[Lingtyp] comparative concepts

Martin Haspelmath haspelmath at shh.mpg.de
Mon Feb 1 09:31:06 UTC 2016

I think we really agree in all major respects here:

– yes, descriptive work has gotten better by integrating typological 
perspectives (so description and typology should continue to exist in 

– no, one shouldn't use ("structuralist") language-specific categories 
for typological comparison (as the generativists often try to do)

– yes, every typological variable "captures" something of interest in a 
language, even if it is only a tiny bit of some category

We probably also agree that language description *needs* to make use of 
language-specific categories, because without them, the description 
would be very unparsimonious in many places and thus impractical (for 
example, to describe German, we need categories such as "Neuter gender" 
and "Weak verb"; otherwise, the description would get hopelessly 

(It seems that this is was at the heart of Randy LaPolla's objection to 
using "subject" or "A" in classifying Chinese word order; he seemed to 
think that word order statements need to be based on the descriptively 
indispensable language-specific categories, even when they are meant 
only for typological purposes.)

But note that I use "comparative concept" in a broader sense than 
"typological variable". Every typological variable is a comparative 
concept (by definition), but some comparative concepts (e.g. "ergative 
case", or "clause", or "high vowel") are not variables by themselves – 
they are crucial ingredients of variables.


On 01.02.16 08:59, Balthasar Bickel wrote:
> I have just now had a chance to read up on this very interesting 
> discussion and noticed a certain trend towards keeping comparative 
> work outside the description of individual languages. I find this 
> problematic. Modern descriptive work has become better precisely by 
> integrating typological perspectives, and I have always found that I 
> have started to understand a phenomenon in a language only once I 
> could tell with some precision how it compares to similar phenomena in 
> other languages.
> It might help to realize that “comparative concepts” --- or 
> typological variables, as I prefer to call them --- contrast with 
> language-specific categories in two very different ways. Only one of 
> these contrasts is a contrast between research enterprises.
> (i) By definition, the categories of a typological variable can recur 
> across languages, language-specific categories can’t. For example, 
> ‘argument with most agent properties in Dowty’s definition’, ‘linearly 
> ordered before’, a Nijmegen-style exlicitation simulus, a Dahl TAM 
> questionnaire context, a translation context etc. can all by 
> definition recur across languages, the Saussurian sign -ed ‘PST’ can’t.
> (ii) Language-specific categories serve a purpose in a 
> Pāṇini-inspired, structuralist analysis where the goal is to re-use 
> categories in a maximally parsimonous way (so you can say, e.g., there 
> is a category of ‘subordinate clauses’ in language X, defined by a set 
> of properties in X, and then find that the same category /also/ 
> captures --- or even ‘causes’ --- the constraints on WH questions in 
> X). By contrast, typological variables don’t serve such a purpose. If 
> you want to explore patterns across variables or even capture the 
> system as a whole, you use stats (and so you might discover for 
> example that certain values on a variable that captures WH 
> possibilities correlate to some extent with certain values on a 
> variable that captures the scope behavior of illocutionary force 
> markers, a correlation caused by information structure principles; 
> Bickel 2010).
> The contrast in (i) merely limits the range of things you can compare 
> typologically (as opposed to reconstruct proto-individuals), but it 
> doesn’t imply anything about the usefulness of typological variables 
> for describing languages, let alone call for a terminological 
> distinction between ‘matching’ vs ‘instantiating’ a category or for 
> different research enterprises. Every well-defined typological 
> variable captures or measures something of interest in a language --- 
> but obviously it is only ‘something’, a very tiny aspect of a very 
> complex phenomenon: “is used for past time event” clearly only 
> captures a tiny bit of, say, the English PST marker, but it does 
> capture something real.  In fact,  a typologial variable might pick up 
> something that is indeed very real because it directly corresponds to 
> electrophysiologically detectable patterns (see e.g. Bickel et al. 
> 2015 in PloS ONE 
> <http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0132819> 
> on the S=A vs S≠A variable).  (And, yes, saying that Chinese is SVO 
> captures something real in Chinese, but I agree with others that the 
> traditional terms and abbreviations here are misleading.)
> The contrast in (ii) is where we get a real opposition between 
> different approaches for analyzing languages, perhaps even 
> sub-disciplines. Given the complexity of language, though, it won’t 
> harm to use both approaches simultaneously. Where things get tricky 
> and confusing is if you want to design typological variables not in 
> the sense of (i) but for comparing the language-specific categories in 
> the sense of (ii). I wouldn’t.
> Balthasar Bickel.
> More on this:
> Bickel, B 2007. Typology in the 21st century: major current 
> developments. /Ling. Typol./ 11. 239–251.
> Bickel, B 2010. Capturing particulars and universals in clause 
> linkage: a multivariate analysis. In I Bril (ed.), /Clause-hierarchy 
> and clause-linking/, 51–101. Amsterdam: Benjamins.
> Bickel, B 2011. Multivariate typology and field linguistics: a case 
> study on detransitivization in Kiranti (Sino-Tibetan). /Proc. Conf. 
> Language Documentation and Linguistic Theory/ 3, 
> 3–13.(http://www.hrelp.org/publications/ldlt3/papers/ldlt3_02.pdf)
> Bickel, B 2015. Distributional typology: statistical inquiries into 
> the dynamics of linguistic diversity. In B. Heine & H. Narrog (eds.), 
> /The Oxford handbook of linguistic analysis, 2nd edition,/ 901 – 923. 
> Oxford: Oxford University Press.
> _______________________________________________
> Lingtyp mailing list
> Lingtyp at listserv.linguistlist.org
> http://listserv.linguistlist.org/mailman/listinfo/lingtyp

Martin Haspelmath (haspelmath at shh.mpg.de)
Max Planck Institute for the Science of Human History
Kahlaische Strasse 10	
D-07745 Jena
Leipzig University
Beethovenstrasse 15
D-04107 Leipzig

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/lingtyp/attachments/20160201/aaec48d8/attachment.htm>

More information about the Lingtyp mailing list