Professional Documents
Culture Documents
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Springer is collaborating with JSTOR to digitize, preserve and extend access to Natural
Language & Linguistic Theory
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Nat Lang Linguist Theory (201 3) 3 1 :409-45 1
DOI 10.1007/sl 1049-013-9188-6
Ángel J. Gallego
Received: 27 July 2009 / Accepted: 10 October 201 1 / Published online: 28 March 2013
© Springer Science+Business Media Dordrecht 2013
Abstract This paper discusses the properties of VOS sentences in Romance and their
bearing on Case assignment, verb movement, parametric variation, and the structure
of the vP periphery / low IP area. The literature on Romance VOS has argued that
this order is derived either through object shift (Ordóñez 1997, 1998, 2000) or VP
fronting (Belletti 2001, 2004; Zubizarreta 1998), providing empirical evidence in
support of both derivations. In this paper, I focus on various aspects of VOS sen-
tences in Romance languages. First, I argue that both object shift and VP-fronting
strategies are actually available, but subject to a very specific parametric cut: Western
Romance languages (Galician, European Portuguese, and Spanish) resort to object
shift, whereas Central-Eastern varieties (Catalan and Italian) fail to do so, requiring
the VP-fronting derivation instead (López 2009a). Second, I put forward a previously
unnoticed generalization that reveals that only those varieties licensing object shift
based VOS can generate VSO sentences, which I refer to as the VOS-VSO Gener-
alization. Finally, I claim that object shift in VOS sentences of Western Romance
languages displays a cluster of unnoticed properties that pattern with Scandinavian-
style object shift (Bobaljik and Jonas 1996; Bobaljik and Thráinsson 1998; Collins
and Thráinsson 1996; Holmberg 1986, 1999; Vikner 2006), thus obeying Holmberg' s
Generalization (Holmberg 1986, 1999). If on track, the present account not only re-
veals interesting syntactic similarities between Scandinavian and Romance object
shift, but also reinforces a well-known micro-parameter that disentangles Western
from Central-Eastern Romance languages.
ô Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
410 Á. J. Gallego
1 Introduction
This paper argues that VOS sentences in Romance languages are generated by means
of two independent strategies, namely "object shift" (or "scrambling", as argued by
Ordonez 1997, 1998, 2000) and "VP fronting" (as argued by Belletti 2001, 2004
and Zubizarreta 1998). Although the literature on this topic has provided evidence in
favor of both strategies (see López 2009a for a summary and much relevant discus-
sion), I claim that both are not only technically possible, but empirically correct and
parameterized. I propose that each strategy is adopted by a group of languages of the
Romance family: Western Romance languages (i.e., Galician, European Portuguese,
and Spanish) resort to object shift, whereas Central-Eastern Romance languages (i.e.,
Catalan and Italian) resort to VP fronting. The outcomes of the relevant derivations
are depicted in (1) and (2).
0)
[t[vW]T]^^
OBJECT vP
tv toBJECT
(2)
VP vP
V Wt
di Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 4 1 1
(3) [tp T [vp OBJECT [vp SUBJECT V [vp V towECT ]]]] MINIMALITY EFFECT
•-Hg)
(4) a. [tp Ayer T estaba [vp Juan [vp v [vp leyendo un libro ]]]]
yesterday be-PST.3.SG Juan reading a book
'Juan was reading a book yesterday.'
b. *[tp Ayer T estaba [vp un libro [vp Juan v leyendo tun libro ]]]
yesterday be-PST.3.SG a book Juan reading
'Juan was reading a book yesterday.'
Just like in Scandinavian, the verb does not escape the vP in Romance, which pre-
vents the object un libro 'a book' from undergoing object shift. Although Romance
and Scandinavian share this movement constraint, I would like to suggest that, in
the case of Romance languages, the generalization is not restricted to pronouns,
and does not appear to obey phonological restrictions either (as argued by Holm-
berg 1999; see also Fox and Pesetsky 2005). In this paper, I take the Romance
VOS and VSO data to support the hypothesis that some instances of verb move-
ment are truly syntactic (as independently argued by Donati 2006; Chomsky 2008;
Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
412 Á.J.Gallego
This section explores two derivational routes that can be taken in order to derive VOS
sentences in Romance. The literature on this topic features a debate on whether VOS
is obtained through movement of the object (as argued for by Ordonez 1997, 1998,
2000) or through movement of the entire VP (as Belletti 2001, 2004 and Zubizarreta
1 998 contend). I claim that both options are correct, and present a concrete parametric
setting that teases apart Western from Central-Eastern Romance languages. I first
review the data that support the two approaches to VOS sentences (Sect. 2.1), and
then I focus on the specific derivations involved in those approaches (Sect. 2.2). 1
As is well known, most Romance languages license SVO and VOS sentences, while
VSO is more restricted. This is illustrated by the Catalan data in (5), taken from
Picallo (1998) (see also Ordóñez 1997, 1998, 2000, 2007; López 2009a; Solà 1992;
Vallduvf 1990 and references therein). As can be seen, Catalan generates SVO and
VOS, but not VSO.
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shiñ in Romance 4 1 3
2 An anonymous reviewer asks me to clarify the status of French. As has been noted in the literature (see
Ordóñez 1997, 2000, and references therein), French does not license VOS and VSO orders. Clearly, as
the reviewer suggests, this is plausibly related to the fact that French is not a pro-drop language. If the
analysis put forward in Sect. 5 is tenable, then the specific properties of French follow naturally, since the
relevant parameter will be related to the morphological composition of v.
â Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
414 Á.J.Gallego
The conclusion to be drawn from these data is that VOS can be derived in most
Romance varieties (i.e., European Portuguese, Galician, Spanish, Catalan, Romanian,
and to some extent Italian). What remains to be determined is whether the same
derivation is at stake in all of them. In what follows I submit that the analyses of
Romance VOS proposed by Belletti (2001, 2004) and Ordonez (1997, 1998, 2000)
are both correct, and apply in a parametrically dedicated fashion.
