Left and Right Dislocation Asymmetries in Spanish∗
Jerid Francom
University of Arizona
1 Introduction
The literature on dislocation in Romance has been predominately based on
data from Left Dislocation(LD) (1) to the exclusion of Right Dislocation
(RD) constructions (2) until recently.1
(1) Esa
that
pel´ıculai,
film
yo
I
no
no
lai
it-ACC
hab´ıa
had
visto
seen
nunca.
never
‘That film, I hadn’t ever seen it.’
(2) Yo
I
no
no
lai
it-ACC
hab´ıa
had
visto
seen
nunca,
never
esa
that
pel´ıculai.
film
‘I hadn’t ever seen it, that film.’
Kayne’s (1994) influential proposal that much of what had been under-
stood as purely motivated in syntax is rather a conflation of processes related
directly to the output interface with Phonetic Form(PF) has had enormous
consequences on the line of inquiry of modern syntax, and has attracted
a renewed interest in contrastively evaluating left and right dislocate con-
structions in Romance. In this work (Linear Correspondence Axiom(LCA))
Kayne makes the prediction that all word order variation can be reduced
to hierarchical structure based on asymmetric c-command and leftward dis-
placement. Under Kayne’s (1994) formulation of the LCA the process that
derives the surface position of right dislocates is viewed as the covert coun-
terpart to that deriving left dislocates and therefore occupying argument
position at PF and the same position as LD at Logical Form(LF).
In more recent work based on Italian (Cecchetto 1999) and Catala´n (Vil-
lalba 1999) syntactic, semantic and phonetic evidence is given suggesting
∗Many thanks to my committee Simin Karimi, Sandiway Fong and Mike Hammond. I
am also grateful to Antxon Olarrea, Mercedes Tubino, Yosuke Sato and Stacey Oberly for
helpful comments and criticism. Also thanks to my informants, Claudia Valdez, Evelyn
Dura´n, Maite Correa and Miriam Dı´az for all the help.
1Bold type-face is used throughout this article for presentation purposes only.
1

that Kayne’s implementation of the LCA for dislocation cannot be main-
tained for Romance. The arguments contend that Unboundedness, Recon-
struction and Prosodic Contour Effects reveal important asymmetries be-
tween LD and RD. Villalba’s (1999) ‘Split-Topic Analysis’(STA) proposes
an LCA licensed account directed towards 1) unifying dislocation as a overt
operation and 2) distinguishing the syntactic and semantic properties of dis-
location as directly related to syntactic position at PF. Under this proposal
dislocation uniformly involves movement of arguments of the matrix clause
outside VP, left and right dislocates occupy unique structural positions at
PF and LF is disposed with as needed in deriving dislocation constructions.
Although the data convincingly show that LD and RD are distinct, the STA
fails to provide a principled account for 1) the unboundedness of left dislo-
cate elements, 2) the semantic properties of multiple dislocates and 3) the
seemingly optional prosodic break from the matrix clause for left dislocates.
The current proposal aims to further disassociate Left and Right Dis-
location contra Kayne (1994). However, I suggest that although basically
on the right track, the formulation of Villalba’s (1999) STA is not sufficient
to account for the evidence presented from Complex NPs, Bare Nouns and
the semantics of dislocation strategies from Spanish. The data points to
the hypothesis that LD patterns with two distinct preverbal dislocations:
Contrastive Left Dislocation that is 1) derived by movement from a clause
internal position and 2) contrastive and Hanging Left Dislocation (HLD)
that is 1) base-generated 2) non-contrastive and 3) prosodically dislocated.
Therefore the asymmetric view proposed here highlights two strategies for
left dislocates and only one for right dislocates. This approach asserts that
unboundedness data and prosodic contours present in LD that were previ-
ously unaccounted for under the STA are properties of the HLD and not
CLD. In addition, I argue that the dual nature of left dislocates provides
a distinction useful in discerning the Information Structure of Spanish and
claim that the critical distinction between left and right dislocates generated
by movement is the same property that marks Topics in Spanish. Thus, the
impoverished asymmetry supported by the STA is augmented to include a
new position for HLD, and reformulated to account for the semantics and
pragmatics of the periphery in syntax.
The organization of the paper is as follows. First, I will outline the the-
oretical context and assumptions that will be needed to critically evaluate
dislocation. Second, the empirical arguments previously weighed against
Kayne’s proposal that RD is the covert counterpart of LD is provided look-
ing at Unboundedness, Reconstruction and Prosodic Contour data. Next, in
§4 both the syntactic and semantic motivations are given that support these
earlier claims culminating into the ‘Split-Topic Analysis’(STA) of Villalba
(1999). Although the STA in its basic structure provides empirical coverage
2

for much of the data, gaps are left in the paradigm. Turning to the current
analysis in §5 evidence from Complex NPs, Bare Nouns and the semantics
of dislocation is given that further disassociates left and right dislocates. As
a corollary this same data points to a further asymmetry within the left
periphery splitting left dislocation into two syntactic and semantic items.
In §5.3 syntactic and semantic framework is elaborated to provide a well-
founded theoretical ground from which the current data and that previously
unaccounted for may both be accommodated. After the main proposal I
address the relevant extensions of this view for Information Structure in
Spanish §5.4. I conclude with a brief discussion of the further implications
of this proposal for syntactic theory and suggest how further lines of inquiry
as to the nature of dislocation may proceed.
2 Empirical Background
Dislocation is a syntactic construction in which a verbal argument appears
detached or ‘dislocated’ from the matrix clause. This occurs in two basic
ways 1) the detached phrase appears left of the matrix clause as in (1)
repeated here as (3) or 2) right of the matrix clause as in (2) repeated here
as (4).
(3) Esa
that
pel´ıculai,
film
yo
I
no
no
lai
it-ACC
hab´ıa
had
visto
seen
nunca.
never
‘That film, I hadn’t ever seen it.’
(4) Yo
I
no
no
lai
it-ACC
hab´ıa
had
visto
seen
nunca,
never
esa
that
pel´ıculai.
film
‘I hadn’t ever seen it, that film.’
Coindexation of a clitic, here la, is always required. Without the occur-
rence of this clitic and coindexation the grammatical (5) and (6) become
ungrammatical.2
(5) Esa
that
pel´ıculai,
film
no
no
*(lai)
(it-ACC)
hab´ıa
had
visto
seen
nunca.
never
‘That film, I hadn’t ever seen (it).’
(6) No
no
*(lai)
(it-ACC)
hab´ıa
had
visto
seen
nunca,
never
esa
that
pel´ıculai.
film
‘I hadn’t ever seen (it), that film.’
2The example in (5) does have a grammatical reading, however this is read as Con-
trastive Focus. This construction is characteristically marked with emphatic stress, quan-
tificational in nature and is never coindexed with other elements of the clause. This
construction falls outside of the scope of this paper, and I only make brief reference to it
§5.3.
3

Yet the referential expression esa pel´ıcula ‘that film’ is not required for
a well-formed sentence with the same propositional meaning, as seen in (7).
(7) No
no
lai
it-ACC
hab´ıa
had
visto
seen
nunca.
never
‘I hadn’t ever seen (it).’
In part due to the data in (5 - 7) that shows that the argument struc-
ture itself stands alone, dislocates have been argued to be clause external
(Vallduv´ı 1992), (Cinque 1990). Further evidence comes from the internal
argument structure of a verb like dar ‘to give’ which appears fixed when
introduced in an ‘out-of-the-blue’ context; the direct object(DO) precedes
the indirect object(IO) in canonical forms.3
3Of note here is that the grammaticality of examples (8) and (9) depends on ‘informa-
tional equivalence’ rather than ‘propositional equivalence’. Ordo´n˜ez (1998) suggests that
this alternation is evidence that Spanish is a Scrambling language which is sensitive to the
semantics of the internal arguments, specifically definiteness. The evidence supporting
the DO/IO structure in (8) as the canonical structure comes from the indefinite/definite
contrast in dative arguments, seen in (i) and (ii).
(i) El
the
director
director
(les)
them
entrego´
gave-3sg
[las
the
notas
grades
a
to-DAT
unos
some
estudiantes].
students
‘The director gave the grades to some students.’
(ii) El
the
director
director
*?(les)
them
entrego´
gave-3sg
[a
to-DAT
los
the
estudiantes
students
las
the
notas].
grades
‘The director gave the grades to the students.’
(Ordo´n˜ez 1998)
It is also shown by Zubizarreta (1998) that structural change in internal arguments
lacking clitic counterparts, namely locatives, is more telling when prosodic contour effects
are held constant.
(iii) Escondio´
hid
el
the
nin˜o
boy
el
the
libro
book
debajo
under
de la
the
cama.
bed
‘The boy hid the book under the bed.’
(iv) * Escondio´
hid
el
the
nin˜o
boy
debajo
under
de la
the
cama
bed
el
the
libro.
book
(Zubizarreta 1998)
This is not the case in some other Romance languages such as Catala´n which possess
a richer clitic system (including locative clitics). Here the DO/IO ordering is uniformly
ruled ungrammatical.
(iii) Fiquem
put-1pl
el
the
ganivet
knife
al
in-the
calaix.
box
‘We put the knife in the box.’
4

When IO precedes DO an unfelicitous utterance results. This contrast
is shown in (8) and (9).
(8) Dimos
gave-1pl
el
the
cuchillo
knife
a
to a
un
boy
nin˜o.
‘We gave the knife to a boy.’
(9) # Dimos
gave-1pl
a
to a
un
boy
nin˜o
the
el
knife
cuchillo.
‘We gave a boy the knife.’
However, coindexation of the clitic (10) and a prosodic break between el
cuchillo ‘the knife’ and boy ‘boy’ restores grammaticality the IO/DO form
(9),
(10) Loi
it-ACC
dimos
gave-1pl
a
to a
un
boy,
nin˜o,
the
el
knife
cuchilloi.
‘The knife, we gave it to a boy.’
while the clitic blocks the application of our original canonical form in which
the IO and DO are in argument position.
(11) * Loi
it-ACC
dimos
gave-1pl
el
the
cuchilloi
knife
a
to a
un
boy
nin˜o.
‘We gave the knife to a boy.’
This suggests that the argument el cuchillo ‘the knife’ appears in an
A’-Position and is in fact right dislocated from the matrix clause. A parallel
description of (10) accounts for left dislocation seen in (12).
(12) El
the
cuchilloi,
knife
loi
it-ACC
dimos
gave-1pl
al
to the
nin˜o.
boy
‘The knife, we gave it to the boy.’
In addition, left and right dislocation may occur recursively. Multiple
dislocations occur to the left or right of the matrix clause seemingly in free
distribution.
(13) El cuchilloi al nin˜oj sej loi dimos.
(14) Al nin˜oj el cuchilloi sej loi dimos.
(iv) * Fiquem
put-1pl
al
in-the
calaix
box
el
the
ganivet.
knife
(Vallduv´ı 1992)
5

