by Noam Chomsky
JM: [Switching a bit,] what's your view now of the status of LSLT? Is the work you've been doing recently a return to the project you find in LSLT? How do you see the historical relationship between that work and the Minimalist Program?
NC: It's different. LSLT was caught between many conflicting impulses. One was to do a distributional analysis, for methodological reasons, for some entity called “language,” whatever that is; and [that impulse] had in the back[ground] motivation[s] like reducing a corpus to a grammar. Another was the biological framework that was just beginning to be thought about. It was discussed in the Skinner review, which was about the same time – maybe a little later – and in other works that were around. LSLT was sort of caught between them: now drop the first and turn to the second, and [try to] see what you're looking at.
Some of the specific proposals that came along in LSLT – well, it has been possible to resurrect them in different form, such as generalized transformations. On the other hand, as Howard Lasnik pointed out to me recently – and I'd completely forgotten – LSLT does not have in it a rule of movement. That's been a core topic in generative grammar since the sixties – how to formulate the movement rules. I started talking about it in Current Issues in Linguistic Theory; and we were talking about it even before that – around 1960. The idea of trying to formulate the principles of a rule of movement and find the conditions on it – WH-island conditions, and this, that, and the other thing – that's turned into a central topic of the field. There's a volume coming out of papers of a conference on WH-movement and chains. It's a very sophisticated picture of where all this has come to, and one aspect of the rules of movement and the concept of Merge. Internal Merge is an effort to show that it is natural – and indeed, inevitable – for language to have such a process. [Well,] it's not there in LSLT. I'd completely forgotten about it, I'm so used to it by now, but Howard went back and pointed out that in fact what's there are notions of permutation and deletion, which give you the effect of movement, but there's no movement there. So that's a substantial change. And you can see where in the semi-methodological approach of that time, rules of permutation and deletion seemed to make sense, whereas from a more fundamental and principled point of view, they shouldn't exist, and you should only have Merge, which yields movement as a special case.
So it's a matter of constant rethinking, and some old things come back, and some don't. And if they do come back, it's usually in different ways.
14 Chomsky and Nelson Goodman
JM: Can we talk about Nelson Goodman for a while? Your relationship to him as an undergraduate and after is not often documented, while your relationship to Zellig Harris is, in several places, although not always accurately. But as for Goodman: there's not much discussion of what you got from him and what you think is valuable in his work. He was your teacher at Penn. He . . .
NC: We stayed quite close for many years later.
JM: And you must have been his protégé; certainly he must have gone to considerable effort to ensure that you got the position as a Harvard Junior Fellow, which made a very big difference in your life. I know that there are important philosophical differences between you, but there are also respects in which, it seems to me, you owe a debt to him. His conception of constructional systems, for example . . . What about your conception of simplicity – is that in any way owed to Goodman?
NC: My interest in it was certainly stimulated by his work. And you'll see the occasional footnote in his writings where we talked about . . .
I met Goodman when I was about 17 or so. I had never had any background in philosophy and I started taking his graduate courses with people who had a serious background, and he was very accommodating and helpful, and he didn't consider it inappropriate in any way that I didn't know anything. He'd direct me to read things, and the like. He was teaching at the time what became the Structure of Appearance, and the later courses that I took with him were teaching what became Fact, Fiction, and Forecast. What struck me particularly was that, whether you agreed with his conclusions or not, whatever he was doing was pursued with absolute intellectual integrity – and that's unusual, and striking. And it was worth pursuing in detail, even if you didn't agree, because you were witnessing a serious mind at work, taking what he's doing very, very seriously, pursuing the difficulties – trying to find the difficulties, and seeing if he could overcome them – all of it clearly on issues of very considerable significance. And he had a very ambitious project – more so, I think, than my interpretation of it at first. The Structure of Appearance was supposed to be a preliminary study to the Structure of Reality – and, of course, that never came. It got sidetracked into Fact, Fiction, and Forecast.
His nominalism, of course, was also an expression of extreme intellectual integrity. He thought that the concept of a set was incoherent. If things are composed of the same elements, they can't end up being different if you arrange them differently. He wanted to pursue that extremely parsimonious assumption to the limits – anything that didn't fall within it was somehow illegitimate. You can see the merits of that. Actually, my earliest work – which disappeared; it didn't go anywhere – was in fact an effort to give a nominalistic account in his sense of the basic foundations of syntactic structure. There's an article in the Journal of Symbolic Logic about it back then, fifty years ago.
And the other work that I was doing was along the same lines. And yes, the notion of a constructional system and its significance comes straight out of that. Also, the concern for simplicity.
