GG re-education camp

Every summer I go back to Generative Grammar (GG) re-education camp. I pick up an old classic (or two) and reread it/them to see what I failed to understand when I read it/them last and what nuggets there remain to mine. This year, prompted by a project that I will tell you about soon (in order to pick your collective brains) I re-read Syntactic Structures (SS), Topics (T) and Language and Mind (L&M) (well, I’m in the middle of the last two and have read the first twice). At any rate, several things struck me and they seemed like good blog fodder, so let me share.

Before I got into linguistics (when I was still but a starry eyed philo major (no cracks please, too easy)) I though that deep structure was, well, deep. After all why call it deep structure if it was just another level, without particular significance. Wasn’t it, after all, the place where Gs met semantics (at least in both SS and the standard theory) and wasn’t meaning deep?

Moreover, I was not the only one who thought this. The popular press circulating around GG always seemed to zero in on “deep structure,” surface structure being so, well, surfacy. Like any philosopher, given a choice between plumbing the depths and skimming the surface I was all for going down and deep.

As I grew more sophisticated I came to realize the error of my ways and how terminology had mislead me. I would sneer at terminologically naived neophytes who failed to appreciate that “deep” did not mean “fundamental.” I would knowingly intone that deep structure was just another level and of no more intrinsic significance that any other level. I would also glibly point out that meaning was not restricted to deep structure as the Katz-Postal hypothesis was slowly giving way to interpretive theories of semantic interpretation where surface structure fed some aspects of meaning (Jackendoff 1972 being the seminal text).[1]

And I was wrong. Sophistication be damned, deep structure really was/is deep, even if not in the way that I originally thought. That’s what my summer rereading of the big three above showed me. So, why is deep structure deep in the sense of terrifically significant to the GG enterprise? Here’s why in one phrase: linguistic creativity (LC).

Chomsky noted that the fact of LC was underappreciated. Humans are able to appropriately produce and easily understand a (practically) infinite number of linguistic expressions. Or, as a matter of course, humans produce or parse linguistic expressions they have never before encountered. The capacity to do this requires that they have internalized a system of rules. What kinds? Rules that tightly couple a linguistic expression’s meaning with that linguistic expression’s articulation (sound, gesture). Absent this kind of theory (aka a grammar that generates an infinite number of sound meaning pairings) there is no possible account for this easily observed fact that humans are linguistically creative.

Moreover, as Chomsky argues in SS and L&M and T, the structure required to code for articulations are insufficient to represent core aspects of meaning. Here’s Chomsky in Topics (17):

It is clear…that deep structures must be quite different from this surface structure. For one thing, the surface representation in no way expresses the grammatical relations that are…crucial for semantic interpretation. Secondly, in the case of ambiguous sentences such as, for example, (5), only a single surface structure may be assigned but the deep structures must obviously differ. Such examples …are sufficient to indicate…that deep structures cannot be identified with surface structures. The inability of surface structures to indicate semantically significant grammatical relations (i.e., to serve as deep structures) is one fundamental fact that motivated the development of transformational generative grammar…

Thus any account of LC which wants to account for the human capacity to use an unbounded number of linguistic expressions (i.e. linguistic expressions a given native speaker has never before encountered) must include a system of rules that recursively generate sound meaning pairings based on different kinds of representations that are G related. Given LC and the fact meaning structures are different from sound structures there really is no other logical option than something like a transformational GG.

Before proceeding, I want to make an unpaid political announcement. GG has been regularly accused of dissing meaning. For example, the autonomy of syntax is often misunderstood as the irrelevance of semantics. As you all know, this is completely bogus. The autonomy of syntax thesis is the very very weak claim. It notes that syntactic properties are not reducible to semantic (or phonetic) ones. It does not deny that meaning (and sound) facts are G irrelevant.

