Short Answers: Ellipsis, Connectivity, and Island Repair
Taisuke Nishigauchi Tomohiro Fujii
Kobe Shoin Graduate School University of Maryland
February 13, 2006
Running head: Short Answers: Ellipsis, Connectivity, and Island Repair
Article type: Theoretically oriented work on syntax and syntax-semantics in-terface, comparative syntax
Correspondence: Taisuke NishigauchiKobe Shoin Graduate School1–2–1 Shinohara-Obanoyama, Nada-ku, Kobe 657–0015 JapanPhone: +81–(0)78–882–8762 • Fax: +81–(0)78–882–5032
Tomohiro FujiiDepartment of Linguistics1401 Marie Mount Hall, University of MarylandCollege Park, MD 20742–7505Phone: 301–405–7002 • Fax: 301–405–7104
1
Short Answers: Ellipsis, Connectivity, and Island Repair
Abstract
The theme of the present article is the nature of short answers. On this analysis,
short answers are derived from the result of focus movement, followed by deletion
of everything except the focus. Various aspects of connectivity associated with short
answers are captured in terms of focus movement. This analysis also shows that the
derivation of short answers involve ellipsis. On the other hand, not all short answers
exhibit properties related with movement and ellipsis: We show that short answers have
another source, which we will call the bare-copular frame. However, functional answers
and pair-list answers cannot be derived from the latter source and exhibit connectivity
and diagnostics of ellipsis, MaxElide in particular.
Issues of wh-questions involving (apparent) violations of the relative clause island
constraint will be re-examined at length. It is shown that short answers to such wh-
questions derive from focus constructions which indeed involve the relevant violations
but are saved by island-repair due to ellipsis. Functional answers provide compelling
evidence for this analysis.
1 Introduction
The theme of the present article is the nature of short answers. By ‘short answers’ we mean
answers to questions exemplified by (1b), as opposed to full sentential answers like (1a).
(Merchant 2004 uses the term ‘fragment answers’.)
(1) A. Where did John buy the book?
B. a. He bought it in Washington DC.
b. In Washington DC.
We are going to show that at least some species of short answers exhibit properties
associated with a complete sentence fully reflecting the syntactic and semantic properties
of the question sentence purporting to solicit the answer. In this sense, we proceed on the
premise that there is a significant parallelism between question and answer both in structure
and meaning.
1 Introduction 3
Thus, one subtheme of the nature of short answers must be captured by the notion of con-
nectivity. While the notion of connectivity has been discussed in a variety of contexts, we will
be exclusively concerned with this notion in the sense that Higgins (1973), Akmajian (1970),
among others, discussed it in their analyses of specificational (pseudo-)cleft constructions.
Using Hiraiwa and Ishihara’s (2002) analysis as theoretical apparatus, in which specificational
clefts can be subsumed under focus constructions, our analysis captures various aspects of
connectivity in terms of focus movement.
On this analysis, short answers are derived from the result of focus movement, followed
by deletion of everything except the focus.
(2) XPi [ . . . ti . . . ]︸�������︷︷�������︸
⇓∅
In this light, our conception of short answers is that they involve movement and ellipsis. But
does this process really show the properties of movement and ellipsis?
We will provide several crucial arguments in favor of the supposition that short answers
are derived by movement and ellipsis:
1. Case-matching phenomena
2. Quantifier scope
3. Binding phenomena
4. MaxElide
5. Island repair
In this connection, discussion of the covert syntax of wh in-situ languages is very impor-
tant. On the one hand, these languages allow the occurrence of a wh-phrase inside a relative
clause, which is universally considered to form an island. If covert movement applies in such
a way as to move this wh-phrase outside the relative clause, it must be free from the relative
clause island effect, as was claimed by Huang (1982). Consider the following example from
Japanese.
4 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
(3) Minna-ga [Akira-ga doko-de tor-ta syasin]-o mi-ta ka osiete.
all-Nom Akira-Nom where-at take-Past picture-Acc see-Past Q tell me
‘Lit. Tell me: Everyone saw a picture that Akira had taken where?’
A short answer like (4) apparently supports this idea.
(4) Tokyo(-de) desu.
Tokyo-at Cop
‘(In) Tokyo.’
However, in Nishigauchi (1990) and subsequent work, it has been shown that these languages
have a strategy, called (Large Scale) LF-pied-piping (LFPP), which moves the entire relative
clause as if it were a wh-constituent. Nishigauchi (1990) points out the acceptability of the
following type of answer as an argument for this claim.
(5) (Kare-ga) Tokyo-de tor-ta {syasin / no}(-o) desu.
he-Nom Tokyo-at take-Past picture / one Cop
‘The picture/one that he had taken in Tokyo.’
If such a strategy exists, the conclusion that the covert syntax of wh in-situ is free from the
relative clause island constraint is not warranted.
A substantial portion of the present article is devoted to a re-examination of this issue.
We suggest that there is a possibility that the acceptability of (4) is due to island-repair, a
hallmark of ellipsis. However, a short answer like (4), involving an individual name, might
have some other sources and derivations, which are independent from connectivity and
ellipsis.
Functional answers eliminate such indeterminate factors. These answers must derive
from focus constructions and show genuine properties of ellipsis.1 The behavior of functional
answers indeed presents very important evidence that the acceptability of (4) is backed by
island repair, which implies that wh-movement in covert syntax is free from the relative
clause island constraint.2
1Pair-list answers also require strict connectivity. We discuss pair-list answers at length in our other owrk[reference suppressed].
2In our other article [reference suppressed] it is shown that some processes in LF are in fact subject to theconstraint in question.
2 Focus and connectivity 5
The present article is organized as follows. First, in section 2 we will discuss the syntax
of focus and cleft constructions using Hiraiwa and Ishihara’s (2002) analysis, which plays
a very important role as analytical machinery. After that, in sections 3 through 5, we will
attempt to demonstrate the two subthemes of the nature of short answers: connectivity and
ellipsis. That is to say, we will show the various aspects of connectivity exhibited by short
answers, and then try to establish that ellipsis is indeed involved in the derivation of short
answers. MaxElide plays an important role here. In sections 6 and 7, we discuss various
issues having to do with the relative clause island and LFPP, showing that island-repair plays
a crucial role.
2 Focus and connectivity
The main claim of this article is that derivation of short answers involves focus movement
(which underlies cleft formation) and ellipsis, drawing on relevant data from Japanese. As a
preliminary to this goal, we discuss focus and cleft constructions in this section.
2.1 Focus and cleft constructions
In the analysis of short answers that we will present here, the analysis of focus and cleft con-
structions proposed by Hiraiwa and Ishihara (2002) plays a very important role as analytical
machinery.
Hiraiwa and Ishihara (2002) claim that cleft constructions are one species of focus con-
structions. On this analysis, cleft constructions are derived from what they call ‘No da’ in-situ
focus constructions, sentences ending in the nominalizer (or complementizer) no followed
by the copula da, with one or more of their constituents receiving focus interpretation (both
semantic and phonological).
(6) Taro-ga kono ringo-o kaw-ta no da.
Taro-Nom this apple-Acc buy-Past C Cop
‘It is that Taro bought this apple.’ or ‘It is this apple that Taro bought.’
Hiraiwa and Ishihara (2002) are not specific on this matter, but we assume that the focused
element carries the feature [+Foc], which has obvious consequences in PF, although our
attention in the present article is focused on its nature in the narrow syntax.
6 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
Assuming Rizzi’s (1997) theory of articulated CP architecture, Hiraiwa and Ishihara
(2002) consider the complementizer no as the head of the Fin(ite) projection, and copula
da as the head of the Foc(us) projection. Cleft constructions are derived from ‘No da’ in-
situ constructions via the following two steps. First, the focused constituent is moved to
SpecFocP.
(7) TopP
(Topic)
FocP
(Focus)
FinP
TP
. . . XP[+Foc] . . .
Fin(-no)
Foc(-da)
Top
This is obviously to check the [+Foc] feature associated with XP. If we stop the derivation
here, we obtain the following focus construction.
(8) Kono ringo-o Taro-ga kaw-ta no da.
this apple-Nom Taro-Nom buy-Past C Cop
‘This apple, Taro bought.’
This is indistinguishable from a sentence which has undergone scrambling. Alternatively, the
following, in which the constituent fronted by focus movement is marked by the nominative
ga, which also serves as the focus marker, can be thought of as the output of focus movement.3
3For some reason which we do not fully understand, the complementizer /nominalizer -no is not congenialhere, although it does not lead to full ungrammaticality. A nominal expression like mono ‘thing’ is much preferred.
(i) Kono ringo-ga Taro-ga kaw-ta mono da.this apple-Nom Taro-Nom buy-Past thing Cop‘It is this apple that Taro bought.’
Hiraiwa and Ishihara (2002) exclude this type of construction, where the presupposition portion is headed by anelement other than no, from their domain of focus constructions. However, sentences like the following showthe property of connectivity, a hallmark of specificational statements.
2 Focus and connectivity 7
(9) Kono ringo-ga Taro-ga kaw-ta no da.
this apple-Nom Taro-Nom buy-Past C Cop
‘It is this apple that Taro bought.’
Then, the remnant of the FinP, headed by -no, is moved to SpecTopP. Although Hiraiwa and
Ishihara (2002) do not mention this, we take it that the copula -da, the head of FocP, is raised
to Top prior to the movement of FinP to SpecTopP — this makes SpecTopP equidistant to
SpecFocP with respect to FinP. This operation yields the presupposition portion of a cleft
sentence.
(10) TopP
(Topic)
FocP
(Focus)XP FinP
TP
. . . tXP . . .
Fin(-no)
Foc(-da)
Top
Thus, we get the following cleft sentence, with (6) as the source.
(11) Taro-ga kaw-ta no wa kono ringo(-o) da.
Taro-Nom buy-Past C Top this apple-Acc Cop
‘It is this apple that Taro bought.’
Given this analysis, the second part of the procedure is only necessary to derive a ‘regular’
cleft construction, one kind of focus constructions.
(ii) Zibun(zisin)-no syasin-ga Taro-ga ki-ni-it-te iru (yuitu-no) mono da.self-Gen photo-Nom Taro-Nom fond be only thing Cop‘A photo of himself is the (only) thing he likes.’
We will discuss this aspect of focus constructions shortly.
8 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
2.2 Binding connectivity
We can make several empirical arguments that cleft formation involves Focus movement,
i.e. focalized material starts off inside the TP and moves to SpecFocP, and not, for example,
that the process involves base-generation of the focused element in the focus position. Here
we present two observations (see Hiraiwa and Ishihara 2002 for other arguments). First, the
cleft construction displays connectivity related with binding.
