Sunday, November 8, 2015

Media linguistics

Language sells. How else to explain the number of consistently uninformed articles in the major media purporting to bring scientific expertise to one or another language phenomenon.  Most recently, there have appeared a trio of articles in The Age, The Guardian and Newsweek whose aim is to delight he masses with scientific insights into language. I was intending to write about these, especially the horrible piece by Ibbotson and Tomasello. However, I have been (very very happily) been spared the chore. I hate cleaning out stables, and I don't have to do it because Asya Pereltsvaig has done the unpleasant lifting (here).  Let me add three comments to her excellent discussion.

First, I like the tone. It recognizes that this stuff is junk and says so. It also provides a handy set of heuristics for those wanting to identify probable junk. I hope that the popular press takes note and acts accordingly, but I am not going to hold my breath.

Second, Asya links to many detailed comments of relevance to the (ahem) substantive points made. The fact is that there are no substantive points made. The Ibbotson and Tomasello piece, as Asya and others point out, is either vacuous or false (sound familiar?). Yes, social interaction probably has something to do with language, and no this tells us very nothing about linguistic structure in any domain. In fact, it is not clear that the claims made are intended to be taken seriously given how little thought it takes to show how inadequate they are (e.g. see the excellent boxed comment by David Pesetsky on Asya's post). What's evident is that editors allow clearly inflated and uninformed claims to be made without any skepticism whatsoever. I used to find this amazing, but again these are the same people who bring us weapons of mass destruction, lights at ends of tunnels,  the immanent collapse of social security and rampant voter fraud. Why expect better when it comes to linguistics? For these venues, language is a little like animal planet: forget biology, look for cute, fuzzy or odd. This said, kudos to those that engage.

Third, though infuriating, I am suspect that this sort of junk in the popular press does not matter much to the standing of linguistics. What matters a whole lot more is that the absence of reasonable linguistic coverage in the popular science press and the absence of respectable linguistics in the research of our scientific neighbors in computer science, psychology and neuroscience. Let me discuss each briefly in turn.

Not having decent stuff in this venue, I believe, matters a lot. Why? Because this is where scientists cover the science that are not their specialties. That should be our intended audience. Of course, I have nothing against correcting the journalistic unwashed. But, I have doubts that their views would carry much weight in the science community if good stuff were constantly being discussed in venues like Science Daily, Scientific American, or New Scientist and aggregators like Aeon or Nautilus. We should be figuring out how to get the attention of these venues.

Second, we ignore the bad linguistics prevalent in psych, CS and cog-neuro at our peril. These are far more important politically than linguistics is. If we fail to engage with this audience I fear that we will have a short half life. Many linguists have stopped seeing linguistics as part of cognition and biology. Indeed, it often looks like many linguists consider the cognitive/biolinguistics perspective on GG to be windy sentiment. This hurts us far more than a few stupid pieces in the popular press.

So, thanks to Asya for carrying the can here and to others who have jumped in to shovel the garbage. I thank you from the booth of my blogging heart.


  1. I responded in the Guardian comments section. Since no sensible person reads those for blood-pressure reasons, I've pasted below what I wrote:
    "This article is astoundingly one-sided and badly misrepresents modern linguistics---but I guess neither author is actually a linguist. Let's just take one example: the theories associated with Chomsky don't lead one to expect universal grammar to pop up in an obvious way when you look at the "diversity of human languages", contrary to what is implied here. Universal grammar is a set of abstract principles that work together, like laws in chemistry, to predict what languages are possible and what ones aren't. For example, it predicts (via an abstract principle called "Structure Dependence of Rules") that languages care about structure in their grammatical patterns, so when you say "It's in the garage that the people who are singing are dancing", the phrase "in the garage" tells you about where the dancing is going on, not the singing (it roughly means "The people who are singing are dancing in the garage" and not "The people who are singing in the garage are dancing"). You can feel this difference in meaning, even though if you're dancing in the garage and you're singing, you're singing in the garage too! Why do you have this sense of what the sentence can mean and what it can't? It's because the verb "singing" is structurally buried down inside the whole subject of true sentence ("The people who are singing") so that it's structurally quite far away from "in the garage", even though it looks quite close in terms of just words in the sentence. In fact, the relevant universal grammar principle correctly predicts that no language allows a phrase like "in the garage", which is structurally outside the subject of the sentence, to modify a verb inside a clause modifying that subject (Phew!). There's universal grammar for you. Now notice that to say this, you need to know a few things about how language actually works - it's abstract and you need quite technical concepts (subject, modifier, relative clause and so on)---apologies Guardian commentariat. But there's precious little evidence in this article, or indeed in their published works, that the authors are willing to spend time finding out about how language works. The level of discussion about language in the media is painfully low, and this article depresses it (and me) yet further.

