In this appendix, I will discuss some traditional examples of ‘conflict’ between antaraṅga and bahiraṅga rules, and present my opinion on them. Before we begin, let us revise the basic definition of antaraṅga. According to the Paribhāṣenduśekhara1, ‘antaraṅga is (a rule) the causes (of the application) of which lie within (or before) the sum of the causes of a bahiraṅga
rule’.2 An antaraṅga rule is stronger than and thus defeats a bahiraṅga rule.3
However, note that Kātyāyana and Patañjali, despite talking about antaraṅga and bahiraṅga, do not define these terms and consequently do not explain why a certain rule is to be regarded as antaraṅga. In vt. 8 on 1.4.2 vipratiṣedhe paraṁ kāryam, Kātyāyana says: antaraṅgam ca. On this vārttika, Patañjali elaborates: antaraṅgam ca balīyo bhavatīti vaktavyam ‘It should also be said that [an] antaraṅga [rule] is stronger [than a bahiraṅga rule]’. Let us examine some
examples discussed by Patañjali (Mbh I.304.10 onwards) while commenting on various vārttikas on 1.4.2.
(1) Let us follow Patañjali’s method to derive syona ‘a stitched item i.e., a sack’. First, we add na to siv ‘to stitch’ by 3.3.1 uṇādayo bahulam.4 By 6.4.19 chvoḥ śūḍ anunāsike ca (which teaches that ch and v are replaced with ś and ūṬH, respectively, when an affix beginning with a nasal, or affix KvI, or one beginning with jhaL i.e., a non-nasal stop or a fricative, and marked with K or Ṅ, follows), we get siū + na. According to Patañjali, two rules are simultaneously applicable to siū + na:
s i ū + na
6.1.77 7.3.86
6.1.77 iko yaṇ aci: iK (i, u, r̥, l̥) is replaced with yaṆ (y, v, r, l) when aC (any vowel) follows.
229
7.3.86 pugantalaghūpadhasya ca: guṇa replaces iK of a verbal base which ends in the augment pUK or which has a laghu ‘light’ vowel as its penultimate sound when a sārvadhātuka or ārdhadhātuka affix follows.
According to Patañjali, the rule teaching substitution with yaṆ (6.1.77) is antaraṅga with respect to the rule teaching guṇa (7.3.86). This is corroborated by the definition of antaraṅga given by the commentary on Pbh 50 of the Paribhāṣenduśekhara: the cause of application of 6.1.77 (i.e., ū) lies before i.e., to the left of the cause of application of 7.3.86 (i.e., na). Let us use this example to speculate about how Kātyāyana might have defined antaraṅga and bahiraṅga. Note that the cause of application of 6.1.77 lies inside (antar) the aṅga siū, while the cause of application of 7.3.86 lies outside (bahir) it. Thus, the term antaraṅga could stand for aṅgasya antaḥ and the term bahiraṅga for aṅgād bahiḥ.
The antaraṅga rule 6.1.77 wins, and thus the derivation proceeds as follows: siū + na 🡪 syū + na (6.1.77) 🡪 syona (7.3.84 sārvadhātukārdhadhātukayoḥ).
Now let me present my opinion about this example. There is no evidence that Pāṇini has composed the Uṇādi sūtras. Therefore, this derivation, which requires us to add na to siv as per an Uṇādisūtra (289) is not Pāṇinian at all.
(2) Let us use Patañjali’s method to derive the form dyaukāmi ‘male offspring of dyukāma’. We start by adding the taddhita affix iÑ to the bahuvrīhi compound made up of div and kāma by 4.1.95 ata iÑ (which teaches that the taddhita affix iÑ occurs to denote an offspring after a syntactically related nominal stem which ends in a). After deleting the inflectional affixes inside the compound by 2.4.71 supo dhātuprātipadikayoḥ, we get div + kāma + iÑ. Here, by 6.1.131 diva ut (which teaches that the final sound of the pada div is replaced with uT), we get diu + kāma + iÑ. At this stage, according to Patañjali, two rules are simultaneously applicable:
d i u + kāma + iÑ
6.1.77 7.2.117
6.1.77 iko yaṇ aci: same as above.
7.2.117 taddhiteṣv acām ādeḥ: the first vowel of the base undergoes vr̥ddhi when an affix marked with Ñ or Ṇ follows in taddhita derivations.