One nice consequence of the analysis portrayed in the previous subsection, not mentioned
by Hiraiwa and Ishihara (2002) themselves, is that it captures straightforwardly the two
salient properties of the specificational (pseudo-)cleft constructions (Akmajian 1970, Higgins
1973): Connectivity and reversibility .
One of the important properties of cleft constructions often discussed in the literature
since Akmajian (1970), Higgins (1973), among others, is that sentences like the following are
ambiguous, having the specificational interpretation and the predicational interpretation.
(12) a. What John is is important.
b. What John ate for supper was the cat’s food.
On the predicational reading, (12a) means that John is a certain X (he has a position or
occupation, etc.) and that X or being X is important. On the specificational interpretation,
this is a statement about John, not about the position or occupation he has, and it simply
says that John is important. On the predicational reading of (12b), John might have eaten a
tuna steak for supper, whose leftover was fed to the cat. On the specificational reading, John
opened a can and ate the cat’s food for supper.
It has been noted since Akmajian (1970), Higgins (1973), among others, that reversing the
(surface) subject-predicate portions of (12) yields only the specificational interpretation.
(13) a. Important is what John is.
b. The cat’s food was what John ate for supper.
These sentences can only mean John is important and John ate the cat’s food for supper,
respectively.
2 Focus and connectivity 9
The distinction between the predicational and specificational uses of cleft constructions is
reflected on the various phenomena related to connectivity, such as binding, agreement, and
a host of other syntactic and semantic properties. Simply put, specificational constructions
exhibit many if not all of the properties associated with connectivity, while predicational
constructions do not show connectivity.
Thus, the following sentences, with the indicated coindexation, do not show the ambigu-
ity in question.
(14) a. What Johni is is important to himselfi. (Specificational)
b. What Johni is is important to himi. (Predicational)
These binding phenomena suggest that specificational cleft sentences involve some syntactic
operations on a ‘source’ structure Johni is important to himselfi to account for the connectivity of
binding, while the predicational sentences are essentially what they look like on the surface,
with a ‘headless relative clause’ in the subject position.
Since the sentences in (14) are unambiguous, reversing the (surface) subject-predicate
order yields different results.
(15) a. Important to himselfi is what Johni is.
b. *Important to himi is what Johni is.
Now, Hiraiwa and Ishihara’s (2002) analysis of focus constructions can be utilized to
account for the syntactic properties of cleft constructions involving connectivity and re-
versibility of specificational cleft constructions. We will show how.
Let us start out with the FinP of the form: John is important to himself, where the bold
faced portion bears the feature [+Foc(us)]. Restricting ourselves to its nature in the narrow
syntax, the constituent bearing [+Foc] can be moved to SpecFocP to check the feature of its
head. Since FocP is a layer of the CP domain, movement occurring in this domain will have
the A′-properties, among them being successive cyclic. Thus, this constituent will move, first
to SpecFinP (via adjunction to VP and/or vP), and then to SpecFocP.
10 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
(16) FocP
(Focus)
FocCopula
FinP
←(what) IP
John is important to himself[+Foc]︸�����������������������������︷︷�����������������������������︸
This is how we get (15): Important to himself is what John is. We assume, as a possibility, that
what in SpecFinP is a spell-out of the copy left by movement of the focused constituent along
the way to SpecFocP. (Similarly, for -no in Japanese.)
We may choose to stop the derivation here, yielding the ‘reverse’ pseudo-cleft construction
(15), or choose to move on. If we choose to move on, what we can do is to move the remnant
of FinP, let us assume that this is marked for a topichood feature, to SpecTopP.
(17) TopP
(Topic)
Top FocP
(Focus)
important to himself[+Foc]
FocCopula
FinP
(what) IP
John is t[+Foc]
While this movement is not to the closest possible landing site, viz. SpecFocP, movement of
Copula to the head of TopP makes SpecTopP equidistant to SpecFocP. Thus this movement
is legitimate in light of the MLC.
2 Focus and connectivity 11
This analysis is close in spirit to Heggie (1988), who takes the ‘reverse’ cleft as the source
structure for the ‘regular’ cleft: We regard the ‘reverse’ cleft construction as closer to the
source structure than the ‘regular’ cleft, for our analysis says that the derivation of the reverse
cleft involves only one syntactic operation, viz. Focus movement, while the derivation of the
‘regular’ cleft involves additional movement of Topic.
The connectivity and reversibility properties also show up in focus constructions in
Japanese. Connectivity in cleft constructions has been discussed at length by Kizu (2005),
among others. This property is observed in sentences like the following.
(18) Minna-ga sonkei-si-teiru no wa zibun(-zisin)-no hahaoya(-o) da.
all-Nom admire-be-Pres C Top self-Gen mother-Acc Cop
‘{??What / the person} everyone admires is his or her mother.’
Although on the surface, the quantifier minna ‘all, everyone’ does not c-command the reflex-
ive zibun(-zisin), the latter can be coindexed with the quantifier.
In keeping with Hiraiwa and Ishihara’s (2002) analysis of focus constructions, we derive
(18) from the source sentence (19), in which the quantifier does c-command the reflexive
within the same clause.
(19) Minna-ga zibun(-zisin)-no hahaoya-o sonkei-si-teiru no da.
all-Nom self-Gen mother-Acc admire-be-Pres C Cop
‘It is that everyone admires his or her mother.’
With movement of the bold-faced constituent of (19) to SpecFocP, we obtain either of the
following. (See footnote 3 relating to the pre-copular element in (20b).)
(20) a. Zibun(-zisin)-no hahaoya-o minna-ga sonkei-si-teiru no da.
self-Gen mother-Nom all-Nom admire-be-Pres C Cop
b. Zibun(-zisin)-no hahaoya-ga minna-ga sonkei-si-teiru {??no /hito} da.
self-Gen mother-Nom all-Nom admire-be-Pres C person Cop
Notice that these are the reversed versions of the cleft sentence (18), and they can be paralleled
by either of the following English sentences.
12 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
(21) a. It is his or her mother that everyone admires.
b. His or her mother is {??who / the person} everyone admires.
Thus, under Hiraiwa and Ishihara’s (2002) analysis of focus constructions, reversibility of
specificational cleft constructions comes with connectivity, simply by focus movement. The
cleft sentence (18) is obtained by movement of the remnant of FinP, the portion headed by
-no in (20).
2.3 Scope reconstruction
Scope reconstruction presents another set of facts indicating connectivity exhibited by fo-
cus/cleft constructions.
In (22), three songs can be interpreted as having narrow scope with respect to the subject
QP, so each of the boys might have sung different three songs from the other.
(22) Hutari-no syoonen-ga samba-o san-kyoku utaw-ta.
2-CL-Gen boy-Nom samba-Acc 3-CL sing-Past
‘(The) two boys sang three sambas.’
In this light, consider the cleft sentence (23).
(23) Hutari-no syoonen-ga utaw-ta no wa samba-o 3-kyoku desu
2-CL-Gen boy-Nom sing-Past C Top samba-Acc 3-Cl Cop
‘It is three sambas that (the) two boys sang.’
Sentence (23) allows the interpretation in which three sambas takes scope narrower than two
boys. The point is made clear when this sentence is contrasted with a sentence involving
head-external relativization. As pointed out by Hoshi (2004), head-external relativization in
Japanese does not allow scope reconstruction, (see also Bianchi 1999, Aoun and Li 2003).
(24) Boku-wa [hutari-no syoonen-ga utaw-u] samba-o 3-kyoku kiita.
I-Top 2-CL-Gen boy-Nom sing-Pres samba-Acc 3-CL hear-Past
‘I listened to (the) three sambas that the two boys sang.’
Sentence (24) does not allow the narrow scope reading of three sambas; that is, either the same
three sambas were sung by the boys or sung in chorus. If the scope interaction is determined
3 Short answers as ellipsis 13
by c-command at LF, the availability of the narrow scope reading of the object QP in (23)
strongly argues for a derivation that we are proposing. The focus material is reconstructed
into the TP at LF, which is possible because it has a trace somewhere below the subject at
some point of the derivation.
3 Short answers as ellipsis
3.1 The canonical case
In the present analysis, we consider short answers as concealed focus constructions, derived
by (i) focus movement of the constituent corresponding to the wh-phrase in the question —
the first step in the derivation of a cleft construction in Hiraiwa and Ishihara’s (2002) analysis,
and (ii) deletion of the clause headed by no, FinP in the articulated CP system, also following
Hiraiwa and Ishihara (2002). The idea that short answers derive from focus has also been
advocated by Kuwabara (1997) and Saito (2004), though our technical implementation of the
idea will differ from theirs.
(25)
⇓∅
(i)
(ii)
FocP
(Focus)
FinP
TP
. . . XP[+Foc] . . .
Fin(-no)
Foc(-da)
A similar analysis is adopted by Merchant (2004), with some difference in the phrase structure
organization.
There is a further possibility that deletion targets (the remnant of) FinP after it moves to
SpecTopP, thereby effecting a cleft construction, along the lines of Hiraiwa and Ishihara (2002).
We have no empirical reason to deny the possibility that movement to the SpecTopP applies
before deletion takes place. Intuitively, this literally means that short answers derive from
cleft constructions, as has been claimed by Saito (2004). We simply assume that movement of
14 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
FinP to TopP is optional, but it is worth stressing that the difference between the two possible
analyses is surprisingly small.
3.2 Identity in ellipsis
Consideration of the notion of identity is indispensable in a theory of ellipsis.
For the present purposes, we assume the influential view of Merchant (2001), making
crucial reference to “semantic isomorphism”. Along the lines advocated by Merchant (see
also Schwarzschild 1999 and references cited therein), ellipsis of FinP is licensed only if the
semantic content of the antecedent clause and that of the FinP entail each other. For example,
the FinP in (26) can be elided.
(26) A. [FinPMari-ga dare-o karakaw-ta] no?
Mari-Nom who-Acc make-fun-Past Q
‘Who did Mari make fun of?’
B. <Mari-ga e karakaw-ta no wa> otooto desu.
Mari-Nom make-fun-Past C Top brother Cop
‘It was (her little) brother that Mari made fun of.’
In this instance, the inference (27) must obtain as a prerequisite for the deletion of the FinP
of the antecedent clause, where the existential quantifier binds the argument position of
the verb make-fun-of, ignoring tense. (At this point, it does not matter whether the in-situ
wh-phrase actually moves or is interpreted without movement. But see section 7.)
