The Syntax of Objects: Agree and Differential Object Marking

Miguel Rodríguez-Mondoñedo

B.A., Pontificia Universidad Católica del Perú, 1992

M.A., University of Arizona, 2002

M.A., University of Connecticut, 2006

A Dissertation

Submitted in Partial Fulfillment of the

Requirements for the Degree of

Doctor of Philosophy

At the

University of Connecticut

2007
Copyright by

Miguel Rodríguez-Mondoñedo

2007
 APPROVAL PAGE

Doctor of Philosophy Dissertation

The Syntax of Objects: Agree And Differential Object Marking

Presented by

Miguel Rodríguez-Mondoñedo, B.A, M.A.

Major Advisor ______________________________________________

Željko Bošković

Associate Advisor ___________________________________________

Jonathan Bobaljik

Associate Advisor ___________________________________________

Diane Lillo-Martin

Associate Advisor ___________________________________________

Jairo Nunes

University of Connecticut

2007
 Acknowledgements

Being a graduate student at UConn has been an amazing intellectual journey, an immense

honor, and the most remarkable experience of my adult life. No words could ever express

the depth of my gratitude to all the people that have made this endeavor possible. And yet,

words, even when insufficient, are necessary.

I am very grateful to the members of my committee, for their patience, their generosity,

and their good disposition toward me during the writing of this dissertation. All of them

have being extraordinary.

Above all, Željko Bošković. I could not have dreamed of a better advisor. He has been a

constant supporter and a tireless reviewer of my work. This dissertation would be the

shadow of what it actually is without his comments and suggestions. He has anticipated

my needs, not only before I express them, but before I even notice them. He has pushed

me to think and rethink my ideas, to never be satisfied with a result, to look for more and

more. His seminars, always a joyful linguistic feast, have been the cornerstone of this

dissertation. He has been a role model as a linguist, as a teacher, and as a person. There

will be no way that I could ever repay the enormous personal and intellectual debt that I

owe to him.

I also owe a great deal to Jonathan Bobaljik, not only for his crucial suggestions to

improve my work, but also for his wonderful classes, which have helped me to realize the
limitations of my own perspectives, and to try new ways to look at a particular theory or

phenomenon. I have always enjoyed talking to him; every time, he has a question or a

piece of data that illuminates the discussion and leaves me thinking for a long time. The

issues he has brought to my attention will keep me busy for a long time.

I am also very grateful to Diane Lillo-Martin, for her support, her kindness and her

always relevant comments. I have enjoyed her classes, and I have learned from her to

look at the issues in a broad perspective, to avoid getting lost in the technicalities and to

seek deeper reasons—and, by the way, she has been a wonderful Head of the Department

too.

It has been an honor to meet Jairo Nunes in the last year. His advice has been precise and

illuminating, pointing out new venues and relating my work to other issues. His lectures

at UConn were terrific, with plenty of intriguing data and interesting twists in the

theoretical possibilities. I only wish I could have had more of it.

I would like to express my gratitude to all other members of the faculty, for their

encouragement and their wonderful lectures. In particular, to Susi Wurmbrand, from

whom I have learned that I have to work hard to make my ideas clearer, my assumptions

more explicit, and my conclusions more apparent; her classes have been always crystal

clear and refreshing, and for that I am very grateful. Many thanks also to William Snyder,

whose kindness and generosity I will always remember; he has given me much needed

motivation and meticulous advice—and I am sure he will be a wonderful Head of the

Department too. A special thanks to Harry van der Hulst, for his classes, his confidence

and his constant support.

Many thanks also to Mamoru Saito, Sigrid Beck, Barbara Citko, Yael Sharvit, Andrea

Calabrese and Koji Sugisaki, whose classes have offered me wonderful issues to think

about. I am also glad to have met Jon Gajewski, Mona Anderson, Eva Bar-Shalom, Keiko

Murasugi, Nancy Ritter and Carol Boster. Thank you also to Howard Lasnik for the

enlightening conversations and discussions we had during my first year.

A crucial part of my life in UConn have been the many friends I have made. I am

honored to having shared my student life with people always interested in discussing and

interchanging ideas and information. They have helped me to think about many issues,

including the topics of this dissertation. They have also been very generous with me,

sharing their friendship and their never ending help, every time I needed it.

I want to thank in particular to Jean Crawford, who has been my roommate in the last

couple of years; she has been a great companion and a wonderful friend. Also, to Carlos

Buesa García, who has helped me in many ways (I could not have moved without his

help), and has been a patient endurer of my moods. I will always remember the long

lasting conversations with Bosook Kang, a friend that is far away now, but who will be

with me forever. Takako Iseda has honored me with her friendship and I am grateful for

her constant support and gracious companionship. I also want to thank Shim Sang-Wan

(Pablo), who is a fanatic of Spanish Linguistics, for his friendship and the chance to share
his work. To Ana Bastos, for all these years of discussion and friendship. To Cynthia

Zocca, for her enthusiastic confidence in the inherent goodness of people. To Serkan

Şener and Nilufer Şener, for sharing with me their jokes, their food and their ideas about

almost everything. To Karen O'Brien, for her kindness and good heart. To Lara Reglero,

María Luisa Martí Martínez and Emma Ticio Quesada, the Spanish community in the

department at the time I arrived, for making me feel so welcome. To Elena Koulidobrova

(and Vanessa) for sharing her house, her food and her happiness. To Klaus Abels, Duk-

Ho An, Masashi Nomura, Aritz Irurtzun and Fumikazu Niinuma, for the opportunity to

discuss their ideas and work. To Natasha Fitzgibbons, Pei-Jung Kuo, Krzysztof

Migdalski, Inkie Chung, Sandy Wood, and Hsu-Te Cheng (Johnny) for being always nice

and joyful. To Elaine Grolla, Sarah Felber, Mariana Lambova, Shigeki Taguchi, Simona

Herdan, Jeffrey Bernath, and Takuro Tanaka, for their enthusiasm and discipline. And to

Masahiko Aihara, Nina Radkevich, Sei-Rang Oh, Bum-Sik Park, Miloje Despic, Zhanna

Glushan, I-Ta Hsieh, Toshiko Oda, Koichi Ohtaki, Jill Magee, Natalia Rakhlin, Masahiko

Takahashi, Oksana Tarasenkova, Kyriaki Vassilaki, and Chommanad Intajamornrak, for

contributing to making the UConn linguistic community such a great place. Thank you

also to Judy Marcus and Catalina Ritton for their assistance in so many issues.

Many people outside of the department have helped me in several ways. I cannot

acknowledge all of them here, but I do not want to forget the Peruvian community in

UConn, especially Francisco Quiroz and Teresa Vergara, to whom I owe very much.

Many thanks also to Mario Montalbetti, who first taught me syntax, and to Antxon

Olarrea, Terry Langendoen, Mary Zampini, Jorge Iván Pérez Silva, Alfredo Arnaiz and
Rodolfo Cerrón-Palomino, whose lectures helped me to understand many difficult issues.

And of course, my family and all my friends in Perú, especially my mother, my sister

Mariella and my brother Antonio. I have always had them in my heart.

Colorless green ideas sleep furiously

Noam Chomsky, 1957

Table of Contents

0. Introduction 1

Chapter 1

Case and Agreement in Spanish Existential Constructions 8

1.1 Existential Constructions and Differential Object Marking 12

1.2 A small v in Existentials 22

1.3 Object Agreement: Existentials in SI and SII 30

1.3.1 Restrictions on the internal nominal and Phi-incomplete small v 32

1.3.2 Interpretable [person] in SII and a repair strategy for SI T 44

1.3.2.1 Default Agreement 49

1.3.2.2 Subject Clitics 53

1.3.2.3 T in SII non existential constructions 57

1.3.2.4 The tendency of haber to regularize in SII 58

1.4 Temporal Existential Constructions 63

1.5 More dialectal differences 72

1.6 Agreement with raising verbs 79

1.7 Conclusions 82

Chapter 2

The properties of Spanish DOM 89

2.1 Describing Spanish DOM 91

2.1.1 Specificity and Spanish DOM 93

2.1.2 Animacy and Spanish DOM 123

 2.1.3 Other factors: topicality, event structure 133

2.2 Previous accounts 144

2.2.1 The Functional OT framework 145

2.2.2 The GB/Minimalist framework 152

2.2.2.1 Brugé and Brugger 1996 154

2.2.2.2 Torrego 1998 156

2.3 Agree and DP structure 163

2.3.1 The main mechanism: raising by Agree 163

2.3.2 Inside DP 185

2.4 Summary an conclusions 209

Chapter 3

Beyond the simple transitive clause 211

3.1 DOM in ditransitives 211

3.1.1 Double A 211

3.1.2 The height of the objects 247

3.2 DOM, Small Clauses and Possessor Raising 256

3.3 DOM and Coordination 272

3.4 Remaining issues 279

Chapter 4

A brief cross-linguistic exploration of DOM 285

4.1 Is DOM universal 288

4.1.1 The question of markedness reversal 292

 4.1.2 Object Shift is not DOM 298

4.1.3 The Specificity Condition does not drive DOM 301

4.2 The DP-vP connection: Kannada 302

4.3 Systematic DOM differences 309

References 315
 Introduction

Spanish Existential Constructions (SEC) with haber (“to be-existential”) exhibit at least

two peculiar properties. First, not every type of nominal can occur as an internal nominal

of haber-sentences; for instance John or he/him cannot be the internal nominal of haber.

At first sight, this restriction seems to be the familiar Definiteness Effect that also

characterizes English there-be sentences. However, a closer examination reveals that

Spanish is more restrictive than English in this respect. There-be sentences do admit

definite/specific nominals, but, when this happens, a special reading is triggered, namely,

the so called list-reading. Spanish haber-sentences do not allow list-readings with the

nominals in question; when a nominal like John or he/him is included as an internal

nominal of haber, the sentence is plainly ungrammatical. The second property, widely

recognized in the literature, is that the internal nominal in SECs is an accusative object

and not a nominative subject (see (2b) below). Significantly, ungrammatical SECs with

objects like Juan or him cannot be rescued by using the marker A, which can mark

objects of this type with all transitive verbs (the so called Differential Object Marking).

In other words, the objects that cannot appear in haber-sentences are exactly the ones that

can be marked with A with all transitive verbs. This situation asks for a unified

explanation.

This dissertation, taking the above correlation as a point of departure, investigates the

nature of Differential Object Marking (DOM), using Spanish as a case study. I propose

that DOM can be understood by exploring the interaction between case and agreement; in
particular, by assuming that the operation Agree is sensitive to the feature specification of

probes and goals.

My main theoretical tool is the operation Agree, as defined in Chomsky 2000, 2001 and

much more subsequent work (in particular, Bošković 2005, 2007a,b). Agree predicts a

correlation between case and agreement that depends on the matching of φ-features in the

elements involved in the operation, that is, the probe and the goal. A probe P searches

inside its domain for a goal G that matches its features. For P to be able to probe G, P

needs to have an unvalued feature; otherwise there would be no reason for it to function

as a probe. This unvalued feature in P will be valued by a matching G with valued

features; when P is a traditional case-assigner and G a nominal element, an Agree relation

between them may also result in case-valuation (that is, case licensing) of G. However, if

there is a difference between the feature specifications of probe and goal, valuation may

not occur. In particular, P must match all the φ-features of G in order for it to value the

case feature of G. If a probe is φ-incomplete, it cannot value the case feature of a φ-

complete goal.

A particularly interesting piece of data that supports this conclusion is provided by a

dialectal split that SECs exhibit with respect to agreement. For a subset of Spanish

dialects (which I call SI), agreement with the internal nominal is not possible (1), but in

other dialects (SII), the internal nominal agrees with the verb, despite the fact that the

nominal in question is an accusative object (as shown by the cliticization (2a-b):

2
(1) a. Hubo dos estudiantes en la fiesta SI

Was two students at the party

There were two students at the party.

b. Los hubo

CL-was

(2) a. Hubieron dos estudiantes en la fiesta SII

Were two students at the party

There were two students at the party

b. Los hubieron

CL-were

In this dissertation, I argue that the constructions in (1) and (2) have a small v, which is

responsible for accusative case licensing. Small v, which I argue has only the [number]

feature (that is, it is φ-incomplete), probes the internal nominal in (1)-(2), valuing its own

number feature. Regarding the agreement difference in (1)-(2), I argue that T in SII has

only one unvalued φ-feature, namely [number], its [person] feature being valued. As a

result, T in SII can probe small v, which yields the agreement in (2). On the other hand, in

SI, where I argue T has all its φ-features unvalued, the φ-incomplete small v cannot value

the feature of T, and then T must receive default agreement. Notice also that, if small v is

assumed to be φ-incomplete, then small v cannot value the case feature of nominals that

3
are φ-complete. I show that this straightforwardly accounts for the absolute impossibility

of having nominals like Juan or him as internal nominals of haber, even under list-

readings, under the assumption that the nominals in question are φ-complete. If the small

v in SECs (responsible for checking the accusative case of φ-incomplete nominals) has

the same feature specification as the small v in all other transitive constructions, the same

mechanism accounts for DOM. This means that assuming a φ-incomplete small v for

SECs is not an ad hoc device for a particular construction, but a core property of Spanish

(and perhaps of other DOM languages as well).

In general, DOM is manifested in Spanish by the fact that nominals that are both specific

and animate are marked with A. I contend that these nominals have a complete set of φ-

features. All other nominals are φ-incomplete; in particular, they do not have the feature

[person]. Additionally, as mentioned above, I contend that small v is φ-incomplete, that is,

it does not have the feature [person]; this means that small v cannot match the φ-features

of φ-complete nominals (like John or him), hence is not able to value their case. This

system provides a uniform account for the ban on certain nominals inside SECs and

Spanish Differential Object Marking. Nominals like John or him (which are φ-complete)

cannot be objects in SECs because there is no case checker available at the point where

they need to check case. In regular transitive constructions, there is an additional case

checker, a Dative head, which is able to value the case of φ-complete nominals. As a

result, such nominals get the A-marker.

As is commonly assumed, syntactic derivations proceed in different stages, called phases.

4
The complement of a phase is spelled-out (i.e. shipped to the interfaces) once the phase is

completed. This means that all the elements inside a phase complement must check all

their features before the phase level is reached. The small v in transitive constructions

heads a phase. As a result, if an element is trapped inside the complement of small v

without having checked all its features, the derivation crashes. Following Bošković 2005,

2007, I assume that if a nominal inside a phase complement has not checked all its

features, the nominal must move to the edge of the phase to avoid being spelled-out with

unchecked features (which, incidentally, allows us to dispense with the EPP). In DOM

constructions, the nominal raises to [Spec, vP], from where it is able to establish a new

Agree relation with a Dative head (following Bošković’s system, I argue that the nominal

in fact raises to [Spec, DatP]), which accounts for the presence of the dative marker (in

Spanish as well as in the majority of DOM languages, the marker for DOM objects is

also the dative marker).

The dissertation is divided in four chapters. The first chapter explores the properties of

SECs, including the ban on φ-complete nominals and the aforementioned dialectal

difference with respect to agreement (SI and SII). I extend the analysis briefly outlined

above to other constructions, including the so called temporal existential constructions

and agreement with raising verbs, where parallel dialectal differences are attested.

The second chapter is dedicated to the properties of Differential Object Marking with

Spanish simple transitive verbs. Contrary to the majority of the literature, I claim that A

is not a marker of specificity per se; that is, although it is true that objects that are both

5
specific and animate must be marked with A, animate objects that are not specific can

optionally be marked with A too. I capture this difference by assuming that Spanish

determiners have two versions: D and *D. I contend that *D determiners must be non-

specific, whereas D determiners can be specific or non-specific. Additionally, *D

determiners, which are inherently non-specific, are incompatible with the [person] feature.

This means that nominals with *D determiners end up being φ-incomplete, and then they

are able to check case against small v (so A will never appear with them). D nominals

host a [person] feature (if animate), in which case they are φ-complete, unable to check

case against small v. As a result, they check their case with the Dative head. The above

predicts that A must appear with D nominals (which can be specific or non-specific), but

it is not mandatory with non-specific objects, since the latter have also the *D option. The

chapter presents a thorough discussion of specificity in Spanish DOM (with both definite

and indefinite determiners), which is implemented by using global and local choice

functions. I also discuss several pieces of data that have been considered exceptional in

Spanish DOM, in particular, the presence of A with inanimate objects with some verbs,

and the mandatory presence of A with some clearly non-specific quantifiers (like nadie).

I offer an account of these previously puzzling issues that supports the current framework.

The third chapter analyzes DOM with ditransitives and coordinate structures. I present

data that are not discussed often in the DOM literature (although some of the data have

been identified a while ago). In some cases, in ditransitive constructions, the A marker

for the Direct Object (DO) can drop if an Indirect Object (IO) is also overtly present.

Interestingly, this dropping is sensitive to the heaviness of the objects, the presence of a

6
pause between DO and IO, and word order. I show that the dropping can only happen if

both objects are in the same linearization domain and provide an explanation for the

generalization in question. I also discuss some of the effects of DOM on coordination,

which includes the well-known ban on the coordination of DOM and non-DOM objects,

and certain restrictions on the possibility of pseudo-gapping with DOM and non-DOM

objects.

The fourth chapter briefly discusses some of the properties of DOM beyond Spanish. The

possibility that DOM may be a phenomenon broader than traditionally thought is

explored. I examine the key differences between DOM and several constructions and

operations in other languages, including Object-Shift, antipassive, the ba-construction,

the ban on extraction from specific objects (the Specificity Condition). I also show that

my analysis of Spanish DOM can account for Kannada, another DOM language, without

any substantial modification.

Although much more cross-linguistic comparison is necessary, the suggestion that DOM

is a result of an interaction between case and agreement (that is, that DOM arises because

the initial case-checking head, small v, is unable to check accusative case of certain

nominals) opens up, I believe, a new venue to investigate Differential Object Marking,

since it makes DOM a core grammatical phenomenon. This dissertation aims to establish

this idea.

7
 Chapter 1

Case and Agreement in Spanish Existential Constructions1

In Spanish Existential Constructions (SEC) (1a), the internal nominal (which corresponds

to the so called “associate” in English) receives Accusative Case (ACC), as we can see

from the presence of the ACC clitic (CL) in (1b):

(1) a. Hay un hombre en la habitación

Is a man in the room

There is a man in the room.

b. Lo hay

CL-ACC is

There is a long standing tradition among Spanish grammarians in favor of considering the

internal nominal of SEC to be an object, not a subject—see Fernández-Soriano and

Táboas-Baylin 1999: 1754-1759 for a review of traditional grammar analyses of haber

sentences, and also Suñer 1982b, a, Torrego 1984b, Díaz 2004, Rodríguez-Mondoñedo

2006d, among others. The main arguments concern the post verbal position of the

nominal and the ACC cliticization.2 Following this tradition, in this dissertation, I will

1
Previous versions of this chapter have been published in Rodríguez-Mondoñedo 2006b, c, d.
2
This position has been challenged occasionally; see Cuervo 1881, Luque Moreno 1978, García Llebra
1983, Nicita 1997, Treviño 2003. This literature uses three types of argument against the claim that the
nominal in question is not a subject. First, they characterize the nominal, which is the only argument in the
sentence, as the “psychological subject” (as opposed to “grammatical subject”). Second, they show that the
nominal agrees with the verb (in SII), witch, they argue, shows that it has become a true “grammatical
subject”. Third, they observe the absolute lack of preposition a, which usually appears with some direct

8
assume that the internal nominal is not a subject. 3

Spanish is split in two dialects with respect to SEC. In Spanish Dialect I (SI), the verb has

a default value (3rd person, singular):

(2) Hubo dos hombres en la fiesta SI

was two men in the party

There were two men at the party

But in Spanish Dialect II (SII), the verb agrees with the ACC nominal, even when the

nominal is a clitic:4

(3) a. Hubieron dos hombres en la fiesta SII

were two men in the party

There were two men at the party

objects in Spanish. Obviously, the first argument makes use of a different notion of subject; in this
perspective, subject is a notion independent of structural considerations, a view from which I depart (i.e. I
assume that the subject is defined structurally). I will show in the following section that the nominal does
not actually enter into an Agree relation with T in SECs, eliminating the second argument. I will also give
reasons why the third argument is not strong enough—see footnote 13.
3
There are actually some cases where the nominal precedes the verb overtly in Spanish Existential
Constructions. However, as discussed in Torrego 1984b: 335, Runner 1992: 288, and others, in order to be
grammatical with this word order, the nominal must have special intonation:

(i) UN HOMBRE había en el jardín

A MAN was in the garden

This is taken as an indication that the nominal is fronted (with A’-movement)—see Rodríguez-Mondoñedo
2006d for further discussion.
4
I will use SI for the dialect without agreement, and SII for the one with agreement. When there is no
indication, the sentence is good or bad in both dialects.

9
 b. Los hubieron

CL-ACC were

Given that the nominal [dos hombres] is Accusative the data are a direct challenge for

theories that link together Agreement and Case (like Chomsky 2000, 2001, 2004). In SII,

the nominal seems to be in agreement with T, but T cannot be its Case-licensor because T

does not value ACC. This means that Chomsky’s 2000, 2001 account of English

Existential constructions, according to which the internal nominal in there-be sentences

(the so called “associate”) checks case against T and receives nominative, cannot be

directly extended to Spanish.

In this chapter, I will present an analysis of SEC that nevertheless supports a relation

between Case and Agreement, as well as the operation Agree (Chomsky 2000, 2001,

2004, 2005b, a, 2006), by postulating a small v in SEC. The proposal is compatible with

other studies that have proposed that there are several types of small v, with different

properties (see Boeckx 2003, Legate 2003, Folli and Harley 2004, among others).

