Skip to content

Chomsky chomsky chomsky chomsky furiously

Noam Chomsky elicits a lot of emotional reactions. I’ve talked with some linguists who think Chomsky’s been a real roadblock to research in recent decades. Other linguists love Chomsky, but I think they’re the kind of linguists I wouldn’t spend much time talking with. Many people admire Chomsky’s political activism, but sociologist blogger Fabio Rojas distinguishes “the Chomsky’s of the world who sit around and speechify about the man” from the good guys, “the academics whose work leads to tangible improvements.”

When Thomas Basbøll sent me this note,

I [Basbøll] wonder if you react in the same (sympathetic) way to these remarks by Chomsky [text here] as I do. I think he’s right that something happens to research when “applications” come into view. I like his distinction between two conceptions of science, one of which is based on “big data” in which patterns are found by brute information processing, and the other which requires the construction of simple, elegant models that represent the underlying principles that explain what is going on. There’s an important difference between the sort of science that mapped the genome and the sort of science that discovered DNA.

I replied: My linguistics colleagues whom I respect very much think Chomsky is wrong and annoying about many things.

Surrounded by admirers and haters

Chomsky seems to be surrounded mostly by admirers or his haters. The admirers give no useful feedback, and the haters are so clearly against him that he can ignore them. As with others in that situation, Chomsky can then make the convenient choice to ignore the critics who are non-admirers and non-haters. From an intellectual standpoint, those are the people who require the most work to interact with.

Basbøll replied:

I agree with that interpretation of Chomsky’s situation, both in linguistics and politics actually. But I think he’s onto something in this particular case, perhaps not about the state of AI research, or even its prospects, but on the choice that can be made between two different ways of doing science.

The lazy left meets the grasping right

This all made me think of the political aspects of the scholarly criticism that Basbøll and I have been doing in recent months.

The question came up in our discussion of sociologist Karl Weick, who tells poorly-sourced stories to audiences of scholars and business executives: I see some sort of alliance there between the lazy left (who, I think, secretly believe in the inevitability and scientific correctness of right-wing economics and embrace anti-scientific thinking as a way out) and the grasping right (who are always on the lookout for stories that justify inequalities of wealth and power). What to do with this political angle? I don’t know. We can blog on it, but I don’t have any ideas right now about how to think about this more systematically.

P.S. This post really isn’t about linguistics, but some commenters requested that I link to some specific criticisms of Chomsky’s linguistics work, so here’s something from Bob Carpenter and here’s something from Dominik Lukes.


  1. Z says:

    So, what is your opinion of Chomsky’s views on ‘Big Data’ and science?

    • Andrew says:


      I think we can learn from “big data” data collection and analysis, and also from data collection and analysis that is more closely tied to subject-matter models. No need to do just one or the other.

  2. T Ferriss says:

    I believe the posted picture is of the mathematician Noam Elkies, not Chomsky.

    • Andrew says:

      I went with the only Noam I know.

      • David says:

        What??!?! How do you expect people to take you seriously if you can’t google for the correct persons image.

        • Andrew says:


          It’s hard for me to detect intonation from typed speech, but . . . do you really think I can’t google for Noam Chomsky’s image?? I just preferred to go with the image of the Noam I know.

          But you can feel free to not take my work seriously. After all, I did publish a false theorem once.

          • tom says:

            I think this is the right choice. With any google image search (even with all parental controls switched on) you inevitably get some sort of ‘erotic’ content and with a search term like ‘Noam Chomsky’ I’m not sure that I want to have that sort of image inflicted on me.

          • David says:

            I’m incredulous that you would (a) not post the correct picture in the first place and (b) not bother to fix it after it has been pointed out that its incorrect. If its an inside joke between yourself and Elkies then it needs explaining, but its rude to talk about one individuals life’s work and post a picture of someone else. Would you want someone to do that to you?

            It may seem like a rather inconsequential thing for you but it makes me really question your integrity and academic values. You post a negative opinion of Chomsky based on some conversations with colleagues who don’t like his political activity. Fine. I really don’t care one way or the other, and I would not be surprised if Chomsky’s last serious work in linguistics is many years/decades past. But then you include a picture of the wrong man?!?!

            Perhaps you mean to critic academics who you believe spend too much time on work unrelated to their appointed position. Perhaps you see this as an abuse of tenure. Perhaps you mean to make a subtle jab at Elkies and suggest that he spends too much time composing chess problems and is holding back the advancement of Mathematics?

            I rather hope that is what you are trying to say, because the alternative is that you have so little respect for a fellow academic that you can’t spend 5 seconds to get his picture correct. If getting details correct is of so little importance to you perhaps we need to be looking at your spreadsheets to see if you failed to drag some formulas down.

            Edit the post, and put the correct picture up. Its really that simple.

            • Andrew says:


              1. I was relaying negative things people had told me regarding Chomsky’s statements on linguistics, not on his political activity. It’s Fabio Rojas at orgtheory who was disparaging Chomsky’s political activity.

              2. I put up Noam’s picture because I thought it was amusing, not because I can’t spend 5 seconds finding the other Noam’s picture. One of my goals in blogging is to be entertaining, and it can be entertaining for people sometimes to figure out the relevance of the picture to the story. My posting of Noam’s picture was a joke, not a mistake.

              3. And, no, I was not in any way criticizing Noam. I have no problem at all with him composing chess problems, composing math problems, or doing whatever else he’d like to do with his life. As long as he does his job, I don’t see this as an abuse of tenure.

              4. I appreciate that you put in the effort to comment here, but I’d appreciate it even more if you didn’t try to tell me what to do.

              • David says:

                Whatever the nature of your criticism of Chomsky’s work, I’m simply not well positioned to critique it. I’ve taken one undergraduate course in basic linguistics and read a few of Chomsky’s political books. So I have no intention or ability to address the precise nature of the criticism. What I will address are the following:

                1) Hearsay critiques of an academic, which as a general rule I won’t read. If your friends have negative opinions of Chomsky’s academic or political work, or they just think his hair is unkempt, they should publish a blog and you should feel free to link to it.