(27) ∃x.make-fun-of(m, x)
The interpretation of FinP of (26B) also permits the inference spelled out as (27). Hence the
antecedent entails the elided site, and vice versa.
If this mutual entailment relation is not met, ellipsis is not allowed, as illustrated by (28)
as reply to (26A).
(28) B. a. [FinPMari-ga e tibi-to iw-ta] no wa otooto desu.
Mari-Nom shorty call-Past C Top brother Cop
‘It was (her little) brother that Mari called Shorty.’
4 Short answers and connectivity 15
b. [FinPMari-ga tibi-to iw-ta] no wa otooto desu.
Mari-Nom shorty call-Past C Top brother Cop
‘(It was) (her little) brother (that Mari called Shorty).’
The potential antecedent does not entail the elided site: Calling somebody Shorty entails
making fun of that person, but the other way around does not hold. For more detailed
discussion on this subject, the reader is referred to Merchant (2001, chapter 1).
4 Short answers and connectivity
Merchant (2004) points out the following connectivity effects exhibited by short answers
(‘fragment answers’ in his terms). We will discuss some of those connecivity effects here.
4.1 Case-matching
The morphological case form of a short answer DP is the same as the case found on the
corresponding DP in a fully sentential answer.
(29) Whose car did you take?
a. *John.
b. John’s.
In Japanese, the morphological realization of structural case is retained in short answers.
(30) A. Mari-ga nani-o kaw-ta no?
Mari-Nom what-Acc buy-Past Q
‘What did Mari buy?’
B. Ringo-o 3-ko desu.
apple-Acc 3-Cl Cop
‘Three apples.’
Since accusative case needs to be assigned by a verb, (30B) has to contain the case assigner
in the structure, even though it is not pronounced.
In Japanese, the morphological case markers -ga (nominative) and -o (accusative) are gen-
erally omitted in short answers, the retention of these case markers leading to unacceptability
with varying degrees.
16 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
(31) A: Dare-ga kawari-ni kuru no?
who-Nom instead come Q
‘Who is coming instead?’
B: a. Taro(??-ga) desu.
Taro-Nom Cop
b. Kawari-ni kuru no wa Taro(??-ga) desu.
instead come C Top Taro-Nom Cop
‘It’s Taro that’s coming instead.’
(32) A. Taro-wa nani-o tabe-te-ru no?
Taro-Top what-Acc eat-ing Q
‘What is Taro eating?’
B. a. Susi(*-o) desu.
sushi(-Acc) Cop
b. Taro-ga tabe-te-ru no wa susi(*-o) desu.
Taro-Nom eat-ing C Top sushi-Acc Cop
‘It is sushi that Taro is eating.’
The omission of the case markers is what is observed in the focus constituent in the pre-
copuar position, as seen in the cleft constructions in (31b) and (32b). The retention of the
case-marker in this position leads to unacceptability just to the extent that the corresponding
short answers with the case-marker are unacceptable. The parallel pattern as seen in (31)
and (32) can be taken as telling evidence that short answers are derived from focus (or cleft)
constructions plus deletion.
As has been pointed out in the literature (e.g. Koizumi 2000, Kuroda 1992), inserting an
element such as a floating quantifier between XP-ga/o and the copular improves the status of
the sentence, as is the case in (30).
4 Short answers and connectivity 17
4.2 Binding connectivity
Phenomena having to do with binding in general provide strong evidence for connectivity
shown by short answers.
4.2.1 Bound pronouns
As seen in the following examples from Merchant (2004), short answer (33a) contains a
pronoun bound by a quantifier contained in the question.
(33) A. Who does every Englishmani admire?
B. a. Hisi mother.
b. Every Englishmani admires hisi mother.
This is straightfowardly captured if we derive (33a) from a full sentential form (33b). Also,
(33a) is an instance of functional answer, also referred to as relational answer (Engdahl 1986,
1988). A functional answer, as applied to each individual in the relevant domain of discourse,
yields a different individual as a referential value. For discussion of functional answers, cf.
Chierchia (1991), Hornstein (1995), among many others. Functional answers play a crucial
role later in this article.
The following examples from Japanese show the same point.
(34) A. Minnai-ga dare-o syootai-suru ka osiete.
all-Nom who-Acc invite-Pres Q tell me
‘Tell me who everyone invites.’
B. a. Zibun-zisini-no kazoku-o hutari-zutu desu.
self-Gen family-Acc 2-Cl each Cop
‘Two of his or her family members each.’
b. Minnai-ga syootai-suru no wa zibun-zisini-no kazoku-o
all-Nom invite-Pres C Top self-Gen family-Acc
hutari-zutu desu.
2-Cl each Cop
‘It is two of his or her family members each that everyone invites.’
18 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
(35) A. Kaku syai-ga dare-o syootai-suru ka osiete.
each company-Nom who-Acc invite-Pres Q tell me
‘Tell me who each company invites.’
B. a. Sokoi-no kabunusi-o hutari-zutu desu.
it-Gen stockholder-Acc 2-Cl each Cop
‘Two of its stockholders each.’
b. Kaku syai-ga syootai-suru no wa sokoi-no kabunusi-o
each company-Nom invite-Pres C Top self-Gen stockholder-Acc
hutari-zutu desu.
2-Cl each Cop
‘It is two of its stockholders each that each company invites.’
The reflexive in (34a) and the bound pronominal in (35a) pattern with those occurring in the
cleft constructions (34b) and (35b), respectively. Assuming with the standard practice that
variable binding requires c-command by a quantificational antecedent, the functional answer
given in (34a) and (35a) constitutes evidence for ellipsis. If there is no structure containing
the quantifier in the short answer at any point of the derivation, then the availability of the
bound reading would be mysterious. In our analysis, this binding possibility is accounted
for because the binder is available in the clause to undergo ellipsis.
4.2.2 Condition C and B effects
Merchant (2004) discusses cases like (36) and (37), which show the obviation effects show up
with short answers, as is expected under the ellipsis analysis
(36) Condition C:
Where is hei staying?
a. *In Johni’s apartment.
b. *Hei’s staying in Johni’s apartment.
(37) Condition B:
Who did Johni shave?
4 Short answers and connectivity 19
a. *Himi.
b. *Hei shaved himi.
It is easy to show the condition A effect in short answers and corresponding focus and
cleft constructions in Japanese.
(38) A. Taro-wa dare-o hihan-si-ta no?
Taro-Top who-Acc criticize-Past Q
‘Who did Taro criticize?’
B. a. Zibun(-zisin) desu.
self Cop
b. Kare-ga hihan-si-ta no wa zibun(-zisin) desu.
he-Nom criticize-Past C Top self Cop
‘It was himself that he criticized.’
In Japanese, the situation with obviation involving pronominals is not so straightforward at
first glance. Imagine that we are talking about a puppy Max.
(39) A. Karei-o doko-e ture-te iku no?
it-Acc where-to take go Q
‘Where (are you going to) bring him?’
B. a.??Maxi-no uti-e desu.
Max-Gen home-to Cop
‘To Max’s home.’
b. ?*Karei-o ture-te iku no wa Maxi-no uti-e desu.
he-Acc take go C Top Max-Gen home-to Cop
‘It is to Max’s home that we are taking him.’
The acceptability of short answer (39a) is slighltly higher than that of the corresponding cleft
(39b), at least to some speakers.
A solution to this puzzle may lie in ‘vehicle change’, which we discuss in the next
subsection.
20 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
4.2.3 ‘Vehicle change’ effects
This is apparently a sign of non-connectivity effect, for while (40b) shows a condition C
violation, no corresponding deviance is found in the parallel short answer (40a).
(40) A. Who did you tell t about Billi’s raise?
B. a. Himi.
b. *I told himi about Billi’s raise.
Merchant (2004) claims that this discrepancy is due to the effect called ‘vehicle change’
(Fiengo and May, 1994), which is widely observed in ellipsis phenomena, concluding that
that this effect is found in short answers is a ‘welcome and expected’ outcome, since this can
be taken as additional evidence that ellipsis is involved in the derivation of short answers.
Similar facts can be observed in Japanese.
(41) A. Dare-ni Taroi-no mozi-ga kaidoku dekiru no?
who-Dat Taro-Gen handwriting-Nom decipher can Q
‘Who can decipher Taro’s handwriting?’
B. a. Karei-ni dake desu.
he-Dat only Cop
b. ?*Taroi-no mozi-ga kaidoku dekiru no wa karei-ni dake desu.
Taro-Gen handwriting-Nom decipher can C Top he-Dat only Cop
‘It is only he (him) that can decipher Taro’s handwriting.’
The judgment we obtain in Japanese seems to pattern with that of the English example we
have seen, so the short answer (41a) is considerably better than its fully sentential source
with a condition C violation (41b). A natural explanation here is ‘vehicle change’ by which
the name in (41b) is replaced by a pronoun in the source for ellipsis.
Now consider the following, where he in the question sentence (39) in the previous
subsection is replaced by a name.
(42) A. Maxi-o doko-e ture-te iku no?
Max-Acc where-to take go Q
‘Where (are you going to) bring Max?’
4 Short answers and connectivity 21
B. a. Maxi-no uti-e desu.
Max-Gen home-to Cop
‘To Max’s home.’
b. Maxi-o ture-te iku no wa Maxi-no uti-e desu.
he-Acc take go C Top Max-Gen home-to Cop
‘It is to Max’s home that we are taking Max.’
Our intuition is that (42a) is considerably improved over (39a). Given that the Condition C
effect disappears when the antecedent is a name in Japanese (e.g. Lasnik 1991), the contrast
between (39a) and (42a) can be accounted for if ellipsis takes place; that is, Max in the
former example is bound by the pronoun while Max in the latter is by the name in the cleft
construction (42b).
We consider this as a case of ‘vehicle change’: the pronominal which appeared in the
elided site is now replaced by a name. One might find this reasoning perverse, because in
the usual cases of ‘vehicle change’ discussed in the literature, the name in the elided site is
replaced by a pronominal, so the situation that we are saying is holding in (42) is going in
the opposite direction.
But notice that the way ‘vehicle change’ is stated in Fiengo and May (1994, 218ff.) does
not stipulate directionality:
(43) Nominals can be treated as non-distinct with respect to their pronominal status
under ellipsis.
In Merchant’s (2001) theory, ‘vehicle change’ is subsumed under semantic identity. This
approach does not imply any directionality of ‘vehicle change’ either.
