Two levels of optionality in the Kaccāyana Vyākaraṇa

Abstract

Understanding how option markers operate in Sanskrit vyākaraṇa is crucial in order to determine the scope of grammatical rules. The exact function of option markers such as vā or vibhāṣā, however, is not unproblematic. In his influential monograph on option markers, Kiparsky (1979) postulated different levels of optionality for vā, vibhāṣā and anyatarasyām, each one showing different degrees of preference or frequency in usage. Kiparsky’s theory became highly polarising, and today we find scholars of vyākaraṇa who readily assume the three levels as a fact, and other scholars, like Cardona (1999: 162–79), who dismiss it as a misunderstanding. While it is impossible to find any trace of Kiparsky’s theory in Sanskrit vyākaraṇa literature after Patañjali, something that prima facie reminds of Kiparsky’s claim is found in Pāli vyākaraṇa treatises of the Kaccāyana school. According to this school, there are two basic levels of optionality: one indicates open option (vikappa), and the other indicates an exception to a general rule. In this article, I explain how the two levels work and what option markers are used to indicate them. I also show that, paradoxically, the theory of two levels in the Pāli Kaccāyana tradition seems to refute Kiparsky’s theory, rather than to support it. This is so because the concept of optionality is clearly not related to frequency of usage, but simply to the scope of exceptions, that is, to whether specific forms always (niccam) adhere, or not (na vā), to general (utsarga) grammatical rules.

Keywords: vyākaraṇa, grammar, Kaccāyana, Pāli, optionality, Kiparsky

1 Introduction: optionality, between preference and possibility

1.1 Optionality in Sanskrit grammar

Optionality is one of the most important metagrammatical devices in the Indian vyākaraṇa tradition. Without a proper understanding of how option markers operate, the exact scope of very many grammatical rules cannot be ascertained. In Pāṇinian studies the discussion on optionality has been largely polarised, for the past few decades, between those who accept Paul Kiparsky’s theory of three degrees of optionality and those who stand with the traditional understanding.

As is well known, Kiparsky (1979) hypothetically postulates that the three main option markers must express three different degrees of optionality in Pāṇini’s Aṣṭādhyāyī. He (1979: 1) proposes that vā indicates ‘or rather, preferably’, vibhāṣā ‘or rather not, preferably not’, and anyatarasyām ‘either way’. He also maintains that this distinction was unknown to Kātyāyana, Patañjali and their successors.

Some scholars have readily accepted Kiparsky’s claims. J. D. Smith (1982: 185), for instance, says that “there is no serious possibility that Kiparsky is in error.” Bronkhorst (1982: 273), in his review article, stated that “it can be said that the author has established this his [sic] thesis beyond reasonable doubt.” In a recent publication, he (2019: 24–25) even goes to the extent of presenting the hypothesis as a well-established fact: “Panini [sic] used a number of terms to indicate the optional use of certain formations. … Patanjali [sic] shows no awareness of this distinction.” Deshpande (1984) gives a more cautious assessment.

As critics of Kiparsky have pointed out, however, it may well be the case that Patañjali was not aware of the distinction because there was never a distinction in the first place. George Cardona (1999: 162–79) emphatically dismisses Kiparsky’s hypothesis. He (1999: 173) reiterates his previous assessment stating, “I conclude that Kiparsky’s main claim fails in view of the evidence, and I consider myself fully justified in maintaining, as before (Cardona 1989: 66), that his thesis is ‘neither cogently maintained nor acceptable.”’ In his review of Kiparsky 1979, Palsule (1982: 340) observes, “The problem as to why Pāṇini should have used three different terms to convey one and the same idea did not bother the tradition because of the accepted dictum: Paryāyaśabdānāṁ lāghavagauravacarcā nādriyate.” Devasthali was also among those who found major weaknesses in Kiparsky’s methodological assumptions. He (1983: 148) writes,

I should now like to draw the attention of the readers to one important point which P[aul] K[iparsky] seems to have totally neglected; and that is the general attitude that people like Pāṇini had toward the Veda. It is found expressed in a nut-shell in dṛṣṭānuvidhiś chandasi bhavati. (In chandas [= the Veda] what is seen is to be (only) explained or accounted for). The question of preference or otherwise simply does not exist. Hence, at least so far as Vedic is concerned, no believer in the Veda in ancient India, could have thought of using the word preferable or not preferable, with reference to a word, or phrase, or a statement in the Veda.

1.2 Optionality in Pāli grammar

When it comes to Pāli vyākaraṇa treatises, little work has been done on the role of option markers. In general, it is agreed upon that Pāli grammarians are not particularly systematic and, consequently, their use of option markers is not as precise as that of their Sanskrit counterparts. With reference to Kiparsky’s hypothesis, Mahesh Deokar (2008: 367) has observed, “such minute distinctions are not observed so rigorously by the Pāli grammarians.” Deokar also shows that navā is used in the sense ‘rarely’ in the Kaccāyana grammar (Kacc.) in places such as Kacc. 21. He (2008: 368–69) deals with other occasional technical terms for optionality as well. He seems to locate the origin of the lack of systematic use for optionality in Pāli grammars in the Sanskrit Kātantra grammar because Śarvavarman would not have systematically followed Pāṇini’s option system. Deokar (2008: 369) writes,

In the earliest portion of the Kātantra, i.e. in the Sandhi section, navā is used to denote vikalpa or vibhāṣā. According to the commentators, in the Kātantra the particle vā is used in two different senses, that of conjunction (samuccayārtha) and optionality (vikalpārtha).

Émile Senart (1871: 14) claimed that it is impossible to ascertain the actual meaning of vā in Kacc. He (1871) usually translates it quelquefois ‘sometimes’ or à volonté ‘if one wills, optionally’.¹ In his translation of the Ākhyātakappa, D’Alwis (1863: 25–26) translated vā as ‘optionally’ and kvaci as ‘sometimes’, but unfortunately there are no further explanations. The late Ole Pind (2012: 83) too has maintained that the difference between vā and kvaci in Pāli grammars is negligible. This view is partly true, and thus we find in the most recent integral translation of Kacc. into English, Ashin Thitzana (2016), a scholar-monk trained in the traditional system, levels down all the four option markers in Kacc. to ‘sometimes’, but he occasionally, for example 2016: 173, leaves the option marker in the sutta untranslated.² If we examine the Kaccāyana commentarial tradition, however, we find abundant discussions of and insights into the specific role of option markers. Even though the principle that different option markers designate different degrees of variation does not work in all cases, as Pāli grammarians acknowledge, it is noteworthy, because it helps us to delimit the exact scope of the rules. In the present article, I look at the earliest attestation of this theory and offer a preliminary analysis of its validity.

2 Two degrees of optionality in Kaccāyana

The general principle of the two degrees of optionality in the Kaccāyana Pāli grammar is laid out by Vimalabuddhi (Pind 2012: 71) in his Mukhamattadīpanī (Mmd., 10th c. CE). It can be summarised as follows:

vā (and vibhāsā) indicates an open option (vikappa) where two or more forms are possible.
kvaci (and navā) indicates that only one form is possible.

The metagrammatical axiom, “indeclinables (nipātas) have many meanings,” seems to be the foundation on which Vimalabuddhi’s theory relies; for option markers are all nipātas, and nipātas do not have a definite meaning.

The only explicit formulation of this principle in Vimalabuddhi’s lengthy commentary derives from his analysis of the term navā as a single word or expression (ekam padam) in Kacc. 21 Ivaṇṇo yam navā.³ The laws of recurrence (anuvutti) lead Vimalabuddhi to the conclusion that navā and na vā are not only different expressions but even mark different degrees of optionality. The conflict of priority that triggers the discussion on option markers in Mmd. is between one rule in the Sandhi section and two rules in the Nāma section, whose scope overlap. The rules with examples are as follows:⁴

Kacc. 21 Ivaṇṇo yam navā “The phoneme (vaṇṇo) i/ī [becomes] y [before a vowel], optionally (navā).” Examples: paṭisanthāravuty assa [← paṭisanthāravuti assa] ‘he will be of agreeable nature’, sabbā vity anubhūyate [← sabbā vitti anubhūyate] ‘all the happiness is enjoyed’.
Kacc. 70 Jhalānaṁ iyuvā sare vā “jha (i/ī masc./neut. endings) and la (u/ū masc./neut. endings) [become] iy and uv [respectively] before a vowel.” Examples: tiyantam [← ti antaṁ] ‘thus ending’, pacchiyāgāre [← pacchi-āgāre] ‘basket(?)-house’, bhikkhuvāsane [← bhikkhu-āsane] ‘monk’s seat’, puthuvāsane [pu-thu-āsane] ‘individual seat’.
Kacc. 71 Yavakārā ca “And also [jha (i/ī masc./neut. endings) and la (u/ū masc./neut. endings) become] ya and va [respectively, before a vowel].” Examples: agyāgāraṁ [← aggi-āgāraṁ] ‘firehouse’, cakkhvāyatanaṁ [← cakkhu-āyatanaṁ] ‘eye-base’, svāgataṁ [← su-āgataṁ] ‘welcome’.

