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.
233<br />
For an important class <strong>of</strong> syntactic dependencies, <strong>the</strong>ir formation may be attributed<br />
to PF or LF requirements. A wh-word might only be interpretable if it occurs<br />
in both a <strong>the</strong>matic <strong>and</strong> a peripheral/scopal position, <strong>and</strong> wh-movement satisfies<br />
this requirement. This is particularly so for <strong>the</strong> A'-system, with its<br />
characteristic interpretive consequences (Rizzi 1997, 2006). Movement scopes a<br />
quantifier above a proposition, turns a sentence into a relative clause, creates a<br />
topic-focus articulation. Yet empirically, such dependencies make use <strong>of</strong> <strong>features</strong><br />
that are not interpretable at LF. In (23), each step <strong>of</strong> wh-movement must take place<br />
to satisfy <strong>the</strong> interpretive requirements <strong>of</strong> <strong>the</strong> wh-word, but <strong>the</strong>y are accompanied<br />
by uninterpretable phi-agreement in English or Passamaquoddy (Bruening 2001),<br />
or syntactic category agreement Chamorro, Welsh, <strong>and</strong> Breton (Chung 1998: 6.2,<br />
Willis 1998: 3.4, Rezac 2004b).<br />
(354) Where are <strong>the</strong> boysi whoi Tom think(s) ti Dick believe(s) ti Harry expect(s)<br />
ti to be late?<br />
(Boston English, Kimball <strong>and</strong> Aissen 1971: 246; section 1.3)<br />
It has been proposed that uninterpretable <strong>features</strong> are simply <strong>the</strong> optimal device<br />
to build syntactic dependencies like wh-movement. If so, <strong>the</strong> <strong>features</strong> are motivated<br />
as a class by <strong>the</strong> need <strong>of</strong> PF/LF for <strong>the</strong>re to be syntactic dependencies (cf.<br />
Chomsky 2000a: 120f., 2008: 140f., 148). On st<strong>and</strong>ard assumptions, <strong>the</strong>se uninterpretable<br />
<strong>features</strong> are <strong>the</strong> lexical properties <strong>of</strong> items like <strong>the</strong> interrogative CQ,<br />
<strong>and</strong> thus a departure from <strong>the</strong> Interpretability Condition (27)b that lexical items<br />
consists only <strong>of</strong> PF/LF properties.<br />
If <strong>the</strong> PF/LF conditions that need syntactic dependencies were construable as<br />
Full Interpretation requirements, <strong>the</strong>y could become <strong>the</strong> direct source <strong>of</strong> individual<br />
occurrences <strong>of</strong> uninterpretable <strong>features</strong> through ℜ. At each nonfinal step <strong>of</strong> whmovement<br />
in (23), <strong>the</strong> wh-word fails to scope over a question. If this is a Full Interpretation<br />
problem for LF, ℜ can insert an uninterpretable feature to drive whmovement<br />
without <strong>the</strong>re being one in <strong>the</strong> lexicon. This potential is partly exploited<br />
in Chomsky's antecedents (281)-(282) <strong>of</strong> ℜ, for successive-cyclic <strong>and</strong> terminal A'movements<br />
(cf. Chomsky 1995: 294, 377, 2000a: 109, 2001: 34). If this were feasible,<br />
it would not be necessary to posit uninterpretable <strong>features</strong> on lexical items,<br />
contrary to <strong>the</strong> Interpretability Condition. They could be always viewed as <strong>the</strong> direct,<br />
dynamic response <strong>of</strong> syntax to <strong>the</strong> needs <strong>of</strong> Full Interpretation. 174<br />
However, structural Agree/Case relationships do not seem to occur to meet<br />
LF needs, as section 5.4 has discussed. They exist without <strong>the</strong>matic, scopal, binding<br />
correlates, <strong>and</strong> include DPs without autonomous referential or quantificational<br />
content (cf. (271)). They also do not appear to meet obvious PF needs, unlike for<br />
instance displacements that may be sensitive to PF content (Chomsky 2001; Lasnik<br />
2009; Holmberg 2000, 2005, Jouitteau 2005). Moreover, <strong>the</strong> uninterpretable<br />
174 The choice <strong>of</strong> <strong>the</strong> inserted uninterpretable <strong>features</strong> is a technical matter: specific <strong>features</strong> such<br />
as [wh:] on Cinterrogative <strong>and</strong> [Q:] on a wh-word, or generalized Edge Features (Chomsky 1995:<br />
289ff., 2000a: 109, 128ff., 149 note 91, Rizzi 2006; Chomsky 2001: 6, 34, 2007, 2008: 151).