Right Dislocation is not right dislocation*

Xavier Villalba
Departament de Filologia Catalana
Universitat Autònoma de Barcelona
villalba@lingua.fil.ub.es

Kayne’s (1994) restrictive theory of word order and phrase structure rests on the assumption
that a rigid mapping exists between hierarchical structure and linear order. This mapping is
formulated as the Linear Correspondence Axiom, which ensures that linear order is just a
corollary of asymmetric c-command. Such a strong hypothesis has raised the issue of
rightward movement and right-adjoined sites. One construction that becomes crucial for
testing the empirical coverage of such a proposal is Right Dislocation, or more exactly, Clitic
Right Dislocation (CLRD). Sometimes, this construction —when considered at all— has been
analyzed as the mirror image of Clitic Left Dislocation (CLLD); see, for example, Vallduví
(1990). Kayne (1994) rejects such an analysis on the basis of the Linear Correspondence
Axiom, proposing instead that CLRD is the covert counterpart of CLLD, that is, clitic right-
dislocated elements appear in complement position and move to the left at LF. In section 1 I
will offer a significant bulk of data supporting Kayne’s claim against the symmetric analysis
of CLRD. However, in section 2, it will be shown that his analysis has serious empirical
drawbacks as well and cannot be assumed. Finally, in section 3, a more inclusive and
principled alternative will be offered: the Split-Topic Hypothesis —argued for elsewhere
(Villalba 1996a). This proposal shows how the existence of two different leftward landing
sites for dislocated elements —the specifier of the Internal Topic Phrase in the periphery of
the VP area for CLRD, and that of the External Topic Phrase in the Comp area for CLLD— is
capable of offering a principled account for the whole of empirical data respecting the
conceptual restrictions imposed by the LCA.

1. Differences between CLRD and CLLD

1.1. Upward Boundedness
It is a well-known fact that CLLD is unbounded (like wh-movement):1

*
This paper has benefited from intelligent suggestions by Anna Bartra, Albert Fontich, Richard Kayne and
Gemma Rigau. They are, however, exempt from any error remaining in it.
(1) a. D’això, vaig dir que volia que tothom en parlés.2
of this PAST-1 say that wanted-1 that everybody of-it talk-SUBJ-3
‘I said that I wanted everybody to talk about this.’
b. De què1 vaig dir que volia que tothom parlés t1?
of what PAST-1 say that wanted-1 that everybody talk-SUBJ-3
‘About what did I say that I wanted everybody to talk?’

In contrast, since Ross’ thesis (Ross 1986: 258), it is assumed that CLRD is bounded to its
own sentence. However, standard examples don’t allow us to test this claim directly as long
as the contexts commonly adduced are islands for extraction of a dislocate (namely, relative
clauses or subject sentences). Nevertheless, there is an indirect way of testing Ross’ claim.
Consider the following sentence:

(2) Van suggerir a la Maria que anés cap a casa.

PAST-3PL suggest to the Maria that go-SUBJ-3 toward to home
‘They suggested Maria to go home.’

If we dislocate to the right the italicized constituents, the supposed upward boundedness of
CLRD will result in a fixed order of the dislocates, since each occurs in a different sentence.
If both were in the same level, a free ordering between them would be expected (this is a
typical property of both CLRD and CLLD; see Bartra 1985, Cinque 1990, Vallduví 1990,
Villalba 1996a, b). The data confirm Ross’ claim:

(3) a. *Li1 van suggerir anar-hi2, [a la Maria]1, [cap a casa]2.

to-him/her PAST-3PL suggested go-LOC to the Maria toward to home
b. Li1 van suggerir anar-hi2, [cap a casa]2, [a la Maria]1.
to-him/her PAST-3PL suggested go-LOC toward to home to the Maria
‘They suggested Maria to go home.’

