Rule. A pada-vidhi — any rule that operates on finished words (padas) — applies only to words that are samartha, "semantically connected". The connectedness itself is called sāmarthya. This is a paribhāṣā: it does not form anything, it tells you how every compound rule that follows is to be read.
Example.
kṛṣṇasya samīpam "near Kṛṣṇa" → upa ("near") and kṛṣṇa are connected in meaning → the compound rule may apply → upakṛṣṇam.
paśya kṛṣṇaṃ, śrito rāmaṃ mitram "Look at Kṛṣṇa; the friend has resorted to Rāma" → kṛṣṇa belongs with paśya, śrita with rāma → kṛṣṇa and śrita are not connected → no compound kṛṣṇaśrita here.
Applies when / blocked by. Bhaṭṭoji's commentators distinguish two kinds of sāmarthya. In a sentence such as rājñaḥ puruṣaḥ "the king's man" the words stay separate and merely expect one another; this is vyapekṣā. In a compound such as rājapuruṣa the words fuse into a single meaning-unit; this is ekārthībhāva. It is the second kind that licenses upakṛṣṇam as one word.
Watch out. Because the members of a compound have become one meaning, an outside word cannot reach inside the compound to qualify one member: in ṛddhasya rājapuruṣaḥ "the prosperous one's king-man", ṛddha does not qualify rājan.
A rule (vidhi) that pertains to padas is to be understood as resting upon [words that are] samartha (semantically connected).
samarthaḥ padavidhiḥ — Now, among the compounds (samāsa), the avyayībhāva. Thus, having set forth the meaning of the case-endings (vibhakti), [Bhaṭṭoji], about to set forth the compounds that depend upon that [case relation], states, as an introduction (upodghāta) to it, "samarthaḥ padavidhiḥ." A "vidhi" is that which is enjoined (vidhīyate), i.e., an operation (kārya). "padavidhi" [=] "a vidhi of a pada," the compound being formed with a residual/general genitive (śeṣa-ṣaṣṭhī). This he states: "a rule that pertains to padas." "Resting upon samartha [words]" — the point is that in the sūtra the word "samartha" is used by indication (lākṣaṇika) for "[a vidhi] that rests upon samartha [words]."
sāmarthya is of two kinds: that characterized by vyapekṣā (mutual expectancy) and that characterized by ekārthībhāva (becoming a single meaning-unit). Of these, the mutual connection (anvaya) which — among words that rest in their own respective meanings (svārtha-paryavasāyin) — arises by the force of mutual expectancy (ākāṅkṣā) and the like, is the sāmarthya named vyapekṣā; by the etymology "vyapekṣā = a distinct/qualified expectancy (viśiṣṭā apekṣā)," and by the etymology "samartha = that whose meaning is connected (saṃbaddha-artha)." And this [vyapekṣā-sāmarthya] obtains only in a sentence (vākya), such as "rājñaḥ puruṣaḥ" (the king's man). And there, each single word connects with whatever is proximate (saṃnihita) and suitable (yogya) to it — for example, "rājñaḥ puruṣo 'śvaś ca" (the king's man and horse), "rājño devadattasya ca puruṣaḥ" (the man of the king and of Devadatta), and "ṛddhasya rājñaḥ puruṣaḥ" (the man of the prosperous king).
But the sāmarthya characterized by ekārthībhāva consists in the capacity to convey a single qualified meaning (viśiṣṭa-eka-artha) through the collective denotative power (samudāya-śakti) of words that, at the derivational stage (prakriyā-daśā), were taken separately as each individually meaningful; by the etymologies "samartha = that whose meaning is joined (saṅgata-artha)" [and] "samartha = that whose meaning is combined (saṃsṛṣṭa-artha)." And "joining" (saṅgati) and "combining" (saṃsarga) are precisely a becoming-one (ekībhāva) — as in "the ghee is joined (saṅgata) with the oil," where "become one" is understood; or as in "the fire is combined (saṃsṛṣṭa)," where "become one" is understood — this also from the Bhāṣya. And this [ekārthībhāva] sāmarthya obtains only in a vṛtti (a complex derived formation such as a compound), such as "rāja-puruṣa." Precisely for this reason, in "ṛddhasya rāja-puruṣaḥ" the qualifier "ṛddha" (prosperous) does not connect with "rājan," which is [itself] a qualifier of "puruṣa"; for, the qualified [whole] being a single word-meaning (eka-padārtha), "rājan" is [merely] a part of [that] word-meaning (padārtha-ekadeśa).
