<!doctype html public "-//W3C//DTD W3 HTML//EN">
<html><head><style type="text/css"><!--
blockquote, dl, ul, ol, li { margin-top: 0 ; margin-bottom: 0 }
--></style><title>HPSG vs PP/MP: empirical
differences</title></head><body>
<div>Regarding the comparison of HPSG with GB/P&P/MP, I'd like to
plug my paper:</div>
<div><br></div>
<div><font face="Geneva" size="+2" color="#000000">W</font><font
face="Geneva" color="#000000">ECHSLER<font size="+2">,
S</font>TEPHEN<font size="+2">. 1999. HPSG, GB, and the Balinese Bind.
In Lexical And Constructional Aspects of Linguistic Explanation, ed.
by</font></font><font face="Times" size="+2" color="#000000"> A.
Kathol</font><font face="Geneva" size="+2"
color="#000000">,</font><font face="Times" size="+2" color="#000000">
J.-P. Koenig</font><font face="Geneva" size="+2" color="#000000">
and</font><font face="Times" size="+2" color="#000000">
G.Webelhuth</font><font face="Geneva" size="+2" color="#000000">.
Stanford: CSLI.</font></div>
<div><font face="Geneva" size="+2" color="#000000"><br></font></div>
<div>It's a case study in 'the myth of the notational variant.'
I compare HPSG and GB analyses of binding in a Western Austronesian
language (Balinese), that are, in some intuitive sense 'the same',
under correspondences of the following sort:</div>
<div><br></div>
<div>HPSG:<x-tab>
</x-tab><x-tab>
</x-tab><x-tab>
</x-tab><x-tab>
</x-tab><x-tab>
</x-tab>GB:</div>
<div>ARG-ST list items <x-tab>
</x-tab><x-tab>
</x-tab><x-tab>
</x-tab>theta positions (under VP)</div>
<div>relative obliqueness<x-tab>
</x-tab><x-tab>
</x-tab><x-tab>
</x-tab>relative c-command</div>
<div>VALENCE list items<x-tab>
</x-tab><x-tab>
</x-tab><x-tab>
</x-tab>spec's of functional projections</div>
<div>ARG-ST/VALENCE
structure-sharing<x-tab>
</x-tab>chains</div>
<div><br></div>
<div>The two analyses make the same predictions for simple Balinese
examples. But they diverge radically when the data become more
complex: the HPSG analysis correctly applies without any
alteration, while the GB analysis turns out to be deeply
flawed. It's not a minor problem, fixable by tinkering
with the definition of c-command or something. In fact it seems
that a GB analysis of W. Austronesian binding is impossible under
normal assumptions. It's a very surprising result. The parts of
the machine are very similar in the two frameworks, but crucial
differences arise when the parts are assembled.</div>
<div><x-tab> </x-tab>The
problem for MIT theories arises from the propensity for modeling
different types of abstract relation (argument structure, grammatical
relations, etc.) within a single phrase structure representation-- a
model from the 'Old Stone Age', as Bob Levine put it. </div>
<div><x-tab> </x-tab>To my
knowledge the only GB syntactician to take up my challenge is Lisa
Travis, who proposed a radical overhaul of GB theta theory and binding
theory to allow for these languages.</div>
<div><br></div>
<div>--Steve</div>
</body>
</html>