Vākya-kāṇḍa

2.1-2 · What is a Sentence? The Eight Views unformalized

आख्यातशब्दः सङ्घातो जातिः सङ्घातवर्तिनी ।
एको'नवयवः शब्दः क्रमो बुद्ध्यनुसंहृतिः ॥ १ ॥
पदमाद्यं पृथक् सर्वं पदं साकाङ्क्षमित्यपि ।
वाक्यं प्रति मतिर्भिन्ना बहुधा न्यायवादिनाम् ॥ २ ॥

ākhyātaśabdaḥ saṅghāto jātiḥ saṅghātavartinī |
eko'navayavaḥ śabdaḥ kramo buddhyanusaṃhṛtiḥ || 1 ||
padamādyaṃ pṛthak sarvaṃ padaṃ sākāṅkṣamityapi |
vākyaṃ prati matirbhinnā bahudhā nyāyavādinām || 2 ||

(1) The verb-word; a collection; the universal residing in the collection; the single partless word; sequence; the gathering-together in cognition; (2) the first word; each word severally, as requiring the others — thus is opinion about the sentence divided manifold among the proponents of theory.

The kanda opens not with a thesis but with a catalogue: eight answers to 'what is a vakya?', compressed into two verses that the whole book will spend itself unpacking. In order: (i) akhyata-sabdah — the sentence is the verb, everything else being its satellites (a Mimamsa-adjacent view: the action-word is where sentencehood lives); (ii) sanghatah — the sheer collection of words; (iii) jatih sanghata-vartini — the universal residing in the collection: sentencehood as a type instantiated by word-strings; (iv) eko 'navayavah sabdah — the one partless word: the sentence-sphota, the view this kanda will defend; (v) kramah — the sequence itself, the ordering as the sentence; (vi) buddhy-anusamhrtih — the gathering-together in cognition: the sentence as a mental act of synthesis; (vii) padam adyam — the first word, which sets the frame the rest merely fills; (viii) prthak sarvam padam sakanksam — every word severally, each 'requiring' the others: the sentence as mutual expectancy distributed over its words. The list is engineered as a spectrum: it runs from views that locate the sentence in a privileged part (i, vii), through views that locate it in the parts collectively (ii, v, viii), through type-theoretic ascent (iii), to views that locate it beyond the parts altogether (iv, vi). Positions (iv) and (vi) are the partless-whole family, divided over whether the unity is in the word or in the mind — and their eventual convergence (the sentence-sphota grasped in pratibha) is where the kanda is headed. The doxographic opening is itself doctrine: that eight serious answers exist is the first datum about how hard the question is.
contested: the enumeration and even the count of views varies among readers — some collapse (v) and (vi), some split (viii) in two; the karika gives no numerals, and the traditional count of eight follows the commentators

2.3 unformalized

निघातादिव्यवस्थार्थं शास्त्रे यत् परिभाषितम् ।
साकाङ्क्षावयवं तेन न सर्वं तुल्यलक्षणम् ॥ ३ ॥

nighātādivyavasthārthaṃ śāstre yat paribhāṣitam |
sākāṅkṣāvayavaṃ tena na sarvaṃ tulyalakṣaṇam || 3 ||

What is technically defined in the science for the sake of settling accent-suppression and the like takes 'having parts that require one another' — hence not everything bears the same definition.

A methodological caution before the battle: grammar's own working definition of 'sentence' settles nothing metaphysical. Panini's system needs a sentence-unit to state rules — nighata, the suppression of accent (a finite verb loses its accent when preceded in the same sentence, per Astadhyayi 8.1.28), is the standard case — and for such purposes the sastra defines (paribhasitam) the vakya as sakanksa-avayavam, a thing whose parts mutually require. That is view (viii) of the catalogue, adopted *for bookkeeping*. Bhartrhari's point: a definition framed vyavastha-artham, for regulating specific operations, is fitted to those operations — na sarvam tulya-laksanam, not everything is captured by one mark. The rule-writer's sentence and the philosopher's sentence need not coincide, and citing Panini's practice to settle the ontology would confuse an instrument with a discovery. The verse generalizes 1.28's move (deflating a metaphysical dispute by distinguishing the purposes of definitions) and licenses what follows: the kanda can canvass sentencehood freely because the grammar's operational definition holds only its own ground.

2.4 unformalized

साकाङ्क्षावयवं भेदे पारानाकाङ्क्षशब्दकम् ।
कर्मप्रधानं गुणवदेकार्थं वाक्यमिष्यते ॥ ४ ॥

sākāṅkṣāvayavaṃ bhede pārānākāṅkṣaśabdakam |
karmapradhānaṃ guṇavadekārthaṃ vākyamiṣyate || 4 ||

The sentence is held to be: having mutually-requiring parts, while — under division — its words have no requirement toward what lies outside; principally about the action, with subordinate elements; single in meaning.

The working consensus-definition, assembled from the tradition's marks. Four conditions: (i) sakanksa-avayavam — internally, the parts require one another (the expectancy of view viii); (ii) para-anakanksa-sabdakam bhede — externally, when the sentence is marked off, its words reach for nothing beyond it: the sentence is where expectancy *closes*, a complete circuit of requirement (this pair of marks — inner akanksa, outer nirakanksata — is the enduring definition Indian theory bequeathed: a sentence is a self-satisfying unit of expectancy); (iii) karma-pradhanam gunavat — action-principal, with subordinates: the verb names the artha around which the rest organizes as gunas (view i, demoted from definition to structure); (iv) ekartham — *single in meaning*. The last mark is the loaded one: it can be read innocently (one topic, one purpose) or with full metaphysical weight — the sentence has *one* meaning, an undivided artha of which the word-meanings are analytical shadows. The kanda will argue for the weighted reading (the vakyartha as pratibha, indivisible). The definition is thus a Trojan structure: three marks any theorist signs, and a fourth that, pressed, yields the sentence-holism the partitive theorists never meant to concede.
contested: ekartham is the crux by design — 'having a single purpose', 'about a single thing', and 'possessing one (undivided) meaning' are all available, and which one takes determines whether 2.4 is a neutral consensus or already the siddhanta in disguise

2.5 unformalized

सम्बोधनपदं यच्च तत् क्रियाया विशेषकम् ।
व्रजानि देवदत्तेति निघातोऽत्र तथा सति ॥ ५ ॥

sambodhanapadaṃ yacca tat kriyāyā viśeṣakam |
vrajāni devadatteti nighāto'tra tathā sati || 5 ||

And a word of address is a qualifier of the action: 'Devadatta! go' — the accent-suppression here is thus in order.

The definition immediately earning its keep on a scholastic test case. The vocative — sambodhana-padam, 'Devadatta!' — looks grammatically detached: an address is not an argument of the verb, so is it *in* the sentence? The answer: it is kriyaya visesakam, a qualifier of the action — the address fixes *whose* going is enjoined, orienting the injunction as surely as any karaka. Consequence: in vrajani devadatta ('let me go, Devadatta' / 'Devadatta, go'), the verb's nighata, its accent-suppression, is tatha sati in order — the vocative counts as sentence-mate (the rule of 8.1 applies within one vakya), because qualification of the action is exactly the sentence-membership the definition requires. The verse is deliberately minor: watch the definition of 2.4 adjudicate a real Paninian detail, and note the direction of reasoning — semantic integration (does it qualify the kriya?) decides formal treatment (does nighata apply?), not the reverse. Meaning-structure is already sovereign over form, in miniature, before the great argument for it begins.
contested: vrajani is the first-person imperative ('let me go'), which sits oddly with a second-person address; the example may be quoted from a ritual context (announcing one's going to Devadatta), and readers divide over whether the oddity is the point (address qualifies even a first-person act) or an artifact of the traditional example

2.6 unformalized

यथानेकमपि क्त्वान्तं तिङन्तस्य विशेषकम् ।
तथा तिङ्न्तं तत्राहुस्तिङन्तस्य विशेषकम् ॥ ६ ॥

yathānekamapi ktvāntaṃ tiṅantasya viśeṣakam |
tathā tiṅntaṃ tatrāhustiṅantasya viśeṣakam || 6 ||

Just as even several gerund-clauses are qualifiers of the finite verb, so, they say, a finite verb may there be the qualifier of another finite verb.

Stretching sentence-unity across clause-boundaries. That ktvanta clauses — 'having bathed, having eaten, he goes' — qualify the finite verb, anekam api, however many they are, is uncontroversial: one sentence, one principal action, subordinate members in train (2.4's third mark). The verse then reports the bolder extension (ahuh, 'they say'): a tinanta may qualify a tinanta — one *finite* verb subordinated to another, as in correlative constructions ('he cooks, he eats' as 'when he cooks, he eats'). This dissolves the surface criterion for counting sentences: two finite verbs need not mean two vakyas, if one action stands as visesaka to the other; principality, not finiteness, individuates. The consequence runs deep — sentence-boundaries follow the architecture of *meaning-dependence*, not verb-morphology, so the vakya is a semantic unit through and through, its edges drawn where expectancy closes (2.4) rather than where inflection changes. The partitive theorist who wanted the sentence assembled from formally-marked pieces has already, by 2.6, lost the ability to say where one assembly ends.

2.7 unformalized

यथैक एव सर्वार्थप्रत्ययः प्रविभज्यते ।
दृश्यभेदानुकारेण वाक्यार्थानुगमस्तथा ॥ ७ ॥

yathaika eva sarvārthapratyayaḥ pravibhajyate |
dṛśyabhedānukāreṇa vākyārthānugamastathā || 7 ||

Just as a single cognition comprehending all its objects is divided up in imitation of the differences of the seen, so is the construing of sentence-meaning.

The kanda's positive doctrine sounded for the first time, on the analogy of the visual field. Take one perceptual act — a single glance at a room. The cognition is *one*: eka eva sarvartha-pratyayah, one awareness comprehending all its objects together. Analysis then divides it — 'I saw a jar, a mat, a lamp' — but the division is drsya-bheda-anukarena, in *imitation of* the differences among the things seen: the plurality belongs to the objects and is retrojected into the act, which was itself undivided. Tatha vakyartha-anugamah: so with the construing of sentence-meaning. Understanding a sentence is one cognitive seizure (the anusamhrti of 2.1's sixth view, the pratibha to come); the 'word-meanings' we report are divisions painted onto that unity in imitation of the divisions of its topic. The analogy is chosen with precision — no one thinks a glance is assembled out of jar-perception plus mat-perception; the unity is primitive, the itemization retrospective. Bhartrhari's claim is that sentence-understanding has exactly this structure, and 1.90's phenomenology (part-seizures as scaffolding toward one grasp) already supplied the temporal side of the picture.

2.8-9 unformalized

चित्रस्यैकस्वरूपस्य यथा भेदनिदर्शनैः ।
नीलादिभिः समाख्यानं क्रियते भिन्नलक्षणैः ॥ ८ ॥
तथैवैकस्य वाक्यस्य निराकाङ्क्षस्य सर्वतः ।
शब्दान्तरैः समाख्यानं साकाङ्क्षैरनुगम्यते ॥ ९ ॥

citrasyaikasvarūpasya yathā bhedanidarśanaiḥ |
nīlādibhiḥ samākhyānaṃ kriyate bhinnalakṣaṇaiḥ || 8 ||
tathaivaikasya vākyasya nirākāṅkṣasya sarvataḥ |
śabdāntaraiḥ samākhyānaṃ sākāṅkṣairanugamyate || 9 ||

(8) Just as a picture, though single in its form, is described by exhibitions of its differences — 'blue' and the rest, each with a distinct mark — (9) just so, of a single sentence free of requirement in every direction, description is carried out by other words, which do require.

The picture-analogy, working the same holism from the object side. A citra, a painting, is ekasvarupa, one in form — the composition is a single visible whole. Yet we describe it bheda-nidarsanaih, by exhibitions of difference: 'blue here, a figure there', each descriptor bhinna-laksana, marking off its own feature. Nobody concludes the painting *is* an aggregate of blue-patch plus figure-patch; the descriptors are our instruments, their plurality an artifact of description. So (9) the sentence: in itself nirakanksa sarvatah — requirement-free in every direction, expectancy closed (2.4's outer mark, now revealing its metaphysical face: the *complete* thing needs nothing) — while its samakhyana, its description-and-analysis, is conducted sabdantaraih sakanksaih, by *other* words which do require one another: the grammarian's paraphrases, the padas of parsing. The inversion is the argument: mutual requirement, the mark by which view (viii) defined the sentence, turns out to characterize not the sentence but *our talk about it* — the analysans, not the analysandum. Words are needy; the sentence is not; and needy instruments cannot be the substance of the self-sufficient thing they describe.
contested: samakhyanam we render 'description'; it can carry the technical sense 'naming/designation', making the verses about how the sentence is *referred to* rather than analyzed — the holist point survives either way, but the picture-analogy fits description better

2.10 unformalized

यथा पदे विभज्यन्ते प्रकृतिप्रत्ययादयः ।
अपोद्धारस्तथा वाक्ये पदानामुपवर्ण्यते ॥ १० ॥

yathā pade vibhajyante prakṛtipratyayādayaḥ |
apoddhārastathā vākye padānāmupavarṇyate || 10 ||

As stem, suffix and the rest are divided out within the word, so is the abstraction of words within the sentence set forth.

The apoddhara doctrine — 1.24's buried premise, now stated as the kanda's methodological center. Inside a word, grammar divides prakrti from pratyaya, stem from suffix: pac-ati, root and ending. Every grammarian knows the division is *procedural* — speakers do not utter roots and then endings; the divisions are found by the method of comparing forms, and the 'pieces' exist in the analysis, not in the utterance (the Mahabhasya's own tradition concedes as much about roots and suffixes). Bhartrhari's move: tatha vakye padanam apoddharah — the extraction of *words* from the *sentence* has exactly the same standing. The pada is to the vakya what the suffix is to the pada: an analytical precipitate, upavarnyate, 'set forth in description', real in the grammar and projected onto the speech. The argument's force is ad hominem in the honorable sense — the opponent who accepts stem/suffix abstraction (and every Paniniya must) has already accepted the machinery that dissolves his word-realism one level up. What he treats as building blocks are the *products of carving*, and carving presupposes the uncarved.
contested: none of substance; the verse's whole weight lies in taking apoddhara seriously, and the tradition itself (Vrtti and later Paniniyas) reads it as here

2.11 unformalized

वर्णान्तरसरूपं च वर्णभागेषु दृश्यते ।
पदान्तरसरूपाश्च पदभागा अवस्थिताः ॥ ११ ॥

varṇāntarasarūpaṃ ca varṇabhāgeṣu dṛśyate |
padāntarasarūpāśca padabhāgā avasthitāḥ || 11 ||

Within the parts of phonemes are seen shapes resembling other phonemes; and parts of words stand there resembling other words.

Empirical sabotage of the 'real parts' picture, at both levels. Take a phoneme and inspect its stretches: varnantara-sarupam, shapes resembling *other* phonemes appear in its parts — the long vowel contains what sounds like the short, transitions mimic neighboring articulations. Take a word: pada-bhagah stand there padantara-sarupah, word-parts resembling other words — 'kambala' contains 'kam' and 'bala', 'rajapurusa' contains sequences that elsewhere are words in their own right. If parthood were a fact about the speech itself, these embedded look-alikes would *be* parts — the theory would have to say 'kambala' really contains the word 'kam', and every utterance would teem with accidental constituents. The absurdity shows that being-a-part is not a phonetic property at all: nothing in the sound marks the 'real' divisions off from the spurious ones. Division is imposed by the analysis that finds it useful (2.10), and where analysis has no use, resemblance to a word does not make a word. The next verse names the classic examples and the criterion that actually governs.
contested: what varna-bhagas, 'parts of phonemes', are is delicate — phases of articulation, or sub-moras; the karika needs only that inspection of stretches yields pseudo-units at every grain, which any phonetician grants

2.12 unformalized

भागैरनर्थकैर्युक्ता वृषभोदकयावकाः ।
अन्वयव्यतिरेकौ तु व्यवहारनिबन्धनम् ॥ १२ ॥

bhāgairanarthakairyuktā vṛṣabhodakayāvakāḥ |
anvayavyatirekau tu vyavahāranibandhanam || 12 ||

'Vrsabha', 'udaka', 'yavaka' are joined with parts that mean nothing; it is agreement and difference that are the ground of the practice.

The examples, and then — in one line — the actual epistemology of grammatical analysis. Vrsabha ('bull') contains 'vrsa' (itself a word elsewhere) plus '-bha' which means nothing here; udaka ('water') offers 'uda-' and a dangling '-ka'; yavaka ('a red dye') looks like yava ('barley') plus '-ka' but the resemblance buys no meaning. So: words visibly *contain* anarthaka bhagas, meaningless parts — parthood without semantic office. What, then, licenses the divisions grammar *does* make? Anvaya-vyatirekau: agreement and difference — the method of concomitant variation. Where a form recurs *with* a recurring meaning (pac- across pacati, pacatah, paktum, always cooking-), analysis credits it; where presence and absence of form track presence and absence of meaning, a unit is posited; where (as in '-bha' of vrsabha) the form recurs with no meaning-constancy, no unit. This is vyavahara-nibandhanam, the ground of the entire practice: segmentation is an *inference* from distributional evidence, not a perception of joints. The verse is one of the great early statements of the distributional method — and its philosophical sting is that a method is exactly what it is: the parts are as good as the inferences that posit them, theoretical entities to the core.
contested: yavaka we gloss as the dye (traditional); some read 'gruel'. Trivial — the example needs only a -ka that adds no sense

2.13 unformalized

शब्दस्य न विभागोऽस्ति कुतोऽर्थस्य भविष्यति ।
विभागैः प्रक्रियाभेदमविद्वान् प्रतिपद्यते ॥ १३ ॥

śabdasya na vibhāgo'sti kuto'rthasya bhaviṣyati |
vibhāgaiḥ prakriyābhedamavidvān pratipadyate || 13 ||

There is no division of the word — whence should there be one of the meaning? By means of divisions the unlearned man comes to grasp the difference of derivational procedure.

The doctrine stated at full strength, with the divisions relocated to their true home: pedagogy. First line, flat denial: sabdasya na vibhago 'sti — the word (the real one, the sphota) has no division; kuto 'rthasya — and if the signifier is undivided, on what would a division of the *meaning* ride? The pada-artha, the word-meaning as separable item, loses its host. Second line, the constructive relocation: vibhagaih — by means of divisions, the grammarian's stem-suffix-word apparatus — the avidvan, the not-yet-knowing student, pratipadyate, comes to grasp prakriya-bhedam, the difference of derivational procedure. Divisions are didactic instruments: ladders by which the unlearned climb toward a competence that itself operates on wholes. The word avidvan carries the theory of instruction of 1.85 and 2.10 — the false-parts scaffolding is *for* someone, namely the learner, and its justification is what it produces (the vidvan's undivided grasp), not what it depicts. Grammar on this view is a fiction that teaches truly: prakriya, the derivation, is the itikartavyata of language-learning, discharged and dropped like the breath in 1.115.
contested: prakriyabhedam can be construed 'the difference (constituted) by derivational procedure' (divisions teach the procedure) or 'the divisions of the procedure' — pedagogically equivalent; note also the polemical edge: the *avidvan* grasps by division, implying the vidvan does not need to

2.14 unformalized

ब्राह्मणार्थो यथा नास्ति कश्चिद् ब्राह्मणाकम्बले ।
देवदत्तादयो वाक्ये तथाव स्युरनर्थकाः ॥ १४ ॥

brāhmaṇārtho yathā nāsti kaścid brāhmaṇākambale |
devadattādayo vākye tathāva syuranarthakāḥ || 14 ||

As there is no meaning 'brahmana' present in 'brahmana-kambala', so 'Devadatta' and the rest in a sentence would likewise be meaningless.

An opponent's reductio, stated for answering (the reply comes across 2.15-16). The analogy: in brahmana-kambala — 'brahmana-blanket', a blanket *for* a brahmin, or a proper name of a blanket-kind — the element 'brahmana' does not deliver its ordinary meaning: no brahmin figures in the blanket. Compounds notoriously swallow their members' senses (the tradition's rudhi-compounds are the extreme: 'krsnasarpa' names a species, not any black snake). Now the thrust: if the holist is right that sentence-meaning is undivided and word-meanings are mere abstractions, then 'Devadatta' in *devadatta gam abhyaja* would contribute nothing of its own — tathaiva syur anarthakah, the words would be meaningless, all of them, just as 'brahmana' is idle in the blanket-name. But that is absurd: exchange 'Devadatta' for 'Yajnadatta' and the sentence commands a different man. Word-choice manifestly *matters*, item by item — how, unless words carry meanings item by item? The objection is the strongest the partitive side owns: compositional relevance of parts. The holist owes an account of why substituting words changes sentence-meaning if word-meanings are fictions — and 2.15-18 begin paying exactly that debt.
contested: brahmanakambale is read by some as 'in (the phrase) brahmana's blanket' (genitive construction) rather than a compound/name; the objection is sharpest on the compound reading, which we adopt with the tradition

2.15 unformalized

सामान्यार्थस्तिरोभूतो न विशेषेऽवतिष्ठते ।
उपात्तस्य कुतस्त्यागो निवृत्तः क्कावतिष्ठताम् ॥ १५ ॥

sāmānyārthastirobhūto na viśeṣe'vatiṣṭhate |
upāttasya kutastyāgo nivṛttaḥ kkāvatiṣṭhatām || 15 ||

A general meaning, once obscured, does not stand fast in the particular; how could what has been taken up be abandoned — and, withdrawn, where should it abide?

The holist's counter-battery: the partitive account, run honestly, dissolves in its own hands. Suppose words do carry their own meanings, the general senses the lexicon assigns — 'gauh' means cow-in-general. In the sentence ('bring the white cow'), that samanya-artha is tirobhuta, obscured, superseded by the particular this-cow-here. Then the dilemma, put as two unanswerable questions. If the general meaning *stands* in the particular — na visese 'vatisthate, but it does not: the particular cow is not cowhood; the general, as general, is precisely what the sentence-meaning is not about. If instead it is *abandoned* en route: upattasya kutas tyagah — how is what has been *taken up* thrown away? (The word did its work; its meaning entered the hearer's process; semantic contributions are not retractable.) And if withdrawn — nivrttah kva avatisthatam — *where does it go*? The discarded general meaning has no station in the final cognition, yet it cannot have simply evaporated if it was ever a real ingredient. The trilemma's point: treating word-meanings as real deposits that the sentence then processes generates an untrackable ontology of appearing, superseded, homeless senses. The holist's picture — one undivided vakyartha, word-'meanings' as post hoc abstractions (2.7) — has no such debris: nothing general was ever *deposited*, so nothing needs discarding. The objection of 2.14 is thus met by showing the objector's own machinery jams on the ordinary facts of contextual specificity.

2.16 unformalized

अशब्दो यदि वाक्यार्थः पदार्थोऽपि तथा भवेत् ।
एवं च सति सम्बन्धः शब्दस्यार्थेन हीयते ॥ १६ ॥

aśabdo yadi vākyārthaḥ padārtho'pi tathā bhavet |
evaṃ ca sati sambandhaḥ śabdasyārthena hīyate || 16 ||

If the sentence-meaning is word-less, then the word-meaning too would be such; and this being so, the connection of word with meaning is forfeited.

The opponent's rejoinder, driving at the holist's most exposed joint. Grant the holist that the vakyartha is asabdah — 'word-less', not delivered by the words severally, an undivided content that outruns its verbal vehicle (the pratibha-doctrine in prospect). Then, the opponent presses, by parity the padartha too goes word-less: the word-meaning was extracted from the sentence-meaning (2.10, 2.13), so if the source-content floats free of specific words, the extracts do too. Evam ca sati — and with that, sabdasya arthena sambandho hiyate: the *connection of word and meaning is forfeited* — the very nitya sambandha of 1.23, the foundation of the whole first kanda, the eternal pairing on which Vedic authority itself was made to rest. The objection is beautifully aimed: it plays Bhartrhari's two books against each other. Kanda I sanctified the word-meaning bond; Kanda II seems to dissolve meaning's allegiance to words altogether. The holist must show that undivided sentence-meaning *preserves* rather than severs the word-meaning connection — that words are genuine (if non-exclusive) determinants of the vakyartha even though not its parts. Verses 2.17-18 begin the answer: words as manifestors, not containers, of the one meaning — the sphota-logic of Kanda I redeployed to save exactly the bond the objector says is lost.
contested: asabdah we render 'word-less' (not constituted by words); it has been read as 'non-verbal (in nature)', making the objection target the pratibha-doctrine's alleged mysticism directly — the dialectic is the same: parity plus forfeiture of sambandha

2.17 unformalized

विशेषशब्दाः केषाञ्चित् सामान्यप्रतिरूपाकाः ।
शब्दान्तराभिसम्बन्धाद् व्यज्यन्ते प्रतिपत्तृषु ॥ १७ ॥

viśeṣaśabdāḥ keṣāñcit sāmānyapratirūpākāḥ |
śabdāntarābhisambandhād vyajyante pratipattṛṣu || 17 ||

For some theorists, particular-words have the look of generals; they become manifest among comprehenders through connection with other words.

First installment of the holist's answer to 2.14-16, reported from within the school (kesamcit, 'for some'). Reverse the lexicographer's picture. Words in isolation are not generals awaiting particularization; rather visesa-sabdah — words are *particular-words*, each fitted to the determinate role it plays in determinate sentences — which, cited alone, merely *look* general: samanya-pratirupakah, wearing the general's appearance, as a tool out of use looks like a mere shape. Their real, particular signification vyajyate — is *manifested* (the Kanda I verb, deliberately) — pratipattrsu, in comprehenders, sabdantara-abhisambandhat, through connection with the other words. On this picture 2.15's trilemma never starts: no general meaning is deposited and then homeless, because the 'general meaning' was an artifact of citation, the dictionary's freeze-frame of a word that lives only in sentential motion. And 2.16's forfeiture is answered in the same stroke: the word-meaning bond is intact — each word is bonded to what it (with its fellows) manifests — it was only the *isolationist* version of the bond, word-to-general-sense as a private possession, that has been surrendered. Substitution-sensitivity (2.14) follows naturally: change 'Devadatta' to 'Yajnadatta' and a different particular manifestation ensues, because a different manifestor entered the ensemble. Words matter item by item as *co-manifestors* of one meaning, not as depositors of separable ones.
contested: pratipattrsu ('in/among comprehenders') locates the manifestation in hearers, which some soften to 'for the comprehenders'; the stronger locative reading — meaning happens *in* minds, not between words — anticipates the pratibha doctrine and we keep it

2.18 unformalized

तेषां तु कृत्स्नो वाक्यार्थः प्रतिभेदं समाप्यते ।
व्यक्तोपव्यञ्जना सिद्धिरर्थस्य प्रतिपत्तृषु ॥ १८ ॥

teṣāṃ tu kṛtsno vākyārthaḥ pratibhedaṃ samāpyate |
vyaktopavyañjanā siddhirarthasya pratipattṛṣu || 18 ||

But for them the entire sentence-meaning is brought to completion at each division; the accomplishment of the meaning, in comprehenders, has the manifest as its auxiliary manifestor.

The radical corollary, and the verse where the holist account becomes genuinely strange and genuinely powerful. For these theorists (tesam tu), the krtsnah vakyarthah — the *entire* sentence-meaning — samapyate pratibhedam: is brought to completion *at each division*, at every word. Not: each word contributes a share until the sum is reached at the end. Rather: the whole meaning is present-in-completion again and again, word by word — each pada an occasion on which the one undivided vakyartha stands accomplished in the comprehender, more and more determinately manifest (vyakta-upavyanjana: the already-manifest serving as auxiliary manifestor of what comes clearer). The picture is the seed-and-ripening of 1.84 turned inside out: there, sounds deposited traces until the last one triggered the whole; here, the whole is *at* every stage, each word re-manifesting it at higher resolution — as each glance at the painting (2.8) sees the *painting*, not a painting-fraction, though later glances see it better. This dissolves the residue of 2.14 completely: 'Devadatta' does not add a Devadatta-brick to a growing wall; it re-precipitates the entire command, now Devadatta-determinate. Siddhir arthasya pratipattrsu: the meaning's accomplishment is in comprehenders — semantics has moved decisively into the mind, where Kanda II will house it (pratibha) from here on. The cost, willingly paid: comprehension is not accumulation but progressive revelation — understanding a sentence is watching one thing come into focus, not watching many things pile up.
contested: vyaktopavyanjana is a difficult compound: we resolve it as 'having the (already-)manifest as auxiliary manifestor' — each stage of clarity assists the next; others resolve 'manifestation by clear auxiliary signs'. The verse's doctrine — whole-at-every-word, completion per division — is secure; the mechanism-compound is not

2.19 · Sequence Interrogated unformalized

अव्यक्तः क्रमवात्र्शब्दः उपांश्वयमधीयते ।
अक्रमस्तु वितत्येव बुद्धिर्यत्रावतिष्ठते ॥ १९ ॥

avyaktaḥ kramavātrśabdaḥ upāṃśvayamadhīyate |
akramastu vitatyeva buddhiryatrāvatiṣṭhate || 19 ||

Unmanifest and, as it were, possessed of sequence, this word is recited inaudibly; but where the cognition comes to rest, it is sequenceless, as though spread out whole.

Evidence from inner speech. Recite a verse upamsu, under the breath — no sound, no articulation reaching the air — and the recitation still *takes time*, still runs in order: avyaktah kramavan iva, unmanifest yet sequenced-as-it-were. But attend to the other terminus: yatra buddhir avatisthate, where the cognition settles — the verse as *known*, the prayer as *possessed* — and sequence is gone: akramah vitatya iva, sequenceless, spread out as a simultaneous whole, the way a memorized text stands available all-at-once to its knower. One content, two conditions: deployed in inner time, or held in no time. The verse quietly dismantles the assumption that sequence needs *sound*: even soundless speech has krama when deployed — so krama tracks deployment (vrtti), not acoustics, exactly as 1.75-77 assigned it to manifestation-conditions. And the resting cognition shows the same content krama-free — so sequence cannot be constitutive of the word known. The iva twice-over is Bhartrhari's signature scruple: both the inner sequence and the spread-out wholeness are *presentations* of something that is in itself neither.
contested: the transcription kramavatr(a)- is faulty; kramavan iva is the evident text and we translate so. Vitatya iva ('as though having spread') is adverbial to the resting cognition — some construe it of the word ('as though extended'), same doctrine

2.20-21 unformalized

यथाक्षेपविशेषेऽपि कर्मभेदो न गृह्यते ।
आवृत्तौ व्यज्यते जातिः कर्मभिर्भ्रमणादिभिः ॥ २० ॥
वर्णवाक्यपदेष्वेवं तुल्योपव्यञ्जना श्रुतिः ।
अत्यन्तभेदे तच्वस्य सरूपेव प्रतीयते ॥ २१ ॥

yathākṣepaviśeṣe'pi karmabhedo na gṛhyate |
āvṛttau vyajyate jātiḥ karmabhirbhramaṇādibhiḥ || 20 ||
varṇavākyapadeṣvevaṃ tulyopavyañjanā śrutiḥ |
atyantabhede tacvasya sarūpeva pratīyate || 21 ||

(20) Just as, though the impulses differ in each case, no difference of act is grasped — in the repetition, the universal is manifested by the acts of whirling and the rest; (21) so, in phonemes, words and sentences, hearing has a like mode of auxiliary manifestation: though its object is absolutely distinct, it is apprehended as if of like form.

An analogy from the perception of action, deployed to explain how hearing seems to present parts it does not present. Watch a whirling firebrand or a spinning wheel (bhramana-adi): each revolution rides on its own impulse — aksepa-visesa, distinct casts of force — yet the spectator grasps no karma-bheda, no plurality of acts; what the avrtti, the repetition, manifests is the *universal* of the action, whirling-as-such, one content standing steady over many pushes. (21) applies the model across all three linguistic grains: sruti, hearing, has tulya-upavyanjana, the same auxiliary-manifestation structure (the upavyanjana of 2.18 again) — the many acoustic pulses stand to the heard word/sentence as the many impulses to the one perceived whirling. And the conclusion, stated with both edges: the object manifested is atyanta-bhede, absolutely distinct from the manifesting events — yet sarupa iva pratiyate, it is apprehended *as if* of like form with them: the word seems sound-shaped, seems to share the pulses' seriality, exactly as the whirl seems to inherit the pushes' rhythm. The analogy improves on the mirror (1.49): there, borrowed properties; here, in addition, *repetition manifesting a universal* — many events, one content, apparent form-sharing — which is the precise logical shape the sentence-sphota needs.
contested: 20a's yatha aksepa-visese 'pi is our construal (impulse-differences not registering as act-differences); the line is compressed and has been read as 'though the projection (of the brand) differs'. The jati manifested by repetition points ahead to the jati-sphota option of 1.93 — whether 2.21 endorses it or merely borrows its structure is open

2.22 unformalized

नित्येषु तु कुतः पूर्वं परं वा परमार्थतः ।
एकस्यैव तु सा शक्तिर्यदेवमवभासते ॥ २२ ॥

nityeṣu tu kutaḥ pūrvaṃ paraṃ vā paramārthataḥ |
ekasyaiva tu sā śaktiryadevamavabhāsate || 22 ||

But among eternal things, whence anything prior or posterior in the ultimate sense? It is rather the power of the one that appears thus.

The metaphysical floor put under the whole sequence-discussion. If the linguistic realities — phoneme-, word-, sentence-sphotas, and behind them the sabda-tattva — are nitya, then paramarthatah, in strictness, purva and para have no purchase among them: earlier-and-later is a relation of the produced, and nothing eternal stands in it. Whence, then, the overwhelming appearance of order — 'g' before 'au', subject before predicate? Ekasya eva sa saktih: it is the *power of the one* that so appears. Sequence is not suffered by the eternal items; it is projected by the one word-principle as its mode of self-display (the kala-sakti of 1.3, the vivarta of 1.48, now answering for syntax itself). The verse turns the kanda's local question — is the sentence's krama real? — into the system's global answer: krama is always phenomenal, at every scale, from phoneme-order to world-process; it is how the undivided *presents*, never how it *is*. What follows (2.23-27) descends from this height to argue the case in the hardest empirical arena: the felt difference between long and short vowels, quick and slow speech — precisely where sequence seems most stubbornly real.

2.23 unformalized

चिरं क्षिप्रमिति ज्ञाने कालभेदादृते यथा ।
भिन्नकाले प्रकाशेते स धर्मो ह्लस्वदीर्घयोः ॥ २३ ॥

ciraṃ kṣipramiti jñāne kālabhedādṛte yathā |
bhinnakāle prakāśete sa dharmo hlasvadīrghayoḥ || 23 ||

As in the cognitions 'slow' and 'quick' — though without a difference in their time — things shine forth as if of different time: that is the property of short and long.

The stock objection to timeless words, met on its own ground. 'Slowly' and 'quickly' are how we experience utterances; hrasva and dirgha, short and long vowels, are contrastive *by duration* (1.76's hard datum). Bhartrhari's reply-in-analogy: consider the cognitions themselves. The *thought* 'this was slow' and the *thought* 'this was quick' do not themselves differ in duration — kala-bhedad rte, without any difference of their own time, they present their objects bhinna-kale iva, as if in different time. A cognition of a week is not week-long; awareness of speed is not speedy. Mental acts represent temporal properties without instantiating them. Sa dharmo hrasva-dirghayoh: that is exactly the property-structure of short and long — the word (like the cognition) *presents* duration-difference without *possessing* it; length lives in the presented profile (the prakrta-dhvani, 1.76), not in the presenting sphota. The argument is elegant because its premise is undeniable: everyone grants that cognitions represent times they do not take. Once representing-without-instantiating is admitted anywhere, the word's timeless presentation of temporal contrast is no longer paradoxical — it is one more instance of a structure consciousness exhibits constantly.
contested: we take the verse's subject to be cognitions of slow/quick (jnane as locative dual framing them); an alternative construes 'in knowledge, though there is no time-difference, the two appear at different times' more generally — the analogical force is identical

2.24 unformalized

न नित्यः परमात्राभिः कालो भेदमिहार्हति ।
व्यावर्त्तिनीनां मात्राग्णमभावे कीदृशः क्रमः ॥ २४ ॥

na nityaḥ paramātrābhiḥ kālo bhedamihārhati |
vyāvarttinīnāṃ mātrāgṇamabhāve kīdṛśaḥ kramaḥ || 24 ||

Time, being eternal, does not here deserve division by further measures; when the measures that would demarcate are absent, of what sort could sequence be?

The dilemma turned against the realist about temporal sequence. Suppose sequence in speech is real, grounded in time itself. But kala, on the going account (and Bhartrhari's own, 1.3: time as the one sakti on which division is *superimposed*), is nitya and one — na bhedam iharhati, it does not admit of intrinsic division; the 'measures' (matras) that would carve it into befores and afters are not further times (on pain of regress: what orders the measures?). And vyavartinam matranam abhave — in the absence of demarcating measures — kidrsah kramah: what *kind* of thing could sequence even be? The argument has the form of Bhartrhari's best: the opponent needs krama to be an objective order; objective order needs a divided time; divided time needs dividers that are not themselves temporal divisions; nothing fills that office. So sequence cannot be *found* in time — it must be *brought* to it, and the bringer is cognition (2.25 next) operating with the superimposed divisions of 1.3. The verse completes the pincer of 2.22-23: metaphysically, eternals admit no order; psychologically, cognitions represent order without instantiating it; and now temporally, time itself has no native joints for order to consist in. Krama has nowhere left to live but in the presenting mind.
contested: the transcription matragnam is corrupt (likely matranam); we translate the evident sense. Para-matrabhih 'by further measures' has also been read 'by supreme/other measures' — the regress-argument stands on any reading

2.25 unformalized

ताभ्यो या जायते बुद्धिरेका सा भागवर्तिनी ।
सा हि स्वशक्त्या भिन्नेव क्रमप्रत्यवमर्शिनी ॥ २५ ॥

tābhyo yā jāyate buddhirekā sā bhāgavartinī |
sā hi svaśaktyā bhinneva kramapratyavamarśinī || 25 ||

The cognition that arises from those is one, though it ranges over the parts; for by its own power it is, as it were, divided — it is the recognizer of sequence.

The positive account: where sequence actually resides. From the many manifesting events ('those' — the sounds, the measures) arises a buddhi that is eka, one act — yet bhaga-vartini, ranging over the parts, touring what it presents as stages. And the crucial attribution: sa hi sva-saktya bhinna iva — by its *own* power it stands as-if-divided; the division is the cognition's self-articulation, not an imprint received from divided objects (2.24 showed there are none to receive from). Its title: krama-pratyavamarsini, the *recognizer of sequence* — deliberately re-sounding 1.124's pratyavamarsini: there, the word-form in awareness was what makes any recognition possible; here, its specific office of constituting order-recognition. Sequence, then, is a categorial contribution of the apprehending mind — brought to the manifold, not extracted from it — and the 'parts' of the sentence are the itinerary of one act's self-divided tour. The verse is the kanda's Copernican moment, and its placement matters: only *after* the object-side (2.22), the representation-side (2.23), and the time-side (2.24) were each shown incapable of housing krama is the mind's own sakti allowed to claim it.
contested: tabhyah ('from those', feminine) most naturally picks up the matras of 2.24; some refer it to the srutis. Either way the mothers of the cognition are the manifesting events, and the doctrine is unchanged

2.26 unformalized

क्रमोल्लेखानुषङ्गेण तस्यां यद् बीजमाहितम् ।
तत्त्वनानात्वयोस्तस्य निरुक्तिर्नावतिष्ठते ॥ २६ ॥

kramollekhānuṣaṅgeṇa tasyāṃ yad bījamāhitam |
tattvanānātvayostasya niruktirnāvatiṣṭhate || 26 ||

Through its entanglement with the tracing-out of sequence, a seed is deposited in that cognition; and of that, as between unity and plurality, no settled account stands.

An admission of undecidability at the theory's exact center — rare in any scholastic author and structural here. The one cognition of 2.25, in performing its krama-ullekha, its tracing-out of sequence, is anusanga — entangled, involved — and by that entanglement a bija is deposited in it (the seed-vocabulary of 1.84 and 1.132: consequence lodged for later ripening — here, the residue that makes the *next* apprehension order-wise, the sedimented habit of sequencing). Then the confession: tattva-nanatvayos tasya niruktir na avatisthate — as to whether that (the cognition with its seed; by extension the whole word-mind complex) is *one* or *many*, no nirukti, no settled analysis, stands. Is the sequencing cognition really single (its division mere iva, 2.25) or really plural (its unity the construction)? The question reproduces, inside the mind, the exact antinomy 1.72-74 found in language — and Bhartrhari, consistently, declines it again: both descriptions are available, neither is final, and the undecidability is itself the doctrine (the one appears as many; asking which it 'really' is mistakes appearance-talk for inventory-talk, 1.74). Placing the avowal *here* — where a lesser system would fortify — is the mark of a philosopher tracking his method's limits from inside.
contested: the referent of tasya ('of that') is genuinely open: the seed, the cognition, or the word itself; we keep it at the cognition-complex. Some read niruktir navatisthate as 'the etymology/definition does not hold', a weaker point; the parallel with 1.74's darsana-relativity supports the stronger reading

2.27 unformalized

भावनासमये त्वेतत् क्रमसामर्थ्यमक्रमम् ।
व्यावृत्तभेदो येनार्थो भेदवानपलभ्यते ॥ २७ ॥

bhāvanāsamaye tvetat kramasāmarthyamakramam |
vyāvṛttabhedo yenārtho bhedavānapalabhyate || 27 ||

But at the moment of realization, this capacity-for-sequence is itself sequenceless — that by which the object, its differentiation withdrawn, is apprehended as differentiated.

The paradox owned, polished, and made the resolution. At bhavana-samaya — the moment of realization, when understanding actually happens (bhavana: the bringing-into-being of the meaning in the mind; the term will do heavy work in the pratibha doctrine) — the krama-samarthyam, the very capacity by which the mind sequences, is akramam, itself sequenceless: the power to run through stages is not itself a staged thing (as the power to move is not itself in motion). And by that sequenceless capacity, the artha — whose own differentiation is vyavrtta, withdrawn, absent in the object as it is (2.22-24) — bhedavan upalabhyate, is apprehended *as* differentiated. Every joint of the completed account is now explicit: undivided object, undivided apprehending power, and between them the *presentation* of division — krama existing nowhere but in the presenting, a pure how mistaken for a what. This is 1.85's productive error given its final form: the mind's sequencing is the upaya, sequenceless in itself, generating the ordered appearance through which alone the unordered can be humanly taken. With sequence thus fully accounted, the kanda can return (2.28ff.) to demolishing the phoneme-and-word ontology that assumed sequence was real all along.

2.28-29 unformalized

पदानि वाक्ये तान्येव वर्णास्ते च पदे यदि ।
वर्णेषु वर्णभागानां भेदः स्यात् परमाणुवत् ॥ २८ ॥
भागानामनुपश्लेषान्न वर्णौ न पदं भवेत् ।
तेषामव्यपदेश्यत्वात् किमन्यदपदिश्यताम् ॥ २९ ॥

padāni vākye tānyeva varṇāste ca pade yadi |
varṇeṣu varṇabhāgānāṃ bhedaḥ syāt paramāṇuvat || 28 ||
bhāgānāmanupaśleṣānna varṇau na padaṃ bhavet |
teṣāmavyapadeśyatvāt kimanyadapadiśyatām || 29 ||

(28) If the words in the sentence are just those (independent reals), and the phonemes likewise in the word, then within phonemes there would be division of phoneme-parts, as with atoms; (29) and since the parts have no mutual adherence, neither phoneme nor word could come to be — and since those (parts) cannot be designated, let something else be designated!

The regress-and-dissolution argument: partitive realism refuted by being taken seriously. Suppose (28) the realist's ladder: sentences really contain words, words really contain phonemes. Consistency forbids stopping: phonemes have stretches (2.11), so varna-bhagah, phoneme-parts, must be real too — paramanuvat, as with atoms, whose divisibility-regress is the tradition's standing scandal (the Vaisesika atom under Buddhist attack: whatever has parts dissolves into them; whatever has none cannot aggregate). Then (29) the two-step execution. First: bhaganam anupaslesat — the parts have no *adherence*, no mutual sticking; sounds are episodic, each gone before the next arrives (1.44's original datum), so nothing binds part to part — na varnau na padam bhavet: by rights neither phoneme nor word could ever come to be; the realist's building-blocks cannot even build his bricks. Second: tesam avyapadesyatvat — the ultimate parts are *undesignatable*, beneath every unit language can mention (what is the name of the first third of 'g'?) — kim anyad apadisyatam: then let something *else* be designated!, a taunt with a doctrine inside it: since analysis bottoms out in the unnameable, the real bearer of names — of designation, of meaning — must be sought elsewhere than at the bottom: in the undivided wholes at the *top*, which is where 2.30 immediately goes. The argument recycles the tradition's finest anti-atomist weaponry against linguistic atomism — likely with full irony intended, since the phoneme-realists included exactly the schools that wielded it.
contested: 29d's exclamation kim anyad apadisyatam we read as the ironic challenge ('name something else, then!') pivoting to the siddhanta; it can be read as a genuine question ('what else should be designated?') closing the purvapaksa — the pivot to 2.30 favors the ironic reading

2.30 unformalized

यदन्तः शब्दतत्त्वं तु नादैरेकं प्रकाशितम् ।
तदाहुरपरे शब्दं तस्य वाक्ये तथैकता ॥ ३० ॥

yadantaḥ śabdatattvaṃ tu nādairekaṃ prakāśitam |
tadāhurapare śabdaṃ tasya vākye tathaikatā || 30 ||

But that inner word-essence, one, made manifest by the sounds — that is what the others call the word; and in the sentence its oneness is likewise.

The siddhanta re-entering at the point the regress cleared. Where designation bottomed out (2.29), the apare, the other theorists — the sphota-vadins, Bhartrhari's own line — point inward: yad antah sabda-tattvam, the word-essence *within*, ekam, one, nadaih prakasitam, lit up by the sounds (the manifestation doctrine of Kanda I in one clause: prakasita, illumination-vocabulary, 1.94's lamp) — *that* is the word. Not an aggregate surviving the regress, but the undivided thing the regress never touched, since it was never composed. And the extension that makes this Kanda II's verse rather than a repetition: tasya vakye tatha ekata — its oneness in the *sentence* is the same. The sentence is not many sphotas in a row (which would re-open every regress one level up); it is *one* sphota at sentence-grain (2.1's fourth view, now installed), manifested by the whole utterance-event as the word was by its sounds. The antah, 'inner', quietly relocates the discussion for the rest of the kanda: the real word was never in the air — the air carries manifestors; the manifested lives where 2.18 put meaning, in the comprehender. What remains is to say what the *meaning* of this inner sentence-one is — and 2.31 answers with the system's deepest identification.
contested: apare, 'the others', is Bhartrhari's usual doxographic distancing even for views he holds; whether 2.30 is reported or owned matters little — by 2.31-33 the voice is unmistakably the author's

2.31 unformalized

अर्थभागैस्तथा तेषामान्तरोऽर्थः प्रकाश्यते ।
एकस्यैवात्मनो भेदौ शब्दार्थावपृथकस्थितौ ॥ ३१ ॥

arthabhāgaistathā teṣāmāntaro'rthaḥ prakāśyate |
ekasyaivātmano bhedau śabdārthāvapṛthakasthitau || 31 ||

So too by the meaning-parts is their inner meaning illuminated; word and meaning are the two aspects, standing non-separate, of one and the same self.

The identification toward which both kandas have been converging. First the parallel completed: as sounds manifest the inner word (2.30), so artha-bhagaih — by the meaning-*parts*, the padarthas of analysis, the itemized senses — the antarah arthah, the inner meaning, is illuminated: word-meanings stand to the one vakyartha exactly as sounds to the sphota — manifestors, scaffolding, upaya (2.13, 2.18). Then the thesis that dissolves the remaining dualism: ekasya eva atmanah bhedau sabda-arthau aprthak-sthitau — word and meaning are two aspects, non-separately stationed, of *one self*. Not two relata bonded by a sambandha (the picture 2.16's objector fought to preserve), but one reality self-presented twice: as expressive form and as expressed content — the double luminosity of 1.50 (word shows itself and its meaning) now grounded in identity: it shows both because it *is* both. This is the sabdadvaita cashing its name: at the root, signifier and signified are the sabda-tattva's two faces, and the 'connection of word and meaning' (1.23's nitya sambandha) is eternal precisely because it is not a connection at all — nothing needs tying where nothing was ever apart. The inner sentence-word of 2.30 and its undivided meaning are one thing turned two ways; comprehension (pratibha, soon) is that thing turning.
contested: atmanah we render 'self' in the reflexive-metaphysical sense (the one reality; ultimately the vagatman of 1.112); a psychological reading ('of one cognition, word and meaning are the two non-separate aspects') is available and makes the verse epistemology rather than metaphysics — the kanda supports both registers deliberately

2.32 unformalized

प्रकाशकप्रकाश्यत्वं कार्यकारणरूपता ।
अन्तर्मात्रात्मनस्तस्य शब्दतत्त्वस्य सर्वदा ॥ ३२ ॥

prakāśakaprakāśyatvaṃ kāryakāraṇarūpatā |
antarmātrātmanastasya śabdatattvasya sarvadā || 32 ||

Illuminator-and-illuminated, and the form of cause-and-effect: these belong, at all times, to that word-essence whose self is the inner measure.

The two great relational idioms of the system, folded back into the one. Throughout both kandas two vocabularies have alternated: prakasaka-prakasya (manifestor/manifested — sounds and sphota, word and meaning, 1.94, 2.30-31) and karya-karana (effect/cause — sounds *produced by* the word, 1.46-47; world proceeding from Brahman, 1.1). Which is the true structure? The verse's answer: both, always, and *internally* — prakasaka-prakasyatvam and karya-karana-rupata are standing features sarvada of the sabda-tattva itself, antarmatra-atmanah, whose self is inner measure (the antah of 2.30 made essential: the word-principle is intrinsically the inner term of every such pair). The point is architectural: illumination and causation are not rival theories of the word-sound or word-meaning link between which the school must choose — they are the two aspect-relations under which the one self-differentiating reality perpetually stands to its own manifestations (2.31's bhedau generalized to relational form). What looks like a relation between two things is the one thing's way of being two. This prepares 2.33's still starker claim: even existence and non-existence are its stationed powers.
contested: antarmatratmanah is a hard compound: we resolve 'whose self is the inner measure' (essentially inner); alternatives — 'whose self is mere inwardness', 'of the inner-moment self' — shade the metaphysics without redirecting it

2.33 unformalized

तस्यैवास्तित्वनास्तित्वसामर्थ्ये समवस्थिते ।
अक्रमे क्रमनिर्भासे व्यवहारनिबन्धने ॥ ३३ ॥

tasyaivāstitvanāstitvasāmarthye samavasthite |
akrame kramanirbhāse vyavahāranibandhane || 33 ||

In that same one stand established the capacities for existence and for non-existence — sequenceless, appearing as sequence, the ground of all practice.

The high-water mark of the kanda's metaphysics, after which the argument deliberately descends to semantics (2.34ff.). Three predications of the one word-essence, each a resolution of a standing antinomy. Astitva-nastitva-samarthye samavasthite: the capacities for existence *and* non-existence stand established in it — presence and absence alike are its powers (how else could language speak of the non-existent, negate, counterfactualize? — the capacity to present absence is as positive a power as the capacity to present being; the hare's horn lives in speech). Akrame krama-nirbhase: sequenceless, *appearing* as sequence — the whole dossier 2.19-27 sealed into an epithet. Vyavahara-nibandhane: and precisely as such — powers of being and non-being, timeless ground of temporal show — it is the nibandhana of vyavahara, the fastening of all practice: the term that has tracked 'ground' since 1.13 here takes its widest object; everything transacted, said, done, negated, planned, hangs from this. The verse is Kanda II's answer to 1.1: there, Brahman-as-word from whom the world proceeds; here, analyzed to its working parts — the one whose paired capacities and sequence-show *are* the world's practicability. What such a ground cannot be made to do is behave as one item within the practice it grounds — which is exactly the deflation about to be visited on the padartha-realists.
contested: the locative absolute construction leaves the verse's syntactic completion open (the sentence continues into the discourse rather than closing); we render it as free-standing predication, as the tradition does

2.34 unformalized

संप्रत्ययप्रमाणत्वात् पदार्थास्तित्वकल्पने ।
पदार्थाभ्युच्चये त्यागादानर्थक्यं प्रसज्यते ॥ ३४ ॥

saṃpratyayapramāṇatvāt padārthāstitvakalpane |
padārthābhyuccaye tyāgādānarthakyaṃ prasajyate || 34 ||

Since the means of knowledge is the comprehension itself: in the positing of word-meanings as existent, if a word-meaning is added or subtracted, meaninglessness results.

Return to semantics, with the criterion that governs the remainder of the kanda: samprataya-pramanatvat — the pramana here is the *comprehension itself*; what a sentence means is answerable, finally, to what competent hearers actually understand, not to what a theory computes from parts. Then the test applied against padartha-realism. If word-meanings are existents that compose (padartha-astitva-kalpane), composition should be monotonic: add an item (abhyuccaya), the meaning should grow by it; drop one (tyaga), it should shrink by it. The facts refuse: anarthakyam prasajyate — add or subtract against the sentence's one meaning and you get not more or less meaning but *meaninglessness* (or a different whole altogether): 'bring the cow' plus a stray genitive is not an enriched command but noise; idioms lose everything by one substitution; the same word added here completes, there destroys. Sentence-meaning behaves like a gestalt, not a sum — tolerance zero at its boundary, holistically re-formed within it. The argument is the empirical mirror of 2.15's ontological trilemma, and its premise is the school's deepest methodological commitment: linguistics answers to comprehension (the pratipattr of 2.17-18), the theorist's ledger of parts having no authority against what understanding actually does.
contested: we construe abhyuccaye tyagat as the add/subtract test (either operation, against the fixed comprehension, yields anarthakya); the compressed ablatives admit other pairings — e.g. 'because of abandonment (of the general sense) in the aggregation' — which would echo 2.15 instead; the comprehension-criterion in pada a is secure on any parsing

2.35ab unformalized

राजशब्देन राजार्थो भिन्नरूपेण गम्यते ।

rājaśabdena rājārtho bhinnarūpeṇa gamyate |

By the word 'rajan' the king-meaning is understood under a differentiated form —

A half-verse (the source transmits it split), opening the compound-analysis: in 'raja-purusa', 'king's man', the element rajan- delivers its sense bhinna-rupena, under a differentiated form — not the free-standing king of 'the king dines' but king-as-qualifier, bent toward the purusa it possesses. The lexical 'same word' contributes differently under construction — the isolationist's fixed deposit (2.15, 2.17) failing again, this time inside a single compound.

2.35cd unformalized

वृत्तावाख्यातसद्दशं पदमन्यत्र युज्यते ॥ ३५cd ॥

vṛttāvākhyātasaddaśaṃ padamanyatra yujyate || 35cd ||

— while in the derivational analysis, a word resembling a finite verb is employed for what stands otherwise.

The completion: grammar's vrtti-analysis paraphrases the compound by a clause — 'rajnah purusah', or with a verb-like element supplied ('the man *who belongs to* the king') — employing akhyata-sadrsam padam, a word resembling a finite verb, for what in the compound stands anyatra, otherwise (verblessly, fused). The paraphrase is faithful to the meaning and *false to the form*: the compound contains no verb, yet its analysis requires one. Moral, in the kanda's running key: analysis is translation into the analyst's idiom, not excavation of hidden constituents — the clause 'in' the compound is exactly as real as the phonemes 'in' the sphota (2.10). The pair 35ab/cd thus exhibits, at the smallest scale, both directions of the general doctrine: meaning shifting under construction (ab), and construction-analysis fabricating parts the meaning never had (cd).
contested: the source splits 2.35 across two pages and the halves' connection is loose; some editions read them as one karika on vrtti (compound-formation) generally — our joint reading follows the evident sense of the received halves

2.36 unformalized

यथाश्वकर्ण इत्युक्ते विनैवाश्वेन गम्यते ।
कश्चिदेव विशिष्टो'र्थः सर्वेषु प्रत्ययस्तथा ॥ ३६ ॥

yathāśvakarṇa ityukte vinaivāśvena gamyate |
kaścideva viśiṣṭo'rthaḥ sarveṣu pratyayastathā || 36 ||

As, when 'asvakarna' is said, a certain particular thing is understood without any horse at all — such is the cognition in all cases.

The rudhi-compound as the revealing limit-case, generalized with sudden boldness. Asvakarna, 'horse-ear', names a tree (the sal); when it is said, vinaiva asvena — without any horse whatsoever entering the comprehension — kascid eva visisto 'rthah, a certain quite particular thing (the tree), is understood: the 'members' contribute nothing; the whole word delivers its whole meaning at a stroke. So far, standard: every school parks such compounds in the 'conventional' bin and moves on. Then the last quarter detonates: sarvesu pratyayas tatha — the cognition is like that *in all cases*. What the opaque compound wears openly — meaning attaching to the undivided whole, 'parts' semantically idle — is on this view the true structure of *every* understanding: 'raja-purusa' too is grasped as one word-one meaning, its apparent member-contributions being the analyst's after-the-fact bookkeeping (2.10, 2.13, 2.35cd); the transparent compound differs from the opaque only in *inviting* an analysis, not in being composed. The generalizing move is the kanda's signature (as 1.119 generalized from music): find the case where partition visibly fails, then argue the 'normal' cases differ in salience, not in kind. Comprehension everywhere is asvakarna-comprehension — whole to whole, at once.
contested: none textually; doctrinally, whether sarvesu extends to all *compounds* or all *words and sentences whatsoever* divides readers — the kanda's argumentative arc (toward the undivided vakyartha) demands, and we adopt, the wider scope

2.37 unformalized

वाक्येष्वर्थान्तरगतेः सादृश्यं परिकल्प्यते ।
केषाञ्चिद् रूढिशब्दत्वं शास्त्र एवानुगम्यते ॥ ३७ ॥

vākyeṣvarthāntaragateḥ sādṛśyaṃ parikalpyate |
keṣāñcid rūḍhiśabdatvaṃ śāstra evānugamyate || 37 ||

It is for the sake of understanding another meaning in sentences that the resemblance is constructed; for some, the status of conventional word is acknowledged within the science itself.

Two housekeeping clarifications that protect 2.36's generalization. First: when analysis paraphrases — when grammar 'finds' horse and ear in asvakarna, king and man in rajapurusa — the sadrsya, the resemblance between member-look-alikes and free words, is parikalpyate, *constructed*, arthantara-gateh, for the sake of getting at another meaning: paraphrase is a bridge built from known words toward the unfamiliar unit's sense, a heuristic of glossing (the teaching-function of 2.13 again), not a discovery that the unit was secretly a sentence. Second: kesamcit rudhi-sabdatvam sastre eva anugamyate — for some theorists, being-a-conventional-word is acknowledged *within the science itself*: Panini's own system treats certain formations as unitary and underived (avyutpanna), their 'derivations' declined even by the derivers. The point of citing it: the concession the opponent makes at the margin (some words are just wholes) is the thin edge of 2.36's wedge — once *any* word's meaning attaches whole-to-whole with idle parts, partition has lost its claim to be constitutive of meaning as such, and retaining it elsewhere is a choice of bookkeeping, not a finding of fact.
contested: arthantara-gateh we take as 'for the sake of reaching the other (=the unit's) meaning' — paraphrase as bridge; it can be read 'because (the compound) conveys a different meaning (than its members)' — the two shade into each other and both serve the verse's deflationary office

2.38 unformalized

उपादायापि ये हेयास्तानुपायान् प्रचक्षते ।
उपायानां च नियमो नावश्यमवतिष्ठते ॥ ३८ ॥

upādāyāpi ye heyāstānupāyān pracakṣate |
upāyānāṃ ca niyamo nāvaśyamavatiṣṭhate || 38 ||

Even things taken up that are afterwards to be discarded — these they declare to be means; and a fixed rule about means does not necessarily stand.

The pedagogy of fictions, stated as a general maxim and closing the movement. Upadaya api ye heyah — what is *taken up precisely to be let go*: the ladder-entities of the whole discussion (phoneme-divisions, word-meanings, derivational members, the constructed resemblances of 2.37) — tan upayan pracaksate: these, the tradition declares, are *means*. The formulation compresses 1.85 (error as the very means of grasping), 2.13 (divisions for the unlearned), and every iva of both kandas into a rule of method: a scaffold is not refuted by the finished building; it is *discharged*. And the rider does real work: upayanam ca niyamo na avasyam avatisthate — among means there is no necessary fixity: no single ladder is privileged; different learners, different analyses, different fictions may serve (the darsana-tolerance of 1.74 and 2.26 as pedagogy). The maxim shields the school from a tu-quoque it plainly invites — 'your sphota-talk, your levels, your similes: are they not also constructions?' Answer: yes, and offered as upaya, discardable at attainment; the difference between the grammarian and his opponent is not that one uses fictions and the other facts, but that one knows which of his instruments are ladders. The verse is the kanda's methodological self-consciousness at full candor — and the traditional resonance (upaya-heya structures run through Buddhist method too) is surely heard, and meant.
contested: none of weight; the verse reads as the maxim it is — its interest is entirely in scope: whether Bhartrhari means it reflexively of his own doctrine (we take it so, with 1.74 and 2.26 as warrant) or only of the opponent's apparatus

2.39 unformalized

अर्थं कथञ्चित् पुरुषः कश्चित् संप्रतिपद्यते ।
संसृष्टा वा विभक्ता वा भेदा वाक्यनिबन्धनाः ॥ ३९ ॥

arthaṃ kathañcit puruṣaḥ kaścit saṃpratipadyate |
saṃsṛṣṭā vā vibhaktā vā bhedā vākyanibandhanāḥ || 39 ||

Somehow or other, some person comes to an understanding of the meaning; the differences — whether fused or divided — are fastenings relative to the sentence.

Deflationary realism about the variety of comprehension, extending 2.38's tolerance from theories to *hearers*. Artham kathamcit purusah kascit sampratipadyate: some person, somehow, arrives at the meaning — the double indefinite (kascit, kathamcit) is the doctrine: routes to comprehension vary irreducibly by hearer, occasion, and preparation; there is no single canonical processing-path any more than a single canonical analysis (2.38). And the differences among construals — samsrsta va vibhakta va, the meaning taken fused (whole-wise) or divided (part-wise) — are vakya-nibandhanah: fastenings *relative to the sentence*, ways of tying the one vakyartha down, not structures in it. The learned hearer may pass through elaborate division, the native speaker through none (2.13's avidvan/vidvan gradient); both arrive, and what they arrive at is identical. The verse quietly performs an important disarmament: the holist does not deny that part-wise comprehension *happens* — people do parse, gloss, construe — he denies that its happening reveals the meaning's composition; parsing is one kathamcit among others. With hearers, theories, and analyses all pluralized around a single invariant comprehension, the ground is fully prepared for naming that invariant — pratibha — and 2.40 begins the approach.
contested: vakya-nibandhanah we render 'fastenings relative to the sentence' (bheda-construals anchored to, and answerable to, the one vakya); the compound also admits 'grounded in the sentence' — either way the bhedas are relational, the vakya prior

2.40 unformalized

सोऽयमित्यभिसम्बन्धो बुद्ध्या प्रकम्यते यदा ।
वाक्यार्थस्य तदैकोऽपि वर्णः प्रत्यायकः क्वचित् ॥ ४० ॥

so'yamityabhisambandho buddhyā prakamyate yadā |
vākyārthasya tadaiko'pi varṇaḥ pratyāyakaḥ kvacit || 40 ||

When the connection 'that is this' is set moving by the mind, then sometimes even a single phoneme is the conveyor of the sentence-meaning.

The extreme datum that only the holist can accommodate, offered as the movement's QED. Comprehension consists in an abhisambandha of the form so 'yam — '*that* is *this*': the incoming utterance identified with a content already alive in the hearer (the recognition-structure of 1.124, pratyavamarsa, here at sentence scale). When the mind prakramyate, is set moving toward that identification — context primed, expectation shaped — tada ekah api varnah: then even a *single phoneme* conveys the entire vakyartha, kvacit, on occasion. The phenomenon is common coin: the shared joke's first syllable; the servant who moves at 'ga—'; the pupil for whom the sutra's opening sound unrolls the whole. For compositional theories this is catastrophe — one phoneme has no parts to compose with, no word-meanings to sum; the meaning conveyed (a full sentential content) is unconstructible from the vehicle. For the manifestation theory it is the expected limit-case: if sounds are manifestors of a whole already standing in the comprehender (2.18, 2.30), then *how much* sound is needed is an empirical matter of the hearer's preparedness — normally much, with priming little, in the limit one varna, the fan of 1.116 barely waved. The verse ends the movement by inverting 2.28's regress: the atomist could not build the word from his parts; the holist retrieves the sentence from less than one of them.
contested: kvacit guards the claim ('sometimes, somewhere') and we keep its modesty; the traditional illustrations vary, and some readers take the varna here as a pratyahara-style code-sound (technical abbreviation) rather than ordinary speech — the argumentative use is identical

2.41 · The Rival Semantics Examined unformalized

केवलेन पदेनार्थो यावानेवाभिधीयते ।
वाक्यस्थं तावतोऽर्थस्य तदाहुरभिधायकम् ॥ ४१ ॥

kevalena padenārtho yāvānevābhidhīyate |
vākyasthaṃ tāvato'rthasya tadāhurabhidhāyakam || 41 ||

Only so much meaning as is expressed by the word in isolation — of just that much meaning, they say, is the word when standing in a sentence the expressor.

The abhihitanvaya position — the Bhatta Mimamsa semantics, the holist's most serious rival — stated with scrupulous fairness at the head of a long examination. The thesis: a word's semantic contribution is *invariant*; kevalena padena, in isolation, 'gauh' expresses cow-in-general — and vakyastham, stationed in a sentence, it expresses tavatah eva, exactly that much, no more or differently. Sentence-meaning then arises downstream: words severally denote their meanings (abhihita), and the denoted meanings connect (anvaya) by expectancy, fitness, proximity. The picture is compositionality at full strength: fixed lexical deposits, syntax as the connective operation on them — the very picture 2.15 and 2.17 attacked, now given its own voice for systematic testing. The verses that follow (2.42-56) will let it and its variants state their best cases, and the objections will not be cheap: Bhartrhari's doxographic conscience — argue the rival to its strongest before dissolving it — is nowhere better displayed than in this stretch.

2.42 unformalized

सम्बन्धे सति यत्त्वन्यदाधिक्यमुपजायते ।
वाक्यार्थमेव तं प्राहुरनेकपदसंश्रयम् ॥ ४२ ॥

sambandhe sati yattvanyadādhikyamupajāyate |
vākyārthameva taṃ prāhuranekapadasaṃśrayam || 42 ||

But the surplus that arises when connection has taken place — that, resting on the several words together, they call the sentence-meaning itself.

The rival's account of what the sentence adds. Grant each word its fixed deposit (2.41); when sambandha occurs — the deposits connecting — anyad adhikyam upajayate: a *surplus* arises, something none of the words separately expressed: the relatedness itself, cow-as-being-brought, the state of affairs over and above its ingredient terms. Tam vakyartham eva prahuh: that surplus, they say, *is* the sentence-meaning — aneka-pada-samsraya, resting on the many words jointly, since no one of them delivers it. The account has genuine elegance: it explains compositional productivity (new sentences understood from old words: the deposits are old, only the connection is new) and gives the sentence a meaning of its own without positing a sentence-entity. Its exposed joint, which the sequel probes: the surplus is *generated* — it arises (upajayate) from the connecting of meanings — so its unity is derivative, an assembly-product; and everything Kanda II has argued about assembly (2.15's homeless generals, 2.28's non-adhering parts, 2.34's non-monotonicity) now stands waiting for it.
contested: whether 2.42 states the Bhatta abhihitanvaya specifically or a generic connection-semantics is debated; the adhikya ('surplus') formulation matches the Mimamsa doctrine of the sentence-meaning as vakyartha-over-padarthas closely enough that we read it so

2.43 unformalized

स त्वनेकपदस्थोपि प्रतिभेदं समाप्यते ।
जातिवत्समुदायेऽपि संख्यावत्कल्प्यते परैः ॥ ४३ ॥

sa tvanekapadasthopi pratibhedaṃ samāpyate |
jātivatsamudāye'pi saṃkhyāvatkalpyate paraiḥ || 43 ||

But that (sentence-meaning), though resting on many words, is brought to completion at each division; by others it is construed as like a universal in the aggregate, or like a number.

Two refinements of the surplus-doctrine, each trying to repair the assembly-problem. First (echoing 2.18 verbatim — pratibhedam samapyate): even granting the vakyartha rests on many words, it completes *at each*: the relational surplus is wholly present wherever it is present at all — a concession the connection-theorist is driven to by the phenomenology (comprehension does not feel like watching a relation assemble) and which quietly surrenders half his picture. Second, the type-theoretic repairs (paraih, by others): the sentence-meaning is jativat samudaye — like a *universal* residing in the aggregate: as cowhood is wholly in each cow yet 'of' the herd only via its members, the vakyartha inheres whole in the word-collection; or samkhyavat — like a *number*: as ten is a property of the group that no member has and no member lacks a share in generating, the sentence-meaning is a collective property, real, unitary, and irreducibly of-the-many. These are the opponent's best structures for a one-over-many meaning — and Bhartrhari has chosen them for their instructive failure: universals and numbers are precisely the entities whose one-in-many character is *unexplained* in the rival ontologies (the Vaisesika's own scandals), so the repair grounds the sentence's unity in structures as problematic as itself. The holist's alternative — the unity is primitive, the plurality constructed (2.30-31) — begins to look like the *cheaper* metaphysics.
contested: the referents shift rapidly and paraih is our pivot: we read three positions (per-division completion; jati-model; samkhya-model), the latter two as others' refinements — some read the whole verse as one position with two similes

2.44-45 unformalized

सर्वभेदानुगुण्यं तु सामान्यमपरे विदुः ।
तदर्थान्तरसंसर्गाद् भजते भेदरूपताम् ॥ ४४ ॥
भेदानाकाङ्क्षतस्तस्य या परिप्लवमानता ।
अवच्छिनत्ति संबन्धस्तां विशेषे निवेशयन् ॥ ४५ ॥

sarvabhedānuguṇyaṃ tu sāmānyamapare viduḥ |
tadarthāntarasaṃsargād bhajate bhedarūpatām || 44 ||
bhedānākāṅkṣatastasya yā pariplavamānatā |
avacchinatti saṃbandhastāṃ viśeṣe niveśayan || 45 ||

(44) Others know the sentence-meaning as a general, conformable to all the particulars; through contact with other meanings it takes on the form of the particular. (45) The relation delimits that hovering condition of it — of the general awaiting its particulars — settling it upon the particular.

The anvitabhidhana-adjacent refinement (the direction Prabhakara Mimamsa would systematize): meaning enters *general* and is *specified in construction*. On this view (apare, others), what a word first gives the sentence is a samanya conformable to every particular it might become — 'bring' awaits a bringable, 'cow' awaits a role — sarva-bheda-anugunyam, fitted to all differentiations, committed to none. Arthantara-samsargat, through contact with the other meanings, it bhajate bheda-rupatam, takes on particular form: 'cow' under 'bring' becomes cow-as-patient-of-bringing. (45) adds the dynamics with a remarkable image: before settlement the general exists in pariplavamanata — a *hovering*, a floating unfixity (the word suggests a boat riding loose) — bheda-anakanksatah... which the sambandha avacchinatti, cuts off, delimits, visese nivesayan, installing it in the particular. Comprehension as progressive fixation of hovering indeterminates: each word casts a shifting field of possibility that the others successively pin. Bhartrhari reports this with evident respect — it is the rival picture *closest* to his own (specification-in-context, meanings alive only in construction, 2.17), differing at one joint only: for this theorist the hovering generals are real semantic individuals that get connected; for Bhartrhari the 'generals' and their 'fixation' are both analytical shadows of one undivided vakyartha manifesting (2.18). The next verse probes the connective itself.
contested: the attribution is anachronistic either way (Prabhakara postdates); we mark the view as the ancestor-type of anvitabhidhana without the school-name. Pariplavamanata is the verse's coinage and our 'hovering' its nearest English

2.46 unformalized

कार्यानुमेयः संबन्धो रूपं तस्य न दृश्यते ।
असत्त्वभूतमत्यन्तमतस्तं प्रतिजानते ॥ ४६ ॥

kāryānumeyaḥ saṃbandho rūpaṃ tasya na dṛśyate |
asattvabhūtamatyantamatastaṃ pratijānate || 46 ||

The relation is inferable from its effect; a form of it is not seen. Therefore they acknowledge it to be utterly beyond thinghood.

The connective interrogated, and the rival semantics pays its ontological bill. Every connection-theory (2.42-45) trades on sambandha — but what *is* it? Karya-anumeyah: knowable only by inference from its effect (the unified comprehension); rupam tasya na drsyate — no form of it is ever presented; nobody perceives the relating, only the related-having-been. The theorist must therefore concede (pratijanate, they acknowledge) that the relation is atyantam asattva-bhutam — utterly beyond sattva, beyond thinghood: not an entity among entities but at most a structure attributed backward from its supposed work. Bhartrhari lets the concession stand without immediate comment because its dialectical yield is complete: the rival account of sentence-unity now rests on an imperceptible, non-thing connective inferred from the very unity it was invoked to explain — a circle with an invisible center. Against it, the holist's primitive (one self-manifesting vakyartha, 2.31) posits no machinery at all between items, since there are no items to connect. The verse also plants a flag the third kanda will march to: the Sambandha-samuddesa's full theory of relation as paratantrya, essence-less dependence, begins in this asattva-concession.
contested: who concedes — the rival driven to it, or Bhartrhari stating his own doctrine of relation through their mouths — is nicely undecidable, and the verse works as both refutation and foreshadowing; we read it as the former with the latter intended

2.47 unformalized

नियतं साधने साध्यं क्रिया नियतसाधना ।
स सन्निधानमात्रेण नियमः सन् प्रकाशते ॥ ४७ ॥

niyataṃ sādhane sādhyaṃ kriyā niyatasādhanā |
sa sannidhānamātreṇa niyamaḥ san prakāśate || 47 ||

The thing-to-be-accomplished is fixed relative to its means; the action has fixed means. That fixity, simply through proximity, shines forth as settled.

Deflation of the connective completed: what work was left for sambandha to do, niyama and sannidhi already do. The semantic roles come *pre-fixed*: sadhye sadhane niyatam — what is to be accomplished is fixed relative to its means (a bringing needs a brought, an agent, an instrument); kriya niyata-sadhana — the action-meaning arrives with its argument-places built in (the karaka-structure that Panini's grammar maps). Given meanings so structured, sannidhana-matrena — through *mere proximity*, the words' co-presence in one utterance — sa niyamah san prakasate: the fixity, already real, simply *shines forth as settled*. Note the verb: prakasate, manifestation-vocabulary — nothing is assembled; an antecedent structure is lit up when its expressors stand together. The connection-theorist's sambandha is thus idle: his 'relating operation' names nothing beyond meanings-with-slots meeting slot-fillers in proximity — and 'meeting in proximity' is an occasion for manifestation, not an act of synthesis. This is the holist absorbing the best insight of 2.44-45 (specification-in-construction) minus its ontology: expectancy (akanksa) is real as the *shape* of the one meaning, not as hooks between many. 2.48 immediately cashes this in grammatical coin.

2.48 unformalized

गुणभावेन साकाङ्क्षं तत्र नाम प्रवर्तते ।
साध्यत्वेन निमित्तानि क्रियापदमपेक्षते ॥ ४८ ॥

guṇabhāvena sākāṅkṣaṃ tatra nāma pravartate |
sādhyatvena nimittāni kriyāpadamapekṣate || 48 ||

The noun operates there as requiring, in the capacity of subordinate; the occasioning elements, as things-to-be-accomplished, await the action-word.

The karaka-architecture of the sentence, in one couplet — 2.47's abstract niyama shown as the actual grammar of Sanskrit. Nama, the noun, pravartate gunabhavena sakanksam: operates as *subordinate* (guna, the dependent member, 2.4's third mark) and as *requiring* — a noun is intrinsically unfinished, a role-filler seeking its role. The nimittani — the occasioning elements, the nominal contributors read as sadhyatvena, under the aspect of what-is-to-be-accomplished — kriyapadam apeksate: await the action-word. The verb is the sentence's gravitational center: nouns orbit it, their cases the names of their orbits (karakas), their expectancy the pull. This is why the akhyata-view (2.1's first answer) was *almost* right: the verb is indeed where sentence-unity is organized — wrong only in taking the organizer for the unity itself (the verb too is an abstraction from the one vakyartha; it is the *structure's* center, not the meaning's substance). The verse also explains, retroactively, why 2.4 defined the vakya as karma-pradhana: principality of action is the syntactic face of the meaning's one-ness — a single sadhya with its sadhanas is how an undivided content articulates when projected into word-roles (2.7).
contested: nimittani ('the occasioning elements') is loose — we take it as the nominal karaka-bearers collectively; some construe it narrowly as the pre-verbal conditions (the karakas qua causal factors of the action); the architecture asserted is the same

2.49 unformalized

सन्त एव विशेषा ये पदार्थेषु व्यवस्थिताः ।
ते क्रमादनुगम्यन्ते न वाक्यमभिधायकम् ॥ ४९ ॥

santa eva viśeṣā ye padārtheṣu vyavasthitāḥ |
te kramādanugamyante na vākyamabhidhāyakam || 49 ||

The particulars that stand settled in the word-meanings, really existent, are followed along in order — the sentence is not an expressor.

The opposite extreme now states its case: the sentence eliminated entirely (the padartha-realist's terminus, counterpart to 2.41's opening). On this view the visesas — the particular determinations that comprehension delivers — santa eva, *really existent*, stand already vyavasthita in the word-meanings themselves; comprehension kramad anugamyante, follows them along in sequence, one deposit after another; and the conclusion is drawn without blinking: na vakyam abhidhayakam — *the sentence expresses nothing*; there is no sentence-level semantics at all, only a parade of word-facts that the hearer strings. This is the mirror-image of the holism it opposes — where Bhartrhari dissolves the words into the sentence, this theorist dissolves the sentence into the words — and Bhartrhari grants it the courtesy of full statement because its existence proves something structurally: the middle positions (2.42-45), trying to have real word-deposits *and* a real sentence-surplus, are unstable, sliding under pressure toward one pole or this one. 2.50-52 will let this eliminativism run to its own dissolution: if only word-order strings the deposits, and order is no word (krama being time's property, not speech's), the view ends with no expressor anywhere — an instructive collapse the holist will collect.
contested: santa eva we read with visesah ('really existent particulars', the eliminativist's realism); it can be construed adverbially ('being so, they are followed'); the view's identity — late Mimamsa? a Nyaya strand? — is unrecorded, and the karika keeps it anonymous

2.50-51 unformalized

शब्दानां क्रममात्रे च नान्यः शब्दोऽस्ति वाचकः ।
क्रमो हि धर्मः कालस्य तेन वाक्यं न विद्यते ॥ ५० ॥
ये च संभविनो भेदाः पदार्थेष्वविभाविताः ।
ते सन्निधाने व्यज्यन्ते न तु वर्णेष्वयं क्रमः ॥ ५१ ॥

śabdānāṃ kramamātre ca nānyaḥ śabdo'sti vācakaḥ |
kramo hi dharmaḥ kālasya tena vākyaṃ na vidyate || 50 ||
ye ca saṃbhavino bhedāḥ padārtheṣvavibhāvitāḥ |
te sannidhāne vyajyante na tu varṇeṣvayaṃ kramaḥ || 51 ||

(50) And if (the sentence is) mere order of words, there is no further word to be the expressor; for sequence is a property of time — hence the sentence does not exist. (51) And the possible particulars, unrealized in the word-meanings, are manifested in co-presence; but this order is not in the phonemes.

The eliminativism of 2.49 driven to collapse, then the holist's harvest. (50): if what 'strings the deposits' is krama-matra, mere order, note what order is: kramo hi dharmah kalasya — sequence is a property of *time*, not of speech (2.24 established it is not even securely that); it is no sabda at all. So the eliminativist has: word-deposits, plus a non-verbal, non-entity ordering — nanyah sabdo 'sti vacakah, no further word exists to do the expressing of the sentence-content — tena vakyam na vidyate: the sentence, on his own premises, *does not exist*. The reductio is complete and Bhartrhari states it in the opponent's own voice, letting the position pronounce its own emptiness. (51) then gathers what survives: yes — sambhavino bhedah, the possible particulars, avibhavitah in the isolated word-meanings (the hovering generals of 2.44), vyajyante sannidhane, are *manifested in co-presence* (2.47's sannidhi, the verb again vyaj-): co-utterance occasions the determinate whole. But — na tu varnesu ayam kramah — this manifesting order is not *in* the phonemes: the sequence that matters is the mind's traversal (2.25), not an acoustic property; the utterance's serial body is occasion, never bearer. Between the two verses the last rival has dissolved and every salvageable insight has been re-housed in manifestation-theory: deposits→hovering possibles; connection→co-presence; order→cognitive traversal; sentence→the one manifested whole.
contested: 50's first pada is compressed (sabdanam krama-matre ca) and we supply the conditional; 51's closing na tu varnesv ayam kramah has also been read 'nor does this order (of manifestation) hold among the phonemes' — i.e. manifestation is not phoneme-by-phoneme — which is the same doctrine from the other side

2.52 unformalized

वर्णानां च पदानां च क्रममात्रनिवेशिनी ।
पदाख्या वाक्यसंज्ञा च शब्दत्वं नेष्यते तयोः ॥ ५२ ॥

varṇānāṃ ca padānāṃ ca kramamātraniveśinī |
padākhyā vākyasaṃjñā ca śabdatvaṃ neṣyate tayoḥ || 52 ||

The designation 'word', and the term 'sentence', settle upon mere order of phonemes and of words; wordhood is not accepted of those two.

The nominalist coda, reported as the last word of the partitive tradition — its most austere and most self-undermining form. On this view 'pada' and 'vakya' are mere labels: pada-akhya settles on krama-matra of phonemes, vakya-samjna on krama-matra of words — names for stretches of ordering, convenience-terms like 'forest' for trees or 'army' for men, and sabdatvam nesyate tayoh: *wordhood is not granted* to either — word and sentence are not themselves linguistic realities at all, only the phonemes are. The position is the exact linguistic analogue of the Buddhist prajnapti-doctrine (composites as designation-only), and Bhartrhari's economy in refuting it is total: he has already done so. If only phonemes are real and phonemes are meaningless severally (2.12, 2.28-29), and order is no entity (2.50), then meaning has no bearer whatsoever — the view saves ontology by losing language. The deeper irony he leaves implicit: the nominalist's 'mere designation' (akhya, samjna) is itself a *linguistic* act — his theory denies wordhood by means of words, dissolving the very instrument of its own statement. The catalogue of rivals thus ends where 2.28-29 predicted: partition, pursued honestly, terminates in the unnameable — and the way back up is the holist's (2.53-62, where the argument turns constructive).
contested: whose nominalism this is remains unrecorded; the prajnapti-parallel is structural, not an attribution. Some read the verse not as a rival but as the holist's own concession (word/sentence as designations upon the krama of manifestors) — the sequel's constructive turn tells against that reading

2.53 unformalized

समानेऽपि तु शब्दत्वे दृष्टः संप्रत्ययः पदात् ।
प्रतिवर्णं त्वसौ नास्ति पदस्यार्थमतो विदुः ॥ ५३ ॥

samāne'pi tu śabdatve dṛṣṭaḥ saṃpratyayaḥ padāt |
prativarṇaṃ tvasau nāsti padasyārthamato viduḥ || 53 ||

But even where wordhood is equal, comprehension is seen to arise from the word; it does not arise phoneme by phoneme — hence they know the meaning to belong to the word.

The constructive turn: from the rubble of the rivals, the minimal argument that meaning attaches to wholes. Concede the nominalist his level playing field — samane 'pi sabdatve, phonemes and words equally 'sound', nothing acoustic privileging either. The *facts of comprehension* (2.34's criterion, the school's constant court of appeal) then discriminate: samprataya drstah padat — understanding is observed to arise from the word; prativarnam tv asau nasti — it does not arise per-phoneme: 'g', 'au', 'h' severally deliver nothing, no fraction of cow arrives with each (1.84 already established the all-at-once arrival with the last sound). Atah padasya artham viduh: therefore the meaning is *the word's* — the semantic unit is fixed not by ontology or acoustics but by where comprehension actually happens. The argument's beauty is its neutrality: it uses no sphota-metaphysics, only the observable non-additivity of understanding — and it sets the pattern the next verses iterate upward: exactly as meaning belongs to the word over the phonemes' heads, it will belong to the sentence over the words' (2.60-61), by the same evidence, at the next grain. The ladder of 2.28 is being climbed in reverse, comprehension lighting each rung.
contested: none of weight; the verse is deliberately ecumenical — even a padartha-realist signs it, which is its strategic point

2.54 unformalized

यथाऽस्यावयवा वर्णा विना वाच्येन केनचित् ।
अर्थवन्तः समुदिता वाक्यमप्येवमिष्यते ॥ ५४ ॥

yathā'syāvayavā varṇā vinā vācyena kenacit |
arthavantaḥ samuditā vākyamapyevamiṣyate || 54 ||

As its constituent phonemes, without any expressible content of their own, become meaning-possessed when combined — so also is the sentence accepted.

