How formal linguistics appeared and disappeared from the scene

Linguistic terminology is often confusing, and this may also apply to labels for subcommunities. There is a sizable community of “formal linguists”, and the term has been productive over the last few decades, as can be seen in the list below (the year in parentheses gives the starting date for the series). But what is “formal linguistics”? (Don’t all linguists study the forms of languages?)

West Coast Conference on Formal Linguistics (WCCFL, 1982)

Austronesian Formal Linguistics (AFLA, 1994)

Formal Approaches to Slavic Linguistics (FASL, 1992)

Formal Description of Slavic Languages (FDSL, 1995)

Workshop on Altaic Formal Linguistics (WAFL, 2003)

Berkeley Papers in Formal Linguistics (2018)

When one looks for explanations of what is meant by “formal linguistics” in these venues, one is usually left alone, which is why I felt this blogpost would be useful. For example, the recent WAFL at MIT explains “Altaic”, but otherwise it just hints at “formal aspects”, which does not help:

“Abstracts must be on topics dealing with formal aspects of any area of theoretical Altaic linguistics, including phonology, morphology, syntax, semantics, or pragmatics. The term ‘Altaic’ is understood to include Turkic, Mongolic, and Tungusic languages, as well as Korean, Japanese, Ryukyuan, and Ainu. (https://wafl14.mit.edu/)”

The Wikipedia page for FASL merely notes that it is “is one of the most reputable international conferences”, suggesting that this community does not suffer from a lack of self-confidence, but how does “formal” differ from other kinds of linguistics?

Here is my answer: “Formal linguistics” is a euphemism for “Chomskyan linguistics”, and it implies that linguistic theorizing is based on the idea of a rich set of specific innate mechanisms. But why is a euphemism needed, and why not call it “innatist linguistics” (or “naturalistic linguistics”, to take up Chomsky’s parlance of “naturalistic inquiry”)?

Of course, a kind of linguistics that is primarily defined by its pedigree (Chomsky 1957 for syntax; Chomsky & Halle 1968 for phonology) is not intellectually interesting, so linguists who are united by it may not want to call themselves “Chomskyan” (well, I happily call myself a “Greenbergian”, but maybe because the content of this term is clearer). Chomsky has often used the term “generative grammar” for the kind of research that he and his students have been engaged in, and others have often adopted this term, too (the European and Asian conferences that correspond to WCCFL are called GLOW: Generative Linguistics in the Old World, since 1977).

So how did “formal linguistics” come to be equated with Chomskyan/generative linguistics? It turns out that the term has a fascinating history, and over the last six decades, it has meant two very different things:

(1) c. 1957-1980: formal linguistics = the study of language that is crucially assisted by insights from formal language theory, based on mathematics, logic and computing science (Frege, Ajdukiewicz, Turing, etc.; this is what Chomsky called “formal sciences” in 1995, and what he was deeply interested in in the 1950s).

(2) c. 1981-today: formal linguistics = the study of language in the Chomskyan tradition (which after 1981 did not crucially include reference to formal language theory)

I am not sure how and when the term “formal language” arose in mathematical logic, but in any event, Chomsky used the term “formal” in his well-known writings in the 1950s (e.g. Chomsky 1959, a very mathematical paper that is not read or cited by linguists anymore). The first use of “formal linguistics” was probably in a UPenn book series edited since 1961 by Henry Hiż (who published a book by Zellig Harris in this series, but also some books about tranformational grammar). So the term existed in the 1960s, but not prominently.

I think that there were two key events in the history of the term formal linguistics: the 1977 book “Formal syntax” (Culicover et al. (eds.)), and the 1982ff conference series WCCFL (West Coast Conference on Formal Linguistics). Both of them originally intended to use the term in the sense of formal-language linguistics, but it seems that they ushered in the dramatic semantic change that we have seen since the 1980s.

