Upload
lyhanh
View
219
Download
3
Embed Size (px)
Citation preview
1
Domain mismatches: PF vs. LF, X vs. XP McSIRG: 28 January 2010
Eva Dobler, Mina Sugimura, Lisa Travis I. INTRODUCTION
First, PF vs. LF with X • Mini‐background: Marantz • A conceptual issue: how does head movement affect Marantz’s claims • Data: Tagalog reduplication….
Second, X vs. XP
• Mini‐background: Bobaljik and Wurmbrand • Conceptual issue: why do agreement domains look different • Data: Restructuring and scope in Japanese
II. HEAD MOVEMENT A. ISSUE: Domains within words: Seen in phonology and semantics Within domain Phonology (more destructive?) (v.s. across domains) Semantics (more idiosyncratic) ((Morphology (more idiosyncratic))) (1) BANTU (Simango 1999:72-73) -tsa
Base form Lexical causative Productive causative tuluk-a come out tulutsa bring (sth.)out tuluk-itsa make come out uluk-a fly ulutsa fly (sth.) uluk-itsa make fly kwer-a climb kweza hoist (sth.) kwer-etsa make climb vulal-a be injured vulaza injure (so.) vulal-itsa cause to be injured
other examples: Chichewa statives vs. passives Malayalam lexical causatives vs. productive
B. Marantz: two domains NOT lexical vs syntactic NOT derivational vs. inflectional Marantz (Words in Phases): Chomsky’s suggestive clarification of the cyclic nature of syntactic computation in “Derivation by Phase” (Chomsky 2001) invites a new investigation of locality domains for semantic and phonological processes within a generative grammatical framework. (bold is ours)
2
Determined by syntax and phases Within phase vs. across phases ‘Inner’ and ‘Outer’ determined by position w.r.t. category determining head (2)
(3)
QUESTION: Does PF domain always match the LF domain?? If they must match, then must be assuming that head movement cannot bleed phasal spell‐out C. Theoretical issues PF and LF mismatches theoretically possible Step 1: PF: Can have pre‐spell out head‐movement
3
Tagalog aspectual reduplication The reduplicative morpheme in Tagalog (CVV) may target a number of different positions within a complex verb, with a few exceptions; these variations derive no semantic effects.
(4) Base form: ma- ka- pag- pa- hintay
ABILITY- COMPLETE- CAUSE- E- wait
a. ma-[kaa]-ka-pag-pa-hintay
‘will be able to cause someone to wait’ (unrealized aspect)
b. ma-ka-[paa]-pag-pa-hintay
c. ma-ka-pag-[paa]-pa-hintay
d. ma-ka-pag-pa-[hii]-hintay
e. *[maa]-ma-ka-pag-pa-hintay
f. *ma-ka-pag-pa-hin-[taa]-tay
Rackowski (1999) observes that Tagalog aspectual reduplication is limited to morpho‐syntactic heads that are generated within the lexical domain (e.g. causative and transitivity markers, and verb roots, located in or below vP) and may not target heads in the functional domain (e.g. tense and modals, located above vP). Anything that is generated above the topmost vP is outside the scope of reduplication. Skinner argues that the narrow syntactic position of the reduplicant is fixed because otherwise the variation should have an impact on its semantic interpretation (scope). Hence, the surface variations must be due to operations occurring after Spell‐out Lowering. He proposes that the reduplicative morpheme is base‐generated in the functional domain as Outer Aspect, given that it denotes imperfectivity, and that the observed surface patterns are derived via head‐to‐head Lowering into the lexical domain.
4
(5) Structure AFTER V‐to‐v movement but BEFORE Lowering:
Importantly, Skinner proposes that all head movement is subject to the HAC: (6) Head Adjunction Condition (HAC)
Head movement adjoins a moving head a to any X0‐level projection of its targeted landing site b.
(7) Lowering:
A head X0 may lower to any zero‐level projection of the (potentially complex) head of its complement Y0 after Spell‐out, but before Vocabulary Insertion.
(8)
5
(9) ma‐ka‐[paa]‐pag‐pa‐hintay This analysis requires that the productive causative is monophasal – it only works if all of the morphemes of the complex v‐head are spelled out in the same phase. Yet it still has the hallmarks of a productive causative: no lexical idiosyncrasy. That is, LF still seems to interpret causative morpheme and root in different domains. (problem: do we ever see ‘destructive’ phonology across more than one morpheme??) Step 2: LF: always interpret low (even when pronounced high) Evidence: Head movement never affects semantics (interpreted low) VP ellipsis in Hebrew (Goldberg)
• Verb left behind with VP ellipsis o V moves out of VP before deletion
• BUT Vs must still match o Deleted material must match at LF (low copy counts)
6
Q: Tazmini et Dvora la-mesiba?
invite[Fut2Fsg] ACC Dvora to.the-party
'(Will) (you) invite Dvora to the party?'
A: Kvar hizmanti.
already invite[Past1sg]
(I) already invited [Dvora to the party].'
Q: Rivka hisi'a otax le-beit ha-sefer?