  2. Pereltswaig's rebuttal quite inappropriately conflates the linguistically illiterate claims of Frenkel with issues that, like it or not, are in fact being debated within the field. It takes remarkable chutzpah to dismiss the statistical work of Matthiesen and Coupé as "pseudo-linguistic rubbish" by citing a single counterexample - and even more so when the "counterexample" happens to agree perfectly with his claim! As for the Guardian article, the work on which it's based may be total rubbish for all I know, but dismissing it on the grounds that Tomasello isn't a linguist is thoroughly misleading, given that his research focuses mainly on language acquisition. Linguists cannot credibly exclude psychologists from the study of the psychology of language, whether you agree with some particular psychologist or not.

    1. I will let others address the M&C stuff. However, it is false that Tomasello's work has not been carefully vetted by many many linguists. It is very bad. And it does no good to claim that it is dealing with "psychology" (i.e. acquisition) rather than simple linguistics. As I noted in the post, I consider linguistics to be part of psychology and I find T's work to be wanting on acquisition grounds. He has NEVER indicated how he thinks the child can scale up to adult grammars using the methods he thinks operative. Pesetsky makes this point well in Asya's post. If this is right, and it is, then the problem with T is not merely that it is linguistically ignorant, but that it does not come close to delivering what it claims. Actually, T doesn't even seem to recognize the problem.

      Is pointing this out "chutzpah"? Maybe. I am told that shouting that emperors have no clothes is also chutzpah. If so, we need more of it.

    2. There is a huge difference between arguing that someone's work is bad - as, say, Pesetsky is quite reasonably doing in this case, and as I'm sure you would - and attempting to delegitimise it based on extraneous factors such as their job title, which is what Pereltswaig's point 1 amounts to. Imagine if someone, sharing your conviction that linguistics is part of psychology, proposed that historians and laymen interested in history should systematically distrust historical linguists on the basis of their "lack of historical credentials".

    3. I agree that title should not matter. ANd I think that Asya would agree. What matters is knowing something about the subject matter. Her three points are intended, I believe, as generally reliable indicators of bad work (i.e." three telltale signs for identifying pseudo-linguistic rubbish"). She thinks, and this seems fair, that being credentialed is often a good prima facie indicator that your views are informed. Is this apodictic? Well not at all as there are many credentialed people who are very confused (I've talked about several on this blog) and there are some that are not credentialed that know a lot (some of my best friends do not have degrees in linguistics yet speak knowledgeably about it). So, I think Asya's point is that this is a good prima facie indicator and that if it is not realized then one should take a very careful look at what passes as expertise. You might think, and I would agree, that you should always do this (i.e. take a careful look at the arguments). However, this advice is not meant for you and me but for neophyte newspaper writers and editors who do not generally believe that they are in a position to evaluate the arguments. So, for these people looking at credentials is not a bad place to begin.