As shown in Table 1, these rules provide that a replacement (ādeśa) occurs in place of a substituend (sthānin) with varying degrees of optionality, either exceptionally (navā) or optionally (vā).

Table 1. Degrees of optionality in Kaccāyana rules replacing i with y or iy before a vowel

Sutta	Substituendum	Replacement	Optionality
21	i/ī	y	navā
70	i/ī	iy	vā
	u/ū	uv
71	i/ī	y	vā
	u/ū	v

The conflict, then, is about the sandhi of i/ī followed by vowel. There are three possible results:

exceptional change into y,
optional change into iy,
optional change into y.

According to Vimalabuddhi, the marker navā in Kacc. 21 overrules the vā of Kacc. 70, 71 (the relevant passage is translated in Appendix 2 §7.3 (p. 453)). This interpretation is difficult to accept if Kacc. is analysed as a prescriptive grammar. It makes sense, however, if Kacc. is understood as a grammar that records the usage of a corpus, which in this case consists of the Tipiṭaka and the commentaries thereupon. What Vimalabuddhi intends is that those exceptions under Kacc. 21 are not affected by the alternatives offered in 70, 71. Thus, in most cases i/ī may be replaced with either iy or y: we are here in the realm of vikappa ‘open option’ marked by vā. But in some specific cases, kvaci (or as here, navā), it is replaced with y only. Here kvaci (navā) establishes a restriction to the vikappa, and as we can see, in the sequence of Kacc. suttas, the restriction appears before the general rule.

3 Option markers in Kaccāyana

At first sight, the two different levels of optionality in Kacc. do not seem to indicate any degree of preference. As Devasthali observed regarding Pāṇini’s description of Vedic language, Pāli grammarians are mostly interested in determining which forms are optional in the corpus and which forms are invariable. We shall now examine how the use of the four option markers in Kaccāyana compare with actual usage in the Tipiṭaka. Although a reliably conclusive survey cannot be conducted in the absence of a tagged critical edition of the canonical corpus, I make a preliminary analysis using the GRETIL e-texts.

3.1 (a) navā

The word navā, which according to Mmd. has to be read as one pada, appears four times in the Kacc./KaccV. text. In the vutti ‘gloss’, it appears as one single expression after the verb (e.g. KaccV. ad Kacc. 21 Yakāraṁ pappoti navā). It needs to be distinguished from na vā, which appears separated in the gloss (e.g. KaccV. ad Kacc. 46: iti vuttarūpā na honti vā). Following the principle of optionality described in §2 (p. 435), it expresses an exception in Kacc. 21, 392, and it also seems to mark an exception, but less clearly, in Kacc. 144, 147v. Here the option with navā seems be the preferred one, or at any rate the most common in the canon.

3.2 (b) vibhāsā

If we follow Vimalabuddhi’s interpretation, the word vibhāsā generally marks an open option that may yield two forms. It is equivalent to vā. It is not possible, with the materials available to us, to determine whether it indicates a more or less preferred option. The marker vibhāsā appears only twice in the Kaccāyana, and interestingly both times in the context of compound (samāsa) derivation: Kacc. 154 Samāse ca vibhāsā, a rule that is not based on canonical evidence,⁵ and Kacc. 325 Vibhāsā rukkha-tiṇa-pasu-dhana-dhañña-janapadādīnañ ca.⁶

3.3 (c) vā

The use of vā as a marker of vikappa ‘open option’ is widely confirmed, for instance in rules like Kacc. 137 Nāmhi raññā vā “[The word rāja, including its vibhatti, is replaced with] raññā optionally (vā in the instrumental singular (nāmhi vibhattimhi).” The example given is raññā, which occurs 426 times in the canon, vs. rājena, which occurs much less frequently with just 50 occurrences. Interestingly, in Kacc. 135 Rājassa rañño rājino se, no option is offered. It seems that rañño and rājino are the only accepted forms for the gen./dat. sg. This is not true in our recension of the canon, where the frequency of these words is: rañño 2,112 occurrences, rājassa 229, and rājino 40. But this is explained because rājassa is not found alone, but as the last member of a compound: mahārājassa, devarājassa, migarājassa, etc. The open option is clear even in rules that yield results that are not attested in the canon, like Kacc. 156 Āne simhi vā “āne [is the ending of puma] in loc.sg., optionally,” which yields the alternative ghost forms pumāne or pume.

In some places vā does not mark an open option — or, if it does, it is not quite clear how. For instance, in Kacc. 13 Vā paro asarūpā “optionally, the latter [vowel is elided] after a non-homogeneous [vowel],” the only way to understand vā as vikappa is not by trying to find variants to the specific examples (cattaro ime can only become cattāro ’me, never *cattār ’ime), but by locating variants in general sandhi: when two non-homogeneous vowels meet, either may be elided. Thus, vā marks a general option: when two vowels coalesce, either may be elided. For restrictions to this general optionality, kvaci or navā will be used (e.g. Kacc. 14).

The option taught by the rule with vā coincides at times with a more frequent form, and one could infer that it indicates some sort of preference, e.g. Kacc. 481 Hi lopam vā “hi is optionally elided”. This rule makes the 2nd sg. imperative suffix hi optional: gacchāhi/gaccha, gamehi/gama, gamayāhi/gamaya. A brief survey gives the following occurrences: gaccha 189 vs. gacchāhi 9. The pairs gama/gamehi and gamaya/gamayāhi are not attested. In this rule vā expresses an open option. It could also be understood as a preferred option, but to assess the actual frequency of the use of hi or its deletion a much more detailed survey of the available literature would be necessary.

Similarly, Kacc. 622 Tuntūnatabbesu vā “Optionally, [r in kar ‘to do’ is replaced by t] when tuṁ, tūna and tabba follow.” Examples: kattuṁ kātuṁ, kattūna kātūna, kattabbaṁ kātabbaṁ. Here we have to understand that the first option is kar + t° with regressive assimilation: *kar-tuṁ → kattuṁ 3 vs. kātuṁ 276. This is a case of an open option where vā introduces what seems to be a less frequent option.

Another example where vā does not indicate preference is Kacc. 507 Jaramarāṇaṁ jīrajiyyamiyyā vā “jīra, jiyya replace jara ‘decaying’, and miyya replaces mara ‘dying’, optionally.” The words jarati and jaranti are not attested, except in grammatical works (cf. Sadd. 167.9; 560.8). PED. s.v. jarati gives the form jarayati, but we find jārayetha Ja V.501.14*, v.l. B^e S^e jirayetha (cf. DOP. s.v. *jarati). As to the replacements provided by this rule, the number of occurrences is as follows: jīrati 9, jīranti 8 vs. jiyyati 3, jiyyanti 0, jīyati 16, jīyanti 30; marati 25, maranti 4 vs. miyyati 3, mīyati 20, miyyanti 0, mīyanti 21; marāmi 3 vs. miyyāmi 4.

Sometimes it is not possible to determine whether a certain higher frequency in use represents any preference, because the contrast in numbers is not radical, e.g. marāmi 3 vs. miyyāmi 4. In conclusion, we can say that vā does generally express an open option, where two forms are acceptable, but it is not possible to observe any clear systematic indication of preference.

3.4 (d) kvaci

The word kvaci seems to indicate exception or restriction of a former general rule, but as some critics of Kacc. have noted, the use of this marker is not as systematic as the principle declares. In Kacc. 24 the sutta seems to refer to those cases where only one form is possible, as the sandhi of ko imaṁ, which always results in ko ’maṁ, never *k’ imaṁ. As to frequency of use, kvaci does not necessarily indicate a less preferred option, but it seems to indicate a more restricted option. Consider, for instance, Kacc. 48 Kvaci paṭi patissa “In some places pati is replaced by paṭi.” The example pataggi is not attested vs. paṭaggi 5 occurrences; patihaññati is only attested in the Niddesa vs. paṭihaññati 22 times. Words with paṭi°19,522 vs. pati°2,823. If we read kvaci in its strict sense of exception, that would mean that some words only accept the retroflex option. Given that Pāli manuscripts oscillate in their representation of dentals and equivalent retroflex consonants, it is impossible to determine whether the word kvaci here indicates exception or not. However, it seems that the examples given in KaccV. follow the principle. The example paṭaggi ‘a counteracting fire’ DOP.), for instance, is a result of the restriction of kvaci, and it never appears as *pataggi (see, however, the compound gahapataggi ‘fire of a householder’, with dental, never gahapaṭaggi).

