus and the language's practice are not in correspondence, and Bhartrhari has stopped treating each mismatch as a puzzle to solve (3.14.74's atidure).

3.14.469 unformalized

यैरप्रयुक्तैः संस्कारः प्रधानेषु प्रतीयते ।
ते भेदेऽपि विभक्तीनां निर्दिश्यन्त उपाधयः ॥ ४६९ ॥

yairaprayuktaiḥ saṃskāraḥ pradhāneṣu pratīyate |
te bhede'pi vibhaktīnāṃ nirdiśyanta upādhayaḥ || 469 ||

Those adjuncts by which, though unemployed, the formation is cognized in the principals — they are indicated, even where the case-endings differ.

Yair aprayuktaih samskarah pradhanesu pratiyate: those by which, though *unemployed*, the *formation* is *cognized* in the *principals* — aprayukta: never said (3.14.464's na kadacit prayujyate). Te bhede 'pi vibhaktinam nirdisyanta upadhayah: *they* are *indicated*, even where the *case-endings differ* — nirdisyante: they are pointed to, though not said. So the resolution: the sutra's 'kriya' and 'tena' are *indications* of unemployed adjuncts, not descriptions of forms that occur. The case-clash of 3.14.461 dissolves because the two words are not both describing the base — they are pointing at conditions that never appear. The rule is notation (3.14.78).

3.14.470 unformalized

समुदायेषु वर्तन्ते भावानां सहचारिणाम् ।
शब्दास्तत्त्वाविवक्षायां समुच्चयविकल्पयोः ॥ ४७० ॥

samudāyeṣu vartante bhāvānāṃ sahacāriṇām |
śabdāstattvāvivakṣāyāṃ samuccayavikalpayoḥ || 470 ||

Words operate in the aggregates of accompanying states, when reality is not intended to be expressed — in accumulation and in alternation.

Samudayesu vartante bhavanam sahacarinam sabdah: *words* operate in the *aggregates* of *accompanying states* — sahacarin: what goes along (3.14.31's bhinnesu sahacarisu; 3.14.222's sambandha-sahacarini). Tattva-avivaksayam: when *reality* is *not intended to be expressed* — tattva: the thing as it is; avivaksa: not meant to be said. Samuccaya-vikalpayoh: in *accumulation* and in *alternation* (3.14.191's two of the particle's senses). So words never touch the states themselves — they operate on bundles of what accompanies them, and they do so precisely where reality is not what one means to say. Tattva-avivaksayam is the phrase to keep: speech works by not intending the thing.

3.14.471 unformalized

मुच्चयस्तु क्रियते येषु प्रत्यर्थवृत्तिषु ।
भेदाधिष्ठानया योगस्तेषां भवति सङ्ख्यया ॥ ४७१ ॥

muccayastu kriyate yeṣu pratyarthavṛttiṣu |
bhedādhiṣṭhānayā yogasteṣāṃ bhavati saṅkhyayā || 471 ||

But those, operating in each meaning, in which accumulation is made — they have connection with a number based on difference.

Samuccayas tu kriyate yesu praty-artha-vrttisu: those, operating in *each meaning* — praty-artha-vrtti: distributively, one to each — in which *accumulation* is *made*. Bheda-adhisthanaya yogas tesam bhavati sankhyaya: they have *connection* with a *number* *based on difference* — bheda-adhisthana: resting on difference. So where the words go one-to-one and are then heaped, the number that results is a difference-number: a plural in the ordinary sense, not 3.14.99's abheda-ekatva. The contrast is exact: fuse the members and get non-different unity; distribute and heap them and get a plural. Which one obtains depends on the mode of operation, not on the things.

3.14.472 unformalized

सर्वैर्विशिष्टास्तैरर्थैर्जन्यन्ते सहचारिभिः ।
बुद्धयः प्रतिपत्तॄणां शब्दार्थांस्तानतो विदुः ॥ ४७२ ॥

sarvairviśiṣṭāstairarthairjanyante sahacāribhiḥ |
buddhayaḥ pratipattṝṇāṃ śabdārthāṃstānato viduḥ || 472 ||

The cognitions of comprehenders are produced as specified by all those accompanying meanings; therefore they know those to be the word-meanings.

Sarvair visistas tair arthair janyante sahacaribhih buddhayah pratipattrnam: the *cognitions* of *comprehenders* are *produced* as *specified* by *all* those *accompanying* meanings — every accompanying meaning shapes the cognition; none is left out. Sabdarthams tan ato viduh: *therefore* they *know those* to be the *word-meanings*. So what a word means is what shaped the cognition it produced — and since the accompaniers all shaped it, they are all the meaning. This is a striking criterion: the word-meaning is defined backwards from the cognition, and it is therefore as wide as the cognition's causes. Nothing is excluded for being merely contextual.

3.14.473 unformalized

संसृष्टाः प्रत्ययेष्वर्थाः सर्व एवोपकारिणः ।
तेषां प्रत्ययरूपेण सर्वेषां शब्दवाच्यता ॥ ४७३ ॥

saṃsṛṣṭāḥ pratyayeṣvarthāḥ sarva evopakāriṇaḥ |
teṣāṃ pratyayarūpeṇa sarveṣāṃ śabdavācyatā || 473 ||

In the cognitions the meanings are all conjoined, all assisting; of all of them there is expressibility by the word, in the form of the cognition.

Samsrstah pratyayesv arthah sarva eva upakarinah: in the *cognitions* the meanings are *all* *conjoined*, *all* *assisting* — samsrsta: mingled (3.14.92's samsrsteh; 3.14.367's vyatikirna). Tesam pratyaya-rupena sarvesam sabda-vacyata: of *all* of them there is *expressibility by the word*, *in the form of the cognition* — pratyaya-rupena: under the cognition's own shape. So everything mingled in the cognition is expressed — but only as the cognition has it, not as it is in itself. Sabda-vacyata is granted to all and qualified by pratyaya-rupena: words say everything, and say it all in the shape thought gave it. This is 3.14.470's tattva-avivaksa stated positively.

3.14.474 unformalized

केवलानां तु भावानां न रूपमवधार्यते ।
अनिरूपितरूपेषु तेषु शब्दो न वर्तते ॥ ४७४ ॥

kevalānāṃ tu bhāvānāṃ na rūpamavadhāryate |
anirūpitarūpeṣu teṣu śabdo na vartate || 474 ||

But of bare states, the form is not determined; and in them, whose form is undetermined, the word does not operate.

The conclusion, and it is one of the work's central sentences. Kevalanam tu bhavanam na rupam avadharyate: of *bare* states, the *form* is *not determined* — kevala: alone, uncombined; avadharyate: is ascertained, fixed. A thing by itself has no determinate form, because determination is what its accompaniers do (3.14.350's the principal has no own-form, from non-determination). Anirupita-rupesu tesu sabdo na vartate: in *them*, whose *form* is *undetermined*, the *word* does *not operate*. So words cannot reach anything bare — not because words are weak but because bare things have no shape to be reached. Every word-meaning is a meaning-in-company (3.14.472), and there is no such thing as the thing itself for language to fail at.

3.14.475 unformalized

पूर्वशब्दप्रयोगाच्च समूहान्न निवर्तते ।
वर्ततेऽवयवे वापि नोपात्तं त्यजति क्वचित् ॥ ४७५ ॥

pūrvaśabdaprayogācca samūhānna nivartate |
vartate'vayave vāpi nopāttaṃ tyajati kvacit || 475 ||

And from the employment of the earlier word, it does not withdraw from the aggregate; or else it operates in the member; it never abandons what has been taken up.

Purva-sabda-prayogac ca samuhan na nivartate: from the *employment of the earlier word*, it does *not withdraw* from the *aggregate*. Vartate 'vayave vapi: *or else* it operates in the *member*. Na upattam tyajati kvacit: it *never* *abandons* what has been *taken up* — kvacit: ever. So a word may sit on the whole or on the part, and whichever it took, it keeps: nothing it has once grasped is let go. This is the principle behind 3.14.264-266's paradox (the half-done thing is done and not done): an aggregate-word retains both applications, and the retention is what generates the contradiction. Here it is stated as a law: words do not release.

3.14.476 unformalized

समुदायाभिधायी च यदि भेदं विशेषयेत् ।
तत्रातुल्यविभक्तित्वं पूर्वकायादिवद् भवेत् ॥ ४७६ ॥

samudāyābhidhāyī ca yadi bhedaṃ viśeṣayet |
tatrātulyavibhaktitvaṃ pūrvakāyādivad bhavet || 476 ||

And if what expresses the aggregate were to qualify the difference — there would be non-equal case-endings, as in 'the fore-body' and the rest.

Samudayabhidhayi ca yadi bhedam visesayet: *if* what *expresses the aggregate* were to *qualify the difference* — the whole-word made into a qualifier of a part. Tatra atulya-vibhaktitvam purva-kayadivad bhavet: there would be *non-equal case-endings*, as in *'purva-kaya'* (the fore-part of the body) and the rest — the class of compounds where the members take different cases in the analysis ('kayasya purvam'). So the qualification would force a case-mismatch and turn every aggregate-word into a purva-kaya. The consequence is formal and it blocks the reading. This is the samuddesa's habitual test: propose a semantics, run it through the case-system, see what breaks.

3.14.477 unformalized

समूहे च प्रदेशे च पञ्चावा इति दृश्यते ।
तथा विशेषणं सर्व इत्येतदुपपद्यते ॥ ४७७ ॥

samūhe ca pradeśe ca pañcāvā iti dṛśyate |
tathā viśeṣaṇaṃ sarva ityetadupapadyate || 477 ||

In the aggregate and in the specific instance alike, 'five sheep' is seen; thus the qualifier 'all' is proper.

Samuhe ca pradese ca pancava iti drsyate: in the *aggregate* and in the *specific instance* alike, *'five sheep'* is *seen* — the phrase serves for the flock and for the individual beasts, indifferently: 3.14.475's the word does not abandon either. Tatha visesanam sarva ity etad upapadyate: *thus* the *qualifier* *'all'* is *proper* — 'all five sheep' construes, and it construes because the phrase already holds both applications. So 'sarva' can qualify because there is a plurality in there to be totalized, and there is a plurality because the aggregate-word kept its distributive grip. The little word 'sarva' is evidence for the law of 3.14.475.

3.14.478 unformalized

तथार्धपिप्लीत्यत्र जात्यन्तरनिवृत्तये ।
अर्धं च पिप्पली चेति खण्डे शब्दः प्रयुज्यते ॥ ४७८ ॥

tathārdhapiplītyatra jātyantaranivṛttaye |
ardhaṃ ca pippalī ceti khaṇḍe śabdaḥ prayujyate || 478 ||

So in 'half a pepper': for the exclusion of another class, 'a half, and a pepper', the word is employed for the piece.

Tatha ardha-pippali ity atra: so in *'ardha-pippali'* — half a long-pepper. Jaty-antara-nivrttaye: for the *exclusion of another class* — the half must still be pepper and not something else: cut a pepper in half and each half is pepper (against 3.14.298's dewlapless cow, which was a different animal). Ardham ca pippali ca iti khande sabdah prayujyate: *'a half, and a pepper'* — the *word* is *employed for the piece*. So 'pippali' applies to the fragment, and the compound's analysis makes both members co-refer to it. The class survives the cutting, and it survives because 3.14.475's word does not release its grip on the member when the whole is broken.

3.14.479 unformalized

मात्राणां हि तिरोभावे परिमाणं न विद्यते ।
कुमार्य इति तेन स्यात् कुमार्यां भेदसम्भवात् ॥ १७९ ॥

mātrāṇāṃ hi tirobhāve parimāṇaṃ na vidyate |
kumārya iti tena syāt kumāryāṃ bhedasambhavāt || 179 ||

For in the disappearance of the measures, the measure does not exist; therefore 'kumaryah' would come about, from the possibility of difference in 'kumari'.

This verse is identical to 3.14.179, and our source numbers it '179' here — the second such doubling in this samuddesa (see 3.14.421, which repeats 3.14.121). Its argument belongs to the number-analysis of 3.14.175-179, where it drew the consequence that on the universal-account the plural must come from the individual girls. It has no work to do among the aggregate-and-part verses that surround it here. We print it where the source prints it, and record the repetition.
contested: The doubling at 3.14.421/121 and here at 3.14.479/179 stands at an interval of 300 verses in both cases, which suggests a mechanical fault in the transmission — a repeated exemplar-page or a systematic mis-numbering — rather than two independent scribal slips. A reader with the critical editions to hand should establish whether either verse is attested at its second position.

3.14.480 unformalized

असिद्धास्तु विशेषेण समुदाये व्यवस्थिताः ।
प्रदेशे दर्शनं तेषामर्थप्रकरणादिभिः ॥ ४८० ॥

asiddhāstu viśeṣeṇa samudāye vyavasthitāḥ |
pradeśe darśanaṃ teṣāmarthaprakaraṇādibhiḥ || 480 ||

But not being established specifically, they stand settled in the aggregate; their being seen in the specific instance is by meaning, context and the rest.

Asiddhas tu visesena samudaye vyavasthitah: *not established specifically*, they stand *settled in the aggregate* — a word's default seat is the whole, and the particular application is not fixed by anything in the word. Pradese darsanam tesam artha-prakaranadibhih: their being *seen* in the *specific instance* is by *meaning*, *context* and the rest — the meaning-fixing conditions (3.14.444's arthat prakaranad vapi; 2.316's classical list). So: aggregate by default, particular by context. This is the settled form of 3.14.475's the word never abandons what it took: it holds the whole always and the part when the situation asks.

3.14.481-482 unformalized

यदुपव्यञ्जनं जातेः सहचारि च कर्मसु ।
तत्र वा रूढसम्बन्धं यत् प्रायेणापलक्षितम् ॥ ४८१ ॥
समुदायः प्रदेशो वेत्येवं तस्मिन्ननाश्रिते ।
अर्थात्मन्यविशेषेण वर्तन्ते ब्राह्मणादयः ॥ ४८२ ॥

yadupavyañjanaṃ jāteḥ sahacāri ca karmasu |
tatra vā rūḍhasambandhaṃ yat prāyeṇāpalakṣitam || 481 ||
samudāyaḥ pradeśo vetyevaṃ tasminnanāśrite |
arthātmanyaviśeṣeṇa vartante brāhmaṇādayaḥ || 482 ||

(481) That indication of the universal, and the accompanier in the actions, or what has a conventional connection there, which is for the most part unmarked — (482) 'aggregate or specific instance': thus, when that is not resorted to, 'brahmana' and the rest operate in the meaning-self without specification.

(481): yad upavyanjanam jateh sahacari ca karmasu — the *indication* of the *universal* (upavyanjana: 3.14.83's accompanying sign), and the *accompanier* in the *actions*. Tatra va rudha-sambandham yat prayena apalaksitam: or what has a *conventional connection* there (3.14.440's rudha-yoga), which is *for the most part unmarked* — prayena apalaksita: mostly leaving no mark (3.14.140's alaksitah). (482): samudayah pradeso va ity evam tasminn anasrite — *'aggregate or specific instance'*: when *that* — the choice — is *not resorted to*. Arthatmany avisesena vartante brahmanadayah: *'brahmana'* and the rest operate in the *meaning-self* *without specification*. So where nothing forces the choice, the word does not make it: it sits on the meaning-self, unspecified. Avisesena is the resting state of every word, and specification is always something done to it.

3.14.483 unformalized

यश्च तुल्यश्रुतिर्द्दष्टः समुदाये व्यवस्थितः ।
तेनोपचरितैकत्वं प्रदेशेऽप्युपलभ्यते ॥ ४८३ ॥

yaśca tulyaśrutirddaṣṭaḥ samudāye vyavasthitaḥ |
tenopacaritaikatvaṃ pradeśe'pyupalabhyate || 483 ||

And that which, being of like hearing, is seen settled in the aggregate — by it a figurative oneness is apprehended in the specific instance too.

Yas ca tulya-srutir drstah samudaye vyavasthitah: that which, being of *like hearing* — the same audible word — is *seen settled* in the *aggregate*. Tena upacaritaikatvam pradese 'py upalabhyate: by *it*, a *figurative oneness* is *apprehended* in the *specific instance* too — upacarita-ekatva: a oneness by transfer (3.14.284's upacaryate). So because the word sounds the same on the whole and on the part, one takes the two to be one — figuratively. The identity of whole and part is an effect of the word's identity: the sound is one, and so the thing seems one. This is 3.14.325's reflection in a homelier setting — a likeness of form producing an apparent transfer.

3.14.484 unformalized

संस्कारादुपघाताद् वा वृत्तोऽक्तपरिमाणके ।
तैलादौ जातिशब्दोऽत्र सामर्थ्यादवसीयते ॥ ४८४ ॥

saṃskārādupaghātād vā vṛtto'ktaparimāṇake |
tailādau jātiśabdo'tra sāmarthyādavasīyate || 484 ||

From preparation or from damage, operating in what has unstated measure — the universal-word in 'oil' and the rest is here determined from capacity.

Samskarad upaghatad va vrtto 'kta-parimanake: from *preparation* or from *damage*, operating in what has *unstated measure* — akta-parimanaka: no quantity said. Tailadau jati-sabdo 'tra samarthyad avasiyate: the *universal-word* in *'oil'* and the rest is *determined* *from capacity*. So 'oil' names a kind and no amount, and whether a given use means the pot of it or the drop is settled by capacity — by whether it was prepared or spoiled, which tells you how much was in question. Mass-terms have no built-in individuation and the context supplies it (3.14.480's artha-prakaranadibhih). The verse is a small, exact observation about a class of nouns the grammar had no separate category for.

3.14.485 unformalized

न जातिगुणशब्देषु मूर्तिभेदो विवक्षितः ।
ते जातिगुणसम्बन्धभेदमात्रनिबन्धनाः ॥ ४८५ ॥

na jātiguṇaśabdeṣu mūrtibhedo vivakṣitaḥ |
te jātiguṇasambandhabhedamātranibandhanāḥ || 485 ||

In universal-words and quality-words, difference of bodily form is not intended; they rest on mere difference of connection with universal and with quality.

Na jati-guna-sabdesu murti-bhedo vivaksitah: in *universal-words* and *quality-words*, *difference of bodily form* is *not intended* — murti: the shaped body; and it is avivaksita (3.14.176's device). Te jati-guna-sambandha-bheda-matra-nibandhanah: they rest on *mere difference of connection* with *universal* and with *quality*. Bheda-matra: nothing but the difference in the connecting. So 'cow' and 'white' do not care about bodies: what differentiates their applications is only how the kind or the quality is connected, and the physical individual is not in view. This is why 3.14.484's oil worked — a mass-term has no murti to be differentiated by, and none of these words needed one.

3.14.486 unformalized

कृष्णादिव्यपदेशश्च सर्वावयववृत्तिभिः ।
गुणैस्तेऽप्येकदेशस्थाः पटादीनां विशेषकाः ॥ ४८६ ॥

kṛṣṇādivyapadeśaśca sarvāvayavavṛttibhiḥ |
guṇaiste'pyekadeśasthāḥ paṭādīnāṃ viśeṣakāḥ || 486 ||

And the designation 'black' and the rest is by qualities operating in all the members; those too, residing in one part, are differentiators of cloths and the rest.

Krsnadi-vyapadesas ca sarva-avayava-vrttibhih gunaih: the designation *'black'* and the rest is by *qualities* operating in *all the members* — the cloth is black because every thread is. Te 'py ekadesa-sthah patadinam visesakah: *those too*, residing in *one part*, are *differentiators* of *cloths* and the rest — a quality in a single part still differentiates the whole cloth. So the quality qualifies from everywhere and from anywhere: total presence and partial presence both do the job. Which is 3.14.475's law again — the word does not release either application — now for quality-terms.

3.14.487 unformalized

पटावयववृत्तास्तु यदा तत्र पटादयः ।
तदा तैलादिवत् तेषां जातिशब्दत्वमुच्यते ॥ ४८७ ॥

paṭāvayavavṛttāstu yadā tatra paṭādayaḥ |
tadā tailādivat teṣāṃ jātiśabdatvamucyate || 487 ||

But when 'cloth' and the rest operate in the cloth's members — then, like 'oil' and the rest, their being universal-words is stated.

Pata-avayava-vrttas tu yada tatra patadayah: when *'pata'* and the rest operate in the *cloth's members* — one calls a scrap of cloth 'cloth'. Tada tailadivat tesam jati-sabdatvam ucyate: *then*, like *'taila'* and the rest, their being *universal-words* is stated. So 'cloth' behaves as a mass-term when it applies to its own fragments: any piece of cloth is cloth, as any drop of oil is oil (3.14.484). The parallel is sharp — a count-noun turns into a mass-noun when it distributes over parts, and the grammar registers the shift by calling it a jati-sabda. The observation is precise and modern-sounding, and it is 3.14.478's ardha-pippali generalized.

3.14.488 unformalized

निवृत्त्यर्था श्रुतिर्येषां भेदस्तेष्वनपेक्षितः ।
प्रदेसे समुदाये वा गुणोऽन्येषां निवर्तकः ॥ ४८८ ॥

nivṛttyarthā śrutiryeṣāṃ bhedasteṣvanapekṣitaḥ |
pradese samudāye vā guṇo'nyeṣāṃ nivartakaḥ || 488 ||

Those whose hearing has exclusion for its purpose — in them difference is not required; in the specific instance or in the aggregate, the quality is the excluder of others.

