possessives

[David Beck]

Okay, let me clarify my position a bit.
>
>     In posing this question, you seem to be assuming that that
>the distinction between "possession" and "nominal modifiers like
>...." can be understood in a language-independent fashion - so
>that for each language, it would be possible to answer whether or
>not it expresses this distinction. Yet this is false.

What I am assuming is that semantic relations like "possession" (in the
broad, deictic sense), modification, and attribution do represent universal
semantic relations and in representational terms can be modelled using a
consistent set of configurations. It is starting to look to me (and more so
since everyone on the list has so kindly provded me with examples) like
these can be placed on a sort of continuum (probably having to do with
predicativity) running from "possession" (semantically a deictic relation
between two names/things) through "attribution" (a relation between two
names/things mediated by a predicate which is elided in the syntax) to
"modification" (a relation between a semantic predicate and a name/thing
which is directly its argument).

Languages should always "differentiate" between these semantic
configuration in the sense that they should have ways to express them, but
in some languages the means of expression are the same. That is, certain
semantic distinctions (perceived by speakers) are neutralized in the
syntax. In Chinese, for instance, all three of these relations are
expressed with the relativizing particle, de. Hausa conflates possession
and attribution (using the linker -n/-r) but distinguishes modification
(which is done with a handful of adjectives or--depending on how you define
modification--with the particle mai and a quality noun). Totonac
distinguishes all three very clearly, although there is a type of
impersonal possession that overlaps structurally with attribution.

My interest the possession-by-juxtaposition languages was to see if they
neutralize all three semantic configurations in the syntax or group them in
an odd way.
>
>   Imagine now, a language has only one construction of this
>kind. Does this mean that this language does not differentiate
>between "possession" and "nominal modifiers like ..."? Again,
>depends on what you mean. In a sense, yes: it has only a single
>construction for what may be called nominal modifiers, and this
>construction is likely to cover what may be called possession
>*and* lots of other meanings which would depart more or less
>significantly from "possession" in the strict sense and be more
>or less "attributive".

Here I would want to claim that the language neutralizes an underlying
semantic distinction. Naturally, there is a contiuum of meanings. In
languages that do make the distinction, I would expect there to be
variation as to which side of the dividing line some of the intermediate
meanings fell on.

>But what if this construction cannot be
>applied to render attributive meanings like in "fire truck" or
>"steak knife", i.e., such attributes (or some of them) are not
>expressed by "nominal modifiers"?
>
This is another kind of variation that I am interested in. The attributive
"range" of the continuum (if it is that) is actually subdividable (I would
analyze mai constructions in Hausa as types of attribution, actually), and
I am curious to see how. The fact that attributive elided a smenatic
predicate im plies that there could be a number of attributive
constructions in a languages depending on the nature of the elided
predicate.

> what sort of answer do you expect
>to find in a reference grammar?

Well, I _wouldn't_ expect to find an answer in most, although I have teased
some out of a few particularly good ones that have texts and lots of
examples.

Cheers

David

======================================================================
David Beck
Programme in Linguistics
University of Michigan
105 S. State St.
Ann Arbor, Michigan 48109
USA
e-mail: (as of Sept. 7) dbeck at umich.edu
phone: (734) 978-4029