Moreover, Chomsky emphasizes this point in all three works. In both T and L&M he emphasizes that the BIG problem with earlier Structuralism was its inability to accommodate the simplest facts about meaning (in particular what we now call theta roles (who did what to whom)). Thus, how language delivers meaning was at the center of Chomsky’s novel GG proposals and was the central feature of his critique of Structuralism. And this is not something that it takes sophisticated close textual analysis to discover. This leads me to think that many (maybe most) critics of GG’s syntactocentrism simply did not (and have not) read the work being criticized. Not only is this deeply ignorant, but is is intellectually irresponsible as well. Sadly, this kind of ignorant criticism has become a hallmark of the anti GG literature, something that people like Evans (see here and links provided) and Everett (see here) among others have further personalized. However, what is clear on rereading these classics is that these critiques are not based on even a cursory reading of the relevant texts.

Ok, back to the main programming: So, LC properly described leads quickly to the modern conception of grammar, one with distinctive levels for the coding of articulatory and semantic information (surface structure (S-S) and deep structure (D-S)) and operations that unite these levels (aka, transformations (T)). So what made deep structure deep was the realization that LC required it and once one had D-S and understood it to be structurally distinct from S-S then one needed Ts to relate them and the whole modern GG enterprise is up and running. Here’s Chomsky in L&M (17):

…the speaker makes infinite use of finite means. His grammar must, then, contain a finite system of rules that generate infinitely many deep and surface structures, appropriately related. It must also contain rules that related these abstract structures to certain representations of sound and meaning…

Well actually, most (not all) of the modern GG enterprise is motivated by the fact of LC, in particular the project of specifying the properties of particular human Gs and the enterprise of specifying the properties humans must have for acquiring these Gs. The minimalist program adds an extra dimension: the extra question (already mooted in these early works btw) of separating out the linguistically specific factors underlying these two capacities from the more cognitively and computationally general ones that are underlie capacities but are not specifically linguistically dedicated.

So, a recursive G is part of any theory aspiring to address the fact of LC and given the difference between S-S and D-S this G will have at least a D-S level, an S-S level and a T component to relate them. And this brings us to why deep structure was a deep discovery. Critically, structuralism was ready to recognize something like S-S. What structuralism missed was any level analogous to D-S, the level relevant to semantic interpretation. Again Chomsky in L&M (19):

…[M]odern structural and descriptive linguistics … restricts itself to the analysis of what I have called surface structure, to formal properties that are explicit in the signal and to phrases and units that can be determined from the signal by techniques of segmentation and classification…[S]uch taxonomic analysis leaves no place for the deep structures…[which] cannot be derived…by segmentation and classification of segmented units, nor can the transformational operations relating deep and surface structure…

So, what brought down classical structuralism and the Empiricist/Behaviorist psychology that it embraced? Well, the observation that LC required something like D-S. That, in short, is one reason why D-S really is deep.

I should add that the relevance of this line of thinking to G issues has still not been entirely internalized. There is an industry trying to show that phrase structure can be statistically induced from the signal, thinking that were this so the GG enterprise would be fatally wounded (see Elissa Newport’s work on this for example). There is nary a mention of the problem of relating D-Sish facts and S-Sish facts. The idea seems to be that if we could just get hierarchically structured S-Ss from the signal the whole GG project as envisioned by Chomsky over 60 years ago would be discredited as fundamentally empirically flawed. There is little recognition that the problems for structuralism and its attendant empiricist psychology started from the concession that S-S might be amenable to standard analytic (associationist) techniques.[2] The problem was that structuralism left out half the problem, the D-S part. Things, sadly, are no better today in much of the anti-GG literature.

There is a second reason that D-S was considered deep: it pointed to where language was likely to be invariant. Chomsky notes this in L&M discussing the philosophical grammarians (e.g. Port Royal types). He observes that modern conceptions of GG “make the assumption that languages will differ very little despite considerable diversity in superficial realization” (76). Where will languages be “similar”? “[O]nly at the deeper level , the level at which grammatical relations are expressed and at which the processes that provide for the creative aspect of language use are to be found” (77). Thus, D-S and the attendant operations that deliver a corresponding S-S were the natural locus of invariance given the obvious surface diversity of natural languages. Thus, the other deep property of D-S was that it and the principles mapping it to S-S were likely to be invariant across Gs, these invariances being key features of UG.

