Linguists who study the structures of languages in a systematic way are structuralists (or structural linguists) – so this label basically applies to all linguists who are interested in language structures (not necessarily to those who only study the social roles of languages, or who only study pychological correlates of a narrow range of phenomena, e.g. word meanings).
There are some well-known divisions in linguistics – between innatists and usage-based linguists, between “functionalists” and “formalists” – but these do not affect this basic consensus among all linguists: Studying language structures is interesting, because there are many “deeper” generalizations that are not apparent at first glance. And regardless of the significance of “depth”, studying language structures is necessary simply because they exist – the regularities of syntax that speakers know and obey in language use are far richer than was suspected in the 19th century.
The term “structuralism” is not well-understood by linguists (like so many other terms), and it is often associated with a particular period of the history of linguistics (say, between 1916/1926 and 1957/1965; cf. Peter Matthews’s nice short 2001 book). But that is not enlightening, because at any point in time, there are different people with different ideas. So what is the main idea of structuralism? My claim is that it is the following:
The Main Idea of structuralism
Languages have intricate structures, in phonology, morphosyntax and semantics, which are worth studying for their own sake and which are distinct from substantive aspects of language and language use (acoustic, articulatory, conceptual, truth-functional, pragmatic and social aspects).
This idea is quite uncontroversial in contemporary linguistics, so it is something that is worth highlighting and cherishing, in view of all the other controversies (I think that it is on a par with the general agreement on the biological nature of capacity for language, linguisticality, though the latter is more trivial).
In the 19th century, languages were primarily studied from a historical point of view, and the idea that structures are different from substance was only beginning to be appreciated (e.g. by Georg von der Gabelentz in his 1891 book, but this was not widely read at the time and became famous only later). A greater appreciation of the inherent interest of studying languages structures began only in the 1930s, when American structuralists were debating the nature of the phoneme (crucially different from phonic substance), when Jakobson (1932) started applying ideas of “markedness” to verbal categories, and when some German linguists started to wrote about “semantic fields” (see Geeraerts 2010: Ch. 2). All these phenomena had been studied before, but usually with the goal of elucidating their diachronic developments. What was new in structuralism was the interest in synchronic systems for their own sake (this was also called “autonomy of linguistics”).
How did the term “structuralism” acquire its negative connotations (which I hope to dispel with this blogpost)? Others are more qualified to trace the historical developments in detail, but there are at least three factors:
(i) “structuralism” as a term was used more widely, also in philosophy, literary studies and anthropology, with debates that are typically unrelated to the debates in linguistics (these debates gave rise to the idea of “post-structuralism overcoming structuralism”)
(ii) in linguistics, Chomsky (1964) distanced himself from “taxonomic phonemics” (radically opposing his “transformational” approach to the earlier ideas), and soon thereafter, “taxonomic structuralism” was widely thought to have been superseded by the Chomskyan approach.
(iii) when a new American functionalism arose in the 1980s (associated with names like Givón and Bybee), these authors continued to use the term “structuralism” in a negative sense (e.g. Noonan 1999; Givón 2013)
However, while there are clear strategic differences between usage-based, generative and functional linguists (i.e. they are pursuing different research strategies), there is no difference in “doctrine” – everyone agrees with the Main Idea of structuralism (as formulated above).
Chomsky put it as follows in 1972:
“A central idea of much of structural linguistics was that the formal devices of language should be studied independently of their use. The earliest work in transformational-generative grammar took over a version of this thesis, as a working hypothesis. I think it has been a fruitful hypothesis.“ (Chomsky 1972: 119)
And David Pesetsky (syntactician at MIT) wrote the following for a 2015 conference in Athens:
“The Most Basic Facts (MBF) of our field
The linguistic capacity of every human being is an intricate system, full of surprises but clearly law-governed, in ways that we can discern by scientific investigation. Though we still have much to learn about this system, a great deal has been discovered already.”
This is again not much different from what I describe as “the main idea of structural linguistics”. And indeed, it was a Facebook post by David Pesetsky that gave me the idea for this blogpost. There was some discussion of what to call non-psycholinguistic, non-sociolinguistic research on languages – and David suggested that it should be called “structural linguistics”. In the Athens 2015 handout, Pesetsky explicitly says that the Chomskyan generative grammar approach is a “second order phenomenon, presupposing the [structuralist] Most Basic Fact”.
