Monday, December 14, 2015

Brains do linguistic hierarchy

I was going to do something grand in praise of the paper I mentioned in an earlier post by Nai Ding, Lucia Melloni, Hang Zhang, Xiang Tian and David Poeppel (DMZTP) in
Nature Neurosceince (here). However, wiser heads have beaten me to the punch (see the comments sections here). Still, as Morris Halle once noted, we here discuss not the news but the truth, and with a mitigated version of this dictum in mind, I want to throw in my 2 cents (which in Canada, where I am writing this now, would amount to exactly 0 cents, given the recent abandonment of the penny (all amounts are rounded to nearest 0)). So here is my summary judgment (recall, I AM NO EXPERT IN THESE MATTERS!!!). It is the best neurolinguistics paper I have ever read. IMO, it goes one step beyond even the best neuro-ling papers in outlining a possible (as in ‘potential’) mechanism for a linguistically relevant phenomenon. Let me explain.

The standard good neuroling paper takes linguistically motivated categories and tries to localize them in brain geography. We saw an example of this in the Frankland and Greene paper wrt “theta roles” (see here and here) and in the Pallier et. al. paper for Merge (see here). There are many other fine examples of this kind of work (see comment section here for other many good references)[1]. However, at least to me, these papers generally don’t show (and even don’t even aim to show) how brains accomplish some cognitive task but try to locate where in the brain it is being discharged. DMZTP also plays the brain geography game, but aims for more. Let me elaborate.

DMZTP accomplishes several things.

First, it uncovers brain indices of hierarchy building. How does it do this? It isolates a brain measure of on-line sentence parsing, a measure that “entrains” to (correlates with) to linguistically relevant hierarchy independently of prosodic and statistical properties of the input. DMZTP assume, as any sane person would, that if brains entrain to G relevant categories during comprehension then these brains contain knowledge of the relevant categories and structures. In other words, one cannot use knowledge that one does not have (cannot entrain to data structures that are not contained in the brain). So, the paper provides evidence that brains can track linguistically significant categories and rationally concludes that the brain does so whenever confronted with linguistic input (i.e. not only in artificial experimental conditions required to prove the claim, but reflexively does this whenever linguistic material is presented to it).

Showing this is no easy matter. It requires controlling for all other sorts of factors. The two prominent ones that DMZTP controls for are prosodic features of speech and the statistical properties of sub-sentential inputs. Now, there is little doubt that speech comprehension exploits both prosodic and statistical factors in parsing incoming linguistic input. The majority opinion in the cog-neuro of language is that such features are all that the brain uses. Indeed, many assume that brains are structurally incompatible with grammatical rules (you know, neural nets don’t do representations) that build hierarchical structures of the kind that GGers have been developing over the last 60 years. Of course, such skepticism is ridiculous. We have scads of behavioral evidence that linguistic objects are hierarchically organized and that speakers know this and use this on line.[2] And if dualism is false (and neuro types love to rail against silly Cartesians who don’t understand that there are no ghosts (at least in brains)), then this immediately and immaculately implies that brains code for such hierarchical dependencies as well.[3] DMZTP recognizes this (and does not interpret its results Falkland&Greenishly i.e. as finally establishing some weak-kneed hair brained linguistic’s conjecture). If so, the relevant question is not whether this is so, but how it is, and this resolves into a series of other related questions: (i) What are the neural indices of brain sensitivity to hierarchy? (ii) What parts of the brain generate these neural markers? (iii) How is this hierarchical information coded in neural tissue? and (iv) How do brains coordinate the various kinds of linguistic hierarchical information in online activities? These are hard question. How does DMZTP contribute to answering them?

DMZTP shows that different brain frequencies track three different linguistically relevant levels: syllables, phrases and sentences. In particular, DMZTP shows

that cortical dynamics emerge at all timescales required for the processing of different linguistic levels, including the timescales corresponding to larger linguistic structures such as phrases and sentences, and that the neural representation of each linguistic level corresponds to timescales matching the timescales of the respective linguistic level (1).

Not surprisingly, the relevant frequencies go from shorter to longer. Moreover, the paper shows that the frequency responses can only be accounted for by assuming that the brain  exploits “lexical, semantic and syntactic knowledge” and cannot be explained in terms of the brain’s simply tracking prosodic or statistical information in the signal.