In the preceding section we saw that Italian VOS, though possible, is slightly
degraded. Belletti (2004) actually argues that VOS structures are licensed by a
derivation where the verb-object complex is interpreted as given/background/pre-
supposed information. Adopting a cartographic approach, Belletti claims that the
entire VP - the constituent containing the verb and the object - moves to the spec-
ifier of a topic projection within the left periphery of the vP phase / the low
IP area, the subject having undergone movement to the specifier of a focus pro-
jection. This is a remnant movement based analysis, since the subject leaves its
base position before the VP moves, as indicated in (9) (see also López 2009a;
Zubizarreta 1998):3
(9) a. ^FocP^ b.
subject, Foe' VPj Top'
Foe tj
A cr
the
that
(200
3 In (9) I depart from Belletti's (2004:34-38) implementation with respect to the amount of structure that
is topicalized. For Belletti (2004), it is vP, while the analysis in (9) assumes it is the VP. Nothing crucial
hinges on this. What matters is that c-command between subject and object is impossible.
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 4 1 5
Departing from Belletti's account, Ordonez (1998) proposes that VOS structures are
derived by scrambling the object to a position that c-commands the subject, roughly
as indicated in (12).
(12)
object, v'
subject v
v VP
veib ti
Ordo
jectio
Specv
arise:5
^This analysis resembles Torrego's (1998) approach to Case-marked direct objects, which also move to
SpecvP. López (2009a, 2012) provides empirical evidence that not all Case-marked direct objects target
SpecvP. According to this author, Case-marked objects abandon their base position, but not all of them
end up above the external argument. 1 cannot discuss the evidence provided by López (2009a, 2012) due
to space constraints. Interestingly for my purposes, López (2012) also takes cases like the ones studied by
Ordóñez (1998) to target an outer specifier of the vP.
5 One comment is in order with respect to the object shift account put forward by Ordóñez (1998). Fol-
lowing Ordóñez, in this paper I concentrate on object shift that targets full DPs, not (clitic) pronouns. The
literature on object shift (Holmberg 1986, 1999, 2000; Diesing 1992; Bobaljik and Jonas 1996; Bobaljik
and Thráinsson 1998; Collins and Thráinsson 1996; Vikner 2006; among others) has used this label to
cover both full DP and pronoun movement, although they behave differently (see Holmberg 1999 and ref-
erences therein for discussion). In the following pages, 1 will ignore the pattern involving (clitic) pronoun
movement (see Leonetti 2007, 2008; Suñer 1988, 2000 for discussion), and regard it as part of clitizication
(see Roberts 2010; Raposo and Uriagereka 2005; Uriagereka 1995b and references therein for a summary
of cliticization in Romance).
Ô Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
416 Á J. Gallego
(13) Strategies to de
a. VP Fronting (Be
b. Object Shift (Or
Different pieces of ev
ish case. Let us briefly
post- verbal subjects,
effects (see (15)).
In stark contrast, objects cannot bind into subjects in Catalan (and Italian) (see
Gallego 2010:Chap. 3; López 2009a:Chap. 3), which provides support for the VP-
fronting strategy for VOS in Central-Eastern Romance.
Finally, consider one more datum that supports the VP-fronting strategy f
Central-Eastern Romance. As noted in the literature on VOS language
and Hermon 2008 and Chung 2006, which argue for a VP-fronting deriva
6There are two other analyses of VOS sentences in Romance that I am putting aside. The first
( 1 982) pre-Kaynian approach, where the subject is right-adjoined to the VP. However plausi
that the post-verbal subject bears the main stress (it is a non-contrastive focus) provides inde
dence that either it has remained in situ or else it is the most deeply embedded element (see C
The second account of VOS in Romance 1 do not consider is so-called emarginazione 'mar
(see Cardinaletti 2001). The intonation pattern associated with this pattern indicates that it i
collapsed with standard VOS (see López 2009a:Chap. 3 for discussion).
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 4 1 7
The literature on object shift has also emphasized the fact that the object receives a
dedicated semantic interpretation, referred to as specific or referential (see López
2012 and references therein for discussion). Given this possibility, investigating
whether objects in VOS sequences are interpreted as specific in Spanish is in or-
der. Ordonez (1998:332 and ff.) argues that they are but I disagree. What Ordonez
(1998) does show is that objects in their derived position in Spanish have a wide
scope reading. Hence, whereas according to Ordonez (1998) the indirect object DP
a un profesor 'to a teacher' is specific in (18b), it does not have to be (at least in my
idiolect).7
Ô Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
418 Á J. Gallego
8 If Catalan and Spanish involve different derivations for VOS sentences, one might expect that the objects
be interpreted differently in these languages, as Louise McNally (p.c.) points out. Though possible, there
is no actual interpretive difference between (i) and (ii):
As noted above, the direct object can, but need not, be interpreted as s
however, that it should be easier for the direct object to receive a wide s
not what we find: un llibre can also out-scope dos estudiants in (ii). 1 take
is obtained through QR (covert movement in the Logical Form compone
available (see May 1977, 1985).
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 419
To conclude, there is good evidence that VOS sentences in Spanish invoke object
shift, i.e., object movement to a specifier position above the in situ subject. Al-
though nothing I have to say here hinges on the label of this functional projection,
I take Ordonez's (1998) SpecFP to be an outer specifier of the vP (or a specifier
within an extended vP periphery, endowed with multiple shells, à la Belletti 2004;
Cinque 1999). As we have seen, the semantic effects associated with this process
as well as the categorial nature of the moved element (not only pronouns, full DPs
too) are different from instances of what one might call canonical "object shift"
(Holmberg 1986, 1999, 2000; Diesing 1992; Bobaljik and Jonas 1996; Bobaljik and
Thráinsson 1998; Collins and Thráinsson 1996; Vikner 2006; among others). This
raises the question of how similar these two instances of object shift are. I defer the
answer to this question until Sect. 3.
At this point, I would like to focus on a problem that a derivation involving object
shift raises. For concreteness, the problem arises the moment one entertains the idea
that nominative Case is assigned by T.9 If this is so, and assuming that the object
moves to a position that c-commands the subject, then a minimality effect (in the
sense of Rizzi 1990 et seq.) should emerge. I explore this conflict within the context
of Chomsky's (2000, 2001) analysis of structural Case in this section.
As just pointed out, the derivation of VOS sentences that involves object shift pre-
dicts a minimality scenario. More precisely, VOS sentences should give rise to what
Chomsky (2000:123) refers to as "defective intervention", a variety of minimality
effect triggered by a DP that has already checked its Case. It is important to notice
that such a minimality effect will emerge regardless of whether nominative is as-
signed through Attract/Move-F (Chomsky 1995), long distance Government/Agree
9 1 will assume that T (Infl or Agrs in previous formulations) assigns (alternatively, 'checks' or 'licenses')
nominative Case. For alternative approaches, taking C as the locus of nominative Case, see Bittner and
Hale (1996), Chomsky (2007, 2008), López (2007), and references therein.
Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
420 Á.J. Gallego
According to Chomsky (2000, 2001), the 0-features on the Probe match those
Goal, and valuation takes place. After the Goal values the ^-features on the
these features are deleted and the DP receives structural Case. The entire pr
involving Match, Valuation (plus deletion of ^-features), and Case assignmen
lustrated in (22), where valued and unvalued features are represented as val
unvalued attributes respectively (e.g., [number: PL] vs. [number: ]).10
Let us next consider the intervention effect created by the shifted object in Romance
VOS sentences. The basic idea, as mentioned in the previous section, is that a config-
uration where an object raises to a position c-commanding the in situ subject should
,0Since it is orthogonal to the issues that we are considering, I am ignoring here the step whereby the
subject raises to the SpecTP (EPP) position.
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 42 1
Chomsky (2000
to indicate tha
that the Goalj
as a bona fide
cannot value it.
In order to overcome scenarios like the one in (23), which had been assumed for
cases where the object had to raise to SpecAgroP, bypassing the in situ subject (see
Chomsky 199:183 and ff.), Chomsky put forward an anti-minimality strategy, which
he called "equidistance".
As can be seen, the technical advantage of equidistance is that it allows for some
minimality violations to be ignored within a local domain. In a structure like (23),
for instance, it does not matter whether Goali c-commands Goah: Equidistance col-
lapses both goals for locality purposes. If a mechanism like Chomsky's equidistance
(or any notational variant) is available, T can agree with the in situ subject in VOS
sentences ignoring the object.
With this technical discussion in mind, I would like to go back to Romance VOS
sentences, for they display a minimality scenario that resembles the one shown by
Scandinavian object shift. Notice, crucially, that for a minimality conflict to arise, we
must first make sure that VOS sentences in Romance are derived through object shift,
and not VP fronting. For this reason, I will mainly focus on Spanish, which meets
this requirement (as shown in Sect. 2.2), unlike Central-Eastern Romance languages.
Since I want to compare the behavior of Romance and Scandinavian languages, I will
first discuss the properties of Scandinavian object shift in some detail.
So far I have argued that Western Romance languages that license the VOS order em-
ploy object shift (Ordonez 1997, 1998, 2007), and not VP fronting. I have also argued
that Romance object shift does not force a referential or specific interpretation of the
object, contrary to what has been argued for in the case of Scandinavian object shift.
Ô Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
422 Á.J. Gallego
Paradigms like
to object shift.
abandoning the
(26) (Swedish)
a. [Tp Jag T kysste
I kiss-PST.l.SG her not
'I did not kiss her.'
c. *[cp C att [Tp jag T [vp henne [vp inte [vp tjag ky
that I her not kiss-PST.l.SG
'that I did not kiss her' [from Holmberg 1 999: 1 ]
As can be seen, the verb kysste 'kissed' moves to T in (26a). In (26b) and (26c), it
remains in situ because the complementizer and the auxiliary block movement, and
thus object shift too.
The second take on object shift, which emphasizes PF factors (see Holmberg
1999), relies on the assumption that this operation is allowed if no phonological ma-
terial is left in the vP. As an anonymous reviewer points out, this analysis aligns with
Fox and Pesetsky's (2005) approach to cyclic linearization. The basic idea of Fox
and Pesetsky's (2005) proposal is that linear order is determined at the end of every
cyclic domain (every 'phase', in Chomsky 2000), after the operation of Spell-Out
takes place. Details aside, both Holmberg (1999) and Fox and Pesetsky (2005) regard
object shift as a phenomenon that is subject to PF requirements only, regardless of
1 1 As an anonymous reviewer points out, (25) should not be taken to indicate that X attracts Y. (In Chomsky
1993, movement was conceived of as a greedy operation, triggered by the moving element.)
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 423
(27) (Swedish)
a. *[tp Jag T gav
I give-PST.l.SG it not to-Elsa
'I did not give it to Elsa.'
b. *[Tp Dom T kastadevERB [vP mej [ inte [vp tDoin tkastade
they throw-PST.3.PL me not
UtpARTICLE Wj ]]]]
out
Though plausible, it is not immediately obvious that the data in (27) could be ex-
plained by invoking standard syntactic devices (be it minimality or equidistance),
since even particles (elements that do not compete with objects for Case-checking
reasons) block object shift. Crucially, as Holmberg (1999) observes, if the offending
material is removed from the vP, object shift is fine again. This can be seen in (28),
where the verb, the dative element, and the particle leave the vP edge.
(28) (Swedish)
a. [cp C KysstvE
kissed have-l.SG I her not
'Kiss, I did not did that to her.'
b. [cp Vemi-oBJ C gavvERB [tp du T [vp den [vp in
who give-PST.2.SG you it not
[vP tdu tVERB Wem tden ]]]]]
'To whom didn't you give it?'
c. [cp UTparticle c kastadevERB [tp dom [vp mej [ inte
OUT throw-PST.3.PL they me not
[vP tDom tkastade tUT tmej]]]]]
'OUT they did not throw me.' [from Holmberg 1999:7,
(29) a. [tp T Compró [vp el coche [vp Maria temprò tei coche ]]] (S
buy-PST.3.SG the car Maria
'Maria bought the car.'
b. [tp T Rompio [vp el vaso [vp Pablo trompió tei vaso ]]] (Span
break-PST.3.SG the glass Pablo
'Pablo broke the glass.'
,2Fox and Pesetsky's (2005) approach is, like Holmberg's (1999), phonological in nature, but
from Holmberg's (1999) in taking the phonological restriction on object shift not as a constrain
application of the movement rule itself, but as a domain-final constraint on the output obtained
earization. For ample discussion of Fox and Pesetsky's system, I refer the reader to Richards (2
Ê Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
424 Á.J. Gallego
(30) *
[T [WV] T]^VP^
OBJECT VP
tv VP
t V ^OBJECT
b. VP
VP vP
V t OBJECT
See Kučerova (2007) for an analysis of Czech object shift that also invokes equidistance.
Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 425
The relevant test is (34), where the object DP un libro 'a book' is shifted over the
subject, and the verb remains within the vP. As expected, the outcome is ungrammat-
ical. I take (34) to be evidence in favor of an object shift account of VOS in terms
of Holmberg' s Generalization , and not phonological factors, at least in the case of
Romance.
