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.
184<br />
(282) Optional operations apply only if <strong>the</strong>y have an effect on outcome: in <strong>the</strong><br />
present case, v* may be assigned an EPP feature to permit successivecyclic<br />
A'-movement or Int [an interpretation assigned to <strong>the</strong> EPP position<br />
<strong>of</strong> v*] (under OS [Object Shift]).<br />
(Chomsky 2001: 34)<br />
These proposals go in <strong>the</strong> direction <strong>of</strong> what is needed for last-resort Case licensing,<br />
for <strong>the</strong>y are used for similar last-resort addition <strong>of</strong> uninterpretable <strong>features</strong><br />
as probes. The initial formulation <strong>of</strong> a mechanism to subsume both <strong>the</strong>ir uses<br />
<strong>and</strong> <strong>the</strong> PCC repair is (283):<br />
(283) ℜ (preliminary formulation): An uninterpretable feature may enter <strong>the</strong><br />
numeration only if it has an effect on output.<br />
The proposal limits ℜ to affecting <strong>the</strong> numeration <strong>and</strong> to adding uninterpretable<br />
<strong>features</strong>. On <strong>the</strong> first point, ℜ, along with (280)-(282), differs from <strong>the</strong><br />
earlier global mechanisms <strong>of</strong> Chomsky (1995: chapters 2, 3) such as Shortest<br />
Move in not affecting <strong>the</strong> syntactic computation itself. The second point derives<br />
<strong>the</strong> conservativity <strong>of</strong> PCC repairs. ℜ can activate a potential Agree/Case locus<br />
present in a structure by adding a phi-probe to it, turning T to TERG, v to vACC, PDAT<br />
to Pfull. It cannot however add or delete <strong>the</strong> interpretable content <strong>of</strong> a structure,<br />
unlike (280): add a new potential Agree/Case locus such as a second T, enrich or<br />
replace PDAT with P pour 'for'. Thus ℜ can only activate <strong>the</strong> syntactic dependencies<br />
already potential in a structure. One might entertain principled departures<br />
from this position: <strong>the</strong> addition <strong>of</strong> an expletive-like syntactic terminal to host <strong>the</strong><br />
added uninterpretable feature, or even <strong>of</strong> a contentful terminal if UG principles<br />
specify a pairing between uninterpretable <strong>and</strong> interpretable <strong>features</strong>, for instance<br />
between uninterpretable [wh:] <strong>and</strong> interpretable [CQ] in questions (cf. Chomsky<br />
1995: 4.2.2, 2000a: 128). It seems unnecessary, although <strong>the</strong> issue will reemerge.<br />
These two restrictions are imposed on ℜ by modularity. The status <strong>of</strong> syntax<br />
as a module restricts ℜ to affecting its numeration. ℜ cannot see inside <strong>the</strong> syntactic<br />
computation to change its operations or <strong>the</strong>ir outputs. The modular status <strong>of</strong> <strong>the</strong><br />
lexicon may be responsible for <strong>the</strong> restriction <strong>of</strong> ℜ to uninterpretable <strong>features</strong>. The<br />
lexicon fixes <strong>the</strong> groupings <strong>of</strong> interpretive, realizational, <strong>and</strong> uninterpretable <strong>features</strong><br />
drawn from UG primitives at acquisition. The numeration is its interface<br />
with syntax, selecting items at <strong>the</strong> start <strong>of</strong> <strong>the</strong> syntactic computation. Thereafter ℜ<br />
can affect <strong>the</strong> numeration, but <strong>the</strong> lexicon is a closed module where ℜ cannot seek<br />
for o<strong>the</strong>r lexical items. Uninterpretable <strong>features</strong> can be accessed from UG inventory.<br />
O<strong>the</strong>r <strong>features</strong> are inaccessible because <strong>the</strong>ir addition would dem<strong>and</strong> <strong>the</strong> construction<br />
<strong>of</strong> new lexical items, a procedure not available on-line after acquisition.<br />
The most evident effect <strong>of</strong> modularity is one that has been omitted from <strong>the</strong><br />
preliminary formulation. <strong>Modular</strong>ity greatly constrains <strong>the</strong> "effect on output" that<br />
permits ℜ to apply. In (280)-(282), <strong>the</strong> "effect on output" may appear transmodular,<br />
<strong>and</strong> that should be a cause for concern. Chapter 2 has argued for phonology-<br />
<strong>and</strong> morphology-free syntax, which cannot respond to PF requirements even if