Praudhamanorama p184

p184

To explain: in 'sakhe' there would be no lopa of 'n', because the lopa of a saṃyogānta is regarded as not having taken effect (asiddha). This might be argued: the lopa of a saṃyogānta is not asiddha, because of the inferential clue (jñāpaka) provided by the mention of 'sambuddhi' in the rule "na ṅisambudhyoḥ" (Pāṇ. 8.2.8). Nor would this entail the undesirable consequence of lopa of 'n' in forms like 'gomān' as well; for an inferential clue has a restricted scope, in accordance with the explanation: 'Only where there is immediate proximity between the consonant n and the vibhakti is there validity (siddhatva).' As for what they say—namely, that because it is for the sake of the neuter in 'he brahman', this is not an inferential clue—that is incorrect; for since an affixal characteristic does not apply when there is elision by 'luk', and because "saṃbuddhau vā napuṃsakānām—" (Pāṇ. 7.1.96) is necessary, the mention of 'sambuddhi' cannot have the neuter as its scope, and therefore the stated inferential clue stands sound. If you say: therefore, lopa of 'n' is not the purpose [of the rule]? In that case, in 'abhino 'tra', the substitution of 'u' for 'ru' would not take place. It is this that is stated by 'nalopādiḥ'. Likewise, in 'abibhar bhavān', there would be no lopa of 'ti', because of the restrictive rule "rātsasya" (Pāṇ. 8.2.24).
Here, this should be noted: the mention of 'ṅyāp' qualifies only 'su', and not 'ti' and 'si', on the basis of authoritative interpretation. The Mahābhāṣya on this very rule, stating 'the mention of dvihalapṛkta and the mention of ti and si need not be made', is also authoritative proof here. For if the mention of 'ṅyāp' also qualified 'ti' and 'si', then how could it state that the mention of 'ti' and 'si' need not be made? As for the view that 'ti' and 'si' are simply impossible after ṅī and āp: on that point, it is true that they are impossible after what ends in ṅī; but after what ends in āp, following a form ending in kvip expressing conduct, 'tip' and 'sip' of the laṅ tense certainly occur, as in 'agaṅgāt', 'agaṅgāḥ'. Nor is there an intervention by 'śap', because a single substitute is treated as the final part of the prior element. Nor does treatment like the original apply, because this is an unprecedented rule (apūrvavidhi).
Here, this is the definitive conclusion: the rule should properly be formulated simply as "ṅyāpaḥ saḥ". Nor is there an overapplication to the 's' of a verbal root in forms such as 'surāḥ', 'surāsau'; for by association with ṅī and āp, and by the metarule 'pratyayāpratyayayoḥ' (between what is an affix and what is not an affix, the affix alone is taken), 'su' as an affix alone is grasped. As for the lopa of 'n' in 'rājā' and the like, that is indeed established by the inferential clue "na ṅisambuddhyoḥ". Furthermore, the substitution of 'u' in 'abhino 'tra' is also easily justified, because it is recited in the valid section (siddhakāṇḍa) as 'the lopa of a saṃyogānta takes effect when there is substitution of u for ru', as in 'harivo medinaṃ tvā'. In 'abibhar bhavān' as well, there is no defect; for by coalesce-reading (praśleṣa) 't' as well in "rātsasya", the division is made as 'tsasya'. If it is objected: in 'ukhāsrat' and the like, would there not be lopa of the initial of a conjunct? No, because when the substitution of 'd' has taken place, that does not apply. If it is objected that the substitution of 'd' is asiddha? No, because it is recited in the valid section: 'the substitution of d takes place before vasu, etc.' If it is objected that because it is not the end of a pada, the substitution of 'd' would not take place? No, because we rely on the view that a form is a pada even before 'su'. And the ground for that is the word division "asarvanāmasthāne yaci", which yields the explanation that before an affix beginning with y or a vowel, it is not a pada when it is a sarvanāmasthāna.
Nor should it be said that in this way, when the lopa of 'n' has been performed in 'rājā' and the like, there would occur the substitution of 'ru' and the visarga for 'su'; for since the form would be established by "ito 't sarvanāmasthāne" (Pāṇ. 7.1.86), the prescription of the long vowel ā by "pathimathi—" (Pāṇ. 7.1.85) serves as an inferential clue that the lopa of 'su' occurs first. If you say: this is not an inferential clue, because it has an application in the vocative—'he panthāḥ', 'manthāḥ', 'ṛbhukṣāḥ'—where the long vowel is excluded by "asaṃbuddhau" and the lopa of 'n' is prohibited? If so, then let the rule "na ṅisaṃbuddhyoḥ" itself be the inferential clue for the sense that the lopa of a saṃyogānta occurs prior to the lopa of 'n'. To explain: 'ṅisambuddhyoḥ' is a genitive and not a locative, because when 'ṅi' has been elided by 'luk' there is no affixal characteristic, and it construes in apposition with 'of the nominal stem ending in ṅī'. And without the lopa of a saṃyogānta having occurred first, a nominal stem ending in the vocative singular is not obtained. Therefore, in the formulation "ṅyāpaḥ saḥ", there is no obstacle whatsoever. If, however, the vārttika 'the substitution of d takes place before vasu, etc.' is rejected, since the absence of lengthening in 'ukhāsrat' and the like is established simply by the mention of 'adhātu', then let the rule be "halṅyāpaḥ saḥ". As for the rule as traditionally worded being explained by the older grammarian as 'after what ends in a consonant, after what ends in ṅī, after what ends in āp, and after what is long'—that is incorrect; for ṅī and āp are long, but what ends in them is not. Nor from 'after what is long, after what ends in ṅī' can the meaning 'after that which ends in ṅī which is long' etc. be obtained, because the ablative 'dīrghāt' does not construe with it, and because within a compound, ṅī and āp, being subordinate members (upasarjana), cannot take an external qualifier. For it is stated in the Mahābhāṣya: 'There is no compound formation for words that have external qualifiers', and 'What has entered into a compound cannot be connected with an external qualifier.' By this, the older grammarian's gloss—'that which does not bear the technical name nadī, is short, and ends in the vowels i or u'—is also refuted.