(15) El cuchilloi sej loi dimos, al nin˜oj .
(16) Al nin˜o sej loi dimos, el cuchilloi.
(17) Sej loi dimos, el cuchilloi, al nin˜oj .
(18) Sej loi dimos, al nin˜oj , el cuchilloi.
Of note here is the fact that dative arguments have the same syntactic
distribution as other dislocates in the respects discussed here, yet may op-
tionally cooccur and coindex with a clitic while in argument position as in
(18).4
(19) Lei
him-DAT
dimos
gave-1pl
la
the
manzana
apple
a
to
mi
my
hermanoi.
brother
‘S/he gave the apple to my brother.’
However, (19) can only be the answer to a question such as ‘Who did
you give the apple to?’ and never ‘What did you give to my brother?’.
The opposite is true of right dislocates as in (20) highlighting the distinct
structural properties of these constructions.
(20) Lei
him-DAT
dimos
gave-1pl
la
the
manzana,
apple
a
to
mi
my
hermanoi.
brother
‘ To my brother, s/he gave the apple.’
Hence, the empirical properties of dislocation seen here include detach-
ment of an XP and coindexation with a clause internal clitic pronoun, the
possibility of simultaneous detachments which are not obviously internally
structured and propositional equivalence with corresponding non-dislocate
constructions. These properties provide a general description of the distri-
bution of dislocates, now I turn to the theoretical framework.
4It may also be of interest to point out that the clitic in these constructions, in which
the indirect object is specific/definite, is not optional whereas the referential expression
is.
(i) * Dimos
gave-1pl
la
the
manzana
apple
a
to-DAT
mi
my
hermano.
brother
(ii) Le
him-DAT
dimos
gave-1pl
la
the
manzana.
apple
‘He have him the apple.’
Furthermore, in double object constructions where the direct object is cliticized the
clitic le becomes se.
(ii) Sei
him-DAT
la
it-ACC
dimos
gave-1pl
a
to
mi
my
hermanoi.
brother
‘He gave him the apple.’
This alternation is seen as phonological and will not be relevant to the current analysis.
6

3 Theoretical Framework
3.1 Information Structure
Turning to the theoretical context, dislocation has been discussed in the liter-
ature as a discourse motivated phenomenon (Prince 1985), (Vallduv´ı 1992),
(Cinque 1990) related to the Topic—Comment/Focus/Open Proposition di-
chotomy (Halliday 1967), (Givo´n 1983), (Prince 1986), (Lambrecht 1994),
(Casielles-Sua´rez 2004). Common to the different approaches on clausal
organization of information, or Information Structure is a ‘Informative’/
‘Anchoring’ distinction in which the topical element(s) of the phrase serve
as an addressing frame, or ‘anchor’ for where the informative content com-
ponent should be stored; ultimately for retrieval. What is controversial is
how a given clause is divided into these subparts.5 Although the discussion
of the finer points that contrast these position is outside the scope of the
paper, there is agreement that the verbal phrase and its arguments indicate
to the hearer what information should be regarded as informative and the
sentence initial position plays a special part in anchoring this information to
an addressable location often serving as theoretic and empirical motivation
for the syntactic construction.
Within this simplification of Information Structure, left dislocated phrases
in Left Dislocation have been traditionally considered a/the Topical element
of a given phrase.
(21) Esa
that
pel´ıculai,
film
no
no
lai
it-ACC
hab´ıa
had
visto
seen
nunca.
never
‘That film, I hadn’t ever seen it.’
Therefore, the LDed element esa pel´ıcula is the reference point where
the informative no la hab´ıa visto nunca is stored. In other words, what this
phrase is about is ‘a film’ and ‘not having seen it.’ has been added by the
speaker to the list of properties for this film.
On the other hand, a canonical form of (21) seen in (22) is not about
‘a film’ but rather the person that hasn’t seen the film, of which not having
seen the film added to the list of properties of that person.
(22) No
no
hab´ıa
had
visto
seen
nunca
never
esa
that
pel´ıcula.
film
‘I hadn’t ever seen that film, ’
5There are at least three views on how the Topic-Comment/ Focus-Open Proposition
is encoded: 1) there is only one articulation either Topic-Comment or Focus-Open Propo-
sition ((Halliday 1967), (Givo´n 1983), (Prince 1986)), (Lambrecht 1994)) 2) that there is a
tripartite system which combines both articulations into one (Vallduv´ı 1992) and 3) it has
been proposed for Spanish that there are two independent articulation systems that do
not cooccur (Casielles-Sua´rez 2004). The latter will be addressed later in the discussion.
7

In the case of right dislocated elements, these elements have been an-
alyzed as in some sense the ‘Anti-topic’ ((Chafe 1976), (Lambrecht 1981))
replacing the ‘add to’ address with a ‘replace’ address entry (Vallduv´ı 1992).6
(23) No
no
lai
it-ACC
hab´ıa
had
visto
seen
nunca,
never
esa
that
pel´ıculai.
film
‘That film, I hadn’t ever seen it.’
This brief introduction to Information Structure highlights the concensus
in the literature that there is good reason to believe that the surface structure
of phrases is organized based in part on the presentation of information.
However, it also has been seen that there are still conflicting views on the
correct way to segment the articulation of information. The goal, then, is
to accomodate syntactic and pragmatic considerations into a unified theory.
In this endeavor dislocation constructions provide a particularly adequate
testing ground as I hope to show.
3.2 Syntactic
Kayne’s (1994) Linear Correspondence Axiom(LCA) will serve as the work-
ing syntactic framework of the paper. This hypothesis offers a very restricted
view of language in which the linearization of phonetic output into words
and morphemes is a function of antisymmetric c-command relations in syn-
tax. Thus asymmetric c-command maps to linear precedence in a direct
structure-to-word order manner conspires to produce a syntactically well-
formed phrase. In this view, (24) is structurally unambiguous and leads to
phonetic linearization, however (25) reveals a mutual c-command relation
between M and P thus will not linearize and is excluded from the grammar.
(24) Antisymmetry K
J
j
L
M
see
N
P
John
(25) Symmetry * K
J
j
L
M
see
P
John
6However, we shall see later in fact that this is somewhat controversial.
8

Multiple adjunctions to heads (26) or non-heads (27) may occur under
the condition that the LCA is not violated.
(26) Head Adjunction L
M
K
k
M
Q
q
M
m
P
R
r
S
T
t
(27) Non-head Adjunction P
L
K
k
P
M
Q
q
P
R
r
S
T
t
Key to the asymmetry in (26) and (27) is the distinct treatment of
categories versus segments. Thus although the given segment M in (26) may
c-command a given category such as K, the category, or complete phrasal
projection does not. This distinction also holds for the category P in (27)
which does not c-command any element.
In addition, the grammar is further restricted as it is also held that
rightward displacement is banned and the only relevant structural nexus
for syntactic relations is Spec - Head - Comp. This then predicts that
all languages are underlyingly SVO and surface variations in language are
produced by leftward displacement of constituents from initiating from this
basic canonical form.
This proposal has had major implications for many constructions in-
cluding Right Dislocation. Earlier proposals most notably that of Vallduv´ı
(1992) RD was analyzed as the ‘Mirror’ of LD; rightward raised and adjoined
to the specifier of the inflectional phrase (IP) from which it was merged.
(28) Left Dislocation IP
XPi IP
clitici ti
9

(29) Right Dislocation IP
IP
clitici ti
XPi
Therefore, the predicted implications of the LCA on dislocates as viewed
by Kayne (1994) is to reject the previous analyses that rightward movement
occurs. Under the LCA Kayne suggests that LD and RD are mirror reflexes
of the same process enacted at distinct points in the derivation; specifically
LD is an overt operation in which the XP is dislocated into A’-Position
and coindexed in narrow syntax and on the other hand, right dislocates are
coindexed in syntax yet are not moved from A-Position to A’-Position until
Logical Form(LF). The claim, then, is that LD and RD occupy a common
position in IP or above.7
(30) Left Dislocation at Phonetic Form FP
XPi IP
clitici ti
(31) Right Dislocation at Logical Form FP
XPi IP
clitici ti
4 Previous Analyses
If, then, LD and RD share the same position at LF then the prediction is
that they will have observable symmetries. Thus retracing the arguments
against Kayne’s (1994) account of LD and RD, it will be key to look for
asymmetries in dislocation data. This evidence comes primarily from two
angles: syntactic distribution and semantic interpretation.
4.1 Syntactic Distribution of Left and Right Dislocation
4.1.1 Boundedness
Cecchetto (1999) suggests the first evidence pointing to syntactic asymmetry
between LD and RD comes from a phenomena called upward boundedness.
The principle is that the movement of an XP is bound if it is restricted to the
clause from which it is originally merged. Thus the prediction states that
if an XP is upward bound no clause may intervene between the element’s
7The structure of the left periphery is still very much a hot topic in the literature.
10

surface position and its matrix clause position. If we take the position of la
macchina ‘the car’ in (32) to be clause internal position of this element, it
is seen that the same element in a LD construction in (33) may appear to
the left of an intervening clause, shown here for Italian.
(32) Mi
me-DAT
sembra
seems-3sg
strano
strange
che
that
le
him
presti
lent-3sg
la
the
macchina.
car
‘It seems strange to me that he’s lending him the car.’
(33) La
me-DAT
macchinai,
seems-3sg
mi
strange
sembra
that
strano
him-DAT it-ACC
che
lent-3sg
gielai
the
presti
car
‘The car, it seems strange to me that he’s lending it to him.’
(Cecchetto 1999)
This holds as well in Spanish as seen in (34) and (35).
(34) Me
me-DAT
parece
seems-3sg
extran˜o
strange
que
that
le
him
preste
lend-3sg
el
the
auto.
car
‘It seems strange to me that he’s lending him the car.’
(35) El
the
autoi,
car
me
me-DAT
parece
seems-3sg
extran˜o
strange
que
that
se
him-DAT
loi
it-ACC
preste.
lend-3sg
‘The car, it seems strange to me that he’s lending it to him.’
Besides an NP being able to appear left of an intervening clause, it is
also grammatical for the entire matrix clause (CP) to appear here as well,
seen in Italian (36) and Spanish (37).
(36) Che
that
le
him-DAT
presti
lent-3sg
la
the
macchina,
car
mi
me-DAT
sembra
seems-3sg
strano.
strange
‘That he’s lending him the car, seems strange to me.’
(Cecchetto 1999)
(37) Que
that
le
him-DAT
preste
lend-3sg
el
the
auto,
car
me
me-DAT
parece
seems-3sg
extran˜o.
strange
‘That he’s lending him the car, seems strange to me.’
Curiously, the effects are very different in RD constructions. Sentence
(38) in which the NP la macchina ‘the car’ appears to the right of the inter-
vening phrase mi sembra strano ‘it appears strange to me’ is ungrammatical.
11

