Contrastivity and Ellipsis*
Kirsten Gengel
1. INTRODUCTION
The Pseudogapping construction (in (1)) is generally considered to be a special instance
of VP ellipsis. Like VP ellipsis (in (2)), Pseudogapping involves a finite auxiliary (will), and, like
Gapping (in (3)), a contrastive remnant (Jane). Pseudogapping has previously been attested only
for English (though cf. Kim 1997 for Korean ‘Pseudogapping’, and Gengel to appear for
Scandinavian Pseudogapping).1 It is considered to be a phenomenon of spoken speech, and
occurs most frequently in comparatives (in (4) and (5)).2
(1)
(2)
(3)
(4)
(5)
John invited Sarah, and Mary did _ Jane.
John invited Sarah, but Mary didn’t.
John invited Sarah, and Mary _ Jane.
John gave more books to Susan than Tony did _ to Sarah.
John gave Mary a book and Jane will _ Bill.
This paper aims to provide an alternative to the standard approaches to the analysis of
Pseudogapping (as in John invited Sarah more often than Mary did Jane), and, more generally,
the analysis of other selected types of ellipsis. I develop an analysis based on Merchant’s Efeature (Merchant 2001, 2004), which specifies an element for deletion. Maintaining the
fundamental link of the E-feature to information structure (in terms of Givenness), I claim that
information structure determines the amount of material to be deleted in ellipsis. To this end, I
modify Merchant’s implementation of the E-feature, and combine it with a [+contrastive]-feature
on material that is redundant with respect to the antecedent clause. This [+contrastive]-feature,
placed on syntactic constituents, is responsible for movement operations, if movement is
necessary to derive elliptical structures.
The paper is structured as follows: In section 2, I will present two predominant approaches to
ellipsis and Pseudogapping, namely Lasnik’s (1995, a.o.) Pseudogapping analysis, and
Merchant’s (2001) E-feature analysis of Sluicing, and propose that a combination of the two is
necessary to handle properties of Pseudogapping that Lasnik did not address. Section 3 discusses
* This paper has, in different stages of its development, enormously benefitted from discussions with the following
people: Artemis Alexiadou, Cedric Boeckx, Lisa Cheng, Sandra Chung, Danny Fox, Irene Heim, Norbert Hornstein,
Sabine Iatridou, Kyle Johnson, Howard Lasnik, Audrey Li, Jim McCloskey, David Pesetsky, Norvin Richards, Lisa
Selkirk, and Shoichi Takahashi. Thank you. I also want to thank the audiences at the LingO Conference in Oxford,
the Conference on Ellipsis in St. Etienne, and the Guest Lecture Series in Stuttgart. Special thanks goes to Artemis
Alexiadou, Sabine Iatridou, Howard Lasnik, Winfried Lechner, Tom McFadden, and Sabine Mohr, as well as to
Nadine Aldinger, and Britta Sauereisen. Of course, all remaining errors are my own.
1
Kim (1997) provides a discussion of Korean Pseudogapping. However, the Korean Pseudogapping construction, as
he describes it himself, patterns rather with English VP ellipsis in having an auxiliary and only a single remnant
(either subject or object). Hence, pending further evidence to the contrary, I suggest that Korean Pseudogapping is a
different phenomenon from English and Scandinavian Pseudogapping.
2
Both generalisations go back to Levin (1978, 1986).
the role of the EPP in the Pseudogapping derivation and offers motivation for a feature encoding
contrastivity. In section 4, I present an analysis of different ellipsis types in terms of the E-feature
and a contrastivity feature. Section 5 concludes the paper.
2. A SYNTACTIC ANALYSIS OF PSEUDOGAPPING
2.1. THE ‘STANDARD’ SYNTACTIC ANALYSIS
Jayaseelan (1990) was the first to propose that Pseudogapping is an instance of VPEllipsis which is preceded by movement of the remnant out of the VP. This line of argument has
been further pursued by Lasnik (1995, 1999, and subsequent work) who proposed that this
movement process is not Heavy Noun Phrase Shift, as suggested by Jayaseelan (1990), but
Object Shift. His analysis is spelt out in (6), with the relevant tree structure in (7) (the antecedent
is omitted).
(6)
John invited Sarah, and Mary did [AgroP Janei] [VP invite ti].
