syncretism w/o paradigms

[Martha McGinnis]

Hi Heidi,

Hmmmm indeed.

>Each time the learner hears a verb form that's identical with a form they
>have remembered hearing with a featurally distinct subject, they say,
>aha! that's a place where I have to invent an Impoverishment rule.
>Then, the syncretisms they observe in, e.g. the 'to have' paradigm will
>be predictive of syncretisms they expect to hear in other paradigms
>(though if they're not, as in the 'to be' paradigm, that's no problem
>either, they just go back and annotate their initial Impoverishment
>rule with 'not for "to be"', or 'only for verbs of class X').

Just to clarify -- I don't have JDB's paper (obviously I need to get
it) -- but is the point that none of the English past tense
Vocabulary items make person distinctions, except "is/was/were"?  If
so, then the issue would be whether this is a deep generalization
about English past tense (hence, Impoverishment) or a coincidence
(hence, underspecification of Vocab items).

There's something unnerving to me about the Impoverishment analysis
you sketched above.  I've always felt that Impoverishment rules
should be posited only for 'special cases', i.e. when syntactic
representations and Vocab underspecification can't account for the
facts at hand.  Using it for cases of metasyncretism seems to lose
the distinction between 'general cases' of syncretism (which are also
consistent with Vocab underspecification) and these 'special cases',
e.g. a case from Piedmontese described by Bonet, where the 1pl
reflexive clitic looks like the default 3rd person reflexive, instead
of like the 1sg reflexive.

I also think of Impoverishment as a rule that targets a limited,
well-defined case -- so (speaking broadly) if there's an exception to
a morphological generalization, the exception, not the
generalization, should be due to Impoverishment.  But the English
case you described goes the other way: Impoverishment gives us the
general case (no person agreement in past tense) and then we need to
prevent it from applying in the more specific case (e.g. by saying
"except for 'be'").  Another view would be that "be" constitutes
evidence that we actually *don't* Impoverish person in the past
tense, so the only possible analysis of the absence of person
distinctions elsewhere is Vocabulary-based (i.e., coincidence).  I'm
not sure how to decide between these two views.

How would the learner decide?  The Impoverishment analysis seems to
require the learner's grammar to notice a (semi-systematic) absence
of distinctions, and posit a corresponding rule; while the Vocab
analysis requires the learner's grammar to notice the presence of
morphological distinctions, and posit corresponding Vocabulary items.
The latter view makes more sense to me -- let's see if I can say why.
Suppose UG gives us a multidimensional syntactic/semantic "space" for
making morphosyntactic distinctions, but the features we actually
posit for Vocabulary items are activated by positive evidence in the
form of contrasts.  Under this view, our initial hypotheses about
Vocabulary items would be maximally underspecified, and
specifications would be added as necessary.  For example, we might
originally posit that "her" is the only 3sg.f pronoun, but when we
realize that "she" also exists, we would invoke a case feature: "she"
is 3sg.f.nom.  If nothing else happens, then 3sg.f genitive and
accusative will remain identical.

Why would we take the additional step of impoverishing genitive case
in 3sg.f?  It seems to me that this would require some motivation --
i.e., evidence that the existing grammar is incorrect.  The
Impoverishment analysis maintains that the motivation is
paradigmatic: *elsewhere* English has a gen/acc distinction, so we
expect one for 3sg.fem, and if there isn't one, we need to impoverish
it away.  But if there are no sub-paradigms, i.e. just one huge
paradigmatic space for all morphosyntactic distinctions, then
shouldn't we also have to notice that there are no tense, mood,
mass/count, positive/negative, proximal/distal, ETC distinctions in
3sg.fem pronouns?  If so, we'd need to posit a huge number of
Impoverishment rules to get the fact that there are just 2 forms of
the 3sg.fem pronoun (or 3, including "hers").  I can't think of any
coherent empirical arguments against this, but I do find it highly
implausible.  The problem may well be lack of imagination, though.
Any help is most welcome!

Best,
Martha
--
mcginnis at ucalgary.ca