In a new position paper, Željko Bošković (2022) compares Greenberg-style typology with Baker-style typology and claims that the two approaches are not as different as it might appear. He suggests that a rapprochement is possible, and even that “typology is setting grounds for a potential rapprochement of the functional and the formal approach to language more generally” (p. 1). But are there really reasons to be optimistic? Here I offer a more cautious assessment, though I am of course happy whenever someone writes a paper comparing approaches and enters into some kind of dialogue.
Bošković’s paper has the title “Formalism and, not vs, Functionalism”, and as it is very long and detailed, I will focus on the following eight claims made by Bošković. I will refer to the two approaches as Greenbergian, after Greenberg (1963), and Bakerian, after Baker (2001); Mark Baker is of course taking a generative/Chomskyan perspective, but I take his work as exemplary of generative typology. (Oddly, Bošković calls the two approaches “traditional typology” (TT) and “generative typology” (GT), as if generative grammar were itself not a decades-long tradition; and in the paper’s title, he seems to equate them with “functionalism” and “formalism”, without further explanation.)
Eight selected points of the paper:
(1) The differences between the two approaches (or “camps”, as he says) are largely due to wrong perceptions of radical slogans, and to misunderstandings, not so much to actual incompatibilities.
(2) The idea that each language should be described in its own terms (categorial particularism, emphasized by Haspelmath 2010) is not taken too literally by most of the Greenbergians, and the Bakerians need to take language-particular differences very seriously, too.
(3) Greenbergians have reacted negatively to suggestions (as in Cinque 1999) that clause structure and syntactic structure more generally is universal across languages, but these claims are untenable anyway – in particular, the NP vs. DP parameter (e.g. Bošković 2008) cannot be stated in a different way.
(4) Functionalists emphasize semantics and pragmatics, and formalists’ focus on syntax, does not mean that there is any incompatibility – the two types of explanations should be combined.
(5) Greenbergians focus on what is possible, while Bakerians are interested in where the gaps are – what kinds of language types are unattested and impossible.
(6) Haspelmath’s (2004) and Stegovec’s (2020) proposals for ditransitive person restrictions (“PCC”) converge to some extent, and one needs them both – in fact, Haspelmath talks about “grammaticalization” but does not say what it amounts to, while Stegovec is explicit about this.
(7) Occam’s razor favours functional explanations. More generally, “Occam’s razor disfavours, and requires re-evaluation of, strictly formal mechanisms” (p. 21).
(8) Both Greenbergians and Bakerians believe that there are general properties of language, and it does not matter much whether one calls them “UG” or not (p. 30).
The hope that we may eventually see a rapprochement is old – I remember hearing a talk by Bernard Comrie in 1987 (at the International Congress of Linguists in Berlin) where he noted an increasing interest in typology among generative grammarians (after his pessimistic assessment in his 1981 book). When Hawkins (1994) was published in Linguistic Inquiry, I hoped that generative linguistics would become more open, and again when Optimality Theory (with its clearly functionalist bent) was becoming popular (cf. my 1999 paper). So is typology “rapidly gaining traction within the generative camp”, as Bošković says (p. 34)? I am not sure – the examples he gives are mostly from two decades ago, or from work by his colleagues and students at the University of Connecticut (such as Roy Messick, Jungmin Kang, Miloje Despić, Adrian Stegovec, Hsu-Te Cheng, Marcin Dadan, Aida Talić, Neda Todorović).
It seems to me that it is far more typical for generative syntacticians to study a particular phenomenon in a particular language, and to draw far-reaching conclusions from this – which of course makes sense if one assumes a rich innate blueprint for grammar (but not otherwise, cf. Haspelmath 2021b). In fact, Bošković himself notes that “many practitioners of [generative grammar] use [detailed investigations of individual languages] (generally … of an understudied language in this case) to argue against proposed crosslinguistic generalizations”. Of course, it is easier to study one language and find a weird phenomenon that does not fit into the received wisdom (and then develop some ideas about what to say about it) than to carry out a large-scale cross-linguistic investigation. So this actual resesarch practice is not surprising, but does it amount to “typology”?