                2) Something evidently meant as a joke, but with no obvious indication of it being a joke, in a posting that is overwhelming negative. If this were a light-hearted and positive posting talking about how Noam Chomsky has a green thumb and grows his own tomatoes, then it wouldn’t be so bad. If this hypothetical article about Chomsky growing tomatoes had a caption: “Perhaps Chomsky should worry about garden Noams” I might find that rather witty.

                But this is NOT a light-hearted posting, this is negative and hostile. You don’t post negative things about people an include the wrong picture. Did you think this cover from the NY Post was acceptable?:

                You may not be accusing Noam Elkies of murder, but I really don’t get any part of this posting. Being the mouthpiece for anonymous criticism is bad enough, then you have to add a joke in rather poor context. This is all well below what I thought were your standards. Needless to say I won’t bother reading this blog in the future.

                Did you even asked Elkies for permission to include him in this joke? Do you know what he thinks about this posting? Is this to be construed as blanket permission for anyone to include your picture in any random article that happens to mention an “Andrew” whether or not its you? No matter what the content?

              • Andrew says:


                All I can say is, if you think a caption such as “Perhaps Chomsky should worry about garden Noams” is “rather witty,” your sense of humor is so different than mine that I don’t think there’s any possibility of communication. Indeed, you’ll have to find some other blogs to read.

            • Entsophy says:


              You can accuse a Statistician of being lazy, incompetent, umimaginative, useless, lying, rude, wrong, inconsiderate, nefarious, illegitimate, big headed, small headed, smelly, and a pox on mankind, without anyone taking it the wrong way and arousing passions. But you can’t just go around accusing them of using Excel. That sort of talk results in “pistols at dawn”.

              And since humor is no laughing matter for you, let me just clarify that I’m completely serious. It literally results in duels. At dawn. That’s how Alexander Hamilton died (his picture is on the $10 bill in case you don’t know what he looks like)

      • Sam says:

        Reminds me of Donald Rumsfeld. “There are known Noams.”

  3. KD says:

    All I see here is hearsay – not a very good thing to do if you don’t really know the field. I have heard this incoherent comment about “X being an impediment to research” before, especially about Chomsky. I am not sure how this could ever be the case, unless X used some sort of force. He presented ideas and arguments, which a lot of people find/found compelling, so they chose to work with the ideas as guiding lights. Those who complain about X having been an impediment are really showing sour grapes, as their ideas haven’t become as popular. This view is really childish, if you ask me.

    About Chomsky’s view on science – the dichotomy seems completely reasonable. It is also important to ask the question what does one mean by “Big Data”? If by it, one means, a lot of observations in reasonably controlled situations, sure it might be useful. But, if by it, one means, data that is there and was collected with no purpose or direction, then I am not sure how useful it will be beyond perhaps initial hypotheses about where to look (perhaps).

    To clarify the point, could we have done physics better with “big data” in the 16th century? Let’s try it out on gravitation in the real world, and see where it gets us. Before saying it is useful, it would be good to show it on classic cases. If not, Chomsky’s point is very well taken. Now, if the point is, “Big data” has its uses, again, it would be good to show exactly how it improves over the other technique. I can understand it being used for predictions (temporary, at least), but I cannot understand how it can be used for “understanding” the world.

    • expr says:

      Kepler used “big data” to discover his laws of planetary motion

      • KD says:

        This point is irrelevant. Note: “…If by it, one means, a lot of observations in reasonably controlled situations, sure it might be useful…”

        The relevant question is: what exactly does one mean by “big data”?

    • Nathanael says:

      Chomsky has an *extremely* bad reputation among historical linguists.

      It’s really more his followers who are the problem, rather than him. There’s a Cult of Chomsky, and they don’t do any actual linguistics — you know, stuff involving learning actual languages — they instead just faff over theories which don’t fit reality. They’ve been driving the actual linguists out of many departments.

      Chomsky himself stopped doing any linguistics decades ago, so he probably hasn’t noticed that his early theories were proven wrong (the brain does not process grammar according to transformational-generative grammars)… but the Cult kept pushing them anyway.

      It’s a bit like string theory in physics, another thing which started out as a perfectly reasonable hypothesis, was disproven, but continued on as a cult and has been damaging physics departments.

      I could go on. There are examples like this all over academia. Deconstructionism has trashed some English departments (even though there were some good ideas in deconstructionism). Perhaps the worst examples are in economics, which is 90% debunked hypotheses being taught cargo-cult style.

      • Nathanael says:

        In short, the problem with Chomsky followers in lingustics is that they are committing the *cardinal sin for scientists*. That is failing to listen to the empirical data, which has concluded in recent decades that transformational-generative grammars do NOT describe the way human brains process language.

  4. Tom Fid says:

    “the lazy left (who, I think, secretly believe in the inevitability and scientific correctness of right-wing economics and embrace anti-scientific thinking as a way out)”

    I doubt this. Most grasping at pseudoscience that I see occurs for absence of a quality alternative – climate denial or anti-vaccine arguments, for example. But there are plenty of valid critiques of right-wing economics. Most economists don’t believe in perfect foresight or complete markets or other magical foundations of right-wing thinking (though ironically, a lot of economic models that make it to the policy arena do embody such assumptions). A lot of right-wing economics IS pseudoscience, at least in its application. You’d have to be very lazy lefty indeed to be unable to discover abundant reasons to doubt that markets will solve environmental problems for which there are no markets, for example.

    Maybe your definition of “right-wing economics” differs from mine.

    • Andrew says:


      I don’t think all the left is the lazy left, nor do I think all the right is the grasping right. “The lazy left” is a subset of “the left,” and “the grasping right” is a subset of “the right.” There are plenty of non-lazy, non-grasping arguments on both sides, I just wasn’t talking about them here.

