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.
54<br />
2.4 The limits <strong>of</strong> a modular signature<br />
From this survey emerges a strong modular architecture in which syntax <strong>and</strong><br />
morphology each have <strong>the</strong>ir distinctive informational <strong>and</strong> computational properties,<br />
<strong>and</strong> morphology is invisible to syntax <strong>and</strong> interpretation. By <strong>the</strong> clustering <strong>of</strong><br />
<strong>the</strong>se properties in <strong>the</strong> modular signatures (41), linguistic phenomena are individuated<br />
as morphological or syntactic. The counterpart <strong>of</strong> <strong>the</strong> morphological signature<br />
<strong>of</strong> this chapter is <strong>the</strong> syntactic one <strong>of</strong> those that follow, <strong>the</strong> more striking<br />
because it characterizes phenomena superficially similar to <strong>the</strong> ones seen here.<br />
The strength <strong>of</strong> <strong>the</strong> conclusions reached in this chapter must be tempered in two<br />
ways: <strong>the</strong> phenomena left aside <strong>and</strong> <strong>the</strong> nature <strong>of</strong> <strong>the</strong> evidence. DP-internal concord<br />
illustrates <strong>the</strong> first caveat. It has not been investigated here, <strong>and</strong> so might be a<br />
domain where morphology transmits phi-<strong>features</strong> a 'moderate' distance (cf. <strong>the</strong><br />
remarks on adjacency in Wechsler <strong>and</strong> Zlatić 2000: 819). The French gens 'people'<br />
illustrates a possible state <strong>of</strong> affairs (Grevisse <strong>and</strong> Goosse 2008: §490). It is masculine,<br />
including for adjectival concord, pleins in (80)a, save for preceding determiners<br />
<strong>and</strong> modifiers if adjacent: vieilles <strong>and</strong> bonnes in (80)a, toutes <strong>and</strong> petites in<br />
(80)b, bonnes in (80)c but not in (80)d where et '<strong>and</strong>' intervenes. This would <strong>the</strong>n<br />
be transmission <strong>of</strong> phi-<strong>features</strong> sensitive to linearity <strong>and</strong> adjacency, <strong>and</strong> thus in <strong>the</strong><br />
morphology, yet iterative or unbounded. However, beyond a few set collocations,<br />
speakers ei<strong>the</strong>r treat gens as masculine or lack judgments. This phenomenon <strong>the</strong>refore<br />
looks like <strong>the</strong> prescriptive 'viruses' <strong>of</strong> Sobin (1997), Lasnik <strong>and</strong> Sobin (2000),<br />
or like idiomatic exceptions that do not refer to phi-<strong>features</strong> at all.<br />
(80) a. de vieilles bonnes gens pleins de saveur antique et frustré<br />
old.PLF good.PLF people full.PLM <strong>of</strong> taste antique <strong>and</strong> frustrated<br />
b. toutes les petites gens<br />
all.PLF <strong>the</strong>.PL little.PLF people<br />
c. Quelles honnêtes et bonnes gens!<br />
What.PL[F] honest.PL <strong>and</strong> good.PLF people<br />
d. Quels bons et honnêtes gens!<br />
What.PL[M] good.PLM <strong>and</strong> honest.PL people<br />
(Grevisse <strong>and</strong> Goosse 2008: §490; [] agreement purely orthographic)<br />
quently displace. However, it would seem that at <strong>the</strong> Vedic stage where Sanskrit was a first language,<br />
<strong>the</strong> analytic perfect is too rare conclude anything; after <strong>the</strong> grammarian Panini's time<br />
when <strong>the</strong> foregoing dissociated analytic perfects occur, Sanskrit was only a second language; <strong>and</strong><br />
Panini describes <strong>the</strong> analytic perfect as a morphological alternative to <strong>the</strong> syn<strong>the</strong>tic with its parts<br />
inseparable (Staal 1967: 30; for Panini's Sanskrit, Cardona 1999: 2.1.4.8). (Kiparsky also discusses<br />
a couple <strong>of</strong> roots that allow <strong>the</strong> analytic perfect because <strong>the</strong> syn<strong>the</strong>tic one has an irregular<br />
meaning, like veda 'knows < has seen'; but <strong>the</strong>se are treated by Whitney (op.cit.) under (iii) as<br />
roots that need a diacritic specifying <strong>the</strong> availability <strong>of</strong> <strong>the</strong> analytic perfect, also needed for (ii)<br />
because roots with <strong>the</strong> same shape may behave differently.)