The parity-ladder, ascending. The word's phonemes are severally anarthaka (2.12, 2.53) — vina vacyena kenacit, with no expressible of their own — yet samuditah, combined, they are arthavantah: the combination is where meaning lives. Vakyam apy evam isyate: accept the same of the sentence — its words, severally, stand to the vakyartha as phonemes to the padartha: contributing no independent deposits, meaning-possessed only as the ensemble. The argument is strictly ad hominem in structure and the more binding for it: every school grants the phoneme-to-word case (no one thinks 'g' means a third of cow); parity of evidence (comprehension arising only at the whole, 2.53) then forces the word-to-sentence case on pain of arbitrariness — what non-question-begging difference marks the levels? The partitive theorist who resists must find in words a semantic self-sufficiency he already denies phonemes, against identical distributional facts (words 'meaning something alone' being citation-artifacts, 2.17). The holism thus propagates up the grains by the opponent's own concessions — the strategy 2.10 inaugurated (apoddhara accepted for stems forces it for words) here reaching the sentence.
contested: isyate ('is accepted/desired') marks this as doctrine being *ratified*, not merely reported; some hear in it lingering doxography — but 2.55-56 continue in the same ratifying voice and we read the siddhanta as speaking

2.55 unformalized

अनर्थकान्युपायत्वात् पदार्थेनार्थवन्ति वा ।
क्रमेणोच्चरितान्याहुर्वाक्यार्थं भिन्नलक्षणम् ॥ ५५ ॥

anarthakānyupāyatvāt padārthenārthavanti vā |
krameṇoccaritānyāhurvākyārthaṃ bhinnalakṣaṇam || 55 ||

Meaningless (severally), by way of being means; or meaning-possessed through the word-meaning: uttered in sequence, they declare them to express a sentence-meaning of distinct character.

The two available descriptions of the words, and the invariant they share. Within the ratified holism, what shall we *say* words are? Either anarthakani upayatvat — meaningless severally, their standing that of *means* (the upaya-doctrine of 2.38: taken up to be discharged, semantic scaffolding); or arthavanti padarthena — meaning-possessed, if one insists, but through the word-meaning as analytically assigned (the abstraction of 2.10 credited back to them as a facon de parler). The choice is bookkeeping (1.74, 2.39): on either description, kramena uccaritani — uttered in their sequence — they express a vakyartha *bhinna-laksanam*: of *distinct character*, differently-marked — a meaning whose nature is other than the word-meanings' own, not their sum, aggregate, or relational surplus (contra 2.42) but a content of a different logical kind. The phrase is the movement's yield in two words: sentence-meaning differs from word-meanings not in *size* but in *kind* — which is why no operation on padarthas (adding, connecting, specifying) was ever going to construct it, and why the kanda must next say positively what kind of thing it is. The answer — indivisible, instantaneous, pratibha — is now one movement away.
contested: bhinna-laksanam we press as 'of distinct (logical) character'; a flatter reading — 'having a different definition' — is available; the argumentative arc from 2.34 onward (non-constructibility in principle) supports the pressed reading

2.56 unformalized

नित्यत्वे समुदायानां जातेर्वा परिकल्पने ।
एकस्यैकार्थतामाहुर्वाक्यस्याव्यभिचारिणीम् ॥ ५६ ॥

nityatve samudāyānāṃ jātervā parikalpane |
ekasyaikārthatāmāhurvākyasyāvyabhicāriṇīm || 56 ||

Whether the aggregates are held eternal, or the universal is what is construed, they declare the single sentence's single-meaningedness to be without deviation.

The invariance-result, sealed across the surviving metaphysical options. Two construals of the sentence-unit remain live from the wreckage: nityatve samudayanam — the aggregates themselves eternal (each sentence-whole a nitya unit, the samudaya elevated from heap to real); or jater va parikalpane — the universal-construal (the jati-sphota line of 1.93, 2.43: the sentence as type). Bhartrhari's point: choose either — ekasya ekarthatam ahur vakyasya avyabhicarinim — the *one* sentence's *one*-meaningedness holds without deviation (avyabhicarin, the term for a mark that never strays from what it marks). The eka-artha of 2.4's definition, flagged there as the loaded fourth condition, is here discharged at full metaphysical weight and shown *robust*: unitary meaning is not hostage to the token/type dispute, exactly as Kanda I's manifestation-doctrine was not hostage to the substrate-dispute (1.107) or the nitya/karya dispute (1.28). The method is constant across the work: fight to the death over the structure (undividedness), stay serene over the ontology beneath it. With unity thus secured against all comers, the kanda can afford its most disarming gesture — 2.57-58 next, where the whole war is recast as two legitimate directions of description.
contested: none textually; the pairing samudaya-nityatva / jati-parikalpana is our resolution of the compressed alternatives, following the earlier appearances of both options (2.43, 1.93)

2.57-58 unformalized

अभेदपूर्वका भेदाः कल्पिता वाक्यवादिभिः ।
भेदपूर्वानभेदांस्तु मन्यन्ते पददर्शिनः ॥ ५७ ॥
पदप्रकृतिभावश्च वृत्तिभेदेन वर्ण्यते ।
पदानां संहिता योनिः संहिता वा पदाश्रया ॥ ५८ ॥

abhedapūrvakā bhedāḥ kalpitā vākyavādibhiḥ |
bhedapūrvānabhedāṃstu manyante padadarśinaḥ || 57 ||
padaprakṛtibhāvaśca vṛttibhedena varṇyate |
padānāṃ saṃhitā yoniḥ saṃhitā vā padāśrayā || 58 ||

(57) The differences are constructed as posterior to non-difference by the sentence-theorists; the word-seers, however, hold the non-differences to be posterior to difference. (58) And the word-as-base relation is described according to the difference of functioning: either the continuous utterance is the womb of the words, or the continuous utterance rests upon the words.

The whole controversy, mapped from above — Bhartrhari's most explicit statement that the war of the kanda is between two *directions of construction*, not two sets of facts. (57): the vakya-vadins (his own side) hold bhedah abheda-purvakah kalpitah — differences are constructed, and construction runs *from* non-difference: the undivided sentence is primitive, words are carved (2.10). The pada-darsins, word-seers, hold abhedan bheda-purvan — non-differences (the sentence-unities) posterior to difference: words primitive, sentences built. Each side's primitives are the other's constructs; the datum — speech understood — is common. (58) then gives the dispute its technical grammatical form, where it had lived all along: samhita, the continuous utterance (the sandhi-fused stream actually spoken), versus the padas (the word-forms isolated by analysis, as in the pada-patha of Vedic recitation against its samhita-patha) — is the stream the *yoni*, the womb, of the words (padanam samhita yonih: analysis extracts them from it), or does the stream *rest on* the words (samhita va padasraya: recitation fuses them into it)? Vrtti-bhedena varnyate: described either way *according to the difference of functioning* — the reciter building samhita from memorized padas, and the analyst deriving padas from heard samhita, are both real procedures, and each grounds one metaphysics. The echo of 1.74 is deliberate and structural (what is primary for some is for others the reverse); but where 1.74 left the antinomy standing, 2.59-62 will now argue that the two directions are *not* symmetric after all — eternality breaks the tie, and it breaks toward the stream.
contested: 58's alternatives compress the samhita/pada-patha background, which we supply as the tradition does; pada-prakrti-bhava is the technical name of the issue (the Rk-Pratisakhya's own phrase), and whether prakrti there means 'base' or 'source' is itself the ancient form of the same dispute

2.59 unformalized

पदाम्नायश्च यद्यन्यः संहिताया निदर्शकः ।
नित्यस्तत्र कथं कार्यं पदं लक्षणदर्शनात् ॥ ५९ ॥

padāmnāyaśca yadyanyaḥ saṃhitāyā nidarśakaḥ |
nityastatra kathaṃ kāryaṃ padaṃ lakṣaṇadarśanāt || 59 ||

And if the word-recension is something else, an indicator of the continuous text — where eternality holds, how could the word be a product, given that its character is (mere) indication?

The tie of 2.58 broken, on the tradition's own deepest commitment. Consider the pada-amnaya, the word-recension — the pada-patha, that venerable apparatus in which the Veda is transmitted word-by-word, sandhi dissolved. What is its standing? Anyah samhitaya nidarsakah: it is *other* than, and an *indicator* of, the samhita — a pointer toward the continuous text, a laksana (the verse's own word: its character is that of showing). Now the argument: nitye — where eternality holds, and for the Veda it holds by the whole tradition's confession (1.132, 1.144) — katham karyam padam: how could the word be a *product*? If the eternal object is the samhita (and it is the samhita that is *recited*, the pada-patha being a human teaching-instrument attributed to named authors — Sakalya's, the tradition says, and what has an author is karya, 1.132), then the padas are downstream: extracted, indicated, didactic — exactly the apoddhara the holist claimed (2.10). The word-seer's direction of construction would make the eternal rest on the authored; the sentence-seer's makes the authored point at the eternal. Only one direction is consistent with the tradition's own reverence — and it is the holist's. The argument is ad verecundiam only in dress; structurally it is 2.38's upaya-doctrine given a philological body: the pada-patha is the paradigm ladder, and everyone already climbs it as one.
contested: the syntax of laksana-darsanat is compressed: we take it as 'given that its (the pada's) character is seen to be indication'; it can be read 'because (only) the laksana (the indicator) is seen (to be karya)' — same force. That the verse presupposes the pada-patha's authored status is the traditional understanding and the argument needs it

2.60 unformalized

प्रतिवर्णमसंवेद्यः पदार्थप्रत्ययो यथा ।
पदेष्वेवमसंवेद्यं वाक्यार्थस्य निरूपणम् ॥ ६० ॥

prativarṇamasaṃvedyaḥ padārthapratyayo yathā |
padeṣvevamasaṃvedyaṃ vākyārthasya nirūpaṇam || 60 ||

As the cognition of the word-meaning is not separately felt at each phoneme, so in the words the discerning of the sentence-meaning is not separately felt.

The phenomenological parity-argument, stated in its final polished form — the experiential twin of 2.54's structural ladder. At the word-level: prativarnam asamvedyah padartha-pratyayah — the cognition of the word's meaning is *not felt phoneme by phoneme*; introspect as one may, there is no experience of one-third-of-cow at 'g' (1.83's anupakhyeya cognitions: real perhaps as process, absent as experience). Padesu evam — in the words, likewise: asamvedyam vakyarthasya nirupanam — the discerning of the sentence-meaning is not separately felt word-by-word. This is an empirical claim about comprehension's phenomenology, and it is carefully calibrated: not that nothing happens at each word (2.18 said the whole re-completes there; 1.90 mapped the part-seizures), but that no *separable installment of the sentence-meaning* is experienced as such — what one feels at 'Devadatta' mid-sentence is the still-resolving whole, not a detachable Devadatta-meaning banked and held. The hearer's experience is of one content clarifying, never of contents accumulating — and since comprehension is the pramana here (2.34, 2.53), how it presents *is* the primary evidence of how it is. The rival must now explain away the phenomenology; the holist merely describes it.
contested: asamvedya ('not separately felt') is the load-bearing term: the objector may grant unfelt processing while denying that unfelt entails unreal — Bhartrhari's implicit reply is the pramana-point (2.34): what never shows in comprehension cannot be *comprehension's* structure, only a theory's

2.61 unformalized

वाक्यार्थः संनिविशते पदेषु सहवृत्तिषु ।
यथा तथैव वर्णेषु पदार्थः सहवृत्तिषु ॥ ६१ ॥

vākyārthaḥ saṃniviśate padeṣu sahavṛttiṣu |
yathā tathaiva varṇeṣu padārthaḥ sahavṛttiṣu || 61 ||

The sentence-meaning enters into the words as they function together; just so does the word-meaning into the phonemes functioning together.

The positive image that replaces deposit-and-assembly, completing 2.60's negative. Vakyarthah samnivisate padesu saha-vrttisu: the sentence-meaning *enters into* — installs itself in, comes to reside in — the words, and the words so entered are saha-vrttisu, *functioning together*: the locative absolute carries the doctrine — only in joint functioning, never severally, are the words the meaning's seat. And downward the same: padarthah into the phonemes saha-vrttisu. The direction of predication is the whole point, and it inverts the rival's at every joint: meaning is not *built up from* the units; it *descends into* them — the whole, primitively real (2.30-31), takes lodging in its manifestors as they operate in concert, the way (the system's oldest images) fire lodges in the churned wood or the reflection in the water. A word 'has meaning', on this picture, in the derivative sense in which a windowpane 'has' the light that fills it while the shutter is open: really present, wholly borrowed, gone when the joint functioning stops (citation, 2.17). Samnivisate also quietly answers where the 'surplus' of 2.42 went: nothing arises over the words; something antecedent *arrives in* them.
contested: none of substance; the verse's force is entirely in its verb and its locative — both are as we render them

2.62 unformalized

सूक्ष्मं ग्राह्यं यथाऽन्येन संसृष्टं सह गृह्यते ।
वर्णोऽप्यन्येन वर्णेन संबद्धो वाचकस्तथा ॥ ६२ ॥

sūkṣmaṃ grāhyaṃ yathā'nyena saṃsṛṣṭaṃ saha gṛhyate |
varṇo'pyanyena varṇena saṃbaddho vācakastathā || 62 ||

As a subtle graspable, conjoined with another, is grasped along with it — so too a phoneme, bound up with another phoneme, is an expressor.

A concessive coda: even the phoneme recovers a kind of expressorhood — the right kind — under the completed doctrine. The model: a suksma grahya, a subtle perceptible — the tradition's examples are the faint scent, the thin thread, the soft tone — too slight to be grasped alone, is grasped anyena samsrstam saha, conjoined-with-and-alongside another: perception carries the subtle in on the back of the salient. So the varna: samsbaddho 'nyena varnena, bound up with its fellows, vacakas tatha — it *is* an expressor, in that conjoined mode: not the bearer of a meaning-fragment (2.53 stands), but a genuine co-carrier of the whole that only the ensemble delivers. The verse does quiet justice-work at the close of the movement: 2.40 showed one phoneme sufficing (primed manifestation); 2.53 showed no phoneme sufficing (no per-phoneme deposits); the apparent tension resolves in the conjunction-model — expressive power is real at every grain and *inalienably collective*, the phoneme expressing as a finger lifts: only as the hand's. With this, the descent-and-parity movement (2.53-62) closes: meaning belongs to wholes, installs itself downward through joint functioning, and every lesser unit shares expressorhood exactly to the degree it shares the functioning. The kanda now stands ready for its center: what the one meaning is, and how it flashes.
contested: the simile's base (what the 'subtle graspable' is) is unstated and the traditional examples vary; nothing turns on the choice — the samsarga-structure (grasped-along-with) is the entire cargo

2.63 unformalized

पदस्योच्चारणादर्थो यथा कश्चिन्निरूप्यते ।
वर्णानामपि सान्निध्यात्तथा सोऽर्थः प्रतीयते ॥ ६३ ॥

padasyoccāraṇādartho yathā kaścinnirūpyate |
varṇānāmapi sānnidhyāttathā so'rthaḥ pratīyate || 63 ||

As some meaning is discerned from the uttering of the word, so from the proximity of the phonemes that same meaning is apprehended.

A bridging verse, securing the levels against a last confusion before the great exegetical excursus begins. Meaning is discerned 'from the word's utterance' — and the word's utterance just *is* the phonemes in sannidhya, co-presence (2.47, 2.51). The two descriptions name one event: no stage exists at which phoneme-proximity has done its work and a further word-act follows; apprehension from the word and apprehension from the concerted phonemes are the same apprehension under coarser and finer description. The verse blocks a reifying misstep the ladder-arguments (2.53-62) might invite — as if the word were a second station where meaning is *again* received. There is one reception, describable at any grain; the grains are the analyst's (2.10), the reception is the hearer's. With the levels thus flattened into one manifestation-event, Bhartrhari turns (2.64ff.) to the hardest test his holism faces: the Mimamsa science of ritual exegesis, whose entire apparatus appears to run on exactly the word-level meanings he has dissolved.

2.64-65 · The Exegetical Test: Word-Meanings in Ritual Science unformalized

प्राप्तस्य यस्य सामर्थ्यान्नियमार्था यतः श्रुतिः ।
तेनात्यन्तं विशेषण सामान्यं यदि बाध्यते ॥ ६४ ॥
यजेतेति यदा द्रव्यं प्राप्तं सामर्थ्यलक्षण्म् ।
व्रीहिश्रुत्या निवर्त्तेत न स्यात्प्रतिनिधिस्तदा ॥ ६५ ॥

prāptasya yasya sāmarthyānniyamārthā yataḥ śrutiḥ |
tenātyantaṃ viśeṣaṇa sāmānyaṃ yadi bādhyate || 64 ||
yajeteti yadā dravyaṃ prāptaṃ sāmarthyalakṣaṇm |
vrīhiśrutyā nivartteta na syātpratinidhistadā || 65 ||

(64) When something is already obtained by capacity, and the explicit text serves (only) restriction — if by that utterly particular (text) the general were annulled, (65) then, when 'let him sacrifice' has obtained substance in general by the mark of capacity, the substance would be excluded by the word 'rice-grains' — and there would then be no substitute.

The test-case, drawn from the deep machinery of Mimamsa: the doctrine of pratinidhi, ritual substitution. The injunction vriihibhir yajeta, 'let him sacrifice with rice-grains'. Analysis: yajeta alone, by samarthya, capacity (a sacrifice *needs* some substance), already obtains dravya-in-general; the explicit sruti 'vriihi' then serves niyama, restriction — it selects *which* substance, among the antecedently-possible, shall be used. Now the argument (64): if the particular utterance *annulled* the general (atyantam badhyate) — if 'rice' obliterated the obtained substance-in-general rather than restricting within it — then (65) when rice is unavailable, na syat pratinidhih: no substitute could be licensed. But the ritual science *does* license substitutes (nivara-grains when vriihi fails — the standing doctrine): therefore the general survives inside the particular; 'rice' specifies substance-in-general without destroying it. The exegetical practice thus *presupposes* a layered semantics — general meaning obtained by capacity, particular meaning by explicit text, the second restricting, not replacing, the first. The relevance to the kanda's war is exact and double-edged, and Bhartrhari will work both edges: the layering looks like word-meaning realism (the Mimamsaka's reading), yet the *general's survival under specification* is precisely the holist's hovering-samanya (2.44-45) — and 2.66 begins extracting that second edge.
contested: the transcription's samarthya-laksanam (65) we read as 'obtained by the mark of capacity'; the argument is the classical pratinidhi-nyaya of the Mimamsa (cf. the sastra's treatment of substitution), which the karikas compress to near-stenography — our expansion follows the traditional case

2.66 unformalized

तस्माद् व्रीहित्वमधिकं व्रीहिशब्दः प्रकल्पयन् ।
द्रव्यत्वमविरुद्धत्वात् प्राप्त्यर्थः सन्न बाधते ॥ ६६ ॥

tasmād vrīhitvamadhikaṃ vrīhiśabdaḥ prakalpayan |
dravyatvamaviruddhatvāt prāptyarthaḥ sanna bādhate || 66 ||

Therefore the word 'vrihi', positing riceness as the surplus, being (itself) for the sake of what is (already) obtained, does not annul substancehood, since there is no conflict.

The resolution of the test-case, and its terms are the holist's. What does 'vrihi' in the injunction *do*? It prakalpayati adhikam vriihitvam — posits riceness as the *addition*, the specific difference — while dravyatvam na badhate: substancehood, already obtained by capacity (2.64), stands unannulled, avirudhatvat, since genus and specification do not conflict: rice *is* a substance; the particular arrives as a determination *of* the general, not a rival to it. The exegetical rule falls out: where the explicit text specifies within the obtained, substitution survives (the general remains to be re-specified when rice fails); where it *conflicts*, annulment (badha) operates. Note what this concedes and what it wins. Conceded: semantic stratification is real — comprehension does distinguish the obtained-general from the text-specific (the Mimamsaka's layering, 2.64). Won: the strata behave exactly as the hovering-account (2.44-45) said — the general persisting *in* its specification, delimited, not discarded (2.15's trilemma dissolved in ritual practice itself: nothing was thrown away; the hovering was pinned). The opponent's own science, run on his own cases, exhibits the holist's semantics. The next verses (2.67-69) generalize the mechanism: specification-without-annulment, and the principled *non-intention* (avivaksa) of what a word could but does not mean here.

2.67-68 unformalized

तेन चापि व्यवच्छिन्ने द्रव्यत्वे सहचारिणि ।
असम्भवाद् विशेषाणां तत्रान्येषामदर्शनम् ॥ ६७ ॥
न च सामान्यवत्सर्वे क्रियाशब्देन लक्षितः ।
विशेषा नहि सर्वेषां सतां शब्दोऽभिधायकः ॥ ६८ ॥

tena cāpi vyavacchinne dravyatve sahacāriṇi |
asambhavād viśeṣāṇāṃ tatrānyeṣāmadarśanam || 67 ||
na ca sāmānyavatsarve kriyāśabdena lakṣitaḥ |
viśeṣā nahi sarveṣāṃ satāṃ śabdo'bhidhāyakaḥ || 68 ||

(67) And even when, by that (word), the co-occurrent substancehood is delimited, other particulars are not seen there, because they are not possible; (68) nor are all the particulars indicated by the action-word, as the general is — for a word is not the expressor of everything that exists.

Two boundary-theorems for the specification-mechanism of 2.66. First (67): when 'vrihi' delimits its sahacarin, its co-occurrent substancehood, the *other* particulars that substance-in-general might have taken (barley, millet) are simply *not seen* there — asambhavat, because after specification they are no longer possible; they do not linger as suppressed candidates needing annulment. Specification closes possibilities without expending badha on each: the pinned general (2.45) just *is* general-as-riced, the alternatives dead by delimitation, not by execution. Second (68): the converse boundary — the action-word (yajeta), which obtains the general by capacity, does *not* thereby indicate all the particulars under it: na hi sarvesam satam sabdo 'bhidhayakah — *a word is not the expressor of everything that exists* (a maxim with reach far beyond ritual: expression is selective by essence; to mean the genus is not to mean, even faintly, each of its members). Between them the theorems fix the semantics' shape: upward, specification does not drag the whole genus-tree into meaning; downward, the genus-word does not scatter into its instances. Meaning moves at the grain the utterance sets — which is the anti-compositionalist moral in exegetical dress: content is not inherited along ontological lines (genus-member, whole-part) but fixed by the utterance's own act.
contested: 68's samanyavat we read 'as the general (is indicated)' — the genus is indicated, its members are not; some construe 'like (words for) universals' as a comparison-class. The maxim of 68d is textually secure and is the verses' exportable yield

2.69 unformalized

शुक्लादयो गुणाः सन्तो यथा तत्राविवक्षिताः ।
तथाऽविवक्षा भेदानां द्रव्यत्वसहचारिणाम् ॥ ६९ ॥

śuklādayo guṇāḥ santo yathā tatrāvivakṣitāḥ |
tathā'vivakṣā bhedānāṃ dravyatvasahacāriṇām || 69 ||

As the qualities, white and the rest, though present, are unintended there — so is there non-intention of the particulars that co-occur with substancehood.

The doctrine of avivaksa — principled non-intention — completing the boundary-theorems with the concept the whole exegetical tradition runs on. The rice-grains actually used are *white* (or red, or old); these gunas are santah, really present — and avivaksitah, *unintended*: the injunction 'with vrihi' says nothing of them; a sacrificer with discolored rice has not deviated. Presence in the object is one thing; presence in the meaning another — vivaksa, speaker's intention (the term that settled 1.155), is what admits features into the said, and its refusals are as constitutive as its grants. So too (the verse generalizes) all the particulars riding with substancehood: co-occurrent, real, unmeant. For the kanda's argument this is the last piece of the exegetical semantics and the most holist of all: what a sentence means is fixed neither by its words' full lexical potential nor by its objects' full property-set, but by the one intentional act the utterance embodies — the vakyartha as a *selection*, whole and indivisible, from which the analyst's 'word-meanings' inherit exactly the determinations the act granted and no others (sato 'vivaksa, 2.136 of Kanda I already named the canon). The Mimamsa exegete, master of avivaksa, has been practicing sentence-holism all along.
contested: none of substance; the verse applies 1.136's canon and the tradition's standing doctrine — its work is consolidation

2.70 unformalized

असंन्निधौ प्रतिनिधिर्मा भून्नित्यस्य कर्मणः ।
काम्यस्व वा प्रवृत्तस्य लोप इत्पपदिश्यते ॥ ७० ॥

asaṃnnidhau pratinidhirmā bhūnnityasya karmaṇaḥ |
kāmyasva vā pravṛttasya lopa itpapadiśyate || 70 ||

Lest there be a substitute in the absence (of what the text names), for the obligatory rite — or for an undertaken optional one — 'omission' is (instead) prescribed.

A precision from the casuistry, marking where substitution's license ends — and thereby what its license was. Some textual elements tolerate no pratinidhi: where what the text names is absent and the element was of the annulling kind (2.66's contrast-class — specification that *conflicts* rather than restricts, or an element textually essential), the science prescribes not substitution but lopa, omission — for the nitya rite (obligatory, which must proceed: it proceeds *without* the element) or the kamya rite already undertaken (optional, but binding once begun). The theoretical cargo: substitution was never a general permission to improvise; it was licensed *exactly by* the survival of the general under specification (2.64-66) — where no general survives (the element being sruti-essential, not capacity-obtained), nothing remains to re-specify, and the only options are omission or abandonment. The casuistry thus confirms the semantics from its negative side: the exegetical operations track, case by case, *which layer of meaning the text's authority entered at* — capacity-obtained layers flex (substitute), text-fixed layers do not (omit). A hermeneutics this fine-grained about layers is the kanda's best witness that comprehension itself is layered manifestation, not flat deposit-summing.
contested: the verse's compressed alternatives (ma bhut pratinidhih / lopa upadisyate) we resolve per the standard Mimamsa doctrine of lopa vs. pratinidhi; the transcription's ity(a)padisyate we read as upadisyate

2.71 unformalized

विशिष्टैव क्रिया येन वाक्यार्थः परिकल्प्यते ।
द्रव्याभावे प्रतिनिधौ तस्य तत्स्यात् क्रियान्तरम् ॥ ७१ ॥

viśiṣṭaiva kriyā yena vākyārthaḥ parikalpyate |
dravyābhāve pratinidhau tasya tatsyāt kriyāntaram || 71 ||

For one by whom the sentence-meaning is construed as the qualified action itself — for him, when substance fails and a substitute stands in, that would be a different action.

The exegetical dossier turned, at last, into a direct argument between the kanda's live semantic theories. Consider the theorist for whom the vakyartha is visista eva kriya — the *qualified action as such*: not action-plus-qualifications assembled, but the single qualified whole (rice-sacrificing as one indivisible act-meaning). For him, the verse observes, substitution has a startling consequence: dravyabhave pratinidhau — rice failing, nivara standing in — tat syat kriyantaram: the performed rite is *a different action* (nivara-sacrificing, not rice-sacrificing-with-replacement). The whole being indivisible, vary anything and you have another whole. Yet the ritual science treats the substitute-performance as *the same rite* discharged (the sacrificer has fulfilled vriihibhir yajeta, not violated it) — the practice sides with layering (2.64-66): the rite's identity rides the capacity-obtained stratum (sacrifice-with-substance), which the substitute preserves. The verse is Bhartrhari at his most honest: it is *his own* holism, in its naive form (the undifferentiated qualified-whole), that the practice embarrasses — and the embarrassment is instructive, forcing the refinement the kanda has been building anyway: the one vakyartha is not a *blank* unity but a unity with internal svabhava-structure (obtained strata, specifications, avivaksa: 2.64-70) — undivided, yet not undifferentiated; one act of meaning whose articulation is real *as* articulation-of-one. Holism that cannot say 'same rite, different rice' is too crude; the layered manifestation-holism now on the table can.
contested: whether 2.71 is an objection Bhartrhari answers (as we read: it refines his holism) or a position he endorses against the substitution-practice (making lopa the 'correct' outcome) is a genuine crux; the dossier's drift (2.64-70 validating the practice) supports the refining reading

2.72 unformalized

विज्ञातार्थं पदं यच्च तदर्थे प्रतिपादिते ।
पिकादि यदविज्ञातं तत्किमित्यनुयुज्यते ॥ ७२ ॥

vijñātārthaṃ padaṃ yacca tadarthe pratipādite |
pikādi yadavijñātaṃ tatkimityanuyujyate || 72 ||

And when a word whose meaning is known has had its object conveyed, one like 'pika', not known — 'what is that?' — is inquired after.

A small phenomenological datum with a long reach: the experience of the unknown word. In a sentence otherwise understood — vijnata-artham padam, the known words having conveyed their object — an unknown item ('pika', the tradition's stock example: a Vedic bird-name opaque to later speakers) arrests the hearer: kim iti anuyujyate, 'what is that?' is asked. The datum cuts several ways and Bhartrhari wants them all. Against flat holism: the inquiry is *word-shaped* — the hearer localizes his ignorance to 'pika', which he could not do if comprehension had no word-articulated structure (2.71's lesson again: the one meaning has real joints). Against deposit-atomism: the rest of the sentence *was* understood — comprehension did not wait on the missing deposit but completed around the gap, holding a determinate *place* open (a something-the-sentence-predicates, awaiting species): exactly the hovering-general structure (2.44), here generated by the sentence for a word, inverting the usual direction. The question 'what is pika?' is thus itself evidence of the layered vakyartha: the sentence has already given 'pika' its role-meaning (some bird-agent, say); what is sought is only the residue. Word-meaning as *role-first, filler-later* — the sentence assigning, the lexicon merely filling — is the datum's moral, and it is the holist's.
contested: none textually; the pika-example's traditional setting (nirukta discussions of unknown Vedic words) is assumed rather than stated by the karika

2.73-74 unformalized

सामर्थ्यप्रापितं यच्च व्यक्त्यर्थमनुषज्यते ।
श्रुतिरेवानुषङ्गे सा बाधिका लिङ्गवाक्ययोः ॥ ७३ ॥
अप्राप्तो यस्तु शुक्लादिः सन्निधानेन गम्यते ।
स यत्नप्रापितो वाक्ये श्रुतिधर्मविलक्षणः ॥ ७४ ॥

sāmarthyaprāpitaṃ yacca vyaktyarthamanuṣajyate |
śrutirevānuṣaṅge sā bādhikā liṅgavākyayoḥ || 73 ||
aprāpto yastu śuklādiḥ sannidhānena gamyate |
sa yatnaprāpito vākye śrutidharmavilakṣaṇaḥ || 74 ||

(73) And that which, obtained by capacity, is attached for the sake of the individual — in such attachment it is the explicit text that annuls both indicatory mark and syntactic connection; (74) but a 'white' and the like, not (antecedently) obtained, understood through proximity — that, brought in by effort into the sentence, has a character quite distinct from the explicit text.

The hierarchy of evidences — the Mimamsa's celebrated ranking of sruti, linga, vakya (explicit text, indicatory mark, syntactic connection) — cited at its decisive joint and read semantically. (73): where meanings obtained by capacity are being attached to their individuals (anusanga, carrying-over of elements across clauses), conflicts are settled by rank: srutir eva badhika lingavakyayoh — the explicit text annuls mark and connection; what is *said* outranks what is *indicated* outranks what is merely *construed together*. (74) then marks the other grade: a qualifier like 'white', aprapta, not obtained by any capacity, gathered sannidhanena, by mere proximity — yatna-prapita, brought in by effort — is sruti-dharma-vilaksana, of a character quite distinct from explicit-text status: present in the comprehension, lowest in authority, first to yield. For the kanda, the ranking is one more stratigraphy of the single vakyartha (2.71's refinement compounding): comprehension is one, but its contents carry *grades of textual warrant* — said, marked, construed, proximate — and the exegete's conflict-rules are a map of those grades. A semantics of flat deposits has no room for warrant-grades (every word a deposit, all deposits equal); the manifestation-holism has exactly room — the one meaning arrives with its determinations differentially anchored, and badha among evidences is the grammar of that anchoring. The opponent's crown jewels keep testifying for the defense.
contested: 73 compresses the sruti-linga-vakya doctrine (the classical exposition is the Mimamsa's krama-sutra context) and our reading supplies that frame; 74's yatna-prapita ('brought in by effort') is read by some as pejorative ('dragged in'), by others as neutral ('secured by interpretive labor') — the grade-distinction is the point either way

2.75 unformalized

अभिन्नमेव वाक्यं तु यद्यभिन्नार्थमिष्यते ।
तत्सर्वं श्रुतिभूतत्वान्न श्रुत्यैव विरोत्स्यते ॥ ७५ ॥

abhinnameva vākyaṃ tu yadyabhinnārthamiṣyate |
tatsarvaṃ śrutibhūtatvānna śrutyaiva virotsyate || 75 ||

But if the sentence is accepted as undivided with undivided meaning, then, since all of that has the standing of explicit text, nothing will be in conflict with the text itself.

The dossier's first summit: the holist collects the entire hierarchy at a stroke. If — as the kanda has argued — the vakya is abhinnam eva, undivided, its artha abhinna, one, then tat sarvam sruti-bhutam: *everything* in the sentence-meaning has the standing of sruti, of the explicitly said — na srutya eva virotsyate, nothing in it can conflict with the text, being text through and through. The ranking of 2.73-74 (said > marked > construed > proximate) is thereby revealed as an artifact of *analysis*: grades of warrant attach to the analyst's extracted items (his lingas, his construals, his carried-over qualifiers) — to the padartha-layer, the scaffolding (2.10, 2.38); the meaning itself, taken whole as uttered, is all equally *said*. The exegete's conflicts are conflicts among his abstractions; the utterance never disagreed with itself. This is simultaneously a compliment and a demotion of the Mimamsa science: its rules are the correct *statics of the scaffold* — indispensable for the avidvan-analyst working part-wise (2.13) — and idle at the level of the undivided meaning, where there are no ranked parts to collide. The verse thus closes the test the excursus opened (can holism survive exegetical practice?) with the strongest possible verdict: the practice is holism's *own* casuistry, conducted one level down.
contested: none of weight; the verse's sweep — sarvam sruti-bhutam — is as bold as we render it, and the tradition reads it so

2.76 unformalized

वाक्यानां समुदायश्च य एकार्थप्रसिद्धये ।
साकाङ्क्षावयवस्तत्र वाक्यार्थाऽपि न विद्यते ॥ ७६ ॥

vākyānāṃ samudāyaśca ya ekārthaprasiddhaye |
sākāṅkṣāvayavastatra vākyārthā'pi na vidyate || 76 ||

And where there is an aggregate of sentences with mutually-requiring parts, for the establishing of a single meaning — there a (separate) sentence-meaning does not exist either.

The holism propagated one level *further up*, with perfect consistency and startling consequence. Sentences themselves aggregate: the mantra-sequence, the extended injunction, the narrative — vakyanam samudayah — whose members are sakanksa-avayava, mutually requiring (the kanda's own mark of unity, 2.4), jointly establishing eka-artha-prasiddhaye, one meaning. There, the verse rules, vakyarthah api na vidyate: the *individual* sentence-meaning does not exist — it dissolves into the discourse-meaning exactly as padarthas dissolved into it (2.13). The sentence, so long defended as the real unit against the word, was never an absolute: it is the unit *where expectancy closes* (2.4) — and where expectancy spans sentences, the unit is the span. The 'sentence' of the kanda's title was always a variable: the maha-vakya, the great sentence, is whatever whole comprehension actually takes undivided — a verse, a chapter, in the limit (the tradition would say, and Kanda I's 1.1 implies) the one Veda, the one Word. The verse guards the system from making a new atom of the vakya — the same relativism of grain (1.88, 2.10) that dissolved the parts dissolves any absolute whole short of the whole itself. Analysis is scaffolding at *every* level; only the sabda-tattva is no one's abstraction.
contested: api ('either/also') carries the argument — as word-meanings do not exist (as independent reals), so sentence-meanings likewise, within the larger unity; some soften to 'the sentence-meaning is not (complete) there' — the radical reading is the consistent one and we keep it

2.77 · The Exegete's Catalogue unformalized

प्रासाङ्गिकमिदं कार्यमिदं तन्त्रेण लभ्यते ।
इदमावृत्तिभेदाभ्यामत्र बाधसमुच्चयौ ॥ ७७ ॥

prāsāṅgikamidaṃ kāryamidaṃ tantreṇa labhyate |
idamāvṛttibhedābhyāmatra bādhasamuccayau || 77 ||

'This operation is incidental'; 'this is obtained by single performance'; 'this, by repetition and differentiation'; 'here, annulment and combination' —

The opening of a remarkable passage (2.77-86): ten verses of pure quotation-catalogue, the working idiom of the ritual exegete recited at speed — each phrase a term of art from the Mimamsa toolbox, strung in oratio recta as if overheard at the sacrificial session. Here: prasangika, the incidental operation (discharged en passant by a rite performed for another purpose); tantrena labhyate, obtained by tantra, single-performance-serving-many (one threshing serves all the grains); avrtti and bheda, repetition and differentiation (when must an element be re-performed for each?); badha and samuccaya, annulment and combination (when do provisions exclude, when cumulate?). The rhetorical strategy unfolds across the catalogue and is stated at 2.87: every one of these operations is *word-meaning-dependent* — each presupposes that 'this element', 'this provision', can be isolated, weighed, transferred; the exegete's entire practice runs on padartha-realism. Bhartrhari recites the catalogue *fairly and at length* precisely because the concession is real and the resolution (2.87-88) must carry its full weight.

2.78 unformalized

ऊहोऽस्मिन् विषये न्याय्यः सम्बन्धोऽस्य न बाध्यते ।
सामान्यस्यातिदेशोऽयं विशेषोऽत्रातिदिश्यते ॥ ७८ ॥

ūho'smin viṣaye nyāyyaḥ sambandho'sya na bādhyate |
sāmānyasyātideśo'yaṃ viśeṣo'trātidiśyate || 78 ||

'Modification is proper in this domain'; 'the connection of this is not annulled'; 'this is extension of the general'; 'here the particular is extended' —

The catalogue continues with the transfer-operations. Uha, modification — the adaptation of mantras when a rite is transposed (new deity, new substance: the formula's 'rice' re-worded to 'nivara'); the exegete judges where uha is nyayya, proper. Atidesa, extension — the carrying of provisions from the archetype-rite (prakrti) to its derivatives (vikrti): sometimes samanyasya atidesa, the general rule extended; sometimes visesa atidisyate, a particular detail carried across. Each operation is a controlled *edit of meaning under identity of frame* — the rite remains 'the same' while its determinations are re-filled — and thus each presupposes the layered semantics of 2.64-71: only where a general survives its specifications can specification be re-done (uha), only where structure detaches from filling can structure travel (atidesa). The exegete manipulating transfer is manipulating exactly the strata the kanda's refined holism recognizes — a point banked silently for 2.88.

2.79 unformalized

अर्थित्वमत्र सामर्थ्यमस्मिन्नर्थो न भिद्यते ।
शास्त्रात् प्राप्ताधिकारोऽयं व्युदासोऽस्य क्रियान्तरे ॥ ७९ ॥

arthitvamatra sāmarthyamasminnartho na bhidyate |
śāstrāt prāptādhikāro'yaṃ vyudāso'sya kriyāntare || 79 ||

'Here, being-in-need is the qualification'; 'in this, the meaning is not divided'; 'this one's competence is obtained from the teaching'; 'his exclusion is in the other rite' —

The catalogue's third stretch: the doctrine of adhikara, competence — who may perform, who is addressed. Arthitva, being-in-need (desire for the fruit) as the qualifying condition; the scope-rulings ('the meaning is not divided' — one provision, not distributed); competence sastrat prapta, conferred by the teaching itself; and vyudasa, exclusion — this performer barred from that other rite. The exegete here is doing *semantics of address*: determining, from the injunction's words, the class of its addressees — reading subject-position out of sentence-meaning. That address is part of meaning (an injunction *to someone*) while no word in the sentence names the addressee is itself quiet evidence for meaning outrunning deposits — the vakyartha including, as structure, what no padartha supplies.

2.80 unformalized

इयं श्रुत्या क्रमप्राप्तिरियमुच्चारणादिभिः ।
क्रमोऽयमत्र बलवानस्मिंस्तु न विवक्षितः ॥ ८० ॥

iyaṃ śrutyā kramaprāptiriyamuccāraṇādibhiḥ |
kramo'yamatra balavānasmiṃstu na vivakṣitaḥ || 80 ||

'This order is obtained from the text'; 'this, from the order of utterance and the rest'; 'this order here has force'; 'in this one, order is not intended' —

The krama-doctrine: in what sequence must the rite's elements proceed? The sources are ranked (another hierarchy, sibling to 2.73's): order from explicit sruti; order from utterance-sequence (the mantras' own order projected onto acts) and the other lesser sources (performance-order of the archetype, and so on); orders that are balavan, forceful, binding; and orders avivaksita, not intended at all — mere textual accident, no rule (1.136's canon and 2.69's avivaksa, applied to sequence itself). For the kanda this last is the resonant entry: even the exegete, sequence's greatest taskmaster, must distinguish *meant* order from *merely present* order — conceding that krama as such carries no authority until intention ratifies it. Sequence unintended is noise even in scripture: the ritual science's own avivaksita-krama is 2.24-27's doctrine (order as contributed, not found) speaking in the opponent's dialect.

2.81 unformalized

इदं पराङ्गैः संबद्धमङ्गानामप्रयोजकम् ।
प्रयोजकमिदं तेषामत्रेदं नान्तरीयकम् ॥ ८१ ॥

idaṃ parāṅgaiḥ saṃbaddhamaṅgānāmaprayojakam |
prayojakamidaṃ teṣāmatredaṃ nāntarīyakam || 81 ||

'This is connected with the limbs of another'; 'this does not occasion the subsidiaries'; 'this occasions them'; 'this here is inseparable' —

The catalogue's structural stretch: anga-relations, the rite's part-whole engineering. Elements connected para-angaih, with another's limbs (serving a different principal than proximity suggests); elements aprayojaka, not occasioning the subsidiaries (present without commanding a retinue); their opposites, prayojaka; and the nantariyaka, the inseparable — what cannot be omitted without the whole failing. The exegete is here mapping *dependence-structure*: which parts serve which wholes, what can detach, what cannot. It is the sentence's own architecture (2.48's karaka-orbits, 2.4's guna-pradhana) rediscovered in the rite — unsurprisingly, since the rite is *known only through its injunctions*: ritual structure is sentence-structure enacted, and the exegete's anga-map is a syntax written in offerings.

2.82 unformalized

इदं प्रधानं शेषोऽयं विनियोगक्रमस्त्वयम् ।
साक्षादस्योपकारीदमिदमाराद् विशेषकम् ॥ ८२ ॥

idaṃ pradhānaṃ śeṣo'yaṃ viniyogakramastvayam |
sākṣādasyopakārīdamidamārād viśeṣakam || 82 ||

'This is the principal, this the remainder'; 'this, moreover, is the order of application'; 'this assists it directly'; 'this qualifies it from afar' —

Continuation of the dependence-map, now graded by intimacy. Pradhana and sesa, principal and remainder (the Mimamsa's master-pair: everything in the rite is either end or auxiliary-to-end); viniyoga-krama, the order of application (which auxiliary is deployed to which principal, and in what precedence); and the two grades of service — saksat upakari, the direct assistant, touching the principal's own accomplishment, versus arad visesaka, the qualifier *from afar*, contributing at a distance through intermediaries. The two grades are the exegete's version of the semantic distinction the kanda drew at 2.73-74 (warranted nearly versus remotely): contribution is not flat; it has metric structure, near and far, and the science measures it. A theory of meaning as flat deposit-summing has no natural home for 'assistance from afar'; a theory of one meaning differentially articulated (2.71, 2.75) does.

2.83 unformalized

शक्तिव्यापारभेदोऽस्मिन् फलमत्र तु भिद्यते ।
संबन्धाज्जातभेदोऽयं भेदस्तत्राविवक्षितः ॥ ८३ ॥

śaktivyāpārabhedo'smin phalamatra tu bhidyate |
saṃbandhājjātabhedo'yaṃ bhedastatrāvivakṣitaḥ || 83 ||

'In this, the difference lies in the operation of the capacity, while here the fruit is what differs'; 'this difference is born of the connection'; 'there, the difference is not intended' —

The individuation-stretch: when is there *one* element, when two? The exegete's tests: sometimes sakti-vyapara-bheda, difference in the capacity's operation (one item, two workings); sometimes phala-bheda, difference of fruit (two results, hence two rites); sometimes bheda sambandhat jata, difference generated merely by the connection (relational, not intrinsic — the item differs only qua related); and sometimes, again, aviviaksita — present difference, unintended, hence null for the science. This is a working ontology of *aspects*: the exegete individuates by operation, by outcome, by relation, and by intention — never by bare form. Individuation answering to function and intention rather than to phonetic or physical identity is precisely how the kanda individuated its own objects (the word one under many utterances, 1.71; the sentence one under many construals, 2.39) — the catalogue keeps testifying that expert practice everywhere runs on the semantics the theory articulates.

2.84 unformalized

प्रसज्यप्रतिषेधोऽयं पर्युदासोऽयमत्र तु ।
इदं गौणमिदं मुख्यं व्यापीदं गुरु लध्विदम् ॥ ८४ ॥

prasajyapratiṣedho'yaṃ paryudāso'yamatra tu |
idaṃ gauṇamidaṃ mukhyaṃ vyāpīdaṃ guru ladhvidam || 84 ||

'This is negation of the predicated, while this here is exclusionary negation'; 'this is secondary, this primary'; 'this is pervasive'; 'this weighty, this light' —

The logical stretch, headed by the tradition's most consequential distinction: prasajya-pratisedha versus paryudasa — negation that *denies a predication* ('he does not eat' — the act negated as such) versus negation that *excludes toward an affirmation* ('he eats not-this' — a positive selection wearing negative dress). The pair, born in ritual exegesis (does 'na' prohibit the act or redirect it?), became common property of every Indian logical school; its presence here marks the catalogue's reach into semantics proper. With it: gauna and mukhya, secondary and primary senses (metaphor's grammar, to which the kanda returns at length later); vyapin, the pervasive provision (holding across the whole domain); guru and laghu, weighty and light (the maxim-calculus of interpretive economy: prefer the lighter assumption). Negation's two kinds are the entry to watch: they are *indistinguishable in wording* — 'na' plus verb — and distinguished wholly by the one intended sentence-meaning (prohibition versus redirected injunction). No deposit-semantics reads them apart; comprehension of the undivided vakyartha does, every time an exegete rules.

2.85 unformalized

भेदेनाङ्गाङ्गिभावोऽस्य बहुधेदं विकल्प्यते ।
इदं नियम्यतेऽस्यात्र योग्यत्वमुपजायते ॥ ८५ ॥

bhedenāṅgāṅgibhāvo'sya bahudhedaṃ vikalpyate |
idaṃ niyamyate'syātra yogyatvamupajāyate || 85 ||

'The limb-and-limbed relation of this is by differentiation'; 'this is optioned manifoldly'; 'this is restricted'; 'here fitness arises for it' —

The modal stretch: how provisions *bind*. Anga-angi-bhava bhedena, the part-whole relation obtaining under differentiation (the same element limb to one principal, not another); vikalpa, option — where texts permit alternatives, bahudha, manifoldly (and the science's deep discomfort with vikalpa, its rule that option is a last resort, is itself a preference for determinate meaning); niyama, restriction (the alternative closed); yogyatva, fitness arising — the element becoming eligible in context. Option, restriction, fitness: the deontic modalities of the rite, and, transposed, of the sentence (what may, must, can fill this role) — the karaka-fitness (yogyata) that sentence-comprehension weighs (2.44-47) here governing offerings. The catalogue is nearly complete; its cumulative effect — the sheer *density* of part-wise operations in expert practice — is about to be conceded in full (2.87) and absorbed (2.88).

2.86 unformalized

अस्य वाक्यान्तरे दृष्टाल्लिङ्गाद्भेदोऽनुमीयते ।
अयं शब्दादपोद्धृत्य पदार्थः प्रविभज्यते ॥ ८६ ॥

asya vākyāntare dṛṣṭālliṅgādbhedo'numīyate |
ayaṃ śabdādapoddhṛtya padārthaḥ pravibhajyate || 86 ||

'Of this, the difference is inferred from a mark seen in another sentence'; 'this word-meaning is divided out, having been abstracted from the word' —

The catalogue's last and most self-referential entries. First: difference established vakyantare drstat lingat — from a mark seen *in another sentence*: the exegete reads across texts, one passage's indication individuating another's provision (intertextual inference, the linga of 2.73 working at distance). Second — and with this the catalogue closes on the kanda's own key-term: ayam sabdad apoddhrtya padarthah pravibhajyate — 'this word-meaning is divided out, having been *abstracted* from the word'. The exegete himself says apoddhrtya: his practice's own idiom confesses that padarthas are extracts (2.10's doctrine in the practitioner's mouth). The catalogue thus ends by quoting the opponent performing the holist's analysis: ten verses of part-wise operations, and the last operation named is the abstraction that generates the parts. Nothing remains but to state the moral both ways — the objection (2.87: all this collapses if words express nothing) and the resolution (2.88: all this stands, as the statics of abstraction, within the undivided meaning).

2.87-88 unformalized

इति वाक्येषु ये धर्माः पदार्थोपनिबन्धनाः ।
सर्वे ते न प्रकल्पेरन् पदं चेत् स्यादवाचकम् ॥ ८७ ॥
अविभक्तेऽपि वाक्यार्थे शक्तिभेदादपोद्धृते ।
वाक्यान्तरविभागेन यथाक्तं न विरुध्यते ॥ ८८ ॥

iti vākyeṣu ye dharmāḥ padārthopanibandhanāḥ |
sarve te na prakalperan padaṃ cet syādavācakam || 87 ||
avibhakte'pi vākyārthe śaktibhedādapoddhṛte |
vākyāntaravibhāgena yathāktaṃ na virudhyate || 88 ||

(87) Thus, the properties in sentences that are fastened to word-meanings — all of them would fail to be construed, if the word were non-expressive; (88) but when, though the sentence-meaning is undivided, abstraction is made according to the difference of capacities, what was stated is not contradicted — by way of the division into other (subordinate) sentences.

The excursus resolved, both edges honored. (87) states the objection at the full weight the ten-verse catalogue earned it: the dharmas operative in sentences — prasanga and tantra, uha and atidesa, adhikara, krama, anga-structure, the negations, the modalities: everything 2.77-86 recited — are padartha-upanibandhanah, *fastened to word-meanings*; sarve te na prakalperan — all of them would fail, the entire science with them, padam cet syad avacakam, if the word expressed nothing. The concession is total: no exegesis without word-meanings. (88) then resolves without retreat: avibhakte 'pi vakyarthe — the sentence-meaning remaining undivided — apoddhrte sakti-bhedat — abstraction being performed *according to the difference of capacities* (the saktis: the real articulations of the one meaning, 2.71's structure, 2.83's operations) — yatha-uktam na virudhyate: nothing stated is contradicted, the whole practice standing, vakyantara-vibhagena, by way of division into subordinate sentences — the analyst's re-casting of the one utterance as many part-statements (2.35cd's paraphrase-clauses, the exegete's provision-sentences), each a legitimate *station of the scaffold*. The settlement, in one sentence: word-meanings are real as capacities of the undivided meaning and fictional as its components — and the exegetical science is the complete, correct engineering manual of the scaffold, true of the vakyartha *as analyzed*, silent on it as it is. Both kandas' method is hereby vindicated at once: analysis everywhere (or the sciences die, 87), constitution nowhere (or comprehension lies, 88). The excursus that began as holism's hardest test ends as its most detailed confirmation.
contested: 88's vakyantara-vibhagena admits two construals — 'by division into other sentences' (the analyst's part-statements, our reading) or 'by the divisions belonging to other sentences' (transferring articulations across contexts); the settlement's substance — undivided meaning, capacity-wise abstraction, practice preserved — is identical on either

2.89 · Similarity, Error, and the Undivided unformalized

यथैवैकस्य गन्धस्य मेदेन परिकल्पना ।
पुष्पादिषु तथा वाक्येऽप्यर्थमेदो विधीयते ॥ ८९ ॥

yathaivaikasya gandhasya medena parikalpanā |
puṣpādiṣu tathā vākye'pyarthamedo vidhīyate || 89 ||

Just as, of a single fragrance, a construction of difference is made across flowers and the rest — so too in the sentence a difference of meaning is prescribed.

A new movement opens with a delicate analogy about projected difference. One gandha — fragrance as such, or a single perfume — is met in flower, in oil, in unguent; and the mind parikalpana, constructs difference: jasmine-scent, oil-scent, as if the fragrance itself were many. The differences belong to the substrates and the occasions; the projected plurality is a construction upon one thing variously housed. Tatha vakye 'pi: so in the sentence — artha-bheda vidhiyate, difference of meaning is *prescribed*, laid on: the padarthas, the itemized senses, are the jasmine-scent and oil-scent of one undivided vakyartha, differentiated by its housings (the words) rather than in itself. Vidhiyate is drily chosen: prescription is the analyst's act (2.10), not the meaning's structure. The analogy also opens the movement's real theme — the psychology of similarity and misconstrual (2.90-94): how one thing partially apprehended generates confident pluralities, which is both how analysis works and how it errs.

2.90 unformalized

गवये नरसिंहे वाप्येकज्ञानावृते यथा ।
भागं जात्यन्तरस्यैव सदृशं प्रतिपद्यते ॥ ९० ॥

gavaye narasiṃhe vāpyekajñānāvṛte yathā |
bhāgaṃ jātyantarasyaiva sadṛśaṃ pratipadyate || 90 ||

As in the gavaya, or in the man-lion, enveloped in a single cognition, one apprehends a part as simply similar to another kind —

The similarity-psychology, opened with the tradition's two best specimens. The gavaya, the wild ox, is the upamana-example of every epistemology primer: the townsman told 'a gavaya is like your cow' meets one and *recognizes via similarity*. The narasimha, the man-lion — Visnu's terrible avatara — is the fused kind: one being, leonine above, human below. In both cases the object arrives eka-jnana-avrta, wrapped in a *single* cognition — one seeing of one thing — and within that single seeing a *part* presents as sadrsam jatyantarasya, similar to another kind altogether: the gavaya's dewlap recalls the cow, the man-lion's mane the lion. The datum Bhartrhari wants: similarity-to-another is apprehended *inside an undivided grasp of one* — the comparison does not dismember the object into a cow-part plus a remainder; the man-lion is not half a man. Partial resemblance, wholly registered, within unbroken unity: precisely the structure the sentence needs (2.93-94 will apply it), where 'shared words' between two sentences are dewlaps — real resemblances of undivided wholes, not shared components.
contested: the verse runs syntactically into 2.91 (yatha ... 2.92 tatha); we mark the suspension with a dash. Narasimhe is read by some as 'in (the word) narasimha' (the compound as example) — the perceptual reading fits eka-jnana-avrte better

2.91 unformalized

अप्रसिद्धं तु यं भागमदृष्टमनुपश्यति ।
तावत्यसंविदं मूढः सर्वत्र प्रतिपद्यते ॥ ९१ ॥

aprasiddhaṃ tu yaṃ bhāgamadṛṣṭamanupaśyati |
tāvatyasaṃvidaṃ mūḍhaḥ sarvatra pratipadyate || 91 ||

— but the part that is unfamiliar, unseen before, he beholds; and to that extent the bewildered man everywhere arrives at non-comprehension.

The error-half of the psychology. The provincial meeting the man-lion knows lion-manes; the human torso in that setting is aprasiddham, adrstam — unfamiliar, never seen *so* — and before it he is mudhah, bewildered: tavati asamvidam pratipadyate, to just that extent he arrives at non-comprehension, and (sarvatra) the bewilderment spreads over the whole: not knowing the part, he cannot close his grasp of the thing. The observation is exact: comprehension of wholes tolerates familiar-part-resemblance smoothly (2.90) and is *globally* disrupted by one unfamiliar region — the unknown does not politely subtract itself; it unsettles the entire object (as the pika-word did the sentence, 2.72, the parallel Bhartrhari is building toward). Confidence in wholes, fragility at unfamiliar parts: both facts will transfer to sentences in 2.92-94 — where they explain, respectively, why analysis works on familiar material and why 'similarity of parts' between sentences can seduce the analyst into inventing shared components (the fallacy 2.92 names).
contested: asamvidam ('non-comprehension') is our reading of the transcription; the line has been construed 'he attributes (to it) whatever he has not comprehended' — the psychology (local unfamiliarity, global unsettling) is common to both

2.92 unformalized

तथा पिकादियोगेन वाक्येऽत्यन्तविलक्षणे ।
सदृशस्येव संज्ञानमसतोऽर्थस्य मन्यते ॥ ९२ ॥

tathā pikādiyogena vākye'tyantavilakṣaṇe |
sadṛśasyeva saṃjñānamasato'rthasya manyate || 92 ||

So, through the presence of a 'pika' and the like, in a sentence utterly distinct, one fancies — as of something similar — the cognition of a meaning that is not there.

The fallacy diagnosed, closing the triptych. A sentence containing the opaque 'pika' (2.72) is atyanta-vilaksana, utterly distinct — *its* meaning is its own, undivided, partly unmanifest to this hearer. But the hearer, like 2.90's provincial, finds familiar dewlaps: the *other* words, met before in other sentences. And by that yoga, that co-presence, he manyate — fancies — sadrsasya iva samjnanam asato 'rthasya: a cognition, as it were of the similar, of a meaning *that is not there* — he projects into this sentence the meanings those look-alike words 'had' elsewhere, assembling a pseudo-comprehension out of remembered fragments. The diagnosis cuts deep against deposit-semantics: what the atomist calls 'the word's standing meaning, carried into each sentence' is here anatomized as the *hearer's transfer by resemblance* — real psychology, illegitimate semantics; the analogical shadow of other wholes mistaken for a component of this one. Word-meanings, once more (2.17), are citation-artifacts — but now with the error-theory supplied: they are *memories of other sentences*, reified. What guards against the fallacy is what guarded the provincial: completing the acquaintance — hearing the whole as the whole it is.
contested: asato 'rthasya ('of a meaning that is not there') is pointed: not-present-in-this-sentence, rather than non-existent absolutely; some press the stronger reading (fancied meanings are unreal as such) — the diagnosis stands on the weaker

2.93-94 unformalized

एकस्य भागे सादृश्यं भागे मेदश्च लक्ष्यते ।
निर्भागस्य प्रकाशस्य निर्भागेणैव चेतसा ॥ ९३ ॥
तथैव भागे सादृश्यं भागे मेदोऽवसीयते ।
भागाभावेऽपि वाक्यानामत्यन्तं भिन्नधर्मणाम् ॥ ९४ ॥

ekasya bhāge sādṛśyaṃ bhāge medaśca lakṣyate |
nirbhāgasya prakāśasya nirbhāgeṇaiva cetasā || 93 ||
tathaiva bhāge sādṛśyaṃ bhāge medo'vasīyate |
bhāgābhāve'pi vākyānāmatyantaṃ bhinnadharmaṇām || 94 ||

(93) Of one thing, similarity is marked in one part and difference in another — of a partless illumination, by a mind itself partless; (94) just so is similarity settled in one part and difference in another even of sentences whose properties are utterly distinct — though parts there are none.

The movement's resolution: partial similarity *without parts* — the apparent contradiction dissolved by the psychology just built. (93) states the general form with maximal paradox: one nirbhaga prakasa, one partless illumination (a color-expanse, a face, any unitary presentation), is marked similar-*here*, different-*there* — by a cetas, a mind, itself nirbhaga, partless (2.25-27: the sequenceless capacity). Neither term of the comparison has parts; yet 'part-wise' comparison happens — because *aspect-discrimination is the mind's act upon wholes*, not a reading-off of components: the partless cognition self-articulates (sva-saktya, 2.25) in registering respects of likeness. Two faces are alike 'about the eyes' though neither face is an assembly. (94) then delivers the application the movement was built for: so with sentences — atyanta-bhinna-dharmanam, of utterly distinct properties (each vakyartha its own undivided whole, 2.92), bhaga-abhave 'pi, *though parts there are none*: similarity settles 'in one part' (the shared-seeming words), difference 'in another'. Inter-sentential resemblance — the deposit-theorist's master-datum: 'the same word recurs with the same meaning across sentences' (1.71) — is thus accounted whole-to-whole: sentences resemble *in respects*, as faces do, and the 'recurrent word-meaning' is a respect of resemblance among undivided meanings, not a shared brick. The atomist's strongest evidence is hereby not denied but re-parented: recurrence is real, and it is a similarity-fact, not a composition-fact.
contested: 93's simile-base (the partless prakasa) is unspecified — light, cognition, or the sphota itself; all three work and the tradition wavers. The doctrine of 94 — respect-wise resemblance without composition — is the movement's secured yield on any choice

2.95 unformalized

रूपनाशे पदानां स्यात् कथं चावधिकल्पना ।
अगृहीतावधौ शब्दे कथं चार्थो विवच्यते ॥ ९५ ॥

rūpanāśe padānāṃ syāt kathaṃ cāvadhikalpanā |
agṛhītāvadhau śabde kathaṃ cārtho vivacyate || 95 ||

If the form of the words is destroyed, how could there be the construction of a boundary? And when no boundary is grasped in the word, how is the meaning intended to be discerned?

An objector's double question, sharp enough that Bhartrhari devotes the next movement (2.96-111) to the machinery it demands. If — as the holism now says — the words' own forms are dissolved into the sentence (rupa-nasa: no independent word-reals), then: katham avadhi-kalpana — how is a *boundary* constructed at all? Where does one sentence (or word, in analysis) end and the next begin? And agrhita-avadhau sabde — the word/utterance having no grasped boundary — katham artho vivacyate: how is the meaning to be discerned/intended? Comprehension surely requires *delimitation*: the hearer must take *this much* as the unit to be understood (2.4's expectancy-closure presupposed a where-it-closes). The objection is the practical shadow of 2.76: if units dissolve upward and downward, the theory owes an account of segmentation — how speech is *portioned* by hearers and analysts if the portions are not objectively there. Bhartrhari's answer, unfolding across 2.96-111, will be drawn from grammar's own devices: tantra (single-utterance-serving-many), ekasesa (one-remainder), the sandhi-doctrine — cases where the science itself manufactures and dissolves boundaries at need, showing delimitation to be operational, not ontological.
contested: vivacyate we read with the transcription as from vi-vac (be discerned/expressed); a variant tradition reads vivaksyate ('is intended') — the objection is the same: no boundary, no determinate unit of meaning

2.96 unformalized

संसर्ग इव रूपाणां शब्देऽन्यत्र व्यवस्थितः ।
नानारूपेषु तद्रूपं तन्त्रेणापरिमिष्यते ॥ ९६ ॥

saṃsarga iva rūpāṇāṃ śabde'nyatra vyavasthitaḥ |
nānārūpeṣu tadrūpaṃ tantreṇāparimiṣyate || 96 ||

Like a commingling of forms, it stands settled in the word elsewhere; that form, among the various forms, is not measured out by single-performance.

The answer begins, with grammar's strangest boundary-phenomena as evidence. Forms samsarga iva, as-if-commingled: in the continuous utterance (samhita), word-shapes interpenetrate — sandhi fuses edges, one sound serving two words at once (the pararupa of 2.101 ahead) — the 'boundary' being genuinely *shared territory*. And such a form, standing among many forms, na tantrena parimisyate — is not measured out by tantra, single-performance: one cannot always utter a shared form once and have it count for each sharer; the commingled is not the common. The verse's work is to put on the table cases where wordhood's edges are *objectively indeterminate* in the utterance — fusion, sharing, overlap — so that the objector of 2.95 cannot pretend boundaries are simply given in speech and merely reported by grammar. They are not given; they are *managed* — by devices (tantra here glimpsed, ekasesa and the rest to come) whose very existence shows segmentation to be the grammar's engineering upon a continuum, exactly as the holist said (2.58: samhita as womb).
contested: the verse is among the kanda's most compressed and our construal — commingled forms resisting tantra-measurement — follows the movement's arc into 2.97-101; alternative readings take tantrena with what follows ('by the single utterance, that form is not exhausted among its various (occurrences)'), same evidentiary point

2.97 unformalized

तस्मिन्नभेदे भेदानां संसर्ग इव वर्तते ।
रूपं रूपान्तरात्तस्मादनन्यत् प्रविभज्यते ॥ ९७ ॥

tasminnabhede bhedānāṃ saṃsarga iva vartate |
rūpaṃ rūpāntarāttasmādananyat pravibhajyate || 97 ||

In that non-difference, there operates as it were a commingling of differences; therefore the form, not being other than the other form, is (nonetheless) divided out.

The metaphysical gloss on 2.96's phenomena, stated as the movement's theorem. Tasmin abhede — in that non-difference, the fused continuum of actual utterance — bhedanam samsarga iva vartate: a commingling *of differences* operates, as-it-were: the analyst's word-forms are 'in' the stream the way currents are in a river — distinguishable flows of what is one water. Therefore: rupam rupantarat ananyat — the form is *not other* than the other form (at the fusion, the 'end' of one word and the 'start' of the next are numerically one stretch of sound) — and yet pravibhajyate, it *is divided out*: analysis assigns the shared stretch twice, to each word, without scruple and without falsity. Division-of-the-undifferent: grammar's daily sandhi-practice performs, at the phonetic joint, exactly the operation the holist claims for all analysis (2.10, 2.89) — real distinction drawn where no real otherness stands, valid as procedure, silent as ontology. The objector's boundary (2.95) is thus answered in kind: boundaries are *drawn*, expertly and reproducibly, upon the boundariless — and their reproducibility (the science's rules) is what their objectivity consists in.
contested: ananyat ('not other') we take at the fusion-point specifically; read globally ('the form is nothing other than the other form') the verse asserts full identity of all word-forms in the one utterance — the stronger claim is congenial to the kanda but the sandhi-context suggests the local one

2.98-99 unformalized

शास्त्रे प्रत्यायकस्यापि क्वचिदेकत्वमाश्रितम् ।
प्रत्याय्येन क्कचिद्भेदो ग्रहणग्रह्ययोः स्थितः ॥ ९८ ॥
ऊ इत्यभेदाश्रित्य यथासंख्यं प्रकल्पितम् ।
लृङ्लृटोर्ग्रहणे भेदो ग्राह्याभ्यां सह कल्पितः ॥ ९९ ॥

śāstre pratyāyakasyāpi kvacidekatvamāśritam |
pratyāyyena kkacidbhedo grahaṇagrahyayoḥ sthitaḥ || 98 ||
ū ityabhedāśritya yathāsaṃkhyaṃ prakalpitam |
lṛṅlṛṭorgrahaṇe bhedo grāhyābhyāṃ saha kalpitaḥ || 99 ||

(98) In the science, sometimes oneness is relied upon even for the conveyor, while sometimes its difference from the conveyed is established — of grasper and grasped. (99) In 'u' oneness is relied upon and the respective-correspondence constructed; in the mention of lrn and lrt, difference is constructed together with the two graspables.

The evidence made technical: Panini's own metalanguage manipulates the identity and difference of *its own signs* at convenience. (98) states the practice: the sastra sometimes relies on the *oneness* of a pratyayaka, a conveying sign (one mention serving many conveyed items), sometimes on its *difference* from and among the conveyed — the grahana-grahya relation (mention and mentioned) engineered per rule. (99) gives the worked examples from the grammar's depths: the single technical 'u' relied on as one and distributed yatha-samkhyam, by respective correspondence (one sign, many assignments in order); the mention of lrn and lrt (the conditional and second-future affix-names) constructed as *different*, each with its own graspable. The argument: if word-forms were self-delimiting reals, the science could not treat one sign as one here and as many there, could not fuse and split mention at need — but it does, ubiquitously and without paradox, because signs *have* the individuation their use assigns (2.83's aspect-individuation, now in the metalanguage). Grammar's meta-practice is the finest-grained confirmation available: even where language talks about language, boundaries and identities are operational postings, not discoveries — and what holds of the mention holds of the mentioned (1.65's regress collapsed into evidence).
contested: the examples are stenographic and our unpacking (u-pratyahara distributed; lrn/lrt differentiated) follows the grammatical tradition's cases; specialists locate the precise sutras variously — the meta-linguistic moral is unaffected

2.100 unformalized

यस्येत्येतदणो रूपं संज्ञिनामभिधायकम् ।
नहि प्रतीयमानेन ग्रहणास्यास्ति सम्भवः ॥ १०० ॥

yasyetyetadaṇo rūpaṃ saṃjñināmabhidhāyakam |
nahi pratīyamānena grahaṇāsyāsti sambhavaḥ || 100 ||

The form 'yasya' (in the rule) is the an-form expressing its name-bearers; for there is no possibility of mention by way of what is (merely) being understood.

A precision on how mention works, guarding the meta-practice of 2.98-99 from a misreading. In a rule like Panini's (the tradition points to 6.4.148, yasya iti ca), the cited form functions as the *an-form* — the technically-shaped sign-body — abhidhayaka of its samjnins, expressing the items it names. The principle in the second line: na hi pratiyamanena grahanasya asti sambhavah — mention cannot proceed by way of the merely-understood: what a rule *grasps* (grahana) must be a *form given in the rule*, not a content the reader would understand anyway; the metalanguage operates on exhibited shapes, not on inferred senses. This looks anti-holist — mention riding forms! — and is in fact the division of labor the settlement (2.88) drew: within the scaffold, forms are the currency (the science individuates by shape, cites by shape, 2.98-99 engineers shape-identity); the scaffold's forms are themselves postings within the undivided (2.97). Rule-craft is form-craft; comprehension is whole-craft; the kanda's care is to keep each sovereign in its own house.
contested: the sutra-context of yasya/an is disputed among the grammarians themselves and our identification follows the common gloss; the verse's principle — grahana requires exhibited form — is clear regardless of the rule intended

2.101 unformalized

ऊ इत्येतदभिन्नं च भिन्नवाक्यनिबन्धनम् ।
भेदेन ग्रहणां चास्य पररूपमिव द्वयोः ॥ १०१ ॥

ū ityetadabhinnaṃ ca bhinnavākyanibandhanam |
bhedena grahaṇāṃ cāsya pararūpamiva dvayoḥ || 101 ||

And this 'u', undivided, is the fastening of different sentences; and its mention is made under difference — like the single subsequent-form belonging to both.

The boundary-sharing phenomenon named at last in its classical form. The undivided 'u' (the technical sign of 2.99, or the sandhi-vowel) serves as bhinna-vakya-nibandhanam — the *fastening of different sentences/rules*: one form, anchoring plural contexts, mentioned bhedena, under difference, as many as its employments. The simile seals it: pararupam iva dvayoh — like the pararupa, the 'subsequent form' of sandhi doctrine, where two adjacent vowels resolve into *one* sound that grammatically *belongs to both* words (the single 'e' of a fusion being the end of one word and the beginning of the next, by rule). One phonetic body, two grammatical citizenships: the limit-case of everything the movement has argued. Boundaries not only can be drawn on the undifferent (2.97) — the science *assigns the same stretch to two units at once* when the economy of description demands. After pararupa, the objector's 'where does the word end?' (2.95) has its answer: where the grammar, for its purposes, posts the end — including, when useful, in two places at once.
contested: whether the 'u' here is the pratyahara-sign continued from 2.99 or the sandhi-vowel is undecidable and immaterial — the pararupa-simile carries the doctrine either way

2.102 unformalized

प्लुतस्याङ्गविवृद्धिं च समाहारमचोस्तथा ।
व्युदस्यता पुनर्भेदः शब्देष्वत्यन्तमाश्रितः ॥ १०२ ॥

plutasyāṅgavivṛddhiṃ ca samāhāramacostathā |
vyudasyatā punarbhedaḥ śabdeṣvatyantamāśritaḥ || 102 ||

And by the one who sets aside the lengthening of the base before pluta, and likewise the coalescence of the two vowels, difference among words is relied upon to the utmost.

The counter-practice acknowledged: grammar also *maximizes* boundaries when its purposes so run. Certain operations are vyudasyat, set aside — blocked: the pluta (protracted) vowel does not trigger the base-lengthening (anga-vivrddhi) that its shape might invite; vowel-coalescence (samahara acoh) is barred at certain junctures — and such blockings are effected precisely by *relying on difference to the utmost* (atyantam asritah bhedah): treating the words as utterly separate, their edges rigid, so that cross-boundary operations cannot see across. Read with 2.101, the pair completes the demonstration: the same science that fuses boundaries (pararupa: one form, two words) also *hardens* them (blocking: two words, no interaction) — per rule, per need. Boundary-strength is a *parameter of description*, tuned case by case — which is unintelligible if boundaries are facts of the speech, and exactly intelligible if they are postings of the analysis (2.97). The objector's segmentation (2.95) has now been shown, from the science's own casebook, to be manufactured in both directions: dissolved where description profits, absolutized where it must.
contested: the specific blockings compressed here are variously identified (the pluta-contexts of Panini 6.1, the samahara-exceptions); our rendering keeps the operations generic — the tuning-of-boundaries moral is the verse's evident office

2.103 unformalized

अर्द्धर्चादिषु शब्देषु रूपभेदः क्रमाद्यथा ।
तन्त्रात्तथैकशब्दत्वे भिन्नानां श्रुतिरन्यथा ॥ १०३ ॥

arddharcādiṣu śabdeṣu rūpabhedaḥ kramādyathā |
tantrāttathaikaśabdatve bhinnānāṃ śrutiranyathā || 103 ||

As, in words like ardharca, difference of form (obtains) according to sequence — so, by single-performance, when there is one-wordhood, the hearing of the different is otherwise.

A final pair of boundary-cases, from the lexicon's own oddities. Ardharca, 'half-verse' — the traditional head of a gana (the ardharcadi class) of words wavering between genders and, emblematically here, between *analyses*: read krama-wise, as sequence (ardha + rca), its form divides one way; read tantra-wise, as one word uttered once, the sruti — the hearing — of the 'different' components is anyatha, otherwise: the parts are not heard *as* parts at all. One phonetic body; two hearings; the difference made not by the sound but by the descriptive frame (sequence-analysis versus unit-citation) under which it is taken. The compound-word — language's own boundary-dissolving device, where yesterday's phrase is today's word — is the natural last witness: lexicalization is boundary-*history*, edges fading as analysis-frames shift, and no acoustic fact marks the transition. Segmentation, once more, follows the taking, not the sound (2.58's vrtti-bhedena, now at the lexical scale).
contested: the verse is elliptical to the edge of riddle; our construal — ardharca-class words as frame-dependent in their part-hearing — follows the gana-tradition and the movement's argument; readers who take ardharcadi merely as 'words with variable gender' get a thinner but compatible point (form-difference without meaning-difference, frame-assigned)

2.104 unformalized

संहिताविषये वर्णाः स्वरूपेणाविकारिणः ।
शब्दान्तरत्वं यान्तीव शक्त्यन्तरपरिग्रहात् ॥ १०४ ॥

saṃhitāviṣaye varṇāḥ svarūpeṇāvikāriṇaḥ |
śabdāntaratvaṃ yāntīva śaktyantaraparigrahāt || 104 ||

In the domain of the continuous utterance, the phonemes, unchanging in their own form, go as it were to the condition of other words, through the assumption of another capacity.

The movement's doctrine crystallized in one verse — the samhita as the place where identity and difference trade places. In samhita-visaya, the continuous utterance, the varnas are svarupena avikarinah: unchanged in their own form — the sounds are what they are, the stream is one. And yet sabdantaratvam yanti iva: they *go, as it were, to being other words* — the same stretches, under sandhi and re-segmentation, belong now to this word, now to that (2.101's double citizenship) — sakty-antara-parigrahat, through the assumption of *another capacity*. Nothing in the sound changes; its office changes. Sakti, the kanda's deepest word (2.88: abstraction follows capacity-difference), here does its finest work: wordhood is a capacity *worn* by stretches of the invariant stream, and re-segmentation is capacity-exchange, not sound-change. The iva, as ever, holds the ontology honest: they do not *become* other words (there are no word-things to become); they present under other offices. The objector of 2.95 receives here the theory his question deserved: boundaries are the seams of capacities, and capacities are assigned by the whole (the one vakya, the one samhita) to its own presented articulations.
contested: none of weight; the verse is the movement's clearest and the tradition treats it as summary

2.105 unformalized

इन्द्रियादिविकारेण दृष्टं ग्राह्येषु वस्तुषु ।
आत्मत्यागादृते भिन्नं ग्रहणं स क्रमः श्रुतौ ॥ १०५ ॥

indriyādivikāreṇa dṛṣṭaṃ grāhyeṣu vastuṣu |
ātmatyāgādṛte bhinnaṃ grahaṇaṃ sa kramaḥ śrutau || 105 ||

It is seen, among graspable things, that through modification of the sense-faculty and the rest, the grasping differs without the thing abandoning its own self; that is the way with the heard.

The epistemological warrant for 2.104, drawn from perception at large. Among grahya vastuni, graspable things generally, grasping varies indriya-adi-vikarena — with modifications of the faculty and its conditions: the same white shell grasped yellow by the jaundiced eye, the same taste sweet or flat by the fevered tongue — bhinnam grahanam atma-tyagad rte: the grasping differs *without the thing's abandoning its self*. Perceptual variation is routinely receiver-side; no one infers a new shell from a new seeming. Sa kramah srutau: that is the pattern with the heard — the one utterance-stream grasped now as these words, now as those (2.103-104), under different analytic 'faculties' (the frames, the trainings, the purposes of hearers and analysts), remains self-same throughout; the variation of segmentation is variation of grasping. The verse generalizes the kanda's constant move (1.75-77: delivery-properties are grasping-conditions; 2.25: sequence is the cognition's act) into a maxim: *difference of apprehension never by itself certifies difference in the apprehended* — the principle on which every reduction of parts to postings has quietly run, here stated as common epistemology.
contested: srutau we take as 'with respect to the heard (utterance)'; some take it 'in the Veda' (sruti), making the verse about scriptural construal — the maxim covers both and the movement favors the acoustic reading

2.106 unformalized

अभिधानक्रियायोगाच्छब्देष्वविकृतेष्वपि ।
रूपमत्यन्तभेदेन तदेवैकं प्रकाशते ॥ १०६ ॥

abhidhānakriyāyogācchabdeṣvavikṛteṣvapi |
rūpamatyantabhedena tadevaikaṃ prakāśate || 106 ||

Through connection with the act of denoting, even among words unmodified, the form — one and the same — shines forth with utter difference.

The converse of 2.104-105, completing the chiasmus: there, invariant sound wearing different word-capacities; here, invariant word-form shining as utterly different *through its denoting-employments*. Abhidhana-kriya-yoga — connection with the act of denoting: the same form 'gauh', unmodified (avikrtesu sabdesu), deployed of the beast, of the constellation, in citation, in metaphor — tad eva ekam rupam atyanta-bhedena prakasate: the *very same one* form radiates with utter difference. Neither sound nor lexical shape changes; the act it serves in changes, and with it everything comprehension receives. Prakasate keeps the register honest — manifestation, not transformation: the differences are modes of the one form's shining, occasioned by employment (the vyapara-individuation of 2.83). Between 2.104 and 2.106 the full doctrine of linguistic identity stands: sameness of sound underdetermines wordhood (capacity decides); sameness of word-form underdetermines meaning (employment decides) — identity at every level is office-identity, and office is conferred by the undivided act of speaking. The Vedic recitation-sciences supply the next verse's confirmation.
contested: none of substance; the verse's placement (before the saman-example) marks it as the principle 2.107 instantiates

2.107 unformalized

ऋचो वा गीतिमात्रे वा साम द्रव्यान्तरं न तु ।
गीतिभेदाद्विगृह्यन्ते ता एव विकृता ऋचः ॥ १०७ ॥

ṛco vā gītimātre vā sāma dravyāntaraṃ na tu |
gītibhedādvigṛhyante tā eva vikṛtā ṛcaḥ || 107 ||

The saman is (nothing but) the rks, or mere melody — not, however, a further substance; by difference of melody, those very rks, modified, are severally grasped.

The confirmation from sacred musicology — the tradition's own limit-case of one-form-many-shinings. What *is* a saman, a chant of the Samaveda? The doctrine the verse ratifies: the samans are the rks — the Rgvedic verses themselves — sung; or say, mere giti, melody, laid upon them; na tu dravyantaram: *not a further substance*, no third thing over verse and tune. Giti-bhedat, by difference of melody, ta eva rcah, those very verses, vikrtah, modified (stretched, repeated, ornamented by the chant's syllable-insertions), vigrhyante — are severally grasped, counted as distinct samans, catalogued, named, transmitted as separate sacred items. An entire Veda — its canon, its schools, its ritual employments — individuated by *mode of rendition* upon unchanged texts: office-identity (2.106) practiced at scripture's own scale, with the tradition's full authority behind the na dravyantaram. If the samans can be many while being the rks, the analyst's words can be many while being the one utterance (2.104): the sacred case licenses the semantic one.
contested: rco va gitimatre va compresses a debate the Mimamsa also conducts (is the saman the verse, the melody, or the composite?); the karika's own verdict — no further dravya — is what the argument needs, and both remaining options serve it

2.108 unformalized

उपायाच्छ्रुतिसंहारे (रो) भिन्नानामेकशेषिणाम् ।
तन्त्रेणोच्चारणे तेषां शास्त्रे साधुत्वमिष्यते ॥ १०८ ॥

upāyācchrutisaṃhāre (ro) bhinnānāmekaśeṣiṇām |
tantreṇoccāraṇe teṣāṃ śāstre sādhutvamiṣyate || 108 ||

When, by a device, the utterances of different words having a single remainder are gathered — in uttering them by single-performance, correctness in the science is accepted.

The ekasesa doctrine arrives — grammar's most audacious boundary-device, and the movement's crowning exhibit. Panini's rule (1.2.64ff.): when several identical forms would occur together ('ramas ca ramas ca'), *one remains* — ekasesa, single-remainder — 'ramau', 'ramah': one uttered form standing for many words. The verse: such gatherings, upayat, by device — sruti-samhare, in the gathering-up of the (plural) utterances — uttered tantrena, by single-performance (the ritual term of 2.77 now naturalized in grammar): sadhutvam isyate, correctness is accepted in the science. Weigh what the science thereby countenances: *several words* — really several, by its own count: each with its case, its reference, its role — *whose utterance is one form uttered once*. The deposit-theorist's minimal assumption — one word-token, one meaning-contribution — is breached by rule: here are meanings without severally-uttered vehicles, vehicles bearing plural citizenship not at their edges (2.101) but in toto. Only office-identity absorbs this smoothly: the one form serves many word-offices, the offices conferred by the construction's needs — the device is 'correct' because correctness was always adjudicated at the whole (2.109 next states exactly that condition).
contested: the transcription wavers (samhare/samharo); the ekasesa-identification is the tradition's and certain. Upayat ('by device') quietly re-invokes 2.38: ekasesa as upaya, the scaffold's own scaffolding

2.109 unformalized

परिगृह्य श्रुतिं चैकां रूपभेदवतामपि ।
तन्त्रेणोच्चारणं कार्यमन्यथा ते न साधवः ॥ १०९ ॥

parigṛhya śrutiṃ caikāṃ rūpabhedavatāmapi |
tantreṇoccāraṇaṃ kāryamanyathā te na sādhavaḥ || 109 ||

And having taken up a single utterance even for those possessing difference of form, the uttering must be done by single-performance; otherwise they are not correct.

The ekasesa-doctrine extended and its condition stated with legislative force. Extended: even rupa-bhedavatam api — even where the gathered words *differ in form* (the dvandva-adjacent cases the grammarians debate: 'pita-mata' gathered as 'pitarau', 'parents', the mother uttered nowhere yet meant) — a single sruti is parigrhya, taken up, and the uttering *must* be tantrena, by single-performance. Then the condition, unblinking: anyatha te na sadhavah — otherwise (uttered severally where the device is prescribed) they are *not correct*. Mark the inversion this completes: correctness — sadhutva, Kanda I's dharma-bearing property (1.27, 1.141) — here *requires* the non-utterance of words that are nonetheless fully meant and grammatically counted. The severally-spelled-out 'correct' forms would be solecism; the fused device is the norm. Against any semantics that pins meaning to uttered vehicles, the science's own norm of correctness legislates meaning-without-vehicle wherever ekasesa runs. The scaffold does not merely tolerate the holist's ontology; at these joints it *mandates* it.
contested: which form-different gatherings fall under the rule is the grammarians' own long dispute (the ekasesa of 'pitarau'-type is contested in the tradition); the karika legislates the principle and leaves the casebook to the sastra

2.110 unformalized

सरूपाणां च वाक्यानां शास्त्रेण प्रतिपादतम् ।
तन्त्रेणोच्चारणादेकं रूपं साधूपलभ्यते ॥ ११० ॥

sarūpāṇāṃ ca vākyānāṃ śāstreṇa pratipādatam |
tantreṇoccāraṇādekaṃ rūpaṃ sādhūpalabhyate || 110 ||

And of sentences of like form, it is taught by the science: from utterance by single-performance, a single correct form is obtained.

