On restructuring infinitives in Japanese: …ling.umd.edu/assets/publications/Takahashitoappear.pdfAdjunction, clausal architecture, and phases Masahiko Takahashi* University of Maryland,

+ ModelsLINGUA-1946; No. of Pages 27

Available online at www.sciencedirect.com

www.elsevier.com/locate/linguaLingua xxx (2012) xxx--xxx

On restructuring infinitives in Japanese:Adjunction, clausal architecture, and phases

Masahiko Takahashi *

University of Maryland, Department of Linguistics, 1401 Marie Mount Hall, College Park, MD 20742-7505, United States

Received 2 May 2010; received in revised form 5 August 2012; accepted 14 August 2012

Abstract

This paper investigates the syntax of Japanese restructuring verbs and makes two major claims: (i) there are (at least) three types ofrestructuring infinitives in Japanese, which is consistent with Wurmbrand's (2001) approach to restructuring infinitives and (ii) there is ageneral ban on adjunction to complements of lexical restructuring verbs, which is best explained by an interaction of spell-out domainsand Case-valuation. It is also shown that this ban regulates adverb insertion, adjective insertion, and quantifier raising.© 2012 Elsevier B.V. All rights reserved.

Keywords: Adjunction; Case; Japanese; Phases; Restructuring

1. Introduction

‘Restructuring’ (i.e. clause-downsizing) has been extensively discussed in the generative literature, with a variety ofapproaches proposed to capture the phenomenon. Thus, while Cinque (2006) argues that all ‘restructuring’ verbs arefunctional heads, researchers like Hoshi (2006) and Saito and Hoshi (1998), for example, argue that ‘restructuring’ involvescomplex predicate formation via direct merger of the verbs. Wurmbrand (2001) pursues yet another approach and arguesthat there are degrees of ‘restructuring’, which are determined by the size of infinitival complements (CP, TP, vP, VP, forexample), where the various sizes of infinitival complements correlate with various (non-) ‘restructuring’ phenomena.

One of the goals of this paper is to resolve this tension from the perspective of Japanese. There is considerable literatureon ‘restructuring’ in Japanese (see Asano, 2007; Bobaljik andWurmbrand, 2007; Hoshi, 2006; Kageyama, 1993; Koizumi,1994, 1995, 1998; Kuno, 1973; Matsumoto, 1996; Miyagawa, 1987; Nakatani, 2004; Nishigauchi, 1993; Nomura, 2003,2005; Saito andHoshi, 1998; Shibatani, 1978; Sugioka, 1984; Tada, 1992; Takezawa, 1987; Terada, 1990; Tomioka, 2006;Tsujimura, 1993; Ura, 1996, 1999, 2000; Yumoto, 2004; Zushi, 1995, 2008; Wurmbrand, 2001, among many others).However, to the best ofmy knowledge,most of the important paradigms in the context of restructuring have been addressedonly partially in this literature. Hence, previous studies on Japanese restructuring constructions have not drawn acomprehensive picture that should have emerged from the observed data. I take up this issue seriously and provide a morecomprehensive description of ‘restructuring’ constructions in Japanese. I show that Japanese data lead us to posit a three-way distinction in ‘restructuring’ configurations, which is broadly consistent withWurmbrand's (2001) proposals concerningrestructuring infinitives.

A theoretical concern of this paper is a restriction on adjunction found in restructuring contexts. I argue that there is aban on adjunction to complements of lexical verbs, which is derived through an interaction of the contextual emergence of

Please cite this article in press as: Takahashi, M., On restructuring infinitives in Japanese: Adjunction, clausalarchitecture, and phases. Lingua (2012), http://dx.doi.org/10.1016/j.lingua.2012.08.006

* Corresponding author. Tel.: +1 301 405 7002.E-mail addresses: [email protected], [email protected].

0024-3841/$ -- see front matter © 2012 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.lingua.2012.08.006

http://dx.doi.org/10.1016/j.lingua.2012.08.006

http://www.sciencedirect.com/science/journal/00243841


mailto:[email protected]

mailto:[email protected]


M. Takahashi / Lingua xxx (2012) xxx--xxx2


spell-out domains (see Bobaljik and Wurmbrand, 2005; Boskovic, 2010; den Dikken, 2007; Takahashi, 2010, 2011,among others) and obligatory late insertion of adjuncts within spell-out domains (cf. Stepanov, 2001). I also argue that theconstraint is a general constraint, which yields a unified account of the distribution of adverbs, adjectives, and quantifiers.

This paper is organized in the following way. In section 2, mainly based on the observations made in the literature (seeBobaljik andWurmbrand, 2007;Matsumoto, 1996;Miyagawa, 1987;Nakatani, 2004;Tsujimura,1993, amongmanyothers),I provide a detailed description of several restructuring constructions in Japanese and show that they do not exhibit uniformbehavior. In section 3 I provide an analysis of the generalization made in section 2 and propose that there is a ban onadjunction to certain restructuring infinitives. In section 4 I extend the analysis of the ban on adjunction to other ‘restructuring’constructions and show that the ban is actually a general constraint. In section 5 I critically discuss alternative approaches inthe literature and show that they face various empirical problems. Section 6 concludes the paper.

2. Restructuring infinitives in Japanese and adverbs

This section offers a detailed description of three types of restructuring constructions and shows that they are differentfrom each other regarding the distribution of adverbs. In particular, I discuss two types of restructuring motion verbconstructions and the potential construction. I start with examples involving two types of motion verb constructions,namely, the Purpose Expression (PE) construction, and the Sequential Expression (SE) construction1:

(1)

Pleaarch

1 SeMatsumconstrucompre

2 Tedefineof -te.

Taroo-ga

se cite thisitecture, an

e Matsumoto (oto (1996), Nctions have ahensive descis pronounced-te as a gerund

gakkoo-ni

article ind phases.

1996), Miyagakatani (200variety of inteription.as de whenive marker. I

sono

pressLingua

awa (194), Shibrpretat

the formwill not

hon-o

as: Takah(2012), htt

87), Tsujimuratani (2007)ions, which I d

er is precedediscuss the n

kai-ni

ashi, Mp://dx.d

a (1993)and Tsujio not disc

d by a veature of -

it-ta.

., On restoi.org/10.1

and Wurmbrmura (1993)uss here. Se

rb stem withte in this pap

(PE)
Taro-NOM school-to the book-ACC buy-NI go-PAST ‘Taro went to school to buy the book.’
(2)
Taroo-ga gakkoo-de sono hon-o yon-de it-ta.
ruc01

an, ae

aer.

(SE)
Taro-NOM school-at the book-ACC read-TE go-PAST ‘Taro read the book at school and went (somewhere).’
The infinitive in (1) is followed by -niwhile the one in (2) is followed by -te.2 As we will see below, both constructions involveoptional clause-union effects (i.e. restructuring). However, the two constructions show different syntactic behavior in otherrespects.

An indication of clause-union effects comes from nominative marking of objects (see Bobaljik and Wurmbrand, 2007;Koizumi, 1994, 1995, 1998; Kuno, 1973; Nomura, 2003, 2005; Saito and Hoshi, 1998; Tada, 1992; Takezawa, 1987; Ura,1996, 1999, 2000, among many others). Consider the following sentences:

(3)
Taroo-ga eego-o/ga hanas-e-ru. Taro-NOM English-ACC/NOM talk-can-PRES ‘Taro can speak English.’
(4)
Boku-ga Mary-ni Taroo-ga eego-o/*ga hanas-u-to
turing infin6/j.lingua.2

d (2001), amomong othersNakatani (20

voiced consoSee Nakatan

i-e-ru.
I-NOM Mary-DAT Taro-NOM English-ACC/NOM speak-PRES-that say-can-PRES ‘I can say to Mary that Taro speaks English.’
In (3), the object is marked nominative in the presence of the potential morpheme -e ‘can’. In (4), on the other hand, there isa clausal boundary between -e ‘can’ and the object and the object cannot be marked nominative. This shows thatnominative Case-licensing of objects is clause-bounded.

Both the PE construction and the SE construction can involve (optional) restructuring. This is supported by the fact thatthe embedded objects in these constructions can be nominative when the matrix verbs are accompanied by the potentialmorpheme (see Miyagawa, 1987; Tsujimura, 1993, among others). I assume that this (apparent) non-local Casedependency is an indication of restructuring following a number of researchers (see Bhatt, 2005; Bobaljik andWurmbrand, 2005, among others). This is further supported by typical distributional properties of restructuring, namely the

itives in Japanese: Adjunction, clausal012.08.006

ng others, for discussion of PEs, and Kuno (1973),, for discussion of SEs. Note that V-te-motion verb04) and Teramura (1984), among others, for a more

nant (Kuno, 1973). Kuno (1973) and Martin (1975)i (2004) and references cited therein for discussion


M. Takahashi / Lingua xxx (2012) xxx--xxx 3


requirement that restructuring infinitives must be adjacent to the matrix verbs (see Miyagawa, 1987; Tsujimura, 1993;Wurmbrand, 2007; Wurmbrand and Bobaljik, 2005, among others):

(5)

Pleaarch

3 Onassumfuture

4 Nomorphe

5 Noconsidcan be

a.

se citectu

e reviee thatreseartice thme iste thaterationinser

Boku-ga

ite this are, and p

wer finds (the nominatch, and focat (8) and (9an atelic pa clear prois a non-re

ted between

tosyokan-ni

rticle in preshases. Lingu

5b) with a nomive object in sucus on the judgm) contain threeredicate, the adsodic break betstructuring conthe two verbs

hon-o/ga

s as: Takahashi,a (2012), http://d

inative object acceph cases is analyzed aent given in the texpredicates, namely tverbs used here, wween the motion verbstruction (the prosod.

kari-ni

M., On rex.doi.org/10

table, especias amajor subt.he potential mhich only modand the emb

ic requiremen

ik-e-ru.

structuring infi.1016/j.lingua.

lly with a pitch aject (see Kuroda,

orpheme, the moify telic verbs, caedded verb is neet blocks the adjac

(PE)
I-NOM library-to book-ACC/NOM borrow-NI go-can-PRES ‘I can go to the library to borrow a book.’
b.
Boku-ga [hon-o/*ga kari-ni] tosyokan-ni ik-e-ru. I-NOM book-ACC/NOM borrow-NI library-to go-can-PRES ‘I can go to the library to borrow a book.’
(6)
a. Hanako-ga atode tosyokan-de zassi-o/ga kari-te
nitiv201

ccen1988

tionnnodedency

ik-e-ru.

es in Japanes2.08.006

t on the nominati), which I ignore in

verb, and the emt modify the potenin (9) to make surequirement). To

(SE)
Hanako-NOM later library-at magazine-ACC/NOM borrow-TE go-can-PRES ‘Hanako can borrow a magazine at the library and go (somewhere) later.’
b.
Hanako-ga [tosyokan-de zassi-o/*ga kari-te] atode ik-e-ru.
v

b

r

Hanako-NOM
library-at magazine-ACC/NOM borrow-TE later go-can-PRES ‘Hanako can borrow a magazine at the library and go (somewhere) later.’
When the infinitival clause is adjacent to the matrix verb, the embedded object can be marked nominative, as shown in(5a) and (6a). The clausal boundary between the matrix and the embedded clauses is transparent for nominative Case-licensing. On the other hand, in (5b) and (6b), the infinitival clause is not adjacent to the matrix verb and the embeddedobject cannot be nominative, which indicates that restructuring is not possible.3

In the remainder of this section, I will show based on the previous literature that the three restructuring constructionsdiffer regarding the distribution of adverbs. In particular, it will be shown that (i) the potential construction with a nominativeobject allows both matrix and embedded modification, (ii) restructuring PEs allow only matrix modification, and(iii) restructuring SEs allow only embedded modification.

Let us first consider the potential construction. This construction allows two durative adverbs.

(7)
Taroo-wa terebi-ga 1-nen-kan 3-zikan mi-rare-ru. Taro-TOP TV-NOM one-year-for three-hour watch-can-PRES ‘For one year, Taro can watch TV for 3 hours.’
(Bobaljik and Wurmbrand, 2007:33; slightly modified)

The intended interpretation of this sentence is the one in which the adverb 1-nen-kan ‘for one year’modifies the potentialmorphemewhile the adverb 3-zikan ‘(for) three hours’modifies the embedded verbmi ‘watch’. This example clearly showsthat the potential construction allows both matrix and embedded modification.

