In the late 20th century, the general view of generative syntax was that it made claims about the innate universal grammar, and that by investigating the principles and parameters of grammar, it did three things simultaneously: (i) explain the possibility of language acquisition despite the poverty of the stimulus, (ii) explain limits on cross-linguistic diversity, and (iii) provide a framework for the elegant description of particular languages (e.g. Baker 2003).
The confidence in the existence of a rich UG and parameters seems to have waned in the 21st century (e.g. Boeckx & Leivada (2013) rejecting parameters, and this blogpost on Chomsky’s skeptical stance concerning a rich UG). So what is left for generative syntax?
I have been struggling to understand this, looking for cues in what generative linguists say. There seems to be a lot of disagreement where the larger generative enterprise is headed (understandably), but what is surprising to me is that I keep encountering the idea that generative syntax can simply be used as a descriptive tool (basically, only for (iii) above).
Thus, in their reply to my critical questions on their textbook (“Syntax”, in this blogpost), Koeneman & Zeijlstra say the following:
“although we build up the theory using technicalities that are adopted from current minimalism, we do not adhere to or to try to persuade students about most of its philosophical or biological underpinnings, such as innateness claims” (https://dlc.hypotheses.org/1082)
So they act as if it made sense to use the high-tech mechanisms simply to describe English syntax (the only language they deal with in their book). I wonder if they really believe this, but it seems a very strange idea to me. English syntax can be described in much simpler terms, as for example in Huddleston & Pullum’s (2002) grammar of English. One can say many things about minimalist syntax, but it is certainly not a tool for the “elegant” description of a particular language (regardless of wider concerns).
Similarly, Federica Cognola replied to my critique of the intro chapter of a book coedited by her (on Null subjects in generative grammar, Cognola & Casalicchio (eds.) 2018) in the following way:
“the term „UG“ only appears once in our paper (in a citation) and we use the categories CP/TP/VP/EPP etc in our chapter as purely descriptive devices to describe the sentence’s structure and position – and I think this is uncontroversial in Generative Grammar.” (https://dlc.hypotheses.org/1340#comments)
It may be uncontroversial in generative grammar to use highly specialized terms such as CP/TP/EPP in the notation, but I thought that it was also uncontroversial that these concepts are part of the generative analysis in terms of UG categories, and not necessary for a generally accessible description of the phenomena. Many generative papers (at least during the principles-and-parameters period) included a “data/facts” section, which laid out the phenomena in straightforward terms, followed by an “analysis” section, which re-described the same phenomena in much more complex terms, using the high-tech notation in terms of entities like vP and EPP (plus movements, feature checking, and so on). All of this makes good sense if one believes that we are born with such categories and devices, and that children actually learn such systems.
But why should they be used for simple description? It seems to me that there is an embarrassing misunderstanding here: Teaching generative syntax has become a routine phenomenon in many corners of linguistics, and more and more people who are simply interested in languages (not in limits on worldwide diversity, and not in innateness explanations for acquisition or universals) got the impression that the notation was somehow established knowledge of the discipline of linguistics, maybe like the notation of chemical formulas has become an established part of chemistry. But of course, the proposals of current mainstream generative grammar are just one out of a very large number of possibilities (cf. McCawley’s (1982) book title “Thirty million theories of grammar”, which has lost nothing of its relevance). They happen to be popular, and to the extent that they are making general claims about Human Language with empirical import, they are scientific theories. But I do not see how one can justify the use of generative grammar technology without the innateness claim.
The fading of the innateness idea can also be seen clearly in Julie Anne Legate’s current view, explained in a blogpost comment earlier this year (by “JAL”, discussing dependent case theory):
“Let me take this opportunity to reveal my current thinking on case. I have now convinced myself of a framework whereby merge is innate, but any (other) language-specific innate properties are highly suspect and require significant evidence. (This is due to Noam’s writings on evolution finally sinking in, and due to my accumulating knowledge about the extent of language variation.) Case, both the distribution of noun phrases and the case morphology, is not universal, varies considerably across languages, and so must be learned. (See Charles Yang’s 2016 book on how.)” (Legate 2018, on Norbert Hornstein’s blog)
Legate (a very influential generative linguist due to her position as chief editor of NLLT) continues to use the minimalist notation (e.g. in her 2015 book on Acehnese, and also in recent presentations at Leipzig University), but apparently for purely descriptive purposes (though she is not explicit about this).
