More on subjacency

Peter Svenonius (once again) asks the right question and Omer (once again) has interesting things to say about it (see here). Take a look. Here is my take on the issues. Please chime in with yours.

I agree with Omer that there may not be much of a consensus right now about how to deal with successive cyclicity in detail. However, so far as I can tell, there is general agreement that it has something to do with the PIC (this is the current analogue of the subacency principle, Bounding nodes and  the domains they create). 

As you know, there are two extant versions of the PIC, the more favored one being the one wherein a complement of a phase head is rendered inaccessible at the NEXT phase (usually when the next phase head is accessed). This comes close to coding the old idea of subjacent domain (as was early observed one has access to the domain one is in and the next one (i.e. no counting)). The strong version of a phase is less in favor, but it has some charms for it would force something like a "no edge skipping" requirement on the grammar (e.g. the old Rizzi idea that one could "skip" the most immediate Comp would be ruled out). There are purported arguments against this stronger version, but they never struck me as dispositive (and there are Legate style arguments against it). At any rate, there is consensus that phases should derive cyclicity.

How closely is this tied to features, uninterpretable or otherwise? Logically speaking, not that closely so far as I can tell. The issue of features is tied to whether movement is optional or obligatory. If Greed drives movement or uninterpretability then C features might be needed. Yes if Greed is strong and no if something like uninterpretability suffices to drive one to the phase edge as a last resort (Phase balance or Boskovic's take on the same idea). There is some evidence that intermediate Cs can have features, as we know. The generalization to all languages is a standard GG move. So, the idea is not empirically nuts. Of course, WHY this should be true is unclear in the absence of something like Greed, but then maybe this is an argument for a strong version of Greed.

This goes against the current fashion. It seems that nowadays movement is free again (free at last, free at last, thank the lord, free at last!). But then there is no requirement that there be intermediate features to drive movement, nor that there be uninterpretable features on WH to force it to move. The WH moves or it does not. If it does, then it must move to the intermediate C for PIC reasons. If it doesn't then no convergence. More specifically, what one needs are language specific requirements that force a given G to have a WH up top overtly in some languages (something like the old strong feature) or some kind of Rizzi Criterion that is fulfilled in G variable ways. This seems generally assumed in current technology, so no biggie here.

There is one last idea that has been tied to successive cyclicity: Chomsky's current idea about labels. Oddly, for Chomsky, the fact that there are languages where there appears to be agreement in non WH Cs with a moving WH is a big problem. Agreement should obviate further movement. Of course one can get fancier here with different features having different effects on labeling (and so movement), but this begins to hand code in the property we want explained (not a good thing to do).

That's the way things look from where I sit. So, there are several ways of getting edge to edge movement all involving the PIC in some fashion and thereby recoding the old subjacency criterion. I want to emphasize this: this is not a new explanation but a recoding of the old one (not that this is a bad thing).

Two last points: what is less clear to me is how this all hooks up with islands. Chomsky, it seems to me, is reluctant to take islands as G-real phenomena. He seems inclined to take the view he once criticized, viz: that islands are performance residues of complexity. I am skeptical myself, but it is a logical possibility. The Sprouse stuff has convinced me that it is likely false. 

This leaves the question of how to code Islands in phases? That's easy (as anyone who has tried will attest). The problem is that the coding follows from nothing (why are D edges different from C edges? why weak PIC rather than stropping? Why transfer when next pause head chosen rather than next phase completed? Why C and D and v as phases? Why week vs strong phases?). In fact, the coding just recapitulates the machinery in classical Subjacency theory. Or, Minimalism has not given us any insight into the details of subjacency as of this date. So, islands stand as having no good deep explanation beyond the one that Chomsky already provided for Subjacency that I quoted in the body of the earlier post.

Second: we really would love to tie island effects with ECP effects as Barriers and Cinque-Rizzi tried to do. Why? Because the domains for bounding and ECP are so damn similar. It would be really odd (IMO too odd to be tolerable) were these driven by different mechanisms given that their domains are virtually identical. So, we need to find a way of finally addressing ECP questions within MP. In particular we need to find a way of unifying them in ways more conceptually acceptable than the Barriers/Lasnik-Saito theory did.