The tracking is actually very sensitive. One of the nicest features of DMZTP is that it shows how “cortical responses” change as phrasal structure changes. Bigger sentences and phrases provide different (yet similar) profiles to shorter ones (see figure 4). In other words, DMZTP identifies neural correlates that track sentence and phrase structure size as well as type.

Second, DMZTP identifies the brain areas that generate the neural “entrainment” activity they identified. I am no expert in these matters, but the method used seems different from what I have seen before in such papers. They used “intracranial cranial” electrodes (i.e. inside brains!) to localize the generators of the activity. Using this technique (btw, don’t try this at home, you need hospitals with consenting brain patients (epileptics in DMZTP’s case) who are ready to allow brain invasions), DMZTP shows that the areas that generate the syllable, phrase and sentence “waves” spatially dissociate.

Furthermore, they show that some areas of the brain that respond to phrasal and sentential structure “showed no significant syllabic rate response” (5). In the words of the authors:

In other words, there are cortical circuits specifically encoding larger, abstract linguistic structures without responding to syllabic-level acoustic features of speech. (5)

The invited conclusion (and I am more than willing to accept the invitation) is that there are neural circuits tuned to tracking this kind of abstract linguistic information. Note: This does not imply that these circuits are specifically tuned to exclusively tracking this kind of information. The linguistic specificity of these brain circuits has not been established. Nor has it been established that these kinds of brain circuits are unique to humans. However, as DMZTP clearly knows, this is a good first (and necessary) step towards studying these questions in more detail (see the DMZTP discussion section). This, IMO, is a very exciting prospect.

The last important contribution of the DMZTP lies in a speculation. Here it is:

Concurrent neural tracking of hierarchical linguistic structures provides a plausible functional mechanism for temporally integrat­ing smaller linguistic units into larger structures. In this form of concurrent neural tracking, the neural representation of smaller linguistic units is embedded at different phases of the neural activity tracking a higher level structure. Thus, it provides a possible mechanism to transform the hierarchical embedding of linguistic structures into hierarchical embedding of neural dynamics, which may facilitate information integration in time. (5) [My emphasis, NH]

DMZTP relates this kind of brain wave embedding to mechanisms proposed in other parts of cog-neuro to account for how brains integrate top-down and bottom-up information and allows for the former to predict properties of the latter. Here’s DMTZP:

For language processing, it is likely that concurrent neural tracking of hierarchical linguistic structures provides mechanisms to generate predictions on multiple linguistic levels and allow interactions across linguistic levels….

Furthermore, coherent synchronization to the correlated linguistic structures in dif­ferent representational networks, for example, syntactic, semantic and phonological, provides a way to integrate multi-dimensional linguistic representations into a coherent language percept just as tempo­ral synchronization between cortical networks provides a possible solution to the binding problem in sensory processing. (5-6)

So, the DMZTP results are theoretically suggestive and fit well with other current theoretical speculations in the neural literature for addressing the binding problem and for providing a mechanism that allows for different kinds of information to talk to one another, and thereby influence online computation.

More particularly, the low frequency responses to which sentences entrain are

… more distributed than high-gamma activity [which entrain to syllables, NH], possibly reflecting the fact that the neural representations of different levels of linguistic structures serve as inputs to broad cortical areas. (5)

And this is intriguing for it provides a plausible way for the brain to use high level information to make useful predictions about the incoming input (i.e. a mechanism for how the brain uses higher level information to make useful top-down predictions).[4]

There is one last really wonderful speculation; the oscillations DMZTP has identified are “related to intrinsic, ongoing neural oscillations” (6). If they are, then this would ground this speech processing system in some fundamental properties of brain dynamics. In other words, and this is way over the top, (some of) the system’s cog-neuro properties might reflect the most general features of brain architecture and dynamics (“the timescales of larger linguistic structures fall in the timescales, or temporal receptive windows that the relevant cortical networks are sensitive to”). Wouldn’t that be amazing![5] Here is DMZTP again:

A long-lasting controversy concerns how the neural responses to sensory stimuli are related to intrinsic, ongoing neural oscillations. This question is heavily debated for the neural response entrained to the syllabic rhythm of speech and can also be asked for neural activity entrained to the time courses of larger linguistic structures. Our experiment was not designed to answer this question; however, we clearly found that cortical speech processing networks have the capacity to generate activity on very long timescales corresponding to larger linguistic structures, such as phrases and sentences. In other words, the timescales of larger linguistic structures fall in the timescales, or temporal receptive windows that the relevant cortical networks are sensitive to. Whether the capacity of generating low-frequency activity during speech processing is the same as the mechanisms generating low-frequency spontaneous neural oscilla­tions will need to be addressed in the future. (6)

Let me end this encomium with two more points.