Similarly, in Kacc. 250 Kvaci to pañcamyatthe “In some places the affix to [is] in the sense of ablative,” the rule cannot possibly mean that the affix to is less frequent in general: sabbato 12 vs. sabbasmā 0,⁷ sabbamhā 0;⁸ yato 758 vs. yasmā 216, yamhā 7; tato 1,646 vs. tasmā 1,331, tamhā 243; ato 9 vs. asmā 40, amhā 69; ito 1,363 vs. imasmā 22, imamhā 51. If kvaci here means exceptionally, it must be in the sense that the ablative sense is generally not denoted with this affix, but with nā and other vibhattis.

For a case where kvaci means exception, consider Kacc. 308 Kvaci dutiyā chaṭṭhīnam atthe “In some places the accusative (dutiyā) [is used] in the sense of a genitive [or dative] (chaṭṭhīnaṁ).” In the sentence, api ssu maṁ aggivessana tisso upamāyo paṭibhaṁsu “Moreover, Aggivessana, three similes occurred to me” (M. I.240.2.9), maṁ means me or mama ‘of/to me’. This usage is certainly exceptional, and kvaci suits the context. Furthermore, the set of paribhāsā rules traditionally known in Burma as mahāsuttas ‘great rules’ (Kacc. 405, 406, 519)⁹ mark exceptions, particularly those connected with indeclinable forms, and here kvaci clearly plays the role assigned by Vimalabuddhi.

4 Development of Vimalabuddhi’s principle

Vimalabuddhi’s theory of optionality evolved in subsequent commentarial interpretations of the Kaccāyana. Buddhappiya’s Rūpasiddhi (Rūp., 12th century, Coḷa kingdom) represents an attempt to correct Mmd. in many ways, including the use of option markers. Although this may not be explicit in Rūp., it is explained in the Rūpasiddhi-ṭīkā (RūpṬ.), traditionally ascribed to the same author.¹⁰ The ambivalent nature of the word kvaci is discussed in RūpṬ. ad Rūp. 35 (= Kacc. 24) sare kvaci, where Buddhappiya criticizes Vimalabuddhi and makes clear that the use of option markers is related to restriction:

Yaṁ pana ñāse sarā sare lopan ti ādīni vatvā sare kvacī ti vuttattā kvaciggahaṇena vinā pi imass’ āniyatabhāve viññāyamāne pi puna kvaciggahaṇakaraṇe payojanaṁ pana na katthaci hoti, katthaci na hotī ti idam eva ñāpanatthaṁ, atha kho ekekassa rūpadvayuppādanatthan ti vuttaṁ,¹¹ taṁ na yujjati, aniyatabhāve siddhe puna aniyatabhāvāya kvaciggahaṇassa niratthakattā. yañ ca tatth’ eva purimasuttavaṇṇanāyaṁ sare kvacī ti ettha kvacisaddo vāttho ti ca vuttaṁ,¹² tañ ca na gahetabbaṁ. evaṁ hi sati vā paro asarūpā ti ettha vāggahaṇam akatvā sarā sare lopan ti etth’ eva vāggahaṇaṁ kareyya, na ca kataṁ. imañ ca suttaṁ ācariyo n’ ārabheyya, āraddhañ ca. tena viññāyati na c’ āyaṁ kvacisaddo vāttho ti.¹³

In the Nyāsa (= Mmd.) we find the following statement: “Since sare kvaci is formulated after having stated sarā sare lopaṁ [Kacc. 12] and so forth, even without the word kvaci the lack of restriction (= the optionality) [of sare] would be understood; nevertheless, the word kvaci is still used, not to express the sense ‘in some places it obtains, in some places it does not obtain’, but in order to allow for the derivation of two forms for each (ekekassa) [concept].” Now this statement is not appropriate, because if the non-restriction was so clear, it would be pointless to use the word kvaci to express further non-restriction. And what is stated in the same commentary [i.e. Nyāsa] on the previous sutta, namely, “in sare kvaci, the word kvaci has the meaning of vā,” this should not be accepted either. For, if we did accept it, we would not find vā in vā paro asarūpā; we would find it instead in sarā sare lopaṁ, and that is not the case; similarly, the Master [Mahā Kaccāyana] would not have introduced the present rule [sare kvaci], but he has actually done it. From this it is inferred that this word kvaci does not have the meaning of vā.

Buddhappiya added precision to Vimalabuddhi’s principle by using a tripartite classification of rules: (1) nicca ‘mandatory’, (2) anicca ‘not mandatory’ and (3) asanta ‘not applicable’. In his vutti on the Kacc. rules, Buddhappiya specifies the exact scope of vā and kvaci, adding layers of restriction when it is needed.¹⁴

In the 13th century, the Burmese grammarian Guṇasāgara of Pagan summarised Vimalabuddhi’s principle in his Mukhamattasāra (ca. 13th c. CE) as follows:

Kvaci navā ca ekatthā yebhuyyen’ ekarūpakā;
vāvibhāsā samānatthā pāyen’ obhayarūpakā ti.¹⁵

Kvaci and navā have the same meaning, they generally (yebhuyyena) express one form; vā and vibhāsā have a common meaning, mostly (pāyena) expressing two forms. (Ruiz-Falqués 2015: 142).

This stanza was quoted in major Kaccāyana commentaries. Chapaṭa Saddhammajotipāla’s Suttaniddesa (15th century, Pagan) quotes the stanzas, without naming the source, when discussing the word vibhāsā in Kacc. 325. This commentary, we shall remind, was the official commentary on Kacc. in early modern monastic education in Burma (Dhammasami 2004: 321). Mahāvijitāvī too, in his Kaccāyanavaṇṇanā (16th century, Pinya) quotes the same stanza, without explicit attribution, when commenting on Kacc. 14 (KaccVaṇṇ. 27.8–9.) This theory of optionality has reached our days through the Burmese commentarial tradition. It is discussed in Ven. Janakabhivamsa’s Kaccāyana bhāsā-ṭīkā, a widely circulating Burmese textbook on the Kaccāyana. It is also found in English translations that derive, totally or partially, from the Burmese Kaccāyana tradition, notably the studies of Ven. Malai (1997: 105) and Bhikkhu Nandisena (2005: 48).

5 Optionality in Kaccāyana and Sanskrit

A thorough study of optionality in the Pāli vyākaraṇa tradition remains a desideratum. But even if the present survey does not exhaust all the available material,¹⁶ it shows that Pāli grammarians do have a theory of optionality and they apply it in their interpretation of the suttapāṭha. In most cases the words kvaci and navā mean ‘exceptionally, rarely’, and they establish some sort of restriction to a more general rule, whereas the words vā and vibhāsā generally mean ‘optionally’ and they express an open option (vikappa). This is how Pāli grammarians, from Vimalabuddhi onwards, have read the Kaccāyana text through the centuries, but this is not reflected in most modern translations and studies of the text.

If we compare Vimalabuddhi’s theory of the two levels with Kiparsky’s thesis, we can see that they show certain similarities. Indeed, Pāli grammarians accept that some option markers prevent variation, rather than allowing for it. They restrict the scope of certain rules. But there is a major difference between the Pāli grammarians’ view and Kiparky’s. Pāli grammarians are clearly not concerned with issues of preference. What Devasthali states regarding Vedic Sanskrit can also be applied to canonical Pāli. The role of the grammarian is simply to describe. Hence, the restrictive effect of markers kvaci and navā is directly related to the dialectics between general (utsarga) and particular or exceptional (apavāda) rules. The restriction of kvaci and navā can only operate within a major level of optionality established by the marker vā/vibhāsā. Strictly speaking, therefore, there is only one form of optionality, which is open and allows for two or more forms as long as they are attested in the literature. Whenever the rule needs to be refined, markers such as niccam ‘always’ or kvaci ‘exceptionally’ are used. In conclusion, then, the theory of the two levels in Pāli does not support Kiparsky’s thesis and rather agrees with the view of Cardona, Devasthali and those scholars who have sought to explain the variety in option marker terminology following the evidence from the commentaries and avoiding conclusions that are impossible to substantiate.

