Phi-features and the Modular Architecture of - UMR 7023 - CNRS
Phi-features and the Modular Architecture of - UMR 7023 - CNRS
Phi-features and the Modular Architecture of - UMR 7023 - CNRS
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
235<br />
The conditions on <strong>the</strong>se dependencies refer to <strong>the</strong> external systems: scope<br />
commutativity <strong>and</strong> string adjacency belong to LF <strong>and</strong> PF, not syntax. It is possible<br />
to formulate <strong>the</strong>m as requirements on <strong>the</strong> input to <strong>the</strong>se systems, Full Interpretation,<br />
<strong>and</strong> thus visible to ℜ (Reinhart 2006: 1.3, 2.7, Chomsky 1995: 377, 2000:<br />
109 for quantifier raising). However, <strong>the</strong>y encroach on ground that belongs properly<br />
within <strong>the</strong> external systems <strong>and</strong> is invisible to syntax <strong>and</strong> to ℜ, like allomorphy<br />
or overlapping reference. This is particularly so if <strong>the</strong> same conditions also<br />
constrain mechanisms properly within <strong>the</strong> external systems, as Fox (2000: chapter<br />
4) shows for variable binding. It is possible to leave <strong>the</strong>m <strong>the</strong>re. 176 The syntax <strong>of</strong><br />
terminal A'-dependencies <strong>the</strong>n need not refer to <strong>the</strong> external systems <strong>and</strong> <strong>the</strong>y may<br />
remain driven by lexically fixed properties. Evidence that <strong>the</strong> latter are uninterpertable<br />
<strong>features</strong> comes from morphology as in (23), from synchronically arbitrary<br />
parametrization independent <strong>of</strong> PF <strong>and</strong> LF, <strong>and</strong> from 'constructional' uses <strong>of</strong> A'dependencies<br />
without <strong>the</strong>ir expected interpretation, if <strong>the</strong>se exist (Culicover <strong>and</strong><br />
Jackend<strong>of</strong>f 1999, 2005, Jackend<strong>of</strong>f 2002).<br />
These considerations suggest that ℜ may not be <strong>the</strong> right way to think about<br />
nonglobal syntactic dependencies. They make use <strong>of</strong> <strong>the</strong> properties <strong>of</strong> lexical<br />
items. Among <strong>the</strong>m are <strong>the</strong> lexicalized uninterpretable <strong>features</strong> <strong>of</strong> <strong>the</strong> Agree/Case<br />
system, not motivated by PF or LF <strong>and</strong> so an 'imperfection' for <strong>the</strong> Strong Minimalist<br />
Theses. Their origins lie in <strong>the</strong> synchronic residue <strong>of</strong> diachrony, consequences<br />
<strong>of</strong> acquisition, <strong>and</strong> in domains still unknown (cf. Bever 2009, Chomsky<br />
2009 on <strong>the</strong> EPP, <strong>and</strong> generally chapter 6). Outside <strong>the</strong> Agree/Case domain, <strong>the</strong><br />
situation is unclear. It is conspicuous that <strong>the</strong>oretically, it seems possible to replace<br />
dedicated A'-<strong>features</strong> with a generalized Edge Feature (Chomsky 2008:<br />
151), <strong>and</strong> empirically, that <strong>the</strong> morphological reflexes <strong>of</strong> A'-probes are <strong>of</strong>ten phi<strong>features</strong>,<br />
as in (23), perhaps under a disguise elsewhere.<br />
For global dependencies, ℜ has <strong>the</strong> right properties. Over <strong>the</strong> development <strong>of</strong><br />
<strong>the</strong> Minimalist Program, <strong>the</strong>re has been a move away from global mechanisms<br />
such as Minimize Structure or Fewest Steps for <strong>the</strong> core Agree/Move dependencies,<br />
while analogues <strong>of</strong> ℜ have remained for o<strong>the</strong>rs such successive-cyclic<br />
movement (cf. Reinhart 1995, 2006, Yang 1997, Collins 1997, Jacobson 1997,<br />
Johnson <strong>and</strong> Lappin 1999, Frampton <strong>and</strong> Gutmann 1999, Chomsky 2000a, Potts<br />
2002). This partitions syntactic phenomena into two intuitively significant groups:<br />
nonglobal Agree/Move dependencies that depend on <strong>the</strong> lexical properties <strong>of</strong> target,<br />
goal, <strong>and</strong> <strong>the</strong> path between <strong>the</strong>m, <strong>and</strong> global ones that depend on <strong>the</strong> global<br />
convergence <strong>of</strong> an entire structure. The latter are <strong>the</strong> province <strong>of</strong> ℜ as an interface<br />
algorithm at <strong>the</strong> edges <strong>of</strong> syntax, relating <strong>the</strong> numeration to legibility at PF <strong>and</strong><br />
LF, within <strong>the</strong> bounds <strong>of</strong> modularity.<br />
The place <strong>of</strong> ℜ in <strong>the</strong> larger architecture <strong>of</strong> language remains all but unexplored.<br />
Reinhart (2006) proposes a unification <strong>of</strong> different global mechanisms as<br />
176 For instance, ra<strong>the</strong>r than quantifiers moving only for new scope, <strong>the</strong>y move freely in syntax,<br />
<strong>and</strong> it is interpretation that decides to use <strong>the</strong> lowest copy unless a higher one gives a different<br />
reading (D. Fox, p.c., citing M. Brody, p.c.; indepedently proposed by Adger 1994: 93-5, 1996).