The device propagated to the sentence-level — completing the movement at the grain where the kanda lives. Sarupanam vakyanam: sentences of like form — the shared refrain, the repeated mantra serving multiple ritual stations, the common formula meant severally in several connections. The science teaches (sastrena pratipaditam): uttered tantrena, once-for-many, ekam rupam sadhu upalabhyate — *one* correct form is obtained, and it discharges them all. Whole sentences, then — each with its own vakyartha, its own station, its own fruit — borne by one uttered body: the ekasesa-logic (2.108-109) at full scale. The theoretical harvest is the movement's QED: if one utterance can *be* many sentences (as office), then sentencehood was never in the utterance — it was always the office conferred by the employment-whole (the rite, the discourse, 2.76's larger unity) upon the uttered form. The boundary-question of 2.95 dissolves in its final form: not only are edges posted, whole *units* are posted — one body, many sentences; many bodies (the substitute-cases), one rite; the mapping engineered by the science at every grain, answerable only to the undivided purposes it serves.
contested: none of weight; the sentence-level tantra is the ritual-exegetical practice (one recitation serving many applications) and the verse cites it as settled

2.111 unformalized

एकस्यानेकरूपत्वं नालिकादिपरिग्रहात् ।
यथा तथैव तन्त्रात् स्यात् बहूनामेकरूपता ॥ १११ ॥

ekasyānekarūpatvaṃ nālikādiparigrahāt |
yathā tathaiva tantrāt syāt bahūnāmekarūpatā || 111 ||

As one thing comes to have many forms through the taking-up of the reed and the like — just so, through single-performance, many may come to have a single form.

The movement's closing simile, chosen for its perfect symmetry. One breath, taken up through nalika-adi — the reed-pipe and its kin (the flute's bores, the instruments' channels) — becomes many-formed: one wind, many notes, the plurality conferred by the apertures it is passed through. Yatha — and tatha, inversely: through tantra, single-performance, bahunam eka-rupata — the many (words, sentences, rites) come to *one* form: plurality gathered into a single uttered body, the unity conferred by the device it is passed through. One-to-many by instrumental taking-up; many-to-one by performative gathering: the two operations the whole movement has documented (2.96-110), here shown to be the same operation run in opposite directions — form and office trading multiplicities under devices. Beneath both, unmoved, the invariant: the breath; the utterance-stream; the undivided meaning. The simile quietly re-sounds 1.116's fan and the vamsa-flute of the old cosmologies: manifestation as one wind through many stops — the kanda's technical grammar of boundaries ending on the system's oldest music.
contested: nalika-adi we take as the wind-instrument's reed/pipe (the traditional gloss); some read 'the (moulding-)tube' of metallurgy — one metal, many castings; either instrument serves the one-through-channels figure

2.112 · Word-Meaning and Sentence-Purpose unformalized

यथा पदसरूपाणां वाक्यानां सम्भवः पृथक् ।
तथा वाक्यान्तराभावे स्यादेषां पृथागर्थता ॥ ११२ ॥

yathā padasarūpāṇāṃ vākyānāṃ sambhavaḥ pṛthak |
tathā vākyāntarābhāve syādeṣāṃ pṛthāgarthatā || 112 ||

As there is separate occurrence of sentences whose words are of like form, so, in the absence of another sentence, these would have separate meaningfulness.

A hinge-verse, turning the movement's results toward the kanda's next great topic (the abhidheya/prayojana distinction of 2.113). Sentences pada-sarupanam — of like word-forms — occur prthak, separately: the same wording, uttered in two settings, is *two sentences* (office-individuation, 2.110 inverted: one form, many units — here each unit its own occurrence). And vakyantara-abhave — where no other sentence stands (no larger discourse absorbing them, 2.76) — syad esam prthag-arthata: they *would have* separate meaningfulness, each its own vakyartha, notwithstanding identity of wording. Same words, different sentences, different meanings: the last nail in form-semantics, driven with the movement's accumulated hammer — if meaning rode wording, like wording would compel like meaning; it does not (the repeated formula means per station, 2.110; the quoted sentence means otherwise than its original; the rehearsed line otherwise than the avowal). What fixes each occurrence's meaning is its *employment-whole* — and the anatomy of that fixing, the difference between what a sentence's words denote and what the sentence is *for*, is precisely where the text now turns.
contested: the double conditional (occurrence separate; meaning separate *in the absence of a larger sentence*) we read as stated; some construe vakyantara-abhave as 'since there is no (real) other sentence' — a denial that re-wordings are ever distinct — against the movement's drift, and we decline it

2.113 unformalized

अभिधेयः पदस्यार्थो वाक्यास्यार्थः प्रयोजनम् ।
यस्य तस्य न संबन्धो वाक्यानामुपपद्यते ॥ ११३ ॥

abhidheyaḥ padasyārtho vākyāsyārthaḥ prayojanam |
yasya tasya na saṃbandho vākyānāmupapadyate || 113 ||

The expressible is the word's meaning; the purpose is the sentence's meaning — whoever (holds otherwise), for him no connection among sentences is intelligible.

The distinction on which the kanda's remaining semantics is built, legislated in a half-line each. Abhidheyah padasya arthah: the *expressible* — the denotable, the lexical content — is the word's meaning: what analysis assigns the extracted units (2.10), the scaffold's currency (2.88). Vakyasya arthah prayojanam: the sentence's meaning is *purpose* — prayojana: what the utterance is deployed to accomplish, its point, its fruit-in-view. The category-difference is the doctrine: word-meanings are contents; sentence-meaning is *telos* — not a bigger content assembled from small ones but a different kind of thing altogether (2.55's bhinna-laksana, now named): the one act-of-meaning whose service the contents are abstracted as serving. And the argument appended: whoever holds otherwise — meaning as content all the way up — tasya na sambandho vakyanam upapadyate: for him *connection among sentences* is unintelligible. Discourse-coherence (the samudaya of 2.76, the paragraph, the rite's liturgy) is teleological through and through: sentences connect by subserving ends within larger ends (2.82's viniyoga at discourse-scale); mere contents, stacked, connect not at all — no proposition reaches for its successor; purposes do. The kanda's two questions — what unifies the sentence? what unifies sentences? — receive here one answer: prayojana, the undivided point, at every scale the principle of unity that content-semantics forever lacks.
contested: yasya tasya we render 'whoever (holds otherwise), for him' with the standard supplement; the ellipse can be filled differently ('of whatever [word], connection... is not intelligible [without purpose]') — the abhidheya/prayojana legislation, the verse's substance, is textually secure

2.114 unformalized

तत्र क्रियापदान्येव व्यपेक्षन्ते परस्परम् ।
क्रियापदानुषक्तस्तु सम्बन्धोऽतः प्रतीयते ॥ ११४ ॥

tatra kriyāpadānyeva vyapekṣante parasparam |
kriyāpadānuṣaktastu sambandho'taḥ pratīyate || 114 ||

There, it is the action-words themselves that mutually depend; and the connection, attached to the action-word, is thereby understood.

The teleology of 2.113 given its grammatical mechanics — how purpose-connection actually shows in the texture of discourse. Tatra — there, among connected sentences — kriya-padani eva vyapeksante parasparam: it is the *action-words* that mutually depend: verb reaches for verb across sentence-edges ('having-bathed... he-offers... he-recites...' — the rite's liturgy; 'she left... he followed' — the narrative's spine); the inter-sentential expectancy that 2.113 said content cannot ground runs concretely along the verbs. And the sambandha, kriya-pada-anusakta, attached to the action-word, atah pratiyate — is understood therefrom: discourse-connection is *read off* verbal dependence. The mechanics ratify the metaphysics: verbs are grammar's purpose-bearers — each names a sadhya, a to-be-accomplished (2.48) — so purposes-within-purposes (2.113's coherence) surface as verb-upon-verb dependence; the karaka-architecture that organized the single sentence (nouns orbiting the verb) iterates upward as sentences orbiting principal actions (2.6's tinanta qualifying tinanta, now at discourse-scale, its promise redeemed). One structure, three scales — noun to verb, clause to clause, sentence to sentence — and at every scale the same principle: unity flows from the end in view, and the end wears the verb.
contested: none of substance; the verse applies 2.6 and 2.48 at discourse-scale as we render, and the tradition reads it so

2.115 unformalized

आवृत्तिरनुवादो वा पदार्थव्यक्तिकल्पने ।
प्रत्येकं तु समाप्तोऽर्थः सहभूतेषु वर्तते ॥ ११५ ॥

āvṛttiranuvādo vā padārthavyaktikalpane |
pratyekaṃ tu samāpto'rthaḥ sahabhūteṣu vartate || 115 ||

On the construction of word-meaning as the individual: either repetition, or restatement; but the meaning, complete in each, operates in them as co-present.

A dilemma for individual-denotation, resolved by completion-per-each. If words denote vyaktis, individuals, then in 'gauh gauh' (or in a plural: 'the cows') the second occurrence is either avrtti, repetition — the same individual denoted twice, redundantly — or anuvada, restatement — reference back, adding nothing. Either way the second token is semantically idle, which affronts the economy every school professes. The holist's escape re-uses 2.18: pratyekam tu samapto 'rthah — the meaning is *complete in each* occurrence — and sahabhutesu vartate, operates in them as co-present: neither token is a fraction or an echo; each is a station of the whole meaning, their plurality the utterance's mode of presenting one content under co-presence (as each glance sees the whole painting, 2.8). The dilemma dissolves because it presupposed deposit-arithmetic — count the tokens, count the contributions — and the whole-at-each principle refuses the count.

2.116 unformalized

अविकल्पेऽपि वाक्यार्थे विकल्पा भावनाश्रयाः ।
अत्राधिकरणे वादाः पूर्वेषां बहुधा गताः ॥ ११६ ॥

avikalpe'pi vākyārthe vikalpā bhāvanāśrayāḥ |
atrādhikaraṇe vādāḥ pūrveṣāṃ bahudhā gatāḥ || 116 ||

Though the sentence-meaning is free of construction, the constructions rest upon the bringing-into-being; on this topic the doctrines of the ancients have gone many ways.

The pivot into the kanda's center. The vakyartha itself is avikalpa — free of conceptual construction: one, undivided, given whole (everything since 2.7 has argued it). The vikalpas — the constructions, the itemized word-meanings and relations analysis reports — bhavana-asrayah: rest upon *bhavana*, the bringing-into-being — the cognitive realization-process in the hearer (2.27's bhavana-samaya, where the sequenceless capacity works). Constructions belong to the coming-to-understand, not to the understood: process-artifacts, real as process, projected as content. And the honest doxographic register resumes: atra adhikarane, on this very topic — the relation of constructed articulation to unconstructed meaning — the purvesam vadah, the ancients' doctrines, bahudha gatah, have gone many ways. The next verses (2.117-135) canvass them: pratibha by practice, convention, the denotation-catalogue — the tradition's whole semantics laid out before Bhartrhari's own resolution (2.143ff.). The avikalpa/vikalpa pair also silently aligns the kanda with the perception-debate (nirvikalpaka/savikalpaka): sentence-understanding as the grammarian's pratyaksa, given whole before construction — against the Buddhist, for whom construction *is* the verbal; the battle of 1.123 rejoined on new ground.
contested: bhavana we render 'bringing-into-being' (the realization-process); the term also means 'meditative cultivation' and 'the efficacy-toward-being' of Mimamsa verb-theory — Bhartrhari plausibly hears all three; the Mimamsa resonance matters at 2.124-125

2.117 unformalized

अभ्यासात् प्रतिभाहेतुः शब्दः सर्वोऽपरैः स्मृतः ।
बालानां च तिरश्चां च यथार्थप्रतिपादने ॥ ११७ ॥

abhyāsāt pratibhāhetuḥ śabdaḥ sarvo'paraiḥ smṛtaḥ |
bālānāṃ ca tiraścāṃ ca yathārthapratipādane || 117 ||

By others, every word is held to be a cause of intuition through practice — as in the conveying of things as they are to children and to animals.

Pratibha enters — through a reported view, as the kanda's great terms always do. Aparaih smrtam: others hold that sarvo 'pi sabdah, every word, is pratibha-hetuh abhyasat — a cause of *pratibha*, intuitive flash, through *abhyasa*, practice, repeated exposure (the term from 1.35: the jeweller's untransmittable expertise). Their evidence: balanam ca tirascam ca — children and *animals*: the calf finds the udder at a call, the dog answers its name, the infant obeys before it can parse — yathartha-pratipadane, in being brought to things as they are. Comprehension without grammar, lexicon, or inference: word acting on prepared life directly, waking a knowing that was never taught propositionally. The view's radicalism: it makes *practice-formed intuition* the entire mechanism of understanding — no deposits, no assembly, just trained flash — dissolving semantics into a natural history of responsiveness. Bhartrhari will not leave it here (2.118-119 immediately press its edges), but he has chosen his reporters carefully: the animal-and-infant datum is *unanswerable* by deposit-semantics (what lexicon does the calf consult?), and pratibha, entering as another school's word, will by 2.143-152 have become his own doctrine's name — abhyasa-formed intuition generalized into the very form of sentence-understanding, and (at 2.146-152) of skill and instinct across the living world.
contested: whose view 2.117 reports is unrecorded; the abhyasa-mechanism suggests naturalist strands the tradition preserves namelessly. That Bhartrhari ultimately *adopts* pratibha (2.143) while here attributing it is his standard doxographic choreography (cf. 2.30), not evidence against adoption

2.118 unformalized

अनागमश्च सोऽभ्यासः समयः कैश्चिदिष्यते ।
अनन्तरमिदं कार्यमस्मादित्युपदर्शनम् ॥ ११८ ॥

anāgamaśca so'bhyāsaḥ samayaḥ kaiścidiṣyate |
anantaramidaṃ kāryamasmādityupadarśanam || 118 ||

And that practice, by some, is held to be convention, without received tradition — the showing that 'immediately after this, this is to be done'.

The practice-doctrine given its conventionalist reading, and quietly its most subversive one. Kaiscit: some construe the abhyasa of 2.117 as samaya — *convention*, agreement — and anagama, *without received tradition*: not the beginningless transmission Kanda I sanctified (1.27-31, 1.141) but mere settled usage, upadarsanam, an ostensive showing: 'anantaram idam karyam asmat' — after this (word), this is to be done. Language-learning as drill: word paired with expected act, the pairing exhausted by regularity — no eternal sambandha (1.23), no sista-lineage, just coordination grown stable. The view is the ancient ancestor of every use-theory of meaning: to understand a word is to know what to do next. Bhartrhari reports it without immediate rebuttal because its half-truth is one he needs (comprehension *is* trained responsiveness; the calf of 2.117 proves it) while its frame — convention *instead of* tradition — collides with his deepest commitments; 2.119 lets the collision surface on the conventionalist's own terrain: the words whose 'expected act' no drill could ever exhibit.
contested: anagamah we read as qualifying the abhyasa ('a practice without agama') — the conventionalist's denial of tradition; it has been read as a separate charge ('and that practice is un-traditional') in Bhartrhari's own voice, making the verse rebuttal rather than report — the sequel favors report

2.119 unformalized

अस्त्यर्थः सर्वशब्दनामिति प्रत्याय्यलक्षणम् ।
अपूर्वदेवतास्वर्गैः सममाहुर्गवादिषु ॥ ११९ ॥

astyarthaḥ sarvaśabdanāmiti pratyāyyalakṣaṇam |
apūrvadevatāsvargaiḥ samamāhurgavādiṣu || 119 ||

'All words have a meaning' — such is the mark of the conveyable; for 'apurva', 'deity' and 'heaven', they say, it is the same as for 'cow' and the rest.

The conventionalist's ostension-model tested against the words that break it — and the tradition's own semantics stated in the testing. The maxim: asti arthah sarva-sabdanam — all words have meaning; meaningfulness is the pratyayya-laksana, the defining mark of the conveyable (a 'word' that conveyed nothing would not be a word, 2.53). Now the test-set: apurva — the unseen ritual potency, the efficacy that bridges rite and remote fruit (Mimamsa's central posit); devata — the deity as injunction addresses it; svarga — heaven, the promised fruit no living speaker has seen. For these, they say, the semantics is samam gavadisu — *the same as for 'cow'*: meaning attaches in the ordinary way. But no upadarsana (2.118) ever paired 'svarga' with an exhibited heaven; no drill, no ostension, no expected-next-act exhausts these meanings. If they mean as 'cow' means — and the whole exegetical science (2.64-88) requires it — then meaning-as-such cannot be constituted by exhibition-drill: the conventionalist's mechanism fails precisely where scripture's core vocabulary lives. The verse thus turns the maxim of universal meaningfulness into an argument: semantics must be of a kind that covers the unexhibitable — which trained-flash pratibha (waked by words, not shown by objects) and eternal-sambandha tradition can claim, and samaya-drill cannot.
contested: whether 2.119 is the conventionalist's own boast (our reading: 'the same as for cow', which 2.120 then dismantles) or already the rebuttal is delicately poised; either way the apurva-set is the fulcrum, and the tradition reads the verse as the test it plainly is

2.120 unformalized

प्रयोगदर्शनाभ्यासादाकारावग्रहस्तु यः ।
न स शब्दस्य विषयः स हि यत्नान्तराश्रयः ॥ १२० ॥

prayogadarśanābhyāsādākārāvagrahastu yaḥ |
na sa śabdasya viṣayaḥ sa hi yatnāntarāśrayaḥ || 120 ||

But the grasping of the form through practice in observing usage — that is not the word's object; for that rests upon a different effort.

The discrimination that saves pratibha from its conventionalist reading — task 2.118 set, here discharged with one clean cut. Two things practice yields, and they must not be confused. Prayoga-darsana-abhyasat, from practice in observing usage, comes akara-avagraha — the grasping of the *form*: the learner's growing command of shapes, collocations, the sound-habits of the community (2.118's drill, real and necessary). But: na sa sabdasya visayah — that is not the *word's object*, not meaning; sa hi yatna-antara-asrayah, it rests on a *different effort* — the effort of form-acquisition, distinct in kind from the semantic waking the word performs (2.117). Learning the language and understanding the utterance are two achievements: drill builds the instrument; meaning happens when the built instrument is played (the fan of 1.116 must exist — and its existing is not the breeze). The conventionalist's error is a level-confusion: he watched the form-drill and called it semantics. With the cut made, both truths stand in order: abhyasa conditions comprehension (against intellectualism), meaning outruns every drill (against conventionalism, 2.119's test-words now accounted: their *forms* are drilled like any; their meanings wake as all meanings do — by pratibha, whose theory now has the field). The catalogue of denotation-doctrines (2.121ff.) follows as the last survey before that theory is named in full.
contested: yatna-antara ('a different effort') we take as form-learning versus meaning-grasp; some construe 'dependent on another's effort' (the teacher's) — the level-cut survives either construal

2.121 · The Catalogue of Denotation unformalized

केचिद् भेदाः प्रकाश्यन्ते शब्दैस्तदभिधायिभिः ।
अनुनिष्पादिनः कांश्चिच्छब्दार्थ इति मन्वते ॥ १२१ ॥

kecid bhedāḥ prakāśyante śabdaistadabhidhāyibhiḥ |
anuniṣpādinaḥ kāṃścicchabdārtha iti manvate || 121 ||

Some differences are illuminated by words as their expressors; certain others, arising-in-consequence, are (also) thought to be word-meaning.

The denotation-catalogue opens with its governing distinction: the *expressed* versus the *entailed*. Kecid bhedah — some determinations — prakasyante sabdaih tad-abhidhayibhih: are illuminated by words as their proper expressors — what the word is *for*, its abhidheya (2.113). Kamscit — others — anunispadinah: *arising-in-consequence*, following along — determinations that come to cognition with the expressed without being expressed: say 'cow' and the hearer has also, willy-nilly, singular number, presence, generic shape. Are these consequences *meaning*? Manvate — some think so, stretching sabdartha to cover the whole cognitive wake; others (the verses following) draw the line tight. The distinction is the catalogue's instrument: each doctrine surveyed (2.122-135) is a different placement of the expressed/entailed boundary — jati-theorists parking individuals as anunispadin (2.122), form-theorists parking shapes (2.123), action-theorists parking usage (2.124) — and Bhartrhari's eventual verdict (2.134-135: comprehension varies per hearer; the boundary is drawn by vivaksa and construal, not found) is already latent in the fact that the boundary *can* be disputed at all.
contested: anunispadin ('arising-in-consequence') is the catalogue's coinage; 'concomitant', 'supervenient', 'resultant' all serve — the term marks the entailed-but-unexpressed, and our rendering keeps its processual flavor

2.122 unformalized

जातेः प्रत्यायके शब्दे या व्यक्तिरनुषङ्गिणी ।
न तद् व्यक्तिगतान् भेदान् जातिशब्दोऽवलम्बते ॥ १२२ ॥

jāteḥ pratyāyake śabde yā vyaktiranuṣaṅgiṇī |
na tad vyaktigatān bhedān jātiśabdo'valambate || 122 ||

When the word conveys the universal, the individual comes along as an adjunct; the universal-word does not lean upon the differences residing in the individual.

The jati-doctrine, catalogued first as the grammarians' default (1.93's kin, Vajapyayana's line in the tradition's memory). The word expresses the universal: 'gauh' means cowhood. The vyakti, the individual — since cowhood stands nowhere but in cows — anusangini, comes along as adjunct: entailed, not expressed (2.121's boundary drawn at the genus). Consequence, stated as the doctrine's signature: na tad-vyakti-gatan bhedan jati-sabdo 'valambate — the universal-word does not *lean on* the differences residing in the individual: its semantic weight never rests on this cow's color, size, or station; those enter comprehension (if at all) as the adjunct's baggage, available to context, invisible to the word. The doctrine's strength is 2.69's avivaksa made systematic (unmeant particulars, principled); its strain shows where language *does* reach individuals ('bring *this* cow') — deixis, names, the whole apparatus the vyakti-theorists (Vyadi's line) built their rival case on. The catalogue holds both in suspension; the kanda's own resolution (individuation by employment, 2.104-106) has already dissolved the question's presupposition — that the word, rather than the utterance-act, is where denotation is fixed.

2.123 unformalized

घटादीनां न चाकारान् प्रत्ययति वाचकः ।
वस्तुमात्रनिवेशित्वात् तद्गतिर्नान्तरीयकी ॥ १२३ ॥

ghaṭādīnāṃ na cākārān pratyayati vācakaḥ |
vastumātraniveśitvāt tadgatirnāntarīyakī || 123 ||

Nor does the expressor convey the shapes of pots and the like; because it is installed upon the bare thing, the cognition of those is inseparably-consequent.

The boundary-instrument applied to a subtler case: shape. Say 'ghata', pot. The hearer's cognition arrives *shaped* — a bulging-bellied, narrow-necked something stands in it. Did the word convey the akara? The doctrine here catalogued answers no: na akaran pratyayati vacakah — the expressor does not convey the shapes; vastu-matra-nivesitvat, it is installed upon the *bare thing* — the pot as such, unconfigured — and tad-gatih nantariyaki: the shape-cognition is *inseparably-consequent* (nantariyaka, 2.81's term for the rite's inseparable elements): you cannot cognize pot bare, so shape rides in of necessity — entailed by the cognitive act's own conditions, expressed by nothing. The analysis is fine: it distinguishes what the word semantically carries from what cognition must supply to host the carriage — the imagery, the default configuration, the mind's obligatory furnishing. A hearer's rich pot-image is thus no evidence of a rich pot-meaning: semantics thin, hosting thick. The verse arms the catalogue's eventual moral (2.134-135): since hosting varies per mind while the word stands one, the seeming variety of 'what the word brings' is the variety of its receptions.
contested: vastu-matra ('the bare thing') is the doctrine's load-bearing posit — thing-sans-configuration — and its coherence is precisely what rival schools (Buddhist apoha-theorists above all) deny; the karika catalogues without defending it

2.124 unformalized

क्रिया विना प्रयोगेण न दृष्टा शब्दचोदिता ।
प्रयोगस्त्वनुनिष्पादी शब्दर्थ इति गम्यते ॥ १२४ ॥

kriyā vinā prayogeṇa na dṛṣṭā śabdacoditā |
prayogastvanuniṣpādī śabdartha iti gamyate || 124 ||

An action enjoined by word is never seen apart from employment; the employment, however, arising-in-consequence, is understood as the word's meaning.

The boundary-instrument at the verb — where Mimamsa lives. Kriya sabda-codita, action word-enjoined ('let him sacrifice'), is never seen vina prayogena, apart from *employment*: an enjoined action exists only as deployed — by an agent, with means, in performance; the bare act-type never walks alone. So what does the verb mean? The doctrine catalogued: the prayoga, the employment — arising anunispadi, in-consequence (the act-type cannot be hosted without it, 2.123's structure at the verb) — *is understood as* the word's meaning: comprehension takes the verb to mean the deployed act, agent-and-means included, because that is the only form in which action is cognizable. The placement is the inverse of 2.122-123: there the entailed was excluded from meaning (individual, shape); here the entailed is *absorbed into* meaning (employment) — the catalogue exhibiting, by design, that the expressed/entailed boundary migrates doctrine by doctrine. The Mimamsa resonance is loud: bhavana, the verb's efficacy-toward-being, with its trio of what-whereby-how, is exactly an employment-semantics of the verb — and 2.125 next catalogues its restriction.
contested: we read prayoga as 'employment/deployment' (the act as performed); the word also means 'usage' (the linguistic employment), yielding 'usage is understood as meaning' — a use-theory entry in the catalogue; the kriya-context favors the performative sense

2.125 unformalized

नियतास्तु प्रयोगा ये नियतं यच्च साधनम् ।
तेषां शब्दभिधेयत्वमपरैरनुगम्यते ॥ १२५ ॥

niyatāstu prayogā ye niyataṃ yacca sādhanam |
teṣāṃ śabdabhidheyatvamaparairanugamyate || 125 ||

But the employments that are fixed, and the means that is fixed — of these, expressibility by the word is followed by others.

The employment-semantics restricted to its lawful domain. Not every deployment enters the verb's meaning — only niyatah prayogah, the *fixed* employments, and the niyatam sadhanam, the *fixed* means: what the injunction's own economy determines (the rite's constitutive performances, its textually-settled instruments — 2.47's niyama, 2.64-70's obtained-and-specified strata). Of *these*, aparaih anugamyate, others hold, sabda-abhidheyatvam — expressibility by the word: the verb means its act *cum* its fixed circumstantials; the variable ones (this priest, that morning) stay outside as mere hosting (2.123). The restriction rescues 2.124 from absurdity (else 'sacrifice' would mean every accident of every performance) and completes a small system: meaning = act-type + textually-fixed frame; entailment = whatever performance adds. It is the exegete's sruti/linga stratigraphy (2.73-74) reappearing as verb-semantics — and its lesson for the catalogue is the constant one: every workable denotation-doctrine ends by indexing meaning to *fixity*, to what the utterance-economy settles — fixity being precisely what the undivided vakyartha distributes (2.88), not what word-atoms possess.

2.126ab unformalized

समुदायोऽभिधेयो वाऽप्यविकल्पसमुच्चयः ।

samudāyo'bhidheyo vā'pyavikalpasamuccayaḥ |

Or the aggregate is the expressible — a cumulation free of alternative-construction; or else an unreal commingling is, by some, held to be the word's meaning.

The catalogue reaches the composite-doctrines — and, in its second line, crosses a threshold. First: samudayo 'bhidheyah — the *aggregate* is what is expressed: word-meaning as the whole bundle (thing + universal + shape + number...), taken avikalpa-samuccaya, as a cumulation *without* alternative-construction: all-at-once, no either/or among aspects — the maximalist placement of 2.121's boundary (everything the cognition hosts, meaning owns). Then the leap: asatyo va api samsargah — or an *unreal commingling*: the word's meaning is a samsarga, a mixture — of thing and form, referent and concept — that is *asatya*, not real: a constructed composite answering to nothing's actual structure, thrown up by language itself. With that entry the catalogue admits semantic *irrealism*: meanings as useful unrealities — and Bhartrhari's own deep teaching (the vikalpas as bhavana-artifacts, 2.116; the constructed circle of 1.129) stands suddenly among the catalogued, wearing another school's name. The next verse completes the irrealist family and adds its subtlest member.
contested: the source's numbering staggers here (126ab with 127cd printed together); we follow the evident karika-boundaries. Asatya-samsarga is glossed by some as the Buddhist-leaning entry (meaning as constructed fusion), by others as a grammarian's own limit-doctrine — the catalogue's silence on attribution is likely deliberate NOTE: wisdomlib paginates 2.126-127 as half-verses with a duplicated '2.127cd' label; commentary for each verse is attached to its 'ab' page and covers the full verse.

2.127ab unformalized

असत्योपाधि यत्सत्यं तद्वा शब्दनिबन्धनम् ।

asatyopādhi yatsatyaṃ tadvā śabdanibandhanam |

Or that real which has the unreal as its adjunct is the word's fastening; or the word itself, having come to the condition of identifying-speech, attains expressibility.

The irrealist family completed by its two profoundest members. First: asatya-upadhi yat satyam — the *real with unreal adjunct*: the word fastens (nibandhana) to something genuinely real — but only as dressed in unreal upadhis, the limiting-conditions construction supplies (the satya here inevitably recalling 1.1's Brahman under avidya's upadhis: word-meaning as reality-wearing-constructions, semantics as the microcosm of vivarta). Second, stranger and destined for a future: sabdo va api abhijalpatvam agatah yati vacyatam — or the *word itself*, come to the condition of *abhijalpa*, identifying-speech, attains expressibility: the word means... itself-as-fused-with-its-object; meaning is the word in a special reflexive condition. The term abhijalpa is defined in the very next verse — the catalogue pausing, uniquely, to unpack one of its own entries: a signal of weight. Between them the two doctrines bracket the deepest question the catalogue holds: whether the unreality in meaning belongs to the *object-side* (real thing, constructed dress) or to the *word-side* (the word itself, self-fused, doing duty as its own meaning) — and Bhartrhari's system, which identifies word and meaning at the root (2.31), has stakes in both.
contested: the source's segmentation of 126-127 is disturbed (halves printed under both numbers); our reconstruction follows the sense-units. Abhijalpa is rare enough that 2.128's definition is the tradition's anchor for it

2.127cd unformalized

असत्यो वाऽपि संसर्गः शब्दार्थः कैश्चिदिष्यते ॥ १२६ ॥
शब्दो वाप्यभिजल्पत्वमागतो याति वाच्यताम् ॥ १२७ ॥

asatyo vā'pi saṃsargaḥ śabdārthaḥ kaiścidiṣyate || 126 ||
śabdo vāpyabhijalpatvamāgato yāti vācyatām || 127 ||

[Second half-lines of 2.126 and 2.127 respectively — see the commentary at 2.126ab and 2.127ab, which treat each verse entire.]

A note on the source rather than on the text. The wisdomlib edition paginates verses 2.126 and 2.127 as half-verses, and — evidently in error — assigns the identical label '2.127cd' to two different pages: one carrying the second half-line of 2.126 (asatyo vapi samsargah sabdarthah kaiscid isyate), the other the second half-line of 2.127 (sabdo vapy abhijalpatvam agato yati vacyatam). Because this edition keys its commentary to the source's own page labels, a single entry necessarily serves both pages. The substantive commentary for each verse is given complete at 'Verse 2.126ab' and 'Verse 2.127ab', where each karika is translated and discussed in full. Readers consulting the source directly should be aware that its labelling here cannot be relied upon.

2.128 unformalized

सोऽयमित्यभिसम्बन्धाद् रूपमेकीकृतं यदा ।
शब्दस्यार्थेन तं शब्दमभिजल्पं प्रचक्षते ॥ १२८ ॥

so'yamityabhisambandhād rūpamekīkṛtaṃ yadā |
śabdasyārthena taṃ śabdamabhijalpaṃ pracakṣate || 128 ||

When, through the connection 'that is this', the word's form is made one with the meaning — that word they declare to be abhijalpa, identifying-speech.

The promised definition, and one of the kanda's psychological gems. Take the recognition-structure so 'yam iti abhisambandhat — 'that is this' (2.40's own formula: the incoming identified with the possessed): when by it the sabdasya rupam, the word's form, ekikrtam arthena, is *made one with* the meaning — when 'cow' and cow have fused in the speaker's cognition so thoroughly that the word is transparently its thing, thing wordlike, word thinglike — tam abhijalpam pracaksate: that fused condition of the word is abhijalpa. The definition names ordinary semantic consciousness: the mature speaker does not experience word and meaning as paired items (only the analyst, and the learner mid-drill, do); he experiences *one* — the meant-word, the worded-thing (1.50's double shining, lived as fusion; the identification samjna/consciousness pun of 1.126 enacted). The abhijalpa-doctrine of 2.127 then reads: what a word expresses is itself-in-fusion — meaning as the word's achieved transparency. And the fusion explains the catalogue's whole disarray (2.129-130 draw exactly this): because word and meaning present as one, every school's boundary-dispute (which side owns what) is a quarrel over joints that lived speech does not exhibit — the joints appear only under vivaksa-driven analysis, which is where the next verse takes the ledger.
contested: none of substance in the definition; its reach — whether abhijalpa names a special condition (citation-like) or ordinary comprehension as such — divides readers, and we take the ordinary-condition reading with the movement's drift

2.129-130 unformalized

तयोरपृथागर्थत्वे रूढेरव्यभिचारिणि ।
किञ्चिदेव क्कचिद् रूपं प्राधान्येनावतिष्ठते ॥ १२९ ॥
लोकेऽर्थरूपतां शब्दः प्रतिपन्नः प्रवर्तते ।
शास्त्रे तूभयरूपत्वं प्रविभक्तं विवक्षया ॥ १३० ॥

tayorapṛthāgarthatve rūḍheravyabhicāriṇi |
kiñcideva kkacid rūpaṃ prādhānyenāvatiṣṭhate || 129 ||
loke'rtharūpatāṃ śabdaḥ pratipannaḥ pravartate |
śāstre tūbhayarūpatvaṃ pravibhaktaṃ vivakṣayā || 130 ||

(129) Since their non-separate meaningfulness is unswerving by established currency, some one form, somewhere, stands as principal; (130) in the world, the word operates having assumed the form of the meaning; in the science, however, the two-formedness is divided out according to intention.

The fusion-doctrine made stratigraphic, and the catalogue's quarrels dissolved in two strokes. (129): tayoh aprthag-arthatve — word and meaning being non-separately meaningful (the abhijalpa-fusion), and that non-separation avyabhicarin, unswerving, secured by rudhi, established currency (2.128's 'that is this' hardened by usage, 1.128) — still, kimcid eva kvacid rupam pradhanyena avatisthate: some one form, in some context, stands as *principal* — now the thing-face (ordinary talk), now the word-face (quotation, grammar): salience rotates within the fusion without breaking it. (130) then draws the two-register conclusion: loke — in the world — the word operates artha-rupatam pratipannah, *having assumed the meaning's form*: transparent, self-effacing (1.54's unnoticed instrument; the world sees through words to things). Sastre tu — in the science — ubhaya-rupatvam pravibhaktam vivaksaya: the *two*-formedness is divided out, by intention: the analyst's vivaksa splits the fusion, holds word-face and thing-face apart, and works the seam (svam rupam, 1.59-68; the grahana-craft of 2.98-100). Between loka and sastra the whole catalogue is adjudicated: every doctrine surveyed was a sastra-posture — a way of splitting for a purpose — while lived meaning is the unsplit fusion; the boundary of expressed and entailed (2.121) is drawn by vivaksa, wherever the analysis needs it, and *that is all the drawing there is*.
contested: 129's syntax is dense (aprthag-arthatve rudher avyabhicarini as locative frame) and our construal follows the movement; some take rudhi as subject ('currency being unswerving in their non-separation'). The loka/sastra division of 130 is textually pristine and is the passage's yield

2.131 unformalized

अशक्तेः सर्वशक्तेर्वा शब्दैरेव प्रकल्पिता ।
एकस्यार्थस्य नियता क्रियादिपरिकल्पना ॥ १३१ ॥

aśakteḥ sarvaśaktervā śabdaireva prakalpitā |
ekasyārthasya niyatā kriyādiparikalpanā || 131 ||

Whether from incapacity or from omnipotence, the fixed construction of action and the rest, for one single meaning, is constructed by words alone.

The ledger's radical closing entry before the perception-turn: even the categories are wordwrought. The parikalpana — the construction of kriya-adi, action-agent-instrument-and-the-rest (the karakas, the very architecture of 2.48): for ekasya arthasya, one single meaning (the undivided vakyartha — one event, say), a *fixed* construal into doer-doing-done is laid down — sabdaih eva prakalpita, constructed *by words alone*. The event itself is one; that it comes carved as agent-acting-on-patient is the contribution of the sentence-form that presents it (active or passive, this the karta, that the karman — carvings the same event tolerates interchangeably). And the alternative etiologies, offered with a shrug: asakteh — from *incapacity*: language cannot present the undivided except through role-fictions (the mind's part-scaffolding, 1.85, 2.13); sarva-sakteh va — or from *omnipotence*: the word-principle (whose sakti binds the universe, 1.118) freely articulates the one as many. Weakness or sovereignty — the two readings of vivarta itself (1.1's contested edge) — and Bhartrhari declines to choose here as there. Either way the theorem stands: the categorial grid — action, agent, instrument, the furniture of every sastra including his own — is of language, not read off the world; Kanda III's samuddesas (padartha by padartha) will spend six hundred verses honoring exactly this theorem: each category examined *as what words posit*.
contested: asakteh/sarva-sakteh we read as genuine alternatives (incapacity of speakers/language, or omnipotence of the word-power); some assign the first to language and the second to the meaning — the theorem (categories as word-constructions) is unaffected

2.132 unformalized

यो वार्थो बुद्धिविषयो बाह्यवस्तुनिबन्धनः ।
स बाह्यवस्त्विति ज्ञातः शब्दार्थः कैश्चिदिष्यते ॥ १३२ ॥

yo vārtho buddhiviṣayo bāhyavastunibandhanaḥ |
sa bāhyavastviti jñātaḥ śabdārthaḥ kaiścidiṣyate || 132 ||

Or: the meaning that is the object of cognition, fastened to the external thing — that, known as the external thing, is by some held to be the word's meaning.

The catalogue's epistemological entry — the doctrine that meanings are *mental contents mistaken for things*. The artha that words convey is buddhi-visaya, an object *of cognition* — a content in the mind — bahya-vastu-nibandhana, fastened to (occasioned by, answerable to) the external thing; and it is jnatah bahya-vastu iti — *known as* the external thing: comprehension systematically takes the inner content for the outer object. What 'cow' gives you is a cognition-shape, taken as the beast. The doctrine (kaiscid isyate — the Buddhist resonance is unmistakable: meaning as buddhi-pratibhasa, the apoha-tradition's ancestor) is catalogued without polemic because Bhartrhari needs its insight while rejecting its frame: that comprehension traffics in cognition-side contents is his own teaching (2.18's siddhi in comprehenders; 2.123's hosting; pratibha ahead) — that this convicts meaning of *error* (mistaking inner for outer) presupposes the very inner/outer ledger his word-monism dissolves (2.31: word and meaning two faces of one; 1.129: the 'external' circle itself word-built). The next verse presses the entry to its idealist limit — and there the kanda's own metaphysics looks back at the objector.
contested: the doctrine's attribution (Buddhist vijnana-leaning strands, or a grammarian's own buddhyartha-view) is the classical question; the karika's kaiscit preserves the ambiguity and we do too

2.133 unformalized

आकारवन्तः संवेद्या व्यक्तस्मृतिनिबन्धनाः ।
ये ते प्रत्यवभासन्त संविन्मात्रं त्वतोऽन्यथा ॥ १३३ ॥

ākāravantaḥ saṃvedyā vyaktasmṛtinibandhanāḥ |
ye te pratyavabhāsanta saṃvinmātraṃ tvato'nyathā || 133 ||

Those that appear in cognition, possessed of form, knowable, fastened to distinct memory — (so far as they so appear); but beyond that, it is mere awareness, otherwise.

The idealist limit, stated with a precision that leaves the catalogue behind. Whatever presents — the akaravantah, the formed ones: samvedyah, knowable contents, vyakta-smrti-nibandhanah, fastened to distinct memory (recognizable, re-identifiable — the whole apparatus of stable objects) — ye pratyavabhasante: they *appear*, and their being formed-knowable-memorable is exactly their manner of appearing. Atah anyatha: beyond that — behind, beneath the appearing — samvin-matram: *mere awareness*, awareness-only, unformed, unfastened. The verse's economy is severe: objecthood analyzed without remainder into appearance-structure; what is not appearance-structure is bare samvit. Whether this is the catalogued opponent's terminus (vijnana-vada naming itself) or Bhartrhari's own floor showing through (the samvit here and the sabda-tattva of 1.1 differing, on his account, in name only — awareness whose essence is word, 1.126) is genuinely undecidable and probably meant to be: at the limit, the grammarian's word-monism and the idealist's awareness-monism face each other as mirror-doctrines, quarreling over whether the ultimate light is verbal — the one dispute (1.123-124 already staked it) neither can cede. The catalogue ends here, on its deepest entry; 2.134 steps back into daylight with the observation that dissolves catalogues as such.
contested: the verse's voice — opponent's conclusion or author's floor — is the crux stated above; the transcription's pratyavabhasanta we read as present middle. Samvin-matram is the only occurrence of the term in the kanda and its weight is accordingly disputed

2.134 unformalized

यथेन्द्रियं सन्निपतद्वैंचित्र्येणोपदर्शकम् ।
तथैव शब्दादर्थस्य प्रतिपत्तिरनेकधा ॥ १३४ ॥

yathendriyaṃ sannipatadvaiṃcitryeṇopadarśakam |
tathaiva śabdādarthasya pratipattiranekadhā || 134 ||

As the sense-faculty, coming into contact, is a displayer according to its variety, just so from the word the comprehension of the meaning is manifold.

The daylight-observation that adjudicates the whole catalogue: *display varies with the displayer*. The indriya, meeting its object, upadarsakam vaicitryena — displays according to its own variety: the same lamp shows differently to the keen eye and the dim, the jaundiced and the clear (2.105's maxim, now aimed at semantics); no one concludes the lamp is many. Tathaiva sabdat — just so from the word: pratipattir anekadha, comprehension is *manifold* — the same utterance waking richer and poorer, thicker and thinner uptakes per hearer (the child, the expert, the exegete, the provincial of 2.91). The application to the catalogue is immediate and lethal: each denotation-doctrine (2.121-133) canonized *one uptake-profile* — the genus-grasper's, the shape-hoster's, the employment-reader's, the idealist's — and legislated it as 'the word's meaning'. But uptake-profiles are receiver-facts (2.123's hosting, 2.135 next: division per own-cognition); the word, like the lamp, is one, and the doctrines were a physiology of hearers mistaken for a semantics of words. What remains constant under all the variety — the one thing every uptake is an uptake *of* — is the undivided vakyartha; and the flash in which any hearer, at his own grain, seizes it whole now finally receives its name (2.143: pratibha).

2.135 unformalized

वक्त्रान्यथैव प्रक्रान्तो भिन्नेषु प्रतिपत्तृषु ।
स्वप्रत्ययानुसारेण शब्दार्थः प्रविभज्यते ॥ १३५ ॥

vaktrānyathaiva prakrānto bhinneṣu pratipattṛṣu |
svapratyayānusāreṇa śabdārthaḥ pravibhajyate || 135 ||

Set going by the speaker in one way entirely, among differing comprehenders the word's meaning is divided out, each according to his own cognition.

The catalogue's verdict completed on the speaker's side, and semantics resituated once for all between two minds. Vaktra anyathaiva prakrantah: launched by the speaker *in one way entirely* — the utterance leaves as one determinate act, the speaker's vivaksa whole in it (1.155's criterion; 2.130's intention). Bhinnesu pratipattrsu: among differing comprehenders it sva-pratyaya-anusarena pravibhajyate — is divided out, each according to his own cognition: every hearer carves the one launch at his own grain, with his own furnishing (2.123), his own training (2.117), his own purposes — the manifold of 2.134 given its mechanism. Between the one launching and the many carvings, where does 'the meaning' live? The kanda's whole architecture answers: in neither ledger as the doctrines kept it — not in word-atoms (dissolved), not in hearer-constructions (variable), but in the undivided act that launching and uptake both answer to: the vakyartha as the invariant of communication, seized whole (never carved the same way twice) by the flash each comprehension is. The two-mind picture also quietly prices error and depth: uptake can fall short of launch (the mudha of 2.91), exceed it (the exegete's construals), or meet it — and meeting it is not matching carvings but sharing the whole: the samepratibha waked. The stage for 2.143 is fully set; 2.136-142 first generalize the receiver-relativity to *seeing itself*, so that when pratibha is named it will be named for perception and speech alike.
contested: anyathaiva ('in one way entirely / quite otherwise') is pointed: some read it as 'otherwise (than the hearers take it)' — launch and uptake never coinciding; our reading ('in one determinate way') keeps the invariant the movement needs, and the two shade together in any case

2.136 unformalized

एकस्मिन्नपि दृश्येऽर्थे दर्शनं भिद्यते पृथक् ।
कालान्तरेण चैकोऽपि तं पश्यत्यन्यथा पुनः ॥ १३६ ॥

ekasminnapi dṛśye'rthe darśanaṃ bhidyate pṛthak |
kālāntareṇa caiko'pi taṃ paśyatyanyathā punaḥ || 136 ||

Even regarding a single visible object, seeing divides, each separately; and with the passage of time, even the same one sees it otherwise again.

Receiver-relativity generalized to perception — the first of the movement's two supports (2.137 gives the word-side) for reading comprehension-variety as unalarming. Ekasminn api drsye 'rthe: one visible thing, publicly there. Darsanam bhidyate prthak: the seeings divide, observer by observer — the painter, the timber-merchant and the botanist before one tree; the lover and the anatomist before one face. Kalantarena ca eko 'pi: and across time even the *same* seer — tam pasyaty anyatha punah: sees it otherwise again; yesterday's tree is not today's to the same eyes grown older. If seeing itself — the pramana of pramanas — is thus per-seer and per-time without anyone doubting the tree's unity, then comprehension's per-hearer variety (2.134-135) convicts language of nothing: it exhibits the common condition of all uptake of the one by the many. The verse is 2.105's maxim (difference of grasping certifies no difference in the grasped) turned from defense into framework — and it prepares the movement's harder edge (2.138-141): if all seeing is perspectival, the authority of *any* seeing, including perception's, needs auditing — which the seers' case (2.139) and the sky's false floor (2.140) then conduct.

2.137 unformalized

एकस्यापि च शब्दस्य निमित्तैख्यवस्थितैः ।
एकेन बहुभिश्चार्थो बहुधा परिकल्प्यते ॥ १३७ ॥

ekasyāpi ca śabdasya nimittaikhyavasthitaiḥ |
ekena bahubhiścārtho bahudhā parikalpyate || 137 ||

And of even a single word, by the fixed occasioning conditions, the meaning is constructed manifoldly — by one (hearer) and by many.

The word-side support, completing the pair. Ekasya api sabdasya: one word, self-same. Nimittaih avasthitaih — by the occasioning conditions, themselves *fixed* (the contexts, constructions, employments: 2.148's nimitta-bheda, the settled apparatus of use) — arthah bahudha parikalpyate: the meaning is constructed manifoldly — ekena bahubhis ca: by one hearer (across occasions, moods, ages: 2.136's diachrony imported) and by many (across the room: 2.135's synchrony). Note the care: the nimittas are avasthita, *fixed* — the variety is not anarchy; construals are occasioned, rule-shadowed, answerable (the same fixities the exegete's science maps, 2.64-88) — and yet within the fixities, parikalpyate: construction, per mind, irreducibly. Law without uniformity: the shape of every practice the kanda has audited (grammar, exegesis, perception) and now of meaning itself. With both supports standing — seeing divides (2.136), construal divides under fixed occasions (2.137) — the movement can draw its severe epistemological conclusions: what *never* divides is not given to ordinary seers and speakers at all (2.138-139), the given teems with false floors (2.140), and the wise must therefore audit even perception by reason (2.141) while leaving the world's working namings undisturbed (2.142). Then, against that chastened background, the one positive absolute the kanda will assert: the flash (2.143).

2.138-139 unformalized

तस्माददृष्टतच्वानां सापराधं बहुच्छलम् ।
दर्शनं वचनं चापि नित्यमेवानवस्थितम् ॥ १३८ ॥
ऋषीणां दर्शनं यच्च तच्वे किञ्चिदवस्थितम् ।
न तेन व्यवहारोऽस्ति न तच्छब्दनिबन्धनम् ॥ १३९ ॥

tasmādadṛṣṭatacvānāṃ sāparādhaṃ bahucchalam |
darśanaṃ vacanaṃ cāpi nityamevānavasthitam || 138 ||
ṛṣīṇāṃ darśanaṃ yacca tacve kiñcidavasthitam |
na tena vyavahāro'sti na tacchabdanibandhanam || 139 ||

(138) Therefore, for those who have not seen reality, seeing and speech alike are at fault, full of traps, forever unsettled; (139) and the seers' vision, which is somewhat settled in reality — by it there is no worldly practice, nor is it fastened to words.

The movement's severe theorem: between fallible practice and settled vision there is no overlap. (138): for adrsta-tattvanam, those who have not seen reality — all of us in the ordinary case — both darsana and vacana, seeing and saying, are sa-aparadham, at fault, bahu-cchalam, full of traps (chala: the debater's snare — our very perceptions and phrasings bait us), nityam eva anavasthitam: *forever unsettled* — the per-seer, per-time, per-hearer divisions of 2.136-137 now pronounced as chronic condition, not occasional noise. (139): and the rsis' darsana — which is kimcid avasthitam tattve, somewhat settled in reality (the vision-knowledge of 1.37-38, its authority intact) — na tena vyavaharah asti: by it *no worldly practice runs* — the settled seeing does not transact, buy, teach, or litigate; and na tat sabda-nibandhanam: nor is it fastened to words — what the seer sees whole does not parcel into speech (1.145's dream-mode showing; the param padam beyond the three, 1.142). The theorem cuts both ways with equal edge: practice cannot borrow vision's settledness (the market runs on trapped seeing and unsettled saying — and runs); vision cannot fund practice (settled, it falls silent). Between them stands everything the two kandas have built: language as the unsettled-but-lawful medium (2.137) in which unseeing beings transact — redeemable not by importing the seers' silence but by the discipline of the word itself (1.14, 1.131). 2.140 next gives unsettled seeing its emblem; 2.141-142 give the wise their protocol for living in it.
contested: kimcid ('somewhat') in 139 is a studied qualification — even the rsis' vision is only *somewhat* settled, or settled *in some respect*; readers who take it as 'a certain vision, settled' remove the hedge; we keep it — the theorem is harsher and more Bhartrharian with it

2.140 unformalized

तलबद् दृश्यते व्योम खद्योतो हव्यवाडिव ।
नैव चास्ति तलं व्योन्मि न खद्योतो हुताशनः ॥ १४० ॥

talabad dṛśyate vyoma khadyoto havyavāḍiva |
naiva cāsti talaṃ vyonmi na khadyoto hutāśanaḥ || 140 ||

The sky is seen as if it had a floor; the firefly, like a fire-carrier; and there is no floor in the sky at all, nor is the firefly fire.

The emblem: perception's confident deliverances of what is not there. Talavat drsyate vyoma — the sky *looks floored*: the vault, the dome, the blue ceiling every eye reports and no sky possesses; khadyotah havyavad iva — the firefly like a fire-bearer: a spark, to all seeing, and no fire at all. Na eva ca asti talam vyomni, na khadyotah hutasanah: flatly — no floor, no fire. The examples are chosen as *universal and incorrigible* illusions: not the sick eye's jaundice (2.105) or the fool's fancy (2.91) but percepts every healthy observer shares and no correction dispels — one still *sees* the dome after learning there is none. Perception at its best, unanimous and stable, can thus be false to fact: unanimity is no truth-mark, stability no floor (the pun is the verse's own). This is 1.129's alata-cakra radicalized — there, construction where nothing was; here, construction *canonized in every normal seeing* — and it discharges the movement's last debt: if even agreed perception needs auditing, the variety of comprehension (2.134-137) was never the scandal; the scandal would be trusting any uptake, however shared, without the yukti that 2.141 now prescribes.
contested: none; the verse is limpid and the tradition quotes it widely — its work is emblematic, and the floor/settledness pun (tala/avasthita) may be heard or not without loss

2.141 unformalized

तस्मात् प्रत्यक्षमप्यर्थं विद्वानीक्षेत युक्तितः ।
न दर्शनस्य प्रामाण्यात् दृश्यमर्थं प्रकल्पयेत् ॥ १४१ ॥

tasmāt pratyakṣamapyarthaṃ vidvānīkṣeta yuktitaḥ |
na darśanasya prāmāṇyāt dṛśyamarthaṃ prakalpayet || 141 ||

Therefore the wise man should examine even a perceptible object by reasoning; he should not posit the object seen merely on the authority of seeing.

The protocol, drawn with tasmat-force from the emblem. Pratyaksam api artham — even the *perceptible*, even what stands before the eyes — vidvan ikseta yuktitah: the wise man should examine by reasoning; na darsanasya pramanyat drsyam artham prakalpayet: he should not posit the seen object on seeing's authority alone — the dome and the firefly's fire being exactly what such positing yields. Mark the placement of this rationalism: Kanda I demoted tarka against agama (1.30-42: reasoning cannot found dharma); here tarka is *promoted* against unaudited perception — and the two rulings are one epistemology, not a reversal: reasoning is corrective, never foundational (1.135-137: the eye for the sightless, sovereign nowhere, employed everywhere) — it audits the given (perceptual or textual) without replacing it. The verse's quiet radicalism is its scope: pratyaksam api — the pramana every school treats as archetype is made a defendant; where 2.140 showed *that* seeing errs incorrigibly, 2.141 draws the duty: conviction is earned by cross-examination, not conferred by vividness. For the kanda's own subject the transfer is immediate: comprehension (speech's 'perception') likewise binds only as audited — the exegete's science (2.64-88) being exactly that audit, institutionalized.
contested: none of substance; the verse is the tradition's standing citation for grammarian rationalism, and its harmony with 1.30-42 (corrective vs. foundational tarka) is as we state

2.142 unformalized

असमाख्येयतत्त्वानामर्थानां लौकिकैर्यथा ।
व्यवहारे समाख्यानं तत्प्राज्ञो न विकल्पयेत् ॥ १४२ ॥

asamākhyeyatattvānāmarthānāṃ laukikairyathā |
vyavahāre samākhyānaṃ tatprājño na vikalpayet || 142 ||

The naming that worldly speakers, in practice, give to things whose reality cannot be told — the truly wise man should not construct alternatives against it.

The protocol's companion, guarding it from misuse — audit the given, but do not harass the working namings. Things are, in their tattva, asamakhyeya — *untellable*: reality outruns designation (the seers' settled vision itself sabda-free, 2.139; the untellable expertise of 1.35). Yet laukikaih, worldly speakers, vyavahare, in practice, samakhyanam — name them, and the namings *work*: the market's 'cow', the household's 'fire', the whole serviceable taxonomy of the unseeing. Tat prajnah na vikalpayet: against *that*, the truly wise man constructs no alternatives — he does not correct the world's speech toward metaphysical exactitude, does not replace 'the sky's dome' with disquisitions at the well. The pairing with 2.141 is exact and deliberate: examine even the perceived (rigor inward, where conviction is formed); leave the world's namings standing (charity outward, where practice runs) — for the namings never claimed tattva; they claimed serviceability, and by that claim they are true (vyavahara-satya, the working truth the whole system has honored since 1.28's vyavastha). The wise man's silence before working speech is thus not condescension but accuracy about what kind of instrument language is — and it is the exact temper in which Kanda III will examine the categories: each padartha audited as *posited by words for practice*, none arraigned for failing to be the untellable.
contested: na vikalpayet we render 'should not construct alternatives against (it)'; the verb also yields 'should not subject it to (skeptical) analysis' — the charity-ruling is the same; note the deliberate echo and inversion of 1.29's nanarthika vyavastha (the sista's codification purposeful) — here the *world's* codification receives the same protection

2.143 · Pratibha: The Flash of Meaning unformalized

विच्छेदग्रहणेऽर्थानां प्रतिभाऽन्यैव जायते ।
वाक्यार्थ इति तामाहुः पदार्थैरुपपादिताम् ॥ १४३ ॥

vicchedagrahaṇe'rthānāṃ pratibhā'nyaiva jāyate |
vākyārtha iti tāmāhuḥ padārthairupapāditām || 143 ||

When the meanings have been grasped in their separation, an intuition altogether other is born; that, brought about by the word-meanings, they call the sentence-meaning.

The kanda's summit: the vakyartha named at last, and its name is *pratibha* — the flash. The mechanics first, stated with all the movement's accumulated care: viccheda-grahane arthanam — when the meanings (the padarthas, the scaffold's items) have been grasped *in their separation* (the analyst's and learner's part-wise passage: 1.90's seizures, 2.13's ladder) — pratibha *anyaiva* jayate: an intuition *altogether other* is born — other in kind, not in degree; not the parts' sum, echo, or relational closure but a new indivisible arising (2.55's bhinna-laksana cashed): the understanding itself, the click, the whole dawning at once. Tam vakyartha iti ahuh: *that* they call the sentence-meaning — not a proposition the flash grasps: the flash *is* the meaning, comprehension and content one event (2.31's identity, lived). And the genetic clause, holding both truths without strain: padarthaih upapaditam — *brought about by* the word-meanings: the parts are real causes (upapadita — occasioned, ushered) of what they neither contain nor compose; scaffolding that precipitates the edifice (1.84's ripening; 2.18's per-word re-completion; 2.38's discharged means — every earlier structure converging here). This is the doctrine the opening catalogue promised under 'buddhy-anusamhrti' (2.1) and the whole kanda has been assembling: sentence-meaning as an *event in the comprehender* — indivisible, occasioned, other — the flash in which the trained mind (2.117's abhyasa) meets the launched whole (2.135) and the undivided shows itself through its own dismantled ladder. The verses following will defend its ineffability (it cannot be told to another — only waked), distribute it across the living world (the bird's nest-craft, the beast's know-how: pratibha as the very form of practical intelligence), and ground it finally in the word-essence itself: the flash as the sabda-tattva's self-recognition in a mind — 1.142's sabda-purva yoga arriving at its terminus.
contested: anyaiva is the verse's hinge and we press it fully ('altogether other' — the flash as novel indivisible); a softer reading ('quite different [from the padarthas]') is possible but under-reads the kanda's architecture. That pratibha is here *defined* as vakyartha (not merely compared to it) is the tradition's own reading and ours

2.144 · Pratibha: The Flash of Meaning unformalized

इदं तदिति सान्येषामनाख्येया कथञ्चन ।
प्रत्यात्मवृत्तिसिद्धा सा कर्त्रापिन निरूप्यते ॥ १४४ ॥

idaṃ taditi sānyeṣāmanākhyeyā kathañcana |
pratyātmavṛttisiddhā sā kartrāpina nirūpyate || 144 ||

It cannot in any way be told to others as 'this is that'; established by its own operation within each self, it is not describable even by the one who has it.

Pratibha's first mark: ineffability, and of a peculiarly strict kind. It cannot be told to others idam tad iti — under the 'this is that' formula, the very structure of identifying-communication (2.40, 2.128's so 'yam): to convey a content one must pair it with something the hearer already holds, and the flash pairs with nothing — it *is* the pairing-event, not a term in one. Pratyatma-vrtti-siddha: established by its own operation *within each self* — self-certifying, self-locating, irreducibly first-personal. Then the harder clause, which distinguishes this from ordinary privacy: kartrapi na nirupyate — it is not describable *even by its own possessor*. Not merely 'I cannot convey it to you' but 'I cannot state it to myself': the flash is not an inner object one inspects and then reports badly; there is nothing standing at a distance to inspect. This is the exact structure of 1.35's connoisseur — the jeweller who knows and cannot say what he knows, asamakhyeya — now generalized from expertise to *every act of understanding a sentence*. Every comprehension is connoisseurship. Note what this costs and buys: it costs the theory any hope of *displaying* the vakyartha (one can only occasion it in another, by uttering — which is what speech is); it buys the explanation of why every attempt to state a sentence's meaning yields another sentence, needing its own flash, forever.
contested: kartra ('by the agent/possessor') we read as the one in whom the pratibha arises; some construe 'even by the speaker' (the utterer cannot describe what he wakes) — both readings make ineffability structural rather than merely practical

2.145 unformalized

उपश्लेषमिवार्थानां सा करोत्यविचारिता ।
सार्वरूप्यमिवान्ना विषयत्वेन वर्तते ॥ १४५ ॥

upaśleṣamivārthānāṃ sā karotyavicāritā |
sārvarūpyamivānnā viṣayatvena vartate || 145 ||

Unexamined, it effects as it were a close conjoining of the meanings; and, as though all-formed, it operates as their objecthood.

Pratibha's second mark: it does the work that connection-theories invented sambandha to do (2.42-46), and does it without machinery. Upaslesam iva arthanam sa karoti — it effects, as it were, a *close conjoining* of the meanings: in the flash the items cohere — the cow is brought, the agent acts, the parts hold together — precisely the relatedness the abhihitanvaya theorist had to purchase with an imperceptible, non-thing connective (2.46). And avicarita — *unexamined*: the conjoining is not a judgement one makes but a coherence one finds already made; scrutiny arrives afterward, upon a whole already unified. Sarva-rupyam iva — as though *all-formed*, wearing every relevant form at once (the whole complete in each aspect, 2.18) — visayatvena vartate: it operates *as their objecthood* — the flash is not a cognition *of* objects standing apart; it is that in which the meanings have their standing-as-objects at all. Both iva's are exact: the conjoining is real-as-presented, not an assembly performed; the all-formedness is completeness, not a sum of forms. Where the rivals posited relations between semantic atoms, pratibha *is* the relatedness — self-arising, prior to inspection, constituting rather than linking its terms.

2.146 unformalized

साक्षाच्छब्देन जनितां भावनानुगमेन वा ।
इतिकर्तव्यतायां तां न कश्चिदतिवर्तते ॥ १४६ ॥

sākṣācchabdena janitāṃ bhāvanānugamena vā |
itikartavyatāyāṃ tāṃ na kaścidativartate || 146 ||

Whether generated directly by the word, or through the following of realization — in the matter of what-is-to-be-done, no one at all transgresses it.

Pratibha's third mark: it is *practically sovereign*. Two genetic routes are conceded: saksat sabdena janitam — generated directly by the word (the hearer's flash on being spoken to: the kanda's home case); or bhavana-anugamena — through the following of bhavana, realization and cultivation (2.116's process, 2.117's abhyasa). Either origin, one authority: itikartavyatayam — in the domain of what-is-to-be-done (Kanda I's term for practical know-how, 1.121, there declared wholly word-dependent) — na kascid ativartate: *no one transgresses it*. Not 'no one should' — no one *does*: when the flash says what is called for, action follows its dictate; deliberation neither precedes nor overrides it, because deliberation is itself conducted in flashes. This is the doctrine's turn from semantics toward life. Comprehension, know-how and conduct converge on a single phenomenon: the immediate, untellable, self-certifying seizure by which any agent knows what is to be done. The next verses press exactly this convergence outward — first to the whole human world, then, far more radically, to the animal one.

2.147 unformalized

प्रमाणात्वेन तां लोकः सर्वः समनुपश्यति ।
समारम्भाः प्रतायन्ते तिरश्चामपि तद्वशात् ॥ १४७ ॥

pramāṇātvena tāṃ lokaḥ sarvaḥ samanupaśyati |
samārambhāḥ pratāyante tiraścāmapi tadvaśāt || 147 ||

The whole world regards it as a means of knowledge; and through it the undertakings even of animals are extended.

The sovereignty of 2.146 given its two-fold evidence. First, epistemic standing: pramanatvena tam lokah sarvah samanupasyati — the *entire* world regards pratibha as a pramana, an authoritative means of knowing: not a private feeling but the thing everyone in fact acts on and treats as knowledge — the schools' formal lists (perception, inference, testimony) notwithstanding, the working epistemology of every agent is the flash. Second, and with a jolt: samarambhah pratayante tirascam api tadvasat — undertakings *even of animals* are extended by its power. Tiryanc — beasts, 'the horizontal ones' — has appeared before (1.126: sentience in all kinds; 2.117: children and animals comprehending), but here it becomes the argument's centre of gravity: if the flash is what animals act by, then pratibha is not an achievement of language-users at all — it is the form of practical intelligence as such, and human sentence-comprehension is one dialect of a capacity running through everything alive. The verses following make good on this with a series of examples remarkable for their observational precision.

2.148 unformalized

यथा द्रव्यविशेषाणां परिपाकैरयत्नजाः ।
मदादिशक्तयो दृष्टाः प्रतिभास्तद्वतां तथा ॥ १४८ ॥

yathā dravyaviśeṣāṇāṃ paripākairayatnajāḥ |
madādiśaktayo dṛṣṭāḥ pratibhāstadvatāṃ tathā || 148 ||

As, from the ripening of particular substances, powers such as intoxication are seen to arise without effort — so are the intuitions of those who possess them.

The naturalist model for pratibha's arising, chosen with care. Certain substances, on paripaka — ripening, fermentation — develop mada-adi-saktayah, powers such as intoxication: the grape left to itself becomes wine; nobody *makes* the power, and no ingredient contains it. Ayatna-jah: they arise *without effort* — unwilled, unassembled, the outcome of a thing's own maturation under conditions. Pratibhah tadvatam tatha: so with the intuitions of those who have them. Every clause carries doctrine. Ripening: the flash comes of maturation (abhyasa, 2.117; the seed-and-paripaka of 1.84), not of construction. Effortless: it is not produced by the will that awaits it — one cannot decide to understand; understanding dawns (2.145's avicarita). And *emergent*: intoxicating power is not found in the grape's parts, as the vakyartha is not in the padarthas (2.143's anyaiva) — the fermented whole has a capacity its ingredients lack. Bhartrhari here naturalizes the flash without deflating it: pratibha is a real power of a matured being, as wine's potency is of a matured substance — which is precisely why it can belong to birds as well as to brahmins, as the next verses insist.
contested: tadvatam ('of those who possess them') is neutral as to the possessors' kind — deliberately, given 2.147's animals and 2.149's cuckoo; some construe 'of those who possess (the requisite ripening)', which tightens the analogy further

2.149 unformalized

स्वरवृत्तिं विकुरुते मधौ पुंस्कोकिलस्य कः ।
जन्त्वादयः कुलायादिकरणे केन शिक्षिताः ॥ १४९ ॥

svaravṛttiṃ vikurute madhau puṃskokilasya kaḥ |
jantvādayaḥ kulāyādikaraṇe kena śikṣitāḥ || 149 ||

Who modifies the voice-function of the male cuckoo in spring? By whom are creatures like the spider taught in the making of nests and the like?

Two questions, and their form is the argument. Spring comes; the kokila's song changes — its svara-vrtti, its very voice-function, transformed, on time, without instruction: kah vikurute, *who* does this? The spider and its kin build — kulaya-karane kena siksitah: by *whom* taught? The expected answers are all unavailable. Not imitation (the hatchling spider has watched no one). Not inference (2.34's reasoners, deliberating, could not derive a web). Not agama in the narrow sense (spiders keep no lineage of sistas). Not trial and error (the first web holds). The tradition around these verses reaches for karma and vasana — ripened residues of prior lives — and Bhartrhari's own bhavana-vocabulary (2.146, 2.151) accommodates that. But the verse's philosophical force does not depend on the metaphysics of rebirth: it depends on the *phenomenon*, undeniable and undomesticated by any epistemology of the day — determinate, complex, immediately-known, unlearned competence. Naming it pratibha is less an explanation than a placement: whatever the cuckoo has, the hearer of a sentence has too, and the schools admitting only perception, inference and testimony have no cell for either. A modern reader will recognize the questions as those that eventually produced ethology; Bhartrhari's use of them is to widen the concept of knowing until it is large enough to hold both the web and the word.
contested: jantvadayah we take with the tradition as spiders and such web/nest-builders (kulaya covers both); the list is deliberately open — the argument wants the whole class of unlearned crafts

2.150 unformalized

आहारप्रीत्यपद्वेषल्पवनादिक्रियासु कः ।
जात्यान्वयप्रसिद्धासु प्रयोक्ता मृगपक्षिणाम् ॥ १५० ॥

āhāraprītyapadveṣalpavanādikriyāsu kaḥ |
jātyānvayaprasiddhāsu prayoktā mṛgapakṣiṇām || 150 ||

In the activities of eating, affection, aversion, flight and the rest — established by kind and by descent — who is the employer, for beasts and birds?

The interrogation widened from craft to the whole repertoire of animal life. Not merely nest-building but the basic modes: ahara, feeding; priti and apadvesa, affection and aversion (the animal's evaluative world: what it goes toward, what it flees); plavana-adi, springing and flight. And the specification that matters: jaty-anvaya-prasiddhasu — *established by kind and by descent*, running true in the lineage, each species with its settled repertoire, transmitted without teaching. Kah prayokta: *who is the employer* — who deploys these activities, sets them going at the right moment toward the right object? The question is put in the grammarian's own idiom: prayoktr is the term for the speaker, the one who employs words (1.130's inner word standing in the prayoktr's self; 1.147's speaker intending gauh). The choice is exact and discloses the design: as the speaker's inner word deploys itself into audible speech, so *something* deploys the beast's inheritance into timely action — and the answer is not a further agent but the pratibha itself, the ripened power (2.148) that is its own employer. Life's know-how and speech's comprehension are one structure at two grains — which is why 2.151 can say, with a straight face, that this too comes from agama.
contested: the transcription's apadvesa/plavana we read as aversion and springing-flight; the exact list varies in the manuscripts and nothing turns on its members

2.151 unformalized

भावनानुगतादेतदागमादेव जायते ।
आसत्तिविप्रकर्षाभ्यामागमस्तु विशिष्यते ॥ १५१ ॥

bhāvanānugatādetadāgamādeva jāyate |
āsattiviprakarṣābhyāmāgamastu viśiṣyate || 151 ||

This arises precisely from tradition, which is followed by realization; but tradition is distinguished according to nearness and remoteness.

The audacious answer to 2.149-150, and the verse where the kanda's naturalism and Kanda I's traditionalism become a single doctrine. Whence the spider's web, the cuckoo's spring song, the beast's fear? Bhavana-anugatad agamad eva jayate: from *agama* — received tradition — followed by bhavana, realization. The claim sounds absurd only if agama is confined to texts. Bhartrhari's move is to widen it: what is handed down and ripened *in a lineage* is agama, whether the vehicle is a recension of the Veda or the inheritance of a species (2.150's jaty-anvaya-prasiddha was already the tell). Then the calibration that makes the widening precise rather than glib: asatti-viprakarsabhyam agamas tu visisyate — tradition is *distinguished by nearness and remoteness*: the brahmin's Veda is proximate agama, recited yesterday, present to consciousness; the spider's is remote — sedimented past all recollection, ripened into the body's own certainty, and for that reason *more* unerring, not less (the web never fails; the recitation may). This is Kanda I's deepest thesis (1.30: even the seers' knowledge is agama-purvaka; nothing is known without inheritance) extended to its limit: *all* competence is inherited and ripened, and the difference between scripture, expertise and instinct is one of distance, not of kind. The six-fold classification follows immediately, and this verse is its ground.
contested: reading agama here as 'tradition/inheritance' in the wide sense is required by the context (spiders have no scripture) and is the tradition's own gloss; a narrower reading forces the verse into incoherence, which is itself an argument for the wide one

2.152 unformalized

स्वाभावचरणाभ्यासयोगादृष्टोपपादिताम् ।
विशिष्टोपहितां चेति प्रतिभां षड्विधां विदुः ॥ १५२ ॥

svābhāvacaraṇābhyāsayogādṛṣṭopapāditām |
viśiṣṭopahitāṃ ceti pratibhāṃ ṣaḍvidhāṃ viduḥ || 152 ||

From nature, from conduct, from the practice of yoga, from the unseen, from what is brought about, and from what is conveyed by the distinguished — thus they know intuition to be sixfold.

The classification, and with it the movement's close: pratibha in six kinds, ordered — not accidentally — from the animal to the sage. Svabhava, nature: the cuckoo's song, the spider's web (2.149-150) — flash as the ripened endowment of a kind. Carana, conduct: the flash bred by a way of life, by observance and habit of living. Abhyasa, practice: the jeweller's eye, the grammarian's ear (1.35, 2.117) — expertise. Yoga: the flash of disciplined contemplation, the yogin's seeing (1.37's unclouded mind). Adrsta, the unseen: what karma ripens unaccountably — the sudden knowing no biography explains. And visistopahita, conveyed by the distinguished: the flash *transmitted* by an eminent one, teacher to pupil, seer to hearer — the only member of the six that passes between persons, and it does so by *waking* rather than telling (2.144's ineffability strictly preserved: the master does not describe the flash, he occasions it). Read against 2.151, the six are the spectrum of agama by distance: nature is tradition at its most remote and most certain, transmission from the distinguished its nearest and most fragile, the middle terms marking the intervals. Read against 2.143, the whole is the kanda's answer to its opening question — what is a sentence's meaning? A flash of the sixth kind, waked in a hearer by a speaker; the same flash, at bottom, by which the bird knows its nest. The Vakyakanda's central argument is complete: sentence-meaning is not a thing assembled out of words but an event in a living mind, continuous with all knowing-how, occasioned by speech and ineffable in it.
contested: the six members are stable in the tradition though their glosses vary; visistopahita is the contested one — 'conveyed by (contact with) the distinguished' (our reading, making it transmitted pratibha) versus 'endowed by a distinguished cause', which loses the transmission-sense the sequence seems to want

2.153-154 · What the Word Does Not Convey unformalized

यथा संयोगिभिर्द्रव्यैर्लक्षितेऽर्थे प्रयुज्यते ।
गोशब्दो न त्वसौ तेषां विशेषाणां प्रकाशकः ॥ १५३ ॥
आकारवर्णावयवैः संसृष्टेषु गवादिषु ।
शब्दः प्रवर्तमानो'पि न तानङ्गीकरोत्यसौ ॥ १५४ ॥

yathā saṃyogibhirdravyairlakṣite'rthe prayujyate |
gośabdo na tvasau teṣāṃ viśeṣāṇāṃ prakāśakaḥ || 153 ||
ākāravarṇāvayavaiḥ saṃsṛṣṭeṣu gavādiṣu |
śabdaḥ pravartamāno'pi na tānaṅgīkarotyasau || 154 ||

(153) As the word 'cow' is employed with respect to an object marked out by the substances conjoined with it, and yet is not the illuminator of those particulars — (154) so, though the word operates upon cows and the like which are commingled with shape, colour and limbs, it does not take those in.

A new movement opens, and its method is subtractive: to fix what the word conveys by cataloguing what it does not. The object 'cow' is met samyogibhih dravyaih laksite — marked out by the substances conjoined with it: the beast comes with horns, dewlap, tether, field, all present in the encounter, all serviceable for pointing it out. And yet: na tv asau tesam visesanam prakasakah — the word is *not the illuminator of those particulars*. (154) makes the point structural rather than incidental: akara-varna-avayavaih samsrstesu — shape, colour and limbs being *commingled* in every cow one meets (there is no unconfigured cow to encounter) — sabdah pravartamano 'pi na tan angikaroti: the word, though it operates upon such commingled objects, *does not take them in*, does not accept them into what it says. The distinction being drawn is between the object's furniture and the word's cargo: everything the thing has is present to the hearer, and almost none of it is *meant* (2.69's avivaksa, 2.123's inseparable hosting, now made the organizing principle of a whole movement). The verses following (2.155-163) will test the principle at its edges: parts, quantity, measure, the conditions under which usage lapses — mapping precisely how little the word carries and how much the encounter supplies.

2.155 unformalized

संस्थानवर्णावैवर्विशिष्टे यः प्रयुज्यते ।
शब्दो न तस्यावयवे प्रवृत्तिरुपलभ्यते ॥ १५५ ॥

saṃsthānavarṇāvaivarviśiṣṭe yaḥ prayujyate |
śabdo na tasyāvayave pravṛttirupalabhyate || 155 ||

The word that is employed with respect to a thing distinguished by configuration, colour and the rest — its operation is not found in that thing's limb.

The subtraction's first theorem: word-meaning does not distribute to parts. 'Cow' is employed of a thing that is samsthana-varnadaih visistam — distinguished by configuration and colour: a whole individual, determinate in every respect. Yet na tasya avayave pravrttir upalabhyate: the word's operation is *not found in its limb* — the cow's leg is not a cow, and 'cow' has nothing to say about it. The observation looks trivial and is not: it establishes that denotation is grain-fixed and non-inheriting, the semantic counterpart of 2.68's maxim (a word is not the expressor of everything that exists). If word-meaning flowed to parts, 'cow' would half-mean 'hoof', and every term would leak into its object's mereology. It does not: the word settles at exactly one grain — the grain its employment sets (2.104-106) — and both the whole's furniture (2.153-154) and the parts beneath it lie outside. The theorem also guards the sphota-doctrine's flank: undivided meaning attaching to undivided object, with no partitive echo on either side.

2.156 unformalized

संख्याप्रमाणसंस्थाननिरपेक्षः प्रवर्तते ।
बिन्दौ च समुदाये च वाचकः सलिलादिषु ॥ १५६ ॥

saṃkhyāpramāṇasaṃsthānanirapekṣaḥ pravartate |
bindau ca samudāye ca vācakaḥ salilādiṣu || 156 ||

Indifferent to number, measure and configuration, the expressor operates — with respect to a drop and to an aggregate alike — in the case of water and the like.

The theorem tested on mass-terms, where it shows most cleanly. 'Salila', water: the word is samkhya-pramana-samsthana-nirapeksah — *indifferent* to number, to measure, to configuration; and it operates equally bindau ca samudaye ca — upon a *drop* and upon an *ocean*. One term, spanning every quantity and shape its stuff can take, with no adjustment and no strain. For the atomist semantics this is awkward (what fixed deposit is common to drop and sea?); for the subtraction-doctrine it is the expected result: quantity, extent, shape are the *encounter's* furniture, never the word's cargo — the drop and the ocean differ in everything except what 'water' says of them. The example also quietly indicts the vyakti-theorist (2.122's rival): if words denoted individuals, 'water' would need an individual, and neither drop nor sea nor the sum of them presents one. What the word settles on is the stuff as such — and 'as such' is precisely the residue left when number, measure and configuration have been subtracted.
contested: salilādisu ('water and the like') marks the class of mass-terms without theorizing it; whether Bhartrhari means to distinguish mass from count nouns systematically, or merely to exhibit indifference-to-quantity in a favourable case, is not stated

2.157 unformalized

संस्कारादिपरिच्छिन्ने तैलादौ यो व्यवस्थितः ।
आहैकदेशं तत्त्वेन तस्यावयववर्तिता ॥ १५७ ॥

saṃskārādiparicchinne tailādau yo vyavasthitaḥ |
āhaikadeśaṃ tattvena tasyāvayavavartitā || 157 ||

Of that which stands settled in oil and the like, delimited by processing and the rest, one says that a single portion is that very thing — its residing in the parts (is thus admitted).

The counter-case, honestly admitted, and it complicates 2.155's theorem. Take taila, oil, samskara-adi-paricchinne — delimited by processing and such (this pressing, this vessel, this batch: the stuff individuated by its history). Of such a mass one says ekadesam tattvena — a single portion *is that very thing*: any ladleful of the oil is oil, wholly and without remainder — tasya avayava-vartita, its residing in the parts is thereby granted. Mass-terms *do* distribute to their parts, exactly where count-terms (2.155's cow and its leg) do not. Bhartrhari records the asymmetry without smoothing it, and it repays notice: what a word reaches is not fixed by any general rule of mereology but by the *kind of standing* its object has under that employment — divisible stuffs distributing, articulated wholes not. This is the movement's method working honestly: the subtraction (what the word omits) yields different residues for different terms, and the residue is settled by the employment, never by the ontology alone. Later semantics would call this the mass/count distinction and make heavy weather of it; the verse simply lets the two behaviours stand side by side.
contested: the syntax is compressed and our construal (oil's portion being oil, hence part-residence granted for such terms) follows the contrast with 2.155-156; some read the verse as reporting an opponent's over-extension of part-residence, to be corrected — the asymmetry it records is real on either reading

2.158 unformalized

येनार्थेनाभिसंबद्धमभिधानं प्रयुज्यते ।
तदर्थापगमे तस्य प्रयोगो विनिवर्तते ॥ १५८ ॥

yenārthenābhisaṃbaddhamabhidhānaṃ prayujyate |
tadarthāpagame tasya prayogo vinivartate || 158 ||

The naming that is employed as bound up with a particular meaning — on the departure of that meaning, its employment ceases.

The subtraction given its operational test: what a word *really* carries is what its usage cannot survive losing. Yena arthena abhisambaddham abhidhanam prayujyate — the naming is employed as bound to a certain meaning; tad-artha-apagame tasya prayogo vinivartate: that meaning departing, the employment *ceases*. This is the criterion the whole movement needs, and it is a dependency-test of exactly the modern kind: vary the object and watch the word — where usage lapses, meaning was; where usage persists, meaning was not. Lose the cow's colour and 'cow' stands (so colour was never carried, 2.153-154); lose the cowness and 'cow' falls (so cowness was). The test converts the doctrine of avivaksa from a claim about intention into an observable of usage — and 2.159-160 apply it immediately, first to the features that come and go without disturbing the word, then to the one whose loss silences it.

2.159 unformalized

यांस्तु संभविनो धर्मानन्तर्नोय प्रयुज्यते ।
शब्दस्तेषां न सान्निध्यं नियमेन व्यपेक्षते ॥ १५९ ॥

yāṃstu saṃbhavino dharmānantarnoya prayujyate |
śabdasteṣāṃ na sānnidhyaṃ niyamena vyapekṣate || 159 ||

But those possible properties which the word does not take within itself when employed — it does not, as a rule, require their presence.

The test's negative half. Sambhavino dharman — the *possible* properties: those a thing of this kind might have, and this one does (the cow's horns, hair, hooves, the colour of the day) — the word, when employed, does *not take them within*: antarniya na prayujyate, they are not carried inside the naming. Hence: tesam sannidhyam niyamena na vyapeksate — the word does not, *as a rule*, require their presence. The consequence is exactly the usage-behaviour 2.158 made criterial: a hornless cow is a cow, a shorn one is a cow, the word does not falter — because it never asked. Niyamena, 'as a rule', is precise and not a hedge: these features may be relevant on occasion (a context may make horns matter), but no rule of the word requires them; their bearing is the employment's business, not the term's. What we are watching is the emergence, from usage-observation, of something like the necessary/accidental distinction — arrived at not by metaphysics but by asking which losses stop the mouth.

2.160 unformalized

यथा रोमशफादीनां व्यभिचारेऽपि दृश्यते ।
गोशब्दो न तथा जातेर्विप्रयोगे प्रवर्तते ॥ १६० ॥

yathā romaśaphādīnāṃ vyabhicāre'pi dṛśyate |
gośabdo na tathā jāterviprayoge pravartate || 160 ||

As the word 'cow' is seen (to operate) even where hair, hooves and the like deviate — not so does it operate when the universal is dissociated.

The test's positive half, and the movement's sharpest single stroke. Roma-sapha-adinam vyabhicare 'pi — hair, hooves and the rest *deviating*: absent, atypical, lost — drsyate go-sabdah: 'cow' is still seen in use, unembarrassed. Na tatha jater viprayoge pravartate — *not so* when the jati is dissociated: let cowness go and the word will not follow; no usage survives it. Two features, two fates, one test (2.158): the deviating ones were never carried; the one whose loss silences the word was its whole cargo. This is the jati-doctrine (2.122) not asserted but *derived* — reached by watching where speech breaks — and it is the movement's yield: after everything has been subtracted (the conjoined substances, the shape and colour, the limbs, the number and measure, the possible properties), the residue is the universal, and the residue is the meaning. Vyabhicara, deviation, is the test's technical name and would become the standard tool of Indian logic for isolating the invariable; here it is turned on semantics with complete clarity: meaning is what does not deviate.

2.161 unformalized

दुर्लभं कम्यचिल्लोके सर्वावयवदर्शनम् ।
कैश्चित्त्ववयवैर्दृष्टैरर्थः कृत्स्नोऽनुमीयते ॥ १६१ ॥

durlabhaṃ kamyacilloke sarvāvayavadarśanam |
kaiścittvavayavairdṛṣṭairarthaḥ kṛtsno'numīyate || 161 ||

Difficult it is for anyone in the world to see all the parts; but from certain parts seen, the whole object is inferred.

