September 22, 2006

Ineffability again

Several readers sent in complaints about my treatment of Nick Piombino in a recent post on "Ineffability". The longest and most eloquent complaint, from a philosopher, is given in full past the jump.

I'm a long-time reader of language log, but I've never felt compelled to write in. I love your posts in general, but I thought you were too harsh to Nick Piombino in "Ineffability"; in fact, his sentiment seems closer to the truth than you give him credit for.

I take it that Nick-- and others who have similarly expressed this feeling of ineffibility-- meant that sentences in English, regardless of their complexity, literally cannot (completely) convey much of the contents of our attitudes and the subtleties of our emotions. This is the case in spite of the fact (as linguists are so fond of pointing out) that natural language is recursive and we therefore have an infinite number of propositions we could assert. 'Infinite' does not mean all. Some conceptual distinctions cannot be got at by logical operations on concepts we already have-- e.g. philosophers argue that the concepts of folk psychology or counterfactuals, though supervenient on the physical, cannot be translated into lower-level vocabulary, not because there aren't an infinite number of things to be said in the vocabulary of neuroscience or particle physics, but because that infinitude of thought doesn't cover every cut in possibility space.

Surely there are things we think and feel that cannot be expressed in words. Part of the problem with assigning truth-conditions to attitude ascriptions is that the meaning of the content-clause only roughly maps on to our actual attitude state. This is not because we speak sloppily, but rather because there are distinctions we cannot draw with words (likely because a word for the right concept would be for the most part useless). Granny's folk psychology only draws so many distinctions to begin with, many of them muddled or hopelessly confused. Would this be remedied if we had a thousand times the words in the dictionary, all concerning human psychological states that previously could not be referred to with our existing vocabulary, however combined? Perhaps. But we would need at least that number.

I think some lingusists (and I genuinely don't mean 'some linguists' to be a referential indefinite, picking out you) are afraid to admit that our lexicon + a compositional semantics doesn't get us everything, because this seems like inviting the Whorfians to have a field day. But this need not be so: we can think things we cannot (given our lexicon) say, and we *could* think things we cannot (given our current conceptual repertoire) think now. All Nick seems to be saying, if I read him right, is that it would be very difficult to say most of the things we think, even given a language with a greater vocabulary in all its compositional glory.

I hope this email is clear and not too argumentative. They stop feeding you in philosophy grad school so you'll be more vicious in the ring. I just wanted to defend Nick's side-- what I take to be a common side, perhaps touching, but not stupid.

I agree, mostly.

Certainly my post was hurried and careless. In particular, Nick Piombino said that more words wouldn't solve the problem, and I agreed with him, but I complained in a misleading way about the way he put it. People too often equate a language with its vocabulary, and talk as if you can't express a concept if your language lacks a single word that denotes it. But Piombino didn't say that, and I shouldn't have attributed to him an argument he didn't make.

I also agreed with Jonathan Mayhew's point that it was distracting for Piombino even to bring up the question of vocabulary size, since the expressivity of poetry doesn't increase monotonically with the size of the vocabulary that the poet draws on. I didn't like Jonathan's child:toys::writer:words analogy, but I didn't explain my problems with it in a coherent way.

The philosopher's letter objects that my analogy sentences:concepts::digit-sequences:numbers is a false one, because "there are distinctions we cannot draw with words". On the other hand, there are numbers (e.g. pi) that we cannot express with (finite) digit sequences in a positional number system. My point was not that digit sequences can express all numbers, but that increasing the base of the number system to a larger integer doesn't change its expressivity.

Something similar, though less clear, happens with words. Sometimes adding new words is just a way to save space and time, because the new word is just a convenient short reference to a longer explanation. But other new words are not reducible in this way to definitions in terms of existing words, and how people learn their meaning is more mysterious.

You could try to reconstruct the analogy between numbers and concepts by reference to the fact that given a finite number of axioms, there are an infinite number of mathematical truths that you can't prove. If words are like mathematical axioms rather than like digits in a positional number system, and expressing a concept is like proving a theorem, then vocabulary size is relevant to the problem of ineffability (because adding axioms increases the set of provable truths) but doesn't solve it (because infinitely many unprovable truths will always remain). Perhaps that's what Piombino meant, at some level. Certainly it's a kinder construal, and therefore to be preferred.

One comment about "Granny's folk psychology", and its limited, "muddled or hopelessly confused" distinctions: after a few weeks of reading the oeuvre of Leonard Sax, M.D., Ph.D., and Louann Brizendine, M.D., I find that Granny is looking smarter all the time.

Posted by Mark Liberman at September 22, 2006 07:05 AM