2 8 Examples of SOI

Having discussed examples of DOI, I will now show, through the following examples, that my solution (i.e., the more specific rule wins) helps deal with cases of SOI. Note that we find very few examples of RE 5 (SOI non-conflict) in Pāṇinian derivations. These cases are neither particularly challenging nor of interest to the tradition. Thus, I will only discuss cases of conflict here. To avoid redundancy, I will refrain from reiterating or proving the existence of conflict in these examples. I will also develop a systematic procedure to identify which rule is more specific. At the end of each example, I will mention the traditional solution.

Type 1 (SOI)

  • unidirectional blocking (RE 1)
    • SOI conflict
  • mutual blocking (RE 2)
    • SOI conflict
  • no blocking (RE 5)
    • SOI non-conflict

My solution: the rule which is more specific (i.e., the exception rule) wins

[[67]]

rāmēbhyaḥ

(1) rāma + bhyas – ‘Rāma’ (masculine), dative plural

(7.3.102, 7.3.103→) rāma + bhyas

7.3.102 supi ca (ato dīrgho yañi): the a at the end of a nominal base is replaced with its long equivalent when followed by a declensional affix starting with yaÑ (i.e., y, v, r l, jh, bh or any nasal).

7.3.103 bahuvacane jhaly et (ataḥ supi): the a at the end of a nominal base is replaced with e when followed by a plural declensional affix starting with jhaL (any non-nasal stop or fricative).

Note that the sets of operands of both rules are exactly the same, namely the final a of a nominal stem.

final a of a nominal stem = operands of 7.3.102 = operands of 7.3.103

However, the sets of contexts of the two rules are different. Neither set is a subset of the other. Instead, the two sets intersect each other.

[[68]]

{other case affixes starting with y, v, r, l, ñ, m, ṅ, ṇ, n, jh, bh- contexts of 7.3.102} ∩ {plural case affixes starting with other sounds of jhaL - contexts of 7.3.103} = {plural case affixes starting with jh or bh}

So how do we decide which rule is ‘more specific?’ Let us develop a procedure that we can use to deal with all examples of SOI.

Each Pāṇinian rule actually represents a collection of one or more sub-rules. For example, consider 7.3.102 which teaches that

a& + yaÑ! 🡪 ā& + yaÑ!
[& = end of nominal stem; ! = beginning of case affix]

[[69]]

7.3.102 represents the following collection of sub-rules:

  1. a& + y!🡪 ā& + y!
  2. a& + v!🡪 ā& + v!
  3. a& + r!🡪 ā& + r!
  4. a& + l!🡪 ā& + l!
  5. a& + ñ!🡪 ā& + ñ!
  6. a& + m!🡪 ā& + m!
  7. a& + ṅ!🡪 ā& + ṅ!
  8. a& +ṇ!🡪 ā& + ṇ!
  9. a& + n!🡪 ā& + n!
  10. a& + jh!🡪 ā& + jh! 11. a& + bh!🡪 ā& + bh!

Pāṇini teaches these 11 sub-rules together in the form of the rule 7.3.102, using his pratyāhāra system, purely for the sake of brevity. Similarly, let us deconstruct 7.3.103 which teaches:

a& + jhaL!# 🡪 e& + jhaL!#

[& = end of nominal stem; ! = beginning of case affix; # = plural]

7.3.103 can be represented by the following collection of sub-rules:

  1. a& + jh! #🡪 e & + jh! # 2. a& + bh!#🡪 e & + bh! #
  2. a& + gh! #🡪 e & + gh! #
  3. a& + ḍh! #🡪 e & + ḍh! #
  4. a& + dh! #🡪 e & + dh! #

…and so on.

Note that two sub-rules from the collection represented by 7.3.102, namely 11 and 2, which I have underlined, look similar to their respective underlined counterparts in the collection represented by 7.3.103. The actual SOI takes place between these two pairs of sub-rules. In fact, when I say that the more specific rule prevails in case of SOI, I mean, the more specific ‘subrule’ prevails.