Now clearly, Greenbergian typology is not wildly popular either, so maybe typological work of both styles will continue to exist on the margins of the field. But are the differences primarily due to misunderstandings (B’s point (1))? I do not think so, and I fear that Bošković has overlooked the crucial stumbling block: Bakerians simply assume that all languages are made up of the same categories, and they find it normal to work with entities like “[ForceP [TopP [FocP [IP ]]]]”, while Greenbergians do not see any value in them because different languages can have different categories (including different categories of information structure, cf. Matić & Wedgwood 2013).
Bošković treats the need for each language to be described in its own terms as a “mantra” (his point (2)), and he says that it is too radical, but it simply follows from the fact that we do not know what the innate categories of the grammar blueprint are (one need not reject this idea, but it remains speculative at the moment). His text betrays a crucial misunderstanding:
“[Haspelmath] treats them [i.e. the comparative concepts] as artificial linguistic constructs and not real (in fact not part of the grammar of individual languages), but there is really no deep reason (and, more importantly, nothing in the general TT worldview) why they should not be considered real.” (p. 3)
The deep reason is that language-particular categories are identified by language-particular criteria, while comparative concepts are defined in the same way for all languages (Haspelmath 2018). This means that it is logically impossible to identify and define them in the same way. There are of course many similarities, and language-particular categories in different languages often have the same name, for good mnemonic reasons (Haspelmath 2020a). For example, English the is called Definite Article (though it is defined as the segment sequence [ðə/ði]), and French le/la/les (defined as three related segment sequences) is also called Definite Article – they exhibit striking similarities in their use, but also clear differences. They could not be the same category (because they belong to different languages), but they can be compared using a comparative concept “definite article” (e.g. Becker 2021). Bošković suggests that all language-particular categories can be composed of a set of universal categories, e.g. category X in French could be composed of Y+Z, while category U in English could be composed of Y+W – they would still both represent Y, which would be simultaneously a comparative concept and a universal category. While this is possible in principle (and David Adger has made similar suggestions, cf. this blogpost), it does not (yet?) work in practice. We simply do not know (yet?) what these atomic features (Y, Z, W) might be. For phonology, there are certain concrete proposals for the “natural kinds” of phoneme inventories, but for morphosyntax, we are groping in the dark.
For this reason, we use all kinds of different comparative concepts (including token-based comparative concepts such as parallel-text passages). Bošković seems to think that each category has one comparative concept corresponding to it (“a distinction between a language particular descriptive category (X) and its crosslinguistically applicable comparative concept (Y)”, p. 3), but this is wrong – the set of possible comparative concepts is completely open-ended (the examples in my 2010 paper may have given a wrong impression; I focused on category-like comparative concepts in that paper because it is these comparative concepts that are most often confused with language-particular categories).
Bošković is right that not all linguists working in the Greenbergian tradition have officially embraced the comparative-concept framework (initiated by Dryer 1997; Croft 2001; Haspelmath 2010), and some have been quite critical (e.g. LaPolla 2017, see this blogpost; and Himmelmann 2021, see this blogpost). However, it is easy to see that descriptions by authors such as Dryer (e.g. his grammar of Kara-Lemakot, 2013), LaPolla (e.g. his grammar of Qiang, LaPolla & Huang 2003), and Himmelmann (e.g. his grammar of Toratán, Himmelmann & Wolff 1998) are very similar and do not make any reference to generative-style notation of the type seen above (“[ForceP [TopP [FocP [IP ]]]]”). So we may not all have the same philosophical understanding of what we are doing (see also my discussion of Nick Evans’s puzzling 2020 comments concerning language comparison), but what we are doing is very similar, and rather different from what generative linguists ar doing.