In the 1970s, the older term “transformational grammar” was felt to be problematic because advocates of “Generative Semantics” had extended the notion of transformation in ways that Chomsky did not approve of (and transformations were no longer so crucial anyway), and in 1976, Peter Culicover, Tom Wasow and Adrian Akmajian organized a conference in Newport Beach, CA, for which they chose the title “Formal syntax”. Peter Culicover recalls:

“We were sitting around and complaining about how all the conferences were being organized by George [Lakoff], Paul [Postal] and Haj [Ross] for generative semantics (at least, that was our perception), and thought, why don’t we have our own conference about things that we found interesting and invite the people who we want to hear from. ‘Formal syntax’ seemed like a good title that would distinguish us from them, although we didn’t analyze it at all. And I think that it was Tom [Wasow] who suggested that we could apply for funding from NSF through the Mathematical Social Sciences Board,  taking advantage of the Stanford connection with Stan Peters. The ‘formal syntax’ title made it clear that the proposal was appropriate to their mission. We did get the funding.”

Then five years later, Ivan Sag organized the first WCCFL at Stanford (maybe with Tom Wasow), after discussions with Frederick Newmeyer. Here is what Newmeyer told me about the reasons for choosing the term “West Coast Conference on Formal Linguistics”:

“1. By 1980, the first American school of functional linguistics (T. Givón, Sandy Thompson, Wally Chafe, et al.) had formed, primarily on the West Coast. In fact, at the time, people often thought of ‘West Coast linguistics’ as functionalism. Ivan and I wanted to make it clear that these people were not part of the conference we were planning.

2. Ivan and Joan Bresnan had recently been hired at Stanford. Both considered themselves to be doing ‘real formal linguistics’, as opposed to mainstream generative grammar. Ivan insisted on the word ‘formal’ in the name of the conference as a coded way of making it clear that people who were doing GPSG and LFG were warmly welcomed.”

Thus, in 1976, “formal” was used as a way of excluding the generative semanticists, and in 1982, as a way of excluding the functionalists and giving special prominence to two approaches (LFG and GPSG/HPSG) that thought of themselves as closer to the spirit of formal-language linguistics than Chomsky’s work of the 1980s. So around 1980, “formal” was still being used in the old sense, but after the publication of Chomsky’s Lectures on government and binding (1981), Chomsky’s less formal way of doing generative grammar became more and more popular. This has not stopped, even after the “minimalist programme” deprived this approach of the status of a “theory” (at least officially), and the label “formal” became more and more associated also with GB/minimalist communities – perhaps because it sounded less pedigree-based than “Chomskyan” or “generative”, though in fact it now has little (if any) independent content.

By 1989, it had become fully clear to advocates of formal-language linguistics that Chomsky and the Chomskyans were no longer interested in the explicitness and rigour that comes with the sort of algebraic formalization that was practiced earlier. Pullum (1989) is a desperate and funny account of how Chomsky gradually abandoned formal-language linguistics in the 1980s, even treating it with scorn. Pullum concluded (1989: 141-142):

“Government-binding syntax no longer makes any pretense at being formally intelligible. It is set to develop into a gentle, vague, cuddly sort of linguistics that will sit very well with the opponents of generative grammar… Current MIT syntax will blend very nicely with the work of the Fuzzies… Students of the 1990s, I predict, will write dissertations on topics like how point of view of participants affects relative strength of barriers, and their mixed GB/Fuzzy thesis committees will be delighted.”

In this sense, formal linguistics has disappeared from the scene of linguistics. Well, it survived in LFG and HPSG circles, but Joan Bresnan later developed interests in Bantu linguistics, stochastic OT and probabilistic syntax, while HPSG is apparently mostly being practiced by computational linguists. (There is still a conference series called “Formal Grammar”, but few linguists seem to know about it. Wikipedia redirects a search for “formal linguistics” to its “formal grammar” page, which is about formal-language studies, not about anything in linguistics in the usual sense.)

Stefan Müller, editor of the forthcoming HPSG handbook, tells me that while HPSG is formal linguistics, this goes without saying, so members of this community are more likely to refer to their conferences as “constraint-based” or “model-theoretic”. The term “formal linguistics” as a community label is now restricted to the Chomskyans, who are not formal in sense (1), but only in sense (2).

