In the 1960s, a view of language became famous according to which key aspects of grammatical structures are innate and “grow” in the child (rather than being learned). This came in two prominent versions: the “formal and substantive universals” of Chomsky (1965), and later the “principles and parameters” of Chomsky’s (1981) Government and Binding (GB) approach. But in the 21st century, there seems to be less and less certainty about the idea of innate grammatical structures (called “universal grammar”, or “innate grammar blueprint”), and in this blogpost I would like to highlight a few remarkable shifts of opinion among prominent generative grammarians. (Of course, there have always been skeptics outside of the generative school; these are not mentioned here.)
Perhaps the most famous defender of innateness idea (apart from Chomsky) was Steven Pinker, whose 1994 book “The language instinct” (written for general readers, in a non-technical way) was very important in advertising the Chomskyan programme. At the time, Pinker thought that much of grammar was innate, and that acquisition could be reduced to setting a few parameters, along the lines of Chonsky’s GB approach: “If this theory of language learning is true, it would help solve the mystery of how children’s grammar explodes into adultlike complexity in so short a time. They are not acquiring dozens or hundreds of rules; they are just setting a few mental switches.” (Pinker 1994: 112). However, Pinker no longer believes so – the following is from a 2020 video interview:
“Language acquisition consists of nothing than setting a small number of parameters (e.g. is the language head-first or head-last; are subjects obligatory or can they be dropped)… Now it turned out that nobody suceeded in reducing language acquisition to the setting of a few parameters… It’s only if there’s a short list of parameters that the theory can explain language acquisition… If the list of parameters can keep growing indefinitely, I don’t think it’s going to work very well. And indeed it has been abandoned, even by people in the Chomskyan camp who first proposed the parameter-setting theory… We know that there’s an enormous amount of learning of vocabulary… Ray Jackendoff has proposed a theory…
So Pinker now seems to be endorsing the kind of “constructionist” view that has been prominently defended by Adele Goldberg, Joan Bybee, Ray Jackendoff and others, where there is a large lexicon forming a network with links of various sorts.
Another prominent author is Julie Ann Legate (editor of the influential journal NLLT), whose recent (2021) paper on the typology of passive-like voice constructions is very explicit. She notes that various proposed restrictions on possible (passive-like) voice constructions in the earlier literature have not held up, and concludes that there are few restrictions that could be attributed to innate structures:
“Overall, our theory must allow these properties of the (canonical) passive to vary independently, rather than positing necessary dependencies between them. This is indeed achieved on an approach wherein there is little or no innate information specific to voice in the language faculty, but rather the properties of each voice construction must be learned on the basis of evidence. For an impoverished language-specific innate component, see, for example, Fitch et al. (2005), who argue on independent grounds that Universal Grammar consists only of the operation Merge, the operation that builds binary-branching structure …” (Legate 2021: §4).