Nivrtty-artha srutir yesam bhedas tesv anapeksitah: those whose *hearing* has *exclusion* for its *purpose* — in them *difference* is *not required*: where a word is there to exclude, one need not settle which thing it applies to (3.14.142's yad yad asriyate tat tad anyasya vinivartakam). Pradese samudaye va guno 'nyesam nivartakah: in the *specific instance* *or* in the *aggregate*, the *quality* is the *excluder of others*. So the exclusion works either way — on the part or on the whole, indifferently — because excluding does not require knowing what one is excluding *from*. This is why 3.14.482's avisesena was tolerable: an unspecified word still excludes.

3.14.489 unformalized

ब्राह्मणाध्ययेन तत्र वर्तते ब्राह्मणश्रुतिः ।
सादृश्यं तत्र दृष्टं हि क्षत्रियाध्ययनादिभिः ॥ ४८९ ॥

brāhmaṇādhyayena tatra vartate brāhmaṇaśrutiḥ |
sādṛśyaṃ tatra dṛṣṭaṃ hi kṣatriyādhyayanādibhiḥ || 489 ||

There, the word 'brahmana' operates by the brahmin's studying; for similarity is seen there with the ksatriya's studying and the rest.

The vati-comparison, resumed from 3.14.431. Brahmana-adhyayanena tatra vartate brahmana-srutih: the word *'brahmana'* — in 'brahmanavat', in the brahmin's manner — operates *by the brahmin's studying*. Sadrsyam tatra drstam hi ksatriya-adhyayanadibhih: for *similarity* is *seen* there with the *ksatriya's studying* and the rest. So the word reaches the brahmin's studying and the comparison is between the two studyings. This is 3.14.443's structure — one action, two seats — and the question 3.14.490 asks is what makes the two studyings alike.

3.14.490 unformalized

ब्राह्मणाध्ययने वृत्तिर्यदि स्याद् ब्राह्मणश्रुतेः ।
वक्तव्यं केन धर्मेण तुल्यत्वं क्रिययोरिति ॥ ४९० ॥

brāhmaṇādhyayane vṛttiryadi syād brāhmaṇaśruteḥ |
vaktavyaṃ kena dharmeṇa tulyatvaṃ kriyayoriti || 490 ||

If the word 'brahmana' operated in the brahmin's studying — it would have to be stated by what property the two actions are similar.

Brahmana-adhyayane vrttir yadi syad brahmana-sruteh: *if* the word *'brahmana'* *operated in* the *brahmin's studying*. Vaktavyam kena dharmena tulyatvam kriyayor iti: it would have to be *stated* *by what property* the *two actions* are *similar*. Vaktavyam: it must be said — and it cannot be (3.14.442: same class, nothing to compare; different class, nothing shared). So locating the word in the action leaves the resemblance groundless. The demand kena dharmena is unanswerable, and 3.14.491 responds by moving the word off the action altogether.

3.14.491 unformalized

अध्येतरि यदा वृत्तिरुच्यते ब्राह्मणश्रुतेः ।
निमित्तत्वं तदोपैति क्रियैवाध्येतरि स्थिता ॥ ४९१ ॥

adhyetari yadā vṛttirucyate brāhmaṇaśruteḥ |
nimittatvaṃ tadopaiti kriyaivādhyetari sthitā || 491 ||

When the word 'brahmana' is said to operate in the studier — then the action residing in the studier is itself what attains causality.

The resolution. Adhyetari yada vrttir ucyate brahmana-sruteh: when the word *'brahmana'* is said to operate in the *studier* — the person, not the studying: which is what the word means anyway. Nimittatvam tada upaiti kriya eva adhyetari sthita: *then* the *action* *residing in the studier* *itself* attains *causality* — nimittatva (3.14.113's word). So the word sits on the man and the action, seated in him, does the occasioning. 3.14.490's unanswerable question does not arise, because the two actions are never compared directly: each is in its man, and the men are compared. This is 3.14.431's opening claim, arrived at by exhausting the alternative.

3.14.492 unformalized

सिंहशब्देन संबन्धे शौर्यमात्राभिधायिना ।
चैत्रात् षष्ठी प्रसज्येत योगे शौर्यादिभिर्यथा ॥ ४९२ ॥

siṃhaśabdena saṃbandhe śauryamātrābhidhāyinā |
caitrāt ṣaṣṭhī prasajyeta yoge śauryādibhiryathā || 492 ||

In connection with the word 'lion' as expressing mere valour — the genitive would follow after 'Caitra', as in connection with 'valour' and the rest.

Simha-sabdena sambandhe saurya-matrabhidhayina: in connection with the word *'lion'* as expressing *mere valour* — if 'simha' in 'caitra-simha' just means bravery. Caitrat sasthi prasajyeta yoge sauryadibhir yatha: the *genitive* would *follow* after *'Caitra'*, as in connection with *'valour'* and the rest. So one would have to say 'caitrasya simhah' — Caitra's lion — as one says 'caitrasya sauryam', Caitra's valour: abstract nouns take a genitive possessor. And one does not. So 'simha' does not mean mere valour: it means the animal, and the man is called it (3.14.428's the word standing for the man-meaning). The case-system settles a semantic question in one line.

3.14.493 unformalized

ब्राह्मणायेव दातव्यं वैश्यायेत्येवमादिषु ।
सम्प्रदानादियोगश्च क्रियामात्रे न कल्पते ॥ ४९३ ॥

brāhmaṇāyeva dātavyaṃ vaiśyāyetyevamādiṣu |
sampradānādiyogaśca kriyāmātre na kalpate || 493 ||

'It should be given to the vaisya as to a brahmin' — in such cases, connection with a recipient and the rest is not conceivable in a mere action.

Brahmanaya iva datavyam vaisyaya ity evam adisu: *'it should be given to the vaisya as to a brahmin'* and the like — the vati-comparison with a dative. Sampradanadi-yogas ca kriya-matre na kalpate: connection with a *recipient* and the rest is *not conceivable* in a *mere action* — sampradana: the karaka of the one given to (3.7's the karakas). So the dative in 'brahmanaya iva' needs a person to be the recipient, and a bare action cannot be one. Hence again the word must be sitting on the man (3.14.491). The karaka-system, like the case-system at 3.14.492, will not let an action stand where a person is wanted.

3.14.494 unformalized

क्रियामात्राभिधायित्वादव्ययेषु वतेर्न च ।
पाठः कदाचित् कर्तव्यस्तुल्यौ पक्षावुभौ यतः ॥ ४९४ ॥

kriyāmātrābhidhāyitvādavyayeṣu vaterna ca |
pāṭhaḥ kadācit kartavyastulyau pakṣāvubhau yataḥ || 494 ||

And from its expressing mere action, the listing of 'vati' among the indeclinables should never be made; since both alternatives are equal.

Kriya-matrabhidhayitvad avyayesu vater na ca pathah kadacit kartavyah: from its expressing *mere action*, the *listing* of *'vati'* among the *indeclinables* should *never* be *made* — patha: the recitation, the entry in the gana. Tulyau paksav ubhau yatah: since *both alternatives* are *equal*. So: if 'vati' expressed a bare action it would be an avyaya and belong in the list; and it does not, so it should not be listed. And the reason offered is that the two positions are *equal* — tulyau: neither is better. So Bhartrhari declines to list it because the question is undecidable, not because he has decided it. The Vrtti-samuddesa's long examination of the vati-rule ends with a shrug that is also a principle: where the alternatives are equal, do not legislate.

3.14.495 unformalized

जहाति जातिं द्रव्यं वा तस्मान्नावयवे स्थितः ।
क्रियायास्तु श्रुतिर्यस्मात् तद्वत्यर्थेऽवतिष्ठते ॥ ४९५ ॥

jahāti jātiṃ dravyaṃ vā tasmānnāvayave sthitaḥ |
kriyāyāstu śrutiryasmāt tadvatyarthe'vatiṣṭhate || 495 ||

It abandons the universal or the substance; therefore it does not reside in the member. But since it is the hearing of an action, it stands in the meaning possessing that.

Jahati jatim dravyam va tasman na avayave sthitah: it *abandons* the *universal* or the *substance*; *therefore* it does *not reside* in the *member* — jahati: lets go (3.14.94's tyaga, the jahat-svartha doctrine). In 'brahmanavat' the word has dropped both what it named and what it meant, so it is not sitting on the compound's member. Kriyayas tu srutir yasmat tadvaty arthe 'vatisthate: *but* since it is the *hearing of an action*, it stands in the meaning *possessing* that. Tadvati arthe: in the *possessor*. So having let go of universal and substance, the word lands on the one who has the action — 3.14.491's conclusion, restated as the word's own movement.

3.14.496 unformalized

आक्रियाणां निवृत्त्यर्था यतश्चात्र क्रियाश्रुतिः ।
क्रियोपलक्षिते तस्मात् क्रियाशब्दः प्रतीयते ॥ ४९६ ॥

ākriyāṇāṃ nivṛttyarthā yataścātra kriyāśrutiḥ |
kriyopalakṣite tasmāt kriyāśabdaḥ pratīyate || 496 ||

And since here the hearing of the action has the exclusion of non-actions for its purpose — therefore the action-word is cognized in what is indicated by the action.

Akriyanam nivrtty-artha yatas ca atra kriya-srutih: since here the *hearing of the action* has for its *purpose* the *exclusion of non-actions* — nivrtty-artha: it is there to turn things away (3.14.488's nivrtty-artha sruti). The rule's word 'kriya' is not describing; it is fencing out. Kriya-upalaksite tasmat kriya-sabdah pratiyate: *therefore* the *action-word* is cognized in what is *indicated by the action* — upalaksita: marked out by it, not characterized (3.14.170). So 'kriya' in the rule excludes non-actions and indicates a possessor, and it never means an action at all. The rule's most puzzling word (3.14.434) turns out to be a fence and a pointer.

3.14.497 unformalized

होतव्यादिषु यस्माच्च क्रियान्या ब्राह्मणादिवत् ।
अपेक्षणीया शुद्धेऽर्थे तस्माद् वृत्तिर्न कस्यचित् ॥ ४९७ ॥

hotavyādiṣu yasmācca kriyānyā brāhmaṇādivat |
apekṣaṇīyā śuddhe'rthe tasmād vṛttirna kasyacit || 497 ||

And since in 'hotavya' and the rest another action, like 'brahmana' and the rest, has to be required — therefore, in the pure meaning, there is no operation of anything.

Hotavyadisu yasmac ca kriya anya brahmanadivat apeksaniya: in *'hotavya'* and the rest, *another action*, *like 'brahmana'* and the rest, has to be *required* — one must reach for a second action, as one reaches for a second person. Suddhe 'rthe tasmad vrttir na kasyacit: *therefore*, in the *pure meaning*, there is *no operation of anything* — na kasyacit: of nothing whatever. So nothing operates in a bare, unaccompanied meaning. This is 3.14.474's kevalanam tu bhavanam na rupam avadharyate, arriving as the conclusion of the vati-analysis: purity is not a state anything can be in and be spoken of.

3.14.498 unformalized

सर्वं वाप्येकदेशो वा यस्मिन्नाश्रीयते क्वचित् ।
विशेषवृत्तिं तं सर्वमाहुर्भेदे व्यवस्थितम् ॥ ४९८ ॥

sarvaṃ vāpyekadeśo vā yasminnāśrīyate kvacit |
viśeṣavṛttiṃ taṃ sarvamāhurbhede vyavasthitam || 498 ||

Whether the whole or a part is what is resorted to somewhere — that whose operation is in the particular, all call settled in difference.

Sarvam vapy ekadeso va yasminn asriyate kvacit: *whether* the *whole* or a *part* is *resorted to* somewhere — either way (3.14.475's the word does not release either). Visesa-vrttim tam sarvam ahur bhede vyavasthitam: that whose operation is *in the particular*, *all* call *settled in difference*. So the criterion of being 'in difference' is operating in a particular, regardless of whether the particular is a whole or a part — the whole/part question is orthogonal to the general/particular one. This clears an ambiguity that has been latent since 3.14.475, and it does so by noting that everyone (sarvam ahuh) already uses the criterion this way.

3.14.499 unformalized

समुच्चयो विकल्पो वा प्रकाराः सर्व एव वा ।
विशेषा इति वर्ण्यन्ते सामान्यं वाऽविकल्पितम् ॥ ४९९ ॥

samuccayo vikalpo vā prakārāḥ sarva eva vā |
viśeṣā iti varṇyante sāmānyaṃ vā'vikalpitam || 499 ||

Accumulation, or alternation, or all the modes — are described as particulars; or else the general, unconstructed.

Samuccayo vikalpo va prakarah sarva eva va: *accumulation*, or *alternation*, or *all the modes* (3.14.191's senses of the particle). Visesa iti varnyante: are described as *particulars*. Samanyam va avikalpitam: *or* the *general*, *unconstructed* — avikalpita: not built up, not made into an alternative (3.14.144's vikalpita, negated). So the modes of combination are themselves either particulars or an unconstructed general, and there is nothing else they could be. The samuddesa is placing its own machinery on the general-particular scale it has been using throughout: accumulation and alternation are particulars, and what is left when one stops constructing is the general.

3.14.500 unformalized

न हि ब्राह्मण इत्यत्र भेदः कश्चिदुपाश्रितः ।
अपाकृतो वा तेनायं समुदाये व्यवस्थितः ॥ ५०० ॥

na hi brāhmaṇa ityatra bhedaḥ kaścidupāśritaḥ |
apākṛto vā tenāyaṃ samudāye vyavasthitaḥ || 500 ||

For in 'brahmana' no difference is resorted to, nor is any rejected; therefore this stands settled in the aggregate.

Na hi brahmana ity atra bhedah kascid upasritah apakrto va: in *'brahmana'* *no* difference is *resorted to*, *nor* is any *rejected* — neither taken up nor thrown out: the word simply does not engage the question. Tena ayam samudaye vyavasthitah: *therefore* this stands *settled in the aggregate*. So the default seat (3.14.480's asiddhas tu visesena samudaye vyavasthitah) is where a word rests when it has done nothing about difference at all — not chosen the whole, but not chosen anything, which comes to resting on the whole. Neither upasrita nor apakrta: the word's indifference is its position.

3.14.501 unformalized

क्रिया त्वाश्रीयते यस्मिन् स भेदोऽध्यवसीयते ।
तथान्यथा सर्वथा चेत्यप्रयोगे न विद्यते ॥ ५०१ ॥

kriyā tvāśrīyate yasmin sa bhedo'dhyavasīyate |
tathānyathā sarvathā cetyaprayoge na vidyate || 501 ||

But that in which the action is resorted to — that difference is determined; 'thus', 'otherwise', 'in every way' — in non-employment it does not exist.

Kriya tv asriyate yasmin sa bhedo 'dhyavasiyate: that in which the *action* is *resorted to* — *that* difference is *determined*: the action's seat is what fixes which particular is meant, where 'brahmana' alone fixed nothing (3.14.500). Tatha anyatha sarvatha ceti aprayoge na vidyate: *'thus'*, *'otherwise'*, *'in every way'* — in *non-employment* it does *not exist*. So the manner-words: unless one of them is *said*, no manner is there. Aprayoge na vidyate — where it is not employed, it does not exist: 3.14.128's what withdraws has no being, applied to modes. The action seats the word, and only an uttered manner-word gives it a manner.

3.14.502 unformalized

उपमाने क्रियावृत्तिमुपमेये क्रियाश्रुतिः ।
प्रत्याययन्ती भेदस्य करोतीव पदार्थताम् ॥ ५०२ ॥

upamāne kriyāvṛttimupameye kriyāśrutiḥ |
pratyāyayantī bhedasya karotīva padārthatām || 502 ||

The hearing of the action, conveying in the comparandum the action's operation in the comparans — as it were makes the difference a word-meaning.

Upamane kriya-vrttim upameye kriya-srutih pratyayayanti: the *hearing of the action*, *conveying* — in the *comparandum* — the *action's operation* in the *comparans*. Bhedasya karoti iva padarthatam: *as it were* *makes* the *difference* a *word-meaning*. Iva: as though. So the vati-word transfers the standard's action onto the compared thing, and in doing so it *seems* to make the difference itself into something meant. But only iva: the difference is not a word-meaning, and the appearance that it is comes from the transfer. This is 3.14.465's finding (the relation is said by no word) protected against the illusion that 'brahmanavat' finally says it.

3.14.503 unformalized

व्यापारेणैव सादृश्ये व्यापारस्य विवक्षिते ।
क्रियावद् वचनाच्छब्दात् प्रत्ययः प्रतिपाद्यते ॥ ५०३ ॥

vyāpāreṇaiva sādṛśye vyāpārasya vivakṣite |
kriyāvad vacanācchabdāt pratyayaḥ pratipādyate || 503 ||

When the similarity of an operation with an operation is intended — the affix is taught from a word expressing something possessing action.

Vyaparena eva sadrsye vyaparasya vivaksite: when the *similarity* of an *operation* *with an operation* is *intended* — the case the whole vati-analysis is about. Kriyavad vacanac chabdat pratyayah pratipadyate: the *affix* is *taught* from a word expressing something *possessing action* — kriyavat: an action-*possessor*, not an action. So even where two operations are being compared, the affix attaches to a word for a possessor. The rule never touches an action directly, and 3.14.434's paradox — a nominal stem expressing an action — never arises, because the stem expresses a possessor and the action rides in him (3.14.491).

3.14.504 unformalized

क्रियावतोऽपि सादृश्ये वक्तुमिष्टे क्रियावता ।
अध्येता ब्राह्मण इव प्रत्ययो न निवर्तते ॥ ५०४ ॥

kriyāvato'pi sādṛśye vaktumiṣṭe kriyāvatā |
adhyetā brāhmaṇa iva pratyayo na nivartate || 504 ||

Even when the similarity of an action-possessor with an action-possessor is desired to be stated — 'the student is like a brahmin' — the affix does not withdraw.

Kriyavato 'pi sadrsye vaktum iste kriyavata: even when the similarity of an *action-possessor* with an *action-possessor* is desired to be stated. Adhyeta brahmana iva: *'the student is like a brahmin'* — persons compared to persons, no actions in view. Pratyayo na nivartate: the *affix* does *not withdraw*. So the vati-affix serves this case too: it does not require actions to be the compared terms. Between 3.14.503 and this verse the affix's range is fixed — it attaches to possessor-words always, and what is compared may be their actions or themselves. The rule is simpler than the debate about it.

3.14.505 unformalized

अधीते तुल्य इत्येवं पुंल्लिङ्गेन विशेषणम् ।
क्रियावति क्रियायां तु तुल्यशब्दे नपुंसकम् ॥ ५०५ ॥

adhīte tulya ityevaṃ puṃlliṅgena viśeṣaṇam |
kriyāvati kriyāyāṃ tu tulyaśabde napuṃsakam || 505 ||

'He studies similarly' — thus the qualifier is masculine, for the action-possessor; but in the action, the word 'tulya' is neuter.

A grammatical test, and it is decisive. Adhite tulya ity evam pumllingena visesanam: *'adhite tulyah'* — he studies, similar — the *qualifier* is *masculine*, agreeing with the *action-possessor*: the man. Kriyavati kriyayam tu tulya-sabde napumsakam: but in the *action*, the word *'tulya'* is *neuter* — 'tulyam adhite', he studies similarly, where the neuter agrees with the studying. So the gender of 'tulya' shows which term the similarity is predicated of, and both are available. The language distinguishes what the theorists were arguing about, by one vowel. This is the samuddesa's best instrument: where a doctrine is in doubt, look at the concord.

3.14.506 unformalized

प्रकृत्यर्थे विशिष्टेऽपि प्रत्ययार्थाविशेषणात् ।
पुत्रेण तुल्यः कपिल इति वृत्तिः प्रसज्यते ॥ ५०६ ॥

prakṛtyarthe viśiṣṭe'pi pratyayārthāviśeṣaṇāt |
putreṇa tulyaḥ kapila iti vṛttiḥ prasajyate || 506 ||

Even though the base-meaning is specified, from the affix-meaning's not being qualified — 'Kapila is like a son' — the derivation would follow.

Prakrty-arthe visiste 'pi pratyayartha-avisesanat: even though the *base-meaning* is *specified*, from the *affix-meaning's* *not being qualified*. Putrena tulyah kapila iti vrttih prasajyate: *'Kapila is like a son'* — the *derivation* would *follow*. So one would get 'putravat kapilah' meaning Kapila resembles a son — in what respect, nothing says, because the affix-meaning was left unqualified. The over-generation is real: the affix will attach wherever a comparison is wanted, and it does not require the respect to be specified. This is the vati-rule's looseness, and 3.14.507-508 explain why it is tolerable.

3.14.507 unformalized

याः पुत्रे रूढसम्बन्धाः क्रियालोके विवक्षिताः ।
ताभिः क्रियावतः पुत्राद् गुणतुल्ये वतिर्भवेत् ॥ ५०७ ॥

yāḥ putre rūḍhasambandhāḥ kriyāloke vivakṣitāḥ |
tābhiḥ kriyāvataḥ putrād guṇatulye vatirbhavet || 507 ||

Those actions conventionally connected with a son, intended in the world — by those, 'vati' would come after 'putra', of an action-possessor, in what is similar in quality.