First, a challenge: Norbert, why aren’t you critical of the hype that has been associated with this paper, as you were of the PR surrounding the Frankland & Greene (F&G) piece (see here and here)? The relevant text for this question is the NYU press release (here). The reason is that, so far as I can tell, the authors of DMZTP did not inflate their results the way F&G did. Most importantly, they did not suggest that their work vindicates Chomsky’s insights. So, in the paper, the authors note that their work “underscore the undeniable existence of hierarchical structure building operations in language comprehension” (5). These remarks then footnote standard papers in linguistics. Note the adjective ‘undeniable.’

Moreover, the press release is largely accurate. It describes DMZTP as “new support” for the “decades old” Chomsky theory that we possess an “internal grammar.” It rightly notes that “psychologists and neuroscientists predominantly reject this viewpoint” and believe that linguistic knowledge is “based on both statistical calculations between works and sound cue structures.” This, sadly, is the received wisdom in the cog-neuro and pysch world, and we know why (filthy Empiricism!!!). So, the release does not misdescribe the state of play and does not suggest that neuroscience has finally provided real evidence for a heretofore airy-fairy speculation. In fact, it seems more or less accurate, hence no criticism from me. What is sad is the noted state of play in psych and cog-neuro, and this IS sad, very very sad.

Second, the paper provides evidence for a useful methodological point: that one can do excellent brain science using G theory that is not at the cutting edge. The G knowledge explored is of Syntactic Structures (SS) vintage. No Minimalism here. And that’s fine. Minimalism does not gainsay that sentences have the kinds of structures that SS postulated. It suggests different generative mechanisms, but not ones that result in wildly different structures. So, you out there in cog-neuro land: it’s ok to use G properties that are not at the theoretical cutting edge. Of course, there is nothing wrong with hunting for Merge (go ahead), but many questions clearly do not need to exploit the latest theoretical insight. So no more excuses regarding how ling theory is always changing and so is so hard to use and is so complicated yada yada yada.

That’s it. My 2 cents. Go read the paper. It is very good, very suggestive and, oddly for a technical piece, very accessible. Also, please comment. Others may feel less enthralled than I have been. Tell us why.

[1] I would include some recent papers by Lyna Pylkkanen on adjectival modification in this group as well.
[2] These are two different claims: it could be that the linguistic knowledge exists but is not used online. However, we have excellent evidence for both the existence of grammatical knowledge and its on-line usage.  DMZTP provides yet more evidence that such knowledge exists and is used online.
[3] P.S. Most who rail against dualism really don’t seem to understand what the doctrine is. But, for current purposes, this really does not matter.
[4] Note, the paper does not claim to explain how hierarchical information is coded in the brain. It might be that it is actually coded in neural oscillations. But DMZTP does not claim this. It claims that these oscillations reflect the encoding (however that is done) and that they can be used to possibly convey the relevant information. David Adger makes this point in the comments section of the earlier post on the DMZTP paper. So far as I can tell, DMZTP commits no hostages as to how the G information is coded in brains. It is, for example, entirely consistent with the possibility that a Gallsitel like DNA coding of this info is correct. All the paper does is note that these oscillations are excellent indices of such structure, not that they are the neural bases of this knowledge.
[5] Here’s a completely wild thought: imagine if we could relate phases to the structure of these intrinsic oscillations? So the reason for the phases we have is that they correspond to the size of the natural oscillations which subvene language use. Now that would be something. Of course, at present there is zero reason to believe anything like this. But then again, why exactly phases exist and are the ones there are is theoretically ungrounded even within linguistics. That suggests that wild speculation is apposite.


  1. I'm happy to join you in praising the paper. I can't say the same for the press release, though. The opposition between linguists who argue that syntax exists and psychologists who argue that instead language "is based on both statistical calculations between words and sound cues to structure" is not even false - it's nonsensical. What was the writer even trying to say? Sound cues to structure? So those evil psychologists do believe in structure after all?