But you have to distinguish two different notions of simplicity. One of them was his interest; I'm interested in it too, but of course focused differently. His notion of simplicity was absolute; he wanted to find a concept of simplicity that's going to hold for everything. He argued, with some plausibility, that the search for simplicity in theory is no different than the search for theory. What you're trying to do is find principles that are true – and their only access to truth is through some form of internal coherence and internal elegance and other properties that he tried to find. Now, there is a different notion of simplicity that is theory-internal. Maybe a different word should be used; sometimes I called it “evaluation.” What you try to do is from within a theoretical framework that you're assuming – maybe you want to change it, but you're working within it – and you want to find the optimal manifestation of it, given particular evidence. That's language acquisition.
JM: That appears clearly in Aspects; does it appear in Logical Structure . . .
NC: It appears in Logical Structure of Syntactic Theory; there's a section in there trying to distinguish notions of simplicity and saying that the simplicity measures . . . actually, it's in The Morphophonemics of Modern Hebrew – 1949 – where there is an internal notion. In fact, most of that is a detailed effort to show that the particular system of grammar that's developed is a relative maximum in terms of simplicity, an internal simplicity measure – meaning that if you change any two rules, it would be less simple. There wouldn't be any way to show that it's an absolute maximum. And a good deal of linguistics since then has been – you can conceive of it as being – concerned with that problem, until it gets to the stage of principles and parameters, around 1980. Every approach to general linguistic theory that I know of that was worked out with any concern or care was looking for essentially a relative fixed maximum within a fixed format. And these notions – though they're not Goodman's notion of simplicity – they are inspired by his search for an absolute notion of simplicity, which was rare. He tried, Pat[rick] Suppes did some work on it, a few others. It's something that runs through all the sciences. Hermann Weyl writes on symmetry; that's a similar kind of conception.
It goes straight back to Galileo. I don't know if Goodman would have seen it this way, but for Galileo, it was a physical point – nature is simple, and it's the task of the scientist to, first of all, discover just what that means, and then to prove it. From the tides to the flight of birds, the goal of the scienti
st is to find that nature is simple; and if you fail, you're wrong. It was a major issue right through – still is – in the history of the sciences. Newton, for example, had an approach to chemistry that was different from Dalton's – they weren't contemporaries, but the Newtonian approach and what became the Daltonian approach were quite different. The Newtonians wanted to reduce chemistry, the phenomena of chemistry – the fields weren't really separated – to fundamental, elegant principles, rather like the principle of gravitation. That was a principled approach to chemistry that a lot of people worked on through the eighteenth century. Dalton simply said, let's assume whatever level of complexity we need in order to account for the phenomenon; that's his atomic theory – anathema to the Newtonians. It was Dalton's approach that proved successful, at least for a long period – although you could argue that contemporary chemistry is returning to a kind of Newtonian approach with the quantum theoretic interpretation of chemistry. In any event, these issues have always been very much alive, and still are. They show up everywhere in the sciences, and Goodman had sufficient integrity not only to notice them, but to try to do something about them.
JM: He almost made it a necessary condition of doing science that one seek this kind of simplicity . . . What was the status of simplicity, so far as he was concerned? Was it a methodological principle?
NC: That's one thing that we discussed for all the years that I knew him, and I never understood it [what his view was]. And it was the same in every other aspect of his work. So take, say, Fact, Fiction, and Forecast. The part of that work that became influential in philosophy was the part that he was less interested in. If you take a look at the book, it's in two parts. The first part is the new riddle of induction; the second part is the solution to it. What he was interested in was the solution to it. I think you can barely find a paper on that in later literature. There are tons of references to grue and green and bleen and blue and this, that, and the other thing. But as far as he was concerned, that was an observation. Even if you could figure out a solution to green and grue, you could find another example just like it. What he was interested in was his solution to the problem – projectibility. And I remember back in 1949 or 1950 – approximately then – discussing with him (because I didn't understand it) how it would possibly work unless you assumed innate principles. So, take a pigeon – or a person, for that matter. It or she sees that emeralds are green, and that gets entrenched, and becomes a projectible predicate, and so forth. But the same person sees that emeralds are grue; so why is that not being entrenched? The only possible answer that I can think of is that green is somehow a part of the nature of the pigeon and the person, and grue isn't. So now we're back to the kind of innatism that he objected to. Quine and Goodman were working together pretty closely, but Quine just cut the Gordian knot at that point. He just assumed that the properties that are going to be projectible are innate; that's his quality space. Goodman didn't want to do that; he wanted an argument for it. The projectibility theory was supposed to be an argument for it, but I don't see how it can possibly work. And the same question arises about simplicity – where is it coming from? Is it a metaphysical fact – the way that, say, Galileo thought? Or is it a cognitive fact – is it something about the way that we look at the world? Well, if it's a cognitive fact, then it's not going to be the way to find out about the truth of the world, but it's going to be a way to discover our best conception of the world, given our cognitive abilities and limitations. He would never have accepted that interpretation; and he also wouldn't have accepted the metaphysical one, I'm sure. And therefore he didn't understand a word I said – and he ended up with neither one nor the other. And that I didn't understand, and still don't. And I can't find any clarification of it in his work, and couldn't find any in our discussions.
