The Science of Language

by Noam Chomsky

  NC: No, we don't accept it consciously; but of course consciously we don't accept the fact that motion requires contact. On the other hand, our commonsense intuitions tell us that that's nonsense. So we're kind of living with two worlds, the world of our commonsense intuition, and some other world that we've been able to construct, in which we have a different conception of rationality. And it's a hard one to hang onto. We probably fail all the time in ordinary life.

  JM: What about authority? Are we somehow set up in such a way that it is going to be something to which we will respond with obedience?

  NC: We can make up stories in both directions. We don't know. There haven't been human societies in which there isn't some kind of authority. And among other primates it's certainly commonly the case – maybe universally the case – that there is some authority structure. It could be. You could make up a reason for that too. We all grow up in an authoritarian atmosphere by definition; we couldn't survive otherwise. Maybe there's something built-in about submission to authority. But you can make up the opposite story too. Children grow up in that framework, and they reject it. Maybe there's something about us that says that you reject authority. You can have it either way; it's one of the joys of evolutionary psychology. You can have it any way you like.

  JM: But the enlightenment conception of the human being that you seem to hold – that is not a conception of human beings that is programmed to respond to authority . . .

  NC: . . . the opposite; we are somehow free. But that's just a hope.

  JM: Just a hope?

  NC: Yes. We can't say we know it. Rousseau argued, “look at savages, animals struggling for their freedom, if we Europeans can't see that that's a part of our nature, it's just because we're so degraded.” OK, that's a kind of an argument, but not a very convincing one.

  JM: But then you don't hope for some scientific evidence in favor of the view that human beings . . .

  NC: . . . have innately an instinct for freedom? I'd like to. I don't anticipate it.

  JM: So you don't think that there's evidence now?

  NC: Not scientific evidence. There's conflicting evidence from history, experience, anthropology, and so on – but conflicting. If it were to be proven that some people are simply born to be slaves, and they will never be fulfilled and satisfied unless they're slaves – if that were proven, I wouldn't like the conclusion, but I couldn't show that it's false.

  JM: Well, that's consistent.

  NC: We can't say now; we don't know. We only hope that things will turn out a certain way, and we can work to try to achieve it because our moral intuition tells us that the world is better off that way, but we may be pushing it to [against] impenetrable barriers.

  JM: Well, the evidence is far from conclusive, but at least at this stage, there's little in the way of firm evidence that people are born to be slaves, I take it.

  NC: No evidence – no firm evidence at all, other than the fact that societies that have existed have authoritarian structures. There was a time that societies that existed had kings and princes; but that didn't show that that was a necessary form of social organization.

  JM: But there might have been some societies in some environments in the past where authority might have been reasonably easily justified – perhaps just due to the need for survival of a particular group or community.

  NC: Or it might be deeper. Dominance hierarchies in animals probably have selectional value; at least, you can make up reasonable stories about that.

  JM: But they seem also to be – to a certain extent – plastic. I read an article recently about a baboon population. Typical baboon populations have male dominance, with a lot of aggression shown on the part of the dominant male against any possible usurpers of that position. But this studied population of baboons had been located near a garbage dump where apparently the dominant males were the only ones to feed, because they excluded all the other members of that population. There was poison in this food, and the dominant males died out; the females took over. The remaining males did not try to assume dominant roles. It became a far more pacifist society and, interestingly, apparently when male newcomers came into this population, they didn't try to take over either. They adopted the mores of that particular population. It might be environmental; who knows what it is.

  NC: Interesting. It's consistent with what we know about human society. Slavery seemed the natural condition; how could you exist without it? It's pretty much the way wage slavery seems to be the natural condition today.

  JM: Very quickly, then, to return to the social sciences: what do you think of them?

  NC: I like Gandhi on Western civilization: it sounds like a good idea . . . [Both laugh]

  Well, there are rudiments of social science, and some of them are interesting. It's hard to criticize researchers just because they can't go beyond what anybody can now understand. So take out of it what's interesting.

  What does merit criticism is posturing and pretentiousness – the pretense of having some kind of significant science when in fact it's just the superficial trappings of science.

  JM: Well, [calling themselves scientists] serves the interests of some social scientists, of course, because they can act as consultants . . .

  NC: . . . and also it looks better in the academic setting. Physicists are using all these complicated words; we'll use the complicated words. There are all kinds of reasons for it, but . . .

  JM: They introduce ever more sophisticated statistical programs . . .

  NC: . . . and it tends to make the economists look like physicists, and then the political scientists want to look like economists. But you have to show that there's a point to that. Probably there is, for some things. If you want to figure out the effect of changing the interest rate on the purchase of cars, then yes, you probably have some sophisticated models for that. But if you want to figure out how the economy works, it doesn't tell you a lot. It tells you very little about where computers came from.

  JM: So it might yield reasonable descriptions of various interrelationships . . .

  NC: Yes, and even explanations of some things that you probably wouldn't have known. But it's way out on the periphery of the economy.