(7)
AgrSP
Mary AgrS’
Agr TP
T’
T
AgrOP
did Jane
AgrO’
AgrO
VP
V’
V
NP
invite
tObj
The object moves into the specifier position [Spec,AgrO]. This movement is taken to be
independent of case, as indicated by the fact that instances of prepositional Pseudogapping
remnants (as in (4)) exist. Instead, it is triggered by an EPP-feature on AgrO (Lasnik 1995, 1999),
causing object shift to [Spec,AgrO]. With respect to the verb, Lasnik adopts Koizumi’s (1995)
assumption that all verbs raise overtly in English.3 In Pseudogapping, however, according to
Lasnik, it does not raise but stays inside the VP. While its features are moved and checked, its
3
Johnson (1991) was the first to propose this in connection with object shift. Koizumi (1995) combines this with a
split VP structure, also adopted by Lasnik.
lexical content fails to be pied-piped. Hence, in order to prevent the derivation from crashing at
PF, the VP, including the offending material, is deleted.
The Pseudogapping construction displays optionality, in the sense that non-pseudogapped
versions of Pseudogapping sentences are mostly fine. This is shown in (8) below, where no
deletion has taken place.4
(8)
John invited Sarah more often than Mary invited Jane.
Lasnik’s analysis accounts for this optionality by linking it to the optionality of the EPP feature
(for objects). The EPP feature itself is optional, but if it enters the derivation, it triggers
obligatory movement of the object.
However, there are two questions that Lasnik’s account leaves open.
(i) The Puzzle of the Origin of the EPP
The account fails to motivate the introduction of an optional EPP in
the derivation of the Pseudogapping cases.
(ii) The Focus Puzzle
The approach cannot capture the focal properties of the Pseudogapping remnant, and
its inherent contrastivity.
2.2. THE E-FEATURE ANALYSIS
Merchant (2001) introduces the E-feature in his analysis of Sluicing (in (9)), which is later
extended to other ellipsis types (cf. Merchant 2004). The primary importance of this account for
the purposes of this paper is that it provides a step towards an answer to the Focus Puzzle
mentioned in the previous section.
After a short review of how this feature works in its original definition, I show how the
mechanism can be modified in order to subsume other types of E-features Merchant assumes in
his (2004) paper.
(9)
Jack bought something, but I don’t know what ø.5
The E-feature is defined such that it instructs the grammar to delete the lexical material
situated below the head bearing the E-feature. This deletion takes place on the PF level. The Efeature is based on the notion of E-Givenness, modelled on Schwarzschild’s (1999) Givenness.
E-Givenness introduces the notion of focus in ellipsis structures (cf. also Rooth 1992 in this
respect). The definition of E-Givenness is given in (10), along with the condition on VP ellipsis
in (11) (for details see Merchant 2001).
4
There are exceptions to this generalisation, however. Kennedy & Merchant (2000: 99, fn. 9) point out examples of
attributive comparative deletion, where the non-pseudogapped counterpart was rated less acceptable than the
pseudogapped version:
i.
They sell better shirts than they do ties.
ii.
??They sell better shirts than they sell ties.
5
Merchant (2001), p. 3, his (1a). ø indicates deleted material.
(10) E-Givenness:
An expression E counts as e-GIVEN iff E has a salient antecedent A and,
modulo -type shifting,
(i) A entails F-clo(E), and
(ii) E entails F-clo(A).
(11) Focus condition on VP ellipsis
A VP  can be deleted only if  is e-GIVEN.
Merchant (2001) proposes that the E-feature for Sluicing works as illustrated in (13) and (14). He
assumes that Sluicing is TP deletion, with the elided structure given in (12). The Focus condition
on Sluicing could then be stated as in (11).6
(11’) Focus condition on TP-Deletion (Sluicing)
A TP  can be deleted only if  is e-GIVEN.
(12) Jack bought something, but I don’t know what [Jack bought t].
(13)
...
C
[wh,Q]
TP
T’
Jack
T[E]
VP
V’
V
bought
NP
(what)
(14)
CP
what
C’
C[E]
[wh,Q]
<TP>
Jack bought t
The E-feature in Sluicing is specified as being [+wh,+Q]. It is generated in T (cf. (13)) and moves
up to C (cf. (14)), where its wh-feature is checked. In C, it then instructs its sister node, <TP>, to
delete, yielding the Sluicing example in (12).
