[Lingtyp] Right node raising cross-linguistically

[David Gil]

Dear all,

I strongly agree with Adam Tallman's latest post, and also with much of 
Martin's latest, though I do take issue with one of Martin's assertions:

On 29/09/2021 22:18, Martin Haspelmath wrote:
> As Adam Tallman has noted (here and in various other venues), 
> constituent structure is not always determinate because different 
> phenomena may argue for different structures – this is of course an 
> experience familiar to typologists. So we'd need comparative concepts 
> that do not make reference to abstract and language-dependent notions 
> such as constituent structure.

I do not see why comparative concepts should not "make reference to 
abstract and language-dependent notions such as constituent structure".  
Most of us probably agree that languages have syntactic constituent 
structures, and if we are agreed on this, we we should surely be able to 
posit constituent structure as a comparative concept, even though the 
actual determination of constituencies may be language specific and also 
theory dependent.  The difficulty in determining constituencies was the 
very good reason why, when working on WALS, we did not contemplate, say, 
including a map that would show whether in a transitive clause with two 
overt arguments, one of the arguments forms a constituent with a verb to 
the exclusion of the other. (The old "Does it have a VP"? question.)  
But just because it would have been (and indeed still is) impractical to 
conduct a large cross-linguistic survey of this feature in an objective 
and consistent fashion, does not mean that it is not an important 
question to think about.

Now throw in coordination, a relatively uncontroversial comparative 
concept, and we are in a position to formulate meaningful questions 
about languages from a typological comparative perspective, such as the 
following:

(1) Does it allow coordination of strings that are not constituents?

Martin seems to be suggesting that because we do not have (and may never 
have) a universally valid toolkit of diagnostic properties that we can 
apply automatically and objectively to each and every language, then 
constituency cannot serve as a useful comparative concept, and therefore 
question (1) above is not a meaningful question for typologists to be 
asking.  On the other hand, I would maintain that it is perfectly 
legitimate for us to take one set of criteria for constituency in 
language L, another set of criteria for language M, yet another set for 
language N, and then, for each analysis of each language, pose the above 
question in (1), and thereby come up with a meaningful comparative 
concept, which, in this case, will end up relating to the notion of 
right node raising.

David


-- 

David Gil

Senior Scientist (Associate)
Department of Linguistic and Cultural Evolution
Max Planck Institute for Evolutionary Anthropology
Deutscher Platz 6, Leipzig, 04103, Germany

Email: gil at shh.mpg.de
Mobile Phone (Israel): +972-526117713
Mobile Phone (Indonesia): +62-81344082091