Another example of an apparent misunderstanding comes from a recent dissertation on the Brazilian language Tuparí by Adam Singerman (2018). Singerman is aware of different views about the best way of describing a language (either in its own terms, or through the use of UG categories), and he tries to somehow reconcile them. In most chapters of his work, he makes no reference to generative notation and uses concepts that will be understood for many decades in the future, but in Chapter 5, he uses opaque terms like “AgrS”, “T heads” and “vP” as well as movements. He justifies his choice of minimalist technology as follows:
“These phenomena are ones which current syntactic theory can help to describe and explain; indeed, they demand an analytic framework that assumes more structural nuance than LaPolla & Poa’s (2006) or Haspelmath’s (2010) do. While I of course will not claim that the framework chosen here is the only one capable of handling the Tuparí facts, I am convinced that an approach that denies the possibility of crosslinguistic universals in syntactic structure will not be able to make satisfactory sense of this language’s grammar” (Singerman 2018: 13-14).
This is based on the misunderstanding that my “framework-free grammatical theory” somehow limits the amount of “structural nuance” that describers can posit. On the contrary, the view of description that I defend in that paper is that anything goes – if a new abstract notion is needed to describe a language-internal generalization, it can (and should) be posited. Description is not the same as explanation, and restrictivness (UG) should play no role for the describer. The last sentence of this passage is also strange: I do not “deny the possibility of cross-linguistic universals” (in fact, there are only a handful of other linguists who have written about grammatical universals as extensively as I have), and I would be the last to deny that inspiration from cross-linguistic studies and from other languages can help the describer. So what is the basis of this misunderstanding? Does Singerman think that inspiration from typology must imply the use of minimalist technology? Or that the rejection of the use of innate categories in language description (= the view that each language should be described in its own terms) means a rejection of universality? I am puzzled.
These seem to be difficult times for linguists interested in grammar. The younger ones are often taught that they should use minimalist notations, but those that think about the big picture (such as Noam Chomsky, Julie Anne Legate, Norbert Hornstein, Frederick Newmeyer) tell them that there is (apparently) no rich UG. The situation is very confusing indeed.
(P.S. After my last blogpost about a Chomsky and UG, a generative linguist commented on Facebook: “Sometimes it seems linguistics is divided into generative and anti-generative linguistics – where the former investigates language, and the latter investigates the ways to attack the former.” But maybe generative linguists have simply become too sanguine about the apparent success of their framework. As Julie Anne Legate put it in her statement for the 2015 Athens conference: “while the field has many excellent practitioners in the younger generations, few students seem to have internalized the foundational goals of the field” – and I would add that the foundational goals are quite unclear by now, at least to me.).
Baker, Mark C. 2003. Linguistic differences and language design. Trends in Cognitive Sciences 7(8). 349–353. doi:10.1016/S1364-6613(03)00157-8.
Boeckx, Cedric & Evelina Leivada. 2013. Entangled parametric hierarchies: Problems for an overspecified Universal Grammar. PLOS ONE 8(9). e72357. doi:10.1371/journal.pone.0072357.
Cognola, Federica & Jan Casalicchio (eds.). 2018. Null subjects in generative grammar: A synchronic and diachronic perspective. Oxford: Oxford University Press.
Haspelmath, Martin. 2010. Framework-free grammatical theory. In Bernd Heine & Heiko Narrog (eds.), The Oxford Handbook of Linguistic Analysis, 341–365. Oxforf: Oxford University Press.
Huddleston, Rodney & Geoffrey K. Pullum (eds.). 2002. The Cambridge grammar of the English language. Cambridge: Cambridge University Press.
Koeneman, Olaf & Hedde Zeijlstra. 2017. Syntax. Cambridge: Cambridge University Press.
LaPolla, Randy J. & Dory Poa. 2006. On describing word order. In Felix K. Ameka, Alan Dench & Nicholas Evans (eds.), Catching language: The standing challenge of grammar writing, 269–295. Berlin: Mouton de Gruyter.
Legate, Julie Anne. 2015. Voice and v: Lessons from Acehnese. Cambridge, MA: MIT Press. http://dx.doi.org/10.7551/mitpress/9780262028141.001.0001
McCawley, James D. 1982. Thirty million theories of grammar. Chicago: University of Chicago Press.