      • Chris G says:

        I’ll add that, as someone whose opinions fall well left of center of most issues, I’ve had plenty of conversations with people nominally on my side where I came away thinking, “Dear lord, what are they thinking?” It’s awkward. Politics is a team sport. You need team players. But what do you do when people come to the right conclusion for the wrong reason? Do you correct them? Do you tell them, “Your reasoning is crap. You need to go back and think things through.”? You don’t want to alienate them but what if your adversaries are holding them up as an example of the idiocy exhibited by your team? What if they’re making a bad impression on people who are making up their minds on an issue?

        An example: I periodically talk with people who think the DoD should be pared back by some huge percentage. While their anti-war inclination is admirable there a) is a need for national defense and b) pruning DoD spending back by a huge amount in the current economic climate would likely have a very negative impact on the economy. (NB: The discussion tend not to be about reallocating DoD resources but reducing the overall federal budget.) To what extent should I engage them about the foolishness of fiscal austerity in the midst of a depressed economy and impress upon them that DoD spending is presently a major source of jobs in our region?

        • Chris G says:

          My post above is a tangent. Sorry. It’s great to see a thoughtful discussion of Chomsky’s impact on linguistics. (A field with which I’m not particularly familiar.)

          Somewhat related, Brad DeLong posted a link to a conference, Public Intellectualism in Comparative Context. I immediately thought of Chomsky. (DeLong’s post is ) In contrast to his work as a linguist I am fairly familiar with Chomsky’ political activism. On the basis of his political activism I see him as one of the original (US) public intellectuals.

        • Nathanael says:

          “I periodically talk with people who think the DoD should be pared back by some huge percentage. While their anti-war inclination is admirable there a) is a need for national defense and b) pruning DoD spending back by a huge amount in the current economic climate would likely have a very negative impact on the economy. (NB: The discussion tend not to be about reallocating DoD resources but reducing the overall federal budget.)”

          I’m one of those who talks about cutting the DoD back to reasonable levels, which are about 10% of the current levels. (That would put US military spending on par with China.)

          This is partly a matter of pushing the Overton Window. The “national defense is important” nuts have managed to push the Overton Window so far over that we have a military budget larger than the entire rest of the world, and the cuts *need* to be made — if they aren’t, eventually the military, in search of something to do, will start invading our own cities. Which is already happening to some extent.

          Of course, I do have the sense to say that the spending should be reallocated to a giant Civilian Conservation Corps, so I don’t really fall into the catogory you’re describing.

  5. I think Chomsky sets up a false dichotomy between “data” and “science”. Science has always been about the data. That’s largely what makes it science. I think Chomsky led linguistics down the wrong path with Aspects of the Theory of Syntax (1965) when he introduced the competence/performance distinction (see the section “the goal of linguistic theory” ). It allowed the field to become an armchair endeavour of studying “competence” without looking at data, because that could all be swept under the rug as a “performance” issue.

    For some thoughtful commentary, check out Peter Norvig’s reply to this big-data vs. science meme of Chomsky’s, which is apparently not yet played out:

    Chomsky’s notion of competence and how the field came to evaluate research in terms of “explanatory adequacy” is related to Daniel Dennet’s notion of “vim” (aka “qualia;” see his “Vim: How much is that in real money?”).

    • Rahul says:

      Interesting. Thanks for filling in on the substantive details.

      Thought I hate to think professionals in any one field can be so easily misled by any one academic and that too for decades. If indeed that has happened, I blame the followers down wrong paths as much as Chomsky. After all, these are professional linguists not some naive dilettantes.

  6. Norbert Hornstein says:

    This was a tough post to read. I have been an occasional follower of your stuff and have found it intriguing and informative. I have also been an active linguist working in generative grammar for well over 40 years. In the latter capacity, I have been a big fan of Chomsky’s work. My reason is simple: he identified a coherent and important project for linguists to engage in. Because of him, much of current linguistics is not really about language, but about the mental capacities that make human language acquisition and use possible. Language is the probe to study these underlying capacities, not the primary object of study. And, we have learned a lot about this. We have discovered a slew of “laws” (I know this sounds pretentious, but bear with me) of grammatical form that has purchase over a pretty wide domain of linguistic data. And there have been terrific spillover effects in other domains, e.g. child language development and language parsing and processing. So, my reason for being a Chomsky fan is that he found a really fertile problem and did the leg work necessary to show how to explore it. This said, most of my fellow admirers have disagreed with Chomsky a whole lot about the details. You are wrong to think that there has been no push back and alternatives and disagreements and all the toing and froing one finds in every healthy scientific enterprise. However, this stuff does NOT make it into the NYTs or the New Yorker or The Chronicle. Why? Because it is either too inside baseball or, more likely, because it build on the fundamental insight that Chomsky had about what the nature of the enterprise is.

    To end, I would love to have you name names regarding your “friends” in linguistics. There is a vocal minority that believes that the problem that Chomsky posed is wrongheaded or incoherent. They are simply incorrect. They might be interested in OTHER issues regarding language, and not be interested in the question that Chomsky posed. But there is nothing wrong or trivial or nutty or retrograde about the question that Chomsky asked nor the direction of work that Chomsky initiated and there is plenty of disagreement within the field about the details, both big and small. So, Andrew, sorry to hear you hang out with people who seem, from where I sit, very uninformed. Travel down to NYU sometime and talk to the linguists there. They will give you another perspective. Or, if you have time got to where many of the larger issues are being discussed quite vigorously.

    • Andrew says:


      I think everyone agrees that Chomsky has made huge contributions, and I don’t think anyone is saying that Chomsky’s ideas should not continue to be studied as a part of linguistics. So maybe the argument is ultimately about resources: what fraction of linguistics research should be done using Chomsky’s precepts, and what fraction should be done using other approaches.