The other (non-underlined) subrules just happen to be represented by 7.3.102 and 7.3.103 respectively and are actually completely irrelevant to the SOI at hand.

We know that jh is not present at the beginning of any case affix, so we will focus on the sub rules which apply to the final a of nominal stems when they are followed by bh-initial case affixes.

Relevant subrule of 7.3.102 =
11. a& + bh!🡪 ā& + bh!

Relevant subrule of 7.3.103 =
2. a& + bh! #🡪 e & + bh! #

[[70]]

Note that we find an extra # symbol in case of sub-rule 2 under 7.3.103. This # stands for plural. Therefore, we conclude that subrule 2 under 7.3.103 is more specific than sub-rule 11 under 7.3.102 and thus wins. Henceforth, I shall take the liberty to rephrase this as ‘7.3.103 is more specific than 7.3.102 and thus wins’.

I will discuss this detailed procedure for the next example too. But after that, to avoid redundancy, I will present this procedure in an abbreviated form for all examples of SOI in this thesis. I will now present the abbreviated form of the procedure discussed above for the present example.

Let us consider the conditions in which each of the two rules 7.3.102 and 7.3.103 apply. Note that here I draw a distinction between a rule and a condition: a rule can apply in multiple conditions. This clarification is important insofar as the exact conditions in which a rule applies can vary, as I will show below.

7.3.102 applies to:

  • base ending in a + declensional affix starting with bh
  • base ending in a + declensional affix starting with any other sound of yaÑ

7.3.103 applies to:

  • base ending in a + declensional affix starting with bh (plural)
  • base ending in a + declensional affix starting with any other sound of jhaL (plural)

Notice that I write sounds, for example, a, bh, yaÑ, jhaL etc., outside of brackets and all their characteristics such as being a plural affix, being a neuter base etc. inside brackets. I treat sounds and their characteristics as two distinct sources of information. Broadly speaking, I will follow this convention for all examples of SOI discussed throughout this thesis.

In every case of SOI, only one condition per rule is relevant to the conflict. I mark the relevant conditions by writing them in bold fonts, as can be seen above. I will do the same for the rest of the examples. We compare the two and determine which one is more specific.

Here the rule including the condition ‘in the plural’ (bahuvacane) is more specific than the other rule, which has no restriction based on number. So, the rule teaching the operation reserved for the plural, that is rule 7.3.103, wins, leading to the correct form: rāmebhyaḥ.

71

The Mahābhāṣya1 on 7.3.103 and the Kāśikā on 1.4.2 state that both 7.3.102 and 7.3.103 are sāvakāśa: 7.3.102 applies in derivations of forms like vr̥kṣābhyām and plakṣābhyām, and 7.3.103 applies in derivations of forms like vr̥kṣeṣu and plakṣeṣu. As stated before, Kātyāyana teaches that vipratiṣedha takes place between two sāvakāśa rules. Thus, by the traditional interpretation of 1.4.2 vipratiṣedhe paraṁ kāryam, the rule which comes later in the serial order of the Aṣṭādhyāyī, namely 7.3.103, wins.

rāmaudārya

(2) Now let us consider the sandhi between the two words of the compound rāmaudārya ‘Rāma’s generosity’. We will not look at how this compound is formed, confining ourselves to the relevant step of the derivation:

rām (6.1.87, 6.1.88 →) [a + au]dārya

  • 6.1.87 ād guṇaḥ (aci): guṇa (a, e, o) replaces both a and the vowel immediately following it.
  • 6.1.88 vr̥ddhir eci (āt): vr̥ddhi (ā, ai, au) replaces both a and the eC vowel (e, o, ai, au) immediately following it.2