Pullum’s observations about Chomsky’s personal intellectual development are scathing and apparently accurate, but his predictions concerning the whole field have not really come true. I think this is because the more recent Chomskyan work has preserved another crucial aspect of Chomsky’s thinking: that the categories and architectures are innate, i.e. that one can study all languages with the same concepts. If a concept like “vP” or “head movement” can be motivated in some way for one language, it is automatically available for all other languages. In fact, this is the default assumption (Chomsky’s (2001: 2) “Uniformity Principle”), which explains why it is often so hard for outsiders to understand the motivation of analyses in the Chomskyan framework. One needs to be a full-fledged community member to appreciate the value of much of this work, because it comes from the big picture of the overall theory of UG, not from particular insights into particular languages. (And this is why all this is relevant to diversity linguistics: Chomskyan linguistics is inherently universalistic, and has been paying a lot of attention to Austronesian, Ataic etc. languages.)

From the perspective of an old-style (pre-1981) formalist, innateness is quite separate from algebraic formalization, and in fact, linguists like Geoff Pullum and Ivan Sag have never claimed universality for their categories. Ivan Sag sat in on one of my 2005 Linguistic Institute classes at MIT, and he told me that he thought that my proposals for functional(-adaptive) explanations of universals were quite compatible with his (HPSG-oriented) view of the world. In the same vein, Stefan Müller’s textbook “Grammhttp://langsci-press.org/catalog/book/195atical theory” (2018) contains an entire chapter (Ch. 13) that critically discusses innateness.

Thus, the term “formal linguistics” in the sense in which it is mostly used nowadays is a confusing misnomer. The original “formal linguistics” has disappeared from the mainstream scene (surviving mostly in HPSG), and the new “formal linguistics” is not “formal” (in any sense that could be clearly described).

But this confusing usage is still spreading. In 2017, Ljutikova & Tatevosov published a paper about “formal linguistics” in Russia, where they explain how it really took off in the 1990s (though they do mention Mel’čuk’s work from the 1960s-1970s, which was non-Chomskyan in that it made no innateness assumption, much like LFG/HPSG). They give no definition of “formal linguistics”, but they talk about “a level of explicitness that allows to regard linguistics as a strict science”, and that it “limits the theorist’s imagination through a framework of clearly formulated initial assumptions” and “logical procedures”. Geoff Pullum would not be impressed, and I think that this paper is typical of the view of the younger generation who simply does not know anymore what algebraic formalization is. All they know is the innateness assumptions of the GB/minimalist period, and they cannot easily think of alternatives.

When I recently asked on Facebook about the meaning of “formal(ist) linguistics”, Yakov Testelets (Russia’s most knowledgeable syntactician, with extensive pre-1991 and post-1991 experience) answered:

“Judging from the current use of term, they use “formal linguistics” for an approach that accepts (explicitly or tacitly) Noam Chomsky’s philosophy of language, i.e. human language is a natural phenomenon and therefore can be studied by the method of the natural sciences.”

Testelets is exactly right, I think – and it should be clear by now that “formal” is the wrong label for this approach. The post-1981 meaning shift was a confusing development.

Unfortunately, linguistics is often confusing – for scholars within the field of linguistics, and of course even more for scientists in other disciplines. If we want to become a more widely respected discipline (and attract more funding, for example), we should probably try to make our contributions more transparent (this also applies to “typology”, which is not a transparent label – so I have suggested that it should be replaced by comparative linguistics). For Chomskyan linguistics, I think that “naturalistic linguistics” would be the best label.

P.S. Using the term “formal linguistics” in the post-1981 sense is not the worst practice. Many Chomskyan syntacticians do not use a label at all for their approach, and they talk just about “(current) syntax” or “syntactic theory”, as if their approach were somehow well-established within the entire field (rather than cutting-edge high-risk research). For example, the journal Syntax only publishes work in generative syntax, and a recently published textbook is called “Introducing syntax”, even though it introduces readers to aspects of generative syntax (of English). This is a little bit like calling your internet address “energy.gov”, as if there were only one country in the world that has a government department for energy (or like calling your language “language”, as if there were only one language in the world).