(34) *[jp Ayer T estaba [vp un libro [vp Juan v leyendo tun libro ]]] (Spanish)
yesterday be-PST.3.SG a book Juan reading
'Juan was reading a book yesterday.'
It must be emphasized that what licenses object shift is not verb movement per se ,
but rather verb movement to a position above the shifted object (in principle, T/Inf).
To see this, consider (35). If manner adverbs like lentamente 'slowly' occupy the
specifier of some vP layer (Cinque 1999), then the verb must have moved in (35a),
but object shift is still ungrammatical. Only verb movement above the object, as in
(35b), yields a licit outcome.
,4For reasons that I fail to understand, estar is the best auxiliary verb to obtain the order AUX-SVO in
Spanish. I have not tested the different possibilities systematically, but ser 'be' and haber 'have' do not
successfully allow such a pattern, which is a necessary condition to test my hypothesis. As an anonymous
reviewer points out, modal verbs, like poder 'can' - and deber 'must' too - , also allow the AUX-SVO
order, as shown in (iMii):
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
426 Á.J. Gallego
What the data in (iii) tell us is that the quantifier todo 'all' blocks nominative
above the subject, unless the verb moves (as in (iv)).
£ļ Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 427
Let us recap. In this section, I have explored how similar the derivations of VOS
sentences in Romance and Scandinavian are, paying attention to the problem that an
object shift based derivation poses for nominative Case assignment. I have argued
in favor of the hypothesis that the object can move if the verb does as well, largely
following Chomsky's (1993, 1995, 2004) idea that head movement can give rise to
equidistance effects. For this to be tenable, one must assume that object shift in Ro-
mance is not subject to phonological factors, verb movement being a truly syntactic
effect. The data provided by Ordonez (1997, 1998, 2000) support a syntactic ap-
proach to object shift, while the examples from (35) to (38) reinforce the necessity
for the verb to move if the object does.
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
428 Á.J. Gallego
4 Alternative accounts
In this section, I want to consider some strategies that could be invoked to account
for the fact that nominative Case is assigned in object shift scenarios. In particular, I
want to consider the possibilities in (39):
Below I show why these options fall short of explaining how the in situ subject in
VOS sentences receives nominative Case. To the extent that the options in (39) can
be safely dismissed, the proposal put forward in the previous section is reinforced.
The first strategy in (39) argues for the possibility that the object moves not to an
outer SpecvP, but to SpecTP. An account along these lines is proposed by Hiraiwa
(2001), who builds on Holmberg (2000) in taking T to be able to satisfy its EPP
feature by pure Merge (without Agree). According to this analysis, when a derivation
reaches the stage in (40a), T is merged, attracting the closest element to satisfy the
EPP requirement, as depicted in (40b).
If the object is in SpecTP, and the 0-Probe operates from T, (40b) predicts that in-
tervention effects should not arise, since there is no intervener sandwiched between
T and v. Let us now return to Romance VOS, focusing on Spanish. Examples like
(41), where the adverb rápidamente 'quickly' occupies the rightmost position, could
in principle be taken to indicate that subject and object have vacated the vP (assuming
adverbs signal the vP periphery; see Cinque 1999).
Be that as it may, it is highly unlikely that this analysis is correct for Romance VOS,
as it would require making ad hoc assumptions. First, objects should undergo A-
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 429
(42) Dicen [cp que repartía las cartas Juan rápidamente] (Spanish)
say.3.PL that deal-PST.3.SG the cards Juan quickly
'They say that Juan dealt the cards quickly.'
If que 'that' occupies the C head, then the verb repartía 'dealt' cannot be above T.18
From this I conclude that, in Romance VOS, neither the object alone, nor both the
object and the subject can be specifiers of TP. Therefore, the option in (39a) can safely
be dismissed.
In Sect. 2.2, it was argued that Central-Eastern Romance differs from Western Ro-
mance in resorting to a VP fronting strategy to generate VOS. Examples like (10),
repeated below as (43), were taken as evidence to show that VOS objects do not
c-command the in situ subjects.
Interestingly enough, some data provided by Cardinaletti (2001) cast doubt on such
a clear-cut asymmetry between Italian and Spanish. As can be seen, the examples
in (44) seem to indicate that objects can c-command the in situ subject in Italian too.
16 At least, it should be the case that objects move beyond T (or the relevant nominative Case-as
head). Plausibly, such a position could be Uriagereka's (1995a) F or Rizzi's (1997) Fin.
17 An anonymous reviewer regards this argument as theory-internal. Taking Case assignment to b
to 0- feature valuation is indeed a theory-internal decision. (There are different approaches to Case
is just one; see Pesetsky and Torrego 201 1 ; Legate 2008 and references therein for alternatives.)
I think the argument has a clear empirical weight; abstracting away from clitics, verb agreement i
oriented in Romance, which means that T (or Infl) can only agree with subjects. If objects could
to T, then we would expect for these dependents to agree with the verb in active sentences, wh
what we find.
1 ^ Here I am assuming a restrictive scenario. Things would be different under a more fine-grain
graphic) approach to the CP field (see Rizzi 1997 et seq.).
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
430 ÁJ. Gallego
b. Ha visitato Gi
have-3.SG visited
'A colleague of his
[from Car
Things are also murky in Spanish. Consider the data in (46). To my ear, plural agree-
ment across a singular object, as in (46b), yields deviance if binding is forced by
a quantificational object. Singular agreement across a plural object in (46a), on the
other hand, is not degraded.19,20
Taken together, the facts suggest that, when raised to an outer SpecvP, objects
potentially interfere with nominative Case assignment unless they share the same
19That is to say, when binding is not forced (because the object DP is not quanti fìcational), plural verb-
subject agreement across a singular object is perfect:
20 As Louise McNally (p.c.) notes, the difference in acceptability in (46) may be due to the possibly non-
quantificational status of todos los N 'all the N' DPs. She further observes that cada N 'each N' and cada
uno de ios N 'each one of the N' may behave differently, since los N in the latter will presumably be
associated with its own discourse referent, while N in the former will not. Although this prediction is
sensible, I find no remarkable difference between in (i) and (ii).
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 43 1
This analy
it is not im
yield a mi
of the so-
therein fo
4.3 Altern
A third po
lar, one co
whereby a
From this perspective, it is the </>-bundle (or little pro) that checks nominative Case.