I will contend that small v is φ-incomplete. This small v probes the internal nominal and

checks its Case (Accusative) only if the nominal is also φ-incomplete. This proposal

predicts that nominals that are φ-complete, for instance proper nouns or some personal

pronouns, will be banned from haber sentences, given the inability of this φ-incomplete

small v to check the Case of φ-complete nominals; I demonstrate that the prediction is

borne out—the elements in question are disallowed in SEC even with list readings, in

10
contrast to English.5

I will argue that the difference in agreement between the two dialects comes from a

difference in the specification of φ-features in Tense (T). We will see that as a result of

this difference, in SII, the operation Agree can value the φ-features of T using the small v

as a goal, but in SI Agree fails to do so, and therefore T in this dialect has to resort to a

mechanism of default agreement. Interestingly, the verb is irregular in SI, but it tends to

regularize in SII, a fact that will be attributed to the above difference in the process of φ-

valuation.

I will show that the current analysis enable us to explain a number of additional

properties and procedures in Spanish, including some correlations between SI and SII

with respect to agreement in temporal existential constructions, raising verbs with haber

(have, be-existential) and certain issues regarding the syntax of estar (be-locative) and

parecer (seem).

Furthermore, I will show that, if extended to all Spanish small vs in the relevant respects,

my proposal explains the syntax of Spanish ACC objects, in particular, it predicts that, if

the φ-incompleteness of the small v prevents this element from valuing the Case of direct

object (DO) when the DO is φ-complete, then the DO must raise to check its Case with

5
The clitic lo in (1b) will be argued not to be specified for [person]. On the other hand, the pronouns in (i),
which are unacceptable even in the list reading, are specified for [person] (see the discussion in section
1.3.1):
(i) *Me/Te/Nos había
Me/Te/Us was
There was me/you/us

11
another head (a Dative head). I will show that this explains the presence of the element

“a” that precedes the objects in question (the so called Differential Object Marking),

illustrated in (4):6

(4) Juan besó a María

John kissed A Mary

John kissed Mary

The rest of the chapter is organized as follows. In Section 1.1, I will establish the

fundamental intuition regarding the correlation between Existential Constructions and

Differential Object Marking, and its relation with the Definiteness Effect. In Section 1.2,

I will present the proposal that there is a small v in SEC. In Section 1.3, I will refine the

idea by proposing that the small v in haber-sentences is φ-incomplete and that, in SII but

not in SI, T has an interpretable φ-feature. I will discuss some consequences of this idea

with respect to temporal existential constructions in section 1.4, with respect to additional

dialectal differences in section 1.5, and with respect to raising verbs in section 1.6.

Section 1.7 presents the conclusions for this chapter.

1. 1. Existential sentences and Differential Object Marking

Like English there-be sentences, Spanish haber-sentences display a Definiteness Effect

(DE, Milsark 1974). However, Spanish existential sentences seem to be more restricted,

since they do not allow nominals that are specific and animate, not even with list-
6
The full technical details of the DOM system will be discussed in chapter 2.

12
readings:7

(5) a. Había un hombre en la habitación

was a man in the room

There was a man in the room

b. * Había Juan en la habitación

was the man in the room

* There was the man in the room

(6) a. Lo/La había

CL-MAS/CL-FEM was

b. *Me/Te/Nos había

Me/Te/Us was

There was me/you/us

Zucchi 1995, McNally 1997, Keenan 2003, among several others, argue that the DE

should receive a purely semantic/pragmatic explanation. In that sense, we could argue

that (5b) and (6b) merely shows the DE at work, and in fact this has been proposed (see

Saab 2006). I will show here that this cannot be the whole story (see also Rodríguez-

Mondoñedo 2006a).

7
See below for discussion of (6a).

13
For McNally 1997, an account of the DE should include an explanation for the list-

readings that we find with specific nominals inside existential sentences (7b):

(7) a. —Is there any taxi driver in this town?

b. —Well, there is John.

According to her, the DE arises from the requirement of a novel discourse referent to

support the truth of the existential assertion; however, this requirement can be overridden

if the existential sentence is presupposed to be true. This is established in condition (8):

(8) If an existential sentence is presupposed to be true in the context in which is

uttered, its re-introduction into the context is not accompanied by the instantiation

of any discourse referent.

(McNally 1997: 327)

Condition (8) correctly predicts list-reading sentences in the right context. These

sentences, as in (7), generally appear after questions, with raising intonation, with no

prepositional phrase (Milsark 1974, Rando and Napoli 1978, Hannay 1985, Safir 1985,

Abbott 1992, McNally 1997).

Significantly, Spanish does not accept this type of sentences with the verb haber, which

is the verb used in Existential Construction:

14
(9) a. —Hay algún taxista en este pueblo?

Is there any taxi driver in this town?

b. —* Bueno, hay Juan

—Well, there is John.

Sentences like (9b) do not have a list-reading, they are plain ungrammatical. It is

important to stress that we cannot attribute this fact to the Definiteness Effect alone

(contra Saab 2006), because in (9b), given the list-reading context, the DE is expected to

be cancelled by (8), as in (7b). This does not happen in Spanish, which raises a potential

problem for (8). I contend, however, that the reason for the ungrammaticality of (9b) is

syntactic, and that Condition (8) can be maintained, even for Spanish.

What is particularly interesting—and I believe this is the core difference between the

Spanish DE and the English DE—is that the nominals that are not allowed even under

list-readings with haber are specific and animate. To obtain a list-reading with these

nominals, Spanish must use estar:8

(10) a. Está/*Hay Juan

Is John

There is John

8
This restriction survives in non-restrictive haber-relatives (to form a relative here, we need estar):
(i) Juan, quien *hay/está en la esquina
John, who is in the corner

15
 b. Estás/*Has tú

are you

There is you.

A very similar effect obtains with Spanish unmarked direct objects. In Spanish, specific

animate Direct Objects (DO) are normally marked by the preposition a “to”, which is the

dative preposition:

(11) a. Besó una mujer UNMARKED ANIMATE OBJECT

kissed a woman (non specific)

He kissed a woman

b. * Besó María

kissed Mary

He kissed Mary

(12) a. Besó a una mujer MARKED ANIMATE OBJECT

kissed PREP a woman (specific)

He kissed a woman

b. Besó a Mary

kissed PREP Mary

He kissed Mary

16
I will explain the details of this phenomenon, which is an instance of the so called

Differential Object Marking (DOM, see Bossong 1982, 1983-1984, 1985, 1991, among

many others) in chapter 2. For now, to capture these facts, I will use the generalization in

(16)—which is very similar to the one provided by Heusinger and Kaiser 2003: 53:9

(13)

[+animate]
Marked Object
[+specific]

[+animate]
[-specific]

[-animate]
Unmarked Object
[+specific]

[-animate]
[-specific]

Differential Object Marking has been observed in a number of languages—Bossong 1991,

Torrego 1998, Aissen 2003 and the references therein. Among Romance languages,

Spanish shares DOM with Rumanian,10 but we also find DOM in languages as diverse as

Hindi and Turkish. Let me illustrate the phenomenon with respect to Turkish, which is

not exactly like Spanish because it does not have an animacy constraint. However, what

is important for us is that the accusative suffix –yı is dropped if the object is non specific

(just like the Spanish preposition a is dropped with non specific animate objects):

9
As we will see in chapter 2, this generalization has severe limitations, which have not prevented it from
being the standard view on DOM in the literature (see Zagona 2002: 13). I will give a detailed description
of this issue in chapter 2.
10
In Spanish and Rumanian, DOM is very robust. But DOM can also be found in non-standard varieties of
Italian (Southern Dialects in particular), Gallego, Portuguese and French (see Pensado 1995a: 14-16, de
Jong 1996: 53-93).

17
(14) a. Bir araba isti-yor-um UNMARKED OBJECT

a car want-PRES-1SG

I want a car (non specific)

b. araba-yı isti-yor-um MARKED OBJECT

car-ACC want-PRES-1SG

I want the car or

I want a car (specific)

Furthermore, with Turkish existential sentences, the internal DP cannot have the suffix –

yı. As in English and Spanish, in this case, the internal DP has a non specific reading:11

(15) a. Buranda bir araba var

Here a car exist

There is a car here

b. * Buranda araba-yı var

Here car-ACC exist

There is the car here

In other words, it seems that there is a correspondence between being a marked object

and some particular interpretation, which I will refer to as INT (this is the interpretation
11
I thank Serkan Şener and Nilufer Şener for discussion of these data.

18
marked objects have). In addition, the unmarked object receives the complement of INT

(which I will call INT’). For Turkish, INT is [+specific] and therefore INT’ will be [-

specific]. To the extent to which (13) is correct, for Spanish, the situation seems to be as

follows:12

(16)

[+animate] a
Marked Object INT
[+specific]

[+animate] b
[-specific]

[-animate] c
Unmarked Object INT’
[+specific]

[-animate] d
[-specific]

This is similar to what we have observed with respect to the restrictions on the internal

nominal of haber sentences, which I will now express as the following:

(17) The nominal of haber always receives INT’ and never INT.

This means that its object will always be unmarked (without preposition a), which is, of

course, true.13 There is nothing particularly new here. But, by combining (16c) and (17),

12
As we will see in chapter 2, one major problem for this generalization is that marked objects actually do
not need to be [+specific]. It is, however, true that unmarked objects cannot be both [+animate] and
[+specific], which means that (17) holds, and this is the relevant generalization with respect to Existential
Constructions.
13
In principle, this removes the possibility of using the lack of preposition to suggest that we are not
dealing here with an ACC object, which has been proposed occasionally (see, for instance García Llebra
1983, Groat 1999)—see footnote 2.

19
we can make an additional prediction:

(18) Objects that are [-animate] but [+specific] can be haber objects.

Notice, however, that (18) must be restricted by the Definiteness Effect, which bans

[+specific] nominals from existential sentences (for semantic reasons, as explained). This

means that [-animate] nominals that are [+specific] can appear under haber with list-

readings (which are the readings left over once we factor in the DE):

(19) a. —Hay alguna diversión en este pueblo?

Is there any entertainment in this town?

b. —Bueno, hay la piscina

Well, there is the pool

This also means that objects that are both [+animate] and [+specific] cannot be under

haber, no even with list-readings. This is exactly what we found when we discussed (9b).

In fact, some researchers have used data like (19) to argue that there is no Definiteness

Effect in Spanish existential constructions, which is not true. For example, Suñer 1982a:

70 denounces what she calls “the myth of Definiteness Restriction in hay sentences”.

Consider these sentences, where the definite article (el=the) is allowed in haber-

environments:

20
(20) …y allí no hay el problema de…

…and there, there-isn’t the problem of…

[Suñer 1982a: 70]

These nominals are [-animate]. If we shift to a [+animate] object, the resulting sentence

becomes ungrammatical:

(21) *…y allí no hay el estudiante de…

…and there there-isn’t the student of…

In fact, Suñer 1982a: 82 notes that haber-sentences are degraded with specific and

animate objects, but she does not attach any significance to this fact, arguing that this is a

side effect from the requirement of preposition with this kind of object in Spanish. I claim

that this is not a “side effect”, but the main effect we need to explain, given the following

generalization that we can now postulate:

(22) If a nominal is allowed in haber-sentences, then it will be allowed as an unmarked

object with any other transitive verb.

Unfortunately, the generalization has at least one strong exception. Some negative

quantifiers are allowed under haber but they require A with other verbs:14

14
Throughout this dissertation, I will use capital A to indicate the “a” (lit. “to”) used in Spanish Differential
Object Marking.

21
(23) a No hay nadie en casa

Not is nobody in house

There is nobody at home

b No maté *(a) nadie

Not killed (to) nobody

I didn’t kill anybody

I will show in chapter 2, however, that the presence of the preposition with nadie in (23)

is accidental, depending on independent properties. If (22) is correct, the DE in existential

sentences and the DE with unmarked animated objects are the same phenomenon, at least

for Spanish (and Turkish, mutatis mutandi). Then, we may entertain the idea that what

unifies both DEs is a grammatical property that is common to both of them, which in turn

correlates with (16). I will discuss this property in the next section.

1.2. A small v in existentials

In this section I will present the idea that haber-sentences in Spanish have a small v that

is responsible for the ACC Case. Actually, under the Agree system, given that we have an

ACC object (cf. 1), there is no choice but to postulate a functional head with φ-features

and the ability to probe the object, that is, small v:15

15
See Deal 2006 for the idea that, even English there-be sentences need a small v. For German, which also
exhibits ACC in the existential construction, Czinglar 2001 proposes a Voice projection (in addition to T).

22
(24) Hay un hombre en la habitación

Is a man in the room

There is a man in the room

(25) [TP T-φ [vP v-φ [VP hay [SC [un hombre] [en la habitación]]]]]

Under the Agree system (cf. Chomsky 2000, 2001, 2004), T and v have a set of

uninterpretable φ-features (person, number, gender) that are unvalued. They probe into

their c-command domain, looking for goals with valued φ-features—a nominal, for

instance. Once they match a goal, the goal will value their uninterpretable φ-features, and,

as a reflex of this, the goal will get its uninterpretable Case-feature valued. It is a matter

of controversy how this “reflex” happens. For Chomsky 2000, 2001, 2004 it just happens

as a stipulation and in situ; for Bošković 2005, 2007b this is obtained by positioning a

valued [case] feature in the probe (that would be ACC in small v), and making the goal

move to c-command the probe to check case (the relevant DP thus becomes a probe,

probing small v, which then serves as a goal, under Agree). The checking of ACC case in

Spanish Existential Sentences does not provide evidence to tease apart these possibilities.

For that reason, I will outline the details of the Agree system using Chomsky’s

assumption, which will suffice as expository device. However, we will see in chapter 2

that, when we extend the analysis to Spanish DOM, the second alternative makes

interesting predictions and proves to be useful. The reader should keep this in mind.

23
In the above system, a probe cannot be left with its φ-feature unvalued, since Spell Out

will not be able to delete them, producing a crash. A goal must value its Case-feature for

the same reason. If T is the probe, the goal will be NOM; if v is the probe, the goal will be

ACC. This is achieved by the operation Agree. For now, let me say that the operation

Agree applies as shown in (26). The small v probes the internal nominal, and the nominal

values small v’s φ-features under Agree; as a reflex of this, the nominal gets ACC:16

(26) a. Match
v un hombre
[φ: ] [φ: 3p, sing]
[Case: ]

b. Valuation
v un hombre
[φ: 3P, SING] [φ: 3p, sing]
[Case: ACC]

Given this proposal, some problems immediately arise. According to Chomsky 1995,

small v has the lexical property of having an external argument and valuing the ACC of

the internal one. This is the way Chomsky encodes Burzio’s Generalization, according to

which:

(27) All and only the verbs that can assign a θ-role to the subject [i.e. an external

argument] can assign accusative Case to an object.

[Burzio 1986:178]

The small v that I propose for SEC can value the ACC of the nominal but it does not have

16
I am abstracting away from [gender] in this dissertation.

24
an external argument. Since, for Chomsky, BG is a lexical property of v, nothing prevents

other kinds of v from having different lexical properties. In fact, if the lexical properties

[external argument] and [valuing ACC] are assumed not to necessarily correlate, we

expect four types of v:

(28) a. [+external argument] and [+valuing ACC]

b. [+external argument] and [-valuing ACC]

c. [-external argument] and [+valuing ACC]

d. [-external argument] and [-valuing ACC]

Notice that at least two of them are attested: (28a) is Chomsky’s 2000 *v for transitive

verbs, and (28d) is the unaccusative v. In addition, (28b) could be associated with

unergative verbs (assuming that there is no cognate object, or that the cognate object does

not receive ACC in narrow syntax).17 If this line of reasoning is correct, the lack of a v like

(28c) would be a gap in the paradigm. Of course, (28c) is perfectly suited for Spanish

existentials. Thus, I conclude that BG does not hold empirically. Actually, we know that

BG is violated in several languages and constructions (see Haider 1985, Yip et al. 1987,

Marantz 1991, Harley 1995, Reuland 2000, López 2001, Woolford 2003 and many

others); in fact, most researchers (including Burzio 2000) consider BG to be an

epiphenomenon (see the papers in Reuland 2000).

17
Under Hale and Keyser 1993 framework, a transformation from a combination like [V [N walk]] into [V
walk] would take place in the pre-syntactic component (a sort of lexical syntax). In this scenario, an
incorporation N-to-V could render the Case feature inactive. Since this type of verbs does not have an
object position in narrow syntax, then we need something like (28b). However, it could be argued that the
process described above happens in the syntax, and then unergative constructions would simply be
transitive constructions. I leave this issue open.

25
There is another problem with (25)-(26): apparently, the φ-features of T remain unvalued,

so, under Chomsky’s system, (24) should be ungrammatical, since the nominal does not

enter into an Agree relation with T, and therefore the φ-features of T remain unvalued.

We cannot say that a null expletive EXPL takes care of the φ-features of T because

expletives are base-generated in Spec, TP (Chomsky 2000, 2001, Bošković 2002a), so T

cannot probe EXPL, under the assumption that probing happens only under c-command

(that is, the probe has to c-command the goal). More importantly, EXPL is supposed to

be φ-incomplete, that is, it has only [person] and not [number], thus, it is unable to value

the features of a φ-complete T.18

Interestingly, according to Chomsky 2000, in English existential sentences, once T has

valued its φ-features by probing the nominal, the expletive THERE can probe T and

value its own [person] feature. This means that valued heads can be goals:

(29) There T are two men in the garden

[person] [person] [person]
[number] [number]

If any head with valued φ-features may be the goal for a probe with unvalued φ-features,

we can value the φ-features of T in sentence (24) using the small v (whose φ-features

have been valued by the internal nominal) and Agree:

18
Notice that a difference needs to be established between THERE-expletives and IT-expletives. The
former are defective and cannot value T; the latter (which also include, for instance, French il) can value T.
As should be apparent, there cannot be an IT-expletive in SEC, since this would predict that T would
always have the same φ-features, which is not true (compare (1) vs (3)). I thus assume that there is no null
expletive in SEC (as argued by Picallo 1998 and Rodríguez-Mondoñedo 2006d; for a different view see
Torrego 1984b).

26
(30) a. Match
T v
[φ: ] [φ: 3p, sing]

b. Valuation
T v
[φ: 3P, SING] [φ: 3p, sing]

This is possible because the φ-features of v are still there after it undergoes Agree with

the nominal, which values its φ-features, since the deletion procedure applies only at the

point of Spell-Out. Although they have already been valued, the φ-features have not been

deleted because Spell Out applies just at the end of a strong phase. I assume that the small

v in question is not the head of a strong phase, because it does not have an External

argument. Notice that this implies that there is a mechanism of mediated agreement: the

nominal values small v and small v in turn values T (under the conditions discussed in the

next section).19

Notice also that the small v in question does not have any unvalued [Case: ] to value.20

At first sight, this is trivial (v just doesn’t have to check Case), but it allows us to make a

crucial observation: a head (in this case T) can get its φ-features valued without assigning

any Case.21 This seems to be true for T in both SI and SII:

19
For related proposals in other languages and constructions see Butt 1995, Legate 2005, Frank 2005,
among others (see also Bhatt 2005 and Bobaljik 2006 for discussion of the limitations of this mechanism).
20
The situation is still the same if we assume, with Bošković 2005, 2007b, that there is a valued [case: ACC]
in v: in this framework, the feature comes with small v from the lexicon, so it does not come from Agree.
More details on this system will be provided in chapter 2.
21
Notice that this does not appeal to Chomsky’s 2000, 2001 notion of defective head, according to which a
head (for instance, an infinitival T) can get a value for its φ-features but it is unable to value the case
feature of the relevant nominal. In (30), it is not the case that T cannot value a case feature (so, it is not
defective); instead what happens is that small v does not have any case feature to be valued.

27
(31) a. Llueve

rains

It rains

b. [TP T-φ [VP Llueve]

In Spanish (both SI and SII), T with φ-uninterpretable features can get a default value [3p,

SING] if no head with interpretable φ-features is available in its c-command domain. This

default value could be implemented via a pronominal subject clitic (SCL) that values the

φ-features of T, rendering inactive its ability to value NOM. The morphological

manifestation of this subject clitic could be the [3p, SING] suffix that we find in these

verbs, in accordance with similar ideas developed in Alexiadou and Anagnostopoulou

1998, Picallo 1998, Kato 1999, 2000, Díaz 2004, Ticio 2004.22

Let me emphasize another assumption to be in position to fully assert my proposal. As

has been observed (Boeckx 2003), the requirement for a head with uninterpretable φ-

features (the probe) to value its features in order to value the Case-feature of the goal is a

way to encode the Inverse Case Filter—that is, the requirement for a traditional Case-

assigner to discharge its Case to an assignee (Bošković 1997b, 2002a)—in the Agree

system. This also gives us a way to derive a weak version of the BG (actually Chomsky

2000 suggests this too), that is, it is possible to assume that there is a type of small v

which must value accusative and assign an external theta role, without precluding the

22
In principle, this subject clitic could replace the idea of a null expletive as responsible for valuing the φ-
features of T, and it also could render unnecessary the idea of default agreement. I will come back to this
issue later in the chapter.

28
possibility of other types of small v (as in (28)).

Assuming that T is always present in a sentence (nothing really new here), if a second

lower φ-head (small v) is present, there are only two (grammatical) options:

(32) i. There is another argument between T and the second φ-head (the External

argument), and T values its φ-features by probing this argument.

ii. There is no other argument between T and the second φ-head, then, either:

a. T values its φ-features by probing the second head (provided that this

head has already valued its own φ-features by probing the internal

argument), or else

b. A subject clitic/default agreement values the φ-features of T.

According to (32i), we need an external argument to value the φ-features of T, since the

internal argument is in the domain of another probe. Obviously, this makes the BG an

epiphenomenon, since it depends on the necessity of T to value its φ-features, but BG can

be violated if the φ-features are valued using a goal different than the External argument.

According to (32ii) a probe can in turn be visible as a goal for another c-commanding

probe, even if it has no Case feature, presumably violating the Activation Condition

(AC)—according to which a goal must have an uninterpretable feature to undergo

agreement.23 A solution has to be developed, even under stricter Chomsky’s assumptions:

otherwise, his idea that the expletive probes T in English existentials constructions like

(29) would violate the AC, since T has no Case-feature. Notice, however, that for
23
See Bošković 2007b for arguments that the AC does not hold for Agree.

29
Chomsky, AC depends on the presence of any uninterpretable feature. T is then still

active since its φ-features are uninterpretable. Since the same is true in (32ii), there is no

problem with respect to AC here.