  Take economics, which is the most advanced of the social sciences. It has a lot of things to say about some topics, but it tells you virtually nothing about the contemporary economy. Its principles – entrepreneurial initiative, consumer choice, markets, and so on – most have marginal relation to core elements of the economy, like, say, computers; they didn't come out of those things. Actually, they came out of the labs that I was working in fifty years ago, all happily paid for by the Pentagon.

  JM: As, indeed, a lot of technological innovations.

  NC: Yes.

  JM: What can one hope for from the social sciences?

  NC: You can hope for insight and understanding. There is work in them that seems to me significant. It sometimes uses fairly sophisticated statistics. Take, say, what Tom Ferguson (1995) calls the “investment theory of politics.” That is a significant thesis, and trying to justify and argue for it takes hard work and regression analyses, good statistics, and so on. And I think that out of it comes evidence that a significant factor – not the whole story, but a significant factor – in the direction of the political system is in fact how groups of investors coalesce to invest to control the state. It's not the answer to everything, but it's a significant social science thesis that took hard work.

  JM: Will the social sciences ever overcome the conceptual scheme of common sense, or people as agents – as thinkers and doers and deciders?

  NC: Maybe not. Maybe in fact that's the right scheme. Maybe they'll show that there's a scientific basis for that. It's not just our intuitive way of doing things, but it's our intuitive way because it's true.

  JM: Because it's true, or it's our only way?

  NC: Maybe our only way. You can't predict the course of science. The kinds of questions where real progress has been made are typi
cally very simple ones. That's part of the reason that physics has made such progress.

  JM: Galileo's inclined planes . . .

  NC: . . . yes: keep it really simple. If a molecule gets too big, give it to the chemists. It's hard to get around that.

  JM: Simplification just doesn't seem to work in the case of human behavior?

  NC: If it does, we haven't found the way. But you never know. Look at linguistics. Fifty years ago it looked as if – as the linguists around then said – there's nothing general you can say about languages, except maybe feature reality. Languages differ in every imaginable way. That's just how things appeared to everyone – also to some of the more technical linguists, such as Martin Joos. And it didn't seem unreasonable; it was what I was learning when I was a student. It looked like that. Now it doesn't look like that.

  JM: But on that way of looking at it, the social sciences might make progress and become serious scientists if they ‘turn inside’ – as linguistics did.

  NC: Well, they can't do better than humans are capable of doing. You have to constantly be thinking of what are the right ways of looking at complex topics, so that you can extricate some things from them that are significant, so that they can be studied in greater depth, leaving the residue of complexity aside. Again, linguistics is the same. So the points of progress have almost nothing to do with why most people are interested in language. I get a flood of letters from all over the world – students and others – asking me for help in giving them ideas for their project on language. They found my name on the internet, or something. And 99 percent of them are topics that are perfectly sensible to be worried about, but there's nothing to say about them. They're mostly questions about sociolinguistics, power and dominance . . . fine questions, but they're not the questions where linguistics has made, or is likely to make, much progress.

  JM: Right. But if the complexity of language arises in cases where we use language, where we . . .

  NC: I'm not sure it's complexity; it's that so far as we know, there's nothing very general we can say about it, other than our commonsense observations which, maybe, can be dressed up a little. Either that's because we don't understand, or because there's nothing to understand.

  JM: But still, a parallel to linguistics would suggest what the social scientist would have to do would be to look to whatever conceptual structures we employ when we make decisions, when we . . .

  NC: It's one way for someone to proceed, but it's not the only way. Take the investment theory of politics. That proceeds without asking why investors coalesce to control the state, or why, when they do so, they act in such a way as to advance certain narrowly conceived interests. It just works within a different framework of factors, leaving these factors out. Those factors are ok; it's the same in . . .

  Look, for example, at most political analysis. Why does the government do so-and-so? Almost all of political analysis tries to explain it in terms of the personal characteristics of the leaders. I don't think that that is very useful, but I have nothing very illuminating that you can look at that will give you much more insight. It's got to be shown.

  JM: You might get insight by doing that, but would you get a science – without looking inside the head?

  NC: You might, because it might turn out that what's going inside in your head is – though a factor – a peripheral factor in the choices you're making. For example, let's take something that we do more or less understand. Take the CEOs of corporations. Here, we kind of understand the institutional framework in which they function. If a CEO doesn't increase profit and market share, he or she's not going to be the CEO of a corporation, for various institutional reasons. By looking at that, we can explain a lot about how businesses operate. We haven't yet asked why he or she does that, but it's enough to know that if he or she doesn't, he or she's not going to be in that position. It's an institutional position that requires that behavior. As for why, who knows? Perhaps childhood made him or her agree to keep the institutional structures he or she keeps to; it might be an interesting question about his or her personal life, but it tells you essentially nothing about the business. So there are plenty of cases where you can say something about what's happening in the world without inquiring into extremely difficult, perhaps impenetrable, questions about why people do things.

