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.
205<br />
with it (Rezac 2006: 3.7, 2008b: 722). Nei<strong>the</strong>r incorporation nor number/gender<br />
Agree provide [+person] licensing, so S/O cannot be [+person] in Mohawk or Nahuatl.<br />
Thus <strong>the</strong> applicative IO thus creates <strong>the</strong> PCC in <strong>the</strong> most straightforward incarnation<br />
<strong>of</strong> <strong>the</strong> Agree/Case approach. As a plain DP it pre-empts clausal Agree,<br />
<strong>and</strong> <strong>the</strong>re is no o<strong>the</strong>r means to license [+person] S/O.<br />
(318) a. TheyEA baked herIO two cakesO<br />
b. SheIO was baked tIO two cakesO<br />
c. *Two cakesS were baked herIO tS.<br />
(319) Ni-mitz-im-maca in huē-hue'xōlo-'.<br />
1SG.SU-2SG.O-3PL.O2-give IN RED-turkey-PL<br />
IEA give youIO <strong>the</strong> turkeysO.<br />
(Nahuatl, Baker 1996: 240 note 12, from Launey 1981: 174)<br />
Chinook seems to belong to this group. Its IO behaves exactly like <strong>the</strong> S/O <strong>of</strong><br />
nonapplicative unaccusatives/transitives for agreement <strong>and</strong> case. The remaining<br />
S/O <strong>of</strong> applicative constructions agrees for number/gender, but is ruled out as<br />
[+person]. Systems <strong>of</strong> this type should raise <strong>the</strong> applicative IO to satisfy <strong>the</strong> EPP<br />
<strong>of</strong> T, since it is closest to T <strong>and</strong> not distinguishable by being dative. This is so in<br />
English <strong>and</strong> apparently in Mohawk. Chinook would <strong>the</strong>n have <strong>the</strong> structure in<br />
(317)b. In Basque (317)a, S is in <strong>the</strong> range <strong>of</strong> T to receive last-resort ergative, but<br />
in Chinook (317)b, <strong>the</strong> applicative IO is. The mechanics are as in Basque, save<br />
that it is <strong>the</strong> IO that must escape <strong>the</strong> vP. The only convergent derivation is one<br />
where it does so prior to (person) Agree with vABS, which remains available for<br />
S. 152<br />
The discussion <strong>of</strong> Chinook may end on a sketch <strong>of</strong> <strong>the</strong> third asymmetric pattern,<br />
because it sheds fur<strong>the</strong>r light on <strong>the</strong> ergative as dependent Case. As in <strong>the</strong><br />
second one, IO is not oblique <strong>and</strong> participates in <strong>the</strong> Agree/Case system <strong>of</strong> <strong>the</strong><br />
core arguments. However, applicative unaccusatives are always treated as transitives,<br />
not just when S is [+person]. The IO is ergative, or <strong>the</strong> O accusative. Baker<br />
(1996: 9.6.2) discusses this system in Ainu <strong>and</strong> Mayali, but <strong>the</strong>ir case <strong>and</strong> agreement<br />
do not indicate whe<strong>the</strong>r transitives are ergative or accusative. O<strong>the</strong>r languages<br />
do, but do not make for minimal comparison with Chinook. The Chukchi<br />
alternation in (320) comes close. In (320)a is an unaccusative with an absolutive<br />
experiencer. In (320)b an absolutive <strong>the</strong>me is added. As a result, <strong>the</strong> experiencer<br />
switches to <strong>the</strong> ergative, independent <strong>of</strong> <strong>the</strong> person value <strong>of</strong> ei<strong>the</strong>r argument.<br />
(320) a. ətɬəɣən (pəčɣ-etə) korɣav-ərkən<br />
152 Some Basque varieties this <strong>and</strong> <strong>the</strong> foregoing types: a dative controls absolutive agreement,<br />
<strong>and</strong> an extra number-only probe appears for <strong>the</strong> remaining S/O, with <strong>the</strong> PCC: see Rezac (2006,<br />
2008a, 2008b) for Basque <strong>and</strong> cross-linguistically. S still raises to satisfy <strong>the</strong> EPP <strong>and</strong> ergativizes,<br />
(Rezac 2008c: 80 note 13, 85 note 14, forthc). This continues to follow from whatever<br />
property renders specifically datives unavailable to satisfy <strong>the</strong> EPP in Basque.