Notice that ‘vehicle change’ as conceived this way does not save the condition C violation
as seen in (36) in the previous subsection, because English does not allow a name to be bound
by a name as in Japanese.
(44) Where is hei staying?
a. *In Johni’s apartment.
b. *Hei’s staying in Johni’s apartment.
22 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
c. *Johni’s staying in Johni’s apartment.
d. *It is in Johni’s apartment that Johni’s staying.
4.3 Scope
We have seen in section 2.3 that cleft sentences (or more precisely examples involving Focus
movement) exhibit scope reconstruction. Short answers also exhibits this property, which is
naturally expected under our approach to short answers. Short answers allow for the same
kind of scopal ambiguity found in fully sentential answers.
(45) A. How many diplomats did every translator greet?
B. a. Three.
b. Every translator greeted three (diplomats).
In (45a), B’s answer has both the scopal possibilities attested in (45b).
The same point can be shown in short answers and cleft constructions in Japanese.
(46) A. Minna-ga samba-o nan-kyoku utaw-ta ka osiete.
all-Nom samba-Acc how many songs sing-Past Q tell me
‘Tell me how many sambas everyone sang.’
B. a. 3-kyoku desu.
3 songs Cop
b. Minna-ga utaw-ta no wa samba-o 3-kyoku desu.
all-Nom sing-Past C Top samba-Acc 3-songs Cop
‘It was three sambas that everyone sang.’
Short answer (46a) has the same scopal ambiguity as the cleft construction (46b) does. So
both of these have the interpretation on which there were three songs that all the relevant
people sang (maybe as a group), or every member of the group sang three songs, possibly
different from other members of the group sang.
If no ellipsis is involved in (46a), namely if it does not derive from a full sentential
structure, it would be hard to imagine how three sambas could interact scopally with the
quantified phrase everyone.
4 Short answers and connectivity 23
4.4 Ellipsis and MaxElide
“MaxElide”, as has been referred to in the current literature (Merchant 2001, Fox and Lasnik
2003, Takahashi and Fox 2005, among others) constitutes an argument that helps establish
that ellipsis is involved in the derivation of short answers.
As has been observed since Lasnik (2001), VP-ellipsis does not apply naturally where
Sluicing can apply. Here is one example.4
(47) They said they heard about a Balkan language, but I don’t know
a. which they heard about.
b. which.
c. *which they did.
The following generalization essentially captures the idea of this concept (see Merchant 2001,
Fox and Lasnik 2003, Takahashi and Fox 2005 for further discussion).
(48) Where XP is a constituent to be elided and YP is also a possible target for deletion,
YP must not properly contain XP.
In (47), the TP is the constituent elided by Sluicing and the VP is a target of deletion
by VP-ellipsis. Since the TP properly contains the VP, (47c), in which VP is deleted in the
presence of a larger constituent TP which properly contains it, is ruled out by (47).
The effect of MaxElide is also observed in ellipsis which is considered to be responsible
for the derivation of short answers in Japanese.5 Consider the following.
(49) A. Hanako-wa [kyoozyu-ga nani-o koogi-su-ru tokoro]-o rokuon sita no?
Hanako-Top prof.-Nom what-Acc lecture-Pres C-Acc tape-rec do-Past Q
‘Lit. What did Hanako tape-record the professor lecturing?’
B. a. Gengogaku-o desu.
linguistics-Acc Cop
4See section 7, especially footnotes 11 and 12.5To our knowledge, Susumu Kuno (Kuno, 1978, 1980) is the first to point out the effect of what is now familiar
as MaxElide (the “Ban against Partial Discourse Deletion” in his terms). The range of data he looked at is muchwider than that we cover here.
24 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
b. Hanako-ga [kyoozyu-ga koogi-su-ru tokoro]-o rokuon sita
Hanako-Nom prof.-Nom lecture-Pres C-Acc tape-rec do-Past
no wa gengogaku-o desu.
C Top linguistics-Acc Cop
‘It was linguistics that Hanako tape-recorded the professor lecturing.’
c. *Gengogaku-o Hanako-ga ∅ rokuon sita no desu.
linguistics-Acc Hanako-Nom tape-rec do-Past C Cop
d. *Hanako-ga ∅ rokuon sita no wa gengogaku-o desu.
Hanako-Nom tape-rec do-Past C Top linguistics-Acc Cop
While (49a), which on our analysis derives from cleft sentence (49b) (which involves no island
violation), is a normal short answer to question (49A), (49c), derived by focus movement and
deletion, is so bad that it is hard to give an English translation. Sentence (49d), a cleft
construction obtained from (49b) by Topicalization of FinP, is equally bad. If anything, they
can only mean ‘Hanako tape-recorded linguistics,’ which hardly makes sense. This shows
the extent to which a violation of MaxElide causes a severe degradation.
To see the point, let us consider how (49a–d) are derived. First, focus movement moves
gengogaku ‘linguistics’.
(50) [FocPlinguisticsi [FP[TPHanako recorded [CPthe prof lecture ti tokoroC]] no] desuCop]
Given this, deletion of FinP leads to the normal short answer (49a).
(51) [FocPlinguisticsi [FinP[TPHanako recorded [CPthe prof lecture ti tokoroC]] no]︸��������������������������������������������������������������������������︷︷��������������������������������������������������������������������������︸
desuCop]
⇓∅
The almost gibberish (49b) is derived as in the following.
(52) [FocPlinguisticsi [FinP[TPHanako recorded [CPthe prof lecture ti tokoroC]︸��������������������������������︷︷��������������������������������︸
] no] desuCop]
⇓∅
4 Short answers and connectivity 25
This is bad, because CP now purporting to be elided is properly contained in FinP, another
potential deletion target. Subsequent movement of FinP to SpecTop yields the cleft (49d),
ending up in an equal near-gibberish.6
Thus the contrast as seen in (49a) and (49c) is readily explained by MaxElide — such an
account is possible only if one supposes that ellipsis is involved in the derivation of short
answers.
Notice that it is not the case that deletion or ellipsis involving circumstantial clauses
headed by tokoro (literally meaning ‘place, spot’) is prohibited. To see this, consider the
following as a possible answer to (49A).
(53) ?Saa. Demo, Hanako-ga ∅ satuei sita no wa
don’t know but Hanako-Nom cam-rec do-Past C Top
gengogaku(-o) desu.
linguistics-Acc Cop
‘Don’t know, but it was linguistics that Hanako cam-recorded (as being lectured).’
We consider this as a significant improvement in comparison with (49d) on the intended
reading on which (the professor’s lecture of) linguistics was cam-recorded. The reason for
this improvement lies in the presence of a verb distinct from the one used in the question.
In fact, this verb needs to be pronounced with stress to obtain the intended interpretation.
The derivation of this answer involves focus movement of gengogaku ‘linguistics’ within the
complement clause, followed by deletion of the remnant clause. MaxElide is irrelevant, and
the sentence is improved as expected.
6In the analysis of Takahashi and Fox (2005), it is crucially necessary to assume that there is no intermediatetrace at the periphery of the tokoro-complement clause, for otherwise this would not constitute a re-bindingenvironment, to which the effect of MaxElide is restricted in their analysis. The fact appears to indicate thatthere is an intermediate trace. As we will discuss in section 5.2, the presence of an intermediate trace makes itpossible for the matrix subject to be the antecedent for the reflexive zibun-zisin which derives from a positionin the complement clause. An answer such as the following to (49A), in which zibun-zisin can be interpreted asHanako, shows that re-binding does not take place here.
(i) Zibun-zisin-no repooto-o desu.self-Gen report-Acc Cop‘Her own term paper.’
However, the intermediate trace itself is not bound within the complement clause, and it is possible to assumethis as causing re-binding.
26 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
5 Short answers and focus movement
Given the familiar assumption that ellipsis only targets constituents, most of the arguments
for ellipsis that we have made in the previous section constitute arguments for a movement
analysis of short answers as well. For instance, our bound variable binding paradigm tells us
that if the carrier of the bound variable were base-generated outside the elided site and, say,
binds a null pronoun in the object position, there would be no chance for the quantificational
antecedent to A-bind into the accusative-marked phrase.
(54) *[FocP selfj’s motheri [FinP everyonej V proi]]
In this section, we will present two more arguments for movement.
5.1 Polarity items
Merchant (2004) discusses the distribution of negative polarity items (NPIs), but this is not
in the context of his discussion of connectivity retained by short answers. As (55) shows, the
NPI any is unable to appear as a short answer.
(55) A. What didn’t Max read?
B. *Anything.
Rather, he discusses this issue as part of his argument for the derivation of short answers
from left-dislocation, for NPIs are unable to appear in a left-dislocated position.
(56) *Anything, Max didn’t read.
In this respect, NPIs in Japanese appear to behave differently. We use sika, whose best
approximate in English would be ‘but’ as in ‘He eats nothing but hamburgers.’7
7Watanabe (2004) extensively discusses negative concord items like nani-mo ‘anything’ in Japanese and theirinteractions with ellipsis, arguing for the semantic isomorphism condition of Merchant (2001). (iB) (from Watan-abe 2004) is a representative example of negative concord, where the antecedent clause does not contain overtnegation but deletion of “I didn’t see” is allowed.
(i) A: Nani-o mita no?what-Acc saw Q‘What did you see?’
B: Nani-mo mi-nakat-tawhat-Mo see-Neg-Past‘(I saw) nothing.’
5 Short answers and focus movement 27
(57) A. Kono neko-wa nani-sika tabe-nai no?
this cat-Top what-sika eat-not Q
Lit. ‘This cat eats nothing but what?’ or ‘Only what does this cat eat?’
B. *Maguro-sika desu.
tuna-sika Cop
Intended: ‘Only tuna.’
Although (57B) is as bad as (55B), left-dislocation of the same NPI is not so bad.
(58) Maguro-sika kono neko-wa tabe-nai no desu.
tuna-sika this cat-Top eat-not C Cop
Lit. ‘This cat eats nothing but tuna.’
We take this as due to the equivocal status of left-dislocation as seen in (58). That is, left-
dislocation in this example may either be focus-movement or scrambling, and it has been
claimed by Ishii (1997) that the restriction on scrambling is less strict than on canonical
varieties of A′-movement, of which we consider focus-movement is an instantiation. It is
highly likely that the acceptability of (58) is due to this aspect of scrambling. On the other
hand, the following cleft construction is very low in acceptability.8
(59) *Kono neko-ga tabe-nai no wa maguro-sika desu.
this cat-Nom eat-not C Top tuna-sika Cop‘What this cat does not eat is but tuna.’