6.1.87 which teaches that

a + aC 🡪 a / e / o

can be rewritten as the following collection of sub-rules:

a + a 🡪 a
a + i 🡪 e
a + u 🡪 o
a + r̥ 🡪 a

[[72]]

a + l̥ 🡪 a
a + e 🡪 e
a + o 🡪 o
a + ai 🡪 e
a + au 🡪 o

6.1.88 which teaches that

a + eC 🡪 ā / ai / au

can be rewritten as the following collection of sub-rules:

a + e 🡪 ai
a + o 🡪 au
a + ai 🡪 ai
a + au 🡪 au

Note that the four underlined subrules under 6.1.87 correspond with the four underlined subrules under 6.1.88 respectively. However, both groups of underlined sub-rules are applicable in exactly the same four conditions, namely a + e, a + o, a + ai and a + au, respectively. In such a case how can we decide which one is more specific? Since we cannot use sub-rules alone to make this decision, we need to look at the rules themselves. Even though 6.1.87 already deals with these four conditions, among other conditions, Pāṇini composed 6.1.88 exclusively to deal with these four conditions. This tells that Pāṇini wants us to apply 6.1.88, and not 6.1.87 in this example.

In the remaining examples I will present only abbreviated versions of this procedure, as follows:

6.1.87:

a + e/ai/o/au
a + any other vowel

6.1.88:

a + e/o/ai/au

[[73]]

Unlike example 1, where one condition was slightly different from the other (by virtue of being marked with the grammatical restriction ‘plural’), in this example, both conditions highlighted in bold are exactly the same i.e., a + e/ai/o/au. In such a case, we go a step further and compare the two rules themselves. 6.1.88 applies only to a + e/ai/o/au whereas 6.1.87 also applies to a + any other vowel. Thus 6.1.88 is more specific and wins the SOI, giving us the correct form: rāmaudārya.

On 6.1.88, the Kāśikā says that 6.1.88 is an apavāda of, and thus wins against, 6.1.87. Even though the tradition does not explicitly define apavāda, I think that the tradition uses the apavāda tool in cases of SOI, when, for example, the conditions in which one rule, here 6.1.88, applies (cf. a + e/ai/o/au), clearly constitute a subset of the conditions in which the other rule, here 6.1.87, applies (cf. a + any vowel). In many such cases, the apavāda rule is taught in the close vicinity of, and often immediately after, the utsarga rule, in the serial order of the Aṣṭādhyāyī. For example, the apavāda sūtra 6.1.88 is taught right after the utsarga sūtra 6.1.87.

tavānandam

(3) Let us look at the sandhi between two padas, i.e., words, tava ‘your’ and ānandam ‘happiness’. This example is similar to example 2. Two rules are simultaneously applicable to a + ā:

tav (6.1.87, 6.1.101→)[a + ā]nandam

6.1.87 ād guṇaḥ (aci): guṇa (a, e, o) replaces both a and the vowel immediately following it.

6.1.101 akaḥ savarṇe dīrghaḥ: a long vowel replaces both aK (a, i, u, r̥, l̥) and the immediately following savarṇa ‘homogeneous’ vowel.

6.1.101:
a + savarṇa
i / u / r̥ / l̥ + savarṇa

6.1.87:
a + savarṇa
a + any other vowel

[[74]]

Here too, like in example 2, the conditions highlighted in bold are exactly the same. So, we have to compare the two rules themselves.

However, here, this too seems difficult. Thus, we have to eliminate those conditions which are completely irrelevant to the SOI at hand, namely ‘i / u / r̥ / l̥ + savarṇa’. It is likely that Pāṇini combined this condition with ‘a + savarṇa’ purely for the sake of brevity. So, we omit it from the comparison.

6.1.101 applies only to cases of ‘a + savarṇa’ whereas 6.1.87 applies to ‘a + any other vowel’ as well. Thus, we conclude that 6.1.101 is more specific and wins, thereby leading to the correct form: tavānandam.3 To the best of my knowledge the tradition does not mention this conflict. I suppose it would use its interpretation of 1.4.2 to resolve it.

tisr̥ṇām

(4) Now let us consider an example similar to example 2. We will derive the genitive plural of the feminine form of tri ‘three’.