  It's the same with studying insects. You can study a bee colony and come to understand a lot about what they're doing – what the waggle dance is, what role the queen plays; you can say an awful lot about these things. In doing so, you're assuming that bees don't have choices, that they're automata. But you're assuming that without any evidence.

  JM: When you're investigating that, what are you investigating?

  NC: You're investigating complex systems at a certain level of abstraction. When you study the corporation, you're not studying how the person [the CEO] got to be six feet tall. He or she did, say, and perhaps that has some marginal role in what he or she does, but . . . Anything you study, you're abstracting away from a mass of complexity that you take to be irrelevant for the purpose at hand.

  That's why scientists do experiments, after all. Why not just take videotapes of what's happening in the world? An experiment is an abstraction: this is the very narrow slice of the world that I want to look at. And it can be one that you create, that didn't exist in nature – as, say, with superconductivity, setting up conditions that – I'm told – never existed in nature. That's what you study, because you think it'll tell you something.

  JM: Would this be like investigating third factor considerations in linguistics?

  NC: Well, if you took a look at that . . . This is something I can talk about, because I'm aware of it. It was pretty obvious to a few of us who were interested in this in the fifties that there were going to be these three factors, and the great battle at the time was to show that the first factor – genetic endowment – actually was a factor. We had to struggle against the belief that everything was the result of generalizations from behavior, and so forth. So there wasn't much talk about the third factor – it might be mentioned, but nothing was done with it. Finally, it got to be more or less accepted – at least among people who had been bothered to address the issue – that yes, there is some innate component, there's a genetic component. The first factor is there, and in fact, the first factor determines what has been called the second factor, experience. Experience is constructed; it's constructed by our genetic endowment on the basis of data. So you get to the point where you grant that there is a genetic component, there is experience that is the result of the way that the genetic component deals with data, and there is that third factor there. And it's got to be there. But it could only be mentioned.

  For a long time, it was implicit at a methodological level – at a kind of ‘best explanation’ level. So if you find that you have written up some rules that are overlapping in their predictive outcomes, you try to think up some other rule system that will not involve redundancy, typically put as and thought of as a methodological point. But at some stage of understanding, it becomes a third factor component. You're saying that, well, we're suggesting that there's a property of the world – not language, maybe not even organisms – that says that efficient computation works in a certain kind of way, whether it's language or organization of distribution of neurons [Cherniak's work], foraging strategies, or whatever. There are just certain laws of nature that are applying, and they apply in such a way as to impose the following structure on systems that meet certain criteria – being accessible to the sensory-motor system, for example. OK, at that point, you're switching a methodological discussion to an empirical discussion, and that's always a step forward, because in the case of methodological discussions we can just appeal to our intuition about what seems to make sense. But when you can turn it into an empirical discussion – here's an economy principle – you can investigate it empirically elsewhere. You can see whether that is the way the world really works: I'll look at something else, the distribution of a
rteries and veins in the body, and see if that meets similar conditions. And you can also hope that you might find a more fundamental theory of efficiency that would give some mathematical substance to the principles that you detect empirically in many parts of the world. And if you can get that far, you can show that it really does apply to, say, eliminating redundant rules in language. OK, then you have a deep explanation, and now in terms of the third factor. That's been hard to do.

  In the early years, it was barely possible to think about it. For one thing, the main battlefield was somewhere else. And secondly, it was hard enough to try to show that there was anything regular about languages – that there was something similar among them. Finally, it got to the point where you had some sort of sense of universals and principles about them that go beyond the complexities of individual languages, but there still remained a fundamental conceptual barrier that no one really understood, and that I think is not much understood now. The guiding framework for linguistic theory . . . let's go back, say, to the 1950s. The basic theories in linguistics, such as they were, were procedural. That's European and American structuralism, which were basically the same in that respect. You had a corpus of data, you apply some procedures; you get units; you get some organization – and that's it. You can maybe believe that for phonemes, although it's tricky. But if you grant that, what are the next bigger units? Well, the next bigger units in terms of hierarchy are morphemes. But morphemes just can't be found by procedures, because procedures will allow you basically to find beads in a string – bigger collections of them, and so on – and morphemes just aren't like that. So the past tense in English isn't a bead on a string, it's more abstract; it fits into the system somehow at a more abstract level by some sort of generative process. So that forces you to take a different point of view, to abandon the procedural approach, and what seemed the natural assumption – or at least, my assumption – was that what universal grammar provides is a format for possible rule systems and a measure that picks the best instantiation of them. Given the data and given the format and given the measure, you can fix on a particular language. Well, that framework made it almost impossible to study the third factor, and the reason is that the format has to be rich and restrictive enough, and highly articulated enough, so that it will yield only a small number of potential grammars, given data. So it's got to be a very complex, language-specific format; and if it's language-specific and rich and highly articulated, the third factor isn't doing much. That looked like an impassible barrier.