Having established the plausibility of (30) in Chomsky’s Agree system, in the next

section, I turn to the dialectal difference noted above with respect to agreement in SEC. I

will slightly adjust (30) to account for it.

1.3 Object Agreement: existentials in SI and SII

In this section, I will present an analysis of the dialectal difference in Spanish with

respect to agreement in existential constructions notice in section 1.1. I will propose that

the small v in SEC is φ-incomplete (it does not have [person]). In addition, an

interpretable [person] in SII-T will account for the dialectal difference in agreement.

Recall that there are two dialects of Spanish with respect to agreement in existential

constructions:

(33) Hubo dos hombres en la fiesta SI

was two men in the party

There were two men at the party

(34) Hubieron dos hombres en la fiesta SII

were two men in the party

There were two men at the party

30
This difference has been attested by several researchers, and it seems to be present in

Spanish at least from the XIV century (Bello 1847, Cuervo 1881, Kany 1951, Lapesa

1980, Montes 1982, Suñer 1982b, a, Obediente 1984, Bentivoglio 1989, Miles 1990, De

Mello 1991, Navarro Correa 1991, De Mello 1994, Nicita 1997, Domínguez et al. 1998,

Fernández-Soriano and Táboas-Baylin 1999, Díaz-Campos 1999-2000, 2003, Treviño

2003, Rodríguez-Mondoñedo 2006c, d, Saab 2006). In the present times, SI is

predominant in Peninsular Spanish (PS), whereas SII is predominant in Latin American

Spanish (LAS)—see, for instance, Bello (1847:§781), who denounces this agreement as

an “almost universal vice”. But SII is also present in PS as a non standard variety, and SI

is usually imposed as a prescriptive rule in LAS—so, in LAS, it is not hard to find

educated speakers with both SI and SII.24

Given that the social and geographical distribution of these varieties is complex, 25 I will

24
It is necessary to keep this in mind. That a given speaker can accept both versions of existentials
constructions does not mean that the agreement is optional—against some suggestions to the contrary:
Contreras 1976: 142, Westphal 1980: 61, Treviño 2003: 179. At best, speakers who accept both are bi-
dialectal. Being a native speaker of SII myself, I remember my surprise when I was first introduced to SI as
the “correct” dialect. Now, however, I can accept SI haber-sentences. See also footnote 46.
25
The distribution of the phenomenon has received considerable attention in the sociolinguistic literature,
where it is usually called “the personalization of haber” (Silva-Corvalán 2001: 30)—see Montes 1982,
Obediente 1984, Bentivoglio 1989, Navarro Correa 1991, De Mello 1991, 1994, Domínguez et al. 1998,
Díaz-Campos 1999-2000, 2003, among others. Notice, however, that this literature generally considers the
agreement in SII as evidence that the nominal is a subject and no longer an object, but without discussion
of the lack of other subject properties, and no explanation is provided for the presence of the ACC clitic.
Some authors explicitly belittle this fact: “La posible realización de este SN como un pronombre
correferencial no parece ser un argumento que interese a los hablantes” [The possible realization of this NP
as a coreferential pronoun does not seem to be an argument that calls the attention of the speakers]
(Domínguez et al. 1998: 33). These opinions, however, must be considered in their real dimension, since
they show something very important. The presence of agreement in the data is so pervasive that the
researchers are willing to consider the nominal as a subject, putting aside other properties. Then, we can
consider that the sociolinguistic explorations confirm that the agreement between the nominal and haber is
widely spread, despite the fact that the nominal is an object or even an ACC clitic. This should not be so
strange, since agreement with the verb does not need to be a property of subjects, as minimal cross-
linguistic examination immediately shows.

31
simply use SI and SII to refer to the varieties without and with agreement, respectively.

Keep in mind, however, that SI and SII are really shorthands for a family of dialects that

share the corresponding phenomenon. Although I will present some correlations between

the difference in (33)-(34) and other properties of SI and SII, nothing prevents the

dialects inside SI or SII to differ in other aspects of their grammars.

Under the current assumptions, the explanation for SII (34) is straightforward: the

nominal values the φ-features of v (as a result of which the nominal gets ACC), and v

values the φ-features of T. T, then, ends up indirectly agreeing with the nominal (i.e. it

ends up having the φ-features of the nominal, which is responsible for the agreement

patterns). However, (33) shows that in SI T cannot probe its goal, and therefore it gets a

default value. Given that in both cases we have v (because there is ACC: recall that both

dialects accept (1)), the question is what could prevent T from valuing its φ-features by

probing v in SI? To answer this question, let’s return to the restrictions on the internal

nominal in haber-sentences.

1.3.1 Restrictions on the internal nominal and φ-incomplete small v

I contend that nominals that are specified with [person] are not possible inside haber-

sentences. This means that there are nominals that do not have [person] feature, and they

are precisely the ones that are allowed under haber. There is a long standing tradition,

which can be traced back to Benveniste 1966, arguing that the category of Person should

be restricted to the participants in the speech act, that is, the First Person and the Second

32
Person. Under this perspective, the non-participant, the Third Person, is a non-person.26

In a feature-based system, being non-person would mean that the item in question lacks a

[person] feature, that is, it is φ-incomplete. I will adopt the view that some nominals have

a [person] feature, but I will also assume that other nominals lacks this feature. This

means that some nominals are φ-complete, but others are φ-incomplete (they lack

[person]).

I assume the following classification of nominals with respect to [person]—see Harley

and Ritter 2002, Adger and Harbour 2003, Alexiadou and Anagnostopoulou 2004, among

others, for related proposals in relation with other languages:

(35) Nominals

participant non-participant

[person:1] [person:2] [person:3] ----

As in the traditional view, if a nominal refers to a participant in the speech act, it will be

marked with [person]. Notice that these nominals are very few, in fact, they are the First

and Second Person pronouns. Notice further that First and Second Person pronouns are

always interpreted as both specific and animate. That is, only non-participants display all

combinations of semantic properties [±animate] and [±specific]. We then have the

26
This traditional view has been challenged, however. For instance, relying on data from an extensive
cross-linguistic survey, Siewierska 2004 argues that excluding the Third Person “would severely skew our
understanding of a number of facets of the category of person” (Siewierska 2004: 8), although she
acknowledges that the Third Person has a distinct nature from the First Person and the Second Person. See
also Nevins 2007, who voices similar concerns.

33
following:

(36)
[+animate]
participants
[+specific]

[+animate]
[-specific]
non-participants
[-animate]
[+specific]

[-animate]
[-specific]

It is natural to assume that the subset of non-participants that are similar to the

participants can also have the [person] feature that the participants uncontroversially hold.

I thus suggest that, at least in some languages, non-participant nominals that are both

specific and animate receive a [person] feature, slightly departing from the traditional

view.27

A language then may link the possibility of having a [person] feature with the semantic

properties of being animate and specific, given that these properties are always connected

with nominals that necessarily hold the [person] feature (the participants). Then the

language would grant the feature [person] to all nominals that are both animate and

specific, even if they refer to non-participants. That may not be the only possibility.

27
It is not clear, however, how far we are from a Benvenistean position. If we take the idea of “non person”
to be related with the value of [person], and not with the category itself, we could reconcile the two
positions. If First Person is [person:+1-2] and Second Person is [person:-1+2], Third Person could be
[person:-1-2], which leaves room for [person:+1+2], a pronoun that refers to both participants (the inclusive
we), a possibility very well attested among languages. I have no space here to explore all the consequences
of this move. For expository reasons, I will continue to use [3p] to refer to non-participants nominals that
carry [person].

34
Another language could interpret the opposition between participants and non-

participants with respect to [person] only on the basis of animacy, or only on the basis of

specificity. Such languages would grant a [person] feature only to animates (or to a

subset of them, depending on how close they are understood to be with respect to the

participants) or only to specific nominals (or to a subset of them, under the same

conditions)—see chapter 4 for examples of languages that instantiate these possibilities.

The above allows an interpretation of the [person] feature that is uniform across

participants and non participants, but changes across languages (see also the discussion in

chapter 4): in a given language being [person] means being both specific and animate,

and the particular value of the feature informs the corresponding participation in the

speech act (see footnote 27). I also suggest that only nominals that do not receive the

feature [person] will be φ-incomplete, that is, they will have no [person] feature. I

contend that this is what happens in Spanish—for other possibilities see chapter 4. In that

sense, for instance, a proper noun like Juan, which is specific and animate, has an

interpretable [person:-1-2] feature. On the other hand, a nominal like azúcar “sugar”,

which is both non animate and non specific, has no [person] feature.

An interesting situation arises with the ACC clitic lo. This form is allowed under haber-

sentences, as we have already observed:

(37) a. Hay una casa en el parque

is a house in the park

There is house in the room

35
 b. La hay

CLITIC is

In this case, lo is not animate, since it stands for “a house”. We assume, then, that here lo

does not have a [person] feature—and in this case it is not specific, since it is inside an

existential sentence. However, when lo is animate, it must have a [person] feature:28:

(38) Lo vi a Juan

CLITIC saw to John

I saw John

Then, I assume two different forms for lo: one with [person] (as in (38)) and the other

without [person] (as in (37)). Interestingly, in some Spanish dialects, the ACC clitic with

[person] has a morphological expression (le) different from lo without [person]. 29 An

interesting prediction then arises. In leísta dialects (the ones that distinguish lo/le with

respect to the direct object), the form le, that is, the one that has a [person] feature, should

not be possible with haber-sentences. This prediction is borne out:

28
Notice also that here, in a clitic doubling structure, lo must be specific (as argued by Suñer 1988, among
others).
29
The relation between these phenomenon (usually called leísmo) and the Differential Object Marking (the
presence of preposition with some direct objects) that Spanish shows has been noticed long ago. For
instance, Lapesa 1968: 549), quoted by Pensado 1995: fn4, characterized the leísmo as a process that
“tiende a distinguir gramaticalmente las categorías de persona y cosa” [tends to grammatically distinguish
the categories of person and thing]. The dialectal differences with respect to these clitics are more complex,
however. For instance, there are dialects where le substitutes lo completely. See Fernández-Ordoñez
1999for a thorough description.

36
(39) *Le había

Him was

There was him.

Furthermore, in non-leísta dialects, the ones that have a single form (lo), the clitic lo,

cannot have an animate interpretation when used in haber-sentences:

(40) a. *Hay Juan

Is John

There is John

b. *Lo hay [if lo=Juan]

him you

There is him.

This is evidence that Spanish can express morphologically the split between nominals

that are animate and specific, and nominals that are not. I claim that it does so by using

the feature [person]. The exact process and conditions that allow the insertion of the

[person] feature will be spelled-out in chapter 2, since they are mostly relevant for DOM-

marking.30 The important thing here is that [person] nominals are precisely the kind of

30
We will see that it is actually only animacy that controls the insertion of [person], but that this feature
gets deleted if a nominal cannot be specific, so [-specific,+animate] nominals may still end up without
[person]—I will delay the technical details until chapter 2. Notice that this does not affect the current
argument regarding existentials.

37
nominals that are banned from Spanish haber-sentences:31

(41) a. *Me/Te/Nos había

Me/Te/Us was

There was me/you/us

b. *Has tú

are you

There is you.

It is crucial to stress that [person] nominals are not possible here even with list-readings.

It is not the case that in (41) we have a different reading, these sentences are simply

ungrammatical, which is different from English, where the corresponding sentences are

accepted with a list-reading (an observation that can be traced back to Milsark 1974).

Therefore, as explained above, the restriction in question cannot be accounted for by

using the Definiteness Effect (contra Saab 2006). I propose that the reason for the ban in

question is that small v appears without the feature [person], i.e. only with [number]:32

(42) v

[number]

31
Notice that this implies that some specific nominals are allowed under haber (with list-readings), if they
are inanimate, as discussed above. Recall also noting that in dialects that distinguish between lo and le (as
ACC clitic), le is banned from haber-sentences, as predicted.
32
As mentioned above, I am abstracting away from other features like [gender], since these do not play any
role in the system I am developing here. For other analysis where syntactic heads can be underspecified,
see Béjar 2003: 37 and the references therein.

38
This means that only objects that are not specified for [person] will be allowed in the

structure in question. If an internal nominal is specified for [person], small v will be able

to probe it and to value its own [number] feature, but it will not be able to value the [case]

feature of the object, given Chomsky’s 2000, 2001 suggestion that incomplete φ-features

cannot value [case], expressed in a principle to maximize matching effects:

(43) Maximize matching effects

[Chomsky 2001: 15]

This requirement for probes must be relativized to the features of the goal. Small v is still

“incomplete” (in the absolute sense) when it probes objects with no [person] feature, but

it is complete with respect to the features of the goal. In other words, Chomsky’s

suggestion can be interpreted in this way: 33

(44) Condition on Case-valuation

Only a probe P that matches all the relevant features of a goal G can value the

[case] feature of G

So far, this is common to SI and SII. A problem arises here. Recall that I have suggested

above that v serves as a goal for T, valuing its φ-features. Notice however that small v

should not be able to value the φ-features of T because it is φ-incomplete. Here we are

looking at the other side of valuation: from the goal to the probe. This situation is not

33
Notice that I am assuming that T is not able to value the [case] feature of a nominal dominated by vP.
The reason for that is that T needs to probe v and cannot escape its interference, failing to probe across v. In
other words, we are dealing here with a minimality effect, since small v is the closest goal to T.

39
covered by (44). Let me then assume a version of Béjar ’s 2003 Condition on valuation

(where, f=features):

(45) Condition on φ-valuation

G(oal) values P(robe) if and only if all uninterpretable φ-f'(P) match φ-f(G)

[Adapted from Béjar 2003: 65]

In other words, all the uninterpretable features of the probe P must match the φ-features

of the goal G, in order for the uninterpretable features of the probe P to get valued. This

basically means that there are two process of valuation under Agree: Case-valuation

(which must satisfy (44)) and φ-valuation (which must satisfy (45)). In other words, the

Agree operation has three steps:34

(46) Agree
a. Match
P G
[person: ] [person: α]
[number: ] [number: β]
[case: ]
b. φ-valuation
P G
[person: α ] [person: α]
[number: β] [number: β]
[case: ]
c. [case]-valuation
P G
[person: ] [person: α]
[number: ] [number: β]
[case: γ ]

34
This will be particularly important when we discuss default agreement in the next subsection.

40
Therefore, although (42) allows us to explain the [person] restriction on internal nominals

in haber-sentences (given (44)), it gives us a new problem, because now we don't have a

way to value the φ-features of T (given (45)). Let's put this aside for a moment.

There is another difference between SI and SII, in addition to the agreement difference

noted above. In SII SEC, but not in SI SEC, the T can have a person value that is

different from the one in the nominal. In fact, as discussed above, the nominal in SEC

does not have [person] (it is non specific), but T shows [person] inflection:

(47) Habemos dos estudiantes en la clase SII

Habéis

Are-1P-PLU two students in the class

Are-2P-PLU

Lit: We there are two students in the class

The same sentence is not acceptable in SI:

(48) * Habemos dos estudiantes en la clase SI

* Habéis

Are-1P-PLU two students in the class

Are-2P-PLU

It is important to notice that (47) still is an existential sentence. First, it is subject to the

41
Definiteness Effect:

(49) *Habemos los estudiantes en la clase SII

Are-1P-PLU the students in the class

Second, it is still subjectless, in fact, it is impossible to use an overt subject (either

preverbal or postverbal):

(50) a. *Nosotros habemos dos estudiantes en la clase SII

We are-1P-PLU two students in the class

b. *Habemos nosotros dos estudiantes en la clase SII

Are-1P-PLU we two students in the class

This behavior allows us to confirm that (47) is not the Spanish equivalent of any of these

English sentences (where the speaker is a student):35

(51) a. We have two students in the class

b. We are two students in the class

35
In fact, Spanish has also these possibilities, using the verb tener (“to have”) or ser (“to be”)—notice that
here we have an overt subject:
(i) Nosotros tenemos dos estudiantes en la clase
We have two students in the class
(ii) Nosotros somos dos estudiantes en la clase
We are two students in the class

42
(47) is, then, truly an existential sentence. Given that [person] nominals are not allowed

under haber, the question is where do [1p] and [2p] come from? The situation is even

more puzzling if we take into consideration the fact that this dissociation between the

person in T and the person in the nominal agreeing with T is possible with other verbs in

both SI and SII. As discussed in Hurtado 1984, Fernández-Soriano 1989, Contreras 1991,

Olarrea 1996, Ordóñez 1997, Ordóñez and Treviño 1999, Ortega-Santos 2004, Ticio

2004, and others, Spanish (SI and SII) subject nominals can trigger first or second person

in T when they refer to a group that includes the first or the second person:

(52) Los estudiantes asistimos a la clase regularmente

asististeis

The students attended-1P-PLU the class regularly

attended-2P-PLU

We, the students, attended the class regularly

You, the students, attended the class regularly

The question is why SI does not allow this dissociation with haber. This is the same

question regarding why haber in SI does not agree with the internal nominal but SII does.

The standard analysis for (52) is to assume that the overt nominal is left-dislocated and

that there is a pro in [Spec, TP] which is responsible for the features in T (Contreras 1991,

Olarrea 1996, among others):36

36
It is worth noticing that the idea that the subject is in a A-bar position, above [Spec, T], is not limited to
cases like (52). See Barbosa 1995, Hulk and Pollock 2001, Ortega-Santos 2004, among others, for some

43
(53) [Los estudiantes [TP pro-1p asistimos a la clase regularmente ] ]

The students attended-1P-PLU the class regularly

We could try to apply this analysis to haber-sentences in SII, assuming an internal a pro

with [1p] or [2p], which would be responsible for the features of T, with the overt

nominal right dislocated to some position:

(54) [ [ Habemos pro-1P ] dos estudiantes en la clase] SII

Are-1P-PLU two students in the class

But if this is correct, we cannot explain (41): nominals specified for [person] are not

allowed in haber-sentences. It seems, thus, that we do not have anything that could value

the φ-features of T in SII SEC.

I think we can solve all these questions at the same time, by assuming an interpretable

[person] feature in T. I will discuss this proposal in the next subsection.

1.3.2 Interpretable [person] in SII T and a repair strategy for SI T

I propose that in SII the effects just discussed are triggered by an interpretable [person]

feature in T. This will solve all the issues noted above. First, given that the [person]

feature in T is interpretable in SI, we don’t need to value it (it already has a value). So

discussion.

44
only the [number] feature needs to probe small v, which is fine because small v has only

[number], satisfying (45):

(55) a. Match
T v
[person: 1p]
[number: ] [number: plu]

b. φ-valuation
T v
[person: 1p]
[number: PLU] [number: plu]

Second, given that now the [person] feature is interpretable in SII, it can vary freely,

resulting in [1p], [2p] or [3p], without agreeing with the internal nominal (as in (47)).

This has the effect that we do not even need to resort to a pro to explain mismatches

between the φ-features of T and the overt subjects. Third, given that in SI T has

uninterpretable [person] and [number], the small v won’t be able to value the [person]

feature (because small v is incomplete, as discussed). This is actually a welcome result

for SI, since we want to produce a situation where T in this dialect receives a default

value in SI haber-sentences. We then have a situation similar to (31), repeated here as

(56), where T has no element to value its features:37

(56) a. Llueve

rains

It rains

b. [TP T-φ [VP Llueve]

37
Notice that the internal nominal in SI haber-sentences cannot value the features of T, because it has
already checked its [case] feature, hence it is no longer active for Agree; furthermore, it is also incomplete,
as discussed above.

45
(56) clearly shows that SI (and also SII, but this is not applicable to haber-sentences for

the reasons discussed above) has a last resort strategy to repair this situation, which I will

discuss in the next section.

Before doing that, let me close this section by discussing some possible objections to the

solution we have just seen for SII. Saab 2006, discussing a preliminary version of this

chapter, has suggested that the ungrammaticality of (50) is somehow a consequence of

the Overt Pronoun Constraint (OPC), which bans overt pronouns that are in contrastive

distribution with empty ones from acting as a bound pronoun (Montalbetti 1984). It is

difficult to see, however, how the OPC can apply here since the very issue of contrastive

distribution between overt and empty pronoun is at stake here. In other words, the OPC

does not ban overt pronouns per se (not even from being bound), but only bound readings

in overt pronouns when both overt and empty pronouns are syntactically possible

(Montalbetti 1984: 89-94). In addition, if we accept that the OPC shows that pro is

somehow less specific than its overt counterparts, and therefore that the sentences in (50)

are excluded by the Definiteness Effect (as Saab 2006 suggests), then it should be

possible to obtain list-readings with overt pronouns, under Condition (8). This is,

however, not possible:

(57) a. —Hay algún taxista en este pueblo?

Is there any taxi driver in this town?

46
 b. —* Bueno, habemos nosotros

—Well, there-are we

So, we disregard the possibility of having pro in (47).

On the other hand, if the [person] feature of T in SII is already valued, as I have proposed

above, it means that the feature must be associated with each tense morpheme (each

instance of T). Therefore, different tense morphemes could adopt different settings, that

is, a morpheme for a particular tense may have a valued [person] feature, but another

morpheme for a different tense may not. This prediction is also borne out. Díaz-Campos

2003 shows that, for Venezuelan Spanish, the possibility of agreement is sensitive to the

tense morpheme. In Venezuelan Spanish, imperfect and present perfect favor

pluralization but preterit does not (Díaz-Campos 2003: 9). In principle, other SII varieties

could be different. We will discuss in the following sections other consequences

regarding the tendency of haber to regularize.

A possible objection is that the interpretable [person] feature of T in SII only combines

with plural, being impossible in singular:

(58) * Habías un estudiante en la clase

Is-2P-SIN a student in the class

Lit: You there is a student in the class

47
However, it can be argued that this is a result of a semantic restriction. The interpretable

[person] feature in T forces an inclusive relation with the set defined by the internal

nominal. In that sense, the situation is similar to other cases of inclusive relations. For

instance:

(59) We went to the movies, and you bought the pop corn.

A possible reading for (59) is that the hearer (you) is part of the group defined by we.