      You write, “You are wrong to think that there has been no push back and alternatives and disagreements and all the toing and froing one finds in every healthy scientific enterprise.” I have no idea why you think I think there has been no push back etc. I have never written such a thing, nor do I hold that view.

      Again, this is second-hand. I refer you to Bob’s comment for more background. Thanks for supplying references to other perspectives.

    • Norbert Hornstein says:

      I took the following to suggest that there was no pushback intellectually: “The admirers give no useful feedback.” There is a lot of very vigorous discussion concerning the “right” way of describing the Faculty of Language, Chomsky’s proposed object of study. As it has turned out, to date, the methods used to study this have not been particularly heavy on statistical methods, though this is changing. Charles Yang, Bill Idsardi, Bob Berwick, and many others are combining statistical methods (btw, many proposed by CHomsky in the mid 50s in his Logical Structure of Linguistic Theory) to study various aspects of language acquisition and what they tell us about FL. However, these methods are combined with quite detailed theories of what the stats are counting. What Chomsky opposes, and I would have thought you did too given your paper with Cosma Shalizi, is the idea that big data could bypass theory. There is an idea out there that get enough data and the principles will just “pop out.” Chomsky demurs. As do many others. Everyone since the start of the modern generative enterprise has recognized that counting matters, even sophisticated counting. However, counting is not enough: you need to know what to count. Chomsky’s view is that counting surface distinctions won’t cut it. He is not alone. Sydney Brenner has had virtually the same reaction. And, as I said, I took you and Shalizi to be making a similar point. It’s Kant’s old observation: Data without theory is blind. I take this to mean that collecting data, even very large amounts without some idea of which data to collect is not a very good way to find out about the world. Sadly there is no substitute for thought. And that I think is Chomsky’s real point. You disagree?

      • Andrew says:


        Fair enough. I think in the entire scientific enterprise, we need design of data collection, we need data, we need data analysis, and we need theory. I also think that individual scientists can work in only some of these areas and still make useful contributions.

        • Norbert Hornstein says:

          This is my last comment and thanks for being so gracious. There is a mom and apple pie tone to your last comment and, of course, at one level, I agree. And there has been a LOT of empirical work within the Chomsky Program. We have learned more about different languages in the last 40 years than we ever learned before. THere has been tons of of psycholinguistic work and lots of CS modeling of great value. However, (you know that was coming right?), within generative grammar one of the bug bears has been a kind of mindless empiricism (it flies under the name of “associationism” in psych). This view has tended to attach itself (improperly in my view) to stat methods, thinking that counting surface distinctions was a route to grammatical competence. This fight has been long and it repeats every couple of years. I believe that part of the reaction to BIG DATA has come from the view that this will serve as another push in the same direction. It won’t. It cannot, as I and others have argued at length. But, that’s the way it often plays. That explains part of the heat in the reaction. At any rate, thx again for the give and take and I await the day of your linguistic enlightenment. I am sure it will come. NYU is not that far away and the Village has great espresso.

          • Andrew says:

            1. I don’t know much about linguistics, but I can see the analogy between the “mindless empiricism” that you discuss in linguistics, and various confusing path-analysis studies that I’ve seen in social science. I do, however, still think that a nearly theoryless data analysis can have great value, as long as it’s not overinterpreted. Red State Blue State had lots of theoryless data analysis, and I think it was a useful contribution, in part in supplying stylized facts that can be incorporated into theory.

            2. Num Pang’s my favorite sandwich shop and it’s on 12th St. But I don’t plan to enter any NYU buildings for awhile.

          • Mark Johnson says:

            It’s always amazes me that otherwise intelligent people equate statistical inference with a tabula rasa empiricism. Bluntly put, somehow you have to know what to count. Many of us actually building statistical models of language are building models that generalise over (i.e., involve counting) abstract hierarchical structures.

            I think it is possible to view the acquisition of functional items in e.g. Stabler’s formalisation of Minimalist Grammar as a fairly standard kind of statistical inference. I would actually work on this if I thought it would prove anything.

            • Norbert Hornstein says:

              Yes, there are stats worth doing.mi mentioned Yang and Berwick, but I could have mentioned Stabler and even Johnson. But much of what gets done is NOT like this. Christiansen, to name just one person. So, of course stats can be done right. Nobody ever denied this or argued against it. But would Mark deny that lots of this kind of work is intended to circumvent structural analysis rather than incorporate it? Is the whole big ram trigram industry a mirage?

              • Mark Johnson says:

                Well, I guess which Mark you’re talking about. (smile) The aims of much of the empiricist work is practical (often in the service of delivering advertisements more effectively), and I believe its claim to fame is more on economic rather than scientific grounds.

                For example, one can learn the “sentiment polarity” of many words from reviews (e.g., “tasty” is a positive term in restaurant reviews, “boring” is a negative term in movie reviews, etc.). One could construct such a “sentiment lexicon” manually, but its cheaper to do it by “data mining” reviews and associated “star” ratings.

                Economically this is very important (it’s a multi-billion dollar business, doing things that would be impractical any other way), even if it’s scientifically not so important — nobody claims kids learn words by associating them with the number of stars in on-line reviews. (*)

                I agree that the empiricist approaches have produced less in the way of impressive insights about language than the Chomskyian approaches. I think it’s very impressive the way that several basic principles interact to make a surprising empirical prediction that no one has noticed before (e.g. the alternation in the embedded subject coreference in “Who do you want to talk to (him)?”). Chomksy has always had many such examples (from “flying planes” on), and I think that it’s a good sign that a theory is on the right track.

                I’d like to be able to point to similar kinds of surprising empirical phenomena that empiricist work has lead to, but I can’t think of any. (BTW, are there examples of surprising new empirical phenomena that Minimalist analyses lead us to discover?)