230This example is similar to the previous one: the cause of application of 6.1.77 (i.e., u) lies before, namely to the left of the cause of application of 7.2.117 (i.e., iÑ). Here too, Patañjali says that 6.1.77 is antaraṅga and thus wins. The derivation proceeds as follows: diu + kāma + iÑ 🡪 dyu + kāma + iÑ (6.1.77) 🡪 dyau + kāma + iÑ (7.2.117) 🡪 dyaukāmi (6.4.148 yasyeti ca5).
In my opinion, no such conflict arises in the first place. We want to derive a word that means: dyukāmasya apatyam pumān ‘male offspring of dyukāma’. Since we are talking about dyukāma’s offspring, and not (div + kāma)’s offspring, the derivation should start with dyukāma and not with div + kāma. Thus, we have: dyukāma + Ṅas + iÑ. Ṅas is deleted by 2.4.71 supo dhātuprātipadikayoḥ and we get dyukāma + iÑ. Here two rules are simultaneously applicable:
dyukāma + iÑ
7.2.117 6.4.148
7.2.117 taddhiteṣv acām ādeḥ: same as above.
6.4.148 yasyeti ca: same as above.
This is a case of DOI. By my interpretation of 1.4.2, we apply the RHS rule 6.4.148 and get dyukām + iÑ. Then we apply 7.2.117 and get dyaukāmi, which is the correct form. 6
Several other examples discussed by Patañjali in his comments on different vārttikas on 1.4.2, such as sautthatiḥ, kādraveyaḥ, stairṇiḥ, khaṭvīyati, kāmaṇḍaleya, cauḍi etc. are similar to this example. For instance, in the derivation of the nominal base sautthati, Patañjali starts with su + utthita, whereas one should actually start with sūtthita.
(3) Let us follow Patañjali’s method to derive the form dudyūṣati ‘desires to shine’. We start by adding the desiderative affix saN to the root div ‘to shine’ by 3.1.7 dhātoḥ karmaṇaḥ samānakartr̥kād icchāyāṁ vā (which teaches that the affix saN is optionally introduced after a verbal stem, the action denoted by which is the object of a verbal stem expressing desire and both actions have the same agent). Thereafter, by 6.4.19 chvoḥ śūḍ anunāsike ca (see
231
translation in example 1), we get diū + saN. Here, according to Patañjali, two rules are simultaneously applicable:
{d [i] } ū + saN
6.1.77 iko yaṇ aci is applicable to i while 6.1.9 sanyaṅoḥ7 is applicable to di. Notice that the cause of application of 6.1.77 (i.e., ū) lies to the left of the cause of application of 6.1.9 (i.e., saN). Patañjali says that 6.1.77 is antaraṅga and thus wins, thereby giving: dyū + saN. Thereafter, 6.1.9 applies and we get dyūdyū + saN. After applying other rules, we get the correct form dudyūṣati.
In my opinion, such a conflict does not arise in the first place. I interpret sanyaṅoḥ as a genitive form, not as a locative form8. So, in my view, 6.1.9 sanyaṅoḥ teaches that a verbal base ending in saN or yaṄ, which has not undergone reduplication, is reduplicated9. Note that diū + saN is not a verbal base ending in saN, but instead two separate items, namely diū and saN. So, 6.1.9 is not applicable here. However, 6.1.77 is applicable here, and on applying it, we get dyū + saN. Now, since no other rules can be applied here, we can fuse the two items dyū and saN into a single item dyūṣa, which we can call a verbal base ending in saN. Therefore, 6.1.9 applies here and we get dyūdyūṣa. After applying other rules, we get the correct verbal base dudyūṣa (and the correct final form dudyūṣati).
The examples jujñaudanīyiṣati and ātestīryate discussed by Patañjali are similar to this one.
(4) Patañjali says that in the string ayaja + i + indram ‘I worshipped Indra’, two rules are simultaneously applicable: 6.1.87 ād guṇaḥ, which is applicable to a + i and 6.1.101 akaḥ savarṇe dīrghaḥ, which is applicable to i + i. He adds that 6.1.87 is antaraṅga and thus win, thereby giving the correct form: ayaje indram.