A pivot from what words carry to how *objects* are given — the movement's second half, on the gap between what is presented and what is taken. Durlabham kasyacil loke sarva-avayava-darsanam: seeing *all* the parts is beyond anyone — no one ever sees the whole cow (the far side, the interior, the moment before and after are never given); perception is always partial. Kaiscit tv avayavair drstaih arthah krtsno 'numiyate: yet from *some* parts seen, the whole object is inferred — and effortlessly, without felt inference, the cow simply *there* in her entirety. The observation is doing two jobs. It secures the earlier subtractions: if the object is never wholly given anyway, the word's indifference to its furniture (2.153-156) is no defect but a fit with the epistemic situation. And it prepares 2.163's principle: comprehension routinely delivers wholes on partial warrant — which is what hearing a word does. The verse belongs with 1.90's part-seizures and 2.90's man-lion: Bhartrhari's persistent interest in how the mind completes what is shown, and how much of the given is in fact supplied.

2.162 unformalized

तथा जात्युत्पलादीनां गन्धेन सहचारिणाम् ।
नित्यसंबन्धिनां दृष्टं गुणानामवधारणम् ॥ १६२ ॥

tathā jātyutpalādīnāṃ gandhena sahacāriṇām |
nityasaṃbandhināṃ dṛṣṭaṃ guṇānāmavadhāraṇam || 162 ||

So too, in the case of jasmine, lotus and the like, whose qualities are constant companions of their scent, the determination of the qualities is observed.

The completion-principle exemplified, with the tradition's favourite evidence: scent. The jasmine's fragrance reaches one in the dark; and with it, not fragrance alone but the flower — its whiteness, its form, its whole quality-complex — avadharanam, *determined*, settled, present to the mind. The mechanism: gunanam sahacarinam nitya-sambandhinam — the qualities are constant companions of the scent, invariably linked; so one companion given, the company follows. This is co-occurrence hardened into cognitive delivery: what has never been met apart is never taken apart. The example serves the movement precisely because the delivery is *not* the scent's doing — the scent conveys scent; the flower arrives by the mind's completion along invariable association (2.161's inference-without-inferring, made specific). Transferred to language, as 2.163 immediately transfers it: the word conveys its narrow cargo (the jati, 2.160), and everything else the hearer finds in his comprehension — the shape, the colour, the whole imagined cow — arrives as the scent's company does: by association, not by the word.

2.163 unformalized

तस्मात् संभविनोऽर्थस्य शब्दात् संप्रत्यये सति ।
अदृष्टविप्रयोगोऽर्थः सम्बन्धित्वेन गम्यते ॥ १६३ ॥

tasmāt saṃbhavino'rthasya śabdāt saṃpratyaye sati |
adṛṣṭaviprayogo'rthaḥ sambandhitvena gamyate || 163 ||

Therefore, when there is comprehension from the word of the possible meaning, the object whose dissociation has not been seen is understood by way of its being connected.

The movement's conclusion, gathering both halves. Tasmat: from the subtraction (words carry only the non-deviating, 2.158-160) and the completion (minds deliver wholes from parts by invariable association, 2.161-162) it follows: sambhavino arthasya sabdat sampratyaye sati — when the word has conveyed its possible meaning — adrsta-viprayogo 'rthah sambandhitvena gamyate: the object *whose dissociation has never been seen* is understood by way of *being connected*. Two channels, cleanly separated. The word gives what it gives: the invariant (2.160). Everything else in the hearer's comprehension — the cow's shape, her four legs, the world of default furnishing — comes by sambandhitva: it rides in because it has never been met apart, exactly as the flower rides in on its scent. Note the criterion's elegance: adrsta-viprayoga, 'not seen dissociated' — not *impossible* to dissociate, merely never yet met apart; the completion is defeasible, an association-effect, not an entailment (the hornless cow of 2.159 defeats it and the word survives). The movement thus ends with a two-channel semantics that a modern reader can recognize: strict semantic content, thin and invariant; and a generous pragmatic penumbra supplied by the hearer's world-knowledge. The kanda's next topic — number and the case-endings — is the natural test of exactly this boundary.
contested: sambhavinah arthasya we take as 'the possible meaning' — that which the word can convey (recalling 2.159's sambhavino dharmah); some construe 'the existent object'. The two-channel result is unaffected

2.164-165 unformalized

वाचिका द्योतिका वा स्युर्द्वित्वादीनां विभक्तयः ।
स्याद्वा संख्यावतोऽर्थस्य समुदायोऽभिधायकः ॥ १६४ ॥
विना संख्याभिधानाद्वा संख्याभेदसमन्वितान् ।
अर्थान् स्वरूपभेदेन कांश्चिदाहुर्गवादयः ॥ १६५ ॥

vācikā dyotikā vā syurdvitvādīnāṃ vibhaktayaḥ |
syādvā saṃkhyāvato'rthasya samudāyo'bhidhāyakaḥ || 164 ||
vinā saṃkhyābhidhānādvā saṃkhyābhedasamanvitān |
arthān svarūpabhedena kāṃścidāhurgavādayaḥ || 165 ||

(164) The case-endings might be either expressive or manifesting of duality and the rest; or the aggregate might be the expressor of a numbered object. (165) Or, without any expression of number, they say that certain words such as 'go' convey, by difference of their own form, objects already accompanied by difference of number.

Number — the perfect test of the two-channel doctrine (2.163), since number is precisely a determination that sometimes seems said and sometimes seems merely there. Three accounts are canvassed. (i) The vibhaktis, the case-endings, are vacikah — *expressive* of duality and the rest: 'gavau' says two, the dual ending denoting the number outright. (ii) They are dyotikah — *manifesting*: the ending does not denote number but lights up a number already in the meaning — the illuminator/expressor distinction (Kanda I's whole apparatus) applied to inflection, and destined for a long career in Indian semantics (the dyotaka-doctrine would become the standard account of particles). (iii) The samudaya, the aggregate — stem-plus-ending taken as one unit — is the expressor of the numbered object: no division of labour at all, the whole form denoting the whole content (2.36's asvakarna-principle at the level of inflection). Then (165) the fourth and strangest: without any number-expression whatever, some words (the tradition instances 'go') convey numbered objects sva-rupa-bhedena — by *difference of their own form*: the form-difference simply *is* the number-difference, nothing denoting anything. The four exhaust the logical space, and the exhaustiveness is the point: number's status is genuinely undecidable by inspection, which is what a two-channel semantics predicts for exactly such borderline determinations.
contested: 165's kamscid ahur gavadayah is grammatically awkward and our construal (certain words convey number by their own form-difference) follows the sense; the verse has also been read as reporting that certain *theorists* say this. Either way the fourth option is a form-difference account without denotation

2.166 unformalized

ये शब्दा नित्यसम्बन्धा विवेके ज्ञातशक्तयः ।
अन्वयव्यतिरेकाभ्यां तेषामर्थो विभज्यते ॥ १६६ ॥

ye śabdā nityasambandhā viveke jñātaśaktayaḥ |
anvayavyatirekābhyāṃ teṣāmartho vibhajyate || 166 ||

Those words which have eternal connection, whose capacities are known in isolation — their meaning is divided out by agreement and difference.

The method for adjudicating 2.164-165, and it is the method of 2.12: distributional analysis. Where words are nitya-sambandha — eternally connected with their meanings (Kanda I's thesis, 1.23) — and vaiveke jnata-saktayah — their capacities *known in isolation*, established independently — then anvaya-vyatirekabhyam tesam artho vibhajyate: their meaning is divided out by *agreement and difference*: covary the form and watch the meaning; where they track, credit a unit; where they do not, do not. The verse is doing careful methodological work. It states the *conditions* under which the distributional method may be trusted — independently-known capacities, stable connection — and thereby marks that the method is not universally applicable: it is an instrument with a domain, not a metaphysics. This matters immediately, because 2.167 will exhibit exactly the cases where the conditions fail, and the whole point of the pair is that the analyst's chief tool has an edge past which it does not cut.

2.167 unformalized

यत्र चाव्यभिचारेण तयोः शक्यं प्रकल्पनम् ।
नियमस्तत्र न त्वेवं नियमो नुट्शबादिषु ॥ १६७ ॥

yatra cāvyabhicāreṇa tayoḥ śakyaṃ prakalpanam |
niyamastatra na tvevaṃ niyamo nuṭśabādiṣu || 167 ||

And where the construing of those two can be made without deviation, there is a fixed rule; but there is no such fixed rule in the case of nut, sap and the like.

The method's limit, exhibited in the grammar's own machinery. Where anvaya and vyatireka can be applied avyabhicarena — without deviation, tracking perfectly — niyamah, a fixed rule obtains: the segmentation is secure, the unit real (as 'pac-' is, recurring with cooking wherever it recurs). But: na tv evam niyamo nut-sab-adisu — no such fixity for nut, sap and their kind. These are Panini's augments and technical elements — nut (the n-augment), sap (the class-marker of the first conjugation) — formatives that appear and vanish by rule with *no* constant meaning tracking them: sap marks a conjugation class and means nothing at all. The method finds them by distribution and finds no meaning to credit them with. So the analyst's own inventory contains units that are real *as units* and empty *as meanings* — 2.12's anarthaka parts (the '-bha' of vrsabha) now met not in the lexicon's oddities but at the heart of the derivational system. The moral for 2.164's dispute: one cannot settle whether the dual ending *denotes* two merely by noting that it covaries with twoness — covariance certifies a unit, never a meaning; sap covaries with everything in its class and means nothing. And the moral for the kanda: the tool that generates padarthas cannot, by itself, tell which of them are semantic — which is precisely why the meanings had to be sought at the whole (2.143).
contested: the transcription's nut-sab-adisu we read as the grammatical augments/markers nut and sap; some editions read differently, but the class intended — meaningless formatives isolable by distribution — is not in doubt

2.168 unformalized

सम्भवे वाभिधानस्य लक्षणं तु न कल्प्यते ।
आपेक्षिक्यो हि संसर्गे नियताः शब्दशक्तयः ॥ १६८ ॥

sambhave vābhidhānasya lakṣaṇaṃ tu na kalpyate |
āpekṣikyo hi saṃsarge niyatāḥ śabdaśaktayaḥ || 168 ||

When direct expression is possible, indication is not to be assumed; for the capacities of words, in combination, are fixed relatively.

Two maxims, the second considerably deeper than the first. (i) Sambhave va abhidhanasya laksanam tu na kalpyate: where abhidhana — direct expression — is available, do not posit laksana, indication (secondary or transferred meaning). This is the standard interpretive economy of the tradition (do not reach for metaphor while the literal serves), and it will govern the coming discussion of secondary meaning. (ii) The ground offered for it, and it is the movement's most compressed insight: apeksikyah hi samsarge niyatah sabda-saktayah — the capacities of words, *in combination*, are fixed *relatively*. Sakti — the word's power, the thing every semantics has been trying to locate — is here declared apeksiki: relational, index-bound, determinate only against the combination it stands in. Not: each word has a fixed power and combination merely displays it. Rather: powers are *fixed by* relation — the same word carrying different capacity in different company, and carrying determinate capacity only *because* of company (2.17's particular-words looking general in citation; 2.104's capacity-assumption in the stream). The economy-maxim then follows naturally: if capacity is settled relationally, then in any given combination the capacity simply *is* what the combination fixes — and there is no room to import laksana while the relationally-fixed abhidhana is doing the work.

2.169 unformalized

न कूपसूपयूपानामन्वयोऽर्थस्य दृश्यते ।
अतोऽर्थान्तरवाचित्वं संघतस्यैव गम्यते ॥ १६९ ॥

na kūpasūpayūpānāmanvayo'rthasya dṛśyate |
ato'rthāntaravācitvaṃ saṃghatasyaiva gamyate || 169 ||

No syntactic connection of meaning is seen in the case of 'kupa', 'supa', 'yupa'; therefore it is understood that the aggregate alone expresses a different meaning.

The relational-capacity doctrine of 2.168 exhibited in a case that has become famous. Kupa (well), supa (broth), yupa (sacrificial post) — three words sharing '-upa', differing in one phoneme, and their meanings share *nothing*: na anvayah arthasya drsyate, no syntactic or semantic connection whatever is seen. The rhyme is total and the sense is null. Hence: sanghatasya eva arthantara-vacitvam gamyate — it is the *aggregate alone*, the whole word, that expresses its wholly different meaning. The example does the work of a hundred arguments. Distributional method (2.166) would find '-upa' as a recurrent segment; the segment means nothing; the three words are not variations on a theme but three unrelated wholes that happen to sound alike. Phonetic overlap is thus proved semantically inert — which is 2.11's point (word-parts resembling other words) driven home with maximum economy, and which vindicates 2.36's generalization (comprehension is whole-to-whole everywhere, the transparent compound differing from the opaque only in inviting analysis). Kupa-supa-yupa is the kanda's emblem of the fact that meaning does not live in the pieces.

2.170 unformalized

अन्वाख्यानानि भिद्यन्ते शब्दव्युत्पत्तिकर्मसु ।
बहूनां सम्भवेऽर्थानां निमित्तं किञ्चिदिष्यते ॥ १७० ॥

anvākhyānāni bhidyante śabdavyutpattikarmasu |
bahūnāṃ sambhave'rthānāṃ nimittaṃ kiñcidiṣyate || 170 ||

Explanations differ in the operations of deriving words; among the many possible meanings, some one occasion is accepted.

Etymology entered — and immediately relativized. Anvakhyanani bhidyante sabda-vyutpatti-karmasu: *explanations differ* in the business of deriving words; the etymologist's craft (the nirukta-tradition's whole enterprise) does not converge — the same word receives incompatible derivations from different authorities, each defensible. And the reason: bahunam sambhave arthanam nimittam kimcid isyate — many meanings being *possible*, some one occasion is *accepted*: the derivation does not discover the word's meaning; it selects one nimitta, one occasioning ground, from among several the form permits. Etymology is thus a choice under-determined by evidence — the analyst picking which possible history to credit, as the exegete picks which analysis to run (2.38's non-fixity of means: 'a fixed rule about means does not necessarily stand'). This is not a complaint. Bhartrhari's tradition prizes nirukta (1.10's angas), and he is not dismissing it; he is placing it — as upaya, a discardable instrument whose value is what it teaches, not what it reveals about the word's real constitution (2.13's prakriya-bheda for the avidvan). The next verses exhibit the multiplicity concretely.
contested: nimittam kimcid isyate we read as the etymologist's selection among possible grounds; some construe 'some particular occasion is (what is) desired (by the deriver)' — the under-determination is the point on either reading

2.171 unformalized

वैरवासिष्ठगिरीशास्तथैकागारिकादयः ।
कैश्चित्कथाञ्चिदाख्याता निमित्तावधिसङ्करैः ॥ १७१ ॥

vairavāsiṣṭhagirīśāstathaikāgārikādayaḥ |
kaiścitkathāñcidākhyātā nimittāvadhisaṅkaraiḥ || 171 ||

Vaira, Vasistha, Girisa, and likewise Ekagarika and the rest, have been explained somehow, by some, through confusions of ground and limit.

The multiplicity exhibited, with a rueful precision. Four specimens whose derivations the tradition contests: vaira (enmity — from vira? from a lost base?); vasistha (the seer's name — from vasin, from vasistha as superlative?); girisa (Siva, mountain-dweller — giri-sa, but sa from which root, and why?); ekagarika (a term the tradition glosses variously — 'burglar', 'one who keeps a single house'). Kaiscit kathamcid akhyatah: explained *by some*, *somehow* — the two indefinites doing withering work. And the diagnosis: nimitta-avadhi-sankaraih — by *confusions of ground and limit*: the etymologist mixes up which element is the occasioning ground and where the analysis is to be cut, and by shuffling those two variables generates as many derivations as he pleases. Bhartrhari has isolated exactly the degrees of freedom that make etymology unfalsifiable: choose a base, choose a boundary, and any form yields any sense. The verse is his most economical demonstration that the analyst's tools generate their own results — and it lands with special force here, since the whole kanda's argument has been that the parts analysis finds are the analysis's own postings (2.10, 2.97, 2.167).

2.172 unformalized

यथा पथः समाख्यानं वृक्षवल्मीकपर्वतैः ।
अविरुद्धं गवादीनां भिन्नैश्च सहचारिभिः ॥ १७२ ॥

yathā pathaḥ samākhyānaṃ vṛkṣavalmīkaparvataiḥ |
aviruddhaṃ gavādīnāṃ bhinnaiśca sahacāribhiḥ || 172 ||

As the designation of a road by trees, ant-hills and mountains is not incompatible — so with 'cow' and the rest, by different accompanying things.

The movement closes with a simile that redeems what 2.170-171 seemed to demolish. A road is designated by its landmarks: 'the way past the great tree', 'the road by the ant-hill', 'the mountain road' — three designations, three grounds, one road, and aviruddham: *not incompatible*. No one asks which landmark is the road's true name; each picks the road out, from where the speaker stands, by what lies to hand. Tatha gavadinam bhinnaih sahacaribhih: so with 'cow' and the rest, by *different accompanying things* — a word may be occasioned by any of its object's companions (its gait, its lowing, its horns), and rival etymologies are rival landmarks, all serviceable, none constitutive. The simile completes the movement's whole argument with a single image. The word does not carry its object's furniture (2.153-160); the mind supplies the rest by association (2.161-163); the analyst's derivations pick among associations for their occasioning grounds (2.170-171) — and none of this impugns the word, because designation was never a matter of *containing* the thing. It is a matter of *pointing successfully from where one stands*. That the pointing succeeds while every account of how it succeeds diverges is not a scandal; it is the ordinary condition of language, and 2.142 has already told the wise man to leave the world's namings undisturbed.

2.173 unformalized

अन्यथा च समाख्यनमवस्थामेददर्शिभिः ।
क्रियते किंशुकादीनामेकदेशावधारणात् ॥ १७३ ॥

anyathā ca samākhyanamavasthāmedadarśibhiḥ |
kriyate kiṃśukādīnāmekadeśāvadhāraṇāt || 173 ||

And a designation is made otherwise by those who see difference of state, from the determination of a single portion — as in the case of the kimsuka and the like.

The landmark-simile (2.172) given its human dimension: designators differ because *seers* differ. Avastha-bheda-darsibhih — by those who see a difference of state: the kimsuka tree (the flame-of-the-forest, a stock example precisely because it is a different sight in each season — bare, budded, blazing red) receives its designation ekadesa-avadharanat, from the determination of a *single portion*: whoever names it names it from the aspect he has met. The traveller who saw it in flower and the one who saw it bare have named the same tree by different grounds, and neither has erred. This completes the etymological relativism (2.170-171) by locating its source not in the etymologist's carelessness but in the structure of encounter itself: nobody meets the whole (2.161), everybody names from a portion, and the portions differ. The kimsuka also carries an old resonance the audience would hear — it is the tree of the parable in which several men describe it from their several seasons and each is mocked by the others — which is exactly the moral: the mockery is misplaced.
contested: the kimsuka's traditional parable-associations are assumed rather than stated by the karika; the verse needs only that its appearance varies by state and that names are given from states

2.174 unformalized

कैश्चिन्निर्वचनं भिन्नं गिरतेर्गर्जतेर्गमेः ।
गवतेर्गतदेर्वापि गौरित्यत्रानुदर्शितम् ॥ १७४ ॥

kaiścinnirvacanaṃ bhinnaṃ giratergarjatergameḥ |
gavatergatadervāpi gaurityatrānudarśitam || 174 ||

By some, a different derivation has been shown here for 'gauh' — from 'girati', from 'garjati', from 'gam', from 'gavate', or indeed from 'gata'.

The relativism made concrete on the tradition's most-derived word. 'Gauh' — cow — has been referred to: girati, she swallows (the grazing beast); garjati, she bellows (the lowing beast); gam, to go (the wandering beast); gavate, a rare root the etymologists press into service; gata, gone/motion. Five derivations, five grounds, one animal — each a landmark on the road (2.172), each picking the cow out by a real companion of hers (her feeding, her voice, her going). Kaiscid, 'by some': the anonymity is now familiar and is the point — the derivations have authorities, and the authorities do not agree, and the disagreement has never once impeded anyone's use of the word. That is the datum. Etymology is a genuine discipline producing incompatible results and costing nothing, because the word's working never depended on it: derivation reconstructs a possible occasion for a name whose life is entirely in its employment (2.104-106). The verse is the kanda's cheerful demonstration that the search for a word's 'true meaning' in its origin is a search for a landmark, not a foundation.

2.175 unformalized

गौरित्येव स्वरूपाद्वा गौशब्दो गोषु वर्तते ।
व्युत्पाद्यते न वा सर्वं कैश्चिच्चोभयथेष्यते ॥ १७५ ॥

gaurityeva svarūpādvā gauśabdo goṣu vartate |
vyutpādyate na vā sarvaṃ kaiściccobhayatheṣyate || 175 ||

Or the word 'gauh' operates upon cows simply from its own form, as 'gauh'; either everything is derived, or not — and by some it is accepted in both ways.

The alternative to derivation, and then the tradition's characteristic settlement. First: gaur ity eva svarupad va — 'gauh' operates on cows simply *from its own form*, underived, the word standing on itself with no history to lean on (the avyutpanna, the underivable, that Panini's own system concedes for certain forms, 2.37). Then the general question posed in its starkest form: vyutpadyate na va sarvam — is *everything* derived, or nothing? The two schools are the tradition's real ones: the vyutpatti-vadins who hold every word analyzable to root and affix, and the a-vyutpatti-vadins who hold that stems are simply given. And the settlement: kaiscid ca ubhayatha isyate — by some it is accepted *both ways*. Not a compromise but a principled dual-standing: the same word is underived as *used* (its life needing no history) and derived as *taught* (its analysis serving the learner, 2.13). This is 1.74 and 2.38 arriving in etymology — the two directions of description, each valid for its purpose, neither the fact of the matter — and it is why the kanda can honour the nirukta-tradition (2.170's craft) while denying it any authority over meaning.

2.176 unformalized

सामान्येनोपदेशश्च शास्त्रे लघ्वर्थमाश्रितः ।
जात्यन्तरवदस्यापि विशेषाः प्रतिपादकाः ॥ १७६ ॥

sāmānyenopadeśaśca śāstre laghvarthamāśritaḥ |
jātyantaravadasyāpi viśeṣāḥ pratipādakāḥ || 176 ||

And instruction by way of the general is resorted to in the science for the sake of brevity; as with another universal, its particulars are (still) the conveyors.

A methodological note that will govern the movement on roots and preverbs now beginning. Samanyena upadesah — instruction by way of the *general* — is what the sastra resorts to laghv-artham, *for brevity*: the grammar teaches 'the root pac', 'the affix tin', stating rules over general categories because stating them over every instance would be endless (laghava, economy, is the Paninian science's governing virtue). But the brevity is a device, and the verse says so: jaty-antaravad asyapi visesah pratipadakah — as with any other universal, its *particulars* are the conveyors: what actually conveys, in any utterance, is the particular form in its particular employment; the general is the teacher's shorthand for a family of particulars, not an entity doing semantic work. The point matters for what follows: the grammar will be seen to treat dhatu and upasarga as separate general items, and the whole question (2.180-190) will be whether that separation is more than pedagogical economy. Bhartrhari has answered before asking: instruction is general for brevity's sake; conveyance is always particular.

2.177 unformalized

अर्थान्तरे च यद् वृत्तं तत्प्रकृत्यन्तरं विदुः ।
तुल्यरूपं न तद् रूढावन्यस्मिन्ननुषज्यते ॥ १७७ ॥

arthāntare ca yad vṛttaṃ tatprakṛtyantaraṃ viduḥ |
tulyarūpaṃ na tad rūḍhāvanyasminnanuṣajyate || 177 ||

And what has occurred in another meaning, they know as a different base; though its form is alike, it is not carried over to another in established currency.

A principle of lexical identity, drawn from the same soil. Artha-antare yad vrttam — what has occurred with a *different meaning* — tat prakrty-antaram viduh: they know it as a *different base*: sameness of meaning-behaviour, not sameness of shape, individuates a stem. And the consequence: tulya-rupam na tad rudhau anyasmin anusajyate — though the form be alike, it is not carried over into another word's established currency: the identical syllables occurring in a rudhi-word (a conventionalized term, 2.37) do not drag their other life along. This is kupa-supa-yupa (2.169) stated as a rule of the lexicon rather than exhibited as an oddity: homophony is not identity, and the etymologist who reasons from shared shape to shared base has confused a landmark with a road (2.172). The principle also anticipates the movement's central case: if 'pra-' in one verb means one thing and in another means nothing recoverable, then by 2.177 they are not the same item at all — which is precisely the reef on which preverb-semantics will run.
contested: vrttam we take as 'has occurred/operated' (the form used with another sense); the compressed syntax also permits 'what has been formed (as a compound) in another meaning', narrowing the rule to vrtti-formations — the individuation-principle survives either way

2.178 unformalized

भिन्नाविजियजी धातू नियतौ विषयान्तरे ।
कैश्चिद् कथञ्चिदुद्दिष्टौ चित्रं हि प्रतिपादनम् ॥ १७८ ॥

bhinnāvijiyajī dhātū niyatau viṣayāntare |
kaiścid kathañciduddiṣṭau citraṃ hi pratipādanam || 178 ||

The two roots 'ji' and 'yaj', distinct and fixed to different domains, have been somehow set forth by some as one; for conveying is a various thing.

A specimen of the etymologist's over-reach, chosen for its audacity. The roots ji (to conquer) and yaj (to sacrifice) are bhinnau — *distinct*, niyatau visayantare — each fixed to its own domain, sharing nothing but a formal resemblance that phonological ingenuity can manufacture. And yet kaiscid kathamcid uddistau: by *some*, *somehow*, they have been set forth as one — the derivers, having the two freedoms 2.171 identified (choose a base, choose a boundary), unified what usage keeps forever apart. Bhartrhari's comment is a single wry clause: citram hi pratipadanam — 'for conveying is a various thing', or more bluntly, exposition is a wonderful business. The dryness is the argument. He does not refute the derivation; he lets its existence testify. If the method can identify conquering with sacrificing, the method's outputs are not constraints on meaning — and 2.177 has just given the rule the derivers violated: different meaning-behaviour, different base, whatever the shapes suggest.
contested: the identification of the two roots as ji and yaj follows the transcription's aviji-yaji; the specific etymological claim being mocked is not recoverable, and the verse's force is entirely in the mockery

2.179 unformalized

एवं च बालवायादिजित्वरीवदुपाचरेत् ।
भेदाभेदाभ्युपगमे न विरोधोऽस्ति कश्चन ॥ १७९ ॥

evaṃ ca bālavāyādijitvarīvadupācaret |
bhedābhedābhyupagame na virodho'sti kaścana || 179 ||

And thus one should proceed as with 'bala', 'vaya' and 'jitvari' and the like; in accepting both difference and non-difference, there is no contradiction whatever.

The settlement of 2.175 formalized as a maxim, with a handful of examples the tradition disputes (bala, vaya, jitvari — words whose analyses divide the grammarians). Evam upacaret: proceed thus — accepting bheda-abheda-abhyupagame na virodho 'sti kascana: in accepting *both* difference and non-difference there is *no contradiction whatever*. Two words may be one base and two bases, according to the purpose of the description: one, for the economy of a rule (2.176's samanyena upadesa); two, for the fidelity of a usage-account (2.177's individuation by meaning-behaviour). The maxim is the kanda's constant (1.74, 2.38, 2.58, 2.175) at its most explicit — bhedabheda not as a metaphysical doctrine to be defended but as a *methodological permission* to describe the same material twice, incompatibly, without incoherence, because descriptions answer to purposes and the material answers to neither. It is the exact license the coming analysis of root-and-preverb will need: they are two (the grammar says so, and rules require it) and one (usage says so, and meaning requires it), and the verse has pre-authorized both.

2.180 · Root and Preverb unformalized

अडादीनां व्यवस्थार्थं पृथक्त्वेन प्रकल्पनम् ।
धातूपसर्गयोः शास्त्रे धातुरेव तु तादृशः ॥ १८० ॥

aḍādīnāṃ vyavasthārthaṃ pṛthaktvena prakalpanam |
dhātūpasargayoḥ śāstre dhātureva tu tādṛśaḥ || 180 ||

The construing of root and preverb as separate, in the science, is for the sake of settling 'at' and the rest; but the root itself is of that very kind.

The movement's thesis, stated at once. In the sastra, dhatu and upasarga — root and preverb ('pra-', 'sam-', 'ud-') — are prthaktvena prakalpanam, construed as *separate* items. Why? Ad-adinam vyavastha-artham: for the sake of settling the augment 'at' and the rest — that is, for strictly formal purposes: the augment a- of the past tense must be positioned (does one say pra-agacchat or pra-a-gacchat?), sandhi must be regulated, accent assigned, and none of that can be stated without treating the preverb as a separable unit. The separation is *rule-driven*, exactly as 2.176 said instruction is general for brevity's sake. And then the counter-claim that governs everything following: dhatur eva tu tadrsah — *the root itself is of that very kind*: what actually conveys is the root already-as-so-qualified, the preverbed root as one meaning-unit; the separation that the grammar's formal operations require does not correspond to any division in the meaning. Two items for the rule-writer; one for the speaker. It is 2.10's apoddhara-doctrine arriving at the innermost joint of the word — and the reason it must arrive here is that if analysis *anywhere* found a real semantic seam, this is where it would be.

2.181 unformalized

तथाहि संग्रामयतेः सोपसर्गाद्विधिः स्मृतः ।
क्रियाविशेषाः संघातैः प्रक्रम्यन्ते तथाविधाः ॥ १८१ ॥

tathāhi saṃgrāmayateḥ sopasargādvidhiḥ smṛtaḥ |
kriyāviśeṣāḥ saṃghātaiḥ prakramyante tathāvidhāḥ || 181 ||

For thus, a rule is taught for 'samgramayati' as having its preverb; and such particular actions are undertaken by whole complexes.

The evidence for 2.180, from the grammar's own practice. Samgramayati — 'he fights, wages battle': the form is derived from samgrama, itself sam + grama, and the grammar teaches its rules sopasargat — *with its preverb*: the sastra cannot state the derivation by taking the root alone and adding the preverb's contribution; it must treat the preverbed complex as the base from the start. The grammarian's own procedure thus concedes what 2.180 claimed: at the point where a rule must actually apply, the unit is the complex. And the generalization: kriya-visesah sanghataih prakramyante tathavidhah — such particular actions are undertaken *by whole complexes*: the specific action a verb names (waging-battle, not going-together-with-a-village) is delivered by the sanghata, the complex as a whole, not composed from separable donations. Note that 'wage battle' is precisely *not* recoverable from sam + grama by any calculus — the meaning is the complex's own (2.169's kupa-supa-yupa at the verbal root). The grammar keeps root and preverb apart for the augment's sake (2.180) and keeps them together for the rule's sake here: the two treatments are 2.179's licensed double description, and only the second tracks meaning.

2.182 unformalized

कार्याणामन्तरङ्गत्वमेवं धातूपसर्गयोः ।
साधनैर्याति संबन्धं तथाभूतैव सा क्रिया ॥ १८२ ॥

kāryāṇāmantaraṅgatvamevaṃ dhātūpasargayoḥ |
sādhanairyāti saṃbandhaṃ tathābhūtaiva sā kriyā || 182 ||

Thus the operations relating to root and preverb are internal; the action, being of that very kind, comes into connection with the means.

The doctrine given its technical seat. Karyanam antarangatvam evam dhatupasargayoh — the operations concerning root and preverb are *antaranga*, 'internal': the Paninian term for rules that apply at the innermost layer, before the outer ones get a chance (antaranga rules take precedence over bahiranga, and the metarule governing this is among the system's most consequential). Bhartrhari's use is exact and shrewd: the preverb's union with its root happens *inside*, in the derivation's core, complete before the word meets anything else. Hence: sadhanaih yati sambandham tathabhuta eva sa kriya — the action, being *already of that kind* (preverbed, particular, its sense settled), then comes into connection with the means, the karakas — the agent, instrument, object, everything the sentence supplies. The order is the argument. By the time the verb reaches the sentence, there is no root-plus-preverb to be summed; there is one action-meaning, which the sentence then furnishes with participants (2.48's architecture presupposes exactly this: the verb arrives whole and the nouns orbit *it*). The grammar's own layering of rules thus certifies the semantics: what is settled antaranga is settled before semantics at sentence-scale begins.
contested: antarangatva is the technical priority-notion and we render it 'internal'; whether Bhartrhari invokes the full paribhasa-machinery or the plainer sense ('these operations belong to the inner constitution') is undecidable, and the argument works on the plainer sense alone

2.183 unformalized

प्रयोगार्हेषु सिद्धः सन् भेत्तव्योऽर्थो विशिष्यते ।
प्राक् च साधनसम्बन्धात् क्रिया नैवोपजायते ॥ १८३ ॥

prayogārheṣu siddhaḥ san bhettavyo'rtho viśiṣyate |
prāk ca sādhanasambandhāt kriyā naivopajāyate || 183 ||

Being accomplished among things fit for employment, the object to be split is (further) specified; and prior to connection with the means, the action does not arise at all.

A pair of claims about how verbal meaning comes to be determinate. First: prayogarhesu siddhah san bhettavyo 'rtho visisyate — the object-to-be-split (bhettavya: the standard example of a verb's patient, 'what is to be split') is *already accomplished* among things fit for employment, and is then further specified: the log exists before the splitting is spoken of; the verb's participants are drawn from a world of usable things and then narrowed by the utterance. Second, and it is the striking one: prak ca sadhana-sambandhat kriya naiva upajayate — *prior to connection with the means, the action does not arise at all*. There is no bare action waiting to be supplied with participants. 'Splitting' as such, unagented, unpatiented, uninstrumented, is not a meaning that exists and then gets furnished; the action-meaning *comes into being* with its karaka-connections. Read against 2.182, the picture is precise: the preverbed root arrives with its *kind* fixed internally (waging-battle, not going-with-village), and its *being an action at all* is constituted in the sentence, by the means-connection. Kind from within, existence from without — which is why the verb is the sentence's centre (2.48) and yet not the sentence's meaning (2.1's first view, refused).
contested: bhettavyo 'rthah we take as the illustrative patient ('the to-be-split thing'); some read it abstractly ('the meaning that is to be analysed is specified'), making the verse about semantic decomposition — the second claim (no action prior to means-connection) is textually secure and is the load-bearing one

2.184 unformalized

धातोः साधनयोगस्य भाविनः प्रक्रमाद्यथा ।
धातुत्वं कर्मभावश्च तथान्यदपि दृश्यताम् ॥ १८४ ॥

dhātoḥ sādhanayogasya bhāvinaḥ prakramādyathā |
dhātutvaṃ karmabhāvaśca tathānyadapi dṛśyatām || 184 ||

As root-hood, and the condition of being an object, belong to a root by the commencement of its coming connection with the means — so let the other case be seen as well.

The retroactive constitution made explicit, and then extended. Dhatoh sadhana-yogasya bhavinah prakramat — from the *commencement* of the root's *coming* (bhavin: future, prospective) connection with the means — dhatutvam karma-bhavas ca: root-hood itself, and the condition of being an object, accrue. This is remarkable: a root is a root *in virtue of* the means-connection it is heading toward; the item's very categorial status is conferred by a relation not yet realized. (The grammar behaves exactly so: an element is called a dhatu because of what it will govern; a stem is called a karman because of the action it will stand under.) Tatha anyad api drsyatam: *so let the other case be seen too* — the invitation is to read the root/preverb relation by the same light: as root-hood is conferred prospectively by connection, so the preverbed root's unity is conferred by the meaning it is going to deliver, not assembled from what its pieces antecedently were. Categories in this grammar are functional and forward-looking — an item is what its role will make it (2.104's capacity-assumption, now temporalized) — which is why no analysis of the pieces will ever yield the whole they were destined to constitute.

2.185-186 unformalized

बीजकालेषु सम्बद्धा यथा लाक्षारसादयः ।
वर्णादिपरिणामेन फलानामुपकुर्वते ॥ १८५ ॥
बुद्धिस्थादभिसम्बन्धात्तथा धातूपसर्गयोः ।
अभ्यन्तरीकृतो भेदः पदकाले प्रकाशते ॥ १८६ ॥

bījakāleṣu sambaddhā yathā lākṣārasādayaḥ |
varṇādipariṇāmena phalānāmupakurvate || 185 ||
buddhisthādabhisambandhāttathā dhātūpasargayoḥ |
abhyantarīkṛto bhedaḥ padakāle prakāśate || 186 ||

(185) As lac-juice and the like, applied at the time of the seed, benefit the fruits by transformation of colour and the rest — (186) so, from the connection standing in the intellect, the difference of root and preverb, having been made internal, shines forth at the time of the word.

The movement's great simile, and one of the most delicate images in the work. The horticultural practice: lac-juice and other dyes applied *at the time of the seed* — bija-kalesu sambaddhah — work their effect not then but far downstream, varna-adi-parinamena phalanam upakurvate: benefiting the *fruits* by a transformation of colour and the rest. Nothing is visible at sowing; the whole efficacy is latent, carried invisibly through the plant's growth, and emerges in the fruit as its colour. So (186): buddhistat abhisambandhat — from the connection *standing in the intellect* (the union of root and preverb accomplished inwardly, in the speaker's mind, antaranga in the fullest sense) — abhyantarikrto bhedah — the difference *having been made internal*, taken inside, absorbed — pada-kale prakasate: shines forth *at the time of the word*. The preverb's contribution is real and is never separately present: it was mixed in at the seed, and what appears at the word is one colour. The simile answers precisely the question that would otherwise sink the doctrine — if root and preverb are semantically one, why does the preverb make a difference? — and answers it without conceding a seam: the difference is *in* the unified meaning as the dye is in the fruit's redness, constitutive and unlocatable. It also re-sounds the kanda's deepest structure: latency ripening into manifestation (1.84's seeds, 2.148's fermentation), here at the scale of a single word's inner history.
contested: laksarasa-adayah we take as lac-juice and similar dyes used in horticulture (the traditional gloss); the exact practice is obscure to us, but the simile's logic — applied at the seed, manifest in the fruit — is fully carried by the text

2.187 unformalized

क्वचित् संभविनो मेदाः केवलैरनिदर्शिताः ।
उपसर्गेण सम्बन्धे व्यज्यन्ते प्रपरादिना ॥ १८७ ॥

kvacit saṃbhavino medāḥ kevalairanidarśitāḥ |
upasargeṇa sambandhe vyajyante praparādinā || 187 ||

Sometimes possible differences, not exhibited by the roots alone, are manifested in connection with a preverb — 'pra', 'para' and the rest.

The phenomenon the doctrine must account for, stated fairly. Sambhavino bhedah kevalair anidarsitah — *possible* differences which the bare roots do not exhibit: har- alone gives nothing determinate; pra-har- is to strike, a-har- to fetch, sam-har- to gather, vi-har- to sport. The preverbs pra, para and the rest *do* something, undeniably and systematically. And the verb chosen for what they do is the movement's whole answer in one word: vyajyante — they are *manifested*. Not produced, not added, not composed: manifested — the technical term of Kanda I's entire apparatus (the sounds manifesting the sphota, 1.44-49; the flower's scent released by crushing, 1.79). The differences were *sambhavin*, possible, latent in the root's own semantic range; the preverb does not import them from outside but occasions them — the dye at the seed (2.185-186), the fan raising the ever-present air (1.116). This is why the doctrine of 2.180 does not have to deny the obvious: the preverb's contribution is total and the seam is still fictional, because manifesting is not contributing. The next verse canvasses the alternatives with characteristic fairness.

2.188 unformalized

स वाचको विशेषाणां सम्भवाद् द्योतकोऽपि वा ।
शक्त्याधानाय धातोर्वा सहकारी प्रयुज्यते ॥ १८८ ॥

sa vācako viśeṣāṇāṃ sambhavād dyotako'pi vā |
śaktyādhānāya dhātorvā sahakārī prayujyate || 188 ||

That (preverb) is the expressor of the particulars; or, since it is possible, it is the manifestor; or it is employed as a co-operating factor for the sake of infusing capacity into the root.

The three accounts of what a preverb does, catalogued without immediate adjudication — and the catalogue is the tradition's live one. (i) Vacaka: the preverb *expresses* the particular — 'pra' denotes forth-ness and the verb's sense is compounded of root-sense plus preverb-sense. This is the compositional account, and it is the one 2.180-186 have been dismantling. (ii) Dyotaka: it is the *manifestor* — sambhavat, 'since it is possible', i.e. because the particulars are already possible in the root, the preverb need only light them up (2.187's vyajyante given its technical name; the dyotaka-doctrine, first canvassed for case-endings at 2.164, here at its most consequential home — this is the account the grammatical tradition would ultimately make its own, and this verse is among its earliest clear statements). (iii) Sahakarin: a *co-operating factor*, employed sakty-adhanaya, for the *infusing of capacity* into the root — a third way between the first two: the preverb neither denotes nor merely reveals but *conditions*, adding potency to the root's own power (as heat conditions the seed without being the plant). Bhartrhari lets all three stand; but the middle one is his (2.187's verb was vyajyante), and 2.189-190 will supply the evidence that only manifestation-or-conditioning can absorb.
contested: which of the three Bhartrhari holds is a classical question; our reading takes (ii) as favoured on the strength of 2.187's vyajyante and 2.185-186's simile, with (iii) as a friendly variant — the tradition after him divides, and the dyotaka-account's eventual dominance is not itself evidence about the karika

2.189 unformalized

स्थादिभिः केवलैर्यच्च गमनादि तु गम्यते ।
तत्रानुमानाद् द्विविधात्तद्धर्मा प्रादिरुच्यते ॥ १८९ ॥

sthādibhiḥ kevalairyacca gamanādi tu gamyate |
tatrānumānād dvividhāttaddharmā prādirucyate || 189 ||

And where going and the like is understood from 'stha' and the rest taken alone, there, by inference of two kinds, the preverb 'pra' and the rest is said to have that same property.

Evidence for the manifestation-account, from the roots that need no preverb. Sthadibhih kevalaih — by roots such as stha *alone*, without any preverb — gamanadi tu gamyate: going and the like is (nonetheless) understood: bare roots already deliver directional and modal particulars that one might have credited to preverbs. So the particulars are in the root's range natively — which is (i)'s embarrassment: if 'pra' *denotes* forth-ness, why does the bare root sometimes deliver it? Then the inference: tatra anumanad dvividhat tad-dharma pradir ucyate — from a two-fold inference, the preverb pra and the rest is said to have *that same property*: reasoning from the bare root's capacity (it can deliver the particular unaided) together with the preverbed form's behaviour (it delivers it reliably), one concludes that the preverb shares the root's kind of work rather than contributing a separate content — it is a manifestor of what was there, tad-dharma, of the same character. The argument is genuinely evidential rather than doctrinal: it turns on a distributional fact (bare roots covering preverb-territory) that only the dyotaka or sahakarin accounts predict. And the next verse supplies the mirror-image evidence.
contested: anumanad dvividhat ('by a two-fold inference') is unexplained by the karika; our reconstruction (from the bare root's capacity, and from the preverbed form's behaviour) is one plausible filling among several the commentators offer

2.190 unformalized

अप्रयोगेऽधिपर्योश्च यावद् दृष्टं क्रियान्तरम् ।
तस्याभिधायको धातुः सह ताभ्यामनर्थकः ॥ १९० ॥

aprayoge'dhiparyośca yāvad dṛṣṭaṃ kriyāntaram |
tasyābhidhāyako dhātuḥ saha tābhyāmanarthakaḥ || 190 ||

And in the non-employment of 'adhi' and 'pari', so far as a different action is seen — the root that expresses that action is, apart from those two, meaningless.

The mirror-evidence, and the movement's closing stroke — the preverbs that cannot be removed. Some roots simply do not occur bare: adhi-i (to study), pari-dha (to don) and their kind. In the *non-employment* of adhi and pari — aprayoge — yavad drstam kriyantaram, so far as a different action is seen (i.e., the distinct actions these complexes name), tasya abhidhayako dhatuh saha tabhyam anarthakah — the root that expresses that action, *apart from those two, is meaningless*: strip 'adhi' from 'adhi-i' and no residue survives that means studying, or anything at all in that domain. Here the compositional account (2.188's (i)) fails completely and visibly: one cannot sum root-sense and preverb-sense when the root has no sense to contribute. And here 2.180's thesis is vindicated at its strongest: dhatur eva tu tadrsah — the root *itself is of that very kind*, i.e. the real unit is and always was the preverbed complex, the grammar's separation being for the augment's sake (2.180) and nothing more. The obligatory preverb is thus to the movement what kupa-supa-yupa (2.169) was to the earlier one: the case where the analysis's own units are exposed as postings. Root and preverb are two for the rule-writer, one for the speaker — and in adhi-i they are not even two for the rule-writer, since neither piece can be given a meaning to be summed.
contested: saha tabhyam is awkward ('together with those two... meaningless') and we construe it as 'apart from those two' following the argument's evident need; some read 'the root, together with those two, expresses that action — (alone it is) meaningless', which yields the same doctrine by a smoother route

2.191 · Particles, Suffixes, and the Meaningless unformalized

तथैव स्वार्थिकाः केचित् सङ्घातान्तरवृत्तयः ।
अनर्थकेन संसृष्टाः प्रकृत्यर्थानुवादिनः ॥ १९१ ॥

tathaiva svārthikāḥ kecit saṅghātāntaravṛttayaḥ |
anarthakena saṃsṛṣṭāḥ prakṛtyarthānuvādinaḥ || 191 ||

Just so, certain self-meaning suffixes, occurring within another complex, are commingled with what is meaningless and merely restate the base's meaning.

The preverb-doctrine (2.180-190) extended to the svarthika suffixes — those affixes that, by the grammar's own account, add *nothing*: the -ka of many a form, the affixes taught as sve arthe, 'in their own meaning', which is a courteous way of saying they mean what the stem already meant. Prakrty-artha-anuvadinah: they *restate* the base's meaning (anuvada, 2.115's term for idle repetition). And they are anarthakena samsrstah — *commingled with the meaningless*: the complex contains formatives that neither denote nor manifest. This is the movement's most awkward datum honestly placed: language is full of morphology doing no semantic work at all (2.167's nut and sap already showed it in the derivational core). The verse does not explain it away; it registers that the analyst's inventory contains items real as units and null as meanings, and that the word nonetheless means. Which is precisely the kanda's thesis from another side: if meaning were assembled from the pieces, meaningless pieces would be subtractions — and they are not; they are simply present, and the whole means what it means.

2.192 unformalized

निपाता द्योतकाः केचित् पृथगर्थाभिधायिनः ।
आगमा इव केऽपि स्युः संभूयार्थस्य वाचकाः ॥ १९२ ॥

nipātā dyotakāḥ kecit pṛthagarthābhidhāyinaḥ |
āgamā iva ke'pi syuḥ saṃbhūyārthasya vācakāḥ || 192 ||

Some particles are manifestors; some express a separate meaning; and some might be like augments — expressors of a meaning (only) in combination.

The nipatas — particles — canvassed in three accounts, the same triad the preverbs received (2.188) and for the same reason: particles are the hard case for any denotational semantics. (i) Dyotakah, manifestors: 'eva', 'ca', 'hi' do not denote anything but light up a determination already in the sentence-meaning (the account that would become the tradition's standard for particles, and the one Bhartrhari's system favours). (ii) Prthag-artha-abhidhayinah: expressors of a *separate* meaning — the particle denotes a content of its own (conjunction, emphasis, limitation) which then combines. (iii) The third and most interesting: agama iva — *like augments*: sambhuya arthasya vacakah, expressors of meaning *only in combination* — as an augment (the nut of 2.167) has no independent existence or sense and yet the form containing it means, so the particle contributes only as absorbed, never as a term. The three placements track exactly the possible relations of a formative to meaning: revealing, denoting, or constituting-only-in-combination — and it is characteristic that Bhartrhari lists them without decision. The next verses supply the evidence that only the first and third can absorb.

2.193 unformalized

उपरिष्टात् पुरस्ताद्वा द्योतकत्वं न भिद्यते ।
तेषु प्रयुज्यमानेषु भिन्नार्थेष्वपि सर्वथा ॥ १९३ ॥

upariṣṭāt purastādvā dyotakatvaṃ na bhidyate |
teṣu prayujyamāneṣu bhinnārtheṣvapi sarvathā || 193 ||

Whether above or before, the manifestor-hood is not differentiated, when they are employed — even in the case of different meanings, in every way.

Evidence for the dyotaka-account, from positional freedom. Uparistat purastad va — whether the particle stands *after* or *before* (the preverb pra- prefixed, the particle ca postposed, the Vedic freedom of placement) — dyotakatvam na bhidyate: its manifestor-hood is *not differentiated*: the work it does is the same work regardless of where in the utterance it falls. And sarvatha, in every way, bhinnarthesv api — even where the meanings differ. The argument turns on a fact about denotation: a denoting element contributes a content that must be *composed* with the rest, and composition is order-sensitive (this is why word order and case are doing labour). A manifesting element illuminates a whole that is already there, and illumination is order-indifferent — the lamp lights the room from any corner. So the particle's positional freedom is evidence of its character: it does not enter the semantic sum (there would be a place for it in the sum), it lights up the sentence-meaning from wherever it stands. The observation is delicate and quite exact about how particles behave; it also explains why the tradition eventually settled on dyotakatva for them.

2.194 unformalized

चादयो न प्रयुज्यन्ते पदत्वे सति केवलाः ।
प्रत्ययो वाचकत्वेऽपि केवलो न प्रयुज्यते ॥ १९४ ॥

cādayo na prayujyante padatve sati kevalāḥ |
pratyayo vācakatve'pi kevalo na prayujyate || 194 ||

'Ca' and the rest are not employed alone, though they have word-hood; and a suffix, though expressive, is not employed alone.

Two dependencies, exhibited as a pair. Cadayah — 'ca' and its kind — na prayujyante kevalah, are never employed *alone*, padatve sati: *although* they possess word-hood (the grammar counts them as padas, full words, by its own criteria). And the mirror: pratyayah vacakatve 'pi kevalo na prayujyate — a suffix, *though expressive* (the grammar assigns it a meaning: the -ti of pacati denotes agency-and-present-time, no one doubts it), is likewise never employed alone. Two categories, each with a property that on the atomist's terms should confer independence — wordhood, expressiveness — and neither can stand by itself. The pairing is the argument: if being a word meant being a semantic atom, 'ca' would be utterable; if being expressive meant carrying a deposit, '-ti' would be. Neither is. What both cases show is that the grammar's categories (word, expressive element) are *rule-postings* (2.3, 2.176), tracking the science's operational needs, and that they underdetermine semantic standing entirely. Dependence is the norm; independence, the citation-artifact (2.17).

2.195 unformalized

समुच्चिताभिधानेऽपि व्यतिरेको न विद्यते ।
असत्त्वभूतो भावश्च क्रियाऽन्येनाभिधीयते ॥ १९५ ॥

samuccitābhidhāne'pi vyatireko na vidyate |
asattvabhūto bhāvaśca kriyā'nyenābhidhīyate || 195 ||

Even where there is expression of cumulation, no exclusion is found; and the state of being, not being a thing, is an action — expressed by something else.

Two observations about what particles and abstract terms fail to deliver. First: samuccita-abhidhane 'pi vyatireko na vidyate — even where cumulation is expressed ('Devadatta *and* Yajnadatta come'), *no exclusion* is found: the conjunction cumulates without thereby excluding anything; 'and' does not contain a 'not-others'. The point registers that a particle's contribution is thin and one-directional — it lights one determination and carries no negative freight; the exclusions a hearer may draw come from elsewhere (the penumbra of 2.163). Second: asattva-bhuto bhavas ca kriya anyena abhidhiyate — bhava, the state-of-being, *not being a thing* (asattva: not a substance, the standing classification of what verbs mean as against what nouns mean), is an action, and is expressed by something *else* — by the verbal apparatus, not by the nominal. The two remarks share a moral relevant to the movement: what the categories of the grammar deliver — cumulation, being, action — are not object-like contents that could be deposited and summed; they are aspects that some element or other lights up within an already-whole meaning.

2.196 unformalized

समुच्चिताभिधानेऽपि विश्ष्टार्थाभिधायिनाम् ।
गुणैः पदानां सम्बन्धः परतन्त्रास्तु चादयः ॥ १९६ ॥

samuccitābhidhāne'pi viśṣṭārthābhidhāyinām |
guṇaiḥ padānāṃ sambandhaḥ paratantrāstu cādayaḥ || 196 ||

Even where cumulation is expressed, the connection is of words expressing qualified meanings, by way of their subordinate elements; but 'ca' and the rest are dependent.

The syntax of conjunction analysed, and the particle put in its place. When 'ca' cumulates, what is actually connected? Visista-artha-abhidhayinam padanam sambandhah gunaih — the connection is of *words expressing qualified meanings*, by way of their gunas, their subordinate elements: the conjunction joins already-determinate contents through the structure they already have (2.47's pre-fixed roles, 2.48's karaka-architecture); it does not manufacture the joining. And the particle itself: paratantras tu cadayah — 'ca' and the rest are *dependent*, paratantra: not standing on their own, contributing only within a structure that supplies them their office (2.62's finger-and-hand; 1.62's paratantra guna). This completes 2.194's observation with its explanation: 'ca' cannot be uttered alone not because of some accident of usage but because dependence is its nature — it is an operator on a whole, and where there is no whole there is nothing for it to be. The verse also quietly closes the door on the second account of 2.192 (particles as separate-meaning-expressors): a separate meaning could be uttered; a dependent manifestor could not.

2.197 unformalized

जनयित्वा क्रिया काचित् सम्बन्धं विनिवर्तते ।
श्रूयमाणे क्रियाशब्दे सम्बन्धो जायते क्वचित् ॥ १९७ ॥

janayitvā kriyā kācit sambandhaṃ vinivartate |
śrūyamāṇe kriyāśabde sambandho jāyate kvacit || 197 ||

Some action, having generated the connection, withdraws; sometimes the connection arises while the action-word is being heard.

A new topic — the genitive and the syntax of relation — opened with a temporal observation about how relations come to be. Two patterns. Janayitva kriya kacit sambandham vinivartate: some action, *having generated* the connection, then *withdraws* — the action that produced a relation is no longer present when the relation is spoken of: 'the king's man' (rajapurusa) records a relation that some act of appointing established and which now stands without it; the verb has done its work and gone. Srtiyamane kriya-sabde sambandho jayate kvacit: and sometimes the connection *arises while the action-word is being heard* — the relation is present in the very utterance of the verb, live in the sentence ('he cooks the rice'). The distinction between relations that outlive their generating action and relations constituted in the saying is about to do real grammatical work: it is what separates the domain of the genitive (relations standing without a present verb) from the domain of the karakas (relations constituted by a present verb). Bhartrhari's semantics of relation always runs this way — through the temporality of how the relation came to stand (Kanda III's Sambandha-samuddesa will take it up at length).

2.198 unformalized

तत्र षष्ठी प्रतिपदं समासस्य निवृत्तये ।
विहिता दर्शनार्थं तु कारकं प्रत्युदाहृतम् ॥ १९८ ॥

tatra ṣaṣṭhī pratipadaṃ samāsasya nivṛttaye |
vihitā darśanārthaṃ tu kārakaṃ pratyudāhṛtam || 198 ||

There the sixth case is prescribed word by word, for the exclusion of the compound; but the karaka has been cited as a counter-example, for the sake of showing.

The technical seat of 2.197's first pattern: the genitive. Where the generating action has withdrawn, the relation is expressed by sasthi, the sixth case — and it is prescribed pratipadam, word by word, samasasya nivrttaye: *for the exclusion of the compound* — that is, the rule teaches the genitive individually so as to block automatic compounding (rajnah purusah rather than rajapurusah, where the analysis is to be kept visible). The grammar's motive here is purely operational: the case-ending is posted where the science needs the relation *displayed*. And then: darsanartham tu karakam pratyudahrtam — the karaka has been cited as a *counter-example*, for the sake of showing: where the action *is* present (2.197's second pattern), the relation is a karaka-relation and the genitive does not apply — the counter-example teaches the boundary. Two devices, two domains, one distinction — and the distinction is the temporal one 2.197 drew. The verse is a small model of the kanda's method: a semantic observation about relations, then the grammar's machinery shown to be its faithful engineering.

2.199 unformalized

स चोपजातः सम्बन्धो विनिवृत्ते क्रियापदे ।
कर्मप्रवचनीयेन तत्र तत्र नियम्यते ॥ १९९ ॥

sa copajātaḥ sambandho vinivṛtte kriyāpade |
karmapravacanīyena tatra tatra niyamyate || 199 ||

And that arisen connection, when the action-word has withdrawn, is restricted, in this case and that, by the karmapravacaniya.

The third device in the system, and the most delicate. Consider a relation that *arose* from an action (2.197's first pattern) and now stands with the verb gone. It is vague — a relation, but which? Here the grammar deploys the karmapravacaniya: the class of words (anu, pari, adhi, prati and the rest) that Panini defines as former preverbs now serving to *specify* a relation rather than to qualify an action. Tatra tatra niyamyate: the standing connection is *restricted*, case by case, by this element — 'anu' fixing it as following-upon, 'prati' as directed-toward. The name itself is the doctrine: karma-pravacaniya, 'that which speaks *about* the action' — it does not name the action, it speaks *of* it, from outside, retroactively specifying the relation the withdrawn action left. This is a category that exists precisely because the grammar needed a place for words that are neither preverbs (no live verb to qualify) nor case-endings (they are not endings) but which nonetheless pin down relations — and Bhartrhari's interest is that its very existence testifies to relation-specification as a distinct semantic office (2.204 will state its character exactly).

2.200 unformalized

येन क्रियापदाक्षेपः स कारकविभक्तिभिः ।
युज्यते विर्यथा तस्य लिखावनुपसर्गता ॥ २०० ॥

yena kriyāpadākṣepaḥ sa kārakavibhaktibhiḥ |
yujyate viryathā tasya likhāvanupasargatā || 200 ||

That by which an action-word is implied is construed with the karaka case-endings — as, for example, the non-preverb-hood of 'vi' in the case of 'likh'.

The boundary-test between the categories. Yena kriya-padaksepah — that element *by which an action-word is implied*, which brings a verb into the construal even though none is uttered — sa karaka-vibhaktibhih yujyate: *that* one is construed with the karaka case-endings, and hence belongs on the live-action side of 2.197's divide (the case-endings being the machinery of verb-governed relations, 2.198). The example is a grammarian's: vi- with likh (to write, to scratch) exhibits anupasargata, non-preverb-hood — vi- there is not qualifying a live action in the way a preverb does, and its behaviour with the case-endings shows it. The details of the specific Paninian ruling matter less than the method being displayed: the categories (preverb, karmapravacaniya, particle) are distinguished by *how they behave with the case-apparatus*, which is to say, by whether the relation they touch is verb-constituted or verb-abandoned. Bhartrhari is showing that the grammar's finest categorial cuts are made on exactly the semantic distinction he drew at 2.197 — and this is the recurring pattern of the kanda's argument from grammar: the science's postings are not arbitrary, they are the semantics engineered.
contested: the transcription is faulty at likhav anupasargata and the intended Paninian case is not securely recoverable; we translate the evident sense and note the corruption

2.201 unformalized

तिष्ठतेरुपयोगश्च दृष्टोऽप्रत्यजयन्निति ।
सुन्बन्तीत्याभिमुख्येऽभिः केवलोऽपि प्रयुज्यते ॥ २०१ ॥

tiṣṭhaterupayogaśca dṛṣṭo'pratyajayanniti |
sunbantītyābhimukhye'bhiḥ kevalo'pi prayujyate || 201 ||

And the use of 'stha' is seen in 'apratyajayan'; and in 'sunbanti', in the sense of facing-toward, 'abhi' is employed even alone.

Two attested usages cited as data — and their state in the source is poor. The first illustrates a root's employment in a context where the expected element is absent; the second, more pointedly, exhibits abhi employed kevalah, *alone*, in the sense of abhimukhya, facing-toward: a preverb standing by itself and delivering a determinate sense. That second datum matters for the movement, because it cuts against 2.194's dependency-claim for particles and shows that the classes are not tidy: some elements the grammar files as preverbs can, in attested (probably Vedic) usage, function independently. Bhartrhari's practice throughout has been to record such untidiness rather than legislate it away (2.157's mass-terms, 2.175's both-ways settlement), and the honest moral is 2.179's: difference and non-difference both admissible, the categories being purpose-relative postings and not natural kinds. The verse is also a reminder that this grammar's categories were built to fit a corpus, and the corpus keeps producing residue.
contested: the transcription of both examples (apratyajayan, sunbanti) is corrupt and the underlying Vedic citations are not recoverable from it; our rendering translates what the text gives and we flag that the specific attestations are uncertain

2.202 unformalized

कर्मप्रवचनीयत्वं क्रियायोगे विधीयते ।
षत्वादिविनिवृत्त्यर्थं स्वत्यादीनां विधर्मणाम् ॥ २०२ ॥

karmapravacanīyatvaṃ kriyāyoge vidhīyate |
ṣatvādivinivṛttyarthaṃ svatyādīnāṃ vidharmaṇām || 202 ||

Karmapravacaniya-hood is prescribed in connection with an action, for the sake of excluding the retroflexion of 's' and the like, in the case of 'sva-ati' and the rest, which are of contrary property.

The karmapravacaniya category shown to be, in part, a phonological posting. Its being *prescribed* kriya-yoge, in connection with an action, has a purpose the verse names bluntly: satva-adi-vinivrtty-artham — *for the exclusion of the retroflexion of s and the like*. Sanskrit's sandhi rules retroflex an s after certain preverbs; classify an element as karmapravacaniya rather than upasarga and the retroflexion is blocked. So the category earns part of its keep by getting the sound-changes right — vidharmanam, for elements 'of contrary property', whose phonological behaviour deviates from what their apparent class would predict. This is 2.180's lesson recurring at the highest level of the movement: a categorial distinction that looks semantic is, at least partly, posted for the augment's and the sandhi's sake. Bhartrhari is not debunking the category (2.199 and 2.204 give it a real semantic office); he is showing, with his usual exactness about the science's motives, that its boundaries are drawn where the *rules* need them — which is why no inference runs from the grammar's categories to the meaning's joints.
contested: the transcription's svatyadinam we read as sva-ati-adinam (the elements ati etc.); the specific list is uncertain but the satva-blocking motive is clear and is the verse's cargo

2.203 unformalized

हेतुहेतुमतोर्योगपरिच्छेदेऽनुना कृते ।
आरम्भाद्धाध्यते प्राप्ता तृतीया हेतुलक्षणा ॥ २०३ ॥

hetuhetumatoryogaparicchede'nunā kṛte |
ārambhāddhādhyate prāptā tṛtīyā hetulakṣaṇā || 203 ||

When the delimitation of the connection of cause and effect has been effected by 'anu', the third case, whose mark is causality, though obtained, is annulled from the outset.

A worked case, showing the karmapravacaniya displacing a case-ending. A causal relation would ordinarily be expressed by the third case (trtiya, the instrumental: hetu-laksana, 'whose defining mark is causality' — 'by reason of X'). But when 'anu' has already effected the yoga-pariccheda — the *delimitation of the connection* between hetu and hetumat, cause and effect — then arambhat badhyate prapta trtiya: the instrumental, though obtained by rule, is annulled *from the outset*. The relation having been pinned by the particle, the case-ending that would pin it is pre-empted; two devices for one office, and the science lets the first-comer hold it. The verse is a small but exact illustration of the badha-machinery (2.66's annulment, 2.73's hierarchy) operating between *devices* rather than between meanings — and of the economy that governs the whole system: no relation is specified twice. It also shows the karmapravacaniya doing genuine semantic work (delimiting a relation), which 2.204 is about to state in its most precise form.

2.204 unformalized

क्रियाया द्योतको नायं सम्बन्धस्य न वाचकः ।
नापि क्रियापदाक्षेपी सम्बन्धस्य तु भेदकः ॥ २०४ ॥

kriyāyā dyotako nāyaṃ sambandhasya na vācakaḥ |
nāpi kriyāpadākṣepī sambandhasya tu bhedakaḥ || 204 ||

This one is not a manifestor of the action, nor an expressor of the connection; nor does it imply an action-word — but it is a differentiator of the connection.

The karmapravacaniya defined by four exclusions and one positive — and the definition is a small masterpiece of categorial precision. It is *not* kriyaya dyotakah, a manifestor of the action (so it is not a preverb, whose office is precisely that: 2.187-188). It is *not* sambandhasya vacakah, an expressor of the connection (so it is not a case-ending, which denotes the relation: 2.198). It does *not* kriya-padaksepin, imply an action-word (so it fails the test of 2.200 and does not belong on the live-verb side). What it *is*: sambandhasya bhedakah — a *differentiator of the connection*: it takes a relation already standing (2.197's withdrawn-action relation, 2.199) and specifies *which* relation it is, without naming it, without lighting up any verb, without bringing a verb into play. The office is real, distinct from every neighbouring one, and — this is the point for the kanda — it is an office of *operating on an already-present whole*: differentiating, not contributing. That is the same office 2.187 gave the preverb (manifesting a latent particular) and 2.192-193 gave the particle (illuminating from any position). Three categories, three ways of touching a meaning that is already there in one piece. Not one of them adds a brick.

2.205 · Complexes and Their Meanings unformalized

अनर्थकानां सङ्घातः सार्थकोऽनर्थकस्तथा ।
वर्णानां पदमर्थेन युक्तं नावयवाः पदे ॥ २०५ ॥

anarthakānāṃ saṅghātaḥ sārthako'narthakastathā |
varṇānāṃ padamarthena yuktaṃ nāvayavāḥ pade || 205 ||

A complex of meaningless things is meaningful; and likewise meaningless. The word of phonemes is joined with meaning; the parts in the word are not.

A summarizing quartet — the kanda's positions on complexes, stated with almost algebraic compression. Anarthakanam sanghatah sarthakah: a complex of *meaningless* items is meaningful — the phonemes, severally null (2.12, 2.53), constitute a meaning-bearing word: emergence, and the standing datum every semantics must face. Tatha anarthakah: and *likewise meaningless* — some complexes of meaningless items remain meaningless (the augment plus the marker, the syllable-strings that are not words). So complexity alone does not confer meaning; something else must. And then the settlement, in the plainest words the kanda ever uses: varnanam padam arthena yuktam — the word of phonemes *is joined with meaning*; na avayavah pade — the *parts in the word are not*. Meaning attaches at the word and does not distribute inward: everything the movement since 2.180 has argued (roots and preverbs, particles, suffixes, augments — none of them a semantic atom) here reduced to a single line. The next verses take the same question one level up, to complexes of *meaningful* things.
contested: the second clause (tatha anarthakah) is elliptical and readers divide over whether it means 'likewise (a complex of meaningless things may be) meaningless' (our reading) or 'and (a complex) of meaningful things (may be) meaningless'; 2.206-207 canvass both, which is why the compression here is tolerable

2.206 unformalized

पदानामर्थयुक्तानां सङ्घातो भिद्यते पुनः ।
अर्थान्तरावरोधेन सम्बन्धविगमेव च ॥ २०६ ॥

padānāmarthayuktānāṃ saṅghāto bhidyate punaḥ |
arthāntarāvarodhena sambandhavigameva ca || 206 ||

But a complex of words joined with meaning is differentiated again, by the exclusion of a further meaning and by the departure of the connection.

One level up: complexes of *meaningful* items. The words, each artha-yukta (2.205's line, granted at their own grain), form a sanghata — and that complex bhidyate punah, is *differentiated again*: it acquires a character of its own, and the verse names the two ways. Arthantara-avarodhena: by the *exclusion of a further meaning* — the complex bars a sense its members would have admitted (compounds notoriously narrow: 'rajapurusa' does not cover every man related to a king). And sambandha-vigamena: by the *departure of the connection* — the relation that generated the compound withdraws and leaves the whole standing on its own (2.197's first pattern, now at the level of word-formation: the compound outlives the syntax that made it). Both mechanisms are subtractive — the complex is *not* the sum plus something; it is the sum minus the freedom of its members. This is a genuinely acute observation about lexicalization, and it prepares the two disputes the next verses report: whether meaningful-plus-meaningless can combine, and whether a complex has a meaning beyond its parts.

2.207 unformalized

सार्थकानर्थकौ मेदे सम्बन्धं नाधिगच्छतः ।
अधिगच्छत इत्येके कृटीरादिनिदर्शनात् ॥ २०७ ॥

sārthakānarthakau mede sambandhaṃ nādhigacchataḥ |
adhigacchata ityeke kṛṭīrādinidarśanāt || 207 ||

A meaningful and a meaningless, being different, do not attain connection; they do attain it, say some, on the evidence of 'kutira' and the like.

The first dispute, and it is exactly the question 2.191's svarthika suffixes forced. Can a meaningful element and a meaningless one *connect*? The negative case is principled: sarthaka-anarthakau bhede sambandham na adhigacchatah — being of different kinds, they do not attain connection: a semantic relation needs two terms with semantic standing, and a null formative offers none; you cannot relate something to nothing. The affirmative case is empirical: adhigacchata iti eke, some say they do — kutira-adi-nidarsanat, on the evidence of 'kutira' and its kind: words whose analysis yields a meaningful stem plus a formative that means nothing, and which nonetheless behave as unified derivatives, the null element visibly doing formal work. The dispute has no resolution offered, and its irresolution is the point: the analyst's inventory contains meaningful and meaningless items whose 'combination' is either impossible or ubiquitous depending on which grain one credits — which is precisely what one expects if the items are postings and the combination a description (2.10, 2.88). The whole is a word; its pieces are the grammar's bookkeeping; asking how the pieces connect is asking how entries in a ledger touch.

2.208 unformalized

अर्थवद्भ्यो विशिष्टार्थः सङ्घात उपजायते ।
नोपजायत इत्येके समासस्वार्थिकादिषु ॥ २०८ ॥

arthavadbhyo viśiṣṭārthaḥ saṅghāta upajāyate |
nopajāyata ityeke samāsasvārthikādiṣu || 208 ||

From meaningful elements a complex arises with a distinct meaning; it does not so arise, say some, in the case of compounds, self-meaning suffixes and the like.

The second dispute, and it is the kanda's own thesis contested from within. Arthavadbhyo visistarthah sanghata upajayate — from meaningful elements, a complex arises *with a distinct meaning*: the whole means something the parts did not (2.143's anyaiva at the word's scale; 2.169's kupa-supa-yupa; 2.181's samgramayati). Nopajayate iti eke: it does *not* so arise, say others — samasa-svarthikadisu, in the case of compounds and self-meaning suffixes: precisely where the grammar teaches that the formation adds nothing (the svarthika means 'in its own meaning', i.e. the base's; many a compound is held to convey just its members' relation). The two parties are reading the same data in opposite directions — one sees emergence everywhere and calls the grammar's 'adds nothing' a bookkeeping convenience; the other sees the grammar's own testimony that some complexes are semantically inert. Bhartrhari reports without deciding, and by now the reader knows why: both are correct at their grain (2.179's licensed double description), and what settles the question in any instance is not the pieces but the comprehension the whole occasions (2.34's pramana). The catalogue of complexes will continue; its moral has been fixed since 2.143.

2.209 unformalized

केचित्तु युतासिद्धार्था भेदे निर्ज्ञातशक्तयः ।
अन्वयव्यतिरेकाभ्यां केचित् कल्पितशक्तयः ॥ २०९ ॥

kecittu yutāsiddhārthā bhede nirjñātaśaktayaḥ |
anvayavyatirekābhyāṃ kecit kalpitaśaktayaḥ || 209 ||

But some elements have their meaning inseparably established, their capacities known in isolation; others have their capacities merely constructed, by agreement and difference.

A distinction among the analyst's own units, and it grades them by evidential warrant. Some are yuta-siddha-arthah — their meaning established as *conjoined*, inseparable from the whole they occur in — and yet bhede nirjnata-saktayah, their capacities *known in isolation*: independently attested, met elsewhere on their own, so that crediting them in the complex rests on real evidence (2.166's conditions for trusting distributional method). Others are kalpita-saktayah — their capacities *merely constructed*, and constructed anvaya-vyatirekabhyam, by agreement and difference alone: the method finds a recurrence and posits a power to explain it, with no independent attestation whatever. The two grades are not equal, and the verse's quiet work is to say so: some of grammar's units are discovered, others are fabricated by the very procedure that reports them (2.167's sap, meaningless and undeniable). Since the whole preceding movement has shown that meaning attaches at wholes, this grading tells the analyst which of his posits he may lean on and which are his own shadow.

2.210 unformalized

शास्त्रार्थ एव वर्णानामर्थवच्वे प्रदर्शितः ।
धात्वादीनां विशुद्धानां लौकिकोऽर्थो न विद्यते ॥ २१० ॥

śāstrārtha eva varṇānāmarthavacve pradarśitaḥ |
dhātvādīnāṃ viśuddhānāṃ laukiko'rtho na vidyate || 210 ||

It is only the scientific meaning of the phonemes that has been shown, in the matter of their meaningfulness; of roots and the rest, in their pure form, no worldly meaning exists.

The distinction between two kinds of meaning, and it governs everything the movement has been reporting. When the grammar treats an element as arthavat — meaningful — the meaning in question is sastrartha eva, *only the scientific meaning*: the value assigned within the system, for the system's operations. Dhatvadinam visuddhanam laukiko 'rtho na vidyate: of roots and the rest *in their pure form* — the root as the grammar isolates it, stripped of every affix and preverb — no *worldly* meaning exists at all. Nobody ever means 'pac' at anyone; the pure root is not a thing one can say. Its 'meaning' is an entry in a derivational ledger, real as an entry and null as an utterance. This is the sharpest formulation the kanda gives of its central methodological claim (2.10, 2.88): the units of analysis have meanings *in the analysis*, and to slide from that to meanings in speech is to mistake the ledger for the world. The verses following extend the diagnosis to the affixes.
contested: sastrartha we render 'scientific meaning' (the value within the science); the compound also permits 'the meaning taught in the sastra' — the deflationary point, that it is not laukika, is explicit either way

2.211 unformalized

कृत्तद्धितानामर्थश्च केवलानामलौकिकः ।
प्राग्विभक्तेस्तदन्तस्य तथैवार्थो न विद्यते ॥ २११ ॥

kṛttaddhitānāmarthaśca kevalānāmalaukikaḥ |
prāgvibhaktestadantasya tathaivārtho na vidyate || 211 ||

And the meaning of primary and secondary suffixes, taken alone, is non-worldly; likewise, prior to the case-ending, the meaning of what ends in them does not exist.

The diagnosis of 2.210 extended in two steps. First: krt and taddhita suffixes — the primary and secondary affixes, the grammar's great derivational engines — kevalanam alaukikah: taken *alone*, their meaning is non-worldly. The -ana of 'karana', the -tva of 'gotva' mean nothing anybody ever meant; they mean within the derivation. Second, and this goes further: prag vibhaktes tad-antasya tatha eva artho na vidyate — *prior to the case-ending*, the meaning of what ends in those suffixes likewise does not exist: even the derived stem, suffix and all, has no meaning until inflected. 'Karana-' is not yet a thing said; 'karanena' is. This is the strictest possible statement of where meaning enters: not at the root, not at the derived stem, but at the *padа* — the inflected word, the smallest thing utterable (2.205's line: 'the word of phonemes is joined with meaning; the parts in the word are not'). Everything the derivation traverses on its way there is scaffolding, meaningful only in the ledger. Panini's own system says as much in its way, teaching that a form is not a pada until it bears its ending.

2.212 unformalized

अभिव्यक्ततरो योऽर्थः प्रत्ययान्तेषु लक्ष्यते ।
अर्थवत्ताप्रकरणादाश्रितः स तथाविधः ॥ २१२ ॥

abhivyaktataro yo'rthaḥ pratyayānteṣu lakṣyate |
arthavattāprakaraṇādāśritaḥ sa tathāvidhaḥ || 212 ||

The meaning that is more manifest, which is observed in forms ending in a suffix — that, being of such a kind, has been resorted to on account of the section on meaningfulness.

Why, then, does the grammar talk as though suffixed forms had meanings? Because they *more manifestly* do: abhivyaktataro yo 'rthah pratyayantesu laksyate — the meaning observed in forms ending in a suffix is more manifest, closer to the surface, nearer to something one could say. And the science asritah — has *resorted* to that appearance — arthavatta-prakaranat, on account of its own section on meaningfulness: because a rule-system needs to classify its elements as meaningful or not, it takes the most manifest available meaning and posts it. Tathavidhah: 'being of such a kind' — that is, being a matter of degree of manifestness rather than of presence or absence. The verse completes 2.210-211 with an account of the grammar's motive that is neither debunking nor credulous: the science's assignments of meaning to sub-word elements track a real gradient (some pieces are nearer to sayability than others) and posit, at a convenient point on that gradient, a boundary that speech does not draw. Postings again — but *motivated* postings, which is why the science works.
contested: arthavatta-prakarana refers to the grammar's treatment of meaningfulness (the arthavat-section); precisely which discussion is meant is not recoverable from the karika, and our rendering keeps it general

2.213 unformalized

आत्मभेदो न चेत् कश्चिद्वर्णाभ्यः पदवाक्ययोः ।
अन्योन्यापेक्षया शक्त्या वर्णाः स्यादभिधायकः ॥ २१३ ॥

ātmabhedo na cet kaścidvarṇābhyaḥ padavākyayoḥ |
anyonyāpekṣayā śaktyā varṇāḥ syādabhidhāyakaḥ || 213 ||

If there is no difference of self of word and sentence from the phonemes, then the phonemes, by a capacity dependent on one another, would be the expressor.

A conditional put to the phoneme-realist, and it is a trap he cannot escape by either exit. Suppose atma-bhedo na kascit — there is no *difference of self* between the phonemes on one side and word-and-sentence on the other: the word just *is* its phonemes, nothing over and above (2.52's nominalism, 2.72's atomism). Then what expresses? Not any phoneme severally, since none means (2.53). So: anyonya-apeksaya saktya varnah syad abhidhayakah — the phonemes, by a capacity *dependent on one another*, would be the expressor. But notice what has just been conceded: a capacity that exists only in mutual dependence is not a property any phoneme has; it is a property of their standing-together — which is to say, of the whole, under another name. The realist who refuses a distinct word-self must relocate expressive power into a relational capacity, and the relational capacity is the word-self he refused. The argument is a fine specimen of the kanda's habit: not refuting the atomist but letting him restate holism in his own vocabulary (2.213's apeksika sakti is exactly 2.168's relationally-fixed power).

2.214 unformalized

वर्णोन केनचिन्नयूनः सङ्घातो योऽभिधायकः ।
न चेच्छब्दान्तरमसावन्यूनस्तेन गम्यते ॥ २१४ ॥

varṇona kenacinnayūnaḥ saṅghāto yo'bhidhāyakaḥ |
na cecchabdāntaramasāvanyūnastena gamyate || 214 ||

The complex that is the expressor is not deficient by any phoneme; if it is not a different word, it is understood by that as not deficient.

The consequence of 2.213's concession, and it is stringent. The sanghata yo 'bhidhayakah — the complex that does the expressing — na kenacid varnena nyunah: is *not deficient by any phoneme*: every one is required; drop one and this complex is not that complex. Then the dilemma's second horn: na ced sabdantaram asau — if the phoneme-diminished string is *not a different word* (if the realist wants to say it is the same word, imperfectly uttered), anyunas tena gamyate: it is understood by that very (claim) as non-deficient — i.e. he must hold that the shortened string still expresses, which contradicts the requirement he just granted. The realist is caught between saying that every phoneme is constitutive (in which case the word is a whole with no tolerance, and its unity is doing the work) and saying that some are dispensable (in which case he cannot say which, or why). Non-deficiency is a property of wholes: this is 2.34's non-monotonicity (add or subtract, and you get not more or less meaning but meaninglessness) driven down to the phonemic grain.

2.215 unformalized

स तस्मिन् वाचकें शब्दे निमित्तात् स्मृतिमादधत् ।
साक्षादिव व्यवहितं शब्देनार्थमुपोहते ॥ २१५ ॥

sa tasmin vācakeṃ śabde nimittāt smṛtimādadhat |
sākṣādiva vyavahitaṃ śabdenārthamupohate || 215 ||

That one, depositing a memory from its occasioning ground in the expressive word, brings the mediated meaning near, as it were directly, through the word.

How the mediation actually feels — and the verse describes the phenomenon that makes all the atomist's mistakes tempting. Something (the phoneme, the part, the occasioning element) nimittat smrtim adadhat — *depositing a memory* from its occasioning ground in the expressive word (the seed-and-trace apparatus of 1.84, the association-channel of 2.163) — saksad iva vyavahitam sabdena artham upohate: brings the *mediated* meaning near *as though directly*, through the word. Two words carry it. Vyavahita: the meaning is mediated, reached through a relay (2.152's doctrine: the corrupt form expressing with the correct one interposed; here generalized). Saksad iva: *as if* directly — the mediation is invisible from inside; the hearer experiences immediacy. That is precisely why the parts seem to carry meaning: their contribution is real, it is mediated, and the mediation does not show. The whole analytic error the kanda has been correcting is the error of reading the phenomenology's *saksat* off the machinery's *vyavahita* — of taking the felt immediacy of comprehension for evidence that what delivered it was simple.

2.216-217 unformalized

पदवाच्यो यथा नार्थः कश्चत् गौरखरादिषु ।
सत्यपि प्रत्ययेऽत्यन्तं समुदाये न गम्यते ॥ २१६ ॥
समन्वित इवार्थात्मा पदार्थैर्यः प्रतीयते ।
पदार्थदर्शनं तत्र तथैवानुपकारकम् ॥ २१७ ॥

padavācyo yathā nārthaḥ kaścat gaurakharādiṣu |
satyapi pratyaye'tyantaṃ samudāye na gamyate || 216 ||
samanvita ivārthātmā padārthairyaḥ pratīyate |
padārthadarśanaṃ tatra tathaivānupakārakam || 217 ||

(216) As no meaning expressible by a word exists in 'gaukhara' and the like — though a cognition does occur, it is not at all understood from the aggregate; (217) so the self of the meaning which is understood, as though fully construed with the word-meanings — there, the seeing of word-meanings is likewise of no assistance.

The strongest case for the analyst, turned against him. 'Gaukhara' (cow-donkey, and its kin: the tradition's dvandvas of incongruous animals) — here, patently, there is no *word-expressible* meaning: no such beast, no unified sense. And yet: saty api pratyaye — *though a cognition does occur*: the hearer does get something, an image, a construal, a suspicion of sense. But atyantam samudaye na gamyate: it is *not at all* understood from the aggregate — the cognition that occurs is not delivered by the compound as a compound; it is the hearer's own manufacture, cobbled from the members' other lives (2.92's fallacy exactly: projection by resemblance, taken for comprehension). Then (217) the generalization that stings: samanvita iva arthatma yah padarthaih pratiyate — the meaning-self that is understood *as though fully construed* with the word-meanings — tatra padartha-darsanam tathaiva anupakarakam: there too, the seeing of word-meanings is *of no assistance*. The felt construal-with-parts is the same illusion in the ordinary case as in gaukhara; in both, the appearance of parts-delivering-sense is retrospective, and in neither did the parts do the delivering. The freak compound is not an anomaly to be quarantined — it is the normal case with its mechanism visible.
contested: gaukhara is our reading of the transcription's gaur-akhara-adisu; the class intended (incongruous dvandva-like formations yielding no unified sense) is clear from the argument

2.218 unformalized

समुदायावयवयोर्भिन्नार्थत्वे च वृत्तिषु ।
युगपद् भेदसंसर्गौ विरुद्धावनुषङ्गिणौ ॥ २१८ ॥

samudāyāvayavayorbhinnārthatve ca vṛttiṣu |
yugapad bhedasaṃsargau viruddhāvanuṣaṅgiṇau || 218 ||

And if the aggregate and its members have different meanings, then in the derivative formations difference and commingling, being contradictory, would attach simultaneously.

A reductio against the middle position — the one that wants both real member-meanings and a real aggregate-meaning. Grant samudaya-avayavayoh bhinnarthatve: the aggregate and its members have *different* meanings, each real. Then in the vrttis — the derivative formations, compounds and the like — yugapad bheda-samsargau: *simultaneously*, difference and commingling — the members must be semantically distinct (else there is no aggregate *of* them) and semantically fused (else there is no aggregate-meaning at all) — viruddhau anusanginau: and these two, *contradictory*, would both attach. The compound would have to be, at one and the same time, a plurality of meanings and a single meaning other than them. This is the antinomy 2.72-73 found for parts and wholes, now sprung on the compromiser who thought he could hold both ends. Bhartrhari's own escape has been available since 2.88: the difference is *apoddhrta*, abstracted according to capacities, and the unity is real — one meaning, articulable, not composed. The contradiction arises only for someone who reifies both the members and the whole; it dissolves for someone who reifies only the whole and lets the members be its capacities.

2.219 unformalized

कश्च साधनमात्रार्थानध्यादीन् परिकल्पयेत् ।
अप्रयुक्तपदश्चार्थो बहुव्रीहौ कथं भवेत् ॥ २१९ ॥

kaśca sādhanamātrārthānadhyādīn parikalpayet |
aprayuktapadaścārtho bahuvrīhau kathaṃ bhavet || 219 ||

And who would construct 'adhi' and the rest as having a mere means for their meaning? And how could a meaning whose word is not employed occur in a bahuvrihi?