The 1964 idea that a “transformational” approach is radically different from a “taxonomic” approach has long disappeared. Of course, differences remain between authors such as Bloomfield and authors such as Jackendoff, but these relate to issues that are quite independent of the Main Idea of structuralism (in particular, to the confidence that we can have in attributing mental reality to the linguistic structures that we discover; regardless of this, issue linguists continue to study language structures with the same methods as did Bloomfield, Trubetzkoy and others in the first half of the 20th century).
When Givón, Bybee and others diverged from the American generative mainstream in the 1980s, they came to emphasize the dynamic nature of language structures, in an interesting rapprochement with pre-structuralist linguists like William Dwight Whitney (1827-1894) and Hermann Paul (1846-1921). Grammaticalization was thought to be incompatible with structuralism, with its rigid, discrete, abstract structures. Bybee (2010: 112) puts it as follows:
“After observing the process of grammaticalization and the mechanisms that propel the process, as well as the gradualness of change and variability of grammatical constructions, my view of the nature of grammar changed completely from what I had been taught. The structuralist and generativist view of discrete, abstract structures and rules is simply not compatible with the dynamic and variable facts of grammaticalization. In contrast, a grammatical theory based on constructions and allowing for usage-based variability among the instances of constructions …, is well suited to the representation of ongoing grammaticalization.”
But note that the Main Idea is not that the structures and rules of languages should be “discrete” and “abstract” – the Main Idea is that the structures are distinct from the phonic and conceptual substance, and worth studying in their own right. This is not something that Bybee, Givón, Croft, Goldberg, or any other functionalist would disagree with, though they have been more interested in dynamic explanations than in synchronic analyses. (I should say that I do not quite understand what Bybee means by “abstract” in this passage – she does not deny the existence of constructions, which are by definition “abstract”, I would think.)
Notice that Bybee, Givón and others do not disagree with Chomsky, Jackendoff and others about the mental reality of the phenomena they study – whereas they do seem to disagree (at least concerning research strategies) with Bloomfield and de Saussure (who were not mentalists). So the differences can be described as follows:
structuralists: e.g. Bloomfield, Jackendoff (but not Hermann Paul)
mentalists: e.g. Jackendoff, Bybee (but not Bloomfield)
dynamicists: e.g. Hermann Paul, Bybee (but not Jackendoff)
It is true that some functional linguists seem to have gone as far as denying the reality of synchronic grammars, in particular Hopper (1987); see the discussion of “autonomy” in Croft (1995) and Newmeyer (1998). But I think that some of the radical claims at the time were not clear, because the simple undisputed fact that many languages have complex inflectional paradigms with irregularities suffices to demonstrate that there are autonomous structures that are worth studying, regardless of what one might perhaps say about phonological structures (“there is no such thing as phonology, it’s all usage”) or semantic structures (“there is no difference between semantic knowledge and conceptual/encyclopedic knowledge”).
At this point, I should mention another inspiration for this blogpost, my close colleague and friend Matthew S. Dryer. In an important 2006 article on the nature of linguistic theories, he writes about the connection between approaches to grammar writing and structuralism:
“Although structuralism had faded from centre-stage during the early 1960’s as a significant approach to theoretical questions, especially in the United States, it had remained entrenched in many places among people whose central interests were descriptive, and as a result, it continued to have a major impact on language description well into the 1970’s.” (Dryer 2006: 209)
This is also reflected in Matthew’s own tajectory, having been trained by structuralist linguists at the University of Toronto, before moving to UCLA for a PhD (with Edward Keenan, who was himself a reluctant universalist, with no great trust in innateness of categories). Matthew told me that he never thought of what he was doing (language description, plus worldwide grammatical comparison, with an interest in functional explanations where possible) as any different from the sort of “structuralism” that was taught at the University of Toronto around 1970.
Thus, it seems clear that what changed after the 1960s was not how we analyze particular languages (as autonomous systems, with categories that are independent of phonic or conceptual substance), but how we think we can explain the systems that we find. So if we distinguish carefully between (structural) analysis and (system-external) explanation (as urged by Dryer 2006), we can agree to disagree about explanations, but we have no reason to disagree about analyses.