Among those generative grammarians who have moved away from the earlier standard view is also my first syntax professor, Hubert Haider (who was teaching generative syntax at the University of Vienna in the early 1980s, when I was an undergraduate student there). Haider is famous for his contributions to German generative syntax (e.g. Haider 1993; 2010), but in a recent paper about “universal grammar”, he writes:
“Chomsky (1975) contemplated this kind of decision problems for a learner, but his original solution – an innate language acqusition device resting on an innately preconfigured grammar … – has remained unfounded. Nobody has ever been able to produce immediate and compelling evidence in favour of the string nativist hypothesis. Eventually, even its proponent (Chomsky 2011) prefers to abandon it.” (Haider 2019: 369)
Next, Amy Rose Deal is a prominent syntactician at UC Berkeley, who has made important constributions to the study of the Sahaptian language Nez Perce as well as to general syntax (in a Chomskyan framework). In a 2016 paper on split ergativity, she concludes that Nez Perce split ergativity is “syntactic”, rather than the “morphological” split ergativity that was highlighted by work such as Legate (2014). She recognizes that the different types of split ergativity obey the same scales, but she concludes: “What is prima facie the same type of split ergativity may arise by morphological means in some languages but by syntactic means in others “ (2016: §6). But this leaves her without a Chomskyan explanation of the scale effects (or “hierarchy effects”). Quite reasonably, she accept that these effects have a functional explanation (as summarized in my recent paper, Haspelmath 2021):
“Why should the same distribution of ergative and nominative arise by different mechanisms in different languages? Why should some lan- guages do by morphological means the exact same thing that other languages do with person-sensitive assignment of abstract case features? A deeper fact must be at stake on the relative markedness of the various person features, outside of the particular vocabulary of any one grammatical module. It therefore seems to me quite reasonable to conclude that hierarchy effects ultimately must arise external to the grammar itself, from the organization of human cognition and communication – a conclusion in line with various approaches that locate the origin of these effects extra-grammatically (i.a. Silverstein 1976, Dixon 1979, DeLancey 1981, Newmeyer 2002, Haspelmath 2008).” (Deal 2016: §6)
By “external to the grammar itself”, Deal means that the causal factor is not the innate grammar blueprint. Of course, the regularities are grammaticalized in the particular languages, but the explanation is efficiency of coding, as has long been proposed for another prominent scale effect, differential object marking. Deal concludes that there is no role for an innate grammar blueprint in explaining these generalizations, in line with Chomsky’s 21st-century views:
“… this approach to the status of hierarchies echoes Chomsky (2005)’s view of language design as arising from the confluence of an extremely simple UG component with a range of Language-independent, “third factor” effects, some of them representing aspects of general human cognition.”
An even more prominent author is Richard Kayne, one of the editors of Oxford’s Handbook of comparative syntax, who made an interesting comment on my earlier blueprint-skeptical paper (Haspelmath 2007) in a footnote of his 2013 paper on (generative) comparative grammar:
“Haspelmath (2007) expresses skepticism as to the cross-linguistic validity of notions like adjective, affix, clitic, passive, pronoun, word and others. Although he fails to distinguish with sufficient clarity between “pre-established” and “universal”, his skepticism is to some extent congenial to work in the generative syntax tradition that has questioned the primitive status of such notions. Cross-linguistically valid primitive syntactic notions will almost certainly turn out to be much finer-grained than any that Haspelmath had in mind.” (Kayne 2013: 136, n. 20)
By “fine-grained“, Kayne presumably means “abstract“, so that actually existing categories are composite, composed of elements of a much smaller set of abstract features (cf. Adger’s 2021 response to my 2021 paper, where he claims that “one can build richer sets of categories out of poorer ones”). Kayne is very vague here, but he clearly intends to say that UG (universal grammar) is much poorer than was assumed by many in the 20th century.
A prominent philosophically-inclined author is Norbert Hornstein, who reacted in 2018 to an earlier blogpost of mine, where I noted that 20th century generativists adopted a natural-kinds approach with building-block uniformity, but that they have “no clear criteria for assessing whether a feature should be assumed to be a natural kind”, and that there didn’t seem to be many convergent results. He writes on his blog (February 2018):
“I found the whole discussion [in the blogpost https://dlc.hypotheses.org/1012] to be disconcertingly convincing and believe me when I tell you that I did not expect this. Haspelmath and I do not share a common vision of what linguistics is all about. I am a big fan of the idea that the faculty of language is richly structured and contains at least some linguistically proprietary information…”
But Hornstein ends up saying that while he has great hopes that structural universals will be confirmed, “our theoretical understanding of features and categories is much poorer than our understanding of structural universals”. Hornstein has been an anthusiastic supporter of Chomsky’s Minimalist programme, but he does not think that those linguists who still (maybe implicitly) assume a rich set of innate categories will contribute substantially to this. Thus, 21st century generative linguistics does not need a rich innate blueprint for grammar.
Another philosophically inclined generative linguist is José-Luis Mendívil-Giró (of the University of Zaragoza). In a 2020 blogpost, he writes:
“In fact, generative grammarians do not assume that there is “a rich innate grammar blueprint”. (Where would it have come from?) Nor do they assume that languages are composed of innate building blocks of grammar.”
