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
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
234<br />
<strong>features</strong> behind <strong>the</strong>se systems seem partly fixed by <strong>the</strong> parametric properties <strong>of</strong><br />
lexical items that do not transparently reduce to PF/LF legibility, so that languages<br />
differ in <strong>the</strong> inventory <strong>of</strong> lexical items with phi-probes in ways that are synchronically<br />
arbitrary. French <strong>and</strong> Icel<strong>and</strong>ic both have <strong>the</strong> past participle morphology to<br />
reflect agreement with fronted objects, but only some varieties <strong>of</strong> French do so.<br />
This is a parameter that appears irreducible to PF or LF, but depends on <strong>the</strong> arbitrary<br />
presence <strong>of</strong> a phi-probe on <strong>the</strong> participle. It is technically feasible to construct<br />
a system without lexically specified uninterpretable <strong>features</strong> for <strong>the</strong>se systems,<br />
but it does not seem to lead to deeper insight into <strong>the</strong>m. 175<br />
Even in <strong>the</strong> domain <strong>of</strong> A'-dependencies, it is not clear that ℜ is <strong>the</strong> right<br />
mechanism. PCC repairs have motivated ℜ in <strong>the</strong> domain <strong>of</strong> global dependencies<br />
that are not easily stated as properties <strong>of</strong> <strong>the</strong> target, goal, <strong>and</strong> <strong>the</strong> path between<br />
<strong>the</strong>m. Successive-cyclic movement belongs here as well. It requires look-ahead to<br />
<strong>the</strong> terminal dependency without which it cannot occur, so it is global (Bošković<br />
2002: 136-8, 2007). At <strong>the</strong> same time, it is driven by <strong>the</strong> probe-goal system, reflected<br />
even at <strong>the</strong> intermediate steps through agreement morphology, as in (23)<br />
(McCloskey 2002, Chung 1998; Kimball <strong>and</strong> Aissen 1971, Bruening 2001, Carstens<br />
2005). Chomsky (2000a: 107f., 2001: 34) reconciles <strong>the</strong>se two properties<br />
through <strong>the</strong> antecedent <strong>of</strong> ℜ in (281)-(282). Under ℜ, each phase with a wh-word<br />
not in its terminal position might be posited to fail Full Interpretation, licensing<br />
<strong>the</strong> insertion <strong>of</strong> a probe for it (cf. Heck <strong>and</strong> Müller 2003, Bošković 2007).<br />
Extending ℜ to <strong>the</strong> remaining, terminal, A'-dependencies is more dubious.<br />
Some <strong>of</strong> <strong>the</strong>m do appear to refer to PF/LF directly, ra<strong>the</strong>r than to <strong>the</strong> fixed featural<br />
content <strong>of</strong> lexical properties. Extraposition is a convenient example. It avoids<br />
Condition C in (355)a, so it occurs in <strong>the</strong> syntax, but (355)b shows that it can only<br />
apply if not string-vacuous, so it refers to a PF property (Fox 2000: 75f.; but see<br />
Fox 2002: 75 for reanalysis as a parsing preference, cf. Sportiche 2005).<br />
(355) a. I introduced him1 (??to <strong>the</strong> woman that John1 likes) yesterday (to <strong>the</strong><br />
woman that John1 likes).<br />
b. ??I introduced him1 [to <strong>the</strong> woman that John1 likes].<br />
(Fox 2000: 76)<br />
O<strong>the</strong>r examples are quantifier raising <strong>and</strong> scrambling only to allow new scope<br />
in Fox (2000), Reinhart (2006), Miyagawa (2005, 2006), scrambling only to allow<br />
focus assignment in Krifka (1998), Icel<strong>and</strong>ic Object Shift only to avoid newinformation<br />
interpretation in Chomsky (2001: 34ff.), wh-movement only if not<br />
string-vacuous in Chomsky (1986a: 48-54), Fox (2000: 75f.).<br />
175 It may seem that <strong>the</strong> difference between Obligatory (lexically specified) <strong>and</strong> Dependent (ℜcreated)<br />
Agree/Case relations would be lost if both reduced to ℜ. However, if Agree/Case reflected<br />
PF/LF requirement, Obligatory Case loci could be supposed to have an interpretable requirement<br />
to relate to a DP, while <strong>the</strong> Dependent Case loci to satisfy ra<strong>the</strong>r <strong>the</strong> requirement <strong>of</strong> a<br />
DP, paralleling <strong>the</strong> difference between interrogative <strong>and</strong> successive-cyclic C for wh-movement.