So, deep structure had some important features that arguably made it deep. But I can sense all you minimalists out there developing an uncomfortable intellectual itch that can be characterized roughly as follows: how deep could deep structure be given that contemporary theories have dispensed with it. Good itch. Let me scratch.

First, we have retained much of the point of deep structure in contemporary theory. So, for example, nobody now thinks that the syntactic structure relevant to surface phonetic form is the same as required to code for underlying grammatical function/thematic form. Indeed, given the predicate internal subject hypothesis there is almost no sentence, no matter how simple in which the underlying semantic subject (the external argument) starts in surface subject position (e.g. Spec T). The structure relevant to phon interpretation is understood as different from the syntax relevant for sem interpretation and it is taken for granted that any adequate G will have to generate an infinite number of phon-sem pairs. In other words, the moral of D-S has been completely internalized.

So too has the idea that D-S is G invariant. Contemporary syntactic theory does not tolerate variation in the mapping of theta roles to initial phrase marker positions. We are all UTAHers now! Thus, we do not expect Gs just like English but where affected objects are underlying subjects and agents are underlying objects. This is not a dimension of permissible variation. Nor do we expect the mapping principles that deliver CI (and possibly AP) interpretable objects to differ significantly. Operations are constrained by universal principles like phase impenetrability (aka subjacency), the ECP, minimality, etc. When we think of universals in GG, this is the kind of thing we are assuming. GG makes no claim about surface invariances. We expect the overt surface properties of language to vary dramatically. We expect little CI/LF variation and no variation in the principles of UG. Thus, invariance lives in the forms/derivations that feed CI, not in the surface realizations of these derivations. Again, this endorses much of the D-S conceptions outlined in SS, L&M and Topics.

So where the difference? Modern syntactic theory, minimalism, has largely abandoned the technology of D-S, not its grammatical point. Minimalism no longer assumes that there is a G level like deep structure or D-structure, i.e. a level at which GFs are determined by something like a phrase structure rules. This was part of every prior (Chomskyan) GG theory. The rejection of D-S (and its analogues) has been more or less complete.

We have given up the idea that D-S is the product of PS rules which all apply prior to displacement operations. In fact, thoroughly modern Minimalists don’t recognize a formal distinction between E-merge and I-merge, both just being instances of the same underlying Merge operation. Furthermore, Bare Phrase Structure has eliminated the distinction between Structure building and lexical insertion so critical to earlier D-S conceptions. In modern theory there is strictly speaking nothing like a PS rule anymore and so not much left of the idea that the Grammatical Functions relevant to semantic interpretation are coded via PS rules.

There remains one last residue of the old technical D-S idea. Some (e.g. Chomsky and PRO lovers everywhere) still hold onto the view that the logical GF roles are products of E-merge exclusively. Others (e.g. moi) do not restrict GF marking to E-merge, but allow that marking via I-merge. However, this is really the last place where the technical notion of D-S has life. I, of course, believe that I am right and Chomsky is wrong here (though I would not bet on myself, at least not a lot). However, this is really the last residue of the older conception of D-S as a level. The technical conception seems largely gone, though the empirical and conceptual points D-S served have been completely internalized.

Last point: here’s something else that struck me in rereading this literature: why is it that S-S and D-S don’t perfectly match. One can imagine a world in which these two had to coincide. In such a world there would be two articulations of flying planes can be dangerous (one corresponding to each interpretation) and passives (where surface and underlying grammatical relations do not coincide) would not exist. This is a perfectly conceivable universe, but it is not ours. Why not? Why is D-S distinct from S-S? Why don’t they match one to one? Might the mere fact that these two kinds of information are differentially encoded support Chomsky’s recent suggestions that the mapping to articulation is a late accretion and that the primary mapping is from something like D-S to SI? I don’t know, but it is curious that our world is not more neatly arranged. And that it is not, should be something we think about, and maybe one day address.