                Still, I think it’s worth studying what you’d call associationalist approaches to learning. I think an associationalist approach is still basically the only game in town for learning much of the lexicon (syntactic bootstrapping just can’t provide enough information, as there are far more words than syntactic frames). We have learnt a bit more about these associationalist approaches in recent years, including that they aren’t all made equal. So even at the associationalist level, the learning problem and the learning algorithm interact in interesting ways. In other words, it’s not just tabula rasa here either (if it were, there wouldn’t be a field of statistics or machine learning).

                (*) I think it’s basically irrelevant that commercially important technology doesn’t rely on the latest scientific insights about language (from Minimalism or elsewhere). Nobody would claim that it’s a failure of Einstein’s theory of relativity that it doesn’t help us build better bridges. And while the fact that something is technologically useful may be suggestive about its possible role in biology, there are lots of cases where artifacts that are incredibly useful seem to play a negligible role biologically (e.g., wheels for locomotion).

                PS. How do you get the time to blog so much? Have you retired?

    • Nathanael says:

      The trouble, Norbert, is that if you actually study language as she is performed — proper neurolinguistics — what you find from the neurology evidence is that human brains do NOT use transformational-generative grammar, certainly not to parse sentences, and probably not to generate them either.

      Chomsky’s work was a *good try* at answering the question of how the brain handles grammar (an ancient question), but it *failed*. Time to move on to a new theory of how the brain processes grammar. Actually, it was time to move on *30 years ago*, which is the problem with the cult of Chomsky — as I mentioned above, this is the same as the problem with string theory in physics departments. It was a nice try, it failed, move on!

      Several of the “laws” of grammar which are promulgated by Chomskyists are simply wrong — there are obscure languages which don’t follow them. In some cases, there are fairly well-known languages which don’t follow them. However, this hasn’t stopped the Cult of Chomsky. The Cult of Chomsky has continued to grow by discouraging students from doing the basic work of linguistics — which is to learn huge amounts about large numbers of different languages so that they *can* compare and contrast.

      There are “linguists” who follow Chomsky who only speak one language! Think about that!

      Chomsky’s work has been fairly useless to the CS work of natural language processing (which I have followed in detail), and due to being simply wrong has not been particularly useful in child development either.

      • Norbert Hornstein says:

        I guess we will have to agree to disagree about most matters. But so that where we disagree is out there let me say several things:
        (1) We have no idea how brains do what they do. Right now what we have some handle on (not a big one however) is where the brain does things. How it does thme is really up for grabs. Not surprisingly, there is a lot of work correlating brain “imaging” rechniques (Fmri/MEG/EEG) with linguistically informed distinctions (e.g. islands (Helen Neville and Marta Kutas have stuff on this)). There is also some very excting work on phonological coding in the brain (David Poeppel and Bill Idsardi have done some of this). However, HOW this gets done, the actual computations involved is at the very best obscure. Thus any conclusion regarding whether Chomsky like grammars are used by the brain is currently very much a moot point. However, there is zero evidence that it doesn’t and there is a lot of reason to believe that the connectionsit rage in brain modeling is really off on the wrong track (c.d. Gallistel and King’s book on this for a thorough, and to my mind, convincing debunking of this approach). So, contrary to your claims, I see NO current reaons for thinking that the brain sceinces have anything deep to add to the discussion concerning language at this time. Of course, we all hope that this will change, but that’s the current state of play. BTW, my dept has put its money that this will change as we have a very active neuro group working on these issues and I can report that they do NOT find the CHomsky vision an impediment to their inquiries.
        (2) Generative grammarians have as a group studied MANY different kinds of languages in great depth. Why any given linguist needs to know more than one language eludes me. Is E.O. wilson disqualified form talking about large issues in biology becuase he is only an ant guy? Nope. So why should a monolingual linguist not be in the same boat. The aim is to look at a variety of languages, not to do so solo.
        (3) Which are wrong? I know of no evidence against, say, structure dependence being refuted by any language. You are no doubt thinking Piraha, as this has gotten lots of play. Sadly, the work is both empirically questionable and conceptually misdirected. As I’ve discussed this elsewhere (see facultyof for extensive discussion and references) I will not bore go over it again here. Suffice it to say, your judgment about these matters is not one that I think holds up to scrutiny.
        (4) As to the CS community, it depends on which parts you trawl. Most of the CS community is not that interested in issues of linguistic competence and so not surprisingly have not found work concentrating on it of much value. However, there are parts of the community that do care and this work has been influential. Berwick’s work and Marcus’s on efficient parsing for example. As I mentioned earlier, even skeptics are discovering that looking for structure can be very useful: Noah Smith’s work is notable. Even Fred Jelinek came to the conclusion that headedness and phrases structure was very much worth tracking even for practical purposes. So, again, we must agree to disagree.
        (5) Let me end with my common refrain: Chomsky’s work is aimed at addressing a very particular issue: what powers underlie human linguistic facility. Answering this question need not answer others that interest you. However, critiques that don’t address this (e.g. Norvig’s) are simply besides the point.

        • Norbert Hornstein says:

          Mark: I find myself in almost total agreement with your assessment. Time? Retired? No. It’s just struck me that there are a lot of misconceptions out there and maybe a little time trying to take them on would be useful.

  7. eric says:

    “who, I think, secretly believe in the inevitability and scientific correctness of right-wing economics”

    Could you expand on this?

  8. JCB says:

    I saw this a while ago and have recently, in fact yesterday, been thinking about Chomsky’s point. It has resonated, you could say. I think it is a good one.

  9. K? O'Rourke says:

    You cannot rule out a hypothesis by the way it was generated -nor even who generated it.

    And everyone is wrong about everything in some sense, but some wrongness is more productive than others (with apologies to George Box).

    My rough sense here is that Chomsky is strugling with the role of purposefullness in cognition and trying to avoid Wittenstein’s (admitted) mistake of neglecting it.

    (Andrew’s posts have a fascinating variety of insights.)

    p.s. Never found Chomski’s work that helpful, but that means little.