    1. I believe that the point was ill-phrased. It does seem to me correct that both psycho and neuro types have resisted the idea that competence involves having an internalized G and so it is accurate to say that they "contend … that our comprehension does not result from an internal grammar." In this context, I think your quote makes sense.

      Last point: I may be wrong, but I have the very strong impression that it has been a point of pride in many cog-neuro-psych circles that brains cannot embody Gs or rules at all and so that anything like a system of knowledge of the kind we think exists is actually impossible for brains to possess. This was the old Connectionist view and the position that Fodor and Pylyshyn and Gallistel have argued so cogently against. Thus showing that Gs matter is a kick in the pants for this view. However, if you are right and that psych and neuro people are changing their minds, then this is excellent news. Great.

    2. I think we've had this discussion before... Perhaps what we need to do is look through the last five years of work on language in journals like Cognition, Cognitive Science, Journal of Memory and Language, etc, and see what proportion of those papers adopt a connectionist or more generally anti-hierarchical view or language. My suspicion is that we'll find that not that many do. You might be right that some people who don't actually work on language are poorly informed enough to think that you can understand Wh-dependencies using "statistical word cues", whatever that means (transition probabilities?).

    3. I do hope you are right, though I have the impression that there is a general consensus out there in the non-ling world that GG has been proven wrong and that even the idea that one can get away with just looking at string properties is considered an idea worth taking seriously. I discussed some of this stuff in the blog (e.g. and I am told that these authors are not nebbishes in the psycho world. And as we both know, there are others. But, that said, I hope you are right and these views are out of date.

    4. I wasn't suggesting that psychologists have embraced the Chomskyan tradition as a whole, I was referring specifically to the DMZTP claim that "there is a grammar in our heads". If your point was that many psycholinguists do not endorse GG more generally then you might be correct.

      For what it's worth, my sense is that the Frank, Bod and Christiansen you linked to that denies the cognitive reality of hierarchy in language does not represent a mainstream view in psycholinguistics - but again, I haven't conducted an empirical survey.

  2. So, you out there in cog-neuro land: it’s ok to use G properties that are not at the theoretical cutting edge. Of course, there is nothing wrong with hunting for Merge (go ahead), but many questions clearly do not need to exploit the latest theoretical insight. So no more excuses regarding how ling theory is always changing and so is so hard to use and is so complicated yada yada yada.

    I think this is a valuable point that probably needs to be made more often. I suspect that some of the blame for the fact that this point is not widely appreciated lies with us linguists, however: granted, it's often pointed out that the latest minimalist theory is just a natural development of the program initiated in the 50s and 60s, but it's also relatively common to read sensational-sounding claims that the modern theories have (for example) "eliminated phrase structure rules", which perhaps unnecessarily obscure Norbert's point.

    1. I agree that this point should be made loud and often. This relates a bit to the Grain issues that Embick and Poeppel discuss. For the kinds of questions cog-neuro-psych techniques are asking, it really doesn't generally matter which vintage theory is adopted. And that is what we should expect. But you may be right that this needs saying to ourselves as well as to others. So:

      It’s ok to use G properties that are not at the theoretical cutting edge. Of course, there is nothing wrong with hunting for Merge (go ahead), but many questions clearly do not need to exploit the latest theoretical insight. So no more excuses regarding how ling theory is always changing and so is so hard to use and is so complicated yada yada yada.

  3. On the matter of the press release, here's what Pinker had to say: "Chomsky's right" NYU press headline reveals harmful polarization in linguistics. Grammat. processing ≠ Chom; shouldn't be so controversial.

    I would agree; and if you really believe the majority of neuroscientists, psycholinguists or otherwise cognitively interested language scientists don't see a role for the kind of basic nested levels of structure investigated in this paper, it seems you're providing an excellent demonstration of how polarization can be harmful. Polarization = putting your blinkers on.

    1. @mark: Ok, you've mystified me. What is Pinker saying here? That linguists believe that Gram Processing does equal Chom? What's this mean?

      As for putting blinkers on: damn right I do. I try to not be influenced by stuff that I am quite confident iw wrong and a waste of time. Included in this are views regarding psyching that find little place for G knowledge. The latter does not exhaust what goes into ling performance (including processing) but is an integral part of that process. Those that deny this are, IMO, so far off the track of the respectable that paying them attention is a serious waste of time, even a bit irresponsible intellectually. So blinker away!

      But, really what is Pinker trying to say? I have no clue.