As Pāli grammatical treatises have been crucial in the textual transmission of the canonical and exegetical literature of the Theravāda school (H. Smith 1928; Hinüber 1983; Pind 2012; Gornall 2020; Ruiz-Falqués 2019), the study of option markers could shed light on the morphology of Pāli as a language. Particularly interesting would be a study of fossilised or “frozen” forms and sandhi ligatures, that have not yet been studied in the broader context of Pāli formulae, prose style and mnemonics. What appears to be a marginal aspect of Pāli grammars, then, could potentially reveal hidden aspects of the early Buddhist literature.

6 Appendix 1: Distribution of technical terms for optionality in Kaccāyana and Kaccāyana–vutti

The following table presents the terms used in the Kaccāyana and the Kaccāyanavutti, the number of the sutta in which the term occurs (marked with ‘v’ if in the vutti), the number of the last sutta to which recurrence (anuvutti ‘Anu.’) extends or ‘ES’ if to the end of the section, the term that blocks further recurrence, and glosses of the term in the Mukhamattadīpanī or Kaccāyana-vutti and other relevant notes.

	Term	No.	Anu.	Blocker	Gloss/Note
	Sandhikappa
	I.1–I.5
1.	vā	13		kvaci	Mmd.: vikappena
2.	kvaci	14	20	navā	Mmd.: kvaci
3.	navā	21	22	ES	Mmd.: navā; Kacc-nidd: navā ti kvacatthaniddeso
4.	kvaci	24	27	ṭhāne	Mmd.: kvaci
5.	ṭhāne	28	29	ES
6.	vā	31	35	kvaci	Mmd.: vikappena
7.	kvaci	36	39	vā	Mmd.: kvaci
8.	vā	40		ca(?)	Mmd.: vā
9.	kvaci	42	43		Mmd.: kvaci
10.	na vā	46	na 47	kvaci	Mmd. explains that this is na vā not navā. Gloss of vā Mmd.: vikappena
11.	kvaci	48	50	paribhāsā 51	Mmd.: kvaci. Anuvutti skips 49 byañjane.
	Nāmakappa
	II.I
12.	vā	64	65		Mmd.: vikappena
13.	vā	68	69	vā	Mmd.: vikappena
14.	vā	70		ca	KaccV.: vā ti vikappanatthaṁ; Mmd.: vikappena
15.	vā	76		ca	Mmd.: vikappena
16.	vā	80	81
17.	vā	93			Mmd.: vikappena
18.	vā	94	95	yosu	Mmd.: vikappena
19.	vā	99		na	Mmd.: vikappena
20.	vā	105	107		Mmd.: vikappena
21.	vā	108	109	na	Mmd.: vikappena
22.	vā	113		ca	Mmd.: vāggahaṇaṁ paggahaṇassa [from Kacc. 112 Pato yā] nivattanatthaṁ
23.	vā	117	119	ES	Mmd.: vikappena
	II.2
24.	vā	123		simhi	Mmd.: vikappena. No anuvutti in 124-126, but KaccV.: vā again in 127
25.	vā	128	129	amussa	Mmd.: vikappena
26.	vā	136		vā	Mmd.: vikappena
27.	vā	137		smimhi	Mmd.: vikappena
28.	navā	144		nāmhi	Mmd.: vikappena
29.	vā	147v			navā is anuvṛtti from 144 by frog’s leap: Mmd.: savibhattiggahaṇañ ca tumhamhaggahaṇañ ca maṇḍukagatyā navāggahaṇañ ca vattate.
30.	vā	150		bahuvacanesu	Mmd.: vikappena
31.	vibhāsā	154		yosu	Mmd.: vikappena
32.	vā	156		hivibhattimhi	Mmd.: vikappena
33.	vā	158	160	ES	Mmd.: vikappena
	II.3
34.	vā	162	163		Mmd.: vikappena
35.	vā	162	163		Mmd.: vikappena
36.	vā	165		na	Mmd.: vikappena
37.	vā	170		ca	Mmd.: vikappena
38.	vā	173v
39.	vā	175	179	na	Mmd.: vikappena
40.	vā	181		ca	Mmd.: vikappena
41.	ca	185v			casaddaggahaṇaṁ kvaci sakārassa’ eva pasiddhatthaṁ (KaccV. 61.3)
42.	vā	186v		sesesu
43.	vā	188v
44.	vā	195		vā	Mmd.: vikappena
45.	vā	196		tu	Mmd.: vikappena
46.	vā	201	203	ca	Mmd.: vikappena
47.	vā	210		ES	Mmd.: vikappena
	II.4
48.	vā	216	217	niccaṁ	Mmd.: vikappena
49.	niccaṁ	218	219	pi
50.	vā	231		niccaṁ	Mmd.: vikappena
51.	niccaṁ	232		vā
52.	vā	233		ca	Mmd.: vikappena
53.	vā	238			Mmd.: vāggahaṇaṁ panānadādayo sampiṇḍeti
54.	vā	241v
55.	kvaci	244	245
56.	vā	248		ES	Mmd.: vikappena
	II.5
57.	kvaci	250
58.	vā	252v
59.	vā	262		ca	Mmd.: vikappena
	II.6 kāraka
60.	vā	273			Disjunctive
61.	vā	276			Mmd.: vāggahaṇaphalaṁ sayamevavakkhati
62.	vā	278			Disjunctive
63.	ca	279			KaccV. casaddaggahaṇaṁ vikappanatthaṁ
64.	vā	281			Disjunctive. Mmd.: vāsaddo samuccayattho
65.	vā	302		sāmismiṁ	Mmd.: vikappamatthaṁ
66.	kvaci	308	311		Mmd.: kvaci
	II.7 samāsa
67.	vibhāsā	325		dvipade tulyādhikaraṇe	Mmd.: vikappena KaccVaṇṇ.: 272-4-5: vibhāsā ti vikappanattha [sic]. Sadd. 127.12ff.: read loc. sg. not acc. pl.
68.	kvaci	339		ca (?)	Mmd.: kvaci
69.	kvaci	343v
	II.8 taddhita
70.	vā	346	348		Mmd.: vikappena
71.	vā	349	354	ca	KaccV.: vikappanatthena; vā in 352 vikappanatthena according to KaccV.; Mmd.: vikappena only by anuvṛtti of ca
72.	tu	362
73.	vā	376		ī	Mmd.: vikappena
74.	niccaṁ	378
75.	tu	382
76.	vā	383v	384
77.	vā	385			Mmd.: vikappena
78.	navā	392			Mmd.: navā
79.	tu	400
80.	vā	402		ṭhāne	Seems disjunctive, not glossed by KaccV. as vā hoti
81.	ṭhāne	403	404	kvaci
82.	kvaci	405	406	ca	Mmd.: kvaci
	Ākhyātakappa
	III.1
	III.2
83.	vā	435	436		Mmd.: vikappena. E^e om.
84.	vā	444	445	kattari	Mmd.: vikappena
	III.3
85.	kvaci	460		abbhāso	Mmd.: kvaci
86.	kvaci	460		abbhāso	Mmd.: kvaci
87.	vā	465			Mmd.: vikappena
88.	vā	467	472	vā	Mmd.: vikappena
89.	vā	473		chappaccayesu	Mmd.: vikappena
90.	vā	478		vacassa	Mmd.: kvaci
91.	vā	481	483	ES	Mmd.: vikappena. KaccV. 483 adds the clause niccaṁ for bhavissantī ‘future’ tense
	III.4
92.	vā	486		aññesu	Mmd.: vikappena
93.	vā	490		ghe	Mmd.: vikappena
94.	vā	493v	494	tthattaṁ
95.	vā	501v	503		Cf. Pind 2013: 169 n.4: E^e dadhātussa dajjaṁ vā = Rūp. 493, cf. Sadd.: 1005: dāssa vā dajjo
96.	vā	507v	509	vā	Mmd.: vikappena
97.	vā	510v	516	kārite	Mmd.: vikappena
98.	kvaci	519v	521	brūto	Mmd.: kvaci
99.	kvaci	523v		vā
100.	vā	524v		lopaṁ	Mmd.: vikappena
	Kibidhānakappa
	IV.1
101.	vā	529			Mmd.: vāggahaṇena kammādimhītimassa vikappanato
102.	vā	540			Mmd.: vā = sampiṇḍana
	IV.2
103.	vā	555	556		Mmd.: vikappena. KaccV. 183.10: vā changes function: itthiyaṁ anitthiyaṁ vā
104.	vā	563		ca?	Mmd.: vāggahaṇaṁ dutiyassapaccayassa saṁpiṇḍanatthaṁ
105.	vā	566
	IV.3
106.	ca	574	576	vā	KaccV. 189.23 ṭhāne
107.	vā	577		ca?	Mmd.: vikappena
108.	vā	581	582		Mmd.: vāggahaṇaṁ avadhāraṇatthaṁ. anekatthā hi nipātā. tena c’ ettha śabdaṁ vuttavidhānaṁ niyataṁ yeva hoti
109.	vā	584		ca?	Mmd.: vikappena
110.	vā	588v			Pind 2013: 193, n.13: Sadd. 1190. This rule is clearly defective, but the readings are confirmed by Mmd. 450.23. Sadd. loc. cit. adds lopaṁ after anto, and B^e and Rūp. 537 substitute tuṁ tabbādisu na for anto
111.	vā	591		ES	Mmd.: vikappena
	IV.4
112.	vā	594			Mmd.: vikappena
113.	vā	596	600		Mmd.: vikappena
114.	vā	602v		ca(?)
115.	kvaci	608		ES	Mmd.: kvaci
	IV.5
116.	vā	615	616	kvimhi	Mmd.: vikappena
117.	vā	622	623		Mmd.: vikappena
	IV.6
118.	vā	629		samādhīhi	Mmd.: vā
119.	vā	631			Mmd.: vikappena
120.	vā	668			KaccV.: kvaci, Mmd.: kvaci