Let us now consider restructuring PEs. The following non-restructuring PE examples in (8) and (9) (cf. the accusativeCase on the object) show that non-restructuring PEs allow both matrix and embedded modification:

(8)
Hanako-wa [10-pun-de robusutaa-o tabe-ni] 3-zikan-de Bosuton-ni ik-e-ru. Hanako-TOP 10-minutes-in lobster-ACC eat-NI 3-hours-in Boston-to go-can-PRES ‘Hanako can go to Boston in 3 hours to eat a lobster in 10 minutes.’
(9)
(?)Hanako-wa 3-zikan-de Bosuton-ni [10-pun-de robusutaa-o tabe-ni] (kuruma-de)
e

et

etiaee

ik-e-ru.
Hanako-TOP 3-hours-in Boston-to 10-minutes-in lobster-ACC eat-NI car-by go-can-PRES
‘Hanako can go to Boston in 3 hours to eat a lobster in 10 minutes (by car).’

The adverb 10-pun-de ‘in ten minutes’ modifies the event of eating a lobster and the adverb 3-zikan-de ‘in 3 hours’modifies the event of going to Boston. Note that the interpretation is identical in the two examples, independently of theposition of the matrix modifiers.4,5

: Adjunction, clausal

Case-marker. I tentativelyhis paper, putting it aside for

dded verb. As the potentiall morpheme.that the construction undernforce this point, an adverb




The situation is different in a restructuring context. As shown in (10) and (11), where the object gets nominative Case,restructuring is impossible with two durative adverbs; only one adverb can appear (cf. (10)). Crucially, the adverb mustmodify the matrix predicate (cf. (11)).

(10) *

Pleasarchit

6 See

Hanako-wa

e cite this artecture, and ph

Matsumoto (1996

3-zikan-de

icle in presases. Lingu

), Nakatani (

Bosuton-ni

s as: Takaha (2012), h

2004), Shibata

10-pun-de

ashi, M., Onttp://dx.doi.org

ni (2007) and Tsu

robusutaa-ga

restructuring i/10.1016/j.lingu

jimura (1993), am

tabe-ni

nfinitivea.2012.

ong othe

ik-e-ru.
Hanako-TOP 3-hours-in Boston-to 10-minutes-in lobster-NOM eat-NI go-can-PRES ‘Hanako can go to Boston in 3 hours to eat a lobster in 10 minutes.’
(11)
Hanako-wa 3-zikan-de Bosuton-ni robusutaa-ga tabe-ni ik-e-ru. Hanako-TOP 3-hours-in Boston-to lobster-NOM eat-NI go-can-PRES ‘Hanako can go to Boston in 3 hours to eat a lobster.’
Some clarifications are in order here. First, the above observation indicates that the restriction on adverbial modification issyntactic in nature, not semantic. Consider again (11). This example is actually felicitous in the context where Hanako cango to Boston in 3 hours to eat a lobsters in 10 minutes, although the adverb 10-pun-de ‘in 10 minutes’ does not appear inthe sentence. This indicates that the ungrammaticality of (10) is not caused by a semantic incompatibility.

The same distribution holds for SEs: (12) shows that two adverbs are possible in the non-restructuring version; (13)and (14) show that only one adverb can appear in the restructuring context. In contrast to PEs, however, the adverb mustmodify the embedded predicate.

(12)
(?)Hanako-wa 3-zikan-de Bosuton-ni [10-pun-de robusutaa-o tabe-te]
s08

rs

(kuruma-de)

in Japanese:.006

, for relevant dis

ik-e-ru.
Hanako-TOP 3-hours-in Boston-to 10-minutes-in lobster-ACC eat-TE (car-by) go-can-PRES
‘Hanako can eat a lobster in 10 minutes and go to Boston in 3 hours (by car).’

(13) *
Hanako-wa 3-zikan-de 10-pun-de robusutaa-ga tabe-te ik-e-ru. Hanako-TOP 3-hours-in 10-minutes-in lobster-NOM eat-TE go-can-PRES ‘Hanako can eat a lobster in 10 minutes and go (somewhere) in 3 hours.’
(14)
Hanako-wa 3-zikan-de robusutaa-ga tabe-te ik-e-ru. Hanako-TOP 3-hours-in lobster-NOM eat-TE go-can-PRES ‘Hanako can eat a lobster in 3 hours and go (somewhere).’
The difference between PEs and SEs is further supported by the distribution of instrumental adverbs.6 As shown in(15a), instrumental adverbs are possible in restructuring SEs. However, they are incompatible with restructuring PEs, asshown in (15b).

(15)
a. Hanako-wa robsutaa-ga hasi-de tabe-te ik-e-ru. (SE) Hanako-TOP lobster-NOM chopsticks-with eat-TE go-can-PRES ‘Hanako can eat a lobster with chopsticks and go (somewhere).’
b. *
Hanako-wa robusutaa-ga hasi-de tabe-ni ik-e-ru. (PE) Hanako-TOP lobster-NOM chopsticks-with eat-NI go-can-PRES ‘Hanako can go to eat a lobster with chopsticks.’
As shown in (16), both constructions are grammatical when the embedded object is changed to accusative (i.e. when thestructure is a non-restructuring configuration).

(16)
a. Hanako-wa [robusutaa-o hasi-de tabe-te] (kuruma-de) ik-e-ru. (SE) Hanako-TOP lobster-ACC chopsticks-with eat-TE car-by go-can-PRES ‘Hanako can eat a lobster with chopsticks and go somewhere (by car).’
b.
Hanako-wa [robusutaa-o hasi-de tabe-ni] (kuruma-de) ik-e-ru. (PE) Hanako-TOP lobster-ACC chopsticks-with eat-NI (car-by) go-can-PRES ‘Hanako can go to eat a lobster with chopsticks (by car).’
Adjunction, clausal

cussion.




Interestingly, exactly the opposite situation holds for matrix modification: while restructuring PEs allow a matrix adverb(cf. (17a)), restructuring SEs prohibit matrix adverbs (cf. (17b)). Again, in non-restructuring configurations, both PEs andSEs allow matrix adverbs.

(17)

Pleasarchit

(19)

numbe

embed

matrix

7 See

a.

e citectur

r of d

ded a

adve

also Z

Hanako-wa

e this article ie, and phases

urative adverb

dverbs:

rbs

ushi (2008) for c

zitensya-de

n press as:. Lingua (20

Potenti

s: two

yes

yes

ritical discussio

robusutaa-ga

Takahashi, M12), http://dx.d

al PE SE

one on

no ye

yes no

n of Cinque (200

tabe-ni

., On reoi.org/10

e

s

6) based

ik-e-ru.

structuring infi.1016/j.lingua.

on Japanese data

(PE)
Hanako-TOP bike-by lobster-NOM eat-NI go-can-PRES ‘Hanako can go to eat a lobster by bike.’
b. *
Hanako-wa zitensya-de robusutaa-ga tabe-te ik-e-ru. (SE) Hanako-TOP bike-by lobster-NOM eat-TE go-can-PRES ‘Hanako can eat a lobster and go (somewhere) by bike.’
(18)
a. Hanako-wa zitensya-de [robusutaa-o tabe-ni] ik-e-ru. (PE) Hanako-TOP bike-by lobster-ACC eat-NI go-can-PRES ‘Hanako can go to eat a lobster by bike.’
b.
Hanako-wa zitensya-de [robusutaa-o tabe-te] ik-e-ru. (SE) Hanako-TOP bike-by lobster-ACC eat-TE go-can-PRES ‘Hanako can eat a lobster and go (somewhere) by bike.’
To summarize, we have seen that (i) the potential construction allows both embedded and matrix modification, (ii)restructuring PEs only allow matrix modification, and (iii) restructuring SEs only allow embedded modification. The resultsare summarized in (19):

(19) strongly suggests that we need a three-way distinction among restructuring configurations, contrary to the claims thatare at least implicitly made in some previous works (Cinque, 2006; Hoshi, 2006; Saito and Hoshi, 1998, among others).Thus, Cinque (2006) argues that all restructuring verbs are functional heads, assuming a single structure for allrestructuring configurations.7 If we take this view, the results in (19) are difficult to account for. It is also difficult to extendHoshi's (2006)/Saito and Hoshi's (1998) analysis to cover all the cases we have discussed because such an approachmay not account for the variations summarized in (19). In their analysis, ‘restructuring’ configurations are obtained bydirectly merging ‘matrix’ verbs and ‘embedded’ verbs (see section 5 for further discussion of this approach). In the nextsection, I therefore propose a new analysis of the pattern in (19).

3. Analysis

To account for the distribution of adverbs in the restructuring infinitives, I essentially adopt Wurmbrand's (2001)approach to restructuring infinitives, though the proposed analysis will depart from her proposals in some crucial respects.I propose that matrix modification is constrained by the thematic properties of verbs and that embedded modification isconstrained by syntactic restrictions on adjunct insertion.

3.1. Matrix modification

Wurmbrand (2001) argues that ‘restructuring’ is not a uniform phenomenon but that there are two types of restructuringconfigurations: lexical and functional restructuring. Lexical restructuring verbs are fully thematic verbs (Vs) (i.e. they assigntheta roles) that take very small complements (bareVPs). Crucially, infinitival complements in this configuration lack aCase-assigning head (v) (they also lack a subject). Embedded objects are thusCase-licensed by a higher functional head (v or T),yielding an (apparent) long-distance Case-licensing, as schematically shown in (20a). If lexical verbs take largercomplements such as CPs, TPs, and vPs, the resulting configurations are non-restructuring configurations (though not

nitives in Japanese: Adjunction, clausal2012.08.006

and Wurmbrand (2004) based on German data.




necessarily full clausal complements), in which embedded objects are Case-licensed by v within the infinitive as shownin (20b).

non-restructuringb.lexical restructuringa.(20)

vP vP

v’ v’

VP v[CASE] VP v

VVP (lexical/thematic) V’

OBJ[CASE] V CP V

C TP

T’

vP T

v’

VP v[CASE]

OBJ[CASE] V

On the other hand, functional restructuring is a direct consequence of the clausal architecture. Functional restructuring verbsare functional heads (in the verbal domain) such asmodals and they take infinitival complements, as shown in (21). Thus, theinfinitival complement is themainpredicateof theclause.Wurmbrand (2001)alsoproposes that functional restructuringverbsneed to be classified into (purely) functional predicates, which do not establish thematic relationshipwith arguments (cf. FP in(21)), and semi-functional predicates, which behave like purely functional heads syntactically but take an external argument(i.e. they assign the subject theta-role) and may assign Case to an internal argument (cf. vP in (20a and b), (21), and (22)).

(semi)-functional restructuring(21)

FP

F’

vP F(functional/non-thematic)

v’

VP v(semi-functional/thematic)

OBJ V

The present study assumes the clausal architecture in (21). I also assume that what is typically referred to as vP consistsof several sub-projections. More specifically, as shown in (22), there is a vP-layer which corresponds to the v-headintroducing an external argument, and a lower Aspect projection (see Cinque, 2006; MacDonald, 2006; Pylkkänen, 2002,2008; Travis, 2010 for relevant discussion).8


8 v in (22) can be further decomposed into several heads (see Pylkkänen, 2002, 2008; Travis, 2010, among others). However, I will not discussthis issue here.




functional categories below vP(22)

vP

SUBJ v ’

AspP v

Asp VP (functional)

V OBJ

The inflectional domain and the verbal domain are thus structured hierarchically as in (23).

(23)

Pleasarchit

9 Thecannot t10 TsujiTsujimur

functional restructuring verb (non-thematic) > semi-functional restructuring verb (thematic) >
aspectual functional head (non-thematic) > lexical verb (thematic)
In what follows, I will argue that (i) restructuring PE verbs are thematic lexical verbs, (ii) restructuring SE verbs are purelyfunctional lower aspect heads, and (iii) the potential morpheme is a semi-functional v head.

To begin with, the three types of restructuring verbs occur in the hierarchical order: potential morpheme > SE verbs(VSE) > PE verbs (VPE). As shown in (24), SE verbs can appear higher than PE verbs, but not vice versa.

(24)
a.
e citectur

orderinake stmuraa's an

John-ga

e this articlee, and phas

g restriction cative predicate(1993) claimsalysis more pr

hon-o/ga

in press as: Taes. Lingua (2012

an be understood ins such as potential mthat restructuring SEecise, considering a

kai-ni-it-te

kahashi, M., O), http://dx.doi.o

terms of semantiorphemes as theverbs are ‘auxiliamore articulated

ko-re-ru.

n restructuring inrg/10.1016/j.lingu

cs. For example, it isir complements. I tharies’, which cannot bevP-internal structure.

(VSE > VPE)
John-NOM book-ACC/NOM buy-NI-go-TE come-can-PRES ‘John can go to buy books and come.’
b. *
John-ga hon-o/ga kat-te-iki-ni ko-re-ru. (VPE > VSE) John-NOM book-ACC/NOM buy-TE-go-NI come-can-PRES ‘lit. John can go to buy books and come (somewhere).’
Furthermore, as shown in (25) and (26), the potential morpheme can appear higher than both PE verbs and SE verbs, butnot vice versa.