Yah putre rudha-sambandhah kriya loke vivaksitah: those *actions* *conventionally connected* with a *son*, *intended in the world* — rudha-sambandha (3.14.440's rudha-yoga): what everyone associates with sonhood — obedience, service, care in old age. Tabhih kriyavatah putrad gunatulye vatir bhavet: *by those*, *'vati'* would come after *'putra'*, of an *action-possessor*, in what is *similar in quality*. So 'putravat' means: doing what sons conventionally do. The looseness of 3.14.506 is filled by convention — the respect is not stated because everyone knows it (3.14.415's the moon's beauty and not its eclipses). Loke vivaksitah: what the world means by it.

3.14.508 unformalized

अन्तर्भूतं निमित्तं च रूढिशब्देषु यद्यपि ।
क्रियास्तु सहचारिण्यो रूढाः सन्ति पदार्थवत् ॥ ५०८ ॥

antarbhūtaṃ nimittaṃ ca rūḍhiśabdeṣu yadyapi |
kriyāstu sahacāriṇyo rūḍhāḥ santi padārthavat || 508 ||

Although the cause is included in conventional words — yet the accompanying actions, conventionally established, exist like word-meanings.

Antarbhutam nimittam ca rudhi-sabdesu yady api: although the *cause* is *included* — antarbhuta: swallowed inside (3.14.311's abhyantarikrta) — in *conventional words*: a rudhi-word has digested whatever occasioned it and shows nothing (3.14.55's saptaparna). Kriyas tu sahacarinyo rudhah santi padarthavat: *yet* the *accompanying actions*, *conventionally established*, *exist* *like word-meanings* — padarthavat: as good as meanings. So even in a word whose ground is buried, the associated actions are there and available, functioning as if they were the meaning. This is what makes 3.14.507's convention usable: it is not vague background, it is as determinate as a word-meaning, and it is why 'putravat' conveys something exact.

3.14.509 unformalized

क्रमं तु यदि बाधित्वा प्रत्ययार्थविशेषणम् ।
प्रधानानुग्रहात् साम्याद् विभक्तेश्चावतिष्ठते ॥ ५०९ ॥

kramaṃ tu yadi bādhitvā pratyayārthaviśeṣaṇam |
pradhānānugrahāt sāmyād vibhakteścāvatiṣṭhate || 509 ||

But if, setting aside the sequence, the qualification of the affix-meaning stands — from the favouring of the principal, from equality, and from the case-ending —

Kramam tu yadi badhitva pratyayartha-visesanam avatisthate: if, *setting aside* the *sequence* — badhitva: having annulled it (3.14.216's badhaka) — the *qualification of the affix-meaning* *stands*. And three reasons why it might: pradhana-anugrahat — from the *favouring of the principal*; samyat — from *equality*; vibhakteh — from the *case-ending*. So one could let the affix-meaning be qualified out of order, and there would be grounds for it. The verse is a protasis and 3.14.510 supplies what follows — which is a mess.

3.14.510 unformalized

प्रकृतेरविशिष्टत्वात् क्रियातुल्ये प्रसज्यते ।
पुत्रादौ गुणशब्देभ्यः पूर्वोक्तस्य विपर्यये ॥ ५१० ॥

prakṛteraviśiṣṭatvāt kriyātulye prasajyate |
putrādau guṇaśabdebhyaḥ pūrvoktasya viparyaye || 510 ||

From the base's not being specified, it would follow in what is similar in action; in 'putra' and the rest, from quality-words, in the reverse of what was stated before.

The apodosis of 3.14.509, and it is a reductio. Prakrter avisistatvat kriya-tulye prasajyate: from the *base's* *not being specified*, it would *follow* in what is *similar in action*. Putradau guna-sabdebhyah purvoktasya viparyaye: in *'putra'* and the rest, from *quality-words*, in the *reverse* of what was *stated before*. Viparyaya: the inversion. So setting aside the sequence inverts 3.14.506-507's result exactly: the affix would attach to quality-words and mean action-similarity, where before it attached to action-words and meant quality-similarity. Everything comes out backwards. So the sequence must not be set aside — and the argument for keeping it is that abandoning it turns the rule inside out.

3.14.511 unformalized

स्थूलेन तुल्यो यातीति बहिरङ्गा क्रियाश्रुतिः ।
अनिमित्तं वतेस्तुल्यं यातीत्यत्रेष्यते वति ॥ ५११ ॥

sthūlena tulyo yātīti bahiraṅgā kriyāśrutiḥ |
animittaṃ vatestulyaṃ yātītyatreṣyate vati || 511 ||

'He goes like a stout man' — there the hearing of the action is external, no occasion for 'vati'; in 'he goes similarly', 'vati' is accepted.

Sthulena tulyo yati iti bahiranga kriya-srutih: *'he goes like a stout man'* — the *hearing of the action* is *external*: bahiranga (3.7.182's word) — the going is outside the comparison, which is between a man and a stout man. Animittam vateh: *no occasion* for *'vati'* — so 'sthulavat' will not do here. Tulyam yati ity atra isyate vati: in *'he goes similarly'*, *'vati'* *is* accepted. So the affix comes only when the manner of going is what is compared, not when the persons are. The contrast with 3.14.504 is precise: there persons were compared and vati stood, because the actions were latent in them; here the action is stated and external, and vati falls.

3.14.512 unformalized

द्वयं विशेष्यते तेन यदेकत्र विशेषणम् ।
तुल्यशब्दो हि तं धर्ममुभयस्थमपेक्षते ॥ ५१२ ॥

dvayaṃ viśeṣyate tena yadekatra viśeṣaṇam |
tulyaśabdo hi taṃ dharmamubhayasthamapekṣate || 512 ||

By that, two are qualified — that which is a qualifier in one place; for the word 'tulya' requires that property as residing in both.

Dvayam visesyate tena yad ekatra visesanam: *two* are *qualified* by that which is a *qualifier* *in one place* — the property is stated once and qualifies twice (3.14.386's one is heard, the other is cognized; 3.14.424's sakrc-chrutau). Tulya-sabdo hi tam dharmam ubhaya-stham apeksate: for the word *'tulya'* *requires* that property as *residing in both* — ubhaya-stha: in the two. So 'tulya' is a word that demands a two-seated property and is satisfied by one saying. That is what makes it the comparison-word: it does the doubling itself, and the speaker need not.

3.14.513 unformalized

एकः समानो धर्मश्चेदुपमानोपमेययोः ।
तुलया संमितं तुल्यमिति तत्रोपपद्यते ॥ ५१३ ॥

ekaḥ samāno dharmaścedupamānopameyayoḥ |
tulayā saṃmitaṃ tulyamiti tatropapadyate || 513 ||

If there is one common property of comparans and comparandum — 'tulya' means 'measured out by a balance' — that is proper there.

Ekah samano dharmas ced upamana-upameyayoh: *if* there is *one common property* of *comparans and comparandum*. Tulaya sammitam tulyam iti tatra upapadyate: *'tulya'* means *'measured out by a tula'* — a *balance*, a pair of scales — and *that* is *proper* there. So the word's own etymology is the argument: 'tulya' is what has been weighed against something, and a balance needs two pans and one weight. The word is built out of the very structure 3.14.512 described. This is Bhartrhari doing what the Nirukta tradition did — reading the doctrine out of the word — and here it works perfectly.

3.14.514 unformalized

सूत्रे श्रुतश्च द्विष्ठोऽसावभेदेन प्रतीयते ।
न च सामान्यशब्दत्वादश्रुता गम्यते क्रिया ॥ ५१४ ॥

sūtre śrutaśca dviṣṭho'sāvabhedena pratīyate |
na ca sāmānyaśabdatvādaśrutā gamyate kriyā || 514 ||

And, being heard in the sutra, that two-seated property is cognized by non-difference; and from its being a general word, an unheard action is not understood.

Sutre srutas ca dvistho 'sav abhedena pratiyate: *heard in the sutra*, that *two-seated* property is cognized *by non-difference* — the rule states it and one takes it as one (3.14.362's asrita-ekatva). Na ca samanya-sabdatvad asruta gamyate kriya: and from its being a *general word*, an *unheard action* is *not understood*. So a general word does not smuggle in an unstated action: 'tulya' says likeness and does not tell you the respect. Which is why 3.14.506's over-generation was real and why 3.14.507's convention had to be called in. The rule is honest about what it does not supply.

3.14.515 unformalized

अश्रुताश्च प्रतीयन्ते निदेशस्थायितादयः ।
ये धर्मा नियतास्तेषां पुत्रादिशु न विद्यते ॥ ५१५ ॥

aśrutāśca pratīyante nideśasthāyitādayaḥ |
ye dharmā niyatāsteṣāṃ putrādiśu na vidyate || 515 ||

And unheard, the properties 'standing in command' and the rest are cognized; those properties which are fixed — of them, none exists in 'putra' and the rest.

Asrutas ca pratiyante nidesa-sthayitadayah: *unheard*, the properties *'standing in command'* and the rest *are* cognized — some properties do arrive unsaid (3.14.507's rudha-sambandha: what everyone associates). Ye dharma niyatas tesam putradisu na vidyate: those properties which are *fixed* — of *them*, *none* exists in *'putra'* and the rest. Niyata: settled, determinate. So: some unstated properties come along, and none of them is *fixed* — 'putravat' brings a cloud of associations and no definite one. This qualifies 3.14.508's padarthavat: the conventional actions are as good as meanings, and they are not determinate meanings.

3.14.516 unformalized

अनाश्रितक्रियस्तस्मान्न तुल्योऽस्ति क्रियावता ।
क्रियायाः श्रवणे सापि क्रियावत्ता प्रतीयते ॥ ५१६ ॥

anāśritakriyastasmānna tulyo'sti kriyāvatā |
kriyāyāḥ śravaṇe sāpi kriyāvattā pratīyate || 516 ||

Therefore one who has not resorted to an action is not similar to an action-possessor; and in the hearing of the action, that action-possessing too is cognized.

Anasrita-kriyas tasman na tulyo 'sti kriyavata: therefore one who has *not resorted to an action* is *not similar* to an *action-possessor* — likeness requires both parties to have the action; an actionless thing resembles no agent. Kriyayah sravane sapi kriyavatta pratiyate: and in the *hearing of the action*, that *action-possessing* too is *cognized* — say the action and the possessing comes with it. So the two are inseparable in the hearing: name an action and you have named someone's having it (3.14.474's nothing bare is reachable by a word). Which is why the vati-rule can afford to speak of actions and mean possessors.

3.14.517 unformalized

द्वयोः प्रतिविधानाच्च ज्यायस्त्वमभिधीयते ।
नित्यासत्त्वाभिधायित्वात् प्रत्ययार्थविशेषणे ॥ ५१७ ॥

dvayoḥ pratividhānācca jyāyastvamabhidhīyate |
nityāsattvābhidhāyitvāt pratyayārthaviśeṣaṇe || 517 ||

And from the provision for two, superiority is stated; from its always expressing non-substance, in the qualification of the affix-meaning.

Dvayoh pratividhanac ca jyayastvam abhidhiyate: from the *provision for two*, *superiority* is *stated* — jyayastva: the greater merit; where a reading handles both cases it is preferred. Nitya-asattva-abhidhayitvat pratyayartha-visesane: from *always expressing non-substance*, in the *qualification of the affix-meaning*. So the reading that makes the affix-meaning qualified is superior because it covers both cases and because the affix always expresses asattva (3.14.435's tinam asattva-vacitvat). The samuddesa is adjudicating between two readings of the rule on the grounds of coverage — an unusually decisive moment.

3.14.518 unformalized

असत्त्वभूतो व्यापारः केवलं प्रत्यये यतः ।
विद्यते लक्षणार्थत्वं नास्ति तेन क्रियाश्रुतेः ॥ ५१८ ॥

asattvabhūto vyāpāraḥ kevalaṃ pratyaye yataḥ |
vidyate lakṣaṇārthatvaṃ nāsti tena kriyāśruteḥ || 518 ||

Since in the affix there is only the operation, which is non-substantial — therefore the action-hearing has no indicative purpose.

Asattva-bhuto vyaparah kevalam pratyaye yatah: since in the *affix* there is *only* the *operation*, which is *non-substantial* — the affix carries an activity and no thing (3.14.193's asattva-bhuta, of the particle). Vidyate laksanarthatvam nasti tena kriya-sruteh: *therefore* the *action-hearing* has *no indicative purpose* — laksanarthatva: the pointing-function of 3.14.451 and 3.14.496. So on *this* reading the rule's 'kriya' does not point at a possessor: the affix already has the operation and needs no indication. Which is the opposite of what 3.14.496 concluded — and both readings are on the table, because the samuddesa is showing what each costs.

3.14.519 unformalized

क्रियावतस्तु ग्रहणात् प्रकृत्यर्थविशेषणे ।
क्रियामात्रेण तुल्यत्वे सिद्धाऽसत्त्वाभिधायिता ॥ ५१९ ॥

kriyāvatastu grahaṇāt prakṛtyarthaviśeṣaṇe |
kriyāmātreṇa tulyatve siddhā'sattvābhidhāyitā || 519 ||

But from the taking-up of the action-possessor, in the qualification of the base-meaning — where similarity is by mere action, the expressing of non-substance is established.

Kriyavatas tu grahanat prakrty-artha-visesane: from the *taking-up of the action-possessor*, in the *qualification of the base-meaning* — the other reading: the possessor is what the rule takes up, and it qualifies the *base*. Kriya-matrena tulyatve siddha asattva-abhidhayita: where similarity is by *mere action*, the *expressing of non-substance* is *established*. So this reading too delivers asattva-abhidhayita — the thing 3.14.517 wanted. Both readings reach it, by different routes. The adjudication of 3.14.517 was therefore premature, and the samuddesa is undoing it — which is its habit (3.14.494's tulyau paksav ubhau).

3.14.520 unformalized

यदा क्रियानिमित्तं तु सादृश्यं स्यात् क्रियावतोः ।
क्रियावतोऽभिधेयत्वात् तदाऽद्रव्याभिधायिता ॥ ५२० ॥

yadā kriyānimittaṃ tu sādṛśyaṃ syāt kriyāvatoḥ |
kriyāvato'bhidheyatvāt tadā'dravyābhidhāyitā || 520 ||

But when the similarity of two action-possessors has the action for its cause — then, since an action-possessor is what is expressed, there is non-substance-expressing.

Yada kriya-nimittam tu sadrsyam syat kriyavatoh: when the *similarity* of *two action-possessors* has the *action* for its *cause* — 3.14.503's case: two persons, alike by reason of their actions. Kriyavato 'bhidheyatvat tada adravya-abhidhayita: *then*, since an *action-possessor* is what is *expressed*, there is *non-substance-expressing* — adravya-abhidhayita: it does not express a thing. So even where a possessor is expressed, what is expressed is not a substance: the possessor is reached under the action, and the action is asattva. This closes the sequence 3.14.517-520: every reading of the rule yields asattva, and the affix cannot express a thing however one construes it.

3.14.521 unformalized

अव्ययेषु वतेः पाठः कार्यस्तत्र स्वरादिवत् ।
ब्राह्मणेन समोऽध्येतेत्यत्र च प्रत्ययो भवेत् ॥ ५२१ ॥

avyayeṣu vateḥ pāṭhaḥ kāryastatra svarādivat |
brāhmaṇena samo'dhyetetyatra ca pratyayo bhavet || 521 ||

There, the listing of 'vati' among the indeclinables should be made, like 'svar' and the rest; and in 'the student is equal to a brahmin' the affix would occur.

Avyayesu vateh pathah karyas tatra svaradivat: *there* — on the reading just reached — the *listing* of *'vati'* among the *indeclinables* should be *made*, like *'svar'* and the rest: the svaradi-gana. Brahmanena samo 'dhyeta ity atra ca pratyayo bhavet: and in *'the student is equal to a brahmin'* the *affix* would *occur*. So if the affix always expresses asattva (3.14.520), it belongs in the indeclinable list — reversing 3.14.494's refusal to list it. The samuddesa has argued itself to the opposite conclusion in twenty-seven verses, and it does not conceal the reversal: 3.14.494 declined because the alternatives were equal, and they have stopped being equal.

3.14.522 unformalized

सामानाधिकरण्यं च वत्यर्थेनापदिश्यते ।
तुल्यमित्यन्यथा कल्प्यो वाक्यशेषोऽश्रुतो भवेत् ॥ ५२२ ॥

sāmānādhikaraṇyaṃ ca vatyarthenāpadiśyate |
tulyamityanyathā kalpyo vākyaśeṣo'śruto bhavet || 522 ||

And co-reference is designated with the vati-meaning; otherwise, for 'tulyam', an unheard sentence-remainder would have to be imagined.

Samanadhikaranyam ca vaty-arthena apadisyate: *co-reference* is designated with the *vati-meaning* — the affix's sense co-refers with what it attaches to. Tulyam ity anyatha kalpyo vakya-sesa asruto bhavet: *otherwise*, for *'tulyam'*, an *unheard* *sentence-remainder* would have to be *imagined* — vakya-sesa: what must be supplied to make the sentence construe (3.14.463's sopaskara). So the co-reference reading saves one from inventing words: on any other reading, 'tulyam' needs a supplement nobody says. The argument is economy — prefer the reading that requires no invention — and it is the same argument as 3.14.469's, that unemployed elements should be indicated rather than supplied.

3.14.523 unformalized

क्रियावतोश्च सादृश्ये प्रत्ययार्थविशेषणे ।
अध्येत्रा सदृशोऽध्येतेत्यत्र नास्ति वतेर्विधिः ॥ ५२३ ॥

kriyāvatośca sādṛśye pratyayārthaviśeṣaṇe |
adhyetrā sadṛśo'dhyetetyatra nāsti vatervidhiḥ || 523 ||

And in the similarity of two action-possessors, in the qualification of the affix-meaning — in 'the student is like a student', there is no rule for 'vati'.

Kriyavatos ca sadrsye pratyayartha-visesane: in the *similarity of two action-possessors*, in the *qualification of the affix-meaning* — the reading of 3.14.517. Adhyetra sadrso 'dhyeta ity atra nasti vater vidhih: in *'the student is like a student'*, there is *no rule* for *'vati'*. So the affix will not attach where both terms are the same kind of agent — because there is nothing to compare (3.14.394's the cause in the cognition 'cow' exists in another cow; 3.14.442's an action of the same class is not compared). The rule's silence here is not a gap: it registers that identical terms make no simile.

3.14.524 unformalized

तुल्यार्थैरिति या तस्यास्तृतीयाया न भिद्यते ।
अर्थो भेदेऽपि सर्वाभिरितराभिर्विभक्तिभिः ॥ ५२४ ॥

tulyārthairiti yā tasyāstṛtīyāyā na bhidyate |
artho bhede'pi sarvābhiritarābhirvibhaktibhiḥ || 524 ||

That third case which comes by 'with words meaning tulya' — its meaning is not differentiated, even where there is difference, by any of the other case-endings.

Tulyarthair iti ya tasyas trtiyayah: that *third case* which comes by *'tulyarthaih'* — Panini's rule prescribing the instrumental with words meaning 'similar'. Na bhidyate artho bhede 'pi sarvabhir itarabhir vibhaktibhih: its *meaning* is *not differentiated*, even where there is *difference*, by *all the other* case-endings. So: whatever case the compared thing would have taken in the underlying construction, the 'tulya'-construction puts it in the instrumental, and the meaning does not vary. The claim is strong and 3.14.525-527 proves it by running through the cases one by one.

3.14.525-527 unformalized

भुज्यते ब्राह्मणेनेव तुल्यं भुक्तं द्विजातिना ।
पश्यति ब्राह्मणमिव तुल्यं विप्रेण पश्यति ॥ ५२५ ॥
ब्राह्मणेनेव विज्ञातं तुल्यं ज्ञातं द्विजातिना ।
दीयतां ब्राह्मणायेव तुल्यं विप्रेण दीयताम् ॥ ५२६ ॥
ब्राह्मणादिव वैश्यात् त्वमधीष्वाध्ययनं बहु ।
इत्येवमादिभिर्भेदस्तृतीयाया न कश्चन ॥ ५२७ ॥

bhujyate brāhmaṇeneva tulyaṃ bhuktaṃ dvijātinā |
paśyati brāhmaṇamiva tulyaṃ vipreṇa paśyati || 525 ||
brāhmaṇeneva vijñātaṃ tulyaṃ jñātaṃ dvijātinā |
dīyatāṃ brāhmaṇāyeva tulyaṃ vipreṇa dīyatām || 526 ||
brāhmaṇādiva vaiśyāt tvamadhīṣvādhyayanaṃ bahu |
ityevamādibhirbhedastṛtīyāyā na kaścana || 527 ||

(525) 'It is eaten as by a brahmin' — 'it is eaten similarly-with a twice-born'. 'He sees, as a brahmin' — 'he sees similarly-with a brahmin'. (526) 'It is known, as by a brahmin' — 'it is known similarly-with a twice-born'. 'Let it be given, as to a brahmin' — 'let it be given similarly-with a brahmin'. (527) 'Study much, as from a brahmin, from a vaisya' — by these and the like, there is no difference at all of the third case.