    2. I take Pinker to be saying that framing the "linguistically modest" contribution of this paper in terms of some deep-seated dispute between Chomsky and the rest (as the press release does) is an example of harmful polarization. The work is relevant to all linguists; the Chomsky clickbait technique, whether pro or anti, is getting old.

    3. I guess I disagree with the great man (and I guess with you). There is a lot of noise out in the cog-neuro world announcing the demise of Generative Grammar. I have reviewed many of these obits in the pages of FoL. I am quite certain that this view is widespread. Thus, noting that these death notices are, at best, premature is of some importance if the cog-neuro community is to take this kind of work seriously.

      Now, are the results being discussed special to Chomsky's current theories? Nope. As the authors indicate, the material relies on the Syntactic Structures theory, and this, you are quite right to observe, has been assimilated into every branch of linguistics, though not, as I have noted, into the general thinking of the cog-neuro community (alas). This the article makes clear, as does the press release quite explicitly in the third paragraph. So, it is not particularly misleading.

      There is one more possible point you are trying to make: that Chomsky should not be associated with his own views when they are being explicitly referred to for it makes it sound like it is his ideas that are being investigated. As I write this, it sounds absurd. Why doesn't it sound absurd to you and Pinker? Neither the paper nor the PR release suggest that ONLY Chomsky's views are at issue here. But clearly views not dissimilar to his ARE. But I agree with one sentiment. I would have preferred that the authors and the release note that it is Hornstein's views that are vindicated, or Kayne's or Grimshaw's or Polinksy's or Postal's or all of our views for that matter. Maybe I read things wrongly, but neither the paper nor the release invidiously distinguishes among various stripe of GG LINGUIST. The objects of criticism are quite explicitly "neuroscientists and psychologists" as they "predominantly reject this [i.e. Syntactic Structures] viewpoint." As I find this to be a pretty accurate reflection of THOSE fields, I find Pinker's comment silly. As he is not a silly man, I also wonder what led him to make it, but I will refrain from speculation here.

    4. Chomsky's right, but so are Zellig Harris (1946) From Morpheme to Utterance (reasonably explicit PSGs with some informalisms), Bloomfield (1942) Outline of Ilocano Syntax (verbally rather than notationally presented PSG) and of course Reed and Kellogg from the late 19th century (a kind of RG with hierarchy but no coherent story about linearization).

    5. @avery. I think Harris and the others would be spinning in their graves at the notion of a `grammar in the head'! Sure they believed there were hierarchical structures, but not that these were mentally `real'. That was definitely a contribution made by Chomsky, and was a huge source of controversy at the time, at least if the various papers and reviews are to be trusted (cf also first few chapters of Randy Harris's Linguistics Wars book).

    6. I wouldn't be so sure, 'the mind' and 'mentalism' were not quite for them what they are for us, since they lacked well understood notions of physically realizable computations, and their concerns were more of methodology (don't speculate about things that you have no way to find out about) than of what might or might not be happening in the head, where they lacked tools to look.

    7. I don't see why Harris et al. would have a problem accepting that new-fangled methods now allow us to show that when people hear speech, their observable brain activity oscillates in ways that correspond to nested levels of linguistic structure like syllable/mora, phrase, and sentence. But that's a long way away from Chomsky's signature rationalist claim that language is a "mental object" (which they would have more trouble with).

      When the press release invokes Chomsky and exclaimes "we do have a grammar in the head", it frames the paper as if it directly supported a rationalist view of language as a mental object, obscuring the fact that the actual findings are perfectly compatible with many empiricist accounts of language structure, which do of course care for levels like syllable, phrase and sentence and which would not be surprised at all to see those correspond to observable brain patterns in language processing.

      What's really interesting about this paper is that its results are of equal interest to rationalists and empiricists (though some rationalist detractors sensing a slippery slope could be seen on the prior thread!). That alone shows that "Chomsky was right" is both overselling and underselling the real contribution of the paper.

    8. And, the study doesn't really how that there's a 'grammar' in the head, thinking of a grammar as something like anything we right down ... what it shows is that there are neurological events corresponding to aspects of the structures that (various kinds of) grammars assign to sentences. Not the same thing.

