A recent issue of the journal “Linguistic Typology” contains a number of articles on the usefulness of typology, among others one by Nikolaus Himmelmann on the usefulness of typology for language documentation (2016). Himmelmann bluntly criticizes the theoretical stance of separating description from comparison:
“Recently, it has become fashionable to emphasize the difference between language description and typological comparison, with authors such as … Haspelmath (2010) actually conceiving of language description as an enterprise completely separate from typological … inquiry. For practitioners of language description, I would think, this view is ill-conceived and does not reflect current practice, which values the typologically informed model of a descriptive grammar most highly.” (p. 475)
But what exactly does it mean for a description to be “typologically informed”? Surely everyone agrees that a newly discovered language which has a special case form for the transitive agent should be described as “ergative” rather than, say, as exhibiting a special passive-like behaviour of its transitive clauses. Thus, when a language has a pattern that is very similar to well-known patterns from other languages, it would not be helpful to introduce completely new terminology. As I noted in Haspelmath (2010: 674), “language-particular labels that are very similar to the corresponding comparative concepts clearly facilitate communication among specialists of different languages”.
So where is the disagreement? Himmelmann notes that “all elicitation, as well as the annotation of textual data, is greatly enhanced by being familiar with grammatical structures and discursive strategies found in other languages of the world” – and he is obviously right in this.
But while the language documenter’s and describer’s work is no doubt “greatly enhanced” by knowing about typology, are description and comparison also part of the same enterprise? I have argued that they are not, even though they are of course mutually beneficial (Haspelmath 2016). The difference is that description relies exclusively on language-internal distribution (Croft 2001), while comparison relies on substantively defined semantic and/or formal concepts. The Georgian Ergative case can be defined as the case that ends in -ma (with no need to refer to agentivity or transitivity), while the “ergative case” as a comparative concept must be defined with reference to the agent of a typical physical-effect verb (Haspelmath 2011).
Himmelmann seems to see things differently, apparently assuming that description and comparison can make use of the same concepts:
“descriptivists and typologists work with the same set of analytical techniques and concepts, including complementary distribution, articulation place features, semantic scope effects, omissibility in control and coordination constructions, constraints on linear ordering, etc., etc. (compare Bickel 2007: 242). The better, i.e., the more fine-grained and better operationalized these techniques and concepts, the better both description and typology.” (p. 475)
But when one goes into the details of all these concepts, one realizes that comparative concepts are different from descriptive categories. For descriptive purposes, phonologists have made use of abstract place features such as [anterior], [coronal] or [back], while typologists would use more fine-grained place descriptions such as “dental”, “alveolar”, “velar” and “uvular” (e.g. Maddieson 2005). For linear order, descriptivists would distinguish strictly between rigid order and preferred order, while typologists use broader notions such as “dominant order” (Dryer 2005). It is simply not true that “the more fine-grained, the better”, because description often needs to work with “big” categories like “verb” (rather than fine-grained semantic categories such as temporary state, process, activity, telic action), and likewise, comparison often needs to work with “coarse-grained” categories like “ergative/accusative”, because otherwise no clear pattern is discernible. Of course, all our concepts need to be clear (and “well operationalized”), but they can be at various levels of generality.
“category controversies such as the continued debate regarding the applicability and usefulness of the concept “subject” in crosslinguistic comparison have been enormously fruitful and productive in bringing to the fore these more fine-grained and crosslinguistically applicable concepts. Thus, it is probably more correct to conceive of this common ground between typology and language documentation and description as one of mutual concern rather than as a unidirectional provision of “services” from typology to the latter.” (p. 476)
Again, nobody argues that typology should unidirectionally provide services to language description (on the contrary – while one can in principle describe a language without knowing typology, it is plainly impossible to do typology without having any descriptions of languages). But the “subject” debate is another good example for the disparateness of typology and description: While typology has successfully dispensed with the notion of “subject” (reducing the attested diversity to different patterns of alignment of S, A and P), it is impossible to describe English without the descriptive notion of “(English) Subject”.
Maybe one of the reasons why the distinction between comparative concepts and descriptive categories is not obvious to all observers is that one can get pretty far in describing most languages by using concepts that are familiar from other languages. After all, languages exhibit great similarities. But we also know that most languages also exhibit some strange quirks, i.e. phenomena that hardly occur elsewhere (Emphatic Obstruents in Arabic, Obviation in Algonquian, Strong/Weak Adjectives in German, and so on). It is in such cases that the need for language-specific concepts becomes particularly clear.
Evans (2013: 254) notes that
“many linguistic phenomena have a “long tail distribution”: a small number of regular structural solutions account for most languages. However, to arrive at a full account, we need a much larger design space.”
In generative linguistics, the attitude seems to be to hope that this “long tail” will somehow go away, that it will turn out to be reducible to the core phenomena in one way or another. This possibility cannot of course be ruled out, but in the functional-typological community, we tend to be more open to the possibility of languages simply differing (and showing rara, discussed at conferences and displayed in a Raritätenkabinett). But if the “long tail” is real, then it means that “typologically informed” description cannot do justice to it. Instead, we need autonomous describers who do justice to the peculiarities of each individual language and go beyond the information provided by typology.
Bickel, Balthasar. 2007. Typology in the 21st century: Major current developments. Linguistic Typology 11(1). 239–251.
Croft, William. 2001. Radical construction grammar: Syntactic theory in typological perspective. Oxford: Oxford University Press.
