Linguistic agnotology continued

David Pesetsky, eloquent as always, says the following in remarks that I lift from the comments section here:

We are not talking about instances of honorable disagreement over high-level scientific hunches and interests. We are talking about a plague of work that also violates the most minimal standards of factual accuracy and logical thinking at every level of discussion — low-level as well as high-level. This is a point that has surfaced in some of Norbert's columns, but I think has not been sufficiently stressed. The work that's getting tagged here with the j-word doesn't merely fail to appreciate, say, the subtle logic of Poverty of the Stimulus arguments, but also screws up simple facts about the languages it mentions, misrepresents the literature, fails to support claims with argument, and worse — yet gets published in a high-profile (usually field-external) venue and blurbed by the press. I don't think there's a slippery slope to worry about. The line seems quite clear, and if we care about the future of our field, we can't afford to pretend otherwise.



That said, it's not clear how best to fight the plague, nor how to restore the health of the field in its wake. 

The real cure is education. The public does not have to know or care about the details of the latest research, but they should at least know that words and sentences have structure, know the difference between letters and sounds (how many times have undergraduate intro students told me "Chinese is not a phonetic language"?), that language acquisition is an intricate puzzle, and that there are smart people called linguists who study this stuff for a living. They don't, which is why they are easy prey for linguistic nutsiness. Sadly, "public" in this case includes not only the average person on the street but also our colleagues in other fields. And we're not likely to see linguistics in every high school (where it truly belongs) any time soon.

Which brings us back to the discussion in progress on this blog, but hopefully with the notion of a slippery slope put to rest. I can be sympathetic to Dan's worries, but I think they are the least of our problems at the moment.

I agree almost completely with David P’s remarks (though I personally do not find the logic of PoS arguments particularly “subtle”). Oddly however, I find myself more sanguine than David P concerning how to rectify matters. It may take a little time (ok, maybe a lot of time) and a little work (ok, maybe a lot of work), but I think that the path forward is pretty clear and the goal attainable, at least in large part. But before I provide suggestions, let me start with a bit of diagnosis.

There are three different “publics” that need addressing in different ways. There is the general public (GP), the wider scientific public (WSP) and what we might term the “near abroad” (NA), linguistics’ scientific neighbors (e.g. psych, neuro, computational).

The general public could care less about our parochial disagreements or the filigree details of our latest proposals, as David P notes above. However, IMO, they are actually quite open to the big general GG ideas. GP is fine with the idea that kids come pre-packaged with all sorts of cognitive stuff that makes them little geniuses (The Onion provides great evidence for this here). Certainly people like Liz Spelke are big hits in the popular press and she is a pretty heavy nativist.  So, as far as GP is concerned, nativism of the GG variety is not considered conceptually out of bounds. Indeed, what parent (well at least one that is not a psychologist) watching a child spring into language could doubt that the process is backed by a heavy dose of innate knowledge.

Moreover, the GP is fascinated by language and its peculiarities. This is most evident in the fascination with words, but it is far more general. The GP loves the idea that languages show all sorts of quirks and that there are things that are perfectly coherent conceptually that are nonetheless very hard to say. I speak from personal experience here. Try ‘wanna’ contraction out on friends and neighbors or ‘fanfuckintastic.’ The GP eats this stuff up.

So, there is no natural hostility towards big GG conclusions or the kind of detailed work we like to play with. However, there is a third factor that may put GP off. In a word, ‘Chomsky,’ but not because of his scientific views. It’s just a fact that Chomsky is primarily known to GP as a political figure, and not a popular one with many. Given that many don’t like the views he defends (and this is especially true for many public intellectuals) many are predisposed not to like any of his views, including his scientific linguistic ones.  Nor is this a problem merely with the GP. The GSP is similarly inclined to be wary of any views associated with Chomsky if they don’t like his political views (and there are many people like this out there). I personally do not think that there is much that we can do about this (save observe that the two are quite separate, as Chomsky himself often stresses), but it is worth keeping in mind when strategizing.

What more can we cay about the GSP? They too are receptive enough, especially if GG ideas can be associated with a non-Chomsky personality, like, for example, Steven Pinker. The great success of The Language Instinct demonstrated, at least to me, that the wider public, both GP and WSP, are perfectly happy to jump on the language nativist bandwagon and consider and be amazed and entertained by the flora and fauna of language structure.

If this is right, then there is no predisposition in the wider public to dislike our findings or our basic research. Quite the contrary. The question then is how to package our stuff for this audience.  There is one great model for how to do this that we can steal from another area. Stephen Jay Gould was a master at combining Animal Planet cuteness with evolutionary instruction. Pinker did a very credible imitation of this for language and linguistics in his book, as did Ray Jackendoff in his short and useful Patterns book. However, we linguists tend not to do this much, or at least not enough. It would be nice, very nice, if the Language Log did more of this kind of thing (e.g. combine some neat facts from language X and show how they bear on how linguistic minds are structured), or if this could appear in some other venue with a wide-ish scientific and lay readership. This sort of pop science writing could and should be completely “positive,” no vigorous criticism of junk views required or desired. Just show off what we do so that a non-expert could follow it. The GP and WSP would lap this stuff up if done right.