The proof of 3.14.524, and it is a demonstration rather than an argument. Five constructions, in five different cases — instrumental (bhujyate brahmanena iva), accusative (pasyati brahmanam iva), instrumental again (brahmanena iva vijnatam), dative (diyatam brahmanaya iva), ablative (brahmanad iva vaisyat) — and each is rendered by *tulyam* with an *instrumental*, and none of the meanings shifts. Ity evam adibhir bhedas trtiyayah na kascana: by these and the like, there is *no difference at all* of the *third case*. So the instrumental prescribed with 'tulya' is case-neutral: it absorbs agent, object, recipient and source alike, and expresses none of them. It is not a karaka-marker at all — it is the mark of the compared term, and the karaka it would otherwise have borne is simply gone. The examples are laid out and one sees it; Bhartrhari adds nothing, which is the right decision.

3.14.528 unformalized

तुल्यं मधुरयाधीये मात्रा तुल्यं स्मरामि ताम् ।
मधुरायाश्च मातुश्च कथं सादृश्यकल्पना ॥ ५२८ ॥

tulyaṃ madhurayādhīye mātrā tulyaṃ smarāmi tām |
madhurāyāśca mātuśca kathaṃ sādṛśyakalpanā || 528 ||

'I study similarly-with Madhura', 'I remember her similarly-with mother' — how is a construction of similarity of Madhura and of mother to be had?

The difficulty in 3.14.524's doctrine. Tulyam madhuraya adhiye: *'I study similarly-with Madhura'* — where the underlying sense is 'I study as one does in Madhura': a locative absorbed into the instrumental. Matra tulyam smarami tam: *'I remember her similarly-with mother'* — 'as I remember mother': an object absorbed. Madhurayas ca matus ca katham sadrsya-kalpana: *how* is a *construction of similarity* of *Madhura* and of *mother*? So the instrumental of 3.14.525-527 has swallowed a place and a person, and now literally says that a city and a mother are being compared to something. Katham: how can that be? The absorption has produced nonsense, and 3.14.529 must explain it away.

3.14.529 unformalized

मधुराविषयः पाठः स्मरणं मातृकर्मकम् ।
मधुरामातृशब्दाभ्यामभेदेनाभिधीयते ॥ ५२९ ॥

madhurāviṣayaḥ pāṭhaḥ smaraṇaṃ mātṛkarmakam |
madhurāmātṛśabdābhyāmabhedenābhidhīyate || 529 ||

The recitation whose domain is Madhura, and the remembering whose object is mother — are expressed by the words 'Madhura' and 'mother' by way of non-difference.

Madhura-visayah pathah: the *recitation* whose *domain* is *Madhura* — the studying done there. Smaranam matr-karmakam: the *remembering* whose *object* is *mother*. Madhura-matr-sabdabhyam abhedena abhidhiyate: are *expressed* by the *words* *'Madhura'* and *'mother'* *by way of non-difference*. So the words do not name the city and the woman: they name the *studying-at-Madhura* and the *remembering-of-mother*, by non-difference — the action and its seat taken as one (3.14.483's upacaritaikatva). So the comparison is between two studyings and two rememberings after all, and 3.14.528's absurdity was an artifact of reading the words as naming their nominal referents.

3.14.530 unformalized

उष्ट्रावयवतुल्येषु मुखेषूष्ट्रश्रुतिर्यथा ।
वर्तते गृहतुल्ये च प्रासादे मधुराश्रुतिः ॥ ४३० ॥

uṣṭrāvayavatulyeṣu mukheṣūṣṭraśrutiryathā |
vartate gṛhatulye ca prāsāde madhurāśrutiḥ || 430 ||

As the word 'camel' operates in faces resembling a camel's part; and the word 'Madhura' in a palace resembling a house.

This verse repeats 3.14.430 verbatim, and our source numbers it '430' — the third such doubling in this samuddesa, after 3.14.421 (=121) and 3.14.479 (=179). Here the repetition at least has a context: 3.14.529 has just been about the word 'Madhura' operating in what is not the city, and the camel-and-palace verse illustrates exactly that. Whether the redactor placed it here deliberately or the transmission duplicated it, we print it where the source does.
contested: This and the two verses following it (3.14.531 = 3.14.431, 3.14.532 = 3.14.432) form a block of three consecutive verses repeated from an earlier position, retaining their original numbers 430-432 in our source. Taken with the isolated doublings at 3.14.421 and 3.14.479, the pattern suggests a mechanical fault in the transmission — a repeated exemplar-leaf — rather than authorial repetition. The critical editions should be consulted.

3.14.531 unformalized

यथाध्ययनयोः साम्यमध्येत्रोरपदिश्यते ।
तथा क्रियागतैर्धर्मैरुच्यन्ते साधनाश्रयाः ॥ ४३१ ॥

yathādhyayanayoḥ sāmyamadhyetrorapadiśyate |
tathā kriyāgatairdharmairucyante sādhanāśrayāḥ || 431 ||

As the similarity of two studyings is designated of the two students — so the substrata of the means are spoken of by properties belonging to the action.

A verbatim repetition of 3.14.431, numbered '431' in our source (see the note at 3.14.530). Its argument — that an action's properties are predicated of the agents who bear them — is the same one 3.14.529 has just used to dissolve the Madhura-and-mother difficulty, so the block's placement here is at least intelligible. But it adds nothing that 3.14.431 did not already establish, and the retained numbering marks it as the transmission's doing rather than the redactor's.

3.14.532 unformalized

इवार्थे यच्च वचनं पूर्वसूत्रे च यो विधिः ।
क्रियाशब्दश्रुतौ भेदो न कश्चित् विद्यते तयोः ॥ ४३२ ॥

ivārthe yacca vacanaṃ pūrvasūtre ca yo vidhiḥ |
kriyāśabdaśrutau bhedo na kaścit vidyate tayoḥ || 432 ||

The statement in the sense of 'iva', and the rule in the previous sutra — in the hearing of an action-word there is no difference at all between those two.

A verbatim repetition of 3.14.432, numbered '432' in our source, and the last of the three-verse block (see 3.14.530). Its subject — the coincidence of the iva-rule and the previous sutra where an action-word is heard — is in fact what 3.14.535-538 are about to take up in detail, so a reader coming to this stretch fresh will find the verse doing useful work here. That does not make it any less a repetition.

3.14.533 unformalized

यद्यप्युपाधिरन्यत्र नियतो न प्रयुज्यते ।
रूपाभेदात् त्वनिर्ज्ञाता क्रियात्र श्रूयते पुनः ॥ ५३३ ॥

yadyapyupādhiranyatra niyato na prayujyate |
rūpābhedāt tvanirjñātā kriyātra śrūyate punaḥ || 533 ||

Although the fixed adjunct is not employed elsewhere — from non-difference of form, an unknown action is here heard again.

Yady apy upadhir anyatra niyato na prayujyate: although the *fixed adjunct* is *not employed* elsewhere — the conditioning element that does not appear (3.14.464's na kadacit prayujyate; 3.14.469's aprayuktaih). Rupabhedat tv anirjnata kriya atra sruyate punah: from *non-difference of form*, an *unknown* action is *here* *heard again* — anirjnata: not known, unidentified. So because the forms are alike, an action one cannot identify gets heard a second time. The verse describes the mechanism by which an unemployed adjunct nevertheless makes itself audible: form-identity smuggles it back in (3.14.483's tulya-sruti producing a figurative oneness).

3.14.534 unformalized

यथा व्युत्परयऋ पुच्छौ क्यङ्न्ते सुदरादयः ।
सत्यपि प्रत्ययार्थत्वे भेदाभावादुदाहृताः ॥ ५३४ ॥

yathā vyutparayaṛ pucchau kyaṅnte sudarādayaḥ |
satyapi pratyayārthatve bhedābhāvādudāhṛtāḥ || 534 ||

As 'sudara' and the rest, ending in 'kyan' — although there is an affix-meaning, they are cited from the absence of difference.

Kyan-ante sudaradayah: *'sudara'* and the rest, ending in *'kyan'* — the denominative affix. Saty api pratyayarthatve bhedabhavad udahrtah: *although* there is an *affix-meaning*, they are *cited* from the *absence of difference*. So these forms have an affix that means something, and they are nevertheless cited as cases where no difference obtains — the affix's contribution has vanished into the whole (3.14.44's ekarthibhava). This is the precedent 3.14.533 needed: an affix-meaning present and undetectable is a familiar thing, and the unknown action of 3.14.533 is another instance.
contested: Pada a is transmitted as 'yatha vyutparayr pucchau', which we cannot construe and take to be corrupt beyond safe conjecture. We translate from padas b-d, whose sense is clear. The lost example presumably belonged with the kyan-forms of pada b.

3.14.535 unformalized

एवञ्च सति पूर्वेण सिद्धोऽत्रापि वतेर्विधिः ।
नियमे वाभिधाने वा भिद्यते न क्रियाश्रुतिः ॥ ५३५ ॥

evañca sati pūrveṇa siddho'trāpi vatervidhiḥ |
niyame vābhidhāne vā bhidyate na kriyāśrutiḥ || 535 ||

And this being so, the rule for 'vati' here too is established by the previous sutra; whether in restriction or in expression, the action-hearing is not differentiated.

Evam ca sati purvena siddho 'trapi vater vidhih: this being so, the rule for *'vati'* *here too* is *established by the previous* sutra — the later rule is unnecessary; the earlier one already covers the case. Niyame va abhidhane va bhidyate na kriya-srutih: whether in *restriction* or in *expression*, the *action-hearing* is *not differentiated* — niyama: a rule that restricts what would otherwise be free; abhidhana: one that states. So the 'kriya' in the rule does the same thing on either construal, and the second rule is redundant. This is 3.14.432's finding (= 3.14.532's), now argued rather than asserted.

3.14.536 unformalized

इवे द्रव्यादिविषयः प्रत्ययः पुनरुच्यते ।
क्रियाणामेव सादृश्ये पूर्वसूत्रे विधीयते ॥ ५३६ ॥

ive dravyādiviṣayaḥ pratyayaḥ punarucyate |
kriyāṇāmeva sādṛśye pūrvasūtre vidhīyate || 536 ||

In the rule for 'iva', the affix whose domain is substance and the rest is stated again; in the previous sutra it is prescribed only for the similarity of actions.

Ive dravyadi-visayah pratyayah punar ucyate: in the rule for *'iva'*, the affix whose *domain* is *substance* and the rest is *stated again* — punah: over again. Kriyanam eva sadrsye purva-sutre vidhiyate: in the *previous* sutra it is prescribed *only* for the *similarity of actions*. So the two rules divide the field: the earlier covers actions, the later substances and qualities — and if that is right, 3.14.535's redundancy-charge fails, because the later rule has territory of its own. The samuddesa has stated the objection to its own conclusion immediately and is about to test it.

3.14.537 unformalized

मधुरायामिव गृहा ब्राह्मणस्येव पाण्डुराः ।
इत्यत्र द्रव्यगुणयोः पूर्वेण न वतिर्भवेत् ॥ ५३७ ॥

madhurāyāmiva gṛhā brāhmaṇasyeva pāṇḍurāḥ |
ityatra dravyaguṇayoḥ pūrveṇa na vatirbhavet || 537 ||

'Houses as in Madhura', 'white as a brahmin's' — here, for substance and for quality, 'vati' would not come by the previous rule.

The evidence for 3.14.536. Madhurayam iva grha: *'houses as in Madhura'* — a substance-comparison. Brahmanasya iva panduarah: *'white as a brahmin's'* — a quality-comparison. Ity atra dravya-gunayoh purvena vatir na bhavet: here, for *substance* and *quality*, *'vati'* would *not* come by the *previous* rule — since that rule speaks only of actions. So the later rule is needed, and 3.14.535's economy was false. The two examples are exactly chosen: one substance, one quality, neither with any action in view, both taking 'vati' in ordinary usage.

3.14.538 unformalized

आरम्भस्याक्रियार्थत्वे नार्थो योगेन विद्यते ।
ऋते क्रियाया ग्रहणात् पूर्वयोगेन सिध्यति ॥ ५३८ ॥

ārambhasyākriyārthatve nārtho yogena vidyate |
ṛte kriyāyā grahaṇāt pūrvayogena sidhyati || 538 ||

If the undertaking is not for the sake of the action, there is no purpose in the rule; apart from the taking-up of 'kriya', it is established by the previous rule.

Arambhasya akriyarthatve narthah yogena vidyate: if the *undertaking* is *not* for the sake of the *action*, there is *no purpose* in the *rule* — arambha: the rule's being framed at all. Rte kriyaya grahanat purva-yogena sidhyati: *apart from* the *taking-up* of *'kriya'*, it is *established* by the *previous* rule. So the later rule earns its place only by its word 'kriya': strike that word out and everything it does is done already. This settles the question 3.14.535-537 opened — the second rule is not redundant, and what saves it is one word. Bhartrhari has been reading Panini this way throughout (3.14.225's artha-grahanam krtam; 3.14.446's atin-grahanam).

3.14.539 unformalized

मधुरावयवे वृत्तिर्व्याख्याता मधुराश्रुतेः ।
ब्राह्मणावयवान् दन्तान् वक्ष्यति ब्राह्मणश्रुतिः ॥ ५३९ ॥

madhurāvayave vṛttirvyākhyātā madhurāśruteḥ |
brāhmaṇāvayavān dantān vakṣyati brāhmaṇaśrutiḥ || 539 ||

The operation of the word 'Madhura' in a part of Madhura has been explained; the word 'brahmana' will state the teeth, which are parts of a brahmin.

Madhuravayave vrttir vyakhyata madhura-sruteh: the *operation* of the word *'Madhura'* in a *part* of Madhura has been *explained* — 3.14.529-530: the word reaches what is in the city, and what resembles a part of it. Brahmanavayavan dantan vaksyati brahmana-srutih: the word *'brahmana'* *will state* the *teeth*, which are *parts* of a *brahmin* — 'brahmanavat dantah', teeth like a brahmin's: the word reaches the teeth. So the same part-reaching happens with persons: 'brahmana' says the teeth. This is 3.14.487's law (a word distributes over its parts) applied where it is least expected, and it is what licenses 3.14.537's brahmanasya iva panduarah.

3.14.540 unformalized

न काचिदिवयोगे तु बाह्यात् सम्बन्धिनो विना ।
षष्ठी विधीयते तत्र पूर्वेण प्रत्ययो भवेत् ॥ ५४० ॥

na kācidivayoge tu bāhyāt sambandhino vinā |
ṣaṣṭhī vidhīyate tatra pūrveṇa pratyayo bhavet || 540 ||

But in connection with 'iva' no genitive is prescribed without an external relatum; there the affix would come by the previous rule.

Na kacid iva-yoge tu bahyat sambandhino vina sasthi vidhiyate: in connection with *'iva'* *no* genitive is *prescribed* *without* an *external relatum* — bahya: outside the construction (3.14.207's word). So 'brahmanasya iva' needs someone outside — whose teeth? — before the genitive can stand. Tatra purvena pratyayo bhavet: *there* the *affix* would come by the *previous* rule. So where the external relatum is absent, one falls back to the earlier rule and gets 'vati' anyway. The two rules interlock rather than compete, and the division of 3.14.536 is confirmed by watching where each one is forced.

3.14.541 unformalized

आधिक्यं तुल्यशब्देन सम्बन्ध उपजायते ।
षष्ठीतृतीये तत्र स्तस्तुल्यशब्दो हि वाचकः ॥ ५४१ ॥

ādhikyaṃ tulyaśabdena sambandha upajāyate |
ṣaṣṭhītṛtīye tatra stastulyaśabdo hi vācakaḥ || 541 ||

With the word 'tulya', an excess of connection arises; there the genitive and the instrumental both exist, for 'tulya' is an expresser.

Adhikyam tulya-sabdena sambandha upajayate: with the word *'tulya'*, an *excess* of *connection* *arises* — adhikya: something more, a surplus relation. Sasthi-trtiye tatra stah: *there* the *genitive* and the *instrumental* both *exist* — both cases are available with 'tulya', where 3.14.524 said only the instrumental was. Tulya-sabdo hi vacakah: for *'tulya'* is an *expresser* — a vacaka, not a dyotaka. So 'tulya' *says* the likeness, and because it says it, it can govern a case of its own, and a surplus connection appears. The vacaka/dyotaka distinction (3.14.194, 3.14.203) is doing the work again, and it is the samuddesa's most reliable tool.

3.14.542 unformalized

इवशब्दप्रयोगे तु बाह्यात् सम्बन्धिनो विना ।
नाधिक्यमुपमानेऽस्ति द्योतकः सम्प्रयुज्यते ॥ ५४२ ॥

ivaśabdaprayoge tu bāhyāt sambandhino vinā |
nādhikyamupamāne'sti dyotakaḥ samprayujyate || 542 ||

But in the employment of the word 'iva', without an external relatum, there is no excess in the comparans; an indicator is employed.

The contrast that 3.14.541 set up. Iva-sabda-prayoge tu bahyat sambandhino vina: in the employment of the word *'iva'*, *without* an *external relatum*. Na adhikyam upamane 'sti: there is *no excess* in the *comparans*. Dyotakah samprayujyate: an *indicator* is *employed* — 'iva' is a dyotaka. So 'iva' brings no surplus because it says nothing: it only reveals what the words already had (3.14.191's the indicator restricts the latent powers). And that is precisely why 'iva' governs no case and 'tulya' governs two. The whole difference between the two comparison-words is that one speaks and one points.

3.14.543 unformalized

इवे यो व्यतिरेकोऽत्र स प्रालादादिहेतुकः ।
तुल्ये तद्विषयापेक्षमाधिक्यमुपजायते ॥ ५४३ ॥

ive yo vyatireko'tra sa prālādādihetukaḥ |
tulye tadviṣayāpekṣamādhikyamupajāyate || 543 ||

The difference which is here in 'iva' is caused by 'palace' and the rest; in 'tulya', an excess dependent on that domain arises.

Ive yo vyatireko 'tra sa prasadadi-hetukah: the *difference* which is here in *'iva'* is *caused by* *'palace'* and the rest — 3.14.530's palace resembling a house: whatever differentiation 'iva' seems to carry comes from the words around it, not from itself (3.14.542: it says nothing). Tulye tad-visaya-apeksam adhikyam upajayate: in *'tulya'*, an *excess* *dependent on that domain* *arises*. So 'tulya' generates its own surplus, in dependence on the same domain that 'iva' merely borrows from. The pair is now fully drawn: 'iva' takes all its content from its neighbours; 'tulya' adds. And the long examination of the comparison-words ends here, on the distinction it began with.

3.14.544 unformalized

गवयेन समोऽनड्वानिति वृत्तिस्तदा भवेत् ।
न त्वस्ति गौरिवेत्यत्र व्यतिरेक इवाश्रयः ॥ ५४४ ॥

gavayena samo'naḍvāniti vṛttistadā bhavet |
na tvasti gaurivetyatra vyatireka ivāśrayaḥ || 544 ||

'The ox is equal to a gavaya' — then there would be a derivation; but in 'like a cow' there is no difference to serve, as it were, as a substratum.

Gavayena samo 'nadvan iti vrttis tada bhavet: *'the ox is equal to a gavaya'* — then there would be a *derivation*: 'gavayavat', since the two are different species and there is something to compare. Na tv asti gaur iva ity atra vyatireka iva asrayah: but in *'like a cow'* — said of an ox — there is *no difference* to serve, *as it were*, as a *substratum*. So the gavaya-comparison stands and the cow-comparison does not, because ox and cow are too close: the difference the comparison needs as its footing is missing (3.14.394's the cause in the cognition 'cow' exists in another cow). Vyatireka iva asrayah: the difference is what the simile *stands on*, and here there is no floor.

3.14.545 unformalized

उपमेयेन सम्बन्धात् प्राक् तु प्रासादादिहेतुके ।
व्यतिरेके वतेर्भावो न तु तुल्याप्थहेतुके ॥ ५४५ ॥

upameyena sambandhāt prāk tu prāsādādihetuke |
vyatireke vaterbhāvo na tu tulyāpthahetuke || 545 ||

But prior to connection with the comparandum, in a difference caused by 'palace' and the rest, 'vati' occurs — not in one caused by the tulya-meaning.

Upameyena sambandhat prak tu prasadadi-hetuke vyatireke vater bhavah: *prior to* connection with the *comparandum*, in a *difference caused by 'palace'* and the rest (3.14.543's prasadadi-hetukah), there is the *occurrence* of *'vati'*. Na tu tulyartha-hetuke: *not* in one caused by the *tulya-meaning*. So the affix attaches to a difference that the surrounding words made, before the comparandum is even connected — and not to a difference that 'tulya' itself generates. The distinction of 3.14.541-543 is thus doing rule-technical work: 'iva'-differences take vati, 'tulya'-differences do not, because the latter are the word's own surplus and not the situation's.

3.14.546 unformalized

इवशब्देन सम्बन्धे न तृतीया विधीयते ।
प्रकृतां तामतस्त्यक्त्वा विभक्त्यन्तरमाश्रितम् ॥ ५४६ ॥

ivaśabdena sambandhe na tṛtīyā vidhīyate |
prakṛtāṃ tāmatastyaktvā vibhaktyantaramāśritam || 546 ||

In connection with the word 'iva' the third case is not prescribed; therefore, abandoning the one under discussion, another case-ending is resorted to.

