Is there such a thing as a breakthrough in linguistics that, for example, can be likened to the discovery of gravitational waves in physics such that it gives a great boost in understanding how language functions/is organized, what is universal among languages, what causes variation, etc.?

  • How recent? ---
    – Keelan
    Commented Jan 13 at 19:17
  • 1
    Transformers and generative AI? Not intelligence, and not really about understanding, but truly amazing if you think back 10 years ago. Commented Jan 13 at 20:08
  • @Keelen, let's say 5 years. 10 years will also be okay.
    – Shpekard
    Commented Jan 13 at 21:06
  • @AdamBittlingmayer, 1) do you mean transformers as in BERT? Do you know what linguistic knowledge was used for it? 2) I'm specifically interested in theoretical knowledge (≠ practical implementation in my opinion)
    – Shpekard
    Commented Jan 13 at 21:11
  • 2
    10 years is extremely recent in linguistics. You can't expect paradigm shifts to occur and reach wide acceptance in social sciences as they do in physical sciences. There are always multiple ways to analyse something, and much less that can be done to rule out an analysis.
    – curiousdannii
    Commented Jan 15 at 12:04

2 Answers 2


One thing I had to think of is the parallelism between categories, primarily between verb and noun. This is typically attributed to Chomsky (1970, 'Remarks on nominalization'). Chomsky discusses sentences and phrases of the type:

John refused the offer (verb)
John's refusing the offer (gerundive nominal)
John's refusal of the offer (derived nominal)

Chomsky gives evidence that unlike gerundive nominals, derived nominals behave according to the lexicalist hypothesis, meaning they are nouns throughout the derivation; they are stored as separate lexical entries from the verb. However, there are still interesting parallels between refuse and refusal. For example, both take by-phrases. They can also be modified by the same elements, though these are adverbs with verbs and adjectives with nouns:

The offer was refused by John.
The refusal of the offer by John.

John quickly refused the offer.
John's quick refusal of the offer.

Chomsky argued that if the noun and the verb have a separate lexical entry, the similarities in behavior most likely stem from parallels in syntactic structure. Thus, in 1970 he introduced X-bar theory, in which verb phrases and noun phrases have the same internal structure:

enter image description here enter image description here

(In the right-hand side, of is omitted; it can be seen as a case marker. The parallel is a bit muddy because of the use of S and past tense in the left example; for details on this see p. 54 of the 1970 paper.)

I think this idea qualifies as a breakthrough for various reasons. It is not exactly new, but has stood the test of time and continues to yield new results. The DP level that was added on top of the NP for determiners (the/that/... [book]) has a parallel in the verbal domain, namely CP, where we find e.g. complementizers (that/whether/... [he refused]). Tsoulas (1996, 'The nature of the subjunctive and the formal grammar of obviation'), for example, demonstrated parallels between finite CPs and definite DPs. The idea that nouns and verbs share the same structure thus enables us to reduce different notions (finiteness, definiteness) to a single, more abstract notion.

Another more recent example of this is Ritter & Wiltschko (2019, 'Nominal speech act structure: Evidence from the structural deficiency of impersonal pronouns'). It is commonly assumed that above the CP there is a number of functional projections, which are used for example to mark things like topic and focus, and more generally the relation of Speaker and Addressee to what is being said. Ritter & Wiltschko suggest that there are similar projections above the DP. We can see such projections in the fact that the use of definite articles and demonstratives depends in part on whether referents have already been mentioned. This structure is thus being used to describe the relation of Speaker and Addressee to DP referents in the same way the structure above the CP is used to describe the relation of Speaker and Addressee to the CP's propositional content.

However, the idea of parallelism between verb and noun is also used a lot outside the Chomskian tradition. There are semantic properties shared by events (verb-like) and entities (noun-like); they have been discussed in ways that do not rely in any way on X-bar theory but still assume the parallelism between these categories. Here I have to think for example of Champollion (2015, 'Stratified reference: The common core of distributivity, aspect, and measurement'). For example, a well-known property of events is telicity (the having of an endpoint); it can be tested with for and in adverbials (which are atelic and telic, respectively):

John walked for five minutes (atelic)
*John walked to the store for five minutes (telic)
*John walked in five minutes (atelic)
John walked to the store in five minutes (telic)

This can be extended to entities:

The crack gradually widens for 5 meters (atelic)
*The crack gradually widens by 2cm for 5 meters (telic)
*The crack gradually widens in 5 meters (atelic)
The crack gradually widens by 2cm in 5 meters (telic)

Talmy (1988, 'The relation of grammar to cognition') has similar observations. He makes distinctions that are more commonly thought about in the nominal domain, but shows that they apply in the verbal domain, too. I will only give the schematic representation of his ontology here, with examples from the nominal and verbal domain; see the paper for details:

enter image description here

Discrete Continuous
Multiplex, unbounded timber, furniture
(to) breathe
(to) sleep
Multiplex, bounded (a) family
(to) molt (The bird molted)
(a) sea, (a) panel
(to) empty (The tank emptied)
Uniplex (a) bird
(to) sigh

These examples show, I think, that the parallelism between verb and noun tells us something about many different aspects of language. It has significantly simplified our view on the structure of language, universally (at least for those working within the generative tradition), and thus tells us something about the computational model that would be part of the innate capacity for language according to the Universal Grammar hypothesis. But it also enables us to think about ontologies for entities and events in a more abstract manner, which is more about how humans conceptualize the world around them.

I find it difficult to think of more recent potential breakthroughs, at least in theoretical linguistics. There is, of course, a lot of interesting work being done, but it is hard to tell in advance whether new proposals have the generality that many future researchers can feed off it. This may be different in the more applied subdomains.


I think a major breakthrough is the wide availability of digital language resources, many of them under open licences with few restrictions (non-commercial use being a typical one). We have now many freely available corpora of different sizes and annotated for different purposes, as well as the Virtual Language Observatory (VLO) as a search portal for language resources.

There are manually curated CLARIN Resource Families of corpora and other resources sharing some characteristic features (e.g., parliamentary corpora, newspaper corpora, L2 learner corpora, to name just a few).

There is a growing family of uniformly annotated dependency treebanks of many languages, namely the Universal Dependencies Treebanks hosted at LINDAT-CLARIN in Prague.

I think that the wider availability of high quality resources is just catching on and fostering a new kind of linguistic research, and I consider this as one of the major achievements in linguistics in the last decade.

DISCLAIMER: The author of this answer worked in the German CLARIN-D project and is now involved in the German National Research Data Infrastructure text+. He was also involved in building some corpora that are now part of this infrastructure.

  • Thanks for the bounty! A pity it didn’t generate more answers.
    – Keelan
    Commented Jan 26 at 14:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.