The description-comparison approach and the audacious Chomskyan approach (or: how to frame better)

According to George Lakoff, we should frame ourselves, because otherwise we will be framed by others. This is an interesting thought, and indeed, linguists of different orientations constantly frame others and themselves. I have not been happy with many of these framings, and here I’ll make another attempt.

So here’s my new label for the approach to particular and general linguistics that I have found the most productive and interesting: the description-comparison approach. To add some historical associations, one can also call it the Boasian-Greenbergian approach – everyone knows the descriptivist Franz Boas (1858-1942) and the comparativist Joseph Greenberg (1915-2001).

The basic two ideas of this approach are very simple: First, just as children do not need to know other languages to acquire a language, linguists can describe languages without knowledge of typology or innate structures. Languages are described in their own terms (but of course, typological knowledge serves as inspiration and makes the descriptions more transparent).

Second, we need to compare languages and find universals in order to learn about variable aspects of Human Language. Language use is quite uniform across languages (e.g. regularities of turn-taking, as famously shown by Stivers et al. 2009), but language structures – not just the words – differ markedly across the world’s languages. Comparative research can be based on a variety of comparative concepts – category-like comparative concepts, passages of parallel texts, structured elicitation, and so on.

Thus, this approach can be given a very simple name: the description-comparison approach. It contrasts with the Chomskyan approach, where description and comparison are thought to be a single enterprise. Languages are described/analyzed in terms of hypothesized innate building blocks of universal grammar (UG), and the universal claims are in turn based on these descriptions/analyses. In a recent paper on the importance of UG for generative grammar, Charles Reiss states that

“it is by postulating UG that [generative] linguists can treat Japanese and English as relevant to each other”

(I discussed this paper in a recent Twitter thread).

Now while “description-comparison approach” is a conscious reframing of my own approach, “Chomskyan” is of course an old label – one that many Chomskyan linguists are not very happy with, for understandable reasons (“Chomskyan” sounds too much as if there were a single guru that others are following). But what would be a better alternative? If they do not choose an explicit label, the Chomskyans will be framed by others (as per Lakoff’s insight – see above). Linguists used to call the approach “transformationalist” (but then Chomsky abolished transformations), and another common term was “generative” (apparently used by Chomsky himself at some point), but this latter term is too broad, as it also includes approaches such as HPSG which are (in part) openly anti-UG (and/or anti-Minimalist and thus anti-Chomskyan, e.g. here).

A term pair that is definitely not suitable is “LINGUistics” vs. “LANGUistics”, used by Norbert Hornstein on a number of occasions (in 2019, he talks about “LANGers” and “LINGers”). It may be that linguists who adopt the description-comparison approach tend to be more interested in languages and can thus be dubbed “LANGuists”, while Chomskyans tend to be more interested in the human mind, but then languistics should be contrasted with mentalistics or similar. But maybe Hornstein’s usage was not meant quite seriously.

Other commonly used labels are “formal(ist)”, but this is clearly too broad, and “structuralist”, which is sometimes used as an external framing for Chomskyan linguistics, although the early generations of Chomskyans explicitly distanced themselves from this label. Thus, I do not know of a good alternative to “Chomskyan”, and I hope that Chomskyans will come up with one. (Well, maybe “mainstream generative grammar”, or MGG, is a good label for the Minimalist framework. However, it is not so good for the general methodological approach, and it does not work for phonology at all.)

Until recently, I hesitated about “Chomskyan” also for other reasons: I felt that it might be inappropriate because many Chomskyans have given up the idea of UG – perhaps even Chomsky himself. However, it is difficult to interpret Chomsky’s philosophical pronouncements of the last 25 years, and they do not seem to have had much relevance to linguistics. In Chomskyan linguistics, it seems, the existence of a set of universally available (pre-established) features (and categories) is generally presupposed, as reflected in the following quotation from a recent paper:

“what is the inventory of available phi features? Is this inventory universal? … [there is a] general consensus that there is an inventory of features, but languages vary in which features they draw from this inventory.” (Fenger & Kouneli 2023)

Now what about possible alternative labels for the description-comparison approach? There is of course a long tradition of contrasting “formal(ist)” with “functional(ist)” approaches (e.g. Newmeyer 1998; Thomas 2020), and the latter label is sometimes also used for work that does not have particular explanatory goals – but this is confusing. De facto, it may be that there is a community of linguists who are happy with a label such as “functional-typological” (popularized by Givón 1984), or maybe “cognitive-functional” (Tomasello 2003: 5), but most linguistics is descriptive, and “functional” considerations do not play an important role at the level of describing a particular language. Language systems ARE autonomous (Newmeyer 2003), and when “usage-based” considerations come into play, it is at the typological and diachronic level (Bybee 2006). Descriptive grammars (e.g. Haspelmath 1993) describe the forms that languages have for the functions that speakers need to express, so they are necessarily both formal and functional.

To avoid the “functional(ist)” label, I have sometimes contrasted “Chomskyan” with “Greenbergian”, but Greenberg’s name only stands for the comparative side. If we add Boas’s name for the descriptive side, then I have no problem with “Boasian-Greenbergian”, as noted above. These scholars belong to the past (and are not known for controversial politics), so their names can be used in a label of a contemporary approach.