6
Concerning the status of TP deletion as deletion of syntactic structure, I refer the reader to the discussion of
possible counterarguments in Chung, Ladusaw & McCloskey (1995).
Essentially, the E-feature thus captures both the semantic requirement on ellipsis (i.e. the
identity condition) and the syntactic locality conditions. In addition, the E-feature has the effect
of phonological deletion; the elided categories contain syntactic structure.
2.3. COMBINING THE TWO APPROACHES
The aim of this section is to show how the E-feature can be combined with the EPP-feature
Lasnik assumed to be at work in Pseudogapping. (The EPP-feature in Pseudogapping will be
argued to be a [+contrastive] feature later.)
Throughout this paper, I presume that the E-feature and the EPP-feature are separate features.
The EPP-feature is placed on a projection above VP (on the Agr head in Lasnik’s terms). On the
most straightforward assumption, I argue that the position for the E-feature is the V head, since
the ‘redundant’ V makes the VP available for deletion.
Placing the E-feature on V in Pseudogapping has exactly the same effects as Merchant’s
account of Sluicing. Here, the E-feature is base-generated in V and moves up to AgrO to get
checked, similar to the Sluicing case, where the E-feature moves from I to C. In AgrO, it is
checked against the EPP feature, as the Sluicing E-feature specified for [+wh] was checked
against the wh-feature present in the C-domain. This results in the desired checking relations.
Moreover, the placement of the E-feature on V derives an important property Lasnik argues to
hold for Pseudogapping, namely, the fact that verbs don’t raise in Pseudogapping. Assuming that
the verb either raises or deletes, we obtain the pattern in (15), where the ‘V-feature’ is the feature
that triggers verb raising in general.7
(15) V-feature
+V
-V
E-feature
-E
+E
The complementarity of V-Raising and VP ellipsis could be derived on the assumption that [E] is
located on the V head and prevents movement of the verb out of the VP. If this movement were
not prohibited, i.e. the connection between [+V] and [-E] not valid, the Pseudogapping
configuration could not be derived. As shown in (16), the verb in AgrO would still permit VP
ellipsis, but the ellipsis site would not include the verb.
Hence, I propose that the Pseudogapping derivation looks as in (17), with the EPP-feature on the
AgrO head and the E-feature on V. The EPP-feature triggers object movement into the specifier
position of AgrO, whilst the E-feature blocks verb movement to the AgrO head, permitting VP
ellipsis.
Unfortunately, I have nothing specific to say about the nature of this ‘V-feature”. Still, the correlation between the
two features seems worth pointing out.
7
(16)
TP
T’
T
AgrOP
AgrO’
did Jane
AgrO
VP
V’
V
invite
V
NP
tObj
(17)
TP
T’
T
AgrOP
did Jane
AgrO’
AgrO
[EPP]
VP
V’
V[E]
NP
invite
tObj
2.4. IS FEATURE RAISING NECESSARY?
One remaining question concerning the structure in (17) is how the EPP-feature and the E-feature
interact, i.e. how the features are checked in the course of the derivation. Merchant (2001, 2004)
assumes featural movement of the E-feature, which would amount to the movement of the Efeature from V to AgrO in the Pseudogapping configuration. In Sluicing, the feature would be
checked against a [+wh,Q] feature (residing in C), and in Pseudogapping, accordingly, against a
[+EPP]-feature in AgrO. However, under current assumptions, it is not necessary to assume
feature movement, since the checking relation could be achieved via an Agree relation.
One argument for moving the E-feature from V to a higher head is the following. If the Efeature stays on V, then, following Merchant, its complement will be deleted. However, this is
not possible in Pseudogapping since the complement would be the object, and deletion of the
object would leave the rest of the VP intact. The obvious way to solve this problem is to follow
Merchant in assuming that the feature moves higher up.
Note, though, that this movement-and-checking account of the E-feature has one
important disadvantage. Merchant (2004) assumes different types of E-features for different
kinds of ellipsis, checking them against different features. Hence, despite the argument for the
movement of the [E]-feature I have just outlined, I would like to pursue a different approach,
which provides a first step towards an analysis that renders this diversification of the E-feature
obsolete.
