Doubled prepositions in NPs

[Prof. Dr. Wolfgang Schulze]

Dear Bernhard, dear Frans, dear Caspar, dear others,
MANY thanks for all the very helpful indications! Let me briefly comment 
upon the given data:

The fact that in Hungarian

> [t]here is agreement only with demonstratives, not with attributive
> adjectives, as with cases, and some sources say that the agreement of
> demonstratives with postpositions is optional [Bernhard]

in fact supports my assumption that doubled prepositions in NPs are only 
possible if both segments marked by pre/postpositions are overtly or 
covertly referential in nature. The same seems to hold for those cases 
Caspar has quoted, especially:

> Piros-at   kettö-t      vett      Mari  cipö-t.
> red-ACC two-ACC bought Mary shoe-ACC
> 'Mary bought two red shoes.'
>  
> Mari kettö piros cipö-t         vett.
> Mary two  red    shoe-ACC bought
> 'Mary bought two red shoes.'

In my interpretation, the first example would (diachronically) read: 
'Red (ones), two (*ones) - bought Mary shoes' (ins't this a focus 
construction, by the way?). I havn't yet checked all the references that 
Frans has mentioned, nevertheless they seem to confirm that doubled 
pre/postpositions normally occur only if the qualifying element has 
acquired referential properties (adnominal > nominal). This assumption 
explains for instance the following pair in Udi (East Caucasian) - 
confirmed by informants:

(1)  sa     kol-l-a                  qosh     k'ic'k'-ot'-un         
          qosh
       one  bush-SA-GEN   behind  small-REF:OBL-GEN  behind
       'Behind a small bush' (recte: Behind a bush, behind a small one')

vs.

(2) sa     k'ic'k'e   kol-l-a                qosh     [group inflection]
      one  small     bush-SA-GEN  behind
      'Behind a small bush'

(3) *sa k'ic'k'e qosh kolla qosh

In (1) the qualifying segment k'ic'k'e 'small' is clearly marked for a 
morpheme (-ot'-) that signals referentiality (here in the oblique case). 
Consequently, the Udi referents in (1) stand in an appositional 
relation. Personally, I strongly believe that the same holds for the 
grammaticalization background of a number of patterns that involve 'case 
agreement', e.g. Latin amic-a bon-a etc., especially if the (case 
marked) attribute (or: adnominal) can be used referentially, e.g. bon-am 
video 'I see the good one'. Another good example is Classical Arabic 
showing 'article agreement' + 'case agreement', e.g. ar-ra(d)zhul-u 
l-kabi:r-u 'the tall man' (< *the man, the tall one'). Here, the article 
added to kabi:r- clearly indicates that the term has referential 
properties, at least from a diachronic perspective. Nothing new - I 
guess....
Best wishes,
Wolfgang

-- 
Prof. Dr. Wolfgang Schulze
Institut fuer Allgemeine und Vergleichende Sprachwissenschaft
Ludwig-Maximilians-Universitaet Muenchen
Geschwister-Scholl-Platz 1
D-80539 Muenchen
Tel.: ++49-(0)89-2180-2486 (Sekr.)
Tel.: ++49-(0)89-2180-5343 (Office)
Fax : ++49-(0)89-2180-5345
E-mail: W.Schulze at lrz.uni-muenchen.de
Web: http://www.ats.lmu.de./index.php

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/lingtyp/attachments/20060715/7a52806a/attachment.htm>