The idea that each language has its own categories is of course very old, and not merely an overreaction to some extreme claims by some generativists (as Bošković suspects). For example, Whitney (1875) said the following (cited from Joseph 2022: 24):
“Every single language has thus its own peculiar framework of established distinctions, its shapes and forms of thought, into which, for the human being who learns that language as his ‘mother-tongue’, is cast the content and product of the mind, his store of impressions, however acquired, his experience and knowledge of the world. This is what is sometimes called the ‘inner form’ of language – the shape and cast of thought, as fitted to a certain body of expressions.” (Whitney 1875: 21-22)
In the 19th century, linguists influenced by the Humboldian tradition were perhaps particularly intrigued by the different semantic categories of languages, but it has long been clear that morphosyntactic categories are cross-linguistically variable, too. In contrast to what Bošković implies (cf. his point (3)), Dryer’s (1997) paper (which influenced both Croft and Haspelmath) did not arise in reaction to “universal sentence structure” proposals (à la Cinque), but rather in reaction to the “opportunistic” (Croft 2009) choice of diagnostics for syntactic functions such as “subject” and “object” (Dryer’s paper is foreshadowed in Dryer 1986, which discusses types of object relations, and criteria for identifying them). And this is the key issue: Can one use different criteria in different languages to identify the same phenomenon? In generative grammar, this is generally accepted (leading to a lot of “moving pieces”), but Greenbergian typologists are fully aware of the need to identify comparable entities in uniform ways. Bošković does not mention the problem of opportunistic choice of criteria (called “diagnostic fishing” in Haspelmath 2018; see also Tallman 2021) at all, so maybe he is simply not aware that it is a deep problem.
Regarding his point (4), there is really no disagreement that syntax, semantics and pragmatics are all important, and that none can be reduced to the other. Maybe some authors who work only on one language still hope that some of the core syntactic phenomena really have a semantic-pragmatic explanation, but every typologist knows that nonsemantic regularities are found all over the place in language systems – languages wouldn’t be so different if semantics-pragmatics ruled supreme. But the question is which aspects of the syntax of a language are historically accidental, and which accents are amenable to an explanation – and if so, whether the explanation is of the functional-adaptive kind or of a biocognitive-representational kind (making reference to innate building blocks which can represent some kinds of generalizations but not others; Haspelmath 2019). Assuming a rich set of innate biocognitive tools (commonly called “UG”) makes a big difference here, and it is odd that Bošković should downplay this aspect by almost equating “UG” with “general properties of language” (his point (8); p. 30), because two pages later he treats UG “as a property of FL (= the domain-specific faculty of language)”. (Unfortunately, the terms “language faculty” and “universal grammar” have been used in confusingly diverse ways; see Haspelmath 2020b for some discussion of these terms.)
Bošković’s point (5) about typological gaps is odd, because the notion of an implicational universal and a corresponding tetrachoric table (with one gap in a 2-by-2 table) is a central part of typological theory (Croft 2003: Chapter 3). Bošković is right that when it comes to the description of particular languages, the systematic identification of unacceptable patterns has been an important contribution of generative grammarians (probably the most important contribution of Chomsky to linguistics; see my recent Inference text about this). But in typology, finding gaps is equally characteristic of both approaches.