Two hard cases for member-meaning realism, put as rhetorical questions. First: kas ca sadhana-matrarthan adhy-adin parikalpayet — who would posit that 'adhi' and its kind have a mere *sadhana*, a bare means or instrumental relation, for their meaning? The preverb in a compound does not denote a relation-token that could be summed with the rest; nobody analysing 'adhyayana' finds an instrument sitting there (2.190's obligatory preverbs already showed the impossibility). Second, and it is the classic: aprayukta-padas ca artho bahuvrihau katham bhavet — how could a meaning *whose word is not employed* occur in a bahuvrihi? The bahuvrihi ('citragu' — one whose cows are dappled) means precisely the *possessor*, and no word for the possessor occurs in the compound: the meaning that the compound is *about* is the one member that is unuttered. For member-meaning realism this is fatal — you cannot sum a meaning that has no vehicle. For the kanda's holism it is expected: the compound is one word with one meaning, its 'members' being the analyst's postings, and there is no more difficulty in its meaning a possessor than in 'asvakarna' meaning a tree (2.36).

2.220 unformalized

प्रज्ञसंज्ञ्वाद्यवयवैर्न चास्त्यर्थावधारणाम् ।
तस्मात् संघात एवैको विशिष्टार्थनिबन्धनम् ॥ २२० ॥

prajñasaṃjñvādyavayavairna cāstyarthāvadhāraṇām |
tasmāt saṃghāta evaiko viśiṣṭārthanibandhanam || 220 ||

Nor is there determination of meaning by such members as 'prajna' and 'samjna'; therefore the complex alone, being one, is the ground of the distinct meaning.

The movement's conclusion, drawn from the last specimens. Words like prajna and samjna are analysable — pra + jna, sam + jna, the pieces attested elsewhere and glossable — and yet na asti artha-avadharanam: there is *no determination of meaning* by those members: no amount of attention to 'pra' and 'jna' settles what prajna means; the analysis, once the meaning is known, looks apt, and it could never have yielded it. Hence, tasmat: sanghata eva ekah visistartha-nibandhanam — the *complex alone*, being one, is the ground of the distinct meaning. Every term is chosen. Eka: one, not a sum. Nibandhana: the *ground*, the kanda's word for what meaning fastens to since 1.13. Visistartha: the distinct meaning, the one that is not the members' (2.208's disputed emergence, here affirmed). The movement that began at 2.205 with a compressed line ends with it demonstrated: meaning attaches to the whole word and does not distribute inward — proved not by metaphysics but by walking through the grammar's own inventory, case by case, and finding that at no level below the utterable word does anything mean.

2.221 · The Dvandva and the Distribution of Action unformalized

गर्गा इत्येक एवायं बहुष्वर्थेषु वर्तते ।
द्वन्द्वसंज्ञोऽपि सङ्घातो बहूनामभिधायकः ॥ २२१ ॥

gargā ityeka evāyaṃ bahuṣvartheṣu vartate |
dvandvasaṃjño'pi saṅghāto bahūnāmabhidhāyakaḥ || 221 ||

'Gargah' — this single one operates upon many meanings; and the complex called dvandva likewise is the expressor of many.

A new movement, on the compounds that mean *pluralities* — and it opens with the cleanest specimen. 'Gargah', the patronymic plural: eka eva ayam — this is a *single* word, one form, one unit — bahusu arthesu vartate: and it operates upon *many* meanings, the whole clan of Garga's descendants. One word, many things meant, and no plurality in the word. Dvandva-samjno 'pi sanghato bahunam abhidhayakah: and the complex called dvandva likewise ('rama-krsnau', 'plaksa-nyagrodhau') expresses many. The pairing sets the movement's problem: these are the cases where multiplicity is unmistakably *in the meaning* while the word remains one. If ever a word were going to be a plurality, it would be here. And the movement's argument will be that even here it is not: what is many is what is meant, and the manner in which the action distributes over the many is a matter of how the *one* meaning articulates — not of the word coming apart into members. The dvandva is the test the holism must pass at its hardest.

2.222 unformalized

यथैकदेशे भुज्यादिः प्रत्येकमवतिष्ठते ।
क्रियैवं द्वन्द्ववाच्येऽर्थे प्रत्येकं प्रविभज्यते ॥ २२२ ॥

yathaikadeśe bhujyādiḥ pratyekamavatiṣṭhate |
kriyaivaṃ dvandvavācye'rthe pratyekaṃ pravibhajyate || 222 ||

As eating and the like stand, each severally, in a single portion — so the action, in the meaning expressed by a dvandva, is divided out severally.

The distribution-phenomenon named. Yatha ekadese bhujy-adih pratyekam avatisthate: as eating and such stand *severally* in a single portion — the eating of a heap of food is, at any moment, the eating of a mouthful; the action realizes itself portion by portion while remaining one action. Kriya evam dvandva-vacye 'rthe pratyekam pravibhajyate: so the action, in the meaning a dvandva expresses, is divided out *severally* — 'he sees Rama and Krsna' involves a seeing that reaches each. The observation is exact and the analogy is chosen to defuse it: distributive realization is not decomposition. Nobody thinks the eating of the heap is *composed of* mouthful-eatings summed; it is one action that meets its object portion-wise, because that is how eating meets heaps. Likewise the seeing distributes over Rama and Krsna without the seeing coming in halves, and without the dvandva coming apart. The verse establishes the movement's key move: distribution belongs to the *meeting of action and object*, not to the structure of the word.
contested: ekadese ('in a single portion') we take with the eating (the action realized portion-wise); it can be read 'in one part (of the dvandva)', making the analogy tighter but the argument circular — we prefer the independent reading

2.223 unformalized

यच्च द्वन्द्वपदार्थस्य तच्छब्देन व्यपेक्षणा ।
साऽपि व्यावृत्तरूपेऽर्थे सर्वनामसरूपता ॥ २२३ ॥

yacca dvandvapadārthasya tacchabdena vyapekṣaṇā |
sā'pi vyāvṛttarūpe'rthe sarvanāmasarūpatā || 223 ||

And the mutual expectancy which the dvandva's word-meaning has with a pronoun — that too, in a meaning whose form has been excluded, is a likeness to a pronoun.

A refinement, from anaphora. A dvandva's meaning can be picked up by a pronoun ('Rama and Krsna — *they* went'), and the pronoun's reaching back constitutes a vyapeksana, a mutual expectancy with the compound's meaning. The observation: sa api vyavrtta-rupe 'rthe sarvanama-sarupata — that expectancy too, in a meaning *whose form has been excluded* (the pronoun's own content being empty: a sarvanaman has no descriptive form of its own, only the reaching), is a *likeness to a pronoun*. The point is that the dvandva's meaning behaves, for anaphoric purposes, as a single antecedent — one thing to be picked up, however many things it comprises. Pronominal reference is the standard test for unity of a semantic item, and the dvandva passes it: 'they' takes the compound's meaning whole, not its members severally. So the plurality in the meaning does not fragment the meaning *as an item* — which is exactly what the movement needs, and what a member-realist cannot explain (his compound has no single antecedent to offer).
contested: the verse is among the kanda's most compressed and our construal follows the anaphoric reading; some take sarvanama-sarupata as the dvandva's members acquiring pronoun-like emptiness in the compound — a possible but less motivated reading

2.224 unformalized

यथा च खदिरच्छेदे भागेषु क्रमवांश्छिदिः ।
तथा द्वन्द्वपदार्थस्य भागेषु क्रमदर्शनम् ॥ २२४ ॥

yathā ca khadiracchede bhāgeṣu kramavāṃśchidiḥ |
tathā dvandvapadārthasya bhāgeṣu kramadarśanam || 224 ||

And as, in the cutting of khadira-wood, the cutting is sequential in the parts — so is the seeing of sequence in the parts of the dvandva's word-meaning.

The distribution given its temporal dimension, with a woodcutter's example. Khadira-cchede — in cutting khadira, the hardest of woods — bhagesu kramavan chidih: the cutting is *sequential* in the parts: the axe takes the log stroke by stroke; the action unrolls through its object in order. Tatha dvandva-padarthasya bhagesu krama-darsanam: so is *sequence seen* in the parts of the dvandva's meaning — 'he honoured Rama and Krsna' is heard as an honouring that reached them one after another. And the verb is darsanam, *seeing*, not 'being': sequence is what is seen in the parts, and the kanda has established at length where such seeing comes from (2.25: the mind's own sequencing power; 2.27: sequenceless capacity generating ordered appearance). The cutting of the khadira is one action; its sequentiality is the shape of its engagement with a resistant object, not a plurality of cuttings. The dvandva's meaning is one; the sequence in which its members come into the action's reach is the shape of the engagement, not a plurality in the word. Same doctrine, third angle.

2.225 unformalized

संघैकदेशे प्रक्रान्तान् यथा सङ्घानुपातिनः ।
क्रियाविशेषान् मन्यन्ते स द्वन्द्वावयवे क्रमः ॥ २२५ ॥

saṃghaikadeśe prakrāntān yathā saṅghānupātinaḥ |
kriyāviśeṣān manyante sa dvandvāvayave kramaḥ || 225 ||

As they consider particular actions, commenced upon a single portion of a group, to be conformable to the whole group — that is the sequence in the dvandva's member.

The movement's resolving simile, and it is a legal-cum-social one. Sangha-ekadese prakrantan — actions *commenced upon a single portion* of a group (the village entered by one household, the assembly addressed by beginning with one man, the herd counted from the first beast) — sangha-anupatinah kriya-visesan manyante: such particular actions are considered *conformable to the whole group*, reckoned as done to the group entire. The action touches a part and counts as the group's. Sa dvandva-avayave kramah: *that* is the sequence in the dvandva's member — the sequential engagement (2.224) is the action's portion-wise contact with a whole that it nonetheless has as its object entire. This resolves the movement precisely. The dvandva means many, and means them as one object; the action reaches them in order, and reaches them as one group; nothing anywhere requires the word to be many. Distribution, sequence, and anaphoric unity are all accounted for by one meaning met portion-wise — the eating of the heap, the cutting of the khadira, the entering of the village. The hardest case for holism has been carried.

2.226 · The Grammarian's Devices and Their Costs unformalized

प्रतिपादयता वृत्तिमबुधान् वाक्यपूर्विकाम् ।
वृत्तौ पदार्थभेदेन प्राधान्यमुपदर्शितम् ॥ २२६ ॥

pratipādayatā vṛttimabudhān vākyapūrvikām |
vṛttau padārthabhedena prādhānyamupadarśitam || 226 ||

By one who teaches derivative formation as preceded by the sentence, to the unlearned, a principality has been shown in the formation, by way of difference of word-meanings.

The kanda turns to the grammar's own teaching-devices and the price they exact. A vrtti — a compound or derivative — is taught vakya-purvika: as *preceded by a sentence*, derived from an analytic phrase ('rajapurusa' from 'rajnah purusah'). Two things about this. It is done abudhan pratipadayata — by one *teaching the unlearned*: the analytic sentence is a pedagogical bridge (2.13's divisions for the avidvan). And in doing it, vrttau padartha-bhedena pradhanyam upadarsitam: a *principality* is exhibited within the formation, by way of the difference of word-meanings — the teacher must say which member is principal (the purusa, in rajapurusa) and which subordinate, and to say that he must treat the members as having meanings that stand in a hierarchy. The device thus imports exactly the member-realism the kanda has spent two movements dismantling. Bhartrhari's position is consistent and has been since 2.38: the device is legitimate *as* a device, discardable at attainment, and its ontological commitments are not the school's. But the next verses show what happens when the device is pressed.

2.227 unformalized

अभेदादभिधेयस्य नञ्समासे विकल्पितम् ।
प्राधान्यं बहुधा भाष्ये दोषास्तु प्रक्रियागताः ॥ २२७ ॥

abhedādabhidheyasya nañsamāse vikalpitam |
prādhānyaṃ bahudhā bhāṣye doṣāstu prakriyāgatāḥ || 227 ||

Because of the non-difference of what is expressed, in the nan-compound the principality has been variously constructed in the Bhasya; but the faults are those belonging to the derivational procedure.

The device pressed to its breaking point, on the compound that has always broken it. The nan-samasa is negation-in-compound: 'abrahmana', not-a-brahmin. Which member is principal? The negation? The brahmin? Neither is comfortable: the compound means a *person* (so the noun looks principal), but what it says of him is the negation (so the nan looks principal), and the thing meant — some non-brahmin — is not a brahmin at all, so the member that names brahmins is naming what the compound denies. The Mahabhasya, Bhartrhari notes, constructs the principality bahudha vikalpitam — in *many ways*, variously: Patanjali turns it over and settles it differently in different places (the tradition's discussions of nan are famously unresolved). The diagnosis is the verse's last three words and they are worth the whole movement: dosas tu prakriyagatah — *the faults belong to the derivational procedure*. Not to the compound, which means what it means without difficulty; not to the language, which has used nan-compounds forever; to the *procedure* — to the device of deriving compounds from sentences with principality assigned among members. The device generates the problem, and the problem is the device's, not the meaning's. Abhedad abhidheyasya: because what is expressed is *undivided*, and the machinery insists on dividing it.
contested: reading dosah prakriya-gatah as locating the faults in the procedure (rather than 'the faults arising in the derivation are (dealt with) there') is our construal and the one the argument requires; the tradition's own reading of the Bhasya's nan-discussions supports it

2.228 unformalized

जहत्स्वार्थविकल्पे च सर्वार्थत्यागमिच्छता ।
बहुव्रीहिपदार्थस्य त्यागः सर्वस्य दर्शितः ॥ २२८ ॥

jahatsvārthavikalpe ca sarvārthatyāgamicchatā |
bahuvrīhipadārthasya tyāgaḥ sarvasya darśitaḥ || 228 ||

And in the option of abandoning one's own meaning, by one who wishes the abandonment of the whole meaning, the abandonment of the entire bahuvrihi's word-meaning has been shown.

The device's other terminus: the doctrine of jahat-svartha, 'abandoning one's own meaning'. The grammarians observed that in some compounds the members do not keep their senses, and framed a doctrine accordingly: the members *abandon* their own meanings (jahat-svartha), as against ajahat-svartha where they keep them — a distinction that has to be optioned (vikalpa) because neither answer works across the board. And in the bahuvrihi, pressed by one who wishes sarvartha-tyagam — the abandonment of the *whole* meaning — the result shown is the abandonment of the *entire* word-meaning: everything the members meant is surrendered, and what the compound means (the possessor: 2.219) was never among their meanings anyway. The device has thus reached the position of saying that the members contribute by giving up everything they had. At which point, the kanda's reader is entitled to observe, the honest description is 2.220's: the complex alone, being one, is the ground of the distinct meaning — and 'the members abandon their meanings' is what member-realism has to say when it wants to say that.

2.229 unformalized

शास्त्रे क्कचित् प्रकृत्यर्थः प्रत्ययेनाभिधीयते ।
प्रकृतौ विनिवृत्तायां प्रत्ययार्थश्च धातुभिः ॥ २२९ ॥

śāstre kkacit prakṛtyarthaḥ pratyayenābhidhīyate |
prakṛtau vinivṛttāyāṃ pratyayārthaśca dhātubhiḥ || 229 ||

In the science, sometimes the base's meaning is expressed by the suffix; and when the base has withdrawn, the suffix's meaning is expressed by the roots.

The devices' arbitrariness exhibited in the derivational core, where the assignments simply trade places. Sometimes prakrty-arthah pratyayena abhidhiyate — the *base's* meaning is expressed by the *suffix*: the grammar, needing a form to come out right, credits the affix with what the stem meant. And prakrtau vinivrttayam — when the base has *withdrawn* (elided, dropped by rule: luk and its kin) — pratyayarthas ca dhatubhih: the *suffix's* meaning is expressed by the *roots*. Whatever is left standing carries whatever needs carrying. This is not a criticism of Panini; it is a description of what a derivational calculus must do to keep its books balanced when its formal operations delete and rearrange the very elements to which it assigned meanings. And it is decisive for the kanda's thesis: if 'the base's meaning' can be borne by the suffix, and 'the suffix's meaning' by the root, then these are not meanings in any sense that could be summed — they are ledger-entries transferable between accounts (2.210's sastrartha). The word means; the pieces are bookkeeping.

2.230 unformalized

यमर्थमाहतुर्भिन्नौ प्रत्ययावेक एव तम् ।
क्वचिदाह पचन्तीति धातुस्ताभ्यां विना क्वचित् ॥ २३० ॥

yamarthamāhaturbhinnau pratyayāveka eva tam |
kvacidāha pacantīti dhātustābhyāṃ vinā kvacit || 230 ||

The meaning which two distinct suffixes express, one alone sometimes expresses; 'they cook' — and sometimes the root, without those two.

The final specimen, and it closes the movement with an example anyone can check. In 'pacanti', 'they cook': the grammar analyses two distinct affixes doing distinct jobs — the class-marker (sap) and the personal ending (jhi) with its plurality and agency. Yam artham ahatur bhinnau pratyayav ekah eva tam kvacid aha — the meaning that the two express, *one alone* sometimes expresses: elsewhere in the paradigm, a single affix carries what took two here. And further: dhatus tabhyam vina kvacid — sometimes the *root*, without either of them, says it. The same semantic freight, borne by two elements, by one, or by none — across forms of the same verb. There is no stable mapping from pieces to meanings anywhere in the system; there is a system that gets the forms right and assigns meanings wherever the getting-right requires. This is the kanda's case, complete, made entirely out of the grammar's own practice: the analysis is a magnificent instrument for deriving forms and a rotten witness about meaning — because meaning was never in the pieces to begin with. It is in the word, and above the word, in the sentence; and what it is there, the kanda said at 2.143: a flash.
contested: the specific affixes in pacanti are not named by the karika; sap and jhi are the standard identification and nothing turns on it — the phenomenon (same content, variable morphological bearers across a paradigm) is the argument

2.231 unformalized

अन्वाख्यानस्मृतौ ये च प्रत्ययार्था निबन्धनम् ।
निर्दिष्टार्थ (ष्टास्ते) प्रकृत्यर्थाः स्मृत्यन्तर उदाहृताः ॥ २३१ ॥

anvākhyānasmṛtau ye ca pratyayārthā nibandhanam |
nirdiṣṭārtha (ṣṭāste) prakṛtyarthāḥ smṛtyantara udāhṛtāḥ || 231 ||

And those suffix-meanings which are the ground in the received analytic tradition — those very meanings, designated (there), are cited as base-meanings in another tradition.

The arbitrariness of 2.229-230 shown to be inter-traditional as well as intra-systematic. What one anvakhyana-smrti — one received analytic tradition — takes as pratyaya-artha, the suffix's meaning, and makes the nibandhana, the ground of its derivation, another smrti cites as prakrty-arthah, the base's meanings. Same forms, same senses, opposite bookkeeping — and both traditions derive correctly, since both get the words out. The verse completes the movement's demonstration by removing the last refuge: one might have hoped that although assignments shift within a system (2.229), at least the system as a whole is answerable to something. It is not. Two competent analytic traditions distribute meaning across the pieces in incompatible ways, and no fact adjudicates, because the pieces never had the meanings either tradition assigns — they had *offices*, and offices are conferred by the description that needs them (2.104's capacity-doctrine at the level of whole schools).
contested: the transcription is disturbed at nirdista-artha(s te) and we render the evident sense; the specific traditions contrasted are not named and are probably not recoverable

2.232 unformalized

प्रसिद्धेरुद्वमिकरीत्येव शास्त्रेऽभिधीयते ।
व्यवहाराय मन्यन्ते शास्त्रार्थप्रक्रिया यतः ॥ २३२ ॥

prasiddherudvamikarītyeva śāstre'bhidhīyate |
vyavahārāya manyante śāstrārthaprakriyā yataḥ || 232 ||

It is out of established currency that a thing is spoken of in the science in just such a manner; they hold that the derivational procedure of the science's meanings is for the sake of practice.

The grammar's own self-understanding, cited in its defence. Why does the sastra say things the way it does? Prasiddheh — from *established currency*: the science's formulations follow the usage already settled among the learned (1.27's sistas, 1.141's charter), not some independent access to how meanings really lie. And the purpose: vyavaharaya manyante sastra-artha-prakriya — the derivational procedure of the science's meanings is *for the sake of practice*, vyavahara: it exists to make the science workable, teachable, applicable. The whole movement's diagnosis is here given its charitable statement. The assignments are arbitrary as ontology and non-arbitrary as instruments: they are what practice, over generations of learned use, has found serviceable. Bhartrhari has never been debunking the grammar; he has been locating it (2.75's verdict: the science is the correct statics of the scaffold). And with the location complete, the kanda turns — as it does at every such juncture — from the instrument to what the instrument is for.

2.233 unformalized

शास्त्रेषु प्रक्रियाभेदैरविधैवोपवर्ण्यते ।
अनागमविकल्पा तु स्वयं विद्यापवर्तते ॥ २३३ ॥

śāstreṣu prakriyābhedairavidhaivopavarṇyate |
anāgamavikalpā tu svayaṃ vidyāpavartate || 233 ||

In the sciences, by differences of derivational procedure, only what is untrue is described; but knowledge itself, free of the alternatives of tradition, proceeds of its own accord.

The kanda's most startling sentence about its own discipline, and it must be read exactly. Sastresu prakriya-bhedaih avidya eva upavarnyate — in the sciences, by the differences of derivational procedure, *only avidya* is described: what the procedures depict is not how things are but the constructed order (avidya here in its technical sense: not stupidity but the domain of constructed difference — 1.1's Vrtti had Brahman with its two aspects, vidya and avidya, and the procedures live wholly in the second). The claim is not that grammar is false; it is that grammar's subject-matter *is* the constructed, and its faithful description of the constructed is a faithful description of avidya. Then: anagama-vikalpa tu svayam vidya pravartate — vidya, knowledge, *free of the alternatives* that tradition offers (the rival procedures of 2.231, the optioned principalities of 2.227-228), proceeds *of its own accord*, svayam. The procedures option because they are procedures; knowledge does not option, because it is not a procedure. This is 1.14's adhividya (grammar as the science above sciences) meeting 1.131's terminus (the word's perfecting *is* the supreme attainment): the discipline is a ladder through avidya's own territory, and what stands at the top is not a better procedure.
contested: reading avidya as the technical term (the domain of constructed difference) rather than as 'ignorance/falsity' flatly is required by the verse's contrast with vidya and by 1.1's usage; the flat reading makes Bhartrhari denounce his own science, which the next verse forbids

2.234 unformalized

अनिबद्धं निमित्तेषु निरुपाख्यं फलं यथा ।
तथा विद्याऽप्यनाक्येया शास्त्रेपायेव लक्ष्यते ॥ २३४ ॥

anibaddhaṃ nimitteṣu nirupākhyaṃ phalaṃ yathā |
tathā vidyā'pyanākyeyā śāstrepāyeva lakṣyate || 234 ||

As a fruit, not fastened to its occasioning causes, is indescribable — so knowledge too, untellable, is discerned as though by the science as its means.

The relation of science to knowledge, in one simile. The fruit is anibaddham nimittesu — *not fastened* to its occasioning causes: nothing in the seed, the soil, the rain is the mango; the causes are not the fruit and the fruit is not among them — nirupakhyam, indescribable in their terms. So vidya api anakhyeya — knowledge too is *untellable* (the anakhyeya of 2.144: pratibha's ineffability, now predicated of vidya itself) — and yet sastra-upaya iva laksyate: it is *discerned as though by the science as its means* — the ladder occasions what it cannot describe, as the causes occasion the fruit they do not contain. Every element of the kanda's architecture is in this: the flash is not composed of what produced it (2.143's anyaiva, upapaditam); the means is discharged in its attaining (2.38's upaya); the description belongs to avidya and the attainment does not (2.233). And the iva keeps it honest — *as though* by the science: even the instrumentality is a manner of speaking, since what ripens does so of its own accord (2.233's svayam; 2.148's effortless fermentation). Grammar is the horticulture of understanding: everything depends on it, and it grows nothing.

2.235 unformalized

अ(य)थाभ्यासं हि वागर्थे प्रतिपत्तिं समीहते ।
स्वभाव इव चानादिर्मिथ्यासो व्यवस्थितः ॥ २३५ ॥

a(ya)thābhyāsaṃ hi vāgarthe pratipattiṃ samīhate |
svabhāva iva cānādirmithyāso vyavasthitaḥ || 235 ||

It is according to practice that one seeks comprehension in the meaning of speech; and, like a nature, beginningless, this falsity stands established.

Why the ladder is unavoidable, stated without consolation. Yatha-abhyasam hi vag-arthe pratipattim samihate: it is *according to practice* — each according to his training (2.117's abhyasa, 2.152's six-fold gradient) — that one seeks comprehension in speech's meaning; no one approaches meaning except through the drill he happens to have had. And the second line names the condition: svabhava iva ca anadih mithya sah vyavasthitah — like a *nature*, *beginningless*, this falsity stands established. Mithya — the constructed, the false-taking, the projected division of the undivided (2.233's avidya) — is not an error we fell into and might climb out of by resolve: it is anadi, without beginning (1.28's vyavastha-nityata: what has no first term is not the sort of thing that was caused), and svabhava iva, as good as a nature. This is the kanda's darkest verse and its most consistent: if all cognition is word-interwoven (1.123) and all wording is sequence-contaminated (1.86), then the constructed condition has no outside from which we entered it. The upaya-doctrine's whole point is that one climbs *within* it. What follows (2.236-238) shows the same beginninglessness at work in the mind's most ordinary constructions.

2.236 unformalized

उत्प्रेक्षते सावयवं परमाणुमपण्डितः ।
तथावयविनं युक्तमन्यैरवयवैः पुनः ॥ २३६ ॥

utprekṣate sāvayavaṃ paramāṇumapaṇḍitaḥ |
tathāvayavinaṃ yuktamanyairavayavaiḥ punaḥ || 236 ||

The unlearned man imagines the atom as having parts, and again imagines the whole as joined with other parts.

The construction-compulsion exhibited in metaphysics itself. The apandita, the unlearned — and the target is not the ignorant but the *theorist* — utpreksate savayavam paramanum: *imagines* the atom as having parts. The paramanu is defined as partless; that is its office in the system; and the mind that entertains it cannot help giving it a near side and a far side, an inside. And symmetrically: avayavinam yuktam anyaih avayavaih punah — the whole, the avayavin, is imagined as joined with *further* parts, the analysis never terminating. The compulsion runs both directions and stops nowhere, which is precisely the regress of 2.28-29 — there deployed against linguistic atomism, here diagnosed as a habit of mind rather than a flaw in a doctrine. Utpreksate is the word: not 'concludes' but *imagines*, projects, over-sees. The mind that constructs divisions in the undivided word (1.85's productive error) is the same mind constructing parts in the partless atom — and it is not doing philosophy badly; it is doing what mind does (2.235's beginningless nature).

2.237-238 unformalized

घटादिदर्शनाल्लोकः परिच्छिन्नोऽवसीयते ।
समारम्भाच्च भावानामादिमद् ब्रह्म शाश्वतम् ॥ २३७ ॥
उपायाः शिक्षमाणानां बालानामपलापनाः ।
उसत्ये वर्त्मन स्थित्वा ततः सत्यं समीहते ॥ २३८ ॥

ghaṭādidarśanāllokaḥ paricchinno'vasīyate |
samārambhācca bhāvānāmādimad brahma śāśvatam || 237 ||
upāyāḥ śikṣamāṇānāṃ bālānāmapalāpanāḥ |
usatye vartmana sthitvā tataḥ satyaṃ samīhate || 238 ||

(237) From seeing pots and the like, the world is concluded to be delimited; and from the commencement of things, the eternal Brahman is concluded to have a beginning. (238) These are means for children being taught — deceptions; having stood on the untrue path, one thereafter seeks the true.

The compulsion at cosmic scale, and then the doctrine that redeems it. (237): ghata-adi-darsanat — from seeing pots — lokah paricchinnah avasiyate: the *world* is concluded to be delimited, bounded, a big thing among things; and samarambhac ca bhavanam — from the commencement of things (each pot has a potter, each thing a beginning) — adimad brahma sasvatam: the *eternal* Brahman is concluded to have a beginning. The mind generalizes from the pot to the cosmos and produces, inevitably, a bounded world and an originated absolute — both false, both irresistible, both the exact shape of 2.236's atom given parts. Then (238), the verse that makes the whole kanda's method explicit: upayah siksamananam balanam apalapanah — these are *means* for children being taught: *deceptions*, apalapana, things said to be unsaid. And the principle: asatye vartmani sthitva tatah satyam samihate — *having stood on the untrue path, one thereafter seeks the true*. This is the most famous formulation in Bhartrhari and one of the great sentences of Indian philosophy — the doctrine of upaya (2.38) stated with complete candour: the false is not the enemy of the true but its only approach; the pot-derived cosmos, the parted atom, the meaning-bearing suffix, the sentence made of words, are the untrue path, and there is no other road. It is the answer to 2.235's despair: the beginningless falsity is not a prison, because it is the way.

2.239 unformalized

अन्यथा प्रतिपद्यार्थं पदग्रहणपूर्वकम् ।
पुनर्वाक्ये तमेवार्थमन्यथा प्रतिपद्यते ॥ २३९ ॥

anyathā pratipadyārthaṃ padagrahaṇapūrvakam |
punarvākye tamevārthamanyathā pratipadyate || 239 ||

Having comprehended the meaning otherwise, by way of first grasping the words, one comprehends that very meaning otherwise again in the sentence.

The untrue path walked, in the ordinary act of understanding. Anyatha pratipadya artham pada-grahana-purvakam: having comprehended the meaning *otherwise* — wrongly, provisionally — by way of *first grasping the words*: the hearer takes the words, gets a construal, and it is not yet the meaning. Punar vakye tam eva artham anyatha pratipadyate: and *again*, in the sentence, one comprehends *that very meaning otherwise* — the construal is superseded by the sentence's own; the same content is taken again, differently, rightly. Note tam eva artham: it was the same meaning throughout — not a wrong meaning replaced by a right one but one meaning taken twice, first through the ladder and then whole. This is 2.238's asatya vartman in a single act of hearing, and it is 1.89-90's provincial (post, then man, then friend) at sentence-scale. The word-grasping stage is not a mistake to be avoided by better technique; it is the deception that teaches, and the flash (2.143) is what stands at the end of it. The kanda's semantics and its soteriology are here revealed as the same doctrine: understanding a sentence *is* the passage from the untrue to the true, performed a hundred times an hour.

2.240 · Negation and the Withdrawal of Meaning unformalized

उपात्ता बहवोऽप्यर्था येष्वन्ते प्रतिषेधनम् ।
क्रियते ते निवर्तन्ते तस्मात् तांस्तत्र नाश्रयेत् ॥ २४० ॥

upāttā bahavo'pyarthā yeṣvante pratiṣedhanam |
kriyate te nivartante tasmāt tāṃstatra nāśrayet || 240 ||

The many meanings that are taken up, upon which at the end a negation is made — those withdraw; therefore one should not rely on them there.

The negation-problem opened, and it is the sharpest test of the whole apparatus. In 'there is no tree here', many meanings are upatta, taken up — tree, here, existence — and then ante pratisedhanam kriyate, at the end a negation is made. What happens to the taken-up meanings? Te nivartante: they *withdraw* — the tree that was entertained is precisely what the sentence says is not. Tasmat tan tatra na asrayet: therefore one *should not rely on them* there. The problem this poses is severe for any deposit-semantics: the word 'tree' deposited its meaning, the negation cancels it, and the sentence's meaning is composed of... what? The cancelled deposit cannot be a constituent (the sentence denies it); its absence cannot be a constituent (absences deposit nothing); and yet the sentence means something perfectly determinate. This is 2.15's homeless-general trilemma at its most acute, and the next verses press it until the deposit-picture breaks. The holist's own answer is available from 2.143: the sentence occasions one flash, whose content is an absence-of-tree-here, and nothing was ever deposited to be cancelled.

2.241 unformalized

वृक्षो नास्तीति वाक्यं च विशिष्टाभावलक्षणम् ।
नार्थेन बुद्धौ सम्बन्धो निवृत्तेरवतिष्ठते ॥ २४१ ॥

vṛkṣo nāstīti vākyaṃ ca viśiṣṭābhāvalakṣaṇam |
nārthena buddhau sambandho nivṛtteravatiṣṭhate || 241 ||

And the sentence 'there is no tree' has the character of a qualified absence; no connection with the object stands in the cognition, on account of the withdrawal.

The negated sentence's content specified, and its problem sharpened. 'Vrkso nasti' is visista-abhava-laksanam: its character is a *qualified absence* — not absence as such (a bare nothing) but a determinate one: the absence *of a tree*, *here*, *now*; absence with a shape. Yet: na arthena buddhau sambandho nivrtter avatisthate — no connection with the object stands in the cognition, *because of the withdrawal*: the tree that would give the absence its shape is precisely what has been withdrawn, so the cognition has no object-connection to be qualified by. The dilemma is exact. The content is qualified (so something must be doing the qualifying); the qualifier has withdrawn (so nothing is). Every semantics that builds meanings out of object-connections founders here — which is why negation became, and remained, the great battleground of Indian semantics: the Buddhist apoha-theorists would make exclusion the *whole* of meaning, the Naiyayikas would posit absence as a real category, and the Mimamsakas would multiply their negation-types. Bhartrhari's contribution is to show that the problem is unanswerable *in the deposit-idiom* — 2.242-245 exhaust the options — which leaves the flash.

2.242 unformalized

विच्छेदप्रतिपत्तौ च यद्यस्तीत्यवधार्यते ।
अशब्दवाच्या सा बुद्धिर्निवर्तेत स्थिता कथम् ॥ २४२ ॥

vicchedapratipattau ca yadyastītyavadhāryate |
aśabdavācyā sā buddhirnivarteta sthitā katham || 242 ||

And in the comprehension of the delimitation, if 'it exists' is determined — that cognition, being not word-expressible, having stood, how would it withdraw?

The first horn. Suppose that in comprehending the sentence one *does* first determine asti, 'it exists' — the tree entertained as there, before the negation lands (the deposit-theorist's only available story: the members deposit, the negation then operates). Then: asabda-vacya sa buddhih nivarteta sthita katham — that cognition, *being not word-expressible* (the flash of it having occurred; and what occurred is not itself a word), *having stood* — how would it withdraw? Once a cognition has arisen it has arisen; cognitions are events, and events do not un-happen. The negation cannot reach backward and unmake the taking-up; at most it can be followed by a further cognition — but then the sentence's meaning is a *sequence of two cognitions*, one affirming and one denying, and no hearer of 'there is no tree' has any such experience: one understands an absence, at a stroke, and never affirms a tree at any moment. The horn is fatal to the sequential-processing picture, and it is fatal on phenomenological grounds — the kanda's standing court of appeal (2.34's samprataya-pramanatva).

2.243 unformalized

अथ यज्ज्ञानमुत्पन्नं तन्मिथ्येति नञा कृतम् ।
नञो व्यापारभेदेऽस्मिन्नभावावगतिः कथम् ॥ २४३ ॥

atha yajjñānamutpannaṃ tanmithyeti nañā kṛtam |
naño vyāpārabhede'sminnabhāvāvagatiḥ katham || 243 ||

Or if the cognition that has arisen is made false by the negative particle — then, in this difference of the particle's operation, how is the apprehension of absence?

The second horn. Suppose instead the nan does not withdraw the cognition but *falsifies* it: 'that cognition which arose — it is false'. Then the particle's operation has changed character: it now operates on the *cognition* rather than on the object, a meta-linguistic or meta-cognitive negation. And the objection: nano vyapara-bhede 'smin abhava-avagatih katham — in *this* difference of the particle's operation, how is the *apprehension of absence*? Falsifying a cognition is not the same as apprehending an absence: 'my tree-thought was mistaken' is a different content from 'there is no tree here' — the first is about me, the second about the spot. The hearer of 'vrkso nasti' learns something about the place, not about his own cognitive history. So the falsification-reading buys coherence at the cost of the content: it explains how the negation could operate and leaves the sentence meaning something nobody means by it. Between 2.242 and 2.243 the deposit-theorist has been given both available construals of what the nan does, and both fail — one to phenomenology, the other to content.

2.244 unformalized

निराधारप्रवृत्तौ च प्राक्प्रवृत्तिर्नञो भवेत् ।
अथाधारः स एवास्य नियमार्था श्रुतिर्भवेत् ॥ २४४ ॥

nirādhārapravṛttau ca prākpravṛttirnaño bhavet |
athādhāraḥ sa evāsya niyamārthā śrutirbhavet || 244 ||

And if its operation is without a substratum, the particle's operation would be prior; but if that very thing is its substratum, then the explicit text would be for the sake of restriction.

The third and fourth horns, closing the exhaustion. Suppose the negation operates niradhara — *without a substratum*, not on any entertained content: then prak pravrttir nano bhavet — the particle's operation would be *prior*: it would have to be doing its work before anything was there to negate, which is not an operation but a mystery (and it would make 'na' meaningful in isolation, which 2.194 denied). Suppose instead sa eva asya adharah — *that very thing* (the entertained tree-content) is its substratum: then niyama-artha srutir bhavet — the explicit text (the word 'tree') would be *for the sake of restriction*, its office being to specify *which* absence, on the model of the vriihi of 2.64-66: substance obtained, rice restricting. This last is the most promising of the four and the verse does not reject it — it names it. The negation's content would then be an absence-in-general specified by the word into a tree-absence, exactly the layered structure the ritual science already runs on. Note that this construal has quietly abandoned deposits: 'tree' no longer contributes a tree; it *delimits* an absence (2.204's bhedaka-office, the differentiator). The horn that survives is the one that has stopped being compositional.

2.245 unformalized

नियमद्योतनार्था वा अनुवादोऽथवा भवेत् ।
कश्चिदेवार्थवांस्तत्र शब्दः शेषास्त्वनर्थकाः ॥ २४५ ॥

niyamadyotanārthā vā anuvādo'thavā bhavet |
kaścidevārthavāṃstatra śabdaḥ śeṣāstvanarthakāḥ || 245 ||

Or it might be for the sake of manifesting the restriction; or it might be a mere restatement; some one word there is meaningful, and the rest are meaningless.

The residues, canvassed to the end. Or the word is niyama-dyotana-artha — for *manifesting* the restriction rather than expressing it (the dyotaka-account of 2.188 applied to negation: 'tree' lights up which absence is meant); or it is anuvada, a *mere restatement* (2.115's idle repetition: the absence already determinate, the word merely echoing it). And then the summary that the whole movement has been driving toward: kascid eva arthavan tatra sabdah sesas tv anarthakah — *some one word there is meaningful, and the rest are meaningless*. That is what deposit-semantics is reduced to in the negated sentence: after every construal has been tried, the analyst must nominate one word as the meaning-bearer and write the others off — an arbitrary nomination, since 2.240-244 have shown no principled ground for choosing. The verse is the reductio's last line. In a sentence everyone understands instantly and identically, the compositional analysis can locate meaning nowhere without declaring most of the sentence null. The alternative has been on the table since 2.143: one flash, whose content is a qualified absence, occasioned by the whole utterance including its 'na' — and no member of which ever meant anything by itself.

2.246 unformalized

विरुद्धं चाभिसम्बन्धमुदहार्यादिभिः कृतम् ।
वाक्ये समाप्ते वाक्यार्थमन्यथा प्रतिपद्यते ॥ २४६ ॥

viruddhaṃ cābhisambandhamudahāryādibhiḥ kṛtam |
vākye samāpte vākyārthamanyathā pratipadyate || 246 ||

And a contradictory connection, made by 'draw water' and the like — when the sentence is complete, one comprehends the sentence-meaning otherwise.

A second class of case, structurally identical to negation and more homely. Commands like udahara — 'draw (water)!' — and their kin generate, mid-utterance, a viruddha abhisambandha, a *contradictory connection*: the construal one is building conflicts with itself, or with what follows (the classic instances are sentences whose early words point one way and whose completion redirects: 'draw water — from the well, not the river', or idioms whose literal construal fights their sense). And the resolution: vakye samapte vakyartham anyatha pratipadyate — *when the sentence is complete*, one comprehends the sentence-meaning *otherwise*. The mid-course construal is not corrected by patching; the completed sentence delivers a different meaning entire, and the contradiction that seemed to be building simply never arises in the delivered content. This is 2.239's structure (comprehend otherwise via words, then otherwise again in the sentence) and it explains why contradictions in the analysis are not contradictions in the meaning: the analysis contradicts itself because it is assembling; the meaning cannot contradict itself because it was never assembled. Every reader has felt this: the garden-path sentence that resolves at its last word, with no felt repair.
contested: udahara-adibhih is obscure — we take it as illustrative imperatives generating mid-course contradictory construals; the specific examples the tradition had in mind are not recoverable from the karika

2.247 unformalized

स्तुतिनिन्दाप्रधानेषु वाक्येष्वर्थो न तादृशः ।
पदानां प्रविभागेन यादृशः परिकल्प्यते ॥ २४७ ॥

stutinindāpradhāneṣu vākyeṣvartho na tādṛśaḥ |
padānāṃ pravibhāgena yādṛśaḥ parikalpyate || 247 ||

In sentences whose principal purport is praise or blame, the meaning is not such as is constructed by the division of the words.

The third class, and the one that seals the movement: figurative and evaluative speech. Stuti-ninda-pradhanesu vakyesu — in sentences whose *principal purport* is praise or blame (the arthavada-passages of 2.8, the eulogies and censures that fill the Veda and ordinary talk alike) — artho na tadrsah yadrsah padanam pravibhagena parikalpyate: the meaning is *not such as* is constructed by the division of the words. 'You are a lion' divides into a false predication; 'his generosity is famous' may be scorn. The word-division yields one thing, the sentence means another, and no reader has ever been confused. Note the term pradhana: the praising or blaming is the sentence's *principal* purport — which by 2.113 is to say its prayojana, its point, and by 2.113 the sentence's meaning *is* its point. So the verse is not reporting an exception; it is exhibiting the rule. Every sentence means its point, and in the praise-sentence the point simply diverges visibly from what the divided words would compose. The figurative case is to sentence-meaning what the opaque compound (2.36) was to word-meaning: not an anomaly but the ordinary structure with its seams showing.

2.248 unformalized

अथासंसृष्ट एवार्थः पदेषु समवस्थितः ।
वाक्यार्थस्याभ्युपायोऽसावेकस्य प्रतिपादने ॥ २४८ ॥

athāsaṃsṛṣṭa evārthaḥ padeṣu samavasthitaḥ |
vākyārthasyābhyupāyo'sāvekasya pratipādane || 248 ||

Now, the meaning that stands in the words uncommingled — that is a means toward the conveying of the one sentence-meaning.

The settlement, offered with the word 'now' as one offers a concession that costs nothing. Grant it: asamsrsta eva arthah padesu samavasthitah — the meaning *stands in the words uncommingled*, each word with its own, unmixed, just as the analyst says. Bhartrhari does not fight the description here; he assigns it its office: vakyarthasya abhyupayo 'sau ekasya pratipadane — *that is a means* toward the conveying of the *one* sentence-meaning. Every load-bearing word is in place. Abhyupaya: a means, the upaya of 2.38 and 2.238 — taken up to be discarded, the untrue path. Ekasya: the sentence-meaning is *one*, and the concession does not touch it. Pratipadane: the words are means *toward the conveying*, not parts of the conveyed. So the uncommingled word-meanings are granted their full reality as instruments and denied any share in the content they deliver — which has been the kanda's position since 2.10 and is now stated as a settlement rather than a polemic. The next verse explains why the concession is safe.

2.249 unformalized

पूर्वं पदेष्वसंसृष्टो यः क्रमादुपचीयते ।
छिन्नग्रथितकल्पत्वान्न विशिष्टतरं विदुः ॥ २४९ ॥

pūrvaṃ padeṣvasaṃsṛṣṭo yaḥ kramādupacīyate |
chinnagrathitakalpatvānna viśiṣṭataraṃ viduḥ || 249 ||

That which, uncommingled in the words at first, is accumulated in sequence — because it is of the nature of the cut-and-strung, they do not know it as anything more distinguished.

Why the means cannot be mistaken for the end — the movement's most precise image. Take the uncommingled word-meanings (2.248) and let them be upaciyate kramat — *accumulated in sequence*, the deposit-theorist's whole procedure. What do you get? Chinna-grathita-kalpatvat — because the result is *of the nature of the cut-and-strung*: pieces severed and then threaded, a necklace of fragments, the join always a knot and never a growth. Na visistataram viduh: they do not know it as anything *more distinguished* — the strung sequence is not the sentence-meaning; it is a string. The image is perfect and devastating: the analysis cuts (apoddhara, 2.10) and then ties (samsarga, 2.42), and the tied-together cuts are what the theory offers in place of a meaning that was never cut. Against it stands the flash: anyaiva, altogether other (2.143), whole because it was never severed. This is the kanda's last word on composition, and it is an aesthetic judgement as much as a logical one — the compositional meaning has the look of repair-work, and comprehension does not feel like a mended thing.
contested: chinna-grathita ('cut-and-strung') is our rendering of the compound and the image is secure; whether kalpatva means 'nature of' or 'mere semblance of' shades the verdict — on the latter, even the strung result is only apparently a meaning

2.250 unformalized

एकमाहुरनेकार्थ शब्दमन्ये परीक्षकाः ।
निमित्तभेदादेकस्य सार्वार्थ्यं तस्य भिद्यते ॥ २५० ॥

ekamāhuranekārtha śabdamanye parīkṣakāḥ |
nimittabhedādekasya sārvārthyaṃ tasya bhidyate || 250 ||

Other examiners say the word is one, with many meanings; by difference of the occasioning grounds, its universal-meaningfulness is differentiated.

A new position enters, and it is the great one — the doctrine toward which the whole work has been tending. Anye pariksakah — other *examiners*, and the term is respectful: those who have looked. Ekam ahuh anekartham sabdam: they say the word is *one*, with *many* meanings — not many words each with a meaning, but *one word* whose meanings are many. And the mechanism: nimitta-bhedad ekasya sarvarthyam tasya bhidyate — by difference of the *occasioning grounds*, the *universal-meaningfulness* of that one is differentiated. Sarvarthya: the condition of having all meanings. The one word is omni-significant, and the nimittas — contexts, employments, the occasions of use (2.137's fixed occasioning conditions, 2.148's landmark-grounds) — differentiate that plenitude into this meaning here, that meaning there. Read against 1.1, the doctrine's identity is unmistakable: the one sabda-tattva, undivided, appearing as the manifold through the differentiating conditions — vivarta, in the semantics of a single word. The kanda's argument has been running toward this since it dissolved the sentence into a flash: if meaning is not built from parts, and words are one, then what a word means is everything, and context is what makes it mean something. The verses following develop it; the position is the school's own.
contested: sarvarthya ('universal-meaningfulness', the state of having all meanings) is the verse's coinage and the crux — some read it weakly ('its applicability to all its meanings'), some strongly (the word as omni-significant, meanings carved by context); the parallel with 1.1's vivarta-structure supports the strong reading and we adopt it

2.251 · One Word, Many Meanings: Primary and Secondary unformalized

यौगपद्यमतिक्रम्य पर्याये व्यवतिष्ठते ।
अर्थप्रकरणाभ्यां वा योगाच्छब्दान्तरेण वा ॥ २५१ ॥

yaugapadyamatikramya paryāye vyavatiṣṭhate |
arthaprakaraṇābhyāṃ vā yogācchabdāntareṇa vā || 251 ||

Transcending simultaneity, it stands settled in succession — either by the object and the context, or by connection with another word.

The mechanism of 2.250's omni-significance, stated as a rule of operation. The one word, whose sarvarthya is a plenitude of meanings, does not deliver them yaugapadyam — *simultaneously*: it transcends simultaneity, atikramya, and paryaye vyavatisthate — stands settled in *succession*, one meaning at a time. What selects? Two channels, and they exhaust the field: artha-prakaranabhyam — by the *object* (what is actually there to be spoken of) and the *context* (prakarana, the standing situation of the discourse); or sabdantarena yogat — by *connection with another word*, the company the word keeps. Everything the movement needs is here. The word is not ambiguous in the sense of harbouring a list from which the hearer picks by guesswork; it is undifferentiated plenitude, and the occasion carves. And carving is not the word's act — the word does nothing but be omni-significant; the nimitta does the work (2.250's nimitta-bheda). This is the semantics of vivarta made operational: one, appearing as many, through conditions — and the appearing is successive because conditions succeed one another.

2.252 unformalized

यथा सास्नादिमान् पिण्डो गोशब्देनाभिधीयते ।
तथा स एव गोशब्दो बाहीकेऽपि व्यवस्थितः ॥ २५२ ॥

yathā sāsnādimān piṇḍo gośabdenābhidhīyate |
tathā sa eva gośabdo bāhīke'pi vyavasthitaḥ || 252 ||

As a mass possessing dewlap and the rest is expressed by the word 'gauh' — so that very same word 'gauh' stands settled also upon a Bahika.

The example that Indian semantics never stopped arguing about, introduced with deliberate flatness. 'Gauh' expresses the beast — sasnadiman pindah, the mass with dewlap and the rest (the standard definition-by-marks). And tatha sa eva go-sabdah bahike 'pi vyavasthitah: *that very same word* 'gauh' stands settled also upon a *Bahika* — the man from Bahika, called an ox: 'gaur vahikah', the tradition's stock case of figurative speech, roughly 'the Punjabi is an ox' (dull, stolid). Note what the verse refuses to say. It does not say the word is *transferred*, or *applied metaphorically*, or *used in a secondary sense*. It says sa eva — *that very same word* — vyavasthitah, *stands settled*: the same word, equally settled, on both. This is 2.250-251 taken with full seriousness: if the word is one and omni-significant, then its standing on the ox and its standing on the man are not two operations, one primary and one derived, but one plenitude carved by two occasions. The next verse says so outright, and in doing so dismantles the primary/secondary distinction that every rival school treats as foundational.

2.253 unformalized

सर्वशक्तेस्तु तस्यैव शब्दस्यानेकधर्मणः ।
प्रसिद्धिभेदाद् गौणत्वं मुख्यत्वं चोपवर्ण्यते ॥ २५३ ॥

sarvaśaktestu tasyaiva śabdasyānekadharmaṇaḥ |
prasiddhibhedād gauṇatvaṃ mukhyatvaṃ copavarṇyate || 253 ||

But of that very word, which is all-capable and has many properties, secondariness and primariness are described by reason of a difference in established currency.

The verse that relocates the entire gauna/mukhya distinction — and it is one of the most consequential in the kanda. Sarva-sakteh tu tasya eva sabdasya aneka-dharmanah: of that very word — *all-capable*, sarva-sakti, and possessed of many properties. Then: prasiddhi-bhedad gaunatvam mukhyatvam ca upavarnyate — secondariness and primariness are *described* by reason of a difference in *established currency*. Every word is load-bearing. Sarva-sakti restates 2.250's sarvarthya: the word can do anything. Upavarnyate — *described*, not discovered (the kanda's verb for postings: 2.10's apoddharah upavarnyate). And prasiddhi-bheda: the difference is one of *currency* — of which use is commoner, more familiar, more settled in the community (2.128's rudhi). So: 'gauh' is primary of the ox and secondary of the man not because the ox-meaning is the word's own and the man-meaning borrowed, but because one use is more current than the other. The distinction is *statistical and social*, not semantic. Every rival school — Mimamsa, Nyaya, the poeticians — builds on gauna/mukhya as a real division in how words mean; Bhartrhari has just made it a fact about usage frequency. If the word is one and omni-significant, there is no other place for the distinction to live.
contested: prasiddhi we render 'established currency' (familiarity of use); reading it as 'well-known-ness' or 'renown' shades toward the same deflation. The verse's force — that the gauna/mukhya distinction is currency-based, not capacity-based — is our reading and it is the one 2.250-252 require

2.254 unformalized

एको मन्त्रस्तथाध्यात्ममधिदैवमधिक्रतु ।
असङ्करेण सर्वार्थो भिन्नशक्तिर्व्यवस्थितः ॥ २५४ ॥

eko mantrastathādhyātmamadhidaivamadhikratu |
asaṅkareṇa sarvārtho bhinnaśaktirvyavasthitaḥ || 254 ||

One mantra, likewise, is settled — with respect to the self, with respect to the deity, with respect to the rite — as having all meanings without commingling, its capacities distinct.

The doctrine confirmed from the tradition's own most authoritative practice. A single mantra is read adhyatmam — with respect to the self (the psychological/spiritual construal), adhidaivam — with respect to the deity (the theological), adhikratu — with respect to the rite (the ritual): the three standard registers of Vedic exegesis, in which one text yields three complete and non-competing meanings. And the description: asankarena sarvarthah bhinna-saktir vyavasthitah — having *all* meanings, *without commingling*, its capacities *distinct*, standing settled. Every term matches 2.250-253: sarvartha (omni-significance), bhinna-sakti (capacities differentiated by the nimitta — here the register of reading), asankarena (no blurring: each construal complete and clean, none the metaphorical shadow of another). The exegetical tradition has always insisted that the adhyatma-reading is not a figurative extension of the adhiyajna-reading; both are the mantra's own. Bhartrhari's point is that the tradition is right, and that its rightness is general: what the mantra does conspicuously, every word does — one plenitude, many carvings, none primary.

2.255 unformalized

गोत्वानुषङ्गो बाहीके निमित्तात् कैश्चिदिष्यते ।
अर्थमात्रं विपर्थस्तं शब्दः स्वार्थे व्यवस्थितः ॥ २५५ ॥

gotvānuṣaṅgo bāhīke nimittāt kaiścidiṣyate |
arthamātraṃ viparthastaṃ śabdaḥ svārthe vyavasthitaḥ || 255 ||

By some, an adherence of cowness in the Bahika is held to follow from the occasioning ground; the mere object is transposed, while the word stands settled in its own meaning.

A rival account of the Bahika-case, reported for contrast. Kaiscid: some hold that go-tva-anusanga bahike — an *adherence of cowness* in the Bahika — follows nimittat, from the occasioning ground (the shared property: dullness, stolidity): the universal cowness is somehow carried over and made to inhere in the man. On this view: artha-matram viparyastam — the *mere object* is transposed, shifted, misassigned — sabdah svarthe vyavasthitah: while the *word* stands settled in its own meaning, unmoved. The word keeps its ox-meaning; what moves is the object it gets applied to. This is the transference-account (the poeticians' laksana in embryo, and the Naiyayika's too): meaning fixed, application stretched. Bhartrhari reports it because it is the natural alternative to his own, and because its very structure concedes his point in miniature — it must posit a *transposition*, an artha-matram viparyastam, some machinery of carrying-over, precisely because it began by fixing the word's meaning. Take 2.250's plenitude instead and no carrying is needed: the word was always able to mean the man; currency alone made the ox familiar.

2.256 unformalized

तथा स्वरूपं शब्दानां सर्वर्थेष्वनुषज्यते ।
अर्थमात्रं विपर्यस्तं स्वरूपे तु स्थितिः स्थिरा ॥ २५६ ॥

tathā svarūpaṃ śabdānāṃ sarvartheṣvanuṣajyate |
arthamātraṃ viparyastaṃ svarūpe tu sthitiḥ sthirā || 256 ||

So too the own-form of words adheres in all meanings; the mere object is transposed, but the standing in the own-form is firm.

The rival account (2.255) generalized — and, in generalizing, revealed. Tatha svarupam sabdanam sarvarthesu anusajyate: the *own-form* of words adheres in *all* meanings — whatever a word is applied to, its svarupa (1.50's self-presenting face, 1.59-68's grammatical object) goes along: the word 'gauh' is the same word in every use, and its identity is not touched by what it is applied to. Artha-matram viparyastam — the mere object is transposed (2.255 verbatim) — svarupe tu sthitih sthira: but the standing in the own-form is *firm*. Read as the rival's doctrine, this is the transference-account's foundation: word-identity is form-identity, applications wander, meanings shift, the word abides. But read against 2.250-253 the same sentence says something else entirely: what is invariant across every use is the word's *form*, and nothing else — not a primary meaning, not a core sense, not a fixed deposit. Which is precisely the omni-significance thesis: the word is one *as form*, and its meanings are as many as its occasions. The rival's own foundation, stated exactly, is the holist's conclusion.

2.257 unformalized

एकत्वं तु स्वरूपत्वाच्छब्दयोर्गौणमुख्ययोः ।
प्राहुरत्यन्तभेदेऽपि भेदमार्गानुदर्शिनः (नुसारिणः) ॥ २५७ ॥

ekatvaṃ tu svarūpatvācchabdayorgauṇamukhyayoḥ |
prāhuratyantabhede'pi bhedamārgānudarśinaḥ (nusāriṇaḥ) || 257 ||

But those who trace out the path of difference say that the oneness of the two words, secondary and primary, is on account of their own-form — even where there is utter difference.

The doxographic placement of 2.256, and it is precisely calibrated. Bheda-marga-anudarsinah — those who *trace out the path of difference*: the analysts, the theorists whose method is division (2.57's pada-darsins, whose primitives are the parts). Their doctrine: sabdayoh gauna-mukhyayoh ekatvam svarupatvat — the *oneness* of the two words, the secondary and the primary, is *on account of their own-form*; atyanta-bhede 'pi — even where there is *utter difference* (the meanings having nothing whatever in common: ox and man). So even the difference-tracers concede unity — but only formal unity: for them, 'gauh'-of-the-ox and 'gauh'-of-the-man are two words that happen to share a shape, held together by nothing but svarupa. Bhartrhari's own position agrees on the datum and reverses the direction: they are not two words sharing a form but *one word* whose plenitude two occasions have carved (2.250-253). The difference between the doctrines is exactly the difference between the two darsanas of 1.74 and 2.57 — parts first, or whole first — and here, as always, the verse states the rival fairly and lets the reader see that both accounts save the phenomena while only one avoids positing a transference (2.255).

2.258 unformalized

सामिधेन्यन्तरं चैवमावृत्तावनुषज्यते ।
मन्त्राश्च विनियोगेन लभन्ते भेदमूहवत् ॥ २५८ ॥

sāmidhenyantaraṃ caivamāvṛttāvanuṣajyate |
mantrāśca viniyogena labhante bhedamūhavat || 258 ||

And thus another samidheni-verse adheres in the repetition; and mantras obtain difference by their application, as in modification.

Ritual confirmation of the nimitta-doctrine, from the fire-kindling verses. The samidhenis are the verses recited while kindling the sacrificial fire; the rite requires a fixed number, and the number is made up by *repeating* verses. The question the exegetes faced: when a verse is repeated to make the count, is the repetition the *same* verse or *another*? The verse's answer: samidhenyantaram anusajyate avrttau — *another* samidheni adheres in the repetition: the repeated verse counts as a distinct one. Same words, same form, two verses — individuated by their ritual office. And the generalization: mantrah viniyogena labhante bhedam uhavat — mantras obtain *difference by their application*, as in uha (the modification of mantras when transposed to a derivative rite, 2.78). Vinyoga — application, deployment to a ritual station — is what individuates. This is 2.104's capacity-doctrine and 2.110's tantra-cases returning at the semantic level: identity of form settles nothing; office decides how many there are. And the ritual science, which must count exactly, counts by office.

2.259 unformalized

तान्याम्नायान्तराण्येव पठ्यते कश्चिदेव तु ।
अनर्थकानां पाठो वा शेषस्त्वन्यः प्रतीयते ॥ २५९ ॥

tānyāmnāyāntarāṇyeva paṭhyate kaścideva tu |
anarthakānāṃ pāṭho vā śeṣastvanyaḥ pratīyate || 259 ||

Those are just other recensional texts; or some one alone is recited; or there is recitation of meaningless things, and the remainder is understood as something else.

The alternatives canvassed for the repetition-problem, and their disarray is the point. (i) Tani amnayantarani eva: the repeated verses are just *other recensional texts* — different amnayas, so no repetition at all, merely distinct verses that happen to coincide. (ii) Pathyate kascid eva tu: only *some one* is recited — one verse, the repetitions being mere re-utterances of it. (iii) Anarthakanam patho va: or there is *recitation of meaningless things* — the repeated tokens carrying no meaning, present only to make the count; sesas tu anyah pratiyate — and the remainder is understood as *something else*. Three accounts of what happens when the same words are said twice in a rite, and they disagree about the most basic question: how many things are there? The disarray is not incompetence. It is what 2.250-258 predict: identity and number are conferred by office, and where offices are contested (is the second recitation a new station or the same one served twice?), the count is contested — with nothing in the words to settle it, because the words never carried the count.

2.260 unformalized

शब्दस्वरूपमर्थस्तु पाठोऽन्यैरुपवर्ण्यते ।
अत्यन्तभेदः सर्वेषां तत्सम्बन्धात्तु तद्वताम् ॥ २६० ॥

śabdasvarūpamarthastu pāṭho'nyairupavarṇyate |
atyantabhedaḥ sarveṣāṃ tatsambandhāttu tadvatām || 260 ||

By others, the own-form of the word is described as the meaning, and the recitation as (something) else; there is utter difference of all, but (identity) of the possessors, from connection with that.

A further account, and it draws the movement's threads to a point. Aparaih: by others — sabda-svarupam arthah — the *own-form* of the word is (its) meaning: the position 2.256 stated, now held as doctrine: what a word means, at bottom, is itself (2.127's abhijalpa-doctrine at its limit; 1.50's self-presentation made the whole story). Pathah anyaih upavarnyate: the *recitation* is described as something else — the utterance-event distinguished from the form recited. Then the summary: atyanta-bhedah sarvesam — there is *utter difference* of all of them (form, meaning, recitation, application: every item the analysis distinguishes stands utterly apart from every other) — tat-sambandhat tu tadvatam: but (identity/unity accrues) to *the possessors* from connection with that. The construal is difficult, and the difficulty is doctrinal: once analysis has cut form from meaning from recitation from office, nothing reassembles them except a sambandha — the connective whose emptiness 2.46 exposed. The verse is a last exhibit of the same pattern the kanda has traced from its first movement: cut, and you will need a knot; the knot will be a nothing; and the thing you cut was never cut.
contested: the syntax of the second line is genuinely obscure and our construal is one of several defensible; the transcription may also be at fault. What is clear is the contrast: utter difference among the analysed items, unity supplied only by connection

2.261 unformalized

अन्या संस्कारसावित्री कर्मण्यन्या प्रयुज्यते ।
अन्या जपप्रबन्धेषु सा त्वेकैव प्रतीयते ॥ २६१ ॥

anyā saṃskārasāvitrī karmaṇyanyā prayujyate |
anyā japaprabandheṣu sā tvekaiva pratīyate || 261 ||

One savitri is for the sacrament, another is employed in the rite, another in the sequences of muttered prayer — yet it is understood as one and the same.

The movement's most beautiful confirmation, and its subject is the most familiar verse in the tradition. The Savitri — the Gayatri — is deployed in three quite different offices: samskara-savitri, the one whose recitation makes the boy twice-born at initiation; the one employed karmani, in the rite; the one used japa-prabandhesu, in the sequences of muttered repetition. Three offices, and by 2.258's doctrine of vinyoga three verses. Anya... anya... anya: *another*, another, another — the verse says it three times. And then: sa tu eka eva pratiyate — *yet it is understood as one and the same*. Both halves are true and the verse asserts both without reconciling them. Individuated by office, one in itself: this is not a paradox to be resolved but the structure the whole kanda has been describing — the one appearing as many through conditions (2.250-251), the many being the one carved by nimittas, and comprehension holding both without strain, as every brahmin holds that the Savitri he mutters is the Savitri that made him. The verse is 1.2 (the one, resting on its differentiated powers, functioning as if divided) come home to the tradition's own daily practice.

2.262 unformalized

अर्थस्वरूपे शब्दानां स्वरूपाद्वृत्तिरिष्यते ।
वाक्यरूपस्य वाक्यार्थे वृत्तिरन्यानपेक्षया ॥ २६२ ॥

arthasvarūpe śabdānāṃ svarūpādvṛttiriṣyate |
vākyarūpasya vākyārthe vṛttiranyānapekṣayā || 262 ||

The operation of words upon the meaning's own form is held to be from their own form; the operation of the sentence-form upon the sentence-meaning is without dependence on anything else.

The self-sufficiency thesis, stated at both grains. Artha-svarupe sabdanam svarupad vrttih isyate: the operation of words upon the meaning's own form is *from their own form* — a word reaches its meaning by being what it is, not by way of some intermediary connection (2.46's inferred non-entity, dismissed). And the sentence: vakya-rupasya vakyarthe vrttih anya-anapeksaya — the sentence-form's operation upon the sentence-meaning is *without dependence on anything else*: not on its words, not on their meanings, not on a connective. Anya-anapeksaya is emphatic and it is the movement's terminus. The sentence does not reach its meaning *through* the words — this is why the flash is anyaiva (2.143), why the words are abhyupaya (2.248), why the strung sequence is not the meaning (2.249). Form and meaning stand to each other as the two faces of one thing (2.31), at word-grain and at sentence-grain alike, and nothing mediates because there are not two things to mediate between. The verse is the kanda's most compact statement of what it has been arguing since 2.7: the sentence means directly, and everything the analysis inserts between the sentence and its meaning is the analysis's own furniture.

2.263 unformalized

अनेकार्थत्वमेकस्य यैः शब्दसायनुगम्यते ।
सिद्ध्यसिद्धिकृता तेषां गौणमुख्यप्रकल्पना ॥ २६३ ॥

anekārthatvamekasya yaiḥ śabdasāyanugamyate |
siddhyasiddhikṛtā teṣāṃ gauṇamukhyaprakalpanā || 263 ||

Those by whom the many-meaningedness of a single word is admitted — for them the construction of secondary and primary is made by attainment and non-attainment.

The doxographic ledger for the gauna/mukhya distinction, opened with the position the kanda holds. For those who admit anekarthatvam ekasya — the many-meaningedness of a *single* word (2.250's examiners, and Bhartrhari with them) — the gauna-mukhya-prakalpana is siddhy-asiddhi-krta: made by *attainment and non-attainment*. That is: a use is primary where the meaning is attained — arrives, lands, is simply there for the hearer — and secondary where it is not attained directly but reached with effort, through the occasion's help. The distinction survives, but as a fact about *how the meaning arrives*, not about what the word owns: 2.253's currency-doctrine given its cognitive face. Note that this leaves the word entirely undivided: the same plenitude, arriving easily here (the ox: well-worn channel) and with labour there (the man: channel less worn). The verses following canvass the rivals' versions of the same distinction, and the contrast will be instructive — each of them locates in the *word* what this one locates in the *arrival*.

2.264 unformalized

अर्थप्रकरणापेक्षो यो वा शब्दान्तरैः सह ।
युक्तः प्रत्याययत्यर्थं तं गौणमपरे विदुः ॥ २६४ ॥

arthaprakaraṇāpekṣo yo vā śabdāntaraiḥ saha |
yuktaḥ pratyāyayatyarthaṃ taṃ gauṇamapare viduḥ || 264 ||

Or: that which, depending on the object and the context, or joined with other words, conveys the meaning — others know as secondary.

The rivals' first definition of the secondary, and it is a good one. Apare: others know as *gauna* that which conveys its meaning artha-prakarana-apeksah — *depending* on the object and the context — or sabdantaraih saha yuktah — *joined with other words*. Secondary use, on this account, is context-dependent use: the word that cannot deliver alone and needs the occasion's help. Note that the two channels named are exactly 2.251's — the same nimittas that, on Bhartrhari's account, carve *every* meaning out of the plenitude. So the rival's definition of the secondary is the holist's description of the ordinary: what this theorist marks as the deviant case (meaning determined by context and company) is, for the kanda, how words always work — the isolated word being the citation-artifact (2.17), not the norm. The definitions are diverging exactly where the theories diverge, and the divergence is visible in what each treats as needing explanation.

2.265 unformalized

शुद्धस्योच्चारणे स्वार्थः प्रसिद्धो यस्य गम्यते ।
स मुख्य इति विज्ञेयो रूपमात्रनिबन्धनः ॥ २६५ ॥

śuddhasyoccāraṇe svārthaḥ prasiddho yasya gamyate |
sa mukhya iti vijñeyo rūpamātranibandhanaḥ || 265 ||

That whose well-established own meaning is understood on the mere uttering of it, unmixed, is to be known as primary — grounded in the mere form.

The rivals' complementary definition of the primary. Suddhasya uccarane — on the uttering of the word *unmixed*, bare, without context or company — svarthah prasiddhah yasya gamyate: whose *well-established* own meaning is (nonetheless) understood — sa mukhya iti vijneyah, that is to be known as primary; rupa-matra-nibandhanah, grounded in the *mere form*. So the primary is the meaning the bare form delivers, and the test is citation: say the word alone and see what comes. It is a clean criterion and it is exactly the one the kanda has spent two hundred verses undermining. Prasiddhah is the tell — 'well-established', the same prasiddhi that 2.253 identified as *currency*: what the bare word delivers is not its own meaning but its commonest one, the channel most worn by use. The rival takes the citation-response for the word's nature; Bhartrhari takes it for a statistic (2.17: words look general in citation because citation is a context too, and a thin one). Both agree on what happens when you say 'gauh' to a Sanskritist. They disagree about whether that tells you anything about the word.

2.266 unformalized

यस्त्वन्यस्य प्रयोगेण यत्नादिव नियुज्यते ।
तमप्रसिद्धं मन्यन्ते गौणार्थाभिनिवेशिनम् ॥ २६६ ॥

yastvanyasya prayogeṇa yatnādiva niyujyate |
tamaprasiddhaṃ manyante gauṇārthābhiniveśinam || 266 ||

But that which is, as it were, forcibly yoked by the employment of another — they consider not-established, as settling upon a secondary meaning.

The rivals' account of the secondary completed, and its vocabulary is revealing. Yah tu anyasya prayogena yatnad iva niyujyate — that which is *forcibly*, as it were with effort, yoked by another's employment: the word conscripted, pressed into a service not its own by the speaker's design — tam aprasiddham manyante gauna-artha-abhinivesinam: they consider it *not-established*, settling upon a secondary meaning. Two words carry the account's whole picture. Yatnad iva: *as if by force* — the secondary use is violence done to the word, a wrenching. Abhinivesin: *settling upon*, with a connotation of clinging or insisting, the word made to squat on a meaning that is not its home. This is metaphor understood as coercion, and it is the standard view — the poeticians would build an entire theory on the word's 'incapacity' in its primary sense forcing a secondary. Bhartrhari's alternative needs no violence: the word could always mean the man (2.250's sarvarthya); the occasion carved that meaning rather than the worn one; nothing was forced, and the felt effort (2.263's asiddhi) is the effort of a channel not yet worn, not the strain of a word being bent.

2.267 unformalized

स्वार्थे प्रवर्तमानस्य यस्यार्थं योऽवलम्बते ।
निमित्तं तत्र मुख्यं स्याद् निमित्ती गौण इष्यते ॥ २६७ ॥

svārthe pravartamānasya yasyārthaṃ yo'valambate |
nimittaṃ tatra mukhyaṃ syād nimittī gauṇa iṣyate || 267 ||

Where one word, operating in its own meaning, has its meaning depended upon by another — there the ground would be primary, and the grounded is held secondary.

The rivals' most precise formulation, and it is a relational one. Consider two words: one svarthe pravartamanasya — *operating in its own meaning*; the other yasya artham yah avalambate — the one that *depends upon*, leans on, the first's meaning. Then: nimittam tatra mukhyam syat — the *ground*, the one leaned on, would be primary; nimitti gauna isyate — the *grounded*, the leaner, is held secondary. In 'the Bahika is an ox', the ox-meaning is the ground (it supplies the dullness); the man-meaning leans on it; so ox is primary, man secondary. The definition is elegant and it makes the distinction structural rather than statistical — which is precisely its vulnerability, and Bhartrhari has already placed the charge: the leaning is a fact about *this* use, in *this* company; reverse the currency (let 'gauh' be commonly said of stolid men and rarely of cattle) and the leaning reverses with it. Nothing in the word ordains which of its meanings is the ground. The rival has described the phenomenology of a particular figurative act with real precision and mistaken it for the anatomy of the word.

2.268 unformalized

पुरारादिति भिन्नेऽर्थे यौ वर्तेते विरोधिनि ।
अर्थप्रकरणापेक्षं तयोरप्यवधारणम् ॥ २६८ ॥

purārāditi bhinne'rthe yau vartete virodhini |
arthaprakaraṇāpekṣaṃ tayorapyavadhāraṇam || 268 ||

The two which operate in different, contradictory meanings — as with 'pura' and 'arat' — their determination too depends on the object and the context.

The auto-antonyms, and they finish the movement's case. Pura means both 'formerly' and 'hereafter'; arat means both 'near' and 'far'. Bhinne 'rthe virodhini: the two meanings are not merely different but *contradictory* — each word carries its own opposite, and no shared property, no leaning, no ground-and-grounded relation (2.267) can connect them. So which is primary? The question has no answer, and the verse says so: tayoh api avadharanam artha-prakarana-apeksam — *their* determination too depends on the object and the context. Nothing but the occasion decides, and the occasion decides completely. For the rival accounts this is a hard case demanding special treatment; for 2.250-251 it is simply the ordinary structure with the primary/secondary illusion stripped away — one word, a plenitude including contradictories, carved by nimitta, and no meaning more its own than another. The auto-antonym is to the gauna/mukhya distinction what the freak compound (2.216) was to composition: not an anomaly but the general case, visible because nothing obscures it.
contested: pura and arat are our reading of the transcription's purarad iti; both are standard examples of words bearing opposite temporal/spatial senses, and the argument requires only that the two senses be contradictory

2.269 · The Sentence as the Unit of Connection unformalized

वाक्यस्यार्थात् पदार्थानामपोद्धारे प्रकल्पिते ।
शब्दान्तरेण सम्बन्धः क (त) स्यैकस्योपपद्यते ॥ २६९ ॥

vākyasyārthāt padārthānāmapoddhāre prakalpite |
śabdāntareṇa sambandhaḥ ka (ta) syaikasyopapadyate || 269 ||

When the abstraction of word-meanings from the sentence's meaning has been constructed, how is connection with another word intelligible for any one of them?

A question that closes a circle opened at 2.10. Grant the apoddhara: vakyasya arthat padarthanam apoddhare prakalpite — the abstraction of word-meanings *from the sentence's meaning* having been constructed (the kanda's own doctrine, stated as accomplished). Then: sabdantarena sambandhah kasya ekasya upapadyate — how is *connection with another word* intelligible for any *one* of them? The force is this: the analyst's padarthas are extracts from one meaning; extracts do not connect — they were never apart. To speak of one word-meaning connecting with another is to speak as if they had been independently there and then joined, which is exactly what the abstraction-doctrine denies. So the whole vocabulary of syntactic connection (2.42's samsarga, 2.47's expectancy, every account of how words combine) is revealed as belonging to the scaffold and not to the meaning: within the analysis, the extracts must be re-tied, because the analysis cut them; in the meaning, nothing is tied because nothing was cut (2.249's chinna-grathita). The verse is the kanda's final disposal of the connection-problem — not solved but dissolved, by noticing that only the analysis ever had it.
contested: the transcription wavers between kasya and tasya; the interrogative reading (our choice) makes the verse the reductio it evidently is

2.270 unformalized

यच्चाप्येकं पदं दृष्टं चरितास्तिक्रियं क्कचित् ।
तद्वाक्यान्तरमेवाहुर्न तदन्येन युज्यते ॥ २७० ॥

yaccāpyekaṃ padaṃ dṛṣṭaṃ caritāstikriyaṃ kkacit |
tadvākyāntaramevāhurna tadanyena yujyate || 270 ||

And that single word which is seen somewhere, its verb-of-existence having been discharged — they call that another sentence entire; it does not connect with anything else.

The one-word sentence, and it seals the movement. Ekam padam drstam kvacit carita-asti-kriyam: a *single word* seen somewhere with its asti-kriya — its verb of existence — *discharged*, already performed, absorbed (the elliptical 'Fire!', 'Water!', the injunctions and exclamations in which one word carries a whole predication). What is it? Tad vakyantaram eva ahuh: they call it *another sentence entire* — not a word but a sentence, complete, whose verb is done. Na tad anyena yujyate: and *it does not connect with anything else* — being complete, it has no expectancy left (2.4's nirakanksata, the mark of the sentence), so there is nothing for connection to do. The verse is the perfect complement to 2.269. There, the analyst's extracts cannot connect because they were never separate; here, the complete utterance does not connect because it needs nothing. Between the two, the entire connective apparatus has been shown to have no domain: not among the abstractions (which were one), not among the wholes (which are self-sufficient). And the one-word sentence makes the kanda's title-doctrine visible to the naked eye: sentencehood is not a matter of how many words there are. It is a matter of whether the expectancy has closed — which is to say, whether a flash has been occasioned.

2.271 unformalized

यच्च कोऽयमिति प्रश्ने गौरश्व इति चोच्यते ।
प्रश्न एव क्रिया तत्र प्रक्रान्ता दर्शनादिका ॥ २७१ ॥

yacca ko'yamiti praśne gauraśva iti cocyate |
praśna eva kriyā tatra prakrāntā darśanādikā || 271 ||

And when, at the question 'who is this?', 'a cow' or 'a horse' is uttered — there the question itself is the action commenced: seeing and the like.

The one-word sentence (2.270) explained by its context. Asked ko 'yam, 'who is this?', one answers gauh — a single word, and unmistakably a complete sentence. Where is its verb? Prasna eva kriya tatra prakranta darsanadika: *the question itself* is the action commenced — the seeing and the like. The interrogative has already deployed the verb; the answer's single word steps into a predication that the question built and left open. So the one-word utterance is not a word doing a sentence's work but a *sentence whose remainder was supplied by the discourse* — 2.76's larger unity in miniature: the unit is the exchange, and 'gauh' is complete within it because expectancy closed there. This is also the cleanest possible demonstration of 2.113's thesis: sentence-meaning is prayojana, purpose — and the purpose here was set by the asking. What the answer means is fixed by what was wanted.

2.272 unformalized

नैवाधिकत्वं धर्माणां न्यूनता वा प्रयोजिका ।
आधिक्यमपि मन्यन्ते प्रसिद्धेर्न्यूनतां क्वचित् ॥ २७२ ॥

naivādhikatvaṃ dharmāṇāṃ nyūnatā vā prayojikā |
ādhikyamapi manyante prasiddhernyūnatāṃ kvacit || 272 ||

Neither excess of properties nor deficiency is what occasions (secondariness); some hold excess also, and sometimes deficiency, on account of established currency.

A demolition of the standard criteria for figurative use, and it is brisk. When 'gauh' is said of the Bahika, the man has *fewer* of the cow's properties (no dewlap, no horns) — so deficiency, nyunata, looks like the mark of secondariness. But some hold adhikatvam, *excess*, to be the mark instead (the case where the figure attributes more than the primary bears: 'he is a lion' claiming a courage beyond any lion's). Bhartrhari's verdict: naiva adhikatvam dharmanam nyunata va prayojika — *neither* is what occasions it; and the reason for the disagreement is given in the last word: prasiddheh — from *established currency*. The theorists disagree because they are reading off usage, and usage varies. The distinction they are trying to ground in a property-count has no property-count to stand on; it stands, as 2.253 said, on how worn the channel is. Every attempt to make gauna/mukhya a structural matter runs into this: the structures point both ways, and only currency is constant.

2.273 unformalized

जातिशब्दोऽन्तरेणापि जातिं यत्र प्रयुज्यते ।
सम्बन्धिसदृशाद्धर्मात् तं गौणमपरे विदुः ॥ २७३ ॥

jātiśabdo'ntareṇāpi jātiṃ yatra prayujyate |
sambandhisadṛśāddharmāt taṃ gauṇamapare viduḥ || 273 ||

Where a universal-word is employed even without the universal, on account of a property resembling that of the relatum — others know that as secondary.

Another rival definition, and a careful one. A jati-sabda — 'gauh', a word whose office is a universal — is employed antarena api jatim, *even without the universal* being present: the Bahika has no cowness. What licenses it? Sambandhi-sadrsad dharmat — a property *resembling* that of the relatum: the man's stolidity resembling the ox's. So the account is resemblance-based: the universal absent, a similar property present, the word extended on the strength of the resemblance. Apare vidus tam gaunam: others know *that* as secondary. It is the definition most readers would give unprompted, and Bhartrhari's objection is not stated here because it has already been made: 2.90-94 showed that resemblance is respect-wise apprehension of undivided wholes, not shared components — so 'a resembling property' is not a thing two objects both have but a way one cognition takes two. Build metaphor on resemblance and you have built it on a cognitive act, which is where the kanda locates meaning anyway (2.143). The rival is closer to the holist than he knows.

2.274 unformalized

विपर्यासादिवार्थस्य यत्रार्थान्तरतामिव ।
मन्यन्ते स गवादिस्तु गौण इत्युच्यते क्वचित् ॥ २७४ ॥

viparyāsādivārthasya yatrārthāntaratāmiva |
manyante sa gavādistu gauṇa ityucyate kvacit || 274 ||

Where, as if by a transposition of the meaning, one thinks there is as it were an otherness of meaning — that use of 'cow' and the like is somewhere called secondary.

A third rival definition, and its hedges are its confession. Viparyasad iva arthasya — as if by a *transposition* of the meaning (2.255's artha-matram viparyastam) — yatra arthantarataam iva manyante: where one *thinks* there is *as it were* an otherness of meaning. Count the qualifications: iva, iva, manyante, kvacit — 'as if', 'as it were', 'one thinks', 'somewhere'. The definition cannot state its own content without four hedges, because what it is trying to name is not a fact about the word but an impression in the hearer: the sense that something has been shifted. Bhartrhari records the hedges rather than smoothing them, and they make his case for him. Secondariness, on every rival account canvassed (2.264-267, 2.273-274), turns out to be describable only in terms of how the use *strikes* someone — forced (2.266), leaning (2.267), resembling (2.273), transposed (here) — which is to say: in terms of the arrival, not the word. Which is 2.263's position, held by the very theorists who deny it.

2.275-277 unformalized

नियताः साधनत्वेन रूपशक्तिसमन्विताः ।
यथा कर्मसु गम्यन्ते सीरासिमुसलादयः ॥ २७५ ॥
क्रियान्तरेण चैतेषां भवन्ति नहि शक्त्यः ।
रूपादेव तु तादर्थ्यं नियमेन प्रतीयते ॥ २७६ ॥
तथैव रूपशक्तिभ्यामुत्पच्या समवस्थितः ।
शब्दो नियततादर्थ्यः शक्त्याऽन्यत्र प्रयुज्यते ॥ २७७ ॥

niyatāḥ sādhanatvena rūpaśaktisamanvitāḥ |
yathā karmasu gamyante sīrāsimusalādayaḥ || 275 ||
kriyāntareṇa caiteṣāṃ bhavanti nahi śaktyaḥ |
rūpādeva tu tādarthyaṃ niyamena pratīyate || 276 ||
tathaiva rūpaśaktibhyāmutpacyā samavasthitaḥ |
śabdo niyatatādarthyaḥ śaktyā'nyatra prayujyate || 277 ||

(275) As plough, sword, pestle and the like are understood, in their operations, as fixed in the character of instruments, endowed with form and capacity — (276) and their capacities do not arise through a different action; from their form alone their being-for-that-purpose is understood as a rule — (277) just so, the word, standing established by its form and capacity from its production, having a fixed purpose, is (nonetheless) employed elsewhere by capacity.

The instrument-simile, and it is the movement's most careful piece of work. Consider the plough, the sword, the pestle: each is niyata sadhanatvena — *fixed* in its instrumental character, rupa-sakti-samanvita, endowed with form and capacity. And the two facts about them. (276): their capacities do not arise kriyantarena, through some *other* action — the sword's cutting-power is not conferred on it by being used to cut; rupad eva tadarthyam niyamena pratiyate: from its *form alone* its being-for-that-purpose is understood, as a rule. Look at a sword and you know what it is for. (277) applies it: tathaiva rupa-saktibhyam utpattya samavasthitah sabdah niyata-tadarthyah — just so the word, established by form and capacity from its very production, has a *fixed purpose* — saktya anyatra prayujyate: and yet *by capacity* it is employed *elsewhere*. That last clause is where the simile turns. A sword has a fixed purpose and can still be used to prop a door; the use is not the sword's purpose and is entirely within the sword's power. So with the word: fixed purpose (the current meaning, prasiddhi), and a capacity that outruns it (2.250's sarvarthya). The 'secondary' use is neither a violation (2.266's forcing) nor a different word — it is the instrument's own power, deployed off its accustomed office. The simile grants the rivals everything they wanted about fixity and takes back everything they built on it.
contested: 277's utpattya ('from its production') is difficult — we take it as 'from the moment of its arising', i.e. the word's form-and-capacity are constitutive rather than acquired; some read it with niyata-tadarthyah, 'having a purpose fixed by its derivation', which would make the point etymological and sit oddly with 2.170-175

2.278 unformalized

श्रुतिमात्रेण यत्रास्य ता(या) दर्थ्यमवसीयते ।
मुख्यं तमर्थं मन्यन्ते गौणं यत्रोपपादितम् ॥ २७८ ॥

śrutimātreṇa yatrāsya tā(yā) darthyamavasīyate |
mukhyaṃ tamarthaṃ manyante gauṇaṃ yatropapāditam || 278 ||

Where its being-for-that-purpose is determined by the mere hearing, they consider that meaning primary; secondary, where it is (only) brought about.