Another term that I used (in a 2010 paper) was “framework-free theory” – this paper has sometimes been cited, but the label has not caught on, perhaps because it does not sound sufficiently positive. The purpose of that paper was to emphasize (following Dryer’s brilliant 2006 paper) that the description of a language can be theoretical without adopting a “theoretical framework” (and that in fact the latter notion is quite problematic). Linguists have the habit of talking in terms of “theoretical approaches”, but this is quite odd – theories are sets of hypothetical ideas that are in principle testable, but linguists typically “adopt” their (theoretical) frameworks, rather than trying to test them. The approaches or orientations (or frameworks) that we adopt should rather be called methodological approaches.

Finally, in a 2014 paper, I used the terms restrictivist (for the Chomskyan approach) and non-aprioristic (for the description-comparison approach). I still find them quite appropriate, because the assumption of a rich set of UG features, categories and architectures (a “grammar blueprint”) makes Chomskyan analyses aprioristic in a way that Boasian descriptions and not. And restrictiveness is frequently cited as a virtue of the UG approach: Chomskyans typically say that only a small subset of logically possible grammars are actually acquirable, which is said to explain both language acquisition and grammatical universals.

But again, like “framework-free”, the label “non-aprioristic” is a negative characterization, and it is better to frame one’s approach in positive terms. Thus, I am now calling my favourite approach the description-comparison approach. This may sound bland, but that’s precisely what is intended: It’s a conservative approach that does not rely on highly unlikely assumptions. It has often been observed that the idea of a rich set of innate features, categories and architectures is not easy to reconcile with the evolvability of human linguisticality. But the idea of a rich UG is certainly not absurd, and one may of course try it out. But basing one’s analyses on pre-established categories is an audacious approach, maybe one that only a bold radical thinker such as Chomsky would come up with. Claims about UG are not easy to test, but they may ultimately turn out to be right in some way. This would certainly be a major discovery.


Bybee, Joan L. 2006. From usage to grammar: The mind’s response to repetition. Language 82(4). 711–733.

Dryer, Matthew S. 2006. Descriptive theories, explanatory theories, and basic linguistic theory. In Ameka, Felix K. & Dench, Alan & Evans, Nicholas (eds.), Catching language: The standing challenge of grammar writing, 207–234. Berlin: Mouton de Gruyter. (

Fenger, Paula & Kouneli, Maria. 2023. Words and features. LingBuzz. (

Givón, Talmy. 1984. Syntax: A functional-typological introduction. Vol. Vol. 1. Amsterdam; Philadelphia: Benjamins Pub. Co.

Haspelmath, Martin. 1993. A grammar of Lezgian (Mouton Grammar Library 9). Berlin: Mouton de Gruyter.

Hornstein, Norbert. 2019. The stupendous success of the Minimalist Program. In Kertész, András & Moravcsik, Edith & Rákosi, Csilla (eds.), Current approaches to syntax: A comparative handbook. Berlin: De Gruyter.

Newmeyer, F. J. 2003. Grammar is grammar and usage is usage. Language 79(4). 682–707.

Newmeyer, Frederick J. 1998. Language form and language function. Cambridge, MA: MIT Press.

Reiss, Charles. 2023. Research methods in armchair linguistics. LingBuzz. (
Stivers, Tanya & Enfield, N. J. & Brown, Penelope & Englert, Christina & Hayashi, Makoto & Heinemann, Trine & Hoymann, Gertie et al. 2009. Universals and cultural variation in turn-taking in conversation. Proceedings of the National Academy of Sciences 106(26). 10587–10592. (doi:10.1073/pnas.0903616106)

Thomas, Margaret. 2020. Formalism and functionalism in linguistics: The engineer and the collector. New York: Routledge.

Tomasello, Michael. 2003. Constructing a language: A usage-based theory of language acquisition. Cambridge, MA: Harvard University Press.

Cite this blog post
Martin Haspelmath (2023, October 12). The description-comparison approach and the audacious Chomskyan approach (or: how to frame better). Diversity Linguistics Comment. Retrieved June 13, 2024, from

8 thoughts on “The description-comparison approach and the audacious Chomskyan approach (or: how to frame better)

  1. Finally, in a 2014 paper, I used the terms restrictivist (for the Chomskyan approach) and non-aprioristic (for the description-comparison approach)

    Perhaps one could contrast the “restrictivist” approach with the “constructivist” approach, if you want a more positive term to replace “non aprioristic”?

    In the restrictive approach, when children learn a language, they must figure out how the patterns in the language they are learning can be connected to the innate categories/features that languages are restricted to choosing from.

    In contrast, in the non-a-prioristic approach, there is no universal, innate set of categories which children must choose from as they figure out the grammar of their language; instead, children *construct* language specific categories based on the experience they have with their language, hence my suggestion of “constructivist”.