I propose that the E-feature in Pseudogapping works differently from what Merchant
(2001) proposed for Sluicing: it triggers optional PF deletion of the maximal projection (VP) of
the head where it is placed (V). The consequences of this assumption for other ellipsis types (in
particular, VP ellipsis, Pseudogapping, Sluicing, and Gapping) will be discussed in the following
sections.
To summarize, I assume the following:
(18) REVISED E-FEATURE
The E-feature, placed on the head X of a phrase XP, deletes the full XP.
3. THE EPP AND CONTRASTIVITY
3.1. THE ROLE OF AGROP AND VP
This section discusses the movement operation that moves the object remnant out of the VP. I
will argue that the EPP-feature for objects Lasnik (1995, 1999) proposed in his analysis is
difficult to transfer into a Agr-less system, i.e. a more minimalist phrase structure. I will show
that a functional projection above the vP layer is still necessary, but that the EPP-feature cannot
account for the contrastivity of the remnant in Pseudogapping. I propose that a different feature is
present in Pseudogapping, namely a [+contrastive]-feature, that triggers movement of elements
contrasted with the antecedent out of the phrase specified for deletion.
Consider the structure given in (17), translated into a phrase structure with vP, (19).
[Spec,vP] hosts the EPP feature.
(19)
vP
[Spec,vP]
v’
[EPP] v
VP
V’
V
NP
invite tObj
The main difference between the two structures lies in one basic hypothesis put forward in the
minimalist framework, namely that the subject of the sentence is base-generated in [Spec,vP], as
opposed to [Spec,VP] in Lasnik’s analysis. Hence, movement of the object to [Spec,vP] is illicit,
since [Spec,vP] hosts the trace of the moved subject (cf. (20)).
(20)
vP
tSubj
v’
v
VP
V’
X
V
NP
invite tObj
This problem could be solved with multiple specifiers, as in Chomsky (2001). The object moves
to the outer specifier of vP (21).
(21)
vP
Object
Subject
v’
v VP
V’
V
NP
invite tObj
Unfortunately, while this configuration would be licit to account for the object position in
Pseudogapping, the presence of the object in the outer specifier of vP now poses a problem for
the movement of the subject.
Given that both object and subject are at the vP phase edge, this has undesirable consequences:
(i)
(ii)
It is not clear why the subject should move, and not the object, since both could
potentially be attracted by the probe T.
If the subject moves, the presence of the object in the outer specifier would cause
intervention effects.
Chomsky (2001) proposes a solution to this problem, by assuming that the evaluation of the
lower phase (vP) only takes place in the next higher phase, namely, CP. In combination with a
second assumption, namely, that elements that have no phonological content neither block
movement nor cause intervention effects (as tObj in the outer specifier in (22)), movement of the
subject to [Spec,TP] is no longer problematic.
(22) (guess) whatObj [JohnSubj T [vP tObj [tSubj read tObj]]]
3.2. CONSEQUENCES FOR THE PSEUDOGAPPING DERIVATION
In Chomsky’s account, the crucial factor allowing the derivation to go through is the fact that the
object has moved higher up, in fact, even past the subject in (22).
This argumentation, however, cannot be replicated for Pseudogapping. Firstly, the object does
not move higher than the subject, i.e. does not appear to the left of the subject in the surface
structure. Secondly, a theoretical issue arises: according to Lasnik’s analysis, movement of the
object out of the phrase containing the EPP-feature is not necessary. In fact, it should even be
illicit, due to the fact that the EPP-feature is checked once the object is in [Spec,AgrO] position.
To be able to accommodate the Pseudogapping configuration in the current system, then,
the object needs to vacate the outer specifier of vP, and move higher up in the clause. The most
straightforward assumption is to establish a position above vP, as Boeckx (2000) suggested, an
‘FP’ (functional projection) in his terms, illustrated in (23).
(23)
TP
Subj
T’
T
FP
F’
Obj
F
vP
tObj
v’
tSubj
v
VP
V’
V
NP
invite
tObj
One consequence of assuming such a functional projection that hosts the object remnant in
Pseudogapping is that the EPP-feature for objects is not sufficient to derive the Pseudogapping
configuration, since it is already checked in the [Spec,AgrO] position.