1
I will assume without any justification that both CLRD and CLLD are instances of movement. See Villalba
(1996a) for a justification.
2
For the sake of clarity, I will italicize both the dislocate and the resumptive clitic thorough the article. The
glosses are to be interpreted as follows: 1, 2, 3 = 1st, 2nd, 3rd person, FEM = feminine, LOC = locative, MASC =
masculine, NEG = negation, PL = plural, SUBJ = subjunctive.
It seems clear that CLRD is really upward bounded, an unexpected contrast for the symmetric
analysis.

1.2. Creation of Island Effects

It is a well-known fact that CLLD creates island effects:3

??
(4) a. Qui1 creus que, de Cuba, t1 en parla al seu llibre?
whom believe-2 that of Cuba of-it talk-3 in-the his/her book
‘Who do you believe talks about Cuba in his/her book?’
b. *[Amb qui]1 creus que, de Cuba, en parla t1 Chomsky?
with whom believe-2 that of Cuba, of-it talk-3 Chomsky
‘With whom do you believe that Chomsky talks about Cuba?’
c. *Com1/??On1/??En quin llibre1 creus que, de Cuba, en parla t1 Chomsky?
how/where/in which book believe-2 that of Cuba of-it talk-3 Chomsky
‘How/Where/In which book do you believe that Chomsky talks about Cuba?’
(5) *Conec un noi [amb qui]1 (crec que), d’aquest tema, se n’ha de parlar t1.
know-1 a boy with whom (think-1 that) of-this subject, SE of-it-have-3 of talk
‘I know a boy with whom (I think that) someone has to talk about this subject.’

In contrast, CLRD creates no island effect, against the prediction the symmetric analysis
raises:

(6) a. Qui1 creus que t1 en parla al seu llibre, de Cuba?

b. Amb qui1 creus que en parla t1 Chomsky, de Cuba?
c. Com1/on1/en quin llibre1 creus que en parla t1 Chomsky, de Cuba?
(7) Conec un noi amb qui1 (crec que) se n’ha de parlar t1, d’aquest tema.

1.3. Licensing of Negative Polarity Items

Consider the following examples:
(8) a. La Maria *(no) és responsable de res/ningú.
Maria not is responsible of nothing/nobody
‘Maria is not responsible of anything/anybody.’
b. La Maria *(no) és amiga de cap lingüista.
Maria not is friend of no linguist
‘Maria is not friend of any linguist.’
c. La Maria *(no) confia en els consells de gaire gent.
the Maria not trust-3 in the advice of many people
‘Maria does not trust the advice of many people.’

The negative polarity items (NPIs) —in boldface for the sake of clarity— are licensed by
negation. What would happen if the phrase containing the NPI were dislocated?4 CLLD
destroys the context for licensing NPIs:

(9) a. *(De) responsable de res/ningú, la Maria no ho és.

of responsible of nothing/nobody the Maria not it is
b. *(D’)amiga de cap lingüista, la Maria no ho és pas.
of friend of no linguist the Maria not it is
c. *En els consells de gaire gent, la Maria no hi confia.
in the advice of any people the Maria not LOC trusts

On the contrary, CLRD maintains the context for the licensing of NPIs, against the
expectations of a symmetric analysis:

(10) a. La Maria no ho és, (de) responsable de res/ningú.

b. La Maria no ho és, (d’)amiga de cap lingüista.
c. La Maria no hi confia, en els consells de gaire gent.

1.4. The Interaction between CLRD and CLLD

3
This is extensible to similar processes in other languages. See, among others, Rochemont (1989) for Italian and
English, Borer (1995) for Hebrew, Lasnik and Saito (1992) for English, and Müller and Sternefeld (1993) for
German.
4
Dislocation of the NPI itself is deviant as long as dislocation of [-specific] elements is strongly restricted. See
Roca (1992) and Quer (1993).
An interesting asymmetry follows when we take into account the interaction between CLRD
and CLLD. Whereas we can always left-dislocate a constituent from within a right-dislocate,
the converse is never true:

(11) a. [Del meu avi]2, me les1 han explicat totes, [les (seves) històries t2 ]1.
of-the my grandfather to-me them-FEM have-3PL told all the his stories
b. *[Les (seves) històries e2]1, me les1 han explicat totes, [del meu avi]2.
the his stories to me them-FEM have-3PL told all of-the my grandfather
‘They have told me all the stories of my grandfather.’