Iva-sabdena sambandhe na trtiya vidhiyate: in connection with the word *'iva'*, the *third case* is *not prescribed* — 'iva' governs nothing (3.14.542: it is a dyotaka and brings no surplus, so it can demand no case). Prakrtam tam atas tyaktva vibhakty-antaram asritam: *therefore*, *abandoning* the one *under discussion* — the instrumental of 3.14.524's rule — *another* case-ending is *resorted to*. So with 'iva' one drops the prescribed instrumental and takes whatever case the underlying construction wanted: 'brahmanam iva pasyati', accusative (3.14.525). The whole absorption of 3.14.525-527 was 'tulya''s doing, and 'iva' undoes it.

3.14.547 unformalized

सप्तम्यपि न तत्रास्ति ज्ञापनार्था तु सा कृता ।
इष्टा सा शेषविषये नियतासु विभक्तिषु ॥ ५४७ ॥

saptamyapi na tatrāsti jñāpanārthā tu sā kṛtā |
iṣṭā sā śeṣaviṣaye niyatāsu vibhaktiṣu || 547 ||

The seventh is also not there; but it has been made for the sake of indication; it is desired in the domain of the remainder, among the fixed case-endings.

Saptamy api na tatra asti: the *seventh* — the locative — is *also not* there. Jnapanartha tu sa krta: but it has been *made* for the sake of *indication* — jnapana: the grammarians' term for a rule that exists to signal something rather than to operate (3.14.130's the elision taught to demarcate a domain). Ista sa sesa-visaye niyatasu vibhaktisu: it is *desired* in the *domain of the remainder*, among the *fixed* case-endings. So the locative in the rule is not prescribing a locative: it is a jnapaka, and what it indicates is that the sesa-domain is where the fixed cases operate. The rule's cases are notation about cases (3.14.78's the science made of its own devices).

3.14.548 unformalized

यदि तु व्यतिरेकेण विषयेऽस्मिन् विभक्तयः ।
प्रवर्तेरंस्तृतीयैव व्यभिचारं प्रदर्शयेत् ॥ ५४८ ॥

yadi tu vyatirekeṇa viṣaye'smin vibhaktayaḥ |
pravarteraṃstṛtīyaiva vyabhicāraṃ pradarśayet || 548 ||

But if the case-endings operated in this domain by way of difference — the third alone would show the deviation.

Yadi tu vyatirekena visaye 'smin vibhaktayah pravarteran: *if* the case-endings *operated* in this *domain* *by way of difference* — each case marking a distinct relation, as cases normally do. Trtiya eva vyabhicaram pradarsayet: the *third alone* would *show* the *deviation* — vyabhicara: straying, the failure of invariable concomitance (3.14.131's word). So the instrumental would be the odd one out: every other case would mark its own relation and the instrumental would be marking all of them (3.14.525-527). The instrumental's absorption is thus visible as an anomaly within the case-system, and the anomaly is the evidence.

3.14.549-550 unformalized

व्यभिचारे तथासिद्धे सप्तमीग्रहणाद् विना ।
सप्तम्येवोच्यते सर्वा न सन्त्यन्या विभक्तयः ॥ ५४९ ॥
अत्यन्तस्य विषये सप्तम्या ज्ञापनार्थया ।
बाधिता विनिवर्तेत षष्ठी सा गृह्यते पुनः ॥ ५५० ॥

vyabhicāre tathāsiddhe saptamīgrahaṇād vinā |
saptamyevocyate sarvā na santyanyā vibhaktayaḥ || 549 ||
atyantasya viṣaye saptamyā jñāpanārthayā |
bādhitā vinivarteta ṣaṣṭhī sā gṛhyate punaḥ || 550 ||

(549) The deviation being thus established, then, without the taking-up of the seventh, the seventh alone would be stated and none of the other case-endings would exist. (550) In the domain of the utter one, the genitive, annulled by the seventh which is there for indication, would withdraw; therefore it is taken up again.

(549): vyabhicare tatha asiddhe saptami-grahanad vina saptamy eva ucyate sarva na santy anya vibhaktayah — the *deviation* being thus *established*, *without* the *taking-up of the seventh*, the *seventh alone* would be *stated* and *none* of the *other* case-endings would *exist*. So without the locative's explicit mention, the rule would over-reach and swallow every case — the mirror of the instrumental's over-reach at 3.14.548. (550): atyantasya visaye saptamya jnapanarthaya badhita vinivarteta sasthi — in the *domain of the utter one*, the *genitive*, *annulled* by the *seventh* which is there for *indication*, would *withdraw*. Sa grhyate punah: *therefore* it is *taken up again*. So the genitive has to be re-stated to survive the locative's annulling. Three cases, each over-reaching or being over-reached, and each held in place by an explicit mention. The rule is a truce.

3.14.551 unformalized

यदा त्वाश्रयभेदेन भेद एव प्रतीयते ।
आकृतेर्द्रव्यपक्षेण तदा भेदो न विद्यते ॥ ३३१ ॥

yadā tvāśrayabhedena bheda eva pratīyate |
ākṛterdravyapakṣeṇa tadā bhedo na vidyate || 331 ||

But when difference itself is cognized through difference of substratum — then, of the akrti, on the substance-view, there is no difference.

This verse repeats 3.14.331 verbatim, and our source numbers it '331' — the fifth doubling in this samuddesa (after 3.14.421, 3.14.479, and the block 3.14.530-532). Unlike the 530-532 block, it has no evident bearing on the case-technical argument of 3.14.546-550 that surrounds it, and no bearing on the king-comparisons that follow. Its home is the akrti-analysis of 3.14.329-331. We print it where the source prints it and record the repetition.
contested: This is the fifth verse in the Vrtti-samuddesa which our source repeats verbatim from an earlier position while retaining the earlier number (121/421, 179/479, 430-432/530-532, 331/551). Five such faults in one samuddesa, all of the same kind, point to a defect in this witness rather than to the text. Readers should treat any of these five as unreliable evidence for the karika-count.

3.14.552 unformalized

तेन तुल्यमिति प्राप्तेः क्रियोपाधिः प्रसिध्यति ।
राजवद् वर्तते राजेत्यत्र भेदे विवक्षिते ॥ ५५२ ॥

tena tulyamiti prāpteḥ kriyopādhiḥ prasidhyati |
rājavad vartate rājetyatra bhede vivakṣite || 552 ||

From the applicability of 'similar to that', the action-adjunct is established; as in 'the king behaves like a king', when difference is intended.

Tena tulyam iti prapteh kriyopadhih prasidhyati: from the *applicability* of *'tena tulyam'* — the rule's own words — the *action-adjunct* is *established*. Rajavad vartate raja ity atra bhede vivaksite: as in *'rajavad vartate raja'* — *the king behaves like a king* — when *difference* is *intended*. So the strange sentence is the test-case: a king compared to a king, which 3.14.544 said was impossible for want of a difference. It works only if a difference is *intended* — this actual king against kingship's exemplars. And 3.14.553 names them.

3.14.553 unformalized

राजत्वेन प्रसिद्धा ये पृथुप्रभृतयो नृपाः ।
युधिष्ठिरान्तास्तेऽन्येषामुपमानं महीक्षिताम् ॥ ५५३ ॥

rājatvena prasiddhā ye pṛthuprabhṛtayo nṛpāḥ |
yudhiṣṭhirāntāste'nyeṣāmupamānaṃ mahīkṣitām || 553 ||

Those kings well-known for kingship, from Prthu down to Yudhisthira — they are the comparans for other rulers of the earth.

Rajatvena prasiddha ye prthu-prabhrtayo nrpah yudhisthirantah: those kings *well-known* for *kingship* — prasiddha (3.14.360, 3.14.373's word: famous, established) — *from Prthu* — the first king, who levelled the earth — *down to Yudhisthira*. Te 'nyesam upamanam mahiksitam: *they* are the *comparans* for *other* rulers of the earth. So 'rajavad vartate raja' means: he behaves as Prthu and Yudhisthira did. The difference 3.14.552 required is the difference between a living king and the exemplary dead, and the comparison is possible because fame has made a class of them (3.14.373's the comparans exceeds everywhere, from its being well-known).

3.14.554 unformalized

सिद्ध्यसिद्धिकृतो भेद उपमानोपमेययोः ।
सर्वत्रैव यतोऽसिद्धं प्रसिद्धेनोपमीयते ॥ ५५४ ॥

siddhyasiddhikṛto bheda upamānopameyayoḥ |
sarvatraiva yato'siddhaṃ prasiddhenopamīyate || 554 ||

The difference of comparans and comparandum is made by establishment and non-establishment; since everywhere, the unestablished is compared with the well-known.

The principle, stated for the last time and in its plainest form. Siddhy-asiddhi-krto bheda upamana-upameyayoh: the *difference* of comparans and comparandum is *made by* *establishment and non-establishment* — 3.14.397's siddhy-asiddhibhyam, and 3.14.398's sampurnatvat: the general becomes a particular by being complete. Sarvatra eva yatah asiddham prasiddhena upamiyate: since *everywhere*, the *unestablished* is *compared with* the *well-known*. Sarvatra eva: everywhere, without exception. So there is no comparison between equals: one term is always famous and the other always is not, and that asymmetry *is* what makes them two. The whole comparison-doctrine rests on fame.

3.14.555 unformalized

राजवद् रूपमस्येति राजन्येव विवक्षिते ।
अक्रियार्थेन योगेन द्वितीयेन भविष्यति ॥ ५५५ ॥

rājavad rūpamasyeti rājanyeva vivakṣite |
akriyārthena yogena dvitīyena bhaviṣyati || 555 ||

'His form is like a king's' — when a king alone is intended, it will occur by the second rule, which is not for the sake of the action.

Rajavad rupam asya iti rajany eva vivaksite: *'his form is like a king's'* — when a *king alone* is *intended*: no action in view, only the man and his bearing. Akriyarthena yogena dvitiyena bhavisyati: it will occur by the *second* rule, which is *not for the sake of action* — 3.14.536-538's division: the earlier rule covers actions, the later substances and qualities. So this case falls to the later rule, and the division holds. The samuddesa is checking its own settlement against a fresh example, and it passes.

3.14.556 unformalized

उपमानाविवक्षायां नियमार्थोऽयमुच्यते ।
धर्मोऽर्हतिक्रियाकर्ता तदर्थं वचनं पुनः ॥ ५५६ ॥

upamānāvivakṣāyāṃ niyamārtho'yamucyate |
dharmo'rhatikriyākartā tadarthaṃ vacanaṃ punaḥ || 556 ||

When the comparans is not intended, this is stated as having restriction for its purpose; the property is the agent of the action 'befits'; the statement is again for that purpose.

Upamana-avivaksayam niyamartho 'yam ucyate: when the *comparans* is *not intended*, this is stated as having *restriction* for its purpose — niyama: a rule that narrows rather than enables (3.14.535's niyame va abhidhane va). Dharmo 'rhati-kriya-karta: the *property* is the *agent* of the action *'arhati'* — befits, is worthy of: and what does the befitting is a property, not a person. Tad-artham vacanam punah: the *statement* is *again* for that purpose. So where no comparison is meant, the rule restricts, and what it restricts is which properties may be said to befit. The vati-affix has a second life outside comparison altogether.

3.14.557 unformalized

कृतहस्तवदित्येतत् प्रसिद्धेष्वेव दृश्यते ।
राजत्वेन प्रसिद्धे च राज्ञि राजवदित्यपि ॥ ५५७ ॥

kṛtahastavadityetat prasiddheṣveva dṛśyate |
rājatvena prasiddhe ca rājñi rājavadityapi || 557 ||

'Like a trained hand' is seen only in the well-known; and 'like a king' too, in a king well-known for kingship.

Krta-hastavad ity etat prasiddhesv eva drsyate: *'krta-hastavat'* — like a trained hand, said of a skilled archer — is *seen* *only in the well-known*: one says it only of the acknowledged master. Rajatvena prasiddhe ca rajni rajavad ity api: and *'rajavat'* too, in a *king* *well-known for kingship*. So both idioms require their subject to be already famous — the comparison does not confer the standing, it presupposes it. This is 3.14.554's principle at work on the comparandum's side: the compared thing must be established enough to be worth comparing, and only then does fame's surplus have anywhere to land.

3.14.558 unformalized

अराज्ञि येषां धर्माणां दृष्टोऽत्यन्तमसम्भवः ।
ते राजनि नियम्यन्ते त्यज्यन्ते व्यभिचारिणा ॥ ५५८ ॥

arājñi yeṣāṃ dharmāṇāṃ dṛṣṭo'tyantamasambhavaḥ |
te rājani niyamyante tyajyante vyabhicāriṇā || 558 ||

Those properties of which utter impossibility is seen in a non-king — they are restricted to a king; the deviating ones are abandoned.

Arajni yesam dharmanam drsto 'tyantam asambhavah: those properties of which *utter impossibility* is *seen* in a *non-king* — the things nobody but a king can do or be. Te rajani niyamyante: *they* are *restricted* to a *king* — niyamyante: fixed on him. Tyajyante vyabhicarina: the *deviating* ones are *abandoned* — vyabhicarin: what strays, what a non-king might also have (3.14.548's vyabhicara). So 'rajavat' means: with the properties no one but a king could have — and the ones a commoner might share are dropped. The selection is by exclusivity, and 3.14.415's problem (which of the moon's properties does 'moon-faced' take?) here gets a criterion: whatever nothing else could have.

3.14.559 unformalized

अर्हतेश्च क्रिया कर्त्री या तस्यां वतिरिष्यते ।
राजानमर्हतिच्छत्रमिति न त्वेवमादिषु ॥ ५५९ ॥

arhateśca kriyā kartrī yā tasyāṃ vatiriṣyate |
rājānamarhaticchatramiti na tvevamādiṣu || 559 ||

And that action which is the agent of 'befits' — in it 'vati' is accepted; but not in cases like 'the umbrella befits the king'.

Arhates ca kriya kartri ya tasyam vatir isyate: that *action* which is the *agent* of *'arhati'* — where an action does the befitting — in *it* *'vati'* is *accepted*: 'rajavad arhati', it befits as it befits a king. Rajanam arhati chatram iti na tv evam adisu: *but not* in *'the umbrella befits the king'* and the like — where the umbrella is the agent and a person the object. So the affix comes only when what befits is an action, not when it is a thing. The distinction is fine and it is the last of the vati-rule's boundaries: the affix lives on actions and their possessors, and a befitting umbrella is neither.

3.14.560 unformalized

प्रयुक्तानां हि शब्दानां शाश्त्रेणानुगमः सताम् ।
छत्राद्यर्थे तु वचने प्रत्याख्यानं न सम्भवेत् ॥ ५६० ॥

prayuktānāṃ hi śabdānāṃ śāśtreṇānugamaḥ satām |
chatrādyarthe tu vacane pratyākhyānaṃ na sambhavet || 560 ||

For the science follows out words that exist and are employed; but if there were a statement in the sense of 'umbrella' and the rest, rejection would not be possible.

Prayuktanam hi sabdanam sastrena anugamah satam: the *science* *follows out* words that *exist* and are *employed* — prayukta and sat: said, and there. Anugamah: it follows *after* (3.14.81's anugamyate; 3.14.154's akhyayate ca sastrena). Chatrady-arthe tu vacane pratyakhyanam na sambhavet: but if there were a *statement* in the sense of *'umbrella'* and the rest, *rejection* would *not* be possible. So one cannot reject a rule that covers actually-occurring usage: if 'the umbrella befits the king' (3.14.559) were provided for, the provision could not be thrown out. The science's authority over its rules stops where usage begins.

3.14.561 unformalized

तदर्हमिति नारब्धं सूत्रं व्याकरणान्तरे ।
सम्भवत्युपमात्रापि भेदस्य परिकल्पना ॥ ५६१ ॥

tadarhamiti nārabdhaṃ sūtraṃ vyākaraṇāntare |
sambhavatyupamātrāpi bhedasya parikalpanā || 561 ||

The sutra 'tad-arham' is not undertaken in another grammar; and even in the mere comparison, the construction of difference is possible.

Tad-arham iti na arabdham sutram vyakaranantare: the sutra *'tad-arham'* is *not undertaken* in *another grammar* — vyakaranantara: a rival system, which manages without the rule. Sambhavaty upamatrapi bhedasya parikalpana: and even in the *mere comparison*, the *construction of difference* is *possible* — parikalpana: a construction (3.14.75's word). So the rule is dispensable: other grammars do without it, and the difference it was meant to secure can be constructed anyway. Bhartrhari cites a rival system's economy against Panini's — rarely, and here without apology.

3.14.562 unformalized

एकस्य कार्यनिर्ज्ञानात् सिद्धस्य विषयान्तरे ।
तद्धर्मत्वविवक्षायां बुद्ध्या भेदः प्रकल्प्यते ॥ ५६२ ॥

ekasya kāryanirjñānāt siddhasya viṣayāntare |
taddharmatvavivakṣāyāṃ buddhyā bhedaḥ prakalpyate || 562 ||

From knowing the operation of one thing, established in another domain — when possession of that property is intended, difference is constructed by the intellect.

Ekasya karya-nirjnanat siddhasya visayantare: from *knowing the operation* of *one* thing, *established* in *another domain* — one knows what a king does, and the king belongs to another sphere than this man. Tad-dharmatva-vivaksayam buddhya bhedah prakalpyate: when *possession of that property* is *intended*, *difference* is *constructed* *by the intellect*. Buddhya: by the intellect (3.14.188's buddhya eva; 3.14.15's the intellect is the ordainer of the meaning). So the difference the comparison stands on (3.14.544's asraya) is not found in the things: it is built by the mind when it wants to say that this man has that man's property. This is the thesis the next eight verses unfold, and it is the samuddesa's last great argument.

3.14.563 unformalized

सूत्रारम्भान्न चैतस्मादिवशब्दस्य विद्यते ।
प्रयोगः सोऽपि चैतस्य विषये विद्यते वतेः ॥ ५६३ ॥

sūtrārambhānna caitasmādivaśabdasya vidyate |
prayogaḥ so'pi caitasya viṣaye vidyate vateḥ || 563 ||

And from this sutra's undertaking, no employment of the word 'iva' obtains; and that too exists in the domain of this 'vati'.

Sutrarambhan na ca etasmad iva-sabdasya vidyate prayogah: from *this sutra's* *undertaking*, no *employment* of the word *'iva'* obtains — where the rule gives 'vati', 'iva' is not used. So 'py etasya visaye vidyate vateh: and *that too* — 'iva' — *exists* in the *domain* of this *'vati'*. So the two do coexist after all: the affix does not banish the particle from its territory. The observation matters because 3.14.542-546 had drawn them sharply apart, and here they overlap — which is what one should expect of a dyotaka and a vacaka that reveal and say the same thing.

3.14.564 unformalized

दस्युहेन्द्र इवेत्येतदेन्द्रे मन्त्रे प्रयुज्यते ।
अन्यत्र दृष्टकर्मेन्द्रो यथेत्यस्मिन् विवक्षिते ॥ ५६४ ॥

dasyuhendra ivetyetadendre mantre prayujyate |
anyatra dṛṣṭakarmendro yathetyasmin vivakṣite || 564 ||

'Slayer of dasyus, like Indra' — this is employed in a mantra to Indra; elsewhere, when 'as Indra, whose deeds are seen' is intended.

Dasyu-hendra iva ity etad aindre mantre prayujyate: *'slayer of dasyus, like Indra'* — this is employed in a *mantra to Indra*: the god is compared to himself, in his own hymn. Anyatra drsta-karmendro yatha ity asmin vivaksite: *elsewhere*, when *'as Indra, whose deeds are seen'* is *intended* — of a man, with the god's famous exploits in view. So the same phrase does two quite different things: in the mantra Indra is likened to Indra, and outside it a man is likened to Indra. The first is 3.14.552's rajavad vartate raja — a thing compared to itself — and the Veda does it, which settles that it can be done.

3.14.565 unformalized

पूर्वामवस्थामाश्रित्या यावस्था व्यपदिश्यते ।
सदृशस्त्वं तवैवेति तत्रैवमभिधीयते ॥ ५६५ ॥

pūrvāmavasthāmāśrityā yāvasthā vyapadiśyate |
sadṛśastvaṃ tavaiveti tatraivamabhidhīyate || 565 ||

Resorting to a former state, the state which is designated — 'you are like your own self' — there it is thus expressed.

Purvam avastham asritya ya avastha vyapadisyate: *resorting to* a *former state*, the *state* which is *designated*. Sadrsas tvam tava eva iti tatra evam abhidhiyate: *'you are like your own self'* — *there* it is thus *expressed*. Tava eva sadrsah: like *yourself alone* — the highest praise the language has, and it is a comparison of a thing to itself. And the analysis: what it compares is two *states* of one person, the present to the former, and the difference is temporal. So 3.14.564's self-comparison is not paradoxical: the intellect has divided one thing into its states (3.14.562's buddhya bhedah prakalpyate), and the states are two.

3.14.566-567 unformalized

प्रसिद्धभेदं यत्रान्यदुपमानं न विद्यते ।
उपमेयस्य तत्रात्मा स्वबुद्ध्या प्रविभज्यते ॥ ५६६ ॥
योऽपि स्वाभाविको भेदः सोऽपि बुद्धिनिबन्धनः ।
तेनास्मिन् विषये भिन्नमभिन्नं वा न विद्यते ॥ ५६७ ॥

prasiddhabhedaṃ yatrānyadupamānaṃ na vidyate |
upameyasya tatrātmā svabuddhyā pravibhajyate || 566 ||
yo'pi svābhāviko bhedaḥ so'pi buddhinibandhanaḥ |
tenāsmin viṣaye bhinnamabhinnaṃ vā na vidyate || 567 ||

(566) Where no other comparans of established difference exists — there the comparandum's own self is divided by one's own cognition. (567) Even the difference which is natural — that too is grounded in cognition; therefore in this domain there is neither the different nor the non-different.