(38) * Che
that
glielai
him-DAT it-ACC
presti,
lent-3sg
mi
me-DAT
sembra
seems-3g
strano,
strange
la
the
macchinai
car
‘That he to her lends it to me sounds strange the car.’
(Cecchetto 1999)
Again, this is also the case in Spanish.
(39) * Que
that
se
him-DAT
loi
it-ACC
preste,
lend-3sg
me
me-DAT
parece
seems-3g
extran˜o,
strange
el
the
autoi.
car
‘That he to her lends it to me sounds strange the car.’
Therefore Cecchetto argues that an XP can be long-distance LDed but
not RDed. In effect suggesting that RD elements are bound to their matrix
clause whereas LDed elements are not; a clear asymmetry between these
constructions.
4.1.2 Reconstruction Effects
Further evidence diverging LD and RD comes from Reconstruction effects.
This notion stems from the claim that binding principles may seemingly ap-
ply between a pronominal and its displaced antecedent as if the antecedent
were in a position other than that occupied at PF. Consider (40) the follow-
ing example from Italian.
(40) L’opera
the first
prima
work
di
of
uno
a
scrittorei
writer
pro∗i
(he)
la
it
scrive
writes
sempre
always
(voluntieri).
with pleasure
‘The first work of a writer is always written with pleasure (by her/him)’.
(Cecchetto 2000)
The coindexation of scrittore ‘writer’ and the null subject pro is unavail-
able. Under the hypothesis that the antecedent ‘writer’ lowers at LF this
conflict can be explained by a violation of Principle C.
It is observed that this holds for Spanish as well.8
8The reading in (40) in Italian allows for an arbitrary subject or ‘ghost-writer’ inter-
pretation where the comprobable example in Spanish (41) does not allow for this reading.
This interpretation is expressed in (i).
12

(41) * La
the
opera
first
prima
work
de
of
un
a
escritori
writer
pro i
(he)
la
it
escribe
writes
siempre
always
(con
with
ganas).
pleasure
‘The first work of a writer is always written with pleasure (by her/him)’.
Notice, however, that coindexation is possible with an overt subject in
postverbal position as in (42) and (43).
(42) L’opera
the first
prima
work
di
of
uno
a
scrittorei
writer
la
(he)
scrive
it
sempre
writes
luii.
always he
‘The first work of a writer is always written by the writer himself’.
(43) La
the
opera
first
prima
work
de
of
un
a
escritori
writer
la
(he)
escribe
it
siempre
writes
e´li.
always he
‘The first work of a writer is always written by the writer himself’.
The fact that a postverbal subject can be bound by the antecedent
whereas a preverbal subject cannot suggests that there is in fact a lowering
site for dislocates between preverbal and postverbal positions (Cecchetto
2000).9
Furthermore if it is assumed that LD and RD are mirror processes one
would expect that RD should in fact show similar reconstruction effects to
those observed in (41) and (43). This prediction is not borne out. RD shows
no reconstruction effects as seen in (44) and (45).
(44) * pro i
(he)
La
it
escribe
writes
siempre
always
(con
with
ganas),
pleasure
la
the
opera
first
prima
work
de
of
un
a
escritori
writer
.
‘The first work of a writer is always written with pleasure (by her/him)’.
(45) * La
(he)
escribe
it
siempre
writes
e´li,
always
la
he
opera
the
prima
first
de
work
un
of
escritori.
a
writer
‘The first work of a writer is always written by the writer himself’.
Although it is not clear where exactly left and right dislocates reconstruct
to, it is concluded minimally that they do not reconstruct to the same posi-
(i) La
the
opera
first
prima
work
de
of
un
a
escritori
writer
se
is
escribe
written
siempre
always
(con
with
ganas).
pleasure
‘The first work of a writer is always written with pleasure’.
9Against Cinque (1990) who suggests that left dislocates lower to argument position.
13

tion at LF; again, a clear distinction between the two constructions contra
Kayne (1994).10
4.1.3 Prosody
Finally, both Villaba and Cecchetto make reference to the phonetic content
associated with LD and RD; more specifically prosody. It is noted that
LD receives an intonational break or juncture between the first and second
dislocate in (46), seemingly optional as seen in (47).11
(46) Ese
that
reloji,
watch
a
to-DAT
mi
my
mujerj,
woman
sej
her-DAT
loi
it-ACC
robaron.
stole-3pl
‘They stole that watch from my wife.’
(47) Ese
that
reloji
watch
a
to-DAT
mi
my
mujerj
woman
sej
her-DAT
loi
it-ACC
robaron.
stole-3pl
‘They stole that watch from my wife.’
On the other hand, RD must receive an intonational break between both
the first dislocate and the matrix clause and again between each dislocate;
otherwise the structure is ungrammatical.12
(48) Sej
her-DAT
loi
it-ACC
robaron,
stole-3pl
ese
that
reloji,
watch
a
to-DAT
mi
my
mujerj .
woman
‘They stole that watch from my wife.’
(49) * Sej
her-DAT
loi
it-ACC
robaron
stole-3pl
ese
that
reloji
watch
a
to-DAT
mi
my
mujeri.
woman
‘They stole that watch from my wife.’
10The data presented here from Cecchetto (2000) builds on his analysis of Left Dis-
location and Clitic Doubling Constructions(CDC) against the popular notion that left
dislocates lower to argument position at LF Cinque (1990). Cecchetto (2000) suggests
that left dislocates and the double in clitic-doubling constructions do not reconstruct to
a clause internal argument position but rather to an ‘intermediate’ position where both
dislocates and CDCs raise to. This notion will be discussed further later on in reference
to the ‘Big-DP’ (Torrego 1992).
11I will mark intonational breaks with commas ‘,’. Also, there will be very little serious
discussion of the more complex nature of prosody beyond intonation breaks or junctures
in this analysis and more in depth treatment is only mentioned later in the discussion
dealing with further investigations on the nature of dislocation in Romance.
12It is of note here that the example in (49) is not uniformly ruled out in all dialects
of Spanish. The Porten˜o (Argentina) and Sonoran (northern Me´xico) dialects are excep-
tional as they allow clitic doubling of direct objects in contrast to the great majority of
standard dialects of Spanish which only admit clitic doubling of indirect objects. It has
been suggested by Cecchetto (2000) that the reason for this is understood best under
current trends as an economy principle. That is, clitic-doubling in an intuitive sense is
redundant. In this work, Cecchetto (2000) makes reference to the Doubly Filled Voice
Filter of Sportiche (1995) as a mechanism to rule out unnecessary morphophonological
overtness in these cases.
14

Consequently it has been demonstrated that the claim that LD and RD
are not mirror processes is empirically supported from boundedness, recon-
struction and prosodic data.
To motivate LD and RD syntactic distribution Villaba and Cecchetto
both propose similar structures that invoke two distinct maximal projections
for each construction. Coined by Villalba (1999) the ‘Split-Topic Analysis’
(STA) claims that LDed elements surface in the TP periphery whereas RDed
elements are pronounced in the VP periphery position. In essence, left dis-
locates surface in a structurally higher position than right dislocates. These
positions correspond to the diagram in (50) in which ExtTopP is understood
as an external Topic phrase and IntTopP an internal Topic phrase.
(50) CP
. . . ExtTopP
LD TP
. . . . . .
. . . IFocP
. . . IntTopP
RD VP
. . .
(Villalba 1999)
Although the general framework in (50) holds for both Cecchetto (2000)
and Villalba (1999) the steps that motivate movement of all dislocates out of
the verbal phrase and the generation of the corresponding resumptive clitic
differ for these authors. According to Cecchetto (2000) clitics and dislocates
are generated in argument position within a special type of DP known as
the ‘Big-DP’ schematically.13
(51) BIG DP
Double Clitic
The Big-DP, comprised of the dislocate in the Specifier and clitic in the
Head moves outside the verbal phrase as a complete unit. Once outside the
13The Big-DP analysis attributed to the work of Torrego (1992).
15

verbal phrase the dislocate and clitic move independent of one another.14
On the other hand, Villalba (1999) suggests that dislocates are generated
under normal conditions (i.e. generated in argument position as an NP, DP,
etc.) and move outside the verbal phrase leaving a copy that contains the
relevant phi-features for morphological requirements. This copy is ultimately
pronounced as the resumptive clitic.15
Nevertheless, either under a Big-DP or copy approach left and right
dislocates are derived through the VP periphery position at some point in
overt syntax. Yet in the case of left dislocates, the XP must then take a
further step moving out of the VP periphery position into a higher Topic
position, located at the TP periphery. The derivation of LD proceeds from
[Spec, IntTopP] to [Spec, ExtTopP] deriving the surface structure seen in
(52).
(52) Ese
that
reloji,
watch
se
her-DAT
loi
it-ACC
robaron.
stole-3pl
‘That watch, they stole it from her.’
However at this point the derivational histories of both LD and RD
would give the same linear effect, i.e. both would precede the verb. To
account for the linear distribution of the RD, Villaba invokes a type of VP
remnant movement in which all dislocate-linked items, i.e. dislocates and
their respective resumptive elements have raised out of the verbal phrase,
and the VP must move into a projection that immediately dominates the
14Cecchetto (2000) suggests that dislocation constructions and ‘Clitic-Doubling’ con-
structions (CDC) (i) are derivationally related and subscribes to the Big-DP analysis in
part due to the difficulties in deriving the surface position of clitics in these constructions
in Romance without some crucial violation of the Head Movement Constraint.
(i) Lo
him
vi
saw-1sg
a
to
e´l.
him
‘I saw him’.
The employment of the Big-DP thus provides an escape-hatch for the clitic to move
as a maximal projection overtly over intervening heads. Another benefit of the Big-DP
approach comes from the fact that left dislocates seem to reconstruct to an ‘intermediate’
position, non-argument position, that is structurally lower than preverbal subjects but
higher than postverbal ones. Cecchetto (2000) argues that this reconstruction site is the
landing site for the Big-DP.
15This approach crucially depends on notions formulated in Minimalist accounts of the
Copy Theory of movement in which copies are not complete duplicates of their antecedents,
rather they are equipped with the minimal feature requirements needed by the lexical
layer of syntax. In addition, in contrast to the Big-DP approach, the copy approach
makes a clear and conscious distinction between CDCs and dislocation constructions.
The claim is that each of these constructions differ in critical interpretive properties that
would otherwise be shared under a Big-DP-type hypothesis. Further elaboration of the
arguments can be found in Villalba (1999).
16

inner Topic position. This projection is seen to be a type of Informational
Focus position following the work of Vallduv´ı (1992).
(53) CP
. . . ExtTopP
LD TP
. . . . . .
. . . IFocP
VP IntTopP
RD VP
. . .
In this way the linear order of RD (54) falls out.
(54) [IFocPSe
her-DAT
loi
it-ACC
robaron,
stole-3pl
[IntTopPese
that
reloji.]]
watch
‘That watch, they stole it from her.’
Furthermore, to account for the free-word order of multiple LDed (55)
and RDed (56) phrases Villalba proposes that the IntTopP and ExtTopP
projections allow recursion through specifier adjunction as seen in (57) for
RD.
(55) Ese
that
reloji,
watch
a
to-DAT
mi
my
mujeri,
woman
sej
her-DAT
loi
it-ACC
robaron.
stole-3pl
‘They stole that watch from my wife.’
(56) Sej
her-DAT
loi
it-ACC
robaron,
stole-3pl
ese
that
reloji,
watch
a
to-DAT
mi
my
mujeri.
woman
‘They stole that watch from my wife.’
(57) Recursive Dislocation IntTop
XP
ZPi XPj
IntTop’
IntTop VP
. . .
ZP c-commands XP the segment XPj where XP does not c-command ZP,
thus ZP asymmetrically c-commands XP providing the surface ordering of
17