Intended: ‘What this cat eats is nothing but tuna.’
We do not have much to say about this interesting topic here, but two things might be worth mentioning. First,sika-phrases differ from negative concord items in that they are barred as an answer to (i).
(ii) B: Hebi(*-sika)snake-sika
Second, ellipsis of the kind investigated by Watanabe seems to be a different phenomenon from one we find inshort answers. Compare (i) and (iii).
(iii) B: *Nani-mo desuwhat-Mo Cop
When the copula is attached to the fragment seen in (iB), no acceptable answer to the question is obtained. If thepresence of the copula is an indication that clefting is involved, the status of (iiiB) suggests that fragments thatWatanabe is looking at do not involve cleft or focus movement in our sense.
8Facts like this are also discussed in Kizu (2005).
28 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
We take this as a strong piece of evidence that the focus element of cleft constructions is
what short answers derive from. Given that the cleft construction is a subspecies of focus
constructions, we continue to assume that short answers derive from focus.
Notice that a short answer without sika is possible as an answer to (57A).
(60) Maguro desu.
tuna Cop
But this answer is arguably from a source other than a focus construction, and must be from
the bare-copular frame discussed in 5.3, which does not observe connectivity. We have two
arguments for this. One is that (57A) does not allow a functional answer.
(61) *Zibun-no emono desu.
self-Gen catch, game Cop
‘Its own catch.’
As we will see in subsection 5.4, a short functional answer must be derived from a focus
construction, not from a bare-copular frame.
Second, short answers exemplified by (60) do not retain a postposition used in the ques-
tion, another hallmark of the absence of connectivity.
(62) A. Sono mise-wa dono miti-kara sika ike-nai no?
that store-Top which road-from sika go-Not QLit. ‘One can go to the shop from no way but which way?’
‘Only from which way can you go to the store?’
B. Kita-gawa-no miti(??-kara) desu.
north-side-gen road -from Cop
‘(From) the north side way.’
For some reason which we do not understand, B’s answer sounds even better without the
postposition. This might be taken as suggesting that this answer is only derived from a
bare-copular frame.
5 Short answers and focus movement 29
5.2 Binding intermediate positions
In section 2, we posited focus movement as an essential ingredient in the derivation of
focus and cleft constructions. This is a movement of a focused constituent to SpecFocP, a
layer of the CP domain. One expectation arising from this is that focus movement, being a
movement in the CP domain, has the properties associated with A′-movement, among them
being successive cyclic. While this was already shown to be the case in section 2, we will
consider additional cases to confirm this point.
Consider the following discourse, in which the question sentence involves a wh-phrase
in the embedded clause:
(63) Minnai-ga [Akiraj-ga dare-o sonkei-si te-iru to] iw-ta ka osiete.
all-Nom Akira-Nom who-Acc admire be that say-Past Q tell me
‘Tell me who everyone says Akira admires?’
(64) a. Mother Teresa-o desu.
Mother Teresa-Acc Cop
‘Mother Teresa.’
b. Zibun-zisini/j-no hahaoya-o desu.
self-Gen mother-Acc Cop
‘His/her mother.’
Of interest to us is the functional answer in (64b). The reflexive that appears in this short
answer can be coindexed either with the matrix subject or the complement subject.
Notice that the anaphor zibun-zisin requires local binding, and does not allow binding by
an antecedent across the clause.
(65) Minnai-ga [Akiraj-ga zibun-zisin∗i/j-no hahaoya-o sonkei-si te-iru to] iw-ta.
all-Nom Akira-Nom self-Gen mother-Acc admire be that say-Past
‘Everyone said Akira admires his own mother.’
So the ambiguity of the reflexive in (64b) would be unexpected if we suppose, for example,
that this short answer derives directly from (65) by deletion. In that case, we would expect
that the anaphor would be bound only by the embedded subject.
30 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
On the other hand, along the lines of the present analysis in which short answers are de-
rived from focus/cleft constructions, which in turn involve focus movement in their deriva-
tion, the ambiguity in question is just what is expected. The following cleft construction,
which we claim is the source of (64b), is also ambiguous.9
(66) Minnai-ga [Akiraj-ga sonkei-si te-iru to] iw-ta no wa zibun-zisini/j-no
all-Nom Akira-Nom admire be that say-Past C Top self-Gen
hahaoya-o desu.
mother-Acc Cop
‘It was his own mother that everyone said Akira admires.’
We take the following, obtained by focus movement, as the representation for the possible
functional answer.
(67) [[selfi/j’s mother] everyonei said [([selfi/j’s mother]) Akiraj admires ([selfi/j’s mother)]]]
In this structure, the reflexive can be licensed either in the complement object position, in
which case it is bound by Akira, or in the intermediate SpecCP position, where it can be
bound by the matrix subject.
This result is consistent with our hypothesis that focus movement, involved in the deriva-
tion of focus/cleft constructions, shares one property associated with A′-movement: succes-
sive cyclicity.
5.3 Interlude: Non-movement source of short answers
Up to this point, we have developed an analysis of short answers in terms of focus movement
and ellipsis. However, there are cases where we find some discrepancies between short
answers and fully sentential focus/cleft constructions in Japanese. In this subsection, we will
consider why.
One case of discrepancy has to do with postpositions. The postpositions are retained
9Kizu (2005) claims that only the matrix subject can be the antecedent for the anaphor in cleft constructionslike (66). While we agree that binding by the matrix subject is dominant, we consider binding by the complementsubject also possible.
Recall that we have independently shown in section 4.3 that focus movement can start off inside the complementclause, based on scope reconstruction.
5 Short answers and focus movement 31
in the focus position — the same postposition is found in a short answer as seen in a fully
sentential answer.
(68) A. Doroboo-ga doko-kara okane-o nusun-da no?
thief-Nom where-from money-Acc steal-Past Q
‘Where did the thief steal the money from?’
B. a. Kono ginkoo(-kara) desu.
this bank-from Cop
‘From this bank.’
b. Doroboo-ga okane-o nusun-da no wa kono ginkoo(-kara) desu.
thief-Nom money-Acc steal-Past C Top this bank-from Cop
‘It was from that bank that the thief stole the money.’
As the parentheses in answer (68a) suggest, the postposition -kara ‘from’ appears to be
optional in short answers. This is also reflected on the cleft construction (68b), in which the
postposition on the focus constituent appears to be optional. This runs counter to our basic
premise that a specificational cleft retains connectivity.
Hoji (1990) argues that a cleft construction with its focus constituent lacking a postposition
has a derivation distinct from those involving focus with a postposition. Saito (2004) takes
the following, with pro as the gap in the presupposition clause, as the structure for clefts with
focus without a case-marker or postposition.
(69) [NP[TP . . . pro . . . ]]-wa NP-da.
-Top -Cop
Since, on the usual understanding, pro is not generated by movement, construction (69),
which Saito (2004) refers to as ‘bare NP cleft’, is base-generated. From this, it is expected that
this construction lacks the type of connectivity observed in (specificational) cleft construc-
tions. Thus, the cleft construction without a postposition (68b) can be thought to derive from
(69).
However, later in the paper, Saito (2004, 43) refutes the use of structure (69) for his analysis
of short answers involving bare NPs, because the subject of the short answer not deriving
32 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
from a specificational cleft is indeterminate.
Instead, Saito (2004) suggests that a ‘bare NP’ short answer has an alternative source,
which has a general form:
(70) pro XP da/desu
Cop
in which pro is an empty version of the pronominal sore ‘it’ vaguely referring to the circum-
stance related to the event depicted in the sentence. We call (70) the bare-copular frame.
We will not attempt to present a precise analysis of the structure of (70), except to note
that it is presumably a case of deep anaphora in the sense of Hankamer and Sag (1976),
where pro is base-generated in SpecTopP. The fact that this frame exhibits a certain degree of
connectivity in such a way that, for instance, the identical postposition is used in the question
and answer in (68), exclusive of other postpositions which may be semantically synonymous
(if such exists), suggests that some copying operation at the discourse-level takes place in the
formation of XP in (70).
5.4 Controlling variability
In the previous subsection, we have seen that short answers have at least two possible sources
and derivations. One is from focus or specificational cleft plus ellipsis, the other from bare-
copular forms. The former type is faithful to connectivity and shows the behavior of ellipsis,
while the latter does not observe connectivity and shows no sign of syntactic movement. But
are all short answers ambiguous having variable sources?
Our answer is no. Functional answers and pair-list answers require that they derive from
focus/cleft constructions plus deletion. This point can be straightforwardly demonstrated by
the following examples, where functional answers and pair-list answers are ungrammatical
in the bare-copular frame (70).
(71) A. Minna-ga dare-kara meeru-o uke-tor-ta ka osiete.
all-Nom who-from e-mail-Acc receive-Past Q tell me
‘Tell me who everyone received an e-mail from.’
5 Short answers and focus movement 33
B. a. Sore-wa Koizumi-san-kara desu.
it-Top Mr. Koizumi-from Cop
b. *Sore-wa soitu-no zyoosi-kara desu.
it-Top the guy-Gen superior Cop
‘It was from his or her boss.’
c. *Sore-wa Taro-ga Koizumi-san-kara, Mari-ga Abe-san-kara desu.
it-Top Taro-Nom Mr. Koizumi-from Mari-Nom Mr. Abe-from Cop
‘It was, Taro from Mr. Koizumi, Mari from Mr. Abe.
That pair-list answers are incompatible with the bare-copular frame (70) is also noted by
Saito (2004).
In section 4.1, we pointed out that the retention of the morphological case-marker in
a short answer is a hallmark of connectivity, subject to deletion in pre-copular positions.
Postpositions are generally retained in short answers (and focus positions of focus/cleft
constructions as well) even in pre-copular positions. However, answer (72a) to question
(71A) suggests that retention of the postposition is optional, also noted in the previous
subsection. The fact is, the optionality of retention of the postposition only applies to
individual answers, and the functional answers and pair-list answers show the obligatory
retention of the postposition.
(72) B. a. Koizumi-san(-kara) desu.
Mr. Koizumi-from Cop
b. Soitu-no zyoosi*(-kara) desu.
the guy-Gen superior Cop
‘From his or her boss.’
c. Taro*(-ga) Koizumi-san(-kara), Mari*(-ga) Abe-san*(-kara) desu.