It is worth observing that there is currently an LSA initiative to support a pop science writer to produce these kinds of articles for dissemination to the general public. I am delighted to report that UMD is helping with this money-wise. I think that every department should participate, so please encourage yours to donate what it can to this worthy effort.

Ok, this brings us to the “near abroad,” the NAs. This is the most hostile audience for the kinds of things we have done and discovered. IMO, you may have guessed, this is because these areas are still hotbeds of vulgar empiricism. They just know that the kind of nativism that our results support must be false so there is no reason to take what we do seriously. It is important to appreciate how little of the “debate” with the NA is substantive. The problem with much of the language work in this area is that it is, at best, at right angles to what we are interested in (at worst it is, well, to use David P’s polite locution, “j-word”). So, for example, the work on language acquisition forever shies away from the hard problems where PoS issues gain traction. There is never much of a discussion on how, e.g., ECP effects or island effects or fixed subject effects are “acquired.” There is endless discussion on word learning (a great topic, but not one where there is an obvious dearth of data).[1] The problem then is not that we present one set of solutions to a problem and they another. The problem is that the problems that we give solutions to are not even addressed by them, ever! This would not be so bad were this recognized. But it is not. Despite offering solutions to problems that are entirely different from the ones we identify, the NAs regularly and confidently assure themselves and the world at large that the problems they are “solving” also crack those we are interested in. This is just plain false, and the only way to deal with it is to demonstrate this again and again and again.  Pointed criticism, now a lost art among linguists it seems, needs to be directed against this stuff in order to expose its weakness and draw the opposition into the field of debate.  We cannot overpower them politically. But we can do so intellectually, and it can have an effect over time. Forcing people to defend the indefensible is a good way of discrediting their views. This worked before, and with patience, it can work again. At any rate, this stuff is not going away. There seems to be an inexhaustible market for this stuff among the NAs (for the latest coming attractions see here) and it gains credibility when it is not rebutted.[2]

The NA attitudes noted above also serve to underwrite other more severe kinds of junk, and here I mean stuff like Evans and Everett within linguistics. How so? Well, there they are a ready audience for the kind of “GG (viz. Chomsky) has failed” message that they love to traffic in. The NAs need GG to be a failed program. If it is, there is no reason to pay attention. There is thus a ready market for the real junk stuff and so its eternal recurrence is hardly a surprise.  The fact that it is scientifically disreputable in the ways that David P has noted does not stop it from functioning in a way congenial to a large class of scholars. And this is why it seeps consistently into the NA literature and mind set. To NAs, it’s comforting to hear that there are no universals, that linguistic data is suspect, and that we can reduce all of syntax to some nebulous conception of information structure because then it can be safely ignored with a good conscience.

So how to deal with the NA? If the diagnosis above is right, namely that is based on a deep commitment to Empiricism and Associationism, the only way to deal with this is frontal attack. These ideas need to be discredited. They need to be exposed as hopeless for dealing with he problems we have identified. Note that this does not mean that we need reject the technology they deploy. There is nothing inherently wrong with stats and probabilities and fretting over experimental design details. However, too often this technology is coupled with a very impoverished theory of mind (and science). Linguistics is one of the poster children for demonstrating how sterile the Empiricist-Associationist axis is. This point, it appears, cannot be repeated enough. We need to engage in the polemics that Chomsky and Fodor and Bever and Gleitman and many others perfected in the early days of GG.

Nor is this only to advance the big picture ideas that I favor, I should add. GG’s impressive empirical results rely on a picture like this being more or less correct. GG has built a large empirical body of doctrine which relies on the adequacy of certain technical tools. This work has demonstrated that syntactic structure matters.  When details are considered, it is not a staggering leap to the obvious conclusion that children come packed with lots of innate knowledge that makes it possible for them to acquire their native Gs. Indeed, it is not hard to produce arguments suggesting that a chunk of this knowledge is domain specific. Our best empirical work presupposes that something along these lines is correct. We debate details (hence local skepticism is almost always warranted) but we all accept an overarching framework at odds with the accepted wisdom of (much of) the NAs, so this framework needs to be vigorously defended and its opposite vigorously criticized.

Will such criticism work? Frankly, I don’t see why not. It did once before, after all. More recently, those who have followed the debate about Bayes in the psych journals over the last couple of years will have noticed that frontal attack invites defense and public airing of contentious issues. IMO, this debate has been very fruitful. I think the same airing would be effective more widely. It is only to our advantage to provoke a response from the NA. It’s far better than being ignored or it being taken for granted that GG is effectively dead.

---

[1] Actually, this is not quite true. The kind of data that Chomsky mentions every now and then concerning lexical meaning is very subtle. But this is not the kind of data that psychologists or machine learning theorists or neuro types worry about. They worry about how we categorize hit as a transitive verb. There are hard issues here, but the absence of relevant data is not obviously one of these. The contrast with, e.g. Principle C effects, is clear.

[2] I am told that Frontiers is not the worst journal in psycho so this is not being done at some backwater venue. Nope. It’s the best and the brightest that gobble this stuff up.