RDed constituents, and LDed constituents in the case of ExtTopP, correctly
licensed by the LCA.
As a result we can summarize what predictions the STA makes. First, it
suggests movement for all dislocate elements to a position outside VP and
that there are two distinct Topic positions that directly correspond to LD
and RD. Next, it falls out that if LD and RD are distinct positions that
there be asymmetries concerning reconstruction effects.16Another correct
prediction that if we continue to follow Vallduv´ı (1992) and Zubizarreta
(1998); Zubizarreta (1999) who provide evidence that prosodic contours in
languages like Spanish and Catala´n accent the rightmost constituent in the
matrix clause marking it as the informative content of the clause17, it may in
fact follow that the obligatory prosodic juncture present in RD in contrast
to LD is a direct result of VP remnant movement marking the matrix clause
boundary.
Assuming that this is on the right track, we are still left with some
problematic data not covered in this proposal. First, there still seems to be
no direct support for the apparent asymmetries of boundedness between LD
and RD. The STA does not make explicit the reason why LD can appear left
of an intervening clause whereas RD cannot. And second, although it may be
claimed that there is some motivation for the obligatory prosodic juncture
present in RD constructions it is not clear how this proposal deals with
the evidence that LD constructions can also seemingly have an intonational
break.
As we have seen there is probable cause to believe that LD and RD are
syntactically distinct. From this observation the account of Villaba and Cec-
chetto succeeds in covering some but not all the syntactic data introduced.
Leaving aside this purely syntactic discussion, let us turn to the second key
angle from which to further test the notion that LD and RD differ minimally
by the point when each is leftward raised.
4.2 Semantic Distribution of Left and Right Dislocation
Villalba (1999) notes that the semantic reading of dislocates may vary.
The first point is that LD can introduce a new Topic18 into discourse (59)
and RD cannot as seen in (60).
16More specifically the claim is that LD reconstructs to a higher position than RD in
effect distinguishing their behavior concerning binding Principle C. This position holds
under the specifics of double-clitic generation of Villalba (1999) and Cecchetto (2000).
17The informative status of dislocates will be discussed in more depth in §5.4
18Again as discussed §3.1Topic will be taken to mean, the ‘point-of-departure’/‘anchor’
of the sentence or ‘what the sentence is about’. A more complete discussion can be found
in Reinhart (1982), Vallduv´ı (1992) and Lambrecht (1994).
18

(58) ¿Do´nde
where
puso
put-3sg
las
the
cosas?
things
‘Where did (s)he put the things?’
(59) Los
the
librosi,
books
losi
them-ACC
puso
put-3sg
en
in
el
the
despacho.
office
‘The books, she put them in the office’
(60) # Losi
them-ACC
puso
put-3sg
en
in
el
the
despacho,
office
los
the
librosi.
books
‘She put them in the office, the books’
However Villalba (1999) also acknowledges that both LD and RD share
the fact that they are presuppositional to some degree. That is to say that
at some point the material can be recovered from the discourse. Along these
lines, the previous example in (58) contained a set of arbitrary items not
explicitly recoverable from the discourse, whereas (61) on the other hand
makes explicit reference to the defined set of items linked into discourse,
specifically los libros ‘the books’.
(61) ¿Do´nde
where
puso
put-3sg
los
the
libros?
books
‘Where did (s)he put the books?’
(62) # Los
the
librosi,
books
losi
them-ACC
puso
put-3sg
en
in
el
the
despacho.
office
‘The books, (s)he put them in the office.’
(63) Losi
them-ACC
puso
put-3sg
en
in
el
the
despacho,
office
los
the
librosi.
books
‘The books, (s)he put them in the office’
The grammaticality of (62) and (63) suggests that when an element has
been introduced into the discourse both LD and RD are fine.
To motivate LD and RD semantic interpretation Villalba (1999)proposes
that there are two syntactic features [new] and [prominent] which drive the
pragmatic effects previously mentioned. In this framework the feature value
[-new] activates the movement of all elements associated with this feature
out of the verbal phrase, specifically to the VP periphery position IntTopP
as seen in (64). Returning to the syntactic motivation briefly, this, in effect,
includes all dislocate items.19 This assertion has a desired secondary result
as it complies with the notions elaborated by Heim (1982) and Diesing (1992)
that mark the VP as a semantic domain for items that are novel in discourse
and/or non-presuppositional.
19Dislocates and resumptive elements.
19

(64) CP
. . . ExtTopP
. . . TP
. . . IFocP
. . . IntTopP
LD/RD
[-new]
VP
. . .
Once these elements marked [-new] raise outside the VP, the feature
[prominent] then takes those elements marked [+prominent] up to the TP
periphery as seen in (65).
(65) CP
. . . ExtTopP
LD
[-new, +prominent]
TP
. . . IFocP
. . . IntTopP
RD
[-new]
VP
. . .
Consequently the feature distribution takes the shape of that in (66).
20

(66) CP
. . . ExtTopP
LD
[-new, +prominent]
TP
. . . IFocP
. . . IntTopP
RD
[-new, -prominent]
VP
. . .
In summary Villalba (1999) motivates the syntactic movement with the
binary features [new] and [prominent]. This has the effect of driving all
non-novel information out of the VP. Once in outside the VP the feature
[+prominent] is attracted to the [spec, ExtTopP] position to fulfill discourse
requirements.
Consequently, the surface structure of left and right dislocates is seem-
ingly accounted for through well-founded semantic features. However, al-
though the presuppositional tie between LD and RD is shown, we are still
left with an inconclusive semantic explanation for why LDed elements can
seemingly introduce a new topic into discourse but RDed elements cannot?
And furthermore whether the features [new] and [prominent] alone can ad-
equately cover the semantic distribution of dislocates.
To this point in the discussion we have seen some examples of ways in
which LD and RD differ syntactically as well as semantically against the
predictions made by Kayne’s (1994) account. In addition, we have laid out
the previous proposals that attempt to account for the syntactic and se-
mantic distribution of these two processes. It has been asserted here that
the previous syntactic analysis of LD and RD by Villalba (1999) and Cec-
chetto (1999) in fact handle well the reconstruction effects and to some
degree the prosodic differences we encounter, on one hand, and on the other
these authors give a telling account of the semantic basis for this syntac-
tic distribution. Nonetheless, it has been noted that the syntactic as well
has semantic and pragmatic data proposed to distance LD from RD can
be only partially accounted for in previous analysis; syntactically we still
have a latent problem accounting for the unboundedness of left dislocates in
comparison to bound right dislocates. On the semantic front, the question
has been raised to whether the features [new] and [prominent] can account
21

for the pragmatics of dislocates, specifically the introduction of new Topics
into discourse.
5 Analysis
Turning to the current analysis I will assume that Villalba (1999) and Cec-
chetto (1999) are on the right track in that LD and RD are distinct processes
based on the empirical evidence from reconstruction effects, boundedness
data and noted differences in prosodic contours, against Kayne’s account of
dislocation. I will also follow portions of their syntactic and semantic analy-
sis, specifically the notion that there are in fact at least two distinct syntactic
projections for LD and RD into which elements marked with appropriate
semantic features and feature values move to before Spell-out.
In this spirit, the main goal of this section will be to further disassociate
LD from RD. Evidence provided here comes from Islands, multiple disloca-
tion constructions and the distribution of Bare Nouns in Spanish. From this
data I will show that LD differs from RD in that LD is really two distinct
processes, one of which is not the result of movement from a clause inter-
nal position and subsequently is not available to RD a processes uniformly
found to be the result of movement. Consequently, the claim is made that
looking at LD as two processes, one the result of movement the other crit-
ically not, facilitates a theory that may indeed capture the empirical data
left unaccounted for in previous analysis, namely the unboundedness, the
optional prosodic break and the introduction of new Topics in LD. In ad-
dition, I argue that the dual nature of LD provides a distinction useful in
discerning the Information Structure of Spanish and claim that the critical
distinction between left and right dislocates generated by movement is the
same property that marks Topics in Spanish.
5.1 Syntactic Distribution
5.1.1 Movement in Spanish
Given the prediction by the Split-Topic Analysis that dislocates are uni-
formly generated by movement from a clause internal position, it is im-
portant to observe the behavior of constituents in constructions classically
analyzed as involving DP movement in Spanish (Zagona 2000) as a diag-
nostic. The two common constructions are passives,20 as seen in (67) and
‘raising constructions’ as seen in (68).
20It is of note that there are in fact two forms of the passive in Spanish; morphological
passive and passive ‘se’. In some dialects of Spain the latter is often more commonly
employed. However it is important to underline that the morphological passive is indeed
grammatical in these dialects. Thus, the preference of some dialects to use one form over
another is considered strictly an issue of language use and will not be discussed in the
main body of this paper.
22

Passive Construction
(67) Esa
that
pelota
ball-FEM
fue
was
robada.
stolen-FEM
‘That ball was stolen.’
Raising Construction
(68) Ese
that
chico
boy-MASC
parece
seems
estar
to be
drogado.
drugged-MASC
‘That boy seems to be on drugs.’
The structure in (67) has been analyzed as parallel to English in the
relevant aspects in that the overt subject esa pelota ‘that ball’ is externally
merged in object position after which it is internally merged into subject
position. And again in (68) the surface subject ese chico ‘that boy’ is inter-
nally merged from its base-position in the specifier of the verbal phrase ‘to
be’.21 A diagnostic to test the validity of the movement analysis, we may
embed the matrix clause in a Complex NP, strong Islands in Spanish, as
seen in (69).
Passive embedded within a Complex NP
(69) * Esa
that-DEM
pelota
ball-FEM
odio
hate-1sg
el
the
hecho
fact
de
of
que
that
fuera
was-1sg
robada.
stolen-FEM
‘That ball, I hate the fact that it was stolen.’
For good measure, WH-extraction shows again that movement is blocked
by Complex NPs (70).
WH-extraction from a passive embedded within a Complex NP
(70) * ¿Que´
what
odias
hate-2sg
el
the
hecho
fact
de
of
que
that
fuera
was
robada?
stolen-FEM
‘What do you hate the fact that it was stolen?’
Now looking at raising constructions, subjects cannot raise outside of
Complex NPs (71).
Raising embedded within a Complex NP
21The reason for NP to raise into overt subject position will not be crucial in this paper
but has been understood primarily to fulfill Case and EPP requirements. However this
assumption based on English is challenged by Casielles-Sua´rez (2004) where preverbal
subject position is seen as the Topic position in Spanish and movement into this position
is entirely discourse motivated by the formal feature [-focus] to avoid being interpreted
as part of the Focus articulation. While on the right track essentially this hypothesis is
argued not to be entirely correct §5.4.
23