And in a 2019 interview with me, Gillian Ramchand (of the University of Tromsø) said:
“I think you overstate the innateness thing. Nothing I actually do in practice would change if I discovered tomorrow that everything that all languages have in common is highly abstract, cognitive and not specific to language.”
I also heard from several generative linguists in Twitter discussions, which I cite here to enrich the picture. Radek Šimík (Charles University) told me last year on Twitter: “Mainstream generative grammar need not assume universal categories. There need not be a DP in every language (many people argue there is none in some languages). Similarly so for CP, TP (many generative grammarians assume tenselessness, which pretty much maps to no TP).”
Similarly, Dennis Ott (University of Ottawa) said in 2020 on Twitter: “Generative grammar requires no firm commitment beyond the obvious (some biological difference between linguistic vs. non-linguistic creatures), which is what everybody assumes.” And he continued:
“There’s plenty of work that entertains precisely the opposite idea [to categorial universalism], namely that these categories are not universal. What the truly universal categories are, and at what level of abstraction they can be found, is just an empirical question.”
So I do think that there is a widespread tendency for generative grammarians to reject or at least downplay the earlier idea of rich innate structures. I had noted this in an earlier blogpost, but I am repeating the point here with more quotations, because some people have questioned whether there has been a 21st-century shift among generativists. In particular, Adger (2021) claims that the natural-kinds programme has not been abandoned, and Mendívil-Giró (2021) even says that Chomsky has not given up the assumption that there are many substantive universals (such as the distinctive features in phonology), because he never made these assumptions.
However, I am certainly not alone in saying that Chomsky changed his mind. In a 2016 article, Tecumseh Fitch described Chomsky’s shift as follows:
“In the 1960s, Chomsky’s conception of the biological basis of language and Universal Grammar was that quite complex and human-specific rules must be, somehow, innately wired into the human brain… But today, in another revolutionary reinvention of linguistics, he has reconceptualized the core of language as consisting of a single very powerful but abstract capacity that is unique to humans.”
According to this new view (Hauser et al. 2002; and also Fitch et al. 2005, which was already cited by Legate above; plus Chomsky 2005; 2011; and passim), there is still something that is innate and specific to Human Language, but it says nothing about the components of grammar (e.g. syntax vs. morphology), the rule types, or about the kinds of features or categories that languages can have. In other words, most of the concepts that linguists have been discussing since the 1960s in innatist terms are not thought to be innate anymore. So at the very least, the innate component is greatly “impoverished” (as Legate 2021: §4 puts it).
And such an impoverished view of innateness is in striking contrast with what many generative grammarians are actually doing. Norbert Hornstein puts it like this: “Much of what interests Chomsky (and me) is not currently high up on generative grammar’s research agenda”, he writes in a 2018 blogpost. And Hornstein even agrees with my earlier statement that “Chomsky’s thinking since at least 2002 is not really compatible with the practice of mainstream generative grammar”.
So what is it, then, that justifies this practice? Why can we still assume that categories like DP, TP, vP, PP (or features like [+wh]) are the same across languages? Are these linguists who are abandoning most of Chomsky’s programme from the 1960s through 1990s still “generative grammarians”, or are they ex-generative grammarians? How can they continue to work with the assumption of uniform building blocks, if these are not innate? I am as puzzled as was back in 2018.
Adger, David. 2021. On doing theoretical linguistics: A reply to Haspelmath. Theoretical Linguistics (to appear). (https://ling.auf.net/lingbuzz/005691)
Chomsky, Noam. 1975. Reflections on language. New York: Random House.
Chomsky, Noam. 2011. Language and other cognitive systems. What is special about language? Language Learning and Development. Routledge 7(4). 263–278. (doi:10.1080/15475441.2011.584041)
Chomsky, Noam A. 1965. Aspects of the theory of syntax. Cambridge, MA: MIT Press.
Chomsky, Noam A. 1981. Lectures on government and binding. Dordrecht: Foris.