Newmeyer, Frederick J. 2004. Against a parameter-setting approach to typological variation. Linguistic Variation Yearbook 4(1). 181–234. doi:10.1075/livy.4.06new.
Singerman, Adam Roth. 2018. The morphosyntax of Tuparí, a Tupían language of the Brazilian Amazon. Chicago: University of Chicago PhD dissertation. doi: 10.6082/ns2a-cv36
Yang, Charles. 2016. The price of linguistic productivity: How children learn to break the rules of language. Cambrudge, MA: MIT Press.
Many years ago I read an excellent comparative article about V2 in the Germanic languages (unfortunately I don’t remember its provenance). The paper divided Germanic languages (other than English) into CP-V2 and IP-V2 groups, but explained that it was not necessary to understand the origin of these terms in order to understand the point of the paper. A CP-V2 language for the paper’s purposes is one that is V2 in main clauses only, whereas an IP-V2 language is V2 in all clauses.
At that time I didn’t know CP or IP from a hole in the ground, and I hardly do yet, but I certainly understood the distinction between “Es ist schwer, Jude zu sein” and “Es iz shver tsu zeyn a Yid”. So perhaps the purely descriptive use of generative terminology does no real harm, at least to the naive reader.
Maybe it does no harm in this case, but my question is: Does it do any good? That’s not clear to me, and the harm it does in more general terms is that it may lead people to think that they have to learn all this complex notation if they want to understand German or Yiddish – which is not the case (it seems).
It probably does no good at all, but I hold that in matters of terminology it is sufficient to do no harm. The fact that the language-specific category and comparative concept “A/accusative” results from a Latin author’s misunderstanding of the Greek LSC “Aitiatike” hardly matters now, though a better translation would have given us “C/causative” instead (with resulting future confusions). Surely it is the etymological fallacy to suppose otherwise.
Per contra, it does do harm to conflate LSCs and CCs, and your capitalization convention helps make the distinction clearly. (What do you use when writing in German, by the way? Letterspacing perhaps?)
I was struck by this comment in the PS:
“Sometimes it seems linguistics is divided into generative and anti-generative linguistics – where the former investigates language, and the latter investigates the ways to attack the former.”
Interesting to see the persecution complex from the other side. In fact, I have argued that generative linguistics seems to be interested in a metalanguage reified as language rather than language itself. And the extent I spend wasting my time being anti-generative is only in response to the damage I see it do to the general understanding of language among the undergraduate intro to linguistics attendees. I’ve argued here that Chomsky is much better off thought as a mathematician than a linguist: http://metaphorhacker.net/2010/08/why-chomsky-doesnt-count-as-a-gifted-linguist.
I was struck early on – when I worked on Categorial Grammar – that Bar-Hillel (1953) introduced it as a useful notation and explicitly warned against seeing any ‘reality’ in it. But by the 1980s that was mostly lost. Chomsky was also seduced by the combinatorial power of the early generative syntax to start assigning unwarranted explanatory powers.
On the other hand, to be honest, I would defend the minimalists here:
“One can say many things about minimalist syntax, but it is certainly not a tool for the “elegant” description of a particular language (regardless of wider concerns).”
This very much depends on the definition of elegance. I’d say that it’s probably more formally elegant and its opaqueness to the rest of the world is a Cartesian feature – you have to learn a lot of foundation before you can say anything in that language. The more ‘accessible’ terms of linguistics are much fuzzier (as you have argued elsewhere). I personally think this is a benefit but I see the other side’s point. Also, these terms are only ‘accessible’ by virtue of having a long tradition of use. As someone who has had to teach some of them to non-experts, I can vouch for them seeming quite opaque to most people. And the terminology around alignment is its own world of hurt.
I followed the link at the beginning of the above post and was similarly struck by this charming statement:
“We can either ignore the many stupid, misguided, or simply fraudulent attacks from outside generative grammar proper, or we can try to fight them.”