    9. @mark: It shows that processing the linguistic token is sensitive to structure that is not in the signal of that token. It requires that structure to be built. In other words, it needs a G in the head. Now, the open question is what the relevant G structures are; are they syntactic or phonological phrases or semantic units or… But what is clear is that what is being built as the sentences are attended to is NOT being built in virtue of prosodic or statistical information IN the input, as these have been controlled for. So, you need to assume that people are building the structure that these measures indicate that they are sensitive to.

      Now, how did this G get there? You might argue that the rules that ARE in the head have found their way there by inductively tracking overt properties of the earlier inputs without the benefit of UG, say. The paper did not argue against this. It simply argued that the hierarchy that was being tracked was a constructed one, and this is, of course, the Syntactic Structures position: Gs as mental objects that generate hierarchical linguistic structures of various kinds.

      I find it interesting that it is so hard for you to agree that this paper argues for the conclusion that Chomsky was right about THIS. He was. Empiricists of his generation (behaviorists) didn't think that there was need for mental representations of ANY sort. This position is still held by some. I even referred you to some of these. Given this, why the reluctance? THis does not mean that Chomsky is right about everything, nor does the paper argue that he is/was. But he is right on the money here and it is worth noting because many (you?) are reluctant to think that some basic ideas in GG are alive and kicking. Like I said before, there is a large industry arguing that GG is dead. Well maybe it is, but some if its ideas appear to be quite fruitful to pursue.

      So why exactly the reluctance to recognize this? Why not just say that Chomsky had this right and move on? Maybe it's because you sense that this does touch on THE vital issue, and though it does not solve the matter, it surely argues in favor of the view that human linguistic specific competence rests on having internalized a G is basically correct and that this has consequences. What exactly is eating you?

    10. @Avery: very fancy philosophy here. There is a G in the head in the way that people usually understand this, which is not that we will find a little book with a set of rules there. In the same way that there is WORD in my operating system, or there are solar positions in bees brains, there are rules in my head.

    11. I don't think there actually is any 'way in which people usually understand this'. The way I understand is 'my proposed G is pretty bad but hopefully not completely useless description of something in the head', but the leap from a formalized description of behavior to the structure of the cause of that behavior is a pretty big one, and not entirely innocent.

      Some contemporary people, such as I believe Geoff Pullum, claim not to understand the concept of G's in the head in any way whatsoever, so perhaps we will find out at some point what he thinks of this result; I'd be astonished if it horrified him.

    12. Nothing is ever innocent. The question is whether this is more "guilty" than what we would say about bees and the sun or birds and their food caches. My view is that we should say about Gs what we say about these, and for these the simple view that such info is coded in brians and this explains why the animals do what thye do is what we regularly find. So...

      As for what Pullum would say, I dont see the point. It's not what he would say, but what he should say, and on this matter he is no more expert than you or me.

    13. @Norbert: As Avery also points out, it is unlikely that empiricists (of Chomsky's generation) didn't think there was a need for any mental representation; it was more that their evidentiary requirements pointed them towards more directly observable behaviour first.

      No matter how much fun it may be to set fire to straw men empiricists, I don't think there is any serious question that even Chomsky's contemporaries would have been happy to accept that new methods at some point might actually give us a peek into mental processes — as they do now.

      Once again, the "linguistically modest" (your term) result of the paper is a long way away from Chomsky's signature rationalist view of language as a mental object, and perfectly compatible with many other current accounts of linguistic structure — and that is the reason the "Chomsky was right"-framing both oversells and undersells the broad interest of the paper.

    14. I'm finding Mark and Avery's positions here a little odd. I think there's no doubt that insistence on abstract linguistic structure being mentally represented in some fashion is most strongly associated with Chomsky. It's pretty anachronistic to say that Harris and others wouldn't have had a problem with mental representation - their methodological commitments completely determined their ontology, and made mental representations no-goers for them. The fact that we, these days, are pretty ok with such an interpretation of linguistic representations is to a large extent a side effect of Chomsky's (and others' who followed him and developed generative grammar) insistence on it. The press release is crystal clear that it is a decades old theory of Chomsky's that is at issue and the content of that theory is the cognitive reality of abstract linguistic representations, nothing to do with the aetiology of such representations. So this isn't an issue of rationalism vs empiricism in the aetiology of linguistic knowledge, it's an issue of abstract representations being imposed on incoming data which don't have certain kinds of informational content. Sure, if you are a construction grammarian or another kind of usage-based theorist you can just say `well, we do have an abstract grammar, we just abstracted it from statistical properties of the input and have stored it in some way that it can be used without appeal to such properties'. But my hunch is that that isn't a comfortable position for at least some advocates of such positions (Tomasello, Bybee, etc), although it is consistent with some of what they say. The reason it isn't a comfortable position, is that it requires such theorists to provide an explanation of how we got to such representations from the bare distributional/statistical input, what the constraints on the representations are, and what the final form of the system that licenses the representations is. Now some excellent people do this kind of work (Alex Clark, for example), but it's not mainstream.