References

Chomsky, Noam A. 1957. Syntactic structures. ’s-Gravenhage: Mouton.
Chomsky, Noam & Morris Halle. 1968. The sound pattern of English. New York: Harper & Row.
Culicover, Peter W, Thomas Wasow & Adrian Akmajian (eds.). 1977. Formal syntax. New York: Academic Press.
Ljutikova, E. A. & S. G. Tatevosov. 2017. Russkji jazyk i formal’naja lingvistika. [The Russian language and formal linguistics.] Rhema 2017(4). 9–22.
Müller, Stefan. 2018. Grammatical theory: From transformational grammar to constraint-based approaches. 2nd edition. (Textbooks in Language Sciences 1). Berlin: Language Science Press. http://langsci-press.org//catalog/book/25.
Pullum, Geoffrey K. 1989. Formal linguistics meets the Boojum. Natural Language and Linguistic Theory 7(1). 137–143.

 

 

 


8 thoughts on “How formal linguistics appeared and disappeared from the scene

  1. My sense of what ‘formal linguistics’ is/was is a bit different: I see the term as declaring a common interest between people who want formalize sooner (ideally right now) rather than later or never, regardless of attitudes towards the full Chomskyan line on innateness, psychological reality & i-language, a task-specific language-faculty, a Great Leap Forward 60k years ago, etc. etc.. So Montague Grammar is in, in spite of disagreeing with Chomsky on all of these issues. (Old school Montague also raises the issue of formalization by means of notation vs being mathematical; the type system was formalized as a notation, but the rules were written in mathematicalese. This seemed to work for smallish fragments of English, but looks worse as they get bigger (e.g Bennett’s thesis), and is probably unsustainable for typologically diverse languages. For whatever reason, people seem to have stopped trying to pursue this approach). The possibility of attaching Formal Semantics to various kinds of Generative Syntax is an illustration of this, since FS comes from a background where the full suite of Chomskyan assumptions are not accepted. Richard Larson once discussed this in a talk in 2005 about GS+FS as an ‘interfaith marriage’ where there were disagreements about many basic things, but it seemed to work anyway, for now, at least.

    And I am not aware that ‘mainstream’ Minimalists have been appropriating the term Formal Linguistics (unlike Generative Grammar, which many of them do), although much of their current work really does seem formal enough to me (exhibit A perhaps being the formalization by Collins and Stabler. (Stabler’s MG’s aren’t mainstream, I would say, although very interesting).

  2. Martin has pointed out a very good point. The same idea has been continuously advocated in China by a small group of young linguists, most of which have backgrounds or interests in maths, logic and computation. In China, the generative/Chomskyan approach has overwhelmingly taken up the title of ‘formal linguistics’ and other approaches are hardly taught or passed on to the students due to the lack of competent teachers. Formal linguistics is not formal. What a sarcasm.

  3. Martin, you didn’t mention Chomsky’s reply to Pullum. Chomsky actually gives kind of a definition of “formal linguistics”: “…it is true that as pursued, formal linguistics largely concerned itself with weak generation of E-languages, for basically two reasons: the background in metamathematics and automata theory; the fact that these are simple notions, readily investigated by available tools. But formal linguistics is not *defined* in this manner. Rather, it is the study of formal models in abstraction from application; it is formal *linguistics* rather than some other branch of mathematics insofar as one can show (or one hopes) that these models have an application in the study of natural language or provide, however, indirectly, some insight into properties of natural language.” So this (in 1990) is still very much the old notion.

    I am confused not only by the term “formal linguistics” but also the use of “form” in general in linguistics. I would propose to use the term “mathematical linguistics” for formal linguistics in the old sense, “algorithmic linguistics” for generative linguistics (in the sense of “formalized description of language”) and maybe “non-functional linguistics” to replace today’s “formal linguistics”.

    Chomsky, Noam. “On Formalization and Formal Linguistics.” Natural Language & Linguistic Theory 8, no. 1 (1990): 143-47. http://www.jstor.org/stable/4047755.

  4. Although recent (mainstream) generative approaches to syntax are less “formal” in the “formal languages” sense, I think this post misses the extent to which (at least some) practitioners of this approach define themselves specifically in contradistinction to functionalists. That is, the linguistically interesting insights (and maybe “explanations”) are to be derived from the “formal” properties of languages (viz. their *forms* or structures), as opposed to their “functional” properties (viz. how they are used communicatively). There are certainly some people who’ve done work that tries to bridge that divide (Newmeyer, inter alia), and modern phonology seems to be trending toward a more functionalist approach.