  10. Brian says:

    Aren’t those “two conceptions of science” basically just another way of framing positivism vs. realism?

    Positivism being about observation and description, and realism being about explanation, process models, and mechanisms (unobservables)?

  11. konrad says:

    I think Chomsky is right when he says that two approaches exist – one can distinguish between scientific model-based approaches and prediction-oriented approaches that are based either on very crude models or on none at all. The choice between these two philosophies is a central theme in Machine Learning as a discipline, and certainly seems to create faultlines within the discipline. The same seems to be true in Statistics, but perhaps not as visibly.

    He may also be right (at least some of the time, and perhaps most of the time) when he associates big data approaches with the latter category.

    But he is dead wrong when he thinks the boundary lies between statistical and non-statistical approaches, and lumps all statistical approaches into the “dumb” category.

  12. This makes sense if Chomsky is talking about something like the distinction in physics between theoretical and experimental. I am also inclined to believe that theory is most respected where there are no good data, like in some areas of macroeconomics:

  13. JP says:

    Two comments. First, there is a rather large difference between Chomsky’s thinking as a linguist, and his thinking as a political scientist. These are really the proverbial apples and pears. I happen to be professionally suspicious of his linguistic contributions, but a (non-professional) admirer of his work on political science. Second, the idea that Chomsky did not engage his critics is simply wrong. He engaged with many of his critics, and often quite convincincly. Sometimes he even flew to other continents to do so. This, for instance, is a famous one:

    • Nathanael says:

      I am another person who finds Chomsky’s contributions to political theory much more insightful than his contributions to linguistics.

  14. Steve Sailer says:

    Chomsky is 84 years old and we’re still arguing over his strengths and weaknesses. That impresses me about Chomsky.

  15. jrkrideau says:

    “Chomsky is 84 years old”

    Strange, that picture makes him look much younger.

  16. If Noam Chomsky were really such a terrible person, the attacks
    on him wouldn’t be quite so lame.

  17. Rahul says:

    So what exactly is wrong about Chomsky’s linguistics? And how exactly does a single academic become a “roadblock” in any serious scientific field? Has peer review failed the field? Is linguistics mere ideology?

    “Other linguists love Chomsky, but I think they’re the kind of linguists I wouldn’t spend much time talking with”

    What is this nasty “kind” of linguist?

    I don’t get this post. I wish the critique was more meaty and substantial. As it stands it tastes like a toxic stew brewed from hearsay and ad hominems.

    PS. I don’t like Chomsky’s activism nor his political views. Maybe his research is total crap too. But I’d loved to be told a bit more than “my friends don’t like him”.

    • Andrew says:


      1. I don’t know linguistics. I’m just reporting what I’ve heard. Bob’s comment gives some background.

      2. I don’t think I talked about anyone being nasty. That’s your word, not mine.

      • JP says:

        Andrew, could you explain *why* you want to repeat what anonymous others mutter about Chomsky? Why don’t you read and cite some actual published criticisms of his position instead? Or, if this is outside your field of expertise, refrain from commenting on Chomsky’s popularity in linguistics and just discuss the big data issue from your own point of view? As a linguist, I’d be very interested in concrete criticisms of Chomsky’s position, but now I don’t even know what and who I’m arguing with.

        • Andrew says:


          Bob Carpenter and Peter Norvig know a lot more about linguistics than I do, so you can see their comments. My impression is that there are scientific disputes regarding the relevance of certain quantitative data to the development and understanding of theories in linguistics, and there are also disputes regarding the allocation of scholarly resources in the field.

    • Nathanael says:

      “What is this nasty “kind” of linguist? “

      Linguists who don’t know very many languages, for one thing.

  18. Steve Sailer says:

    Chomsky’s big breakthrough came in the 1950s (!). Yet, he remains productive in the 2010s. This is one of the longest reigns at the top of any scientific field. I’m sure there are some younger linguists, however, who wouldn’t mind him retiring.

    • TGGP says:

      I haven’t heard of him doing much in linguistics recently, but on the other hand what I do know of linguistics tends to come from the folks at Language Log who write more about funny mistakes than academic papers (I suppose there’s some sort of anthropic principle about me not bothering to read a blog unless it is accessible to laymen). Contrast Freeman Dyson, who I hear about making a new contribution to the prisoner’s dilemma quite recently at an old age. There are other prominent academics who make a major achievement and then don’t do much more, James Watson basically became an administrator after his DNA breakthrough. I might have added something about Krugman having a similar reputation among academics, but he wrote a paper with Eggertsson recently.

      • Norbert Hornstein says:

        Chomsky has been very active all through his life. Syntactic Structure started it all. But Aspects of a Theory of Syntax was one of the more important ling books ever and Lectures on Government and Binding is a classic as well. There have been many other books and papers on technical topics, most recently The Minimalist Program and a series of papers further developing the ideas therein. Those interested in these topics can check out, where many Chomskyans issues are discussed at length. I hate to say this, but it seems that a good part of the Chomsky “critique” is rooted in complete ignorance of his work. Too bad.

        • TGGP says:

          Interesting to know.

        • Nathanael says:

          I know Chomsky’s lingustics work quite well. It is best described as “obsolete” in light of what we currently know about lingustics. (This doesn’t make it *bad* — it was good *at the time it was published*.)

          Unfortunately Chomsky has a cult who worships his writing. This has been an absolute disaster because it has suppressed actual linguistics research — I believe there are two, maybe three universities left in the US with strong linguistics departments (U of Indiana Bloomington being at the top of the list), while the rest are just doing Chomskyist nonsense.

          As I say, this isn’t unique to lingustics. It’s happened to a lot of other fields. String theory has wrecked physics departments.

    • Nathanael says:

      Chomsky hasn’t done any useful lingusitics research since the 1980s. That’s fine. Many people switch fields when they get older and it’s a fine thing to do.