(25)
a. John-ga hon-o/ga kai-ni ik-e-ru. (potential > VPE) John-NOM book-ACC/NOM buy-NI go-can-PRES ‘John can go buy books.’
b. *
John-ga hon-o/ga ka-e-ni ik-u. (VPE > potential) John-NOM book-ACC/NOM buy-can-NI go-PRES ‘lit. *John goes to can buy books.’
(26)
a. John-ga hon-o/ga kat-te ik-e-ru. (potential > VSE) John-NOM book-ACC/NOM buy-TE go-can-PRES ‘John can buy books and go (somewhere).’
b. *
John-ga hon-o/ga ka-e-te ik-u. (VSE > potential) John-NOM book-ACC/NOM buy-can-TE go-PRES ‘lit. John can buy books and go (somewhere).’
The ordering restrictions among the potential morpheme, PE verbs and SE verbs provide support for the assumption thatthese verbs are base-generated in different positions in the clausal architecture, as laid out in (23).9

We are now ready to provide an account of the (im)possibility of modifying a restructuring predicate. I propose that onlythematic projections can be modified by VP adverbs. This immediately accounts for why restructuring SEs, but not thepotential morpheme and restructuring PE verbs, prohibit matrix modification (see Cardinaletti and Giusti, 2001; Cinque,2006; Napoli, 1981; Rochette, 1988, 1990; Rosen, 1989, 1990, and especially, Tsujimura, 1993, among others).10 This

finitives in Japanese: Adjunction, clausala.2012.08.006

plausible that there is a semantic restriction that SEsnk one reviewer for this suggestion.modified. The analysis presented in the text makes




proposal is further supported by the fact that restructuring PE verbs can take arguments, while restructuring SE verbscannot take arguments.

Let us first consider SE verbs. In a non-restructuring sentence, the SE verb can take a locative argument. However, in arestructuring sentence, the SE verb cannot take a locative argument (cf. Shibatani, 2007):

(27)

Pleasarchit

11 Evid12 Bobaagnosticwhen thverbal coclarifying

Hanako-ga

e cite this artiecture, and ph

ence for this claimljik and Wurmbraregarding the statey are selected bymplements are athis point.

tosyokan-ni

cle in pressases. Lingua

will be providend (2005) call thus of non-verbalexical verbs. S

lways spell-out

zassi-o/*ga

as: Takahashi, M., O(2012), http://dx.doi.or

d later in this paper.e relevant domains (i.e. (verl complements such as DPsee the discussion of light v

domains. I refer the reader to

kaesi-te-ik-e-ru.
Hanako-NOM library-to magazine-ACC/*NOM return-TE-go-can-PRES ‘Hanako can return a magazine and go to the library.’
Tosyokan-ni ‘to the library’ is disallowed when the object is nominative. Based on this observation, I conclude thatrestructuring SE verbs do not take arguments. The following example shows that the restructuring PE verbs can take alocative argument:

(28)
Hanako-ga tosyokan-ni zassi-o/ga kaesi-ni-ik-e-ru. Hanako-NOM library-to magazine-ACC/NOM return-NI-go-can-PRES ‘Hanako can go to the library to return a book.’
Here, the matrix verb can take a locative argument regardless of the Case of the embedded object.To summarize, I have proposed that (i) the potential morpheme is a semi-functional verb, (ii) restructuring PE verbs are

lexical restructuring verbs, and (iii) restructuring SE verbs are functional aspectual heads below vP. The distinctionbetween restructuring PE verbs and restructuring SE verbs is supported by the fact that while the former can takearguments, the latter cannot take arguments.

3.2. Embedded modification and restructuring verbs

In this section, I provide an account of the distribution of embedded adverbs. The observation we have to account for isthat while the potential morpheme and restructuring SE verbs allow embedded modification, restructuring PE verbs donot. We cannot extend the account of the impossibility of modifying restructuring SE verbs based on the lack of thematicproperties to restructuring PEs because all the embedded verbs under consideration take a (nominative) object, whichindicates that these verbs take arguments (i.e. they are lexical verbs). This means that these verbs do have thematicproperties.

FollowingWurmbrand (2001), I assume that restructuring PE verbs are lexical restructuring verbs. In other words, theyare lexical verbs (Vs). However, contrary to Wurmbrand's (2001) claim that complements of lexical restructuring verbsmust be bare VPs, I assume that complements of these verbs are headed by v, which does not assign Case (see Bhatt,2005; Takahashi, 2010, 2011, for relevant discussion).11 I also assume with Takahashi (2010, 2011) that v works as aphase head only if it assigns Case. Given this assumption, the complement vPs of lexical restructuring verbs are notphases. The core assumptions that I put forth for the ban on embedded modification in restructuring PEs are given below:

(29)
Lexical verbs (Vs) are phase heads.
(30)
Adjunction to XP is impossible if XP contains an unvalued Case-feature.
These two assumptions in tandem explain the ban on adjunction to complements of restructuring PEs in a principledway.

(29) is inspired by a proposal in Bobaljik and Wurmbrand (2005), which provides an analysis of what they call anti-reconstruction effects, which are observed cross-linguistically (see below for discussion).12 While I follow their insights,I interpret them in terms of the phase theory advanced by Chomsky (2000, 2001, 2004, 2008) and propose that the lexicalverbs under consideration are phase heads. In other words, the matrix VP is a phase. This in turn indicates that the vPcomplement of a lexical verb is a spell-out domain. Spell-out domains are domains across which Agree is blocked(see Chomsky, 2000, 2001, 2004, 2008 for Agree). Thus, if there are any elements in a spell-out domain that are still not

n restructuring infinitives in Japanese: Adjunction, clausalg/10.1016/j.lingua.2012.08.006

bal) complements of lexical verbs for them) agreement domains. They are. I argue that some non-verbal complements can also be spell-out domainserb constructions in section 4.2. However, I don’t intend to claim that non-Takahashi (2011) for relevant discussion. I am greatful to one reviewer for




Case-licensed, they must move out of the domain to avoid a derivational crash (cf. Bobaljik and Wurmbrand, 2005;Boskovic, 2007a). We then have a derivation like the following for restructuring PEs13:

(31)

VPgo

OBJ[CASE] V’ go

vP Vgo

PRO v’

VP v

OBJ[CASE] V

As complements of lexical verbs are spell-out domains, the vP-complement of the lexical verb in (31) is a spell-out domain.However, the object in the spell-out domain cannot be Case-licensed within the spell-out domain.14 The object then has tomove to the Spec, VP to avoid a derivational crash at the point of the introduction of the higher V. Once the matrix v isintroduced into the derivation, the moved object is Case-valued in the Spec, VP:

(32)

vP

VPgo v[CASE]

OBJ[CASE] V’go

vP Vgo

PRO v ’

VP v

OBJ[CASE] V

Here, the object in the Spec, VP is Case-valued by the matrix v and the derivation converges.15

Turning now to (30), (30) states that adjunction to a phrase is constrained by Case. In particular, according to (30),adjunction to a phrase becomes possible only after all the Case-features in the phrase are valued, which forces acyclicinsertion of adjuncts (cf. Stepanov, 2001). This is going to be crucial for the account of the ban on adjunction inrestructuring PEs.16

Having laid out the crucial assumptions for the current analysis, let us now see how they interact to exclude adjunctionto complements of restructuring PE verbs:


13 I assume that the infinitival marker -ni is inserted at PF.14 This is the crucial point of the derivation that distinguishes the derivation of restructuring PEs from that of transitive sentences. I will come backto this point below.15 If the object is marked nominative, the Case of the object may come from T (see Koizumi, 1994, 1995, 1998; Nomura, 2003, 2005; Takezawa,1987, among others). Then, the object in Spec, VP is Case-valued by T via Agree (see Nomura, 2003, 2005 for discussion). I will return to thispoint below.16 For a possible deduction of (30), see Takahashi (2011).




(33)

vP

adjunction VPgo v[CASE]

OBJ[CASE] V’ go

vP Vgo

PRO v’

VP*adjunction v

OBJ[CASE] V

The complement vP is a spell-out domain in (33) because it is selected by a lexical verb, which is a phase head. Thus, atthe point of the introduction of the matrix V, the object must move to the matrix Spec, VP. I assume that the condition onadjunction is evaluated only based on the elements in a spell-out domain. Thus, the vP complement, which is a spell-outdomain, contains a ‘trivial’ chain, i.e. the copy of the object. Then, adjunction to this embedded vP is impossible due to(30). The proposed analysis predicts that matrix modification should be possible in restructuring PEs because theembedded object is Case-licensed in the matrix clause. When this higher VP is spelled-out, the ‘trivial’ chain is now thehigher copy of the object that is Case-valued. Since adjuncts are inserted counter-cyclically after Case-valuation underthe current analysis, adjunction is predicted to be allowed. Note that counter-cyclic adjunction to the complement vP isimpossible, since this vP is already spelled-out when the object is Case-valued.

A question remains as to how the derivations converge under the proposed analysis given that the spell-out domaincontains a copy of the moving element that is not Case-valued. I assume, essentially following Nunes (2004), thatunvalued features of lower copies of the object are deleted at the point of Transfer to the interfaces.17 My intention here isto implement Nunes's (2004) formal feature (FF) deletion under the model that assumes Multiple Spell-out (cf. Chomsky,2000, 2001, 2004, 2008; Uriagereka, 1999), which Nunes does not assume. Nunes (2004) assumes that the FF-deletionprocess takes place in the phonological component to avoid a PF crash. Slightly departing from his original proposal, Iassume that unvalued features of lower copies in spell-out domains are always deleted by FF-deletion at the point ofTransfer. This is shown below:

(34)

VPgo

OBJ[CASE] V’ go

vP Vgo

PRO v’

VP v

OBJ[CASE] (Deletion by FF-deletion)V

TheunvaluedCase-feature in (34)disappearsbefore it reaches theLFandPF interfaces.Thederivation thusdoesnot crash.However, we have to make sure that FF-deletion takes place only if the unvalued features are those in the copy left

behind by movement. In other words, if FF-deletion were always possible, the object under consideration may not have tomove out of the spell-out domain. I assume, essentially following Boskovic (2007b), that the computational component


17 I thus depart from Chomsky (2001) and assume that lower copies of a chain in a spell-out domain can be deleted independently of featurechecking on the top of the chain. In other words, unlike Chomsky's (2001) system, in a non-trivial chain X1, X2, X3, deletion of a feature in X1 doesnot affect the feature on the lower copies.




looks at the whole phase (i.e. VP phase in (34)) at the point of Transfer and propose that the availability of FF-deletion in aspell-out domain can be determined on the basis of the elements at phase edges. Once the moving element in the phaseedge and the copy of the moving element in a spell-out domain are detected, the unvalued feature of the lower copy isdeleted at the point of Transfer by FF-deletion. The derivation then does not crash. In (34), the unvalued Case-feature ofthe object in the spell-out domain is deleted because there is a copy of the object in the VP edge. An important point here isthat the unvalued feature of the lower copy in the spell-out domain is not deleted before Transfer. As a result, theadjunction operation under consideration (i.e. insertion of adverbs), which is performed counter-cyclically in the syntax,but crucially before Transfer to the interfaces, is blocked by (30).

Let us now consider how the current analysis captures matrix modification in restructuring PEs. The relevant exampleis repeated below:

(35)

Pleasarchit

Hanako-wa


zitensya-de

icle in pressases. Lingu

robusutaa-ga

as: Takahasa (2012), http:

tabe-ni-ik-e-ru.
Hanako-TOP bike-by lobster-NOM eat-NI-go-can-PRES ‘Hanako can go to eat a lobster by bike.’
In (35), the adverb zitensya-de ‘by bike’ is intended to modify the matrix verb ik ‘go’ and the object is marked nominative.
I assumewith Koizumi (1994, 1995, 1998), Nomura (2003, 2005), and Takezawa (1987), among others, that the nominativeobject is Case-licensed by T. Specifically, I assumewith Nomura (2003, 2005) that the nominative object is Case-valued byT via Agree. Here, the moved object in the Spec, VP is Case-valued by the matrix T. Furthermore, I assume with Takahashi(2010, 2011) that v is a phase head only if it assigns Case. Consider the following step of the derivation:
(36)

vPcan T[CASE]

SUBJ v’can

VPadjunction go vcan

OBJ[CASE] V’ go

vP Vgo

PRO v’

VP v

OBJ[CASE] V

(36) illustrates that the embedded object moves to the Spec of the matrix VP once the restructuring PE verb is merged. Thepotential morpheme, which is represented as vcan, is then merged but crucially, this vcan does not assign Case (see Ura,1996, 1999, 2000). Given the second assumption that v is a phase head only if it assigns Case, vPcan is not a phase. This inturn indicates that there is no spell-out domain between the object and T. Therefore, the nominative object can be Case-valued by T via Agree. Adjunction is then predicted to be allowed. Nothing changes if the nominative object moves to Spec,TP (as in Koizumi, 1994, 1995, 1998; Nomura, 2003, 2005) because the object can move after Case-valuation. This isdifferent from thecaseof theembedded vP in (34)where theobjectmustmovebeforeCase-valuation.Thecrucial point in thederivation of the embedded vP in (34) is that there is a phase before the introduction of the matrix Case-licenser.