Therefore, it is the 0-bundle that controls subject-verb agreement. The key idea is
2 Unlike Spanish, Italian behaves as expected under a Multiple Agree / Co-valuation analysis:
A reviewer suggests that (i) may be ruled out because the variable contained in the post-verbal subject is
not bound by the moved quantifier. If I am correct, variable binding is possible, but it forces an object shift
based derivation, which is illicit in Italian.
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
432 Á.J. Gallego
Although such a proposal solves the Case conflict posed by VOS sentences, it make
incorrect predictions. One such prediction concerns negative quantifiers, which ca
be subjects in VOS sentences (see 51a), but fail to be doubled by a(n overt) clitic
object position not only in Spanish (see 51b), but in all Romance languages. That
is to say, given that doubling is barred with negative quantifiers, it is unlikely th
doubling solves the minimality scenario we are considering.22
Û Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 433
I would like to conclude by considering one final analysis for VOS in Romance. The
key to this account has been entertained in the literature (Atkinson 2001; Broekhuis
2007; Richards 2004) and consists in denying the existence of defective intervention,
the type of minimality effect that concerns us here. In Richards (2004:161), it is
argued that the idea that inactive DPs create intervention effects is an unpleasant and
unnecessary complication of the theory: "Ideally, a DP should either be a potential
goal or it should not, and inactive nominais should be just that: inactive, for goalhood
and intervention alike." Stated differently, given that the shifted object has already
been assigned Case, it should not count as an intervener. As Richards (2004) observes,
there are three types of evidence in favor of defective intervention effects, some of
which have already been mentioned.
Richards (2004) discusses the cases in (53), noting that they force Chomsky (2000,
2001) to make unclear assumptions, such as the idea that the intervention effect cre-
ated by quirky subjects (53a) and expletives (53b) only follows if expletives have a
person feature and quirky subjects bear "inherent Case with an additional structural
Case" (Chomsky 2000: 127, 2001 :43, fn. 8, an idea that goes back at least to Jónsson
1996). As for (53c), Richards (2004) argues that Chomsky's (2001) analysis of nom-
inative Case assignment across wh-objects goes through under two odd assumptions
as well: first, traces must be regarded as invisible and, second, evaluation of opera-
tions takes place at the phase level. Richards (2004) discards (53c) entirely and pro-
poses a unified analysis for (53a) and (53b) whereby quirky subjects are treated like
DPs with an additional layer hosting an expletive with the minimal ^-specification: a
person feature. This is shown in (54).
(54) gives rise to an otherwise unnatural conception of quirky subjects and expletives
as having some properties in common for the purposes of Agree. Viewed this way,
partial agreement displayed with quirky subjects follows from the T head agreeing
first with the expletive shell in person, and then with the downstairs DP (the so-
called associate) in number. As Richards (2004) puts it, there is no intervention, but
split agreement. Accordingly, the fact that the nominative object cannot be 2nd/ 1st
person (see (55) below) follows from the non-distinctness condition on agreement:
"An object with a lexical value or l-/2-person is distinct from the T-probe's 3 person,
and therefore fails to be matched by T. Since Agree (T, object) thus fails, Case on the
Ô Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
434 Á.J. Gallego
Let us now go back to object shift in Romance VOS. If there are no defective in-
tervention effects, as Richards (2004) contends, the question is: what rules (35) out,
repeated here as (56)?
(56) *[tp Ayer T estaba [vp un libro [vp Juan v leyendo tun libro ]]] (Spanish)
yesterday be-PST.3.SG a book Juan reading
'Juan was reading a book yesterday.'
Marc Richards points out through personal communication that (56) is out due to the
fact that the object moves across the verb, which is impossible, since Spanish is a
SVO language.23 If I interpret him correctly, Richards is arguing that a linearization
parameter (or a linear order preservation constraint, à la Fox and Pesetsky 2005) is
responsible for (56)'s degradation. I am not convinced. Consider the sentence in (57),
which is possible in my Spanish idiolect, and for other Spanish speakers I have con-
sulted too. I would like to take this example as direct evidence that (56) is out not
because of object shift alone.24
(57) ?[tp T Estaba [vp los libros [vp pro v leyendo tios ubros ]]]•••
be-PST.3.SG the books reading
. . . cuando, de pronto, llegó María. (Spanish)
when of soon arrive-PST.3.SG Maria
'He was reading the books. . . when, suddenly, Maria showed up.'
The sentence in (57) may not sound entirely natural to some speakers. It has indeed
somewhat rhetorical style, whence the question mark, but it is not out. The important
example is (58), almost identical to (57), but fully ungrammatical. Importantly fo
my purposes, even those speakers that disagree with (57) observe a contrast betwe
(57) and (58), the latter being considerably worse. (Judgments are comparative, rath
than absolute.)25
23 Of course, the claim that the object cannot precede the verb in a SVO language is empirically fal
for it can in cases of topicalization or focalization, which are not relevant to the present study. I thank
anonymous reviewer for pointing out this issue to me.
24The odd status of (57) certainly follows from having moved the object to a specifier of a non-finite form,
which does not normally allow syntactic objects to stay in its edge, for morphological reasons. As expec
from everything I have said, the Catalan counterpart of (57) is much worse.
25 An anonymous reviewer suggests that (58) be approached in terms of linearization parameters. Althou
such an approach is appealing, considering it in detail is beyond the scope of this paper. For reade
Û Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 435
Here I would like to claim that the contrast between (57) and (58) has to do precise
with defective intervention. In (58) Juan cannot check its Case feature because lo
libros 'the books' intervenes. In contrast, (57) is fine because there is no interventio
the subject (a little pro, I assume) has moved to SpecTP. Since it is not c-commande
by los libros , Case assignment to pro succeeds. Hence, I modify (57) as follow
taking pro to occupy the EPP/subject position SpecTP:
(59) ?[tp pro T Estaba [vp los libros [vp tpro v leyendo tios libros ]]]•••
be-PST.3.SG the books reading
. . . cuando, de pronto, llegó María. (Spanish)
when of soon arrive-PST.3.SG Maria
'He was reading the books. . . when, suddenly, Maria showed up.'
(60) [tp (?Juan) T estaba [vp los libros [vp (*Juan) v leyendo tios libros ]]]...