[[75]]

(7.1.53, 7.2.99 →) tri + ām

7.1.53 tres trayaḥ (āmi): the base tri is replaced with traya when ām follows.

7.2.99 tricaturoḥ striyāṁ tisr̥catasr̥ (vibhaktau): tri and catur are replaced with tisr̥ and catasr̥ respectively in the feminine when a declensional affix follows.

7.1.53:

tri + ām

7.2.99:

  • tri (feminine) + ām
  • tri (feminine) + any other declensional affix
  • catur (feminine) + any declensional affix

I have written the conditions that are relevant to the conflict in bold. 7.2.99 is applicable only to the feminine tri base, whereas 7.1.53 is applicable to the base in all genders. 7.2.99 is more specific and thus wins, thereby giving us the correct form: tisr̥ṇām.

To the best of my knowledge the tradition does not mention this conflict. I suppose it would use its interpretation of 1.4.2 to resolve it.

SOI-L and SOI-M

Note that in the four examples above, (1) and (4) are similar to each other, and (2) and (3) are similar to each other.

In both (1) and (4), the two conditions (in bold) involved in the SOI are not exactly the same. One operation is conditioned by a grammatical specification (‘plural’ in example 1 and ‘feminine’ in example 4), while the other is not. The operation conditioned by the grammatical specification (which is often morphological) wins.

On the other hand, in the case of examples (2) and (3), the conditions highlighted in bold are exactly the same, and thus we have to go a step further and compare the two rules themselves.

For clarity, let us give names to these two types: we will call examples 1 and 4 SOI-L and examples 2 and 3 SOI-M. The primary definition of SOI-L is that it can be resolved at the first step of comparison: the conditions highlighted in bold are not exactly the same, and so the one which has a specific restriction or marker (e.g., plural) wins. The choice of the winning rule can be made at the first step of comparison itself, i.e., by comparing conditions.

[[76]]

On the other hand, SOI-M cases are defined as those where the conditions highlighted in bold are exactly the same, and so we cannot decide which one is more specific. We need to go a step further and compare the two rules themselves to determine the winning rule.

bhānūdaya

(5) Now let us look at the sandhi-related step of the derivation of the compound bhānūdaya ‘sunrise’:

bhānu + udaya

Here the following two rules are applicable:

6.1.77 iko yaṇ aci: iK (i, u, r̥, l̥) is replaced with yaṆ (y, v, r, l) when aC (any vowel) follows.

6.1.101 akaḥ savarṇe dīrghaḥ (aci): a long vowel replaces both aK (a, i, u, r̥, l̥) and the following savarṇa ‘homogeneous’ vowel.

However, the problem is that they do not have exactly the same operand. Here I use round brackets to indicate the operand of 6.1.77 and square brackets to indicate the operand of 6.1.101:

bhān[(u) + u]daya

The operand of 6.1.77 is inside the operand of 6.1.101. How do we solve such an example? I propose that we treat u + u as the operand of both rules. This means that we have to reanalyse rule 6.1.77: instead of saying that iK is replaced with yaṆ when aC follows, we say that iK + aC is replaced with yaṆ + aC.4

Now that both rules have the same operand, we can choose the rule that is more specific. 6.1.77:

  • u + savarṇa
  • u + any other vowel
  • i / r̥ / l̥ + any vowel

[[77]]

6.1.101

  • u + savarṇa
  • a / i / r̥ / l̥ + savarṇa

The conditions in bold are exactly the same. This is a case of SOI-M. Thus, we need to compare the two rules.

Note that the conditions ‘i / r̥ / l̥ + any vowel’ (under 6.1.77) and ‘a / i / r̥ / l̥ + savarṇa’ (under 6.1.101) are not relevant here because our operand is u + u. So, we won’t take these conditions into account. 6.1.77 applies to cases in which u is followed by any vowel. On the other hand, 6.1.101 applies only to those cases in which u is followed by a savarṇa. 6.1.101 is more specific and thus wins, leading to the correct form: bhānūdaya.