The proposed analysis also correctly predicts that embedded modification should be possible in restructuring SEs andthe potential construction. The relevant data are repeated below:

(37)
Hanako-wa robusutaa-ga hasi-de
hi, M//dx.

tabe-te-ik-e-ru.

., On restructuring indoi.org/10.1016/j.lingu

(SE)
Hanako-TOP lobster-NOM chopsticks-with eat-TE-go-can-PRES ‘Hanako can go to eat a lobster with chopsticks.’
(38)
Taroo-ga robusutaa-ga hasi-de tabe-rare-ru. (potential) Taro-NOM lobster-NOM chopsticks-with eat-can-PRES ‘Taro can eat a lobster with chopsticks.
finitives in Japanese: Adjunction, clausala.2012.08.006




In both of the above examples, the adverbsmodify the embedded verbs. Consider the following derivation for the potentialconstruction and restructuring SEs (recall that restructuring SE verbs (cf. Aspgo in (39)) are located below the potentialmorpheme (cf. vcan in (39))):

Restructuring SEs/the potential construction(39)

TP

T’

vPcan T[CASE]

SUBJ v’can

AspPgo vcan

AspVP go

OBJ[CASE] V

As in the case of the potential construction discussed above, there are no spell-out domains between the object and T.The object can then be Case-valued in situ via Agree. VP adjunction is thus predicted to be possible.

At this point, it is worth discussing some examples that are potentially problematic for the analysis I am pursuing. Let usfirst consider a simple transitive sentence:

(40)

Pleasarchit

18 I am19 See(2004:68

John ate sushi with chopsticks.

Here, the object is Case-valued by v and VP adjunction is apparently allowed, given that the adjunct with chopsticks doesnot cause ungrammaticality. What is important here is that VP is a spell-out domain (Chomsky, 2000, 2001, 2004, 2008).The question is then why VP adjunction is possible here under the proposed analysis. Consider the following derivation:

(41)
John [v’ v-atei [VP ti susi[CASE] [with chopsticks]]].
As noted above, I assume with Takahashi (2010, 2011) that vP constitutes a derivational phase only if v assignsCase. This in turn implies that spell-out domains, which are complements of phase heads, emerge only at the point ofCase-valuation. The object in the above example is then Case-valued in the spell-out domain, given that the object isbase-generated within the VP. On the other hand, in the case where adjunction is banned, hence adverbial modificationis impossible, the spell-out domain emerges before the introduction of v. Consider the following derivation including arestructuring PE verb18:

[(42) VP go [VP eat sushi[CASE] [with chopsticks ]]].

Assuming that the derivation proceeds in a bottom-to-top manner, the embedded VP becomes a spell-out domain at thepoint of merger of the matrix V. Thus, the Case-feature of the embedded object necessarily remains unvalued if the objectdoes not move out. The object then moves out to avoid a derivational crash.

Next, let us consider the following ECM construction, which I assume involves overt object shift19:

(43)
a.
e citectur

adoptChom2) for

I believe him to have been hit with a stick.
b. [CP C [TP1 I [vP1 v-believei [VP1 himj [ti [TP2 tj to [VP2 have [VP3 been [vP2 [v-hitk [VP4 tk tj with a stick]]]]]]]]]]]
e this article in press as: Takahashi, M., On restructuring infinitives in Japanese: Adjunction, clausale, and phases. Lingua (2012), http://dx.doi.org/10.1016/j.lingua.2012.08.006

ing the head-initial structure for the sake of exposition.sky (2008), Koizumi (1995), Lasnik (1999), Lasnik and Saito (1991) and Postal (1974), among others. See also Boskovica summary of empirical arguments for overt object shift.




Here, the ECMed subject, which is base-generated in VP4, is Case-licensed by the matrix v and adjunction to VP4 isapparently allowed.This dataposesaquestionbecause theproposedanalysismight predict that adjunction to the lowestVPshould be impossible. If the ECM verb believe is a lexical verb, then the TP complement (TP2) is a spell-out domain. As aresult, the object in VP4 would have to move out of this domain for Case. Though I do not have a complete answer to thisquestion, I would like to suggest two potential solutions. First, wemay follow Chomsky (2008) and assume that TP can be aspell-out domain only if it is selected by C. As a result, the TP complement of the ECM verb is not a spell-out domain. TheECMed subject is then Case-valued before movement so that adjunction to the embedded VP is allowed. The otherpossibility is to modify the definition of spell-out domains. While complements of ECM verbs are TPs, those of restructuringPEsare vPs. This difference in the size of the infinitival complementmight be related to the definition of spell-out domains. Inother words, complements of lexical verbs must be ‘small enough’ (i.e. vP, not TP) to qualify as spell-out domains.

To summarize, I have argued in this section that the ban on adjunction to complements of certain restructuring verbscan be explained in terms of spell-out domains and the timing of adjunction within spell-out domains. In the next section, Iexplore further ramifications of this analysis and show that other adjunction operations (adjective insertion and quantifierraising) also obey the proposed constraint.

4. Further extensions

I have so far argued that there is a ban on adjunction to complements of restructuring PE verbs and that the ban followsfrom the two assumptions:

(44)

Pleasarchit


(45)
XP cannot be a target of adjunction if it has an unvalued Case-feature.
If these assumptions are correct, we would expect to observe a similar ban on adjunction in other constructions. In thissection, I argue that this is indeed thecase. Thediscussionconcerns infinitival complementsofwasure ‘forget’andJapaneselight verb constructions. The former gives us a case of QR and the latter gives us a case of adjective insertion.

4.1. Infinitives with wasure ‘forget’: the Case of QR

In this section I discuss infinitives with wasure ‘forget’ to demonstrate that the ban on adjunction is observed in theserestructuring infinitives. Wasure-infinitives show the anti-reconstruction effect, which motivates the analysis in terms ofagreement domains in Bobaljik and Wurmbrand (2005) briefly noted above. A relevant example is given in (46):

(46)
Taroo-wa
e cite thisecture, and

ringo-dake-o

article in pressphases. Lingua

tabe-wasure-ta.

as: Takahashi, M(2012), http://dx.

Taro-TOP
apple-only-ACC eat-forget-PAST ‘Taro forgot to eat only apples.’ (only > forget, *forget > only)
., On restructuring infinitives in Japanese: Adoi.org/10.1016/j.lingua.2012.08.006

(Koizumi, 1995:56)

(46) shows that the embedded object must take scope over the matrix predicate (see Koizumi, 1995; Yumoto, 2004).Under the current proposal, the example has the following derivation:

(47) vP

SUBJ v

VPforget v[CASE]

OBJ[CASE] V’ forget

vP Vforget

PRO v’

VP v

OBJ[CASE] V

djunction, clausal




The complement ofwasure ‘forget’ is a spell-out domain hence the object has to move to the matrix VP. Themoved objectis Case-licensed by the matrix v after merger of the matrix V. This object takes scope over wasure ‘forget’ because it islocated above wasure ‘forget’.20

What is particularly interesting here is the fact that even non-Casemarked elements show the anti-reconstruction effect(cf. Saito and Hoshi, 1998):

(48)

Pleasarchit

20 Theand Yum21 For a(2012),22 I asssingle cy23 Evid

Taroo-wa


analysis predoto (2004). Irguments thaFitzgibbons (2ume with Bobcle. I also asence for this

hon-o

article in pphases. L

icts that the bawill discuss Tt there is QR i010), Oh (20alijk (1995) asume that QRclaim will be p

Mary-dake-kara

ress as: Takahaingua (2012), htt

n on adjunction to comioka's (2006) ann languages that are05) and Wurmbrandnd Saito (2005), amis a movement oprovided later in this

kari-wasure-ta.
Taro-TOP book-ACC Mary-only-from borrow-forget-PAST ‘Taro forgot to borrow books only from Mary.’ (only > forget, *forget > only)
(48) involves a PP argument, which does not receive Case from the matrix v. Dake ‘only’ contained in the PP musttake scope over wasure ‘forget’. A number of authors have argued that dake ‘only’ undergoes QR (see Futagi, 2004;Goro, 2007; Sano, 1985; Shoji, 1986, among many others), which I assume in this paper.21 I also assume thatQR of dake ‘only’ must target a propositional node (i.e. vP) (Bobaljik and Wurmbrand, 2007; Goro, 2007). Assumingthat QR is a syntactic adjunction operation (May, 1985), we now have an account of (48). Consider the followingderivation22:

(49)

vP

SUBJ v’

VPforget v[CASE]


*adjunction (QR) vP Vforget

PRO v’

VP v

OBJ[CASE] spell-out domainVP

V(QP)PP

(49) shows that the embedded vP, which is a spell-out domain, contains a copy of the object with an unvalued Case-feature. The proposed condition on adjunction predicts that adjunction to this vP should be impossible even though the vPis a propositional node (see also Takahashi, 2010, 2011).23 Then, the quantifier cannot adjoin to this vP. Given that theobject needs Case and that dake ‘only’ must undergo QR, the only possible derivation is the one in which the PP, aswell as the object, move to the matrix VP via scrambling, the quantifier in the PP then undergoing QR (adjunction) to thematrix vP:

shi, M., On restructuring infinitives in Japanese: Adjunction, clausalp://dx.doi.org/10.1016/j.lingua.2012.08.006

omplements of wasure should be at work, which is indeed the case. See Tomioka (2006)alysis in section 5.2.considered to be scope rigid (e.g., German and Japanese), see Bobaljik andWurmbrand(2008).

ong others, a model where ‘overt’ operations and ‘covert’ operations take place within aeration where the tail of the chain is pronounced (Fox and Nissenbaum, 1999).section.




(50)

vP (node of type t)

SUBJ v’

VPforget v[CASE]


V’(QP)PP forget

*adjunction (QR) vP Vforget

PRO v’

VP v

OBJ[CASE] domain spell-outVP

V(QP)PP

After scrambling of the PP to the matrix VP, the quantifier in the PP undergoes QR to the matrix vP, which is a node of typet. We thus capture the anti-reconstruction effect with PPs. It is interesting to note that Koizumi (1995) briefly considers thepossibility of QR in the context of restructuring infinitives and states as follows: ‘‘what we have to say in the case of theControl construction . . . is that QR may not be too short. Current syntactic theories, including GB theory, do not have anytheoretical device to cope with such a situation’’ (Koizumi, 1995:81; emphasis mine).24 What I have tried to do here is tocapture the effect in question.

One potential alternative analysis of the data in question would be an analysis in terms of lexical compounds in thesense of Kageyama (1993). Under this analysis, complex predicates with wasure ‘forget’ do not have any internalsyntactic structure. Hence, embedded quantification is predicted to be impossible because there should be no‘‘embedded’’ clause.25 One might also wonder if this anti-reconstruction effect with PPs could follow from the assumptionthat complements lack positions where quantifiers can potentially adjoin to (e.g. vP) (cf. Tomioka, 2006; Wurmbrand,2001). Thus, if complements of lexical verbs lack a vP projection, quantifiers in the complements must adjoin to matrixvPs, which yields the anti-reconstruction effect.