Rivka drive[Past3Fsg] ACC.you[Fsg] to-house the-book
(Did) Rivka drive you to school?'
A: (Ken,) hi *hevi'a / √hisi'a.
Yes she bring[Past3Fsg] drive[Past3Fsg]
‘(Yes,) she *brought/_drove [me to school].'
THEREFORE: Can get a ‘superphase’ at PF but not at LF Head movement: e.g. productive causative Without pre‐spell‐out head‐movement With pre‐spell‐out head‐movement
2 III. DOMAIN MISMATCHES BETWEEN XP MOVEMENT AND AGREE
Japanese productive causative shows a rather ‘wide’ domain for Case‐agreement. A. ISSUE:
Agreement across ‘domains’ (the structure from Harley 2006)
7
(10) 3
vP F 3
CAUSER v’ 3
vP v CAUSATIVE 3 –sase (
CAUSEE 3 √P(VP) v BECOME
3 3 XPOBJ √(V)
We will see cases where you get Agreement across these domains… Question:
What is a domain for Agreement? How does it correlate with other domains? B. BACKGROUND: CASEAGREEMENT in JAPANESE
Non‐stative verbs assign accusative case to their objects, whereas stative verbs assign nominative case (Kuno 1973):
(11) a. * Emi‐ga nihongo‐o deki‐ru
Emi‐NOM Japanese‐ACC be.capable‐PRES ‘Emi speaks Japanese’
b. Emi‐ga nihongo‐ga dekiru Emi‐NOM Japanese‐NOM be.capable‐PRES ‘Emi speaks Japanese’
Case assignment is controlled by the matrix predicate in restructuring environments (Wurmbrand 2001). In other words, if the higher predicate is a stative verb, it licenses nominative case on the object in the embedded clause, even if the embedded verb is not stative (e.g. Koizumi 1995)
8
(12)
Object can also be assigned ACC in (12b) (13) Emi‐ga ringo‐o tabe‐rare‐ru.
Emi‐NOM apple‐ACC eat‐can‐PRES ‘Emi can eat apples’
For Wurmbrand, (13) is a case of non‐restructuring, where there is an independent case‐assigner (i.e. v) in the embedded clause. We’ll come back to that point later.
C. CORRELATION BETWEEN CASEASSIGNMENT AND SCOPEFREEZING
EFFECTS: BOBALJIK & WURMBRAND (2005) • A nominative object must take scope in the higher domain when the object
agrees with the matrix predicate. (14)
??only >> can
(Tada 1992) NOTE: I disagree with the data….and allow high scope reading of the object in (14a). In fact, B & W note that the scope facts are subject to variations among speakers. According to Nomura, the low‐scope reading of nominative objects is possible, contrary to what has been assumed in the literature.
9
But let’s assume for now that the scope facts in (14) hold… How do B & W explain the correlation between Case and Interpretation?
10
(15) Structures for restructuring from Wurmbrand (2001)
TP 3
NOM T’ Johnga 3
vP T 3 ‐ru
NOM v’ ‐ga 3
tSUBJ v’ 3
VP v 3 [+stative]
V’ 3
VP V 3 re ‘can’
OBJ V migimedake tubu
‘right eye‐only’ ‘close’ B & W’s account:
Nom‐case is assigned by v by Movement. • The object cannot reconstruct to get low scope because Case‐assignment must
be visible at LF (i.e. AGREE must be evaluated at LF). • Case‐assignment cannot be implemented by AGREE when the agreement goes
across an agreement domain. (16) The induced domain generalization (B & W 2005:20) The (verbal) complement to a lexical verb delineates an agreement domain.
Thus, agreement across a domain that is the complement of a lexical verb (i.e. the lowest VP in (15)) is not permitted.
However, this locality constraint only constrains AGREE, and not MOVE, contrary to standard views where domains for AGREE and MOVE coincide.
There is a domain mismatch between AGREE and MOVE (17) Movement across an (agreement) domain for XP
3
XPOBJ 3
VP L 3
3 XPOBJ V Boundary for AGREE
11
Thus, if AGREE fails to happen, OBJ must move to get Case. This A‐movement is not successive cyclic: i.e. movement through an intermediate position (e.g. VP Spec) is not allowed.
…“A‐movement is subject to a feature‐checking requirement at the point in the derivation at which it occurs” (B& W 2005: 839)
• In (14a), since B & W assume that ‐rare ‘can’ is a lexical restructuring predicate, the above scope facts follow the OBJ must move outside of the embedded clause.
However, we assume –rare is a functional restructuring predicate, contra Tada (1992).
D. LEXICAL RESTRUCTURING VERB ik ‘go’ AND ITS AGREEMENT DOMAIN (18) Movement across an (agreement) domain for XP
3 XPOBJ
3 VP L
3 3
XPOBJ V Boundary for AGREE
(19) Lexical Restructuring Predicate ik ‘go’ (20) Taroo‐ga [Kanda‐ni hon‐o kai‐ni] itta ‐NOM Kanda‐to book‐ACC buy went ‘Taro went to Kanda by bicycle to buy a book’ (see Miyagawa 1987 for evidence for restructuring behaviours of ik)
12
(21) Structure from Miyagawa (1987)
VP 3 V’
3 vP V
3 ik ‘go’ VP v’
3 OBJ V Agreement domain hon kaini ‘book’ ‘buy’
We’ll come back to the validity of this structure later…
If ik creates an agreement domain for its complement clause, then the OBJ must move when it is assigned NOM case from the higher predicate.