To the best of my knowledge the tradition does not mention this conflict. I suppose it would use its interpretation of 1.4.2 to resolve it.

vanam

(6) vana + sU – ‘forest’ (neuter), nominative singular

vana + (7.1.23, 7.1.24 →) sU

7.1.23 svamor napuṁsakāt (luk): affixes sU and am occurring after a neuter base are replaced with LUK.

7.1.24 ato’m (svamor napuṁsakāt): affixes sU and am occurring after a neuter base ending in a are replaced with am.

7.1.23

  • a (neuter) + sU / am
  • any other sound (neuter) + sU / am

7.1.24

-a (neuter) + sU / am

[[78]]

The conditions in bold are exactly the same. This is a case of SOI-M. Thus, we now compare the rules. Both rules are meant for sU and am affixes added to neuter bases, but 7.1.24 is specifically meant for those cases in which sU and am are preceded by a base ending in a. 7.1.24 is more specific and thus wins, leading to the correct form: vanam.

On 7.1.24, the Kāśikā says that 7.1.24 is an apavāda of, and thus wins against, 7.1.23.

yuṣmabhyam

(7) yuṣmad + bhyas – ‘you’ (any gender), ablative plural

yuṣmad + (7.1.30, 7.1.31→)bhyas

7.1.30 bhyaso bhyam (yuṣmadasmadbhyām): the affix bhyas which occurs after the bases yuṣmad and asmad is replaced with bhyam.

7.1.31 pañcamyā at (yuṣmadasmadbhyām bhyaso): the ablative affix bhyas which occurs after the bases yuṣmad and asmad is replaced with at.

7.1.30

yuṣmad / asmad + bhyas

7.1.31

yuṣmad / asmad + bhyas (ablative)

Note that bhyas is a plural affix used for both dative and ablative forms. 7.1.31 is specifically about the ablative bhyas. This is a case of SOI-L. 7.1.73 is more specific because it mentions the ablative, and thus wins, leading to the correct form: yuṣmat.

On 7.1.31, the Nyāsa says that 7.1.31 is an apavāda of, and thus wins against, 7.1.30.

[[79]]

ēkasmai

(8) eka + Ṅe – ‘one’ (masculine), dative singular

eka + (7.1.13, 7.1.14 →) Ṅe

7.1.13 ṅer yaḥ (ataḥ): the affix Ṅe, when occurring after a base ending in a, is replaced with ya.

7.1.14 sarvanāmnaḥ smai (ṅer yaḥ ataḥ)5: the affix Ṅe, when occurring after a pronominal base ending in a, is replaced with smai.

7.1.13

a + Ṅe

7.1.14

a (pronoun) + Ṅe

This is a case of SOI-L. 7.1.14 concerns only pronominal bases. Thus, it is more specific and wins, leading to the correct form: ekasmai.

To the best of my knowledge the tradition does not mention this conflict. I suppose it would use the apavāda tool to solve it.

harī

(9) hari + au – ‘green’ (masculine) nominative dual

hari + au

The two rules that are applicable here are:

6.1.77 iko yaṇ aci: iK (i, u, r̥, l̥) is replaced with yaṆ (y, v, r, l) when aC (vowel) follows.

6.1.102 prathamayoḥ pūrvasavarṇaḥ (aci akaḥ dīrghaḥ): aK (a, i, u, r̥, l̥) and the following vowel which constitutes the first sound of nominative and accusative affixes, are both replaced with a long vowel which is homogeneous with the sound on the left-hand side.

Note that, here too, like in example 5 of this section, the operand of one rule is inside the operand of another. We overcome this problem just as we did in example 5.

[[80]]

6.1.77

i / u/ r̥ / l̥ + any vowel

6.1.102

a + any vowel (nominative / accusative)

i / u/ r̥ / l̥ + any vowel (nominative / accusative)

This is a case of SOI-L. 6.1.102 is more specific and thus wins, leading to the correct form: harī.