Obviously, this reading is not available if both pronouns are in singular:38

(60) He went to the movies, and you bought the pop corn.

Although a full-fledged theory of inclusive relations may be necessary (I have no space

here to elaborate it), the intuition is clear enough to suggest that we are on the right track,

and that the restriction of the interpretable [person] feature in T to sentences with plural

internal nominals is not a problem to my solution (contra Saab 2006)—see also Lyons

1999: 313-315 for arguments that the restriction of this phenomenon to pluralities has a

semantic origin.39

38
Another inclusive relation that requires a plurality is partial control (see Landau 2001 for discussion):

(i) John decided PRO to meet at the coffee shop.

In (i), John controls PRO only partially, since the subject of meet must be plural (John is just one of the
people meeting). If this condition is not met, partial control cannot arise:

(ii) John decided PRO to walk home alone.

39
Notice that this problem is not limited to my solution. A pro analysis faces the same challenge (see
Olarrea 1996, Ordóñez 1997). Saab’s 2006 solution is peculiar: he stipulates that the relevant elements here
are “inherently plural”. I could certainly do the same thing without changing anything else in my system,

48
The remaining question is how the unvalued [person] feature in T gets valued in the SI

dialect. As briefly pointed out when (31) was discussed, there are plausible candidates for

this strategy: default agreement or subject clitic. We will discuss each of them now.

1.3.2.1 Default agreement

Recall that T in SI has an uninterpretable [person] feature, which cannot be valued by

probing small v, since the latter is φ-incomplete. A possibility is to value it by default.

The idea of default agreement has been proposed by several authors to deal with

situations like the one just discussed—for a sketch of a theory of default agreement in the

Agree system, see Béjar 2003: 76-80. A problem that always arises is how to restrict it. If

default agreement were allowed to repair all instances of unvalued φ-features we would

predict no crashes in the Agree system (when it comes to φ-features) and, possibly,

multiple instances of disagreement. However, the idea of default agreement is necessary

in the system we are developing here. Minimally, we can define default agreement as the

presence of a feature in a head that cannot be explained by Agree.

Recall that the Condition on φ-valuation (45) restricts φ-valuation to instances where the

Goal is complete with respect to the Probe. In addition, I am assuming that some

nominals (precisely the ones that are allowed under haber) are φ-incomplete (they have

but I think that an exploration of inclusive relations shows some promise to derive this restriction. I have
left that for future research. Saab adopts my general solution for the situation in Spanish Existentials (a
small v that can check ACC), but rejects the claim that it is φ-incomplete, claming instead that the effects of
incompleteness can be derived from the Definiteness Effect, together with the Overt Pronoun Constraint.
As we have just seen above, this cannot be correct.

49
only [number], no [person]). Given the combination of these assumptions, we would

expect that these nominals would not be allowed as subjects of any T in SI, because, by

hypothesis, T is φ-complete in SI but these nominals are not.40 Notice that there is no

problem with the [case]-feature of these nominals, which can be valued, under the

Condition on Case-valuation (44). The problem is the φ-valuation of T. Notice further

that we cannot assume that there is an unconstrained mechanism of default agreement

that repairs this situation. This will predict a default value in T, which does not happen—

these nominals, when possible, agree with T:

(61) Llegaron dos autos

Arrived-3P-PLU two cars

There arrived two cars.

In (61), if there is a small v, it is φ-less and caseless, that is, it does not have φ-features

and it cannot assign any case; consequently, it does not interfere with the probing of the

nominal by T.41 No default agreement (3P-SING) is possible in (61). In addition, in this

sentence we have a situation similar to existential constructions: recall that the φ-

incomplete v cannot value the φ-features of the φ-complete T in existential constructions

in SI:

40
This is not a problem in SII, where T has only one uninterpretable [number], which can be valued by the
interpretable [number] of these nominals, under (45).
41
Alternatively, there could simply be no small v in (61). See, however, Legate 2003 and Deal 2006 for the
idea that unaccusatives do need small v.

50
(62) Existential constructions SI
T v
[person: ]
[number: ] 8 [number: plu]

(63) φ-incomplete subjects

T nominal
[person: ]
[number: PLU ] 9 [number: plu]
[case: NOM ]

Recall that the Condition on Case-valuation (44) does not prevent the [case]-feature of

the nominal from being valued by T (since T does match all the φ-features of the

nominal). On the other hand, if we do not allow the nominal to value the [number] feature

of T in (63) we will create a mismatch between Case and Agreement: the Goal will get a

[case]-value, even if it is not able to value any φ-feature of the Probe. We need to allow

sentences like (61), where the subject is φ-incomplete but it still agrees and checks case

against a φ-complete T. Obviously, we need to allow φ-valuation in this case. I propose

the following condition:42

(64) If a Goal G receives a [case]-value from a Probe P, it must value all the

uninterpretable φ-features that the P is matching.

This allows the valuation of the [number]-feature in T, and leaves the [person]-feature

unvalued. Then, we can propose a mechanism of default value that may repair heads that

have incomplete valuation, in other words, if one feature needs to be valued, because of

42
Notice that this situation suggests that case-valuation happens under Match, although it produces φ-
valuation.

51
(64), the remaining unvalued features will receive default agreement—see Béjar and

Rezac 2003: 54 for a similar claim. In (61) and (63), this means that T will receive

default [3p]. This is fine, because [1p] and [2p] nominals are always φ-complete, so

default agreement will be unnecessary in those cases. Notice further that the above does

not affect our discussion of small v and the ban on [person]-nominals in haber-sentences;

in this situation, the problem was Condition (44), that is, the [case]-value of these

nominals. The Condition on φ-valuation (45) is satisfied there.

Existential constructions, that is (62), cannot fall under Condition (64), since here the

Condition of φ-valuation (45) applies, because there is no [case]-valuation involved. This

means that small v cannot value the [number]-feature of T in (62), under (45), which

corresponds to the situation in SI SEC (recall that the [person] feature of T in SI is

uninterpretable). Therefore the mechanism of default agreement must be even richer than

suggested, being able to repair heads that are totally unvalued, as long as they do not

value the [case] of any goal. A general characterization may then be that default

agreement applies to φ-features that are not involved in [case]-valuation,43 if they fail to

meet Condition (45). This is what happens in the situation illustrated in (62), which

corresponds to SEC in SI. As stated, (45) could be in conflict with (64), however. I will

therefore modify (45) as follows (where f=features):

43
Notice that even in (63), which corresponds to (61), the feature that receives a default value, that is,
[person], is not involved in [case]-valuation, since it does not match anything in the nominal.

52
(65) Condition on φ-valuation (modified)

If a Goal G does not receive a [case]-value from a Probe P, G values P if and

only if all uninterpretable φ-f'(P) match φ-f(G)

The modified version makes (65) the caseless counterpart of (64).

1.3.2.2 Subject clitics

In this section, I will consider one implementation of the default agreement mechanism,

which is intended to be the sole means of resolving situations where default agreement in

T is called for (i.e. no other mechanism for giving default agreement to φ-features of T

will be necessary under the analysis about to be presented).44 In particular, I will consider

the possibility that a subject clitic values the φ-features of T in both (56) and haber-

sentences in SI. This possibility would be the morphological instantiation of the default

agreement mechanism. The morphological manifestation of this clitic is a [3p, SING]

suffix, a solution that is in line with several analysis for subject agreement in null subjects

languages—see Alexiadou and Anagnostopoulou 1998, Kato 1999, 2000, Díaz 2004,

Ticio 2004, among others. Following this tradition I will call this CL a subject clitic

(SCL), although it will be attached to a subjectless T. We can see the process here:

44
The discussion below is somewhat tentative. If the subject clitic analysis to be developed is not adopted,
we can still fall back to the more traditional default agreement mechanism without affecting the rest of the
analysis developed in this chapter.

53
(66) a. Había dos autos en el garage SI

was two cars in the parking lot

There were two cars in the parking lot

b’. [TP T-φ [vP v-φ [VP había [SC dos autos en el garage] ]]]

Agree/Case

b’’. [TP T-SCL [vP v-φ [VP había [SC dos autos en el garage] ]]]

Agree/Case

In (66), the agreement takes place as explained before. What is different is that T receives

a SCL with a specified value: [3p, sing] (let me call it invariable SCL), which surfaces as

a verbal suffix. As a result, T values its φ-features. In other words, since in SI T cannot

value its φ-features by probing v (i.e. probing v would not result in valuation of the φ-

features of T in SI given (65)), it does so by using the SCL, a morphological resource. If

this is correct, the difference between SI and SII can be stated as the difference between

the mechanisms to value the φ-features of T. In SII, where the [person] feature of T is

interpretable, Agree between T and v is possible, whereas SI, where the [person] feature

of T is uninterpretable, must resort to a SCL, because Agree won’t be able to value the φ-

features of T, given that small v is φ-incomplete.

Let me explain the distribution of the invariable SCL. We don’t want this SCL to be

freely attached to any verb in SI. This would predict that any sentence could have [3p,

SING], which is contrary to the facts. On the other hand, we do have instances of invariant

SCL in SII too, as showed by (56). Let me assume that SCL must appear only if T has no

54
other way to value its φ-features. To be more precise, a SCL must be used in SEC with

haber-sentences (in SI), otherwise, the derivation will crash, for the reasons discussed

above. It is worth noticing that the verb in (56), i.e., llover (to rain), allows a nominal to

trigger agreement with T in Spanish, under certain circumstances in both dialects; in this

case, the SCL is avoided, since T can establish a relation with a nominal, which gets NOM

(no accusative cliticization is possible in this sentence):

(67) a Llovieron insultos sobre el árbitro

rained-PLU insults over the referee

Lit: “Insults rained over the referee”

(to mean something like “The people insulted the referee a lot”)

b. * Los llovieron

CL rained-PLU

In existential constructions, an additional φ-head is present (the small v). Since, as

discussed, v has already valued its φ-features, in SII, T (which has interpretable [person]

but uninterpretable [number]) can probe v, valuing its uninterpretable [number]. However,

SI requires the presence of SCL because here T also has an uninterpretable [person],

which cannot be valued by small v, because small v has no [person], as discussed above.

If we further assume that this SCL is φ-complete, with a value of [3p, SING], we explain

why the [number] feature of T is SING. As just mentioned, if φ-v can value the φ-features

of T, no SCL will be necessary. This is precisely what happens in SII. Notice that this

55
amounts to saying that the verbal suffixes in existential haber-sentences in SII are not the

manifestation of some SCL, but the morphological counterpart of the φ-valuation. There

is some evidence that this is indeed the case, which I will discuss in the next section.

We may push the SCL analysis tentatively by taking into consideration the claim made

by several researchers (Alexiadou and Anagnostopoulou 1998, Kato 1999, 2000, Díaz

2004, Ticio 2004), that Spanish also has SCLs that license null subjects in non existential

sentences. If we add to this picture the suggestion (made by Manzini and Savoia 2002)

that the inflection (or the SCLs) is also able to receive the external θ role, and that,

therefore, there is no pro, we have an interesting parallelism between both forms of

SCL.45 The consequences of these suggestions are beyond the goals of this dissertation. It

is important to keep in mind that my analysis does not depend on them. If the suggestions

regarding SCL are ultimately proven wrong, that is, if we need a mechanism of default

agreement different from SCLs, the system developed here will still work.

So far, we have seen that, for SII, Agree is enough to explain the situation in SEC.

Provided that small v has only uninterpretable [number] and that T has interpretable

[person] in addition to uninterpretable [number], T can probe small v, valuing its [number]

feature (the only one that needs to be valued). With this mechanism, I have explained the

ban on [person] nominals, the agreement with the verb, and the ACC Case. There is no

need for any default agreement mechanism in SII SEC. I have proposed that in SI T has a

complete set of uninterpretable φ-features, which is the only difference with respect to

45
Notice that the invariable SCL for SI SEC will be the only one (with [3p, SING]), but for non existential
sentences we need a whole set of SCLs. This is not a disadvantage with respect to the pro analysis, since, if
we use pro to value the φ-features of T, we will also need a whole set of different pros.

56
SII. Given this, under Condition (65), the φ-features of T cannot be valued by probing

small v (although everything else remains the same); accordingly, T does not agree with

the nominal in SI, but always surfaces with [3p, SING]. Since Agree is not responsible for

this value, an additional mechanism is necessary. We have seen that, when φ-features are

not involved in the [case]-valuation of a goal and, in addition, Condition (65) is not met,

the φ-features must get a default value.

1.3.2.3 T in SII non existential constructions

A final question I want to address here is what happens with T in non existential

constructions in SII. There are two possibilities. It has a [person] feature that is

interpretable (as in SEC) or not. Suppose that it has not, that is, that T in SII non

existential constructions has both uninterpretable [person] and [number]. The first

question to address is why we cannot use this T, which has complete uninterpretable φ-

features, in SEC too. If we could, we would predict the agreement in SEC to be optional

in SII. The T with all uninterpretable φ-features will give us a default value (as in SI), and

the T with interpretable [person] will give us agreement with the nominal. But we have

already seen that this is unlikely (see footnote 24). If we say that the T with two

uninterpretable features is banned from SII SEC because there will be no way to value its

[person] feature (given that v has no [person]), under the assumption that there is no

default agreement or SCL in SII, we will face the problem that one of these mechanisms

must be present in SII too (to explain (56)). I conclude, then, that SII has a unique T with

57
an interpretable [person] and an uninterpretable [number]. 46

Given that in non existential constructions there is a subject and there is no restriction

regarding the [person] value of the subject, we could ask how the [case] feature of the

subject is valued. T in SII has a full set of φ-features; it just happens that one of them is

interpretable. Therefore, it will be able to match all the relevant features of the goal,

valuing its case.47 This is different from small v, which is truly incomplete, even with

respect to a potential goal with full φ-features, since it lacks [person]. This allows me to

retain the conclusion that there is a unique T for all sentences in SII.

1.3.2.4 The tendency of haber to regularize in SII

I will suggest in this subsection that the process of valuing the φ-features of T with an

invariable SCL or with default agreement (in SI) may leave a morphological form

different from the one that results from valuing by Agree (in SII). Remember that in

Spanish (both dialects) it is possible to have a mismatch between the overt subject and

the features in T:

46
It is worth noticing that, if the agreement proves to be actually optional, the idea of two Ts, one with
interpretable [person] and the other without it, will be a viable solution. If the agreement is optional in SII,
but ungrammatical in SI, this will mean that the only difference between SI and SII is that, in addition to
the normal T with full uninterpretable φ-features, SII has a T with an interpretable [person] feature.
47
This means that Match is an operation that involves the category as a whole, not feature by feature—but
the φ-valuation itself can be feature by feature. Here it is worth mentioning that there are proposals that the
so called EPP property can be satisfied only under Match, without valuation (Boeckx 2003). Given that
under the Inverse Case Filter Hypothesis (Bošković 1997b, 2002a), the EPP property and the necessity of
checking Nominative are the same (to be more precise, the EPP in T is deduced from the later requirement),
we can extend Boeckx’s idea by saying that Match is enough to value Nominative. This is not the place to
explore all the consequences of this suggestion.

58
(68) Los estudiantes hemos asistido a la clase regularmente

habéis asistido

The students have-1P-PLU attended the class regularly

have-2P-PLU attended

We, the students, have attended the class regularly

You, the students, have attended the class regularly

For SII (but not for SI, as discussed), this situation can be replicated with existential

haber-constructions. For illustration, I repeat here (47) as (69):

(69) Habemos dos estudiantes en la clase SII

Habéis

Are-1P-PLU two students in the class

Are-2P-PLU

Lit: We there are two students in the class

Surprisingly, in the first person plural, the morphological form of the verb is habemos

and not the expected hemos (compare (68) with (69))—see Fernández-Soriano and

Táboas-Baylin 1999: 1758 for discussion of these data. Interestingly, the form habemos is

the regular form, whereas hemos is the normal irregular form that is used in any other

occurrence of the verb haber in first person plural (for instance, as an auxiliary in (68)).

In other words, in (69) the verb haber has been regularized (the irregular form hemos is

ungrammatical in these constructions). I suggest that this very strange fact is a

59
consequence of the already mentioned distinction between the strategy of valuing the φ-

features of T by using default agreement or an SCL, which gives us the form hemos, or

the strategy of using Agree, which gives us the form habemos.

In (68), the presence of SCLs could explain the apparent disagreement; in fact, dialects

that do not have these SCLs—like Puerto Rican Spanish—do not show this phenomenon

(see Ticio 2004 for an analysis in this direction). 48 We have already discussed what

happens in (68) in section 1.3.1. The overt subject is in the left periphery. The

morphological merging of a [1P, PLU] SCL with haber renders the irregular form hemos.

As seen, in (69), which is possible only in SII, the operation Agree values the φ-features

of T. T can get valued only by probing v. An invariable SCL (or default agreement) is not

necessary in SII SEC. Since this time no relevant morphological operation will occur

(because there is no SCL or default agreement), but just the pure morphological

interpretation of T’s φ-values set by Agree, haber surfaces in its regular form.49 This

explanation, however, is not totally conclusive, because the regularization does not affect

the whole paradigm.

Contrary to what happens in the present tense discussed in (69), in the preterite tense, SII

haber-sentences do not regularize the verb:50

48
This means that we would not need pro to account for the disagreement: SCLs will suffice.
49
I must say that I don’t claim that all regular/irregular differences among verbs should be analyzed like
this. SCLs are perfectly compatible with regular verbs (in fact, in (68) the second person is regular). I
suggest that what causes the regularization is that this is a sort of “new process” (the probing of v by T),
which forces the morphological component to reanalyze the inflection.
50
In order to be regular, the verbal forms of haber should be habimos (1p-PLU) and habieron (3p-PLU),
using the unmodified root hab-.

60
(70) Hubimos dos estudiantes en la clase

Hubieron

Were-1p-PLU two students in the class

Were-3p-PLU

It is worth noticing, however, that some dialects of SII—remember that SI and SII are

just shorthands for different families of dialects—show another case of regularization. In

the two dialects the present form of haber (hay) does not show agreement morphology.51

Fernández-Soriano 1999: 133) suggests that this is because the morpheme –y (a remnant

from an old locative clitic) blocks the presence of agreeing morphemes. This locative

clitic can be found in French (y) and Catalan (hi), and it could be related to Italian

expletive ci (Fernández-Soriano 1999: 133). In these languages, the clitic is productive,

but hay is the only form that has this element in Spanish. That means that this is a

morphological fossil, with no role in the construction, except blocking the morphological

presence of the agreement. The exceptional form hay has these morphological

components (roughly speaking):

(71) a. SINGULAR hay

ha -PRESENT-3PERSON-SINGULAR -MORPHOLOGICAL FOSSIL

ha ø y

51
I am going to call hay an exceptional form, as oppose to normal. The familiar regular/irregular
distinction does not fit so well here, since some normal forms of haber are also irregular. The form hay can
only appear in the [3p] present tense (singular or plural) of existential constructions; it’s an exceptional
form of haber. The normal form in third person singular present should be ha (as it appears in the auxiliary
position of all verbs). As auxiliary or in the imperfect tense in SEC, however, haber is regular (and not just
normal) in both dialects.

61
 b. PLURAL hay

ha -PRESENT-3PERSON-PLURAL -MORPHOLOGICAL FOSSIL

ha ø y

Note that the normal forms for third person present with haber are the following (see also

footnote 51):

(72) a. SINGULAR ha

ha -PRESENT-3PERSON-SINGULAR

ha ø

b. PLURAL han

ha -PRESENT-3PERSON-PLURAL

ha n

Notice, however, that strictly speaking, these are not regular forms (but they are normal

in the sense that they do not receive -y), since true regular forms should be formed by

using an unmodified root (in this case hab-) and the present suffix; since the [1p] present

suffix is a zero morpheme, a true regular form should surface as habe, with the theme

vowel -e, following the model of comer (“to eat”), come (“I eat”), which is a true regular

verb. However, in some dialects of SII, the exceptional form hay [áj] becomes haen [áen]

or hayn [ájan], that is, it allows the plural -n to be suffixed to the verb, unblocking the

62
effect of -y, as reported by Kany 1951: 257 for rural Argentinean, Lapesa 1980: § 133 for

substandard Venezuelan, Montes 1982: 384 for Colombian Antioqueño—see also

Fernández-Soriano and Táboas-Baylin 1999: 1758. As in the case of hemos > habemos, I

want to suggest that this process of normalization is triggered because T is valuing its φ-

features by probing v, without intervention of any SCL or default agreement process. If

this line of reasoning is correct, we have evidence here that T indeed probes v in SII, and

that an invariable SCL is responsible for the lack of agreement in SI. However, since, as

noted above, this regularization does not affect the whole paradigm, this conclusion

should be taken with some reservation.

1.4 Temporal existential constructions

In this subsection I will present data from the so called “temporal existential

constructions” (TEC) (Rigau 2001), which seem to have a behavior similar to SEC, but

also present some interesting differences.52 Some verbs that indicate elapsed time can

also combine with a nominal, in particular, hacer (lit. “to make”) and ir (para) (lit. “to go

(for)”):

(73) Hace un mes que Juan murió

makes a month that John died

It makes a month since John died

52
TEC and SEC are similar in that they are both subjectless: the internal nominal does not enter into an
Agree relation with T—see Fernández-Soriano and Táboas-Baylin 1999 and Rigau 2001 for some
discussion.

