Round & Corbett’s (2020) paper on “comparability and measurement in typological science” is an interesting contribution to the question of what kinds of concepts linguists (should) use to compare languages. While comparative linguistics may well have a bright future, we often talk past each other at present (and sometimes we gather in subcommunities that do not talk to each other at all), so it is interesting to ask what methodological choices there are. I found myself agreeing overall with the paper, even though I did not fully understand all the points, and even though in n. 25, Round & Corbett (R & C) seem to be saying that my views are incompatible with theirs (this is a misunderstanding; see below).
The most important new point here is that the notion of “measurement” is put at the centre of the discussion. Linguists have traditionally thought of what they are doing as setting up classes or categories and trying to map these across languages, but mapping of language-particular categories such as “subject” or “word” or “sonority” across languages is often questionable, so a fresh perspective with novel terminology is highly welcome. Some linguists remain optimistic that we will eventually find a smallish set of innate building blocks of a hypothesized genetically encoded grammar blueprint, and that comparison will eventually be possible on this basis. But most of those who have experience with large-scale comparison have come to the conclusion that we need “multivariate” approaches, with “fine-grained” variables, and we need to think of cross-linguistic research as “measuring similarities and differences” across languages. As far as I am aware, the “measuring” terminology was first used by Bickel (2007) (e.g. “In order to capture and test distributions, typologists develop variables that measure similarities and differences between languages”, p. 241), and while it is still often associated with numerical measures, the basic idea of an independent yardstick that is distinct from the entities we compare is very important and more broadly applicable.
R & C note that the sciences can be divided into those that can rely on well-established measurement yardsticks (like physics and chemistry, where notions like “carbon and oxygen, mass and length” can be presupposed to be sound in most circumstances, because they are natural kinds), and those where the concepts needed for comparison are “still under contention”, e.g. psychology, anthropology, economics and also linguistics. In these latter disciplines, we need to think carefully about our concepts, and conceptual-methodological work and empirical work has to go hand in hand. For comparison, one needs “points of reference”, and “canons” are “methodological devices”. In other words, they are independent yardsticks (or comparative concepts, as such methodological devices for comparison are also called in linguistics).
Some linguists have proposed that in the absence of well-established innate building blocks, we can work with prototype categories, but R & C are rightly critical of prototypes (§1.5, §2.3). They are not “stable”, and they may be similar to “stereotypes” in that they pick out an arbitrary phenomenon that just happens to be familiar from well-known languages, e.g. inflectional classes of the Russian type (see n. 16). Instead, we should start from “logical end points”, and use “idealized extremes”. So what would this mean for concepts like “subject” or “word”? It seems that a subject must be a verbal argument of some kind, and if we take the concepts of “verb” and “argument” as somehow given, we can ask what would be a “logical extreme”. Perhaps we can say that an argument that has zero flagging and maximal person indexing is an example of a logical extreme, and in most languages (with accusative alignment), this comes pretty close to what one might call the “subject” (though there are some languages with no flagging or indexing of core arguments, so these would have at least two “extreme” arguments). It seems to me that this approach could indeed serve as a basis of comparison, and in its idealization, it bears clear resemblances to my (2011) proposals (based on Comrie’s and Lazard’s work). As for the “word” concept, it seems that the logical extreme is the “morph” (mentioned briefly by R &C), which is a minimal form (i.e. again an extreme case, see Haspelmath 2020). Traditional (stereotypical) “words” often include multiple morphs, but in the clearest cases, these consist of a single root plus bound forms which are extreme in that they occur only with roots of one root class (i.e. bound forms which are not promiscuous clitics).
R & C also note that “language-particular categories can be highly language-specific” (§1.4), and while the attributes that underlie the space of variation are “often shared across languages”, this is not necessary. Rather than having clear cases of “genitives” (as described for Russian and Archi by R &C), languages may have no flagging of a possessor at all, or they may have broader “oblique” cases which do not map clearly onto the dimension discussed in the paper. As we learn more and more about small languages from around the world (cf. Seifart et al. 2018), we become aware that even though we often see more of the same, we also often see quite unique phenomena. More generally, language systems are structurally unique, as Nick Evans has often emphasized.
Thus, even though language-particular description must make use of “comparison” (and can perhaps be called “typological”, so that all of linguistics would be a “typological science”), the language-particular categories themselves cannot serve as a basis of comparison, but we need yardsticks that work for all languages. This was the point of my earlier papers (Haspelmath 2010; 2016; 2018), and R & G’s paper supports this overall approach.