Here neither CLRD nor CLLD move outside their sentence, so then in ((11)b) there is no
violation of the boundedness condition. The contrast remains without explanation for the
symmetric analysis.

1.5 Nonfinite clauses

As Solà (1992) points out, CLLD cannot freely apply in nonfinite clauses (but we’ll see a
partial exception in section 3 below):5

(12) a. *Confiava a, d’aquest tema, parlar-ne.

expected-1 to of-this subject talk-of-it
a’. *Confiava, d’aquest tema, a parlar-ne.
expected-1 of-this subject to talk-of-it
‘I expected to talk about this subject.’
b. *La Joana anava cap a casa, d’aquest tema, parlant-ne.
Joana went-3 toward to house of-this subject talking-of-it
‘Joana went home talking of this subject.’

CLRD, however, can freely apply in the same contexts:

5
Note, however, that CLLD is fine if the dislocate is in the periphery of the main clause:

(i) D’aquest tema, confiava a parlar-ne.

of-this subject expected-1 to talk-of-it
‘I expected to talk about this subject.’
(13) a. Confiava a parlar-ne, d’aquest tema.
expected-3 to talk-of-it of-this subject
b. La Joana anava cap a casa parlant-ne, d’aquest tema.
Joana went-3 toward to house talking-of-it of-this subject

Once more, the contrast remains a mystery for the symmetric analysis.

1.6. Informational status

Until here we have just taken into account the syntactic differences between CLRD and
CLLD. However, the proposal that they are the mirror image of each other is weakened as
well when we analyze their pragmatic properties. It has been pointed out by several authors
(Benincà 1988, Vallduví 1990, 1995, among others) that whereas CLLD can introduce a new
topic in the discourse, CLRD cannot:

(14) A: On va posar les coses? B: Em sembla que...

where PAST-3 put the things to-me seems that
‘Where did (s)he put the things?’ ‘It seems to me that...’
a. els llibres, els va posar al despatx.
the books them-MASC PAST-3 put in-the study
b. #els va posar al despatx, els llibres.
them-MASC PAST-3 put in-the study the books
‘...(s)he put the books in the study.’

Here CLRD is not felicitous because els llibres ‘the books’ has not been previously
introduced in the discourse. Compare with the following:

(15) A: On va posar els llibres? B: Em sembla que...

where PAST-3 put the books to-me seems that
‘Where did (s)he put the books?’ ‘It seems to me that...’
a. els llibres, els va posar al despatx.
the books them-MASC PAST-3 put in-the study
b. els va posar al despatx, els llibres.
them-MASC PAST-3 put in-the study the books
Here CLRD becomes perfect as long as els llibres ‘the books’ has already been introduced in
the discourse.
A similar informational contrast arises with shift-topics (also called ‘contrast topics’).
Consider the following contrast:

(16) Hi havia un home i una dona.

‘There was a man and a woman.’
a. A ell, li van regalar un cotxe però a ella, li van comprar un vestit.
to him to-him/her PAST-3PL give a car but to her to him/her PAST-3PL buy a dress
b. #A ell, li van regalar un cotxe però li van comprar un vestit, a ella.
to him to-him/her PAST-3PL give a car but to him/her PAST-3PL buy a dress to her
c. #Li van regalar un cotxe, a ell, però li van comprar un vestit, a ella.
to-him/her PAST-3PL give a car to him but to him/her PAST-3PL buy a dress to her
‘They gave him a car but they bought her a dress.’

Shift topics are only possible with CLLD.