So island effects are currently no better understood within minimalist theory than they were within GB. The GB story can be smoothly translated into technologically acceptable minimalist terms, but doing so provides no insight. Moreover, some parts of the old theory, the ECP part dealing with adjuncts vs arguments and their differing locality conditions, really has not good minimalist counterparts (does anyone really thing gamma marking is part of FL/UG?). That's how I see things. You? 

Nice one line joke regarding falsification

University closed today due to some snow. While consoling myself reading some blogs I came across this great one liner in Andrew Gelman's blog. Enjoy.

The wonder of subjacency

I am currently teaching our Grad syntax 2 course and, not surprisingly, it focuses on the Minimalist Program (MP). Given my predilections (and the influence of Howard Lasnik) I find that one can best appreciate MP by starting with Government-Binding Theory (GB). Jairo Nunes, Kleanthes Grohmann and me used GB as backdrop to MP in our intro book (here). But every time I teach this course I become more and more impressed with the virtues of GB. It is a pretty neat little theory, and, for my money, it still provides the best set of analytical tools in linguistics. In fact, were I charged with the task of describing a new construction or writing the G of a language I would render it in a GB idiom, minimalist technology be damned. However, this is not what I wanted to write about here. Rather, I wanted to sing the praises on one particular sub-part of GB that dealt with a topic that has largely fallen out of research favor but that stands as one of the great scientific accomplishments of Generative Grammar (GG). The topic? Islands and Subjacency. What follows is why I consider it such an achievement.

As everyone knows, Chomsky’s aim in developing the theory of Subjacency (S) was to unify Ross’s islands, the latter having been discovered and described about a decade earlier. The locus classicus of this effort is On Wh Movement (OWM) where Chomsky lays out the story in gory detail.  Here’s a question: what did the unification add to Ross’s original discussion?

One thing it added was unification. Looked at theoretically, Ross’s islands are a motely, a list of domains opaque to movement. From the get-go, it was hard to believe that this list was what FL/UG coded. There has to be some underlying method. Chomsky’s goal was to find it. I do not actually recall his theoretical discontent being widely shared across the GG community (but, in my experience GG hardly ever suffers from the mental unease that poor theory regularly generates in Chomsky). At any rate, in unifying Ross’s islands, OWM tries to explain why the islands we find are the islands we have. In fact, OWM tries to tie the existence of islands to general computational considerations thereby providing what is, in retrospect, an excellent paradigm of Minimalist thinking. Here is what OWM says:

… the island constraints can be explained in terms of general and quite reasonable computational properties of formal grammar (i.e. subjacency, a property of cyclic rules that states, in effect, that transformational rules have a restricted domain of potential application; SSC, which states that only the most prominent phrase in an embedded structure is accessible to rules relating it to phrases outside; PIC, which stipulates that clauses are islands subject to the language specific escape hatch..). If this conclusion can be sustained, it will be a significant result, since such conditions as CNPC and the independent wh-island constraint seem very curious and difficult to explain on other grounds. (p. 89; On WH Movement, my emphasis).

So the list like nature of the islands becomes comprehensible when viewed from a more general computational perspective. And this is indeed a virtue.

But, and I want to emphasize this, this unification is not, as it stands, an empirical argument in favor of S. Taken at face value, what OWM demonstrates is that it is possible to unify Ross’s islands on a more rational basis, but just unifying them does not show that this unification is empirically fecund or justified.

Happily, the unification proved to be empirically very fertile indeed. OWM provides two ways that S logic could lead to the discovery of novel data.

First, it provides a general method for discovering which kinds of dependencies should be subject to island effects. OWM has a long and interesting discussion of comparative constructions and notes that given the nature of the unification proposed, comparatives should be formed by movement. This was somewhat unconventional at the time (though the work is based on some earlier work by Richie Kayne that argued for this conclusion). In fact, the most carefully worked out theory of comparatives (due to Bresnan) treated comparatives as products of a deletion operation, rather than as products of movement. If memory serves, there was quite a bit of very vigorous debate on this topic over the next little while, including at the UCSD conference where OWM was originally presented. This debate became quite heated and gave lowly grad students like me an appreciation of the old adage: when elephants fight what gets hurt is the grass. At any rate, this was one consequence of the unification that OWM emphasizes. 