232
I do not think that such a conflict arises at all. I think that, in the Pāṇinian system, all possible rules that can be applied while constructing a word ought to be applied before the word is considered within the context of the sentence. In other words, these rules, which contribute towards the construction of a word, cannot be applied after the word enters the sentence. Here, the rule 6.1.87 applies to ayaja + i, giving the form ayaje. Now that the word is ready, it enters the sentence: ayaje indram10.
Other examples of this nature discussed by Patañjali include agnir indraḥ, pacatv atra.
(5) Let us derive the form vānīya ‘should be weaved’ using Patañjali’s method. We add the affix anīyaR to veÑ ‘to weave’ by 3.1.96 tavyattavyānīyaraḥ. Here, according to Patañjali, two rules are simultaneously applicable:
ve + anīyaR
6.1.78 6.1.45
6.1.78 eco’yavāyāvaḥ: the sounds represented by eC (e, o, ai, and au) are replaced with ay, av, āy, and āv, respectively, when a vowel follows.
6.1.45 ād eca upadeśe’śiti: the final sound of a verbal root which ends in eC (e, o, ai, and au) in the Dhātupāṭha is replaced with ā, when an affix which is not marked with Ś follows.
Patañjali says that 6.1.45 is antaraṅga with respect to 6.1.78 and thus wins. Note that this contradicts what the commentary on Pbh 50 tells us. We would expect the cause of application of the antaraṅga rule to be within or before that of the bahiraṅga rule. But here, the cause of application of the bahiraṅga rule 6.1.78 (i.e., a at the beginning of anīyaR) lies inside the cause of the antaraṅga rule 6.1.45 (i.e., anīyaR). This exemplifies the fact that the antaraṅga tool is poorly defined and not always useful.
According to me, this is a case of SOI, and we do not need the antaraṅga tool to deal with cases of SOI. In case of SOI, the more specific rule wins. Let us compare the two rules:
233
6.1.78
e / o / ai / au + vowel
6.1.45
e / o / ai / au (end of verbal root) + vowel (beginning of affix not marked with Ś) e / o / ai / au (end of verbal root) + non-vowel (beginning of affix not marked with Ś)
6.1.45 is more specific because it applies only when the affix is not marked with Ś. Thus, it wins, giving us the correct form vā + anīya 🡪 vānīya (6.1.101 akaḥ savarṇe dīrghaḥ).
Other examples discussed by Patañjali such as glācchatram, agnicid idam are similar to this one.
Finally, Patañjali does not simply say that antaraṅga rules defeat bahiraṅga rules in case of conflict. He goes a step further to claim: asiddhaṁ bahiraṅgam antaraṅge ‘a bahiraṅga rule is asiddha with respect to an antaraṅga rule’. Thus, he implies that an antaraṅga rule cannot see a bahiraṅga rule, and therefore cannot see the outcome of the application of the bahiraṅga rule either. This is true not only for cases of Same Step Rule Interaction (including conflict) but also for any pair of antaraṅga-bahiraṅga rules which are not simultaneously applicable. Consider the following example.
(6) Consider pacāva + idam. By 6.1.87 ād guṇaḥ, we get pacāvedam. Here, Patañjali claims that by 3.4.93 eta ai (which teaches that eT, which is a substitute of the first-person replacement of LOṬ, is replaced with ai), the e in pacāvedam could get replaced with ai, thereby giving the incorrect phrase *pacāvaidam. He says that this is prevented by the fact that the rule 6.1.87 is bahiraṅga and thus asiddha with respect to the antaraṅga rule 3.4.93. Thus, 3.4.93 cannot apply to e, which is the outcome of the application of 6.1.87. This ensures that we get the correct phrase: pacāvedam.
I do not agree with Patañjali. As stated before, according to me, in the Pāṇinian system all possible rules that can be applied while constructing a word ought to be applied before the word is considered within the context of the sentence. In other words, these rules, which contribute towards the construction of a word, cannot be applied to the word after it enters the sentence. Note that, 3.4.93 eta ai is a rule which helps the construction of a word (e.g., edhāvahai) and, therefore, it is not applicable at sentence level.
234
In conclusion, I think that the antaraṅga tool is completely unnecessary in both SSRI and non SSRI contexts. Most examples (like 1, 2, 3, 4 and 6) which it allegedly solves are not problematic in the first place. Some examples (like 5) it deals with are actually ordinary cases of SOI which can be solved by choosing the more specific rule.
235