The criterion restated in the instrument-simile's terms, and now it is exact. Sruti-matrena — by the *mere hearing*, the word alone, nothing else deployed — yatra asya tadarthyam avasiyate: where its being-for-that-purpose is determined: mukhyam tam artham manyante, that meaning is considered primary. And gaunam yatra upapaditam: secondary where it is *brought about* — where the meaning has to be produced, occasioned, worked for by the context. Upapadita is the verse's precision: it is 2.143's own word for what the padarthas do to the sentence-meaning (padarthaih upapaditam — 'brought about by the word-meanings'). So the 'secondary' meaning stands to its occasion as the sentence-meaning stands to its words: occasioned, not contained. Which means the distinction primary/secondary is a distinction between meanings that need less occasioning and meanings that need more — a gradient, not a boundary. The sword recognized at sight versus the sword recognized as a door-prop once you see the door: same sword, same powers, different amount of context required.

2.279 unformalized

गोयुष्मन्महतां च्व्यर्थे स्वार्थादर्थान्तरे स्थितौ ।
अर्थान्तरस्य तद्भावस्तत्र मुख्योऽपि दृश्यते ॥ २७९ ॥

goyuṣmanmahatāṃ cvyarthe svārthādarthāntare sthitau |
arthāntarasya tadbhāvastatra mukhyo'pi dṛśyate || 279 ||

In the sense of the cvi-affix, 'go', 'yusmat' and 'mahat' stand in a meaning other than their own; there the becoming-that of the other meaning is seen as primary even so.

A grammatical case that complicates the gradient, and Bhartrhari lets it. The cvi-affix forms the 'becoming' verbs: agni-bhavati, 'it becomes fire'; suklibhavati, 'it becomes white'. In such formations go, yusmat, mahat — cow, you, great — stand svarthad arthantare sthitau: in a meaning *other than their own*, since the form no longer names a cow but a becoming-cow. And yet: arthantarasya tad-bhavah tatra mukhyah api drsyate — the *becoming-that* of the other meaning is seen *as primary even so*. The construction's whole point is transformation into the other, so the 'other' meaning is exactly what the form is for — its primary office, not a stretch. The case matters because it shows the gradient is not a simple function of distance from the citation-meaning: here the word is used at maximum distance from its bare sense and the use is fully primary, because the *construction* fixes the purpose (2.277's tadarthya, conferred by form). Grammar can build a form whose primary meaning is the departure. Which is one more demonstration that primariness is an office, not a possession.
contested: the cvi-construction is standard; whether Bhartrhari cites it as a genuine counter-case to his gradient or as a case the gradient absorbs (the construction being its own nimitta) is undecidable — we read it as the latter

2.280 unformalized

महच्वं शुक्लभावं च प्रकृतिः प्रतिपद्यते ।
भेदेनापेक्षिता सा तु गौणत्वस्य प्रयोजिका ॥ २८० ॥

mahacvaṃ śuklabhāvaṃ ca prakṛtiḥ pratipadyate |
bhedenāpekṣitā sā tu gauṇatvasya prayojikā || 280 ||

The base attains greatness and the state of being white; but that (base), being depended upon by way of difference, is what occasions secondariness.

The cvi-case analysed, and the analysis locates secondariness precisely. In suklibhavati the prakrti, the base ('sukla'), pratipadyate the state of being white — the base does deliver its own content; there is no mystery about what it contributes. But: bhedena apeksita sa tu gaunatvasya prayojika — that base, *being depended upon by way of difference*, is what occasions the secondariness. The phrase bhedena apeksita is the key: the base is leaned on *as differentiated*, as a term standing apart within the construction, and it is that standing-apart-and-being-leaned-on which generates whatever secondariness the form has (2.267's nimitta/nimitti relation, now located in the derivational morphology rather than between two words). The verse's interest is methodological: it shows that even where a construction's primary purpose is 'becoming other' (2.279), the analyst can find a secondariness inside it — by dividing. Divide, and a leaner and a leaned-on appear; and where they appear, gauna appears. Secondariness is a shadow cast by apoddhara.

2.281 unformalized

अग्निसोमादयः शब्दा ये स्वरूपनिबन्धनाः ।
संज्ञिभिः संप्रयुज्यन्तेऽप्रसिद्धेस्तेषु गौणता ॥ २८१ ॥

agnisomādayaḥ śabdā ye svarūpanibandhanāḥ |
saṃjñibhiḥ saṃprayujyante'prasiddhesteṣu gauṇatā || 281 ||

Words such as 'agni' and 'soma', which are grounded in their own form, are employed together with their name-bearers; secondariness in them is on account of non-establishment.

Proper names as the test case for the currency-doctrine. Agni, Soma: words svarupa-nibandhanah, grounded in their own form — used as *names* (of a man called Agnidatta, of a boy called Soma), samjnibhih samprayujyante, employed together with their name-bearers. When 'Agni' names a man, is that secondary? The verse: aprasiddheh tesu gaunata — secondariness in them is *on account of non-establishment*: because the naming-use has not (yet) become current for this hearer. Not because a fire-meaning is being stretched to a man; not because of resemblance (the boy is not fiery) or leaning or transposition — none of the rivals' machinery applies to proper names, which is exactly why they are the test. What makes the use feel secondary is only that the channel is unworn. And once it is worn — once you know the man — 'Agni' delivers him with no felt stretch at all, sruti-matrena (2.278). The name-case isolates the currency-variable from every other, and the currency-variable alone accounts for the phenomenon. This is 2.253's thesis at its cleanest.

2.282 unformalized

अग्निदत्तस्तु योऽग्निः स्यात् तत्र स्वार्थोपसर्जनः ।
शब्दो दत्तार्थवृत्तित्वाद् गौणत्वं प्रतिपद्यते ॥ २८२ ॥

agnidattastu yo'gniḥ syāt tatra svārthopasarjanaḥ |
śabdo dattārthavṛttitvād gauṇatvaṃ pratipadyate || 282 ||

But where an Agnidatta is himself a fire — there the word, having its own meaning subordinated, attains secondariness on account of operating on the meaning of 'datta'.

The mirror-case, constructed with some wit. Take a man named Agnidatta ('fire-given'), and now suppose he *is* a fire — or rather, consider the word 'agni' inside the compound-name when the compound is being used of its bearer. There: svartha-upasarjanah sabdah — the word has its own meaning *subordinated*, made upasarjana (the technical term for the dependent member, 1.62, 2.226); datta-artha-vrttitvat gaunatvam pratipadyate — it attains secondariness *because it operates on the meaning of 'datta'*: 'agni' inside 'Agnidatta' is not naming a fire, it is qualifying a giving. The verse pairs with 2.281 to complete the demonstration: the same word 'agni' is secondary in one direction (name of a man: unworn channel) and secondary in another (member of a name-compound: subordinated office), and the two secondarinesses have nothing in common — one is a fact about currency, the other about grammatical rank. The category 'gauna' is not one category. It is a bag into which theorists have swept every case where a use does not arrive with the ease of the commonest one.

2.283 unformalized

निमित्तभेदात् प्रक्रान्ते शब्दव्युत्पत्तिकर्मणि ।
हरिश्चन्द्रादिषु सुटो भावाभावौ व्यवस्थितौ ॥ २८३ ॥

nimittabhedāt prakrānte śabdavyutpattikarmaṇi |
hariścandrādiṣu suṭo bhāvābhāvau vyavasthitau || 283 ||

When the operation of deriving words has been commenced, on account of difference of occasioning ground, the presence and absence of the augment 'sut' stand settled in 'Hariscandra' and the like.

The doctrine confirmed from the grammar's own augment-rules, and the case is exquisitely apt. Hariscandra: the name takes the augment sut (the s in hari-s-candra) — or does not, in otherwise parallel compounds. Why? Nimitta-bhedat, from *difference of occasioning ground* — the augment's presence or absence tracks which derivational ground the analyst credits, which is to say, which story he tells about the compound's formation. Sabda-vyutpatti-karmani prakrante: when the operation of deriving has been *commenced* — the augment's fate is settled inside the derivation, by the derivation's own choice of nimitta. So even the phonological shape of the word depends on which of the possible occasioning grounds the grammar elects (2.170's under-determination, 2.171's confusions of ground and limit — now shown to have audible consequences). This is as far as the point can be pressed: the analyst's choice of nimitta does not merely re-describe the word, it *changes the form* the rules produce. Postings all the way down, and the postings have teeth.

2.284 unformalized

ऋष्यादौ प्राप्तसंस्कारो यः शब्दोऽन्येन युज्यते ।
तत्रान्तरङ्गः संस्कारो बाह्येऽर्थे न निवर्तते ॥ २८४ ॥

ṛṣyādau prāptasaṃskāro yaḥ śabdo'nyena yujyate |
tatrāntaraṅgaḥ saṃskāro bāhye'rthe na nivartate || 284 ||

A word that has obtained its refinement in the case of a seer's name and the like, when it is joined with another — there the internal refinement does not withdraw in respect of the external meaning.

A rule about the persistence of grammatical form across changes of reference. A word that has prapta-samskarah — obtained its *refinement*, its grammatically-formed shape — rsy-adau, in the case of a seer's name and such (names having their own formation-rules), and which is then anyena yujyate, joined with another (used of something else, transferred to a new bearer): tatra antarangah samskarah bahye 'rthe na nivartate — there the *internal* refinement does not withdraw in respect of the *external* meaning. The form, once made, stays made: the word carries its morphology into every new office, and the new referent does not unmake it. This is 2.256's svarupa-invariance given its grammatical statement, and it is the exact complement of everything the movement has argued: what is fixed about a word is its form; what is free is its meaning. The rivals had it backwards — they fixed the meaning (primary sense, the word's own) and let the application wander (2.255's transposition). Bhartrhari fixes the form (antaranga, made once, never withdrawing) and lets the meaning be carved by occasion. The grammar's own antaranga-doctrine testifies for him.
contested: antaranga/bahya we read as internal (formal) versus external (referential); the technical antaranga-paribhasa (inner rules take precedence) is surely being invoked, but the verse's own contrast is the form/meaning one

2.285 · Truth, Error, and the Word unformalized

अत्यन्तविपरीतोऽपि यदा योऽर्थोऽवधार्यते ।
यथासंप्रत्ययं शब्दस्तत्र मुख्यः प्रवर्तंते ॥ २८५ ॥

atyantaviparīto'pi yadā yo'rtho'vadhāryate |
yathāsaṃpratyayaṃ śabdastatra mukhyaḥ pravartaṃte || 285 ||

Even when the meaning determined is utterly contrary, the word operates there as primary, according to the comprehension.

The movement's conclusion, and it is radical. Atyanta-viparito 'pi yada yo 'rtho avadharyate: even when the meaning determined is *utterly contrary* — the man taken for a post, the shell for silver, the rope for a snake, the word's referent as wrong as it can be — yatha-sampratyayam sabdas tatra mukhyah pravartate: the word operates there *as primary*, *according to the comprehension*. Say 'snake!' of a rope in the dusk, and 'snake' is not being used figuratively. It is being used primarily, of what the speaker took to be a snake. Error does not make speech figurative; it makes it false. And that distinction — which the whole movement has been clearing ground for — is what separates the gauna from the mithya: secondariness is a matter of currency and occasioning (2.253, 2.278), falsity is a matter of the world. Yatha-sampratyayam, 'according to the comprehension', is the standing criterion (2.34): what the word does is fixed by what was understood, not by what was there. The verses following ask the harder question this opens — how, then, do we ever tell the rope from the snake?

2.286 unformalized

यद्यपि प्रत्ययाधीनमर्थतत्वावधारणम् ।
न सर्वः प्रत्ययस्तस्मिन्नसिद्ध इव जायते ॥ २८६ ॥

yadyapi pratyayādhīnamarthatatvāvadhāraṇam |
na sarvaḥ pratyayastasminnasiddha iva jāyate || 286 ||

Although the determination of the meaning's reality depends on cognition, not every cognition arises in respect of it as if unestablished.

The guard against the obvious misreading of 2.285. Yes: arthatattva-avadharanam pratyaya-adhinam — the determination of the meaning's *reality* depends on cognition; there is no access to what is real except through some cognition of it (2.132-133's ledger). But: na sarvah pratyayah tasmin asiddha iva jayate — *not every* cognition arises in respect of it *as if unestablished*: cognitions are not all equally shaky; some arrive with the character of the settled, some with the character of the doubtful, and the difference is *in* them, not imposed afterward. This blocks the slide from 'all access is cognitive' to 'no cognition is better than another' — the slide that 2.285's tolerance might seem to license. Bhartrhari has already refused it in Kanda I (1.39's undoubted cognition; 1.141's audit) and refuses it again here. The dependence of reality-determination on cognition is not the collapse of the distinction between veridical and erroneous; it is the statement that the distinction must be drawn *within* cognition. The next verses draw it.

2.287 unformalized

दर्शनं सलिले तुल्यं मृगतृष्णादिदर्शनैः ।
तुल्यत्वे दर्शनादीनां न जलं मृगतृष्णिका ॥ २८७ ॥

darśanaṃ salile tulyaṃ mṛgatṛṣṇādidarśanaiḥ |
tulyatve darśanādīnāṃ na jalaṃ mṛgatṛṣṇikā || 287 ||

The seeing in the case of water is like the seeings of a mirage and such; the seeings being alike, water is not a mirage.

The distinction drawn, and the verse's structure is its argument. First line: darsanam salile tulyam mrgatrsnadi-darsanaih — the *seeing* in the case of real water is *like* the seeing of a mirage: phenomenally indistinguishable; the thirsty man's experience is the same in both, which is why he runs in both. Second line: tulyatve darsananam na jalam mrgatrsnika — the seeings being alike, *water is not a mirage*. The flat refusal is the point. Indistinguishability of the experiences does not entail identity of the objects; the sceptic's inference (we cannot tell them apart, therefore there is no difference) is invalid, and the verse simply declines it. This is the exact counterweight to 2.140's emblem (the sky's false floor): there, perception's confident deliverance of what is not there; here, the reminder that its equally confident deliverance of what *is* there is not thereby impugned. Phenomenal parity is a fact about us. The difference between water and mirage is a fact about them — and the next verse says how we get at it.

2.288 unformalized

यदसाधारणां कार्यं प्रसिद्धं रज्जुसर्पयोः ।
तेन भेदः परिच्छेद्यस्तयोस्तुल्येऽपि दर्शने ॥ २८८ ॥

yadasādhāraṇāṃ kāryaṃ prasiddhaṃ rajjusarpayoḥ |
tena bhedaḥ paricchedyastayostulye'pi darśane || 288 ||

The non-common effect that is well established for a rope and a snake — by that, the difference of the two is to be discriminated, even though the seeing is alike.

The criterion, and it is a pragmatic one. Rope and snake, in the dusk, are one seeing. What distinguishes them? Yad asadharanam karyam prasiddham rajju-sarpayoh: the *non-common effect* well established for each — the snake bites, the rope does not; the rope ties, the snake will not. Tena bhedah paricchedyah — *by that* the difference is to be discriminated — tayoh tulye 'pi darsane: even though the seeing is alike. The test is differential efficacy: not how it looks but what it does, and specifically what it does that the other cannot. This is the tradition's arthakriya-criterion (2.33's own vocabulary, the Buddhists' central test of the real) deployed against phenomenal scepticism — and it is exactly the audit 2.141 prescribed: examine the perceptible by reasoning, do not posit the seen on seeing's authority. Note that the criterion is available *within* cognition (one cognizes the biting, or its absence) — so 2.286's constraint is honoured: the distinction between the veridical and the erroneous is drawn inside cognition, by cognition's own differential yields, without appeal to an outside we do not have.

2.289 unformalized

प्रसिद्धार्थविपर्यासनिमित्तं यच्च दृश्यते ।
यस्तस्माल्लक्ष्यते भेदस्तदसत्यं प्रचक्षते ॥ २८९ ॥

prasiddhārthaviparyāsanimittaṃ yacca dṛśyate |
yastasmāllakṣyate bhedastadasatyaṃ pracakṣate || 289 ||

And where an occasioning ground for the transposition of a well-established meaning is seen — the difference that is marked from that, they declare to be the unreal.

The definition of the unreal, arrived at by the same route. Prasiddha-artha-viparyasa-nimittam yac ca drsyate — where an *occasioning ground* for the transposition of a well-established meaning is *seen*: where one can identify what produced the misjudgement (the dusk, the coiled shape, the fear, the distance) — yah tasmat laksyate bhedah tad asatyam pracaksate: the difference marked from *that* is what they declare *unreal*. The unreal is not defined by how it appears (2.287: appearances are parity) nor by decree, but by *diagnosability*: an error is a taking whose occasioning ground can be exhibited. The rope-snake is unreal because one can say why it looked that way; the water is real because no such account is forthcoming — one looks for the mirage's cause and finds instead a pond. This is a criterion of remarkable modernity — error identified by the availability of a debunking explanation — and it completes the epistemology of 2.285-288 without ever leaving the circle of cognition. It also, quietly, explains 2.140's dome: we can say why the sky looks floored, and that is why it is not.

2.290 unformalized

यच्च निम्नोन्नते चित्रे सरूपं पर्वंतादिभिः ।
न तत्र प्रतिघातादि कार्यं तद्वत् प्रवर्तते ॥ २९० ॥

yacca nimnonnate citre sarūpaṃ parvaṃtādibhiḥ |
na tatra pratighātādi kāryaṃ tadvat pravartate || 290 ||

And where, in a painting, high and low are of like form with mountains and the like — there the effect of obstruction and the rest does not operate as it does with those.

The criteria applied to the case that will matter most, and it is chosen with care. In a citra, a painting, nimna-unnate — the depths and heights, the modelled recession — are sarupam parvatadibhih: of *like form* with real mountains: phenomenal parity again (2.287), and the painter's whole art consists in achieving it. But: na tatra pratighatadi karyam tadvat pravartate — there the effect of *obstruction* and the rest does not operate as it does with those: you cannot climb the painted mountain, it does not block the wind, it stops nothing. Asadharanam karyam (2.288) is absent; the debunking ground is available (it is a painted surface, 2.289); therefore the painted mountain is unreal, however like. The example is the movement's terminus and it is doing more than it says. The kanda has repeatedly used the picture-analogy for the sentence (2.8-9: the painting one in form, described by its differences; 2.52: the figure executed in stages) — and here the painting is the emblem of the real-seeming unreal. The suggestion, left unstated, is the one Kanda I made at 1.129 with the fire-brand circle: what language builds has exactly this status — vivid, coherent, useless for climbing — and knowing it is what 2.238's untrue path means by standing on the false to seek the true.

2.291-293 unformalized

स्पर्शप्रबन्धो हस्तेन यथा चक्रस्य सन्तः ।
न तथाऽलातचक्रस्य विच्छिन्नं स्पृश्यते हि तत् ॥ २९१ ॥
वप्रप्राकरातल्पैश्च स्पर्शनावरणे यथा ।
नगरेषु न ते तद्वद् गन्धर्वनगरेष्वपि ॥ २९२ ॥
मुगपश्वादिभिर्यावान् मुख्यैरर्थः प्रसाध्यते ।
तावान्न मृन्मयेऽप्यस्ति तस्मात्ते विषयः कनः ॥ २९३ ॥

sparśaprabandho hastena yathā cakrasya santaḥ |
na tathā'lātacakrasya vicchinnaṃ spṛśyate hi tat || 291 ||
vapraprākarātalpaiśca sparśanāvaraṇe yathā |
nagareṣu na te tadvad gandharvanagareṣvapi || 292 ||
mugapaśvādibhiryāvān mukhyairarthaḥ prasādhyate |
tāvānna mṛnmaye'pyasti tasmātte viṣayaḥ kanaḥ || 293 ||

(291) As the continuity of touch by the hand exists for a real wheel — not so for a firebrand-circle; for that is touched as interrupted. (292) And as, in cities, there is touching and enclosure by ramparts, walls and terraces — those are not so in gandharva-cities. (293) As much purpose as is accomplished by real deer, cattle and the like — that much does not exist even in a clay one; therefore they are the domain of the 'ka' suffix.

Three applications of the effect-criterion (2.288), and the third is a small joke with a doctrine inside it. The firebrand-circle (1.129's emblem): a real wheel gives sparsa-prabandha, a *continuity of touch* to the hand laid on it; the whirling brand gives interrupted contact — vicchinnam sprsyate — the hand finds gaps where the eye found a circle. Touch audits sight, and the circle fails. The gandharva-city, that shimmering mirage-town of the tradition: real cities enclose and resist by rampart and wall; the phantom city does not — no obstruction, no shelter. And the clay deer: it accomplishes less than a living one — na tavat mrnmaye 'pi asti — and the verse's conclusion is a grammatical pun. Tasmat te visayah kanah: *therefore they are the domain of the 'ka' suffix* — the affix that forms diminutives and likenesses (mrga-ka, 'a little deer', 'a mere deer-thing'). The clay deer is deficient in efficacy, and the grammar marks the deficiency morphologically: language has a suffix for the not-quite-real. The point is serious under the wit — the effect-criterion is not the philosopher's invention; it is already sedimented in the language's own machinery for marking simulacra.
contested: kanah we take as the affix ka (the diminutive/similitude-forming suffix, Panini 5.3.96ff.); the reading makes the verse's conclusion a grammatical observation, which fits the kanda's habit of finding its doctrines in the sastra's own postings

2.294 unformalized

महानाव्रयते देशः प्रसिद्धैः पर्वतादिभिः ।
अल्पदेशान्तरावस्थं प्रतिबिम्बं तु दृश्यते ॥ २९४ ॥

mahānāvrayate deśaḥ prasiddhaiḥ parvatādibhiḥ |
alpadeśāntarāvasthaṃ pratibimbaṃ tu dṛśyate || 294 ||

A great region is covered by well-established mountains and the like; but a reflection is seen as standing within a small extent of space.

The effect-criterion applied to the reflection — the image the work has used at every load-bearing joint (1.49, 1.99-100). A real mountain avriyate mahan desah — *covers a great region*: occupies, occludes, takes up the space its size requires. The reflected mountain is alpa-desantara-avastham drsyate — seen as standing within a *small extent*: the whole range in the palm-sized mirror, in the sword's blade, in the eye. Size, which for the real thing is an efficacy (it covers), is for the reflection a mere appearance (it does not). This is 1.100's argument (the mountain in the gem-facet: viruddha-parimana, contradictory measures) now stated as an effect-test rather than a reductio: not merely that duplicate-entities would be absurd, but that the reflection *fails the criterion* — it does not do what mountains do. The verse quietly completes a long arc. The reflection has served throughout as the model for the sphota's borrowed properties; here it is examined in its own right and found unreal by the same test that found the painted mountain unreal (2.290). The instrument of the great similes has been audited by the criterion the similes' user gave us.

2.295 unformalized

मरणादिनिमित्तं च यथा मुख्या विषादयः ।
न ते स्वप्नादिषु स्वस्य तद्वदर्थस्य साधकाः ॥ २९५ ॥

maraṇādinimittaṃ ca yathā mukhyā viṣādayaḥ |
na te svapnādiṣu svasya tadvadarthasya sādhakāḥ || 295 ||

And as real poisons and the like are the occasion of death and such — those, in dreams and the like, are not accomplishers of that same purpose for oneself.

The criterion's last and most intimate application. Mukhyah visadayah — *real* poisons — are marana-adi-nimittam, the occasion of death and the rest: that is what poison does, and doing it is what makes it poison. Na te svapnadisu svasya tadvad arthasya sadhakah — those, in dreams and such, are not *for oneself* accomplishers of that purpose: the dreamt poison kills the dreamt man and leaves the dreamer breathing. The example completes the series (firebrand, phantom city, clay deer, reflection, painting) with the case where the stakes are one's own life, and the effect-criterion holds there as everywhere: what does not do it, is not it. Note svasya — *for oneself*: the qualification is exact and it concedes what the dream-world genuinely has. Within the dream the poison works; the dreamt man does die. What the dream lacks is efficacy across the boundary, for the dreamer. Efficacy is thus always efficacy-within-a-domain — which is precisely the vyavahara-relativity the kanda has honoured throughout (2.142's working namings; 1.28's vyavastha), and which keeps the criterion from collapsing into a naive realism it never wanted.

2.296 unformalized

देशकालेन्द्रियगतैर्भेदैर्यद् दृश्यतेऽन्यथा ।
यथा प्रसिद्धिर्लोकस्य तथा तदवसीयते ॥ २९६ ॥

deśakālendriyagatairbhedairyad dṛśyate'nyathā |
yathā prasiddhirlokasya tathā tadavasīyate || 296 ||

That which, through differences residing in place, time and sense-faculty, is seen otherwise — it is determined according as the world's established currency has it.

The final adjudication, and it hands the verdict to a court the kanda has been deferring to since 1.27. Desa-kala-indriya-gataih bhedaih — through differences residing in *place*, *time*, and *sense-faculty* (the three sources of perceptual variation catalogued at 2.105 and 2.136) — yad drsyate anyatha: what is seen *otherwise*, variously, by different observers at different places and times. How is it settled? Yatha prasiddhir lokasya tatha tad avasiyate: *according as the world's established currency has it*, so is it determined. Prasiddhi — the same currency that decided primary and secondary (2.253, 2.281) — now decides real and apparent. This is neither scepticism nor dogmatism but the position 2.142 staked: the world's working namings are not to be second-guessed by the wise, because they are not claims about tattva; they are the settled outcome of countless audits by countless observers with hands to touch the wheel (2.291) and walls to walk into (2.292). Loka-prasiddhi *is* the effect-criterion, applied over generations and sedimented. The epistemology ends where the kanda's authority-doctrine began: in the unbroken practice of the many.

2.297 unformalized

यच्चोपघातजं ज्ञानं यच्च ज्ञानमलौकिकम् ।
न ताभ्यां व्यवहारोऽस्ति शब्दा लोकनिबन्धनाः ॥ २९७ ॥

yaccopaghātajaṃ jñānaṃ yacca jñānamalaukikam |
na tābhyāṃ vyavahāro'sti śabdā lokanibandhanāḥ || 297 ||

And the cognition that is born of an impairment, and the cognition that is non-worldly — by neither of those two is there any practice; words are grounded in the world.

The boundary of the linguistic, drawn with a firm hand. Two kinds of cognition fall outside. Upaghata-jam jnanam — cognition born of *impairment*: the jaundiced eye's yellow, the fevered tongue's bitterness, the madman's certainty — private, undersharable, failing the world's audit (2.296). And jnanam alaukikam — the *non-worldly* cognition: the seer's vision, settled in reality (1.139: 'by it there is no worldly practice, nor is it fastened to words'). Na tabhyam vyavaharo 'sti — by neither is there *practice*: the impaired cannot transact because nobody shares his yellow; the seer cannot transact because what he sees does not parcel into speech. And the conclusion, which is the kanda's charter: sabdah loka-nibandhanah — *words are grounded in the world*. Language lives in the middle band — above the private, below the ineffable — and its home is the loka, the shared, audited, currency-settled domain of ordinary transaction. This is why the sciences describe avidya (2.233) and why that is no disgrace: the constructed middle is exactly where speech works, and a word that reached past it would not be a word.

2.298-299 · The Lamp and the Unintended Meaning unformalized

घटादिषु यथा दीपो येनार्थेन प्रयुज्यते ।
ततोऽन्यस्यापि साचिव्यात् स करोति प्रकाशनम् ॥ २९८ ॥
संसर्गिषु तथाऽर्थेषु शब्दो येन प्रयुज्यते ।
तस्मात् प्रयोजकादन्यानपि प्रत्याययत्यसौ ॥ २९९ ॥

ghaṭādiṣu yathā dīpo yenārthena prayujyate |
tato'nyasyāpi sācivyāt sa karoti prakāśanam || 298 ||
saṃsargiṣu tathā'rtheṣu śabdo yena prayujyate |
tasmāt prayojakādanyānapi pratyāyayatyasau || 299 ||

(298) As a lamp, employed for whatever purpose among pots and the like, effects the illumination also of what is other than that, by way of accompaniment — (299) so, among commingled objects, the word, employed for whatever purpose, conveys also others than that occasioner.

A new movement, on the meanings a word delivers without being meant to — and the lamp returns to carry it. A lamp is brought in for the pot: yena arthena prayujyate, for whatever purpose it is employed. But it lights the mat, the wall, the cat: tato 'nyasya api sacivyat — of what is *other* than that, by way of *accompaniment*, sa karoti prakasanam. Illumination has no way of restricting itself to what it was wanted for; it falls on everything in reach. Tatha: so the word. Samsargisu arthesu — among *commingled* objects, meanings that come mixed together (2.153-154's inescapable furniture) — the word employed for one purpose tasmat prayojakat anyan api pratyayayati asau: conveys *others* than the occasioner too. This is the mechanism behind the whole penumbra of 2.161-163 (the flower arriving on its scent) now given its proper simile: not association merely, but *illumination overspilling its errand*. The lamp-image has served the work at every scale (1.46's fire in the wood, 1.94's unchanging thing lit by changing light, 2.134's faculty displaying by its variety); here it explains why speech always says more than it means to.

2.300-301 unformalized

निर्मन्थन यथाऽरण्योरग्न्यर्थमुपपादितम् ।
धूममप्यनभिप्रेतं जनयत्येकसाधनम् ॥ ३०० ॥
यथा शब्दोऽपि कस्मिश्चित् प्रत्याय्यार्थे विवक्षिते ।
अविवक्षितमप्यर्थं प्रकाशयति सन्निधेः ॥ ३०१ ॥

nirmanthana yathā'raṇyoragnyarthamupapāditam |
dhūmamapyanabhipretaṃ janayatyekasādhanam || 300 ||
yathā śabdo'pi kasmiścit pratyāyyārthe vivakṣite |
avivakṣitamapyarthaṃ prakāśayati sannidheḥ || 301 ||

(300) As the churning of the two fire-sticks, undertaken for the sake of fire, being a single means, generates smoke also, which was not intended — (301) so the word too, when some conveyable meaning is intended, illuminates even an unintended meaning, from proximity.

The same doctrine with a sharper simile, and the sharpening matters. The lamp (2.298) lights the mat because the mat is *there*; nothing about the lamp requires a mat. But the fire-drill: nirmanthanam aranyoh agny-artham upapaditam — the churning of the fire-sticks, undertaken *for fire's sake* — dhumam api anabhipretam janayati eka-sadhanam: generates *smoke* as well, unintended, *being a single means*. Smoke is not an accident of the surroundings; it is what fire-making does. The unintended is here the *same operation's* other yield — you cannot have the fire without it. Tatha sabdo 'pi: so the word — kasmimscit pratyayye 'rthe vivaksite, some conveyable meaning being *intended* — avivaksitam api artham prakasayati sannidheh: illuminates the *unintended* meaning too, from proximity. The pair of similes brackets the phenomenon exactly: some overspill is circumstantial (the lamp's mat) and some is constitutive (the fire's smoke) — and speech has both. Avivaksita, the unintended, has been the movement's technical term since 1.136 and 2.69, always as what meaning *excludes*; here it returns as what meaning cannot help delivering. Intention governs what is said and does not govern what arrives.

2.302 unformalized

यथैवात्यन्तसंसृष्टस्त्यक्तुमर्थो न शक्यते ।
तथा शब्दोऽपि सम्बन्धी प्रविवेक्तुं न शक्यते ॥ ३०२ ॥

yathaivātyantasaṃsṛṣṭastyaktumartho na śakyate |
tathā śabdo'pi sambandhī pravivektuṃ na śakyate || 302 ||

Just as an object utterly commingled cannot be abandoned — so a word that is a relatum cannot be separated out.

The reason the overspill cannot be stopped, stated as a symmetry. Atyanta-samsrstah artho na sakyate tyaktum — an object *utterly commingled* cannot be abandoned: you cannot fetch the cow and leave her colour behind; the furniture comes because it is not detachable (2.154's samsrsta, now given its consequence). Tatha sabdo 'pi sambandhi pravivektum na sakyate: so too a word that is a *relatum* — that stands in relation, as every uttered word does — cannot be *separated out*. The symmetry is the argument: as objects come in bundles, so words come in company, and neither the bundle nor the company can be dissolved by wanting. The speaker who wishes to say just this much and no more is in the position of one who wishes to bring the cow without her colour. Pravivektum — 'to separate out' — is the apoddhara-vocabulary (2.10's extraction), and its appearance here is precise: the analyst can extract a word in his ledger; the speaker cannot extract it in the world. What analysis does on paper, utterance cannot do in the air.

2.303 unformalized

व्यर्थानां सन्निधानेऽपि सति चैषां प्रकाशते (ने) ।
प्रयोजकोऽर्थः शब्दस्य रूपाभेदेऽपि शस्यते ॥ ३०३ ॥

vyarthānāṃ sannidhāne'pi sati caiṣāṃ prakāśate (ne) |
prayojako'rthaḥ śabdasya rūpābhede'pi śasyate || 303 ||

Even when meaningless things are in proximity, and even when these are illuminated, the occasioning meaning of a word is commended, though there be no difference of form.

A qualification that keeps the doctrine from dissolving into noise. Grant that words illuminate the unintended (2.299-301) and that proximity cannot be prevented (2.302). Does the word then simply mean everything in its vicinity? No: prayojakah arthah sabdasya sasyate — the *occasioning* meaning of the word is *commended*, singled out, held up — even where vyarthanam sannidhane, meaningless things stand in proximity and are themselves lit; and even rupa-abhede 'pi, where there is *no difference of form* to mark the intended from the incidental. The distinction between what a word is *for* and what it merely *reaches* survives, and it survives without any formal marker: nothing in the sound tells you which of the lit things was the errand. What tells you is the prayojaka — the occasioning purpose (2.113's prayojana, the sentence's meaning). So the overspill is real and the intention is sovereign: the lamp lights the cat, and the lamp was brought for the pot, and no inspection of the light will tell you which — only knowing what was wanted.

2.304-305 unformalized

क्वचिद् गुणप्रधानत्वमर्थानामविवक्षितम् ।
क्वचित् सान्निध्यमप्येषां प्रतिपत्तावकारणम् ॥ ३०४ ॥
यच्चानुपात्तं शब्देन तत्कस्मिंश्चित् प्रतीयते ।
क्वचित् प्रधानमेवार्थो भवत्यन्यस्य लक्षणम् ॥ ३०५ ॥

kvacid guṇapradhānatvamarthānāmavivakṣitam |
kvacit sānnidhyamapyeṣāṃ pratipattāvakāraṇam || 304 ||
yaccānupāttaṃ śabdena tatkasmiṃścit pratīyate |
kvacit pradhānamevārtho bhavatyanyasya lakṣaṇam || 305 ||

(304) Sometimes the subordinate-and-principal status of meanings is unintended; sometimes even their proximity is no cause in the comprehension. (305) And what is not taken up by the word is understood in some case; sometimes the principal meaning itself becomes the indicator of another.

Four observations that dismantle any tidy mapping from form to content. (304): sometimes the *hierarchy* of meanings — which is guna, which pradhana — is avivaksita, unintended: the sentence does not say which of its contents is central, and the hearer must find it. And sometimes proximity itself is akaranam pratipattau — *no cause* in comprehension: things stand adjacent and contribute nothing, the lamp's light falling on what nobody notices. (305): and yad anupattam sabdena — what is *not taken up* by the word at all — tat kasmimscit pratiyate: is understood in some case (the bahuvrihi's unuttered possessor, 2.219; the question's unstated verb, 2.271). And kvacit pradhanam eva artho bhavaty anyasya laksanam: sometimes the *principal* meaning itself becomes the *indicator* of another — the central content demoted to a mere signpost toward what was actually meant. Take the four together and there is no rule connecting what is uttered, what is proximate, and what is comprehended: content can be present without being meant, meant without being present, central without being the point. Which is exactly what one expects if comprehension is a flash occasioned by an utterance (2.143), and exactly what no compositional calculus can tolerate.

2.306 unformalized

आख्यातं तद्धितार्थस्य यत्किञ्चिदुपदर्शकम् ।
गुणाप्रधानभावस्य तत्र दृष्टो विपर्ययः ॥ ३०६ ॥

ākhyātaṃ taddhitārthasya yatkiñcidupadarśakam |
guṇāpradhānabhāvasya tatra dṛṣṭo viparyayaḥ || 306 ||

The finite verb is what somehow exhibits the meaning of a secondary suffix; there the reverse of the subordinate-and-principal relation is seen.

A grammatical instance of 2.305's inversion. The taddhita, the secondary suffix, carries a meaning; and sometimes it is the akhyata, the *finite verb*, that exhibits it — yat kincid upadarsakam, somehow displaying what the suffix would have said. And the consequence: guna-apradhana-bhavasya tatra drsto viparyayah — there is seen *the reverse* of the subordinate-and-principal relation: the element the grammar files as principal is doing the subordinate's work, and vice versa. The rank the system assigns is not the rank the utterance enacts. This is 2.229-231's ledger-shuffling (base-meaning borne by suffix, suffix-meaning by root) reaching the syntactic categories themselves — and it lands with particular force because pradhana/guna is the axis on which the entire analysis of the sentence turns (2.4's definition, 2.48's architecture, 2.82's ritual hierarchy). If even *that* inverts under employment, then the categories are offices conferred by use, from top to bottom, with nothing underneath them but the flash they serve.
contested: the verse is compressed and the specific construction it has in mind is not recoverable; our reading takes it as a general observation about verbs discharging taddhita-meanings, which the guna/pradhana inversion in the second line requires

2.307(ab) unformalized

निर्देशे लिङ्गसंख्यानां सन्निधानमकारणम् ॥ ३०७ab ॥

nirdeśe liṅgasaṃkhyānāṃ sannidhānamakāraṇam || 307ab ||

In the designation of genders and numbers, proximity is no cause.

A half-verse, and it applies 2.304's second observation to the grammar's own metalanguage. When the science *designates* genders and numbers — when a rule mentions 'the feminine', 'the dual' — sannidhanam akaranam: the *proximity* of the actual gendered or numbered things is no cause: the rule is not made true by there being feminine things about, and the mention does not reach out to them. Metalinguistic designation floats free of its subject matter's presence, exactly as 2.100 said (mention proceeds by exhibited form, not by what is understood). The half-verse is doing housekeeping before the movement's last topic — the semantics of measure-words — but its placement is not idle: it reminds the reader that everything about to be said concerning 'short' and 'long' is being said in a metalanguage whose relation to its objects is itself a posting.
contested: the source paginates 2.307 as two half-verses; we key commentary to each half as the corpus labels them

2.307(cd) unformalized

प्रमाणमेव ह्रस्वादावनुपात्तं प्रतीयते ॥ ३०७cd ॥

pramāṇameva hrasvādāvanupāttaṃ pratīyate || 307cd ||

It is the measure itself, in 'short' and the like, that is understood, though not taken up.

The complement, and it opens the movement's final case. Say 'short' — hrasva — of a vowel. What is understood? Pramanam eva — the *measure itself*: a determinate quantity, one matra. And: anupattam pratiyate — *though not taken up*: no word for the measure occurs; 'short' does not contain 'one matra'; the quantity arrives unuttered (2.305's anupatta, now in the grammar's most technical vocabulary). Measure-terms are thus the purest specimens of the movement's thesis: what they convey is precisely what they do not say. And they are chosen for a further reason. Vowel-quantity was the datum that most embarrassed the timeless-word doctrine back at 1.76 — the one acoustic difference that is genuinely contrastive — and the resolution there was that duration belongs to the manifesting sound and is transferred to the word figuratively. Here the same territory is re-entered from the semantic side: what does the *word* 'short' mean? The next verses answer, and the answer is that it means by indicating, not by containing.

2.308-309 unformalized

ह्रस्वस्यार्द्धं च यद् दृष्टं तत्तस्याऽसन्निधावपि ।
ह्रस्वस्य लक्षणार्थत्वात् तद्वदेवाभिधीयते ॥ ३०८ ॥
दीर्घप्लुताभ्यां तस्य स्याद् मात्रया वा विशेषणम् ।
जातेर्वा लक्षणा यस्मात् सर्वथा सप्तपर्णवत् ॥ ३०९ ॥

hrasvasyārddhaṃ ca yad dṛṣṭaṃ tattasyā'sannidhāvapi |
hrasvasya lakṣaṇārthatvāt tadvadevābhidhīyate || 308 ||
dīrghaplutābhyāṃ tasya syād mātrayā vā viśeṣaṇam |
jātervā lakṣaṇā yasmāt sarvathā saptaparṇavat || 309 ||

(308) And that half which is seen of a short (vowel) — that, even in its absence, is expressed just as such, because the short serves as the indicator. (309) Or its qualification would be by the long and the prolated, or by the matra; or there is indication of the universal — since in every way it is like 'seven-leaf'.

The measure-word's semantics, canvassed to exhaustion, and the last simile is the movement's signature. The problem: the pluta (three matras), the dirgha (two), the hrasva (one) are defined by reference to one another, and half a short is spoken of though no such vowel exists. (308): hrasvasya ardham — the half of a short — is expressed *even in its absence*, tasya asannidhav api, because hrasvasya laksanarthatvat: the short serves as the *indicator*, the measuring-rod; one speaks of half of it as one speaks of half a cubit, without there being any such thing lying about. (309) offers the alternatives: qualification by the long and prolated (defining each by the others, the system closing on itself); or by the matra (an abstract unit doing the measuring); or jater va laksana — *indication of the universal*. And then: sarvatha sapta-parnavat — *in every way, like 'seven-leaf'*. The saptaparna is a tree whose name says seven leaves and whose leaves are not seven: the name indicates without describing, points without containing, and everyone finds the tree. That is the measure-word, and — the simile's reach is deliberate — that is every word: sarvatha, in every way. Language names as saptaparna names.
contested: saptaparna (the Alstonia tree, 'seven-leaved' and not seven-leaved) is the traditional example of a name whose components misdescribe its bearer; whether sarvatha extends the comparison to all words (our reading, with 2.36's precedent) or only to measure-terms is undecidable from the karika

2.310 unformalized

गन्तव्यं दृश्यतां सूर्य इति कालस्य लक्षणे ।
ज्ञायतां काल इत्येतत्सोपायमभिधीयते ॥ ३१० ॥

gantavyaṃ dṛśyatāṃ sūrya iti kālasya lakṣaṇe |
jñāyatāṃ kāla ityetatsopāyamabhidhīyate || 310 ||

'One must go — look at the sun!' — in indicating the time, this, 'let the time be known', is expressed together with its means.

The movement closes on a small scene, and it gathers everything. Someone says: gantavyam, drsyatam suryah — 'we must go; look at the sun!' What is meant? Kalasya laksane — in *indicating the time*: jnayatam kalah — 'let the time be known' — sopayam abhidhiyate: expressed *together with its means*. Nobody is being told to look at the sun for the sun's sake; the sun is the means, and the errand is the hour. And notice how much of the kanda is in this one utterance. The meaning is the purpose (2.113: vakyartha is prayojana) — knowing the time, which no word in the sentence names (2.305's anupatta). The sun is the indicator, meaning by pointing rather than containing (2.309's saptaparna). What is uttered is the means and what is meant is the end, delivered together, undivided, in a single flash to any hearer who is standing there (2.143). And the flash is practical through and through — itikartavyata, what is to be done (2.146: 'in the matter of what is to be done, no one transgresses it'). The Vakyakanda's whole doctrine, in a sentence someone might say at a doorway: language points at the sun and delivers the hour.

2.311 · Context Determines Meaning unformalized

विध्यत्यधनुषेत्यत्र विशेषेण निदर्यते ।
सामान्यमाश्रयः शक्तेर्यः कश्चत्प्रतिपादकः ॥ ३११ ॥

vidhyatyadhanuṣetyatra viśeṣeṇa nidaryate |
sāmānyamāśrayaḥ śakteryaḥ kaścatpratipādakaḥ || 311 ||

In 'he pierces with a non-bow', a general is indicated by way of the particular — a general which is the substratum of the capacity, and which is somehow the conveyor.

A curious idiom opens the movement on contextual determination. 'Vidhyati adhanusa' — he pierces with a non-bow: the phrase denies the instrument and yet an instrument is understood; something does the piercing, and what is conveyed is instrumentality-in-general, samanyam asrayah sakteh — a general that is the *substratum of the capacity*: the bare role of instrument, waiting for a filler. Visesena nidarsyate: the general is indicated *by way of the particular* — one names a bow, negates it, and what survives is the role the bow would have played. This is the hovering-general of 2.44-45 generated by an unlikely route, and the verse's yah kascit pratipadakah — 'which is *somehow* the conveyor' — is the honest shrug the case deserves. What conveys the role? Not the word 'dhanus', which is negated; not the negation, which negates nothing else. The role arrives from the construction, and no word in it says so. Every subsequent verse in this movement is a variation on that discovery.

2.312 unformalized

काकेभ्यो रक्ष्यतां सर्पिरिति बालोऽपि चोदितः ।
उपघातपरे वाक्ये न श्वादिभ्यो न रक्षति ॥ ३१२ ॥

kākebhyo rakṣyatāṃ sarpiriti bālo'pi coditaḥ |
upaghātapare vākye na śvādibhyo na rakṣati || 312 ||

'Guard the ghee from the crows' — even a child, so instructed, in a sentence whose purport is harm, does not guard it from dogs and the like.

The movement's most famous example, and one of the great examples in the world's philosophy of language. Kakebhyo raksyatam sarpih — 'guard the ghee from the crows'. The word says crows. The child — balo 'pi coditah, *even a child*, so instructed — guards it from dogs. Why? Upaghata-pare vakye: the sentence's *purport* is harm; crows were named because crows were the salient threat, and what was meant was: keep the ghee safe. And na svadibhyo na raksati — the litotes is exact: it is not the case that he fails to guard it from dogs. No inference is performed, no rule of pragmatics consulted; the child simply understands, and would be astonished to be told that 'crows' meant crows. The example demolishes literal-meaning semantics with a single household scene, and it is chosen for the child on purpose: this is not sophisticated interpretation but the baseline competence of anyone who understands anything (2.117's balanam tirascam ca — children and animals). The sentence's meaning is its purpose (2.113), and the purpose was never crows.

2.313 unformalized

प्रक्षालनं शरावाणां स्थालीनां मार्जनं तथा ।
अनुक्तमपि रूपेण भुज्यङ्गत्वात् प्रतीयते ॥ ३१३ ॥

prakṣālanaṃ śarāvāṇāṃ sthālīnāṃ mārjanaṃ tathā |
anuktamapi rūpeṇa bhujyaṅgatvāt pratīyate || 313 ||

The washing of the cups and likewise the scouring of the pots, though unstated in explicit form, is understood, from being an auxiliary of eating.

The companion example, from the other side. Say 'prepare the meal' and the praksalanam saravanam, the washing of the cups, and the marjanam sthalinam, the scouring of the pots, are anuktam api rupena — *unstated in explicit form*, no word for them anywhere — and pratiyate, understood: bhujy-angatvat, *from being an auxiliary of eating*. Where 2.312 showed a named thing not meant, this shows unnamed things meant. Together they close the field: what is said and what is meant come apart in both directions, and neither direction troubles a child. The mechanism named here is angatva — being an *auxiliary* — the Mimamsa's own relation (2.81-82's anga-structure): the meal's meaning carries its subsidiaries with it because a meal *is* a structure of subsidiaries, and understanding the end is having the means. This is 2.146's practical sovereignty: comprehension delivers the whole itikartavyata, the entire what-is-to-be-done, and the words are the occasion of that delivery rather than its inventory.

2.314 unformalized

वाक्यात् प्रकरणादर्थादौचित्याद् देशकालतः ।
शब्दार्थाः प्रविभज्यन्ते न रूपादेव केवलात् ॥ ३१४ ॥

vākyāt prakaraṇādarthādaucityād deśakālataḥ |
śabdārthāḥ pravibhajyante na rūpādeva kevalāt || 314 ||

Word-meanings are differentiated by the sentence, by the context, by the object, by propriety, by place and time — not from the form alone.

The movement's thesis, stated flatly and comprehensively. Sabdarthah pravibhajyante — word-meanings are *differentiated*, carved out (the vocabulary of 2.250-251: the plenitude divided by nimittas) — and the list of carvers is given: vakyat, by the *sentence* (the word's meaning fixed by its host); prakaranat, by the *context* (the standing situation of discourse); arthat, by the *object* (what is there to be talked about); aucityat, by *propriety* (what it would be fitting to mean); desa-kalatah, by *place and time*. And the denial that gives the verse its force: na rupad eva kevalat — *not from the form alone*. The bare word-form determines nothing. This is the direct contradictory of 2.265's rival criterion (the primary meaning is what the bare form delivers, rupa-matra-nibandhanah), and its placement after 2.312-313 is not accidental: the crows and the unwashed pots are the evidence, this is the thesis. The next two verses give the full inventory of what does the carving, and it is the most systematic list of contextual factors in the ancient world.

2.315-316 unformalized

संसर्गो विप्रयोगश्च साहचर्यं विरोधिता ।
अर्थः प्रकरणं लिङ्गं शब्दस्यान्यस्य सन्निधिः ॥ ३१५ ॥
सामर्थ्यंमौचिती देशः कालो व्यक्तिः स्वरादयः ।
शब्दार्थंस्यानवच्छेदे विशेषस्मृतिहेतवः ॥ ३१६ ॥

saṃsargo viprayogaśca sāhacaryaṃ virodhitā |
arthaḥ prakaraṇaṃ liṅgaṃ śabdasyānyasya sannidhiḥ || 315 ||
sāmarthyaṃmaucitī deśaḥ kālo vyaktiḥ svarādayaḥ |
śabdārthaṃsyānavacchede viśeṣasmṛtihetavaḥ || 316 ||

(315) Connection, dissociation, association, opposition, the object, the context, an indicatory mark, the proximity of another word, (316) capacity, propriety, place, time, gender, accent and the rest — these are the causes of the recollection of the particular, where the word-meaning is undelimited.

The catalogue — fourteen factors, and it has no rival in antiquity. Samsarga, connection (what the word is joined with); viprayoga, dissociation (what it is explicitly kept from); sahacarya, association (the company it customarily keeps); virodhita, opposition (what it is contrasted against); arthah, the object; prakaranam, the context; lingam, an indicatory mark; anyasya sabdasya sannidhih, the proximity of another word; samarthyam, capacity (what could do the job); aucitii, propriety; desah, place; kalah, time; vyaktih, gender; svaradayah, accent and the rest. And the frame that makes the list a doctrine rather than an inventory: sabdarthasya anavacchede — *where the word-meaning is undelimited* — these are visesa-smrti-hetavah, the *causes of the recollection of the particular*. Every term is exact. The word-meaning is *undelimited* in itself (2.250's sarvarthya, the plenitude); these factors *delimit*. And they do so by causing *smrti*, recollection — not by adding content but by waking what the hearer already has (2.117's abhyasa, 2.152's ripened endowment). The list is the operational face of the kanda's entire semantics: one omni-significant word, fourteen kinds of occasion, and a trained mind in which the particular flashes.
contested: the fourteen are stable in the tradition though their glosses vary; vyakti we render 'gender' (the grammatical sense, given the company of svara) though it can mean 'individual' — both are contextual determinants and nothing turns on it

2.317 unformalized

भेदपक्षेऽपि सारूप्याद्भिन्नार्थाः प्रतिपत्तिषु ।
नियता यान्त्यभिव्यक्तिं शब्दाः प्रकरणादिभिः ॥ ३१७ ॥

bhedapakṣe'pi sārūpyādbhinnārthāḥ pratipattiṣu |
niyatā yāntyabhivyaktiṃ śabdāḥ prakaraṇādibhiḥ || 317 ||

Even on the difference-view, words having distinct meanings, being alike in form, come to manifestation as fixed in comprehensions, by the context and the rest.

The catalogue's results shown to be neutral between the schools — the kanda's habitual generosity at a movement's close. Bheda-pakse 'pi: *even on the difference-view* — even for the theorist who holds that 'gauh'-of-the-ox and 'gauh'-of-the-man are two distinct words that merely share a shape (2.257's bheda-marga-anudarsinah), rather than one word with a plenitude (2.250) — the same machinery is required. Sarupyad bhinnarthah: being alike in form though distinct in meaning, such words niyata yanti abhivyaktim pratipattisu — come to *manifestation*, *as fixed*, in comprehensions — prakaranadibhih, by the context and the rest. The homonymist must still explain how the hearer knows which of his two words was uttered, and his only available answer is the list of 2.315-316. So the contextual determination is not a consequence of Bhartrhari's monism; it is a fact both sides must accommodate, and the monism merely explains it more cheaply (one word carved, rather than two words disambiguated). Note abhivyakti — manifestation: even the difference-theorist's disambiguation is, described accurately, a lighting-up of what the hearer holds.

2.318 unformalized

नामाख्यातसरूपा ये कार्यान्तरनिबन्धनाः ।
शब्दा वाच्याश्च तेष्वर्थो न रूपादधिगम्यते ॥ ३१८ ॥

nāmākhyātasarūpā ye kāryāntaranibandhanāḥ |
śabdā vācyāśca teṣvartho na rūpādadhigamyate || 318 ||

Those words which are alike in form with nouns and finite verbs, and which are grounded in a different operation — and the things expressible in them: there the meaning is not understood from the form.

A hard case that clinches 2.314's denial. Some words are nama-akhyata-sarupah — *alike in form* with nouns and finite verbs (the forms that are ambiguous between a nominal and a verbal reading: Sanskrit is full of them, and the grammar's own technical vocabulary compounds the problem) — and karyantara-nibandhanah, *grounded in a different operation*: their real office is elsewhere than their shape suggests. Vacyah ca tesu arthah: and the things expressible in them — na rupad adhigamyate: the meaning is *not understood from the form*. Here form is not merely insufficient (2.314) but actively misleading: the shape says noun, the office is verbal, and only the context reveals it. The case matters because form-based semantics has one last refuge — that form at least fixes the *category*, if not the meaning — and this verse closes it. The form does not even tell you what kind of word you are hearing. Everything below the sentence is up for grabs, and the sentence settles it (2.10's apoddhara, from the top down).

2.319-320 unformalized

या प्रवृत्तिनिवृत्त्यर्था स्तुतिनिन्दाप्रकल्पना ।
कुशलः प्रतिपत्ता तामयथार्था प्रतीयते ॥ ३१९ ॥
विधीयमानं यत्कर्म दृष्टादृष्टप्रयोजनम् ।
स्तूयते सा स्तुतिस्तस्य कतुरेव प्रयोजिका ॥ ३२० ॥

yā pravṛttinivṛttyarthā stutinindāprakalpanā |
kuśalaḥ pratipattā tāmayathārthā pratīyate || 319 ||
vidhīyamānaṃ yatkarma dṛṣṭādṛṣṭaprayojanam |
stūyate sā stutistasya katureva prayojikā || 320 ||

(319) That construction of praise and blame which has as its purpose engagement and withdrawal — the skilled comprehender understands it as not-according-to-the-literal-meaning. (320) The action being enjoined, which has seen and unseen purposes, is praised; that praise is the occasioner of the agent alone.

The arthavada-doctrine, and it is the movement's culmination. Praise and blame in scripture — the eulogies of rites, the horrors attached to omissions — have a purpose: pravrtti-nivrtty-artha, *engagement and withdrawal*, getting the hearer to do or not do. And: kusalah pratipatta tam ayatharthaam pratiyate — the *skilled* comprehender understands it as *not-according-to-the-literal-meaning*, ayathartha. He does not believe the extravagant claims; he does not need to; understanding them as praise *is* understanding them. (320): the enjoined action, having drsta and adrsta purposes, is what gets praised — and stutis tasya kartur eva prayojika: that praise is the occasioner *of the agent alone*. The praise does not describe the rite; it moves the man. This is the abhidheya/prayojana distinction (2.113) applied to scripture's own rhetoric, and its consequence is large: whole tracts of the Veda mean *nothing about the world* and everything about the hearer's will — and the kusala, the skilled, knows this and is not deceived, while the literalist who believes the eulogy has misunderstood the sentence precisely by getting its words right.

2.321-323 unformalized

व्याघ्रादिव्यपदेशेन यथा बालो निवर्त्यते ।
असत्योऽपि तथा कश्चित् प्रत्यवायो विधीयते ॥ ३२१ ॥
न संविधानं कृत्वापि प्रत्यवाये तथाविधे ।
शास्त्रेण प्रतिषिद्धेऽर्थे विद्वान् कश्चित् प्रवर्तते ॥ ३२२ ॥
सर्पेषु संविधायापि सिद्धैर्मन्त्रौषधादिभिः ।
नान्यथा प्रतिपत्तव्यं न दतो गमयेदिति ॥ ३२३ ॥

vyāghrādivyapadeśena yathā bālo nivartyate |
asatyo'pi tathā kaścit pratyavāyo vidhīyate || 321 ||
na saṃvidhānaṃ kṛtvāpi pratyavāye tathāvidhe |
śāstreṇa pratiṣiddhe'rthe vidvān kaścit pravartate || 322 ||
sarpeṣu saṃvidhāyāpi siddhairmantrauṣadhādibhiḥ |
nānyathā pratipattavyaṃ na dato gamayediti || 323 ||

(321) As a child is turned back by the mention of a tiger and the like, so some consequence, even though unreal, is laid down. (322) Yet, even having made provision against a consequence of that kind, no wise man engages in an object prohibited by the science. (323) Even having made provision against snakes with proven mantras, herbs and the like, one should not conclude otherwise — nor should one let oneself be brought (to it).

The doctrine of 2.319-320 pressed to its most uncomfortable point, and Bhartrhari does not flinch. (321): a child is turned from the well by the mention of a tiger; there is no tiger; the deterrent works and the child is safe. Tatha asatyo 'pi kascit pratyavayah vidhiyate: *so too* some consequence, *even though unreal*, is laid down — scripture's threatened horrors may be the tiger, and the admission is startlingly direct. But then (322) the turn that saves the practice: samvidhanam krtva api — even having *made provision*, having seen through the tiger, having satisfied oneself that the threatened consequence is of that kind — sastrena pratisiddhe 'rthe vidvan kascit na pravartate: *no wise man engages* in what the science prohibits. And (323) the illustration that makes it vivid: even with proven mantras and antidotes against snakes, na anyatha pratipattavyam, one should not conclude otherwise — na dato gamayed iti, nor should one let oneself be brought to the bite. The one who has neutralized the venom still does not put his hand in the basket. The reasoning is exact and it is not superstition: seeing that a deterrent is a device does not license acting as though the prohibition were empty, because the prohibition's authority never rested on the deterrent's literal truth (2.320: the praise occasions the agent, not the facts). The tiger is unreal; the well is deep.
contested: 323's na dato gamayed is corrupt in the transcription and our rendering is tentative; the verse's structure (provision made, prohibition still binding) is clear from 322 and the snake-illustration

2.324-325 unformalized

क्वचित् तत्त्वसमाख्यानं क्रियते स्तुतिनिन्दयोः ।
तत्रापि च प्रवृत्तिश्च निवृत्तिश्चोपदिश्यते ॥ ३२४ ॥
रूपं सर्वपदार्थानां वाक्यार्थोपनिबन्धनम् ।
सापेक्षा ये तु वाक्यार्थाः पदार्थैरेव ते समाः ॥ ३२५ ॥

kvacit tattvasamākhyānaṃ kriyate stutinindayoḥ |
tatrāpi ca pravṛttiśca nivṛttiścopadiśyate || 324 ||
rūpaṃ sarvapadārthānāṃ vākyārthopanibandhanam |
sāpekṣā ye tu vākyārthāḥ padārthaireva te samāḥ || 325 ||

(324) Sometimes a designation of the reality is made in the case of praise and blame; there too, engagement and withdrawal are taught. (325) The form of all word-meanings is the fastening of the sentence-meaning; but those sentence-meanings which are dependent are equal to word-meanings.

The arthavada-doctrine completed, and then a hinge. (324): sometimes praise and blame *do* state tattva, the reality — not every eulogy is a tiger; some say what is so. Tatrapi ca pravrttir nivrttis ca upadisyate: and *there too* engagement and withdrawal are taught — the purpose is the same whether the content is true or not, which is the point: purpose is what the sentence means (2.113), and truth of content is a separate question from meaning (2.285's error operating primarily). Then (325) turns to the movement's last topic: rupam sarva-padarthanam vakyartha-upanibandhanam — the form of *all* word-meanings is the *fastening* of the sentence-meaning: everything the padarthas are, they are as the vakyartha's anchorage. And the corollary: sapeksah ye tu vakyarthah padarthaih eva te samah — those sentence-meanings which are *dependent* (which still reach for something outside themselves) are *equal to word-meanings*: an incomplete sentence is, semantically, just a big word. Completion — the closing of expectancy (2.4) — is what makes a sentence a sentence, and dependence demotes it to the status of a part, however many words it has.

2.326 unformalized

वाक्यं तदपि मन्यन्ते यत्पदं चरितक्रियम् ॥ ३२६ ॥

vākyaṃ tadapi manyante yatpadaṃ caritakriyam || 326 ||

That word too, whose action has been discharged, they consider a sentence.

The complement of 2.325, and the pair states the criterion exactly. There, a many-worded but dependent utterance is only a word; here, padam carita-kriyam — a single *word* whose action has been *discharged* (2.270's carita-asti-kriya: the verb already performed, absorbed, supplied by the context) — vakyam tad api manyante: *that too* they consider a sentence. Between them the two verses sever sentencehood from word-count entirely. What makes a sentence is neither length nor grammatical completeness but the discharge of the action and the closing of expectancy — which is to say, the occasioning of a flash (2.143). 'Fire!' is a sentence; a dependent clause of nine words is not. The criterion is functional through and through, and it is the natural terminus of a kanda that began by asking what a vakya is and answered by relocating the question from the utterance to the comprehension.

2.326a unformalized

आख्यातशब्दे नियतं साधनं यत्र गम्यते ।
तदप्येकं समासार्थं वाक्यमित्यभिधीयते ॥ ३२६a ॥

ākhyātaśabde niyataṃ sādhanaṃ yatra gamyate |
tadapyekaṃ samāsārthaṃ vākyamityabhidhīyate || 326a ||

Where a fixed means is understood in a finite verb, that too, being one, having the meaning of a compound, is called a sentence.

A third case, and it completes the set. Akhyata-sabde niyatam sadhanam yatra gamyate — where a *fixed means* is understood *in a finite verb*: the verb whose instrument or agent is built into it, needing no separate word (as some verbs simply imply their instrument). Tad api ekam samasartham vakyam iti abhidhiyate: *that too*, being *one*, having the meaning of a *compound*, is called a sentence. One word, containing its own participants, discharging its own predication — a sentence by every criterion that matters and by none that counts words. Note samasartham: it has the meaning of a compound — i.e. what a compound does (fusing what analysis would spread across words) this verb does with its own morphology. So the three cases now stand: the dependent many-word string that is not a sentence (2.325); the context-completed single word that is (2.326); and the self-sufficient verb that is (here). Sentencehood is expectancy-closure, and morphology, syntax, and context are three interchangeable means to it.
contested: the source paginates this as a separate karika 326a; some editions integrate it with 326. The doctrine is continuous either way

2.327 unformalized

शब्दव्यवहिता बुद्धिरप्रयुक्तपदाश्रया ।
अनुमानात् तदर्थस्य प्रत्यये हेतुरुच्यते ॥ ३२७ ॥

śabdavyavahitā buddhiraprayuktapadāśrayā |
anumānāt tadarthasya pratyaye heturucyate || 327 ||

A cognition mediated by a word, resting on an unemployed word — that is said to be the cause in the comprehension of its meaning, by way of inference.

The mechanism behind 2.326's context-completion, and it re-uses the kanda's oldest machinery. Sabda-vyavahita buddhih — a cognition *mediated by a word* (vyavahita: the relay of 2.152 and 2.215, where the meaning is reached through an interposed term) — aprayukta-padasraya: resting on a word *not employed*, an unuttered one. Such a cognition is anumanat tad-arthasya pratyaye hetuh ucyate: said to be, *by way of inference*, the cause in the comprehension of its meaning. So the one-word sentence works like this: the unuttered word is present *in the hearer's cognition*, supplied by the occasion, and the uttered word reaches its full sentential meaning through that unuttered relay — as the corrupt form reaches its meaning through the correct one it never says (2.152's sadhu-vyavahita). Anumana is named, and named modestly: this is the inferential channel (2.149), the mediated route whose mediation does not show (2.215's saksad iva). The single word feels complete because the relay is invisible, not because there is no relay.

2.328 unformalized

यस्मिंस्तूच्चरिते शब्दे यदा योऽर्थः प्रतीयते ।
तमाहुरर्थं तस्यैव नान्यदर्थस्य लक्षणम् ॥ ३२८ ॥

yasmiṃstūccarite śabde yadā yo'rthaḥ pratīyate |
tamāhurarthaṃ tasyaiva nānyadarthasya lakṣaṇam || 328 ||

In whatever word, when uttered, whatever meaning is understood at whatever time — that they call the meaning of that very word; there is no other criterion of meaning.

The kanda's semantic first principle, stated at last in its bare form. Yasmin uccarite sabde yada yo 'rthah pratiyate — in whatever word, uttered, at whatever time, whatever meaning is understood — tam ahur artham tasya eva: *that* they call the meaning *of that very word*. And the clause that makes it a principle rather than an observation: na anyad arthasya laksanam — there is *no other criterion* of meaning. Not etymology (2.170-175's rival derivations); not the bare form's citation-response (2.265's rival criterion, refused at 2.314); not the primary sense (2.253's currency-artifact); not the analyst's ledger (2.210's sastrartha). What a word means, on this occasion, is what was understood on this occasion — and there is no appeal beyond comprehension because there is no court above it (2.34's samprataya-pramanatva, stated here as the semantics' foundation rather than as a debating move). The principle is radical and the kanda has earned it over three hundred verses: every attempt to locate meaning anywhere but in the comprehender's flash has been shown to end in ledgers, statistics, or fictions.

2.329 unformalized

क्रियार्थोपपदेष्वेवं स्थानिनां गम्यते क्रिया ।
वृत्तौ निरादिभिश्चैव क्रान्ताद्यर्थोऽवगम्यते ॥ ३२९ ॥

kriyārthopapadeṣvevaṃ sthānināṃ gamyate kriyā |
vṛttau nirādibhiścaiva krāntādyartho'vagamyate || 329 ||

Thus, in the case of the upapadas that have an action for their purpose, the action of the substitutes is understood; and in derivative formation, by 'nir' and the rest, the meaning of 'having gone beyond' and so on is understood.

The principle of 2.328 applied to the grammar's own machinery. Kriyartha-upapadesu — in the case of the *upapadas* (the words that condition an affix's application) that have an action for their purpose — sthaninam gamyate kriya: the action of the *sthanins*, the substituends, is understood: what the grammar's substitution-rules replace is nonetheless what the hearer gets. And in vrtti, derivative formation, nir-adibhih: by 'nir' and the rest, krantady-artho 'vagamyate — the meaning of 'having gone beyond' and so forth is understood: the preverb delivers a sense that no word in the formation states. Both halves say the same thing: the meaning understood is not the meaning any element expresses, and by 2.328 the meaning understood *is the meaning*. The grammar's substituends, its elided elements, its preverbs — all of them contribute contents the surface does not carry, and the science tracks this with rules while the hearer simply comprehends. Which is the kanda's settlement (2.88): the science is the correct engineering of the scaffold, and the flash is the meaning.

2.330 unformalized

तानि शब्दान्तराण्येव पर्याया इव लौकिकाः ।
अर्थप्रकरणाभ्यां तु तेषां स्वार्थो नियम्यते ॥ ३३० ॥

tāni śabdāntarāṇyeva paryāyā iva laukikāḥ |
arthaprakaraṇābhyāṃ tu teṣāṃ svārtho niyamyate || 330 ||

Those are just other words — like worldly synonyms; but their own meaning is restricted by the object and the context.

The movement closes by dissolving the grammar's categories one final time. Tani sabdantarani eva — the elements just discussed (the preverbs, the upapadas, the substitutes) are *just other words*: not a special class with special powers, but words among words. Paryaya iva laukikah: *like worldly synonyms* — as the ordinary language has several words for one thing and lets context choose, so the grammar's elements are alternative vehicles. And the closing clause, which is 2.314 and 2.328 in five words: artha-prakaranabhyam tu tesam svartho niyamyate — but their *own meaning* is restricted by the object and the context. Even here, at the technical heart of the science, meaning is fixed by occasion and not by form. The verse's placement is deliberate: after three hundred verses of grammatical argument, the last word on the grammar's own vocabulary is that it works the way everything else works — one plenitude, carved by nimittas, delivered in a flash to a trained mind. Sabdah loka-nibandhanah (2.297): words are grounded in the world, and the grammarian's words are not exempt.

2.331 · Words, Awakening, and the Real unformalized

प्रतिबोधाभ्युपायास्तु ये तं तं पुरुषं प्रति ।
नावश्यं तेऽभिसंबद्धाः शब्दा ज्ञेयेन वस्तुना ॥ ३३१ ॥

pratibodhābhyupāyāstu ye taṃ taṃ puruṣaṃ prati |
nāvaśyaṃ te'bhisaṃbaddhāḥ śabdā jñeyena vastunā || 331 ||

But those words which are means of awakening with respect to this person or that — they are not necessarily connected with the thing to be known.

A verse that quietly separates two things the tradition kept fused. Words are pratibodha-abhyupayah — *means of awakening*: they wake the flash (2.143, 2.152's transmitted pratibha), and they do so tam tam purusam prati, *with respect to this person and that*: what wakes one hearer leaves another cold, since waking depends on what has ripened in him (2.117's abhyasa). And then: na avasyam te 'bhisambaddhah sabdah jneyena vastuna — they are *not necessarily connected* with the thing to be known. The word that wakes need not be tied to what it wakes you to. This is a large concession from a school whose first kanda made the word-meaning bond eternal (1.23): here the bond is loosened at exactly the point where waking is concerned. What wakes is an occasion, and occasions need no resemblance to what they occasion (the fan and the air, 1.116; the seeds and the ripening, 1.84). The next verse restores what this seems to threaten.

2.332 unformalized

असत्यां प्रतिपत्तौ च मिथ्या वा प्रतिपादने ।
स्वैरर्थैर्नित्यसम्बद्धास्ते ते शब्दा व्यवस्थिताः ॥ ३३२ ॥

asatyāṃ pratipattau ca mithyā vā pratipādane |
svairarthairnityasambaddhāste te śabdā vyavasthitāḥ || 332 ||

And when the comprehension is not real, or when the conveying is false — those various words stand established as eternally connected with their own meanings.

The restoration, and it is precisely calibrated. Grant the worst: asatyam pratipattau — the comprehension is *not real* (the hearer takes a rope for a snake and says so); mithya va pratipadane — the conveying is *false* (the sentence says what is not). Even then: svaih arthaih nitya-sambaddhah te te sabdah vyavasthitah — those words stand established as *eternally connected with their own meanings*. The eternal bond of 1.23 is untouched by falsity, because the bond is between word and meaning, not between word and world. 'Snake' means snake whether or not there is one; the error is in the taking, not in the word (2.285: the word operates as primary even when the meaning determined is utterly contrary). This is what lets the kanda hold two things at once that look incompatible: meaning is fixed by comprehension (2.328) and words are eternally bonded to their meanings (here) — because comprehension is where the meaning lives, and the bond is what makes a word *that* word rather than another. Falsity is a relation to the world; meaning is not.

2.333 unformalized

यथाप्रकरणं द्वारमित्यस्यां कर्मणः श्रुतौ ।
बधान देहि वेत्येतदुपायादधिगम्यते ॥ ३३३ ॥

yathāprakaraṇaṃ dvāramityasyāṃ karmaṇaḥ śrutau |
badhāna dehi vetyetadupāyādadhigamyate || 333 ||

In the hearing of the action 'the door!' — according to context — 'shut it' or 'give it' is understood from the means.

The one-word sentence returns with a perfect specimen. Say dvaram — 'the door!' And what is understood is badhana, shut it, or dehi, give (make) it — yatha-prakaranam, *according to context*: the same syllables, an order to close or an order to construct, and no hearer ever hesitates. Upayat adhigamyate: it is understood *from the means* — from the situation, the standing arrangement, what is to hand and what is wanted. This is 2.312's crows in a single word: the utterance names one thing and means an action nobody said. And note what the verse's structure concedes: the *action* is what is heard — karmanah srutau, in the hearing of the action — though no verb was uttered. What arrives is a complete practical directive; what was said was a noun. Between the two stands nothing but the occasion, and 2.326-327 have already explained how: the unuttered verb is present in the hearer's cognition, and the uttered word reaches its sentential meaning through that invisible relay.

2.334 unformalized

यत्र साधनवृत्तिर्यः शब्दः सत्त्वनिबन्धनः ।
न स प्रधानभूतस्य साध्यस्यार्थस्य वाचकः ॥ ३३४ ॥

yatra sādhanavṛttiryaḥ śabdaḥ sattvanibandhanaḥ |
na sa pradhānabhūtasya sādhyasyārthasya vācakaḥ || 334 ||

Where a word whose operation is upon a means is grounded in a substance — that word is not the expressor of the principal, the to-be-accomplished meaning.

The syntax of subordination, stated as a rule. A word sadhana-vrttih — whose operation is upon a *means*, a karaka — and sattva-nibandhanah, *grounded in a substance* (a noun, whose office is to present a thing, not an action: 2.195's asattva-bhuto bhavah drew the contrast) — na sa pradhana-bhutasya sadhyasya arthasya vacakah: is *not the expressor* of the principal, the *to-be-accomplished* meaning. The nouns do not say what is to be done. Only the verb carries the sadhya (2.48's architecture: nouns orbit, the verb centres), and a sentence's point — its prayojana, which by 2.113 is its meaning — lives in the sadhya. So the nouns, however many and however precise, never deliver the sentence's meaning; they furnish participants for an accomplishment they cannot name. This is the grammatical face of the kanda's holism: even within the sentence's own machinery, the parts that name things are structurally incapable of naming the point.

2.335 unformalized

स्वार्थमात्र प्रकाश्यासौ साकाङ्क्षो विनिवर्तते ।
अर्थस्तु तस्य सम्बन्धी प्रकाशयति सन्निधिम् ॥ ३३५ ॥

svārthamātra prakāśyāsau sākāṅkṣo vinivartate |
arthastu tasya sambandhī prakāśayati sannidhim || 335 ||

Having illuminated only its own meaning, it withdraws, still requiring; but its meaning, being a relatum, illuminates the proximity.

What the subordinate word does instead, described with unusual delicacy. Svartha-matram prakasya asau sakanksah vinivartate: having illuminated *only its own meaning*, it *withdraws* — still sakanksa, still requiring, its expectancy unsatisfied. The noun says its thing and steps back, unfinished. But: arthah tu tasya sambandhi prakasayati sannidhim — its *meaning*, being a relatum, illuminates the *proximity*: the content it delivered, standing in relation, lights up its neighbourhood — the other meanings, the roles they might fill, the shape of the accomplishment they are converging on. The word retires; its meaning goes on working. This is the mechanism of 2.47 (fixity shining forth through mere proximity) given its phenomenology, and it explains why comprehension feels like convergence rather than assembly: each word deposits a light rather than a brick, and the lights overlap until the scene is visible. Vinivartate — 'withdraws' — is the same verb 2.240 used for the meanings that negation cancels; here it is not cancellation but self-effacement (1.54's unnoticed instrument, 1.115's breath dissolving into the phonemes).

2.336 unformalized

परार्थ्यस्याविशिष्टत्वान्न शब्दाच्छब्दसन्निधिः ।
नार्थाच्छब्दस्य सान्निध्यं न शब्दादर्थसन्निधिः ॥ ३३६ ॥

parārthyasyāviśiṣṭatvānna śabdācchabdasannidhiḥ |
nārthācchabdasya sānnidhyaṃ na śabdādarthasannidhiḥ || 336 ||

Since being-for-another is not distinguished, there is no proximity of word from word; nor proximity of the word from the meaning, nor proximity of the meaning from the word.

A dense negative that blocks three mistaken pictures of how proximity works. Parartthyasya avisistatvat — since *being-for-another* is not distinguished: every element in the sentence is for the sake of something else (the nouns for the verb, the verb for the point, the whole for the purpose), and this parartha-status is uniform, admitting no ranking that could make one element the anchor from which others take their bearings. Hence three denials. Na sabdat sabda-sannidhih: no proximity of word from word — words do not locate each other. Na arthat sabdasya sannidhyam: nor is the word's proximity given by the meaning. Na sabdat artha-sannidhih: nor the meaning's by the word. Every attempt to make one term of the system the fixed point from which the rest are placed fails, because all of them are equally for-another. What then locates anything? Only the whole they all serve — the vakyartha, which is not among them (2.143's anyaiva). The verse is 2.269's dissolution of connection restated: nothing in the sentence positions anything else, because positions are conferred by the meaning that has no position in the sentence at all.
contested: the three denials are compressed and their exact targets are debated; we read them as blocking word-to-word, meaning-to-word, and word-to-meaning anchoring respectively — the uniform parartha-premise supports all three equally

2.337 unformalized

नष्टरूपमिवाख्यातमाक्षिप्तं कर्मवादिना ।
यदि प्राप्तं प्रधानत्वं युगपद् भावसत्त्वयोः ॥ ३३७ ॥

naṣṭarūpamivākhyātamākṣiptaṃ karmavādinā |
yadi prāptaṃ pradhānatvaṃ yugapad bhāvasattvayoḥ || 337 ||

If, when principality has been attained by being and substance simultaneously, the finite verb — as though its form were destroyed — is implied by the karma-theorist —

The opening of a movement on the verb's status, and it begins with a difficulty. The karma-vadin — the theorist for whom the action is the sentence's principal (2.1's first view, the akhyata-doctrine) — must handle sentences where bhava and sattva, being and substance, attain principality yugapad, *simultaneously*: where the action and its participant seem equally central. His recourse: the akhyata is nasta-rupam iva aksiptam — the finite verb, *as though its form were destroyed*, is merely *implied*: he supplies a verb that is not there and treats its absence as elision. The strategy has a cost the verse's iva signals: nasta-rupam, form-destroyed — the verb is present in the analysis and absent in the utterance, which is to say it is a posting (2.10's apoddhara) and not a fact. Bhartrhari lets the difficulty stand as the premise of a conditional whose apodosis follows: if the verb can be implied wherever the theory needs it, the theory has stopped being answerable to what is said.

2.338 unformalized

तैस्तु नामसरूपत्वमाख्यातस्योपवर्ण्यते ॥ ३३८ab ॥
अन्वयव्यतिरेकाभ्यां व्यवहारो विभज्यते ॥ ३३८cd ॥

taistu nāmasarūpatvamākhyātasyopavarṇyate || 338ab ||
anvayavyatirekābhyāṃ vyavahāro vibhajyate || 338cd ||

But by those, the finite verb's likeness of form to a noun is described; practice is differentiated by agreement and difference.

The reply, and it returns to the movement's oldest tool. Taih tu nama-sarupatvam akhyatasya upavarnyate: by *those* theorists, the finite verb's *likeness of form* to a noun is described — the verb assimilated to the nominal, the categories blurred to make the analysis run (2.318's nama-akhyata-sarupah, words alike in form with nouns and verbs). And the corrective: anvaya-vyatirekabhyam vyavaharo vibhajyate — *practice* is differentiated by agreement and difference: what settles whether an element is functioning as noun or as verb is not its shape but the distributional facts of use — covary the form, watch the meaning (2.12, 2.166). The verse thus refuses the theorist's assimilation without asserting a rigid categorial divide: the categories are real *as offices in practice*, discoverable by method, and not as shapes. It is 2.104's capacity-doctrine applied to the parts of speech themselves — noun and verb are what words are *doing*, and the doing is read off distribution, not off endings.
contested: the source paginates 338 as two half-verses; we read them as one karika, the second line giving the method by which the first line's assimilation is to be tested

2.339 unformalized

न चापि रूपात् सन्देहे वाचकत्वं निवर्तते ।
अर्धं पशोरिव यथा समर्थ्यात् तद्विकल्प्यते ॥ ३३९ ॥

na cāpi rūpāt sandehe vācakatvaṃ nivartate |
ardhaṃ paśoriva yathā samarthyāt tadvikalpyate || 339 ||

Nor, in a doubt arising from the form, does expressive power withdraw; as with 'half of an animal', so it is optioned by capacity.

A principle about ambiguity, and it is more consequential than its brevity suggests. Na ca api rupat sandehe vacakatvam nivartate: in a doubt arising from the *form* — where the shape leaves it open which word or which office is present — expressive power does *not withdraw*: the utterance still means, and means determinately. The word does not go silent because its form is ambiguous; the context decides (2.314). The illustration: ardham pasoh iva — as with 'half of an animal': the phrase does not name any existing half-beast, and yet it is perfectly determinate in use (the butcher's half, the sacrificer's portion) — samarthyat tad vikalpyate, it is *optioned by capacity*: what could be meant, given what is possible here, settles what is meant. So ambiguity is not a defect that suspends meaning; it is the ordinary condition of a form met before its occasion, and capacity resolves it. This is 2.311's adhanusa (the non-bow that still yields an instrument) given its rule.

2.340-341 unformalized

सर्वं सत्त्वपदं शुद्धं यदि भावनिबन्धनम् ।
संसर्गे च विभक्तोऽस्य तस्यार्थो न पृथग्यदि ॥ ३४० ॥
क्रियाप्रधानमाख्यातं नाम्नां सत्त्वप्रधानता ।
चत्वारि पदजातानि सर्वमेतद्विरुध्यते ॥ ३४१ ॥

sarvaṃ sattvapadaṃ śuddhaṃ yadi bhāvanibandhanam |
saṃsarge ca vibhakto'sya tasyārtho na pṛthagyadi || 340 ||
kriyāpradhānamākhyātaṃ nāmnāṃ sattvapradhānatā |
catvāri padajātāni sarvametadvirudhyate || 341 ||

(340) If every pure substance-word is grounded in being; and if, in combination, its meaning is divided but not separate — (341) the finite verb is action-principal, nouns have substance as principal, and there are four classes of words: all this is contradicted.

A reductio against a doctrine that would collapse the categories, and its conclusion is a catalogue of casualties. Suppose sarvam sattva-padam suddham bhava-nibandhanam — every pure substance-word is grounded in *being*: every noun is at bottom a bhava-word, substance reduced to being (a monism of the verbal). And suppose that in combination its meaning is vibhakta but na prthak — *divided but not separate*. Then: sarvam etad virudhyate — *all this is contradicted*: the doctrine that the akhyata is kriya-pradhana, action-principal; that nouns are sattva-pradhana, substance-principal; and — the third casualty, and the one that stings — catvari pada-jatani, the *four classes of words*: noun, verb, preverb, particle, the ancient fourfold division that Yaska's Nirukta opens with and that every grammarian inherits. Reduce nouns to being and you have unmade the fourfold classification of the tradition's first linguistic text. Bhartrhari does not endorse the reduction; he prices it, and the price is the whole taxonomy. The next verse names the men who paid it.

2.342 unformalized

वाक्यस्य बुद्धौ नित्यत्वमर्थयोगं च लौकिकम् ।
दृष्ट्वा चतुष्ट्वं नास्तीति वार्तक्षौदुम्बरायणौ ॥ ३४२ ॥

vākyasya buddhau nityatvamarthayogaṃ ca laukikam |
dṛṣṭvā catuṣṭvaṃ nāstīti vārtakṣaudumbarāyaṇau || 342 ||

Seeing the eternality of the sentence in the intellect, and its worldly connection with meaning, Vartaksa and Audumbarayana said the fourfold division does not exist.

One of the most valuable verses in the work: an attribution, and to men we know almost nothing else about. Vartaksa and Audumbarayana — the latter named in Yaska's Nirukta as holding that speech's forms are not permanent in the way the grammarians suppose — are here credited with a doctrine of remarkable boldness: catustvam na asti, the *fourfold division does not exist*. Their grounds are given, and they are the kanda's own: vakyasya buddhau nityatvam — the sentence's *eternality in the intellect* (the sentence as the real unit, standing whole in cognition: 2.7, 2.30's antah sabda-tattva); and artha-yogam laukikam — its connection with meaning being *worldly* (2.297's sabdah loka-nibandhanah). Seeing these, they concluded that the parts-of-speech are analytical fictions and the sentence alone is real. Bhartrhari reports them as predecessors, and they are: the whole Vakyakanda is the working-out of what these two apparently asserted. The verse also does honest doxographic service — the kanda's most radical thesis is not claimed as an innovation but traced to a lineage, however thin the record of it now is.
contested: Audumbarayana is known from Nirukta 1.1 and Vartaksa is otherwise obscure; what exactly either held cannot be recovered beyond this verse and Yaska's brief notice, and our gloss draws the connection the karika itself makes

2.343 unformalized

व्याप्तिमांश्च लघुश्चैव व्यवहारः पदाश्रयः ।
लोके शास्त्रे च कार्यार्थं विभागेनैव कल्पितः ॥ ३४३ ॥

vyāptimāṃśca laghuścaiva vyavahāraḥ padāśrayaḥ |
loke śāstre ca kāryārthaṃ vibhāgenaiva kalpitaḥ || 343 ||

And practice resting on words is extensive and light; in the world and in the science alike, it is constructed by division, for the sake of the operation.

