“Words mean what they mean”

The title of this post is a reference to a “common-sense” principle that might be called skhalīya-nyāya, since it is attributed to an Associate Justice of the U.S. Supreme Court (and because it tends to trip people up…) But the content comes from a recent workshop in Vienna with Kiyotaka Yoshimizu (covered in anticipation by Elisa here), whose lectures led more or less directly to this post.

As long as there was “Indian philosophy,” there was a lively debate about what words actually mean. But regardless of the answer—Mīmāṃsakas like Kumārila Bhaṭṭa insisted that the meaning of a word was always, in the first instance, a universal (jāti) rather than a particular (vyakti)—a related question remains open: do the meanings of words remain stable across contexts of use? Mīmāṃsakas, like grammarians, maintained that words had a stable and in fact “originary” connection with their meanings. One might assume that, for Mīmāṃsakas, a given word should mean the same thing in every context.

But this assumption is only half right. The meanings of individual words are stable and timeless universals, but they enter into “discourse” (vyavahāra) or “usage” (prayoga) through particular instantiations of those universals. For a given instance of language use, there is not only the question of what is denoted (abhihita), but also of what is intended (vivakṣita). Although we commonly speak of “intention” in psychological terms, i.e. what a particular person “had in mind” in saying something, Kumārila maintains that it is possible to ask of an impersonal text (i.e. the Veda) whether something is “intended” or not. (See Yoshimizu in ZDMG 158 [2008]: 51-71). There are, in other words, criteria for determining “intention” that do not make reference to psychological states. These criteria refer to hierarchical relationships within discourse, which it is the primary purpose of Mīmāṃsā to explicate, as well as to a number of “discursive functions” that Mīmāṃsakas appear to presuppose.

Kumārila and later Mīmāṃsakas allude to two complementary functions (see especially the end of the grahaikatva section of the Tantravārttika, on sūtras 3.1.13-14). They are uddeśya and vidheya, which can be paraphrased as “that in relation to which something else is enjoined,” and “what which is enjoined in relation to something else.” Before Yoshimizu’s article (in Acta Asiatica 90 [2006]: 5-38), these terms were commonly translated by the quasi-Aristotelian terms “subject” and “predicate.” Yoshimizu suggested that they are closer to the roles of “topic” and “comment” theorized in pragmatics: the “topic” identifies a known or given entity, while the “comment” says something new about it. (If he is right, this post will be filed in the “Euro-American linguistic theory reinvents the wheel” category.)

These functions add another level to the “meaning” of a sentence, which more or less corresponds to the level that modern linguists have called “information structure.” Put otherwise, this additional level further determines the “meaning” of a sentence: once the denotations are available (abhidhāna) and arranged in a teleological hierarchy (viniyoga), the discursive functions allow us to determine if particular linguistic elements are intended or not. The basic principle enunciated by Kumārila is that only elements that are expected (ākāṅkṣita) can be intended (vivakṣita), and since there is no expectation for the various delimiters (such as gender, number, and so on) of the topic (uddeśya), they are considered to be unintended. (Professor Yoshimizu referred to Ṭupṭīkā 1205.24-26 in the workshop: kriyā sādhyā sādhanaṃ tatparicchedakāni ca liṅgasaṃkhyādīn apekṣate, ataḥ sarvaviśeṣaṇaviśiṣṭā kriyaiva vidhīyate.)

I had three general questions. First, I purposely phrased this principle without reference to human agents such as “readers” (or “speakers” or “listeners”). I wonder whether it is possible to speak of “discursive dependencies” in this case, similar to “syntactic dependencies,” which do not necessarily refer to an actual psychological state of expectation or anticipation, but rather to the constraints of information structure (or syntax). Part of what makes this theory so elegant, I think, is that intention so defined is not arbitrary, or equivalent to a speaker’s “gist” (abhipreta), but the correlate of a particular configuration of discursive functions.

Second, what are uddeśyatva and vidheyatva, really? How are they defined? We briefly discussed “givenness” (as pragmaticists say) or prāptatva (as Mīmāṃsakas say) as one of the conditions for uddeśyatva. But it seems that this isn’t the whole story; these functions can’t be reduced to old vs. new information.

Third, the most obvious outcome of this discussion is that words don’t always mean what they mean, even for people who believe that linguistic systems are eternal. A word like graham, which denotes a vessel in the accusative singular, doesn’t always refer to a single vessel (in the grahaikatva section, it refers to each of the vessels in a particular ritual context). An interesting further consequence of this discussion is that certain linguistic elements are present in a sentence not because they are intended, or “motivated,” but as it were by default. (A Mīmāṃsaka might call this anuvāda or “repetition.”) They are present only because their absence would violate well-formedness constraints at the level of morphology (e.g., all nouns need a gender, number, and case), syntax, and perhaps information structure.