(71) * Ese
that
chico
boy-MASC
odio
hate-1sg
el
the
hecho
fact
de
of
que
that
parezca
seems
estar
to be
drogado.
drugged-MASC
‘That boy, I hate the fact that he seems drugged (high).’
And again, WH-extraction is also disallowed in (72).
WH-extraction from a raising construction within a Complex NP
(72) * ¿Quie´n
who
odias
hate-2sg
el
the
hecho
fact
de
of
que
that
parezca
seems
estar
to be
drogado?
drugged-MASC
‘Who do you hate the fact that he seems drugged (high)?’
Hence, the ungrammaticality of (69 - 72) highlights the fact that Com-
plex NPs are Islands to movement in Spanish.
5.1.2 Dislocation as Movement
Having established a diagnostic for movement, let us return to the behavior
of dislocates under this new light. Assuming the analysis of Villalba (1999)
and Cecchetto (1999) under which LD and RD is derived from movement we
should expect that left and right dislocates would be sensitive to Complex
NPs. This prediction initially seems to hold for LD as seen in the contrast
between (73) and (74).
Left Dislocation
(73) A
to-DAT
Martai
Marta
lei
her-DAT
prestaron
lent-3pl
ese
that
libro.
book
‘To Marta they lent that book.’
Left Dislocation embedded within a Complex NP
(74) * A
to-DAT
Marta
Marta
el
the
hecho
fact
de
that
que lei
her-DAT
prestaron
lent-3pl
ese
that
libro
book
me
me
parece
seems
extran˜o.
strange
‘To Marta, the fact that they lent her that book seems strange (to
me).’
However curiously when LD is prefixed with hablando de ‘speaking of’ or
en cuanto a ‘as far as’ the grammaticality is reversed as seen in (75) and (76).
Left Dislocation embedded within a Complex NP (Prefix/Prosodic break)
24

(75) Hablando
speaking
de
of
Martai,
Marta
el
the
hecho
fact
de
that
que lei
her-DAT
prestaron
lent-3pl
ese
that
libro
book
me
me
parece
seems
extran˜o.
strange
‘(Speaking of) Marta, the fact that they lent her that book seems
strange (to me).’
(76) En
as
cuanto
far as
a Martai,
Marta
el
the
hecho
fact
de
that
que lei
her-DAT
prestaron
lent-3pl
ese
that
libro
book
me
me
parece
seems
extran˜o.
strange
‘As far as Marta is concerned, the fact that they lent her that book
seems strange (to me).’
Two points of note here, first is the fact that the prefixing of the dislo-
cate in (75) strips the case marking of dative dislocates and second, prefixing
appears to force an obligatory prosodic break between the dislocate and the
matrix clause. However, looking closer this prosodic break seems to be in-
dependent of prefixing as seen in (77).
Left Dislocation embedded within a Complex NP (Prosodic break)
(77) Martai,
Marta
el
the
hecho
fact
de
that
que lei
her-DAT
prestaron
lent-3pl
ese
that
libro
book
me
me
parece
seems
extran˜o.
strange
‘Marta, the fact that they lent her that book seems strange (to me).’
Returning to right dislocation constructions, the fact that left dislocates
seemingly display a dual nature with respect to Complex NPs the intuition
is that this may in fact hold for right dislocates as well. However the data
in (79) reveal that this is not the case; RD, (78), is uniformly blocked by
Complex NPs regardless of prefixing, case stripping or appearing prosodi-
cally separate from the matrix clause, (79).
Right Dislocation
(78) Lei
her-DAT
prestaron
lent-3pl
ese
that
libro,
book
a
to-DAT
Martai.
Marta
‘They lent that book, to Marta .’
Right Dislocation embedded within a Complex NP (prefixed/prosodic break)
(79) * El
the
hecho
fact
de
that
que lei
her-DAT
prestaron
lent-3pl
ese
that
libro
book
me
me
parece
seems
extran˜o,
strange
(hablando
(speaking
de)
of)
Martai.
Marta
25

‘The fact that they lent her that book seems strange (to me), (speak-
ing of) Marta .’
While it is clear that right dislocates are uniformly sensitive to Complex
NPs it seems that LD in fact has a dual nature. On one hand left dislocates
show sensitivity to Complex NPs if the dislocate is Case-marked and/or
prosodically merged with the matrix clause and on the other hand they are
not sensitive to Complex NPs if prefixed and/or prosodically separate from
the matrix clause.
5.1.3 The Left Periphery
Continuing this line of inquiry, it is also the case that the asymmetric prop-
erties of LD not available to RD appear limited to the left periphery, there-
fore here labeled as Hanging Left Dislocation (HLD), seen in the contrast
between (80) and (81).
Left Periphery of Declaratives
(80) (Hablando
speaking
de)
of
tu
your
padrei,
father,
ese
that
libroj
book
sei
him-DAT
loj
it-ACC
dieron
gave-3pl
a
to-DAT
e´li.
him
‘(Speaking of) your father, that book they gave it to him.’
Non-Left Periphery of Declaratives
(81) * Ese
that
libroi,
book
(hablando
speaking
de)
of
tu
your
padrej
father
sej
him-DAT
loi
it-ACC
dieron
gave-3pl
a
to-DAT
e´lj .
him
‘That book, (speaking of) your father, they gave it to him.’
HLD is also relegated to the left periphery in WH-Constructions, as
shown in (82) and (83).
Left Periphery of WH-Constructions
(82) (Hablando
speaking
de)
of
esas
those
sillasi,
chairs
¿cua´nto
how much
quieres
want-2sg
por
for
ellasi?
them
‘(Speaking of) those chairs, how much do you want for them?’
Non-Left Periphery of WH-Constructions
(83) * ¿Cua´nto
how much
(hablando
speaking
de)
of
esasi
those
sillas,
chairs
quieres
want-2sg
por
for
ellasi?
them
‘How much (speaking of) those chairs do you want for them?’
26

Furthermore HLD is limited to one element in WH-Constructions.22
Single element in WH-Constructions
(84) (Hablando
speaking
de)
of
ese
that
libroi,
book
¿quie´n
who
se
him/her-DAT
loi
it-ACC
pidio´?
asked for-3sg
‘(Speaking of) that book, who asked him for it?’
(85) (Hablando
speaking
de)
of
tu
your
hermanoi,
brother
¿quie´n
who
sei
him-DAT
lo
it-ACC
pidio´?
asked for-3sg
‘(Speaking of) your brother, who asked him for it?’
Double element in WH-Constructions
(86) * Ese
that
libroi,
book
tu
your
hermanoj ,
brother
¿quie´n
who
sej
him-DAT
loi
it-ACC
pidio´?
asked for-3sg
‘That book, your brother, who asked him for it?’
This also, of course, shows that HLD elements are structurally higher
in the clause in comparison to WH-words, argued to be in complementary
distribution with Contrastively Focused elements (Rizzi 1997), while curi-
ously here the second left dislocate strategy appears blocked as well.23 A
preliminary sketch reveals the following schema of the left periphery.
Structure of the Left Periphery
22However elements may be complex in the sense that a coordinate structure may appear
here (Contreras 1976), seen in (84 - 86).
(i) (Hablando
speaking
de)
of
Marta
Marta
y
and
Juan,
Juan
no
no
los
them-ACC
he
have-1sg
visto
seen
desde
from
hace
makes
mucho.
a lot
‘(Speaking of ) Marta and Juan, I haven’t seen them in a long time’.
23This observation will be considered in more detail in §5.4.
27

(87) CP
Hanging LD FuncP
WH
Contrastive Focus
ExtTopP
LD TP
. . .
5.1.4 Data Unaccounted For
Boundedness One piece of evidence given by Cecchetto (1999) distin-
guishing LD from RD is that LD appears to be unbounded, repeated here
in (88) and (89).
(88) Me
me-DAT
parece
seems-3sg
extran˜o
strange
que
that
le
him
preste
lend-3sg
el
the
auto.
car
‘It seems strange to me that he’s lending him the car.’
(89) El
the
autoi,
car
me
me-DAT
parece
seems-3sg
extran˜o
strange
que
that
se
him-DAT
loi
it-ACC
preste.
lend-3sg
‘The car, it seems strange to me that he’s lending it to him.’
If, as the data here suggest, there are two types of LD strategies it may
in fact be the case that only one of the two show unbounded properties.
Thus, applying the diagnostics seen earlier to distinguish HLD from LD,
Complex NPs and Prefixing/Prosodic Break, it is shown that accusative
dislocates, non Case-marked in Spanish, can appear as HLDed elements
when accompanied with a prosodic break, (90)
(90) El
the
autoi,
car
odio
hate-1sg
el
the
hecho
fact
de
of
que
that
se
him-DAT
loi
it-ACC
preste.
lend-3sg
‘The car, it seems strange to me that he’s lending it to him.’
However, notice that given a Case-marked element, the construction is
bounded, (91) while unbounded when appearing as a typical HLD element;
prefixed and/or non Case-marked with a prosodic break.24, (92).
24The subject ‘Carlos’ is added in (92) to remove the ambiguous (subject/object) read-
ing that non Case-marked elements create in this example.
28

(91) * A
to-DAT
Juani
Juan
odio
hate-1sg
el
the
hecho
fact
de
of
que
that
lei
him-DAT
preste
lends
el
the
auto.
car
‘To Juan, I hate the fact that he’s lending the car to him.’
(92) (Hablando
speaking
de)
of
Juani,
Juan
odio
hate-1sg
el
the
hecho
fact
de
of
que
that
(Carlos)
Carlos
lei
him-DAT
preste
lends
el
the
auto.
car
‘(Speaking of) Juan, I hate the fact that Carlos is lending the car to
him.’
Hence examples (91) and (92) indicate unboundedness is a property of
HLD while LD is bounded, as in RD. Given that the both LD and RD are
sensitive to extraction from Complex NP, whereas HLD is not, I suggest that
movement is the common property that provides the asymmetry in bound-
edness data. Under this view, only HLD is unbounded precisely because
it is not generated in a clause internal position. Consequently, only clause
internally generated elements are subject to boundedness.
Prosody Having seen the distributional patterns of HLD, the seemingly
optional prosodic contour appears to be more clearly a property of the HLD
construction. Instead of being an optional characteristic of LD, prosodic
breaks are a required feature of this construction.
(93) Juan
Juan
*(,) me
me-DAT
parece
seems
extrano
strange
que
that
le
her-DAT
preste
lends
el
the
auto.
car
‘Juan, it’s strange (to me) that he’s lending it to her.’
5.1.5 Summary
In this section data from extraction from Complex NPs provides evidence
that there are in fact two types of Left Dislocation and only one Right
Dislocation. Consequently, Hanging Left Dislocation, unique to the left
periphery, appears to demonstrate some of the problematic properties left
unaccounted for in previous analyses, including unboundedness and prosodic
contours. In addition it was shown that HLD 1) only appears at the left
periphery 2) is limited to only one element and 3) never appears Case-
marked.
5.2 Semantic Distribution
Bouncing back to the semantic evidence, the contention by Villalba (1999)
that the feature [prominent] solely distinguishes RD from LD is weakened.
29