You characterize such common notions as T, C, and v as “highly specialized,” “opaque,” and “high tech.” They are specialized, but there is a thriving research community for whom they are meaningful, and the literature is well enough referenced that it is not hard to find out what these notions are meant to refer to if you want to know. Technical terms in any field are “opaque” to outsiders. You yourself make use of such terms — would somebody who hadn’t studied the relevant literature know what you mean by T-flagging? I don’t see any point in trying to conduct research without jargon. I think if you are genuinely interested in learning the status of a notion like T or C or v for generative grammar, you should tone down the rhetoric. To answer the question, what is T and in what sense could it be said to be innate? I recommend Martina Wiltschko’s 2014 CUP book _The Universal Structure of Categories_ for a highly idiosyncratic but very interesting and thoughtful perspective on the question. You will find there a perfectly sensible example of how a generative linguist could straightforwardly respond that no, T is not “innate” but yes, I continue to find it insightful to posit T in my analysis of Language L in the pursuit of (i), (ii), and (iii).
Yes, every science needs technical terminology (or even “jargon”, though I never use this expression because of its derogatory overtones). But while some terms can be explained straightforwardly (e.g. “T-flagging”, which is the case- or adpositional marking of T-arguments (= themes) of ditransitive clauses), others cannot be explained straightforwardly, but to begin to understand them one needs to be deeply immersed in a particular research tradition that makes is very difficult for outsiders to get in. Wiltschko’s book looks good, but idiosyncratic – is there some non-idiosyncratic book that explains why we need C, T and v as universal categories of syntax? Most papers or articles either presuppose these notions or motivate them on the basis of some parochial facts from English – but parochial facts from English cannot justify universal claims.
You presuppose that facts from English are parochial, but if there is a high degree of cross-linguistic uniformity, as generative grammar argues, then English is expected to exemplify many non-parochial facts. It stands to reason that if English provides particularly clear evidence of T (modals, do-support, subject-aux inversion, VP ellipsis, VP fronting) then English will frequently be used to motivate T. More generally, if the presentation of the facts is conducted in English, for example in an introductory textbook, it will be more compact to also use English for examples; hence English might also be used to argue for C, even though the evidence for C is much clearer in languages such as German (V2 and V-initial polar questions, in complementary distribution with subordinating complementizers, similar structure below the V2/V-initial position and the complementizers, for example). The less familiar the facts of the language, the more complex the exposition, so an introductory textbook written in a European language is less likely to use Warlpiri to motivate T or Kashmiri to motivate C, when English, French, and German are more pedagogically accessible. There’s also the fact that because of skilled native speaker linguists, our knowledge of the European languages is much deeper and more reliable, and forms a better foundation on which to build than typological facts which have to be filtered through the lenses of individual non-native grammar writers with distinct backgrounds, agendas, and skill sets.
But in your reply to my comment you are also continuing to presuppose that generativists cannot meaningfully use, e.g., C without supposing it to be universal. That is not the case. Wiltschko proposes that some languages have T and others don’t. Similarly, it has been proposed that languages with finite embedded clauses have C with shared properties, but languages without finite embedded clauses lack that category (e.g. Kiparsky 1995, Indo-European origins of Germanic syntax, OUP volume), or that noun phrases in some languages are headed by D but not in others (Bošković 2008, what will you have, DP or NP? Proceedings of NELS).
So, to restate your question, is there some non-idiosyncratic and convincing argument for T, for example, in some well-researched language, that we could use as a basis for exploring whether other languages also have a T and which of T’s properties are shared across languages? Why, yes, for example Chapter 5 of David Adger’s 2003 OUP book _Core Syntax_ summarizes some of the main arguments for T in English in a clear and accessible way. McCloskey 1996 ‘On the scope of verb movement in Irish’ (NLLT) shows why it is insightful to posit T in Irish, and how it is meaningful to use the same term for that head as for the corresponding head in English (McCloskey calls it I, for Inflection). Benmamoun 2000 (OUP book _The Feature Structure of Functional Categories_) shows how T is insightful in the analysis of various Arabic dialects. These analyses provide a small sample of the understanding of T in mainstream generative grammar, and supply parts of the backdrop for Wiltschko’s proposal.
Many thanks for these comments and for the references. Yes, of course it has been argued that not everything is strictly universal, so what I mean is “universally available” – i.e. the idea that T in Irish is the same as T in English or in Arabic: All these languages make use of the same element T, and since different symptoms are used to identify it, I don’t see how it could be *the same* (rather than just similar – of course the languages are similar in some ways) unless the category is innate.