Juan be-PST.3.SG the books Juan reading
. . . cuando, de pronto, llegó María. (Spanish)
when of soon arrive-PST.3.SG Maria
'He was reading the books. . . when, suddenly, Maria showed up.'
From this discussion, I conclude that defective intervention does exist, and that the
deviant status of Aux-OVS sentence is not to be related to PF parameters (Richards
interested in that perspective, see Fox and Pesetsky (2005), Kayne (2010), López (2009b), Richards (2004,
2007), and references therein.
26 A reviewer points out that similar patterns hold in causative constructions. (For ample discussion, see
Guasti 1996; Moore 1996 and Treviño 1994; Zagona 1988; Zubizarreta 1985 and Torrego 2010.) As the
examples below show, the causee can raise over the matrix subject if and only if the infinitival verb has
raised (see Ordóñez 2007, 2009):
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
436 Á.J. Gallego
In the previous sections, it was noted that although object shift in Romance does not
have a direct semantic effect on the object, there are intriguing similarities between
what is found in Romance and Scandinavian. We have just seen, for instance, that
the object can only undergo movement if the verb moves too, which I have related
to Holmberg's Generalization. In this final section, I would like to consider one fur-
ther trait that Romance and Scandinavian languages share. To advance what I will
be discussing here, let me start this section by quoting Bobaljik and Jonas' (1996)
observation about the connection between object shift and subject positions:
The statement in (61) should be clarified (in particular, the status of "SpecaP", to
which I return below). To begin with, note that, at the time Bobaljik and Jonas
(1996) was written, there were only two subject positions: the base one (Spec VP
or SpecvP) and the derived one (SpecAgrsP). Therefore, SpecTP was considered as
a third subject position. Obviously, the point of Bobaljik and Jonas (1996) was that
while Icelandic licenses this third position, English does not. Thus, what must be
kept in mind is that some languages license three subject positions, not two. The
question now is how to translate such empirical observation into an analysis. Techni-
cally, two well-known strategies can be and have been invoked in order to address a
6 Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 437
The interesting datum is (62c), which shows that VSO order is out in Cata
(2004) points out that the same is found in Italian:
Ô Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
438 Á.J. Gallego
'
In the sa
mance la
(see Dobr
Zubizarre
Belletti (2004) considers two possible explanations for the grammatical statu
VSO: either the relevant language licenses an additional subject position or e
has an additional mechanism to assign Case to the object. Belletti (2004) defe
the latter view, revamping an original proposal by Zubizarreta (1999). Cruci
Belletti's (2004) account is the idea that the projection assigning accusative C
outside the vP and that the object must move to it in order to get its Case ch
Though plausible, the analysis is clearly problematic; why does Case assignmen
quire movement to a Case-checking position (as in Chomsky 1993, 1995), onc
distance Agree is available - and, as we have seen, empirically attested? Like
why is the accusative Case assigning head outside the vP? Since this approac
parts from the system I have assumed so far, I cannot adopt it here.
Ordóñez (2007) investigates VSO in Spanish and Catalan, adopting the opti
missed by Belletti (2004), namely, the hypothesis that Spanish, but not Catala
vides itself with an additional (third) subject position. Building on differen
of evidence, Ordóñez (2007) argues that Spanish is systematically provided w
extra subject position, one that seems to be barred in Catalan. Regular VSO tra
sentences (see (67)), small clauses selected by raising verbs (see (68)), structur
volving modais (see (69)), and infinitival contexts (see (70)) illustrate this asym
27 Interestingly, Mexican Spanish precludes VSO sentences, as noted by Gutiérrez- Bravo (2007). This
author argues that VSO may also be slightly marginal in other non-European varieties of Spanish, like
Puerto Rico's. I thank an anonymous reviewer for informing me about this micro-parametric datum.
Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 439
In Ordonez's (2007) account, Spanish has two subject positions for post-verbal sub-
jects, and only one of them is available in Catalan. Ordonez (2007) relates this third
subject position to Bobaljik and Jonas' (1996) discussion of Icelandic, but he does
not relate the third subject position to having VOS through object shift. Here I want
to connect those facts, and, more particularly, the double strategy to generate VOS
(VP fronting vs. object shift) to the availability of VSO. In a nutshell, I claim that if
a language L can resort to an object shift based derivation for VOS, it licenses the
position necessary to generate VSO. I also claim that the key factor has to do with the
morphological richness of v, which is a locus of parametric variation (like other func-
tional heads, as argued by Belletti and Rizzi 1996; Biberauer et al. 2010; Borer 1984;
Kayne 2000, 2005 and others). Suppose we formalize this hypothesis as (71):28
28 A reviewer points out that the correlation has been previously noted in the literature. Although 1 agree
that the observation that Case-marked direct objects and the availability of VSO are somehow connected
has been made (see Belletti 2004; López 2009a and references therein), 1 believe that the present account
offers a novel picture. Here is why. As we have seen, proposals differ with respect to whether VOS is
derived through object shift or VP fronting, but they never consider the possibility that both derivations
are available, and key to the licensing of the VSO pattern- and thus explain parametric cuts. 1 see it as a
genuine contribution of the present proposal that both routes to generate VOS are presented as possible and
crucially responsible for the possibility to license subjects in VSO sentences. The connection is actually
made stronger the moment the position of objects (in VOS) and subjects (in VOS) is the same, as I contend
here.
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
440 Á.J. Gallego
(7 1 ) VOS-VSO Gene
If a Romance langua
VSO.
(72) a. [Agr-sP SUBJECT Agrs [tp T [focP Focus [vp t¡ V . . . ]]]] SVO
b. [Agr-sP Agrs [tp subject T [FocP Focus [vp t¡ V . . . ]]]] VSO
C. [Agr-sP Agrs [tp t [focP subject Focus [Vp t¡ V . . . ]]]] VOS
According to (72), Spanish lacks in situ subjects (they must all abandon the vP). Fur-
thermore, Ordonez (2007) assumes that an AgrP hosts preverbal subjects (in SVO).
Here I would like to provide a slightly different (but nonetheless compatible) account
of the facts, dispensing with the possibility that there be additional projections. I do
this for two reasons: first, because it will allow me to capture (71) in a more nat-
ural way (I explain why below); second, because the difference between having an
'additional projection' or an 'additional specifier' may ultimately be vacuous.29
The analysis I would like to put forward here, though departing in technical as-
pects from Belletti 's (2004) and Ordonez's (2007), can keep their main contributions.