If the dual nature of LD is on the right track we must then assume that
there is some type of distinct semantic or pragmatic nature for this process
that minimally distinguishes LD from HLD.
5.2.1 Contrastive Left Dislocation
Refreshing the reader’s memory, the semantic basis for the feature [promi-
nent] was in fact to motivate the distinction in (61), (62) and (63) here (94),
(95) and (96) where a ‘new’ Topic may be introduced in LD but not in RD.
(94) ¿Do´nde
where
puso
put-3sg
las
the
cosas?
books
‘Where did (s)she put the books?’
(95) Los
the
librosi,
books
losi
them-ACC
puso
put-3sg
en
in
el
the
despacho.
office
‘The books, (s)he put them in the office.’
(96) # Losi
them-ACC
puso
put-3sg
en
in
el
the
despacho,
office
los
the
librosi.
books
‘The books, (s)he put them in the office’
However, on closer examination of the pragmatics of these examples if
we take the sentence in (94) to be a question concerning the set of things in
question we see that (95) can pick a member of the relevant set of things and
bring it into discourse. In effect, this use of LD is not introducing new Topics
into discourse rather it is contrasting set membership of relevant discourse
items already presupposed.
Lo´pez (2003b) makes a similar assertion as to the contrastive nature
concerning LD for Catala´n and proposes that the unavailability of RD to
capture this contrastive use is to be understood as a reflex of the ‘identifi-
cational’ nature of right dislocates. It is seen that this holds for Spanish as
well, (97).
(97) ¿Que´
what
hiciste
do-2sg
con
with
el
the
boli?
pen
‘What did you do with the pen?’
(98) # El
the
bolii,
pen
loi
it-ACC
deje´
left-1sg
en
in
el
the
escritorio.
desk
‘The pen, I left it on the desk’
(99) Loi
it-ACC
deje´
left-1sg
en
in
el
the
escritorio,
desk
el
the
bolii.
pen
‘I left it on the desk, the pen.’
30

Lo´pez (2003b) also notes that the semantically unavailable use of LD in
this context (98) can be recovered. By invoking a continuation requirement
in which the hearer expects some type of contrasting options the ungram-
matical (98) becomes grammatical. Observe (101).
(100) ¿Que´
what
hiciste
do-2sg
con
with
el
the
boli?
pen
‘What did you do with the pen?’
(101) El
the
bolii,
pen
loi
it-ACC
deje´
left
en
in
el
the
escritorio
desk
pero
but
el
the
la´pizj ,
pencil
no
no
loj
it-ACC
he
have
tocado.
touched-PP
‘The pen, I left it on the desk but the pencil, I didn’t touch it.’
This evidence further corroborates the contrastive nature of LD against
the use of feature [prominent]. Yet adopting this proposal raises a new
question: are both HLD and LD interpreted as contrastive?
As seen in (103), HLD elements cannot be contrastive.
(102) En
as
cuanto
far as
al
to-the
reloji,
watch
se
her-DAT
loi
it-ACC
di
gave-3sg
a
to-DAT
mi
my
mujer.
woman
‘As far as the watch is concerned, I gave it to my wife.’
(103) # En
as
cuanto
far as
al
to-the
reloji,
watch
se
her-DAT
loi
it-ACC
di
gave-1sg
a
to-DAT
mi
my
mujer
woman
pero
but
en
as
cuanto
far as
al
to-the
libroj
book
se
him-DAT
loj
it-ACC
di
gave-1sg
a
to
mi
my
hijo.
son
‘As far as the watch is concerned, I gave it to my wife but as far as
the book is concerned, I gave it to my son.’
Furthermore, left periphery elements in general cannot get contrastive
readings as seen in the contrast between (104) and (105).
Inner Left Dislocate
(104) Ese
that
cochei,
car
a
to-DAT
mi
my
hermanoj
brother
sej
him-DAT
loi
it-ACC
prestaron
lent-3pl
(pero
(but
a
to-DAT
mi
my
mujer3
wife
se3
her-DAT
loi
it-ACC
regalaron).
gave-3pl
‘That car, it was lent to my brother but it was given to my wife’
Outer Left Dislocate
31

(105) * Ese
that
cochei,
car
a
to-DAT
mi
my
hermanoj
brother
sej
him-DAT
loi
it-ACC
prestaron
lent-3pl
(pero
(but
ese
that
reloj3
watch
sej
him-DAT
lo3
it-ACC
regalaron).
gave-3pl
‘That car, it was lent to my brother but that watch it was given to
him.’
Therefore contrastiveness seems to be linked specifically to the LD po-
sition, henceforth Contrastive Left Dislocation (CLD).
5.2.2 Coreference
Another semantic asymmetry between HLDed elements and CLDed ele-
ments is that HLD can be coreferential with a variety of syntactic elements,
as seen in (106 - 108),
Full XP
(106) (Hablando
speaking
de)
of
Juani,
Juan
casi
almost
nunca
never
como
eat-1sg
con
with
ese
that
desgraciadoi.
bastard
‘(Speaking of) Juan, I almost never eat with that bastard.’
Pronouns
(107) (Hablando
speaking
de) Juani,
of
no
Juan
me
no
olvido
me-REF
nunca
forget-1sg
de
never
e´li.
of
him
‘(Speaking of) Juan, I’ll never forget him.’
Clitics
(108) (Hablando
speaking
de)
of
Juani,
Juan
no
no
lei
him-DAT
voy
go-1sg
a dar
give
el
the
premio
prize
este
this
an˜o.
year
‘(Speaking of) Juan, I’m not going to give him the prize this year.’
whereas CLD phrases are only coreferential with clitic pronouns, (109 -
111).
Full XP
(109) * Juani
Juan
casi
almost
nunca
never
como
eat-1sg
con
with
ese
that
desgraciadoi.
bastard
‘Juan I almost never eat with that bastard.’
Pronouns
32

(110) * Juani
Juan
no
no
me
me-REF
olvido
forget-1sg
nunca
never
de
of
e´li.
him
‘(Speaking of) Juan, I’ll never forget him.’
Clitics
(111) A
To-DAT
Juani,
Juan
no
no
lei
him-DAT
voy
go-1sg
a dar
give
el
the
premio
prize
este
this
an˜o.
year
‘To Juan, I’m not going to give him the prize this year.’
In this comparison is it of interest to note that right dislocates pattern
with CLDed elements in that right dislocates may only be coreferential with
clitic pronouns as seen in (112 - 114).
Full XP
(112) * Casi
almost
nunca
never
como
eat-1sg
con
with
ese
that
desgraciadoi,
bastard
Juani.
Juan
‘I almost never eat with that bastard, Juan.’
Pronouns
(113) * No
no
me
me-REF
olvido
forget-1sg
nunca
never
de
of
e´li,
him,
Juani.
Juan.
‘I’ll never forget him, Juan.’
Clitics
(114) No
no
lei
him-DAT
voy
go-1sg
a dar
give
el
the
premio
prize
este
this
an˜o,
year
a
to
Juani.
Juan
‘I’m not going to give him the prize this year, to Juan.’
5.2.3 Summary
The semantic evidence further corroborates the proposal that there are two
types of LD, one that is contrastive in nature and one that is not. Although
RD cannot be contrastive, CLD and RD share the property that they are
both limited to coreference with clitic pronouns whereas HLD is able to be
associated with a much wider range of elements including full XPs and tonic
pronouns.
5.3 Augmenting The Split-Topic Analysis
5.3.1 Three Positions For Dislocates
Considering the ability for HLD to be coindexed with various elements, XPs,
pronouns and clitics, in a variety of constituents traditionally understood
as islands to binding, such as complex NPs, in addition to being un Case-
marked and prosodically separate from the main clause, I propose that what
33

syntactically distinguishes CLD/RD from HLD is movement. This basic in-
sight underscores the basic asymmetry between Left and Right Dislocation;
there are two types of LD and only one RD. Hence, I propose that there
are three distinct positions for dislocates. Under the original formulation of
the Villalba’s (1999) Split-Topic Analysis only two positions were slotted for
the surface position of dislocates corresponding to left and right dislocates,
repeated here in (115).
(115) CP
. . . ExtTopP
LD TP
. . . . . .
. . . IFocP
. . . IntTopP
RD VP
. . .
(Villalba 1999)
However, this proposal is unable to capture the observed asymmetries
between HLD and CLD hence I propose the augmented schema in (116).25
(116) CP
HLD ExtTopP
CLD TP
. . . . . .
. . . IFocP
. . . IntTopP
RD VP
. . .
25The diagram in (116) is underspecified for other functional projections for the sake of
clarity.
34

The present formulation assumes the underlying motivation for the orig-
inal STA, namely the ability to derive the surface positions of left and right
dislocates in narrow syntax while still conforming to the strict phrase struc-
ture laid out by the Linear Correspondence Axiom. In addition, I maintain
the surface positions of CLD and RD are the result of movement given
their sensitivity to extraction from Complex NPs, obligatory Case-marking,
etc.26 However, the second left dislocate strategy, i.e. HLD shows properties
of base-generation; no sensitivity to Complex NPs, no Case-marking, only
appears at the extreme left periphery and is necessarily prosodically sep-
arate from the main clause and therefore appears structurally higher than
other dislocates.
5.3.2 Feature Attraction
Under Generative theory, the movement of CLD and RD must be motivated.
Thus to motivate movement and to distinguish the surface positions of LD
versus RD the original STA presented the features [prominent] and [new]
as formal features designated for the projections ExtTopP and IntTopP
respectively; a plausible account given the discourse import of dislocation.
Along these lines, I maintain the hypothesis by Villalba (1999) that elements
marked [-new] move under attract uniformly to [Spec, IntTopP]. This has
the desired effect of moving all non-novel material outside the VP, as seen
in (117).27
26However Case-marking of dislocates derived by movement appears to provide indirect
support for the Big-DP analysis for dislocation structures in that any other account would
have to stipulate the parallel Case specifications for dislocates and their clitic counterparts.
27It is of note that the feature [-new] could be reformulated as any attribute that has
the result of driving presuppositional elements out of the VP before Spell-out, such as
[+presuppositional] in Villalba (1999) or [-focus] in Casielles-Sua´rez (2004). I chose to
remain agnostic to whether this is an important distinction to make, and therefore chose
to maintain the same notational device as elaborated in the original formulation of the
Split-Topic Analysis.
35

(117) CP
HLD ExtTopP
CLD TP
. . . . . .
. . . IFocP
. . . IntTopP
.CLD/RD
[-new]
VP
. . .
However, the evidence does not support a [prominent] feature, but more
clearly shows that contrastiveness is the active property that distinguishes
CLD from RD. To overcome the gap in this paradigm I propose that a third
feature [contrast] must replace [prom(inent)] to account for the semantic
distribution of dislocates.28
(118) CP
HLD ExtTopP
CLD
[+contrast]
TP
. . . . . .
. . . IFocP
. . . IntTopP
RD VP
. . .
This characterization of the feature structure that drives movement of
dislocates appears to more naturally distribute across the various other dis-
28The notion that contrastiveness is the appropriate feature to distinguish CLD from
RD is reconsidered §?? after discussing the implications of this hypothesis for Information
Structure.
36

course strategies derived by movement, seen in table (119), with correspond-
ing examples in (120 - 123).
(119) Feature Structure of Discourse Movement
Contrast Foc CLD Topicalization RD
+new -new +new -new
+cont +cont -cont -cont
Contrastive Focus
(120) UNA
an
MANZANA
apple
s´ı
yes
tengo.
have-1sg
‘An apple, yes I do have (one).’
Contrastive Left Dislocation
(121) Ese
that
coche
car
no
no
lo
it-ACC
vi.
saw-1sg
‘That car, I didn’t see it.’
Topicalization
(122) Cafe´
coffee
dicen
say-3pl
que
that
no
no
tienen.
have-3pl
‘Coffee, they say they don’t have (any).’
Right Dislocation
(123) No
no
lo
it-ACC
vi,
saw-1sg
ese
that
coche.
car
‘I didn’t see it, that car.’
The feature shape of HLD is harder to motivate since it is not derived
by movement. While it is clear that HLD elements are presuppositional and
that they are non-contrastive, this feature shape mirrors that of RD, [-new],
[-contrast]. However, assuming the base-generation analysis, HLD is not
held to the same feature checking requirements as its moved counterparts.
In addition to the elaboration of feature attraction as motivation for
movement given here, I will extend this hypothesis to incorporate the In-
formation Structure of Spanish in §5.4. In particular, this approach reveals
restrictions posed by the articulation of information on the syntax and ulti-
mately the available surface structure configurations in Spanish.
5.3.3 Summary
Hence, I propose here a version of the Split-Topic Analysis (Villalba 1999)
which modifies the original formulation in that a new position for HLD is
37