A digression: could Ross’s analysis been used as a diagnostic of movement? This is, in effect, what OWM does. It assumes that if dependency D obeys islands yet allows unbounded dependency (btw, this second conjunct is a critical yet often ignored part of S reasoning) then the dependency must be the product of movement. Could Ross’s theory be interpreted in the same way? Not really. Recall, that for Ross, what makes an island and island is not the movement (movement out of islands was fine for Ross). Rather what makes an island is chopping the resumptive pronoun that movement leaves behind. In other words, for Ross, islands restrict chopping, not movement. For Chomsky, S restricts the movement and resumption is analyzed as a non-movement dependency precisely because it does not show island effects.[1] Given this, comparatives are a very good place to empirically distinguish Ross’s theory from S-theory. Why? Because comparatives have no apparent resumptive analogues like DP movement cases do (*John is taller than Bill is it/that/such). But if there are no resumptives then there can be no chopping and so no expectation of islands. This would make deletion the natural generative operation sub-serving comparatives. Thus OWM’s argument that comparatives are actually products of movement, was an empirical argument for the S view of islands.[2]

The second empirical argument for the OWM unification came from a crop of new islands. Thus, the OWM story implied that complex DPs should be islands for extraction. This implied that we should find subject islands (which we more or less do: *What do pictures of hang in the National Gallery) but also that objects should be islands (which is far less evident: What did Bill paint pictures of). OWM spends some time trying to get out from under the problems that object extraction creates. To the degree that it succeeds, then the predicted presence of subject islands is an empirical plus for S-theory.[3]

The third empirical argument in favor of S is by far the best and, if my recollection is correct, the most wow-inducing. It’s successive cyclic A’ movement. The unification of islands predicted that unbounded movement (movement that shows no island effects) is nonetheless derivationally bounded in that it is made up of a series of small bounded steps. Ross’s theory made no such prediction, Indeed, prior to S-theory there was no reason to believe it to be true. Unbounded dependencies were considered to be perfectly reasonable operations. S-theory implies that, at least for one class of dependencies (i.e. movement), such unboundedness is an illusion. This was (and is) a hell of an implication. There are many many languages where there is little evidence suggesting that anything like this is true (English being a good example of one). But, as we soon discovered (and by ‘we’ I mean GGers), it was TRUE (insert fireworks and brass bands here).

I was a grad student in Cambridge when the empirical evidence started trickling in. Jean Yves Pollock gave versions of the deservedly famous paper he co-wrote with Richie Kayne on stylistic inversion in French. If memory serves, Esther Torrego’s equally excellent paper was floating around when I was still a Cambridge denizen. As most now know, this trickle soon became a torrential stream of results with many languages providing overt evidence for cyclic Wh movement (Irish, Chamorro a.o.) At any rate, that this implication of S-theory was apparently true (or at least had non-obvious data that could be explained by it) was stunning. This is what good science does: its theories imply something unexpected and the unexpected turns out to be the case. It was great. And this was the evidence that really sold S-theory.

Let me emphasize the important argumentative structure: Unifying islands as in OWM implies that all movement, even that which does not manifest island-like properties, is local. Thus islands imply successive cyclic C to C movement. The discovery that this prediction holds is stunning confirmation of the unification of Ross’s islands in OWM and a strong confirmation of S-theory.

So, if anyone asks you what unifying islands brought to the table, successive C to C movement (or edge of domain to edge of next higher domain) is one of the biggies. It served a bit like the Syntactic Structures analysis of affix-hopping and do-support in that it sold S-theory with its aha effect and thereby made it widely accepted.

One last virtue: S-theory served as a bridge to other parts of cogsci. For example, S-theory had very natural interpretations in the context of parsing theories (E.g. Berwick and Weinberg) and learnability theories (e.g. Culicover and Wexler). S-theory served as a grammatical bridge to, IMO, the richest interaction between GG and other parts of cognition witnessed to date. In fact, like the income of most Americans, GG has receded from this high point, which, is really too bad.

So, what did S-theory add? It unified islands, allowed for a refinement of our understanding of movement, led to the postulation of new islands, implied that long movements were made up of short steps and served as a productive bridge to other parts of cognition.