Why the fourfold division survives its refutation. Vyavahara pada-asrayah — practice resting on *words* (rather than on undivided sentences) is vyaptiman ca laghus ca eva: *extensive* and *light* — it covers everything and costs little: with a few hundred stems and a handful of affixes one can say anything, whereas an inventory of sentences would be infinite. And its status: loke sastre ca karyartham vibhagena eva kalpitah — in the world *and* in the science, it is *constructed by division*, for the sake of the operation. Kalpita — constructed (2.10's postings); karyartham — for the sake of getting things done. So word-based practice is a fiction with the two virtues that matter in an instrument: generality and economy — laghava, the very virtue Panini's system is built to maximize (2.176's brevity). The verse is Bhartrhari's answer to the obvious objection against Audumbarayana: if the parts of speech are unreal, why does everyone use them? Because they work, and working is what fictions are for (2.238's untrue path).

2.344 unformalized

न लोके प्रतिपत्तॄणामर्थयोगात् प्रसिद्धयः ।
तस्मादलौकिको वाक्यादन्यः कश्चिन्न विद्यते ॥ ३४४ ॥

na loke pratipattṝṇāmarthayogāt prasiddhayaḥ |
tasmādalaukiko vākyādanyaḥ kaścinna vidyate || 344 ||

In the world, comprehenders' establishments do not arise from connection with the object; therefore there is nothing non-worldly, other than the sentence.

The conclusion of the movement, and it is a strange and powerful one. Na loke pratipattrnam artha-yogat prasiddhayah: in the world, comprehenders' *establishments* — their settled takings, their prasiddhis — do not arise *from connection with the object*: what a community settles into meaning is not fixed by how things are (2.296 gave prasiddhi the verdict on the real; here prasiddhi is shown to owe nothing to the real). Tasmat alaukiko vakyat anyah kascin na vidyate: *therefore* there is nothing non-worldly other than the sentence. The inference is compressed and its force is this: if the world's semantic settlements float free of the objects, then everything below the sentence — the words, their meanings, the categories — is thoroughly *laukika*, a creature of settlement and currency; and the only item that is *not* a settlement, the only thing that stands as it is regardless of what anyone has agreed, is the sentence itself, standing whole in cognition (2.342's nityatvam buddhau). The sentence is the sole non-conventional item in language. Everything else is what we have agreed to see in it.
contested: alaukikah is doing surprising work here — elsewhere in the kanda it names what falls outside practice (2.297's seer's cognition, excluded from vyavahara); here the sentence is called the only alaukika, which we read as 'the only item not constituted by worldly settlement'. The tension with 2.297 is real and may be deliberate

2.345 unformalized

अन्यत्र श्रूयमाणैश्च लिङ्गैर्वाक्येषु सूचिताः ।
स्वार्था एव प्रतीयन्ते रूपाभेदादलक्षिताः ॥ ३४५ ॥

anyatra śrūyamāṇaiśca liṅgairvākyeṣu sūcitāḥ |
svārthā eva pratīyante rūpābhedādalakṣitāḥ || 345 ||

And the own-meanings, indicated in sentences by marks heard elsewhere, are themselves understood — unnoticed, because of the non-difference of form.

A mechanism-note on how the sentence supplies what it does not say. Svarthah eva pratiyante — the *own-meanings* are themselves understood — sucitah lingaih anyatra sruyamanaih: *indicated* by marks *heard elsewhere*, in other sentences, other contexts (the intertextual linga of 2.86). And the crucial clause: rupa-abhedad alaksitah — *unnoticed*, because of the *non-difference of form*: the indication does its work invisibly, since nothing in the shape marks the indicated content as having come from outside. So a sentence delivers meanings whose warrant lies in utterances the hearer met long ago, and delivers them seamlessly, as though they were in the words. This is 2.92's fallacy (projecting other sentences' meanings into this one) seen from the other side — as a *competence* rather than an error: the same transfer-by-resemblance that misleads the provincial is what lets any of us understand anything, since every word we hear is met with the residue of every prior hearing. The kanda has consistently refused to treat this as either simply veridical or simply erroneous. It is the untrue path (2.238), and it is the only road.

2.346 unformalized

उत्सर्गवाक्ये यत्त्यक्तमशब्दमिव शब्दवत् ।
तद् बाधकेषु वाक्येषु श्रुतमन्यत्र गम्यते ॥ ३४६ ॥

utsargavākye yattyaktamaśabdamiva śabdavat |
tad bādhakeṣu vākyeṣu śrutamanyatra gamyate || 346 ||

What is abandoned in a general rule's sentence — as if wordless, though word-like — that, heard in the annulling sentences, is understood elsewhere.

The doctrine applied to the structure of rule-systems. In an utsarga-vakya, a general rule, something is tyaktam — *abandoned*, left out: asabdam iva sabdavat, as if wordless though word-like — present in the rule's sense without any word for it (the exception the general rule silently reserves). And: tad badhakesu vakyesu srutam anyatra gamyate — *that*, heard in the *annulling* sentences (the exceptions, the apavadas that override the general), is understood *elsewhere* — i.e. the general rule's unstated reservation is what the exception-sentences make explicit, and reading them, one understands retroactively what the general rule was always meaning. Rule-systems are thus *distributed* sentences: no single formulation carries its own content, and each is completed by others it does not mention. This is 2.76's maha-vakya doctrine (the discourse as the real unit) applied to the sastra itself — Panini's grammar is one sentence, and every sutra is a word in it.

2.347 unformalized

ब्राह्मणानां श्रुतिर्दन्धि प्रक्रान्ता माठरं विना ।
माठरस्तक्रसम्बन्धात् तत्राचष्टे यथार्थताम् ॥ ३४७ ॥

brāhmaṇānāṃ śrutirdandhi prakrāntā māṭharaṃ vinā |
māṭharastakrasambandhāt tatrācaṣṭe yathārthatām || 347 ||

'Curds for the brahmins' — the text is commenced, excluding Mathara; Mathara, from his connection with buttermilk, there declares its actual meaning.

The mechanism of 2.346 in a worked scene, and it is a small comedy. Brahmananam srutir dadhi prakranta — the text says 'curds for the brahmins'; and mathara vina, *excluding Mathara* — one brahmin gets buttermilk instead. Now: matharas takra-sambandhat tatra acaste yatharthataam — *Mathara*, from his connection with the buttermilk, *declares the actual meaning* there. The exception reveals the rule. Until Mathara's buttermilk, 'curds for the brahmins' looked universal; his exclusion shows that the general provision always meant 'curds for the brahmins other than Mathara' — the reservation was in it from the start, unworded (2.346's asabdam iva). And the pleasant irony the verse leans on: it is the *excluded* member who tells you what the rule says. The one left out is the interpreter. This is the apavada-doctrine (exception overrides general) recast as semantics: exceptions do not defeat general sentences, they *disclose* them, and a rule's meaning is not available until you know what it excludes — which is 2.344's point about words in another key.
contested: the Mathara-example is from the exegetical tradition's stock and its details vary; nothing turns on the specific figure

2.348 unformalized

अनेकाख्यातयोगेऽपि वाक्यं न्याय्यापवादयोः ।
एकमेवेष्यते कैश्चिद्भिन्नरूपमिव स्थितम् ॥ ३४८ ॥

anekākhyātayoge'pi vākyaṃ nyāyyāpavādayoḥ |
ekameveṣyate kaiścidbhinnarūpamiva sthitam || 348 ||

Even where there is connection of several finite verbs, the sentence of the general rule and its exception is held by some to be one alone, though standing as if of divided form.

The unity of rule-and-exception, stated as doctrine. Aneka-akhyata-yoge 'pi — even where *several* finite verbs are connected (the general rule has its verb, the exception has its own; by every formal criterion, two sentences) — nyayya-apavadayoh vakyam ekam eva isyate kaiscid: the sentence of the general and the exception is held by some to be *one alone* — bhinna-rupam iva sthitam, *standing as if* of divided form. The iva is doing the work it always does in this kanda: the division is apparent, the unity real. So a general rule and its exception, however far apart in the text, however separately verbed, constitute a single sentence — because a single expectancy closes across them (2.4: the sentence is where requirement is satisfied; the general rule alone is sakanksa, still requiring, until the exception limits it). This is the most extreme extension of the vakya-concept in the work: sentencehood spanning discontinuous text, indifferent to verb-count, defined solely by where the meaning completes.

2.349 unformalized

नियमः प्रतिषेधश्च विधिशेषस्तथा सति ।
द्वितीये यो लुगाख्यातस्तच्छेषमलुकं विदुः ॥ ३४९ ॥

niyamaḥ pratiṣedhaśca vidhiśeṣastathā sati |
dvitīye yo lugākhyātastaccheṣamalukaṃ viduḥ || 349 ||

Restriction, prohibition, and likewise the remainder of the injunction — that being so, what has elision of the finite verb in the second is known as its remainder, without elision.

The technical consequences of 2.348's unity. If rule and exception are one sentence, then the exception's various functions — niyama, *restriction* (narrowing the general); pratisedha, *prohibition* (barring a case); vidhi-sesa, the *remainder of the injunction* (completing what the general left open) — are all articulations of one meaning rather than independent assertions. And a grammatical marker of the unity is offered: dvitiye yo lug akhyatah tac-chesam alukam viduh — what has *elision of the finite verb* in the second member is known as *its remainder without elision*: the exception's verb may be dropped precisely because the general's verb is doing duty for both — the ellipsis is possible because the two are one sentence (2.327's unuttered word supplied by the cognition, here supplied by the other half of the same sentence). Elision is thus evidence of unity: you can only omit what the whole already carries.

2.350 unformalized

निराकाङ्क्षाणि निर्वृत्तौ प्रधानानि परस्परम् ।
तेषामनुपकारित्वात् कथं स्यादेकवाक्यता ॥ ३५० ॥

nirākāṅkṣāṇi nirvṛttau pradhānāni parasparam |
teṣāmanupakāritvāt kathaṃ syādekavākyatā || 350 ||

They are requirement-free in their accomplishment, and principal with respect to one another; since they are of no assistance to each other, how could there be single-sentence-hood?

The objection to 2.348, put with full force, and the movement ends on it. Consider the general rule and the exception as sentences: nirakanksani nirvrttau — each is *requirement-free* in its own accomplishment: 'curds for the brahmins' is a complete provision; 'buttermilk for Mathara' is another. Pradhanani parasparam: and they are *principal* with respect to one another — neither is subordinate to the other; each stands as its own injunction, not as a member of the other. Hence: tesam anupakaritvat katham syad eka-vakyata — since they are *of no assistance* to each other, how could there be single-sentence-hood? The objection is exact and it uses the kanda's own criteria: unity requires expectancy, expectancy requires incompleteness, and two complete injunctions have none to offer each other (2.325: dependence is what demotes a sentence to a word). Bhartrhari has stated the difficulty at its strongest and left it standing here — the movement's next verses take it up. It is characteristic: the hardest objection to his own position gets the best statement in the text, and it gets it from him.

2.351 unformalized

विशेषविधिनाऽथिंत्वाद् वाक्यशेषोऽनुमीयते ।
विशेषवन्नवर्त्येऽर्थे तस्मात्तुल्यं व्यपेक्षणम् ॥ ३५१ ॥

viśeṣavidhinā'thiṃtvād vākyaśeṣo'numīyate |
viśeṣavannavartye'rthe tasmāttulyaṃ vyapekṣaṇam || 351 ||

Because of the need created by the particular rule, the sentence-remainder is inferred; since the meaning is to be excluded as qualified, the mutual requirement is therefore equal.

The answer to 2.350's objection, and it turns on where the need arises. The objection was that a general rule and its exception are each complete and so cannot require each other. The reply: visesa-vidhina arthitvat — *because of the need created by the particular rule* — vakya-sesah anumiyate: the sentence-remainder is *inferred*. The exception generates the requirement retroactively: once 'buttermilk for Mathara' is heard, 'curds for the brahmins' is no longer complete as it stood — it now needs its reservation, and the reservation was always its own (2.346's asabdam iva, the unworded exclusion). And the ground: visesavat nivartye 'rthe — since the meaning to be *excluded* is itself qualified, particular — tasmat tulyam vyapeksanam: the mutual requirement is therefore *equal*. Each half needs the other, symmetrically: the general needs the exception to be determinate, the exception needs the general to have something to except from. So nirakanksata (2.350's premise) was an illusion of reading the general rule alone — completeness that only looked complete because the reader had not yet met what limits it.

2.352 · Names, Their Parts, and the Sciences unformalized

संज्ञाशब्दैकदेशो यस्तस्य लोपो न विद्यते ।
विशिष्टरूपा सा संज्ञा कुता च न निवर्तते ॥ ३५२ ॥

saṃjñāśabdaikadeśo yastasya lopo na vidyate |
viśiṣṭarūpā sā saṃjñā kutā ca na nivartate || 352 ||

That which is one part of a name-word — of it there is no elision; that name, having a distinct form, does not withdraw from anywhere.

A new movement opens, on proper names and their internal parts, and its first rule is severe. Samjna-sabda-ekadesah — a *part* of a name-word ('datta' inside 'Devadatta', 'agni' inside 'Agnidatta') — tasya lopo na vidyate: of it there is *no elision*: you cannot drop it, because it is not a semantically detachable element that the whole could survive losing. And: visista-rupa sa samjna kutas ca na nivartate — that name, having a *distinct form*, does not *withdraw* from anywhere: the name is one form, self-identical, and no part of it can be treated as an ingredient that might be subtracted or replaced. This is 2.220's conclusion (the complex alone is the ground of the distinct meaning) applied to names, and names are the ideal case: nobody thinks 'Devadatta' means god-given, everybody knows whom it means, and the members are visibly idle. The movement will use names as a controlled experiment in what parts can and cannot do.

2.353 unformalized

संज्ञान्तराच्च दत्तादेर्नान्या संज्ञा प्रतीयते ।
संज्ञिनं देवदत्ताख्यं दत्तशब्दः कथं वदेत् ॥ ३५३ ॥

saṃjñāntarācca dattādernānyā saṃjñā pratīyate |
saṃjñinaṃ devadattākhyaṃ dattaśabdaḥ kathaṃ vadet || 353 ||

And from another name — 'Datta' and the like — no other name is understood; how should the word 'Datta' express the name-bearer called Devadatta?

The rule of 2.352 argued from the obvious. Samjnantarat dattadeh na anya samjna pratiyate: from 'Datta' — which is itself a name, a perfectly good one, borne by other men — no *other* name is understood. And the question that makes it vivid: samjninam devadattakhyam datta-sabdah katham vadet — how should the word 'Datta' express the name-bearer *called Devadatta*? Call for Datta and Devadatta does not turn. The two names share a form and share nothing else; the presence of 'datta' in 'Devadatta' gives you no purchase on Devadatta whatever. This is kupa-supa-yupa (2.169) in the domain where the intuition is strongest: phonetic containment is semantically inert, and names make it undeniable because names have no penumbra of associated sense to blur the test. If parts contributed, 'Datta' would half-summon Devadatta. It does not summon him at all.

2.354 unformalized

सर्वैरवयवैस्तुल्यं सम्बन्धं समुदायवत् ।
केचिच्छब्दस्वरूपाणां मन्यन्ते सर्वसंज्ञिभिः ॥ ३५४ ॥

sarvairavayavaistulyaṃ sambandhaṃ samudāyavat |
kecicchabdasvarūpāṇāṃ manyante sarvasaṃjñibhiḥ || 354 ||

Some consider the connection of the own-forms of words with all their name-bearers to be equal, by all their parts, as with the aggregate.

The rival account, stated fairly. Kecid manyante: some hold that the connection of sabda-svarupani — the own-forms of words — with sarva-samjnibhih, *all* their name-bearers, is tulyam sarvaih avayavaih: *equal, by all their parts* — samudayavat, as with the aggregate. That is: the whole name and each of its parts are equally connected to the named; the connection distributes. On this view 'datta' *is* connected to Devadatta, just as 'Devadatta' is — the parts being co-owners of the reference. The position is not silly: it explains why we feel that 'Agnidatta' has something to do with fire, why patronymics work, why name-elements recur meaningfully across a family. And it is exactly what 2.353 has just denied. Bhartrhari's practice throughout has been to let the rival state its best case immediately after the datum that embarrasses it — the reader is meant to hold both and feel the pressure, which the next verses release.

2.355 unformalized

वर्णानामर्थंवत्त्वं तु संज्ञानां संज्ञिभिर्भवेत् ।
सम्बद्धोऽवयवः संज्ञाप्रविवेके न कल्प्यते ॥ ३५५ ॥

varṇānāmarthaṃvattvaṃ tu saṃjñānāṃ saṃjñibhirbhavet |
sambaddho'vayavaḥ saṃjñāpraviveke na kalpyate || 355 ||

But the meaningfulness of phonemes would be as that of names with their name-bearers; a connected part is not construed in the analysis of a name.

The reductio of 2.354. If the parts of a name are connected to the named as the whole is, then by parity varnanam arthavattvam — the *phonemes'* meaningfulness — would be samjnanam samjnibhih bhavet: on the same footing as names' connection with their bearers: the 'd' of 'Devadatta' would name him too, and every phoneme would be a name. The regress is the one 2.28-29 ran against linguistic atomism, and it lands here with the same force. Hence the rule: sambaddho 'vayavah samjna-pravive ke na kalpyate — a *connected part* is not construed in the *analysis* of a name: when the grammar analyses a name into elements, it does not thereby credit those elements with the connection the name has. The analysis divides; the connection does not. This is the movement's version of 2.88's settlement (abstraction according to capacities, the practice preserved, the constitution untouched) applied where it is cleanest.

2.356 unformalized

सर्वस्वरूपैर्युगपत् सम्बन्धे सति संज्ञिनः ।
नैकदेशसरूपभ्यस्तत्प्रत्यायनसम्भवः ॥ ३५६ ॥

sarvasvarūpairyugapat sambandhe sati saṃjñinaḥ |
naikadeśasarūpabhyastatpratyāyanasambhavaḥ || 356 ||

When there is connection of the name-bearer with all the own-forms simultaneously, there is no possibility of conveying it from things that are alike to one part.

The rival's position pressed to its collapse. Suppose the samjnin is connected sarva-svarupaih yugapat — with *all* the own-forms *simultaneously*: every part of the name bonded to the bearer at once. Then: na ekadesa-sarupebhyah tat-pratyayana-sambhavah — there is *no possibility* of conveying him from things alike to *one part*. The reasoning: if the bond requires all parts at once, then anything sharing only a part shares none of the bond — so 'Datta' cannot even partially convey Devadatta (which 2.353 observed), but neither can any partial resemblance ever convey anything, and the phenomenon the rival was invoking (the felt pull of name-elements) is unexplained by the very doctrine invoked to explain it. The rival wanted parts to carry reference and has ended with parts carrying nothing. The next verse gives the correct account of what the parts actually do.

2.357 unformalized

एकदेशात्तु सङ्घाते कंषाञ्चिज्जायते स्मृतिः ।
स्मृतेस्तु विषयाच्छब्दात् सङ्घातार्थः प्रतीयते ॥ ३५७ ॥

ekadeśāttu saṅghāte kaṃṣāñcijjāyate smṛtiḥ |
smṛtestu viṣayācchabdāt saṅghātārthaḥ pratīyate || 357 ||

But from one part, in the case of some, a memory arises with respect to the aggregate; and from the word that is the object of the memory, the aggregate's meaning is understood.

The right account, and it is the kanda's standing mechanism. Ekadesat tu sanghate kesamcit jayate smrtih: from *one part*, in the case of some (hearers, cases), a *memory* arises with respect to the *aggregate* — hearing 'datta', one is put in mind of 'Devadatta'. And then: smrteh tu visayat sabdat sanghatartah pratiyate — from the *word that is the object of that memory* — i.e. from the whole name now recalled — the aggregate's meaning is understood. The relay is exact: part → memory of whole → meaning of whole. The part never touches the meaning; it touches the *whole word*, which touches the meaning. So the felt pull of name-elements (2.354's datum) is real and is a fact about recollection, not about reference — 2.315-316's visesa-smrti-hetavah, the causes of the recollection of the particular, operating inside a name. The parts are mnemonic triggers, and the trigger is not the gun.

2.358 unformalized

एकदेशात् स्मृतिर्भिन्ने सङ्घाते जायते कथम् ।
कथं प्रतीयमानः स्याच्छब्दोऽर्थस्याभिधायकः ॥ ३५८ ॥

ekadeśāt smṛtirbhinne saṅghāte jāyate katham |
kathaṃ pratīyamānaḥ syācchabdo'rthasyābhidhāyakaḥ || 358 ||

How does a memory arise from one part with respect to a distinct aggregate? And how could a word that is being cognized be the expressor of a meaning?

Two objections to 2.357, and they are good ones. First: ekadesat smrtih bhinne sanghate jayate katham — how does a memory arise from one part with respect to a *distinct* aggregate? 'Datta' is a part of many names; how does it recall *this* one rather than another? (The answer is implicit in 2.357's kesamcit — 'in the case of some': the recall is occasion-relative, context supplying what the part cannot, per 2.315-316's fourteen factors.) Second, and deeper: katham pratiyamanah syat sabdo 'rthasya abhidhayakah — how could a word that is *being cognized* be the expressor of a meaning? If the recalled whole is itself an object of cognition — something the hearer is aware *of* — then it stands as content, not as vehicle; and a content does not express. This is 2.57's problem (the word must be grasped to convey, yet what is grasped is object not instrument) returning at the movement's heart, and it has no easy answer. Bhartrhari raises it against his own account, which is his habit.

2.359 unformalized

एकदेशसरूपास्तु तैस्तैर्लिङ्गैः समन्विताः ।
अनुनिष्पादिनः शब्दाः संज्ञासु समवस्थिताः ॥ ३५९ ॥

ekadeśasarūpāstu taistairliṅgaiḥ samanvitāḥ |
anuniṣpādinaḥ śabdāḥ saṃjñāsu samavasthitāḥ || 359 ||

But words that are alike to one part, accompanied by these and those marks, arising-in-consequence, stand settled in names.

The reply to 2.358's first objection. Ekadesa-sarupah tu sabdah — words that are *alike to one part* ('datta' as it stands in 'Devadatta') — taih taih lingaih samanvitah: accompanied by *these and those marks* — the indicators that context supplies (2.315's linga) — anunispadinah: *arising-in-consequence* (2.121's term: the entailed, not the expressed) — samjnasu samavasthitah: stand settled *in* the names. So the part-like word inside a name is not a word doing word-work; it is an anunispadin, something that comes along, marked by whatever the occasion marks it with, lodged in the name without office. And the resolution of 2.358's first question falls out: the memory reaches *this* aggregate rather than another because the marks accompanying the part are these and not those — the linga discriminates where the part cannot. The vocabulary is precise: what lives inside a name is not a constituent but a resident.

2.360 unformalized

साधारणत्वात् संदिग्धाः सामर्थ्यान्नियताश्रयाः ।
तेषां ये साधवस्तेषु शास्त्रे लोपादि शिष्यते ॥ ३६० ॥

sādhāraṇatvāt saṃdigdhāḥ sāmarthyānniyatāśrayāḥ |
teṣāṃ ye sādhavasteṣu śāstre lopādi śiṣyate || 360 ||

Being common, they are doubtful; by capacity, they have a fixed substratum; among them, those that are correct — for those, elision and the rest is taught in the science.

The status of these residents, in one verse. Sadharanatvat sandigdhah — being *common* (the same syllables occurring in many names), they are *doubtful*: taken alone, no part tells you which name it belongs to. Samarthyat niyatasrayah — by *capacity*, they have a *fixed substratum*: what they can do, in a given case, fixes what they belong to (2.339's samarthyat vikalpyate: capacity resolving ambiguity). And then the science's interest: tesam ye sadhavah tesu sastre lopadi sisyate — among them, *those that are correct*: for *those*, elision and the rest is taught in the science. Here is the sharp point. The grammar's operations (elision, augment, sandhi) apply to some of these residents and not others — and the criterion is sadhutva, *correctness* (1.27's dharma-bearing property). So the parts inside names are sorted by the science into those it will operate on and those it will not, and the sorting is a posting of the sastra, not a discovery about the name. Which parts of your name the grammar recognizes is a fact about the grammar.

2.361 unformalized

तुल्यायामनुनिष्पत्तौ ज्ये द्रा घा इत्यसाधवः ।
नह्यन्वाख्यायकं शास्त्रे तेषु दत्तादिवत् स्मृतिः ॥ ३६१ ॥

tulyāyāmanuniṣpattau jye drā ghā ityasādhavaḥ |
nahyanvākhyāyakaṃ śāstre teṣu dattādivat smṛtiḥ || 361 ||

The consequence being equal, 'jye', 'dra', 'gha' are incorrect; for there is no analytic teaching in the science with respect to them, as there is a memory in the case of 'datta' and the like.

The sorting of 2.360 exhibited. Tulyayam anunispattau — the *consequence being equal*: 'jye', 'dra', 'gha' arise inside names exactly as 'datta' does, by the same mechanism, with the same standing as residents — and yet asadhavah, they are *incorrect*: the grammar will not operate on them. Why? Na hi anvakhyayakam sastre tesu — there is no *analytic teaching* in the science with respect to them — dattadivat smrtih: as there *is* a memory (a recollection-relation, 2.357) in the case of 'datta' and the like. The difference is not in the syllables and not in their mode of occurrence; it is that 'datta' is independently a word (so a memory can reach a whole through it, 2.357) and 'jye' is not (so no memory, no analytic treatment, no correctness). The verse thus completes the movement's demonstration: what makes a part of a name analysable is not its being a part but its being, elsewhere, a word — which is to say, the analysis of names borrows entirely from the lexicon and finds nothing in the name itself.
contested: the three specimens are our reading of the transcription; the class intended — syllable-sequences occurring in names that the grammar declines to treat as elements — is clear from the contrast with dattadi

2.362 unformalized

कृतणत्वाश्च ये शब्दाः नित्यं खरणसादयः ।
एकद्रव्योपदेशित्वात्तान् साधून् संप्रचक्षते ॥ ३६२ ॥

kṛtaṇatvāśca ye śabdāḥ nityaṃ kharaṇasādayaḥ |
ekadravyopadeśitvāttān sādhūn saṃpracakṣate || 362 ||

And those words which have retroflexion effected, permanently — 'kharanasa' and the like — because they teach a single substance, they call correct.

A further class, and its criterion is startling. Some words have krta-natva nityam — retroflexion *permanently* effected: the sandhi-change is baked into the word, no longer derivable, simply part of its shape (kharanasa, 'ass-nosed', where the n has become n and stays so). And the ground on which the science calls them sadhu: eka-dravya-upadesitvat — *because they teach a single substance*: the word names one thing, individually, as a name does. So correctness here is conferred by *referential singularity* — the word is proper, it picks out one, and therefore its frozen phonology is licensed rather than corrupt. The criterion is worth pausing on. Sadhutva was defined in Kanda I as a dharma-bearing property established by the sistas' usage (1.27, 1.141), and here it turns out to track whether a form names an individual. Names get grammatical amnesty. What would be a corruption in a common noun is correct in a proper one, because the proper noun's business is done.

2.363 unformalized

गोत्रण्येव तु तान्याहुः संज्ञाशक्तिसमन्वयात् ।
निमित्तापेक्षणं तेषु स्वार्थे नावश्यमिष्यते ॥ ३६३ ॥

gotraṇyeva tu tānyāhuḥ saṃjñāśaktisamanvayāt |
nimittāpekṣaṇaṃ teṣu svārthe nāvaśyamiṣyate || 363 ||

But they call those very things gotra-names, from their connection with the capacity of a name; among them, dependence on an occasioning ground is not necessarily accepted in respect of their own meaning.

The doctrine of 2.362 given its category. Those forms are called gotranti — gotra-names, lineage-names (the patronymics and clan-designations, which the grammar treats as a class) — samjna-sakti-samanvayat: from their connection with the *capacity of a name*. And the consequence that matters: nimitta-apeksanam tesu svarthe na avasyam isyate — among them, dependence on an *occasioning ground* is *not necessarily* accepted in respect of their own meaning. A gotra-name does not need a reason. 'Garga' does not have to mean anything about the Gargas; it names them, and the naming needs no ground beyond itself. This is the limit-case of the whole kanda's semantics: everywhere else meaning is carved by nimittas (2.250-251, 2.314-316), and here is a class of words that mean without any nimitta at all — pure reference, ungrounded, working perfectly. Names are where language touches things without saying anything about them.

2.364 unformalized

व्यवहाराय नियमः संज्ञानां संज्ञिनि कचित् ।
नित्य एव तु सम्बन्धो डित्थादिषु गवादिवत् ॥ ३६४ ॥

vyavahārāya niyamaḥ saṃjñānāṃ saṃjñini kacit |
nitya eva tu sambandho ḍitthādiṣu gavādivat || 364 ||

The restriction of names upon a name-bearer is sometimes for the sake of practice; but the connection is eternal indeed, in 'Dittha' and the like, as in 'cow' and the like.

The two-level account of naming, and it resolves an apparent tension. Vyavaharaya niyamah samjnanam samjnini kvacit: the *restriction* of names upon their bearers is *sometimes for the sake of practice* — this child is called Dittha because we needed to call him something; the assignment is conventional, dated, ours. But: nitya eva tu sambandhah ditthadisu gavadivat — the *connection* is *eternal indeed* in 'Dittha' and the like, as in 'cow' and the like. Both are true and they are about different things. The *assignment* (which bearer gets which name) is practice; the *connection* (that this word means this) is eternal, exactly as with 'gauh' — the nitya sambandha of 1.23 holding for proper names no less than common ones. Naming a child does not create a word-meaning bond; it deploys one. Which is why the reference works the instant it is conferred, and why 'Dittha' does not have to be learned twice.

2.365 unformalized

वृद्ध्यादीनां च शास्त्रेस्मिञ्छक्त्यवच्छेदलक्षणः ।
अकृत्रिमोऽभिसम्बन्धो विशेषणाविशेष्यवत् ॥ ३६५ ॥

vṛddhyādīnāṃ ca śāstresmiñchaktyavacchedalakṣaṇaḥ |
akṛtrimo'bhisambandho viśeṣaṇāviśeṣyavat || 365 ||

And in this science, the connection of 'vrddhi' and the rest, whose mark is the delimitation of capacity, is non-artificial — like that of qualifier and qualified.

The doctrine of 2.364 extended to the grammar's own technical terms, and this is a substantial claim. Vrddhy-adinam ca sastre 'smin — of 'vrddhi' and the rest in *this science* (Panini's technical vocabulary: vrddhi, guna, ti, ghu, the artificial names the Astadhyayi coins) — sakty-avaccheda-laksanah abhisambandhah akrtrimah: the connection, whose mark is the *delimitation of capacity*, is *non-artificial* — visesana-visesyavat, like that of qualifier and qualified. The claim is that even coined technical terms are not arbitrary labels: the connection between 'vrddhi' and the vowels it names is akrtrima, not made — it is a delimitation of the word's own capacity (2.250's sarvarthya carved by nimitta, here by the science's definitional act). This is 1.59's point (grammar's technical terms rest on svarupa) given its semantics: the sastra does not invent bonds, it *selects* from the plenitude. Even the most artificial-looking naming in the language is a carving, not a fabrication.

2.366 unformalized

संज्ञास्वरूपमाश्रित्य निमित्ते सति लौकिकी ।
काचित् प्रवर्तते काचिन्निमित्तासंनिधावपि ॥ ३६६ ॥

saṃjñāsvarūpamāśritya nimitte sati laukikī |
kācit pravartate kācinnimittāsaṃnidhāvapi || 366 ||

Resting on the own-form of the name, some worldly (name) operates when the occasioning ground is present; some operates even in the absence of the ground.

The taxonomy of names by their dependence on grounds. Samjna-svarupam asritya — resting on the name's own form — laukiki kacit pravartate nimitte sati: some *worldly* name operates when the occasioning *ground is present*: 'the cook' names him while he cooks; remove the cooking and the name lapses (2.158's criterion: the word's employment ceases when its meaning departs). Kacid nimitta-asannidhav api: some operates *even in the absence* of the ground — 'the cook' as a man's name, sticking to him at rest, in old age, in his grave. The same form, two regimes: ground-dependent and ground-free, and only usage tells you which is running. This is the distinction the tradition would formalize as yaugika (etymologically grounded) versus rudha (conventionalized), and its interest here is that both are *laukiki*, worldly: the difference is not between a real meaning and a lapsed one but between two ways the plenitude is carved — by a live ground, or by settled currency (2.253's prasiddhi doing the work once the ground has gone).

2.367 unformalized

शास्त्रे तु महती संज्ञा स्वरूपोपनिबन्धना ।
अनुमानं निमित्तस्य संनिधाने प्रतीयते ॥ ३६७ ॥

śāstre tu mahatī saṃjñā svarūpopanibandhanā |
anumānaṃ nimittasya saṃnidhāne pratīyate || 367 ||

But in the science, a great name is grounded in the own-form; the inference of the occasioning ground is understood in the case of proximity.

The scientific names contrasted with the worldly ones of 2.366. Sastre tu mahati samjna svarupa-upanibandhana — in the science, a *great* name (the major technical terms) is grounded in the *own-form*: it means by exhibiting its shape (1.59-68's svam rupam, the grammar's whole metalinguistic apparatus). And: anumanam nimittasya sannidhane pratiyate — the *inference* of the occasioning ground is understood in the case of *proximity*: where the term is used near its occasion, one can infer why it was coined so. But only infer, and only there. The technical term does not carry its ground; the ground is reconstructible, sometimes, from context — which is 2.170's etymological under-determination in the sastra's own vocabulary: the grammarians dispute why Panini chose 'ghu' or 'ti' precisely because the terms do not say. Their meaning is their delimitation (2.365); their motivation is guesswork.
contested: mahati samjna ('a great name') is obscure — we take it as the science's major technical terms; some read it as 'a name of wide scope' (covering many items), which fits svarupa-nibandhana less well

2.368 unformalized

आवृत्तेरनुमानं वा सारूप्यात्तत्र गम्यते ।
शब्दभेदानुमानं वा शक्तिभेदस्य वा गतिः ॥ ३६८ ॥

āvṛtteranumānaṃ vā sārūpyāttatra gamyate |
śabdabhedānumānaṃ vā śaktibhedasya vā gatiḥ || 368 ||

Or an inference from repetition is understood there, from likeness of form; or an inference of a difference of words, or the ascertaining of a difference of capacity.

The alternatives for what the 'inference' of 2.367 actually delivers, and the disarray is familiar. Avrtteh anumanam va — an inference from *repetition*: seeing the term recur, one infers its ground from the pattern — sarupyat tatra gamyate, understood there from likeness of form. Or sabda-bheda-anumanam va — an inference of a *difference of words*: the recurrence is not one term but two homonyms, and one infers the split. Or sakti-bhedasya va gatih — or the *ascertaining of a difference of capacity*: one term, two capacities, carved by the contexts (2.250's doctrine). Three accounts of what the grammarian is doing when he reasons from a technical term's occurrences, and they disagree about the most basic question — how many terms are there? This is 2.259's disarray (how many verses in the repetition?) in the metalanguage, and the diagnosis is the same: identity and number are conferred by office (2.104, 2.258), and where the office is disputed, the count is.

2.369 unformalized

क्वचिद्विषयभेदेन कृत्रिमा व्यवतिष्ठते ।
संख्यायामेकविषयं व्यवस्थानं द्वायोरपि ॥ ३६९ ॥

kvacidviṣayabhedena kṛtrimā vyavatiṣṭhate |
saṃkhyāyāmekaviṣayaṃ vyavasthānaṃ dvāyorapi || 369 ||

Sometimes the artificial one stands settled by difference of domain; in the case of number, the settling has a single domain, even for both.

How artificial and worldly names divide their territory. Kvacid visaya-bhedena krtrima vyavatisthate: sometimes the *artificial* (technical) name stands settled by *difference of domain* — the technical sense and the ordinary sense coexist by occupying different territories: 'vrddhi' in grammar names three vowels, 'vrddhi' in the world names growth, and no one collides. But: samkhyayam eka-visayam vyavasthanam dvayor api — in the case of *number*, the settling has a *single* domain, even for both: the technical term for number and the ordinary word for number are about the same thing; the domains coincide, and the two names of the same territory must be disambiguated some other way. The verse marks a real complication in 2.366-368's tidy picture: the standard method (domain-separation) fails wherever the science's subject matter overlaps the world's. And number is exactly such a place — everyone counts, and the grammar counts too.

2.370 unformalized

विषयं कृत्रिमस्यापि लौकिकः क्वचिदुच्चारन् ।
व्याप्नोति दूरात्सम्बुद्धौ तथाहि ग्रहणं तयोः ॥ ३७० ॥

viṣayaṃ kṛtrimasyāpi laukikaḥ kvaciduccāran |
vyāpnoti dūrātsambuddhau tathāhi grahaṇaṃ tayoḥ || 370 ||

Uttering, in some case, even the artificial one's domain, the worldly one pervades; from a distance, in the vocative, the grasping of the two is thus.

The movement's closing observation, and it is a scene. Visayam krtrimasya api laukikah kvacid uccaran vyapnoti: the *worldly* name, uttered, sometimes *pervades* even the *artificial* one's domain — ordinary usage floods into the technical territory and covers it. And the illustration: durat sambuddhau tatha hi grahanam tayoh — *from a distance*, in the *vocative*, the grasping of the two is thus. Someone shouts a name across a field. At that distance, with the wind, the technical and the ordinary are not distinguished — the hearer grasps whichever his situation supplies, and the vocative, which is pure summons, carries no marks to sort them. The image closes the movement on the note the whole kanda has sounded: at the edge of audibility, where form gives out, nothing remains but occasion and the trained mind — and that is not a degenerate case but the ordinary one seen clearly. Every hearing is a hearing from a distance.
contested: the verse is elliptical and our reading of the vocative-at-a-distance as the illustration follows tatha hi; the point (worldly usage overrunning technical domains where marks are unavailable) is secure

2.371 · Distribution of Action over Groups unformalized

सङ्घैकशेषद्वन्द्वेषु केचित्सामर्थ्यलक्षणम् ।
प्रत्याश्रयमवस्थानं क्रियाणां प्रतिजानते ॥ ३७१ ॥

saṅghaikaśeṣadvandveṣu kecitsāmarthyalakṣaṇam |
pratyāśrayamavasthānaṃ kriyāṇāṃ pratijānate || 371 ||

In groups, single-remainders and dvandvas, some assert a standing of the actions substratum by substratum, whose mark is capacity.

A movement returns to the distribution-question of 2.221-225, now with the machinery of the intervening argument available. In sangha, ekasesa and dvandva — the three ways language gathers many into one form (2.108's single-remainder, 2.221's dvandva) — kecid: some assert pratyasrayam avasthanam kriyanam, a *standing of the actions substratum by substratum*: the action lodges separately in each member. And the mark by which they identify such cases: samarthya-laksanam — *capacity* is the criterion: what could the action do here? An action that can only be done to one at a time distributes; one that can take a group whole does not. The criterion is functional rather than formal, and it is the right one: nothing in the dvandva's shape says whether the eating reaches each or the group. Only what eating *is* decides. The verses following work this out with a precision that is among the most careful in the work.

2.372 unformalized

भोजनं फलरूपाभ्यामेकैकस्मिन् समाप्यते ।
अन्यथा हि व्यवस्थाने न तदर्थः प्रकल्प्यते ॥ ३७२ ॥

bhojanaṃ phalarūpābhyāmekaikasmin samāpyate |
anyathā hi vyavasthāne na tadarthaḥ prakalpyate || 372 ||

Eating is completed in each single one, by way of its fruit and its form; for on any other settlement, its purpose is not construed.

The test applied to the clearest case. Bhojanam phala-rupabhyam ekaikasmin samapyate: eating is *completed in each single one*, by way of its *fruit* and its *form* — its form (the taking of food) and its fruit (satisfaction) are both realized per eater, not per group. And the argument: anyatha hi vyavasthane na tadarthah prakalpyate — on any other settlement, its *purpose* is not construed: an eating that fed the group without feeding the members would feed nobody; the purpose fails unless it distributes. This is the samarthya-criterion (2.371) doing exactly what it should — reasoning from what the action is *for* to how it must distribute (2.113's prayojana, now determining syntax). Note that nothing linguistic settles this: 'Rama and Krsna eat' and 'Rama and Krsna surround the house' have the same form, and only the nature of eating and surrounding tells you that the first distributes and the second cannot.

2.373 unformalized

अन्नादानादिरूपां च सर्वे तृप्तिफलां भुजिम् ।
प्रत्येकं प्रतिपद्यन्ते न तु नाट्यक्रियामिव ॥ ३७३ ॥

annādānādirūpāṃ ca sarve tṛptiphalāṃ bhujim |
pratyekaṃ pratipadyante na tu nāṭyakriyāmiva || 373 ||

And all of them attain, each severally, the eating whose form is the taking of food and whose fruit is satisfaction — not, however, as with a dramatic performance.

The contrast that makes 2.372 precise. Sarve pratyekam pratipadyante bhujim — all attain the eating *each severally*: anna-adana-adi-rupam, its form being the taking of food; trpti-phalam, its fruit being satisfaction. Both form and fruit are per-person. Na tu natya-kriyam iva: *not, however, as with a dramatic performance* — and the counter-example is beautifully chosen. A troupe performs a play: the performing is genuinely the group's, no member performs the play, each does a part and the play happens between them; and its fruit (the drama, the rasa) is not divisible among the actors at all. So natya is an action that *cannot* distribute — the group is its only possible substratum — where bhoji cannot help distributing. Two actions, two behaviours, and the difference is in the actions, not in the grammar. The verse gives the tradition a working typology of collective versus distributive predicates, arrived at by asking what each action's fruit is and to whom it accrues.

2.374-375 unformalized

पाद्यवत् सा विभागेन सामर्थ्यादवतिष्ठते ।
भुजिः करोति भुज्यर्थं न तन्त्रेण प्रदीपवत् ॥ ३७४ ॥
दृश्यादिस्तु क्रियैकापि तथाभृतेषु कर्मसु ।
आवृत्तिमन्तरेणापि समुदायाश्रया भवेत् ॥ ३७५ ॥

pādyavat sā vibhāgena sāmarthyādavatiṣṭhate |
bhujiḥ karoti bhujyarthaṃ na tantreṇa pradīpavat || 374 ||
dṛśyādistu kriyaikāpi tathābhṛteṣu karmasu |
āvṛttimantareṇāpi samudāyāśrayā bhavet || 375 ||

(374) Like water-for-the-feet, it stands by division, from capacity; the eating effects the eating-purpose — not by single-performance, as a lamp does. (375) But an action such as seeing, though single, with respect to objects of that sort, may rest on the aggregate even without repetition.

The typology given its two poles and its explanatory principle. (374): padyavat — like *water for the feet*, which must be given to each guest severally (one basin does not wash all) — bhujih vibhagena samarthyat avatisthate: the eating stands by *division*, from capacity. And the crucial denial: na tantrena pradipavat — *not by single-performance, as a lamp does*: one lamp lights a whole assembly at once, discharging its office for all by a single operation (tantra, 2.108's ritual device, here as a natural category). So actions sort into the padya-type (one per recipient) and the pradipa-type (one for all). (375) gives the other pole: drsyadis tu kriya eka api — an action such as *seeing*, though *single* — tathabhutesu karmasu, with respect to objects of that sort — samudaya-asraya bhavet avrttim antarena api: may rest on the *aggregate* even *without repetition*: one seeing takes in the whole group; no re-seeing per member. The principle throughout is the same: how an action distributes over a plurality is fixed by the action's own nature, and language merely reports it. The lamp, which has served the work as the model of manifestation (1.94, 2.298), here serves as the model of collective efficacy — the same property under another aspect.

2.376 unformalized

भिन्नव्यापाररूपाणां व्यवहारादिदर्शने ।
कर्तॄणां दर्शनं भिन्नं संभूयार्थस्य साधकम् ॥ ३७६ ॥

bhinnavyāpārarūpāṇāṃ vyavahārādidarśane |
kartṝṇāṃ darśanaṃ bhinnaṃ saṃbhūyārthasya sādhakam || 376 ||

In the seeing of practice and the like, of those whose forms of activity are distinct — the agents' seeing is distinct, though jointly accomplishing the purpose.

A third pattern, and it complicates the binary. Consider a joint undertaking — vyavahara-adi-darsane, the seeing of a transaction and the like (a lawsuit, a trade, a rite performed by several officiants): the participants have bhinna-vyapara-rupanam, *distinct forms of activity* — each does something different. Their kartrnam darsanam bhinnam: the agents' seeing is *distinct* — nobody's view of the transaction is anybody else's; each sees his own part. And yet: sambhuya arthasya sadhakam — *jointly accomplishing* the purpose: the transaction happens, and it happens by their several distinct activities converging. So here the action neither distributes wholly (each does not do the whole thing) nor collects wholly (there is no single group-act over their heads): the members do different things and the purpose is one. This is the structure of cooperation, and it is a third category the padya/pradipa binary misses — which is exactly why 2.371's criterion was samarthya, capacity, and not a formal rule: capacities are as various as actions, and no dichotomy exhausts them.

2.377-378 unformalized

लक्ष्यस्य लोकसिद्धत्वाच्छास्त्रे लिङ्गस्य दर्शनात् ।
अर्थित्वादैक्षुभेदेन वृद्धिसंज्ञा समाप्यते ॥ ३७७ ॥
शतादाने प्रधानत्वाद् दण्डने शतकर्मके ।
अर्थिनां गुणभेदेऽपि संख्येयोऽर्थो न भिद्यते ॥ ३७८ ॥

lakṣyasya lokasiddhatvācchāstre liṅgasya darśanāt |
arthitvādaikṣubhedena vṛddhisaṃjñā samāpyate || 377 ||
śatādāne pradhānatvād daṇḍane śatakarmake |
arthināṃ guṇabhede'pi saṃkhyeyo'rtho na bhidyate || 378 ||

(377) Because what is to be characterized is established in the world, and because the mark is seen in the science, and from need — the name 'vrddhi' is completed by way of the sugarcane-difference. (378) In the taking of a hundred, from principality; in a punishing whose object is a hundred — though the needers differ in their subordinate elements, the countable object is not differentiated.

Two technical cases testing the distribution-doctrine on quantities and names. (377) is a grammarian's puzzle about how the technical name vrddhi reaches its instances — laksyasya loka-siddhatvat, what is to be characterized being established in the world; sastre lingasya darsanat, the mark being seen in the science; arthitvat, from need — and the resolution comes aiksu-bhedena, 'by way of the sugarcane-difference' (the traditional example of a taddhita-formation whose derivation is disputed). (378) is sharper: sata-adane pradhanatvat — in the *taking of a hundred*, from principality; danda ne sata-karmake — in a punishing whose object is a hundred: arthinam guna-bhede 'pi samkhyeyo 'rtho na bhidyate — though the *needers* differ in their subordinate elements, the *countable object is not differentiated*. Fine a hundred men, and the hundred is one hundred however variously they are fined. The number is a property of the group and does not distribute, even where the punishing does. So distribution can split within a single predication: the action reaching each, the count reaching only the whole.
contested: 377 is among the most compressed and technically opaque verses in the kanda; aiksu-bhedena and the vrddhi-samjna's completion refer to grammatical disputes not recoverable from the karika, and our rendering translates the evident syntax while leaving the case unreconstructed

2.379 unformalized

संघस्यैव विधेयत्वात् कार्यवत्प्रतिपादने ।
तत्र तन्त्रेण सम्बन्धः समासाभ्यस्तसंज्ञयोः ॥ ३७९ ॥

saṃghasyaiva vidheyatvāt kāryavatpratipādane |
tatra tantreṇa sambandhaḥ samāsābhyastasaṃjñayoḥ || 379 ||

Since it is the group alone that is to be enjoined, as in the conveying of an operation — there the connection is by single-performance, for the compound-name and the reduplicated-name.

The tantra-model (2.374's lamp) applied to grammar's own naming. Sanghasya eva vidheyatvat — since it is the *group alone* that is to be enjoined (the rule addresses the collection, not its members) — karyavat pratipadane, as in the conveying of an operation: tatra tantrena sambandhah — there the connection is by *single-performance*: one act of naming reaches all at once. And the instances: samasa-abhyasta-samjnayoh — for the *compound*-name and the *reduplicated*-name: Panini's terms samasa and abhyasta name classes, and the naming does not repeat per member; one posting covers the class. This is the pradipa-pattern in the metalanguage: a definition is a lamp, not a basin. The verse quietly makes a point about rule-systems generally — the economy that makes grammar possible (2.343's laghava) is precisely the availability of tantra: one statement discharging indefinitely many applications. A science in which every naming had to distribute would be as long as its subject matter.

2.380-382 unformalized

लक्षणार्था स्तुतिर्येषां कञ्चिदेव क्रियां प्रति ।
तैर्व्यस्तैश्च समस्तैश्च स धर्म उपलक्ष्यते ॥ ३८० ॥
वृषलैर्न प्रवृष्टव्यमित्येतस्मिन् गृहे यथा ।
प्रत्यकं संहतानां च प्रवेशः प्रतिषिध्यते ॥ ३८१ ॥
सम्भूय त्वर्थलिप्सादिप्रतिषेधोपपादने ।
पृथगप्रतिषिद्धत्वात् प्रवृत्तिर्न विरुध्यते ॥ ३८२ ॥

lakṣaṇārthā stutiryeṣāṃ kañcideva kriyāṃ prati |
tairvyastaiśca samastaiśca sa dharma upalakṣyate || 380 ||
vṛṣalairna pravṛṣṭavyamityetasmin gṛhe yathā |
pratyakaṃ saṃhatānāṃ ca praveśaḥ pratiṣidhyate || 381 ||
sambhūya tvarthalipsādipratiṣedhopapādane |
pṛthagapratiṣiddhatvāt pravṛttirna virudhyate || 382 ||

(380) Those whose praise has characterization for its purpose, with respect to some action — by them, taken severally and taken together, that property is indicated. (381) As in 'sudras must not enter this house', the entry of them severally and of them combined is prohibited. (382) But in the establishing of a prohibition on the seeking of gain and the like jointly — since it is not prohibited severally, engagement is not contradicted.

The distribution-question in prohibitions, where the stakes are practical. (380) states the principle: where praise or blame serves *characterization* — laksanartha stutih, marking out a property — the property is indicated by the terms vyastaih ca samastaih ca: *severally and together* both. (381) illustrates: vrsalair na pravestavyam — 'sudras must not enter this house' — pratyekam samhatanam ca pravesah pratisidhyate: the entry of them *severally* and *combined* is prohibited: the prohibition distributes and collects at once; no sudra may enter, and no group of them may. (382) gives the contrasting case: where a prohibition is established on artha-lipsa-adi *jointly* — on seeking gain *in concert* — prthag apratisiddhatvat pravrttir na virudhyate: since it is *not prohibited severally*, individual engagement is not contradicted: one may seek gain alone if only conspiracy was barred. The pair is a piece of practical jurisprudence, and its lesson is the movement's: whether a provision distributes is not readable from its form. It is readable from what it is *for* — and getting it wrong is, in the second case, the difference between a conspiracy and a livelihood.

2.383 unformalized

व्यवायलक्षणार्थत्वादट्कुप्वाङादिस्तथा ।
प्रत्येकं वा समस्तैर्वाणत्वं न प्रतिषिध्यते ॥ ३८३ ॥

vyavāyalakṣaṇārthatvādaṭkupvāṅādistathā |
pratyekaṃ vā samastairvāṇatvaṃ na pratiṣidhyate || 383 ||

Because the purpose is the characterizing of intervention, 'at', 'ku', 'pu', 'an' and the like — retroflexion is not prohibited either severally or by all of them taken together.

The principle of 2.380-382 applied to the grammar's most notorious rule. Panini's at-ku-pv-an-numvyavaye 'pi (8.4.2) governs natva, the retroflexion of n: the change occurs even when certain sounds *intervene* — the vowels, the gutturals, the labials, and so forth, listed as at, ku, pu, an. The question the verse settles: does the list block retroflexion when its members intervene severally, or only when they intervene all together? Answer: vyavaya-laksanarthatvat — because the purpose is the *characterizing of intervention* — anatvam na pratisidhyate pratyekam va samastaih va: retroflexion is not prohibited either *severally* or *by all of them taken together*. The rule is permissive in both directions: any of them may intervene, in any combination, and the change still runs. The reasoning is 2.380's exactly — the list's purpose is to characterize a class of interveners, not to specify a configuration — and the verse's placement shows why the movement has been so patient: the distribution-question is not scholastic decoration. It decides whether a fundamental sandhi rule fires.

2.384 unformalized

अनुग्रहार्थ भोक्तॄणां भुजिरारभ्यते यदा ।
देशकालाद्यभेदेन नानुगृह्णाति तानसौ ॥ ३८४ ॥

anugrahārtha bhoktṝṇāṃ bhujirārabhyate yadā |
deśakālādyabhedena nānugṛhṇāti tānasau || 384 ||

When the eating is undertaken for the sake of favouring the eaters — it does not favour them, without a difference of place, time and the rest.

A refinement on 2.372's distribution, and it turns on individuation. Anugrahartham bhoktrnam bhujir arabhyate yada — when the eating is undertaken for the sake of *favouring* the eaters (the host's purpose: to benefit them) — desa-kala-adi-abhedena na anugrhnati tan asau: it does *not favour them* without a difference of *place, time and the rest*. To benefit each eater, the eating must be individuated per eater — this eater, here, now, with his own portion. An undifferentiated eating benefits nobody, because benefit accrues to individuals in circumstances. So the distribution 2.372 established by the action's fruit is here shown to require the individuating apparatus (desa, kala — the same factors that carve meanings at 2.315-316, here carving events). The verse is a small but exact point: collective predicates that carry benefit must distribute, because benefit is not a thing a group can receive on its members' behalf.

2.385 unformalized

पात्रादिभेदान्नानात्वं यस्यैकस्योपदिश्यते ।
विपर्यये वा भिन्नस्य तस्यैकत्वं प्रकल्प्यते ॥ ३८५ ॥

pātrādibhedānnānātvaṃ yasyaikasyopadiśyate |
viparyaye vā bhinnasya tasyaikatvaṃ prakalpyate || 385 ||

That single thing whose manifoldness is taught by difference of vessels and the like; or, in the reverse case, the oneness of what is differentiated is constructed.

Individuation shown to run both ways, and the verse is deliberately symmetrical. Patradi-bhedat nanatvam yasya ekasya upadisyate: that *single* thing whose *manifoldness* is taught by difference of *vessels* and the like — one substance in several pots, and the pots make it many (the ghee is 'the ghees' if it is in three jars, for ritual purposes of counting). Viparyaye va bhinnasya tasya ekatvam prakalpyate: or *in the reverse case*, the *oneness* of what is *differentiated* is constructed — many things gathered as one by a single vessel, a single occasion, a single office. Number is conferred by the individuating conditions, in either direction, and neither the one nor the many is the substance's own. This is 2.104's capacity-doctrine applied to countability itself: as wordhood is an office worn by stretches of the stream, so unity and plurality are offices worn by stuff, conferred by vessels, and the grammar's number-apparatus reports the conferral rather than discovering a count.

2.386-387 unformalized

संहत्यापि च कुर्वाणा भेदेन प्रतिपादिताः ।
स्वं स्वं भोज्यं विभागेन प्राप्तं संभूय भुञ्जते ॥ ३८६ ॥
वीप्साया विषयाभावाद् विरोधायन्यसंख्यया ।
द्विधा समाप्त्ययोगाच्च शतं सङ्घेऽवतिष्ठते ॥ ३८७ ॥

saṃhatyāpi ca kurvāṇā bhedena pratipāditāḥ |
svaṃ svaṃ bhojyaṃ vibhāgena prāptaṃ saṃbhūya bhuñjate || 386 ||
vīpsāyā viṣayābhāvād virodhāyanyasaṃkhyayā |
dvidhā samāptyayogācca śataṃ saṅghe'vatiṣṭhate || 387 ||

(386) And those who, though acting in combination, are conveyed by way of difference, eat jointly what has been obtained severally, each his own food by division. (387) Since there is no domain for distributive repetition, and because of contradiction with another number, and because a two-fold completion is not possible — the hundred stands upon the group.

Two verses of exact analysis on the mixed case. (386): people samhatya kurvanah, acting *in combination* — a shared meal — are nonetheless bhedena pratipaditah, conveyed by way of *difference*; and the description is precise: svam svam bhojyam vibhagena praptam sambhuya bhunjate — each *his own* food, obtained *by division*, they eat *jointly*. Joint in occasion, several in substance and fruit: the structure of every shared meal, and language reports both aspects at once without strain. (387) then handles the count: vipsayah visaya-abhavat — there being no domain for *distributive repetition* (no 'a hundred each'); virodhat anya-samkhyaya — from *contradiction* with another number (a hundred distributed would yield some other total); dvidha samapty-ayogat ca — and because a *two-fold completion* is not possible (the count cannot both distribute and collect): satam sanghe 'vatisthate — the *hundred stands upon the group*. Three independent arguments converging on one conclusion, and the conclusion is 2.378's: actions may distribute where their counts cannot.

2.388 unformalized

भुजिद्वन्द्वैकशेषाभ्यां यत्रान्यैः सह शिष्यते ।
तत्रापि लक्षणार्थत्वाद् द्विधा वाक्यं समाप्यते ॥ ३८८ ॥

bhujidvandvaikaśeṣābhyāṃ yatrānyaiḥ saha śiṣyate |
tatrāpi lakṣaṇārthatvād dvidhā vākyaṃ samāpyate || 388 ||

Where, by the eating-dvandva and the single-remainder, it is taught along with others — there too, because the purpose is characterization, the sentence is completed two-fold.

The movement's last technical case, and it admits a complication the earlier analysis suppressed. Bhuji-dvandva-ekasesabhyam yatra anyaih saha sisyate — where, by the eating-dvandva and the ekasesa, something is taught *along with others*: the two gathering-devices operating together (2.108's single-remainder inside a dvandva) — tatrapi laksanarthatvat dvidha vakyam samapyate: *there too*, because the purpose is *characterization* (2.380's principle), the sentence is completed *two-fold*. The sentence closes twice — once distributively, once collectively — and both closures are real. This is 2.387's dvidha samapty-ayoga (a two-fold completion is not possible) apparently contradicted, and the contradiction is only apparent: there, of *number*, which cannot complete twice; here, of *characterization*, which can. The verse is a fine demonstration of how carefully the movement has been discriminating: what completes twice and what cannot depends on what kind of thing is completing, and no general rule about sentences settles it.

2.389 unformalized

वाक्यान्तरणां प्रत्येकं समाप्तिः कैश्चिदिष्यते ।
रूपान्तरेण युक्तानां वाक्यानां तेन संग्रहः ॥ ३८९ ॥

vākyāntaraṇāṃ pratyekaṃ samāptiḥ kaiścidiṣyate |
rūpāntareṇa yuktānāṃ vākyānāṃ tena saṃgrahaḥ || 389 ||

By some, a completion of the several sentences is accepted, each severally; by that, sentences joined with another form are comprehended.

The doctrine of 2.388 generalized into a position. Kaiscid isyate: some accept vakyantaranam pratyekam samaptih — a *completion of the several sentences, each severally*: where a compressed utterance carries several predications (the dvandva with its distributed action, the ekasesa with its several referents), it is not one sentence with a complex meaning but *several sentences*, each complete. And the consequence: tena rupantarena yuktanam vakyanam sangrahah — by that, sentences *joined with another form* are comprehended: the several sentences are grasped through the single form that gathers them. So the ekasesa's one uttered word (2.108-109) is, on this account, several sentences wearing one body — 2.110's doctrine (one form, many sentences) stated as a semantics rather than a ritual convenience. The position is available to Bhartrhari and he does not quite claim it; but it fits everything he has argued about offices and bodies (2.104), and it explains the two-fold completion of 2.388 without paradox: two completions because two sentences, sharing a form.

2.390 unformalized

न वाक्यस्याभिधेयानि भेदवाक्यानि कानिचित् ।
तस्मिंस्तूच्चरिते भेदांस्तथान्यान् प्रतिपद्यते ॥ ३९० ॥

na vākyasyābhidheyāni bhedavākyāni kānicit |
tasmiṃstūccarite bhedāṃstathānyān pratipadyate || 390 ||

The difference-sentences are not things expressible by the sentence, any of them; but when that has been uttered, one comprehends those differences and others so.

The movement's closing correction, and it pulls the doctrine of 2.389 back into the kanda's frame. Na vakyasya abhidheyani bheda-vakyani kanicit: the *difference-sentences* — the several predications the analysis distributes out of the compressed form — are *not things expressible by the sentence*, not any of them: the utterance does not *say* them; they are not its content. But: tasmin tu uccarite bhedan tatha anyan pratipadyate — when *that* has been uttered, one comprehends those differences, *and others*, so. The distributed predications are what the hearer *gets*, not what the sentence *says* — and 'and others' (anyan) concedes that he gets more besides, unpredictably, per hearer (2.299's overspill; 2.135's division per own-cognition). This is the kanda's settled position applied one last time: the utterance occasions a flash whose articulation the analyst can spread across several sentences, and none of those sentences is what was said. What was said was one thing. What arrived was one thing. The several sentences are the ledger.

2.391 · The Word-Meaning's Standing unformalized

येषां समस्तो वाक्यार्थः प्रतिभेदं समाप्यते ।
तेषां तदानीं भिन्नस्य किं पदार्थस्य सत्तया ॥ ३९१ ॥

yeṣāṃ samasto vākyārthaḥ pratibhedaṃ samāpyate |
teṣāṃ tadānīṃ bhinnasya kiṃ padārthasya sattayā || 391 ||

For those for whom the entire sentence-meaning is completed at each division — for them, at that moment, of what use is the existence of a separate word-meaning?

The objection turned back on the holist, and it is a sharp one. If, as 2.18 held, the krtsnah vakyarthah samapyate pratibhedam — the *entire* sentence-meaning completes at *each* division, each word — then tesam tadanim bhinnasya kim padarthasya sattaya: at that moment, what use is the *existence* of a separate word-meaning? The whole being present at every word, the word-meanings are not merely non-constitutive (which the kanda has argued at length) but *idle* — they do nothing, explain nothing, and might as well not be. The objection is a reductio of the holist's own most radical thesis, and it deserves its force: a doctrine on which the parts contribute nothing owes an account of why speech has parts at all. The verses following do not dodge it; 2.394-397 will answer with the sadharana-doctrine, and the answer will be that the word-meanings are real as *shares* in a common thing rather than as contributions to a sum.

2.392-393 unformalized

अथ तैरेव जनितः सौऽर्थो भिन्नेषौ वर्तते ।
पूर्वस्यार्थस्य तेन स्याद्विरोधः सह वा स्थितिः ॥ ३९२ ॥
सहस्थतौ विरोधित्वं स्याद् विश्ष्टाविशिष्टयोः ।
व्यभिचारी तु सम्बन्धस्त्यागेऽर्थस्य प्रसज्यते ॥ ३९३ ॥

atha taireva janitaḥ sau'rtho bhinneṣau vartate |
pūrvasyārthasya tena syādvirodhaḥ saha vā sthitiḥ || 392 ||
sahasthatau virodhitvaṃ syād viśṣṭāviśiṣṭayoḥ |
vyabhicārī tu sambandhastyāge'rthasya prasajyate || 393 ||

(392) Now if that meaning, generated by those very ones, operates upon the differentiated — then, of the prior meaning, there would be either contradiction with it, or co-standing. (393) On co-standing, there would be contradictoriness of the qualified and the unqualified; but a deviating connection follows, on the abandonment of the meaning.

The objection of 2.391 pressed into a dilemma. Suppose the sentence-meaning, generated by the word-meanings, then *operates upon the differentiated* — comes back to bear on the very items that produced it. Then the prior meaning (the word-meaning as it stood before) either contradicts the new one or co-stands with it. (393) takes both horns. On co-standing: viruddhata syad visista-avisistayoh — there is *contradictoriness of the qualified and the unqualified*: the word-meaning as bare (before the sentence) and as qualified (after) cannot both stand, being incompatible determinations of one item. On abandonment: vyabhicari tu sambandhah tyage 'rthasya prasajyate — a *deviating connection* follows: if the bare meaning is abandoned, then the word's connection with its meaning is not invariable, and the eternal bond of 1.23 and 2.332 fails. Contradiction or deviation: the dilemma is real, and it is the same one 2.15 posed against the deposit-theorist (the general obscured in the particular: where does it go?), now aimed at the holist. Bhartrhari has arranged for his own position to face the argument he used against everyone else.

2.394 unformalized

एकः साधारणो वाच्यः प्रतिशब्दमवस्थितः ।
सङ्घे सङ्घिषु चार्थात्मा संनिधाननिदर्शकः ॥ ३९४ ॥

ekaḥ sādhāraṇo vācyaḥ pratiśabdamavasthitaḥ |
saṅghe saṅghiṣu cārthātmā saṃnidhānanidarśakaḥ || 394 ||

A single common expressible, standing in each word, is the meaning-self in the group and in the group's members — indicating the proximity.

The answer, and it introduces the model that will carry the movement: sadharana, the *common*. Ekah sadharano vacyah — a *single common* expressible — pratisabdam avasthitah: standing *in each word*: not divided among them, not summed from them, but present in each as a common thing is present to each of its sharers. Sanghe sanghisu ca arthatma: it is the meaning-self *in the group and in the group's members* — both, without division. And its office: sannidhana-nidarsakah — *indicating the proximity*: it is what makes the words' co-presence into a togetherness rather than a heap. The model dissolves 2.392-393's dilemma at a stroke. There is no prior bare meaning that must be either contradicted or abandoned, because the word never had a meaning of its own to lose — it had a *share* in a common one, and shares are not diminished by being shared. What the word-meaning *is*, on this account, is the sentence-meaning as standing in this word: 2.391's 'idle' padartha turns out to be the vakyartha itself, seen from where one word stands.

2.395-396 unformalized

यथा साधारणे स्वत्वं त्यागस्य च फलं धने ।
प्रीतिश्चाविकला तद्वत् सम्बन्धोऽर्थे न तद्वता ॥ ३९५ ॥
वर्णानामर्थवत्तायां तैनैवार्थेन तद्वति ।
समुदायेन चैकत्वं भेदेन व्यवतिष्ठते ॥ ३९६ ॥

yathā sādhāraṇe svatvaṃ tyāgasya ca phalaṃ dhane |
prītiścāvikalā tadvat sambandho'rthe na tadvatā || 395 ||
varṇānāmarthavattāyāṃ tainaivārthena tadvati |
samudāyena caikatvaṃ bhedena vyavatiṣṭhate || 396 ||

(395) As, in common property, ownership, and the fruit of relinquishing, and the satisfaction, are undiminished in the wealth — so the connection is with the meaning, not with its possessor. (396) In the meaningfulness of phonemes, by that very meaning in the possessor, oneness with the aggregate stands by way of difference.

The sadharana-model given its illustration, and it is drawn from law. Consider sadharana dhana — *common property*, held jointly by brothers. Each has svatva, *ownership* — full ownership, not a fraction of it; each has tyagasya phalam, the *fruit of relinquishing* (the merit of giving it away — and each gets the whole merit, not a share); each has pritih, *satisfaction* — avikala, *undiminished*. The wealth is one, the ownings are several, and no owning is a portion. This is the exact structure the holist needs: the sentence-meaning is one, the words' meanings are several, and each is the whole thing owned from a position, not a slice. And the doctrinal formulation: sambandho 'rthe na tadvata — the connection is *with the meaning*, not with its *possessor*: what a word is bonded to is the meaning itself, not to the word-as-meaning-possessor. (396) applies it downward: phonemes' meaningfulness, likewise, is oneness with the aggregate standing bhedena, by way of difference — each phoneme owning the word's meaning from where it stands. The joint-property analogy is one of the most illuminating in the work: it explains how many can each have all of one thing, which is precisely what the flash-doctrine requires and what every deposit-semantics finds impossible.

2.397 unformalized

एकेनैव प्रदीपेन सर्वे साधारणं धनम् ।
पश्यन्ति तद्वदेकेन सुपा संख्याऽभिधीयते ॥ ३९७ ॥

ekenaiva pradīpena sarve sādhāraṇaṃ dhanam |
paśyanti tadvadekena supā saṃkhyā'bhidhīyate || 397 ||

By a single lamp alone, all see the common wealth; likewise, by a single case-ending, number is expressed.

The lamp returns to seal the sadharana-model, and the pairing of images is exact. Ekena eva pradipena sarve sadharanam dhanam pasyanti: by *one* lamp, *all* see the *common* wealth — one illumination, many seers, one object, and nobody's seeing is a fraction of anybody's. Tadvat ekena supa samkhya abhidhiyate: *likewise*, by a *single case-ending*, number is expressed. The application is precise: the ending does not distribute a number-contribution among the words it governs; it lights the whole and each sees it entire (2.375's collective seeing, 2.374's lamp as the tantra-model, now doing semantic work). Between the joint-property (2.395) and the lamp, the movement has its two figures for how one thing can be wholly many's: ownership without division, and illumination without division. Both were needed. Property answers 'how can each have all?'; the lamp answers 'how does one act reach all at once?' — and together they explain how a single utterance delivers one undivided meaning to every word in it and every hearer of it.

2.398 unformalized

नार्थवत्ता पदे वर्णे वाक्ये चैवं विशिष्यते ।
अभ्यासात् प्रक्रमोऽन्यस्तु विरुद्ध इव दृश्यते ॥ ३९८ ॥

nārthavattā pade varṇe vākye caivaṃ viśiṣyate |
abhyāsāt prakramo'nyastu viruddha iva dṛśyate || 398 ||

Meaningfulness is not thus distinguished in the word, in the phoneme, and in the sentence; but from practice, a different commencement is seen, as though contradictory.

The consequence of the sadharana-model for the whole architecture of grains. Na arthavatta pade varne vakye ca evam visisyate: meaningfulness is *not distinguished* in the word, in the phoneme, and in the sentence — not, that is, in the way the analysis supposes, with each grain owning its own quantum. On the common-property model there is one meaning and three positions from which it is held, and 'meaningfulness' names the same thing at every grain. Then the concession: abhyasat prakramo 'nyah tu viruddha iva drsyate — but *from practice*, a *different commencement* is seen, *as though* contradictory: usage proceeds as if the grains had distinct meanings, and the appearance of conflict between the grains (the word's meaning versus the sentence's) is an artifact of the practice, not of the semantics. Iva, as always, holds the line. The kanda has now answered 2.391's challenge fully: word-meanings exist, are not idle, and are not contributions — they are the one meaning, owned from a place.

2.399 · Intention and the Word's Operation unformalized

विनियोगादृते शब्दो न स्वार्थस्य प्रकाशकः ।
अर्थाभिधानसम्बन्धमुक्तिद्वारं प्रचक्षते ॥ ३९९ ॥

viniyogādṛte śabdo na svārthasya prakāśakaḥ |
arthābhidhānasambandhamuktidvāraṃ pracakṣate || 399 ||

Without application, a word is not the illuminator of its own meaning; they call the connection with the expressing of meaning a door.

A new movement, on what activates a word, and its thesis is stated at once. Viniyogad rte sabdo na svarthasya prakasakah: *without application* — viniyoga, the ritual and grammatical term for deploying an element to an office (2.258's vinyoga individuating mantras) — a word is *not the illuminator of its own meaning*. An unapplied word illuminates nothing; it sits inert, a form without office. And the figure: artha-abhidhana-sambandham uktidvaram pracaksate — the connection with the expressing of meaning they call a *door*: the bond between word and meaning is a doorway, standing open, through which nothing passes until someone walks. The image is exactly calibrated to the kanda's position. The eternal connection (1.23, 2.332) is real and is not sufficient: it is a standing possibility, a door, and utterance with intent is the walking through. This is why the plenitude (2.250) needs nimittas, why the lamp needs bringing (2.397), why the word means everything and says one thing.

2.400 unformalized

यथा प्रणिहितं चक्षुर्दर्शनायोपकल्पते ।
तथाऽभिसंहितः शब्दो भवत्यर्थस्य वाचकः ॥ ४०० ॥

yathā praṇihitaṃ cakṣurdarśanāyopakalpate |
tathā'bhisaṃhitaḥ śabdo bhavatyarthasya vācakaḥ || 400 ||

As the eye, when directed, becomes fit for seeing — so a word, when intended, becomes the expressor of a meaning.

The doctrine of 2.399 given its simile, and the simile is chosen with great care. Yatha pranihitam caksuh darsanaya upakalpate: as the eye, when *directed* — pranihita, aimed, laid upon its object — becomes *fit* for seeing. The eye is always an eye; open, it is capable of every visible thing; and it sees *this* because it was turned this way. Tatha abhisamhitah sabdo bhavaty arthasya vacakah: so a word, when *intended* — abhisamhita, aimed by the speaker's intent — becomes the expressor of *a* meaning. Every element of the kanda's semantics is in the analogy. The word's omni-significance is the eye's openness to all colours (2.250's sarvarthya); the intention is the aiming; the meaning that results is what was in the line of sight — carved from the plenitude by a directing, not selected from a list. And the simile explains, incidentally, why the word's capacity is not diminished by use: an eye that has looked at a pot is no less able to look at a cloth. The vivaksa of 1.155 has become, here, the whole engine of signification.

2.401 unformalized

क्रियाव्यवेतः सम्बन्धो दृष्टः करणकर्मणोः ।
अभिधा नियमस्तस्मादभिधानाभिधेययोः ॥ ४०१ ॥

kriyāvyavetaḥ sambandho dṛṣṭaḥ karaṇakarmaṇoḥ |
abhidhā niyamastasmādabhidhānābhidheyayoḥ || 401 ||

The connection of instrument and object is seen as mediated by an action; therefore the fixing of expression is of the expression and the expressible.

The structural point behind 2.400's simile. Kriya-vyavetah sambandho drstah karana-karmanoh — the connection of *instrument* and *object* is seen as *mediated by an action*: an axe and a tree are not related until there is a felling; the karaka-relations exist only within an action (2.183: prior to connection with the means, the action does not arise). By parity: abhidha niyamah tasmat abhidhana-abhidheyayoh — the *fixing* of expression is therefore of the *expression and the expressible* — i.e. the word-meaning relation, like the instrument-object relation, is constituted within an act: the act of abhidha, expressing. There is no standing relation between word and meaning waiting to be used; there is a door (2.399) that becomes a relation when someone walks through it. This is the metaphysics behind the eye-simile: relations of this kind are not found but *made*, in and by the act that has them as its terms — which is the Sambandha-samuddesa's thesis (Kanda III) in embryo, and which is why the third kanda must exist.

2.402 unformalized

बहुष्वेकाभिधानेषु सर्वेष्वेकार्थकारिषु ।
यत्प्रयोक्ताभिसन्धत्ते शब्दस्तत्रावतिष्ठते ॥ ४०२ ॥

bahuṣvekābhidhāneṣu sarveṣvekārthakāriṣu |
yatprayoktābhisandhatte śabdastatrāvatiṣṭhate || 402 ||

Among many single expressions, all effecting one meaning — the word stands upon that which the speaker intends.

The intention-doctrine's decisive case. Bahusu eka-abhidhanesu sarvesu ekartha-karisu: among *many* single expressions, *all* effecting *one* meaning — the synonyms, several words any of which would do (2.330's paryayah): yat prayokta abhisandhatte sabdah tatra avatisthate — the word stands upon *that which the speaker intends*. Where several words could serve and one is chosen, nothing but the intention distinguishes them — the meaning being identical, the choice is the speaker's and no fact about the language settles it. This is the cleanest possible demonstration that abhisandhi, intention, is doing real work: it is the only variable left when the semantic contribution is held constant. And it explains 2.399's door-figure: the eternal connections stand open for all the synonyms at once, and what determines which door is walked is not the doors.

2.403-404 unformalized

आम्नायशब्दानभ्यासे केचिदाहुरनर्थकान् ।
स्वरूपमात्रवृत्तींश्च परेषां प्रतिपादने ॥ ४०३ ॥
अभिधानक्रियाभेदादर्थस्य प्रतिपादकात् ।
नियोगभेदान्मन्यन्ते तानेवाकत्वदर्शिनः ॥ ४०४ ॥

āmnāyaśabdānabhyāse kecidāhuranarthakān |
svarūpamātravṛttīṃśca pareṣāṃ pratipādane || 403 ||
abhidhānakriyābhedādarthasya pratipādakāt |
niyogabhedānmanyante tānevākatvadarśinaḥ || 404 ||

(403) In the practice of the Vedic words, some say they are meaningless and operate on their own form alone, in the conveying to others. (404) Because of the difference in the act of expression, which conveys the meaning, those who see them as one consider them (different) by difference of application.

The doctrine applied to the recitation of scripture. (403): amnaya-sabdanam abhyase — in the *practice*, the drill, of the Vedic words (the rote recitation by which the Veda is transmitted) — kecid ahuh anarthakan: some say they are *meaningless* — svarupa-matra-vrttin, operating on their *own form alone*: the reciter is not saying anything, he is producing forms; meaning does not enter until the words are deployed in a rite. And (404): the ekatva-darsinah — those who see them as *one* (one word, whether recited or applied) — account for the apparent difference abhidhana-kriya-bhedat and niyoga-bhedat: by the difference of the *expressing act* and of *application*. The two accounts of what happens when a boy chants a hymn he does not understand: either the words are meaningless there (and become meaningful when applied), or they are one and the same words differing only in their office. Both are versions of 2.399: unapplied, a word illuminates nothing — whether one says it is then a different word or a meaningless one is bookkeeping (2.179's licensed double description).

2.405 unformalized

तेषामत्यन्तनानात्वं नानात्वव्यवहारिणः ।
अक्षादीनामिव प्राहुरेकजातिसमन्वयात् ॥ ४०५ ॥

teṣāmatyantanānātvaṃ nānātvavyavahāriṇaḥ |
akṣādīnāmiva prāhurekajātisamanvayāt || 405 ||

Those who deal in manifoldness declare their utter manifoldness — as of dice and the like, from their connection with a single class.

The contrary school, and its analogy is worth examining. Nanatva-vyavaharinah — those who *deal in manifoldness*, whose method is division (2.57's pada-darsins) — tesam atyanta-nanatvam prahuh: declare their *utter manifoldness*: the recited word and the applied word are simply two words, not one word twice. And the simile: aksadinam iva eka-jati-samanvayat — as of *dice* and the like, from their connection with a *single class*: dice are many, and their oneness is only their sharing a kind — there is no one die of which the several are appearances. So words: many tokens, unified by nothing but a common type, no single word standing behind them. The analogy is well chosen for the position and it is exactly what Bhartrhari denies (2.250's one word carved by nimittas). Note that the two schools have now been running in parallel since 2.57, and neither can refute the other by the phenomena — which was 1.74's finding and has never been retracted. What separates them is which primitive they will pay for.

2.406 unformalized

प्रयोगादभिसन्धानमत्पदेषु न विद्यते ।
विषये यतशक्तित्वात् स तु तत्र व्यवस्थितः ॥ ४०६ ॥

prayogādabhisandhānamatpadeṣu na vidyate |
viṣaye yataśaktitvāt sa tu tatra vyavasthitaḥ || 406 ||

Intention arising from employment does not exist in the mantra-words; because their capacity is fixed with respect to their domain, that stands settled there.

A limit-case for the intention-doctrine, and Bhartrhari concedes it. Prayogad abhisandhanam mat-padesu na vidyate: intention arising from employment does *not exist* in the mantra-words — the reciter does not aim them; he utters what is given, and often does not know what it means (2.403's meaningless drill). So 2.400's directed eye is absent. What then makes them mean? Visaye yata-saktitvat sa tu tatra vyavasthitah: because their *capacity is fixed* with respect to their *domain*, it stands settled there — the mantra's power is lodged in it by its office in the rite, independently of any reciter's intention (1.138's mantra-efficacy: form-bound, working whether or not understood). The concession is significant: intention is the engine of ordinary signification (2.400-402) and is *not* the whole story — scripture works without it, its capacities fixed by domain rather than aimed by minds. Which is exactly what a tradition holding an authorless Veda must say (1.144), and Bhartrhari says it rather than forcing the doctrine to cover everything.

2.407 unformalized

नानात्वस्यैव संज्ञानमर्थप्रकरणादिभिः ।
न जात्वर्थान्तरे वृत्तिरन्यार्थानां कथञ्चन ॥ ४०७ ॥

nānātvasyaiva saṃjñānamarthaprakaraṇādibhiḥ |
na jātvarthāntare vṛttiranyārthānāṃ kathañcana || 407 ||

The knowledge of manifoldness alone comes by the object, the context and the rest; there is never, in any way, an operation of words meant for other meanings upon a different meaning.

The movement's closing thesis, and it is a strict one. Nanatvasya eva samjnanam artha-prakaranadibhih: the *knowledge of manifoldness alone* comes by the object, the context, and the rest — the fourteen factors (2.315-316) tell you *which* meaning is in play, and that is all they do. And the denial that follows: na jatv arthantare vrttih anyarthanam kathamcana — there is *never*, in any way, an operation of words *meant for other meanings* upon a *different* meaning. A word aimed at one meaning does not stray into another; the context does not bend a word's sense, it *identifies* which of the word's senses is being used. The distinction is fine and it is the movement's whole point: context is not a force acting on meanings but a light showing which door was walked through (2.399). This closes the door on any account of figurative language as sense-stretching (2.266's forcing) and it protects the eternal bond (2.332): the word never means what it was not, because everything it could mean, it always could.

2.408-410 · The Word as Sentence; Existence and Predication unformalized

पदरूपं तु यद्वाक्यमस्तित्वोपनिबन्धनम् ।
कामं विमर्शस्तत्रायं न वाक्यावयवे पदे ॥ ४०८ ॥
यथैवानर्थकेर्वर्णैविंशिष्टार्थोऽभिधीयते ।
पदैरनर्थकैरेवं विशिष्टार्थोऽभिधीयते ॥ ४०९ ॥
यदन्तराले ज्ञानं तु पदार्थेषूपजायते ।
प्रतिपत्तेरूपायोऽसौ प्रक्रमानवधारणात् ॥ ४१० ॥

padarūpaṃ tu yadvākyamastitvopanibandhanam |
kāmaṃ vimarśastatrāyaṃ na vākyāvayave pade || 408 ||
yathaivānarthakervarṇaiviṃśiṣṭārtho'bhidhīyate |
padairanarthakairevaṃ viśiṣṭārtho'bhidhīyate || 409 ||
yadantarāle jñānaṃ tu padārtheṣūpajāyate |
pratipatterūpāyo'sau prakramānavadhāraṇāt || 410 ||

(408) But that sentence which has the form of a word, grounded in existence — let this reflection be there, granted, and not in the word that is a member of a sentence. (409) Just as, by meaningless phonemes, a qualified meaning is expressed — so, by meaningless words, a qualified meaning is expressed. (410) But the knowledge that arises in the interval, with respect to the word-meanings, is a means to the comprehension, on account of the non-ascertainment of the commencement.

Three verses that state the kanda's architecture one last time, now in terms of existence. (408): a pada-rupam vakyam — a sentence *having the form of a word* (the one-word sentence, 2.270, 2.326) — is astitva-upanibandhanam, *grounded in existence*: its whole predication is the assertion that something is. Grant the reflection *there*, the verse says — kamam vimarsah tatra — and *not* in the word that is a *member* of a sentence: the isolated word that constitutes a sentence has a predication; the word inside a sentence has none. (409) then draws the parity that has been the kanda's constant: as anarthakaih varnaih — by *meaningless phonemes* — a qualified meaning is expressed, so padaih anarthakaih — by *meaningless words*: the words within a sentence are as semantically null as the phonemes within a word, and the sentence means all the same. (410) rescues the intermediate cognitions from that nullity: the knowledge that arises antarale, *in the interval*, concerning the word-meanings, is pratipatteh upayah — a *means* to the comprehension: not a contribution but a ladder (2.38's discardable means; 1.85's productive error), and it is needed prakrama-anavadharanat, because the commencement is not ascertained — because one does not know, mid-utterance, where the whole will land.

2.411 unformalized

पूर्वैरर्थैरनुगतो यथार्थात्मा परः परः ।
संसर्ग एव प्रकान्तस्तथाऽन्येष्वर्थवस्तुषु ॥ ४११ ॥

pūrvairarthairanugato yathārthātmā paraḥ paraḥ |
saṃsarga eva prakāntastathā'nyeṣvarthavastuṣu || 411 ||

As each successive meaning-self is accompanied by the prior meanings, so, in other meaning-things, the commingling alone is what is commenced.