      Is it a good place to end? No, of course not. But for the particular cases Asya discusses, the work has been vetted professionally and been rightly found wanting. Or, let me be careful, The Tomasello stuff that I know of has been carefully vetted and the Asya links to professional assessments of the other pieces as well. So, she does not exclusively rely on her rules of thumb here for her judgments. She relies on what she knows and the opinion of experts she trusts who have provided reasons she finds compelling. And here I have a question: do you think that Asya got the substance wrong? If so, explain how. I would one to discover that my low opinion of Tomasello's language work was based on misunderstanding his results. I have read some of his work and have always found it irrelevant to the conclusions he wishes to defend. I have vetted my conclusions with people who work more closely in these areas and have not been dissuaded that my judgment is well-founded. But…

      So, Asya is harsh and I applaud this for from what I can tell the work deserves a harsh judgment. Are her rules of thumb too flippant. Yes if advice to professionals, not as heuristics for newspaper people. But thx for the thoughts.

  3. I'm with Lameen. The dismissal of the Guardian piece seems to be a knee-jerk reaction to its criticism of UG. You may not be a fan of the particular stance it advertises (I'm not; see below). Indeed, there may be grounds to disagree with it, cf. David Adger's point, which is really nicely written by the way (except for the dig at the end). But the fact remains that the article manages to cover quite a bit of interesting empirical and theoretical work on language development, even citing sources — features that are otherwise quite rare in news outlets covering of language and linguistics.

    Whether you agree with it or not, it is poor form to equate the Guardian article, written by well-informed and respected academics who work on language development, to the absolute drivel by Frenkel that appeared in The Age and went viral subsequently.

    Worse, doing so makes one lose sight of the most obvious answer to the question of how to improve the representation of language and linguistics in the media. You participate in torching a piece written by fellow travellers, and in the same breath you say "We should be figuring out how to get the attention of these venues." Well, not like this.

    Ibbotson and Tomasello have devoted some of their time to writing a piece on their research program for a general audience. Quite likely, they just got the attention of this venue by writing to them with a constructive suggestion for a piece. The same can be done for Scientific American or any of the other venues you mention. These outlets are not unreachable, and they are usually happy when academics are willing to make an effort to explain their work to a broader audience. So there you have it: a recipe for less negative energy and more constructive media coverage of language.

    By the way, I'm not a fan myself of the anti-UG framing of the Ibbotson and Tomasello piece. It detracts from the empirical results and makes them unnecessarily subservient to an ideological fight. It is easy to see their work as mapping out some of the things that have to be in what Jerry Bruner has called a language acquisition support system (LASS), quite distinct from UG or LAD but just as important for explaining how children learn language. Once you see that, it becomes clear that the accounts are not as exclusive as they are usually made out to be.

    1. I think Norbert is taking issue with the claim that people like Ibbotson and Tomasello are "well-informed" when it comes to language. And in doing so, I believe, he is actually giving them the benefit of the doubt, because the alternative would be much more insidious. Let me explain.

      Let's take the example that David Pesetsky brought up. V2 phenomena show evidence for completely productive, dynamically-built constituency, that then forms the input to displacement. (And, as David points out, V2 is found in languages that are genealogically and areally unrelated.) Anyone proposing that the capacity for language arises from formulaic expressions of one sort or another must explain how you get from that to productive, dynamically-built constituency and displacement.

      Now, with respect to I&T, there are two options. Either they are not aware of the implications of V2 for the syntax of human languages, or they are. If it is the former, then lumping them with Frenkel is not so wrong, because they are both making claims about how something arises without being well-informed about that something.

      The alternative is that they are aware of V2 and its implications and are pretending not to know. That would be much worse – hence, benefit of the doubt.

    2. @Mark: I would like to appropriate Omer's comment as my own. T has long argued that his acquisition work argues against the Chomsky program of FL/UG. Indeed, this has been the main repeated conclusion of his research. As such, it is fair to ask whether his work addresses the core issues that the FL/UG program tries to answer. The answer is it does not. This has been repeatedly pointed out, but it has zero effect on the claims. Pesetsky's comment in Asya's post makes the core point again (which, to date T has not addressed): how to get from T's proposed Gs to adult ones. How do the kids using social cues, eye gaze or whatever generalize to the Gs that WE KNOW they acquire. This is not a small point, yet despite its having been made countless times, T stays mum. There are three possibilities. Either T doesn't understand the criticism, or he doesn't understand the facts or he has nothing to say. All three options mean that his remarks are disingenuous at best and…, well Omer is right, much much worse.