Let us now consider examples that involve causatives, which circumvents these possibilities. I briefly discuss thecausative construction and the distribution of binders of zibun ‘self ’ in Japanese. Consider the following causativesentence:

(51)

Pleasarchit

24 Note2000; Sa25 See

Taroo-ga


, however, thaito and MuraKoizumi (199

Hanako-ni

article in presphases. Lingu

t there is a ban osugi, 1999, amon5) for arguments

hon-o

s as: Takaa (2012), h

n movement tg many otheragainst this a

sute-sase-ta.
Taro-NOM Hanako-DAT book-ACC discard-make-PAST ‘Taro made Hanako discard a book.’
(51) is analyzed as having a bi-clausal structure, which is motivated by the following observation concerning zibun ‘self ’(see Kuno, 1973; Kuroda, 1965, among others):

(52)
Tarooi-ga Hanakoj-ni zibuni/j-no
h

hsn

hon-o

ashi, M., Ottp://dx.doi.o

at is too short).alysis.

sute-sase-ta.
Taroi-NOM Hanakoj-DAT selfi/j-GEN book-ACC discard-cause-PAST ‘Taro made Hanako discard his/her book.’
n restructuring infinitives in Japanese: Adjunction, clausalrg/10.1016/j.lingua.2012.08.006

, often referred to as anti-locality (Abels, 2003; Boskovic, 1994; Grohmann,




Here, the reflexive zibun ‘self’ can refer to either Hanako or Taro. Given the standard assumption that the antecedent ofzibun ‘self ’ must be the subject of a clause, this shows that (52) has a bi-clausal structure. Following Harley (2008),Murasugi and Hashimoto (2004), and Saito (2006a), I assume that lower clauses of causative constructions are vPs andsubjects in the relevant sense are elements in Spec, vP.Hanako in (52) thus occupies the Spec of the lower vP. Considernow the following example:

(53)

Pleasarchit

26 Thewhich iswhich cl27 I ass

Senseei-wa


data also provideconsistent with thearly involves a vume with Koizum

gakuseej-ni

icle in pressases. Lingua

independent evide analysis proposP complement. Si (1995) that the

sono-nyuusu-o

as: Takahashi, M(2012), http://dx

ence that complemed in this paper. Thiee section 5.1 fordative causee is a

zibuni/j-no

., On rest.doi.org/10.1

ents of lexicals was shown bsimilar observPP in Japane

ryoosin-dake-ni
teacheri-TOP studentj-DAT that-news-ACC selfi/j-GEN parents-only-to hookoku-sase-wasure-ta. report-cause-forget-PAST ‘The teacheri forgot to make the studentj report the news only to hisi/j parents.’
(only > forget, *forget > only)
(only > forget: it is only to hisi/j parents that the teacheri forgot to make the studentj report the news.) (*forget > only: the teacheri forgot to make the studentj report the news to hisi/j parents but not others(i.e. it was possible that the student reported the news to other people).)
In (53),wasure ‘forget’ takes a causative construction, which involves a vP complement. This is shown by the fact that thedative causee can be the subject of zibun ‘self’. Importantly, we still observe the anti-reconstruction effect with the PP,dake ‘only’ obligatorily taking scope overwasure ‘forget’. If the anti-reconstruction effect were to be obtained by the lexicalcompounding analysis, sentecnes like (53) should be predicted to be impossible, because the complement of wasure‘forget’ contains a vP complement, which indicates that the complement has internal structures. Furthermore, if the anti-reconstruction effect on QR in (48) were to be obtained by assuming a lack of a proper adjoining position (i.e. vP), dake‘only’ should still be able to take scope under wasure ‘forget’, contraty to fact. (53) thus gives further credence to thedomain-based analysis entertained here.26

Let us now consider how the proposed analysis captures the above facts. The crucial assumptions are repeatedbelow:

(54)
(55)
XP cannot be a target of adjunction if it has an unvalued Case-feature.
(56)
vP constitutes a derivational phase only if v assigns Case (Takahashi, 2010, 2011).
Let us first consider how Case-marking takes place in the causative construction:

(57)
a. John-ga eego-?o/ga wakar-u. John-NOM English-ACC/NOM know-PRES ‘John understands English.’
b.
Mary-ga John-ni eego-o/*ga wakar-ase-ru.
ructuring inf016/j.lingua

restructuring vy the fact thatations.se.

Mary-NOM
John-DAT English-ACC/NOM know-cause-PRES ‘Mary makes John understand English.’ (Takahashi, 2010:338)
In (57a), it is only marginally acceptable to have accusative Case on the object. However, in (57b), where the causativemorpheme selects wakar ‘understand’, the accusative Case is fully acceptable while nominative marking is impossible.This indicates that the accusative Case of the object in (57a) comes from --(s)ase ‘cause’, rather thanwakar ‘understand’.This contrast suggests that the embedded accusative objects in causative constructions are Case-valued by matrixpredicates.27 This point is quite important because it shows that the verbal projection of the complement of the causativemorpheme does not constitute a phase, given (56). We can then assume the following derivation for the causativeconstruction. I assume with Harley (2008), among others, that the causative morpheme --(s)ase ‘cause’ is an exponent ofthe matrix v (cf. vcause in (58)), which selects vP:

initives in Japanese: Adjunction, clausal.2012.08.006

erbs such as wasure ‘forget’ can be larger than VPs,wasure ‘forget’ in (53) takes a causative construction,




(58) vPcause

SUBJ v’cause

vP vcause [CASE]

SUBJcausee v’

VP v

OBJ[CASE] V

Here, the Case of the embedded object is licensed by the matrix v. This object does not have to move for Case becausethere is no relevant spell-out domain before Case-valuation of the object. Now, consider the following derivation, whichcorresponds to (53):

(59) vP

SUBJ v’

VPforget v[CASE]


V’(QP)PP forget

*QR vPcause Vforget

PRO v’cause

*QR vP vcause spell-out domain

SUBJcausee v’

VP v

OBJ [CASE] VP

V(QP)PP

The complement ofwasure ‘forget’ in (59) is a causative sentence. Aswasure ‘forget’ is a lexical verb, the complement is aspell-out domain. As a result, the embedded object must move out of this domain for Case. As the spell-out domaincontains an unvalued Case-feature, adjunction (i.e. QR) to this vP node is predicted to be impossible. We thus capture theanti-reconstruction effect in the causative construction.28


28 As pointed out by one reviewer, the analysis predicts that embedded modification should be impossible in examples like (53). This predictionis borne out. Consider first the following example:

(i) Tarooi-wa Hanakoj-ni mata zibuni/j-no hon-o sute-sase-ta.Taroi-TOP Hanakoj-DAT again selfi/j-GEN book-ACC discard-cause-PAST‘Taro made Hanako discard his/her books again.’

This example contains an adverbmata ‘again’.Mata ‘again’ canmodify either the event of discarding (in which case Hanako could have discardeda book previously without being forced by Taro) or the event of causing (in which case Taro did make Hanako discard a book previously). Theexample is thus ambiguous. However, the following example, where the causative construction is selected by wasure ‘forget’, is unambiguous:

(ii) Tarooi-wa Hanakoj-ni mata zibuni/j-no hon-o sute-sase-wasure-ta.Taroi-TOP Hanakoj-DAT again selfi/j-GEN book-ACC discard-cause-forget-PAST‘Taro forgot to make Hanako discard his/her books again.’

Mata ‘again’ in this example can only modify wasure ‘forget’. Hence the example only means that the event of Taro's forgetting happened again,which shows that embedded modification is impossible in this example. See sections 5.1 and 5.2 for discussion.




To conclude, I have argued in this subsection that infinitives with wasure ‘forget’, which show the anti-reconstructioneffect, provide further evidence for the analysis of the ban on adjunction proposed in this paper. In particular, it was shownthat quantifier raising, as well as adverb insertion, obey the adjunction constraint.29

4.2. Japanese light verb constructions as lexical restructuring: a preliminary analysis

In this subsection I briefly discuss Japanese light verb constructions (henceforth LVC) and provide a further argumentfor the analysis proposed in this paper. Examples of LVCs are given below (see Grimshaw and Mester, 1988; Saito andHoshi, 2000; Terada, 1990, among many others):

(60)

Pleasarchit

29 Thisunder th

(i) a. TT‘

b. TT‘

In both eaccusatithe pote2005) su

(ii) TaroTarokusuring.‘I ha

Nomurainvestiganominat30 Any

(i) a. HH‘

b. HH‘

In (ia), th(1982) fo

a.

e citectur

analyse pote

aroo-aro-NTaro c

aroo-aro-NTaro c

xampve andntial mggest

o-ga-NOMriyubifingerve kn

's (200tion, Iive obPP/DP

anakanakHanak

anakoanakoA lette

e posr disc

John-wa

e this article, and phas

is predicts thantial morphem

ga migime-OM right.eyean close only

ga migime-OM right.eyean close only

les, the verbit takes scoporpheme. It has that given a

koyubi-dakpinkie-only-

-dake-ga m-only-NOM cown that Taro

3, 2005) obsleave this issujects (see alsoin a nomina

o-*(no) hono-GEN booko's book’

-kara-*(no)-GENr from Hanak

sessor NP Haussion.

[NP

e ines.

t nome. C

dake-onlyhis r

dake-onlyhis r

tumue unds ben ap

e-gaNOMage-rook-can

ervate opNom

l proj

tegamlettero’

nako

zaisan-no

press as: TakLingua (2012)

inative objects inonsider first (i) (th

-o tumur-e-ru-ACC close-canight eye.’ (?*only

-ga tumur-e-r-NOM close-canight eye.’ (only >

r ‘close’ is accomer the potential men assumed that tpropriate context,

mage-rare-rucrook-can-Pres

rare-ru no-nican-PRES that-Dcrook only his pin

ion is thus consisen in this paper. Sura, 2005).

ection must be m

i

is marked genitiv

bossyuu]-o

ahashi, M., On re, http://dx.doi.org/1

the potential constructe judgments are Tada

.-PRES> can, can > only)

u.-PREScan, *can > only)

panied by the potentiaorpheme. On the otherhe nominative object mnominative objects ca

no-wa sit-te i-ta-that-TOP know-PR

-wa odoro-ita.AT-TOP surprise-PAkie but I am surprised

tent with the analysisee Koizumi (2008) for a

arked with genitive Ca

e, and in (ib), the PP H

si-ta.
John-TOP property-GEN confiscation-ACC do-PAST
b.
??John-wa zaisan-o [bossyuu]-o si-ta. John-TOP property-ACC confiscation-ACC do-PAST
‘lit. John did confiscation of property.’

In (60a), zaisan ‘property’ is theta-marked by the verbal noun bossyuu ‘confiscation’ and is located in the projection ofthe verbal noun bossyuu ‘confiscation’, as shown by the fact that zaisan ‘property’ is genitive-marked (cf. Kitagawaand Ross, 1982).30 On the other hand, in (60b), zaisan ‘property’ receives accusative Case, which indicates that it isCase-licensed outside of the NP. This construction is called light verb construction because the verb su ‘do’ does notseem to assign any theta-roles to its arguments (but see below for further discussion). (60b) is marginal due to the‘surface’ double-o constraint, which roughly states that there cannot be more then one accusative phrase (seeHarada, 1973, 1975; Hiraiwa, 2010; Sells, 1988; Shibatani, 1973, among others, for relevant discussion). Importantly,the violation can be circumvented by some syntactic operations such as clefting. Consider the following example thatinvolves clefting:

structuring infinitives in Japanese: Adjunction, clausal0.1016/j.lingua.2012.08.006

ion, which allows embedded modification, should be able to take scope's (1992)):

(Tada, 1992:94)

l morpheme -e ‘can’. In (ia), the object migime ‘right eye’ is markedhand, in (ib), the object is marked nominative and can take scope overust take scope over the potential morpheme. However, Nomura (2003,n take scope under the potential morpheme:

ga (kare-ga)OG-PAST-but, (he-NOM)

STthat he can also crook only his ring finger.’ (Nomura, 2005:176)

in this paper. As scope facts of nominative objects require a carefulpossible account of the preference for the wide scope interpretation of

se in Japanese:

anako-kara ‘from Hanako’ is marked genitive. See Kitagawa and Ross



(61) [CP Opi [TP John-ga ti bossyuu-o si-ta] no]-wa zaisani-o da.


Pleasarchit

31 Thiswhich ar

(i) MaryMary‘Mar

(ii) [CP

‘It w

Double ain (ii).32 Noteungramm

(i) *JohJoh‘lit.

33 (62dclefting t

(i) [CP [

‘lit. It

There isimpossibattention


‘surface’ doube totally ungra

-ga John--NOM John-y made JohnOpi [IP Mary-g

Mary-Nas a book tha

ccusatives are

that the orderatical.

n-wa zinson-TOP quickJohn did quic

) has a degrado (62c) and (6

TP John-gaJohn-NOM

was property

no violation ofle, which furth.

article in prphases. Lin

le-o constrainmmatical (see

*o/ni hoACC/DAT boread a book.’a John-*o

OM John-ACt Mary made J

banned in the

between the a

ku-na zaisanproper

k confiscation

ed status due2d):

*zinsokuna/zquick/quickl

that John did

the double-o cer confirms th

esgu

t sH

n-ook

/ni

Coh

ca

dj

-oty-of

to

inslyqu

one c

s as: Takahashi,a (2012), http://dx

hould be distinguishedarada, 1975, among

yom-ase-ta.-ACC read-cause-PA

ti yom-ase-ta]

/DAT read-cause-n read.’

usative construction a

ective and the verbal n

[bossyuu]-oACC confiscation-ACproperty.’

the surface double-o c

okuni bossyuu-oconfiscation-A

ick confiscation of.’

straint due to the appliontrast observed in (6

M., On re.doi.org/10

from the coothers):

ST

no]-

PAST that-

s shown in (i

oun does n

si-ta.C do-PAST

onstraint, bu

si-ta]CC do-ACC

cation of clef2). I thank to

structuring.1016/j.lin

nstraint that

wa hon-o

TOP book-

). The senten

ot affect the

t it is clearly

no]-wathat-TOP

ting. While aone reviewe

infinitives ingua.2012.08.

rules out doubl

da.