created for the base-generated dislocated, the feature paradigm is restated
under [contrast] not [prominent] as the active feature relevant in distin-
guishing between the moved dislocates, CLD and RD. This hypothesis is
supported by the following summary of data covered in this analysis.
(124) Support For Augmented STA
Phenomenon HLD CLD RD
Bounded - + +
Prosodic break + - +
Multiple Elements - + +
Case marked - + +
Contrastive - + -
Summarizing the current hypothesis, boundedness and Case-marking is
seen as directly related to clause-internally generated elements. The appar-
ent optional break in LD is now clearly associated with the base-generated
HLD. The obligatory prosodic break of RD however, is maintained as a reflex
of prosodic marking of the left-most constituent of the VP. Under remnant
movement this position will necessarily precede right dislocates. Further-
more, data indicates contrastiveness is key in distinguishing surface position
between moved dislocates. Finally, only HLD appears to disallow multiple
elements however, this notion is challenged in the following section as the
model elaborated here is evaluated on its discourse properties.
5.4 Information Structure
Dislocation is a non-canonical reformulation of clause structure in Spanish.
This type of construction has been argued to be discourse related by Vallduv´ı
(1992), Lambrecht (1994), etc., as well as here. Therefore, I now turn to
how the current proposal supports the discourse-driven articulation in the
framework of Information Structure.
5.4.1 A Dual Articulation Model For Spanish
Clause structure in Spanish is understood to be composed of two basic
halves; demarked by the verbal projection. Those elements that surface
within the VP are interpreted as Informational Focus (Kiss 1998); that is
the informative part of the clause. On the other hand, those elements outside
of the VP have received a less unified account29, although the basic assump-
tion is that they are not informative but act as either Topical or Background
elements in discourse. A salient proposal for Spanish claims that in fact
the distinction between Topical and Background elements is complementary
(Casielles-Sua´rez 2004). Under this view, there are two distinct Information
29See Vallduv´ı (1992) for discussion.
38

Articulations for Spanish; Topic/Comment and Focus/Background, hence-
forth The Dual Articulation Model. The distinction can be seen in (125)/
(126) and (127)/ (128).
Question: Juan is Focused
(125) ¿Que´
what
sabes
know-2sg
de
of
Juan?
Juan
‘What’s up with Juan (lately)?’
Answer: Topic/Comment Structure, Juan is Topical
(126) Juan
Juan
acaba
just-3sg
de llamar.
call
‘Juan just called.’
Question: Proposition is Focused
(127) ¿Quie´n
who
ha
has
llamado?
called
‘Who called?’
Answer: Background/Focus Structure, Juan is in Focus
(128) Ha
has
llamado
called
Juan.
Juan
‘Juan called.’
(Casielles-Sua´rez 2004)
In (126) the subject appears preverbal in Topic/Comment structures
while in the Focus/Background structure (128) it is postverbal. Casielles-
Sua´rez argues that the canonical form for out-of-the-blue sentences adheres
to the Topic/Comment template.
(129) [topic Juan
Juan
[comment tiene
has
mucho
much
talento].
talent
‘Juan is talented.’
In addition, the argument is made that preverbal subject position is
derived by movement into Topic position. However, a unique characteristic
to this proposal is the idea that only preverbal subjects can be Topics and
that all other preverbal elements are dislocates, base-generated and read
as Background elements in the Background/Focus articulation. This idea is
supported by the fact that subjects do not appear preverbally (in other words
they only appear in Background/Focus constructions) when the dislocate is
presupposed, as in (130) and (131).
39

(130) ¿Qui´ıen
who
comio´
ate-3sg
las
the
fresas?
strawberries
‘Who ate the strawberries?’
(131) Las
the
fresasi,
strawberries
lasi
them-ACC
comı´
ate-1sg
yo.
‘The strawberries, I ate them.’
(Casielles-Sua´rez 2004)
This account also makes the prediction that elements that are understood
to be ‘trapped’ inside the verbal projection such as Bare Nouns in Spanish30
should not appear in preverbal subject position as this would necessitate
movement outside VP. This seems to be the case shown in the contrast
between, (132) and (133).
(132) Jugaban
play-3pl
nin˜os
children
en
in
la
the
calle.
street
‘Children were playing in the street.’
(133) * Nin˜os
children
jugaban
play-3pl
en
in
la
the
calle.
street
‘Children were playing in the street.’
(Casielles-Sua´rez 2004)
On the other hand, dislocates are uniformly seen as base-generated and
therefore Bare Nouns should be allowed.
(134) A
to-DAT
mı´
me
dinero
money
Juan
Juan
nunca
never
me
me-DAT
deja.
gives
‘To me, money Juan never gives it to me.’
Then, for Casielles-Sua´rez, Background elements are non-clause gener-
ated and coindexed with a clause-internal pronominal element. In the case
of full DP dislocates these take the form of clitics. However, the argument
is that for Bare Noun dislocates this pronoun is a null pro.
(135) A
to-DAT
mı´i
me
dineroj
money
Juank
Juan
prok nunca
never
mei
me-DAT
deja
gives
proj .
‘To me, money Juan never gives it to me.’
30See (Casielles-Sua´rez 2001; Casielles-Sua´rez 2004) for a complete argument for the
non-movement status of Bare Nouns.
40

What’s attractive about this account, over other accounts (Vallduv´ı
1992), is that the ambiguity in marking items as Topical or Background
is removed. In dislocation constructions, and other structures in which
the subject appears preverbally are uniformly marked as part of the Fo-
cus/Background articulation. On the other hand, those instances in which
the subject is preverbal are Topic/Comment structures.
5.4.2 Movement of Background Elements
The Dual Articulation Model is based on the separation between clause-
generated elements, Topic structures and base-generated elements, Back-
ground structures. Therefore the prediction is that all dislocates will not
be sensitive to Island effects. As it has been shown in the current analysis,
right dislocates and contrastive left dislocates show sensitivity to extraction
from Complex NPs. Applying this same principle to the examples in which
pro is claimed, it is seen that in fact movement seems to be active.
(136) * A
to-DAT
mı´
me
dinero
money
Juan
Juan
odio
never
el
me-DAT
hecho
gives
de que nunca me
deje.
‘To me, money Juan never gives it to me.’
Assuming that only HLD elements are not sensitive to Complex NPs,
the number of elements may in truth be the cause of the ungrammaticality
in (136), however (137) is still ungrammatical with only one element.
(137) * Dinero,
to-DAT
odio
me
el
money
hecho
Juan
de
never
que
me-DAT
nunca
gives
me deje.
‘To me, money Juan never gives it to me.’
Yet when a resumptive clitic appears in the matrix clause coindexed with
the Bare Noun, (138), the grammaticality returns.
(138) Dineroi,
to-DAT
odio
me
el
money
hecho
Juan
de
never
que
me-DAT
nunca
gives
me loi deje.
‘To me, money Juan never gives it to me.’
This indicates first, that Bare Nouns may appear in HLD, a base-generated
position, but only when the matrix clause is satisfied with a pronominal ar-
gument, in this case lo, and second, that the Bare Nouns in (134), (135)
and (136) are not base-generated. The notion that HLD is a non-moved
position again is strengthen by the fact that Bare Nouns cannot appear in
RD constructions, with a corresponding clitic (139), or otherwise (140).
41

(139) * Nunca
never
me
me-DAT
lo
it-ACC
deja,
gives
dinero.
money
‘He never lends it to me, money.’
(140) * Nunca
never
me
me-DAT
deja,
gives
dinero.
money
‘He never lends to me, money.’
I suggest that Bare Nouns are not base-generated dislocates of the type
proposed by Casielles-Sua´rez, rather her items show instances of Topicaliza-
tion31, which leaves a gap, not pro. Hence, Bare Nouns can be HLD and
Topicalized, but not CLD nor RD.
This argument, then, is that pro is not active in dislocation construc-
tions32 can be extended to DPs as well, as seen in the contrast between (141)
and (142), where the verbal morphology supports the left dislocate in (142).
(141) * Su
her
coche,
car
Marta
Marta
odia
hates
el
the
hecho
fact
de
of
que
that
todo
all
el
the
mundo
world
quiera
wants
tener
to have
e/pro.
‘Her car, Marta hates the fact that everyone wants to have.’
(142) Su
her
coche,
car
Marta
Marta
odia
hates
el
the
hecho
fact
de
of
que
that
todo
all
el
the
mundo
world
le
her/him-DAT
guste.
pleases
‘Her car, Marta hates the fact that everyone likes it.’
This indicates that base-generated elements in fact need an overt mor-
phological marker to be licit; pro cannot bind a dislocate.
So then, although the Dual Articulation Model predicts that Bare Nouns
as well as other dislocates should be uniformly base-generated, the evidence
seems to suggest that this is not true for all dislocates. In fact, it appears
to strengthen the hypothesis proposed here that only HLDed elements are
base-generated. In addition, it was shown that Bare Nouns cannot be right
dislocates, which points to incompatibility of Bare Nouns to comply with
the type of movement, or the features that drive movement to the [Spec,
31Therefore, in some sense Bare Nouns can escape VP, against Casielles-Sua´rez (2001);
Casielles-Sua´rez (2004). What actually bars the type of movement that would allow Bare
Nouns to surface in subject position is more difficult to pinpoint. Some mileage may be
found in distinguishing the quantificational force of Topicalized constituents, however this
is speculative. I choose here to be agnostic as to what differentiates the type of movement
that Bare Nouns can undergo.
32Ordo´n˜ez & Trevin˜o (1999) makes the further claim that in fact, pro is not active in
Spanish whatsoever, even for null subjects. The argument is that verbal morphology acts
as the ‘clitic’ parallel to the clitic in dislocation constructions.
42