  19. […] One oversold critique of empirical work is that statistical relationships do not a theory make. The point was made best by Noam Chomsky in a recent interview (via Gelman): […]

  20. I’d like to chime in with my critique of Chomsky. I have a vague skeptical sympathy with his politics but absolutely not time for his linguistics. In fact, I’d say his major achievements were in mathematics and engineering not in linguistics at all. Here’s the argument in full: The key phrase:

    “I honestly cannot think of a single insight he’s had about how language works as language. His main contribution to the study of language (his only one really) was a description of how certain combinatorial properties of English syntax can be modeled using a particular formal system” but the problem is, I note, “everything that followed was predicated on the model being isomorphic with the thing modeled.” This may be vaguely relevant to some of the themes of this blog.

    I also quote Robin Lakoff who said that following Chomsky meant “accepting the impossibility of saying almost everything that might be interesting, anything normal people might want or need to know about language.”

    But ultimately my main beef with Chomsky is his philosophy of science which demands that everything looks like the hard sciences – which means has math. This to me seems more like voodoo…

    • Norbert Hornstein says:

      It would be nice to know what Chomsky skeptics take to be the subject matter of linguistic investigation. Chomsky has been very explicit about what HE takes the topic of investigation to be. His critics? Not so much. There are many things one might want to study. If you don’t like Chomsky’s interests then what he does will not interest you. The scientific issue is whether the question Chomsky poses (Plato’s Problem, Darwin’s Problem, Broaca’s Problem, to name 3) are interesting and worth investigating. If they are, then Chomsky has made great contributions to their statement and solution. If these problems DON’T interest you, then much of what Chomsky says will not interest you. Much of Chomsky’s criticisms of other work starts from asking what this work contributes to answering the problems he has posed. And this seems to me fair enough. It would behoove his critics to specify the problems they are interested in and then showing that how Chomsky has to say bears on that. Till then, it seems to me just name calling, and not very productive name calling at that. BTW, his views about Norvig only make sense when viewing Norvig’s comments as critiques of the program that he has specified. He has nothing to say about engineering problems or whether big data might make web search better or worse. If Norvig’s comment are understood as proposals concerning the cognition of language, then Chomsky’s critique makes sense. He may not be right (though I think he is) but it is on target. If not, if his and your interests are otherwise, then Chomsky’s views are entirely besides the point. Is this really a surprise?

      So, state the problem then we can analyze the relevance of Chomsky’s proposals. Avoid doing this and the “debate” is moot.

      • What evidence do you have for saying that Chomsky skeptics haven’t specified the subject matter of linguistic investigation? Lakoff, Langacker, Hoey, Halliday, and many others have all done that – not always in reference to Chomsky. The problem is that Chomskeans say that this kind of specification doesn’t count. Which is why I think the only real argument you can have with Chomsky is about the philosophy of science. His only argument with his critics is to dismiss them – which he’s only done in interviews – never to my knowledge in his published work (at least no in a substantive way).

        But he hasn’t always been very good about delineating what he takes language to be. I give some quotes from him on which show that far from being very precise, he equates language with what his method can apprehend rather than with the complex phenomenon it really is. I’ve had this argument with Chomskeans many times. In the same breath, they say, Chomsky never claims he describes all of language and Chomsky tells us so much about language. But he doesn’t. He just ignores all the bits that cannot easily be described by his theory. Which is why my argument is that rather than being right or wrong he is mostly irrelevant to the study of language.

        • Norbert Hornstein says:

          Hmm, what evidence do I have that they DON’T specify a program? Well, like I said, I’ve been doing this for about 40 years and though many have pushed back on Chomsky, it’s never been clear to me that they’ve understood what he is up to. The program is to investigate what it is about humans that make them acquire language reflexively. You know, fish swim, birds fly, ants navigate, people talk. Ok, fish don’t go to school (joke) to acquire this capacity, nor birds nor ants. The idea is that this is also true of humans re language. Ok, what’s the fins structure of this reflexive capacity? Here’s an idea: let’s look at the formal and substantive properties of actual natural languages and argue backwards from their properties to the properties of the underlying capacity. That’s the program. Have we learned anything? In my view, tons and tons. We know a lot about the formal properties of grammars: the kinds of dependencies they license, the kinds of rules they employ, the kinds of restrictions they obey. We also have pretty good ideas about how these systems interact with some interfaces, though this work is still less well developed than the pure linguistics work. So, I know and can articulate Chomsky’s program. I cannot articulate the others, beyond saying that they want to describe language. I am fine with this, and taken in some ways it is not inconsistent with the program I outlined. But there are problems as well. First, that “language” turns out not to be a very well defined notion. Second, and this is not a criticism just an observation, given the CHomsky program not all descriptions are of equal interest. This is what I meant when I said that it is important to know the target of inquiry: the question being asked. WIthout this, it is not easy to evaluate the work or the criticism. One thing that seems silly to me, is to say that some questions are acceptable and some not. Rather, some interest you and some don’t. Fine. I am interested in CHomsky’s question and hence find his suggested answers interesting and, dare I say it, often right on the money. Sometimes I disagree with him. This is fun and makes for lively discussion. What is less interesting are disagreements masking divergent interests, and unless one specifies what the program is, these tend to proliferate.

          • I think we’re arguing at cross purposes. I never said that some questions were not acceptable – just the opposite. My criticism of Chomsky is exactly is that both his questions and answers are uninteresting to anybody that would like to study actual language use. The main criticism of Chomsky that this blog post is referring to is that he (and his acolytes) engage in accusing non-generativist research from being unscientific and have set back language research in some regions by doing so. I have seen plenty of evidence for this in the face of strenuous denials of Chomsky’s proponents.