7 Appendix 2: Vimalabuddhi on Kaccāyana 21

Here is presented the text of Mmd. 32.24–34.30 with my translation. I have adjusted the punctuation and put the page numbers to the edition in square brackets.

7.1 Gloss and general purport of the rule

Ivaṇṇo yaṁ na vā. Kimattham idam uccate. Ivaṇṇo sare pare yakāraṁ pappoti navā ti ñāpanatthaṁ. Ivaṇṇo ti ekaṁ padaṁ. Yan ti ekaṁ padaṁ. Navā ti ekaṁ padan ti tipadam idaṁ suttaṁ. Sareggahaṇassa nimittabhāvenānuvattanato ivaṇṇo sare pare yakāraṁ pappoti navā ty attho. I eva vaṇṇo ty atthe nāmānaṁ samāso yuttattho ty adhikicca dvipade tulyādhikaraṇe kammadhārayo ti samāso ca vibhattilopo ca pakatibhāvo ca nāmam iva katvā vibhatyuppattādīni ca [33] vuttanayen’ eva.

[Kacc. 21] The speech-sound i [when a vowel follows, sare pare Kacc. 12]¹⁷ takes the form y, exceptionally (na vā).

What is the purpose of this statement? It is in order to convey that the speech-sound i, when a vowel follows, is replaced by y, exceptionally. Ivaṇṇo is one word. Yaṁ is one word. Navā is one word. This sutta consists of three words. The meaning is that the speech-sound i is replaced by y before a vowel, because of the recurrence (anuvattanato) of the word sare (sareggahaṇassa) [from Kacc. 12] as a condition (nimittabhāvena). With regard to the meaning, “i itself [is] a speech sound,” on the basis of Nāmānaṁ samāso yuttattho, “A compound is that which has the combined meaning of nouns” [Kacc. 318], and Dvipade tulyādhikaraṇe kammadhārayo “When both words have the same substratum, [that tappurisa compound receives the technical term] kammadhāraya” [Kacc. 326], the compound, the elision of the case ending and the nominal stem are made like a noun, and the formation of the case endings and so forth are carried out according to the rule.

7.2 Objection and refutation 1: homogeneity between i and ī

[Codanā:] Ivaṇṇo ti vuttattā īvaṇṇo kathaṁ saṅgahaṁ gacchatī ti.

[Objection:] Since only ivaṇṇo is stated, how can īvaṇṇo be included?

[Parihāra:] Samānarūpattā. Yathā pana ajo ti vutte dīgharassakāḷakodātādibhedabhinnā pi samānarūpattā ajasaṅgahaṁ gacchati, evaṁ sampadam idaṁ.

[Refutation:] Because they share the same form (rūpa). In the same way that when one says ‘goat’, this word comprises all sorts of goats, on account of having something in common, although they are different in shape: some are long, some short, some are black, some white, etc. This is how [the meaning of the sutta] is complete (sampadaṁ).

7.3 Objection and refutation 2: the rule is redundant because Kacc. 70 and 71 already provide for the same sandhi

[Codanā:] nanu jhalānam iyuvā sare vā, yavakāro ceti vuttattā imassāniyatatā pākaṭā va. kasmā idha navāggahaṇaṁ katan ti.

[Objection:] But is it not true that this non-restriction (aniyatatā) is already clear from the rules Jhalānaṁ iyuvā sare vā [Kacc. 70] and Yavakāro ca [Kacc. 71]? Why is the word navā used here?

[Parihāra:] Payojanantarasambhavato. Asati hi navāggahaṇe ivaṇṇassa sare pare iminā yakārādeso hoti. Puna yavakārā ceti vuttattā katthaci tena pi yakāro hoti. Puna jhalānam iyuvā sare vā ti vuttattā katthaci iyādeso hoti. Evaṁ imehi vinimuttā keci payogā kadāci pi na siyuṁ tasmā taṁnivattanatthaṁ idha navāggahaṇaṁ kataṁ. Tena yattha navāggahaṇena yakāro na hoti, tattha aññena pi yakāro na hoti. Ten’ eva hoti muttacāgī anuddhato ti.

[Refutation:] Because it [namely navā in Kacc. 21] serves another purpose. For, if the word navā were not there, the speech-sound i, when a vowel follows, by this rule should be [always] replaced with y. And because the rule Yavakārā ca [Kacc. 71] is stated, in some places the replacement y obtains. Again, because the rule Jhalānaṁ iyuvā sare vā [Kacc. 70] is stated, in some places the replacement iy obtains. Thus, some usages [of the y/iy replacement] that are independent from these [Kacc. 70, 71] would never take place. Therefore in order to prevent that, the word navā is used here [in Kacc. 21]. Because of this (tena), where y does not obtain by navā, it does not obtain by any other rule either. By this method only one obtains [the counter-example (cf. KaccV. 7.7)] muttacāgī anuddhato [and *muttacāgy anuddhato by Kacc. 71 and *muttacāgiy anuddhato by Kacc. 70 are immediately impossible to derive].

7.4 Objection and refutation 3: why use navā instead of kvaci

[Codanā:] Nanu kvaciggahaṇenā pi tadattho viññāyati, kasmā kvaciggahaṇam anapekkhitvā idha navāggahaṇaṁ katan ti.

[Objection:] But is it not true that the same purpose can be served by the word kvaci? Why is kvaci ignored here and navā is used?

[Parihāra:] Navāsaddo pi kvacattho ti dīpanatthaṁ. Samānatthattā yeva hi attanā vattabbatthaṁ.¹⁸ Ettha navāggahaṇam eva vadatī ti katvā kvaciggahaṇaṁ nivattate. Aññathā ivaṇṇaggahaṇe sati pi sareggahaṇaṁ viya bhinnatthattā tenā pi vattitabbaṁ siyā ti.

[Refutation:] The word navā is [used] in order to express the meaning of kvaci as well. It is precisely because they have a common meaning that it [kvaci¹⁹] itself can serve the same purpose. Here, because the word navā itself is used, one understands that the word kvaci is blocked (nivattate). Otherwise (aññathā), if they [kvaci and navā] had different meanings, [the word kvaci] would be still recurring, as the word sare is recurring in spite of the word ivaṇṇa being also there.

7.5 Objection and refutation 4: why kvaci and vā do not have the same technical meaning

[codanā:] Yadi evaṁ vā paro asarūpā ti ettha sati pi vāggahaṇe kvac’ āsavaṇṇaṁ lutte ti ettha kvaciggahaṇaṁ kataṁ. Na ca taṁ disvā tattha vāggahaṇaṁ vattate. Tasmā vāsaddena kvacisaddo pi samānattho ti sakkā mantun ti.

[Objection:] If that is so, one could argue the following (ti sakkā mantuṁ): “In the rule Vā paro asarūpā [Kacc. 13], even though the word vā is there, in the [next] rule, [namely] Kvac’ āsavaṇṇaṁ lutte [Kacc. 14], the word kvaci is used. And having seen this [kvaci], the word vā does not recur there. Therefore [we can infer that] vā and kvaci have the same meaning.”