Dryer, Matthew S. 2005. Determining dominant word order. In Martin Haspelmath, Matthew S Dryer, David Gil & Bernard Comrie (eds.), The world atlas of language structures, 371. Oxford University Press. (http://wals.info/chapter/s6).
Evans, Nicholas. 2013. Language diversity as a resource for understanding cultural evolution. In Peter J. Richerson & Morten H. Christiansen (eds.), Cultural evolution: Society, technology, language, and religion, 233–268. Cambridge, MA: MIT Press.
Haspelmath, Martin. 2010. Comparative concepts and descriptive categories in crosslinguistic studies. Language 86(3). 663–687.
Haspelmath, Martin. 2011. On S, A, P, T, and R as comparative concepts for alignment typology. Lingustic Typology 15(3). 535–567.
Haspelmath, Martin. 2016. The challenge of making language description and comparison mutually beneficial. Linguistic Typology 20(2). 299–303.
Himmelmann, Nikolaus P. 2016. What about typology is useful for language documentation? Linguistic Typology 20(3). 473–478.
Maddieson, Ian. 2005. Uvular consonants. In Martin Haspelmath, Matthew S Dryer, David Gil & Bernard Comrie (eds.), The world atlas of language structures. Oxford: Oxford University Press. (http://wals.info/chapter/6).
Payne & Payne’s recent (2013) grammar of Panare explicitly titles itself *A typological grammar of Panare*, raising the question of what exactly makes it “typological”. I attempt to answer this question in my review of the book in IJAL (Hieber 2016), and my overall impression is that the grammar is typologically-informed in exactly the way Martin suggests is useful here: the authors begin each section with a language-general definition of the relevant topic and brief typological survey, then examine that particular grammatical feature of Panare in light of the typological literature. Ultimately, however, their grammatical description is a language-specific one, with the useful result that it is immediately apparent how conformant or divergent each feature of the grammar of Panare is from known typological patterns. So the book is comparative in the sense that the authors consistently situate and compare the features of Panare with known typologies, but language-specific in its actual grammatical descriptions. As one interesting example, they state that Panare has “split-inverse marking”, a term which to my knowledge has not been used in grammatical descriptions elsewhere, and thus is specific to the grammar of Panare, but which nevertheless is accessible to other linguists because it is framed in terms of existing, well-known comparative concepts (“split alignment” and “inverse alignment”).
This is a very interesting and relevant discussion that more graduate students who work on language description and documentation should follow. I’ve read a few grammars that are labeled as “typologically informed” or “from a typological perspective”. What that often means is that the constructions in the language-specific case gets put into context of what neighbouring or related languages do, and sometimes also what other languages of the world does. It seems that this is often based on function rather than form, “language x does y for comparative constructions, languages of the area are predominately y found to do z for y”.
Most of the time, I have confidence that the language descriptivists are not unduly influenced, i.e. applying top-down categories in such a way that it distorts the description of the language. That’s the worry though, isn’t it? That we’ll get “typologese-translation-grammars”, just like we have “french-translation-grammars” where poor little europeans run around west-africa trying to figure out what Mande languages do for “Subjonctif”..
Btw, there’s a HWRG post on these matters too: http://humans-who-read-grammars.blogspot.com.au/p/help-linguistics-is-hard.html
It seems to me that the disagreement between Martin and Nikolaus is not about whether descriptive grammars should be typologically informed (both agree that they should), but rather with regard to the orthogonal issue of whether language description and cross-linguistic comparisons make use of the same categories and concepts (Nikolaus argues that they should, Martin that they shouldn’t, whereas in my own contribution to the recent Linguistic Typology volume, I attempt to stake out an intermediate position.)
However, I would like to take issue with respect to one passage in Martin’s blog above. Martin writes that …
“… one can get pretty far in describing most languages by using concepts that are familiar from other languages. After all, languages exhibit great similarities. But we also know that most languages also exhibit some strange quirks, i.e. phenomena that hardly occur elsewhere (Emphatic Obstruents in Arabic, Obviation in Algonquian, Strong/Weak Adjectives in German, and so on).”
Well let’s take my analysis of Riau Indonesian as lacking a grammaticalized distinction between predication and attribution as a case in point. (And since the issue here is methodological, you don’t have to buy into my analysis in order to appreciate the force of my comments below.) Since most languages don’t work like Riau Indonesian (though I’ll bet more do than is usually acknowledged), I suppose that Martin would add this property of Riau Indonesian to his list of “strange quirks”. The problem with this is, however, that the notion of “quirk”, as I understand it, bundles together two distinct properties: being rare, and being “small”, ie, something than can be shunted off into one side corner of one’s description. Which presumably is why Martin writes that “one can get pretty far in describing” languages without coming up against their quirky properties.” But can one go far in a description of Riau Indonesian without coming up against its absence of a grammaticalized predication/attribution distinction? Well yes and no. Yes in the sense that you can simply assume that the distinction exists and then, if you’re sufficiently conscientious, note somewhere that many or most actual utterances neutralize it. But no in the sense that by adopting this path, you will be doing violence to the way the language really is.
In other words, Nick Evans’ long tails are distributional tails, however, in terms of grammatical architecture, such tails can actually lie at the heart of the language (not at its periphery).
Thus, my taking issue with Martin’s passage about quirks actually brings me to an even more emphatic support of his final conclusion, to the effect that …
“… we need autonomous describers who do justice to the peculiarities of each individual language and go beyond the information provided by typology.”