While pre-1960s linguists were interested in analyses but not much in further explanations, interest then shifted toward external explanations – either in terms of an innate grammar blueprint, containing uniform building blocks for all languages (starting with Chomsky 1965; Chomsky & Halle 1968, and others), or in terms of dynamic principles of language change and language use (starting with Givón 1979, Bybee 1985, and others, very much inspired by Joseph Greenberg’s work). This should not be an either-or distinction either (because a comprehensive theory will surely include both innate components and adaptive components), but in practice, there are big differences with respect to the strategy that one should adopt. They can be described as follows:
innatists: Chomsky, Jackendoff, Rizzi, Pesetsky
dynamicists: Hermann Paul, Greenberg, Givón, Bybee, Lehmann, Croft
But there is no difference with respect to the Main Idea of structuralism. In fact, one could even add to the broad consensus another idea that is also fairly widespread, but distinct from the Main Idea:
The Second Idea of structuralism
The core structures of languages are broad classes, often defined by binary feature values, which can be revaled by in-depth analyses. Linguists should aim for analyses that unify the apparently disparate phenomena and bring the broad classes to light.
Many linguists will be surprised to see this labeled as an “idea of structuralism”, because it is so deeply entrenched in our discipline. For many decades, virtually every phonology or syntax class has urged students not to “miss a generalization”. Unified analyses are prized everywhere. For example, Dąbrowska (1997) finds that upon closer inspection, the seemingly disparate uses of the Polish Dative have a motivation that unites them (partly following Goldberg’s (1995) work on the English Double-Object construction). In a generative framework but otherwise quite similarly, Hole (2005) proposed a “unified voice account” that “reconciles” possessor and beneficiary datives in German. For phonology, Passino (2013) proposes “a unified account of consonant gemination in external sandhi in Italian” (many more cases of unified analyses in different approaches can be easily found in the Glottolog bibliography; simply type “unified” into the “title” search field).
Why do we all value unified analyses? There does not seem to be good evidence that such “broad classes” have mental reality (in cases where they are not obvious and productive), but mentalists still seem to have confidence in them. (There may often be an implicit idea that analyses should be economical, because the human mind prefers economical storage, but this seems to be quite wrong.) So I do not have a good explanation for the popularity of unified analyses, and personally, I do not endorse the Second Idea. But I do admit that many such analyses have a certain beauty, and many of the papers are interesting to read and think about (regardless of their truth). Maybe it’s simply part of my upbringing as a linguist, and a certain sense of orderliness, that attracts me to them. In any event, the Second Idea does not divide linguists, so even if we say that it, too, is a crucial part of structural linguistics, we are all structuralists.
I close with a quotation from Givón (2013), where he says the same thing: While we may differ in how we propose to go beyond our structural analyses, they are the foundation of everything:
“All of us … acknowledge that to transcend description one had better first learn to describe. Fundamentally, all decent functionalists, cognitivists, historical grammarians, child-language scholars, typologists, anthropological linguists, neuro-psycho-linguists, and evolutionary linguists are structuralists plus.” (Givón 2013: 413)
Bybee, Joan L. 1985. Morphology: A study of the relation between meaning and form. Amsterdam: Benjamins.
Chomsky, Noam A. & Morris Halle. 1968. The sound pattern of English. New York: Harper & Row.
Chomsky, Noam A. 1965. Aspects of the theory of syntax. Cambridge, MA: MIT Press.
Chomsky, Noam A. 1972. Some empirical Issues in the theory of Transformational Grammar. In Goals of linguistic theory, ed. by Stanley Peters, 63–130. Englewood Cliffs, N.J.: Prentice-Hall.
Croft, William. 1995. Autonomy and functionalist linguistics. Language. Linguistic Society of America 71(3). 490–532. https://doi.org/10.2307/416218.
Dąbrowska, Ewa. 1997. Cognitive semantics and the Polish dative. Berlin: Mouton de Gruyter. https://doi.org/10.1515/9783110814781
Dryer, Matthew S. 2006. Descriptive theories, explanatory theories, and basic linguistic theory. In Felix K. Ameka, Alan Dench & Nicholas Evans (eds.), Catching language: The standing challenge of grammar writing, 207–234. Berlin: Mouton de Gruyter.
Geeraerts, Dirk. 2010. Theories of lexical semantics: An introduction to the history and current state of theories of word meanings. Oxford: Oxford University Press.
Givón, T. 2013. Beyond structuralism: Should we set a priori limits on our curiosity? Studies in Language. John Benjamins 37(2). 413–423. https://doi.org/10.1075/sl.37.2.05giv.
Givón, Talmy. 1979. On understanding grammar. New York: Academic Press.
Goldberg, Adele E. 1995. Constructions: A construction grammar approach to argument structure. Chicago: The University of Chicago Press.
