Martha McGinnis: Predictions?

[Martha McGinnis]

I thought I would try to respond to Andrew Carstairs-McCarthy's
request for a summary of correct predictions made by DM.  I
should say first of all that I don't think any one person can
satisfy this request, which seems to call for a summary of
both the current achievements and the as-yet-unrealized potential
of the DM literature..!  A tall order for a Sunday morning.
What I thought I would do instead is summarize what I've tried to
argue in my own work within DM.  I hope others will write in and
do likewise.

In one paper (in the proceedings of a Workshop on Structure
and Constituency in Native American Languages held at the U.
of Manitoba, recently published as a MITOPL) I argued that the
so-called "inverse" morphology in Ojibwa does not reflect
syntactic or morphological "inversion," which is supposed
to involve the logical subject becoming a syntactic object,
and a logical object becoming the subject.  I argued that
such inversion does not occur syntactically.  According
to DM, morphological operations are subject to the same
constraints as syntactic operations, making the strong
prediction is that inversion cannot be brought about
morphologically, either.  Following Halle and Marantz (1993),
I proposed that the apparent inversion in Ojibwa arises from
(a) the fact that a single agreement slot can be sensitive
to features of both the subject and the object, and (b)
competition among vocabulary items for a single slot, such
that (for example) the first-person item might always beat
the second-person item, when both are compatible with the
features of the agreement node.  I argued that the competition
analysis makes the right predictions for the distribution
of first- and second-person agreement in the conjunct form
in Ojibwa, by contrast with the inversion analysis.  For more
details, I refer you to the paper.  (I'll try to post it
on the web tomorrow.)

A second paper (published in the proceedings of the 1996
Seoul International Conference on Generative Grammar)
argued against grouping word formation rules into arbitrary
blocks of mutually exclusive rules, in the manner of
Stephen Anderson's theory of A-Morphous Morphology (AMM).
As you know, AMM captures the insight (adopted as a central
concept of DM) that morphophonological information is
syntactically and semantically underspecified, so a fully
specified syntax can't be "projected" from it.  Rather,
morphophonological "stuff" (for DM, vocabulary items; for
AMM, the phonological information inserted by word-formation
rules) competes for insertion into a fully specified
representation, with the winning stuff blocking the
insertion of other stuff. However, DM differs from AMM
in the details of how competition and blocking takes place.
Specifically, DM has a concept of "spelling out" or
"exhausting" the features of a syntactic representation
by inserting vocabulary items, while AMM treats blocking
solely as the result of various types of interactions
among word-formation rules.  The "spelling out" theory
makes the right predictions for the distribution of
the plural suffix of Georgian verbs, by contrast with
an approach that makes no reference to the representation
to be spelled out.  Again, the details are in the
paper (which I'll also try to post tomorrow).

I also have a third paper, a short ms., in which I tried
to see if there's a specific kind of long-distance selectional
relations among derivational affixes and stems in English,
differing in nature from the local selectional relations among
syntactic categories.  I did an extensive search and didn't find
any, which is at least weak support for the DM view that
relations below the word level follow the same principles as
relations above the word level. (This paper is still a bit
drafty, so I'm not going to post it just yet, but if anyone's
interested, please write to me directly.)

It's worth noting that, like any generative linguistic
theory, DM doesn't directly constrain surface forms.  What
it constrains is the range of structures and derivations
underlying surface forms.  So I can't point to a verb form
and say, "DM could never have generated this!"  The goal of
DM, as I understand it, is instead to clarify the connection
between morphology and syntax in such a way that each can be
used to make predictions about the other.

Best,
Martha