Chomsky, Noam A. 2005. Three factors in language design. Linguistic Inquiry 36(1). 1–22. (doi:10.1162/0024389052993655)
Deal, Amy Rose. 2016. Person-based split ergativity in Nez Perce is syntactic 1. Journal of Linguistics. Cambridge University Press 52(3). 533–564. (doi:10.1017/S0022226715000031)
DeLancey, Scott. 1981. An interpretation of split ergativity and related patterns. Language 57. 626–657.
Dixon, R.M.W. 1979. Ergativity. Language 55. 59–138.
Fitch, W. Tecumseh. 2016. Reinventing linguistics – again (Review article of “What kind of creatures are we?”, by Noam Chomsky). Inference: International Review of Science 2(3). (https://inference-review.com/article/reinventing-linguistics-again)
Fitch, W. Tecumseh & Hauser, Marc D. & Chomsky, Noam. 2005. The evolution of the language faculty: Clarifications and implications. Cognition 97(2). 179–210.
Haider, Hubert. 1993. Deutsche Syntax – generativ: Vorstudien zur Theorie einer projektiven Grammatik (Tübinger Beiträge Zur Linguistik 325). Tübingen: Gunter Narr Verlag.
Haider, Hubert. 2010. The syntax of German. Cambridge: Cambridge University Press.
Haider, Hubert. 2019. An anthropic principle in lieu of a “Universal Grammar.” In Brown, Jessica M.M. & Schmidt, Andreas & Wierzba, Marta (eds.), Of trees and birds: A festschrift for Gisbert Fanselow, 363–381. Potsdam: Universitätsverlag Potsdam. (doi:10.25932/publishup-42654)
Haspelmath, Martin. 2007. Pre-established categories don’t exist: Consequences for language description and typology. Linguistic Typology 11(1). 119–132.
Haspelmath, Martin. 2008. Frequency vs. iconicity in explaining grammatical asymmetries. Cognitive linguistics 19(1). 1–33.
Haspelmath, Martin. 2021. Role-reference associations and the explanation of argument coding splits. Linguistics 59(1). 123–174. (doi:10.1515/ling-2020-0252)
Hauser, Marc D. & Chomsky, Noam & Fitch, W. Tecumseh. 2002. The faculty of language: What is it, who has it, and how did it evolve? Science 298(5598). 1569–1579. (doi:10.1126/science.298.5598.1569)
Kayne, Richard S. 2013. Comparative syntax. Lingua 130. 132–151. (doi:10.1016/j.lingua.2012.10.008)
Legate, Julie Anne. 2014. Split ergativity based on nominal type. Lingua 148. 183–212.
Legate, Julie Anne. 2021. Noncanonical passives: A typology of voices in an impoverished Universal Grammar. Annual Review of Linguistics 7(1). (doi:10.1146/annurev-linguistics-031920-114459)
Mendívil-Giró, José-Luis. 2021. On the innate building blocks of language and scientific explanation: A reply to Haspelmath. Theoretical Linguistics (to appear). (doi:https://ling.auf.net/lingbuzz/005708)
Newmeyer, Frederick J. 2002. Optimality and functionality: A critique of functionally-based optimality-theoretic syntax. Natural Language & Linguistic Theory 20(1). 43–80.
Pinker, Steven. 1994. The language instinct: How the mind creates language. New York: William Morrow.
Silverstein, Michael. 1976. Hierarchy of features and ergativity. In Dixon, Robert M. W. (ed.), Grammatical categories in Australian languages, 112–171. Canberra: Australian National University.
I have a question which maybe shows my ignorance It’s true that most people in US have been taught syntax on the university level using generativist approaches. However, the theory is only 50 years old or so, and it seems there was never an overwhelming evidence for it. Of course, it could be mentioned as a theory, but why was it taken for proven so quickly? If I’m right, it took the theory of evolution much longer to be accepted as mainstream and taught as a basis for biology, including human biology, despite overwhelming evidence and great explanatory power…?