ACC COP

ce is ungramma

ungrammaticalit

better than (62c

zaisan-oproperty-ACC

dverbial modificr for bringing th

John-NOM
confiscation-ACC do-PAST that-TOP zaisan-ACC COP ‘lit. It was John who did confiscation of property.’
This example shows that the surface double-o constraint observed in (60b) is suppressed under clefting. FollowingHiraiwa (2010), among others, I assume that the marginal status of (60b) is due to the surface filter, which disallows twoaccusative phrases in certain syntactic domains.31

In this section I offer a preliminary analysis of LVCs in terms of lexical restructuring infinitives. More precisely, I willdevelop a version of the heavy verb analysis of LVCs (see Kuo, 2009; Terada, 1990; Uchida andNakayama, 1993, amongothers) within the system I have adopted.

Consider the examples in (62). Kurogi (2002) observes that adverbs, but not adjectives, can appear in LVCs withdouble accusatives, which is shown in (62c) and (62d):

(62)
a. John-wa [NP zinsoku-na zaisan-no bossyuu]-o si-ta. John-TOP quick property-GEN confiscation-ACC do-PAST ‘lit. John did quick confiscation of property.’
b.
John-wa zinsoku-ni [zaisan-no bossyuu]-o si-ta. John-TOP quickly [property-GEN confiscation-ACC do-PAST ‘lit. John quickly did confiscation of property.’
c. *
John-wa zaisan-o zinsoku-na [bossyuu]-o si-ta. John-TOP property-ACC quick confiscation-ACC do-PAST ‘lit. John did quick confiscation of property.’
d.
??John-wa zinsoku-ni zaisan-o [bossyuu]-o si-ta. John-TOP quickly property-ACC confiscation-ACC do-PAST
‘lit. John quickly did confiscation of property.’
(Kurogi, 2002:31; slightly modified/glosses and translations by the author)
In (62a and b) the argument of the verbal noun is Case-licensed within the verbal noun projection. In these examples, anadjective can appear in the verbal noun projection (cf. (62a)) and an adverb can modify the matrix verb (cf. (62b)). In (62cand d), on the other hand, the argument of the verbal noun is Case-licensed by the matrix v, as indicated by the fact thatthe argument receives accusative Case. The contrast between (62c) and (62d) shows that the adverb can appear in thisconstruction, but the adjective cannot.32,33 This paradigm receives a straightforward explanation under the proposedanalysis. I assume that su ‘do’ in LVCs is actually a heavy (i.e. lexical) verb as argued by Kuo (2009), Terada (1990), and

Japanese: Adjunction, clausal006

e accusatives in causative constructions,

tical even if clefting takes place, as shown

y of (62c). Thus, the following example is

).This point becomes clear once we apply

da.COP

ation is possible, adjectival modification ise relevance of the examples like (i) to my




Uchida and Nakayama (1993). This assumption is supported by the fact that non-agentive subjects are prohibited in LVCs(see Kuo, 2009; Miyagawa, 1989; Tsujimura, 1990; Saito and Hoshi, 2000 for discussion):

(63)

Pleasarchit

34 I do ndo not corelevant

?*Nimotu-wa

e cite this articleecture, and phas

ot intend to claim thanstitute a spell-out ddiscussion.

Oosaka-ni

in presses. Lingua

t nominal comomain. A ques

tootyaku-o

as: Takahas(2012), http:

plements of lextion remains as

si-ta.
package-TOP Osaka-to arrival-ACC do-PAST ‘The package arrived in Osaka.’ (Kuo, 2009:173; cf. Grimsahw and Mester, 1988)
This example shows that subjects of LVCs cannot be inanimate, which in turn indicates that si ‘do’ in fact assigns a theta-role to the subject (but see Saito and Hoshi, 2000; Saito, 2006b for alternative explanations of the data). Second, I assumethat adjectives undergo adjunction (see Boskovic, 2010 for recent evidence to this effect). The first assumption predictsthat the complement of su ‘do’ is a spell-out domain, which should force the argument of the verbal noun to move out of thenominal projection when the argument cannot get its Case licensed within the spell-out domain. Given the secondassumption, just like adverbs, adjectives are subject to the condition on adjunction I have proposed. Having laid out theassumptions for the analysis, let us go back to the analysis of the paradigm. In (62a and b), the argument of the verbalnoun is Case-licensed within the NP, which is the complement of su ‘do’.The complement NP is a spell-out domain. Sincethe argument does not need to move out of the complement NP for Case, adjunction to the complement is allowed, asshown in (62a). Also, nothing bans adverbial modification of the matrix verb, which is indeed allowed, as shown in (62b).The crucial contrast we have to account for is the one between (62c) and (62d). Let us consider the following derivation:

(64)

VPdo v[CASE]

NPproperty[CASE] V’ do

nP Vdo

n’

NP *adjunction confiscation n

NPproperty[CASE] Nconfiscation

I assume that NP is selected by n (but nothing hinges on this assumption). The nP here is a spell-out domain because it isthe complement of su ‘do’. As there is no Case-assigner within the nP, the object has to move out of this domain for Case,hence it moves to the Spec, VP. As the lower copy of this object has no Case, adjunction to the NP and the nP (i.e.adjective insertion) is impossible. (62c) thus follows. As Case of the moved object is licensed by the matrix v, counter-cyclic adjunction to the matrix VP (i.e. adverb insertion) is allowed, which results in (62d).34

Although the full analysis of LVCs in Japanese is beyond the scope of the present study, I would like to consider herebriefly some examples that are discussed in Kurogi (2002). Kurogi (2002) argues that the adverb in (62d), which isrepeated below, in fact modifies the verbal noun, rather than the verb.

(65)
a. ??John-wa zaisan-o zinsoku-ni
hi, M//dx

icalto w

[bossyuu]-o

., On restructurin.doi.org/10.1016/j.li

verbs are always spell-hen nominal compleme

si-ta.
John-TOP property-ACC quickly confiscation-ACC do-PAST ‘lit. John quickly did confiscation of property.’
b. *
John-wa zaisan-o [zinsoku-na bossyuu]-o si-ta. John-TOP property-ACC quick confiscation-ACC do-PAST ‘lit. John did quick confiscation of property.’

g infinitives in Japanese: Adjunction, clausalngua.2012.08.006

out domains. Thus, in transitive sentences like (41), objectsnts constitute spell-out domains. See Takahashi (2011) for




He provides the following data as evidence to this effect:

(66) *

Pleasarchit

35 I tha36 Therwould lik37 Tomhas layeobjects a

John-wa


nk Susi Wurme are other ine to addressioka's (2006)rs. Thus, the hre distinct fo

zaisan-o

article in presphases. Lingu

brand for suggesteresting data in Kthose data in futudiscussion conceread that is responr Tomioka (2006).

[bossyuu]-o

s as: Takahashi, Ma (2012), http://dx.

ting this analysis.urogi's work that are wore research.ns infinitival complemesible for the semantics oSee also Kratzer (199

zinsoku-ni

., On restdoi.org/10.1

rth investigat

nts of wasuref causation an6) and Pylkkä

si-ta.
John-TOP property-ACC confiscation-ACC quickly do-PAST ‘lit. John quickly did confiscation of property.’

The point of this observation is that the adverb is adjacent to the verb but the sentence is ungrammatical. Kurogi (2002)thus concludes that the adverb does not modify the verb. This interesting observation can be accounted for under thepresent analysis. I suggest that (66) violates Abels's (2003) Stranding Generalization, which prohibits movement ofcomplements of phase heads (see Abels, 2003 for details).35 In (66) the complement which contains bossyuu‘confiscation’ is moved out of the VP phase headed by si ‘do’. The derivation thus violates the generalization.36

To conclude, I have argued in this subsection that the distribution of adjuncts in Japanese light verb constructionsreceives an account under the theory proposed here. In particular, I proposed that light verb constructions in Japanese arebest analyzed as lexical restructuring constructions. If the analysis proposed in this section is correct, it adds another caseto the ban on adjunction, namely the ban on insertion of adjectives.

5. Comparison with alternatives

In this section I consider several proposals that could be extended to account for the ban on embedded modification orthose that are specifically made to account for the ban on embedded modification.

5.1. Tomioka (2006)

Tomioka (2006) proposes that complements of lexical restructuring verbs lack a projection that can host adverbs (i.e.voiceP, which introduces an agent as its Spec (see Kratzer, 1996; Pylkkänen, 2002, 2008 for discussion)).37 Consider thefollowing structures:

b. non-restructuring a.(67) restructuring

adjunction voiceP

Agent voice’

causeP voice *adjunction causeP

VP cause VP cause

V OBJ VOBJ

(67a), which is a non-restructuring construction, involves voiceP, which can host adverbs. On the other hand, (67b), whichis a restructuring construction, does not involve voiceP, hence there is no way to insert adjuncts (and subjects). Thisanalysis seems to correctly capture the fact that complements of certain lexical verbs disallow adjunction. However, itfaces difficulties with more complex cases. I will now introduce a phenomenon that I will call additional ban on adjunctionhereafter. Consider the following example of restructuring PEs, where the motion verb takes vP as its complement:

(68)
Titioyai-ga musukoj-ni zibuni/j-no hirugohan-o/ga tabe-sase-ni
ructuring in016/j.lingua

ing to complet

‘forget’. Noted the one thatnen (2002, 20

ik-e-ru.
fatheri-NOM sonj-DAT selfi/j-GEN lunch-ACC/NOM eat-cause-NI go-can-PRES ‘The fatheri can go to make his sonj eat hisi/j lunch.’
finitives in Japanese: Adjunction, clausal.2012.08.006

e the proposed analysis of light verb constructions. I

also that Tomioka (2006) assumes that ‘vP’ actuallyis responsible for introducing Agents and the Case of08).




The lowest object in (68) can be marked nominative or accusative. Regardless of the Case of the object, the sentence isambiguous; zibun ‘self’ can refer to either titioya ‘father’ or musuko ‘son’. The fact that zibun ‘self’ can refer to the dativecausee indicates that there is a vP complement in the causative construction.

Significantly, adverbial modification of the most embedded verb is impossible when the embedded object is markednominative:

(69)

Pleasarchit

a.

e citectur

Titioyai-ga

e this articlee, and phase

musukoj-ni

in press ass. Lingua (2

zibuni/j-no

: Takahash012), http:/

hirugohan-o

i, M., On re/dx.doi.org/10

hasi-de

structuring infini.1016/j.lingua.20

tabe-sase-ni

tives in Japan12.08.006

ik-e-ru.
fatheri-NOM sonj-DAT selfi/j-GEN lunch-ACC chopsticks-with eat-cause-NI go-can-PRES ‘The fatheri can go to make his sonj eat hisi/j lunch with chopsticks.’
b. *
Titioyai-ga musukoj-ni zibuni/j-no hirugohan-ga hasi-de tabe-sase-ni
e

ik-e-ru.
fatheri-NOM soni-DAT selfi/j-GEN lunch-NOM chopsticks-with eat-cause-NI go-can-PRES ‘The fatheri can go to make his sonj eat hisi/j lunch with chopsticks.’
Here, we have an adverb hasi-de ‘with chopsticks’, which is intended to modify the verb tabe ‘eat’. In (69a), the embeddedobject is marked accusative and the adverb is allowed. On the other hand, in (69b), the embedded object is markednominative and the adverb is now disallowed. Importantly, (69b) shows that there is a ban on adjunction to predicatesembedded in complements of lexical verbs, which I call additional ban on adjunction.