(566): prasiddha-bhedam yatra anyad upamanam na vidyate — where no *other* comparans of *established difference* exists: nothing outside is good enough, as nothing outside is good enough for Indra. Upameyasya tatra atma sva-buddhya pravibhajyate: *there* the *comparandum's own self* is *divided* *by one's own cognition* — pravibhajyate: is cut apart. So one cuts the thing in two and compares the halves. (567): yo 'pi svabhaviko bhedah so 'pi buddhi-nibandhanah — *even* the difference which is *natural* — svabhavika: belonging to the thing's own nature (3.14.186's word) — *that too* is *grounded in cognition*. Tena asmin visaye bhinnam abhinnam va na vidyate: *therefore*, in *this domain*, *neither* the *different* *nor* the *non-different* *exists*. So the division of Indra from Indra is not a special expedient: *every* difference is the mind's, including the ones that look built-in, and once that is seen, neither difference nor identity is available as a fact about things. This is 3.14.142's na nirupyate given its ground.

3.14.568 unformalized

अङ्गदी कुण्डली चेति दर्शयन् भेदहेतुभिः ।
चैत्रमीदृश इत्याह बुद्ध्यवस्थापरिग्रहात् ॥ ५६८ ॥

aṅgadī kuṇḍalī ceti darśayan bhedahetubhiḥ |
caitramīdṛśa ityāha buddhyavasthāparigrahāt || 568 ||

Showing Caitra by the difference-causes 'bracelet-wearing' and 'ear-ringed' — one says 'he is such', from grasping a state of the intellect.

Angadi kundali ca iti darsayan bheda-hetubhih: *showing* by the *difference-causes* — *'bracelet-wearing'*, *'ear-ringed'*: the ornaments by which one picks a man out. Caitram idrsa ity aha: one says of *Caitra*, *'he is such'*. Buddhy-avastha-parigrahat: from *grasping* a *state of the intellect* — buddhy-avastha: a condition of the mind, not of Caitra. So the man is identified by his ornaments and what one has actually grasped is a mental state. Idrsah — *such* — points at a cognition, not a man. This is 3.14.567's thesis in the plainest possible case: even 'that one, with the bracelet' reaches only a buddhy-avastha.

3.14.569 unformalized

एतैः शब्दैर्यथाभूतः प्रत्ययात्मोपजायते ।
तत्प्रत्ययानुकारेण विषयोऽप्युपपद्यते ॥ ५६९ ॥

etaiḥ śabdairyathābhūtaḥ pratyayātmopajāyate |
tatpratyayānukāreṇa viṣayo'pyupapadyate || 569 ||

The cognition-self which arises by these words, as it is — the object too comes about in conformity with that cognition.

Etaih sabdair yathabhutah pratyayatma upajayate: the *cognition-self* which *arises* by *these words*, *as it is* — the words make a cognition, and the cognition has whatever shape they gave it. Tat-pratyaya-anukarena visayo 'py upapadyate: the *object* too *comes about* *in conformity with* that *cognition* — anukara: following after, imitating. So the object *follows* the cognition, and the cognition follows the words: the order of dependence runs word, then thought, then thing. This is 3.14.197's sabdad arthah pratayante — meanings are spread out from the word — and 1.123's no cognition without word, arriving as the conclusion of an argument about jewellery.

3.14.570 unformalized

बुद्ध्यवस्थाविभागेन भेदकार्यं प्रतीयते ।
जन्यन्त इव शब्दानामर्थाः सर्वे विवक्षया ॥ ५७० ॥

buddhyavasthāvibhāgena bhedakāryaṃ pratīyate |
janyanta iva śabdānāmarthāḥ sarve vivakṣayā || 570 ||

By the division of the intellect's states, the difference-operation is cognized; all the meanings of words are, as it were, produced by the intention to speak.

One of the summit verses of the Vrtti-samuddesa, and of the work. Buddhy-avastha-vibhagena bheda-karyam pratiyate: by the *division* of the *intellect's states*, the *difference-operation* is *cognized* — every rule that needs a difference to operate on is fed by the mind's dividing itself (3.14.567's even natural difference is buddhi-nibandhana). Janyanta iva sabdanam arthah sarve vivaksaya: *all* the *meanings* of *words* are, *as it were*, *produced* by the *intention to speak*. Janyante: are *generated*, brought into being. Sarve: all of them. Vivaksaya: by the wish to say. And iva — *as it were* — which is the whole of Bhartrhari's caution and the whole of his boldness: he will not say the world is made by speaking, and he will say it is *as though* made by speaking, and he will not take the iva back. 3.14.197 said the word ordains the differences; this says the wish to speak generates the meanings. Between them the samuddesa's method — analyses are devices, divisions are the mind's, the science stands far off — is not scepticism at all. It is this.

3.14.571 unformalized

तथाविधेऽपि बाह्येऽर्थे भिद्यन्ते यत्र बुद्धयः ।
न तत्र कश्चित् सादृश्यं सदपि प्रतिपाद्यते ॥ ५७१ ॥

tathāvidhe'pi bāhye'rthe bhidyante yatra buddhayaḥ |
na tatra kaścit sādṛśyaṃ sadapi pratipādyate || 571 ||

Even where the external object is of the same kind, wherever the cognitions differ — there no one conveys a similarity, though it exists.

The test of 3.14.570, negative. Tathavidhe 'pi bahye 'rthe: even where the *external object* is *of the same kind* — the things really are alike, out there. Bhidyante yatra buddhayah: wherever the *cognitions* *differ*. Na tatra kascit sadrsyam sad api pratipadyate: *there* *no one* conveys a *similarity*, *though it exists* — sad api: though it is there. So an actual resemblance in the world goes unsaid if the cognitions have not converged: nobody makes the comparison, and nobody could. The external likeness is real and inert. This is as sharp a statement as the work gives of what it means for meanings to be generated by vivaksa: the world's own resemblances do not get into speech unless the mind has already made them.

3.14.572 unformalized

अत्यन्तं विषये भिन्ने यावत् प्रख्या न भिद्यते ।
न तावत् प्रत्यभिज्ञानं कस्यचिद् विनिवर्तते ॥ ५७२ ॥

atyantaṃ viṣaye bhinne yāvat prakhyā na bhidyate |
na tāvat pratyabhijñānaṃ kasyacid vinivartate || 572 ||

Even when the object is utterly different, so long as the presentation is not differentiated — so long does no one's recognition withdraw.

The converse test, and it is the stronger. Atyantam visaye bhinne: even when the *object* is *utterly different* — atyantam: completely; two different things, out there. Yavat prakhya na bhidyate: so long as the *presentation* is *not differentiated* — prakhya: the showing-forth (3.14.187's eka-prakhya; 3.14.403's jati-bhagasraya prakhya). Na tavat pratyabhijnanam kasyacid vinivartate: *so long* does *no one's* *recognition* *withdraw* — pratyabhijnana: the recognition 'this is that'. So one goes on recognizing a thing as the same across an utter difference, provided the presentation has not split. With 3.14.571 the pair is complete: real likeness without mental likeness conveys nothing, and real difference without mental difference does not stop recognition. The mind, not the world, is where sameness and difference are.

3.14.573 unformalized

अयमेव तु सूत्रेण भेदो भेदेन दर्शितः ।
प्रसिद्धमपि दुर्ज्ञानमबुधः प्रतिपद्यते ॥ ५७३ ॥

ayameva tu sūtreṇa bhedo bhedena darśitaḥ |
prasiddhamapi durjñānamabudhaḥ pratipadyate || 573 ||

This very difference is shown by the sutra by way of difference; the unlearned apprehends what is well-known but hard to know.

Ayam eva tu sutrena bhedo bhedena darsitah: *this very difference* — the mind-made one of 3.14.562-572 — is *shown* by the *sutra* *by way of difference*: the rule exhibits it as though it were a difference among things. Darsitah: exhibited (3.14.42's bheda-abheda-nidarsanam krtam). Prasiddham api durjnanam abudhah pratipadyate: the *unlearned* *apprehends* what is *well-known* but *hard to know* — prasiddham api durjnanam: familiar and hard: everyone lives by these differences and nobody can say what they are. So the rule is a teaching-device for the abudha (3.14.49's abudhan praty upayas ca vicitrah), exhibiting as objective a difference that is the mind's. The samuddesa's whole account of grammar, in one verse, applied to its own last doctrine.

3.14.574 unformalized

वैयाकरणवद् ब्रूते न वैयाकरणः सदा ।
वैयाकरणवद् ब्रूष्वेत्यतः सोऽप्यभिधीयते ॥ ५७४ ॥

vaiyākaraṇavad brūte na vaiyākaraṇaḥ sadā |
vaiyākaraṇavad brūṣvetyataḥ so'pyabhidhīyate || 574 ||

One who is not always a grammarian speaks like a grammarian; 'speak like a grammarian' — therefore he too is expressed.

Vaiyakaranavad brute na vaiyakaranah sada: one who is *not always a grammarian* speaks *like a grammarian* — na sada: not always: even a real grammarian is only sometimes grammatical, and so 'vaiyakaranavat' is said of him too. Vaiyakaranavad brusva ity atah so 'py abhidhiyate: *'speak like a grammarian'* — *therefore* *he too* is *expressed*. So one may tell a grammarian to speak like a grammarian, and the injunction is not empty: it asks him to be, now, what he is only sometimes. This is 3.14.565's tava eva sadrsah — like your own self — in a professional key, and there is a joke in it that Bhartrhari, of all people, was entitled to make.

3.14.575 unformalized

केचित् पुमांसो भाषन्ते स्त्रीवत्, पुंवच्च योषितः ।
व्यभिचारे स्वधर्मोऽपि पुनस्तेनोपदिश्यते ॥ ५७५ ॥

kecit pumāṃso bhāṣante strīvat, puṃvacca yoṣitaḥ |
vyabhicāre svadharmo'pi punastenopadiśyate || 575 ||

Some men speak like women, and women like men; in the deviation, one's own property too is again taught by that.

Kecit pumamso bhasante strivat, pumvac ca yositah: *some men* speak *like women*, and *women* *like men* — the vati-affix used of what one is. Vyabhicare sva-dharmo 'pi punas tena upadisyate: in the *deviation*, one's *own property* too is *again* *taught* by that — vyabhicara: straying (3.14.548, 3.14.558's word). So where a man speaks womanishly, the word 'strivat' also teaches what his own property was — you cannot say he speaks like a woman without saying he is not one. The comparison, in marking the straying, re-teaches the norm it strayed from. The observation is exact, and it holds for every 'X-vat' said of someone who is not X.

3.14.576 unformalized

सदृशस्त्वं तवैवेति लोके यदभिधीयते ।
उपमानान्तरं तत्र प्रसक्तं विनिवर्तते ॥ ५७६ ॥

sadṛśastvaṃ tavaiveti loke yadabhidhīyate |
upamānāntaraṃ tatra prasaktaṃ vinivartate || 576 ||

'You are like your own self' — what is said in the world; there, another comparans, which was applicable, withdraws.

3.14.565's phrase given its logic. Sadrsas tvam tava eva iti loke yad abhidhiyate: *'you are like your own self'* — what is *said in the world*: not a grammarian's construct but the language's own idiom. Upamanantaram tatra prasaktam vinivartate: *there*, *another* comparans, which was *applicable*, *withdraws* — prasakta: it was available, and it goes. So the praise works by *dismissing* every other standard: to say a man is like himself is to say that all the famous exemplars have been considered and found short. The self-comparison is not a failure to find a comparans; it is the refusal of the ones on offer, and the refusal is the compliment.

3.14.577 unformalized

युक्तमौपयिकं राज्ञ इत्यर्थस्य निदर्शने ।
उपमानाविवक्षायां तदर्हमिति पठ्यते ॥ ५७७ ॥

yuktamaupayikaṃ rājña ityarthasya nidarśane |
upamānāvivakṣāyāṃ tadarhamiti paṭhyate || 577 ||

'It is fitting, suitable for a king' — in illustrating the meaning; when the comparans is not intended, 'tad-arham' is recited.

Yuktam aupayikam rajna ity arthasya nidarsane: *'it is fitting, suitable for a king'* — in the *illustration* of the *meaning*: what 'rajarham' comes to. Upamana-avivaksayam tad-arham iti pathyate: when the *comparans* is *not intended*, *'tad-arham'* is *recited* — the sutra's name. So the rule has two employments: with a comparans in view it gives comparison, without one it gives fitness (3.14.556's niyamartho 'yam ucyate). The two are distinguished by vivaksa alone, and Panini gave them one rule — which 3.14.561 noted another grammar does without.

3.14.578-579 unformalized

प्रसक्तानुप्रसक्तस्तु वतिशेषोऽभिधीयते ।
उपमानाभिसम्बन्धादस्मिन् वतिरुदाहृतः ॥ ५७८ ॥
प्रधानकल्पनाऽभावे गुणशब्दस्य दर्शनात् ।
उपसर्गाद् वतौ सिद्धा धातौ धात्वर्थकल्पना ॥ ५७९ ॥

prasaktānuprasaktastu vatiśeṣo'bhidhīyate |
upamānābhisambandhādasmin vatirudāhṛtaḥ || 578 ||
pradhānakalpanā'bhāve guṇaśabdasya darśanāt |
upasargād vatau siddhā dhātau dhātvarthakalpanā || 579 ||

(578) The remainder of 'vati', occasioned incidentally, is stated; from connection with the comparans, 'vati' is cited in this. (579) In the absence of any construction of a principal, from the seeing of a quality-word — in 'vati' after a preverb, the construction of the root-meaning in the root is established.

(578): prasakta-anuprasaktas tu vati-seso 'bhidhiyate — the *remainder* of 'vati', *occasioned incidentally*: prasakta-anuprasakta, one applicability dragging in another. Upamana-abhisambandhad asmin vatir udahrtah: from *connection with the comparans*, 'vati' is *cited* in this. (579): pradhana-kalpana-abhave guna-sabdasya darsanat — in the *absence* of any *construction of a principal*, from the *seeing of a quality-word*. Upasargad vatau siddha dhatau dhatv-artha-kalpana: in *'vati'* after a *preverb*, the *construction of the root-meaning* in the *root* is *established*. So where no principal is constructed, the root's meaning has to be constructed into the root — which is 3.14.68's roots with imagined selves, and 3.14.276's the preverb's sense not being in the root. Kalpana at every level, exactly as 3.14.75 said.

3.14.580 unformalized

स्वं रूपमिति चैतस्मिन्नर्थस्यापि परिग्रहः ।
रूपवज्ज्ञापितस्तस्मादसन्नोऽर्थो ग्रहीष्यते ॥ ५८० ॥

svaṃ rūpamiti caitasminnarthasyāpi parigrahaḥ |
rūpavajjñāpitastasmādasanno'rtho grahīṣyate || 580 ||

And in this, by 'its own form', the meaning too is included; being indicated as the form is — therefore the near meaning will be taken.

Svam rupam iti ca etasminn arthasya api parigrahah: by *'svam rupam'* — Panini's rule that a word in the grammar stands for itself — the *meaning* too is *included*. So the rule that makes a word self-referring does not exclude its sense: the form is meant *and* the meaning comes along. Rupavaj jnapitas tasmad asanno 'rtho grahisyate: being *indicated* as the *form* is, *therefore* the *near* meaning will be *taken*. Asanna: the *nearby* one (3.14.242's asanna eva; 3.14.295's asatti). So one takes the closest sense, not any sense — the same nearness-principle that bounded paryudasa's drift. Even self-reference does not escape it.

3.14.581 unformalized

धात्वर्थेनोपजनितं साधनत्वेन साधनम् ।
धातुना कृतमित्येवमस्मिन् सूत्रे प्रतीयते ॥ ५८१ ॥

dhātvarthenopajanitaṃ sādhanatvena sādhanam |
dhātunā kṛtamityevamasmin sūtre pratīyate || 581 ||

The means, produced by the root-meaning as a means — 'made by the root' — thus it is cognized in this sutra.

Dhatv-arthena upajanitam sadhanatvena sadhanam: the *means*, *produced by the root-meaning* *as a means* — the karaka is generated by the action and generated *as* a karaka: it has no prior standing (3.7.3's the sadhanas as capacities the action confers; 3.14.220's the verb makes object and agent). Dhatuna krtam ity evam asmin sutre pratiyate: *'made by the root'* — thus it is cognized in *this sutra*. So Panini's wording is read as saying that the root makes its means. Krtam: *made*. This is the karaka-doctrine's strongest formulation and Bhartrhari finds it in a rule about affixes.

3.14.582 unformalized

यः शब्दश्चरितार्थत्वादत्यन्तं न प्रयुज्यते ।
विषयेऽदर्शनात् तत्र लोपस्तस्याभिधीयते ॥ ५८२ ॥

yaḥ śabdaścaritārthatvādatyantaṃ na prayujyate |
viṣaye'darśanāt tatra lopastasyābhidhīyate || 582 ||

That word which, its purpose being served, is never employed — its elision is stated, from its non-appearance in the domain.

Yah sabdas carita-arthatvad atyantam na prayujyate: that word which, its *purpose being served* — carita-artha: having done its work — is *never* employed. Visaye 'darsanat tatra lopas tasya abhidhiyate: from its *non-appearance* in the *domain*, its *elision* is *stated*. So the grammar declares an elision for a word that was never there — the word had already done its job (as a conditioning presence) and was never audible (3.14.468's upadanam prayogesu tasya atyantam na vidyate). Lopa is thus the technical name for a word that finished before it was spoken, and 3.14.128's what withdraws has no being applies: the elision is bookkeeping.

3.14.583 unformalized

क्रियायां साधने द्रव्ये प्रादयो ये व्यवस्थिताः ।
तेभ्यः सत्त्वाभिधायिभ्यो वतिः स्वार्थे विधीयते ॥ ५८३ ॥

kriyāyāṃ sādhane dravye prādayo ye vyavasthitāḥ |
tebhyaḥ sattvābhidhāyibhyo vatiḥ svārthe vidhīyate || 583 ||

Those 'pra' and the rest which are settled in an action, in a means, or in a substance — from those, as expressing substance, 'vati' is prescribed in its own meaning.

Kriyayam sadhane dravye pradayo ye vyavasthitah: those *'pra'* and the rest — the preverbs — *settled* in an *action*, in a *means*, or in a *substance*: the three places a preverb can lodge. Tebhyah sattva-abhidhayibhyo vatih svarthe vidhiyate: from *those*, as *expressing substance*, *'vati'* is prescribed *in its own meaning* — svarthe: adding nothing (3.14.174's svarthika). So the affix attaches to preverbs when they have gone over to expressing substance, and it adds nothing when it does. The preverb, that most verbal of elements, becomes a substance-word and takes a nominal affix — 3.14.308's nan changing category by composition, happening again.

3.14.584 unformalized

प्रत्ययेन विना प्रादिस्तत्रार्थेन प्रयुज्यते ।
भेदेन तु समाख्याने विभागः परिकल्पितः ॥ ५८४ ॥

pratyayena vinā prādistatrārthena prayujyate |
bhedena tu samākhyāne vibhāgaḥ parikalpitaḥ || 584 ||

There, 'pra' and the rest is employed without an affix, along with the meaning; but in the designation by division, the division is constructed.

Pratyayena vina pradis tatra arthena prayujyate: *there*, *'pra'* and the rest is employed *without an affix*, *with the meaning* — the preverb carries its sense unaided. Bhedena tu samakhyane vibhagah parikalpitah: but in the *designation by division*, the *division* is *constructed* — parikalpita. So the affix of 3.14.583 is not needed and the analysis that requires it is a construct. This is 3.14.35's samudayam upakramya padam tasyam prayujyate, vibhagena samakhyane — having begun with the aggregate, the division is what the designation makes. The samuddesa's oldest finding, in its last technical application.

3.14.585 unformalized

अनङ्गीकृतसत्त्वं तु यदि गृह्येत साधनम् ।
विभक्तिभिर्न योगः स्याद् यथैव तसिलादिषु ॥ ५८५ ॥

anaṅgīkṛtasattvaṃ tu yadi gṛhyeta sādhanam |
vibhaktibhirna yogaḥ syād yathaiva tasilādiṣu || 585 ||

But if the means were grasped without substance being accepted — there would be no connection with case-endings, as in 'tasil' and the rest.

Anangikrta-sattvam tu yadi grhyeta sadhanam: if the *means* were *grasped* *without substance being accepted* — a karaka with no thing in it. Vibhaktibhir na yogah syad yatha eva tasiladisu: there would be *no connection with case-endings*, as in *'tasil'* and the rest — the affix giving 'tatah', 'from there': an indeclinable, taking no case. So a substanceless means would be an adverb. The argument is 3.14.163's dravyam eva asrayah in its karaka-form: strip the substance out and the case-system has nothing to attach to, and one is left with indeclinables.

3.14.586 unformalized

पाठाद् यैरविभक्तित्वं वत्यन्तेष्वनुगम्यते ।
तेषामुद्वत् इत्यत्र वक्तव्या सविभक्तिता ॥ ५८६ ॥

pāṭhād yairavibhaktitvaṃ vatyanteṣvanugamyate |
teṣāmudvat ityatra vaktavyā savibhaktitā || 586 ||

By those who follow, from the recitation, the case-endinglessness of vati-ending forms — for them, in 'udvat', the having-of-a-case-ending must be stated.