[Parihāra:] Tan na. Vaggantaṁ vā vagge ti ito hi yavamadanataralā c’ āgamā ti ettha vāggahaṇe vattamāne pi kvaci o byañjane ti ettha kvaciggahaṇaṁ kubbaṁ ñāpayati bhinnatthā v’ ime ti. Nāpi ubhinnaṁ samānatthabhāvaṁ ñāpetun ti sakkā mantuṁ, dvikkhattuṁ ñāpane payojanābhāvā.

[Refutation:] This is not so. Because even though the rule Yavamadanataralā c’ āgamā [Kacc. 35] carries vā [by anuvutti] after (ito) the rule Vaggantaṁ vā vagge [Kacc. 31], [subsequently] in the rule Kvaci o byañjane [Kacc. 36], by supplying (kubbaṁ) the word kvaci, he [i.e. the author, Mahā Kaccāyana] indicates (ñāpayati) that they [i.e. vā and kvaci] have certainly different meanings. Furthermore (pi), it is not possible to think that these two have a similar meaning, because when it comes to indication (ñāpane), it cannot apply twice [i.e. redundancy is inadmissible].

7.6 Objection and refutation 5: in one case, kvaci and navā do not mean exactly the same

[codanā:] Yadi evaṁ, te navā ivaṇṇe ti ettha navāggahaṇe vijjamāne pi kvaci paṭi patissā ti ettha kvaciggahaṇena ime pi nānatthā ti sakkā mantun ti.

[Objection:] If that is so, it is possible to conclude that navā and kvaci also have different meanings here too, since the word navā is found in the rule Te navā ivaṇṇe [Kacc. 46] and it is also found in the rule Kvaci paṭi patissa [Kacc. 48].

[Parihāra:] Tan na. Tatth’ eva navāggahaṇassābhāvā. Na hi tattha navāggahaṇam atthi. Atha kho natthiggahaṇaṁ vāggahaṇañ ca. Āha ca te ca kho abhi adhi icc ete ivaṇṇe pare abbho ajjho iti vuttarūpā na honti vā ti. Ten’ eva atissa c’ antassā ti suttassa vuttiyaṁ ṭhapetvā vāggahaṇaṁ naggahaṇam eva vattate.

[Refutation:] That is not so, because of the absence of the word navā in that sutta. For clearly it is not the word navā that we find there, but the expression “It does not exist” (n’ atthi) plus (ca) the word vā ‘optionally’. And, indeed he [i.e. the author of the Kaccāyana-vutti] stated: Te ca kho abhi adhi icc ete ivaṇṇe pare abbho ajjho iti vuttarūpā na honti vā ti “And those abhi and adhi, when the speech-sound i follows, do not adopt the respective forms abbho and ajjho, optionally” (KaccV. 15.10–11). Because of this (tena), the word vā is excluded in the vutti of the sutta Atissa c’ antassā [Kacc. 47]; only the word na recurs [cf. KaccV. 15.16 Vuttarūpā na honti].

7.7 Refutation 5 continued

Tasmā [34] sandeham akatvā²⁰ kvacisaddo ca navāsaddo ca ime dve samānatthā. Vāsaddo ca vibhāsāsaddo ca ime dve samānatthā ti gahetabbaṁ. Tattha ādidvayañ hi katthaci hoti, katthaci na hotī ti dīpeti. Itaradvayam ekass’ eva rūpadvayaṁ janeti. Idam pi hi yathā hū bhū sattāyam iti. Ettha bhū t’ idaṁ abhimaddanānubhavanapātubhavanādianekatthe pi yebhuyyena sattatthavācakattā hū bhū sattāyan ti vuttaṁ. Evaṁ yebhuyyavasena vuttaṁ.

Therefore, without any doubt, it has to be accepted (gahetabbaṁ) that both the word kvaci and the word navā have the same meaning, and both the word vā and the word vibhāsā have the same meaning. Herein, the first pair (kvaci and navā) certainly shows (dīpeti) that [a certain grammatical phenomenon] in some places (katthaci) obtains, and in some places it does not obtain. The other pair (vā and vibhāsā) produces (janeti) two forms (rūpadvayaṁ) for one single (ekassa) [concept]. For this is also stated in cases such as: hū bhū sattāyaṁ. Here, even though the root bhū expresses different meanings such as ‘overpowering’, ‘experiencing’, ‘manifesting’, etc., because most commonly (yebhuyyena) it expresses ‘being’, it is defined as “hū bhū [verbal roots] in the sense of being”. In this way it is stated according to what is most common (yebhuyyavasena).

7.8 Conclusion of refutation 5: summary of Vimalabuddhi’s principle of two levels

Tasmā yattha ādidvayaṁ gahitaṁ, tattha vuttavidhānaṁ yebhuyyena katthaci hoti, katthaci na hoti. Ten’ ekassa rūpadvayuppādanam appakaṁ. Yattha pana antadvayaṁ gahitaṁ, tattha vuttavidhānaṁ yebhuyyen’ ekassa rūpadvayaṁ janeti, itaraṁ pana appakan ti daṭṭhabbaṁ. Aññathā sirivaḍḍhako ty ādīni viya nipātā ’nekā ’nekatthavisayā ti vacanamattaṁ siyā ti.

Therefore, it has to be considered (daṭṭhabbaṁ) that where the first pair [kvaci/navā] is found, in most cases, the operation (-vidhānaṁ) that is prescribed (vutta-) applies in some places, and, in some places, it does not apply. By this principle (tena), the double derivation of one word is rare (appakaṁ). However, where the other pair [vā/vibhāsā] is found, the operation that is prescribed, in most cases, generates two forms for one single [word], whereas the other [possibility, namely, that only one derived form is possible], is rare. Otherwise, the principle according to which “various (anekā) indeclinables (nipātā) have various scopes in terms of meaning (anekatthavisayā)” would be a mere verbal [convention], like the name Sirivaḍḍhako (“The one who increases glory”).

7.9 Objection and refutation 6: why not merge this rule with Kacc. 17?

[codanā:] Nanv idam pi vam odudantānan ti suttaṁ viya yam edantassādeso t’ iminā ekayogaṁ katvā yam edantānan ti sakkā vattun ti.

[Objection:] But is it not the case that this sutta too can be made one (ekayogaṁ) with Yam edantass’ ādeso [Kacc. 17], and one can simply say Yam edantānaṁ, much in the same way as we find the sutta Vam odudantānaṁ [Kacc. 18]?

[Parihāra:] Na sakkā, ubhinnam pi bhinnavisayattā. Yam edantassā ti hi antass’ eva yakāro hoti. Tena e ayā ti sutte ādiss’ ekārassa asati antatthe ekārassa yakāro na hoti. Iminā pana avisesena ivaṇṇo ti vuttattā ādissa pi yakāro hoti. Tena mā yūnam āgamo ṭhāne ti ettha i ca u ce ti dvande kate ādiss’ iminā yakāro hoti. Tadatthaṁ bhinnayogakaraṇan ti.

[Refutation:] It is not possible, because these two rules have different scopes. For, in the case of the rule which says Yam edantassa [Kacc. 17], the form y appears only as a replacement of an e that is at the end of a word. By this principle, in the sutta E aya [Kacc. 516], there is no replacement of e for y, because e is not in final position. By this present rule, however, because the word ivaṇṇa has been used non-specifically (avisesena), the replacement y can affect the beginning of a word too. That is why in the sutta Mā yūnam āgamo ṭhāne, “There is no [vuddhi] of i and u [when they are in initial position, but there is the] insertion (āgama) [of vuddhi vowels e and o] whenever suitable (ṭhāne)” [Kacc. 403], y appears as a replacement of initial i in the dvanda meaning ‘i and u’. It is with this purpose that different rules are posited.

7.10 Objection and refutation 7: more on why Kacc. 21 is not redundant

[codanā:] Nanv idam eva na vattabbaṁ. Yavakārā ca pasaññassa ceti iminā va sijjhanato. Yathā c’ etaṁ, tathā yam edantassādeso vamodudantānan ti idaṁ dvayam pi na vattabban ti.

[Objection:] But is it not true that this does not work either? Because by the two rules Yavakārā ca [Kacc. 71] and Pasaññassa ca [Kacc. 72] this [namely the same object as Kacc. 21] is accomplished. And by the same token, the two rules Yam edatass’ ādeso [Kacc. 17] and Vam odudantānaṁ [Kacc. 18] should not be there either.

[Parihāra:] Tan na. Asati hi imasmiṁ sandhivisaye jhalānam iyuvā sare vā ty ādinā iyuvādesā pi siyuṁ. Taṁnivattanatthaṁ siddhe saty ārambho vidhīyate.