63
(74) Va (para) un mes que Juan murió

Goes (for) a month that John died

A month has gone since John died

As expected, in SI no agreement surfaces between the nominal and the verb, which is an

indication that a SCL is present here also. In some dialects of SII (in particular in some

SII substandard varieties), the agreement is possible:

(75) Hacen dos meses que Juan murió SII (substandard)

make two months that John died

It makes two months since John died

(76) Van (para) dos meses que Juan murió SII (standard)

Go (for) two months that John died

Two months have gone since John died

There are, however, two types of temporal-existential verbs, as reported by Fernández-

Soriano and Táboas-Baylin 1999: 1748-1751, among others. The first one—TEC1,

exemplified in (73)-(76)—takes a nominal that expresses time with a proposition that

expresses an event. The second one (TEC2) is a sort of temporal adjunct that behaves like

a preposition:

64
(77) Juan vino hace dos años

John came makes two years

John came two years ago

Only TEC1 can trigger agreement in some dialects of SII. TEC2 never triggers agreement

in any dialect:

(78) *Llegó hacen dos horas SI and SII

came make-PLU two hours

He came two hours ago

[Fernández-Soriano and Táboas-Baylin 1999: 1750]

Only TEC1 allows cliticization of the object (at least for SI):

(79) a. Hace veinte años que nos casamos SI

makes twenty years that we married

It makes twenty years since we married

b. Los hace

CL-ACC makes

[Fernández-Soriano and Táboas-Baylin 1999: 1750]

This is not possible with TEC2:

65
(80) a. Nos vimos hace un mes

Us saw makes a month

We saw each other one month ago.

b. *Nos vimos lo hace

Us saw CL-ACC makes

Fernández-Soriano and Táboas-Baylin 1999: 1750 interpret this contrast as evidence that

in TEC2 the form hace is no longer a verb, but maybe a preposition (see also Rigau 2001).

However, this is not so clear, since the form allows some inflection, for instance, simple

future tense morphology:

(81) Nos vimos hará un mes

Us saw make-FUT a month

We saw each other probably one month ago.

This inflection is, however, severely limited in TEC2. No past or perfect tense is possible:

(82) a. *Nos vimos habrá hecho un mes

Us saw have-FUT made(PARTICIPLE) a month

We saw each other probably one month ago.

66
 b. *Nos vimos hizo un mes

Us saw made a month

We saw each other one month ago.

These possibilities are available for TCE1.

(83) a. El lunes habrá hecho un mes que Juan murió

The Monday have-FUT made a month that John died

On next Monday, it will be a month since John died

b. El lunes hizo un mes que Juan murió

The Monday made a month that John died

On past Monday, it was a month since John died

Let’s accept the suggestion that hace in TEC2 is a preposition. Putting aside the issue of

how a verb is transformed into a preposition, the presence of inflectional morphology in

TEC2, although limited, could be interpreted as the presence of valued features in the

forms hace or hará, that is, there is no uninterpretable φ-features, and the forms are not

able to check structural Case. This could be a problem for its internal nominal, since there

will be no φ-head to value its [case] feature. However, the nominal is an argument of

hace, and it receives a thematic interpretation from it (temporal, we could say). So it

meets the requirement for inherent Case assignment. Let’s assume that this is what

happens in TEC2.

67
On the other hand, TEC1 behaves like haber-sentences, except for the restrictions on its

nominal (it must be a nominal compatible with the thematic interpretation that the verb

assigns). It allows cliticization of its object, and agrees with it, in some dialects of SII.

Then, I propose that there is also a small v in this kind of TEC, and that in the relevant

dialects, T also probes v. Given that the same mechanism can apply here, this reinforces

that idea that a small v with φ-features, which is probed by T, is responsible for Case and

agreement in haber-sentences. In addition, for SI dialects which do not allow agreement

in TEC, we can also adopt the proposal that there is an SCL or a process of default

agreement, as explained.

An interesting question arises here with respect to the SII dialects that do not allow

agreement in TEC 1 (these are in fact the Standard SII dialects). The question is why, if

the probing of v by T is available in SII, these dialects do not make use of it in TEC 1.

We must take into consideration the fact that SCL/default agreement is also available in

SII. The presence or absence of SCL could depend on an idiosyncratic option, according

to the verb. In fact, there is some evidence that this is indeed the case. For instance, in

Standard Peruvian Spanish (which is typically SII), there are two different TEC verbs that

behave differently with respect to agreement. The verb hacer (lit. “to make”) does not

allow agreement in this dialect, but the verb ir (lit. “to go”) does:

68
(84) *Hacen dos años que Juan murió Standard Peruvian

make-PLU two years that John died Spanish

It makes two years since John died

(85) Van ya dos años que Juan murió Standard Peruvian

go-PLU already two years that John died Spanish

Two years have gone since John died

The question is what could be this idiosyncratic characteristic of the verb that prevents

the agreement here. We can make use of the inherent Case property of TEC2. The

thematic situation is the same in TEC1, and inherent Case is indeed an idiosyncratic

property. So, for some dialects of SII, the TEC1 verb hacer assigns inherent Case to its

internal argument. Under the assumption that inherent Case deactivates the φ-features of

the nominal (Chomsky 2000: fn 88, Stepanov 2002), T cannot value its φ-features using

Agree, thus an SCL or a mechanism of default agreement must be used, as in SI haber-

sentences. Other verbs of the same dialect (for instance (85)) do not have inherent Case to

assign, so they allow agreement.

Another interesting issue arises here. Given that the form hace in some SII dialects can

assign inherent Case, and given that nominals with inherent Case cannot feed Agree, we

predict that small v is not possible there, since it would not be able to value its φ-

features—or alternatively, that it receives default values, and that ACC is not the

69
morphological realization of inherent case in Spanish.53 Under the assumption that ACC

clitics are possible only under Agree with the small v,54 no clitics should be allowed for

these constructions. The initial data seem to confirm this prediction. SII Speakers that

cannot accept agreement in hace-sentences also have resistance to accept the accusative

clitic:

(86) a. Hace dos horas que espero Some SII dialects

Makes two hours that I wait

It makes two hours since I’m waiting

b. ?? Las hace

CL-FEM-PLU makes

The same speakers accept the clitic and agreement with other TEC1 verbs:

(87) a. Van dos horas que espero Some SII dialects

Go two hours that I wait

Two hours have gone since I’m waiting.

b. Las van

CL-FEM-PLU makes

53
Notice that this does not change the fact that these sentences are subjectless. The internal nominal does
not receive structural case, that is, it does not enter into an Agree relation with T either.
54
In fact, it has been proposed (see, for example, Franco 1993) that ACC and DAT clitics are the φ-features
in the head responsible for case-valuation (small v, in our case). See Zagona 2002: 184-194 and the
references therein for discussion on this topic with respect to Spanish clitics.

70
The situation is not so clear, however. The judgments are not as steady as we would

expect. But there are interfering factors. First, the sequence Las hace is actually

ambiguous; it can also mean “S/he makes them”, and under this reading it is a perfectly

normal sentence—notice that no ambiguity arises in haber-sentences or with other cases

like (87). In addition, the agreeing hace is identified as substandard, in other words, it is

heavily stigmatized; so it is not implausible to suggest that the speakers are trying to

suppress the agreement, but fail to suppress the clitic.

A more rigorous data collection is needed before we can draw firm conclusions regarding

hace-sentences, in order to properly classify the speakers of relevant dialects and identify

the interfering factors. I will leave this for future research. However, the discussion of

TEC shows a possible scenario where we can observe two mechanisms to value the φ-

features of T, as in haber-sentences: (i) by probing v, and (ii) by default agreement or the

insertion of a SCL. Given the idiosyncratic differences among TEC verbs with respect to

the possibility of inherent Case assignment, it is possible that, even inside SII, sometimes

the agreement gets suppressed together with the cliticization. In fact, it has been reported

that the same happens in some dialects with at least a subset of haber-sentences. We will

discuss this in the next subsection.

71
1.5 More dialectal differences

Treviño 2003 reports a dialect that, at least for a subset of haber-sentences (the locative

ones, that is, the ones that can be paraphrased by estar), does not allow either agreement

or cliticization of the internal nominal.55 According to Treviño 2003, there are three types

of haber-sentences in Spanish: locational (paraphrased by estar “to be (locative)”),

unbounded (paraphrased by existir “to exist”) and contingent (paraphrased by tener “to

have (possessive)”):56

(88) Hay una víbora detrás del sofá locational

Is a snake behind the sofa

There is a snake behind the sofa

PARAPHRASE: Detrás del sofa está/*existe/*tiene una víbora

Behind the sofa is/*exist/*has a snake

(89) Hay tréboles de cuatro hojas unbounded

Be-plu clovers of four leafs

There are clovers of four leafs

PARAPHRASE: Existen/*están/*tienen tréboles de cuatro hojas

Exist/*is/*has clovers of four leafs

55
Unfortunately, the author does not provide any external characterization of this dialect. Actually, she
seems to imply that this behavior is common to all Spanish, or at least all SII. This is, however, not true, as
the discussion until now has shown.
56
For comparable distinctions in English, see Hornstein et al. 2000 and Felber 2002.

72
(90) Hay monarquía en Suecia contingent

Is monarchy in Sweden

There is a monarchy in Sweden

PARAPHRASE: Suecia tiene/*existe/*está monarquía

Sweden has/*exist/*is monarchy

[Treviño 2003: 179-180]

In all SII dialects, included the one reported by Treviño (let me call it SII-T), unbounded

and contingent haber-sentences can replace its object with an ACC clitic and they show

agreement, as discussed. In SII-T, however, a locational haber-sentence “bars or strongly

disfavors verb-agreement and object-clitics” (Treviño 2003: 180). This is a behavior that

resembles the one we just saw with TEC1 in some dialects. If this is correct, the existence

of SII-T may represent evidence in favor of the system I am developing here. That is, we

can suggest that inherent Case is involved here too (actually Treviño suggests Partitive

Case as a possible explanation). We will address this issue later; first let me raise a note

of caution.

Although Treviño claims to “have gathered a considerable body of data” (2003: 178), she

actually presents few examples of agreement violation in locational haber-sentences. She

also warns us that in general, the agreement facts in her data “are quite complex and not

as steady in their behavior, as far as native speakers’ performance and judgments are

concerned” (2003: 184), 57 but she nevertheless thinks that, with respect to locational

57
Interestingly, this is similar to the situation with hace-sentences in the previous subsection.

73
haber-sentences, “agreement on the verb is felt alien” (2003: 184) by the speakers of SII-

T. With respect to the clitic, it could be the case that some interfering factors are

conspiring to obtain the same result. For instance, in both SI and SII, the cliticization of

bare nominals renders some differences in the judgments:

(91) a. ¿Quieres helado?

Want ice cream

Do you want ice cream?

b. % Lo quiero

CL want

I want it

c. (Sí) quiero

yes want

Yes, I want

The same is true for existential constructions:58

(92) a. Hay helado?

is ice cream-MAS

Is there ice cream?

58
Notice that this is a case of “contingent” haber, so the effects of SII-T do not apply here. Besides, the
variability in the judgments seems to be cross-dialectal.

74
 b. % Lo hay

CL is

c. (Sí) hay

yes is

Yes, there is

Given that we are dealing here with a general property of Direct Objects for all verbs

(and not only for existential constructions), in this case it is difficult to assume that there

is no small v if the clitic is not licensed (in other words, we do have structural Case here).

A better alternative will be that the clitic needs something additional to be properly

licensed; it could be a high degree of specificity, as has been proposed in the literature

(for instance, see Suñer 1988), or topicality (Leonetti 2003, in press-b, a). In any case,

there may be abstract agreement between v and the object, even when the latter is a bare

noun. In fact, the object in haber-sentences cannot be doubled by a clitic:

(93) a. Hay un niño en el parque

Is a kid in the park

There is a kid in the park

b. * Lo hay un niño en el parque

CL was a kid in the park

There were some kids in the park

75
This is compatible with the fact that that doubling structures usually need higher

specificity and that objects in haber-sentences have a low degree of specificity. This is

relevant because several of Treviño’s examples are with bare objects. However, she also

shows examples of contingent and unbounded haber-sentences with bare nominals that

do allow cliticization, which represents a contrast with locational haber-sentences. Then,

there is no reason to doubt the existence of SII-T.

In addition, there is another difference with respect to cliticization between locational

haber-sentences and other types, which is not directly addressed by Treviño 2003. More

importantly, this seems to hold for both SI and SII. Locational existential sentences do

not allow cliticization of the nominal alone, that is, they cannot cliticize only the object,

stranding the locative phrase—as observed by Díaz 2004: 36:

(94) a. Hay una niña en el parque locational

Is a kid-FEM in the park

There is a kid in the park

b. ?? La hay en el parque

CL-FEM was in the park

There were some kids in the park

c. La hay

CL-FEM is

76
This is not the case for other haber-sentences:

(95) a. Hay monarquía en Suecia contingent

Is monarchy-FEM in Sweden

There is monarchy in Sweden

b. La hay en Suecia

CL-FEM is in Sweden

[Treviño 2003: 180]

At first sight, this is unexpected in the current system. Why should there be a difference

with respect to cliticization? What is the source of the contrast between (94b) and (95b)?

However, this could turn out to be evidence for the presence of small v in these

constructions, under the assumption that small v is needed to license the internal small

clause in these sentences; I will return to this in chapter 3. These are still several rather

significant questions that we have not yet addressed, which are deeper and more

important: Why is there a small v in existential constructions? What is the source of its

presence? I will deal with these issues in chapter 3.

Now let me address the question regarding how the idea of inherent Case assignment

could help us to explain the lack of agreement, as previously suggested. The only

problematic case is locational haber-sentences in SII-T. Here, the agreement disappears

together with cliticization. The others cases are normal instances of SII SEC. At first

glance, the inherent Case assignment is problematic because there is no obvious thematic

77
relation between haber and its internal nominal. In fact, the internal nominal seems to be

in a thematic relation with the PP. It is standardly assumed that existential constructions

include a Small Clause SC (Stowell 1981, 1983, Safir 1985). Let’s accept Stowell 1983’s

idea that SC is an XP with the nominal in its Specifier and the predicative phrase in its

Complement (for discussion of SC structure, see the papers in Cardinaletti and Guasti

1995). If so, a SC is identical to a vP structure (in fact, Koopman and Sportiche 1988’s

formulation of the Internal Subject Hypothesis assumes that the subject forms a Small

Clause with the VP). For locational haber-sentences, the predicate will be a Prepositional

Phrase (PP):

(96) XP

(nominal) X’

X PP

Notice that the nominal is the external argument of the PP, and then, it will receive a θ-

role (the locatum as opposed to the location, which will be the internal argument of PP).

If we assume that X is involved in the assignment of the external θ role—naturally

extending standard assumptions regarding vP—we have a potential inherent Case

assigner.59

59
Some issues arise with respect to symmetry with other inherent case configurations. Here the assigner
does not c-command the assignee. If we accept that c-command by the assigner is necessary for inherent
case assignment, we could still implement the idea just described by incorporating X into the verb haber.
This will create a c-command relation. In fact, this gives us another possibility to implement the assignment
of inherent case. If X incorporates into haber, it could give the ability to assign inherent case to the
resulting element haber+X; in this case, haber+X will be the inherent case assigner, under standard c-
command. I will leave this issue open. It is worth noting, however, that Lasnik 1999b: 86-87, following
Saito and Hoshi 1994’s analysis of Japanese light verb constructions, makes a similar suggestion.

78
Since inherent Case is a lexical property, it could be an idiosyncratic property of X in SII-

T—it does not need to be a general property of X in all of Spanish. Once the

corresponding nominal has inherent Case, it will be frozen for Agree, as discussed above.

This means that it cannot value φ-features of any head, T or v. Then, neither agreement

nor clitics will be possible, as discussed above. Given that inherent Case blocks Agree,

one could ask if in all instances of SI haber-sentences (with a default value) we have

inherent Case. However, we cannot use the idea that inherent Case blocks Agree to

explain the differences between SI and SII with respect to agreement discussed above. If

a nominal receives inherent Case, we predict default values in T and no clitics, but

although T in SI existential constructions has a default value, such constructions do allow

clitics, as noted above.

1.6 A prediction with respect to raising verbs

Another prediction is made, given the idea that T probes small v. If a defective T (in

Chomsky’s 2000 sense; that is, typical raising infinitival T) is on top of the small v in

haber-sentences, the φ-features of a higher non-defective T can be valued by v in SII but

not in SI. This prediction is borne out. The nominal agrees with the finite verb in SII (this

sentence is ungrammatical in SI):

According to Lasnik, in English, the predicate of the small clause in existential constructions raises to V0
(and then to AgrO0) to assign inherent case (partitive case) to the internal nominal.

79
 (97) Parecen haber dos hombres en el jardín SII

seem-PLU to be two men in the garden

There seem to be two men in the garden

On the other hand, in SI, T is not valued by Agree, as discussed, and then it receives [3p,

SING]:

(98) Parece haber dos hombres en el jardín SI

seems-SING to be two men in the garden

There seem to be two men in the garden

This is illustrated here (using a SCL for SI):60

(99) SII
[ T-φ Parecen [TP T-def [vP v-φ [VP haber dos hombres en el jardín ]]]]

Agree Agree

(100) SI
[ T-SCL Parece [TP T-def [vP v-φ [VP haber dos hombres en el jardín ]]]]

Agree

This contrast between the two dialects—noticed by Suñer 1982a: 100-107 but left

unexplained—is straightforwardly accounted for under the assumption that T can probe

60
Notice that I am abstracting away from the role of T-def (defective T) here. There are two alternatives.
On one hand, it could be that T-def is defective because it has no φ-features (at least in Spanish); in this
case, it is completely outside of the Agree system. On the other hand, it could be that it is defective in the
sense that its φ-features are not able to value Case (after all, the φ-features do surface in some languages,
like Brazilian Portuguese), but still are able to participate in Agree; if the later is true, in (99), T-def could
probe v-φ and then serve as a goal for the higher T-φ. I will leave this issue open.

80
small v. Notice, in addition, that, as in matrix haber-sentences, agreement is maintained

with clitics in SII; on the other hand, SI also allows ACC clitic, but no agreement surfaces,

as predicted: 61

(101) Parecen haber-los SII

seem-PLU to be-CL-ACC-PLU

(102) Parece haber-los SI

seem-SING to be-CL-ACC-PLU

Before turning to the conclusions, let me briefly comment on a hypothesis that has been

proposed with respect to Case in SEC. Demonte 1989: 165-167, Treviño 2003 (for

Spanish), Rigau 1994 (for Catalan), among others, have suggested that the nominal in

SEC receives Partitive Case from the verb, following ideas presented by Belletti 1988,

and Lasnik 1992, 1995a, 1999b.62 My proposal can be seen as a way to implement the

Partitive Case Hypothesis inside the Agree system, although with different technical

details, since my analysis assumes that the nominal checks Case in a lower position,

without relation with T. It seems clear that (given that in Spanish the internal nominal

receives ACC and not NOM) Chomsky’s 2000, 2001 analysis of English existential

constructions, according to which the nominal checks case against T and receives NOM,

cannot be extended to Spanish.

61
Notice that here there is no source for ACC other than the small v of haber. Parecer (to seem) does not
license ACC.
62
See also Martin 1992, Bošković 1997b, 2002a, 2005, Epstein and Seely 1999, Boeckx 2000, among
others, for arguments in favor of the Partitive Case Hypothesis. For discussion of its limitations, see
Chomsky 1995: 288, Vikner 1995: 171-176, Vainikka and Maling 1996, Hornstein 2000, among others.
For a recent argument that, even in English there-be sentences, the internal nominal has to receive case
from a source other than T, see Hornstein 2007.

81
1. 7. Conclusions: beyond existential sentences

I have given an explanation for a dialectal split in Spanish with respect to agreement in

Existential Constructions with haber. In the SI dialect the verb has a default value (3rd

person, singular), but in the SII dialect, the verb agrees with the internal nominal, despite

the fact that the nominal is accusative, as shown by the cliticization data. This state of

affairs seems to be a direct challenge for theories that link Agreement with Case, since, in

SII, the nominal seems to be in agreement with T, but T cannot be its Case-licensor

because T does not value ACC. However, I have presented an analysis of the phenomenon

in question that preserves the relation between Case and Agreement and supports the

operation Agree (Chomsky 2000, 2001, 2004).

I contend that Spanish existential constructions with haber have a small v which accounts

for the presence of ACC (in both dialects). Additionally, I have shown that there is a ban

on nominals with [person], the underlying assumption being that only [1p], [2p] and

some [3p] nominals are marked for [person]. The ban is motivated by the fact that

Spanish Existentials Constructions with [person] nominals, contrary to their English

counterparts, cannot have a presentational (or “list”) reading: they are simply

ungrammatical. Interestingly, in SII, haber can be inflected for [1p] or [2p] even when

the internal nominal is [3p]; furthermore, in these cases, it is not possible to use an

explicit subject (that is, these sentences are still existential). If we postulate a pro with a

[person] value ([1p] or [2p]) as responsible for the valuation of T in SII, we would expect

that it could be replaced by a lexical pronoun, which is contrary to the facts. In addition,

82
pro would violate the ban on [person]-marked nominals in Spanish Existential

constructions.

I have shown that we can solve all these problems by postulating: (A) a small v with only

[number] and no [person] for Existential constructions in both dialects, and (B), for SII, a

T with an interpretable [person] feature, in addition to its uninterpretable [number].

According to (A), if small v has only [number], the ban on nominals with [person] is

explained, given the assumption that a probe must match all the features of its goal.

Being incomplete, small v cannot value the case-feature of nominals with [person]. If the

nominal does not have [person], the probe v will be complete with respect to its goal, then

it will be able to value its [case] feature. In SII, given that T has only one uninterpretable

feature, namely [number], it can probe v, producing the effect of object-agreement—this

is possible because the φ-features of v are still present, since the deletion procedure

applies only at the point of Spell-Out. (B) correctly predicts that in SII haber-sentences,

the verb can be inflected in all persons, even if no nominal has the proper value. In

addition, in SI, where both φ-features ([person] and [number]) in T are uninterpretable,

small v is incomplete with respect to T, thus unable to value its features, and then T must

get a default value—which can be implemented by using a SCL.

Further predictions have been borne out with respect to raising verbs as well as other

differences between SI and SII. We have been able to account for the corresponding

difference in the agreement pattern with respect of raising verbs in SI and SII. We have

seen that appealing to the idea that inherent case bleeds Agree explains why we have

83
neither agreement nor clitics in a subset of SEC in some SII dialects (as well as in some

TEC). In addition, the tendency of haber to regularize in SII suggests that T indeed

probes small v in SII. More precisely, I have argued that the regularization of haber

represents a morphological reflex of valuation of T via probing of small v.

An additional prediction arises if we extend the analysis to all Spanish small v.