Having introduced the additional ban on adjunction, we can go back to the discussion of Tomioka's (2006) proposal.Under Tomioka's (2006) analysis, it is unclear how this ban can be explained. There are two reasons for this conclusion.First, as the analysis is stated in terms of a selectional property (of restructuring verbs), which works locally, it is difficult toexplain the additional ban on adjunction, which takes place in complements that are not directly selected by restructuringverbs. Second, in (69b), the complement of the restructuring verb does seem to have voiceP. We have seen that thedative causee in the causative construction is a subject, which is located in Spec, vP. This in turn indicates that the headthat hosts the dative causee (i.e. the embedded v) should be a voice head. Then, we have an example in which adjunctionis banned even in the presence of voiceP. All of these considerations point to the conclusion that Tomioka's (2006)proposal is insufficient to capture the full range of relevant facts. On the other hand, the additional ban on adjunctionreceives an account under the present analysis, which provides the following derivation for (69b):

(70)

VPgo/forget v[CASE]

OBJ[CASE] V’ go/forget

vPcause Vgo/forget

PRO v’cause

*adjunction vP vcause

SUBJcausee v’ domain spell-out

VP*adjunction v

OBJ[CASE] V

The complement of the lexical restructuring verb in (70) is a causative sentence. This is a spell-out domain because itis the complement of ik ‘go’. Consequently, the embedded object must move out of this domain for Case. As the lowercopy left behind in the spell-out domain has no Case, adjunction within this domain is correctly predicted to beimpossible.

5.2. Complex head analysis

Hoshi (2006) and Saito and Hoshi (1998), among others, propose that (at least some) restructuring constructionsinvolve a complex head, where two heads are directly combined in the syntax (either by direct merger or headmovement).Consider the following structure proposed in Saito and Hoshi (1998):

se: Adjunction, clausal




(71)

X2

XOBJ 2

X1 X2

Here the embedded verb X1 is directly merged to the restructuring verb X2. This analysis can capture the ban onadjunction to complements of restructuring verbs if we assume that adjuncts need to modify phrasal categories. This is sobecause there is no phrasal complement that adjuncts could adjoin to (see Yumoto, 2004 for somewhat relevantdiscussion). Furthermore, if we assume with Saito and Hoshi (1998) that the object that is base-generated above thecomplex head asymmetrically c-commands X2 in (71), the anti-reconstruction effect discussed earlier also follows.

The complex head analysis armed with the assumption concerning phrasal modification made here is also not withoutproblems. In particular, the addtional ban on adjunction I have introduced raises a question for this approach, becausenothing bans the following structure:

(72)

vP

SUBJ v’

VP v

SUBJcausee . VP

VP vcause-Vgo/forget

VOBJ

Here, the causative morpheme and the lexical restructuring verb form a complex head but the complement of thecausative morpheme is a VP, which can in principle host adjuncts. One could postulate the following structure, in which allthe verbs are merged together:

(73)

vP

SUBJ v’

VP v

SUBJcausee VP

V-vOBJ cause-Vgo/forget

In (73) all the verbs are merged together to form a complex head. The additional ban on adjunction can be capturedbecause there are no projections to adjoin adverbs below the restructuring verb (cf. Vgo/forget in (73)). While this structure isnot excluded under the complex head analysis, the real problem is that it is unclear under this analysis how (72) and (73)can be differentiated. In other words, (72) should be ruled out even as an option but it is unclear how this can be doneunder the complex head analysis. Unless this possibility is excluded in a principled way, the complex head analysispredicts that there should be no additional ban on adjunction.





5.3. Tsujimura (1993)

Tsujimura (1993) focuses on morphological differences between PEs and SEs. Tsujimura (1993) assumes that whilethe morpheme -te in SEs has tense specification, -ni in PEs has no tense specification. Based on this assumption,Tsujimura (1993) suggests that adjuncts require [tense] to be interpreted. This condition applies uniformly to bothrestructuring and non-restructuring sentences. This suggestion correctly captures at least part of the generalization I haveobtained. As SEs have a [tense] specification, embedded modification is predicted to be available. Tsujimura (1993)assumes that matrix modification for SEs is unavailable because the matrix predicate becomes an auxiliary. Moreover, asPEs have no tense specification, embedded modification is banned. However, as we have seen above, the differencebetween the two constructions disappears if we force non-restructuring configurations. In particular, it remainsunexplained why non-restructuring PEs allow embedded modification.

In fact, Tsujimura's (1993) motivation for the proposal is her observation that the contrast between PEs and SEs withrespect to embedded modification can be observed even in sentences with an accusative object. However, whatTsujimura (1993) fails to note is that embedded accusative objects do not necessarily entail non-restructuring. In otherwords, examples with an accusative object can involve restructuring as long as the adjacency requirement is respected.This point can be shown by the fact that a clause-boundedNPI -sika in the embedded clausewith an accusative object canbe licensed by matrix negation (see Tanaka, 1997 and references therein for -sika NPIs)38:

(74)

Pleasarchit

38 The(i) Taro

Taro‘Tar

(ii) *HaHa‘Han

In (i) theboundar

a.

e citectur

clauseo-ga-NOMo buynako-wnako-Tako th

objecy betw

Hanako-ga

e this article ine, and phases.

-boundedness ofhon-sikabook-SIKA

only books.’a [ Taroo-gaOP Taro-NOinks that Taro bu

t hon ‘book’ is acceen sika and the

Mary-ni-sika

press as: TaLingua (2012

sika is shown bykawa-na-i.buy-NEG-PRES

hon-sika kM book-SIKA by only books.’

ompanied by sikanegation. The co

zassi-o

kahashi, M., On), http://dx.doi.or

the following contra

a-u-to] omuy-PRES-that thin

and sika is c-commntrast shows that s

watasi-ni

restructug/10.1016

st:

ow-ana-i.k-NEG-PRE

anded by nika must be

ika-na-i.

ring infinitives/j.lingua.2012.08

S

egation in the samelicensed by negati

(PE)
Hanako-NOM Mary-to-SIKA magazine-ACC pass-NI go-NEG-PRES ‘Hanako goes to pass magazines only to Mary.’
b.
Hanako-ga Mary-ni-sika zassi-o watasi-te ika-na-i.
i

o

(SE)
Hanako-NOM Mary-to-SIKA magazine-ACC pass-TE go-NEG-PRES ‘Hanako passes magazines only to Mary and goes (somewhere).’
In (74a and b), the embedded verb is adjacent to the matrix verb. The NPI sika attached to the dative argument of theembedded verb is licensed by the matrix negation. I assume that the accusative Case in these examples is Case-valuedby the matrix v. As Tsujimura's (1993) crucial examples are those in which the two verbs are adjacent, which means thatthey can satisfy the adjacency requirement, we cannot draw any conclusions based on her original examples. Theunambiguous non-restructuring examples discussed in section 2 show that Tsujimura's (1993) suggestion cannot becorrect.

6. Conclusion

I have argued for the following two conclusions: (i) there are (at least) three types of restructuring infinitives inJapanese, which is consistent with Wurmbrand's (2001) approach to restructuring infinitives and (ii) there is a general banon adjunction to complements of lexical restructuring verbs, which is best explained by an interaction of spell-out domainsand Case-valuation. I have also argued that this ban regulates adverb insertion, adjective insertion, and quantifier raising.This paper further confirms contextual approaches to phases on which phasehood of a phrase is determined contextually(see Bobaljik and Wurmbrand, 2005; Boskovic, 2010; den Dikken, 2007; Takahashi, 2010, 2011, among others), and thecounter-cyclic nature of adjunction (cf. Stepanov, 2001).

The proposed analysis also has implications for the status of Case. The analysis indicates that Case plays a crucialrole in the syntax. Case determines phases (see Miyagawa, 2011; Takahashi, 2010, 2011), and Case of arguments insome context forces movement of the arguments. This is inconsistent with the approaches that push Case outside of thesyntax (see Marantz, 1991, among others).

n Japanese: Adjunction, clausal.006

clause. However, in (ii) there is a clausaln in the same clause.




Acknowledgements

I am very grateful to three Lingua reviewers for their extremely helpful suggestions on earlier versions of this paper. Iwould like to thank Jonathan Bobaljik, Zeljko Boskovic, Jairo Nunes, Mamoru Saito, Daiko Takahashi, Susi Wurmbrand,and the audiences at the University of Connecticut and PLC 32 for valuable comments and suggestions. My thanks alsogo to Masahiko Aihara, Miloje Despic, Jon Gajewski, I-Ta Chris Hsieh, Taichi Nakamura, Toshiko Oda, Koichi Otaki,Tsuyoshi Sawada, Serkan Sener, Yoshiyuki Shibata, Hisako Takahashi, Kensuke Takita, and Lyn Shan Tieu for theircomments, criticisms, and/or judgments at various stages. I am responsible for any errors and misunderstandings.

References

Abels, K., 2003. Successive cyclicity, anti-locality, and adposition stranding. Ph.D. dissertation. University of Connecticut.Asano, Y., 2007. Restructuring in Japanese revisited: a phrasal movement analysis of purpose expressions. In: Scheffler, T., et al. (Eds.),

Proceedings of the 30th Annual Penn Linguistics Colloquium. Penn Linguistics Club, University of Pennsylvania, Philadelphia, pp. 15--28.Bhatt, R., 2005. Long-distance agreement in Hindi--Urdu. Natural Language and Linguistic Theory 23, 757--807.Bobalijk, J., 1995. Morphosyntax: the syntax of verbal inflection. Ph.D. dissertation. MIT.Bobaljik, J.D., Wurmbrand, S., 2005. The domain of agreement. Natural Language and Linguistic Theory 23, 809--865.Bobaljik, J.D., Wurmbrand, S., 2007. Complex predicates, aspect, and anti-reconstruction. Journal of East Asian Linguistics 16, 27--42.Bobaljik, J.D., Wurmbrand, S., 2012. Word order and scope: transparent interfaces and the 3/4 signature. Linguistic Inquiry 43, 371--421.Boskovic, Z., 1994. D-structure, theta-criterion, and movement into theta-positions. Linguistic Analysis 24, 247--286.Boskovic, Z., 2004. Be careful where you float your quantifiers. Natural Language and Linguistic Theory 22, 681--742.Boskovic, Z., 2007a. On the locality and motivation of Move and Agree: an even more minimal theory. Linguistic Inquiry 38, 589--644.Boskovic, Z., 2007b. Agree, phases, and intervention effects. Linguistic Analysis 33, 54--96.Boskovic, Z., 2010. Phases beyond clauses. Ms. University of Connecticut.Cardinaletti, A., Giusti, G., 2001. ‘Semi-lexical’ motion verbs in Romance and Germanic. In: Corver, N., Riemsdijk, H.v. (Eds.), Semi-lexical

Categories: The Function of Content Words and the Content of Function Words. Mouton de Gruyter, Berlin/New York, pp. 371--414.Chomsky, N., 2000. Minimalist inquiries: the framework. In: Martin, R., et al. (Eds.), Step by Step: Essays onMinimalist Syntax in Honor of Howard

Lasnik. MIT Press, Cambridge, MA, pp. 89--151.Chomsky, N., 2001. Derivation by phase. In: Kenstowicz, M. (Ed.), Ken Hale: A Life in Language. MIT Press, Cambridge, MA, pp. 1--52.Chomsky, N., 2004. Beyond explanatory adequacy. In: Belletti, A. (Ed.), Structures and Beyond: The Cartography of Syntactic Structures, vol. 3.

Oxford University Press, Oxford, pp. 104--131.Chomsky, N., 2008. On phases. In: Freidin, R., et al. (Eds.), Foundational Issues in Linguistic Theory: Essays in Honor of Jean-Roger Vergnaud.

MIT Press, Cambridge, MA, pp. 133--166.Cinque, G., 2006. Restructuring and Functional Heads. Oxford Studies in Comparative Syntax: The Cartography of Syntactic Structures. Oxford

University Press, Oxford/NewYork.den Dikken, M., 2007. Phase extension: contours of a theory of the role of head movement in phrasal extraction. Theoretical Linguistics 33, 1--41.Fitzgibbons, N., 2010. Licensers and meanings: structural properties of dependent indefinites. Ph.D. dissertation. University of Connecticut.Fox, D., Nissenbaum, J., 1999. Extraposition and scope: a case for overt QR. In: Bird, S., et al. (Eds.), Proceedings of the 18th West Coast

Conference on Formal Linguistics. Cascadilla Press, Somerville, MA, pp. 132--144.Futagi, Y., 2004. Japanese focus particles at the syntax--semantics interface. Ph.D. dissertation. Rutgers University.Goro, T., 2007. Language-specific constraints on scope interpretation in first language acquisition. Ph.D. dissertation. University of Maryland.Grimshaw, J., Mester, A., 1988. Light verbs and theta-marking. Linguistic Inquiry 19, 205--232.Grohmann, K., 2000. Prolific peripheries: a radical view from the left. Ph.D. dissertation. University of Maryland.Harada, S.-I., 1973. Counter equi NP deletion. Annual Bulletin 8, 113--147 (Reprinted in Syntax and Meaning: S.I. Harada Collected Works in

Linguistics. In: Fukui, N. (Ed.), Taishukan, Tokyo, 2000).Harada, S.-I., 1975. The functional uniqueness principle. Attempts in Linguistics and Literature 2, 17--24 (Reprinted in Syntax and meaning: S.I.