To the best of my knowledge the tradition does not mention this conflict. I suppose it would use its interpretation of 1.4.2 to solve it.

vāriṇi

(10) vāri + Ṅi – ‘water (neuter)’ locative singular

Let us look at the rules that apply:

7.3.116 ṅer ām nadyāmnībhyaḥ
7.3.117 idudbhyām 7.3.118 aut 7.3.119 ac ca gheḥ

Kielhorn6 shows that 7.3.117-7.3.119 together originally constituted one sūtra: idudbhyām aud ac ca gheḥ. Kātyāyana split it into two: idudbhyām and aud ac ca gheh, and Patañjali further split the latter into two: aut and ac ca gheḥ. I accept the original version taught by Pāṇini himself: 7.3.117 idudbhyām aud ac ca gheḥ.

Now, in vāri + Ṅi, two rules are applicable:

(7.3.117 →) vāri (7.1.73 ») [] + (7.3.117 →) Ṅi

7.1.73 iko’ci vibhaktau (num napuṁsakasya): augment nUM is attached to a neuter iK-final (ending in i, u, r̥, l) base when a vowel-initial declensional affix follows.

[[81]]

7.3.117 idudbhyām aut ac ca gheḥ (ṅer): after ghi bases, Ṅi is replaced with au, and the final sound of the base is replaced with a.

Note that 7.3.117 is unusual because it teaches two operations together. And curiously, we can say that the operand of 7.1.73 lies between the two operands of 7.3.117. We cannot treat this as a case of DOI, so we have to treat this as a case of SOI.

7.1.73 applies to:

i/u (neuter) + Ṅi
i/u (neuter) + other vowel initial affixes
r̥/l̥ (neuter) + vowel initial affixes

7.3.117 applies to:

i/u + Ṅi

This is a case of SOI-L and the condition which is marked ‘neuter’ is more specific and thus wins, giving us the correct form vāriṇi.

The tradition uses the vārttika, guṇa-vr̥ddhy-auttva-tr̥jvad-bhāvebhyo num pūrvavipratiṣiddham (vt. 1061 on 7.1.96 striyāṁ ca), to solve this conflict. This vārttika teaches that even though the rule teaching the attachment of the augment nUM (7.1.73) comes before the rule teaching auttva (7.3.117 idudbhyām aud ac ca gheḥ) in the serial order of the Aṣṭādhyāyī, the former wins.

82

  1. Mbh III.340.1-5. ↩︎

  2. Note that both 6.1.87 and 6.1.88 belong to the ekādeśa-adhikāra i.e., the section headed by the sūtra 6.1.84 ekaḥ pūrvaparayoḥ which teaches that both the LHS and the RHS item are replaced with a single substitute. ↩︎

  3. I must admit that my method is able to tackle other examples with greater ease as compared to this one. Here, I am compelled to add an extra step i.e., that of excluding the condition ‘i / u / r̥ / l̥ + savarṇa’ from the comparison. Perhaps we could attach greater value to Pāṇini’s use of the term savarṇa and characterize this SOI as follows:

    6.1.101:

    a + vowel (savarṇa)

    i / u / r̥ / l̥ + vowel (savarṇa)

    6.1.87:

    a + vowel

    I have written the conditions that are relevant to the conflict in bold. 6.1.101 is applicable only when the following vowel is a savarṇa. Thus, it is more specific and wins. ↩︎

  4. Another way of comparing the two rules is to simply compare the RHS item of each. For example, for 6.1.77, the RHS item is aC (any vowel) while for 6.1.101, it is specifically a savarṇa sound. This leads us to the correct conclusion that 6.1.101 is more specific than 6.1.77. ↩︎

  5. The base eka is listed in the sarvādigaṇa, referred to in 1.1.27 sarvādīni sarvanāmāni. ↩︎

  6. See Staal’s ‘A Reader on the Sanskrit Grammarians’ (1972: 115). ↩︎