That’s it for now. The discovery of D-S launched the modern GG enterprise. The existence of D-Sish facts and what they means for GG are now part of the common wisdom. It is fair to say that D-S focused scientific attention on LC and Plato’s problem. If that ain’t “deep” I don’t know what could be.

---

[1] A great book btw, one that I would still recommend highly.

[2] Btw, this is almost certainly false once one starts thinking about how to “abstract” out categories that allow for recursion. It is one thing to define the VP in John saw the dog via these simple techniques and another to define the VPs in John saw the dog that Bill thinks that Mary kissed using them. Once we consider categories with recursive subparts the standard analytic techniques quickly fail. Simple phrase structure might be statistically coaxed from surface forms. Interesting ones with complex structure will not be.

It's not really possible that UG doesn't exist

There are four big facts that undergird the generative enterprise:

1.     Species specificity: Nothing talks like humans talk, not even sorta kinda.

2.     Linguistic creativity: “a mature native speaker can produce a new sentence on the appropriate occasion, and other speakers can understand it immediately, though it is equally new to them’ (Chomsky, Current Issues: 7). In other words, a native speaker of a given L has command over a discrete (and for all practical and theoretical purposes) infinity of differently interpreted linguistic expressions.

3.     Plato’s Problem: Any human child can acquire any language with native proficiency if placed in the appropriate speech community.

4.     Darwin’s Problem: Human linguistic capacity is a very recent biological innovation (roughly 50-100 kya).

These four facts have two salient properties. First, they are more or less obviously true. That’s why nobody will win a Nobel prize for “discovering” any of them. It is obvious that nothing does language remotely like humans do, and that any kid can learn any language, and that there is for all practical and theoretical purposes an infinity of sentences a native speaker can use, and that the kind of linguistic facility we find in humans is a recentish innovation in biological terms (ok, the last is slightly more tendentious, but still pretty obviously correct). Second, these facts can usefully serve as boundary conditions on any adequate theory of language. Let’s consider them in a bit more detail.

(1) implies that there is something special about humans that allows them to be linguistically proficient in the unique way that they are. We can name the source of that proficiency: humans (and most likely only humans) have a linguistically dedicated faculty of language (FL) and “designed” to meet the computational exigencies peculiar to language. 

(2) implies that native speakers acquire Gs (recursive /procedures or rules) able to generate an unbounded number of distinct linguistic objects that native speakers can use to express their thoughts and to understand the expressions that other native speakers utter. In other words, a key part of human linguistic proficiency consists in having an internalized grammar of a particular language able to generate an unbounded number of different linguistic expressions. Combined with (1), we get to the conclusion that part of what makes humans biologically unique is a capacity to acquire Gs of the kind that we do.

(3) implies that all human Gs have something in common; they are all acquirable by humans. This strongly suggests that there are some properties P that all humans have that allow them acquire Gs in the effortless reflexive way that they do.

Indeed, cursory inspection of the obvious facts allows us to say a bit more: (i) we know that the data available to the child vastly underdetermines the kinds of Gs that we know humans can acquire thus (ii) it must be the case that some of the limits on the acquirable Gs reflect “the general character his [the acquirers NH] learning capacity rather that the particular course of his experience” (Chomsky, Current Issues; 112). (1), (2) and (3) imply that FL consists in part of language specific capacities that enable humans to acquire some kinds of Gs more easily than others (and, perhaps, some not at all).

Here’s another way of saying this. Let’s call the linguo-centric aspect of FL, UG. More specifically, UG is those features of FL that are linguistically specific, in contrast to those features of FL that are part of human or biological cognition more generally. Note that this allows for FL to consist of features that are not parts of UG. All that it implies is that there are some features of FL that are linguistically proprietary. That some such features exist is a nearly apodictic conclusion given facts (1)-(3). Indeed, it is a pretty sure bet (one that I would be happy to give long odds on) that human cognition involves some biologically given computational capacities unique to humans that underlie our linguistic facility. In other words, the UG part of FL is not null.[1]

The fourth fact implies a bit more about the “size” of UG: (4) implies that the UG part of FL is rather restricted.[2] In other words, though there are some cognitively unique features of FL (i.e. UG is not empty), FL consists of many operations that FL shares with other cognitive components and that are likely shared across species. In other words, though UG has content, most of FL consists of operations and conditions not unique to FL.