Pathad yair avibhaktitvam vaty-antesv anugamyate: by those who *follow*, from the *recitation* — the gana-list — the *case-endinglessness* of *vati*-ending forms: those who list vati among the avyayas (3.14.521's proposal). Tesam udvat ity atra vaktavya savibhaktita: for *them*, in *'udvat'*, the *having-of-a-case-ending* must be *stated* — vaktavya: they must say it. So the listing costs them an extra rule: 'udvat' does take endings, and if vati-forms are indeclinable by the list, an exception must be written. This is the samuddesa's standing audit again: every doctrine is charged for the rules it adds (3.14.318's the prohibition is purposeless; here, a rule becomes necessary).

3.14.587 unformalized

वत्यर्थं नावगाहेते पुंवदित्यस्य दर्शनात् ।
नञ्स्नञावपवादस्य बाधकं तन्निपातनम् ॥ ५८७ ॥

vatyarthaṃ nāvagāhete puṃvadityasya darśanāt |
nañsnañāvapavādasya bādhakaṃ tannipātanam || 587 ||

'Nañ' and 'snañ' do not enter into the vati-meaning, from the seeing of 'pumvat'; that irregular form is the annuller of the exception.

Vaty-artham na avagahete pumvad ity asya darsanat: *'nañ'* and *'snañ'* do *not enter into* the *vati-meaning*, from the *seeing* of *'pumvat'* — the form is there and it has no nañ or snañ in it. Nan-snanav apavadasya badhakam tan nipatanam: that *irregular form* — nipatana: a form Panini simply lists as given — is the *annuller* of the *exception*. So an irregularly-listed form defeats a rule that would otherwise have applied. Nipatana beats apavada: what the grammar simply records overrides what it derives. This is a real principle of the system and it is 3.14.560's in another key — the science follows what is said.

3.14.588 unformalized

एवमुत्क्रामतो नूनं वत्यर्थं नञ्स्नञाविति ।
तयोः प्रवृत्तावुत्सर्गो बाधनान्नोपपद्यते ॥ ५८८ ॥

evamutkrāmato nūnaṃ vatyarthaṃ nañsnañāviti |
tayoḥ pravṛttāvutsargo bādhanānnopapadyate || 588 ||

Thus, 'nañ' and 'snañ' surely transgressing the vati-meaning — in their operation the general rule, being annulled, is not possible.

Evam utkramato nunam vaty-artham nan-snanav iti: thus *'nañ'* and *'snañ'* *surely* *transgressing* the *vati*-meaning — utkramatah: overstepping, going past their bounds. Tayoh pravrttav utsargo badhanan na upapadyate: in *their operation*, the *general rule*, being *annulled*, is *not possible*. Utsarga: the general provision; badhana: its annulment by the special (3.14.216's badhaka). So if the two affixes did overstep into vati's territory, the general rule would be annulled there and could not operate. The consequence is unacceptable, and the argument is the standard one of rule-conflict — special defeats general, and one must therefore keep the special off the general's ground.

3.14.589 unformalized

नञ्स्नञौ विहितौ येन स योगो नावगाहते ।
वतिकरणं तद्धि लिङ्गमेव समर्थ्यते ॥ ५८९ ॥

nañsnañau vihitau yena sa yogo nāvagāhate |
vatikaraṇaṃ taddhi liṅgameva samarthyate || 589 ||

That rule by which 'nañ' and 'snañ' are prescribed does not enter into the vati-making; for that is justified as a mere indication.

Nan-snanau vihitau yena sa yogo na avagahate vati-karanam: that *rule* by which *'nañ'* and *'snañ'* are *prescribed* does *not enter into* the *vati-making* — the two rules do not touch. Tad dhi lingam eva samarthyate: for *that* is *justified* as a *mere indication* — linga: a sign, a jnapaka; eva: nothing more. So the rule that seemed to conflict is not a rule at all in the operative sense: it is a marker (3.14.547's jnapanartha; 3.14.469's nirdisyanta upadhayah). The conflict of 3.14.587-588 dissolves because one of the parties was never a competitor. This is the last of the samuddesa's many findings that a rule is notation.

3.14.590 unformalized

अभेदेनोपमानस्य भिन्नार्थोपनिपातिता ।
ऊहस्तथोपमाननामङ्गवन्नोपलभ्यते ॥ ५९० ॥

abhedenopamānasya bhinnārthopanipātitā |
ūhastathopamānanāmaṅgavannopalabhyate || 590 ||

By non-difference, the comparans' falling upon a different meaning; and thus the modification of comparans-words, as with 'anga', is not found.

Abhedena upamanasya bhinnartha-upanipatita: by *non-difference*, the *comparans'* *falling upon* a *different meaning* — upanipatita: it lands there, uninvited (3.14.428's the word coming to stand for the man-meaning). Uhas tatha upamana-nam angavan na upalabhyate: *thus* the *uha* of *comparans*-words, as with *'anga'*, is *not found*. Uha: the Mimamsakas' term for the modification of a mantra when it is carried to a new rite — one substitutes the new deity's name for the old. So comparans-words do not undergo uha: one cannot swap the standard out and keep the comparison, as one swaps a deity's name. The comparans is welded to its comparison, and the reason is that it landed there by non-difference and not by substitution.

3.14.591 unformalized

गावेधुके चरौ दृष्टा गोविकर्ताक्षवापयोः ।
पशू रुद्र इव ह्येतावित्येकवचनश्रुतिः ॥ ५९१ ॥

gāvedhuke carau dṛṣṭā govikartākṣavāpayoḥ |
paśū rudra iva hyetāvityekavacanaśrutiḥ || 591 ||

In the gavedhuka oblation, seen in 'go-vikarta' and 'aksa-vapa'; 'these two are like Rudra's beast' — thus the singular is heard.

The ritual illustration of 3.14.590. Gavedhuke carau drsta go-vikarta-aksa-vapayoh: in the *gavedhuka* oblation — the offering of coix-seed to Rudra — *seen* in *'go-vikarta'* (the ox-carver) and *'aksa-vapa'* (the dice-thrower): two persons the rite excludes. Pasu rudra iva hy etav iti eka-vacana-srutih: *'these two are like Rudra's beast'* — and the *singular* is *heard*: 'pasu', not 'pasu' dual. So two men are compared to Rudra's beast, singular, and the number does not shift to match them. The comparans keeps its own number — which is 3.14.590's point exactly: it does not undergo modification to fit its new employment. And Bhartrhari proves it from the srauta ritual, where the Mimamsakas' uha lives.

3.14.592 unformalized

उपमानस्य भेदाच्च बहुषु स्यादञो विधिः ।
कश्यपा इति लोपः स्यात् तथा प्रतिकृतिष्वपि ॥ ५९२ ॥

upamānasya bhedācca bahuṣu syādaño vidhiḥ |
kaśyapā iti lopaḥ syāt tathā pratikṛtiṣvapi || 592 ||

And from difference of the comparans, the rule of 'añ' would apply in the plural; there would be elision in 'kasyapah'; and likewise in the case of images.

Upamanasya bhedac ca bahusu syad ano vidhih: from *difference* of the *comparans*, the rule of *'añ'* would apply *in the plural* — if the standard were plural, the patronymic affix would take its plural form. Kasyapa iti lopah syat: there would be *elision* in *'kasyapah'* — the affix would drop, as it does in patronymic plurals. Tatha pratikrtisv api: and *likewise* in the case of *images* — the pratikrti-affix, for statues and likenesses (a comparison-affix of a kind: an image is *like* what it depicts). So the comparans' number is about to become the question, and it runs through both the patronymics and the images.

3.14.593 unformalized

एवं तु युक्तवद्भावादत्रैकवचनं भवेत् ।
लुम्मनुष्ये तथोक्तं स्याल्लिङ्गस्यैकस्य सिद्धये ॥ ५९३ ॥

evaṃ tu yuktavadbhāvādatraikavacanaṃ bhavet |
lummanuṣye tathoktaṃ syālliṅgasyaikasya siddhaye || 593 ||

But thus, from 'treatment as though joined', the singular would occur here; 'luk in the case of a human' would be so stated, for the establishing of a single gender.

Evam tu yuktavad-bhavad atra eka-vacanam bhavet: from *yuktavad-bhava* — the treatment of a derivative as though it were its analysed phrase (3.14.110, 3.14.318) — the *singular* would occur here. Lum manusye tatha uktam syal lingasya ekasya siddhaye: *'luk in the case of a human'* — the rule for images of men — would be so *stated*, for the *establishing of a single gender*. So the rule about human images exists to fix one gender, and yuktavad-bhava would give the singular. Two provisions, both about number and gender in likenesses, and 3.14.594 asks what governs them.

3.14.594 unformalized

उपमेयेषु भिन्नेषु किञ्चिदेकं प्रवर्तते ।
प्रत्ययस्य विधौ तत्र नित्यं युक्तवदिष्यते ॥ ५९४ ॥

upameyeṣu bhinneṣu kiñcidekaṃ pravartate |
pratyayasya vidhau tatra nityaṃ yuktavadiṣyate || 594 ||

When the comparanda are different, some one thing operates; in the prescription of the affix there, 'as though joined' is always accepted.

Upameyesu bhinnesu kincid ekam pravartate: when the *comparanda* are *different*, *some one thing* operates — many things compared, one standard. Pratyayasya vidhau tatra nityam yuktavad isyate: in the *prescription of the affix* there, *yuktavad* is *always* accepted — nityam: invariably. So where one standard serves many, the affix is prescribed on the yuktavad-bhava, and the compound answers to its analysis. The rule is uniform, and 3.14.595 gives the case where it is not.

3.14.595 unformalized

यदा प्रत्युपमेयं तु तदेकैकमवस्थितम् ।
तदा बाह्यार्थभेदेन तद्धितान्तं प्रचीयते ॥ ५९५ ॥

yadā pratyupameyaṃ tu tadekaikamavasthitam |
tadā bāhyārthabhedena taddhitāntaṃ pracīyate || 595 ||

But when it is settled one by one, for each comparandum — then, by difference of external meaning, the taddhita-ending form is multiplied.

Yada praty-upameyam tu tad ekaikam avasthitam: when *that* is *settled one by one*, *for each* comparandum — a standard apiece, not one for all. Tada bahyartha-bhedena taddhitantam praciyate: *then*, by *difference of external meaning*, the *taddhita*-ending form is *multiplied* — praciyate: is heaped up, pluralized. So distributed standards give a plural derivative, and a shared standard (3.14.594) gives a singular. The difference is not in the words: it is in whether one standard or many is intended (3.14.570's vivaksaya).

3.14.596 unformalized

यथा समूहप्रचये द्विगूनां भिन्नसंख्यता ।
पञ्चपूल्यादिषु तथा सुबन्तप्रचये भवेत् ॥ ५९६ ॥

yathā samūhapracaye dvigūnāṃ bhinnasaṃkhyatā |
pañcapūlyādiṣu tathā subantapracaye bhavet || 596 ||

As in the multiplication of aggregates, dvigus have different numbers — in 'five bundles' and the rest — so it would be in the multiplication of case-ending forms.

Yatha samuha-pracaye dvigunam bhinna-samkhyata: as in the *multiplication of aggregates*, *dvigus* have *different numbers* — the numeral-compounds: 'panca-puli', a five-bundle, and several of them. Panca-pulyadisu tatha sub-anta-pracaye bhavet: *so* it would be in the *multiplication of case-ending forms*. So the precedent is the collective numeral: one can have many five-bundles, and the compound's own number is one while the count of them is another. The comparison-derivative of 3.14.595 does the same. The parallel is well chosen — a dvigu is a collection with its own number, exactly as a comparison-derivative is.

3.14.597 unformalized

प्रचये भिद्यमाने तु संख्या पूलेषु भिद्यते ।
अर्थभेदो लुबन्तेषु नैव कश्चन दृश्यते ॥ ५९७ ॥

pracaye bhidyamāne tu saṃkhyā pūleṣu bhidyate |
arthabhedo lubanteṣu naiva kaścana dṛśyate || 597 ||

But when the multiplication is differentiated, the number in the bundles is differentiated; whereas no difference of meaning at all is seen in luk-ending forms.

Pracaye bhidyamane tu samkhya pulesu bhidyate: when the *multiplication* is *differentiated*, the *number* in the *bundles* is differentiated — many five-bundles means many bundles, and the count runs down into them. Artha-bhedo lubantesu na eva kascana drsyate: whereas *no* difference of meaning *at all* is *seen* in *luk*-ending forms — na eva kascana. So the parallel of 3.14.596 breaks: the dvigu's plurality reaches its members, and the luk-form's does not. The images of Kasyapa (3.14.592) are many and there is no plurality of Kasyapa in them. So the two cases are not alike after all, and Bhartrhari retracts his own analogy in the next verse but one.

3.14.598 unformalized

येषूपमेयवचनः शब्दोऽन्यो न प्रयुज्यते ।
उपमानस्य तत्रान्यैः संख्याया भेद इष्यते ॥ ५९८ ॥

yeṣūpameyavacanaḥ śabdo'nyo na prayujyate |
upamānasya tatrānyaiḥ saṃkhyāyā bheda iṣyate || 598 ||

Those in which no other word expressing the comparandum is employed — there, others accept a difference of number in the comparans.

Yesu upameya-vacanah sabdo 'nyo na prayujyate: those in which *no other word* expressing the *comparandum* is *employed* — the compared thing is not separately named: 'kasyapah', images of Kasyapa, with no word for the images. Upamanasya tatra anyaih samkhyaya bheda isyate: *there*, *others* accept a *difference of number* in the *comparans*. So when the compared thing goes unnamed, the number attaches to the standard by default — and that is why 3.14.592's difficulty arose. The plural of 'kasyapah' looks like a plurality of Kasyapas, because there is no other word for it to be a plurality of.

3.14.599 unformalized

यथा गुडतिलादीनां प्रयोगादेकसंख्यता ।
पाकादेरप्रयोगे तु भिन्नसंख्याभिधीयते ॥ ५९९ ॥

yathā guḍatilādīnāṃ prayogādekasaṃkhyatā |
pākāderaprayoge tu bhinnasaṃkhyābhidhīyate || 599 ||

As from the employment of 'jaggery-sesame' and the rest there is a single number; but when 'cooking' and the rest is not employed, a different number is expressed.

Yatha guda-tiladinam prayogad eka-samkhyata: as from the *employment* of *'guda-tila'* — jaggery-and-sesame, the sweetmeat — and the rest, there is a *single number*: the compound names one thing. Pakader aprayoge tu bhinna-samkhya abhidhiyate: but when *'paka'* — the cooking, the preparation — and the rest is *not employed*, a *different number* is *expressed*. So: name the confection and it is one; leave the preparation-word out and the plural of the ingredients shows through. The number tracks which word is present, not what is in the bowl. Which confirms 3.14.598: an unnamed thing cannot carry a number, and the number goes to whatever is named.

3.14.600 unformalized

यः संबन्धिगतो भेदः स प्रयोगे प्रतीयते ।
संबन्धिनामतो भेद उपमेये न गम्यते ॥ ६०० ॥

yaḥ saṃbandhigato bhedaḥ sa prayoge pratīyate |
saṃbandhināmato bheda upameye na gamyate || 600 ||

The difference which resides in the relatum is cognized in the employment; therefore the difference of the relata is not understood in the comparandum.

Yah sambandhi-gato bhedah sa prayoge pratiyate: the *difference* which *resides in the relatum* is cognized *in the employment* — where the related term is *said*, its difference is available. Sambandhinam ato bheda upameye na gamyate: *therefore* the *difference of the relata* is *not understood* in the *comparandum*. So an unsaid relatum's difference does not transfer: the images' plurality is not Kasyapa's, and it does not become his by his being the only word present. This blocks the slide 3.14.598 identified, and it does so by a principle: differences are cognized where they are employed, and nowhere else (3.14.501's aprayoge na vidyate).

3.14.601 unformalized

तस्मात् सामान्यशब्दत्वप्रसङ्गविनिवृत्तये ।
उपमेयगतो भेद उपमानेषु दृश्यते ॥ ६०१ ॥

tasmāt sāmānyaśabdatvaprasaṅgavinivṛttaye |
upameyagato bheda upamāneṣu dṛśyate || 601 ||

Therefore, to avoid the consequence of being a general word — the difference residing in the comparandum is seen in the comparanda.

Tasmat samanya-sabdatva-prasanga-vinivrttaye: *therefore*, to *avoid* the *consequence* of being a *general word* — if the derivative carried no difference at all, it would be a mere general and would not pick out these images (3.14.425's samanya-vacanah katham). Upameya-gato bhedah upamanesu drsyate: the *difference* residing in the *comparandum* is *seen* in the *comparanda*. So the images' difference is *seen in* the standard — displaced there, because the standard is the only word present. This is the resolution: the plural of 'kasyapah' is the images' plurality, seen in Kasyapa's word, and no Kasyapas are multiplied. Drsyate: it is *seen* there, which is not the same as being there.

3.14.602 unformalized

उपमानं समस्तानामभिन्नं श्रूयते क्वचित् ।
भिन्नानामुपमेयानामेकैकं वोपमीयते ॥ ६०२ ॥

upamānaṃ samastānāmabhinnaṃ śrūyate kvacit |
bhinnānāmupameyānāmekaikaṃ vopamīyate || 602 ||

Somewhere the comparans of all of them together is heard undivided; or else each one of the differentiated comparanda is compared.

Upamanam samastanam abhinnam sruyate kvacit: *somewhere* the *comparans* of *all of them together* is *heard undivided* — one standard, for the whole company, and it stays single. Bhinnanam upameyanam ekaikam va upamiyate: *or else* *each one* of the *differentiated comparanda* is *compared*. So two readings, always available: the collective and the distributive. 3.14.594-595 gave the rule for each; here they are set side by side as the two things a comparison can be doing. And 3.14.603-605 give three examples that make the difference vivid.

3.14.603 unformalized

यथा गरुड इत्येतद् व्यूहापेक्षं प्रयुज्यते ।
एकेन यत्र सादृश्यं वैनतेयेन हस्तिनाम् ॥ ६०३ ॥

yathā garuḍa ityetad vyūhāpekṣaṃ prayujyate |
ekena yatra sādṛśyaṃ vainateyena hastinām || 603 ||

As 'garuda' is employed with reference to a battle-array, where the elephants' similarity is with one Vainateya.

Yatha garuda ity etad vyuha-apeksam prayujyate: as *'garuda'* is employed with reference to a *vyuha* — the garuda-vyuha, the eagle-formation of the epic battlefields, in which an army is drawn up in the shape of the bird. Ekena yatra sadrsyam vainateyena hastinam: where the *similarity* of the *elephants* is with *one* *Vainateya* — Garuda, son of Vinata. So: many elephants, one eagle, and the likeness is of the whole formation to the single bird. This is 3.14.602's first reading — samastanam abhinnam, the undivided standard for all together — and the example is perfect: no elephant resembles Garuda, and the army does.

3.14.604 unformalized

एकस्यापि प्रतीयते भिन्ना प्रतिकृतिः सह ।
काश्यपस्येति तेनायं प्रत्येकमवतिष्ठते ॥ ६०४ ॥

ekasyāpi pratīyate bhinnā pratikṛtiḥ saha |
kāśyapasyeti tenāyaṃ pratyekamavatiṣṭhate || 604 ||

Even of one, differentiated images are cognized together — 'of Kasyapa'; therefore this stands distributively.

Ekasya api pratiyate bhinna pratikrtih saha: even of *one*, *differentiated images* are cognized *together* — one Kasyapa, many statues, and one hears them all at once in 'kasyapah'. Tena ayam pratyekam avatisthate: *therefore* this stands *distributively* — pratyekam: one by one. So the images-case is 3.14.602's second reading: the plural distributes over the statues, though the man is one. Set against 3.14.603's garuda — one bird, many elephants, collective — the pair shows that the same grammatical shape does both, and only vivaksa tells which.

3.14.605 unformalized

मेघाः शैल इवेत्युक्ते समस्तानां प्रतीयते ।
सादृश्यं गिरिणैकेन प्रत्येकं तेन भिद्यते ॥ ६०५ ॥

meghāḥ śaila ivetyukte samastānāṃ pratīyate |
sādṛśyaṃ giriṇaikena pratyekaṃ tena bhidyate || 605 ||

When 'the clouds are like a mountain' is said, the similarity of all of them together with one mountain is cognized; therefore it is differentiated distributively.

Meghah saila iva ity ukte samastanam pratiyate sadrsyam girina ekena: when *'the clouds are like a mountain'* is said, the *similarity* of *all together* with *one mountain* is cognized. Pratyekam tena bhidyate: *therefore* it is *differentiated* *distributively*. So this one does both: the massed clouds resemble the single mountain collectively, and the resemblance is then distributed over them. It is 3.14.603 and 3.14.604 at once. The example is the best of the three because it is what one actually sees — a bank of cloud on the horizon, one great shape, and each cloud in it — and it shows the collective and the distributive as two moments of one cognition rather than two doctrines.