JM: I suppose his behaviorism might have played a role. It seems to play a role in his so-called solution to the projectibility problem.
NC: He thought so, but I don't see how. Unless you presuppose some cognitive structure, I don't see any way of distinguishing in the entirety of experience which predicates are going to be projectible, and which not. It won't work just to say that these are the ones that have been used; how do you know which ones can be used?
JM: There are certain parallels there between what he had to say and what Wittgenstein had to say about the “given” – which amounted to “that's just the way we use language.”
NC: Well, I don't really know Wittgenstein in the manner of a true disciple, but one of the reasons I was less impressed with his work was that he evaded the issue. Goodman tried to face it. Why is it “given?” Well, at that point Wittgenstein, so far as I can tell, retreats to saying “We're giving descriptions, not explanations.” It's fine to give descriptions, but to someone looking for answers to “Why?” it's not interesting. Goodman didn't want to do that. He wanted explanations. But it seems to me that the explanations are going to be either cognitive – something to do with our mental structures – or maybe even extra-organic: maybe it has something to do with the ways in which organic creatures can deal with the world, given the laws of physics, or whatever. So either it's going to be cognitive and epistemological. Or it's going to be metaphysical. And I don't see how to be in favor of that, for essentially Kant's reasons – you can't get beyond what we can understand. Or it's [the explanation is] in some kind of methodological domain – which I cannot see how we can ground, except in our cognitive capacities.
JM: Hence, certain aspects, at least, of the science-forming capacity – the cognitive equipment we must have in order to construct a science – must somehow be innate within us. This search for simplicity . . . it's hard to define, of course . . .
NC: Purely personal, but in my own personal development, the ideas [involved in the effort to construct a science of language and, more generally, human nature] kind of grow out of an effort to come to terms with the impossible but exciting project that Goodman was engaged in. And it's a terrific education. The project was a perfect foil, because he was pursuing it with complete integrity and commitment; and it was a very clear project. And if you can respect the project, you can – at least in my own case – you can come to the conclusion that “here's why it can't work.” You are led like me in a different direction which leads to something about universal grammar and, more generally, the universal moral principles, and also to some sort of science-forming capacity – it's [all] just part of our nature.
JM: Kant spoke of [science being guided by] what he called a regulative ideal – I'm not sure what that is, probably a label for something that no one really understands. I take it that your idea that the science-forming capacity is . . .
NC: I prefer Hume's version. He says, look, it's animal instinct. He didn't know where to go from there. But I think that's correct. So his solution to the problem of induction, as distinct from Goodman's, is that it's simply animal instinct. It's our nature; and we can't go beyond that, other than to discover our nature – which leads us into cognitive science. But not for him, of course.
JM: Is there any residue of those early efforts to construct a nominalist program in your work?
NC: I think it's a project for the future. In the work that I've done since The Logical Structure of Linguistic Theory – which just assumes set theory – I would think that in a biolinguistic framework you have to explain what that means. We don't have sets in our heads. So you have to know that when we develop a theory about our thinking, about our computation, internal processing and so on in terms of sets, that it's going to have to be translated into some terms that are neurologically realizable. I don't know how helpful pure nominalism will be, but there is a gap there that the nominalistic enterprise is focused on. It's a gap that has to be overcome. There are a lot of promissory notes there when you talk about a generative grammar as being based on an operation of Merge that forms sets, and so on and so forth. That's something metaphorical, and the metaphor has to be spelled out someday. Whether this is a high prior
ity for today or not, I don't know. But in the 1950s as a student of Goodman's – I was terribly impressed by him, as was everybody that knew him – I was convinced that you had to do it that way. But I came to the conclusion that it's either premature or hopeless, and if we want a productive theory-constructive [effort], we're going to have to relax our stringent criteria and accept things that we know don't make any sense, and hope that some day somebody will make some sense out of them – like sets.
JM: Is there anything you want to add to the discussion of Goodman?
NC: We retained a very close personal relationship, until the point where he realized that I was really serious about talking about innate structures – and for him, it was almost a religious principle that you can't be serious about this. Basically, we had to break relations over it, which was unfortunate.