Finally, what is the relationship between functional explanation, “grammaticalization” and syntactic description (B’s points (6) and (7))? Bošković is right that when I attributed the ditransitive person-role restrictions to “grammaticalization” (Haspelmath 2004), this was a bit vague. He makes the following interesting comment:
“Functionalists often question whether a typological generalization requires a formal analysis of the data, but they also often use the term ‘grammaticalization’ in a way that in fact indicates such a need. More generally, what the functionalists often mean by grammaticalization (and what Haspelmath also means by it in his approach to person restrictions), is that X started out as a tendency due to functional/pragmatic reasons, but has “rigidified” into a hard rule, without really explaining the rigidification. This is a place where a formalist can step in to provide an actual account of the rigidification, including the details of exactly what X has hardened into, and why it has hardened into a particular format rather than a different logically possible format.” (Bošković 2022: 20)
It is of course true that the very claim of grammaticalization (of a discourse tendency) entails the need for grammatical description – but what kind of grammatical description? Many would say that it does not really matter, as long as the description is explicit and clear (Dryer 2006). It is true that we do not really know why this “rigidification” happens, but how would generative grammar contribute to understanding it? Stegovec (2020) provides a generative account, but it is highly technical, involving derivations such as the following:
Stegovec does not take the discourse tendencies into account at all and does not even mention them (unlike Anagnostopoulou 2017; see this conversation). Stegovec’s paper is quite typical of work in generative grammar, and it is not really compatible with Bošković’s point (7) about favouring functional explanations. If there is a functional explanation for the tendencies of person-role restrictions (see also Haspelmath 2021a for a much bigger picture), then we would not need the kind of complex machinery that Stegovec employs. It seems, however, that Stegovec’s paper completely ignores the possibility of a functional explanation for the general tendencies, because it is specifically interested in developing this formal machinery. Thus, it is not a good example of rapprochement – or at least it is very difficult to see because the paper is so hard to understand.
I agree with Bošković that there are too many ideological statements (“slogans”), both negative and positive, and it is certainly not healthy that linguists are working in different “camps” or “traditions” – ideally, we would develop theories and descriptions that are broadly accessible and that can be appreciated by a wide range of colleagues. But this is simply not happening at the moment, and many linguists are comfortable with finding a place in a corner of the discipline, without worrying much how it all fits together. Maybe we shouldn’t be expecting too much – understanding language structures is simply very hard.
Baker, Mark C. 2001. The atoms of language. New York: Basic Books.
Becker, Laura. 2021. Articles in the world’s languages. Berlin: De Gruyter Mouton.
Bošković, Željko. 2008. What will you have, DP or NP? North Eastern Linguistic Society 37(1). 101–114.
Bošković, Željko. 2022. Formalism and, not vs, functionalism. Lingbuzz. (https://ling.auf.net/lingbuzz/006337)
Cinque, Guglielmo. 1999. Adverbs and functional heads: A cross-linguistic approach. New York: Oxford University Press.
Comrie, Bernard. 1981. Language universals and linguistic typology: Syntax and morphology. Oxford: Blackwell.
Croft, William. 2001. Radical Construction Grammar: Syntactic theory in typological perspective. Oxford: Oxford University Press.
Croft, William. 2003. Typology and universals. 2nd edition. Cambridge: Cambridge University Press.
Croft, William. 2009. Methods for finding universals in syntax. In Scalise, Sergio & Magni, Elisabetta & Bisetto, Antonietta (eds.), Universals of language today, 145–164. Dordrecht: Springer.
Dryer, Matthew S. 1997. Are grammatical relations universal? In Bybee, Joan L. & Haiman, John & Thompson, Sandra A. (eds.), Essays on language function and language type: Dedicated to T. Givón, 115–143. Amsterdam: Benjamins.
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. (http://www.acsu.buffalo.edu/~dryer/desc.expl.theories.pdf)
Dryer, Matthew S. 2013. A grammatical description of Kara-Lemakot (Asia-Padific Linguistics Open Access). Canberra: Australian National University.
Evans, Nicholas. 2020. Introduction: Why the comparability problem is central in typology. Linguistic Typology (ahead of print). (doi:10.1515/lingty-2020-2055)
Greenberg, Joseph H. 1963. Some universals of grammar with particular reference to the order of meaningful elements. In Greenberg, Joseph H. (ed.), Universals of language, 73–113. Cambridge, MA: MIT Press.
