26.11.2012 Views

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

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

185<br />

<strong>the</strong>y lead to ineffability. Various interpretive requirements also may be invisible to<br />

syntax, such as Condition B (section 4.6). These must all be excluded from "effects<br />

on output". <strong>Modular</strong>ity suggests that ℜ should only refer to <strong>the</strong> interfaces <strong>of</strong><br />

syntax with <strong>the</strong> external systems, not to <strong>the</strong> internal or external requirements <strong>of</strong> <strong>the</strong><br />

latter. ℜ only sees whe<strong>the</strong>r <strong>the</strong> output <strong>of</strong> syntax meets <strong>the</strong> input requirements <strong>of</strong><br />

<strong>the</strong> external systems, whe<strong>the</strong>r it is legible to <strong>the</strong>m: Full Interpretation. A minimal<br />

requirement for Full Interpretation is <strong>the</strong> absence <strong>of</strong> illegible <strong>features</strong>, like unvalued<br />

[Case:]. Problems that occur when PF attempts to realize <strong>the</strong> output, <strong>and</strong> LF<br />

attempts to interpret it, or o<strong>the</strong>r cognitive systems attempt to use <strong>the</strong> result, are invisible.<br />

However, Full Interpretation need not reduce to <strong>the</strong> absence <strong>of</strong> interpretable<br />

<strong>features</strong>. O<strong>the</strong>r legibility requirements have been proposed <strong>and</strong> may cover<br />

(279)-(281) (see section 5.9): <strong>the</strong>matic interpretability (Chomsky 1995: 347, cf.<br />

2000a: 111f.); linear order (Chomsky 1995: 221); <strong>the</strong> need to receive both QR <strong>and</strong><br />

non-QR structures if <strong>the</strong> two are not logically equivalent (Reinhart 2006: 1.3, 2.7).<br />

Taken toge<strong>the</strong>r, <strong>the</strong>se restrictions make <strong>of</strong> ℜ an interface algorithm outside<br />

syntax <strong>and</strong> lead to <strong>the</strong> formulation in (38), repeated below:<br />

(38) ℜ: An uninterpretable feature may enter <strong>the</strong> numeration only if needed<br />

for Full Interpretation <strong>of</strong> <strong>the</strong> syntactic structure built from it.<br />

ℜ takes <strong>the</strong> numeration <strong>and</strong> <strong>the</strong> structure built from it as input to PF <strong>and</strong> LF,<br />

<strong>and</strong> adds a feature to <strong>the</strong> former if need for Full Intepretation <strong>of</strong> <strong>the</strong> latter. The<br />

syntactic computation <strong>and</strong> <strong>the</strong> external systems are impenetrable to ℜ. All <strong>the</strong> information<br />

<strong>and</strong> effects <strong>of</strong> ℜ are localized in <strong>the</strong> interfaces <strong>of</strong> syntax with <strong>the</strong> lexicon,<br />

LF, PF. The appearance <strong>of</strong> globality or look-ahead <strong>of</strong> <strong>the</strong> PCC repairs or successive-cyclic<br />

wh-movement is deceptive. Syntax itself is not global. This<br />

construal <strong>of</strong> ℜ as an interface algorithm fits Reinhart's (2006) view <strong>of</strong> global<br />

mechanisms as last-resort repairs to meet interface requirements (section 5.9).<br />

So far undiscussed is <strong>the</strong> proposal that ℜ affects <strong>the</strong> numeration, as in (280),<br />

ra<strong>the</strong>r than edge <strong>of</strong> <strong>the</strong> syntactic structure built from it, (281) <strong>and</strong> (282). Both<br />

might be accommodated by <strong>the</strong> modularity <strong>of</strong> syntax, for <strong>the</strong> computation from<br />

<strong>the</strong> numeration is not affected in ei<strong>the</strong>r case. Empirically, <strong>the</strong> two options have<br />

different effects, but only under certain assumptions. The addition <strong>of</strong> a feature to<br />

<strong>the</strong> numeration may lead to <strong>the</strong> construction <strong>of</strong> a radically different structure than<br />

would be possible o<strong>the</strong>rwise, despite <strong>the</strong> two having <strong>the</strong> same interpretable content.<br />

In PCC repairs, it will be manifest in <strong>the</strong> creation <strong>of</strong> a new Agree/Case domain<br />

out <strong>of</strong> a part <strong>of</strong> <strong>the</strong> numeration, containing a part <strong>of</strong> its content that would<br />

o<strong>the</strong>rwise block an Agree/Case relationship for <strong>the</strong> rest. This may occur in <strong>the</strong> interior<br />

<strong>of</strong> <strong>the</strong> structure built up from <strong>the</strong> numeration, not only at it edge. Thus ℜ<br />

opens up a new derivational path for <strong>the</strong> numeration. We will see that ℜ could be<br />

made to operate in <strong>the</strong> manner <strong>of</strong> (281)/(282) at <strong>the</strong> edge <strong>of</strong> already built-up structures,<br />

at <strong>the</strong> cost <strong>of</strong> a certain counter-cyclicity, perhaps allowing elimination <strong>of</strong> its<br />

reference to <strong>the</strong> numeration. If ℜ does operate on <strong>the</strong> numeration, cyclicity is

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!