And it has one last virtue of contemporary relevance. It serves (IMO) as and excellent  (maybe even the best) example we have of what linguistic theory should aim for. It is our poster child for for GGs scientific bona fides, which makes it odd that S-theory appears not to be a central part of the grad syntax curriculum anymore. I say this on the basis of very cursory investigation, actually just one or two discussions with recently minted PhDs. For the reasons noted above, this is too bad. It is a beautiful GG discovery and deserves to be regularly trumpeted as one of GGs great achievements. So, next time you are at a party and there is a lull in the conversation, remember the wonders of S-theory.

---

[1] Note that given current work arguing that resumption involves movement raises interesting questions about S theory. Given my partiality to this excellent idea (Demirdache is a leading exponent of this line of thinking), I think that it is worth revisiting some of the OWM assumptions, though I will refrain from doing so here.

[2] There was even independent dialectal evidence in favor of the movement analysis of comparatives: John is taller than what Bill is. The ‘what’ sure looks like a relative pronoun. This observation was due to Kayne, if I recall correctly (way to go Richie!).

[3] Wh islands another “novel” island, one that in fact Ross argued at length did not exist. As you all know, the status of Wh islands is somewhat variable cross linguistically and even among speakers of the same language. Sprouse’s thesis shows that they more or less display island-like acceptability signatures. However, whatever their status (maybe they are semantic rather than syntactic as some have argued), theoretically, they fell under the OWM unification only if one makes additional assumptions about the structure of C (how many “escape” hatches it contains). This assumption was usefully investigated empirically by Reinhart and Comorovski. They showed that the degree of freedom that S-theory allowed for was in fact empirically consequential, thus providing an indirect argument in favor of the unification along OWM lines.

David Poeppel's second provocation

I want to say a couple of words about David Poeppel’s second lecture (see here); the one on which he argues that brains entrain to G like structures in online speech comprehension. In earlier posts (e.g. here), I evinced some skepticism about whether it will be possible for linguistics to make much contact with brain work (BW) until BW starts finding the neural analogues of classical CS notions like register, stack, index, buffer, etc. Here I want to, at once, elaborate, yet also mitigate, this skepticism in light of David’s lectures.

Whence my skepticism? The reason lies with BW methodology. Most BW focuses on what the brain does while being subjected to a particular stimulus. So for example, how does the brain “react” when peppered with sentences that vary in grammaticality (note: I mean ‘grammaticality’ here and not ‘acceptability’).[1] So, while a brain is processing an actual linguistic object, what does it do? By its very nature, this kind of experiment can only indirectly make contact with the basic claims of GG. Why? Because, GG is a theory of competence (i.e. what speakers know), and what BW probes is not knowledge (linguistic or otherwise) per se but how this knowledge is deployed in real time (e.g. how it is used to perform the task of, e.g. analyzing the incoming speech stream). I strongly believe that what you know is implicated in what you do. But what you do (and how you do it) does not rely exclusively on what you know. That’s why it is useful to make the competence/performance distinction. And, back to the main point, BW experiments are all based on online performance tasks where the details of the performance system matter to the empirical outcomes (or so I would suppose).[2] But if these details matter, then in order to even see the contribution of Gs (or FL/UG) it is imperative to have performance systems of the kind that we believe are cognitively viable, and the ones that we think are such are largely combinations of Gs embedded in classical computing devices with Turing-like architectures.[3]

Nor is this mix an accident if people like Randy Gallistel (and Jerry Fodor and Zenon Pylyshyn and Gary Marcus) are to be believed (and I for one make it a cognitive policy to believe most of what Randy (and Jerry) says). Turing architectures are just what cog-neuro (CN) needs to handle representations, and representations are something that any theory hoping to deal with basic cognition will need. As I have gone over these arguments before, I will not further worry this point. But, rest assured, I take this very seriously and because of this I draw the obvious conclusion: until BW finds the neural analogues of basic Turing architecture the possibility of BW and cognition (including linguistics) making vigorous contact will be pretty low.

That said, one of the interesting features of David’s second lecture is that he showed that it is not impossible. In fact, the general program he outlines is very exciting and worth thinking about from a linguistic point of view. This is what I will try to do here (with all the appropriate caveats concerning my low skill set etc.).