Having these differences in mind, it seems quite plausible to assume Vallduví’s (1990,
1995) proposal that CLRD is a focalization process, in the sense that it removes an element (a
“tail” in Vallduví’s terms) from its original place, allowing another element to receive focus
—we’ll show in section 3 below that something more has to be said. The following example
illustrates the process:

(17) Què va trobar la Joana a casa?

what PAST-3 find Joana at home
‘What did Joana find at home?’
a. #Va trobar un llibre a casa.
PAST-3 find a book at home
‘It was at home that she found a book.’
b. Hi va trobar un llibre, a casa.
there PAST-3 find a book at home
c. A casa, hi va trobar un llibre.
at home there PAST-3 find a book
 ‘It was a book that she found at home.’

Focus is assigned to the last element in the sentence. ((17)a) is not felicitous since it is a
thematic element, a casa ‘at home’, that receives focus interpretation. In order to allow the
correct element, un llibre ‘a book, to receive focus, the disturbing element appears in a right-
dislocated position, as in ((17)b). It is in this very special sense that we can say that CLRD is
a focalization process. CLLD, on the other hand, does also have this role, as ((17)c) shows.
Nevertheless, it also fulfills other tasks in the informational level, namely, introducing a new
or shift topic, as we have just shown. We’ll turn back to this difference in 3.1 below.

2. Against Kayne’s Analysis

The bulk of empirical data shown in section 1 strongly argues against the symmetric analysis:
none of the contrasts described between CLRD and CLLD can easily be accommodated under
such an analysis. Moreover, it also has a serious and obvious theoretical drawback: it doesn’t
comply with the Linear Correspondence Axiom. It is this second point that drives Kayne to a
new proposal: CLRD is the covert counterpart of CLLD. According to Kayne (1994), right-
dislocates are in fact elements in complement position that only at LF would move to the
same position overtly occupied by left-dislocates. His proposal is very appealing, since it
draws a parallelism with other similar overt/covert contrasts, and explains the boundedness
condition on CLRD. However, it has serious drawbacks.
First of all, complements in Catalan show a fixed order. Under Kayne’s analysis, we only
expect CLLD to display free word order: if right-dislocates were in complement position, we
would wrongly predict that they must pattern with complements, displaying a fixed order. The
data are clear in this point, and CLRD patterns with CLLD in showing a free ordering (I make
use of the traditional terms D(irect) O(bject) and I(ndirect) O(bject) as a mere descriptive
label):

(18) a. Els llibres1, al Pere2, els1 hi2 vam donar. DO-IO-V

the books to-the Pere them-MASC to-him/her PAST-2PL give
b. Al Pere2, els llibres1, els1 hi2 vam donar. IO-DO-V
c. Al Pere2, els1 hi2 vam donar, els llibres1. IO-V-DO
d. Els llibres1, els1 hi2 vam donar, al Pere2. DO-V-IO
 e. Els1 hi2 vam donar, els llibres1, al Pere2. V-DO-IO
f. Els1 hi2 vam donar, al Pere2, els llibres1. V-IO-DO
‘We gave Pere the books.’

Moreover, if right-dislocates were in complement position before spell-out, it would be

difficult to explain why they behave as islands for extraction: Kayne’s analysis incorrectly
predicts that extraction from a right-dislocate should be as good as from a complement, since
both are in the same position. Again, data are conclusive, and no difference follows between
CLRD and CLLD:

(19) a. *De què1 creus que, [(de) responsable t1], no ho és pas?

of what think-2 that of responsible not it is NEG
b. *De què1 creus que no ho és pas, [(de) responsable t1]?
‘What do you think (s)he is not responsible of?’
(20) a. *Tinc un amic de qui1 [responsable t1], no m’hi considero pas.
have-1 a friend of who responsible not to-me-LOC consider NEG
b. *Tinc un amic de qui1 no m’hi considero pas, [responsable t1].
‘I have a friend that I do not consider myself responsible of.’

Thirdly, the interaction between CLRD and CLLD shown in paragraph 1.4 gives us the
following paradigm:

A) both CLRD and CLLD are possible from a complement;

B) CLLD is possible from a right- or left-dislocate;
C) CLRD is not possible from a left-dislocate.

