________________
New Horizons of Research in Indology
66
in the Aşțădhyay. Having introduced a term into his system by means of rules, Pāṇini does not then proceed to use another, una fined term -- let alone two -- in seemingly random alternation with it, as he apparently does in the case of the words for "optionally". The well known fact however, is that not all saṁjñāsūtrasare definition in the same sense. Consider the follwing: 1.2.27, 1.4.10, 1.4.45: Achtsstraugai: ero agl 3ERTSTECHRUTHI and 1.1.68 94 9TEGRITETEGYİŞTI. All of these are what the tradition calls sam jñasūtra, and they do all indeed concern conventions of terminology. Nevertheless, they do not all serve the same purpose. Sūtras of the first group are what I call classification rules. They serve to assign class names to specified entities, thereby classifying them: vowels with the duration of u ū, u3, respectively, have the names hrasva, dirgha, pluta; a short vowel (hrasva) has the name laghu; a participant that funcitions as locus with respect to an action has the name adhikaraņa (see also 2.1). And Päņini of course uses these terms with reference to short vowels and so on. 1.1.68, on the other hand, serves only to establish a convention, not to assign any entity to a class: a term (svas rūpam 'own form') other than a technical term of grammar (aśabdasamjna) is a sajña for the term itself (sabdasya), not for its meaning. This obviously establishes the convention that non-technical terms are regularly self-referring. Moreover, there are instances where, for particular reasons, Panini deviates from this convention of self-reference, allowing terms like vȚksa to signify trees instead of the term itself.98 Equally clearly, 1.1.44 does not serve to classify anything. This too concerns a covention. In this connection, consider the rule 4.1.82: T FT CRAIGT. In normal Sanskrit usage, va 'or' can be construed with various constitutents of sentences. Accordingly, a native speaker of Sanskrit, confronted with an utterance such as 4.1.82, could interpret this as stating that a taddhita affix follows the first (prathamat) or some other one (anyasmād vā) of related elements given in subsequent rules; he could make a required suppletion to understand the statement in this way. However, this is obviously not the desired interpretation. Instead, one should understand the sutra in the same way that one understands GTI oft al. That is, vă should be construed, as equivalent to na va, with the supplied verb from asti : The rule should provide that a taddhita affix follows the first of specified related elements or does not do so; that is, it provides for the optional introduction of taddhita affixes. One might suggest, then that the purpose of Pāṇini's rule 1.1.44 is precisely to establish such a convention for all possiblc Sanskrit words that have to do with optionality. It would then be understandable why Panini uses vibhasa in this sūtra, as the most general term for option, derived from bhas 'speak' with the preverb vi just as vikalpa derives from krp with this preverb. This also makes understandable that in Astādhyāyī 7.2.15, Pāṇini uses vibhāṣā: As the general term for optionality, this covers also terms like vå. Under this interpretation, of course, we do not have the problems noted above concerning 7.2.15.99 We also do not have to assume for vā or a highly special meaning that is not at all the usual meaning of the particle in Sanskrit of any cpoch.100
5.4 In sum, although Kiparsky's thesis is both intelligently conceived and generally well argued, in the end I find it neither cogently maintained nor acceptable. The commentarorial tradition, which he is so willing to disparage may, in this case, be perfectly right in not having any inkling of the supposed systematic semantic distinction among the three terms for optionality.
6. Recent years have witnessed an increasd interest in questions concerning Panini's syntactic theory and method of deriving Sanskrit sentences and relating them to each other.