The lecture focuses on what David describes as “an interesting alignment between…systems neuroscience….physics…(and) linguistics” (slide 21). More particularly, he argues that the brain have natural “theta rhythms” of 4-8 Hz, that physically speaking the modulation spectrum of speech is 4-5 Hz and that the mean syllable duration cross linguistically is 150-300 ms, which is the right size to fit into these theta bands. So, if speech processing requires chunking into brains sized units, then we might expect the stream to be chopped into theta band sizes for the brain to examine in order to extract further linguistic information necessary to get one to the relevant interpretation. And what we expect, David claims we in fact find. The brain seems to entrain to syllables, phrases and sentences. Or, more exactly, we can find neural measures that seem to correlate with each such linguistic unit (see slides 29-34).

Now showing this is not at all trivial, and therein lies the beauty of the experiments that David so lovingly reported (it was actually a bit embarrassing to see him caressing those findings so intimately in such a public setting (over 150 people watching!!)). Here’s a post to a discussion of (and link to) the paper that David discussed. Suffice it to say that what was required to make this convincing was controlling for the many factors that likely correlate with the syntactic structure. So the Ding et al (David being the last al) paper stripped out prosodic information and statistical transitional probability information so that only linguistic information concerning phrase structure and sentence structure remained. What the paper showed is that even in the absence of these cues in the occurent stimulus the brain entrained to phrases and sentences in addition to syllables. So, the conclusion: brains of native speakers can track G like structures of their native languages on line. In other words, it looks like brains use Gs in online performance.

Now, gorgeous as this is, it is important to understand that the result is very unsurprising from a linguistic point of view. The conventional position is speakers use their knowledge to parse the incoming signal. On the assumption that humans do this in virtue of some feature of their brains (rather than say their ham strings or pituitary glands) then it is not surprising to find a neural correlate of this performance. Nor do David and friends believe otherwise. Oddly, the finding caused somewhat of a buzz in the lecture hall, and this only makes sense if the idea that people use their Gs in performance is considered a way-out-there kind of proposition.

I should add, that nothing about these results tell us where this Gish knowledge comes from. None of the entrainment results implicate piles of innate brain structure or a genetic basis for language or any of the other “sexy” (viz. conceptually anodyne) stuff that gets a psychologist or neuroscientist hopping madly around flailing his/her arms and/or pulling his/her hair. All it implicates is the existence of something like our old recognizable linguistic representations (representations, btw, which are completely uncontroversial linguistically in that they have analogues in virtually every theory of syntax that I know of). They imply that sentences have hierarchical structures of various kinds (syllables, phrases, sentences) and that these are causally efficacious in online processing. How could this conclusion possibly cause a stir.  It shouldn’t, but it did. That we can find a neural correlate of what we know to be happening is interesting and noteworthy, but it is not something that we did not expect to exist, at least absent a commitment to Cartesian Dualism of the most vulgar kind (and yes, there are sophisticated forms of dualism).

What structures do the neural signals correlate with? Many have noted that the relevant level may not be syntactic. In particular, what we are tracking might be phonological phrases, rather than syntactic ones (I said as much at Nijmegen and Ellen Lau noted this in a comment here). However, this does not really matter for the main conclusion. Why? Because the phonological phrases that might be being tracked are products of an underlying syntax from which they are determined. Here’s what I mean.

Say that what is being tracked are phonological phrases of some sort. What the experiment shows is that they are not tracking these objects in virtue of their phonological structure. The relevant phonological information has been weeded out of the stimuli, as has the statistical information. So, if the brain is entraining to phonological phrases, then it is doing so in virtue of tracking syntactic information. Now, again, every theory I know of has a mapping between syntactic and intonational structure so the fact that one is tracking the latter by analyzing structure according to the former is not a surprise and it remains evidence that the brain can use (and does use) G information in parsing the incoming speech string. So the general conclusion that David draws (slide 57) seems to me perfectly correct:

1.     There are cortical circuits in the brain that “generate slow rhythms matching the time scales of larger linguistic structures, even when such rhythms are not present in the speech input” and this “provides a plausible mechanism for online building of large linguistic structures.”

2.     Such tracking is rule/grammar based.

So what is exciting is not the conclusion (that brains use G information in performance) but the fact that we now have a brain measure of it.

There is also an exciting suggestion: that what holds for the syllable, also holds for phrases and sentences. Here’s what I mean. There is ample evidence that David reviews in lecture 2 that the physics of speech results in chunks that have both a nice physical description and that fit into nice neural bands and that find a linguistic analogue; the syllable. So there is a kind of physical/neural basis for this.

