An interview with Scott DeLancey about language description and cross-linguistic considerations

Martin Haspelmath: Scott, you have been spending a lot of time in Northeastern India over the last years, and you are currently working on a description of Bodo, a Trans-Himalayan language of Northeast India. In a few weeks’ time, you will attend the NEILS conference in Kokrajhar (Bodoland Territorial Region, Assam, India). As an experienced fieldworker, how do you see the relationship between language-particular description and typological comparison? Should each language be described in its own terms?

Scott DeLancey: Yes, I’m very much in the every-language-in-its-own-terms camp. Of course it’s also the case that often looking at a similar phenomenon in other languages will help you understand what you’re looking at, but often the enlightenment comes from noticing how your problem is different from all the others.

Martin: Interesting – differences are of course the flip side of similarities between languages.

Scott: In fact I’m very critical of the practice (sometimes found in grammars of Northeastern Indian languages) of including sections in their grammars for categories imported from English or Hindi that aren’t really there in the languages they are describing. Particularly “relative clauses”, a function which in almost all Trans-Himalayan languages  (and a number of others, most notoriously Japanese, see Shibatani 2018) is simply a matter of using a nominalization as a nominal modifier.

Martin: Yes, there is a constant danger of (perhaps unwittingly) carrying over the categories of familiar languages to less familiar languages, which is something that Franz Boas warned against. But there are also advantages to comparisons with other languages, don’t you think? Isn’t a grammar more transparent that has a section on “relative clauses”?

Scott: That’s just what most modern grammar-writers do. And it has the advantage that it makes the facts easily accessible to readers whose primary interest is in comparative concepts. My problem with that is that it doesn’t follow the structure of the language. Older Tibetanists were often quite explicit that Tibetan, at least, just doesn’t have relative clauses, at least “in our (European) sense”, as Michael Hahn (1985) put it. The modern practice treats some of the uses of nominalization in a chapter on derivation, and the rest in the section on multi-clausal constructions. What I see is a single, unified construction – a nominalized verb or clause is simply a nominal, which like any other can be an argument of a verb, or a modifier of another nominal. In my mind separating these functions out into different chapters is misrepresenting the language, trying to make it look more Indo-European-like than it is. One might object to that characterization, and say that it’s just trying to fit the language into a universal scheme, but in this case, that “universal” scheme is primarily the Indo-European pattern, as there aren’t many other languages with classic “relative clauses” with a relative pronoun and movement and all.

Martin: Doesn’t it depend on how you use the term “relative clause” – either as a term for a common function that all languages need a strategy for, or as a term for a particular strategy type?

Scott: Yes, one way to describe these different approaches is in terms of organization by structure or by function. The Boasian approach is strictly by structure – you inventory the forms and constructions in the language you are describing, and that gives you the organization of the grammar. The modern typological approach, as advocated especially by T. Givón, is that you organize the grammar in terms of (putatively) universal functions. So, all languages seem to have some grammaticalized strategy for restricting the reference of a nominal by invoking another clause one of whose arguments has the same reference. That then is a universal category, or comparative concept. I’ve advocated and used that approach for a long time, but I’m now starting to think differently.

Martin: That’s interesting – my own intellectual trajectory has been similar. In the 1980s and 1990s, I was primarily interested in how similar the languages around the world are despite their superficial differences, but over the last two decades I’ve come to recognize that there are also important language-internal generalizations (of the “emic” type, as emphasized by the structuralists) that one cannot simply ignore.

Scott: Of course, as you have said more than once, it is not a matter of one of these approaches being correct and the other wrong – they have different benefits, for different audiences. I’m thinking now that our general practice has swung to far toward the universalist-functional side, and it’s time once again to pay more attention to structural differences.

Martin: Yes, we should somehow manage to keep in mind both the particular-structural aspects and the universal-functional aspects – I would hope that this can be achieved by recognizing that they both exist but are not identical to each other.

Scott: I’ve always understood, and described, the functionalist program as trying to explain structure as much as possible in terms of function, rather than taking structural categories as ordained by God as the generativists do. (And of course one often can’t explain synchronic structures that way, which is why grammaticalization is an essential part of functionalist explanation). But for some – usually not very explicitly, except for Givón – it means replacing the generativist idea of a universal inventory of structures with a universal inventory of functions, which then motivates the idea of describing any particular language in terms of that universal inventory. Taken (too) seriously this leads to things like the idea of treating object fronting as a “word order passive”, because like a passive it is used to indicate that the object of the clause is more topical than the subject.

Martin: Yes, this is a weird use of the term “passive”, but it’s true that in Givón’s thinking, this is quite consistent. It seems that he is simply not very interested in thinking about language-particular description (though his 2011 Ute grammar is excellent, so he does not always practice what he preaches).