I agree with Belletti (2004) and Zubizarreta (1999) that there is a connection between
having a special device to assign accusative - the dative Case marker a in Spanish;
see Leonetti (2004), López (2012), Torrego (1995, 1998) - and VSO, but I will not
pursue the hypothesis that Spanish has a special way to assign accusative Case. In-
stead, I want to relate the possibility of having a with having a morphologically richer
v. Let me elaborate. I assume that accusative Case is assigned through an Agree de-
pendency between v and the internal argument cross-linguistically. As noted above
(Sect. 2.2), languages like Spanish (and also Romanian, Kiswahili, Hindi-Urdu and
Persian; see López 2009a, 2012) can display a further process of Differential Object
Marking (DOM), whereby the internal argument is preceded by a particle. There is
a general consensus in the literature in taking DOM to involve movement of the ob-
ject to a derived position (Leonetti 2004; López 2009a, 2012; Torrego 1998). What
this position is remains an open matter, but let us nonetheless assume that it is an
outer specifier of vP (see López 2012 for qualifications), which is above the base-
generation position of the external argument. Assuming this much, I propose that the
SpecaP in (61) corresponds precisely to this outer SpecvP, at least for the case of
Romance. Consequently, I reformulate (61) as follows:
29Both views should ideally converge, which will require a better understanding of features, and how they
may become independent lexical items in some languages but not in others (see Cinque and Rizzi 2008;
Van Craenenbroeck 2009 and references therein).
Ô Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 441
b. [CP c [Tp T [vp SUBJECT [vp tsubject v [vp V object ]]]]] VSO
C. [CP C [Tp subject T [vp tsubject v [vp V object ]]]] SVO
If this analysis is on the right track, then we obtain a rather straightforward connec-
tion between the availability of SpecvP for object shift and DOM and the generation
of VSO sentences - since both dependents occupy the same position. The analysis
further allows us to preserve SpecTP as the position of preverbal subjects, dispens-
ing with projections that appear to have a purely theory-internal status or make no
semantic contribution - agreement projections.
There is, however, a potential problem with the structure in (74a), the one corre-
sponding to postverbal subjects in VOS sentences (so-called "free inversion"). As dif-
ferent authors have argued in the recent literature (Alexiadou and Anagnostopoulou
2001, 2007; Chomsky 2001; Mayr 2007 and references therein), some argument of
30Louise McNally asks a very relevant question: How is DOM licensed in a VSO sentence (since both
dependents presumably compete for the same position)? The system I am advocating for here forces me
to assume that the subject and the Case-marked object would occupy outer specifiers of the vP, roughly as
indicated in (i):
(i) [tp [t [v v vio] T] [vp Juan [vp a María [vp tjuan tv [vp tvio *a Maria ]]]]] (Spanish)
see-PST.3.SG Juan to María
'Juan saw María.'
For an alternative analysis, where the Case-marked object could occupy a lower position, see López (2012).
31 See Gallego (2007, 2010) for an alternative approach, assuming that the external argument is generated
below v, following ideas that go back to Fukui and Speas (1986), Kitagawa (1986), Koopman and Sportiche
(1991), Sportiche (1988), and Hale and Key ser (1993). In such approach, the subject in VSO sentences is
in SpecvP, after it moves from its base-generated position.
Ô Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
442 Á.J. Gallego
I know
of no agreed-u
pecially given that th
external argument mu
sky 1981, 1982). The a
support the (75b) optio
interpretation, then i
a FocusP).32 Belletti (2
adverbs, NPI licensing
occupies a low positio
Costa (2004), Ordóñez
absence of no non theo
subject in VOS occupie
To wind up this sectio
As argued, what (76) is saying is that the double strategy to generate VOS (which
concerns the vP field) in Romance has interesting consequences for the licensing o
additional positions in the vP external area (Chomsky's phase edge, Belletti's low
area, etc.). Very interestingly, it seems that this scenario has interesting correlates
the syntax of the CP field (as argued for by Poletto 2006 on independent ground
32 Though legitimate in and of itself, this approach is not forced upon us, unless we take notions li
"focus", "topic", and the like to be akin to lexical items. For problems with this approach, see Lópe
(2009a), Fanselow and Lenertová (2010), and references therein.
* * Of course, future investigation may show that (75b) is the correct analysis (perhaps because argumen
must always leave their theta positions in the surface structure). If this is so, then the present account w
not require dramatic changes. It would be crucial, though, that the XP projection is not related to obje
shift and DOM, for otherwise the connection between VOS (via object shift) and VSO would be undone.
Ô Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 443
As (79) and (80) indicate, the Catalan counterparts of (77) and (78) are
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
444 Á.J. Gallego
6 Conclusions
This paper has investigated different phenomena concerning the vP syntax of Ro-
mance languages. I have focused on three main issues: (i) the derivation of VOS
£ļ Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 445
(82) [vp [VP V OBJECT] [vp SUBJECT V tvp]]] ROMANCE VOS (VIA VP FRON
The key difference between these two analyses concerns the syntactic relat
tween the object and the subject: Only in an object shift based derivation do
object c-command the subject, and can therefore bind it. The main contribu
this paper in this respect has been the claim that both object shift and VP fr
derivations take place, but in a dedicated parametrized fashion. Evidence has s
that Central-Eastern Romance varieties (Catalan, Italian) resort to the VP fr
derivation, whereas Western Romance varieties (European Portuguese, Galici
Spanish) resort to object shift.
It is precisely the latter varieties that pose a problem for nominative Case
ment. Since the object is moved to a specifier position that intervenes between
cus of nominative Case (T/Infl) and the in situ subject, an intervention effect
obtain. Interestingly (and rather puzzlingly), the intervention effect does not
which requires an explanation. In Sect. 3, 1 have argued that verb movement
the shifted object invisible, which makes it possible for T /Infi to assign nom
Case to the subject. I have adopted a version of Chomsky's (1993, 1995) "equi
tance", whereby the specifiers of the vP are collapsed on locality grounds. N
that, for this to be feasible, some instances of head movement must have syn
effects, as argued by Chomsky (2008), Donati (2006), Den Dikken (2006),
(2010), and Roberts (2010) {contra Boeckx and Stjepanovič 2001; Chomsky
As I have shown, an analysis based on verb movement is empirically superior t
natives that rely on movement of the object to T, Multiple Agree/Co-valuation
doubling, or the nonexistence of defective intervention effects.