            The criticism of his language program is also exactly equating of syntax with language – I don’t go in for definitionalism – but a definition that strips away the very nature of what is being studied does not seem much better than no definition at all. There are so many more things going on in language (acquisition of complex lexical patterns, collocations, figurative language, language politics) that go on at exactly the same time. None of those are addressed by Chomskeans other than through positing some sort of massive modularity. I stand by me original definition of Chomsky’s program as “a description of how certain combinatorial properties of English syntax can be modeled using a particular formal system”. These are formal properties of language but nothing about them is substantive. The very existence of multiple overlapping models suggests that. This is essentially a structuralist program. Nothing wrong with that but there is much more to language than its structure.

            • Anonymous says:

              There is indeed much more than syntax. There is semantics, phonology, morphology and then some. And Chomsky and his acolytes have worked on these questions too. What is correct, is that he has not worked much on language USE. there are several reasons for this. One is that he is very skeptical that this can be productively studied without knowing anything about the underlying system. Now there are others working on this a lot: how are grammars used to parse a sentence, how are grammars used to verify visual displays, to name two. However, people interested in use have in mind much bigger fish, and these have not been productively studied by generative grammarians. Why not? Well, they think that this is too complicated, a massive interaction effect that will not productively shed light on the structure of the faculty of language, which let me remind you, is what they are interested in studying. So, they don’t put many eggs in this particular basket. Of course if that’s what interests you…

              Last point: Chomsky tends to counterpunch. He has been a punching bag for a long time. It’s de rigeur (you mig fall into this category too) to define your project by first pooping all over his. Why? Well he is the big fish. Now if you poop on his stuff, I think it’s fine for him to hit right back. And he is very good at this. So, yes there are claims and counter claims. Your position, however, does not seem particularly open minded. And saying that he isn’t either is not a great step forward. I will end our little discussion it’s one more remark, leaving the last word to you. Of late, even the use inclined have decided that knowing someone about grammar would be very useful. Before his death, Jelenek decided structure was useful, for example. I suspect that this is the rule for the future.

              • Nathanael says:

                “Now if you poop on his stuff, I think it’s fine for him to hit right back. And he is very good at this. “

                Not in linguistics. His “counterpunches” in linguistics are pathetic, a series of well-known fallacies in reasoning strung together.

                The point at which I stopped having any use for Chomsky was when I realized that transformational-generative grammars (which I independently invented at the age of 8, by the way — they’re a very simplistic method of analyzing English grammar) do not actually correspond to the way the brain processes incoming sentences, or the way it generates them.

                There’s lots of psychology studies on this, you know! The Chomskyist linguists should go read some of them!

                Now, there are people who are attempting to construct different theories of grammar which are consistent with what we know of psychology and neurology. This is an interesting and useful program. Chomsky cultists are not helpful in pursuing this program.

          • Nathanael says:

            “The program is to investigate what it is about humans that make them acquire language reflexively. “

            Except that Chomskyists haven’t been PURSUING this program, not scientifically. If they were they’d be in neurology departments. And they aren’t.

            “Here’s an idea: let’s look at the formal and substantive properties of actual natural languages and argue backwards from their properties to the properties of the underlying capacity. That’s the program. “

            Yes. And it’s a failed program.

            “Have we learned anything?”

            No. The Chomskyist formal grammars suffer from being WRONG — they do not correspond to how the grammar is processed in the brain!

            • Norbert Hornstein says:

              As I said in reply to earlier material: we must simply agree to disagree. There is currently no evidence that shows that Chomsky like grammars are not used in brain processing and production of sentences. INdeed, there is MEG/EEG and Fmri evidence that it is used. What we don’t have are good accounts of HOW brains implement grammars. But, aside form early vision, we have little idea of how brains do much of anything. We have some ideas of WHERE brains do what they do, but little idea of how wetware implements circuits or what circuits it implements. Till we do, your claims, I fear, have little empirical foundation.

            • To defend Chomsky a bit here, he explicitly claims that the processing part doesn’t matter. Our knowledge of language is separate from our use of language. You can accept that and argue (successfully) that his description of the “knowledge of language” is wholly inadequate on its own – which by now most mainstream linguists have done. But of course you can also reject the entire premise of competence/performance and look for both a cognitively and socially plausible theory of language.

              The thing is most linguist never think about Chomsky in relation to their work but it is tiresome to have to correct drivel from students who come in and talk about innateness of morphology, etc. The other frustrating thing is that many other disciplines (philosophy, neuroscience, some strands of cognitive psychology) still waste time on trying to prove/disprove Chomsky while the bulk of linguistics has moved on to more productive matters. So you can’t really get very far in linguistics “pooping” on Chomsky, although it might give you some column inches in the gullible press. His own reaction to criticism has been to dismiss his opponents as unscientific and largely done through interviews rather than in scientific forums. Many will say that the one time dominance and negative impact of Chomskean linguistics is overstated and they are probably right. But I did come across a “linguistics” graduate from a Chomskean department who had never heard of Roman Jakobson.

  21. LJ says:

    I am not sure whether this is a proper comparison, For me, Chomsky always reminds me of Einstein. With limited knowledge in two fields, I cannot comment on their achievements as a linguist or as a physicist, both of them made tremendous contributions in their own fields yet, what really brought them a big name is their political activities. Based on many physicists’ reactions towards Einstein,I can imagine there could be quite a lot of linguists who don’t totally agree with Chomsky. So, it’s not at all surprising that one can happen to have more linguists friends who don’t admire Chomsky that much. Mostly, I believe people who come and check your blog have their own brain to judge every issues – if one can leave a proper comment on this issue, he/she will take your opinions to Chomsky with the same degree of seriousness that they would have taken for a law professor who talks about eduction. So, there won’t be anyone who stops you getting your favorite sandwich from all the way down to Manhattan :)

    • Andrew says:


      Huh? It was Einstein’s physics, not his (mild) political activities, that brought him fame. Regarding linguistics, I’ve expressed no opinions about Chomsky, I’ve only been discussing the opinions of experts I’ve spoken with.