Kayne’s analysis correctly predicts situations in A) and B), but has no principled way to
account for C). If it were the case that right-dislocates occupied a complement position before
their covert moving to the CLLD position, then there would be no difference between case B)
and C) excepting the overt/covert nature of the movement. This contrast is even more
puzzling for Kayne’s analysis when put together with the lack of contrast between CLRD and
CLLD with respect to wh-extraction, we have just seen. Obviously, something else has to be
said in order to explain the contrast.
 Furthermore, the unavailability of CLLD with non-finite clauses, shown in paragraph 1.5,
cannot easily be explained by means of an overt vs. covert movement analysis: if both CLLD
and CLRD involved movement to the same position, it would be unclear why they should
contrast with respect to nonfinite clauses. This contrast is a mystery for Kayne’s proposal.
Finally, the covert movement of the right-dislocate is very dubious on conceptual grounds.
The left-peripheral position of CLLD has to do with its informational status as a new/shift
topic —it is worth noting that this concrete correlation between informational role and
position is universal: all languages place their new/shift topics in the left periphery. Since
CLRD doesn’t bear this informational role, it is not clear what the motivation would be for
such a covert operation. His assumption of an optional feature forcing both visible —
invisibility for the mechanism assigning focus and a special intonation— and invisible effects
—movement at LF— is just a machinery without clear motivation. On a minimalist view of
language the covert movement of the right-dislocate is suspicious, at least.
To sum up, neither the symmetric analysis nor Kayne’s one can offer a principled
explanation for the set of properties of CLRD. A new analysis is called for.

3. The Split-Topic Analysis

In this section I will argue that the basic difference between CLRD and CLLD lies in the
different position they occupy in the clause. Whereas CLRD moves to the specifier of the
Internal Topic Phrase (IntTopP) just over the VP, CLLD moves to a higher one, that of the
External Topic Phrase (ExtTopP) in the Comp area (see Rizzi 1995). Schematically:

(21) inserir Fig. 1

Let us see how this clause structure can easily derive the whole bulk of empirical contrasts.

3.1. CLRD and Word Order

The basis for the distinction between CLRD and CLLD is the position in the clause structure:
whereas CLLD appears in a functional projection very high in the structure, CLRD appears
quite low. All the differences between these two structures will be derived from this contrast.
Nevertheless, such an analysis raises interesting questions with respect to word order. Once
we assume that the LCA is a principle of UG, if CLRD is movement, it must be left-
movement. However, the resultant linear order must have the right-dislocate in the right
boundary of the sentence. How is this conflict to be solved?
We have seen in paragraph 1.6 that CLRD should be analyzed as a focalization process:
the right-dislocated element is moved in order to allow the constituent carrying new
information to receive focus. If we follow different scholars (see Belletti and Shlonsky 1995,
Uriagereka 1995, among others) in considering focalization the result of movement to the
specifier of a Focus Phrase (FocP), then something more has to be said. CLRD would still be
a necessary condition for focalization, since it extracts the presupposed (and hence non-
focalizable) part of a larger non-presupposed constituent —generally, the VP—, otherwise we
would have the clash of a focalized constituent containing a presupposed subpart.
Nevertheless, CLRD is not a sufficient condition, since movement of the non-presupposed
remnant to [Spec, FocP] is still necessary for it to receive focus (or to check a [+focus]
feature, in minimalist terms). Schematically:

(22) inserir Fig. 2

As long as the partition of the information carried by the sentence has to follow such a
derivation, we expect the right-dislocate to correctly surface on the right edge of the sentence
regardless of it having been moved to the left.