    • But what does it mean to say that “insights/explanations” come from “formal properties”? I never understood this. Whyt I do understand, however, is the idea of an innate universal grammar – the idea that linguistic categories and architectures are given in advance, before any language learning. If there were a rich UG (= a rich set of linguistic features given by nature), then one would have an explanation that could be a competitor to functional explanations. “Form” by itself explains nothing (in contrast to function), but rich innate structures might be explanatory. Hence, “formal linguistics” is best understood as meaning “naturalistic linguistics”.

  5. I only started studying linguistics in 1991 (in Prague) but I always associated ‘formal linguistics’ with the HPSG crowd and would only include GB work tangentially. Montague and Partee were where I would send the interested novice. I’ve since moved into the realm of Pullum’s fuzzies so I’m quite surprised the meaning has moved quite so drastically.

  6. As I have commented to you in email, I think it’s worth saying more about the relation between early MIT linguistics and Formal Language Theory, and the subsequent development of that connection. I copy here my own recollections about these things: I wasn’t part of the earliest MIT generation, but I think I knew something about these origins when I got there.

    I had read a couple of Chomsky’s works on formal language theory while still a student at Chicago, and that stuff was certainly considered at the core of the field when I got to MIT (in 1966). The picture then was that Chomsky had demonstrated mathematically that Context-Free Phrase Structure Grammars, production systems deriving from the work of Post, were incapable in principle of producing (generating) all and only the sentences of a natural language (e.g. English). Postal had then, in his 1964 Constituent Structure Monograph, demonstrated that all of the available Structuralist theories of morphosyntax were equivalent to Phrase Structure Grammars, and thus that none of these could provide an adequate theory of language. Some additional generative power was needed, and that was what Transformations provided. That was why we were interested in exploring the formal properties of a theory of Transformational Grammar.

    I should say that a great many of my fellow students at MIT (and I myself) came from earlier backgrounds in mathematics — not really advanced stuff, although for instance McCawley had an MA in math from Chicago, I’d taken several graduate courses, as had a number of my colleagues. Indeed, it was the impression of mathematical rigor that came from the association with formal language theory that drew many of us to the field as it was developing at MIT at the time. It didn’t hurt that there was this really cute analysis of the English Auxiliary in Syntactic Structures, an analysis that relied on non-Phrase Structure mechanisms, while all of the competing (Structuralist) approaches to language seemed to be spinning their wheels uselessly. Actually, my own first teachers in the field , apart from McCawley, were a radically un-reconstructed structuralist (William M. Austin: “I don’t care that your answer is the ‘right’ one. I want to know HOW YOU GOT IT!”) and a kindly American dialectologist fascinated with the distribution of words for “frying pan” in the upper midwest. When I discovered that I wasn’t really going to make it as an academic mathematician, it was pretty clear where to go, and those weren’t the ways.

    It may well be true, as Pullum (2010: “Creation myths of generative grammar and the mathematics underlying Syntactic Structures”) argues that much of Chomsky’s demonstration and the mathematics of SS were wrong, incomplete, or whatever, and thus that the formal groundwork for taking Transformational Grammars as the object of inquiry in the field was weak to non-existent. As is typical of Geoff, the paper is full of anti-Chomsky venom, some of it rather gratuitous in my opinion, but that doesn’t affect the fact that in hindsight, he’s right about a great deal. But that in turn also doesn’t matter here for my point: what does matter is that the sense in which the early generation of MIT linguists thought they were doing “formal linguistics” was grounded in the (supposed) basis of the field in mathematical formal language theory, and the extension of the original results in that context to a formal characterization of the enrichments involved in moving from Phrase Structure Grammar to Transformational Grammar. The course in “Mathematical Foundations” (yielding ultimately Barbara Partee’s long-delayed textbook) continued to be offered and generally taken at least as long as I was around.

    This grounding in Formal Language theory did indeed fade into the rear-view mirror over time. It certainly persisted through the Aspects period, into the late 60s, but probably got lost in the commotion of the Generative/Interpretive Semantics battle. Anyway, my point in all this is that in your historical account, I think you glide much too quickly over the sense of “formal” that the first generations of MIT linguists had.

Leave a Reply to Martin Haspelmath Cancel reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.