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.
46<br />
bizu 2002). In Albondiga (66)b, <strong>the</strong> impoverishment <strong>of</strong> <strong>the</strong> phi-<strong>features</strong> <strong>of</strong> <strong>the</strong> ergative<br />
agreement simply turns <strong>the</strong> exponent 2pE sue to 3sE ∅, so that ga1pA-itxpAu[+ERG]-sue2pE-spA<br />
becomes ga-itx-u-∅-s. In (66)d shown in (68)c on <strong>the</strong> o<strong>the</strong>r<br />
h<strong>and</strong>, <strong>the</strong> ergative phi-<strong>features</strong> are totally erased so that not even 3sE is present.<br />
This affects <strong>the</strong> choice <strong>of</strong> <strong>the</strong> auxiliary root, which is now <strong>the</strong> [-ERG] root <strong>of</strong> unaccusatives,<br />
with its distinctive allomorphy for absolutive phi-<strong>features</strong>. The phenomenon<br />
is superficially similar to syntactic detransitivization used in Basque for<br />
anticausatives, mediopassives, <strong>and</strong> reflexives, so that (68)c ikusi sara can mean<br />
'You see yourselves' (Hualde <strong>and</strong> Ortiz de Urbina 2003: 4.7, 4.9.1.3). However, no<br />
ergative argument is possible <strong>the</strong>n. When ikusi sara is created by opaque agreement<br />
in (68)c, <strong>the</strong> ergative argument guk remains, <strong>and</strong> remains ergative. Syntax<br />
<strong>and</strong> interpretation do not see <strong>the</strong> agreement impoverishment at all.<br />
These observations also testify to <strong>the</strong> local character <strong>of</strong> opaque agreement, <strong>the</strong><br />
same as for opaque cliticization. Only phi-<strong>features</strong> in <strong>the</strong> agreement complex are<br />
visible, not phi-<strong>features</strong> at a distance. There seems to be no analogue <strong>of</strong> opaque<br />
agreement that would reach out to affect <strong>the</strong> phi-<strong>features</strong> <strong>of</strong> remote independent<br />
pronouns like guk in (68), with or without those <strong>of</strong> agreement affixes. It would appear<br />
as <strong>the</strong> deletion or transfer <strong>of</strong> phi-<strong>features</strong>, like Basque (68) or Chukchi (67),<br />
or <strong>the</strong> opaque cliticization in (55), as arbitrary as <strong>the</strong>y are – but affecting <strong>the</strong> phi<strong>features</strong><br />
<strong>of</strong> pronouns separated by as large phrase-structure distances as agreement<br />
regularly spans. The absence <strong>of</strong> such phenomena bespeaks a significant limit on<br />
morphology, (69).<br />
(69) Morphological mechanisms are restricted to "small" domains: adjacent<br />
elements, nonadjacent ones in <strong>the</strong> extended word, etc.<br />
(69) characterizes many <strong>the</strong>ories <strong>of</strong> morphology, which restrict morphological<br />
mechanisms to small domains explicitly (Carstairs-McCarthy 1992: 3.3.1,<br />
Ackema <strong>and</strong> Neeleman 2003, Embick <strong>and</strong> Marantz 2008), or through <strong>the</strong> nature <strong>of</strong><br />
<strong>the</strong> information which <strong>the</strong>y manipulate (Embick <strong>and</strong> Noyer 2001). O<strong>the</strong>r <strong>the</strong>ories<br />
give realizational morphology <strong>the</strong> power to span phrase-structurally unbounded<br />
domains, some explicitly for agreement <strong>and</strong> case (Marantz 2000, Bobaljik 2008).<br />
Opaque agreement <strong>and</strong> cliticization must not be within <strong>the</strong>ir scope to derive (69). 22<br />
The generalizations accounted for by (69) are one <strong>of</strong> <strong>the</strong> core issues to be addressed<br />
if opaque agreement <strong>and</strong> cliticization belong to syntax ra<strong>the</strong>r than morphology.<br />
In order to model <strong>the</strong>m, syntax must be capable <strong>of</strong> selecting or moving<br />
subsets <strong>of</strong> phi-<strong>features</strong>: 3 rd person but not 1 st /2 nd person, 3 rd person plural but not<br />
22 Strict adjacency does not seem to be <strong>the</strong> right condition on morphological phi-feature manipulations<br />
or even allomorph selection; see Carstairs-McCarthy (1992: 3.1.3), Bonet (1991: 90, 114<br />
note 15, 173 note 29), Noyer (1992: 87), Stump (1998: 34), Ackema <strong>and</strong> Neeleman (2003), Bobaljik<br />
(2000: notes 17, 44). The iteration <strong>of</strong> local processes in morphology does not seem to<br />
break out <strong>of</strong> this small domain condition in general, Noyer (2001: 796), although perhaps such<br />
iteration is available, Harris <strong>and</strong> Halle (2005). If failure to iterate is derivable from <strong>the</strong> way <strong>the</strong><br />
syntactic tree is spelled out, cf. Bobaljik (2000), it might permit exceptions.