3.2. Explaining the Differences

Note that, although the left-movement analysis is forced by conceptual reasons —namely, the
LCA—, the position of both IntTopP and ExtTopP is to be based on empirical grounds. This
paragraph will be devoted to showing that the Split-Topic Hypothesis correctly predicts the
bulk of differences between CLRD and CLLD.
To begin with, let us make the plausible assumption that dislocation is constrained by
economy principles, hence its application must be as cheap as possible. In the case of CLRD,
which, as we have seen in paragraph 1.6, is just a focalization process (in the sense of
Vallduví 1990, 1995), the dislocate —a presupposed element— has to be placed outside the
constituent carrying new information. The shortest position for it to be placed is [Spec,
IntTopP], which qualifies as an A’ position hosting presupposed material. Since this
movement is the shortest, it must be the case that it will be the only one possible, according to
the Minimal Link Condition (see Chomsky 1995). As a result, CLRD must be upward
bounded: movement to the [Spec, IntTopP] of a higher sentence (i.e. unbounded CLRD)
would be less economical since it will fulfill the same function with a longer movement. In
the case of CLLD, things are different. The placement of the left-dislocate is not just a
focalization process: it has to do with other discourse requirements (like being a new/shift
topic; see paragraph 1.6), allowing it to appear in a higher position in the Comp area, namely
the [Spec, ExtTopP] either of the lower sentence or of a superior one, depending on its scope
as a new/shift topic.
Consider now the lack of strong island effects created by CLRD in contrast with CLLD.
Relativized Minimality might be an explanation: [Spec, ExtTopP] is an A’-specifier, so, when
filled, it will create a minimality barrier for A’-movement from within IP, blocking wh-
movement.6 However, something has to be said about CLRD, which also involves a filled A’-
specifier: [Spec, IntTopP]. Shouldn’t it block A’-movement as well? In fact, the two cases are
identical for relativized minimality. The crucial point is that the minimality barrier created by
a filled [Spec, IntTopP] is bypassed when moving the VP to [Spec, FocP] to receive focus:
extraction from within the moved VP will be unaffected by the presence of a right-dislocate.
Obviosuly, we have to make sure that the movement of the VP itself is allowed. The solution
may have to do with L-relatedness: whereas Foc is commonly assumed to be a head with V-
features, this cannot be extended to IntTop. Then, we may say that L-related movement to
FocP is unaffected by the presence of the non-L specifier of IntTop. Schematically:

(23) [L’ ... [L(FocP) [VP ...twh...]... [L’(IntTopP) ... [...tVP... ]]]]

To sum up, CLRD does create a barrier, but its low position permits the bypassing strategy
just suggested. Such an option is not available for CLLD, since ExtTopP is too high in the
structure.
Consider now the licensing of NPIs. We have seen in paragraph 1.3 that, whereas CLLD
destroys the context for the licensing of NPIs, CLRD doesn’t. Our analysis makes the correct
prediction without any additional stipulation. The low position of IntTopP straightforwardly
explains the licensing of NPIs: it is still under the scope of negation. In the case of CLLD,
ExtTopP is too high in the structure to allow licensing by negation.

6
See Lasnik & Saito (1992) for a different explanation based on a (modified) barriers framework and on the
assumption that Topicalization/CLLD is adjunction to IP.
 The interaction between CLRD and CLLD surveyed in paragraph 1.4 is another empirical
issue that follows from our analysis without any additional assumption. Consider again the
possible configurations:7

A) both CLRD and CLLD are possible from a complement;

B) CLLD is possible from a right- or left-dislocate;
C) CLRD is not possible from a left-dislocate.

Leaving aside case A), which is unproblematic, it is obvious that neither B) nor C) pose any
problem to the split-topic analysis, rather they directly follow from it. On the one hand, case
B) is perfect since CLLD will always be movement to a higher [Spec, ExtTopP]. On the other
hand, C) is a clear instance of improper —i.e. downward— movement, given the relative
position of ExtTopP and IntTopP in ((21)).
Finally, something has to be said with respect to the impossibility of having CLLD in
nonfinite clauses (paragraph 1.5). The problem might have to do with clause structure. Let us
assume that nonfinite clauses lack part of the higher functional structure present in finite ones.
If so, there is no problem with CLRD, but this would affect ExtTopP. This is just a
speculation, but examples like the following show that it might be on right track:

(24) a. No sé, d’aquest pernil, on comprar-ne.

not know-1 of this ham where buy-of-it
b. *No sé on, d’aquest pernil, comprar-ne.
not know-1 where of this ham buy-of-it
‘I don’t know where to buy some of this ham.’