3.3. THE EPP FOR OBJECTS IS NOT ‘UNIFORM’
A second argument supporting the claim that the EPP-feature does not suffice to derive the
Pseudogapping configuration comes from a comparison of the different cases where, in English,
the EPP for objects is assumed. If we compare particle constructions (in (24)), ECM cases (in
(25)), and Pseudogapping (in (26)), we find that Pseudogapping differs from the other two cases
in that there is a clear contrastivity requirement on the moved object remnant.
(24) a. Mary made John out to be a fool.
b. Mary made out John to be a fool.
(25) John believes everyone [t not to be there yet].
(26) John invited Jane, and Jill did SUSAN.
This observation permits the following interpretation. The observation that contrastivity seems to
arise only in the ellipsis configuration, i.e. that only Pseudogapping requires a special focus
structure, suggests that Pseudogapping should perhaps not be treated on a par with ECM and
particle constructions.8 However, this take runs counter to a rather ‘uniform’ EPP account
suggested by Lasnik (2001, a.o.). In other words, the contrastivity, as a second argument in
addition to the configurational one pointed out above, also indicates that the EPP-feature does not
play the decisive role in Pseudogapping, and that an additional requirement is involved in the
derivation.
3.4. CONTRASTIVITY
This section deals with contrastivity as such, building on the concept of Givenness, and the
observation (referred to in Merchant) that ‘a deleted constituent will not contain any F-marked
[focused] material; material extracted from the ellipsis site, on the other hand, will often – though
not always – be F-marked.’ (Merchant 2001: p.26, fn. 9).
Recall that the E-feature is based on the notion of E-Givenness (i.e. informally, the
combination of given material and a proper antecedent). I now propose that this link to focus and
Givenness also serves to account for the presence of the functional layer above vP in
Pseudogapping constructions.
More specifically, the functional layer in question may well be a Focus Phrase (FocP, cf. (27)),
of the kind assumed by e.g. Belletti & Shlonsky (1995), and previously implemented in
Jayaseelan (2001), and Kim (1997). This functional layer, due to its relation to information
structure, may serve to encode contrastivity on the object remnant in Pseudogapping.
In order to implement this contrastivity, I propose that there is a [+contrastive]-feature on
the element contrasted with its correspondent in the antecedent. This feature is designed to
capture the impossibility of deleting focused or new material, blocking the deletion of a phrase
that contains an unchecked and [+contrastive]-feature. It does so by triggering movement of
contrasted elements out of the phrase marked for deletion.
8
Note that the examples below might display a different focal structure. In (24), however, the difference is not
perceived to be striking.
i.
He set it up in the corner.
ii.
*He set up it in the corner.
(27)
TP
Subj
T’
T
FocP
Obj[contr.] Foc’
Foc[contr.] vP
tObj[contrastive]
tSubj
v’
v
VP
V’
V
NP
invite tObj[contrastive]
4. THE PROPOSED ANALYSIS
4.1. THE INTERACTION OF CONTRASTIVITY FEATURE AND E-FEATURE
Before I go on to illustrate how a combination of the two features can be used to derive different
types of ellipsis, a brief note on the nature of the contrastivity feature and its interaction with the
E-feature is in order.
Similar to the E-feature, which is based on Givenness, the [+contrastive]-feature also relies on
information structure. Given Merchant’s definition of E-Givenness in (10), we may even conflate
the two features, since contrastivity seems to encode non-Givenness of a constituent, at least on
the sentential level. Presumably, though, the [+contrastive]-feature has a broader range than the
E-feature, since it could also be important for the phonological marking of contrasted elements,
even in cases where no ellipsis in the strict sense is involved (e.g. in deaccentuation contexts).9
4.2. THE PSEUDOGAPPING DERIVATION
Applying both E-feature and [+contrastive]-feature to derive Pseudogapping (28) yields the
structure in (29).
(28) John invited Mary and Jane did _ Bill.
9
As a more specific analysis of this [+contrastive] feature is beyond the scope of this paper, I will keep the two
features apart for the time being. Minimally, I think, that the contrastivity feature can be motivated if it is responsible
for movement of contrasted elements out of the phrases that are deleted afterwards. It seems quite counterintuitive
that a feature marking Givenness (i.e. the E-feature) should induce movement of elements on which this feature is
absent.