But in "devadattasya gurukulam" (Devadatta's teacher's household), a compound [gurukula] [is nonetheless formed], because the subordinate member (upasarjana, i.e., guru) is permanently relational (nitya-sāpekṣa). Or, alternatively: like "guru," "devadatta" too construes only with "kula" — the principal qualificand (viśeṣya). There, "guru" connects with "kula" by the relation of [the kula's] being-produced [by the guru] (utpādyatva-saṃbandha); but "devadatta" connects with "kula" through [the kula's] being-produced-by-his-guru — a connection that has "guru" embedded within it (guru-garbha). And it has been said by Hari (Bhartṛhari): "Every relational word (saṃbandhi-śabda), [though] dependent (sāpekṣa), is always compounded; for that mutual expectancy (vyapekṣā), as in a sentence, is not forfeited even in a vṛtti." And: "Those [external words] whose relation is with the aggregate (samudāya) — e.g. with a 'gurukula' and the like — while touching its members (avayava), yet construe with that which possesses [the aggregate]." By this [principle], in the Bhāṣya on the "ayaḥśūla" sūtra, [the form] "śiva-bhāgavata" and the like is explained in the sense "a devotee of the Lord Śiva (śivasya bhagavato bhaktaḥ)."
And this ekārthībhāva is posited (kalpyate) in the non-worldly analytical sentence (alaukika-vigraha-vākya) — just as the co-referentiality with a non-nominative meaning (aprathamā-sāmānādhikaraṇya) of śatṛ and śānac, which are substitutes for the la[-kāra], is posited for the la-kāra [itself] as well. Precisely for this reason, having raised the objection — "Since the la[-kāra] is not connected with a meaning co-referential with a non-nominative, a substitution [of śatṛ/śānac] for it is impossible; and because it (the la-kāra) is never actually employed anywhere [so its co-reference cannot be directly observed]" — it is resolved in the Bhāṣya on the sūtra "laṭaḥ śatṛ-śānacau [aprathamā- samānādhikaraṇe]": "Having observed co-referentiality in the substitute, one must reach by inference [the conclusion] that this [co-referentiality] belongs to the prakṛti (the la-kāra) as well." Since [Pāṇini's grammar] gives an account (anvākhyāna) of already-established (siddha) words, and since usages such as "pacantaṃ devadattaṃ paśya" (see Devadatta cooking) are observed, the positing of a denotative capacity (bodhakatva) even for the sthānin — the la[-kāra], which is itself posited only for the sake of the derivational process (prakriyā) — is thereby indicated.
And since there is no authority (māna) for abandoning the established denotative power (kḷpta-śakti) of the words heard in the alaukika-vigraha-vākya, the qualified cognition (viśiṣṭa-upasthiti) [arises] through the collective power (samudāya-śakti) assisted by the individual powers (pratyeka-śakti). And therefore this ekārthībhāva is held to be an ajahat-svārthā vṛtti (a formation that does not abandon the component words' own meanings). But the jahat-svārthā vṛtti (a formation that abandons the components' own meanings) — resting on the [alleged] individual meaninglessness (anarthakatva) of the words within the vṛtti — is not fit to be adopted, since in [forms] such as "mahā-bāhuḥ" and "su-panthāḥ" the ā-substitution (āttva) and the like would [then wrongly] fail to obtain: for, the words "mahat" and the like being [on that view] meaningless in the vṛtti, and given that the grasping of [only] the meaningful (arthavat) is possible, the grasping of a meaningless [element] in [rules] such as "ān mahataḥ" would be impossible [by the arthavad-grahaṇa maxim]. As has been said: "The jahat-svārthā [vṛtti obtains] only there where a conventional meaning (rūḍhi) is opposed [to the component meanings]." Its sense is: it [jahat-svārthā] [obtains] only where the meaning of the aggregate is contrary to the meaning of the members — as in "ākarṇa-maṇḍapa" and the like. The full treatment is to be consulted in the Śabdenduśekhara and the [Paramalaghu-]mañjūṣā.
Why "samartha"? [What purpose does the qualification serve?] "paśya kṛṣṇaṃ, śrito rāmaṃ mitram" (Look at Kṛṣṇa; the friend has resorted to Rāma). Here, because "kṛṣṇa" and "śrita" have no mutual connection [kṛṣṇa construing with "paśya," śrita with "rāma"], and hence produce no qualified single-meaning cognition, there is no sāmarthya [and so no compound].
- textual In the "yad vā" alternative on devadattasya gurukulam the source reads "viśeṣye pradāne kula eva," but the sense requires pradhāne ("the principal [qualificand]"); rendered accordingly (ध/द confusion). The example of a jahat-svārthā vṛtti is printed "ākarṇamaṇḍapa" and is left as transmitted; its reference is not certain. Minor OCR normalized silently: mahādādi (read mahad-ādi), vṛttau, etc.