The appearance of the wh-element and hence of a richer functional structure over the
nonfinite verb seems to license CLLD.8 More research is, however, needed.

7
It is still to be explained why dislocates are opaque domains for wh-movement but not for dislocation.
8
Nevertheless, note that a wh-relative doesn’t rescue CLLD:

(i) *No tinc ningú amb qui, d’aquest tema, parlar-ne.

not have-1 nobody with who of-this subject talk-of-it
‘I don’t have anybody to talk about this subject.’
4. Conclusion
This article has offered a comprehensive analysis of dislocation processes in Catalan. It has
been shown that CLRD and CLLD have a set of distinguishing properties which cannot be
explained neither by an analysis taking CLRD to be the mirror image of CLLD nor by one
taking CLRD to be the covert counterpart of CLLD (like that of Kayne 1994). Instead, it has
been shown that the Split-Topic Hypothesis, which postulates two topic positions —Internal
Topic Phrase in the periphery of the VP, and ExtTopP in the Comp area—, is capable of
giving a principled answer to the otherwise surprising asymmetries between CLRD and
CLLD.

References
Bartra, A. 1985. Qüestions de la sintaxi d’ordre en català. Universitat Autònoma de
Barcelona. PhD dissertation.
Belletti, A.; Shlonsky, U. 1995. The order of verbal complements: A comparative study.
Natural Language & Linguistic Theory 13: 489-526.
Benincà, P. 1988. L’ordine delle parole e le costruzioni marcate. In: Lorenzo Renzi (ed.).
Grande grammatica italiana di consultazione. Bologna: Il Mulino, vol. I, 115-225.
Borer, H. 1995. The ups and downs of Hebrew verb movement. Natural Language &
Linguistic Theory 13: 527-606.
Cinque, G. 1990. Types of A’-dependencies. Cambridge, Mass.: MIT Press.
Chomsky, N. 1995. The minimalist program. Cambridge, Mass.: MIT Press.
Kayne, R. 1994. The antisymmetry of syntax. Cambridge, Mass.: MIT Press.
Lasnik, H.; Saito, M. 1992. Move α. Cambridge, Mass.: MIT Press.
Müller, G.; Sternefeld, W. 1993. Improper movement and unambiguous binding. Linguistic
Inquiry 24: 461-507.
Quer, J. 1993. Estructures dislocades i quantificadors. Llengua & Literatura 5: 393-415.
Rizzi, L. 1995. The fine structure of the left periphery. Université de Genève: manuscript.
Roca, F. 1992. On the licensing of pronominal clitics. Universitat Autònoma de Barcelona.
Master Thesis.
Rochemont, M. 1989. Topic islands and the subjacency parameter. Canadian Journal of
Linguistics 34: 145-170.
Ross, J. 1986. Infinite syntax. Norwood, N.J.: Ablex.

I have no explanation for this contrast.

Solà, J. 1992. Agreement and subjects. Universitat Autònoma de Barcelona. Ph.D.
dissertation.
Uriagereka, J. 1995. An F position in Western Romance. In: K. É. Kiss ed.. Discourse
Configurational Languages. New York: Oxford University Press, 153-75.
Vallduví, E. 1990. The informational component, University of Pennsylvania. Ph.D.
dissertation.
Vallduví, E. 1995. Structural properties of information packaging in Catalan. In: K. É. Kiss
ed.. Discourse Configurational Languages. New York: Oxford University Press, 122-
52.
Villalba, X. 1996a. Leftward right dislocation. Universitat de Barcelona. Unpublished
manuscript.
Villalba, X. 1996b. Sobre la dislocació a la dreta. Llengua & Literatura 7: 209-234.