The crucial connection between object shift and verb movement has furthe
me to investigate the similarities between Romance and Scandinavian langua
we have seen, the syntactic behavior of some members of these two families
markably similar. On one side, both groups of languages appear to obey Holm
Generalization (Holmberg 1986, 1999), therefore requiring verb movement f
ject shift to take place. This appears to be correct in Western Romance langua
the Spanish sort, where the restriction is purely syntactic. A second property
shared by these two language families concerns the possibility to resort to an
(third) subject position. In the Romance case, this has been shown in the wo
Ordonez (2007), which convincingly argues that Catalan has fewer subject po
than Spanish. In particular, Catalan (like Italian) cannot generate VSO sentence
like Spanish (and Galician and European Portuguese). As the reader can easily
Ô Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
446 Á.J. Gallego
Ordonez's (2007) analysis of VSO proposes that Spanish uses SpecTP as a subject
position, assuming that SpecAgrsP is reserved for SVO. I have put forward an al-
ternative account whereby the third subject position is connected to having a richer
vP periphery, involving an additional specifier. I have chosen this account for two
reasons: first, I discard agreement projections, following Chomsky (1995); second,
linking the subject position of VSO sentences with the object position of VOS sen-
tences provides us not only with a direct connection between these two patterns, but
also with a cluster of empirical observations that concern the Case-agreement systems
and the CP periphery (clitic doubling, differential object marking, laísmo/leísmo, par-
ticipial agreement, auxiliary selection, etc.).
In sum, the data that I have discussed in this paper not only indicate that the para-
metric differences between Romance languages can be accounted for in terms of a
morphological parameter (i.e., a parameter of the micro type; see Belletti and Rizzi
1996; Biberauer 2008; Kayne 2000, 2005; Holmberg and Roberts 2009) that plausi-
bly has v as its locus, but also indicate that there are deep similarities between Ro-
mance and Scandinavian, which require reconsidering the nature of operations such
as verb movement and its impact on locality conditions.
Acknowledgements Previous versions of this paper were presented at the Going Romance XX (Vrije
Universiteit Amsterdam, December 7-8 2006), the Linguistic Symposium on Romance Languages XXXVII
(University of Pittsburgh, March 15-18 2007), the West Coast Conference in Formal Linguistics 26 (Uni-
versity of California, April 27-29 2007), and the 41st Annual Meeting of North East Linguistic Society
(University of Pennsylvania, October 22-24 2010), whose audiences I thank for questions and comments.
This paper has benefited from discussion with Cedric Boeckx, Noam Chomsky, Ricardo Etxepare, Josep
M. Fontana, Tomohiro Fujii, M.Lluïsa Hernanz, Luis López, Francisco Ordóñez, Gemma Rigau, Jaume
Solà, Xavier Vilallba, and Masaya Yoshida. Thanks especially to José M. Brucart and Juan Uriagereka for
discussion, constant support, and encouragement. I am also very grateful to three anonymous reviewers,
whose observations and suggestions were very useful, and contributed to considerably improve different
aspects of the present piece. Finally, I would like to express my gratitude to John Moore and Louise Mc-
Nally for thorough comments, suggestions, and editorial help. All errors are my own. This research has
been partially supported by grants from the Ministerio de Ciencia e Innovación (FF1201 1 -29440-C03-0 1 )
and the Generalitat de Catalunya (2009SGR-1079).
References
Alboiu, Gabriela. 1999. (De-)focusing and object shift in Romanian. Canadian Journal of Linguistics 44:
1-22.
Alexiadou, Artemis, and Elena Anagnostopoulou. 2001 . The subject-in-situ generalization, and the role of
case in driving computations. Linguistic Inquiry 32: 193-231 .
Alexiadou, Artemis, and Elena Anagnostopoulou. 2007. The subject-in-situ generalization revisited. In
Interfaces + recursion = language ? Chomsky's minimalism and the view from syntax-semantics ,
eds. Uli Sauerland and Hans-Martin Gärtner, 31-59. Berlin: Mouton de Gruyter.
Atkinson, Martin. 2001. Defective intervention effects, die! Essex Research Reports in Linguistics 37:
1-29.
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 447
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
448 Á.J. Gallego
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 449
Marantz, Alec. 1997. No escape from syntax: don't try morphological analysis in the privacy of your own
lexicon. In UPenn WPL 4 , eds. Alexis Dimitriadis, Laura Siegel, Clarissa Surek-Clark, and Alexander
Williams, 201-225. Philadelphia: UPenn Publications.
May, Robert. 1977. The grammar of quantification. PhD dissertation, MIT, Cambridge.
May, Robert. 1985. Logical form: its structure and derivation. Cambridge: MIT Press.
Merchant, Jason. 2001. The syntax of silence: sluicing, islands, and the theory of ellipsis. Oxford: Oxford
University Press.
Moore, John. 1996. Reduced constructions in Spanish. New York: Garland.
Motapanyane, Virginia. 1991. Theoretical implications of the complementation in Romance. PhD disser-
tation, Université de Genève, Geneva.
Motapanyane, Virginia, ed. 2000. Comparative studies in Romanian syntax. Dordrecht: Elsevier.
Müller, Gereon. 1998. Incomplete category fronting. Dordrecht: Kluwer.
Ordóñez, Francisco. 1997. Word order and clause structure in Spanish and other Romance languages. PhD
dissertation, City University of New, York, New York City.
Ordóñez, Francisco. 1998. Post- verbal asymmetries in Spanish. Natural Language and Linguistic Theory
16: 313-346.
Ordóñez, Francisco. 2000. The clausal structure of Spanish: a comparative study. New York: Garland.
Ordóñez, Francisco. 2005. Two specs for postverbal subjects: evidence from Spanish and Catalan. Ms.,
SUNY Stony Brook.
Ordóñez, Francisco. 2007. Cartography of postverbal subjects in Spanish and Catalan. In Romance lan-
guages and linguistic theory 2005: selected papers from 'Going Romance ', eds. Sergio Baauw, Frank
A. C. Drijkoningen, and Manuela Pinto, Utrecht. 8-10 December 2005, 259-280. Amsterdam: Ben-
jamins.
Ô Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
450 Á.J. Gallego
& Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms
Object shift in Romance 45 1
£¡ Springer
This content downloaded from 132.248.9.8 on Fri, 24 Aug 2018 23:16:20 UTC
All use subject to https://about.jstor.org/terms