In the meantime, David Adger has replied in a blogpost of his own: https://davidadger.org/2021/04/08/are-generative-grammarians-abandoning-innateness/ – and I have written another blogpost answering David’s: https://dlc.hypotheses.org/2490
How much of the work in the generative tradition still makes sense if generative linguists choose to explicitly reject innate categories such as DP, TP, vP and so forth? I ask because from this post, it seems like a lot of modern Generative work does not emphasize the importance of such innate categories, and some directly state that innate categories are not necessary for Generative work, so this seems like an important question.
Relatedly, what, if anything, currently unites the Generative tradition, and in particular, what differentiates it from non-Generative traditions? From my understanding, at least historically, the idea of innate linguistic categories was fundamental to the Generative tradition, and the rejection of innate linguistic categories was equally important to many non-Generative approaches. If the Generative tradition is now de-emphasizing the existence of innate, purely linguistic categories, it seems like a major barrier between the different traditions is lost. Might this allow for more cooperation between Generative and non-Generative researchers?
Yes, one would hope that the barriers will disappear, but there is a lot of technical jargon that many people feel they need, because they were taught these techniques in their syntax classes. Syntax teachers usually don’t say how speculative all these notations are. So many linguists are confused, and to be on the safe side, they stay within the bigger community (this is my speculative explanation…).
One important thing that mainstream generative grammar assumes to be innate is the primacy of syntax: that the combinatorial power of language comes from syntax, and that semantics and phonology are derived from syntax. This architecture doesn’t follow from anything in general cognition, and no one to my knowledge has argued that it’s learned. In fact, no one to my knowledge has argued that it’s *true*, and there are plenty of arguments, especially from the constructionist camp, that it’s false.
Many thanks for this comment! Yes, there are many unquestioned presuppositions that linguists make and that they don’t have good reasons for. I hope that my blogposts contribute to making younger linguists aware of these issues. In journal papers, one doesn’t usually see these problems, because the authors mostly address the members of their own subcommunity.
See also comments on Twitter (https://twitter.com/haspelmath/status/1378996722823729154) and public comments on Facebook (https://www.facebook.com/martin.haspelmath/posts/10222972036741780).
As far as the last paragraph goes, the original meaning of ‘generative’ in this context meant defining the set of putative linguistic structures by means of some kind of mathematics, independently of whether or not there was supposed to be a fixed set of building blocks (in the original versions, there wasn’t). This sense still predominates in the ‘alternative generative theories’ which can be agnostic about universal categories, although there are various views whose proponents attempt to support them empirically, such as the endocentric X-bar categories vs. exocentric S, recently discussed by Lowe and Lovestrand in https://jlm.ipipan.waw.pl/index.php/JLM/article/view/247.
Great post as always, though rather tragic subject matter. Seems like there is an opportunity for study here: how is the language faculty of these linguists changing as they abandon UG ….
Interesting post. I wonder to what extent there is just less interest in broad ideological stances in fields like syntax. Linguistics as a field has grown and bifurcated in various ways (as most disciplines do). It seems like nowadays fewer people would claim to have a strong grasp of the various types of evidence from disparate areas to make any big claims about “Language.” A strong ideological stance about innatism ends up being kind of cute, but what does it have to do with the price of tea in China? If you’re deep into work on definiteness or tone or causativity, it doesn’t seem like your work will really touch on the question anyways.
Now, the follow-up is, to what extent do people using OT actually believe in the assumptions they make about the lexicon?
Thanks for the comment – but it isn’t true that if you’re “deep into work on definiteness”, the issue of innateness does not come up. On the contrary, if you want to say something about “definiteness” in general, you will have to make a choice between the natural-kinds approach (assuming that some of your descriptive notions are innate) and the comparative-concepts approach. You cannot be agnostic about innate categories and still reject the Greenbergian comparative-concepts approach. Note that I’m not objecting to the “Bakerian/Cinquean/Hallean” innateness approach – it may well turn out to be reight eventually. But it’s incoherent to work with universal features but be agnostic about innateness.