Recall that the ban on [person] nominals in haber-sentences, which is responsible for the

impossibility even of list-readings with these nominals (that is, their absolute

ungrammaticality), is explained because small v is φ-incomplete, hence it fails to value

the case-feature of φ-complete nominals:63

(103)

[v’ [ v [VP haber… nominal ]]]]]

[unumber: α ] [inumber: α ]
(i)
The nominal values [iperson: β ]
the φ-feature of v. [ucase: ]

8 (ii)
v fails to value the case
of the nominal.

I have assumed that this vP is not a strong phase; which means that the nominal cannot

escape vP to establish a dependency with T. Under Bošković 2005, 2007b’s version of

the PIC, movement is motivated by an uninterpretable feature (uK) in the goal, that is,

63
Notice that the failure of Agree under (44) can also be obtained if the nominal undergoes movement over
v (as in Bošković 2005, 2007b), since v is still incomplete there.

84
with respect to (103), by an uninterpretable case-feature in the nominal. The reason why

uK drives the movement is that a nominal with uK cannot get trapped in the unit that is

sent to Spell-Out, that is, the complement of a phase (we can assume, for ease of

exposition, that uK is PF-incompatible). Since vP in SEC (that is, in (103)) is not a phase

(which means its complement is not sent to Spell-Out), the nominal does not have any

reason to move; it stays in situ with a uCase, even after T is inserted, which is fine so far,

since it is not yet trapped inside a phase complement:

(104)

[TP T [vP [v’ [ v [VP haber… nominal ]]]]]

[unumber: α] [inumber: α ]
(i)
The nominal values the [iperson: β ]
φ-feature of v. [ucase: ]

(iii)
The nominal cannot move. 8 (ii)
v fails to value the case
of the nominal.
8

When T probes small v (which has been in turn given a value for its [number] feature by

the nominal), small v has the ability to check the [number] feature of T. Given this, there

are two possibilities. In SII, this is all we need:64 T, which has an interpretable [person]

feature in this dialect, has all its features valued, so that it can no longer serve as a probe;

as a result the φ-complete internal nominal never gets case. In SI, there is a process to

repair incomplete heads, a version of default agreement, which must be used with all Ts

64
I am disregarding the possibility of checking EPP, in line with Epstein and Seely 1999, 2006, Boeckx
2000, Grohmann et al. 2000, Bošković 2002a, 2007b, Wurmbrand 2006, among others (see also fn. 47).

85
with an φ-incomplete subject. In all sentences of this type, such a subject only values the

[number] feature. The [person] receives default agreement, that is, [3p]. This is fine

because all [1p] and [2p] nominals are φ-complete, only [3p] nominals are φ-incomplete.

However, in SECs small v intervenes between T and the nominal, which prevents T from

probing the nominal.65 So, a φ-complete nominal will never get Case in SI either.

What would happen if we assume the v that is found in normal transitive constructions,

that is, with verbs other than haber, also lacks the [person] feature, just like its haber-

counterpart? The prediction is clear: we will have the same results, namely, the restriction

on [person] objects. Only nominals that are not specified for [person] will be able to

check Case against the small v. If a nominal is specified for [person], the small v will be

able to probe it and to value its own [number] feature (so no φ-features will remain

unvalued); however, it will be incapable of valuing the Case feature of the nominal, given

the assumption that φ-incomplete heads are not capable of valuing Case. I contend that

this is exactly what happens in all Spanish transitive constructions.

The small v with transitive verbs, although φ-incomplete, is a strong phase. Therefore,

again under Bošković 2005, 2007b’s version of the PIC, if the nominal stays in situ, it

will not be able to check Case, because it will be spelled out. Hence, it is forced to move,

to avoid being sent to Spell Out without checking Case, which is possible because now

vP is a phase. If we further assume that there is an additional Dative head that can value

the case feature of the shifted-object, we have an explanation regarding why the object

carries a DAT marker, namely, the preposition “a” (“to”)—which is the so called
65
Notice that I am assuming a strict locality condition, that is, that there is no multiple Agree.

86
Differential Object Marking phenomenon:

(105)

[ DAT [vP nominal [v’ [ v [VP V nominal ]]]]]

[unumber] [inumber] [unumber] [inumber]
[uperson] [iperson] [iperson]
[ucase] [ucase]
8
(i)
Agree fails to value the
(iii)
case of the nominal
The nominal agrees with DAT
and values its [ucase], resulting (ii)
in A-marking The nominal moves to avoid
being sent to spell out with
[ucase]

A possible question is if, when merged, the subject would create any locality problems in

this picture. There are two possibilities. The first one is to adopt a split-vP structure,

similar to the one proposed by Koizumi 1993, 1995 and Lasnik 1995b, where the subject

is base-generated above phrases to which objects move for case-licensing (that is, above

DatP, for Spanish). The second one is to adopt an analysis that assumes that Spanish

subjects are not real subjects (see Barbosa 1995, Olarrea 1996, among others); under such

analysis, the relevant subject features, including θ-features (assuming that θ-roles are

features, as in Bošković 1994, Bošković and Takahashi 1998, Lasnik 1995b, 1999b,

2003a, Hornstein 1999, 2001, among others), could even be checked by agreement

morphemes (see also Manzini and Savoia 2002), assuming that they are clitic-like, as

discussed in the previous sections.66

66
The post-verbal subject remains as a problem. But there is a variety of solutions. For instance, there are
analysis that assume that the post-verbal subject in Spanish is a right specifier (see Zubizarreta 1994,
1999)—in chapter 3, I adopt a similar approach for the DO-IO order. Other analysis are based on lower
copy pronunciation (see Stjepanović 2003 and, for Spanish, Ortega-Santos 2006). I leave this issue for
future research.

87
The picture that arises from these considerations is, I think, interesting. Under the current

analysis, the small v in SEC and the small v in regular transitive constructions are alike:

both are φ-incomplete (and perhaps neither have a subject). The only difference is that

the small v with regular transitive sentences has a thematic role to assign. The small v

with SEC has no thematic role to assign—notice that this could account for the lack of

phasehood with the latter small v (as discussed above). However, Differential Object

Marking in Spanish is a phenomenon that requires much more discussion. I will address

it in the next chapter.

88
 Chapter 2

The properties of Spanish DOM

Spanish is perhaps the most studied of all DOM languages, at least in the Western

tradition. There is in an overwhelming amount of serious literature that deals with this

phenomenon from diverse perspectives and with different goals; there is also a number of

states of the art and commented bibliographies (Niculescu 1959, Ariza Viguera 1989,

Pensado 1995). The vast majority of the works are concerned with the issue of spelling

out the exact descriptive rules that predict the presence/absence of the preposition A—the

most detailed descriptions are Fernández Ramírez 1986: 148-190 and Torrego 1998,

1999. As we will see, total agreement is far from being reached. However, there are at

least two dimensions that everybody has to contemplate in one way or another:

specificity and animacy; some researchers also add a third one: telicity. Topicality,

affectedness and the need to avoid confusion with the subject are also considered as the

driving forces behind Spanish DOM. Not everybody uses this terminology, or even

accept these dimensions as relevant. Curiously enough, there have been few attempts to

integrate all these factors in a comprehensive way—the most successful of these attempts

has been, in my opinion, Torrego 1998. There are even fewer attempts to compare in a

systematic way the descriptions and generalizations regarding Spanish DOM with the

hundreds of other DOM languages—for relevant works, see Bossong 1985, 1991, Isaak

2000, Aissen 2003, Lima 2003, 2006, Carnie 2005.

Given this situation, this chapter has four goals. The first one is descriptive: I will spell

out the reasons why many authors have claimed that specificity, animacy and telicity are

implicated in DOM, and discuss the motivation behind the quest for different or related

factors (in particular, topicality and affectedness, but also person). The second goal is to

review the theoretical literature on Spanish DOM from a critical perspective, showing

that, in spite of their crucial achievements and insights, these theories have enough

descriptive and explanatory limitations to warrant proposing an alternative account. The

third goal is to present what I consider the fundamental contribution of the generative

literature to the DOM phenomenon: that the morphologically marked object is in a

different syntactic position than the morphologically unmarked object. I will endorse this

stance, showing that we can derive this state of affairs from the principles that govern the

interaction between the feature specifications of different DP/NPs and their case-

checking dependencies. This constitutes the gist of my proposal: in accordance with the

theory outlined in chapter 1, there is a φ-incomplete small v in all Spanish transitive

sentences, which is not able to check the case of φ-complete nominals. As a result, they

need to look for a case value outside of vP. Finally, I will show that this point of view not

only leads to considerable empirical gains (virtually eliminating the “exceptional”

character of some DOM sentences), but it also leads to a simplification of the whole

DOM system, which is a valuable explanatory advantage and put us in a position to

address the issue from a comparative perspective. The following sections are dedicated,

respectively, to each goal.

90
2. 1 Describing Spanish DOM

The most common claim regarding Spanish DOM is that objects that are [+specific] and

[+animate] are marked with the preposition A, and all other objects are unmarked. The

following is the basic paradigm behind these claims:67

(1) a. Juan besó a María [+animate, +specific]

John kissed A Mary

John kissed Mary

b. *Juan besó María

John kissed Mary

John kissed Mary

(2) a. María quiere a un abogado [+animate, +specific]

Mary wants A a lawyer

Mary wants a (specific) lawyer

b. María quiere un abogado [+animate, -specific]

Mary wants a lawyer

Mary wants a lawyer (any lawyer)

67
Throughout this dissertation I will use capital A to translate the form “a” (lit. “to”), which is the dative
preposition, but it also have different values—for instance directional (“toward”) and temporal (“at”).

91
(3) a. *Juan destruyó a la ciudad [-animate, +specific]

John destroyed A the city

John destroyed the city

b. Juan destruyó la ciudad [-animate, +specific]

John destroyed the city

John destroyed the city

(4) a. *Juan destruyó a una ciudad [-animate, -specific]

John destroyed A a city

John destroyed a city

b. Juan destruyó una ciudad [-animate, -specific]

John destroyed the city

John destroyed the city

Besides its very well known limitations, which, of course, are acknowledged, the

generalization that emerges from the above paradigm has been assumed by several

researchers to be the core set of data that a satisfactory explanation for the distribution of

Spanish DOM must account for. A correlation like (5) is explicitly endorsed very

frequently—for instance in Zagona 2002: 140, Heusinger and Kaiser 2003: 53, among

several others:

92
(5)

[+animate]
Marked Object
[+specific]

[+animate]
[-specific]

[-animate]
Unmarked Object
[+specific]

[-animate]
[-specific]

We will discuss this generalization now, dissecting each of its components, starting with

specificity.

2.1.1 Specificity and Spanish DOM

As is well known, specificity is not a straightforward notion. As Farkas puts it, “the

notion of specificity in linguistics is notoriously non-specific” (Farkas 2002: 19). In this

dissertation, I do not aim to make any special contribution to this complex issue.

However, given that nature of my proposal, I cannot escape discussing it, especially

because the relevance of this dimension has been questioned, mainly in recent years. For

discussion of the notion of specificity in the context of DOM languages, the reader is

referred to Kliffer 1982, King 1984, Laca 1995, Pensado 1995, Brugé and Brugger 1996,

Torrego 1998, Bleam 1999, Lyons 1999, Torrego 1999, Brugè 2000, Farkas 2002,

Heusinger 2002a, Farkas and Heusinger 2003, Heusinger and Kaiser 2003, Leonetti 2003,

2004, 2005, López 2006, among others. The following discussion relies on their

observations.

93
Almost all the literature on Spanish DOM considers the preposition A to be a marker of

specificity (see Torrego 1999 for an overview).68 There are arguments in favor of this

conclusion, and also arguments against it (although the last ones are not really well

understood). I will show, following Brugé and Brugger 1996, Bleam 1999: 175-186,

Brugè 2000, Leonetti 2003, 2004 and López 2006 that A is not a marker of specificity per

se. We will find that indefinite objects marked with A can be interpreted as specific;

however, the specific reading is not mandatory. While it is true that there is a preference

for the specific reading (in the sense that it is the first reading available without additional

context), this preference can be overridden under appropriate discourse and linguistic

contexts. In addition, since definite objects can have non-specific readings (see Heusinger

2002a and the references therein for discussion), it is predicted that definite objects could

appear without A, which is also correct, an observation that is often ignored (see,

however, Heusinger and Kaiser 2003).69

Evidence in favor of A as a marker of specificity is provided by the following set of data

(the examples mirror similar sentences from Torrego 1999 and Leonetti 2003):

(6) a. Había (*a) un policía en el parque.

there-was A a policemen in the park

There was a policeman in the park.

68
It is important to keep in mind that the expression “marker of specificity” has, for the time being, no
theoretical content, beyond expressing the fact that the presence/absence of the preposition A shows a
degree of alternation with the specific character of the DP. I will come back to this issue.
69
Although it has been recognized that generic definite objects do not require A, I will show that actually
no definite object requires A in all cases.

94
 b. Tenía (*a) un amigo.

he-had A a friend

He had a friend.

(7) a. Busca a una enfermera que habla español.

looks-for A a nurse that speak-INDIC Spanish

He looks for a nurse that speaks Spanish.

b. Busca una enfermera que hable español.

looks-for a nurse that speak-SUBJ Spanish

He looks for a nurse that speak Spanish.

(8) Había besado *(a) varias de sus amigas.

Had kissed A several of his girlfriends

He kissed several of his girlfriends.

(9) Besó *(a) toda chica con sombrero.

kissed A every girl with hat

He kissed every girl with hat.

In (6) we observe that A is incompatible with existential sentences, a typical context

where specific nominals are excluded (as we saw in chapter 1, the internal nominal is

95
actually an object, not a subject, see also Suñer 1982a, among others). In (7) we see a

distinction between an object modified by an indicative clause (7a) and an object

modified by a subjunctive clause (7b); as observed by Rivero 1977, objects modified by

subjunctive relative clauses cannot be specific. In (8) we have a partitive element, which

is typically considered to be specific, and in this case A is mandatory. (9) shows that

strong quantifiers,70 which are also banned from haber-sentences, have to take A. The

data in (6)-(9), then, provide prima facie evidence that A is a marker of specificity.

Nominals with definite determiners seem to provide additional evidence for the claim that

A is a marker of specificity. As Heusinger 2002a explains (see also Lyons 1999: 57-60),

building on observations by Quine 1960: section 31, 146ff, the traditional view of

specificity as a subset of indefiniteness (as in Givón 1978 and many others) cannot be

maintained. The distinction between specific and non-specific items can be extended to

definite nominals too (see also Fodor 1970: 130-215 for discussions of ambiguities with

definite nominals inside intensional contexts):71

70
The difference between strong and weak quantifiers can be tested by the possibility of inserting the
quantifier in an existential sentence. Strong quantifiers are banned from there-be sentences, as well as from
haber-sentences. See Milsark 1977, Barwise and Cooper 1981, Keenan 1987, among several others; and for
Spanish, Suñer 1982a: 61-100.
71
The non-specific/specific readings were originally termed “non referential” and “referential”,
respectively. The shift to “non-specific/specific” is due to Baker 1966 and Fillmore 1967, according to
Heusinger 2002a. With respect to definite NPs, Fodor 1970: 132 distinguishes “referential” and
“attributive” readings, and claims that this difference should not be given the same status as the
specific/non-specific distinction, but she admits that the notion of “having somebody in mind” (that she
uses to characterize specificity) can be applied to definite NPs too.

96
(10) John is looking for the dean…

a. …whoever it might be. [non specific]

b. …namely for Smith, who happens to be the dean. [specific]

[Heusinger 2002b: 248]

Interestingly enough, there are languages whose article system marks specificity and not

definiteness—for instance Maori (Chung and Ladusaw 2004), and according to Lyons

1999: 59, this situation is not unusual outside Indo-European languages. This provides

additional reason to depart from the traditional view that holds that specificity is a subset

of indefiniteness.72

If specificity (and not definiteness) is the notion behind Spanish DOM, the sentences

corresponding to (10a-b) should be, respectively, without A and with A, which is in fact

the case:

(11) a. Juan está buscando el decano, sea quien sea

John is looking-for the dean, be-SUBJ who be-SUBJ

John is looking for the dean, whoever it might be.

b. Juan está buscando al decano, es decir, a Smith

John is looking-for A+the dean, is to say, A Smith

John is looking for the dean, namely for Smith.

72
Heusinger 2002a takes the Turkish accusative marker to be a morphological marker or specificity, also
independent of definiteness (Turkish does not have a definite article); see also Öztürk 2005.

97
It is worth mentioning that (11a) may sound inappropriate for some speakers but this is

just an interference from an interaction between specificity and definiteness. If we

assume that “a definite singular expression unambiguously denotes or refers to one

object, i.e. the object can be identified as the only one that is denoted by the expression”

(Heusinger and Kaiser 2003: 44), we can find the source of the potential confound for

accepting (11a): at first sight there seems to be a conflict between the conditions for the

definite nominal and the possibility that more that one individual satisfies the conditions

for Juan’s object of searching in (11a)—a possibility stressed by the subjunctive clause

sea quien sea (“whoever it might be”). However this nuisance can be overridden when

we introduce an appropriate context—here we use a strategy provided by Heusinger and

Kaiser 2003: 50 to better isolate de dicto (non-specific) readings with definite

expressions:

(12) Juan está buscando el decano que mejor maneje su Departmento

John is looking-for the dean that better run-SUBJ his department

John is looking for the dean that better run his/her department.

It is important to notice that, as expected, the non-specificity of definite nominals is not a

possibility restricted to potential DOM objects (animate and specific, let’s say), but is a

general property of the combination of the definite article and a subjunctive clause that

forces non-specificity (see Quer 1998: 106, Leonetti 1999: 865, Pérez Saldanya 1999:

3265-3266, Ahern and Leonetti 2004, among others):

98
(13) Juan está buscando

John is looking for

a. la casa que esté más cerca de su oficina [non specific]

the house that be-SUBJ more close to his office

b. la casa que está más cerca de su oficina [specific]

the house that be-INDIC more close to his office

(13a) is felicitous if John does not know the house he is looking for (he wants to buy a

house close to his office, for instance). (13b) is felicitous only if John does have some

knowledge about the house (for instance, he just bought it by phone, has the address, and

is looking for the exact place). These data confirm that specificity is not really a subset of

indefiniteness, but an independent property, as argued by Heusinger 2002b, a. Given the

above discussion, we can agree with Heusinger and Kaiser 2003 in that definiteness does

not play a role in Spanish DOM.73 Rather, the relevant notion behind Spanish DOM is

specificity. As we will see later, this conclusion presents a serious problem for OT-based

explanations of DOM, at least as they are presented in Aissen 2003, which predict that all

definite objects should be A-marked. As we has just just seen, this cannot be correct.74

73
According to Heusinger and Kaiser 2003: 42, definiteness might play a role in some American Spanish
dialects, but this is not confirmed by systematic explorations of DOM in these dialects (as in Heusinger and
Kaiser 2005).
74
More generally, this is also a potential problem for the so called Definiteness Hierarchy (Croft 1990,
2003), which assumes a ranking like this:
(i) Definite > Specific > Non-specific [Croft 2003: 132]
Under this view, it is expected that non-specific elements are also non-definite. As I have shown above, this
is not true. I will come back to this issue in the next section, when discussing the OT-model.

99
I would like to add more evidence in favor of this conclusion. Heim has pointed out that

the definite article in superlative constructions like the higher mountain is sort of

“indefinite” (Heim 1999: 8-9)—see also Selkirk 1977, Szabolcsi 1986, Herdan and

Sharvit 2006, among others. If this is correct, we predict that superlative constructions in

an object position can appear without the preposition A. This prediction is also borne out:

(14) Está buscando el decano más alto

is looking-for the dean most tall

He is looking for the tallest dean.

In addition, Herdan 2007 shows that amount relatives (in the sense of Carlson 1977b)

have a covert superlative morpheme; given the above discussion, we then predict that

they should be fine without the A-marker. This prediction is borne out:

(15) Contrató los ingenieros que había en la oficina

hired-he the engineers that there was in the office

He hired the engineers that there were in the office

It is crucial to stress that in (14)-(15) the A-marker is possible. In fact, the generalization

in (5)—under the traditional assumption that all definite determiners are specific—

expects them to have a mandatory A-marker, but this is not the case; in these sentences A

is optional:

100
(16) a. Está buscando (al/el) decano más alto

is looking-for A+the/the dean most tall

He is looking for the tallest dean.

b. Contrató (a) los ingenieros que había en la oficina

hired-he A the engineers that there was in the office

He hired the engineers that there were in the office

Thus, although the behavior of definite nominals does provide evidence in favor of

rejecting the idea that specificity is a subset of indefiniteness, it represents a challenge for

the generalization in (5). A satisfactory solution for DOM should address these data.

An interesting situation arises with nominals denoting kind, which do not need to have A:

(17) a. Antoñito buscaba la mujer rica

Antoñito looked-for the rich women

Antoñito looked-for the rich women

b. …una fuente de vida nueva que purifica el hombre moral

…a source of life new that purifies the man moral

…a source of new life that purifies the moral man.

[Fernández Ramírez 1986]

101
Commenting on these examples, Brugé and Brugger 1996 observe that kind-nominals can

receive A:

(18) …una fuente de vida nueva que purifica al hombre moral.

…a source of life new that purifies A+the man moral

…a source of new life that purifies the moral man.

[Brugé and Brugger 1996: 6]

They claim that (18) is due to the fact that kind-denoting nominals can freely have

[+animate] or [-animate], and therefore that the object has [-animate] in (17), preventing

the insertion of A. As it should be apparent, this represents a major departure from the

idea that animacy is a semantic notion—a step that other researchers are also willing to

take (see Torrego 1998: 55). I reject this move, and contend that the peculiarities of these

nominals should be accounted for by resorting to the idea that kind-denotation is a

reading available independently of referentiality (as in the tradition stemmed from

Carlson 1977a), keeping [animate] as a semantic feature.

