2 possessors: degree of grammaticalization?

[Martin Haspelmath]

Maybe it is possible to correlate the different behavior of possessor
phrases across languages with the degree of grammaticalization of their
construction. Let me elaborate:

Matthew Dryer suggests that two possessors (or "genitives" -- but
perhaps this terms should be reserved for "genitive case") are excluded
when they occur in exactly the same construction, just like languages
don't generally allow two subjects or two direct objects.

But if there is such a "Stratal Uniqueness Law" (the Relational Grammar
term for this constraint), this is certainly not absolute. We can see
this in possessive constructions (cf. Basque, where Luzia-ren zu-re
argazki-a 'Luzia-GEN I-GEN photo-DEF' seems to be much better than
'Luzia's my photo' is in English, according to Alan King), and we can
also see this in clause-level syntactic functions: Some languages do
seem to allow two direct-object arguments, or two indirect-object
arguments.

But things don't seem to be entirely random: Most importantly, the
strength of the Uniqueness Law seems to be governed by the Syntactic
Functions Hierarchy:

Subject < Direct Object < Indirect Object < Oblique

The higher a function is on the hierarchy, the stronger the Uniqueness
constraint. Less strongly grammaticalized functions such as temporal or
locative adjuncts can occur multiply without problems in most languages.
(This is related to Christian Lehmann's grammaticalization parameter of
"paradigmatic cohesion" -- see his 1995 Lincom book).

Similarly, I suggest that more strongly grammaticalized possessive
constructions are less likely to occur doubly, cf.

Spanish    la parte de Juan de la casa
vs. Basque   ?etxe-a-ren Jon-en zati-a (both examples from Alan King)

German   der Teil von Erika von dem Haus
vs.             *der Teil Erikas des Hauses

In Basque, Spanish, and German, the generalization holds that (more
strongly grammaticalized) genitive-case possessors cannot be doubled
(easily), whereas (less grammaticalized) adpositional possessors can. Of
course, since this is a hierarchy, not all languages will conform to
this particular generalization, but I would expect that they all obey
the hierarchy (e.g. if a genitive-case possessor can be doubled, then so
can an adpositional possessor).

This still leaves us with no explanation of Edith's Hungarian facts,
because as Edith has pointed out, Hungarian has two possessive
constructions, and one clearly looks less grammaticalized (the one with
the -nak/-nek case-marker). Note, however, that this latter construction
also requires the possessive indexing suffix on the possessum, and this
(sub-)construction is pretty strongly grammaticalized (in this case,
morphologized). So the Hungarian facts are not really surprising from
this perspective.

Martin
--
Dr. Martin Haspelmath (haspelmath at eva.mpg.de)
Max-Planck-Institut fuer evolutionaere Anthropologie, Inselstr. 22
D-04103 Leipzig (Tel. (MPI) +49-341-9952 307, (priv.) +49-341-980 1616)