[Refutation:] This is not so. Because if this [rule, namely ivaṇṇo yaṁ navā Kacc. 21] were not in the sandhi section (sandhivisaye),²¹ the replacements iy and uv would also take place as the result of Jhalānaṁ iyuvā sare vā [Kacc. 70] and so forth. In order to prevent that, as [the scope of the sutta] is well known, this operation is prescribed.²²

7.11 Examples

Paṭisandhāravuty āssa ty ādīni udāharaṇāni. Tesaṁ pana paṭisandhāravutti assa sabbā vitti anubhūyate ti chedaṁ katvā byañjanaṁ viyojetvā sarā sare lopan ty adhikicc’ iminā yakāraṁ katvā parakkharaṁ netvā rūpasiddhi veditabbā.

Paṭisandhāravuty²³ āssa (“let him be friendly”, Dhmp. 376), etc. are the examples. After establishing the division of the words ([pada] chedaṁ) as follows paṭisandhāravuti assa, sabbā vitti anubhūyate “all the prosperity is enjoyed,” and having separated (viyojetvā) the consonant [from the vowel by Kacc. 10],²⁴ subsequently the rule “vowels are replaced by 0 before a vowel” [Sarā sare lopaṁ Kacc. 12] is applied, and then, by the present sutta [Kacc. 21 Ivaṇṇo yaṁ navā], the replacement y [for 0] applies [and the previous elision is cancelled], and finally we join again the ending with the following speech-sound [by Kacc. 11 Naye paraṁ yutte]. In this way the final derivation should be known.

7.12 Counterexamples

Pañcahaṅgehī ty ādīni kimudāharaṇāni. Tesaṁ pana pañcahi aṅgehi muttacāgī anuddhato ti chedaṁ katvā byañjanaṁ viyojetvā iminā yakāre sampatte navāggahaṇenākatvā²⁵ rūpasiddhi veditabbā.

Pañcah’ aṅgehi ‘with the five limbs’, etc. are the counter-examples. In that case, having done the analysis as pañcahi aṅgehi and muttacāgī anuddhato, and having done the separation of the consonant, by the present rule the replacement y obtains, but due to the word navā it is not applied. In this way the final derivation should be known.

Abbreviations


A.	Aṣṭādhyāyī	Böhtlingk 1887
Dhmp.	Dhammapada	Andersen and H. Smith 1921
DOP.	A Dictionary of Pali	Cone 2001–2020
Kacc.	Kaccāyana	Pind 2013
KaccSNidd.	Kaccāyana-Suttaniddesa	Dharmānanda 1931
KaccV.	Kaccāyana-Vutti	Pind 2013
KaccVaṇṇ.	Kaccāyana-Vaṇṇanā	Vijitāvī 1916
M.	Majjhima-nikāya	Trenckner 1888
MBh.	Mahābhāṣya	Kielhorn 1962–1972
Mmd.	Mukhamattadīpanī	Vimalabuddhi 1933
MmdPṬ.	Mukhamattadīpanī-Purāṇa-ṭīkā	Laṅḥ 1914
MmdSāra.	Mukhamattasāra	Ruiz-Falqués n.d.
PED.	Pali-English Dictionary	Rhys Davids and Stede 1921–1925
Rūp.	Rūpasiddhi	Buddhappiya 2006
RūpṬ.	Rūpasiddhi-Ṭīkā	Buddhappiya 1964
Sadd.	Saddanīti	H. Smith 1930, 1949
Vism.	Visuddhimagga	Rhys Davids and Caroloine 1975

References

Andersen, Dines and Helmer Smith. 1921. The Pāli Dhātupāṭha and the Dhātumañjūsā: edited with indexes. Copenhagen: Royal Academy of Sciences of Denmark.

Böhtlingk, Otto, ed. and trans. 1887. Pāṇinis Grammatik: herausgegeben, übersetzt, erläutert und mit verschiedenen Indices versehen. 2nd ed. Vol. 3. Leipzig: Verlag von H. Haessel.

Bronkhorst, Johannes. 1982. “The variationist Pāṇini and Vedic: a review article.” Indo-Iranian Journal 24: 273–82.

—. 2019. A śabda Reader: language in classical Indian thought. New York: Columbia University Press.

Buddhappiya, Bhadanta. 1964. Padarūpasiddhiṭīkā. Mandalay: Padesā Piṭakat-cā-pe. [Undated reprint.]

—. 2006. Padarūpasiddhi. Yangon: Saccamaṇḍaiṅ Press. [3rd reprint.]

Cardona, George. 1989. “Pāṇinian studies.” New Horizons of research in Indology: silver jubilee volume, ed. by Vashishtha Narayan Jha, pp. 49–84. Pune: Centre of Advanced Study in Sanskrit, University of Poona.

—. 1999. Recent research in Pāṇinian Studies. 1st ed. Delhi: Motilal Banarsidass. xi, 372.

Cone, M. 2001–2020. A dictionary of Pāli. 3 vols. Oxford; Bristol: Pali Text Society.

D’Alwis, James. 1863. An introduction to Kaccāyana’s grammar of the Pali language. Colombo: Williams and Norgate.

Deokar, Mahesh A. 2008. Technical terms and technique of the Pali and Sanskrit grammars. Central Institute of Higher Tibetan Studies, Miscellaneous Series 23. Sarnath, Varanasi: Central Institute of Higher Tibetan Studies.

Deshpande, Madhav M. 1984. “Review of Kiparsky and Joshi 1979.” Language 60.1: 161–64.

Devasthali, G. V. 1983. “Pāṇini and Vedic: a critique.” Annals of the Bhandarkar Oriental Research Institute 64.1: 137–48.

Dharmānanda, Nayaka Sthavira. 1931. Ṁoggallānapañcikā with suttavutti. Colombo: Satya Samuccaya Press.

Dhammasami, Khammai. 2004. “Between idealism and pragmatism: a study of monastic education in Burma and Thailand from the seventeenth century to the present.” PhD. dissertation. Oxford: Faculty of Oriental Studies, University of Oxford.

Gornall, Alastair. 2020. Rewriting Buddhism: Pali literature and monastic reform in Sri Lanka, 1157–1270. London: University College London Press.

Hinüber, O. von. 1983. Notes on the Pāli tradition in Burma: Beiträge zur Überlieferungsgeschichte des Buddhismus in Birma, I. Nachrichten der Akademie der Wissenschaften in Göttingen, I. Philologisch-historische Klasse 3, 68.

Joshi, Shivaram Dattatray and Jouthe Anthon Fokko Roodbergen, eds. and trans. 1986. Patañjali’s Vyākaraṇa-Mahābhāṣya; Paspaśāhnika: introduction, text, translation and notes. Publications of the Centre of Advanced Study in Sanskrit. Class C 15. Pune: University of Poona.

Kielhorn, Lorenz Franz, ed. 1962–1972. The Vyākaraṇa-Mahābhāṣya of Patañjali: revised and furnished with additional readings, references, and select critical notes, rev. by Kashinath Vasudev Abhyankar. 3rd ed. 3 vols. Pune: Bhandarkar Oriental Research Institute.

Kiparsky, Paul. 1979. Pāṇini as a Variationist. Current Studies in Linguistics 7; Publications of the Centre of Advanced Study in Sanskrit B 6. Cambridge; Pune: MIT Press; Centre of Advanced Studies in Sanskrit, University of Poona.

Laṅḥ, Charā Mauṅ, ed. 1914. Mukhamattadīpanī-purāṇa-ṭīkā alias śaṁpyaṅ-ṭīkā-pāṭh. Yangon: Kavi Myat Hman Press.

Malai, P. Th. 1997. “Kaccāyana-vyākaraṇa: a critical study.” Ph.D. dissertation. Department of Sanskrit and Prakrit Languages, University of Pune.

Nandisena, Bhikkhu. 2005. Kaccāyanabyākaraṇaṁ: translated into English by U. Nandisena. Yangon.

Palsule, Gajanan Balkrishna. 1982. “Review of Kiparsky (1982).” Annals of the Bhandarkar Oriental Research Institute 63.1: 340–42.

Pind, Ole H. 1996. “Saddavimala 12.1-11 and its Mūlasārvāstivādin origin.” Saddavimala: la pureté par les mots, ed. by F. Bizot and F. Lagirarde, pp. 67–72. Paris: École française d’extrême-orient.

—. 2012. “Pāli grammar and grammarians from Buddhaghosa to Vajirabuddhi: a survey.” Journal of the Pāli Text Society 31: 57–124.