It seems to me that in order to have what we might call a ‘Universal Descriptive Category’, as opposed to a ‘Simply Descriptive Category’, there needs to be both a diagnostic for its presence, and some further empirical phenomenon that always appear as concomitants of its presence, as determine by the diagnostic. So for nouns, the diagnostic might be ‘a part of speech that contains most names of kinds of people, living things, things, places and stuff’ (the last one for mass nouns), and the universal phenomenon is that such a part of speech always seems to exist, and to be open (classifiers name kinds of people etc, and are closed (with the potential exception of repeaters, which can be explained away as an usual use of nouns), but fail to contain most of the kind-designating words). Note that this does *not* imply that there are any other open classes opposed to noun (or, actually even any closed ones). All of these falling into one, open, part of speech is the claim.
There are a number of issues that need to be nailed to down to finish this story, including subclasses (e.g. mass nouns, and exactly why can’t we say that verbs are a subclass of nouns in various languages) and inflectional features such as gender, which does split up the noun part of speech in wierd ways in many languages. But when we turn to T and C, I don’t see what the universal claims would be, given that it is not always clear that they exist in the way that nouns do, and I’m not aware of anything that they always do (which would have to be different for T and C, to make them both universal), nor of things that only they do. So I’d say that they were ‘simply descriptive’. ‘Often do’ is I think not good enough to meet Martin’s criticism of identifying categories with different diagnostics across languages: what we are basically in search of are things that are like species in biological taxonomy, rather than ‘critters that look sort-of like a moose’ vs ‘critters that look sort-of like a coyote’.
Otoh the somewhat more abstract concept of ‘functional category’ does seem to me to have real universal properties, while the concepts of subject, object etc that I put forth in my 1985/2007 Major Functions of the NP chapter in the Shopen Typology volume were simply descriptive. LFG has proposed making the subject GR universal in my sense by having a stipulation that only subjects can be functional controllees, but this is perhaps a bit dubious.
I don’t think there is anything wrong with working with simply descriptive categories, even though they are not as ambitious as people have been led to aim for. But the explanatory program especially launched with GB does not seem to me to have fared so well, while the ‘better descriptions’ program launched with Syntactic Structures still seems to me to be barrelling along just fine.
“when we turn to T and C, I don’t see what the universal claims would be, given that it is not always clear that they exist in the way that nouns do, and I’m not aware of anything that they always do” – exactly, that’s why I think we don’t need these things in typology, and I find them confusing in descriptions of partocular languages as well. In any event, there’s no clear sense in which “T” has the same meaning in English and in Arabic.
So on more thought I have come up with a possible universal claim for T & C, albeit fairly weak ones, which is that there is at most one functional projection associated with tense marking, for T, and clause-type, for C. This formulation does not imply that they have to be different (T=C would not be an insane proposal for Warlpiri), and could be perhaps be strengthened to the claim that these projections always exist, although they are sometimes different, sometimes the same. But that would require some theoretical development, in languages that don’t appear to have evidence for an Aux (Arandic in Australia, Papuan languages from what little I know of them).
The T claim in particular is fairly weak, especially for people who accept the idea that there are only three tenses, past, present and future, but perhaps something could be pulled out of it for languages such as Modern Greek, where future and past marking happen in competely different places in/on the verb.
This is not that far off from the idea in Bresnan, Asudeh and Toivonen (3016) (the rewrite of Bresnan’s 2001 LFG book) that there can be at most two functional projections for any lexical category. This is a formal idea (and probably not one that Ron Kaplan would approve of on the basis of his 1985/19871995 paper “Three seductions of computational psycholinguistics” in _Formal Issues in Lexical Functional Grammar_), which might tend to produce the effect of C appearing over T on functional/semantic grounds.
“T is not “innate” but yes, I continue to find it insightful to posit T in my analysis of Language L in the pursuit of (i), (ii), and (iii)” – that is not the question. The question is whether T in Halkomelem is the same T as in English, or whether a language should be described in terms of the categories that one needs for it. In other words, whether we are categorial universalists or particularists (and how one can be a categorial universalist without claiming that the categories are innate).
Are meteor craters the same on the moon as on earth? No, in the sense that gravity, atmosphere, and the presence or absence of water cause them to manifest differently, but yes, in the sense that I can define “meteor crater” as a depression caused by impact from an object from space. Commonalities are due to their being meteor craters; differences are caused by independent factors. Innateness is not at issue.
