universality of gfs

[Farrell Ackerman]

for what it's worth, here's my private collection of misunderstandings
concerning rachel's
inquiry:

it's easy to agree with both paul and joan concerning the distinction
between a formalism
and the vision of grammar implemented within a formalism.  given this, it's
possible to
imagine a kind of  lfg without gfs (if it proves that this is empirically
necessary) and it's possible to imagine a kind of cg containing a theory of
universals (if this proves desirable).  (in fact, a variant of the
latter is developed by gert webelhuth and myself in terms of what we refer
to as "grammatical archetypes" (see also harris and campbell on "universal
syntactic
constructions") in our recently published book `a theory of predicates'-
the motivation for this was a recognition similar to joan's that we like she

>.. like the CG focus on irreducible language particularities; but ...
>also like the LFG/OT focus on the universal structure in the
>typological space.

and we tried to develop a typologically responsible universal theory responsive
to markedness considerations within the multiple
inheritance type hierarchy assumptions of both cg and hpsg, since we felt
that neither cg nor hpsg provided such a theory - perhaps oddly,  we felt
we were
developing a kind of lfg in this enterprise as well!)

given this state of affairs, namely, a theory with a commitment to gfs as
universals and a theory without a commitment to characterizing "universal
structure in the
typological space", it is hard to see why a researcher would select
"goldberg's cg" over lfg, except for purely local and utilitarian concerns: if
one is concerned solely with argument selection, then it would be
reasonable to contrast
lfg's lexical mapping theory (bresnan and kanerva etc.) or functional
mapping theory (alsina) with goldberg's argument structure construction
proposal in this domain: personally
over the years I have found both lmt and fmt to be the least compelling
aspects of
lfg, but it seems
strange to me to contrast "goldberg's  construction grammar", i.e., a
hypothesis about
argument selection" with lfg, i.e., a general theory of grammar, as rachel
suggests is
happening -
the former is a theory of argument selection based on empirical
observations which are
relevant to all argument selection stories and addressable by several sorts
of theories
(see levin and rappaport (to appear) where it is argued that argument
structure constructions on the goldberg account may well be a notational
variant
of a lexicalist theory informed by "event structure".)  as I see it,
goldberg's argument structure proposal represents a particularly dexterous
and engaging interpretation of fillmore and kay's construction grammar
within the domain of argument selection: as
paul effectively confirms, goldberg's proposal may share the general cg
lack of commitment to universal gfs, on the other hand, the general lack of
explicit emphasis on defining the typological space of grammar and relating
language particular properties to universals within cg,
as contrasted with lfg, could easily be thought of as representing a good
reason to be cautious
about adopting cg over lfg in general (at this stage of the game.): on this
level we would
appear to be contrasting committments of general theories of grammar.   so,
I guess that on a fairly abstract level, I just don't see why one would
prefer cg to lfg in general, rather than
preferring certain mapping or selection hypotheses to certain others (all
formulable, with
enough "squinting", within pretty much the same family of formalisms.)

on a more concrete level, i'd like to know more about what specific
empirical arguments are being
offered for the claim that gfs are not universal or alternatively, that some
interpretration of argument structure is enough.  as mentioned by a preceding
commenter, aissen has argued pretty compellingly for the role of gfs in
tzotzil (what
mayan languages are being alluded to for which "thematic roles"or "argument
structure" are sufficient?) it is certainly conceivable that gfs may not be
universal and various theories have been
formulated to express this - this clearly has intriguing consequences for
both diachrony and
language acquisition, let alone the evolution of language (see deacon, the
symbolic species): for example, dik's functional grammar is quite explicit
in making this claim (see simon dik, studies in functional grammar 1980),
where it is claimed
that serbo-croatian is one such language.  whether serbo-croatian is truly
a good example -
and I don't think it is - the question naturally arises concerning
the criteria that are being appealed to to characterize "thematic roles"
such that one
can confidently appeal to them in a theory of grammar.  Dik's thematic
roles differ
in some respects from those used by others: this should matter, shouldn't
it?  there is a burgeoning literature on why appealing to atomic thematic
roles may even be a bad bet (see Dowty 1989, 1991, among others).  these
criticisms about thematic roles may prove to be true, and yet it still may
be possible to provide evidence that the analysis of a language need
not appeal to gfs, but rather to lexical semantic entailments of predicates
for grammatical
generalizations. I guess i'm in the dark at the moment about how to react
specifically to the claims about the absence of gfs in the Australian
language Jaminjun  because I don't know
what specific assumptions are being made and how these congeal into an
analysis of
grammatical phenomena in that language: is it just the absence of
universalist claims for
gfs that makes cg a more appropriate framework to formulate an analysis of
this language in?  how does goldberg's story formulated for english and
cached out in terms of surface
structure arrangements of nouns and verbs translate into languages where
there is a lot of
morphology and surface orders are less relevant.  I know that this is
supposed to follow
from the claim that both phrases (including sentences) and morphology are
"constructions",
but i'm not really sure that I know what this means in practice.  in
contrast, lfg has a fairly strong and explicit tradition of characterizing
the relation between syntax and morphology under the recent rubric
"morphology competes with syntax":  mightn't this fairly
well-developed aspect of lfg be more useful than cg presently in a language
where morphology has syntactic force?

finally, a minor detail that I don't understand concerning the terms
active/stative.  I thought
these terms referred to the general typology of word classes and sentence
structures developed by Klimov, specifically  in his book `the typology of
the active system' 1977), which as pointed out by Dixon in his book on
ergativity, corresponds, more or less, to fluid-S marking.  this sort of
system, though clearly displaying semantic influence on surface
encoding, does not itself preclude the existence of gfs, as beautifully
demonstrated in
cathy o'connor's thesis on northern pomo.  is this how the terms were being
used in the
previous messages?

-farrell

Farrell Ackerman                (619) 534-1158  (office)
Dept. of Linguistics            (619) 534-4789  (fax)
UC San Diego                     ackerman at ling.ucsd.edu
La Jolla, CA. 92093-0108