(29)
TP
Jane
T’
T
did
FocP
Bill[contr.] Foc’
Foc[contrastive] vP
v’
tBill
tJane
v
VP => ø
V’
V[E]
invite
NP
tBill[contrastive]
The derivation in (29) proceeds as follows. The verbal head V (invite) is marked with the Efeature, hence specified for deletion. Since the head bearing the E-feature induces the deletion of
its maximal projection, as argued above, the full VP should be deleted. However, since the
maximal projection, VP, contains material bearing the [+contrastive]-feature, deletion has to be
deferred. Instead, the element with the [+contrastive]-feature is moved out of the phrase specified
for deletion. The [contrastive]-feature is checked in a Spec-head relation in FocP. Finally,
deletion of the full phrase, now containing only given material, ensues.
In some sense, the E-feature and the [+contrastive]-feature interact each other. If both features are
present in the derivation, as in Pseudogapping, the E-feature phonologically deletes the full
phrase after the material marked with [+contrastive] has moved out (in the syntax), i.e. once the
contrastivity feature is checked. This pattern is also found in other ellipsis types, as we will see
below.
4.3. EXTENSION TO OTHER CASES OF ELLIPSIS
4.3.1. VP ELLIPSIS
Since Pseudogapping is treated as VP ellipsis with the remnant having moved out of the VP, VPEllipsis (in (30)) receives a straightforward account.
(30) John invited Mary, and Jane did too.
(31) John invited Mary, and Jane did <invite Mary> too.
The derivation proceeds as in the Pseudogapping case in (29), without movement of the remnant.
Again, the redundant head (V) is marked with the E-feature. In contrast to Pseudogapping, the
VP ellipsis remnant is not contrastive, hence, the [+contrastive]-feature is absent, and no
movement is triggered. The maximal projection (VP), then, contains no contrastive material and
can be deleted straightforwardly.
4.3.2. SLUICING
The analysis above could also hold for Sluicing (in (32)). If we assume that in this case, the Efeature is placed on T (as also posited by Merchant 2001, 2004), and triggers the deletion of the
maximal projection of the T head, TP, as suggested above, then it would not be necessary to
move the E-feature higher up to C (as in Merchant’s 2001, 2004 analysis). Since TP does not
contain any contrastive material - the subject Jack is given -, the full TP can be deleted without
problem. The relevant syntactic structure is given in (34) below.
(32) Jack bought something, but I don’t know what [Jack bought t].
Of course, the Sluicing derivation in the terms proposed above hinges on two implicit,
fundamental assumptions. Firstly, the object ‘what’ needs to be contrastive, and bear the
[+contrastive]-feature to be moved up (possibly passing intermediate positions in between, if one
wanted to propose a separate deletion process of the lower part of the clause, i.e. VP). Secondly,
for TP deletion, the subject needs to be non-contrastive. This can easily be verified, however,
since the presence of non-contrastive subject would not result in Sluicing but rather in VP ellipsis
or a sentence with the verb repeated (cf. (33)).
(33) a. *Jack bought something, but I don’t know what SUSAN [bought t].
b. Jack bought something, but I don’t know what Susan did/bought.
(34)
CP
what
C’
C
[wh,Q]
TP
Jack
T’
T[E]
VP
V’
V
bought
NP
twhat[contrastive]
4.3.3. GAPPING
Accounts for Gapping largely treat Gapping as non-deletion (e.g. Johnson’s 1996 ATB-account,
or Agbayani & Zoerner’s 2004 sidewards movement account). Combining the E-feature with a
contrastivity feature allows us to reconsider gapping as ellipsis. More specifically, I propose that
Gapping is TP deletion. The movement and deletion approach outlined above gives a principled
explanation for the need of a contrastive subject in Gapping, an observation illustrated in the
examples in (35).10 (Incidentally, Pseudogapping differs in that respect. Levin (1986) points out
that there is a very strong preference for coreferential subjects in Pseudogapping.)
(35) a. Claire read a book, and Heather a magazine.
b. *Claire read a book, and she a magazine.
c.Clairei read a book, and SHEk a magazine.