Similarly with linguistic phenomena. If two i-languages (an internal, mental grammar; a brain states which exists in all and only minds of speakers of that language) share a set of properties, I can call that set of shared properties T, independently of whether the brain state is hardwired or emerges on the basis of other factors. I can even potentially discover that all languages have T without answering the question of whether T itself is innate.
I don’t feel that I am obliged to inserting myself into either of your two pigeonholes, universalist or particularist.
I have argued that concepts like “meteor crater” are comparative concepts, defined everywhere in the same way (in my forthcoming new paper on comparative concepts: https://zenodo.org/record/1158392, see Table 2). If concepts like C, T and v are concepts of this same kind, it means that they are just methodological tools (like measurements in physics, or plant anatomy terms in botany), not indispensable elements of the causal explanation. That is perfectly fine – but not in line with 20th century generative thinking (as far as I can tell). Moreover, there is no reason why there should be just one set of such comparative concepts, rather than many different overlapping sets (just as there different overlapping measurement tools such as meter and inch, or different nomenclatures for geologic time periods). And finally, there should be no reason why a particular language *needs* to be described in terms of C, T and v – if a simpler description (in language-particular terms) is available, this should actually be preferred. This is how I operate: When I describe a language, I use the terms for widely known comparative concepts as descriptive categories, unless the language requires something idiosyncratic (e.g. German Weak Adjective Forms). Descriptions in terms of C, T and v are typically highly complex and opaque even to experts, if these are not members of the generative subcommunity.
Yes, comparative concepts can be useful for description. I think most generativists use terms like “subject” and “auxiliary” in this way. But they are insufficient for descriptive adequacy in Chomsky’s sense (1965, Aspects, p. 24).
For descriptive and explanatory adequacy, including your (i), (ii), and (iii), I believe that mentalist commitments are needed (see Chomsky 1986, Knowledge of Language, chapter 2, or Katz 1964, ‘Mentalism in linguistics,’ Language).
We all use descriptive devices. But since I am committed to the mental reality of grammars, I can also meaningfully discuss whether or not Irish and Danish, for example, both have a functional head “T”, and whether I am right or not ultimately depends on whether there is or isn’t a mental structure or brain state which speakers of Irish and speakers of Danish have in common and which is causally related to the things I am using as diagnostics for the presence of T in that sense.
Of course we don’t have a good idea right now of what the brain state corresponding to T would look like (see Embick and Poeppel 2014, ‘Towards a computational(ist) neurobiology of language,’ in Language, Cognition, and Neuroscience). But I believe we are philosophically committed to it, and that it is this, and not whether T (or similar notions) is universal or innate, which is the core of the difference between purely descriptive devices like comparative concepts in your sense and the explanatory devices that generativists posit.
I take results like that of Ding et al 2015 ‘Cortical tracking of hierarchical linguistic structures in connected speech’ in Nature Neuroscience as outside vindication of the mentalist approach to linguistics.
Or, to take another example, as mentioned by Katz, the anti-mentalist Hockett wrote in 1954 in Word: “. . . if it is said that the English past tense form baked is “formed” from bake by a “process” of “suffixation”, then no matter what disclaimer of historicity is made, it is impossible not to conclude that some kind of priority is being assigned to bake as against baked or the suffix. And if this priority is not historical, what is it?” But work like Sahin et al 2009 in Science (‘Sequential processing …’) has shown that the generativist conception of a morphological process has detectable correlates in brain functions, vindicating the mentalist position more than 50 years later.
“Meteor crater” is not a comparative concept: it already implies particular causal explanation. “Crater” would be, but many of them turn out to be volcanic instead (a few also from e.g. nuclear testing, and I’m sure even a quarry abandoned to erosion could in a pinch be described as a crater).
A close geological analysis can tell that terrestrial impact craters form a distinct subtype. However, this evidence is not available for formations like lunar craters. In the absense of direct evidence of the fact that asteroids exist and are able to collide with larger celestial bodies, at most we could conclude that lunar craters and terrestrial impact craters have certain morphological similarities, but the mechanism for this remains unknown. (For an alternate hypothesis, we could e.g. imagine a world where impact craters on the Moon mostly result from meteorites, but impact craters on the Earth mostly result from orbital warfare conducted by pre-human civilizations.)
