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.

266<br />

purely syntactic phi-<strong>features</strong>. If <strong>the</strong>re is more to be said, it is about <strong>the</strong> interface <strong>of</strong><br />

syntax <strong>and</strong> PF/LF. One might speculate, for instance, that Agree/Case relations<br />

permit <strong>the</strong> 3SG phi-<strong>features</strong> <strong>of</strong> 1PL on to be deleted, or on to be linked to 1PL phi<strong>features</strong><br />

in <strong>the</strong> clause, as suggested at <strong>the</strong> end <strong>of</strong> section 6.3.<br />

No light has been shed on <strong>the</strong> potential autonomy <strong>of</strong> interpretive from syntactic<br />

phi-<strong>features</strong>. The 1PL <strong>of</strong> on might be absent from syntax, but it might be<br />

needed <strong>the</strong>re, <strong>and</strong> its invisibility to certain mechanisms can be straightforwardly<br />

modelled by making its 3SG more prominent. Similar devices can ensure that <strong>the</strong><br />

masculine <strong>of</strong> French laideron 'ugly woman' is visible only for concord.<br />

There is no overstating <strong>the</strong> tentativeness <strong>of</strong> <strong>the</strong>se conclusions. The inertness<br />

<strong>of</strong> <strong>the</strong> 3SG <strong>of</strong> 1PL on or <strong>the</strong> masculine <strong>of</strong> laideron for interpretive dependencies is<br />

striking. Yet no more so than Chomsky's (1975) argument from (411)a that<br />

syntactic plural marking can be autonomous <strong>of</strong> interpretive entailments. Such 'dependent'<br />

plurals admit <strong>of</strong> a different treatment, unspecified for number, <strong>and</strong> <strong>the</strong> inferences<br />

in (411) depend on world knowledge, not plural semantics (Zweit 2008,<br />

Seuren 2005). It has also been seen in <strong>the</strong> last section that <strong>the</strong>re may be interpretive<br />

coercions that affect syntactic phi-specifications, in perhaps only apparent<br />

contrast to autonomy. Whatever future inquiry brings, phi-mismatches are likely<br />

to prove fruitful in exploring this central ground <strong>of</strong> <strong>the</strong> relationship <strong>of</strong> syntax <strong>and</strong><br />

interpretation (cf. Chomsky 1975, 1977, 2000b, Higginbotham 1985, 1991,<br />

McGilvray 1998, Fox 2000, Jackend<strong>of</strong>f 2002, Pietroski 2005, Hinzen 2006).<br />

(411) a. Unicycles have wheels. (→only one wheel per unicycle)<br />

b. Hotel rooms have bathrooms. (→at least one bathroom per room)<br />

c. Students have supervisors. (→need not be one supervisor per student)<br />

((a) Chomsky 1975, (b, c) Seuren 2005)<br />

The autonomy <strong>of</strong> syntactic phi-<strong>features</strong> from interpretation corresponds to <strong>the</strong><br />

more robust conclusions drawn in chapter 3 for syntax <strong>and</strong> realization. Syntax <strong>and</strong><br />

realizational morphology both access <strong>and</strong> manipulate phi-<strong>features</strong>, but as separate<br />

modules distinct in information <strong>and</strong> computation, <strong>and</strong> limited in interaction. Theories<br />

<strong>of</strong> modularity should derive this as a matter <strong>of</strong> principle. The classical Y/T<br />

model fits well, as do proposals that only partially weaken it <strong>and</strong> those that<br />

streng<strong>the</strong>n it. <strong>Modular</strong>ity does not bar all sharing <strong>of</strong> information or mechanism<br />

<strong>and</strong> rich interactions between modules, nor limit <strong>the</strong> influence that realization<br />

might have on syntax in <strong>the</strong> choice <strong>of</strong> primitives <strong>and</strong> <strong>the</strong> content <strong>of</strong> lexical items<br />

during acquisition (Aron<strong>of</strong>f, Meir <strong>and</strong> S<strong>and</strong>ler 2005). On <strong>the</strong> interpretive side, all<br />

is murkier, but <strong>the</strong> same principles hold.<br />

A partial sharing <strong>of</strong> information types between modules is instantiated by phi<strong>features</strong><br />

<strong>the</strong>mselves, as a common alphabet across syntax, interpretation, <strong>and</strong> realization.<br />

A modular architecture requires a way to share information, <strong>and</strong> so it<br />

must relate alphabets across modules. However, it is surprising to find a common<br />

alphabet, <strong>and</strong> atypical <strong>of</strong> o<strong>the</strong>r primitives that map across <strong>the</strong> modules quite indirectly.<br />

Through phi-<strong>features</strong>, syntax, interpretation <strong>and</strong> realization operate <strong>and</strong> interact<br />

in commensurable ways. Here perhaps is a key to <strong>the</strong> central role <strong>of</strong> phi-

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

Saved successfully!

Ooh no, something went wrong!