Now, the argument outlined above is often taken to be very controversial and highly speculative. It isn’t. That humans have an FL with some unique UGish features is a trivial conclusion from very obvious facts. In short, the conclusion is a no-brainer, a virtual truism! What is controversial, and rightly so, is what UG consists in. This is quite definitely not obvious and this is what linguists (and others interested in language and its cognitive and biological underpinnings) are (or should be) trying to figure out. IMO, linguists have a pretty good working (i.e. effective) theory of FL/UG and have promising leads on its fundamental properties. But, and I really want to emphasize this, even if many/most of the details are wrong the basic conclusion that humans have an FL with UGish touches must be right. To repeat, that FL/UG is a human biological endowment is (or should be) uncontroversial, even if what FL/UG consists in isn’t.[3] 

Let me put this another way, with a small nod to 17^th and 18^th century discussions about skepticism. Thinkers of this era distinguished logical certainty from moral certainty. Something is logically certain iff its negation is logically false (i.e. only logical truths can be logically certain). Given this criteria, not surprisingly, virtually nothing is certain. Nonetheless, we can and do judge that many more or less certain that are neither tautologies nor contradictions. Those things that enjoy a high degree of certainty but are not logically certain are morally certain. In other words, it is worth a sizable bet with long odds given. My claim is the following: that UG exists is morally certain. That there is a species specific dedicated capacity based on some intrinsic linguistically specific computational capacities is as close to a sure thing as we can have. Of course, it might be wrong, but only in the way that our bet that birds are built to fly might be wrong, or fish are built to swim might be. Maybe there is nothing special about birds that allow them to fly (maybe as Chomsky once wrly suggested, eagles are just very good jumpers). Maybe fish swim like I do only more so. Maybe. And maybe you are interested in this beautiful bridge in NYC that I have to sell you. That FL/UG exists is a moral certainty. The interesting question is what’s in it, not if it’s there.

Why do I mention this? Because in my experience, discussions in and about linguistics often tend to run the whether/that and the what/how questions together. This is quite obvious in discussions of the Poverty of Stimulus (PoS). It is pretty easy to establish that/whether a given phenomenon is subject to PoS, i.e. that there is not enough data in the PLD to fix a given mature capacity. But this does not mean that any given solution for that problem is correct. Nonetheless, many regularly conclude that because a proposed solution is imperfect (or worse) that there is no PoS problem at all and that FL/UG is unnecessary. But this is a non-sequitur. Whether something has a PoS profile is independent of whether any of the extant proposed solutions are viable.

Similarly with evolutionary qualms regarding rich UGs: that something like island effects fall under the purview of FL/UG is, IMO, virtually uncontestable. What the relevant mechanisms are and how they got into FL/UG is a related but separable issue. Ok, I want to walk this back a bit: that some proposal runs afoul of Darwin’s Problem (or Plato’s) is a good reason for re-thinking it. But, this is a reason for rethinking the proposed specific mechanism, it is not a reason to reject the claim that FL has internal structure of a partially UGish nature. Confusing questions leads to baby/bathwater problems, so don’t do it!

So what’s the take home message: we can know that something is so without knowing how it is so. We know that FL has a UGish component by considering very simple evident facts. These simple evident facts do not suffice to reveal the fine structure of FL/UG but not knowing what the latter is does not undermine the former conclusion that it exists. Different questions, different data, different arguments. Keeping this in mind will help us avoid taking three steps backwards for every two steps forward.

---

[1] Note that this is a very weak claim. I return to this below.

[2] Actually, this needs to be very qualified, as done here and here.

[3] This is very like the mathematical distinction between an existence proof vs a constructive proof. We often have proofs that something is the case without knowing what that something is.