Hole, Daniel. 2005. Reconciling “possessor” datives and “beneficiary” datives – Towards a unified voice account of dative binding in German. In Maienborn, Claudia and Wöllstein, Angelika (eds.), Event Arguments: Foundations and Applications, 213-242. Berlin, Boston: De Gruyter.
Hopper, Paul. 1987. Emergent grammar. Berkeley Linguistics Society (13). 139–157.
Jakobson, Roman. 1932. Zur Struktur der russischen Verbums. In. Pražskỳ Linguistickỳ Kroužek.
Matthews, Peter. 2001. A short history of structural linguistics. Cambridge: Cambridge University Press.
Newmeyer, Frederick J. 1998. Language form and language function. Cambridge, MA: MIT Press.
Noonan, Michael. 1999. Syntax. In Mike Darnell, Edith A. Moravcsik, Frederick J. Newmeyer, Michael Noonan & Kathleen Wheatley (eds.), Functionalism and formalism in linguistics, vol. I, 11–30. Amsterdam: Benjamins.
Passino, Diana. 2013. A unified account of consonant gemination in external sandhi in Italian: Raddoppiamento Sintattico and related phenomena. The Linguistic Review 30. 313-346. De Gruyter Mouton.
we are taught in our classroom that structuralism has its origin in the linguistics of Ferdinand de sassaure, but the person who is credited with popularising this concept across the Europe is Roman jakobson. and we are also taught that with the rise of deonstruction and the seminal essay at Johns Hopkins university in 1967 by Derrida that the poststructuralist phase began. now this is my question? how do we understand this paradigm shift from structuralism to poststructuralism in linguistics (not in literary criticism first)?
One point that Martin eloquently makes in this blogpost (I think) is that in linguistics there never was any “post-structuralism”—we linguists have all been structuralists the whole time. This sort of gets to the heart of how people trained as “linguists’ pretty reliably differ from (most) scholars in fields like psychology, anthropology, sociology, history, philosophy, and literary criticism, among other fields.
As for Derrida’s famous deconstruction of Saussure, I actually think it is very insightful, but in itself it really is more a contribution to philosophical metaphysics than to linguistics. At least, that’s how it’s been received — linguists don’t typically find much of practical use in Derrida, not least of all because they were not the audience Derrida was speaking to. Also, it’s probably a mistake to assume that just because JD deconstructed FdS, that JD wasn’t essentially a structuralist too.
Here’s Tom Shippey at a Tolkien conference (emphasis mine):
So my story about the preference for unified analyses would be that if the similarities motivating a unified analysis are unlikely to be coincidental, then they are worthy of some kind of explanation, and if they lack a clear functional or historical one, then the representational format of knowledge of language is the next candidate.
NPs having the same overall structure regardless of ‘phi-features’ and case-marking being perhaps the most luridly obvious example. More questionable would be the failure of non-nominative subjects in Icelandic to trigger person-number agreement, which I tried to make a consequence of the framework in my 1982 paper, but it could perhaps be a consequence of the fact that once upon a time they were not subjects. Regardless of what the best explanation for such things turns out to be, if people don’t focus on the suspicious generalizations, no explanations for them will ever be found.
What about historical accidents? Take the case of “reduce/reduction” and “produce/production”. Aronoff claimed that one needs a morpheme “-duce” to explain the similarities of behaviour, but this can be due to historical accident. For subject case and agreement, it seems that this is an entrenched Indo-European pattern, which is not found in Trans-Himalayan (as Bickel has shown nicely in various publications dealing with the languages of South Asia). So again, maybe only a historical accident?
Yes, figuring out what is or isn’t a historical accident can be tricky, much tricker than people have perhaps realized. For example, perhaps the failures of Icelandic oblique subjects to be agreement triggers that I tried to make follow from principles in 1982 are just historical accidents caused by the fact that once they weren’t subjects. Ironically, this position is undercut if Barðdal and Eythórsson are basically right about oblique subjects being an ancient Germanic and even IE heritage, because the phenomena are just too persistent.
In my 2019 LFG paper I back off from the 1982 analysis on the basis of too much discrepant data, but still propose a way to make a ‘don’t agree with an oblique subject’ reasonably easy to state, and therefore plausibly persistent.
Otoh NPs are very solid, since there are plenty of examples of NPs with different combinations of case and phi features going the same way under changes in NP structure.
In general, I think it is a theme that deserves a lot more attention that it seems to me to get (the soft diachronic underbelly of synchronic linguistics?)