The question is whether this analogy extends to larger units?[4] What David shows is that the brain entrains to these units and that it does so using the theta and delta band oscillations to do so. However, is it plausible that phrases and sentences come (at least on average) in certain physical “sizes” that neatly fit into natural brain cyclic bands so that we can argue that there is a physical/neural natural size that phrases and sentences conform to? This seems like a stretch. Is it?

I have no idea. But, maybe we shouldn’t dismiss the analogy that David is pushing too quickly. Say that it is phono phrases that the Ding et al paper is tracking. The question then is what kind of inventory of phono phrases do we find. We know that these don’t perfectly track syntactic phrases (remember: this is the cat, that ate the rat, that stole the cheese, that…). There are “readjustments” that map constituents to their phonological expressions. The question then is whether it is too far fetched to think that something like this can hold for phrases and sentence as well as syllables. Might there be a standard “average” size of a phrase (or a phase that contains a phrase) say? One that is neatly packaged in a brain wave with the right bandwidth? This doesn’t seem too far-fetched to me, but how would I know (remember, I know squat about these matters). At any rate, this is the big idea that David’s second and third lecture hint at. We are looking for natural neural envelopes within which interesting syntactic processing takes place. We are not looking at syntactic processing itself, or at least neither David’s lectures nor the Ding et al paper suggest that we are, but the containers the brain wraps them in when analyzing them online.

That said, the envelope thesis would be interesting were it true, even for a syntactician like me. Why? Because this could plausibly constrain the mapping between (let’s call them) phonological phrase and syntactic phrases and this might in turn tell us something about syntactic phrases. Of course, it might not. But it is a useful speculation worth further investigation, which, I am pretty certain, is what David is going to do.

To wrap up: I am pretty skeptical that current neuro methods can make contact with much of what linguistics does but not because of any problems with linguistics. The problem lies, in the main, with the fact that BW has stopped looking for the structures required to tell any kind of performance story. In a word, it has severed its ties to classical CS as Gallistel has cogently argued.  I believe that once we find classical architectures in the brain (and they are there as the cognitive evidence overwhelmingly shows us) then contact will be rampant and progress will be made understanding how brains do language. This, however, will still mainly tell us about how Gs get used and so only indirectly shine a light on what kinds of representations Gs have. Of course, you can find lost keys even in indirect light so this is nothing to sneeze at, or pooh-pooh, or sneer at or cavalierly dismiss, or…What David’s work has shown is that it might be possible to find interesting linguistically significant stuff even short of cracking the Turing architecture/representation problem. David’s work relies on the idea (truism?) that brains chunk information and the bold hypothesis that this chunking is both based in brain architecture and might (at least indirectly) correlate with significant linguistic units. He has made this conjecture solid when it comes to syllables. It would be a really big deal if he could extend this to phrases and sentences. I hope he is right and he can do it. It would be amazing and wonderful.

One last comment: say that David is wrong. What would it mean for the relation between linguistics and BW? Not much. It would leave us roughly where we are today. The big problem, IMO, is still the one that Gallistel has identified. This is true regardless of whether David’s hope is realizable. Neural envelopes for packaging representations are not themselves representations. They are what you stuff representations into. The big goal of the cog-neuro game should be to understand the neural bases of these representations. If David is right, we will have a narrower place to look, but just as you cannot tell a book by its cover (although, personally I often try to) so you cannot deduce the (full) structure of the representations from the envelopes they are delivered in. So should David fail, it would be really too bad (the idea is a great one) but it would not alter much how we should understand the general contours of the enterprise. In other side, if David wins, we all do. If his conjecture stops at the syllable, this has no implications about the neural reality of phrases and sentences. They are very real, whether or not our current technology assigns them stable neural correlates.

---

[1] A HW problem: what’s the difference?

[2] In fact, that whole point is that these are online tasks that allow you to look into what is being done as it is being done. This contrasts with off line, say, acceptability judgment tasks where all we look at it is the final step of what all concede to be a long and complicated process. Interestingly, it appears that taking a look at this last step allows for a better window into the overall system of knowledge than does looking at how these steps are arrived at. This is not that surprising when one thinks about it, despite the air of paradox. Think forests and trees.

[3] Think Marcus parsers or the Berwick and Weinberg left corner parser (a favorite of mine for personal reasons).

[4] This is what Ellen was skeptical of.