IntTopP] position through which all clause-generated dislocates must pass
before Spell-out.
5.4.3 The Topic in Topic/Comment is CLD
If Background elements are in part derived by movement, then what conse-
quences does this have for the unique Topic position filled only by preverbal
subjects? Under the Dual Articulation Model, the only movement in the ar-
ticulation system was to preverbal subject position yet now evidence shows
that CLD and RD too move from clause-internal position, feasibly making
both CLD and RD candidates for Topichood. Refreshing the reader’s mem-
ory, the critical feature that distinguishes moved elements in the current
proposal is [contrast]. Looking closer at the the original examples provided
by Casielles-Sua´rez for the Topic/Comment structure, (127) and (128) (re-
peated here as (143) and (144)), we see that subjects here can be interpreted
as contrastive.
(143) ¿Quien
who
ha
has
llamado?
called
‘Who called.’
(144) Juan
Juan
ha
has
llamado,
called
no
no
Marta.
Marta
‘Juan called, not Marta.’
Again, just as is the case for CLDed elements, seen in (147) and (148).
(145) ¿Quie´n
who
comio´
ate
las
the
fresas?
strawberries
‘Who ate the strawberries?’
(146) Las
the
fresasi
strawberries
lasi
them-ACC
comı´
ate-1sg
yo
I
pero
but
las
the
uvasj
grapes
no
no
lasj
them-ACC
he
have-1sg
tocado.
touched
‘The strawberries I ate them but the grapes I haven’t touched them.’
Another interesting characteristic that Casielles-Sua´rez points out is that
WH-Constructions are uniformly Background/Focus structures; informative
by nature. Therefore, this makes the prediction that elements that appear
in these constructions must necessarily be Background, or Focus. Follow-
ing this line of inquiry, dislocates that may occur in WH constructions are
necessarily Background, and those that cannot, Topics. Consider examples
(147 - 150), in which we have the various permutations of possible subject
positions.
Subject in Focus Position
43

(147) ¿Que´
what
quiere
wants
Juan?
Juan
‘What does Juan want?’
Subject in HLD position
(148) (Hablando
speaking
de)
of
Juan,
Juan
¿Que´
what
quiere?
wants
‘(Speaking of) Juan, what does he want?’
Subject in preverbal position
(149) * ¿Que´
what
Juan
Juan
quiere?
wants
‘What Juan does he want?’
Subject in RD position
(150) ¿Que´
what
quiere,
wants
Juan?
Juan
‘What does he want, Juan?’
Notice, that subjects may appear in Focus position, HLD and RD po-
sitions however the preverbal slot is disallowed. This is correctly predicted
by the Dual Articulation Model, as preverbal subjects are to be consid-
ered as Topics and therefore not allowed in Background/Focus articulations.
Extending this analysis to WH-Constructions with object displacement, a
similar distribution arises, (151 - 154).
Object in Focus Position
(151) ¿A
to-DAT
quie´n
whom
le
him-DAT
dimos
gave-1pl
ese
that
libro?
book
‘Who did we give that book to?’
Object in HLD position
(152) (Hablando
speaking
de)
of
ese
that
libroi,
book
¿A
to-DAT
quie´n
whom
se
him-DAT
loi
it-ACC
dimos?
gave-1pl
‘(Speaking of) that book, who did we give it to?’
Object in preverbal position
(153) * ¿A
to-DAT
quie´n
whom
ese
that
libroi
book
se
him-DAT
loi
it-ACC
dimos?
gave-1pl
‘Who that book did we give to?’
Object in RD position
44

(154) ¿A
to-DAT
quie´n
whom
se
him-DAT
loi
it-ACC
dimos,
gave-1pl
ese
that
libroi?
book
‘Who did we give it to, that book?’
This evidence suggests that the preverbal position for subjects disallowed
in WH-Constructions, is the same for CLD. This creates a natural class of
preverbal subjects and CLD elements; both are Topics. Thus, I conclude
the two Topics in Spanish, CLD and preverbal subjects, are both underly-
ingly contrastive/Topical ([+contrast]) and both derived by movement. RD
constructions are also derived by movement but the active feature is not
[+contrast], rather the feature [-new].
5.4.4 Section Summary
In this section I provided an overview of Casielles-Sua´rez’s (2004) proposal,
given here as the Dual Articulation Model. This model predicts that there
are two distinct information articulations for Spanish; Topic/Comment and
Background/Focus. The original formulation of this model viewed all dis-
locates as necessarily base-generated and Background elements, while only
preverbal subjects were deemed viable Topics.
However evidence from the current proposal shows that movement does
occur in Background/Focus structures. Evidence suggests that there is no
pro in dislocation constructions and that Bare Nouns, considered dislocates
in Casielles-Sua´Roze’s examples, are instances of Topicalization not disloca-
tion33. Yet, I follow Casielles-Sua´rez in that Bare Nouns cannot undergo the
same type of movement as full DPs. Under given the evidence, this makes
the correct prediction that Bare Nouns may only appear as HLD elements as
this position is base-generated. In addition, I suggest that Topic/Comment
structure can be extended to include CLD as Topical; given evidence from
WH-constructions. Consequently, CLD arguments (including preverbal sub-
jects) fit into Topic/Comment structure, Bare Nouns cannot be DP-moved
thus cannot be subjects, CLD nor RD elements.
Although a better picture of the Information Structure can be gained
from this discussion, we do not have the complete picture. There are still
questions left unanswered: First, what is the discourse function of HLD?
Intuitively, it seems that HLD elements are Topics but at the discourse level
rather sentential, however this is yet to be fleshed out. Second, What is the
discourse function of right dislocate word-order? A finer picture of the left
periphery has been drawn, nevertheless, the seemingly free word-order of
right dislocates gives the impression that there is truly optional structures.
This has yet to be uncovered.
33For a more complex elaboration on the distinction between Topicalization and Dislo-
cation in Spanish see (Rivero 1980)
45

5.5 Analysis Summary
In this analysis I have made the case the basic asymmetry between Left
Dislocation and Right Dislocation is the fact that there are two types of
LD, one generated by movement, one not and one type of RD, uniformly
derived by movement. Movement is motivated by the features [contrast]
and [new], where [contrast] critically distinguishes between CLD and RD
elements attracting them to [Spec, ExtTopP] and [Spec, IntTopP] respec-
tively. I adopt the basic assumptions made in the Split-Topic Analysis and
the Dual Articulation Model, however I augment the first by adding a po-
sition for Hanging Left Dislocates and modify the second allowing for CLD
elements to be considered Topical in addition to preverbal subjects.
6 Conclusion
The larger goal of this paper has been to further disassociate LD from RD
as mirror processes against Kayne (1994) and supporting Villalba (1999) and
Cecchetto (1999). To this effect this analysis provides new evidence from
Complex NPs, Bare Nouns and the semantics of dislocation strategies from
Spanish that supports this conclusion. However, it also has been demon-
strated that on closer scrutiny the proposals by Villalba (1999)and Cecchetto
(1999) cannot cover the empirical data from unboundedness of LD and the
pragmatic distribution of contrastive dislocates. In this investigation it is
shown that understanding LD as one monolithic process simplifies the anal-
ysis of dislocates to the exclusion of the data considered. I argue that in
Spanish it is the case that there are two types of left dislocates; Hanging
Left Dislocation and Clitic Left Dislocation. Therefore the reformulated ac-
count here assumes the underlying framework of the ‘Split-Topic Analysis’
in which both CLD and RD move outside of their merged positions to dis-
tinct projections within the clause, yet a subset of LD, HLD does not share
this derivational history and is base-generated in a clause left-periphery po-
sition. Hence, the availability of HLD to appear only at the extreme-left
of the clause, and elements in this positions ability to be coindexed with
various types of elements, such as full XPs and tonic pronouns, inside con-
stituents traditionally understood as barriers to binding further corroborates
the idea that it is not the result of movement. Furthermore, I adopt the
proposal that there are two distinct information articulations for Spanish;
Background/Focus and Topic/Comment. Nevertheless I contend, first that
Background elements are not uniformly base-generated and second, that
evidence from the distribution of WH-Constructions show that Topic distri-
bution cannot be limited to preverbal subjects and therefore CLD elements
must be considered part of the Topic/Comment articulation. Then adopt-
ing this view of Spanish clause structure and information structure provides
a more principled account for the syntactic and semantic asymmetries ob-
46

served between LD and RD while simultaneously further disassociating the
two as distinct processes.
47

References
Casielles-Sua´rez, Eugenia. 2001. Current Topics in Spanish Syntax
and Semantics, chapter The Syntax and Semantics of Preverbal Topical
Phrases in Spanish, 65–82. Mouton de Gruyter.
——. 2004. The Syntax-Information Structure Interface: Evidence from
Spanish and English. Outstanding Dissertations in Linguistics. Rout-
ledge.
Cecchetto, Carlo. 1999. A comparative analysis of left and right dislo-
cation in romance. Studia Linguistica 53.40–67.
——. 2000. Doubling structures and reconstruction. Probus 12.93–126.
Chafe, W. 1976. Subject and Topic, chapter Giveness, contrastiveness,
definiteness, subjects, topics and point of view, 25–55. Academic Press.
Cinque, Guglielmo. 1990. Types of A’-dependencies. MIT Press.
Contreras, Heles. 1976. A Theory of Word Order with Special Reference
to Spanish. North-Holland, Amsterdam.
Diesing, Molly. 1992. Indefinites. MIT Press.
Givo´n, T. 1983. Topic Continuity in Discourse, chapter Introduction, 5–41.
John Benjamins.
Halliday, M. 1967. Notes on transitivity and theme in english, part ii.
Journal of Linguistics 3.199–244.
Heim, Irene, 1982. The Semantics of Definite and Indefinite Noun Phrases.
Dept of Linguistics South College UMASS Amherst MA 01003: Uni-
versity of Massachusetts, Amherst dissertation.
Kayne, Richard. 1994. The Antisymmetry of Syntax. MIT Press.
Kiss, Katalin E´. 1998. Identificational focus versus informational focus.
Language 74.245–273.
Lambrecht, Knud. 1981. Topic, antitopic and verb agreement in
Non-Standard French. John Benjamins.
——. 1994. Information Structure and Sentence Form: Topic, focus and the
mental representations of discourse referents. Cambridge University
Press.
Lo´pez, Luis. 2003b. Steps for a well-adjusted dislocation. Studia
Linguistica 57.193–231.
48

Ordo´n˜ez, Francisco. 1998. Post-verbal asymmetries in spanish. Natural
Language & Linguistic Theory 16.313–346.
——, & Esthela Trevin˜o. 1999. Left dislocated subjects and the pro-drop
parameter: A case study of spanish. Lingua 107.39–68.
Prince, Ellen. 1985. Fancy syntax and ‘shared knowledge’. Journal of
Pragmatics 9.65–81.
——. 1986. On the syntactic marking of presupposed open propositions. In
CLS, volume 22, 208–222.
Reinhart, Tanya. 1982. Pragmatics and linguistics : an analysis of sen-
tence topics. Philosophica 27.
Rivero, Mar´ıa-Luisa. 1980. On left-dislocation and topicalization in
spanish. Linguistic Inquiry 11.363–393.
Rizzi, Luigi. 1997. The Fine Structure of the Left Periphery, 281–337.
Kluwer Academic Publishers.
Sportiche, Dominique. 1995. Phrase Structure and the Lexicon, chapter
Clitic Constructions. IULC Press.
Torrego, Esther, 1992. Case and argument structure. Manuscript
U/Mass Boston.
Vallduv´ı, Enric. 1992. The Informational Component. Outstanding
Dissertations in Linguistics. Garland.
Villalba, Xavier. 1999. Symmetry and antisymmetry in syntax. Syntaxis
2.
Zagona, Karen. 2000. The Syntax of Spanish. Cambridge.
Zubizarreta, M.L. 1998. Prosody, Focus and Word Order. MIT Press.
—— 1999. Grama´tica descriptiva de la lengua espan˜ola, chapter Las fun-
ciones informativas: Tema y foco, 4216–4302. Real Academia Espan˜ola.
49