    The restrictivist/constructivist terminology I suggest emphasizes the differences in these approaches to child language acquisition, but it also naturally connects to other differences. For example, when writing a grammar of a language, if you aren’t “restricted” to using the hypothesized innate categories of Universal Grammar, it’s natural to create whatever categories seem natural to that language. You don’t have to worry about whether the categories you use to elegantly describe Japanese can be used to elegantly describe English.

    • As an aside, I don’t really understand why the ability to “treat Japanese as relevant to English” could be seen as a good thing. To me, being able to do good work on a particular language without having to be deeply knowledgeable about the structure of every other language in the world is very important, as no single person can plausibly be knowledgeable about all languages. If every language is “relevant” to every other language, you can never be sure that your description of even one language is correct.

      (To me, the idea that the categories of Japanese are NOT relevant to the categories of English is a good thing when writing a descriptive grammar. I remember when I took a Generative Syntax class, we started working on one language, but then we looked at another language which didn’t seem to fit the generalizations we had made. So we invented a new, more abstract rule of universal grammar which was able to handle both languages. Then we went back and redescribed the first language in terms of this more abstract rule, which often didn’t lead to any more correct generalizations in the first language; it was just a hack to “save” the hypothesis of a single universal grammar which made it much harder to understand any individual language grammar.

      I was always very suspicious of this methodology; how can we be confident that any of the abstract rules are right, if whenever some allegedly universal rule seems to be violated in some other language, we can invent another, more abstract rule to save appearances?

  2. Linguists used to call the approach “transformationalist” (but then Chomsky abolished transformations), and another common term was “generative” (apparently used by Chomsky himself at some point), but this latter term is too broad, as it also includes approaches such as HPSG which are (in part) openly anti-UG (and/or anti-Minimalist and thus anti-Chomskyan, e.g. here).

    Chomsky’s views seem to have changed over time, so it might not make sense to talk about a single, unified Chomskyan approach. It might be better to talk about these individual components instead. The transformationalist approach, the generative approach, the Minimalist approach, the Universal Grammar approach, etc.

    • Yes, of course! But on the other hand, there is an amazing amount of homogeneity in the work of Chomsky’s followers – in book after book, dissertation after dissertation, the most recent Chomskyan view is taken as a starting point, and while sometimes modifications to Chomsky’s proposals are made, it is invariably Chomsky’s position that is at the centre and that provides the common ground for the Chomskyan community. This is clearly a social phenomenon as much as a scientific phenomenon, so it’s difficult to avoid the label “Chomskyan”.

  3. Interesting post, as always!

    Regarding the idea of “framework free theory” that you discuss in your 2010 article, what do you think about things like Fluid Construction Grammar? As far as I understand Fluid Construction Grammar, it can be seen as a sort of meta-meta language: it allows you to formalize the constructions of any particular language using a standard format. (Here’s an introduction to FCG if you aren’t familiar with it:

    So for Tagalog, you could create a Tagalog Predicate construction, a Tagalog Topic construction, and a Tagalog Sentence construction; for German, you could create a Prefield Construction, Left Bracket construction, Middle Field construction, Right Bracket construction, and Post-field construction, and similarly for other languages.

    Linguists are free to create whatever constructions they need to describe a particular language using FLC, so it seems to follow the main idea of being “framework free”. But the constructions are formalized in the metalanguage of FLC, which could be seen as being a “framework”?

    • I fully agree that FCG is very close in spirit to what I am saying. FCG practitioners are interested in developing formal systems, something that I don’t do, but I think they are doing it in the right way, with different constructions for different languages. I’m not well-versed in the formalisms, but I know that Remi van Trijp, one of the most prominent FCG exponents, shares many of my views. We were both strongly influenced by Bill Croft, whose (2022) book “Morphosyntax” is another fully compatible approach.

  4. Hey Martin, don’t you think the following passage:

    “it is difficult to interpret Chomsky’s philosophical pronouncements of the last 25 years, and they do not seem to have had much relevance to linguistics…”

    seems a little strong of a statement to provide without citations or even examples of what you mean. It also just sounds like an insult – why are they “pronouncements” rather than “arguments”? (… he has entire books dedicated to philosophy topics and publishes in philosophy journals and ha engaged in debates with philosophers over the past 25 years as well).

    Not that I’m defending Chomsky’s specific views (I might defend some of them), but I honestly found this comment to be brazenly anti-intellectual, even for a blog post. Minimally some citations could be provided to illustrate that at least some portion of the field finds his comments confusing (e.g. Postal has an article about the “virtually conceptually necessary” turn of phrase; Goldsmith discusses Chomsky’s comments about Galileo)


    • Thanks, Adam – yes, maybe that was a bit quick… But I do point to several earlier blogposts where I discuss some of Chomsky’s views. To be honest, though, I didn’t understand his arguments of the last 25 years. And my feeling is that others were confused, too. Look at David Pesetsky’s comment on my recent blogpost (on the occasion of a 2023-09 NYT article where Fitch is interviewed on the Hauser et al. (2001) paper): Pesetsky says that despite appearances, Chomsky actually did assume “a lot of innate grammar besides Merge”: And here’s the crucial quotation: “How do I know? I asked Chomsky myself, because I too was massively confused.” So even Pesetsky didn’t understand Chomsky’s argument.

Leave a 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.