The phenomenology of the interval (2.410) described. Purvaih arthaih anugatah — accompanied by the *prior* meanings — yatha arthatma parah parah: each successive meaning-self, one after another, comes attended by what preceded it: no word arrives bare; each is met by a mind already shaped by the words before it. And the conclusion: samsarga eva prakrantah tatha anyesu artha-vastusu — the *commingling alone* is what is *commenced*, in the other meaning-things as well. What is under way through the interval is not a series of meanings but a *commingling* — a single process of fusion whose stages are not stages of a sum. This is why 2.410 could call the interval-cognitions a means without calling them constituents: what they are doing is not accumulating but converging, and 2.143's flash is the convergence, not its total.

2.412 unformalized

अङ्गीकृते तु केषाञ्चित् साध्येनार्थेन साधने ।
आधारनियमार्थैव साधनानां पुनः श्रुतिः ॥ ४१२ ॥

aṅgīkṛte tu keṣāñcit sādhyenārthena sādhane |
ādhāraniyamārthaiva sādhanānāṃ punaḥ śrutiḥ || 412 ||

But when, by some, the means is admitted along with the to-be-accomplished meaning — the re-hearing of the means is only for the sake of fixing the substratum.

A technical case: why sentences state what their verbs already imply. When a verb is uttered, its means are implied (2.183: the action does not arise apart from means-connection; 2.47: the roles come pre-fixed). So why does the sentence bother to name them? Sadhananam punah srutih — the *re-hearing* of the means: adhara-niyama-artha eva — *only for the sake of fixing the substratum*: the verb implies *some* agent, and the noun tells you *which*. So the nouns are not adding the roles; the roles were there in the verb (2.48's architecture). They are specifying the fillers. This is 2.66's structure exactly — vrihi restricting the substance-in-general that yajeta obtained — and it shows the same layering: what the sentence *says* and what it *fixes* are different offices, and most of a sentence's words are doing the second.

2.413 unformalized

आधारनियमाभावात् तदाक्षेपो न विद्यते ।
सामर्थ्यात्सम्भवस्तस्य श्रुतिस्त्वन्यनिवृत्तये ॥ ४१३ ॥

ādhāraniyamābhāvāt tadākṣepo na vidyate |
sāmarthyātsambhavastasya śrutistvanyanivṛttaye || 413 ||

In the absence of the fixing of the substratum, its implication does not exist; its possibility is from capacity, but the explicit hearing is for the exclusion of others.

The complement to 2.412, and it is a precision about implication. Adhara-niyama-abhavat tad-aksepah na vidyate: in the absence of a *fixing* of the substratum, the *implication* of it does not exist — an unfixed role is not implied at all; there is no floating 'some agent' hovering until named. Sambhavah tasya samarthyat: its *possibility* is from *capacity* — the verb's nature makes an agent possible (an action of that sort could have one). Srutih tu anya-nivrttaye — but the *explicit hearing* is for the *exclusion of others*: naming the agent does not create him, it excludes the alternatives. So the sentence's words are, one and all, exclusion-devices: they narrow a plenitude of possibility that the verb's capacity opened. This is 2.250-251's semantics at the level of syntax — the omni-significant carved by nimittas — and it explains why a sentence's meaning is not the sum of its words: subtraction does not sum.

2.414 unformalized

क्रिया क्रियान्तराद्भिन्ना नियताधारसाधना ।
प्रक्रान्ता प्रतिपत्तॄणां भेदाः सम्बोधहेतवः ॥ ४१४ ॥

kriyā kriyāntarādbhinnā niyatādhārasādhanā |
prakrāntā pratipattṝṇāṃ bhedāḥ sambodhahetavaḥ || 414 ||

An action, distinct from other actions, having a fixed substratum and means, once commenced — the differences are causes of awakening for the comprehenders.

The result of 2.412-413's exclusion-work, and it recasts the whole sentence as a waking-device. Kriya kriyantarad bhinna — an action *distinct from other actions*, individuated by the exclusions the words performed — niyata-adhara-sadhana, having its substratum and means now *fixed* — prakranta: commenced. And the office of everything that did the fixing: bhedah sambodha-hetavah pratipattrnam — the *differences* are *causes of awakening* for the comprehenders. Sambodha: awakening (2.331's pratibodha-abhyupayah, words as means of awakening). The words' distinctions do not build the meaning; they wake it — narrowing until the hearer's ripened capacity (2.152's pratibha) catches, and the flash occurs. The verse gathers the movement: exclusion, fixing, and waking are one process, and the meaning was never among the words that performed it.

2.415 unformalized

अविभागं तु शब्दभ्यः क्रमवद्भ्योऽपदक्रमम् ।
प्रकाशते तदन्येषां वाक्यं वाक्यार्थ एव च ॥ ४१५ ॥

avibhāgaṃ tu śabdabhyaḥ kramavadbhyo'padakramam |
prakāśate tadanyeṣāṃ vākyaṃ vākyārtha eva ca || 415 ||

But to others, the sentence and the sentence-meaning both shine forth as undivided, without sequence of words, from words that possess sequence.

The doctrine stated in its final form, and the symmetry is the point. Sabdebhyah kramavadbhyah — from words *possessing sequence*: the utterance is serial, its words come in order, nothing about the vehicle is undivided — prakasate avibhagam apada-kramam: there shines forth what is *undivided*, *without word-sequence*. And what shines: tad anyesam vakyam vakyartha eva ca — to others, *both* the sentence *and* the sentence-meaning. Both terms of the relation are undivided; only the vehicle has parts. This is 1.48's structure (the sequenceless born as if in sequence) at the sentence's scale, and it completes the parallel the kanda has been drawing since 2.30: as the word-sphota is manifested by sequential sounds, the sentence-sphota is manifested by sequential words, and in neither case does the manifestor's seriality touch the manifested. The one word 'prakasate' carries the whole doctrine: not built, not summed, not connected — *shone forth*.

2.416 unformalized

स्वरूपं विद्यते यस्य तस्यात्मा न निरूप्यते ।
नास्ति यस्य स्वरूपं तु तस्यैवात्मा निरूप्यते ॥ ४१६ ॥

svarūpaṃ vidyate yasya tasyātmā na nirūpyate |
nāsti yasya svarūpaṃ tu tasyaivātmā nirūpyate || 416 ||

That which has an own-form — of it, the self is not described; but of that which has no own-form, of that alone is the self described.

A paradox stated with deliberate sharpness, and it concerns the limits of description. Svarupam vidyate yasya tasya atma na nirupyate: that which *has* an own-form — a determinate nature of its own — of it the *self* is *not described*: what is fully itself cannot be caught in a description, since description would have to stand outside it and there is no outside (2.144's pratibha, not describable even by its possessor; 1.35's untellable expertise). Nasti yasya svarupam tu tasya eva atma nirupyate: but of what *has no* own-form — of *that alone* is the self described. What can be described is exactly what has no nature of its own — the constructed, the relational, the item that is nothing but its place in a scheme (the sambandha of 2.46, asattva-bhuta, beyond thinghood, and knowable only by inference from its effect). So description tracks unreality. The verse is the epistemological ground of 2.233's claim that the sciences describe only avidya: they describe because their objects are constructions, and what is not a construction is not available for describing. It is also the deepest reason the vakyartha must be a flash: it has an own-form, and so it can only be waked, never told.
contested: the verse is aphoristic and admits an idealist reading (only the unreal is describable, hence description as such is of the unreal) and a narrower semantic one (self-standing items resist definition; relational ones are definable); we take the first as the kanda's own, with 2.233 and 2.421-422 as warrant

2.417 unformalized

अशब्दमपरेऽर्थस्य रूपनिर्धारणं विदुः ।
शर्थावभासरूपा हि शब्देभ्यो जायते स्मृतिः ॥ ४१७ ॥

aśabdamapare'rthasya rūpanirdhāraṇaṃ viduḥ |
śarthāvabhāsarūpā hi śabdebhyo jāyate smṛtiḥ || 417 ||

Others know the determination of the meaning's form to be wordless; for the memory that arises from words has the form of an appearance of the meaning.

The consequence of 2.416 for semantics. Asabdam apare arthasya rupa-nirdharanam viduh: others know the *determination of the meaning's form* to be *wordless* — what fixes a meaning is not itself verbal (2.16's objection, once feared, now embraced: the sentence-meaning is asabda). And the reason: artha-avabhasa-rupa hi sabdebhyo jayate smrtih — the *memory* that arises from words has the form of an *appearance of the meaning*: what words produce in a hearer is not a word but a showing — an avabhasa, a shining-forth of the thing (2.415's prakasate). So the terminus of a verbal process is non-verbal: words wake an appearing, and the appearing is what meaning is. This does not sever the word-meaning bond (2.332 secured it) — it locates the bond's *effect*: the eternal connection delivers, at its far end, something that is not language. Which is why 1.123's thesis (all cognition is word-interwoven) and this verse are compatible: the appearing is word-woven in its structure and wordless in its presentation.

2.418 unformalized

अन्यथैवाग्निसम्बन्धाद्दाहं दग्धोऽभिमन्यते ।
अन्यथा दाहशब्देन दाहाद्यर्थः प्रतीयते ॥ ४१८ ॥

anyathaivāgnisambandhāddāhaṃ dagdho'bhimanyate |
anyathā dāhaśabdena dāhādyarthaḥ pratīyate || 418 ||

It is in one way that the burnt man, from connection with fire, experiences the burning; in quite another way that the meaning of burning and the rest is understood by the word 'burning'.

The gap between the meant and the had, put with unusual directness. Anyatha eva agni-sambandhad daham dagdho 'bhimanyate: it is *in one way entirely* that the *burnt man*, from connection with fire, experiences the burning — the searing itself, undergone. Anyatha daha-sabdena dahady-arthah pratiyate: in *quite another* way is the meaning of burning understood by the *word* 'burning'. The word does not burn. What it wakes is an appearing (2.417), and the appearing of burning is not a burning — no hearer of the word is scorched. The verse marks the limit of the whole semantic enterprise with a directness the kanda has earned: language reaches everything and touches nothing; it can wake an appearance of any experience and cannot supply one. This is also why 2.416 said what it said: the burning has an own-form and is therefore indescribable in the strict sense — what the word conveys is not it, but an appearance answering to it.

2.419-420 unformalized

पृथङ्निविष्टतत्त्वानां पृथगर्थानुपातिनाम् ।
इन्द्रियाणां यथा कार्यमृते देहान्न कल्पते ॥ ४१९ ॥
तथा पदानां सर्वेषां पृथगर्थनिवेशिनाम् ।
वाक्येभ्यः प्रविभक्तानामर्थवत्ता न विद्यते ॥ ४२० ॥

pṛthaṅniviṣṭatattvānāṃ pṛthagarthānupātinām |
indriyāṇāṃ yathā kāryamṛte dehānna kalpate || 419 ||
tathā padānāṃ sarveṣāṃ pṛthagarthaniveśinām |
vākyebhyaḥ pravibhaktānāmarthavattā na vidyate || 420 ||

(419) As the effect of the sense-faculties — whose realities are separately lodged, and which severally follow their own objects — is not construed apart from the body; (420) so, of all the words, which severally settle on their own meanings, when they are divided out from sentences, there is no meaningfulness.

One of the finest similes in the work, and it answers 2.391's challenge definitively. (419): the indriyas — the senses — have prthag-nivista-tattvanam, their *realities separately lodged* (eye in the socket, ear in the head: really distinct organs); and they are prthag-artha-anupatinah, *severally following their own objects* (eye to colour, ear to sound, no overlap). So they are as separate as things can be. And yet: yatha karyam rte dehan na kalpate — their *effect* is not construed *apart from the body*: an eye out of a head sees nothing; the organ's separateness is real and its functioning is not its own. (420): tatha padanam sarvesam prthag-artha-nivesinam — so of *all* words, which *severally settle* on their own meanings — vakyebhyah pravibhaktanam arthavatta na vidyate: *divided out* from sentences, there is *no meaningfulness*. The simile grants the atomist everything he asked for — words really are distinct, really do have their own meanings, as eyes really are distinct organs with their own objects — and takes back the only thing that mattered: none of it functions in isolation. The word in the sentence is the eye in the head.

2.421-422 unformalized

संसर्गरूपं संसृष्टेष्वर्थवस्तुषु गृह्यते ।
नात्रोपाख्यायते तत्वमपदार्थस्य दर्शनात् ॥ ४२१ ॥
दर्शनस्यापि यत्सत्यं न तथा दर्शनां स्थित्म् ।
वस्तुसंसर्गरूपेण तदरूपं निरूप्यते ॥ ४२२ ॥

saṃsargarūpaṃ saṃsṛṣṭeṣvarthavastuṣu gṛhyate |
nātropākhyāyate tatvamapadārthasya darśanāt || 421 ||
darśanasyāpi yatsatyaṃ na tathā darśanāṃ sthitm |
vastusaṃsargarūpeṇa tadarūpaṃ nirūpyate || 422 ||

(421) The form of commingling is grasped among meaning-things that are commingled; here the reality is not told, because of the seeing of what is not a word-meaning. (422) Even of the seeing, what is true is not established as the seeing is; that formless thing is described by way of the form of the commingling of things.

Two verses at the edge of what the kanda can say. (421): samsarga-rupam samsrstesu artha-vastusu grhyate — the *form of commingling* is grasped among commingled meaning-things: what one gets is the fusion, not the items. And: na atra upakhyayate tattvam apadarthasya darsanat — *here the reality is not told*, because of the seeing of *what is not a word-meaning*: what is seen is no padartha at all, so no word-meaning-talk reaches it (2.416: what has own-form is not described). (422) goes further and turns on the seeing itself: darsanasya api yat satyam na tatha darsanam sthitam — even of the *seeing*, what is *true* is not established as the seeing *is*: the seeing's own truth is not the seeing as it presents. Vastu-samsarga-rupena tad arupam nirupyate: that *formless* thing is described by way of the *form of the commingling of things*. So the formless is described in the only available idiom — the idiom of commingled things — and the description is therefore not of it. This is the kanda's most rigorous moment of self-limitation: it says that its own central object cannot be said, and says why, and continues.
contested: 422 is among the most difficult verses in the work and our construal is one of several; the transcription's sthitm for sthitam is a clear fault. What is not in doubt is the verse's direction: the formless described through the form of fusion, and the description acknowledged as inadequate

2.423 unformalized

अस्तित्वेनानुषक्तो वा निवृत्तयात्मनि वा स्थितः ।
अर्थोऽभिधीयते यस्मादतो वाक्यं प्रयुज्यते ॥ ४२३ ॥

astitvenānuṣakto vā nivṛttayātmani vā sthitaḥ |
artho'bhidhīyate yasmādato vākyaṃ prayujyate || 423 ||

Since a meaning is expressed either as accompanied by existence, or as standing in a withdrawn self — therefore a sentence is employed.

The definition of the sentence, reached from the side of existence. A meaning is expressed in one of two ways: astitvena anusaktah — *accompanied by existence*: 'the pot is', the predication affirming; or nivrttya atmani sthitah — *standing in a withdrawn self*: 'the pot is not', the predication denying, the meaning present but with existence withheld (2.240-245's negation-dossier, now given its positive statement). Atah vakyam prayujyate: *therefore a sentence is employed*. The inference is the definition: a sentence is what does *this* — what places a meaning in relation to existence, either way. This is why a word cannot be a sentence unless its existence-verb is discharged (2.270's carita-asti-kriya, 2.408's astitva-upanibandhanam), and it is the deepest answer the kanda gives to its title-question. A vakya is not a string of a certain length or structure. It is an assertion of being or non-being.

2.424 unformalized

क्रियानुषङ्गेण विना न पदार्थः प्रतीयते ।
सत्यो वा विपरीतो वा व्यवहारे न सोऽस्त्यतः ॥ ४२४ ॥

kriyānuṣaṅgeṇa vinā na padārthaḥ pratīyate |
satyo vā viparīto vā vyavahāre na so'styataḥ || 424 ||

Without the accompaniment of an action, a word-meaning is not understood — whether real or contrary; therefore it does not exist in practice.

The corollary, and it is severe. Kriya-anusangena vina na padarthah pratiyate: without the *accompaniment of an action*, a word-meaning is *not understood* — satyo va viparito va, whether *real or contrary*: neither a true one nor a false one; not at all. A meaning unattached to any action, any predication, any being-or-not-being, does not present itself to anyone. Atah na so 'sti vyavahare: *therefore it does not exist in practice*. Note the qualification: vyavahare — *in practice*. The bare word-meaning is not declared unreal simpliciter; it is declared absent from practice, which by 2.297 (sabdah loka-nibandhanah: words are grounded in the world) is where words live. So the padartha is an item with no place in the only domain language occupies — the exact status the kanda has assigned it since 2.10: real in the ledger, absent from the transaction.

2.425 unformalized

सदित्येव तु यद्वाक्यं तदभूदस्ति नेति वा ।
क्रियाभिधानसम्बन्धमन्तरेण न मन्यते ॥ ४२५ ॥

sadityeva tu yadvākyaṃ tadabhūdasti neti vā |
kriyābhidhānasambandhamantareṇa na manyate || 425 ||

But that sentence which is just 'sat' — 'it was', 'it is', or 'not' — is not conceived apart from the connection with the expressing of an action.

The limit-case tested: the barest possible sentence. Sad iti eva yad vakyam — a sentence that is just *'sat'*, 'being' — tad abhut asti na iti va: 'it was', 'it is', or 'not'. Even here, at the minimum, kriya-abhidhana-sambandham antarena na manyate — it is not conceived *apart from* the connection with the *expressing of an action*: the existence-verb is an action-word, and 'sat' predicates only because it carries one. There is no assertion so thin that it escapes the verb. The verse forecloses an obvious escape from 2.423-424 — the thought that pure existence-claims might be predication without action — and in doing so it makes the action-centred account of the sentence complete: every sentence has a verb, even the ones that are only a verb, and the verb is what makes the sentence a sentence rather than a naming.

2.426-427 unformalized

आख्यातपदवाच्येऽर्थे साधनोपनिबन्धने ।
विना सत्तवाभिधानेन नाकाङ्क्षा विनिवर्तते ॥ ४२६ ॥
प्राधान्यात्तु क्रिया पूर्वमर्थस्य प्रविभज्यते ।
साध्यप्रयुक्तान्यङ्गनि फलं तस्याः प्रयोजकम् ॥ ४२७ ॥

ākhyātapadavācye'rthe sādhanopanibandhane |
vinā sattavābhidhānena nākāṅkṣā vinivartate || 426 ||
prādhānyāttu kriyā pūrvamarthasya pravibhajyate |
sādhyaprayuktānyaṅgani phalaṃ tasyāḥ prayojakam || 427 ||

(426) In the meaning expressible by a finite verb, grounded in the means, expectancy does not withdraw without the expressing of existence. (427) But from principality, the action is first divided out of the meaning; the auxiliaries are employed for the sake of the to-be-accomplished; its fruit is the occasioner.

The architecture of the sentence, stated completely. (426): in what a finite verb expresses — sadhana-upanibandhane, *grounded in the means* — na akanksa vinivartate vina sattva-abhidhanena: *expectancy does not withdraw* without the *expressing of existence*: the sentence does not close (2.4's nirakanksata, the mark of the vakya) until being has been predicated. So existence-assertion is the closure-condition. (427): pradhanyat tu kriya purvam arthasya pravibhajyate — from *principality*, the *action* is *first* divided out of the meaning: when the analyst carves, the verb comes off first, being the principal (2.4's karma-pradhanam). Sadhya-prayuktani angani: the *auxiliaries* are employed for the sake of the *to-be-accomplished*. Phalam tasyah prayojakam: its *fruit* is the *occasioner*. The chain is complete and it runs backwards from the end: the fruit occasions the action, the action requires the means, the means are named by the nouns — and the sentence closes when existence is predicated of the whole. Purpose at the top (2.113), existence at the bottom, and everything between is auxiliary.

2.428 unformalized

प्रयोक्तैवाभिसन्धत्ते साध्यसाधनरूपताम् ।
अर्थस्य वाऽभिसम्बन्धकल्पनां प्रसमीहते ॥ ४२८ ॥

prayoktaivābhisandhatte sādhyasādhanarūpatām |
arthasya vā'bhisambandhakalpanāṃ prasamīhate || 428 ||

It is the speaker alone who intends the character of to-be-accomplished and means; or he desires the constructing of the meaning's connection.

The architecture of 2.427 handed to the speaker. Prayokta eva abhisandhatte sadhya-sadhana-rupatam: it is the *speaker alone* who *intends* the character of to-be-accomplished and means — which element shall be the action and which the instrument is not read off the world; it is conferred by the intending (2.131's theorem: the karaka-grid is constructed by words alone). Artthasya va abhisambandha-kalpanam prasamihate: or he *desires the constructing* of the meaning's connection. The same event can be carved as a cooking-with-fire or a fire's-burning-under-a-pot; the world does not choose, the speaker does. This is 2.400's directed eye applied to syntax rather than to reference: intention aims not only *at* a meaning but *into* an articulation — deciding what shall count as the action. And the next verse gives the demonstration.

2.429 unformalized

पचिक्रिया करोमीति कर्मत्वेनाभिधीयते ।
पक्तिः करणरूपं तु साध्यत्वेन प्रतीयते ॥ ४२९ ॥

pacikriyā karomīti karmatvenābhidhīyate |
paktiḥ karaṇarūpaṃ tu sādhyatvena pratīyate || 429 ||

In 'I do the act of cooking', the cooking-action is expressed as the object; but the cooking, in the form of the instrument, is understood as the to-be-accomplished.

The demonstration, and it is elegant. Take paci-kriyam karomi — 'I do the cooking'. Here the *cooking* — which is an action — is karmatvena abhidhiyate, expressed *as the object*: it has become the patient of 'do'. And yet paktih karana-rupam tu sadhyatvena pratiyate — the cooking, in the *form of the instrument*, is understood as the *to-be-accomplished*: it is still what is being brought about. One and the same cooking is simultaneously object (of 'do'), instrument (of the doing), and the sadhya (what is accomplished) — three karaka-roles at once, assigned by how the sentence was built. Nothing in the cooking determines which; the speaker's construction does (2.428). This single example refutes any view on which the karakas are read off events: an event's roles are as many as the sentences one can make about it, and the sentence-maker assigns them.

2.430 unformalized

योऽसौ येनोपकारेण प्रयोक्तॄणां विवक्षितः ।
अर्थस्य सर्वशक्तित्वात् स तथैव व्यवस्थितः ॥ ४३० ॥

yo'sau yenopakāreṇa prayoktṝṇāṃ vivakṣitaḥ |
arthasya sarvaśaktitvāt sa tathaiva vyavasthitaḥ || 430 ||

That which is intended by speakers, with whatever assistance — since the meaning is all-capable, that stands settled just so.

The movement closes by folding syntax into the omni-significance doctrine. Yo 'sau yena upakarena prayoktrnam vivaksitah — whatever is *intended* by speakers, with whatever *assistance* (in whatever role, contributing however) — arthasya sarva-saktitvat sa tatha eva vyavasthitah: *since the meaning is all-capable*, that stands settled *just so*. Sarva-saktitva — the same all-capacity 2.253 predicated of the word (sarva-sakti) and 2.250 of its meaning (sarvarthya) — is here predicated of the *meaning* to explain the karaka-freedom of 2.429: the cooking can be object, instrument, or accomplished because the meaning can be anything, and the speaker's vivaksa selects. The doctrine has now been stated at every level of the language: the word means everything and says one thing; the meaning plays every role and takes one; and in both cases the carving is intention's, working on a plenitude. Kanda III will spend twelve hundred verses examining the categories this carving produces — and 2.430 is its charter.

2.431-433 · Connection, Assistance, and the Word's Sovereignty unformalized

आराद् वृत्तिषु सम्बन्धः कदाचिदभिधीयते ।
आश्लिष्टो योऽनुपश्लिष्टः स कदाचित् प्रतीयते ॥ ४३१ ॥
संसृष्टानां विवेकित्वं संसर्गश्च विवेकिनाम् ।
नानात्मकानामेकत्वं नानात्वं च विपर्यये ॥ ४३२ ॥
सर्वात्मकत्वादर्थस्य नैरात्म्याद्वा व्यवस्थितम् ।
अत्यन्तयतशक्तित्वाच्छब्द एव निबन्धनम् ॥ ४३३ ॥

ārād vṛttiṣu sambandhaḥ kadācidabhidhīyate |
āśliṣṭo yo'nupaśliṣṭaḥ sa kadācit pratīyate || 431 ||
saṃsṛṣṭānāṃ vivekitvaṃ saṃsargaśca vivekinām |
nānātmakānāmekatvaṃ nānātvaṃ ca viparyaye || 432 ||
sarvātmakatvādarthasya nairātmyādvā vyavasthitam |
atyantayataśaktitvācchabda eva nibandhanam || 433 ||

(431) In derivative formations, a connection is sometimes expressed from afar; what is unattached is sometimes understood as attached. (432) Of the commingled there is discreteness, and commingling of the discrete; of the many-natured, oneness — and manifoldness in the reverse case. (433) Whether from the meaning's being all-natured, or from its being self-less, it stands settled; because capacity is utterly fixed, the word alone is the ground.

Three verses that state the movement's thesis by first cataloguing the inversions language performs. (431): in the vrttis, connection is expressed arat, *from afar* — relations asserted between things not adjacent; and aslisto yo 'nupaslistah — what is *unattached* is understood as *attached*. (432) makes the inversions systematic and total: samsrstanam vivekitvam — of the *commingled*, discreteness (language separates what is fused); samsargas ca vivekinam — and commingling of the *discrete* (it fuses what is separate); nanatmakanam ekatvam — of the *many-natured*, oneness; nanatvam ca viparyaye — and manifoldness in the reverse. Every possible inversion, and language performs them all. Then (433) the conclusion, with its startling either/or: sarvatmakatvad arthasya nairatmyad va vyavasthitam — whether because the meaning is *all-natured* (able to be anything, 2.430's sarva-saktitva) *or* because it is *self-less* (having no nature of its own to resist) — it stands settled thus. Bhartrhari does not choose between plenitude and emptiness; either grounds the same result. And the result: atyanta-yata-saktitvat sabda eva nibandhanam — because capacity is *utterly fixed*, the *word alone* is the ground. The word is sovereign because the meaning offers no resistance, whether from having every nature or none.
contested: 433's sarvatmakatvad / nairatmyad va is the crux: the first is the kanda's own sarvarthya-doctrine (2.250), the second a nairatmya-formulation with unmistakable Buddhist resonance. That Bhartrhari treats them as interchangeable grounds for one conclusion is remarkable and, we think, deliberate

2.434 unformalized

वस्तूपलक्षणं शब्दो नापकारस्य वाचकः ।
न स्वशक्तिः पदार्थानां संस्प्रष्टुं तेन शक्यते ॥ ४३४ ॥

vastūpalakṣaṇaṃ śabdo nāpakārasya vācakaḥ |
na svaśaktiḥ padārthānāṃ saṃspraṣṭuṃ tena śakyate || 434 ||

The word is an indicator of the thing; it is not the expressor of the assistance. The own-capacity of word-meanings cannot be touched by it.

A precise limitation on what the word does. Vastu-upalaksanam sabdah: the word is an *indicator of the thing* — upalaksana, a mark that points out without describing (2.309's saptaparna: naming without characterizing). Na upakarasya vacakah: it is *not the expressor of the assistance* — of the contribution one meaning makes to another, the relational work that goes on among the items. And: na sva-saktih padarthanam samsprastum tena sakyate — the *own-capacity* of word-meanings cannot be *touched* by it. The word points at things; it does not reach into their powers, does not manipulate what they can do, does not express the assistance they render one another in a construction. This is why the karaka-relations must be conferred by the speaker (2.428) rather than expressed by the words: the words are indicators, and indicators indicate — they do not perform the joinery. The verse limits the word's sovereignty of 2.433 in exactly the right place: sovereign over what is meant, powerless over what the meant things are.

2.435 unformalized

सम्बन्धिधर्मा संयोगः स्वशब्देनाभिधीयते ।
सम्बन्धः समवायश्च सम्बन्धित्वेन गम्यते ॥ ४३५ ॥

sambandhidharmā saṃyogaḥ svaśabdenābhidhīyate |
sambandhaḥ samavāyaśca sambandhitvena gamyate || 435 ||

Conjunction, a property of the relata, is expressed by its own word; connection and inherence are understood by way of being relata.

The taxonomy of relation, and it discriminates carefully. Samyoga — *conjunction*, the Vaisesika category (two substances in contact) — is sambandhi-dharma, a *property of the relata*, and svasabdena abhidhiyate: expressed by *its own word*: 'conjunction' names it, as any word names any thing, because it is a thing — a property residing in its terms. But sambandhah samavayas ca — *connection* as such, and *inherence* (the Vaisesikas' other relation, binding qualities to substances) — sambandhitvena gamyate: are understood *by way of being relata* — i.e. not named directly but grasped only through the terms they relate. The distinction is exact: conjunction is a property and so nameable; relation-as-such is not a property of anything and so is reachable only obliquely (2.46: the relation is inferable from its effect, its form is never seen, it is utterly beyond thinghood). The verse is the third kanda's Sambandha-samuddesa in miniature — and it shows Bhartrhari treating the rival school's categories with technical care before dissolving them.

2.436 unformalized

लक्षणाद् व्यवतिष्ठन्ते पदार्था न तु वस्तुतः ।
उपकारात् स एवार्थः कथञ्चिदनुगम्यते ॥ ४३६ ॥

lakṣaṇād vyavatiṣṭhante padārthā na tu vastutaḥ |
upakārāt sa evārthaḥ kathañcidanugamyate || 436 ||

Word-meanings stand settled by way of characterization, not in reality; from the assistance, that very meaning is somehow followed along.

The status of the padartha, stated flatly. Laksanad vyavatisthante padarthah na tu vastutah: word-meanings stand settled *by way of characterization*, and *not in reality* — laksana, the marking-out by which analysis fixes an item, is what gives a word-meaning its determinacy, and the determinacy is the marking's, not the thing's. And the concession: upakarat sa eva arthah kathamcid anugamyate — from the *assistance*, that very meaning is *somehow* followed along. Kathamcid — 'somehow': the shrug is the verse's honesty. The word-meanings do have a life, traceable through the assistance they render in constructions; but the tracing is of something the characterization posited, and how it manages to be followed is not further explained. This is 2.10's apoddhara-doctrine in its final formulation: the padartha is a creature of laksana, real in the practice, and its reality *in* the practice is not nothing — it is just not vastu.

2.437 unformalized

वाक्यार्थे योऽभिसम्बन्धो न तस्यात्मा क्वचित् स्थितः ।
व्यवहारे पदार्थानां तमात्मानं प्रचक्षते ॥ ४३७ ॥

vākyārthe yo'bhisambandho na tasyātmā kvacit sthitaḥ |
vyavahāre padārthānāṃ tamātmānaṃ pracakṣate || 437 ||

The connection which is in the sentence-meaning — its self does not stand anywhere; in the practice of word-meanings, they call that its self.

The connection dissolved, and its ghost given an address. Vakyarthe yo 'bhisambandhah — the connection that is *in* the sentence-meaning (the relatedness by which the items cohere, 2.145's upaslesa) — na tasya atma kvacit sthitah: its *self does not stand anywhere*. It is nowhere: not in the words (which only indicate, 2.434), not among the meanings (which are characterizations, 2.436), not as a thing (2.46's asattva-bhuta). And then: vyavahare padarthanam tam atmanam pracaksate — *in the practice* of word-meanings, they call *that* its self: the practice has a name for this nothing, treats it as an item, and transacts with it. So the connection is a posting with no referent, indispensable to the practice and absent from the world — which is the sharpest thing the kanda has said about the central concept of every rival semantics.

2.438 unformalized

पदार्थे समुदाये वा समाप्तो नैव वा क्वचित् ।
पदार्थरूपभेदेन तस्यात्मा प्रविभज्यते ॥ ४३८ ॥

padārthe samudāye vā samāpto naiva vā kvacit |
padārtharūpabhedena tasyātmā pravibhajyate || 438 ||

Whether completed in the word-meaning, or in the aggregate, or nowhere at all — its self is divided out by way of the difference of word-meaning-forms.

The three options for the connection's location, and their equivalence. Padarthe — completed in the *word-meaning* (each item carrying its own relatedness); samudaye va — or in the *aggregate* (the relation belonging to the collection); samapto na eva va kvacit — or *nowhere at all* (2.437's verdict). Whichever: tasya atma pravibhajyate padartha-rupa-bhedena — its self is *divided out* by way of the *difference of word-meaning-forms*. The connection, wherever one places it, gets its apparent articulation from the items it supposedly relates — which is circular in exactly the way 2.46 diagnosed: the relation is known only from its terms, and the terms are individuated by the relation. Bhartrhari lists the three placements without choosing because the circularity infects all three. What escapes it is his own account: no relation, one meaning, and the 'items' its capacities (2.394's sadharana-model).

2.439-440 unformalized

अन्वाख्यानाय यो भेदः प्रतिपत्तिनिबन्धनम् ।
साकाङ्क्षावयवं भेदे तेनान्यदुपवर्ण्यते ॥ ४३९ ॥
अनेकशक्तेरेकस्य प्रविभागोऽनुगम्यते ।
एकार्थत्वं हि वाक्यस्य मात्रयाऽपि प्रतीयते ॥ ४४० ॥

anvākhyānāya yo bhedaḥ pratipattinibandhanam |
sākāṅkṣāvayavaṃ bhede tenānyadupavarṇyate || 439 ||
anekaśakterekasya pravibhāgo'nugamyate |
ekārthatvaṃ hi vākyasya mātrayā'pi pratīyate || 440 ||

(439) The difference which, for the sake of analysis, is the ground of comprehension — by that, in the differentiation, something else, having mutually-requiring parts, is described. (440) The division of a single thing possessing many capacities is followed; for a sentence's single-meaningedness is understood even by a fraction.

The analysis's status, and then a striking empirical claim. (439): the *difference* posited anvakhyanaya, *for the sake of analysis*, and serving as pratipatti-nibandhanam, the ground of comprehension (the ladder, 2.410) — by it, tena anyad upavarnyate: *something else* is described: sakanksa-avayavam bhede — 'having mutually-requiring parts', which is 2.3-2.4's definition of the sentence. So the analytic definition of the vakya describes not the sentence but *something else* — the analysis's own construct. (440) then names what is really there: anekasakteh ekasya pravibhagah anugamyate — the *division* of a *single* thing possessing *many capacities* is what is being followed (the sadharana-model, 2.394-396: one meaning, many capacity-positions). And the evidence: ekarthatvam hi vakyasya matraya api pratiyate — the sentence's *single-meaningedness* is understood *even by a fraction*: hear part of a sentence and you already grasp that it has one meaning — the unity is given before the completion, which no summative account can explain (2.40's single phoneme conveying the whole).

2.441 unformalized

संप्रत्ययार्थाद् बाह्योऽर्थः सन्नसन्वा विभज्यते ।
बाह्यीकृत्य विभागस्तु शक्त्यपोद्धारलक्षणः ॥ ४४१ ॥

saṃpratyayārthād bāhyo'rthaḥ sannasanvā vibhajyate |
bāhyīkṛtya vibhāgastu śaktyapoddhāralakṣaṇaḥ || 441 ||

The external meaning, whether existent or non-existent, is divided out from the meaning of the comprehension; but the division, having made it external, has the mark of the abstraction of capacities.

The relation of comprehension to the world, and it is the movement's most careful formulation. Sampratyayarthad bahyo 'rthah san asan va vibhajyate: the *external* meaning — *whether existent or non-existent* — is *divided out from* the meaning of the comprehension. The order is the point: what is primary is the sampratyaya-artha, the meaning-in-comprehension (2.328's criterion: what is understood *is* the meaning); the 'external' object is then carved off from it, by us, and its existence or non-existence is a further question that the carving does not settle. And: bahyikrtya vibhagas tu sakty-apoddhara-laksanah — the division, *having made it external*, has the mark of the *abstraction of capacities*: externality itself is an apoddhara, an abstraction, and what it abstracts are capacities (2.88's settlement: abstraction according to capacity-difference). This is as close as the kanda comes to an explicit idealism, and it stops precisely where 2.132-133 stopped: the external is a division within the comprehended, and whether anything answers to it is not a question the semantics can reach.

2.442 unformalized

बहुष्वपि तिङ्न्तेषु साकाङ्क्षेष्वेकवाक्यता ।
तिङन्तेभ्यो निघातस्य पर्युदासस्तथार्थवान् ॥ ४४२ ॥

bahuṣvapi tiṅnteṣu sākāṅkṣeṣvekavākyatā |
tiṅantebhyo nighātasya paryudāsastathārthavān || 442 ||

Even among many finite verbs that are mutually requiring, there is single-sentence-hood; hence the exclusion of accent-suppression from the finite verbs is also meaningful.

The doctrine of 2.348 (rule and exception as one sentence) given its grammatical proof. Bahusv api tinantesu sakanksesu eka-vakyata: even among *many finite verbs*, if they are *mutually requiring*, there is *single-sentence-hood* — verb-count does not individuate sentences; expectancy does (2.6's tinanta qualifying tinanta). And the evidence from the sastra: tinantebhyo nighatasya paryudasah tatha arthavan — the *exclusion* of accent-suppression *from* the finite verbs is *thus meaningful*: Panini's nighata-rule (8.1.28: a verb loses its accent after a preceding word in the same sentence) has an exception for verbs following verbs — and that exception only makes sense if multiple verbs *can* stand in one sentence. The grammar's exception presupposes the doctrine. This is the kanda's favourite move (2.5's vocative, 2.59's pada-patha): a semantic thesis confirmed by a rule that would be unstatable without it.

2.443 unformalized

एकतिङ् यस्य वाक्यं तु शास्त्रे नियतलक्षणम् ।
तस्यातिङ्ग्रहणेनार्थो वाक्यभेदान्न विद्यते ॥ ४४३ ॥

ekatiṅ yasya vākyaṃ tu śāstre niyatalakṣaṇam |
tasyātiṅgrahaṇenārtho vākyabhedānna vidyate || 443 ||

But that sentence which has one finite verb, whose definition is fixed in the science — for it, there is no purpose in the mention of 'not-tin', because of the difference of sentences.

The alternative position and its cost. Suppose one holds eka-tin yasya vakyam — that a sentence has *one* finite verb, by a definition niyata-laksanam sastre, fixed in the science (the operational definition of 2.3, adopted as metaphysics). Then: tasya atin-grahanena artho na vidyate — for such a theorist, there is *no purpose* in the mention of 'not-tin' (the atin of the nighata-rule's exception) — vakya-bhedat, *because of the difference of sentences*: on his view, two verbs simply *are* two sentences, so the rule's exception is idle — no verb ever follows a verb within a sentence, and the exception excludes nothing. The argument is a clean reductio from grammatical economy: Panini does not legislate vacuously; his exception has work to do; therefore the one-verb definition is false. The verse thus turns the science's own laghava-principle (2.343) against a definition the science itself uses (2.3) — which is exactly what 2.3 warned: rule-definitions are for the rules' purposes and settle nothing beyond them.

2.444 unformalized

तिङन्तान्तरयुक्तेषु युक्तयुक्तेषु वा पुनः ।
मृगः पश्यत यातीति भेदाभेदौ न तिष्ठतः ॥ ४४४ ॥

tiṅantāntarayukteṣu yuktayukteṣu vā punaḥ |
mṛgaḥ paśyata yātīti bhedābhedau na tiṣṭhataḥ || 444 ||

In cases joined with another finite verb, or again in the joined-and-joined — 'the deer! look! it goes' — difference and non-difference do not stand.

The hard case, honestly displayed. Mrgah pasyata yati — 'The deer! Look! It goes' (or 'Look at the deer going'): three predications, or one utterance? Tinanta-antara-yuktesu — in cases joined with *another* finite verb — yukta-yuktesu va punah, or in the *joined-and-joined* (chains of them): bheda-abhedau na tisthatah — *difference and non-difference do not stand*: neither the claim that these are separate sentences nor the claim that they are one can be maintained. The utterance is genuinely indeterminate between one sentence and three, and no criterion settles it — not verb-count (2.442-443 disposed of that), not expectancy (each closes and none does), not the speaker's intent (he intended the whole scene). Bhartrhari lets the indeterminacy stand. It is the same undecidability he confessed at 2.26 (whether the sequencing cognition is one or many: niruktir na avatisthate) and at 1.74, and its recurrence at the kanda's climax is not a failure: it is what a doctrine of undivided meaning carved by analysis *predicts*. Where the carving is optional, the count is.

2.445 unformalized

इतिकर्तव्यताऽर्थस्य सामर्थ्याद्यस्य काङ्क्ष्यते ।
अशब्दलक्षणाकाङ्क्षं समाप्तार्थं तदुच्यते ॥ ४४५ ॥

itikartavyatā'rthasya sāmarthyādyasya kāṅkṣyate |
aśabdalakṣaṇākāṅkṣaṃ samāptārthaṃ taducyate || 445 ||

That whose what-is-to-be-done is desired, from capacity, for the meaning — having an expectancy not characterized by words, that is called 'of completed meaning'.

A definition of completion that severs it from words entirely. Itikartavyata arthasya samarthyad yasya kanksyate — that whose *what-is-to-be-done* (the practical remainder: what one must now do) is *desired*, from *capacity*, for the meaning — asabda-laksana-akanksam: having an expectancy *not characterized by words* — samapta-artham tad ucyate: *that* is called 'of completed meaning'. So a sentence is complete when its verbal expectancy has closed even though a *practical* expectancy remains open: 'the deer is going' is a complete sentence and leaves you wanting to know what to do about it. Asabda-laksana-akanksa — wordless expectancy — is the crucial coinage: the sentence's completion is measured against words, and the residue it leaves is measured against action. This distinguishes semantic closure from practical closure, and it explains why sentences can be complete and yet demand more: the more they demand is not linguistic (2.146's itikartavyata, sovereign and untransgressed).

2.446 unformalized

तत्त्वान्वाख्यानमात्रे तु यावानर्थोऽनुषज्यते ।
विनाऽपि तत्प्रयोगेण श्रुतेर्वाक्यं समाप्यते ॥ ४४६ ॥

tattvānvākhyānamātre tu yāvānartho'nuṣajyate |
vinā'pi tatprayogeṇa śrutervākyaṃ samāpyate || 446 ||

But in the mere analytic statement of the reality, however much meaning is carried along — even without the employment of that, the sentence is completed from the hearing.

The complement of 2.445, for the case of pure statement. Tattva-anvakhyana-matre tu — in the *mere analytic statement of the reality*: description for its own sake, no injunction, no practical residue — yavan arthah anusajyate, however much meaning is carried along: vina api tat-prayogena srutervakyam samapyate — *even without the employment of that*, the sentence is completed *from the hearing*. A descriptive sentence closes with its utterance; nothing further is wanted; there is no itikartavyata to leave open. Between 2.445 and 2.446 the two kinds of sentence are distinguished by what they leave: the practical sentence closes verbally and opens actionally; the descriptive closes altogether. And note which is treated as the special case: it is the *descriptive* one that needs the qualification 'even without employment'. For this school the default sentence is the one that tells you what to do — which is what a tradition whose paradigm text is a body of injunctions would naturally think, and what 2.113's prayojana-doctrine requires.

2.447 unformalized

चङ्क्रम्यमाणोऽधीष्वात्र जपंश्चङ्क्रमणं कुरु ।
तादर्थ्यस्याविशेषेऽपि शब्दाद्भेदः प्रतीयते ॥ ४४७ ॥

caṅkramyamāṇo'dhīṣvātra japaṃścaṅkramaṇaṃ kuru |
tādarthyasyāviśeṣe'pi śabdādbhedaḥ pratīyate || 447 ||

'While walking about, study here'; 'muttering, do the walking-about' — though the purposiveness is not different, a difference is understood from the word.

A beautiful demonstration that syntax means something even when semantics does not change. Two instructions: cankramyamano 'dhisva atra — 'while walking about, *study* here'; and japan cankramanam kuru — 'while muttering, *do the walking-about*'. The same two activities, the same pairing, the same practical upshot: the man walks and recites. Tadarthyasya avisese 'pi — though the *purposiveness is not different* — sabdad bhedah pratiyate: a *difference is understood from the word*. What differs is which activity is principal and which subordinate — and that difference is real, is understood, and is contributed by nothing but the grammatical construction (2.428: the speaker intends the sadhya-sadhana structure). The example is decisive against any view on which sentence-meaning is a function of the states of affairs described: here the states of affairs are identical and the meanings differ. What differs is the *articulation*, and articulation is the speaker's.

2.448 unformalized

फलवन्तः क्रियाभेदाः क्रियान्तरनिबन्धनाः ।
असंख्याताः क्रमोद्देशैरेकाख्यातनिदर्शनाः ॥ ४४८ ॥

phalavantaḥ kriyābhedāḥ kriyāntaranibandhanāḥ |
asaṃkhyātāḥ kramoddeśairekākhyātanidarśanāḥ || 448 ||

The differences of action, having fruits, grounded in other actions, are uncounted; by the designations of sequence, they have a single finite verb for their exhibition.

The verb's compression, stated as a fact about the world it reports. Kriya-bhedah phalavantah kriyantara-nibandhanah asamkhyatah: the differences of action — each *having its fruit*, each *grounded in other actions* — are *uncounted*: any action is indefinitely divisible into sub-actions, each with its own result, each depending on others (the cooking is a fetching, a lighting, a stirring, a waiting). And: kramoddesaih eka-akhyata-nidarsanah — by the *designations of sequence*, they have a *single finite verb* for their exhibition. One verb, 'he cooks', displays an unbounded manifold. This is the sadharana-model (2.394) at the verb: the single verb does not sum its sub-actions, it *is* them, wholly, from where it stands. And it explains 2.415's undivided shining: the vakyartha can be undivided while its analysis is inexhaustible, because analysis of a whole is not decomposition into parts.

2.449-450 unformalized

निवृत्तभेदा सर्वैव क्रियाऽऽख्यातेऽभिधीयते ।
श्रुतेरशक्या भेदानां प्रविभागप्रकल्पना ॥ ४४९ ॥
अश्वमेधेन यक्ष्यन्ते राजानः सत्रमास्ते ।
ब्राह्मणा इति नाख्यातरूपाद्भेदः प्रतीयते ॥ ४५० ॥

nivṛttabhedā sarvaiva kriyā''khyāte'bhidhīyate |
śruteraśakyā bhedānāṃ pravibhāgaprakalpanā || 449 ||
aśvamedhena yakṣyante rājānaḥ satramāste |
brāhmaṇā iti nākhyātarūpādbhedaḥ pratīyate || 450 ||

(449) The whole action, its differences withdrawn, is expressed in the finite verb; the constructing of a division of the differences is not possible from the hearing. (450) 'The kings will sacrifice with the horse-sacrifice'; 'the brahmins sit the session' — a difference is not understood from the finite verb's form.

The consequence of 2.448, with evidence. (449): nivrtta-bheda sarva eva kriya akhyate abhidhiyate — the *whole* action, its differences *withdrawn*, is expressed in the finite verb; sruteh asakya bhedanam pravibhaga-prakalpana: the constructing of a division of the differences is *not possible from the hearing* — listen as hard as you like to 'he cooks' and no sub-action is audible in it. (450) supplies two cases where the distribution question is live and the verb is silent: asvamedhena yaksyante rajanah — 'the *kings* will sacrifice with the horse-sacrifice' (each king his own? or one joint rite?); satram asate brahmanah — 'the brahmins *sit* the session' (the sattra is by definition collective, but the verb does not say so). Na akhyata-rupad bhedah pratiyate: a difference is *not understood from the finite verb's form*. The verb form is identical in the distributive and collective cases; only samarthya (2.371) tells them apart. Language does not encode the distinction that ritual practice most needs, and the exegetes have been supplying it from capacity ever since.

2.451 unformalized

सकृच्छ्रुता सप्तदशस्वनावृत्तापि या क्रिया ।
प्राजापत्येषु सामर्थ्यात् सा भेदं प्रतिपद्यते ॥ ४५१ ॥

sakṛcchrutā saptadaśasvanāvṛttāpi yā kriyā |
prājāpatyeṣu sāmarthyāt sā bhedaṃ pratipadyate || 451 ||

The action which, heard once, is not repeated among the seventeen — in the Prajapatya rites it attains difference, from capacity.

The doctrine at work in a real ritual crux. In the Prajapatya rites, seventeen items are involved and an action is sakrt sruta — *heard once*, uttered a single time, anavrtta, not repeated. Does it apply once, or seventeen times? Sa bhedam pratipadyate samarthyat: it *attains difference*, from *capacity* — the action distributes across the seventeen, though nothing in the utterance says so, because what the action *is* requires it (2.371's criterion; 2.372's eating completed in each). The case is chosen because the stakes are exact: get this wrong and the rite is malperformed. And it demonstrates the movement's thesis in the domain where the tradition is most careful: the verb, uttered once, undivided (2.449), distributes by capacity and not by any linguistic mark — which is why the exegetical science exists and why it reasons from samarthya rather than from grammar.

2.452 unformalized

देवदत्तादिषु भुजिः प्रत्येकमवतिष्ठते ।
प्रतिस्वतन्त्रं वाक्यं वा भेदो न प्रविभज्यते ॥ ४५२ ॥

devadattādiṣu bhujiḥ pratyekamavatiṣṭhate |
pratisvatantraṃ vākyaṃ vā bhedo na pravibhajyate || 452 ||

The eating stands severally with respect to Devadatta and the rest; or the sentence is independent for each — the difference is not divided out.

Two descriptions of the distributed case, offered as equivalent. Devadattadisu bhujih pratyekam avatisthate: the *eating* stands *severally* with respect to Devadatta and the others — one sentence, its action distributing (2.372). Prati-svatantram vakyam va — *or* the sentence is *independent for each*: several sentences, one per eater, sharing a form (2.389's position). Bhedo na pravibhajyate: the *difference is not divided out* — nothing distinguishes the two descriptions; no fact settles whether this is one sentence distributing or several sentences coinciding. This is 2.444's undecidability again, and its recurrence is now clearly systematic: wherever the kanda's own doctrine (undivided meaning, analytic carving) applies, the count of sentences is a function of the carving. Bhartrhari does not treat this as a problem to be solved. It is the doctrine's signature, and its consistency across every application is the best evidence for it.

2.453 unformalized

उच्चारणे तु वाक्यानामन्यद्रूपं निगृह्यते ।
प्रतिपत्तौ तु भिन्नानामन्यद्रूपं प्रतीयते ॥ ४५३ ॥

uccāraṇe tu vākyānāmanyadrūpaṃ nigṛhyate |
pratipattau tu bhinnānāmanyadrūpaṃ pratīyate || 453 ||

But in the uttering of sentences, one form is grasped; while in the comprehension of the differentiated, another form is understood.

The two-sidedness that generates 2.452's undecidability, named. Uccarane tu vakyanam anyad rupam nigrhyate: in the *uttering* of sentences, *one* form is grasped — the utterance is one, a single stretch of sound, however many predications it carries (2.110's tantra: one form, many sentences). Pratipattau tu bhinnanam anyad rupam pratiyate: but in the *comprehension* of the differentiated, *another* form is understood — the hearer's grasp articulates into as many as the case requires. Utterance-side unity, comprehension-side plurality: and since sentencehood is defined by neither the sound (2.104) nor the count of words (2.326) but by the closing of expectancy in a mind, the two sides can disagree without contradiction. The verse explains, at last, why the kanda's central question — what is a sentence? — has been answered with a phenomenon rather than a definition. Sentences are individuated where they happen, and they happen in comprehenders.

2.454 unformalized

एकं ग्रहणकं वाक्यं सामान्येनाभिधीयते ।
कर्तरीति यथा तच्च पश्वादिषु विभज्यते ॥ ४५४ ॥

ekaṃ grahaṇakaṃ vākyaṃ sāmānyenābhidhīyate |
kartarīti yathā tacca paśvādiṣu vibhajyate || 454 ||

One comprehensive sentence is expressed by way of the general — as in 'in the agent' — and that is divided out among the animal and the rest.

The mechanism of generality in rule-sentences. Ekam grahanakam vakyam samanyena abhidhiyate: *one comprehensive sentence* is expressed *by way of the general* — the example is Panini's kartari, 'in (the sense of) the agent' (the rule-fragment governing a whole class of affixations) — and tad ca pasvadisu vibhajyate: *that* is *divided out* among the animal and the rest: the single general provision distributes across all its instances. This is 2.176's economy (instruction by the general, for brevity) shown to be a case of 2.448's compression: one utterance, indefinitely many applications, no division audible. And it closes the circle the movement opened at 2.371: how a single utterance distributes over a plurality — whether of eaters, of kings, of sacrificial items, or of grammatical instances — is settled by capacity and purpose, never by form, and the grammar's own most general rules are the clearest case.

2.455 unformalized

यद्याकाङ्क्षा निवर्तेत तद्भूतस्य सकृच्छ्रुतौ ।
नैवान्येनाभिसम्बन्धं तदुपेयात् कथञ्चन ॥ ४५५ ॥

yadyākāṅkṣā nivarteta tadbhūtasya sakṛcchrutau |
naivānyenābhisambandhaṃ tadupeyāt kathañcana || 455 ||

If the expectancy of that were to withdraw upon a single hearing — then it could not in any way attain connection with another.

A closing argument, and it is a conditional whose antecedent the kanda denies. Yadi akanksa nivarteta tad-bhutasya sakrc-chrutau: *if* the expectancy of that were to *withdraw* upon a *single hearing* — if one utterance discharged the expectancy completely and finally — na eva anyena abhisambandham tad upeyat kathamcana: then it could *not in any way* attain connection with *another*. A fully discharged sentence is inert: nothing can be added to it, no further clause can join it, no exception can limit it (2.348's rule-and-exception unity would be impossible; 2.442's multi-verb sentences would not exist). But such joinings manifestly occur. Therefore expectancy is not discharged by a single hearing — it remains live, reaching across utterances, closing only where the whole discourse closes (2.76's maha-vakya). The verse defends, by reductio, the most expansive consequence of the kanda's doctrine: sentencehood has no fixed upper bound, and what looked like a completed sentence is a member of a larger one whenever a larger one is being said.

2.456 · One Form, Many Meanings: The Womb of Sentences unformalized

एकरूपमनेकार्थं तस्मादुपनिबन्धनम् ।
योनिर्विभागवाक्यानां तेभ्योऽनन्यदिव स्थितम् ॥ ४५६ ॥

ekarūpamanekārthaṃ tasmādupanibandhanam |
yonirvibhāgavākyānāṃ tebhyo'nanyadiva sthitam || 456 ||

Therefore the ground is single in form and manifold in meaning — the womb of the sentences of division, standing as though not other than them.

The conclusion of 2.455's argument, and it names the sentence's source with the kanda's oldest metaphor. The ground — upanibandhanam, the fastening (the term for what meaning rests on since 1.13) — is eka-rupam aneka-artham: *single in form*, *manifold in meaning* (2.250's one word with many meanings, now at the sentence's scale). And its office: yonih vibhaga-vakyanam — the *womb* of the sentences of *division*: the analytic sentences, the several predications the analysis distributes (2.389-390), are born from it. Yoni is 2.58's word (samhita as the womb of the words) and 1.3's (the six modifications as wombs of differentiation) — the vocabulary of birth rather than of composition. And the closing qualification: tebhyah ananyad iva sthitam — standing as though *not other* than them: the source is not a further item behind its progeny; the division-sentences are it, differently seen. Iva, as always, holds the ontology honest.

2.457 unformalized

क्वचित् क्रिया व्यक्तिभागैरुपकारे प्रवर्त्तते ।
सामान्यभाग एवास्याः क्वचिदर्थस्य साधकः ॥ ४५७ ॥

kvacit kriyā vyaktibhāgairupakāre pravarttate |
sāmānyabhāga evāsyāḥ kvacidarthasya sādhakaḥ || 457 ||

Sometimes an action operates in its assistance by way of the individual-portions; sometimes the general-portion alone of it is what accomplishes the meaning.

The distribution-doctrine (2.371-455) restated in terms of portions. Kvacit kriya vyakti-bhagaih upakare pravartate: sometimes an action operates, in its *assistance*, by way of the *individual-portions* — reaching its object through the particulars (the eating, per eater, 2.372). Samanya-bhaga eva asyah kvacid arthasya sadhakah: sometimes the *general-portion alone* of it is what accomplishes the meaning — the action engaging only the type, indifferent to instances (the lamp's lighting, 2.374). The vocabulary of bhaga, portion, is doing careful work: these are not two actions nor two parts of one action, but two *aspects* under which one action engages a plurality. The sadharana-model again (2.394-397): one thing, wholly present at several positions, and which position bears the weight is settled by capacity.

2.458 unformalized

कालभिन्नाश्च भेदा ये ये चाप्युष्ट्रासिकादिषु ।
प्रक्रमे जातिभागस्य शब्दात्मा तैर्न भिद्यते ॥ ४५८ ॥

kālabhinnāśca bhedā ye ye cāpyuṣṭrāsikādiṣu |
prakrame jātibhāgasya śabdātmā tairna bhidyate || 458 ||

The differences that are separated by time, and those in 'ustrasika' and the like — in the commencement of the class-portion, the word's self is not differentiated by them.

The invariance of the word under two kinds of variation. Kala-bhinnah ca bhedah — differences separated *by time*: the same word uttered on different occasions, its instances scattered across the years. And ye ca api ustrasikadisu — those in 'ustrasika' and the like: the traditional class of words whose formation or reference is irregular. Prakrame jati-bhagasya sabdatma taih na bhidyate: in the *commencement of the class-portion*, the word's *self is not differentiated* by them. Whether the word is met now or a century hence, in a regular formation or an odd one, its atman is one — 2.256's svarupa-invariance (what is fixed about a word is its form) given its temporal and morphological range. The verse guards the sadharana-model from an obvious objection: if the word's meaning is carved fresh at every occasion, is it the same word? Yes — the carving varies, the word does not.

2.459 unformalized

एकसंख्येषु भेदेषु भिन्ना जात्यादिभिः क्रियाः ।
भेदेन विनियुज्यन्ते तच्छब्दस्य सकृच्छ्रुतौ ॥ ४५९ ॥

ekasaṃkhyeṣu bhedeṣu bhinnā jātyādibhiḥ kriyāḥ |
bhedena viniyujyante tacchabdasya sakṛcchrutau || 459 ||

Among differences of a single number, actions differentiated by class and the rest are applied by way of difference, on a single hearing of that word.

A refinement: distribution can be selective. Eka-samkhyesu bhedesu — among differences that are *of a single number* (items counted as one group) — bhinnah jaty-adibhih kriyah: actions *differentiated* by class and the rest — bhedena viniyujyante: are applied *by way of difference* — tac-chabdasya sakrc-chrutau: on a *single hearing* of that word. One utterance, one number, and yet the several actions distribute selectively across the group by class — some members getting one treatment, some another, all from a single hearing. Nothing in the utterance encodes the sorting; capacity and class do it (2.371's samarthya-laksanam). The verse is a fine specimen of how much the exegetical tradition asks of a single sentence, and of why the kanda has insisted that the words are not carrying the load: no distribution of deposits could produce a selective sorting that the utterance never mentions.

2.460-463 unformalized

अक्षादिषु यथा भिन्ना भञ्जिभक्षिदिविक्रियाः ।
प्रयोगकालाभेदेऽपि प्रतिभेदं पृथक् स्थिताः ॥ ४६० ॥
अक्षाणां तन्त्रिणां तन्त्रमुपायस्तुल्यरूपता ।
एषां क्रमो विभक्तानां तन्निबद्धा सकृच्छ्रुतिः ॥ ४६१ ॥
द्वावभ्युपायौ शब्दानां प्रयोगे समवस्थितौ ।
क्रमो वा यौगपद्यं वा यौ लोको नातिर्त्तते ॥ ४६२ ॥
क्रमे विभिद्यते रूपं यौगपद्ये न भिद्यते ।
क्रिया तु यौगपद्येऽपि क्रमरूपानुपातिनी ॥ ४६३ ॥

akṣādiṣu yathā bhinnā bhañjibhakṣidivikriyāḥ |
prayogakālābhede'pi pratibhedaṃ pṛthak sthitāḥ || 460 ||
akṣāṇāṃ tantriṇāṃ tantramupāyastulyarūpatā |
eṣāṃ kramo vibhaktānāṃ tannibaddhā sakṛcchrutiḥ || 461 ||
dvāvabhyupāyau śabdānāṃ prayoge samavasthitau |
kramo vā yaugapadyaṃ vā yau loko nātirttate || 462 ||
krame vibhidyate rūpaṃ yaugapadye na bhidyate |
kriyā tu yaugapadye'pi kramarūpānupātinī || 463 ||

(460) As, with dice and the like, the actions of breaking, eating and playing are distinct — though there is no difference in the time of employment, they stand separately for each. (461) The single-performance of dice and of lute-players is a means; likeness of form is the way; their sequence, being divided, is the once-hearing bound up with that. (462) Two means stand settled in the employment of words: sequence, or simultaneity — which the world does not transgress. (463) In sequence the form is differentiated; in simultaneity it is not; but the action, even in simultaneity, follows the form of sequence.

Four verses on the two modes of linguistic employment. (460): aksa — 'dice', and the pun is deliberate, since aksa also names other things — take distinct actions: bhanji (breaking), bhaksi (eating), divi (playing); prayoga-kala-abhede 'pi, though the time of employment does not differ, prati-bhedam prthak sthitah, they stand separately for each. (461) gives the machinery: tantra, single-performance, and tulya-rupata, likeness of form, as the means; the sakrc-chruti, the once-hearing, bound to the divided sequence. Then (462) the doctrine: dvau abhyupayau sabdanam prayoge samavasthitau — *two* means stand settled in the employment of words: kramo va yaugapadyam va — *sequence* or *simultaneity* — yau loko na ativartate: which the world does not transgress. Every utterance is either serial or simultaneous; there is no third. And (463) the asymmetry: krame vibhidyate rupam — in sequence the *form* is differentiated; yaugapadye na bhidyate — in simultaneity it is not. Kriya tu yaugapadye 'pi krama-rupa-anupatini: but the *action*, even in simultaneity, *follows the form of sequence*. Even where the words come at once, what they mean unrolls in order — sequence expelled from the vehicle reappears in the meaning, which is 2.24-27's doctrine (krama as the mind's contribution) confirmed from an unexpected side.

2.464 unformalized

भेदसंसर्गशक्ती द्वे शब्दाद्भिन्ने इव स्थिते ।
यौगपद्येऽप्यनेकेन प्रयोगे भिद्यते श्रुतिः ॥ ४६४ ॥

bhedasaṃsargaśaktī dve śabdādbhinne iva sthite |
yaugapadye'pyanekena prayoge bhidyate śrutiḥ || 464 ||

The two capacities, of difference and of commingling, stand as though distinct from the word; even in simultaneity, in employment with several, the hearing is differentiated.

The two capacities named. Bheda-samsarga-sakti dve — the capacity of *difference* (to be carved into distinct meanings) and of *commingling* (to fuse into one) — sabdad bhinne iva sthite: stand *as though distinct* from the word. The iva is essential: they are not two properties the word has but two aspects of its one omni-capacity (2.433's atyanta-yata-saktitva). And the consequence: yaugapadye 'pi anekena prayoge bhidyate srutih — even in *simultaneity*, in employment with several, the *hearing* is differentiated: the word uttered once, at once, is nonetheless heard as several where several are meant. This is 2.463's asymmetry given its mechanism: simultaneity in the vehicle does not prevent differentiation in the uptake, because the differentiating is not the vehicle's work.

2.465 unformalized

अभिन्नो भेदरूपेण य एकोऽर्थो विवक्षितः ।
तत्रावयवधर्मेण समुदायोऽनुगृह्यते ॥ ४६५ ॥

abhinno bhedarūpeṇa ya eko'rtho vivakṣitaḥ |
tatrāvayavadharmeṇa samudāyo'nugṛhyate || 465 ||

Where one meaning, undivided, is intended in the form of difference — there the aggregate is favoured by the property of the parts.

A case where the whole borrows from its parts. Abhinnah bheda-rupena yah eko 'rtho vivaksitah — where *one* meaning, *undivided*, is intended *in the form of difference*: the speaker means one thing and means it as many (a group as a group-of-these). Tatra avayava-dharmena samudayah anugrhyate: there the *aggregate* is *favoured* — assisted, benefited — by the *property of the parts*: the whole takes on characteristics that belong to its members, and the taking-on is what makes the group-as-many available to speech. This is the reverse of the kanda's usual direction (wholes lending to parts, 2.61's samnivisate); here the parts lend upward. And it is licensed by the same model: on the sadharana-account (2.394), whole and parts are one thing at different positions, so lending can run either way. What is not licensed is composition — the aggregate is *favoured* by the parts' property, not *made of* it.

2.466 unformalized

भेदनिर्वचने त्वस्य प्रत्येकं वा समाप्यते ।
श्रुतिर्वचनभिन्ना वा वाक्यभेदेऽवतिष्ठते ॥ ४६६ ॥

bhedanirvacane tvasya pratyekaṃ vā samāpyate |
śrutirvacanabhinnā vā vākyabhede'vatiṣṭhate || 466 ||

But in the explicit stating of its difference, either it is completed severally, or the hearing, differentiated by the expression, stands in a difference of sentences.

The two descriptions of 2.465's case, and they are 2.452's pair again. Bheda-nirvacane tu asya — in the *explicit stating* of its difference (where the speaker spells out the members) — pratyekam va samapyate: either it is *completed severally* — one sentence, distributing; or srutih vacana-bhinna va vakya-bhede 'vatisthate: the *hearing*, differentiated by the expression, stands in a *difference of sentences* — several sentences, one per member. Undecidable, as always where the carving is optional (2.444, 2.452). The recurrence at this point is worth marking: the kanda is closing, and it has not resolved this question anywhere. It has instead shown, case after case, that the question has no answer *and that nothing depends on it* — which is a philosophical result rather than an evasion, and the strongest possible confirmation that the unit was never in the utterance.

2.467 unformalized

तत्रैकवचनान्तो वा सोऽक्षशब्दः प्रयुज्यते ।
प्रत्येकं वा बहुत्वेन प्रविभागो यथाश्रुति ॥ ४६७ ॥

tatraikavacanānto vā so'kṣaśabdaḥ prayujyate |
pratyekaṃ vā bahutvena pravibhāgo yathāśruti || 467 ||

There, either that word 'aksa' is employed as ending in the singular, or severally, by way of plurality; the division is according to the hearing.

The two options exhibited on the aksa-example (2.460). Either eka-vacana-antah — the word ends in the *singular*, one form for the group; or pratyekam bahutvena — *severally*, by way of *plurality*, the plural form spelling out the members. And the settlement: pravibhagah yatha-sruti — the division is *according to the hearing*: whichever was uttered, the division follows. Yatha-sruti is a modest and exact criterion: the grammar does not decide whether this group is one or many; the speaker's chosen number does, and the hearer follows. Which is 2.385's doctrine (number conferred by vessels) at the level of morphology — the singular and the plural are two vessels, and the same stuff is one or many according to which one it is poured into.

2.468 unformalized

द्विष्ठानि यानि वाक्यानि तेष्वप्येकत्वदर्शिनाम् ।
अनेकशक्तेरेकस्य स्वशक्तिः प्रविभज्यते ॥ ४६८ ॥

dviṣṭhāni yāni vākyāni teṣvapyekatvadarśinām |
anekaśakterekasya svaśaktiḥ pravibhajyate || 468 ||

Even in those sentences which are two-placed, for those who see oneness, the own-capacity of a single thing possessing many capacities is divided out.

The oneness-doctrine applied to sentences with two loci. Dvisthani yani vakyani — sentences that are *two-placed*: predications reaching two subjects, two sites, two applications. Tesu api ekatva-darsinam — even in those, for *those who see oneness* (Bhartrhari's own party, 2.57's abheda-purvaka) — anekasakteh ekasya sva-saktih pravibhajyate: the *own-capacity* of a *single* thing possessing *many capacities* is *divided out*. Not two sentences, not one sentence with two halves: one capacity, carved. The formulation gathers the movement's whole vocabulary — anekasakti (2.440), sva-sakti, pravibhajyate (2.314's carving) — and applies it where the temptation to count is strongest. Two places, one meaning, and the twoness is the division of a capacity rather than the presence of a pair.

2.469-471 unformalized

अत्यन्तभिन्नयोर्वा स्यात् प्रयोगस्तन्त्रलक्षणः ।
उपायस्तत्र संसर्गः प्रतिपत्तृषु भिद्यते ॥ ४६९ ॥
भेदेनावगतौ पूर्वं शब्दौ तुल्यश्रुती पुनः ।
तन्त्रेण प्रतिपत्तारं प्रयोक्त्रा प्रतिपादिताः ॥ ४७० ॥
एकस्यापि विवक्षायामनुनिष्पद्यते परः ।
विनाऽभिसंधिना शब्दः शक्तिरूपः प्रकाशते ॥ ४७१ ॥

atyantabhinnayorvā syāt prayogastantralakṣaṇaḥ |
upāyastatra saṃsargaḥ pratipattṛṣu bhidyate || 469 ||
bhedenāvagatau pūrvaṃ śabdau tulyaśrutī punaḥ |
tantreṇa pratipattāraṃ prayoktrā pratipāditāḥ || 470 ||
ekasyāpi vivakṣāyāmanuniṣpadyate paraḥ |
vinā'bhisaṃdhinā śabdaḥ śaktirūpaḥ prakāśate || 471 ||

(469) Or there might be an employment, characterized by single-performance, of two utterly distinct things; the means there is the commingling, which is differentiated among the comprehenders. (470) Two words, previously apprehended by way of difference, alike in hearing, are again conveyed by single-performance to the comprehender by the speaker. (471) Even when one alone is intended, the other arises-in-consequence; without any intention, the word shines forth in the form of capacity.

Three verses on homonymy under tantra, and the last is the movement's most interesting concession. (469): two *utterly distinct* things may be employed tantra-laksanah, by single-performance — one utterance serving both — and the samsarga, the commingling, is the means, pratipattrsu bhidyate, *differentiated among the comprehenders*: different hearers get different fusions. (470): two words bhedena avagatau purvam — previously apprehended *by way of difference* (known separately, as distinct words) — tulya-sruti punah, *alike in hearing*, are conveyed tantrena by the speaker: one utterance, both words, deliberately. That is punning, and the verse describes it exactly. Then (471) the concession: ekasya api vivaksayam anunispadyate parah — even when *one alone* is intended, the *other* arises-in-consequence (2.121's anunispadin: the entailed, unbidden); vina abhisandhina sabdah sakti-rupah prakasate: *without any intention*, the word shines forth *in the form of capacity*. The second meaning comes whether the speaker wants it or not — because the word's capacity is not under his control (2.400's eye can be aimed; what falls in the periphery cannot be excluded). Intention governs what is said and not what arrives (2.301's smoke), and here that doctrine reaches its sharpest formulation: the pun is not always the punster's.

2.472-473 unformalized

अनेकशक्तिरेकस्य युगपच्छ्रीयते क्वचित् ।
अग्निः प्रकाशदाहाभ्यामेकत्रापि नियुज्यते ॥ ४७२ ॥
आवृत्तिशक्तिभिन्नार्थे वाक्ये सकृदपि श्रुते ।
लिङ्गाद्वा तन्त्रधर्माद्वा विभागेनावतिष्ठते ॥ ४७३ ॥

anekaśaktirekasya yugapacchrīyate kvacit |
agniḥ prakāśadāhābhyāmekatrāpi niyujyate || 472 ||
āvṛttiśaktibhinnārthe vākye sakṛdapi śrute |
liṅgādvā tantradharmādvā vibhāgenāvatiṣṭhate || 473 ||

(472) The many capacities of a single thing are sometimes heard simultaneously; fire is applied, even in one place, by way of both illuminating and burning. (473) In a sentence whose meaning is differentiated by the capacity of repetition, even when heard once, it stands by way of division — either from a mark or from the property of single-performance.

The doctrine's culminating illustration, and it is the work's oldest image put to a new use. (472): aneka-saktih ekasya yugapat sruyate kvacit — the *many capacities* of a single thing are sometimes heard *simultaneously*; and the example: agnih prakasa-dahabhyam ekatra api niyujyate — *fire* is applied, *even in one place*, by way of *both* illuminating *and* burning. One fire, two capacities, exercised at once, in the same act, on the same occasion — and no one thinks the fire is two, or that the burning is a secondary use of the lighting (2.253's gauna/mukhya dissolved by the plainest fact about fire). So with the word: many capacities, exercised at once, one word. (473) then applies it to the sentence heard once and meaning severally — standing vibhagena, by way of *division*, either lingat, from a *mark*, or tantra-dharmat, from the *property of single-performance*. The fire-image is the perfect close to the movement, and it is chosen for a reason: fire has been the work's figure for the word since 1.46 (fire latent in the fire-sticks), and here it is the figure for omni-capacity exercised without division.

2.474 unformalized

सम्प्रसारणसंज्ञायां लिङ्गाभ्यां वर्णवाक्ययोः ।
प्रविभागस्तत्र सूत्रे एकस्मिन्नेव कल्पते ॥ ४७४ ॥

samprasāraṇasaṃjñāyāṃ liṅgābhyāṃ varṇavākyayoḥ |
pravibhāgastatra sūtre ekasminneva kalpate || 474 ||

In the samprasarana-designation, by the two marks of phoneme and sentence, the division is construed there in one and the same sutra.

The doctrine confirmed from the grammar's own metalanguage. Samprasarana is Panini's term for the vowel-substitution of semivowels (y→i, v→u, r→r). The verse observes that in the samprasarana-samjna, the division is effected lingabhyam varna-vakyayoh — by the *two marks* of *phoneme* and *sentence*: two different kinds of indicator, operating at different grains — and tatra sutre ekasmin eva kalpate: the division is construed *in one and the same sutra*. One rule, two grains of marking, both required and both present. This is 2.472's fire in the sastra: a single item exercising several capacities at once, and the grammar's own most technical machinery presupposing exactly the multi-capacity account the kanda has argued. The verse is characteristic of the work's method at its most confident: after four hundred verses of philosophy, the confirmation is a footnote to a sutra.

2.475 unformalized

तथा द्विर्वचनेऽचीति तन्त्रोपायादिलक्षणः ।
एकशेषेण निर्देशो भाष्य एवोपवर्णितः ॥ ४७५ ॥

tathā dvirvacane'cīti tantropāyādilakṣaṇaḥ |
ekaśeṣeṇa nirdeśo bhāṣya evopavarṇitaḥ || 475 ||

So too in 'reduplication before a vowel', the designation by way of single-remainder, characterized by the means of single-performance and the rest, has been described in the Bhasya itself.

A second confirmation, and this one names its authority. Dvirvacane aci — 'reduplication before a vowel' (Panini 6.1.1-2 territory): the rule's formulation employs ekasesa, the *single-remainder* device (2.108-110), and is tantropaya-adi-laksanah, characterized by the means of single-performance and the rest. And: bhasya eva upavarnitah — it has been described *in the Bhasya itself*. Patanjali got there first, and Bhartrhari says so. The citation is doing more than crediting: it establishes that the whole apparatus the kanda has built — tantra, ekasesa, one form serving many offices — is not the author's invention but the tradition's own, present in the Mahabhasya and only made explicit here. Which is exactly the claim 1.30 requires of any teaching: it must be agama-purvaka, preceded by tradition, and even the grammarian's boldest semantics is a drawing-out of what the gurus had already done.

2.476 · The History of the Tradition unformalized

प्रायेण संक्षेपरुचीनल्पविद्यापरिग्रहान् ।
संप्राप्य वैयाकरणान् संग्रहेऽस्तमुपागते ॥ ४७६ ॥

prāyeṇa saṃkṣeparucīnalpavidyāparigrahān |
saṃprāpya vaiyākaraṇān saṃgrahe'stamupāgate || 476 ||

Having, for the most part, reached grammarians who delight in brevity and have grasped little learning — when the Sangraha had gone to its setting —

The famous historical passage begins, and its first word is a lament. The Sangraha — Vyadi's vast lost work, the tradition's great compendium — has *gone to its setting*, astam upagate: the sun-verb, the same used at 2.133 for doctrines that set like stars and at 1.18 for differentiations that set like the sun. The book is gone. And the reason it went: the grammarians it reached were samksepa-rucin — *delighting in brevity*, wanting the short version — and alpa-vidya-parigrahan, having *grasped little learning*. A great work perished because its inheritors preferred summaries. Prayena — 'for the most part': the qualification is scholarly and it is bitter. This is the only extended historical narrative in the work, and it is placed at the very end of the Vakyakanda, after four hundred verses arguing that meaning lives in a flash waked by tradition in a ripened mind — which is precisely what a tradition loses when its heirs want it brief.
contested: the Sangraha's authorship (Vyadi) and its loss are known only from this passage and later report; nothing of it survives, which makes 2.476-482 the principal evidence for a major chapter of the tradition's history

2.477-478 unformalized

कृतेऽथ पचञ्जलिना गुरुणा तीर्थदर्शिना ।
सर्वेषां न्यायबीजानां महाभाष्ये निबन्धने ॥ ४७७ ॥
अलब्धगाधे गाम्भीर्यादुत्तान् इव सौष्ठवात् ।
तस्मिन्नकृतबुद्धीनां नैवावास्थित निश्चयः ॥ ४७८ ॥

kṛte'tha pacañjalinā guruṇā tīrthadarśinā |
sarveṣāṃ nyāyabījānāṃ mahābhāṣye nibandhane || 477 ||
alabdhagādhe gāmbhīryāduttān iva sauṣṭhavāt |
tasminnakṛtabuddhīnāṃ naivāvāsthita niścayaḥ || 478 ||

(477) — and when, thereafter, the Mahabhasya had been composed by the teacher Patanjali, the seer of the ford, as the binding-together of the seeds of all reasonings — (478) that work, unfathomed because of its depth, and as though on the surface because of its excellence — in it, for those whose understanding is unformed, no settled conclusion stood at all.

The second act of the history, and its praise is exact. Patanjali is guruna tirtha-darsina — the *teacher*, the *seer of the ford*: tirtha is the crossing-place, and to see the ford is to know where the river can be passed. His Mahabhasya is sarvesam nyaya-bijanam nibandhane — the *binding-together of the seeds of all reasonings*: not a commentary but a seed-bank, and bija is the work's word for what survives dissolutions (1.132's world-seed). Then (478) the diagnosis of its difficulty, in a phrase of great beauty: alabdha-gadhe gambhiryad — *unfathomed* because of its *depth*; uttana iva sausthavat — and *as though on the surface* because of its *excellence*: the very perfection of its exposition made it look shallow, so that readers waded in and drowned. Tasmin akrta-buddhinam na eva avasthita niscayah: in it, for those whose understanding was *unformed*, *no settled conclusion stood at all*. The book was too good to be read by the unready, and being unread it was misread.

2.479-481 unformalized

बैजिसौभवर्य्यक्षैः शुष्कतर्कानुसारिभिः ।
आर्षे विप्लाविते ग्रन्थे संग्रहप्रतिकञ्चुके ॥ ४७९ ॥
यः पतञ्जलिशिष्येभ्यो भ्रष्टो व्यकरणागमः ।
काले स दाक्षिणात्येषु ग्रन्थमात्रे व्यवस्थितः ॥ ४८० ॥
पर्वतादागमं लब्ध्वा भाष्यबीजानुसारिभिः ।
स नीतो बहुशाख्त्वं चन्द्राचार्यादिभिः पुनः ॥ ४८१ ॥

baijisaubhavaryyakṣaiḥ śuṣkatarkānusāribhiḥ |
ārṣe viplāvite granthe saṃgrahapratikañcuke || 479 ||
yaḥ patañjaliśiṣyebhyo bhraṣṭo vyakaraṇāgamaḥ |
kāle sa dākṣiṇātyeṣu granthamātre vyavasthitaḥ || 480 ||
parvatādāgamaṃ labdhvā bhāṣyabījānusāribhiḥ |
sa nīto bahuśākhtvaṃ candrācāryādibhiḥ punaḥ || 481 ||

(479) When the seer's work had been swept away by Baiji, Saubhava and Haryaksa, followers of dry reasoning, with the Sangraha as their cover — (480) the grammatical tradition which had fallen away from Patanjali's pupils, at that time stood established among the Southerners in the mere book. (481) Having obtained the tradition from the mountain, by those following the seeds of the Bhasya — by Candracarya and the rest — it was again brought to many branches.

The near-death and the recovery, and it is the most cited historical passage in Sanskrit grammar. (479) names three men — Baiji, Saubhava, Haryaksa — as suska-tarka-anusarinah, *followers of dry reasoning*: tarka without agama, exactly what 1.30 and 1.137 declared groundless; and they operated sangraha-pratikancuke, with the Sangraha *as their cover* — invoking the authority of the great work while destroying its teaching. Arse viplavite granthe: the *seer's work was swept away*. (480): the tradition bhrastah, *fallen* from Patanjali's pupils, survived kale daksinatyesu grantha-matre — among the *Southerners*, in the *mere book*: text without living transmission, the letter surviving where the flash had gone out (2.152's transmitted pratibha, broken). And (481) the rescue: parvatad agamam labdhva — having obtained the tradition *from the mountain* — bhasya-bijanusaribhih, by those *following the seeds of the Bhasya* (477's bija) — candracarya-adibhih punah, by *Candracarya and the rest*: sa nito bahu-sakhatvam, it was brought again to *many branches*. The whole doctrine of Kanda I is enacted here as history: tradition is not a text but a living inheritance (2.151's agama ripened by bhavana), it can be broken, and its recovery requires both the book and the mountain.
contested: 'from the mountain' (parvatat) is the great crux: a place-name, a mystical source, a teacher's name, or a lineage-designation have all been proposed, and nothing decides. Candracarya is plausibly the Buddhist grammarian Candragomin. The three named opponents are otherwise unknown

2.482-485 unformalized

न्यायप्रस्थानमार्गांस्तानभ्यस्य स्वं च दर्शनम् ।
प्रणीतो गुरुणाऽस्माकमयमागमसंग्रहः ॥ ४८२ ॥
वर्त्मनामत्र केषाञ्चिद् वस्तुमात्रमुदाहृतम् ।
काण्डे तृतीये न्यक्षेण भविष्यति विचारणा ॥ ४८३ ॥
प्रज्ञाविवेकं लभते भिन्नैरागमदर्शनैः ।
कियद्वा शक्यमुन्नेतुं स्वतर्कमनुधावता ॥ ४८४ ॥
तत्तदुत्प्रेक्षमाणानां पुराणैरागमैर्विना ।
अनुपासितवृद्धानां विद्या नातिप्रसीदति ॥ ४८५ ॥

nyāyaprasthānamārgāṃstānabhyasya svaṃ ca darśanam |
praṇīto guruṇā'smākamayamāgamasaṃgrahaḥ || 482 ||
vartmanāmatra keṣāñcid vastumātramudāhṛtam |
kāṇḍe tṛtīye nyakṣeṇa bhaviṣyati vicāraṇā || 483 ||
prajñāvivekaṃ labhate bhinnairāgamadarśanaiḥ |
kiyadvā śakyamunnetuṃ svatarkamanudhāvatā || 484 ||
tattadutprekṣamāṇānāṃ purāṇairāgamairvinā |
anupāsitavṛddhānāṃ vidyā nātiprasīdati || 485 ||

(482) Having studied those paths of the courses of reasoning, and his own school, this compendium of the tradition was composed by our teacher. (483) Here, of some of the ways, the bare substance has been cited; in the third kanda the examination will be thorough. (484) One obtains discrimination of understanding through differing traditional views; how much can be raised up by one who runs after his own reasoning? (485) For those who imagine this and that without the ancient traditions, and who have not waited upon their elders, knowledge does not become very gracious.

The kanda closes on four verses that gather the whole work's method. (482): the compendium was composed by gurunasmakam — *our teacher* — having studied both the nyaya-prasthana-margan, the paths of reasoning, and svam darsanam, *his own school*: both, and in that order. (483) is a promise and a pointer: here vastu-matram udahrtam, the *bare substance* of some views has been cited; kande trtiye nyaksena bhavisyati vicarana — in the *third kanda* the examination will be *thorough*. The Padakanda is announced. Then (484-485) the closing counsel, and it is the work's own epistemology addressed to its reader. Prajna-vivekam labhate bhinnair agama-darsanaih — one obtains *discrimination of understanding* through *differing* traditional views: the doxography that has filled two kandas was not scholarly display but the method itself; discrimination comes from holding many inherited views, not from any one. Kiyad va sakyam unnetum sva-tarkam anudhavata — *how much can be raised up* by one who *runs after his own reasoning*? And the last word: vidya na atiprasidati — knowledge does *not become very gracious* — to those who imagine without the puranair agamaih, the ancient traditions, and who are anupasita-vrddhanam, *who have not waited upon their elders*. Prasidati: knowledge is not seized but *graced*, and it graces those who have sat with the old. After four hundred verses arguing that meaning is a flash in a ripened mind, the kanda ends by naming what does the ripening: not argument, but sitting with those who know.
contested: 'our teacher' (guruna asmakam) in 482 is disputed — whether Bhartrhari refers to a teacher of his own or, as some hold, to himself in the third person; the Vrtti-authorship question is entangled with it and remains open