So it comes as a surprise that in their final footnote (n. 25), R & G attribute to me “a rather singular philosophical understanding of science”, but this is a misunderstanding (perhaps based on Matthew Spike’s misunderstandings; see also my replies to Spike’s paper here and here). While I may have emphasized the uniqueness of language-particular categories more than R & C do (e.g. in my work on semantic maps, Haspelmath 2003, where I noted that language-particular categories need not be assumed to be polysemous, like Evans 2011 and much other work in this vein), I see the challenges of the science of language in exactly the same way as they do. Even though it is not inconceivable that we will eventually find elementary innate building blocks (or natural kinds), I regard this as quite unlikely (as also discussed recently by Aronoff 2016). So I agree with R & C that the field of linguistics must invest in thinking about the methodology of comparison, but I think that by separating description from comparison, typologists have found the solution (see also Lazard 2005; Croft 2016). The observation that “language-internal and cross-linguistic research can be mutually informative” has guided my work over three decades, and I reemphasized it in my forthcoming paper on the structural uniqueness and the value of comparison for description (see also this earlier blogpost). Clearly, it is the NON-separation of typology from description that has improved grammatical descriptions enormously since the 1980s (see Dryer 2006). I have never said that “the external real world” is “not accessible to science”, nor that linguistics needs to be “completely re-built” (in my 2010 paper, I make very clear that what I am describing there is the mainstream Greenbergian approach, contrasting with the Chomskyan approach that tried “re-build” linguistics, and to base comparison on universally available natural kind categories). I have never said that linguistics should be “isolated” or “unique” – just like R & C, I have placed our field in the context of other disciplines such as economics and anthropology (Haspelmath 2018), with very similar conclusions.
As Bickel (2007: 248) notes: The goals of comparative linguistics “are embedded in a much broader anthropological perspective: to help understand how the variants of one key social institution are distributed in the world, and what general principles and what incidental events are the historical causes for these distributions.” Anthropology does have a tradition of speculating about cognitive universals underlying cross-cultural generalizations (from Adolf Bastian’s “elementary ideas” through Berlin & Kay’s colour term research), but it never occurred to anthropologists that cross-cultural comparison should use such hypothesized basic cognitive elements in their anthropological databases (e.g. Ember & Ember 1998). Thus, what Balthasar Bickel and I have been advocating in our methodological papers is not in any way different from what R & C propose here. The true contrast is between an approach to comparison that is built on uniformity of measurement (like the World atlas of language structures, to which both Bickel and Corbett contributed) and an approach which alternatively hopes to build comparison on uniform innate building blocks (like Baker 2001; 2010).
Aronoff, Mark. 2016. Unnatural kinds. In Ana R. Luís & Ricardo Bermúdez-Otero (eds.), The morphome debate, 11–32. Oxford: Oxford University Press.
Baker, Mark C. 2001. The atoms of language. New York: Basic Books.
Baker, Mark. 2010. Formal generative typology. In Bernd Heine & Heiko Narrog (eds.), The Oxford handbook of linguistic analysis, 285–312. Oxford: Oxford University Press. (6 July, 2015).
Bickel, Balthasar. 2007. Typology in the 21st century: Major current developments. Linguistic Typology 11(1). 239–251.
Croft, William. 2016. Comparative concepts and language-specific categories: Theory and practice. Linguistic Typology 20(2). 377–393. (doi:10.1515/lingty-2016-0012)
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.
Ember, Carol R. & Melvin Ember. 1998. Cross-cultural research. In H. Russell Bernard (ed.), Handbook of methods in cultural anthropology, 647–687. Walnut Creek, CA: AltaMira Press.
Evans, Nicholas. 2011. Semantic typology. In Jae Jung Song (ed.), The Oxford handbook of linguistic typology, 504–533. Oxford: Oxford University Press.
Haspelmath, Martin. 2003. The geometry of grammatical meaning: Semantic maps and crosslinguistic comparison. In Michael Tomasello (ed.), The New Psychology of Language, vol. 2, 211–243. New York: Lawrence Erlbaum. 10.5281/zenodo.831410.
Haspelmath, Martin. 2010. Comparative concepts and descriptive categories in crosslinguistic studies. Language 86(3). 663–687. doi:10.1353/lan.2010.0021.
Haspelmath, Martin. 2011. On S, A, P, T, and R as comparative concepts for alignment typology. Linguistic Typology 15(3). 535–567.
Haspelmath, Martin. 2016. The challenge of making language description and comparison mutually beneficial. Linguistic Typology 20(2). 299–303. doi:https://doi.org/10.1515/lingty-2016-0008.
Haspelmath, Martin. 2018. How comparative concepts and descriptive linguistic categories are different. In Daniël Van Olmen, Tanja Mortelmans & Frank Brisard (eds.), Aspects of linguistic variation: Studies in honor of Johan van der Auwera, 83–113. Berlin: De Gruyter Mouton. https://zenodo.org/record/3519206.
Lazard, Gilbert. 2005. What are we typologists doing? In Frajzyngier, Zygmunt & Hodges, Adam & Rood, David S. (eds.), Linguistic diversity and language theories, 1–23. Amsterdam: Benjamins.
Round, Erich & Greville G. Corbett. 2020. Comparability and measurement in typological science: The bright future for linguistics. Linguistic Typology 2020 (https://doi.org/10.1515/lingty-2020-2060). (Preprint on Academia.edu)
Seifart, Frank, Nicholas Evans, Harald Hammarström & Stephen C. Levinson. 2018. Language documentation twenty-five years on. Language 94(4). e324–e345. doi:10.1353/lan.2018.0070.