Van Fraassen and Natural Language as “the Language of Science”

Bas van Fraassen has been at Western this week. He said a couple of things I found extremely surprising, and I wanted to get help from the Blogosphere about how to understand one of them.

As far as I could make out, one of his main plaints yesterday was that Putnam-style model theoretic arguments cannot work because the language of science already comes interpreted. So, the whole idea that one can “re-interpret” the language of science in a bizarre way, yet preserve truth, is a non-starter. I am happy with the conclusion. But the bit that really surprised me was his grounds for this claim, namely that the language of science is ordinary natural language and it requires no interpretation. He granted that texts in English, German, etc., may not be understood immediately (he mentioned conflicting interpretations of novels), and that we frequently lack an analysis of natural language terms — but he insisted that, nonetheless, the very of idea of “assigning an interpretation” was inappropriate.

I felt sure I had misunderstood, and in the question period I brought up two very familiar points, raised by thinkers as different as Frege, Tarski, Austin/Strawson, Wittgenstein and Chomsky: 1) That sentences of natural language are not typically true or false, even relative to a narrow list of objective parameters. (Think of ‘It’s three o’clock’ or ‘Washington is annoyed at Putin’). 2) That sentences of natural language have use-theoretic properties that fit ill with the needs of the scientist, such as illocutionary force, level of formality/politeness, tone, etc. (Think of the difference between ‘tu’ and ‘vous’, ‘barf’ and ‘vomit’, etc.) In short, natural language lacks some of the crucial “semantic properties” that sentences of a scientific theory would seem to require, and they exhibit “pragmatic properties” that would not be desirable.

Van Fraassen’s own reply seemed to amount to denying out of hand that natural languages are like this. But it’s hard for me to believe that he could be so naive and out of touch, given his own interest in pragmatics. (I do not mean that he made sophisticated moves to set aside the raft of cases, familiar to linguists and philosophers, of lacking-truth and exhibiting-more-than-truth . Folks who do that are, I think, mistaken; but they definitely are not naive. I mean that he just dismissed the idea.)

I came away thinking that there must have been serious cross-talk at work. Thus my question to y’all: As a proponent of the “semantic approach” to scientific theories, what can he have in mind by the idea that, by appealing to ordinary natural language sentences, we solve the problem of providing a proper interpretation for scientific theories?


4 responses

  1. My impression was not that he denied that natural languages “are like this,” but that these features (apparent lack of truth conditions in some contexts, prevalence of pragmatic properties) simply make semantic theorizing tricky. Probably he would deny that NL sentences “typically” are not true or false (but merely that some rare cases are particularly difficult). It seems almost a matter of definition that language is representational in character, that a sentence represents the world as being a certain way, that uttering a sentence is asserting the world to be that way, and so on. That it is difficult to capture this theoretically given the complexity of natural language use cannot imperil the fundamentally representational character of language (any more than the difficulty in describing the physics of cat landings imperils actual cat landings: physics is our problem, not the cat’s.*)

    Given what van Fraasen wrote in “The Empirical Stance,” I can see why he’d want to make such a move: resting firmly in natural language might provide a way to avoid metaphysical entanglement. (Maddy’s “Second Philosophy” sort-of tries to avoid metaphysics similarly.) But it is easy to see that the properties of NL are spectacularly ill-suited to scientific languages, unless you think that NL are, really, at bottom, just a form of logic. But if this is the case, why can’t we push back and say, look, if your view requires that natural languages are just fancy logic, why don’t your objections to model-theoretic and positivist views just come right back and undermine “uninterpreted” NL sentences?


    (*Except in a way it is, since evolution had to figure it out, after all!)

    • Agreed, mostly. But, though you are aware of this, let’s be clear for other readers that “language is representational” may be quasi-definitional, yet “items of natural language typically stand for things, or are true/false tout court” by no means follows from this. For one can mean by the former slogan that language is usable to represent. [As I seem to recall from a long ago dinner, Jason Stanley was once tempted by a transcendental argument for names referring, viz. that there must be a way to talk about objects themselves. Maybe so, but that people can talk about objects per se using language does not get you that there are items of language that stand for objects. That people refer doesn’t entail that words do.]

  2. This is how I understood Bas’ response to the model-theoretic argument: It’s not the job of philosophy of science to explain how theories get interpreted, because theories are stated in natural language, and figuring out how natural language gets to hook up to the world is not the job of philosophy of science. It’s the job of philosophy of mind and language. Philosophy of science can just assume that scientific theories have determinate interpretations, because they’re stated in natural language, and natural language expressions (or specific uses of expressions) are already interpreted. In other words, there’s no *special* problem of scientific representation. I’m not sure if this is exactly what he had in mind, but it is congruent with other things I’ve heard him say in person. And it’s reasonable (though the same worries about interpretation might now be passed on to whoever has the job of explaining how natural language hooks onto the world).

    This doesn’t automatically address the specific concerns you raised in your question. This is how I understood his response to you: I don’t think he flat-out denied that natural languages have the properties you claim they have. What he seemed to be saying was that, in some way or other, natural language items get to be true or false, but it doesn’t matter what exactly those items are, e.g. whether they’re sentences or uses of sentences. As for (2), I don’t think he was at all bothered by it. I can see where all this might be coming from. You might think that since natural language is “good enough” to convey a non-scientific message, then it’s good enough to express a scientific theory. How the details work out (e.g. whether it’s sentences or uses of sentences that are the bearers of truth values) is to be settled by someone else. Again, there’s no special problem of scientific representation.

    • Thanks to both Angela and Nic for their helpful insights.
      Both point to another bit of puzzlement I had. On the view of scientific theories at issue, theories are sets of some sort, whose members stand in logico-semantic relations. The question is, what ontological category do the members belong to? On the “syntactic view”, they are uninterpreted formal items — which raises the issue of unintended interpretations. But what are they on the “semantic view”?
      I would imagine, for the kinds of reasons noted in my original post, that they cannot be either natural language sentence types tout court or an ordered pair consisting of a natural language sentence type and some narrow group of contextual parameters (i.e., what Kaplan calls an occurrence). For the most part, those do not have truth values. (Exceptions might include: ‘Three is smaller than four’ and ‘There have existed black dogs prior to April 5, 2014’, but not ‘Snow is white’ or ‘The earth orbits the sun’.)
      That leaves a variety of options, but none look initially tempting to the Empiricist philosopher of science:
      i) speech acts
      ii) natural language sentence tokens
      iii) natural language sentence types relative to “big context” (i.e., everything about the world, the speech situation, and the participants that in fact fixes truth conditions)
      iv) structured propositions
      v) sets of non-actual possible worlds
      The first three options make the elements of scientific theories into spatio-temporal items, but ones which are deeply infused with human interests and a massive range of worldly facts. The latter two have the right kind of human independence, but are not spatio-temporal in the way Empiricists tend to like.
      I want to distinguish this question, by the way, from the one Angela points to, namely the medium in which actual scientists work when they converse. There is no immediate problem with having (i) be the kind of thing that get exchanged between working scientists. But I suspect — wrongly? — that philosophers of science who want to model the truth of theories in terms of a structured set corresponding to the facts are looking for something more formally tractable.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s