Scott: There’s no real contradiction there. No one is going to try to write a descriptive grammar hewing rigidly to one or the other party line – even if they preach a party line in other contexts. That would be not just silly, but impossible. (Which is why there aren’t many descriptive grammars written in a generative framework). And you always have to keep your intended audience in mind, and Givón’s intended audience explicitly includes the Ute community, who supported his work for years, and whose interest I can assure you is primarily in Ute-specific particulars, not in how their language will contribute to the field of linguistic typology. (This is a point, by the way, that all grammar-writers in the 21st century need to keep in mind). But I imagine that some of the resistance to your advocacy of a Boasian approach is rooted in the idea that if you start from any kind universalist-functional position, explicitly or otherwise, then to present a description which is not organized by function is to neglect one’s responsibility to contribute to general linguistics.

Martin: I recently reread Christian Lehmann’s 1989 paper on the relationship between general comparative grammar and language description, and he expressed it quite well back then: We need both perspectives, the (universalist) function-to-form perspective, and the (particularist) form-to-function perspective. They should somehow be present simultaneously in a descriptive grammar. – In any event, thank you very much for this conversation, Scott!


Givón, T. 2011. Ute reference grammar. Amsterdam: Benjamins.

Hahn, Michael. 1985. Lehrbuch der Klassischen Tibetischen Schriftsprache. Bonn: Indica und Tibetica Verlag.

Lehmann, Christian. 1989. Language description and general comparative grammar. In Graustein, Gottfried & Leitner, Gerhard (eds.), Reference grammars and modern linguistic theory (Linguistische Arbeiten 226), 133–162. Tübingen: Niemeyer. (

Shibatani, Masayoshi. 2018. Nominalization in crosslinguistic perspective. In Prashant Pardeshi & Taro Kageyama (eds.), Handbook of Japanese Contrastive Linguistics, 345-510. Berlin: de Gruyter.

OpenEdition suggests that you cite this post as follows:
Martin Haspelmath (January 29, 2020). An interview with Scott DeLancey about language description and cross-linguistic considerations. Diversity Linguistics Comment. Retrieved July 19, 2024 from

3 thoughts on “An interview with Scott DeLancey about language description and cross-linguistic considerations

  1. Very interesting interview, thank you. I do agree that a linguistic description needs to be “balanced” between a strictly structural and formal approach. In fact, I wrote a paper once with almost that exact title! But, lest we go too far in the Boasian direction, I do believe that there is value in identifying broad typological similarities that may transcend obvious structural differences. Such similarities may be understandable in terms of common “functional” motivations. For example, I did fieldwork in the Philippines before I started my graduate program at UCLA. When I eventually took linguistic field methods as part of my graduate education, we worked with Central Alaskan Yup’ik. Yup’ik and Philippine languages are very different on just about every structural measure (Yup’ik=Verb-final, post-positions, complex morphological case marking, person-number marking on verbs, etc., Tagalog, for ex., =Verb initial, prepositions, no morphological case marking, no person-number marking on verbs . . . ), yet it seemed to me there was an uncanny similarity in the way basic clauses were formed. My master’s thesis essentially made the claim that Yup’ik was a “Philippine type” language. I think there is truth to that, and that such observations are valuable insofar as they lead to a better understanding of how human cognition and communication work. I have always been grateful for the opportunity I’ve had to work in-depth in multiple genetically and typologically distinct languages. Every language has its own “personality,” — its own characteristic way of doing things. But the communicative tasks that languages need to perform are broadly similar. Insights from one language may be valuable to folks working in structurally and genetically very distinct languages. A grammatical description should be able to depict the personality (or “genius” of the described language, while also illuminating it for others who work in very different language communities.

    • Yes, Tom, I agree with all this, BUT: How far do you take it? Some people think that all languages must be described in terms of general concepts, which are presumably part of an innate grammar blueprint. In fact, there is a large industry that does this, and many fieldworkers are taught that you need a general theory before you can describe a language. Usually, this general theory is taken from English. And even if it is taken from some other language, it MAY lead you astray – there is also the danger of looking at all languages from a Bantu perspective, e.g. when it makes you posit an empty Appl node in your tree of English ditransitives. This is why I’v been emphasizing that other languages are of course often INSPIRING, but they never decide on the analysis of a language.

    • I think your example illustrates my point very well. The only reason anyone talks about a “Philippine type” language in the first place is because people (tried and) failed to fit the Philippine languages into any familiar framework of voice and/or alignment. Once we accept that there are languages which don’t fit those frameworks, it is no surprise to find others as well, and not that much of a surprise to find that some of those resemble each other more than they resemble the languages on which the received typology of alignment was based.

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.