I leave the explanation for the presence/absence of A in the above example for section

2.3, but I want to conclude here that, given that definite nominals can drop A even when

they are not kind-denoting (as previously discussed), it seems more reasonable to

assimilate the dropping of A to a more general property of determiners. In other words,

there is no correlation between having A and being a kind-denoting expression. Notice

102
that this also challenges the generalization (5), since it implies that neither specificity nor

animacy is enough to define kinds (that is, the behavior of kind-denoting expressions

with respect to DOM is not captured by (5)).

There are three conclusions to be drawn from the discussion of definite nominals. First,

definiteness by itself does not play a role in Spanish DOM—as Heusinger and Kaiser

2003 also conclude. Second, the Spanish DOM phenomenon provides evidence in favor

of Heusinger 's 2000 notion of specificity, namely the idea that specific nominals are not

a proper subset of indefinite ones; hence definite nominals can also be classified as (non)-

specific. Third, given that non-specific definite nominals can have A optionally, this

challenges the almost universally accepted claim that A is a marker of specificity. As I

will show now, indefinite nominals also provide evidence in favor of the third

conclusion.

It is sometimes observed (see Pensado 1995: 32-33, Brugé and Brugger 1996, Bleam

1999: 175-186, Brugè 2000: 211, Leonetti 2003, 2004, López 2006: 145-149) that

indefinite objects marked with A are not necessarily specific. Rather, as argued by these

authors, A-objects have the possibility of being specific, but they do not need to be so.

Their evidence comes from the fact that A-marked objects are possible (although not

mandatory) with other grammatical devices that are related to non-specificity, in

particular nouns modified by subjunctive clauses (19a) and nominals without determiners

(19b):

103
(19) a. Juan busca a una enfermera que sepa español.

John looks-for A a nurse that speak-SUBJ Spanish

John looks for a nurse that speaks Spanish.

b. El comité busca a estudiantes con buenas calificaciones

The committee looks-for A students with good grades

The committee looks for students with good grades.

As observed by Rivero 1977 for Spanish (see Quer 1998 for other Romance languages),

subjunctive relative clauses are incompatible with a specific reading of the nominal

whose head they modify; (19a) then must be non-specific, in spite the fact that it does

have A. Additionally, in Spanish, determinerless nouns cannot have a referential reading;

thus (19b) must be non-specific, but it does have A. So, it seems that A does not actually

encode specificity.

Additional evidence comes from the fact that both DOM and no-DOM objects can have

de re and de dicto readings in intensional contexts. 75 A typical test to differentiate

between de re and de dicto readings is provided by sentences like this:

75
When reporting propositional attitudes (as beliefs or desires), sentences exhibit an ambiguity with respect
to the quantified expression. Take the sentence:

(i) John believes that a French engineer is efficient.

This could mean either of the following:

(ii) a. John believes that French engineers are efficient. de dicto

b. Some engineer is such that John believes that he is efficient. de re

In the de re reading (ii-b), John directly attributes a property (being efficient) to a particular engineer. In the

104
(20) John wants to hire a French engineer.

If John wants to hire Mary, but he does not know that Mary is French, then the nominal a

French engineer must have a de re reading, and it cannot have a de dicto reading, which

requires that John knows that the person that he wants to hire has to be French. In

Spanish, de dicto readings can be obtained with or without A:

(21) a. Juan quiere contratar a una ingeniera francesa

John wants to hire A a engineer French.

John wants to hire a French engineer.

b. Juan quiere contratar una ingeniera francesa

John wants to hire a engineer French.

John wants to hire a French engineer.

According to native speaker judgments, both (21a) and (21b) are felicitous in a situation

where Juan wants to hire Mary and he does know that she is French; that is, both of them

can have a de dicto reading. With respect to the other reading, in the case of (21a), there

is no problem obtaining a de re reading; in addition, there seems to be no strong

preference for one reading or the other in (21a). But this is not the case for (21b); if

someone utters (21b), there is a strong preference for the de dicto reading. However,

de dicto reading (ii-a), such property is not linked to any particular engineer. Notice too that, in (ii-b), John
does not need to know if the engineer he deems efficient is French.

105
native speakers judge (21b) to be felicitous under circumstances that seem to require a de

re reading. Consider (22):

(22) Peter, Bill and Juan are having lunch together in the cafeteria, discussing candidates

for a position in engineering. Then, Mary, a beautiful girl and a candidate, enters the

cafeteria, and Juan, who does not know that Mary is French, says “I want to hire

her!” To excuse Juan for his inappropriate comment, Peter says to Bill:

(23) Parece que Juan quiere contratar una ingeniera francesa

Seems that John wants to hire a engineer French.

It seems that John wants to hire a French engineer.

Native speakers judge that Peter has uttered a felicitous sentence in this case. 76 We

should not conclude so rapidly that una ingeniera francesa has a “specific” reading. In a

context like (22), Peter knows that John wants to hire Mary; given that he also knows that

Mary is French, for Peter, it is also true that John wants to hire a French engineer: as it is

well known, wide readings entail narrow ones (see Ruys 1992: 6-11 for some interesting

discussion on this issue). In other words, una ingeniera francesa is taken de re by the

speaker, but it does not refer to any particular individual with respect to Juan. This seems

to fit the description of the so-called “non-specific de re” readings (see Fodor 1970,

76
Notice that Juan is the subject of the sentence, but he is not the speaker, that is, Juan is not a participant
in the speech act. This is relevant since specificity is said to be dependent on conditions related to the
participants in the speech act, especially the speaker (see Fodor 1970, Enç 1991, Lyons 1999: 165-178,
Farkas 2002, Heusinger 2002a, among several others)—see also fn 77.

106
Bonomi 1995, von Fintel and Heim 2003, among others). 77 If this is correct, we can

conclude that non-A objects cannot be specific.

This suggests another possibility to deal with non-specific/narrow readings of A-objects:

these readings are obtained under entailment from specific/wide ones. This would mean

that the A-objects in question are not really non-specific. This is, however, not true. The

examples in (19) already point to this direction. In those cases, the specific interpretation

is not available; therefore, their non-specific reading cannot come from the latter. There

is additional evidence here.

Donkey-sentences provide a test for non-specificity (see Heim 1990 for discussion).

Indefinites in donkey-sentences cannot be specific, they need to have a bound reading

with respect to the universal quantifier:

(24) Every farmer that owns a donkey beats it.

A-objects are perfectly fine with donkey-sentences:78

(25) Toda cadena que contrata a un actor famoso gana dinero con él.

Every network that hires A an actor famous, wins money with him

Every network that hires a famous actor, wins money after him.

[Bosque 2001: 25]

77
Notice that “specific” implies de re but de re does not imply “specific”; “non-specific” does imply de
dicto, however (and, of course, de dicto implies non-specific)—see von Fintel and Heim 2003: 6-7 for a
formal implementation of this correlation (and lack thereof).
78
(25) and (27) come from Bosque 2001, who is not discussing DOM, but he does use them to test non-
specificity, with different purposes.

107
More evidence is provided by so called cleft conditional sentences (see Gutiérrez

Ordóñez 1994 for Spanish). Since, in the if-clause, such sentences have free variables

whose content is identified in the copulative clause, they constitute another ground for

testing non-specificity:

(26) If John loves any girl, it is Mary.

A-objects are fine with cleft conditionals:

(27) Si Juan ha conocido a una persona interesante en su vida, ha sido María

If John has met A a person interesting in his life, has been Mary

If John has ever met anyone interesting in his life, it is Mary

[Bosque 2001: 26]

The above discussion leads us to an inescapable conclusion: the preposition A cannot be

a marker of specificity. Although this is in conflict with the vast majority of the literature

on Spanish DOM (see Torrego 1998, 1999 and the references she cites), it is also true that

the literature has put aside the problems presented by the examples discussed above,

limiting itself to explaining the paradigm in (1)-(5), which has been a useful idealization

so far, but it needs to be refined if we ever want to advance our understanding of Spanish

DOM. It is necessary to point out that this idealization is not something particular to

Spanish DOM researchers, but is quite common in the DOM literature in other languages

108
and in different frameworks; for instance, OT researchers make similar idealizations with

respect to the languages they work on—to say it in Aissen words, her model

“oversimplifies in some respects” (Aissen 2003: 471).79 Nevertheless, there has been a

growing line of research on Spanish DOM that has pointed out this problem and has also

concluded that A is not a marker of specificity; see Pensado 1995: 32-33, Brugé and

Brugger 1996, Brugè 2000: 211, Leonetti 2003, 2004, López 2006: 145-149 (a

conclusion already anticipated in the rich descriptive literature on Spanish DOM; see in

particular Fernández Ramírez 1986: 148-190 and Laca 1995). I will side here with this

line of research, in light of the evidence presented so far.80

At this point, and putting aside for the moment the issue of animacy (which I will include

in the next section), we could raise the following question: does the A-marker make any

contribution to object interpretation? I contend that the A-marker does not make any

semantic contribution, and this claim seems to be firmly grounded in the above

discussion. But data that seem to contradict it has already been presented (see (6)-(9)).

On one hand, there are some objects that cannot accept the preposition A, precisely those

that cannot be specific—I repeat (6) here:

79
Although Aissen’s system has been improved to overcome some of her “oversimplifications” (as in de
Swart 2003 and Morimoto and de Swart 2006), the issue of specificity remains the same.
80
Notice that this conclusion also undermines explanations of DOM that are based on the framework of
generalized conversational implicatures (García García 2005)—since they rest on the assumption that A-
marked objects are more informative. In other words, as recognized by García García 2005: fn. 14, this line
of research does not predict cases where the presence/absence of A makes no difference with respect to the
information conveyed by the sentence.

109
(28) a. Había (*a) un policía en el parque.

there-was A a policemen in the park

There was a policeman in the park.

b. Tenía (*a) un amigo.

he-had A a friend

He had a friend.

We already seen in chapter 1 the case of existential sentences (like (28a): [+specific]

objects are not admitted there because the small v in haber-sentences cannot value their

case-feature (because they are φ-incomplete)—as we will see later in this chapter, this

means that haber-sentences do not have what it takes to mark objects with A (a Dative

Phrase). In chapter 3, I will show that possessive verbs like tener (to have) present a

similar case, which will be shown to account for the lack of A in (28b). Additionally, the

contrast between indicative and subjunctive clauses modifying objects—presented in

(7)—does not really show that A is a marker of specificity, since A is optionally possible

in the example with subjunctive clauses (29), as discussed above:

(29) Busca (a) una enfermera que hable español.

looks-for A a nurse that speak-SUBJ Spanish

He looks for a nurse that speak Spanish.

110
Pending an account of the optionality (which I will provide later), we can conclude that

the lack of A in the above examples is not evidence against the view that A is not a

marker of specificity.

On the other hand, there are some cases where A is mandatory, precisely those

constructions where the object must be specific. Two of them have already been

presented in (8)-(9): partitive constructions and objects with strong quantifiers. I repeated

them here in (30a) and (30b) respectively:

(30) a. Había besado *(a) varias de sus amigas.

Had kissed A several of his girlfriends

He kissed several of his girlfriends.

b. Besó *(a) toda chica con sombrero.

kissed A every girl with hat

He kissed every girl with hat.

The other case, provided by Bosque 2001, is related to a very well known alternation

between pre- and post- nominal adjectives—see Ticio 2003: 112-145 for a thorough

discussion of the issues and the relevant literature concerning this phenomenon and its

consequences for understanding the structure of DPs. Prenominal attributive adjectives

force a specific interpretation of indefinites and they must receive A:

111
(31) a. Busco (a) un actor famoso

look-for A a actor famous

I am looking for a famous actor.

b. Busco *(a) un famoso actor

look-for A a famous actor

I am looking for a famous actor.

[Bosque 2001: 27]

I will leave the full explanation of these facts for section 2.3, but let me advance the

following idea, which will allow me to conclude this section.

I will assume that Spanish has two definite, non definite types of determiners, which I

will name D and *D. This distinction cut across definite and indefinite determiners,

across quantifiers (weak or strong), and across kinds, and in fact, across all types of

determiners. That is, definite and indefinite determiners can be D or *D. The crucial

difference between D and *D is that D can optionally be specific or non-specific, whereas

*D cannot be specific:

(32)
D [+specific] or [-specific]
*D [-specific]

112
It is important to spell out what I mean by [+specific]. As I mentioned in the beginning of

this section, I do not aim to contribute anything special to the notion of specificity. I use

[+specific] as stand for a set of properties that may or may not be related. In particular, I

mean by [+specific] two things: (i) a choice-function interpretation, and (ii) a strong

interpretation. This means that only D determiners have the possibility of receiving a

choice function (in the sense of Reinhart 1992, 1997, Kratzer 1998 but also Heusinger

2002b), but they do not need to do so, and only D determiners can be weak or strong; in

turn *D determiners cannot receive a choice function, nor they can be strong.

For indefinite determiners, this distinction should be straightforward, assuming that their

specificity can be accounted for by the availability of a choice function (as in Reinhart

1997, Kratzer 1998; see also Lidz 2006 for an application of choice functions to the

specificity effects induced by DOM-objects in Kannada). A choice function is a function

that maps from a set of individuals to a particular member of that set:

(33) Choice Function

A function f is a choice function CH(f) if it applies to any non-empty set and yields

a member of that set.

[Reinhart 1997: 374]

To see an example, let’s consider this sentence:

(34) Every student reads some book.

113
In the wide-scope reading, there is a particular book that every student reads (The

Minimalist Program, for instance). This reading can be represented by assigning a choice

function to some:

(35) ∃f (CH (f) ∧ ∀z (student (z) Æ z read f(book))

According to (35), there is a function f, such that for every z, if z is a student, then z reads

the book selected by f (namely, The Minimalist Program).81 As extensively discussed by

Reinhart 2006, this sets the cornerstone for a principled explanation of a number of issues

regarding scope and QR (far beyond DOM systems). Following Lidz 2006 (who works

on Kannada), I take choice functions as a tool also relevant to accounting for specificity

in DOM-objects.

Indefinite determiners, then, come in two flavors: D and *D. Indefinite D determiners

have the possibility of optionally receiving a choice function, whereas indefinite *D

determiners do not have that possibility.

Definite determiners, on the other hand, are not usually understood as having choice

functions; in fact, according to Reinhart 1997: 368-370, determiners have to be weak in

order to receive a choice function, although not all weak determiners can receive it—

which means that we have to understand the indefinite *D as a weak determiner that

81
Notice that existential closure is needed to bind the function variable. There is some debate concerning to
where the closure applies, an issue that is only relevant for sentences with intermediate scope. For Reinhart
1997, existential closure can apply at any point in the derivation; for Kratzer 1998, it can only apply at the
root node. See Lidz 2006 for ways to tease these options apart in the context of DOM languages.

114
cannot receive a choice function. But Heusinger 2002b has found an interesting way to

use choice functions to analyze definite determiners, by distinguishing what he calls a

local choice function (the ones analyzed by Reinhart) from a global choice function

(GCF), which can be applied to definite DPs. GCF can be successfully used to represent

the salience structure in the discourse and in this way it determines the actual reference of

definite DPs. To evaluate the technical details of this notion is beyond the scope of this

dissertation, so let me merely illustrate it with an example (taken from Heusinger 2002b).

The lake Constance is a large lake on the Rhine, located between Germany, Switzerland,

and Austria, which contains three islands: Mainau, Reichenau and Lindau. The definite

expression the island on Lake Constance then can refer to three different individuals,

namely, the island Mainau, the island Reichenau and the island Lindau, depending on

different situations:

If we hear the expression [the island on Lake Constance] from a Reichenau

fisherman, he probably means the island Reichenau; if we encounter the same

sentence during a guided tour through Lindau it will rather be the island Lindau

that is meant; however, uttered by the Earl, owner and occasional inhabitant of

Mainau, the sentence is sure to be about the island Mainau.

[Heusinger 2002b: 68]

Assuming that each context i has its own choice function F, a definite DP the N is

interpreted in this way (informally):

115
(36) The selected x in the context i such that x is N or the most salient x in i such that x is

N.

[Heusinger 2002b: 68]

This means that the uniqueness requirement usually associated with definite expressions

can be actually interpreted as the unique availability of the referent in a given context,

that is, a global choice function picks up the salient individual in a particular context.82 If

we assume that definite determiners can also be D or *D and that *D determiners cannot

receive a global choice function, this implies that definite *D cannot receive a global

choice function, meaning that it cannot pick an individual referent, but other non-

referential/non-specific interpretations are available, as discussed: nominals modified by

subjunctive clauses (37a), modified by clauses containing comparative adverbs (37b),

modified by superlatives phrases (37c), or modified by amount relative clauses (37d):

(37) a. Juan está buscando el decano, sea quien sea

John is looking-for the dean, be-SUBJ who be-SUBJ

John is looking for the dean, whoever it might be.

b. Juan está buscando el decano que mejor maneje su Departmento

John is looking-for the dean that better run-SUBJ his department

John is looking for the dean that better run his/her department.

82
Interestingly enough, as observed by Laca 1995: 83, the discourse saliency of a referent highly increases
the likelihood that a definite animate object referring to it will be marked with A, whereas the lack of
saliency decreases it.

116
 c. Está buscando el decano más alto

is looking-for the dean most tall

He is looking for the tallest dean.

d. Contrató los ingenieros que había en la oficina

hired-he the engineers that there was in the office

He hired the engineers that there were in the office

I assume that the determiners in (37) are *D determiners. As discussed above, all these

sentences can have a version with the object marked with A. I contend that, when A is

present, this means that a D determiner is present. This is consistent with the view

(expressed in (32)) that D can have a specific reading, but it does not need to have it.

From the above discussion, the following generalization emerges, putting aside animacy

(until the next section):

(38) Definite Determiners

I II

Can receive a global choice function Must have A

D
(if animate)

*D Cannot receive a global choice function Cannot have A

117
Notice that (38) does not imply any causal relation between I and II, it is just a

descriptive correlation; that is, (38) should not be understood as saying that A is the

carrier of the global choice function. We will explain this correlation in section 2.3.

With respect to indefinite determiners, recall that *D has to be understood as a weak

quantifier that cannot receive a local choice function. D in turn can be strong or weak,

but, crucially, it has the possibility of receiving a local choice function:

(39) Indefinite Determiners

I II

Can receive a local choice function Must have A

D
Can be strong or weak (if animate)

Cannot receive a local choice function Cannot have A

*D
Cannot be strong

Again, this is just a descriptive correlation, which will be explained in section 2.3. As in

the case of definite determiners, this also implies that A is optional with some indefinite

determiners, precisely the ones in the contexts where the choice function is not enforced

(40); but A is mandatory if the indefinite must receive a choice function (41) or if the

determiner is strong (42), as discussed above:

(40) a. Juan contrató a un actor famoso

John hired A a actor famous

John hired a famous actor.

118
 b. Juan contrató un actor famoso

John hired a actor famous

John hired a famous actor.

(41) Juan contrató *(a) un famoso actor

John hired A a famous actor

John hired a famous actor

(42) a. Juan contrató a todo actor famoso

John hired A every actor famous

John hired every famous actor.

b. * Juan contrató todo actor famoso

John hired every actor famous

John hired every famous actor.

Notice that in (40) the A-marker is optional only under the non-specific interpretation; if

(40a) is understood as specific, that is, if (40a) receives a local choice function, A

becomes mandatory:

(43) Juan contrató *(a) un actor famoso, Sean Connery

John hired A a actor famous, Sean Connery

John hired a famous actor, Sean Connery

119
Likewise, weak quantifiers can receive A optionally, under the non-specific interpretation

(that is, without a choice function):

(44) a. Juan contrató a dos actores famosos

John hired A two actors famous

John hired two famous actors.

b. Juan contrató dos actores famosos

John hired two actors famous

John hired two famous actors.

This is exactly what we expect from the generalization in (39): only D determiners can

have choice functions, and only D determiners can be strong. However, D determiners do

not need to have a choice function, as witnessed by the compatibility of A with cases

discussed above, where it is not possible to receive a choice function:

(45) a. Subjunctive clauses

Busca (a) una enfermera que hable español.

looks-for A a nurse that speak-SUBJ Spanish

He looks for a nurse that speak Spanish.

120
 b. Donkey sentences

Toda cadena que contrata (a) un actor famoso gana dinero con él.

Every network that hires A an actor famous, wins money with him

Every network that hires a famous actor, wins money after him.

c. Cleft Conditionals

Si he conocido (a) una persona interesante en mi vida, ha sido María

If has met A a person interesting in his life, has been Mary

If I have ever met anyone interesting in his life, it is Mary

We can summarize these findings in the following tables, which, once again, should be

understood only as descriptive correlations, without any theoretical import with respect to

the A-marker:83

(46)

[no choice function & A] (40a)

[no choice function & no A] (40b)

[choice function & A] (41)

*[choice function & no A] (43)

83
Recall that, as discussed above, local choice functions are not compatible with strong quantifiers, so we
cannot combine (46) and (47), but the reason for this is entirely independent from DOM systems.

121
(47)
[weak & A] (44a)

[weak & no A] (44b)

[strong & A] (42a)

* [strong & no A] (42b)

So far, I have not explained the reasons for these correlations, but, as should be apparent,

according to (46)-(47), if something forces the specificity of the object (that is, if

something forces the nominal in object position to have a choice function or if the object

has a strong quantifier), it must receive A. On the other hand, if the specificity is not

forced, or if the object has a weak quantifier, A is optional. This means that A is not a

marker of specificity, since the presence of A does not force specificity.

A final conclusion must be added with respect to kind-denoting nominals, which, as

discussed above, can have A optionally:

(48) Una fuente de vida nueva que purifica (a) el hombre moral

A source of life new that purifies (A) the man moral

A source of new life that purifies the moral man.

I take this optionality to indicate that kind-denoting nominals can freely be D or *D. This

assumption, combined with the system I will present in section 2.3, will explain the

presence/absence of A in (48), just as in the cases discussed above.

122
Now I will turn to the less complicated issue of animacy in Spanish DOM.

2.1.2 Animacy and Spanish DOM

The relevance of animacy for Spanish can hardly be denied. There are some minimal

pairs that are distinguished just by the feature [± animate]:

(49)

[+animate] [-animate]

Alguien Algo
Somebody Something

Nadie Nada
Nobody Nothing

Quién Qué
Who What

In all [+animate] cases above, objects must receive A:

(50) a. Vi *(a) alguien en el parque

saw A somebody in the park

I saw somebody in the park

b. No vi *(a) nadie en el parque

No saw A nobody in the park

I saw nobody in the park

123
 c. *(A) quién vi en el parque?