      There has been some bad reporting on this, I agree (the piece in Medical Daily is awful) but the actual press release is, as far as press releases go, pretty accurate. I've seen a lot of people who seem to be saying `it's not fair to say Chomsky was right about X, because people might also think that he was right about Y, because he has argued for X and Y, and the paper is only about X'. It's true that logic is thrown out of the window in press reporting, and that some (many?) people who don't know much about this area will come away with a distorted view of the implications of the results because they'll just file it with the other vague associations they have with Chomsky in their heads. So perhaps the press release shouldn't have headlined with `Chomsky was right'. But, to be honest, about this, it looks like he was.

    15. @David: I'll stick with my position that it's reasonable to doubt that Bloomfield would have a problem with the result of this paper, on the basis that, although he certainly was an enthusiatic physicalist and antimentalist (having just quickly reread 'Language and Ideas' from 1936), the reasons were due to the complete lack of substantive content added by the use of mentalistic terms such as 'concept' at that time. Measurable neural events correlating with aspects of linguistic analysis is simply not the kind of thing he was objecting to.

      @Norbert: I think that what somebody like Pullum *would* say is interesting because for thoughtful people, which Pullum is, and both Harris and Bloomfield were, methodological thinking is not some kind of religious activity, but an attempt to make sensible judgements about what kinds of things can vs cannot be investigated with presently available methods. And Pullum seems to me to have a similar take on things to that of mid-20th century figures such as Quine and Suppes, who had issues with Chomskyan mentalism, but can no longer be asked what they make of this paper.

    16. A bit more detail on why I don't think Bloomfield would freak out when confronted by neuro evidence for phrase-structure processing, retrieved by reviewing Hockett's Bloomfield Anthology.

      1. Bloomfield was highly aware of the nature of phrase structure. Not only in the Outline of Ilocano Syntax that I mentioned above, but in smaller snippets such as the reference to 'box-within-box syntactic constructions' [Anthology p. 307, from Linguistic Aspects of Science (1935)], and "the word 'apple' enters into phrases with preceeding adjectives, and, except for certain very special expressions, is always so preceeded ('an apple', 'the apple'); the phrases so-formed enter into larger expressions with following finite verb expression ('the apple tastes sour'); [Anthology p.403, from Meaning (1843)']

      2. The anti-mentalism is well-justified by the absence of any serviceable basis for non-mystical conceptions of mental structures and processes 'they form a rigid system, - so rigid that that without any adquate physiologic information, and with psychology in a state of chaos, we are nevertheless able to subject it to scientific treatment' [Anthology 141-142, review of Jespersen's Philosophy of Grammar (1927; he clearly overestimated the rigidity).

      Furthermore, the notions of computable function etc that are central to our present day ideas about concpets, etc in the 'mind/brain' were completely unavailable at the time of the earliest quote, very recent at the time of the latest, and he suffered a career-ending stroke in 1946, and so certainly did not have enough time to develop any conception of mentality on the basis of that work (note that we only developed terms such as 'mind/brain' for some of the basic concepts in the 1980s or thereabouts).

      Therefore, I think that it's quite reasonable to think that, if resurrected, and given a week or two to get the basic ideas behind late 20th century theories of neuroscience and computation, his response to the phrase-structure results would have been 'wow, cool' (however people said that back in the 30s).

      A final intellectually irrelevant but perhaps amusing observation: he seems to have invented his own version of the Nigerian 419 scam, but not implemented it; in an account of some practical jokes he thought of, on was: "(ii) to write a letter to the Registrar or Bursar of the University, in an 'obviously Slavic' but actually faked language, clearly having to do with large sums of money that the University might get under undecipherable conditions."