Taro-Nom Mr. Koizumi-from Mari-Nom Mr. Abe-from Cop
‘Taro from Mr. Koizumi, Mari from Mr. Abe.’
While the presence of the postposition is optional in the individual answer (72a), its presence
is obligatory in the functional answer (72b), its omission leading to ungrammaticality. In the
34 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
pair-list answer (72c), the nominative case-marker ga, which is normally omitted in focus
constituents and short answers, is obligatorily retained. While the postposition kara ‘from’ is
optional in the first conjunct of the pair-list answer, its presence is obligatory in the second
conjunct. This is a general pattern of ellipsis found in Japanese, such as the equivalent of
Gapping in Japanese.
Short functional answers will play important roles in the remainder of our discussion.
This type of answer is faithful to connectivity, which arguably comes from its syntactic source
and derivation: functional answers unequivocally derive from focus/cleft constructions.10
6 LF Pied-Piping and short answers
6.1 The pied-piping analysis
Question sentences like (73), in which a wh-phrase appears inside a relative clause, have
given rise to an important issue in the current linguistic theory.
(73) Minna-ga [Akira-ga doko-de tor-ta syasin]-o mi-ta ka osiete.
all-Top Akira-Nom where-at take-Past picture-Acc see-Past Q tell me
‘Lit. Everyone saw a picture that Akira had taken where?’
(74) a. (Kare-ga) Tokyo-de tor-ta {syasin / no}(-o) desu.
he-Nom Tokyo-at take-Past picture / one Cop
‘The picture/one that he had taken in Tokyo.’
b. Tokyo(-de) desu.
Tokyo-at Cop
‘(In) Tokyo.’
The acceptability of answers like (74b) has been taken to indicate the lack of the effect of
relative clause islands in Chinese, Japanese, Korean and other ‘wh in-situ’ languages. Huang
(1982) specifically argued that no island constraint applies to LF-movement and subsequently
approaches have been developed under which some non-movement operation licenses wh
in-situ (Reinhart 1998, Tsai 1994, among others).10We discuss syntactic properties of pair-list answers at length in our other work [reference suppressed].
6 LF Pied-Piping and short answers 35
On the other hand, the fact that (74a) is a possible answer to questions like (73) is one of
the reasons that led Nishigauchi (1990) to hypothesize that in Japanese (and other languages
in which wh-questions exemplified by (73) are permissible) the entire complex NP containing
the wh-phrase can be moved and occupy the operator-position at LF. (‘Large-scale Pied-
Piping at LF.’) This idea has later been modified by Watanabe (1992) so that the wh-feature
of the entire complex NP, instead of the whole NP itself, is moved to the operator position
either in overt syntax or at LF. We’ll call this the LF Pied-Piping (LFPP) approach.
According to this latter approach, the large constituent containing the wh-phrase, which
itself moves inside the relative clause island, thereby making it [+wh], is moved to SpecCP
in the covert syntax.
(75) [CP [DP[+wh]a pic that Akira took where[+wh]]︸�������������������������������������������������︷︷�������������������������������������������������︸
[TP everyone saw t]]
The derivation of this structure involves no island violations.
Assuming that short answers involve movement of the answer fragment plus deletion
under identity, the analysis of the question-answer pair, (73)–(74a) would be captured in the
following way: (74a) must be derived by movement of the complex NP out of the FinP to the
Spec-FocP, followed by deletion of the FinP which contains the trace of the large NP. Deleted
material is represented by < . . . > here.
(76) [FocP[DP[+foc]a pic that Akira took in Tokyo[+foc]]︸�����������������������������������������������������︷︷�����������������������������������������������������︸
<[FinP[TP everyone saw t] Cop]>]
The corresponding focus construction, as predicted, is acceptable.
(77) [Akira-ga Tokyo-de tor-ta syasin]-o minna-ga mi-ta no desu.
Akira-Nom Tokyo-in take-Past pic-Acc all-Top see-Past C Cop
‘A picture that Akira took in Tokyo, everyone has seen.’
As expected, corresponding to (74a), the following cleft sentence is also grammatical, again
suggesting the strong parallelism between wh-questions and their answers.
(78) Minna-ga mi-ta no wa [Akira-ga Tokyo-de tor-ta syasin](-o) desu.
all-Nom see-Past C Top Akira-Nom Tokyo-at take-Past picture-Acc Cop
‘Lit. It was a picture that Akira had taken in Tokyo that everyone saw.’
36 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
6.2 Whence short answers?
Given that a LFPP option exists and a ‘pied-piping’ answer (74a) is available, let us ask why
a short answer like (74b) is a possible answer to (73).
Nishigauchi (1990) claimed that short answers like (74b) are obtained by ‘truncating’
the ‘Pied-Piping answers’ like (74a) via some discourse-deletion process that operates in
discourse such as:
(79) Q: Is it the picture that you took in Tokyo?
A: No, in Washington DC.
We’ll turn to this possibility later on.
Now, if we pursue the idea of the present work that short answers are derived by focus
movement plus deletion, it appears that we must derive (74b) from the following focus
construction.
(80) *[Tokyo-de]PP minna-ga [Akira-ga tPP tor-ta syasin]-o mi-ta no desu.
Tokyo-in all-Nom Akira-Nom take-Past pic-Acc see-Past C Cop
‘In Tokyo, everyone has seen a picture that Akira took.’
This sounds like a grammatical sentence, but it is ungrammatical on the intended reading
on which Tokyo was the place where the pictures were taken (not where everyone saw the
pictures), so we cannot view it as the source of (74b), so it seems.
7 Relative clause island repair
7.1 Island repair by deletion
There has been a growing body of literature in which (at least some class of ) island violations
are viewed as PF-phenomena. In this subsection, we will re-examine the problem of short
answers presented in the previous subsection in light of island-repair.
Merchant (2004) develops an analysis of the problem posed by sentences like (81), in
which he invokes the PF theory of islands.
(81) They hired someone who speaks a Balkan language — guess which!
7 Relative clause island repair 37
In this theory, island violations are considered to be due to properties of pronounced syntactic
structure, not to constraints on derivation or representations of LF themselves.
According to this theory, intermediate traces created by crossing syntactic islands are
defective and cannot remain in a structure that gets pronounced. These traces are assigned
*, a PF-uninterpretable feature. Alternatively, * may be assigned to XPs that form islands, as
in Fox and Lasnik (2003). If ellipsis can apply, the structure which contains the * feature(s)
are eliminated from the PF object.
As Merchant (2004) proposes (see also Fox and Lasnik 2003, Merchant 2001), this analysis
not only accounts for (81), but also captures the well-known asymmetry between sluicing
and VP-ellipsis:11
(82) They want to hire someone who speaks a Balkan language, but
a. I don’t remember which.
b. *I don’t remember which (language) they do.
Assuming that wh-movement targets every intermediate maximal projection along the way,
deletion of TP (sluicing) and vP (VP-ellipsis) are shown to have different consequences.
11 Note that MaxElide may be violated in (81b), as clear from the discussion given in section 4.4. We wouldlike to mention here one alternative that analyzes the difficulty of VP-ellipsis in cases that involve no island (i.e.cases like (47c), repeated as (i)) as a case of “failure of island repair”.
(i) *They said they heard about a Balkan language, but I don’t know which they did.
Fox and Lasnik propose that their Parallelism requirement for ellipsis forces the wh-phrase in (i) to move inone-fell-swoop fashion, because there is no intermediate binders in the antecedent clause. They argue that thisone-fell-swoop movement causes a Subjacency violation in Chomsky’s (1986) sense, and that deletion of the VP,unlike deletion of the TP, does not help to remove the cause of the Subjacency violation that results from the‘long’ movement.
A couple of questions arise. Is a principle like MaxElide needed independently from island repair? If it is,when is the condition active and when is it not? Why is it active (or inactive) when it is (or when it is not)?We will not address these issues in detail here. The reader is referred to Takahashi and Fox (2005) for relevantdiscussion and a wider range of data. See also footnote 12.
38 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
(83)
TP-deletion eliminates all *-traces
vP-deletion leaves *t′′i
. . . CP
[DPwhich]i C′
C TP
*t′′i TP
they(do) vP
*t′i vP
want to hire [DPsomeone [CPwho speaks ti]]
With the type of approach to islands just outlined in mind, let us turn to Japanese. This
analysis is plausible as an account of the acceptability of the short answer (74b) as a reply to
(73).
(73) Minna-ga [Akira-ga doko-de tor-ta syasin]-o mi-ta ka osiete.
all-Nom Akira-Nom where-at take-Past picture-Acc see-Past Q tell me
‘Lit. Everyone saw a picture that Akira had taken where?’
(74) b. Tokyo-de desu.
Tokyo-at Cop
‘In Tokyo.’
As we have seen above, the ‘full’ focus construction (80) from which (74b) is supposed to be
obtained by deletion displays the island effect. The following cleft construction shows the
same effect.
(84) *[[Minna-ga Akira-ga tor-ta syasin]-o mi-ta no]-wa Tokyo-de desu.
all-Nom Akira-Nom take-Past picture-Acc see-Past C-Top Tokyo-in Cop
Lit. ‘It is in Tokyo that everyone saw a picture [that Akira had taken t].’
Thus the contrast between (74b) and (80)/(84) follows if ellipsis ameliorates the island viola-
tion.
Furthermore, this analysis also predicts, correctly, that argument deletion, which deletes
the remnant of the relative clause from which the answer fragment has moved, yields a bad
7 Relative clause island repair 39
answer to (73).
(85) a. *Tokyo-de (minna-ga) ∅mi-ta no desu.
Tokyo-in all-Nom see-Past C Cop
‘Lit. In Tokyo, everybody saw ∅.’
b. *(minna-ga) ∅mi-ta no wa Tokyo-de desu.
all-Nom see-Past C Top Tokyo-in Cop
‘Lit. It was in Tokyo that everybody saw ∅.’The answer in these sentences only means Tokyo was the place where they saw the pictures,
never where the pictures were taken.
The following structure shows how this analysis makes the distinction between (74b) and
(85) as an answer to (73).
(86)
FinP-deletion eliminates all *-traces
vP-deletion leaves *t′′i
DP-deletion leaves *t′i and *t′′i
FocP
(Focus)[in Tokyo]i FinP
TP
∗t′′i TP
everyone vP
∗t′i vP
VP
DP
pictures [CPAkira took ti]
Vsaw
v(saw)
Fin-no
Foc--desu
As the structure indicates, deletion of FinP erases all the offending traces, and this explains
why the short answer (74b) is an acceptable answer to (73). In contrast, if (85) is derived by
deletion of DP (argument deletion), then both the offending traces created by the movement
40 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
of the answer fragment, *t′i and *t′′i , remain in the structure, which ends up being ungrammat-
ical. Cf. Saito (2004) for the arguments that argument deletion exists in Korean and Japanese.