—. 2013. Kaccāyana and Kaccāyanavutti. Bristol: Pali Text Society.

Rhys Davids, Thomas William and A. F. Caroloine, eds. 1975. The Visuddhimagga of Buddhaghosa. London: Pali Text Society.

Rhys Davids, Thomas William and Wilhelm Stede. 1921–1925. Pali-English Dictionary. London: Pali Text Society.

Ruiz-Falqués, Aleix. Guṇasāgara’s Mukhamattasāra and the early Pāli scholarship of Burma. Pune Indological Series. Pune: Department of Pali, Savitribai Phule Pune University; printed by Aditya Prakashan, New Delhi. Forthcoming.

—. 2015. “A firefly in the bamboo reed: the Suttaniddesa of Saddhammajotipāla and the grammatical foundations of Theravāda Buddhism in Burma.” Ph.D. dissertation. Cambridge: University of Cambridge.

—. 2018. “Review of Thitzana (2016).” Journal of the Ñāṇasaṁvara Centre for Buddhist Studies 1: 279–304.

—. 2019. “Purifying the Pātimokkha: Pali grammar and Buddhist law in 17th-century Haṁsāvatī.” Buddhism, Law and Society 4: 93–128.

Senart, Émile. 1871. “Kaccāyana et la littérature grammaticale du Pāli.” Journal Asiatique VI Série. T. XVII: 193–351, 361–540.

Smith, Helmer. 1928. Saddanīti: la grammaire palie d’Aggavaṁsa; vol. 1, Padamālā. Lund: Glerup. [Reprint: Oxford: Pali Text Society, 2001.]

—. 1930. Saddanīti: la grammaire palie d’Aggavaṁsa; vol. 3, Suttamālā. Lund: Glerup. [Reprint: Oxford: Pali Text Society, 2001.]

—. 1949. Saddanīti: la grammaire palie d’Aggavaṁsa; vol. 4, Tables 1e partie: textes cités, racines, morphèmes, système grammatical et métrique. Lund: Glerup. [Reprint: Oxford: Pali Text Society, 2001.]

Smith, John D. 1982. “Review of Kiparsky (1982).” Bulletin of the School of Oriental and African Studies 45.1: 185–86. University of London.

Thitzana, Ashin U. 2016. Kaccāyana Pāli Grammar: translated into English with additional notes, simple explanations and tables. Vol. 2. Onalaska: Pariyatti Press.

Trenckner, Viggo, ed. 1888. Majjhima-Nikāya. Vol. 1. London: Pali Text Society.

Vijitāvī, Mahā. 1916. Kaccāyana-vaṇṇanā-pāṭh. Yangon: Zabu Meit Swe Press.

Vimalabuddhi. 1933. Nyāsa-pāṭh: also known as Mukhamattadīpanī. Rangoon: Sudhammavati Press.

See Senart (1871: 24, 25); Senart (1871: 43) notes that the Rūpasiddhi (Rūp.) provides a vā missing in Kacc. 72; (1871: 60) for vā in KaccV. regarding plural forms ending in ā or e; (1871: 69) notes the absence of vā when needed; (1871: 77) for navā; (1871: 87) ca for vā; (1871: 93) wrong use of vā; (1871: 142) vā in Kacc. 281 regarding the karaṇa-kāraka; (1871: 144, 152); (1871: 155) “kvaci could therefore have a double use, but there is nothing simpler than taking it as more or less equivalent to vā” (my translation); (1871: 259, 293). ↩
For a review of Thitzana’s work, see Ruiz-Falqués 2018. ↩
Note that in his critical edition Pind (2013) reads na vā. ↩
Perhaps these rules had the same function originally, but they became redundant once the different independent kappas ‘chapters’ of Kacc. were compiled in one single work. See Pind 1996: 71. ↩
Cf. KaccSNidd. on Kacc. 64 Tassā vā: anukaraṇasuttaṁ hi duvidhaṁ paccakkhāpaccakkhavasena “an imitation-sutta is indeed of two types: based on evidence and not based on evidence” (KaccSNidd. B^e 34.25; C^e 32.8–9). The word anukaraṇa means ‘imitation’ with regard to usage, especially in the Tipiṭaka. Saddhammajotipāla compares the language of the Tipiṭaka with a face reflected in a mirror that is vyākaraṇa, cf. Ruiz-Falqués 2015: 5. Saddhammajotipāla further states: yaṁ pakataṁ tad anukaraṇan “an imitation is that which is natural (pakataṁ, Skt. prakṛtaṁ)”, or perhaps “what is imitated [by grammar] is that which is natural” (KaccSNidd. B^e 32.2.3; C^e 30.1.2). ↩
Cf. A. 2.4.12 vibhāṣā vṛkṣamṛgatṛṇadhānyavyañjanapaśuśakunyaśvavaḍavapurvāparādharottarāṇām. ↩
Pind cites the post-canonical Visuddhimagga (cf. Vism. 651.26). ↩
I could not count exactly how many sabbā masc./neut.abl.sg. there are, because sabbā appears many times as fem. ↩
The other mahāsutta, apart from Kacc. 305, 306 and 519, is Kacc. 393 Yadanupapannā nipātanā sijjhanti. I thank Ven. Kondaññakitti, PhD. cand. at Shan State Buddhist University, for providing this information to me via private communication. ↩
RūpṬ. 17.6; 17.19; 21.5; 26.13; 27.11; 33.22; 35.5; 140.1, see also Ruiz-Falqués 2021. ↩
Mmd. 39.18–21. ↩
Mmd. 38.11. ↩
RūpṬ. 21.5–16. ↩
For a detailed explanation of optionality in Buddhappiya’s Rūpasiddhi, see Ruiz-Falqués 2021. ↩
MmdSāra. 94 = KaccSNidd. B^e 154.10–11; C^e 150.13–14. The Ṁukhamattasāra has not been edited. I am currently working on the editio princeps (Ruiz-Falqués n.d.). Apart from this stanza, KaccSNidd. quotes the MmdSāra. many more times, often without attribution. ↩
For instance, in a very interesting discussion that I have left out, MmdPṬ. 81.1.2 classifies kvaci into two types: vikārasādhaka ‘the means for a modification’ and nisedhasādhaka ‘the means for a negation’. ↩
vā in Kacc. 13 opens a sub-section where sare is blocked; kvaci in Kacc. 14 opens a sub-sub-section where vā is blocked; KaccV. §17 recovers sare but retains kvaci from Kacc. 14. The reason for this is unclear, but sare pare in KaccV. §21 is supposedly anuvutti from Kacc. 12 Sarā sare lopaṁ. ↩
Read with Nyāsa-nissaya I.183.8. Mmd. 33.15, vattabbam atthaṁ, does not make sense. ↩
I also follow the Nissaya (ibid.) for the meaning of attanā = kvacisaddena. ↩
This remark resonates with Patañjali’s paribhāṣā: vyākhyānato viśeṣapratipattiḥ; na hi sandehāt alakṣaṇam “[t]he understanding of something particular (among two or more possibilities should be decided) on the basis of vyākhyāna ‘reasoned explanation’. For (a rule should) not (be regarded as) a bad rule (simply) because of doubt (regarding its meaning).” (MBh. I.6.26; Joshi and Roodbergen 1986: 21, 96 ¶68). ↩
I follow the interpretation of Nyāsa-nissaya 187.27. ↩
MmdPṬ. 73.30-74.3: Tattha asati hi imasmiṁ sarasandhivisaye ti imasmiṁ sutte asatī ti yojanā. Siddhe saty ārambho vidhīyate ti siddhe sati adhippetatthe puna suttārambho vidhīyate karīyate. Another possible reading could be siddhe ’sati (= asati). ↩
Read always paṭisaṇṭhāra. Vimalabuddhi also reads vuty assa = Pind 2013; not vutty assa Dhmp. E^e. ↩
It is by no means clear what Vimalabuddhi is intending here, because in theory there is no longer any consonant (byañjana). My understanding of the derivation is tentative: (0) saṁhitā ‘connected text’: paṭisaṇṭhāravuttiāssa (1) chetvā ‘analysing’: paṭisaṇṭhāravutti-āssa (2) viyojetvā ‘separating’ [Kacc. 10]: paṭisaṇṭhāravutti āssa (3) lopaṁ katvā ‘eliding’ [Kacc. 12]: paṭisaṇṭhāravutt0 āssa [i → 0] (4) yakāraṁ katvā ‘insertion’ [Kacc. 21]: paṭisaṇṭhāravutty āssa [0 → y]. ↩
Read navāggahaṇena akatvā. ↩