The received story behind GB/MP does indeed seem to be coming apart, but there have also long been the ‘alternative generative frameworks’, populated by people who presumably rejected some aspects of it. My thoughts are along the lines that there is a lot to be said for trying to find a formalism that works for all languages, capturing the generalizations that child learners of languages appear to be picking up (basically the ‘level 1.5; of Chris Peacocke and Martin Davis). ‘Trying to find’ is supposed to encompass the views of Minimalists who want to work semi-formally while reserving judgement on various issues required for full formalization.
Generalization-capturing is clearly relevant to explaining the possibility of language acquisition, and it is fortunate that as the P&P approach seems to be running into trouble, Bayesian and related ideas have been developing that can function with only positive evidence and a weaker sort of UG have arisen to restore at least some plausibility to the more stipulative ‘rule based’ approaches, characteristic of the alternative generative frameworks. (e.g. the Empiricism and Language Learnability book).
This project also doesn’t depend on any claims about whether the facilities involved in language use are specific to language, or something that is applied more widely. But it would certainly be interesting to find that some of them are not used in some language families or areas; for example I think it might be the case that Mainland Southeast Asian languages and at least many Austronesian ones such as Bahasa Indonesia and its close relatives don’t use morphosyntactic features such as are so prominent in Indo-European and Australian languages, and even less the ‘inside out zipping’ mechanism presented by myself and Rachel Nordlinger, in slightly different versions, to deal with case-stacking/Suffixaufnahme.
Furthermore, many of these frameworks (e.g. LFG) don’t depend on claims about universal categories or even grammatical relations, although many workers assume them, and LFG does need some universal classifications of GRs, such as governable and nongovernable. The XLE parser, for example, has no built in categories or grammatical relations, though at the front of the grammar you have to make various kinds of declarations about the ones you are going to use.
Aunque los comentarios me parecen bastantes atinados, la teoría minimalista, al igual que la gramática generativa, las he considerado una filosofía de la lingüística la cual nos invita a pensar, a reflexionar de cómo podríamos describir X lengua, a pesar que su pretensión es exponerse como modelo teórico para describir las lenguas del mundo. Con respecto al innatismo, pienso que los hechos empíricos podrían dar argumentos pero hay un celo para considerar los hechos empíricos. En este sentido, los generativistas y minimalistas deberían volcarse a estudiar el desarrollo del lenguaje del niño y tomar datos de las lenguas indoeuropeas y de las amerindias y constratarlos en el sentido popperiano. El inglés como lengua para exponer un modelo de análisis lingüístico es insuficiente y esto lo sabe cualquier estudiante de lingüística. Por otro lado, este planteamiento de Chomsky choca con sus declaraciones y conferencias de justicia social y, las lenguas del mundo responden a la sociedad que las ejecuta. No hay lengua sin sociedad y al estudiar lo dicho en el decir de una sociedad estamos haciendo sociolingüística de la cual se distancian los minimalistas y generaativistas.
“… (such as Noam Chomsky, Julie Anne Legate, Norbert Hornstein, Frederick Newmeyer) tell them that there is (apparently) no rich UG.”
You keep saying this over and over, but that doesn’t make it any more true. As has been pointed out so many times before by many people, you can have a rich UG with little to no language-specificity in the sense of innate language-specific categories. It is certainly not the case that, as you say, “anything goes” in language and at most you could posit, as Charles Yang (who you mention) does, that there is an additional step involved in, for example, acquisition of language-specific categories. But this has never been controversial in generative grammar. A rich UG need not be specific to language but can, for example, provide the design space in which language-specific categories may be formed during acquisition and the details of said design space need not be hard-wired: providing constrains for developmental processes would already suffice. Universality in genetics (i.e. UG aka. the design space) does not imply universality in the nitty-gritty details of the mature individual’s biology and maybe even the specifics of cognitive operations–but this is what you study by looking at individuals’ grammars.
In a nutshell, if such a presentation of generative grammar can bring your point of view closer to this research program, then that’s great. But please stop saying that generative linguists are contradicting themselves when what you point to is just two different aspects of doing generative linguistics (or linguistics in general, for that matter): on the one hand you have those who, just like you, are focused on describing languages, while there is others (admittedly, probably a minority) who are focused on the biological nature of language. For the former group it makes sense to talk about a rich UG because the nature of language-specific categories is not arbitrary at all, but for the latter there is only a highly impoverished UG in the sense that very little about the genetics and biology of language is probably unique to language. But those are two different things and two different ways of using the term UG. For the language-description folk acquisition and evolution do not matter much, whereas they are vital in the work of the those working on the biology of language.