Alternatively, if the analysis in Otani and Whitman (1991) is correct, what we see in (85) is
VP-ellipsis, with head V raised to v or T, so that either VP or vP is deleted. In this latter
case, *t′′i is left in the structure. In either case, deletion as seen in (85) leaves some offending
trace, and this accounts for their ungrammaticality in contrast to (74b). As the reader must
have noticed, the ungrammaticality of (85) also involves a violation of MaxElide, discussed
in section 4.4.12
7.2 Alternative solutions
There are at least two potential interfering factors with the analysis depicted in subsection
7.1. First, it is not self-evident that the ellipsis site does contain an island to begin with.
Merchant (2001) discusses this possibility for sluicing in English.
Let us first review Merchant’s (2001, chapter 5) answer to the puzzle posed by the fact
12 The unacceptability of cases like (85a,b) can be accounted for as a violation of MaxElide, independently fromisland repair (see section 4.4 and footnote 11). This is perfectly compatible with our story about island repair. Asalluded to in footnote 11, the Fox and Lasnik type theory may treat the bad examples in (49b,c), which involveno island, as instances of failure of island repair. We repeat (49A) and (49Bd) as (iA) and (iB) here.
(i) A. Hanako-wa [kyoozyu-ga nani-o koogi-su-ru tokoro]-o rokuon sita no?Hanako-Top prof.-Nom what-Acc lecture-Pres C-Acc tape-rec do-Past Q‘Lit. What did Hanako tape-record the professor lecturing?’
B. Hanako-ga [kyoozyu-ga koogi-su-ru tokoro]-o rokuon sitaHanako-Nom prof.-Nom lecture-Pres C-Acc tape-rec do-Pastno wa gengogaku-o desu.C Top linguistics-Acc Cop‘It was linguistics that Hanako tape-recorded the professor lecturing.’
Fox and Lasnik report that, in English, when overt A′-movement takes place in the antecedent clause, deletionof a smaller constituent becomes easier. Sentences (iia,b) are from Fox and Lasnik (2003), with their judgments(cf. Takahashi and Fox 2005).
(ii) a. *I know John said that Mary read a certain book, but YOU don’t know which one he did.
b. ??I know which book John said that Mary read, but YOU don’t know which one he did.
We observe that the effect in question seems to hold for Japanese as well. The status of (iB) seems to improvewhen it follows the question in (iii), compared to when it follows (iA).
(iii) Hanako-ga [kyoozyu-ga koogi-su-ru tokoro]-o rokuon sitaHanako-Nom prof.-Nom lecture-Pres C-Acc tape-rec do-Pastno wa nani-o desu ka?C Top what-Acc Cop Q‘What was it that Hanako tape-recorded the professor lecturing?’
We cannot discuss exactly how “MaxElide” works in Japanese question-answer pairs for space reasons. Moreresearch and groundwork needs to be done; see Kuno (1978, 1980) for relevant data from Japanese.
7 Relative clause island repair 41
that sluicing can be grammatical even when its source structure involves an island violation.
Consider a well-known case of the CNPC violation, which we repeat.
(81) They hired someone who speaks a Balkan language — guess which!
Sluicing in (86) is acceptable although the sentence from which it can be derived is ungram-
matical because of the CNPC.
(87) *Guess which Balkan languagei they hired someone who speaks ti.
Merchant’s (2001) answer to this puzzle is that the CNPC may not be violated even when
ellipsis yields a grammatical output. Merchant (2001) argues that the source structure of the
sluiced portion in (81) is the following:
(88) Guess [which s/he speaks]
This structure involves a portion within the island which itself does not constitute an island.
Also, this portion contains an E-type pronoun referring to the head of the relative clause. In
the case of (88), the pronoun s/he is an E-type pronoun that refers to the person who speaks a
certain language.
Although this idea appears to work for the range of the sluicing data involving relative
clause island violations, it does not provide a real solution to the problem posed by the fact
that (74b) is a possible answer to (73) — A non-elliptical answer that has an E-type pronoun
referring to the head of the relative clause in the question and does not involve an island in
the answer, which the type of analysis represented by Merchant (2001) would posit for (73),
does not function as an appropriate answer to the question because it does not satisfy some
of the pragmatic conditions imposed on question–answer pairs.
(73) Minna-ga [Akira-ga doko-de tor-ta syasin]-o mi-ta ka osiete?
all-Nom Akira-Nom where-at take-Past picture-Acc see-Past Q tell me
‘Lit. Everyone saw a picture that Akira had taken where?’
(89) #Tokyo-de Akira-ga sorera-o tor-ta no desu.
Tokyo-in Akira-Nom those-Acc take-Past C Cop
‘It was in Tokyo that Akira took them.’
42 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
So even if (89) provides a good antecedent for the ellipsis site in the sense that it does not
involve an island, that must not be the source for (74b), because it is not a good answer.
The second factor that may interfere with the ‘island repair’ analysis of (74b), which we
will defend, has to do with what we call the bare-copular strategy, discussed in section 5.3.
Examining the absence of the island effect in examples like (74b), Saito (2004) proposes the
possibility that the subject of (74b) is occupied by a phonologically empty pronoun (as in
(90a)) and therefore it does not have to involve either movement or deletion. The empirical
basis of this claim is that an overt version of that pronoun is available in Japanese, sore “it”,
which is illustrated in (90b):
(90) a. pro Tokyo-de desu.
Tokyo-at Cop
b. Sore-wa Tokyo-de desu.
it-Top Tokyo-at Cop
(90b) is acceptable as an answer to question (73), as Saito observes based on the same type of
example. It is worth stressing here that when the derivation of (74b) utilizes the bare-copular
strategy, the identity condition for ellipsis is trivially satisfied. The argument for island repair
based on the contrast between (74b) and (80)/(84) is considerably weakened if (90) is a source.
For the sake of discussion, we assume that the null version of sore is available.
Now it is clear from the discussion so far that, in order to show that deletion can fix an
island violation in short answers in Japanese, we have to find an environment where sore/pro
is not available and that short answers do not exhibit the island effect that their non-elliptical
counterpart exhibit in that environment.
Recall, as discussed in section 5.4, functional answers are not compatible with the bare-
copular strategy. Observe that when an answer contains a bound variable inside, the bare-
copular frame with sore cannot be used.
(91) A: Minna-ga John-ga dare-to atta to omotteiru ka osiete.
All-Nom John-Nom who-with meet-Past that think Q tell me
‘Who does everyone think that John saw?’
7 Relative clause island repair 43
B: a. Sore-wa Hanako-to desu.
it-Top Hanako-with Cop
b. *Sore-wa soitu-no hahaoya-to desu.
it-Top the guy-Gen mother Cop
The contrast seen in the a- and b-examples of (91) suggests that the presence of sore prevents
the binder for the bound pronoun from occurring in the structure. Thus functional answers
help us to force the derivation of a short answer to involve deletion.
7.3 Functional answers
Now let us consider (73) again, in light of the possibility of functional answers.
(73) Minna-ga [Akira-ga doko-de tor-ta syasin]-o mi-ta ka osiete.
all-Nom Akira-Nom where-at take-Past picture-Acc see-Past Q tell me
‘Lit. Everyone saw a picture that Akira had taken where?’
(92) a. (Akira-ga) soitui-no kokyoo-de tor-ta {syasin / no} desu.
Akira-Nom the guy-Gen hometown-at take-Past pic / one Cop
‘The picture/one that he had taken in his or heri hometown.’
b. ?Soitui-no kokyoo*(-de) desu.
the guy-Gen hometown-at Cop
‘In his or heri hometown.’
Answer (92a) is a functional answer derived by LFPP, in which, if all the people concerned
are from different areas, each person saw a different picture taken in his or her home town.
The acceptability of this answer corresponds with the acceptability of the following focus
sentence, also derived by ‘large-scale pied-piping’.
44 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
(93) [Akira-ga soitui-no kokyoo-de tor-ta syasin]-o minnai-ga t
Akira-Nom the guy-Gen hometown-at take-Past pic-Acc all-Nom
mi-ta no desu.
see-Past C Cop
‘Lit. It was the picture that Akira had taken in his or heri home town that
everyonei saw.’
The availability of the functional answer (92a) finds its basis on the fact that the pronominal
(epithet) soitu appearing in the complex NP is bound, as desired, by the quantificational
subject minna ‘all, everyone’ in the position designated by t of the cleft sentence (93), from
which (92a) is derived.
The status of short answer (92b), which we find quite good, is very significant. As we will
show shortly, this example provides a very important piece of evidence that island-repair is
the only solution available in the derivation of this short answer.
As the asterisk outside the parentheses indicates, the omission of the postposition -de ‘at’
leads to total ungrammaticality to all speakers. This is a sign of connectivity, and we take it
as indicating that we cannot find the source of (92b) in the bare-copular strategy, discussed
in 7.2. In fact, such a putative source is itself ungrammatical.
(94) *Sore-wa soitui-no kokyoo-de desu.
it-Top the guy-Gen hometown-at Cop
‘It was in his or heri hometown.’
This also shows that the analysis suggested in Nishigauchi (1990), according to which
(92b) is derived from (92a) by a discourse deletion rule, is on the wrong track, for this
approach presupposes what is similar to the bare-copular frame.
The only possible source for (92b) is (80), which we have shown to be ungrammatical for
the violation of the relative clause island even when its focus portion contains an individual
name. As expected, the following focus and cleft sentences are also ungrammatical.
7 Relative clause island repair 45
(95) a. *[Soitui-no kokyoo-de]PP minna-ga [Akira-ga tPP tor-ta syasin]-o
the guy-Gen hometown-at all-Nom Akira-Nom take-Past pic-Acc
mi-ta no desu.
see-Past C Cop
‘In his or her home town, everyone saw a picture that Akira took.’
b. *Minna-ga [Akira-ga tPP tor-ta syasin]-o mi-ta no wa
all-Nom Akira-Nom take-Past pic-Acc see-Past C Top
[soitui-no kokyoo-de]PP desu.
the guy-Gen hometown-at Cop
‘It was in his or her hometown that everyone saw a picture that Akira took.’
