In her recent post on theories of sentence meaning in Prābhākara Mīmāṃsā, Elisa talked about a possible “contemporary Prābhākara” viewpoint about sentence meaning:
…a contemporary Prābhākara might suggest that some preliminary understanding of word-meaning is immediately denoted by each word, but that each new word adjusts the meaning of the previous one through the above mentioned criteria in a hermeneutic circle. This solution is not explicitly discussed, at least in the texts I am aware of, possibly because it implies a preliminary (and therefore epistemologically unsound) step within linguistic communication and could have therefore jeopardised the role of linguistic communication as an instrument of knowledge.
She suggests this approach because Prābhākara thinkers appeal to three factors which generate sentence meaning: proximity (sannidhi), semantic fitness (yogyatā), and expectancy (ākāṅkṣā). But there seems to be something puzzling about their appeal to these factors, since on the anvitābhidhāna view we hear words and from that understand meanings as connected (anvita). (This, of course, is in contrast to the abhihitānvaya view, in which each word first denotes its natural or primary meaning in isolation.) It’s puzzling because semantic fitness doesn’t apply to words, but to their meanings. For instance, there is semantic fitness between the meanings of “sprinkle” and “water” in the sentence, “She sprinkles the garden with water,” but not between the meanings of “sprinkle” and “fire” in “She sprinkles the garden with fire.” However, on the anvitābhidhāna view, it seems like we must evaluate semantic fitness directly with regard to words themselves.
Elisa suggests that Śālikanātha’s solution involves some intermediary stage between words and meanings—that of remembering word meanings before the words give their denotation. In comments at her blog, Elisa wonders if the distinction between these two things (remembering and denoting, or smār- and abhidhā-) may not be a lexical technicality. In some comments on that post at her blog, I remarked that this seems analogous in some ways to François Recanati’s approach. And in this approach, there is more than a lexical technicality involved. I thought I’d expand that idea a little bit here, rather than relegate the remarks to a comment thread. The basic idea is that what is remembered is not the kind of thing which can contribute meaning to a sentence, although it is a form of linguistic meaning. In Recanati’s terms, it is the “wrong format.”
Recanati’s most recent book focusing on his theory of meaning (with the same name) is Truth Conditional Pragmatics (TCP). However, his own view is not merely TCP but “radical contextualism” or “the wrong format view” (on which more below), which he presents in Literal Meaning and other articles. (He thinks the latter view entails the former, but not the other way around.) In what follows, I focus on Recanati’s wrong format radical contextualism, although what he has to say about TCP is also interesting. Let’s take an example sentence:
(1) My horse is in the backyard.
Suppose this is uttered by a child at her home in the country. Here are two plausible ways of making explicit what the child is getting at:
(1a) My toy horse is in the backyard.
(1b) My real horse is in the backyard.
On either disambiguation, the utterance could be used for different purposes.
(2a) On being asked if she wants to bring anything with her in the car: “My (toy) horse is in the backyard (so let’s go get it).”
(2b) On being asked what she wants to do this afternoon: “My (real) horse is in the backyard (so let’s go ride it).”
The parenthetical elements are the contents which are apparently not derived compositionally from the sentence. For this example, let’s assume that we know she has a plastic toy horse that she keeps in the backyard, so she is intending (1a) and (2a).
On the radical contextualist view, we must understand the contextual sense of “horse” before we have understood the entire sentence and evaluated it for its truthfulness. For Recanati, what the child has said is a matter of the speech act in context. On just about all views, the meaning of word “my” will necessarily involve some contextual input, as it is an indexical. It has some covert variable which requires contextual input about the utterance’s speaker. However, for Recanati, “horse” also needs contextual input. This is to evaluate whether (1) is true, we need to understand whether it is a toy horse or real horse or etc. that she is talking about. We can distinguish between three kinds of “word meanings” here:
- The lexical meaning. This is whatever we can, through an inferential theorizing process, abstract as being in common with all uses (senses 1& 2 below) of “horse.” It is a schema or a network of many senses.
- The conventional/protoypical sense. This is the contextual sense which is default for our uses of a word. For “horse,” it might be a real biological horse.
- The modulated sense. This is the contextual sense which is loosened or narrowed through processes which are semantically speaking “optional” (on Gricean views), like free enrichment or modulation. (So, contrast “my” which requires a speaker to understand its sense.)
What’s crucial is that for Recanati, in a sentence “horse” never means its lexical meaning, as he argues that this is the wrong format for a contributor to sentence meaning, as it is a schematic network of different senses, recalled to our memory. Rather, we need some contextually suitable sense of “horse.” What we must do is interpret the utterance in the specific context at hand, sometimes backtracking in our comprehension before we have completely understood the entire sentence. Our idea of lexical meaning is something that we abstract from senses–which are the ways we use “horse” in context. (That the child wants to go get her toy horse is for Recanati, since it is not part of the intuitive truth-conditions of (1), a further implicature.)
You might think that Recanati is an eliminativist about word meanings—and he does in fact talk about meaning eliminativism as one form of contextualism in Literal Meaning. Since on his view above, “horse” does not contribute anything context-invariant to each sentence, it seems the word has no meaning. But he would reply, I take it, that the wrong format view does not deny that words contribute something stable, a kind of meaning, across different contexts. It merely claims that what they contribute does not, without further contextual adjustment, contribute to the content of the sentence.
As for memory, it plays an important role for Recanati, as it does for Śālikanātha. Take his account (Recanati 2017) of the polysemous term “swallow” which now gets conventionally applied to ATMs (they “swallow” your debit card). At one time, “swallow” was used in a creative way for a debit card at an ATM, drawing on some prototypical relationship between the way ordinary swallowing works and the debit card disappearing into the cash machine. So the lexical meaning of “swallow” was the input for further modulation in a sentence like:
(3) The ATM swallowed my card.
Over time, that sense of “swallow” has become conventionalized. For Recanati, that means that when I utter (3) your memory of the way this word is used may include cases of ATMs. Thus the lexical meaning now includes, as part of its “network of senses” the biological sense along with the ATM sense. Still, this lexical meaning is not the right kind of thing to contribute to sentence meaning. If we do not modulate it (by loosening or narrowing), we draw on the prototypical sense, which here is the biological sense. In this way, there is (a) lexical meaning, which never contributes to sentence meaning, (b) prototypical meaning, which may contribute to sentence meaning as a default, and (c) modulated meaning, which is an extension of the prototypical meaning.
Enough of Recanati. My suggestion is just that in considering what a “contemporary Prābhākara” might say about sentence meaning, we want to also consider what sort of thing Śālikanātha is committing to as the content of our memory, in that first step before we understand the contextually appropriate sense of the word. Is the smāraṇa for “horse” something like lexical meaning, an abstracted network of senses from all of our previous uses of the word? A prototypical meaning, a default, if not “primary” but immediately accessible and/or often used sense of the word? Or something else?
François Recanati (2017). “Contextualism and Polysemy,” dialectica 71(3): 379–397.