cross-linguistic categorization

[Balthasar Bickel]

I agree with Dan Everett that abstractions like the IPA are excellent points of departure in fieldwork (what else could we do!), and I also agree with Tom Givon that universals are excellent guideposts in fieldwork. (And so one of the best ways of preparing for fieldwork is still to read a lot of typology literature, in addition to grammars of neighboring languages.)

In fact, analyses of individual languages have become better over the past few decades precisely because when we go to the field, we now carry more and more precise analytical notions with us --- all critically informed by comparative work: instead of hunting for 'words' tout court we now look for various rule and constraint domains in phonology and grammar, instead of 'agreement', we are aware of many different varieties of agreement etc., instead 'subject', we look for coding and behaviour of arguments under various conditions  etc.. The progress in all this is the same as the progress in typology: typology gets better the more it is based not on gross types that sweep many language-specific distinctions under the carpet, but on the kind of fine-grained notions that we also need in descriptive work. From this point of view, refining the tool set for analyzing an individual language and refining the tool set of typology remains essentially the same enterprise. (I make this argument at length in a recent case study on clause linkage available from my web site: "Capturing particulars and universals in clause linkage: a multivariate analysis")

Terms like 'alignment' look like they are not needed in descriptive work. But in fact we use them in descriptive work all the time when we describe the distribution of, say, a case marker and list the arguments that are covered by the marker. It is true that in typological surveys we usually don't talk about case distributions in terms of lists or sets of arguments but use different statements and notations, e.g. 'S=A vs. P' or 'accusative' etc, but these statements and notations are nothing but reformulations --- often unduly simplified --- of lists of arguments; in fact, one can be computed from the other (which is what we indeed do currently in an AUTOTYP database on grammatical relations).

Balthasar Bickel.




_______________________
www.uni-leipzig.de/~bickel
www.uni-leipzig.de/~autotyp
www.uni-leipzig.de/~asw