Keiser on metasemantics

There are many conceptions of linguistic meaning. One approach, that I like, assumes that the semantic values we assign to sounds and scribbles function somewhat like the numbers we assign to certain pieces of paper and plastic when we say that they are a "5 pound note" or a "10 pound note": they are a compact summary of the kinds of activities people can perform with the relevant objects. With a 5 pound note you can buy certain kinds of goods. With the sounds 'it is raining' you can inform people that it is raining.

When people like Lewis (1975) spell out this use-based conception of semantics, they generally focus on assertion and information exchange. Roughly, the semantic value assigned to a declarative sentence is identified with the information that is conventionally conveyed by an utterance of the sentence.

Keiser (2022) argues against this emphasis on information exchange. After all, we can do many other things with declarative sentences – we can engage in "manipulation, social bonding, entertainment, harm, aesthetic reverie, formal proof, and ritual ceremony" (p.25). Why assume that these are irrelevant to the determination of meaning?

Fair point. Keiser also offers an alternative. At the heart of our language use, she suggests, lies the activity of directing the hearer's attention to a particular content. The semantic value we assign to a sentence is the content to which speakers conventionally direct the hearer's attention by uttering the sentence.

When I utter, for example, 'snow is white', then I may be doing all sorts of things. I may or may not be conveying information. Perhaps I'm telling a joke, or giving an example. But whatever I do, I am directing your attention to the content that snow is white. That's why the sounds I've uttered mean that snow is white.

The idea is interesting. But I'm not convinced it works. I'm also not convinced that the common emphasis on assertion and information transfer is a serious problem.

Let me begin with this second issue.

True, we can do a lot of things by uttering declarative sentences. But many of these other activities are plausibly dependent on the possibility of using the relevant sentences to convey information. When we manipulate or harm or bond etc. by using declarative sentences, we often do so at least in part through the information we convey.

Sure, when we tell jokes, we don't convey the information expressed by our sentences. But it is hard to envisage a community of language users who only tell jokes. Intuitively, in order to understand a joke, you have to understand what would have to be the case if the sentences that make up the joke were used to convey information. If language was never used to convey information, and only to make jokes, would anyone be able to understand the jokes?

It's easier to imagine a community in which sentences are only used to give commands, as in some of Wittgenstein's examples. This is enough to show that the common focus on assertion is too narrow. In principle. In practice, however, there are no such languages, and we probably wouldn't reach a different semantics for English if we assumed that meaning is determined not just by conventions for assertions but also by conventions for commands.

What we want, in the end, is a mapping from sentences to semantic values that we can plug into a general theory of language use to get an accurate model (at some interesting level of abstraction) of our entire linguistic behaviour. For normal linguistic communities, part of this behaviour involves sharing information. So we might as well start by tailoring our semantic values to this part of the job description, and then see if we need to add or revise anything to fit the other parts.

Now for the second point. Can we replace the conveying-information basis with an attention basis?

Again, the idea is that there is a convention in every linguistic community to utter sentences with the aim of causing the hearer to attend to a certain content. On this basis, we can associate sentences (and individual words, see below) with content.

One obvious worry is that for the whole account to avoid circularity, we now need a way to explain "attending to content p" that doesn't involve linguistic meaning. If, for example, "attending to p" means attending to a sentence whose meaning is p, we're not getting anywhere.

Keiser doesn't have a lot to say on what "attending to p" involves. It is supposed to be some kind of conscious activity in which the content p is somehow before our mind.

On this account, consciousness is a precondition for linguistic meaning. This seems wrong. Couldn't a community of robots use a language, without ever consciously attending to anything? I don't see why not. They could surely convey information, give instructions, and so on.

Also, the goal isn't merely to assign some kind of content to sounds and scribbles. The idea behind the use-based account is that the assignment of semantic values to sounds and scribbles is part of a high-level systematisation of interesting patterns in our behaviour. And surely one interesting thing we do with sentences is convey information. So there must be a way of recovering the information that is conventionally conveyed by an utterance of, say, 'snow is white', from the content of the relevant state of attention.

I don't really know what the content of a state of attention is. I know that the information conveyed by a sentence can be usefully represented as a set of worlds, or as a context change potential. Am I consciously attending to any of these things when I hear 'snow is white'? I don't think so. In general, it isn't clear to me how the content of a state of attention should be understood. Keiser says next to nothing about this, but the whole project seems to hang on it.

Another problem. Are there are any interesting conventional regularities involving attention? Is there really a convention to utter 'snow is white' with an intention to get the hearer to attend to a specific content? I'm not sure. When I utter, say, 'dinner is ready', I intend that my children come to believe that dinner is ready, because I want them to come to dinner. Do I intend that they have a certain conscious experience? Maybe. But it isn't really important to me what exactly that experience is. I don't mind if my utterance causes a different mental image (say) in my older daughter than in my younger daughter, as long as they both understand that dinner is ready.

The more general point here is that attending to content p isn't entailed by conveying the information p. Keiser's idea is to identify a "thin" activity that takes place whenever we use a sentence with semantic value p, whether or not we use it to convey information, and to recover the content p from this thin activity. This might work if the thin activity is defined in terms of the other activities (for example, as their disjunction). But Keiser doesn't want to do this. She wants to give an independent, direct account of the thin activity. I don't think this is possible.

Keiser also has some interesting things to say about word meaning.

Standard use-based accounts concentrate on sentences, since only complete sentences convey information. Individual words may be assigned a derivative meaning, but even if we hold fixed the general rules of compositional grammar, there might be different ways of assigning meanings to words that generate the same meanings for sentences.

Some see this as a problem. Keiser says that denying that individual words have a determinate meaning "flies in the face of a strong pretheoretical idea that the words of our shared language do have meaning, which can be found in a dictionary, disputed over the dinner table [etc]" (p. 30).

At this point, I'm no longer sure whether Keiser is interested in the same project. The use-based project I find appealing associates sounds and scribbles with semantic values in order to systematise certain robust regularities in the relevant community. It is no part of this job description that the semantic values that are assigned to sounds and scribbles are things we would ordinarily call "meanings". In fact, when we ordinarily talk about meaning, we don't seem to be talking about things at all.

Keiser argues that her attention-based proposal solves the problem of word meaning, since we can attend not only to propositions, but also to meanings of individual words. We might, for example, have a "convention of uttering 'Elizabeth Warren' to direct an audience's attention to Elizabeth Warren" (p.31). If so, Elizabeth Warren would be the meaning of 'Elizabeth Warren'.

It's not clear what these word meanings are meant to be good for. Keiser doesn't provide a reason for thinking that they combine with general rules of compositional grammar to determine sentence meaning.

Overall, I'm not entirely sure whether Keiser is trying to offer an upgrade to the use-based account of meaning that I like, or whether she is trying to define a different conception of meaning, with a different job description.

Keiser, Jessica. 2022. “Language Without Information Exchange.” Mind & Language 37 (1): 22–37.
Lewis, David. 1975. “Languages and Language.” In Language, Mind, and Knowledge, edited by Keith Gunderson, VII:3–35. Minnesota Studies in the Philosophy of Science. Minneapolis: University of Minnesota Press.


# on 12 November 2022, 03:38

I thought the point about implicature is that it "direct[s] the hearer's attention to a particular content" of their own thought about the ongoing situation. So one word is sufficient to generate the appropriate (grammatically complete) sentence in the listener's mind.

Add a comment

Please leave these fields blank (spam trap):

No HTML please.
You can edit this comment until 30 minutes after posting.