3.14.606 unformalized

छापेक्षा तद्विषयता विधेयत्वान्न गम्यते ।
काकतालीयमित्यत्र प्रसिद्धं ह्युपलक्षणम् ॥ ६०६ ॥

chāpekṣā tadviṣayatā vidheyatvānna gamyate |
kākatālīyamityatra prasiddhaṃ hyupalakṣaṇam || 606 ||

The state of having that domain, which depends on 'cha', is not understood, since it is what is to be prescribed; for in 'the crow and the palm-fruit' the indication is well-known.

Cha-apeksa tad-visayata vidheyatvan na gamyate: the state of *having that domain*, which *depends on 'cha'* — the affix — is *not understood*, since it is *what is to be prescribed*: one cannot use the affix's domain to determine the affix, because the domain is what the rule is for (the circularity of 3.14.78's uhyau). Kaka-taliyam ity atra prasiddham hy upalaksanam: for in *'kakataliya'* — the crow-and-palm-fruit, the crow that alights just as the fruit falls, Sanskrit's byword for coincidence — the *indication* is *well-known*. So the compound is understood by everyone and the affix's domain is not thereby fixed: usage runs ahead of the rule (3.14.560's the science follows). The example is chosen for its fame, which is the point: what everyone knows does not tell the grammarian what to write.

3.14.607 unformalized

राजाश्वादिश्च विषयः स्यादन्यो वेत्यनिश्चितम् ।
तेन छस्य विधानात् प्राग् व्यपदेशो न विद्यते ॥ ६०७ ॥

rājāśvādiśca viṣayaḥ syādanyo vetyaniścitam |
tena chasya vidhānāt prāg vyapadeśo na vidyate || 607 ||

And whether the domain is 'rajasva' and the rest or something else is undetermined; therefore, prior to the prescription of 'cha', there is no designation.

Rajasvadis ca visayah syad anyo va ity aniscitam: whether the *domain* is *'rajasva'* and the rest, *or* something *else*, is *undetermined* — aniscita: unsettled. Tena chasya vidhanat prag vyapadeso na vidyate: *therefore*, *prior to* the *prescription of 'cha'*, there is *no designation* — vyapadesa: the technical naming (3.14.135's word). So the domain cannot be named before the rule that creates it, and the rule cannot be framed before the domain is named. The circle is the same one 3.14.78 exhibited — base and affix each conjectured from the other — and here it closes on a single affix. The Vrtti-samuddesa's method-argument and its rule-technical argument have arrived at the same place.

3.14.608 unformalized

द्वयोरिवार्थयोरत्र निमित्तत्वं प्रतीयते ।
एकेनावयवो युक्तः प्रत्ययोऽन्येन युज्यते ॥ ६०८ ॥

dvayorivārthayoratra nimittatvaṃ pratīyate |
ekenāvayavo yuktaḥ pratyayo'nyena yujyate || 608 ||

Here the causality of two 'iva'-meanings is cognized; the member is joined with one, the affix with the other.

The anatomy of 'kakataliya'. Dvayor iva-arthayor atra nimittatvam pratiyate: the *causality* of *two* *'iva'-meanings* is cognized here — two likenesses, not one. Ekena avayavo yuktah pratyayo 'nyena yujyate: the *member* is joined with *one*, the *affix* with the *other*. So the compound carries a double comparison: one of its members answers to one likeness and the affix to another. This is unlike anything else in the samuddesa — a compound with two comparisons inside it — and 3.14.609 unpacks them.

3.14.609 unformalized

चैत्रस्य तत्रागमनं काकस्यागमनं यथा ।
दस्योरभिनिपातस्तु तालस्य पतनं यथा ॥ ६०९ ॥

caitrasya tatrāgamanaṃ kākasyāgamanaṃ yathā |
dasyorabhinipātastu tālasya patanaṃ yathā || 609 ||

Caitra's arriving there is like the crow's arriving; but the bandit's falling upon him is like the palm-fruit's falling.

The two likenesses of 3.14.608, and the image is exact. Caitrasya tatra agamanam kakasya agamanam yatha: *Caitra's arriving* there is *like* the *crow's arriving* — he came for his own reasons, as the crow did. Dasyor abhinipatas tu talasya patanam yatha: but the *bandit's falling upon him* is like the *palm-fruit's falling* — the fruit was ripe and would have dropped anyway. So neither event is for the other's sake, and their meeting is the whole of the misfortune. 'Kakataliya' compounds two independent likenesses and means their coincidence — which no member of it says.

3.14.610 unformalized

संनिपाते तयोर्यान्या क्रिया तत्रोपजायते ।
वधादिरुपमेयेऽर्थे तया छविधिरिष्यते ॥ ६१० ॥

saṃnipāte tayoryānyā kriyā tatropajāyate |
vadhādirupameye'rthe tayā chavidhiriṣyate || 610 ||

In their conjunction, that other action which arises there — killing and the rest, in the comparandum-meaning — by that the prescription of 'cha' is accepted.

Sannipate tayor ya anya kriya tatra upajayate: in their *conjunction*, that *other* action which *arises* there — sannipata: the falling-together; and the third action, which neither party performed. Vadhadir upameye 'rthe: *killing* and the rest, in the *comparandum-meaning*. Taya cha-vidhir isyate: *by that* the *prescription of 'cha'* is *accepted*. So what licenses the affix is the action that came out of the meeting and belonged to neither — the killing that nobody did. Upajayate: it *arises*. This is the finest thing in the last stretch: the compound is licensed by an emergent event, and a coincidence is precisely an event with no agent.

3.14.611 unformalized

क्रियायां समवेतायां द्रव्यशब्दोऽवतिष्ठते ।
पातागमनयोः काकतालशब्दौ तथा स्थितौ ॥ ६११ ॥

kriyāyāṃ samavetāyāṃ dravyaśabdo'vatiṣṭhate |
pātāgamanayoḥ kākatālaśabdau tathā sthitau || 611 ||

The substance-word stands in the inherent action; thus the words 'crow' and 'palm' stand for the arriving and the falling.

Kriyayam samavetayam dravya-sabdo 'vatisthate: the *substance-word* stands in the *inherent action* — the noun lodges in the action that inheres in what it names. Pata-agamanayoh kaka-tala-sabdau tatha sthitau: the words *'kaka'* and *'tala'* stand for the *falling* and the *arriving*. So 'crow' in the compound means the crow's coming and 'palm' means the fruit's dropping: two nouns, and each has become an event. This is 3.14.529's madhura and matr all over again — the word names the action at its seat, not the seat — and it is why 'kakataliya' can compound two happenings out of two things.

3.14.612 unformalized

यदन्वाख्यायकं वाक्यं तदेवं परिकल्प्यते ।
प्रयोगवाक्यं यल्लोके तदेवं न प्रयुज्यते ॥ ६१२ ॥

yadanvākhyāyakaṃ vākyaṃ tadevaṃ parikalpyate |
prayogavākyaṃ yalloke tadevaṃ na prayujyate || 612 ||

The sentence which is expository is constructed in this way; the sentence of usage in the world is not employed in this way.

Yad anvakhyayakam vakyam tad evam parikalpyate: the sentence which is *expository* — anvakhyayaka, the grammarian's paraphrase (3.14.71's anvakhyanaya) — is *constructed* in this way: parikalpyate, built. Prayoga-vakyam yal loke tad evam na prayujyate: the *sentence of usage* in the *world* is *not employed* in this way. So everything 3.14.608-611 has just unpacked belongs to the exposition and to nothing anyone says. Nobody utters 'as the crow's arriving and the palm-fruit's falling'; they say 'kakataliyam', and they mean coincidence. The analysis is a construction and the word is a word — 3.14.49's atyanta-bheda, held to the very end.

3.14.613 unformalized

ययोरतर्किता प्राप्तिर्दृश्यते काकतालवत् ।
तयोः समासप्रकृतेर्वृत्तिरभ्युपगम्यते ॥ ६१३ ॥

yayoratarkitā prāptirdṛśyate kākatālavat |
tayoḥ samāsaprakṛtervṛttirabhyupagamyate || 613 ||

Of those two whose meeting is seen to be unreasoned, like the crow and the palm-fruit — the derivation of the compound-base is accepted.

Yayor atarkita praptir drsyate kaka-talavat: of those two whose *meeting* is seen to be *atarkita* — *unreasoned*, not to be argued out, arriving without a why — *like the crow and the palm-fruit*. Tayoh samasa-prakrter vrttir abhyupagamyate: of *those* the *derivation* of the *compound-base* is *accepted*. So the rule's condition is *atarkita prapti*: an unreasoned meeting. This is one of the loveliest phrases in the samuddesa and one of the most precise: a coincidence is exactly a conjunction for which no reasoning is available, and Sanskrit grammar has an affix for it. Drsyate: it is *seen* to be unreasoned — one sees the absence of a reason, and that seeing is what the compound reports.

3.14.614 unformalized

काकस्य तालेन यथा वधो यस्य तु दस्युना ।
तत्र चित्रीकृतेऽन्यस्मिन्नुपमेये छ इष्यते ॥ ६१४ ॥

kākasya tālena yathā vadho yasya tu dasyunā |
tatra citrīkṛte'nyasminnupameye cha iṣyate || 614 ||

As the crow's killing by the palm-fruit — but whose killing is by a bandit; there, when the other comparandum is made striking, 'cha' is accepted.

Kakasya talena yatha vadho yasya tu dasyuna: as the *crow's* killing *by the palm-fruit* — but *whose* killing is *by a bandit*: Caitra's. Tatra citrikrte 'nyasminn upameye cha isyate: *there*, when the *other* comparandum is *made striking*, *'cha'* is *accepted*. Citrikrta: made vivid, made a picture of, rendered arresting. So the affix comes when the compared event has been made *striking* — and that is an aesthetic condition, not a semantic one. Bhartrhari has been describing a figure of speech all along, and here he says what figures are for: 'kakataliya' is used because it makes the thing vivid, and the grammar accommodates the vividness.

3.14.615 unformalized

चञ्चत्प्रकारश्चञ्चत्को बृहत्क इति चापरे ।
मणिमण्डूकखद्योतान् सादृश्येन प्रचक्षते ॥ ६१५ ॥

cañcatprakāraścañcatko bṛhatka iti cāpare |
maṇimaṇḍūkakhadyotān sādṛśyena pracakṣate || 615 ||

And others: 'cañcatka', of the flickering kind, and 'brhatka' — they describe the gem, the frog and the firefly by similarity.

Cancat-prakaras cancatko brhatka iti ca apare: *'cancatka'*, of the *flickering kind*, and *'brhatka'* — the great-ish one — and *others* say. Mani-manduka-khadyotan sadrsyena pracaksate: they describe the *gem*, the *frog* and the *firefly* *by similarity*. So three small pulsing things, named by a mode-affix, and each of them likened to something. The set is unexpected and Bhartrhari has chosen it: a jewel, a frog, a firefly — things that throb. 3.14.616 says what each is likened to, and the verse is a small poem.

3.14.616 unformalized

तत्रोन्मेषनिमेषाभ्यां खद्योत उपमीयते ।
श्वासप्रबन्धैर्मण्डूकः स्पन्दमानप्रभोमणिः ॥ ६१६ ॥

tatronmeṣanimeṣābhyāṃ khadyota upamīyate |
śvāsaprabandhairmaṇḍūkaḥ spandamānaprabhomaṇiḥ || 616 ||

There the firefly is compared by its opening and closing; the frog by its succession of breaths; the gem, its light throbbing.

One of the most beautiful verses in the kanda, and it is doing analytic work. Tatra unmesa-nimesabhyam khadyota upamiyate: the *firefly* is compared by its *opening and closing* — unmesa-nimesa: the eye's opening and shutting, the standard Sanskrit measure of an instant. Svasa-prabandhair mandukah: the *frog* by its *succession of breaths* — prabandha: an unbroken series; the throat going in and out. Spandamana-prabho manih: the *gem*, its *light throbbing* — spandamana: pulsing, quivering. So the three are alike in *pulsing*, and each is compared by its own kind of pulse. What the mode-affix catches is a rhythm — and 3.14.321's pravrtti said that a thing's general character *is* a rhythm: manifestation, disappearance, persistence. The firefly is that doctrine made visible.

3.14.617 unformalized

प्रविकासिप्रभोऽल्पोऽपि महान् य उपलभ्यते ।
बृहत्क इति तत्रैव मणौ शब्दः प्रयुज्यते ॥ ६१७ ॥

pravikāsiprabho'lpo'pi mahān ya upalabhyate |
bṛhatka iti tatraiva maṇau śabdaḥ prayujyate || 617 ||

That which, though small, is apprehended as great, its light expanding — there, in that very gem, the word 'brhatka' is employed.

Pravikasi-prabho 'lpo 'pi mahan ya upalabhyate: that which, *though small*, is *apprehended as great*, its *light expanding* — pravikasin: unfolding, blooming outward. A little stone that throws a great blaze. Brhatka iti tatra eva manau sabdah prayujyate: *there*, in that very *gem*, the word *'brhatka'* is *employed*. So the affix says: great, in the way this small thing is great. It records the discrepancy between the size and the apprehension — and it is the apprehension that the word takes (3.14.569's the object comes about in conformity with the cognition). A jewel is brhatka because of how one sees it.

3.14.618 unformalized

सादृश्यमेव सर्वत्र प्रकारः कैश्चिदिष्यते ।
भेदेऽपि तु प्रकाराख्या कैश्चिदभ्युपगम्यते ॥ ६१८ ॥

sādṛśyameva sarvatra prakāraḥ kaiścidiṣyate |
bhede'pi tu prakārākhyā kaiścidabhyupagamyate || 618 ||

By some, 'mode' everywhere is accepted as just similarity; by others, the designation 'mode' is accepted even in difference.

Sadrsyam eva sarvatra prakarah kaiscid isyate: by *some*, *'prakara'* — mode, kind, manner — *everywhere* is accepted as *just similarity*: to be of a kind is to be like. Bhede 'pi tu prakarakhya kaiscid abhyupagamyate: by *others*, the designation *'prakara'* is accepted *even in difference*: a mode may be a mode without any likeness. So the last doctrinal question of the samuddesa: is a kind a resemblance? Two answers, both held, neither adjudicated — the work's manner from first to last (3.14.97's the upayas are unsettled among persons).

3.14.619 unformalized

प्रकारवचनः कश्चित् प्रकारवति संस्थितः ।
प्रकारमात्रे वर्तित्वा कश्चिद् तद् व्यतिवर्तते ॥ ६१९ ॥

prakāravacanaḥ kaścit prakāravati saṃsthitaḥ |
prakāramātre vartitvā kaścid tad vyativartate || 619 ||

Some word expressing a mode stands in what possesses the mode; some, having operated in the mode alone, passes beyond it.

Prakara-vacanah kascit prakaravati samsthitah: *some* word expressing a *mode* stands in what *possesses* the mode — it lodges in the thing that has the manner. Prakara-matre vartitva kascid tad vyativartate: *some*, having operated in the *mode alone*, *passes beyond* it — vyativartate: goes past, overshoots. So mode-words do two things: some settle on the possessor, some touch the mode and travel on. The pair is 3.14.354-355's two permutations once more, and it is the last time the samuddesa runs its most characteristic move — set two readings side by side, change one word, and let the reader see that nothing decides.

3.14.620-621 unformalized

सादृश्यग्रहणं सूत्रे सदृशस्योपलक्षणम् ।
तुल्ययोरव्ययीभावे सहशब्दोऽभिधायकः ॥ ६२० ॥
वीप्सासादृश्ययोर्वृत्तिर्या यथार्थाभिधायिनः ।
स चायमव्ययीभावे भेदो भेदेन दर्शितः ॥ ६२१ ॥

sādṛśyagrahaṇaṃ sūtre sadṛśasyopalakṣaṇam |
tulyayoravyayībhāve sahaśabdo'bhidhāyakaḥ || 620 ||
vīpsāsādṛśyayorvṛttiryā yathārthābhidhāyinaḥ |
sa cāyamavyayībhāve bhedo bhedena darśitaḥ || 621 ||

(620) The taking-up of 'similarity' in the sutra is an indication of what is similar; in the avyayibhava of two equals, the word 'saha' is the expresser. (621) That derivation of what expresses the sense of 'as' — in repetition and in similarity: and this difference in the avyayibhava has been shown by way of difference.

(620): sadrsya-grahanam sutre sadrsasya upalaksanam — the taking-up of *'sadrsya'* in the *sutra* is an *indication* of what is *similar*: the rule's word points at the similar thing and does not describe similarity (3.14.170's upalaksana). Tulyayor avyayibhave saha-sabdo 'bhidhayakah: in the *avyayibhava* of *two equals*, the word *'saha'* is the *expresser* — a vacaka, where 'sadrsya' was only a pointer. (621): vipsa-sadrsyayor vrttir ya yatharthabhidhayinah — the *derivation* of what expresses the sense of *'yatha'*, in *repetition* (vipsa: 'gramam gramam', village by village) and in *similarity*. Sa ca ayam avyayibhave bhedo bhedena darsitah: and this *difference* in the avyayibhava *has been shown by way of difference* — darsitah, exhibited (3.14.573's the sutra shows the difference by way of difference). So the rule exhibits as a real distinction what is a distinction of the exposition. The samuddesa's last word about rules is the one it began with.

3.14.622 unformalized

सादृश्यं योग्यता कैश्चिदनावभ्युपगम्यते ।
यत् तु मूर्तिगतं साम्यं तत् सहेनाभिधीयते ॥ ६२२ ॥

sādṛśyaṃ yogyatā kaiścidanāvabhyupagamyate |
yat tu mūrtigataṃ sāmyaṃ tat sahenābhidhīyate || 622 ||

By some, similarity in 'anu' is accepted as fitness; but that similarity which resides in bodily form is expressed by 'saha'.

Sadrsyam yogyata kaiscid anav abhyupagamyate: by *some*, *similarity* in *'anu'* is accepted as *yogyata* — *fitness*, aptness, the capacity to go together (2.317's yogyata, one of the sentence's meaning-conditions). So 'anu' marks not likeness but suitability. Yat tu murti-gatam samyam tat sahena abhidhiyate: but that *similarity* which *resides in bodily form* is expressed by *'saha'* — murti: the shaped body (3.14.485's murti-bheda, which jati- and guna-words did not intend). So the division: 'anu' takes fitness, 'saha' takes bodily likeness. Two particles, two kinds of togetherness, and the difference is whether shape is in question.

3.14.623 unformalized

इत्थंभावेऽपि सादृश्यं बुद्ध्यवस्थानिबन्धनम् ।
ग्रहणे भेदमात्रस्य तत्रान्यैवाभिधीयते ॥ ६२३ ॥

itthaṃbhāve'pi sādṛśyaṃ buddhyavasthānibandhanam |
grahaṇe bhedamātrasya tatrānyaivābhidhīyate || 623 ||

Even in 'being-thus', similarity is grounded in a state of the intellect; where mere difference is grasped, another thing altogether is expressed.

The work's thesis, sounded once more before the end. Itthambhave 'pi sadrsyam buddhy-avastha-nibandhanam: even in *itthambhava* — being-thus, the condition of being in a certain way — *similarity* is *grounded in a state of the intellect*: buddhy-avastha-nibandhana, the exact words of 3.14.567 and 3.14.570. Grahane bheda-matrasya tatra anya eva abhidhiyate: where *mere difference* is *grasped*, *another thing altogether* is *expressed*. So even here, at the last technical question, the answer is the one the samuddesa reached at its summit: the likeness is in the mind's condition, and where the mind grasps only difference, something else is being said. Api — *even* — carries it: there is no case left where this does not hold.

3.14.624 unformalized

गौर्वाहीक इति द्वित्वे सादृश्यं प्रत्युदाहृतम् ।
शुक्लादौ सति निष्पन्ने बाहीकोन द्विरुच्यते ॥ ६२४ ॥

gaurvāhīka iti dvitve sādṛśyaṃ pratyudāhṛtam |
śuklādau sati niṣpanne bāhīkona dvirucyate || 624 ||

'The Vahika is an ox' — in the duality, similarity is cited as a counter-instance; when 'white' and the rest is present and accomplished, it is said twice with 'Vahika'.

The last verse of the Vakyapadiya, and Bhartrhari ends on the metaphor every Sanskrit grammarian ends on. Gaur vahika iti dvitve sadrsyam pratyudahrtam: *'gaur vahikah'* — *the Vahika is an ox* — the Bhasya's own example, the countryman called an ox for his dullness. In the *duality*, *similarity* is cited as a *pratyudaharana* — a *counter*-instance: the case brought to show what the rule does *not* cover. Suklad au sati nispanne bahikena dvir ucyate: when *'white'* and the rest is *present and accomplished*, it is said *twice* with *'Vahika'*. So: 'the white Vahika is an ox' — and 'Vahika' is said twice over, once for the man and once inside the metaphor, because the qualifier has fixed him and the ox is still there. The two do not merge and they do not part. It is the right ending. The work opened at 1.1 with the Word-Brahman, without beginning or end, undivided, appearing as many; and it closes on a man who is an ox, where a metaphor is a duality that will not resolve into one thing or two — 3.14.367's vyatikirna, intermingled past separating. And the last thing the samuddesa does is call it a counter-instance: the case the rule does not reach. Bhartrhari has spent six hundred verses showing that the science stands very far off (3.14.74), that the divisions are the intellect's (3.14.570), that the upayas must not be taken for true (3.14.76) — and he ends by pointing at an ordinary insult that his grammar cannot derive, and stopping there.