Haspelmath, Martin. 2004. Explaining the Ditransitive Person-Role Constraint: A usage-based approach. Constructions 2. (doi:10.5281/zenodo.831408) (https://journals.linguisticsociety.org/elanguage/constructions/article/view/3073.html)
Haspelmath, Martin. 2010. Comparative concepts and descriptive categories in crosslinguistic studies. Language 86(3). 663–687. (doi:10.1353/lan.2010.0021)
Haspelmath, Martin. 2018. How comparative concepts and descriptive linguistic categories are different. In Van Olmen, Daniël & Mortelmans, Tanja & Brisard, Frank (eds.), Aspects of linguistic variation: Studies in honor of Johan van der Auwera, 83–113. Berlin: De Gruyter Mouton. (https://zenodo.org/record/3519206)
Haspelmath, Martin. 2019. Can cross-linguistic regularities be explained by constraints on change? In Schmidtke-Bode, Karsten & Levshina, Natalia & Michaelis, Susanne Maria & Seržant, Ilja A. (eds.), Competing explanations in linguistic typology, 1–23. Berlin: Language Science Press. (http://langsci-press.org/catalog/book/220)
Haspelmath, Martin. 2020a. The structural uniqueness of languages and the value of comparison for description. Asian Languages and Linguistics 1(2). 346–366. (doi:10.1075/alal.20032.has)
Haspelmath, Martin. 2020b. Human linguisticality and the building blocks of languages. Frontiers in Psychology 10(3056). 1–10. (doi:10.3389/fpsyg.2019.03056)
Haspelmath, Martin. 2021a. General linguistics must be based on universals (or nonconventional aspects of language). Theoretical Linguistics 47(1–2). 1–31. (doi:10.1515/tl-2021-2002)
Haspelmath, Martin. 2021b. Role-reference associations and the explanation of argument coding splits. Linguistics 59(1). 123–174. (doi:10.1515/ling-2020-0252)
Hawkins, John A. 1994. A performance theory of order and constituency. Cambridge: Cambridge University Press.
Himmelmann, Nikolaus P. 2021. Against trivializing language description (and comparison). Studies in Language. (doi:10.1075/sl.19090.him) (https://ling.auf.net/lingbuzz/005705) (Accessed January 18, 2022.)
Himmelmann, Nikolaus P. & Wolf, John U. 1999. Toratán (Ratahan) (Languages of the World/Materials 130). Munich: Lincom Europa.
Joseph, John E. 2022. Saussure’s universal grammar, Chomsky’s structuralism. In Cosenza, Giuseppe & Forel, Claire A. & Puskas, Genoveva (eds.), Saussure and Chomsky: Converging and diverging, 19–36. Berne: Peter Lang.
Kiparsky, Paul. 1982. From cyclic phonology to lexical phonology. In van der Hulst, Harry & Smith, Norval (eds.), The structure of phonological representations (Part I). Dordrecht: Foris.
LaPolla, Randy & Huang, Chenglong. 2003. A grammar of Qiang: With annotated texts and glossary. Berlin: Mouton de Gruyter.
LaPolla, Randy J. 2017. Causation as a factor and goal in typological comparisons. Linguistic Typology 21(3). 547–554. (doi:10.1515/lingty-2017-0013)
Matić, Dejan & Wedgwood, Daniel. 2013. The meanings of focus: The significance of an interpretation-based category in cross-linguistic analysis1. Journal of Linguistics 49(1). 127–163. (doi:10.1017/S0022226712000345)
Schiering, René & Bickel, Balthasar & Hildebrandt, Kristine A. 2010. The prosodic word is not universal, but emergent. Journal of Linguistics 46(3). 657–709. (doi:10.1017/S0022226710000216)
Stegovec, Adrian. 2020. Taking case out of the Person-Case Constraint. Natural Language & Linguistic Theory 38(1). 261–311. (doi:10.1007/s11049-019-09443-0)
Tallman, Adam J. R. 2021. Analysis and falsifiability in practice. Theoretical Linguistics 47(1–2). 95–112. (doi:10.1515/tl-2021-2009)
Whitney, William D. 1875. The life and growth of language. London: Henry S. King & Co.