The gapping derivation I propose builds on the assumption that gapping involves a contrastive
topic – contrastive focus structure (as put forward in e.g. Gergel et al. to appear). Following Rizzi
(1997) in general, and van Craenenbroeck (2004) for ellipsis cases in particular, I assume that a
split-CP structure is necessary for the derivation of Gapping.
I suggest that in Gapping, the E-feature is placed on ‘redundant’ elements, namely, T and V.
Then, on the VP level, the derivation could proceed as in Pseudogapping, with the remnant
moving up to the outer specifier of vP and then on to [Spec,FocP]. Note that the latter step would
not be necessary, if long A’-movement is assumed, and the movement in question is regarded as
last resort-type movement.
(36) a. Claire read a book, and Heather a magazine.
Treating the movement as cyclic, the VP-level in Gapping would look as in the Pseudogapping
case in (29). However, if T is also deleted, the E-feature on T would prohibit the presence of
focused material in its complement position, namely FocP in (37). Hence, the complement needs
to move higher up in the clause. Thus, the question arises where and how far the complement can
be moved. Building on proposals by Rizzi (1997), I assume that there are positions available for
movement in the C-domain, illustrated in (38).11
(37) TP level
TP
Heather
T[E]
T’
FocP
a magazineFoc
(38) ForceP... TopP... FocP... FinP
(Rizzi 1997: p. 288)
As the subject moves up to [Spec,TP] in the usual fashion, i.e. A-movement, subject movement
and object movement do not interfere. However, once the subject is in [Spec,TP], the surface
word order has to be derived. In order to exclude movement of the object to a position on the left
of the subject, I tentatively assume parallel movement of subject and object. The particular
information structure (i.e. contrastive topic - contrastive focus) of Gapping could be implemented
via features. I suggest that in Gapping (in contrast to both VP ellipsis and Pseudogapping where
the subject can be coreferential), the subject bears a topic feature, which triggers movement up to
10
Cf. e.g. Lechner (2004: 112, fn. 103), a.o., on disjoint reference in Gapping.
As ForceP and FinP are not crucial to my purposes here, I will henceforth only consider the TopP/FocP
projections.
11
the Topic position in the C-domain. The contrastive object remnant then moves into the specifier
of a Focus phrase, situated below the Topic phrase. Both Topic and Focus phrase serve to check
the contrastivity feature.
The structure in (39) illustrates the configuration after these movement operations have taken
place.
(39) CP level
TopP
Heather Top’
Top
FocP
a magazine Foc’
Foc
....
Assuming the featural layout proposed above is correct, this would account for the general
contrastivity of the subject in Gapping, and the non-coreferential reading if the subject pronoun is
focused, as in (35) (where capitals indicate focus).
Note that the presence of two focus positions in the gapping sentence, standardly assumed to be
illicit, poses no specific problem in my analysis, since the second focus position (above vP) does
not contain a second focused element, and is deleted in the course of the derivation, yielding only
one focus position in the final sentence.
The full gapping derivation of (35) would then look as in (40) (below).
(40)
TopP
Heather Top’
Top
FocP
a magazine Foc’
Foc
TP
tHeather
T’
T[E]
FocP
Foc’
tmagazine
Foc
vP
tObj
v’
tSubj
v
VP
V’
V[E] NP
invite tObj
5. CONCLUSION
In this paper, I have shown that the current analyses of Pseudogapping do not address the
problem of contrastivity on the Pseudogapping remnant. By combining Lasnik’s (1995, 1999)
analysis of the Pseudogapping construction with Merchant’s (2001) E-feature analysis of
Sluicing, I have laid the foundation for the inclusion of this information-structural peculiarity into
the derivation of Pseudogapping. I have put forward evidence that the EPP-feature, as assumed
by Lasnik, is not sufficient to yield the Pseudogapping configuration, at least not in the current
understanding of the general theory. Keeping the E-feature, I have introduced a [+contrastive]feature to capture the required contrastivity on the Pseudogapping remnant. The application of
the E-feature differs from Merchant’s original definition in that it does not depend on sisterhood
but immediately targets maximal projections. This modification of the mechanism invoked by the
E-feature, in combination with a contrastivity feature on contrastive elements, permits a uniform
picture of various types of ellipsis: The interaction between the two features captures the intuition
that given material can be deleted, whilst non-given material cannot be deleted and must be
moved out of a phrase undergoing deletion. The E-feature marks the ‘redundant’ element and
specifies the whole maximal projection of the head that bears the E-feature for deletion. The
[+contrastive]-feature marks the contrastive material and triggers its movement out of the phrase
marked by the E-feature before it undergoes deletion.