      Look, I have beefs with the Chomsky program, as does Adger, as does Omer as does David P as does…However, there is a way to play this game and T has refused to do so. This means that his stuff is not serious. There might be results in his work worth considering (though I am more skeptical than you are about this). However, the general conclusions regarding UG and the success of the general Chomsky program, which is what the Guardian piece highlighted and to which we are responding, is unfounded junk. It is too bad that bright people do this sort of thing, but it is even worse when bright people pretend it is not being done.

      Last point: I too have nothing against looking into LASS. In facet some of my colleagues to this regularly (Jeff and Colin and their students) with insight. But as you noted, the article was "framed" in a very particular way, one that suggests that LASS obviates the need for something like FL/UG. Now nobody that I know thinks that there are extra linguistic systems at play in real time acquisition. Let me repeat NOBODY I KNOW…So, the only contentious issue I&T raise regards the standing of FL/UG. I know this. You know this and I&T know this. Hence the appropriate severe push back. If you want to forget about the junk and just talk about LASS then I am all ears. I should add that it is at least contentious that T is on the right track even about this, but this is for another time.

    3. Look, I'm not here to defend Tomasello and while we disagree on the merit of that line of work, we do agree that the framing isn't optimal. If you ask me, the tired UG/not UG debate is well past its best by date; it is only a slight exaggeration to say that it is kept alive mostly by old and embittered types. Many others vote with their feet and choose to focus on exciting questions not defined or constrained by tribal allegiances.

      Anyway, on the topic of linguistics and the media, I guess I just don't get the penchant for doom and gloom. What attracts cranks like Frenkel to language is the same thing that makes it such a tremendously promising topic for general audiences: it is experience-near and endlessly fascinating to everyone. If there is a need for decent media coverage, there are better ways to fix that than whining about what goes wrong. The biggest question is why not more language scientists jump at the chance to make a positive contribution by, say, writing a piece for The Guardian or pitching a topic at Scientific American.

    4. Good, so we won't defend T. BTW, I agree with the gist of Adger's remarks below (for the record). As for the "tired" debate; here we differ. There are many reasons to study language. But for my money the only reason to study language is to understand FL and UG. You are right that this view seems quaint now. I take this to mean that the field has lost its way. I assume on this we disagree. BTW, people are pitching this to Sci Am and the Guardian as we speak. We shall see.

  4. Just on Tomasello/Ibbitson - I agree that these guys are serious scientists and hence not in the same ballpark as Frenkel. And, to be honest, I'm fine with them pursuing what they think will be fertile research questions - as I've said on may occasions I think language is complex, and am happy people are tackling various aspects of it from various directions. I'm sure we'll learn things from these approaches, and I've always believed that generative grammar is just one part of our understanding of language (though I think it's an important part). But what I don't think is reasonable behaviour is to reprise, in the popular press, misinterpretations of concepts from generative grammar which have been corrected many times; which is what this article does.

    My own view of why many people without any linguistic training, like Tomasello, do this, is that they actually just don't know very much about language at any level of detail, so their abstract generalisations about language are not based on rich theories of empirical facts, but on `common sense' perspectives. I don't think one has to be a linguist to have real engagement with linguistic research, many psychologists, both anti UG (say Brian Macwhinney) and pro ( e.g. Pinker) have incredibly impressive understanding of the issues and deep theoretical knowledge. But I've read a lot of Tomasello now, and I see no evidence that he knows much about how languages work, either at a descriptive or theoretical level. I find it a bit rich that he says of generative grammar that the years haven't been kind to it, when his own proposals are incapable of dealing with well established findings emerging from that programme of research. In fact, that is basically the point of my Frontiers paper with Peter Svenonius ( No amount of reliance on shared intentionality or communication, or whatever is going to get you the cross-linguistic conditions on bound variable anaphora. Common sense just ain't gonna cut it.