Dear anonymous, it would be nice to provide some references (or to give your name). As far as I have understood, the use of extremely specific categories or features such as “vP” and “EPP” implies that these are innate (as part of a rich UG, as traditionally understood). If the term “UG” has several different meanings, maybe it should not be used, and transparent terms should be used instead. (I have started using the term “natural kind”, and making the analogy with other cognitive categories that are often claimed to be innate, such as tastes and emotions.)
I am guessing that one reason for the persistence of generative grammatical theories is that they do address several issues which other approaches have not. In particular, issues of discontinuities (or movement), and of embedding and constituency, have been a rich subject for generative analysis. That is no doubt fueled by the need to analyze such issues in the English language.
I agree with you about the flaws of various generative theories. However, as far as I am aware (and please correct me if I am wrong) the above matters have not yet received much attention from typologists and from framework-free grammarians. Dixon’s compendium spends two pages on constituency (I:232). Shopen has a bit of a discussion, in the chapter on complementation by Noonan, and it uses generative terms (equi-deletion and such). As it is, if one wants to find the tools to completely describe the syntax of question formation in English, the generative literature would be the easiest place to find them, however inadequate they may be.
Yes, it’s true that the non-generative literature has not dealt with movement and constituency so much (this also applies to Relational Grammar and even LFG, where constituency has not been prominent). And I think it’s also true that because of generative grammar’s claim to actually *explain* many of the phenomena that they have become prominent. Unfortunately, it seems that the explanatory promises have not been fulfilled.
I suggest that one reason that there is not so much written about constituency in LFG, in spite of the fact that c-structure rules are its ‘generative component’ (perhaps only a psychological truth rather than anything with real theoretical meaning) is that it is extremely difficult to say what the correct consttuent structures really are when you don’t really believe in a big pile of assumptions to tell you what they have to be, so progress is slow. But we do try to push the wagon down the road.
Yes, LFG has a far more realistic view of constituent structure than MGG. One problem with the post by “Y” is that it seems to presuppose that it’s clear what are “issues of constituency structure”. To a large extent, MGG simply presupposes that everything is constituency structure, but there are very often quite different ways of looking at things. But since trees are visually very pleasant, many linguists don’t consider the alternatives.
I indeed was not expressing myself clearly, so let me try again. The general issue is the ordering of “words” (short for smaller syntactical units) within “sentences” (short for longer syntactical units). Syntactic typology has successfully been addressing such issues for small syntactic structures, -of a bound length-. Generative syntax has occupied itself with the ordering of words in syntactic structures of -unbound length-. Thus, what I am saying, typology has done a much better job in describing the subtleties of basic word order. On the other hand, if I wanted to fully describe German V2 sentence ordering, back issues of Linguistic Typology, or Dixon or Shopen’s manuals, would not prepare me to describe confidently what exactly comes in the position preceding the verb.
Maybe, but what does it mean to “prepare” someone to describe a language? I think one should be prepared to find completely new things, rather than expecting that languages conform to some framework. True, they exhibit many similarities, but if we want to do justice to every language, and avoid Anglocentrism, it’s better to have a distinct framework for easch language.
Umm, I don’t see how it is practical to have a distinct framework for each language, since, if you want to send someone out to do fieldwork in New Guinea or somewhere, you are not so likely to find someone who is willing to go there plus concoct a framework from scratch plus get a thesis handed in by the deadline. I think what is best for that purpose is a loose, semi-formal regional/language-family oriented framework such as what Bob Dixon developed for Australian languages, plus encouragement to look for things that don’t quite fit into it. Such as for example the fact that the possessive and other adnominal cases didn’t really fit into the derivational vs syntactic case categories as proposed by Dixon (as observed by the Dixon students Alan Dench and Nicholas Evans in their 1988 “Multiple Case Marking in Australian Languages” and others, irrc.
It’s even worse with the formal aspects of frameworks (phrase structure rules, categorial reductions, unification, constraint violation ranking, sophisticated type systems …), because nobody is capable of generating this kind of stuff to a level adequate for language description de novo out of nothing, and very few people seem to be capable of even improving them in any but the most trivial ways without making messes such as adding inconsistencies.