Harada Collected Works in Linguistics. In: Fukui, N. (Ed.), Taishukan, Tokyo, 2000).Harley, H., 2008. On the causative construction. In: Miyagawa, S., Saito, M. (Eds.), The Handbook of Japanese Linguistics. Oxford University

Press, Oxford, pp. 20--53.Hiraiwa, K., 2010. Spelling out the double-o constraint. Natural Language and Linguistic Theory 28, 723--777.Hoshi, H., 2006. Functional categories and configurationality. Memoirs of the Faculty of Education and Human Studies, Akita University 61, 1--38.Kageyama, T., 1993. Bunpou to Gokeisei (Grammar and Word Formation). Hituzi Syobo, Tokyo.Kitagawa, C., Ross, C., 1982. Prenominal modification in Chinese and Japanese. Linguistic Analysis 9, 19--53.Koizumi, M., 1994. Nominative objects: the role of TP in Japanese. In: Koizumi, M., Ura, H. (Eds.), MIT Working Papers in Linguistics 24: Formal

Approaches to Japanese Linguistics, vol. 1. Department of Linguistics and Philosophy, MIT, Cambridge, MA, pp. 211--230.Koizumi, M., 1995. Phrase structure in minimalist syntax. Ph.D. dissertation. MIT.Koizumi, M., 1998. Remarks on nominative objects. Journal of Japanese Linguistics 16, 39--66.Koizumi, M., 2008. Nominative object. In: Miyagawa, S., Saito, M. (Eds.), The Handbook of Japanese Linguistics. Oxford University Press, Oxford,

pp. 141--164.Kratzer, A., 1996. Severing the external argument from its verb. In: Rooyck, J., Zaring, J.L. (Eds.), Phrase Structure and the Lexicon. Kluwer,

Dordrecht, pp. 109--137.Kuno, S., 1973. The Structure of the Japanese Language. MIT Press, Cambridge, MA.Kuo, P.-J., 2009. IP-internal movement and topicalization. Ph.D. dissertation. University of Connecticut.Kuroda, S.-Y., 1965. Generative grammatical studies in the Japanese language. Ph.D. dissertation. MIT.





Kuroda, S.-Y., 1988. Whether we agree or not: a comparative syntax of English and Japanese. Lingvisticae Investigationes 12, 1--47.Kurogi, A., 2002. Nihongo keidooshikoobun-no keitaitougoron-teki bunseki (Morphosyntactic analysis of Japanese light verb constructions).

Master's thesis. Tohoku University.Lasnik, H., 1999. Chains of arguments. In: Epstein, S.D., Hornstein, N. (Eds.), Working Minimalism. MIT Press, Cambridge, MA, pp. 189--215.Lasnik, H., Saito, M., 1991. On the subject of infinitives. In: Dobrin, L.M., et al. (Eds.), Papers from the 27th Regional Meeting of the Chicago

Linguistic Society. Chicago Linguistic Society, pp. 324--343.MacDonald, J., 2006. The syntax of inner aspect. Ph.D. dissertation. Stony Brook University, New York.Marantz, A., 1991. Case and licensing. In: Westphal, G., Ao, B., Chae, H.-R. (Eds.), Proceedings of Eastern States Conference on Linguistics, vol.

91. Cornell Linguistics Club, Cornell University, Ithaca, NY, pp. 234--253.Martin, S., 1975. A Reference Grammar of Japanese. Yale University Press, New Haven.Matsumoto, Y., 1996. Complex Predicates in Japanese: A Syntactic and Semantic Study of the Notion ‘Word’. CSLI Publications, Stanford.May, R., 1985. Logical Form: Its Structure and Derivation. MIT Press, Cambridge, MA.Miyagawa, S., 1987. Restructuring in Japanese. In: Imai, T., Saito, M. (Eds.), Issues in Japanese Linguistics. Foris Publications, Dordrecht, pp.

273--300.Miyagawa, S., 1989. Light verbs and the ergative hypothesis. Linguistic Inquiry 20, 659--668.Miyagawa, S., 2011. Genitive subjects in Altaic and specification of phase. Lingua 121, 1265--1282.Murasugi, K., Hashimoto, T., 2004. Three pieces of acquisition evidence for the v-VP Frame. Nanzan Linguistics 1, 1--19.Nakatani, K., 2004. Predicate concatenation: a study of the V-te-V predicate in Japanese. Ph.D. dissertation. Harvard University.Napoli, D., 1981. Semantic interpretation vs. lexical governance: clitic climbing in Italian. Language 57, 841--887.Nishigauchi, T., 1993. Long distance passive. In: Hasegawa, N. (Ed.), Japanese Syntax in Comparative Grammar. Kuroshio, Tokyo, pp. 79--114.Nomura, M., 2003. The true nature of nominative objects in Japanese. In: Kaier, E., Arunachalam, S. (Eds.), Proceedings of the 26th Annual Penn

Linguistics Colloquium. Penn Linguistics Club, University of Pennsylvania, Philadelphia, pp. 169--183.Nomura, M., 2005. Nominative Case and AGREE(ment). Ph.D.dissertation. University of Connecticut.Nunes, J., 2004. Linearization of Chains and Sideward Movement. MIT Press, Cambridge, MA.Oh, S., 2005. Plurality markers across languages. Ph.D. dissertation. University of Connecticut.Postal, P.M., 1974. On Raising: One Rule of English Grammar and Its Theoretical Implications. MIT Press, Cambridge, MA.Pylkkänen, L., 2002. Introducing arguments. Ph.D. dissertation. MIT.Pylkkänen, L., 2008. Introducing Arguments. MIT Press, Cambridge, MA.Rochette, A., 1988. Semantic and syntactic aspects of Romance sentential complementation. Ph.D. dissertation. MIT.Rochette, A., 1990. On the restructuring classes of verbs in Romance. In: Di Sciullo, A.M., Rochette, A. (Eds.), Binding in Romance: Essays in

Honour of Judith McA’Nulty. Canadian Linguistic Association, Ottawa, Ontario, pp. 96--128.Rosen, S.T., 1989. Argument structure and complex predicates. Ph.D. dissertation. Brandeis University.Rosen, S.T., 1990. Restructuring verbs are light verbs. In: Halpern, A. (Ed.), Proceedings of the 9thWest Coast Conference on Formal Linguistics.

CSLI Publications, Stanford, pp. 477--491.Saito, M., 2005. Further notes on the interpretation of scrambling chains. In: Sabel, J., Saito, M. (Eds.), The Free Word Order Phenomenon.

Mouton de Gruyter, Berlin, pp. 335--376.Saito, M., 2006a. Subjects of complex predicates: a preliminary study. In: Kawamura, T., et al. (Eds.), Stony Brook Occasional Papers in

Linguistics, vol. 1. Department of Linguistics, Stony Brook University, pp. 172--188.Saito, M., 2006b. Expletive replacement reconsidered: evidence from expletive verbs in Japanese. In: Brandt, P., Fuss, E. (Eds.), Form, Structure,

and Grammar: A Festschrift Presented to Günter Grewendorf on Occasion of his 60th Birthday. Akademie Verlag, Berlin, pp. 255--273.Saito, M., Hoshi, H., 1998. Control in complex predicates. In: Report of the Special Research Project for the Typological Investigation of

Languages and Cultures of the East and West University of Tsukuba, pp. 15--46.Saito, M., Hoshi, H., 2000. Japanese light verb construction and the Minimalist Program. In: Martin, R., et al. (Eds.), Step by Step: Essays on

Minimalist Syntax in Honor of Howard Lasnik. MIT Press, Cambridge, MA, pp. 261--295.Saito, M., Murasugi, K., 1999. Subject predication within IP andDP. In: Johnson, K., Roberts, I. (Eds.), Beyond Principles and Parameters: Essays

in Memory of Osvaldo Jaeggli. Kluwer, Dordrecht, pp. 167--188.Sano, M., 1985. LF movement in Japanese. Descriptive and Applied Linguistics 18, 245--259.Sells, P., 1988. More on light verbs and theta-marking. Ms. Stanford University.Shibatani, M., 1973. Semantics of Japanese causativization. Foundations of Language 9, 327--373.Shibatani, M., 1978. Nihongo no bunseki (Analysis of Japanese). Taishuukan, Tokyo.Shibatani, M., 2007. Grammaticalization of converb constructions: the case of Japanese -te conjunctive constructions. In: Rehbein, J., et al.

(Eds.), Connectivity in Grammar and Discourse. John Benjamins, Amsterdam, pp. 21--49.Shoji, A., 1986. Dake and sika in Japanese: syntax, semantics and pragmatics. Ph.D. dissertation. Cornell University.Stepanov, A., 2001. Late adjunction and minimalist phrase structure. Syntax 4, 94--125.Sugioka, Y., 1984. Interaction of derivational morphology and syntax in Japanese and English. Ph.D. dissertation. University of Chicago.Tada, H., 1992. Nominative objects in Japanese. Journal of Japanese Linguistics 14, 91--108.Takahashi, M., 2010. Case, phases, and nominative/accusaitive conversion in Japanese. Journal of East Asian Linguistics 19, 319--355.Takahashi, M., 2011. Some theoretical consequences of Case-marking in Japanese. Ph.D. dissertation. University of Connecticut.Takezawa, K., 1987. A configurational approach to Case-marking in Japanese. Ph.D. dissertation. University of Washington.Tanaka, H., 1997. Invisible movement in Sika-Nai and the linear crossing constraint. Journal of East Asian Linguistics 6, 143--178.Terada, M., 1990. Incorporation and argument structure in Japanese. Ph.D. dissertation. University of Massachusetts, Amherst.Teramura, H., 1984. Nihongo-no sintakusn-to Imi II (Syntax and Semantics of Japanese II). Kuroshio Shuppan, Tokyo.Tomioka, N., 2006. The interaction of between restructuring and causative morphology in Japanese. In: Gurski, C. (Ed.), Proceedings of the 2006

Annual Conference of the Canadian Linguistic Association.Travis, L., 2010. Inner Aspect: The Articulation of VP. Springer, Dordrecht.Tsujimura, N., 1990. Ergativity of nouns and Case assignment. Linguistic Inquiry 21, 277--287.





Tsujimura, N., 1993. Adjuncts and event argument in restructuring. In: Choi, S. (Ed.), Japanese/Korean Linguistics, vol. 3. CSLI, Stanford, pp.121--136.

Uchida, Y., Nakayama, M., 1993. Japanese verbal noun constructions. Linguistics 31, 623--666.Ura, H., 1996. Multiple-feature checking: a theory of grammatical function splitting. Ph.D. dissertation. MIT.Ura, H., 1999. Checking theory and dative subject constructions in Japanese and Korean. Journal of East Asian Linguistics 8, 223--254.Ura, H., 2000. Checking Theory and Grammatical Functions in Generative Grammar. Oxford University Press, Oxford.Uriagereka, J., 1999. Multiple spell-out. In: Epstein, S.D., Hornstein, N. (Eds.), Working Minimalism. MIT Press, Cambridge, MA, pp. 251--282.Wurmbrand, S., 2001. Infinitives: Restructuring and Clause Structure. Mouton de Gruyter, Berlin.Wurmbrand, S., 2004. Two types of restructuring---Lexical vs. functional. Lingua 114, 991--1014.Wurmbrand, S., 2007. How complex are complex predicates? Syntax 10, 243--288.Wurmbrand, S., 2008. Word order and scope in German. In: Zwart, J.-W. (Ed.), Groninger Arbeiten zur Germanistischen Linguistik, vol. 46.

University of Groningen, Groningen, pp. 89--110.Wurmbrand, S., Bobaljik, J.D., 2005. Adjacency, PF, and extraposition. In: Broekhuis, H., et al. (Eds.), Organizing Grammar: Linguistic Studies in

Honor of Henk van Riemsdijk. Mouton de Gruyter, Berlin, pp. 679--688.Yumoto, Y., 2004. Fukugoudoushi Haseidoushi-no Imi-to-Tougo (Syntax and Semantics of Complex Verbs and Derived Verbs). Hituzi Syobo,

Tokyo.Zushi, M., 1995. Long-distance dependencies. Ph.D. dissertation. McGill University.Zushi, M., 2008. Some remarks on the lexical nature of restructuring predicates. English Linguistics 25, 340--363.



Documents

On restructuring infinitives in Japanese: …ling.umd.edu/assets/publications/Takahashitoappear.pdfAdjunction, clausal architecture, and phases Masahiko Takahashi* University of Maryland,