Thus, the grammaticality of (92b) can only be accounted for by island-repair, as described
in section 7. We consider this a very important result. This consideration provides a com-
pelling piece of evidence that wh-movement in covert syntax is free from the relative clause
island constraint effect and that the derivation of the relevant answer involves focus or cleft
construction plus ellipsis, where island repair plays a crucial role.
7.4 Scope: another case for island-repair
In section 4.3, we pointed out that retention of relative quantifier scope is an indication
of connectivity effected by movement. Given that this is correct, we can establish another
argument that relative clause island violations may be repaired by ellipsis. Consider (96).
(96) A. Taro-wa [sankasya zen’in-ga nankyoku utaw-u] paatii-o
Taro-Top participant all-Nom how many songs sing-Pres party-Acc
hirai-ta no?
hold-Past Q
Lit. ‘Taro held a party such that all the participants sing how many songs?’
B. a. 3-kyoku desu.
3 songs Cop
46 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
b. *Taro-ga [sankasya zen’in-ga utaw-u] paatii-o hirai-ta no wa
Taro-Top participant all-Nom sing-Pres party-Acc hold-Past C Top
3-kyoku desu.
3 songs Cop
Lit. ‘Taro held a party such that all the participants sing three songs’
c. [Sankasya zen’in-ga 3-kyoku utaw-u] paatii desu.
participant all-Nom 3 songs sing-Pres party Cop
‘A party where all the participants sing three songs.’
The question sentence given in (96A) is to ask the number n such that Taro held a party where
each of the attendees is supposed to sing n-many songs, where the universal quantifier takes
wide scope over the numeral, and the identity of the songs that each attendee chooses does
not matter. The non-elliptical answer in (96b) is hopeless irrespective of the scope relations,
whereas its elliptical version (96b) is fine.
Moreover, the latter answer, like the pied-piping answer in (96c), retains the narrow scope
interpretation of the numeral. If scope reconstruction signals overt movement, the difference
between (96a) and (96b) in acceptability strongly suggests that application of ellipsis remedies
the island violation that would be observed otherwise.
Notice further that here again, the derivation of (96) cannot be from the bare-copular
frame.
(97) Sore-wa 3-kyoku desu.
it-Top 3 songs Cop
This answer cannot have the interpretation on which the numeral quantifier takes scope
narrower than the universal quantifier in the relative clause. It can only be interpreted in
such a way that the total number of songs sung by everyone was three, or the attendees sang
three songs in chorus.
Therefore, short answer (96) must really be derived from a focus or cleft construction
followed by ellipsis, where its acceptability is supported by island repair.
8 Concluding remarks 47
8 Concluding remarks
In this article, we have presented a number of arguments that at least some species of short
answers must be derived from focus or cleft constructions, followed by ellipsis.
Various facts stemming from morphological case-markers and postpositions, binding,
and scope relations provide the evidence of connectivity shown by short answers, showing
that derivation of short answers must come from focus or cleft constructions, in turn derived
by focus movement.
Further, facts related with vehicle change, MaxElide, and island repair have shown that
derivation of short answers involves ellipsis.
It has also been shown that short answers can have at least one possible source other
than focus or cleft plus deletion. Some short answers have as their source what we call the
bare-copular frame, which shows little or no sign of connectivity and ellipsis. However,
it has been shown that functional answers cannot have this latter derivation, and must be
derived from focus and ellipsis.
With this much background, cases involving wh in-situ in relative clause islands were
reconsidered. Short functional answers in these cases have provided compelling evidence
that their derivation in fact involves relative clause island constraint violations (at PF), where
island repair plays a crucial role, although LF pied-piping operation remains as a viable
option.
This paper closes with one very important problem area unaddressed. If we consider
the possibility of getting pair-list answers out of questions like (73) or (98), in which the
universal quantifier is replaced by a wh-phrase, making it a multiple wh-question, the result
is somewhat different.
(73) Minna-ga [Akira-ga doko-de tor-ta syasin]-o mi-ta ka osiete.
all-Top Akira-Nom where-at take-Past picture-Acc see-Past Q tell me.
‘Lit. Everyone saw a picture that Akira had taken where?’
48 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
(98) Dare-ga [Akira-ga doko-de tor-ta syasin]-o mi-ta no?
who-Nom Akira-Nom where-at take-Past picture-Acc see-Past Q
‘Lit. Who saw a picture that Akira had taken where?’
Restricting our attention to pair-list answers, there is a clear contrast between pied-piping
answers and short pair-list answers.
(99) a. Anna-ga (Akira-ga) Moscow-de tor-ta syasin(-o), Johann-ga
Anna-Nom Akira-Nom Moscow-in take-Past pic-Acc Johann-Nom
Berlin-de tor-ta syasin(-o) desu.
Berlin-in take-Past pic-Acc Cop
‘Anna saw the picture that Akira took in Moscow, Johann saw the one that
(Akira) took in Berlin.’
b. *Anna-ga Moscow-de, Johann-ga Berlin-de desu.
Anna-Nom Moscow-in Johann-Nom Berlin-in Cop
‘Anna in Moscow, Johann in Berlin.’
It is impossible to understand (99b) as meaning Moscow and Berlin were the cities in which
the pictures in question were taken, not where they were seen.
We will leave these and other important issues related with LF pied-piping to another
work of ours [reference suppressed].
References
Akmajian, Adrian. 1970. Aspects of the grammar of focus in English. Doctoral Dissertation,
MIT, Cambridge, MA.
Aoun, Josef, and Yen-hui Audrey Li. 2003. Essays on the representational and derivational nature
of grammar: The diversity of Wh-constructions. The MIT press.
Bianchi, Valentina. 1999. Consequences of antisymmetry: Headed relative clauses. Mouton de
Gruyter.
Chierchia, Gennaro. 1991. Functional wh and weak crossover. West Coast Conference on Formal
Linguistics 10:75–90.
References 49
Chomsky, Noam. 1986. Barriers. The MIT Press.
Engdahl, Elisabet. 1986. Constituent questions. Dordrecht: Kluwer.
Engdahl, Elisabet. 1988. Relational interpretation. In Mental representations, ed. R. Kempson,
63–82. Cambridge University Press.
Fiengo, Robert, and Robert May. 1994. Indices and identiy. The MIT Press.
Fox, Danny, and Howard Lasnik. 2003. Successive cyclic movement and island repair: The
difference between sluicing and VP ellipsis. Linguistic Inquiry 34:143–154.
Hankamer, Jorge, and Ivan Sag. 1976. Deep and surface anaphora. Linguistic Inquiry 7:391–
428.
Heggie, Lorie. 1988. The syntax of copular structures. Doctoral Dissertation, University of
Southern California.
Higgins, F. Roger. 1973. The pseudo-cleft construction in English. Doctoral Dissertation,
MIT, Cambridge, MA.
Hiraiwa, Ken, and Shin’ichiro Ishihara. 2002. Missing links: Cleft, sluicing and ‘no da’ con-
struction in Japanese. In The Proceedings of HUMIT 2001. MIT Working Papers in Linguistics,
volume 43, 35–54. Cambridge, MA: MIT.
Hoji, Hajime. 1990. Theories of anaphora and aspects of Japanese syntax. Unpublished,
University of Southern California.
Hornstein, Norbert. 1995. Logical form: From GB to minimalism. Oxford: Blackwell.
Hoshi, Koji. 2004. Remarks on N-final relativization in Japanese. The Hiyoshi Review of English
Studies 44:113–147.
Huang, Cheng-Teh James. 1982. Logical relations in Chinese and the theory of grammar.
Doctoral Dissertation, MIT.
Ishii, Toru. 1997. An asymmetry in the composition of phrase structure and its consequences.
Doctoral Dissertation, University of California, Irvine.
Kizu, Mika. 2005. Cleft constructions in Japanese syntax. New York: Palgrave Macmillan.
50 Short Answers: Ellipsis, Connectivity, and Island Repair (February 13, 2006)
Koizumi, Masatoshi. 2000. String vacuous overt verb raising. Journal of East Asian Linguistics
9:227–285.
Kuno, Susumu. 1978. Danwa-no bunpoo. [Grammar of discourse]. Tokyo: Taishukan.
Kuno, Susumu. 1980. Discourse deletion. In Harvard studies in syntax and semantics, vol. 3, ed.
Susumu Kuno, 1–144. Harvard University.
Kuroda, Shige-Yuki. 1992. Pivot-independent relativization in Japanese. In Japanese syntax
and semantics: Collected papers, 114–174. Kluwer Academic Publishers.
Kuwabara, Kazuki. 1997. On the properties of truncated clauses in japanese. In Researching and
verifying an advanced theory of human languages, ed. Kazuko Inoue, 61–83. Kanda University
of International Studies.
Lasnik, Howard. 1991. On the necessity of binding conditions. In Principles and parameters in
comparative grammar, ed. Robert Freidin, 7–28. The MIT Press.
Lasnik, Howard. 2001. When can you save a structure by destroying it? In Proceedings of the
North East Linguistic Society, volume 31, 301–320. GLSA.
Merchant, Jason. 2001. The syntax of silence: Sluicing, islands, and the theory of ellipsis. Oxford
University Press.
Merchant, Jason. 2004. Fragments and ellipsis. Linguistics and Philosophy 27:661–738.
Nishigauchi, Taisuke. 1990. Quantification in the theory of grammar. Dordrecht: Kluwer
Academic Publishers.
Otani, Kazuyo, and John Whitman. 1991. V-raising and VP-ellipsis. Linguistic Inquiry 22:345–
358.
Reinhart, Tanya. 1998. Wh-in-situ in the framework of the minimalist program. Natural
Language Semantics 6:29–56.
Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of grammar, ed.
L. Haegeman, 281–337. Dordrecht: Kluwer Academic Publishers.
Saito, Mamoru. 2004. Ellipsis and pronominal reference in Japanese clefts. Nanzan Linguistics
References 51
1:21–50.
Schwarzschild, Roger. 1999. GIVENness, AvoidF and other constraints on the placement of
accent. Natural Language Semantics 7:141–177.
Takahashi, Shoichi, and Danny Fox. 2005. MaxElide and the re-binding problem. In SALT
15.
Tsai, Wei-Thein Dylan. 1994. On economizing the theory of A-bar dependencies. Doctoral
Dissertation, MIT.
Watanabe, Akira. 1992. Subjacency and S-Structure movement of wh-in-situ. Journal of East
Asian Linguistics 1:255–291.
Watanabe, Akira. 2004. The genesis of negative concord: Syntax and morphology of negative
doubling. Linguistic Inquiry 35:559–612.