REFERENCES
Belletti, Adriana & Ur Shlonsky (1995). ‘The order of verbal complements: A comparative
study’, Natural Language and Linguistic Theory 13, 489-526.
Boeckx, Cedric (2000). ‘An additional note on Pseudogapping’ in K. Schwabe & N. Zhang
(edd.) Ellipsis in Conjunction. Tuebingen: Niemeyer, 117-132.
Chomsky, Noam (2001). ‘Derivation by phase’ in M. Kenstowicz (ed.) Ken Hale: A Life in
Language. Cambridge, Mass.: MIT Press.
Chung, Sandra, William A. Ladusaw & James McCloskey (1995). ‘Sluicing and Logical
Form’, Natural Language Semantics 3, 239-282.
Craenenbroeck, Jeroen van (2004). Ellipsis in Dutch dialects. LOT publications.
Gengel, Kirsten (to appear). ‘Object Shift and Pseudogapping in the Scandinavian Languages’,
Proceedings of ConSOLE XIV.
Gergel, Remus, Gengel, Kirsten & Susanne Winkler (to appear). ‘Ellipsis and inversion: A
feature-based focus account’, in K. Schwabe & S. Winkler (edd.) On Information Structure,
Meaning, and Form. Amsterdam/Philadelphia: John Benjamins.
Jayaseelan, K. A. (1990). ‘Incomplete VP deletion and gapping’, Linguistic Analysis 20: 64-81.
Jayaseelan, K. A. (2001). ‘IP-Internal topic and focus phrases’, Studia Linguistica 55(1), 39-75.
Johnson, Kyle (1991). ‘Object Positions’, Natural Language and Linguistic Theory 9, 577-636.
Johnson, Kyle (1996). ‘In search of the middle field’, unpublished manuscript, University of
Amherst, Mass.
Kennedy, Christopher & Jason Merchant (2000). ‘Attributive Comparative Deletion’, Natural
Language and Linguistic Theory 18: 89-146.
Kim, Jeong-Seok (1997). Syntactic focus movement and ellipsis: a minimalist approach. Ph.D.
Dissertation, University of Connecticut.
Koizumi, M. (1995). Phrase Structure in Minimalist Syntax, Ph.D. Dissertation, MIT.
Lasnik, Howard (1995). ‘A Note on Pseudogapping’, MIT Working Papers in Linguistics 27:143163.
Lasnik, Howard (1999). ‘Pseudogapping Puzzles’ in E. Benmamoun & S. Lappin (edd.),
Fragments: Studies in Ellipsis and Gapping. New York/ Oxford: Oxford University Press,
141-174.
Lasnik, Howard (2001). ‘Subjects, Objects, and the EPP’. In: W. D. Davies & S. Dubinsky
(edd.) Objects and other subjects: Grammatical functions, functional categories, and
configurationality. Dordrecht: Kluwer, 103-121.
Lechner, Winfried (2004). Ellipsis in Comparatives. Berlin: Mouton de Gruyter.
Levin, N. (1978). ‘Some Identity-of-Sense Deletions Puzzle Me. Do They You?’, CLS 14, 229240.
Levin, Nancy (1986). Main Verb Ellipsis in Spoken English. New York: Garland.
Merchant, Jason (2001). The Syntax of Silence: Sluicing, islands, and the theory of ellipsis.
Oxford: Oxford University Press.
Merchant, Jason (2004). ‘Fragments and ellipsis’, Linguistics and Philosophy 27.6: 661-738.
Rizzi, Luigi (1997). ‘The fine structure of the left periphery’ in L. Haegeman (ed.) Elements of
Grammar. Dordrecht: Kluwer, 281-337.
Rooth, Mats (1992). ‘A Theory of Focus Interpretation’, Natural Language Semantics 1: 75 116.
Schwarzschild, Roger (1999). ‘GIVENness, Avoid F and other Constraints on the Placement of
Focus’, Natural Language Semantics, 7(2): 141-177.