Wednesday, October 27, 2010

Nominalism reconsidered

A while back I declared myself a "nominalist" when it comes to linguistic theory, as opposed to a realist.

Well, hold on a minute. It's not entirely clear what I meant, because it's not entirely clear what the terms mean. In the philosophy of mathematics, "nominalism" generally means the stance that "there's no such thing as an abstract thing." [E.g. the Oxford Handbook of Philosophy of Mathematics and Logic] So on this view, not even a number really "exists," whatever that means. I think it is a little extreme to claim that there are no abstract things in this world. I'm not sure why I think that, but I do.

In the philosophy of linguistics (one of the most starved for literature), "nominalism" would have to mean something else. But it's complicated to figure out exactly what. I'll take a stab by saying that in linguistics, nominalism is a stance that says "elements of linguistic theory are not necessarily literally elements of the cognitive structure of language, although they might be."

This is quite a bit weaker than mathematical nominalism, whatever it all means.


  1. I was re-reading the introduction to the HPSG book, where they take more or less this stance.

    I think if you make stronger ontological claims -- like the "psychological reality" claims of Chomskyan linguists, then you do need some extra evidence, but I think we should nonetheless aim at that stronger position as our ultimate goal, while being modest about the degree of certainty that we can achieve.

  2. Much is said for and against "psychological reality" claims, but I'm not really sure who claims what in this respect. Suppose I'm a practicing syntactician who uses minimalism. What is supposed to be literally in the cognitive architecture of language? The categories? The syntactic features? The trees? How about Merge and Move? Are these mental operations under some view? Shall we do an experiment with probes in the brain where we try to catch subjects in the act of Merging? It starts to seem comical at a certain point.

  3. This seems like a completely different notion. Wouldn't it be clearer (and given standard philosophical nomenclature more appropriate) to leave "nominalism" to mean the denial of the existence of abstract objects, and call the view you talk about "instrumentalism" or something alike?

  4. Well I started to use the terms "nominalism" and "realism" as they seem to be used in the discussion of phonology in Anderson's book. I since found out that he wasn't using the term "nominalism" in a standard mathematical way, so now yes I believe it would be better to recalibrate the terminology to make progress with the discussion specific to the philosophy of linguistics. There is no point in "denying abstract objects" within the philosophy of linguistic theory, because nobody really pins anything to their existence. In linguistic theory one has theoretical constructs and classes, which can function like abstract objects in mathematics I suppose. The quarrel is about whether a linguistic theory should assert that its constructs are "real" in the colloquial sense of that term---not that they exist as abstract things (no linguist cares about that) but that they exist as concrete things. The supposed concrete objects behind linguistic constructs are usually suggested to be "in the mind" in a literal fashion. It is a further philosophical problem to discuss whether mentally real things are concrete things.

  5. It seems obvious that there must be some structures in the brain/mind that allow us to use language; and in the process of using language there must again be changes in brain state that reflect what the person listening (or speaking) has done so far. So there must be some cognitively "real" objects that correspond to grammars and structural descriptions of sentences.

    So it seems bizarre not to take these indubitably real entities as the object of study of linguistics.

    I don't know what abstract and concrete mean in this context.

  6. I suppose that yes, from a certain point of view, one could insist that linguists actually study cognitive (and thus concrete) objects. But I believe that our ability to scientifically study cognitive objects is sorely limited. We just simply don't really know what they are, only that they are in there somewhere. A linguistic theory of some kind, however, may succeed to an extent at modeling or describing facets of language. That's nice, but this alone does not give us the right to then claim that the "objects" manipulated by the theory, which are possibly abstract from the perspective of this theory (I mean, it probably doesn't matter for the theory whether various elements are real in any sense at all), correspond in actuality to real cognitive objects. The mere "success" of a theory as a model of language does not guarantee this sort of one-to-one correspondence between theoretical objects and systems and cognitive objects and systems. But most stances of "linguistic realism" are making this leap of faith, which is why I don't subscribe to this line of thinking. I will be happy enough with a linguistic theory that succeeds purely as a formal model. What's really in the cognitive systems is another question entirely, which I do not see myself as normally addressing in linguistic research.

  7. Yes, absolutely -- I completely share your scepticism. But if we think of linguistics as a science, we have to be able to answer the question:

    (Say you have some theory T, maybe some classic P 'n' P model.)

    What would it mean if T were actually true on its own terms?

    If on the other hand you don't view linguistics as a science, and all you aim to do is to have a model of the set of grammatical sentences or something, and you just want this to be succinct or elegant or satisfy some other methodological criteria, then we don't need to think about what this means; but then picking your theory is like picking a programming language.

    That said -- as you rightly point out -- we don't have the ability to perform the critical experiments to distinguish between competing theories, but at the moment there are no theories that meet minimal levels of adequacy so we can be quite sure that all current theories are wrong, so the problem is moot.

  8. Well, we do have a dozen or so syntactic frameworks that can generate mildly context-sensitive languages. That's sort of a minimal adequacy condition isn't it? But of course I'm sort of kidding. I just wonder what "minimal levels of adequacy" are we really missing in your view.

    Nice paper in Machine Learning, by the way Alex. I found it today.

  9. I think any plausible theory would have produced

    a) explicit generative grammars for several if not many languages,


    b) a mathematically precise explanation of how they might be learned from the primary linguistic data.

    As far as I am aware, no one has produced an explicit generative grammar for any language, not even English. There are GPSG style fragments, and I guess there are LFG and HPSG fragments as well, but these are obviously hugely incomplete. And as for learning these ...

  10. Dear Sean,

    I wouldn't swear that nobody in linguistics is committed to the reality of abstracts objects. At least one major example comes to mind: Paul Postal as recently as 2009 (Biolinguistics 3.1) argued for a sort of Platonism in linguistics.

    After all, it seems that we still need the classic, tripartite classification adopted by Jerrold Katz in his old "Philosophy of Linguistics" reader: nominalism, conceptualism (AKA psychological realism), realism proper (Platonism).

  11. Oh goody, a reference. Thanks for that, I'll have to check it out. It sounds like what I started calling nominalism in linguistics is actually just the denial of conceptualism, which seems weaker than denying abstract objects.