  4. I also love this paper. Although we have a number of neural measures that appear sensitive to the presence vs. absence of sentence structure (the many areas that show fMRI increases for sentences vs. word lists, the work from Liina's group showing reliable MEG responses for nouns preceded by adjectives relative to nouns in isolation), so far we don't have *any* neural measures that clearly track the details of that structure. Discovering a good candidate is much more than anyone else has done and is very exciting. Of course lots more still has to be done (I am less confident that they have ruled out prosodic accounts), but if we go all these decades with no measure and then someone at least gives us a good candidate, I am giving them lots of credit.

    I will say that I am less excited/optimistic about DMZTP's mechanistic speculations that neural oscillations might be in some way encoding the syntactic structure. First, IMO they were not able to do very much to show that the effects they are observing are *not* reflecting evoked responses. In other words, imagine a bunch of neurons fire every time you close off a phrase boundary. This would inherently be an evoked response rather than an oscillatory response--it would only happen when you can close off a phrase boundary, and if you presented the phrases at very irregular intervals, you would see this evoked response at very irregular intervals. But if you presented the phrases at a regular rate, as in the current study, it would be indistinguishable (in these kind of analyses) from an oscillatory response.

    Second, it's not very clear how a structure encoding mechanism based on the kind of frequency effects observed here in the artificial case (phrases presented at a constant rate) could possibly scale up to real world sentences. In other presentations, the authors have pointed to the fact that in most real-world speech, syllable rate is actually pretty constant. But it seems impossible that *phrase* rate is going to be constant--like sometimes you are going to have a noun phrase that has a one-syllable article, a two-syllable adjective and a three-syllable noun, followed by a one-syllable verb phrase (The older computer died), sometimes it might go the other way (Thieves destroyed all of the new houses), etc etc. Put another way, there's going to be no straightforward correlate to the authors' 1Hz 'sentence' power increase in e.g. some discourse with a long sentence followed by a short sentence. (discussions with Laura Gwilliams and Chris Neufeld were what really brought this point home to me)

    Now, does this mean that having oscillations play a key role in encoding structure is ruled out? Could it be the case e.g. that membership of a word in a given NP is encoded by having all of the morphemes in that NP oscillate at the same frequency (e.g. 'the' 'old' 'er' 'computer' 'die' 'ed')? Of course. But that would bear absolutely no relationship to the cortical tracking effects reported by DMZTP--those guys could be oscillating together at 12Hz or 31Hz or whatever, but there's no reason that would need to bear any relation to the rate at which phrases are being presented, and that is a good thing (if you favor that kind of encoding scheme) because as discussed above, in perfectly easy-to-understand sentences and discourses those phrases are not going to be presented at a regular rate.

    Since I am throwing cold water on the mechanism bit, I will end by reiterating my great admiration for this paper. I think it would be *unbelievably* useful to have an implicit neural index of phrase structure for addressing lots of important questions, and for that purpose I could care less if this index is underlyingly an evoked response instead of an oscillatory response, or if it tells us nothing deep about how phrase structure relationships are neurally implemented. And I appreciate that they did not make very strong claims in the paper about either of these things.

  5. Just one point for some might misunderstand: "I am less confident that they have ruled out prosodic accounts." By this you don't mean that DMZYP failed to wipe prosody from the signal. They did. Rather you mean that what is being entrained to is CONSTRUCTED prosodic phrases rather than constructed syntactic phrases. This might be right.

    However, from the little I know, stories about prosodic phrases map them from syntactic representations. So, the prosodic competence seems to rely on the capacity for syntactic hierarchy. If this is so, then even where the subjects entraining to constructed prosodic phrases, we would have neural evidence for syntactic structure. Of course, it would be even nicer if these entrainments were directly indexing syntactic phrases.

  6. I could write a long response but don't have the time or mental energy for it. What I'll do instead is state my position without providing a lot of evidence for my views.

    I believe that Norbert really likes this paper because Ding et al. posit a neurobiological mechanism that might actually connect with a linguistic construct. However, the observed oscillations are predicted on a number of different hypotheses of the neurobiological mechanisms underlying language. So the speculation about oscillations as a mechanism only connects with their data in that the data are not incompatible with an oscillatory mechanism. Many other authors have already speculated on oscillations as mechanisms underlying linguistic representation/computation (e.g., Boeckx, who goes into far more detail about how such a mechanism might work). The data Ding et al. present does not change that picture.

    I enjoy the paper for the correlational neurobiological data it does present, which I think is informative for understanding the relation between language and brain.