Consequently, OBJ must take high scope.
(22) 3
OBJ 3 VP F
3 ‐rare ‘can’ V’
3 vP V
3 ik ‘go’ VP v’
3 tOBJ V Agreement domain kaini ‘buy’ (23) Midoriga sono cafeni aisudakega tabeni ike(re)ru. Midori‐NOM that café‐to ice cream–only‐NOM eat go‐can‐PRES ‘Midori can only go to that café to eat an ice cream.’ *can > only; only > can
13
BACK TO THE MAIN ISSUE: JAPANESE PRODUCTIVE CAUSATIVE
Case Agreement across domains (24) FP
3 vP F
3 -rare CAUSER v’ ‘can’
3 vP v CAUSATIVE
3 –sase ( CAUSEE 3
√P(VP) v BECOME 3
3 XPOBJ
What are the domains here? Two vPs? The scope facts below tell us that AGREE is possible across these domains. (25) Tarooga Hanakoni keekidakega tabesase(ra)reru.
Taroo‐NOM Hanako‐DAT cake–only‐NOM eat‐CAUSE‐can‐PRES. ‘Taroo can make Hanako eat only a piece of cake’
can > only; ??only > can (Native speakers? Possibilities of focus scramblin
If both –rare and ‐sase are both functional restructuring predicates (see Wurmbrand 2001 for the distinction between functional vs. lexical restructuring predicates), then there should be no independent agreement domains, according to B & W’s system.
Consequently, OBJ should be able to stay in‐situ, being assigned NOM Case (presumably by ‐rare) by AGREE.
Therefore, the object should be able to (or must?) take low scope. QUESTIONS:
So far, we have established the generalizations that domains for MOVE are looser than AGREE.
Can AGREE happen across a phase? How do we derive agreement domains for ik ‘go’ but not for sase? In order to answer these questions, we must look at the syntax of –ni‐clause in ik constructions…
14
E. RESTRUCTURING VS. NONRESTRUCTURING Accusative Case assignment is possible in both ‐sase and ‐ik examples above: (26) Midoriga sono cafeni aisudakeo tabeni ike(re)ru.
Midori‐NOM that café‐to ice cream–only‐ACC eat go‐can‐PRES ‘Midori can only go to that café to eat an ice cream.’ can > only; ??only > can
(27) Tarooga Hanakoni keekidakeo tabesase(ra)reru.
Taroo‐NOM Hanako‐DAT cake–only‐ACC eat‐CAUSE‐can‐PRES. ‘Taroo can make Hanako eat only a piece of cake’ can > only;?? only > can
Assuming with Wurmbrand (2001), we analyze (26) and (27) as cases of non‐restructuring where a ACC case assigner is inside the lower domain.
Thus, in restructuring –ik, there is no extra vP layer, and there must be an only bare infinitive.
F. FURTHER ISSUES: LOWSCOPE READING OF NOMOBJECT
Under B & W’s system, A‐movement is not successive‐cyclic. Movement to 2 or 3 does not happen since such movement does not lead to feature‐checking requirement at the point that happens.
However, as B & W points out, it is theoretically possible that movement independent of Case‐agreement proceeds through intermediate positions, and from there, an object should be able to get Case from a higher predicate by AGREE without moving into the matrix clause.
Scrambling of objects as an independent movement operation (28) Midoriga aisudakega [sono cafeni t tabeni] ike(re)ru. Midori‐NOM ice cream–only‐NOM that café‐to eat go‐can‐PRES ‘Midori can only go to that café to eat an ice cream.’ *can > only; only > can
Low scope of the object is impossible (?). What does this tell us? THM: Further refinement of the notion of AGREE is needed to understand ‘domains’.
15
REFERENCES Bobaljik, J & S. Wurmbrand. 2005. The domain of agreement. NLLT 23: 809–865. Harley, H. 2006. On the causative construction. In S. Miyagawa and M. Saito ed., The
Handbook of Japanese Linguistics. Oxford, OUP. Koizumi, Masatoshi. 1995. Phrase Structure in Minimalist Syntax, unpublished
Ph.D. dissertation, MIT, Cambridge, MA Kuno, S. 1973. The structure of the Japanese language. MIT Press, Cambridge, Mass. Miyagawa, S. 1987. Restructuring in Japanese. In Issues in Japanese linguistics pp.
273‐300. , Foris Publications, Dordrecht. Tada, H. 1992. Nominative objects in Japanese. Journal of Japanese Linguistics 14:
91‐108 Wurmbrand, S. 2001. Infinitives: Restructuring and clause structure. Mouton de
Gruyter, Berlin/New York.