morphosyntactic feature geometries

[Heidi Harley]

hey martha!

 OK, I've given this some more thought, and I think there may indeed
 be a way to get disjunctive interpretations without negative feature
 values.  Suppose that all languages have a [Speaker, Addressee]
 category syntactically, though only some have an [Addressee] feature
 morphologically.  If so, we can safely make the simplifying
 assumption that the [Speaker] (only) syntactic category *always*
 means "first person exclusive" (i.e. the meaning of the syntactic
 node is rigid, not determined relative to other syntactic nodes). If
 there's no specific [Speaker, Addressee] vocab item, a [Speaker] item
 -- e.g. English "we" -- can be inserted EITHER into a [Speaker] node,
 OR into a [Speaker, Addressee] node.  Does that make sense?

Well, it certainly could work, but it does carry the somewhat curious
consequence that a language's syntax has a full geometry of features
even when there is no morphological evidence present in the language
showing that they're there. Betsy and I were thinking more of an
incremental acquisition approach, where the tree was elaborated as the
morphological contrasts were observed, on the basis of positive
evidence. Consequently no English speaker would ever get beyond just
having the [Speaker] feature --  they'd never (syntactically) activate
the [Addressee] feature. So any meaning compatible with a [+Speaker]
interpretation would have to be represented with the same
[Adressee]-free geometry in English, whether the intended interp was
inclusive or exclusive, which I guess would result in semantic
underspecification. I now forget what your original point about the
problem with that was... sorry! But you're right that if the whole tree
is used in the syntax of all languages, given by UG, then this would
give you the full semantic specification while still allowing for
morphololgical underspec.

 >W/r to the subgraph/subtree contrast mentioned by Rolf, if Arabic
 't-'
 >does realize just 2, it could do so by being specified for the H&R
 >subtree headed by Hearer (since Arabic has the incl/excl distinction,
 >Hearer is active there). Hearer doesn't dominate number, so there's
 no
 >entailment relationship between 2 and number.

 ... So for H&R,
 [Participant, Addressee] would be a subtree, [Participant, Addressee,
 Individuation, Group]  would be a subgraph, and [Participant,
 Addressee, Class, Animate] would be neither?

Actually, I think Rolf was thinking that any subtree with Participant in
it would necessarily need to have Individuation in it too, so you
couldn't have 't-' realizing 2 without also realizing number, so he was
distinguishing between subtree (which i think on Rolf's terminology
would have to include a root) and a subgraph (which is closer to the
syntactic notion of subtree, I think). (Rolf, is that sort of like what
you were getting at?) Anyway, I think I was trying to suggest that if
you fission off the 'subgraph' headed by Individuation, then you could
realize '2' as a subtree including root without predicting that number
should still be there as well.

the [participant, addressee, class, animate] collection of features
could not exist without an [individuation] node, though, so absolutely
that would be an impossible collection as either a subtree or subgraph.

:) hh

 Cheers,
 Martha
 --
 mcginnis at ucalgary.ca