A who saw in the park

Whom did I see in the park?

On the other hand, none of the [-animate] cases can receive A:

(51) a. Vi (*a) algo en el parque

saw A something in the park

I saw something in the park

b. No vi (*a) nada en el parque

No saw A nothing in the park

I saw nothing in the park

c. (*A) qué vi en el parque?

A whatsaw in the park

What did I see in the park?

Notice that all the cases in (50) are “non-specific”, so they provide further confirmation

that A cannot be a marker of specificity. In fact, such cases have been a particularly hard

issue for proponents of the A-as-a-specific-marker hypothesis, and some proponents of

the hypothesis have even suggested that those items are actually specific—see for

instance Kliffer 1982: 203-204 and Torrego 1998: 175 fn. 156. I contend that this cannot

124
be true, based on the fact that the elements in (49) are possible with existential

sentences:84

(52) a. Había alguien

there-was somebody

There was somebody

b. No había nadie

No there-was nobody

There was nobody

c. Quién había?

Who there-was

Who was there?

As discussed in chapter 1, the lack of A in existential sentences does not have anything to

do with specificity distinctions, but with the case-assignment mechanism.

84
Of course, it is always possible to imagine a systematic ambiguity: for Torrego 1998: 175 fn. 156 the
items in (50) are [+specific] in sentences like (51) but [-specific] in existential sentences. However, if this
were correct, it would mean that [+specific] is not really a semantic/pragmatic feature—see also Torrego
1998: 55 for the idea that not even [+animate] is semantic. Here, I reject this notion.
In addition, for quién (who), we could entertain the possibility that it is D-linked (in the sense of Pesetsky
1987), that is, discourse-linked, and then [+specific]; this, however, can not be true, since aggressively non-
D-linked phrases—like the hell (or its equivalent in Spanish, diablos)—are perfectly compatible with quién,
preserving the obligatory A:

(i) *(A) quién diablos viste en el parquet

A who demons saw in the park
Who the hell did you see in the park?

125
There is a subclass of nominals that it is easy to recognize as animates: pronouns, proper

nouns of persons and animals; all of them must take A (as observed many times; see

Fernández Ramírez 1986, Laca 1995, Torrego 1999 for overviews on the issue of

animacy in Spanish DOM):

(53) a. Lo vi *(a) él

CL saw A he

I saw him

b. Vi *(a) Pedro

saw A Peter

I saw Peter

c. Vi *(a) Fido

saw A Fido

I saw Fido

With other animate DPs, A can be dropped, under the circumstances described in the

previous subsection (in relation with specificity):

(54) a. Juan está buscando (a) la decana que mejor maneje su Departmento

John is looking-for A the dean that better run her department

John is looking for the dean that better run his/her department.

126
 b Busco (a) un actor famoso

look-for A a actor famous

I am looking for a famous actor.

c. El comité busca (a) estudiantes con buenas calificaciones

The committee looks-for A students with good grades

The committee looks for students with good grades.

A particularly convincing example of the relevance of animacy for A-marking is

provided by the behavior of the noun pueblo, which can mean either town (in which case

it is [-animate]) or people (in which case it is [+animate]). Only the [+animate] meaning

triggers the presence of A—as observed by Torrego 1998: 67, among others:

(55) a. Buscan el pueblo de Numancia

looking-for the town of Numancia

They are looking for the town of Numancia

b. Buscan al pueblo de Numancia

looking-for A+the people of Numancia

They are looking for the people of Numancia

There is a subclass of animates, though, that can never receive A: bare nouns without

127
modifiers—as observed by Torrego 1984a, Contreras 1986, Lois 1989, Brugé and

Brugger 1996: 8-10, among others:85

(56). El comité busca (*a) estudiantes

The committee looks-for A students

The committee looks for students.

Notice also that the nominals in (54) do not need to have a specific reading even if A is

present. That is, as discussed in the previous section, A does not preclude a non-specific

reading. So, in principle, we cannot use specificity (or rather the lack thereof) as an

explanation for the mandatory lack of A in (56). I will account for this lack of A in

section 2.3.

On the other hand, non-animate nominals do receive A under certain circumstances, as

has been pointed out by many authors (see Hanssen 1945: 296, Molho 1958, Rebollo

Torío 1976/1977, Luján 1978, Weissenrieder 1985, Fernández Ramírez 1986, García

Martín 1988, Lois 1989, Weissenrieder 1990, 1991, Laca 1995, Contreras 1996, Torrego

1998: 55, Martín 1999, Torrego 1999: 1788, Company Company 2002, Zagona 2002: 13-

14, Barraza Carbajal 2003, Heusinger and Kaiser 2003, Estrada García 2005, Martín

2005, Laca 2006), who have consequently questioned the relevance of animacy to

Spanish DOM:

85
However, bare nouns with modifiers can receive A optionally (see (19)).

128
(57) Los ácidos atacan (a) los metales

The acids attack A the metals

Acids corrode metals.

[Molho 1958: 215]

Notice that (57) is not a case of figurative speech. It is well-known that, when objects are

personalized, as for instance, in cartoon-like situations, they behave (not surprisingly)

like [+animate] objects, and therefore they receive A:

(58) La sartén atacó a la olla

The pan attacked A the pot

The pan attacked the pot.

But in (57) attack means corrode, in fact, if we use the more technical corroer

(“corrode”), the A stays optional:

(59) Los ácidos corroen (a) los metales

The acids corrode A the metals

Acids corrode metals.

Pushed by this kind of examples, some authors have proposed to interpret the notion of

[animate] in a different way. Torrego 1998: 55, using a category borrowed from

Jackendoff 1983: 180-183, suggests to use [ACTOR] instead. It is unclear, however, how

129
this can be applied to (57). If anything, following Jackendoff’s definition (“the one that is

performing the [ACTION]”—Jackendoff 1983: 180), the [ACTOR] in (57) is Los ácidos,

no los metales. Even if we extent the definition to mean “capable of performing an

[ACTION]” or “autonomous” (as Torrego seems to have in mind), it is still unclear what

action are the metals capable of performing or how they are autonomous. Thus, I reject

this notion.

However, it is true that [animate] needs some clarification, and most certainly it cannot

be limited to “living” individuals. I think it is natural to assume that animacy is a notion

related to the autonomy of motion. This includes not only humans and animals, but also

things like vehicles. If so, this means that we have a [+animate] in this case:

(60) …atravesando el Pont Neuf, vi al barquito en cuestión

crossing the Pont Neuf, saw A+the little+ship in question

crossing the Pont Neuf, I saw the little ship in question

[Laca 1995: 83]

This idea of animacy as autonomy of motion implies that things that are understood as

lacking motion cannot be [+animate], and they do not receive A:

(61) * Vi (*a) la casa en cuestión

built A the house in question

I built the little ship in question

130
It is possible to suggest that the semantic [± animate] should be replaced by an alternative

[± motion] as the feature relevant for DOM, but I would continue using [± animate] for

expository reasons.

As it should be apparent, this does not apply to (57), since metals do not satisfy this

criterion. I will come back to the issue of truly non-animate objects that receive A in the

next section. Let me point out for now that there are a number of verbs where A is

mandatory, even when the object is very clearly non-animate:

(62) a. Un adjetivo califica a un sustantivo

An adjective qualifies A a noun

An adjective qualifies a noun

b. Los días siguen a las noches

The days follow A the nights

The days follow the nights

[Torrego 1999: 1788]

Cases like these have been taken by several researchers as crucial evidence that the key

factor behind DOM is actually a principle of distinguishability between the subject and

the object, in the sense that, according to these authors, A marks the object when the

object has the same properties as the subject, i.e., when both are animate and specific (see

Hills 1920, Weissenrieder 1985, 1990, 1991, Contreras 1996, Zagona 2002: 14,

Morimoto and de Swart 2006, among several others)—this is sometimes expressed as the

131
potential for ambiguity between subject and object, given the relatively free order in

Spanish (see Laca 1995: 69-74, Torrego 1999: 1784 for some discussion of this issue).

Since the presence of A with non-animates posses a challenge for the classic OT analysis

of Aissen 2003 (which I will discuss in the next section), some researchers working in

this framework have proposed a formal implementation of this idea that could make it

compatible with OT. Morimoto and de Swart 2006, following a similar idea presented by

de Swart 2003 and De Hoop and Lamers 2005 for other DOM languages, propose a

constraint of minimal semantic distinctness for Spanish, according to which the

arguments of a transitive clause must be minimally distinct, that is, the subject must be

higher than the object on the relevant semantic scales (animacy and definiteness). When

this constraint is violated, the object is marked, giving rise to a DOM system:

(63) DISTINGUISHABILITY

Mark objects that are not outranked by the subject in prominence (animacy,

specificity).

[Morimoto and de Swart 2006: 232]

However, this analysis is easily falsified:

(64) El huracán provocó (*a) la tormenta

The hurricane caused the storm

In this case, both subject and object have the same degree of definiteness and animacy,

132
but the marker is not possible. Thus, I reject explanations based on (63) or any of its

antecedents (like the subject-object ambiguity).

I will show in section 2.3 that the examples discussed above do not constitute a problem

for the generalization regarding Spanish DOM based on animacy, since there is good

evidence to suggest that in these cases the A is accidental. Before doing that, however, it

is necessary to discuss a couple of more notions often associated with Spanish DOM.

2.1.3 Other factors: topicality, event structure

There is a tradition of research that tries to encompass DOM from a discourse-related

perspective (Isenberg 1968, Laca 1987, García 1993, 1995, Laca 1995, Delbecque 1998,

Leonetti 2003, 2004, García García 2005, among others). In this approach, the use of A is

driven by the topic-nature of the nominal. It is important to point out that “topic” in this

line of research is meant to refer to “aboutness”—in that sense, it is related to the notion

of internal or low topic (Cecchetto 2000, Belletti 2001, 2004). It has been difficult for

these researchers to establish that this discourse notion is the decisive factor behind DOM

(as recognized by Delbecque 1998: 400-401), given that many nominals can have a

“topic” interpretation without being A-marked, and that A-marked nominals do not need

to have this interpretation (see also Martín 1999: 472-473).

There are mechanisms in Spanish that are more clearly discourse-related; for instance,

clitic-doubling, whose relation to topicality has been suggested by Núñez del Prado 1997

133
and Suñer 1999.86 From this point of view, the clitic acts like scopal marker to allow a

presuppositional reading, given that assumption that, in order to obtain this reading, the

object must reach the domain of the restriction (Diesing 1992); for Germanic, this is

obtained by moving the DP overtly (Object-Shift), while Spanish does this by using the

doubling clitic (and the doubled object stays in situ, at least overtly). There are alternative

views. For Sánchez 2005, at least for Limeño Spanish, clitic-doubling structures are not

topics, but instances of indentificational focus (in the sense of Kiss 1998). All authors

agree that there is a contrast between clitic doubled and non-doubled structures:

(65) a. Vi a Susana

Saw A Susan

I saw Susan.

b. La vi a Susan

CL saw A Susan

I saw Susan

Everybody agrees the object in (65a) conveys new information (it is informational focus).

For Sánchez 2005, (65b) is not presuppositional, but an instance of identificational focus

(which, as proposed by Kiss 1998: 245, identifies a subset of elements of a contextually

given set). 87 Evidence to this effect comes, for instance, from the fact that it is not

86
Cecchetto 2000 and Belletti 2005 also treat doubling structures in Italian as corresponding to internal
topics.
87
In fact, Núñez del Prado 1997: 238-269 observes that the “presuppositional” character of clitic-doubling
must be understood as “familiarity” and it cannot mean “understood” object, which is different from the

134
possible to continue a sentence that introduces a novel element in the discourse with a

clitic-doubling structure, but it is possible to do so with a clitic-only structure:

(66) a. I saw Mary. …

b. La saludé efusivamente

CL greeted effusively

I greeted her effusively

c. # La saludé a María efusivamente

CL greeted A Mary effusively

I greeted her effusively

[adapted from Sánchez 2005: 24]

This would mean that in (66b) we do have a topic, but not in (66c).

I will not attempt to decide this matter. I only want to observe that DOM is compatible

with all alleged interpretations (compare the sentences in (65). This means that DOM

does not affect the informational structure. Given that clitic-doubling does correlate with

the informational structure, we conclude that clitic doubling and DOM must be kept as

separate phenomena. See also López 2006 for the idea that DOM is not equivalent to

Germanic Object-Shift, which I interpret as indicating that DOM nominals are not

necessarily internal topics, a point of view that I will adopt here. This also means that

presence of the clitic alone (see (66b)).

135
theories that fit well with Germanic Object-Shift (as, for instance, Diesing 1992 and De

Hoop 1992) cannot be applied to DOM, as López 2006: 153-165 also concludes (see also

Lidz 2006 for the same conclusion with respect to DOM in Kannada).88

There is an additional notion that has also been related to DOM, in particular by Martín

2005, 2006. According to Martín, DOM objects have the ability to quantize the event.

Quantization is a concept proposed by Krifka 1989, 1992, 1998 to describe the

quantificational structure of nouns and events, which allows us to clearly spell out the

relation between the direct object (the theme) and the event. In Krifka’s framework, there

is a homomorphism between the structure of the theme and the structure of the event.

Chunks of the theme are assigned to parts of the event and when the theme has been

extinguished, the event is over. A nominal is quantized when it is possible to determine

its quantity, and event is quantized when bounded. The boundedness of the event must

not be confounded with its delimitation (or telicity, as in Tenny 1994). As pointed out by

Kiparsky 1998 (see also Ritter and Rosen 2001, 2005), a bounded event requires that the

participation in the event be absolute, which means that bounded events are not

compatible with adverbials expressing degrees of participation:89

88
In addition, the claim that clitic-doubling and DOM must be treated as separate phenomena casts some
doubts on the relevance of Kayne’s Generalization (Kayne 1975, Jaeggli 1982, 1986, Bleam 1999: 160-201)
for DOM:

(i) Kayne’s Generalization (KG)

An object NP may be doubled by a clitic only if the NP is preceded by a preposition.

In other words, it is not possible to suggest that A and clitic-doubling have the same source (for instance,
specificity), and that KG emerges as an expression of some kind of overt agreement. Additional, as
observed by Suñer 1988: 399-401, there are many cases of KG violations in several Spanish dialects (see
also Sánchez 2005).
89
This is not the same as saying that bound events are not compatible with durative phrases (as the
difference between in one hour/for an hour, which is used to test the delimitation of the event).

136
(67) a. # Mary owned the book very much.

b. # The customer bought the vase a bit.

c. # John recalled the password radically.

[Kiparsky 1998: 270]

Following Krifka, Martín 2005 assume that events can be quantized or cumulative:90

(68) a Cumulative

If a predicate P applies to x and y, it also applies to the sum of x and y,

provided that it applies to at least two distinct entities

b. Quantized

If a predicate P applies to x and y, y cannot be a proper part of x

In this sense, according to Martín 2005: 194, DOM objects (and only DOM objects) can

quantize the event, and thus, they are similar to subjects, which have this ability (Krifka

1989). This, as Martín notices, is a way to reestablish the hypothesis of distinguishability

(63). It is, however, not true that DOM objects quantize the event (but Martín’s

considerations hold for a subset of DOM objects, as we will see).

90
These definitions are informal. They can be formalized in this way:
(i) A predicate P is cumulative iff ∀x, y [ [P(x) ∧ P (y) Æ P(x⊕y)] ∧ card(P)≥2]
(ii) A predicate P is quantized iff ∀ x, y [P(x) ∧ P (y) Æ ¬y > y]
[From Martín 2005: 185, who adopts Krifka 1989]

137
First, DOM objects are perfectly compatible with adverbials expressing degree of

participation:

(69) a. Juan quiere un poquito a María

John wants a bit A Mary

John loves Mary a bit

b. Juan molestó muchísimo a María

John bothered very much A Mary

John bothered Mary very much

c. Juan hirió considerablemente a María

John hurt considerably A Mary

John hurt Mary considerably

This means that DOM objects do not necessarily quantize the event, since events in (69)

must be unbounded.

Secondly, it is perfectly possible to have quantized events with unmarked objects:

(70) a. Encontré (#muchísimo) el libro

Found (#very much) the book

I found the book (#very much).

138
 b. Probé (#un poquito) el teorema

Proved (# a bit) the theorem

I proved the theorem (# a bit).

c. La caja contiene (#considerablemente) dos relojes.

The box contains (#considerably) two watches.

The box contains two watches (#considerably).

This means that, contrary to Martin’s suggestion, unmarked objects can quantize the

event.

Additionally, plural objects can multiply the event (which is a sign of quantization), even

if they are unmarked:

(71) a. Vi los libros

Saw the books

I saw the books

b. Vi a los enfermos

Saw A the ill

I saw the patients

139
Contra Martin’s report (2005: 194), in both cases in (71) there can be more than one

event of seeing (at different times, for instance). Spanish DOM then seems to be

insensitive to quantization. This conclusion is compatible with findings regarding other

DOM languages (e.g. Turkish and Hebrew), which are also claimed to be insensitive to

quantization (as pointed out by Ritter and Rosen 2001: 441-442).

Interestingly, there are some cases where plural DOM objects seem to quantize the event

as well as cases where unmarked objects cannot do so. More interestingly, the former

cases are limited to DOM objects that receive A irregularly, that is, when they are [-

animate], illustrated by (62), repeated here as (72):

(72) a. Un adjetivo califica a un sustantivo

An adjective qualifies A a noun

An adjective qualifies a noun

b. Los días siguen a las noches

The days follow A the nights

The days follow the nights

[Torrego 1999: 1788]

In all these cases, if the object is plural, there must be multiple events. This is in fact

Martin’s original evidence (2005: 194) for his proposal, which, as we saw, cannot be

extended to all DOM objects. In addition, in (72) the sentences are not compatible with

140
adverbials expressing a degree of participation:

(73) a. Un adjetivo califica (# un poquito) a un sustantivo

An adjective qualifies (# a bit) A a noun

An adjective qualifies a noun (# a bit)

b. Los días siguen (# considerablemente) a las noches

The days follow (# considerably) A the nights

The day follow the nights (# considerably)

Moreover, as observed also by Martín 2005: 194, there are some verbs where the event

must always be [-quantize], and in these cases, the object can never get A, which is

further evidence in favor of the idea presented here:

(74) El huracán provocó (*a) las tormentas

The hurricane caused (*A) the storms

In (74), there cannot be multiple events of causation, and, as expected, the sentence is

compatible with adverbs of degree of participation:

(75) El huracán provocó hasta cierto punto las tormentas

The hurricane caused up to certain point the storms

The hurricane caused the storms up to a point.

141
Notice that this eliminates any possibility of using quantization to revive the notion of

distinguishability as the force behind DOM.

In section 2.3, I will explain the presence of A in (72), without giving up the idea that the

nominals in these sentences are [-animate]. I will show that there is a connection between

the presence of A and the mandatory quantization in these sentences.

Another notion that has been argued to be a factor behind Spanish DOM is affectedness

(Torrego 1998: 18-20). The evidence comes from contrasts like the following:

(76) a. Juan buscó (a) un ingeniero

John looked for A an engineer

John looked for an engineer

b. Juan golpeó *(a) un ingeniero

John hit A an engineer

John hit an engineer

In (76a), the A-marker can be dropped (subject to the conditions discussed in the

previous sections), but in (76b) it is mandatory (even if the object is understood as non-

specific). The difference between these sentences is the nature of the verbal predicate.

The object in (76b) is affected by the action of golpear (“to hit”). Affectedness is a notion

142
proposed by Anderson 1979, which has proved useful to capture a number of

grammatical phenomena (see Anderson 2006 for an overview). It implies that the object

undergoes a physical or psychological change, a condition that is not met in (76a).

Notice, however, that the mandatory character of A in (76b) is still subject to the animacy

constraint, that is, even with verbs that affect their object, A is not possible if the object is

inanimate:

(77) Juan golpeó (*a) un escritorio

John hit A a desk

John hit a desk

I contend that to account for these facts it is enough to assume that affecting verbs select

a D nominal as an object. Following the mechanism I will propose in section 2.3, which

is common for all other cases, this predicts a mandatory A with these verbs for [+animate]

nominals, but it does not allow the A with [-animate] nominals.

Before presenting this mechanism, I will review some previous accounts of Spanish

DOM.

143
2.2 Previous accounts

In this section, I will review two kinds of theories that have offered a comprehensive

account for Differential Object Marking in Spanish from a syntactic perspective. They

are Functional-OT and GB/Minimalism.

There is a significant amount of functional and typological literature dealing with Spanish

DOM, which I will not review here (see Molho 1958, Niculescu 1959, Roegiest 1979,

1998, Kliffer 1982, Weissenrieder 1985, 1990, 1991, Lazard 1989, García Martín 1992,

1996, Herrera Caso 1997, Delbecque 1998, García Martín 1998, Clements and Yoon

2006). However, the gist of the functional and typological analyses, that is, the

postulation of a set of hierarchies that correlate with the A-marking, has been

incorporated in a formal fashion in Functional-OT.91 These hierarchies are taken from the

typological literature (Silvernstein 1976, Croft 1988, Comrie 1989, Croft 1990, 2003).

There is also a prolific line of research stemming out from the notion of degree of

transitivity (Hopper and Thompson 1980)—see Næss 2004b for a nice overview and a

careful presentation of the main issues—which has made its way to Spanish (for instance,

Weissenrieder 1991, Clements and Yoon 2006, among others). This literature has made

an impact in formal frameworks like Optimality Theory, starting a productive venue to

describe DOM languages. For these reasons, I will choose functional OT as the

representative of this approach. The limitations we find here, discussed below, are also

91
It is worth noticing that this move is not always welcome in the typological literature—see Haspelmath
2004 for some criticism.

144
applicable to other models of this family.

2.2.1 The Functional-OT framework

Aissen 2003 presents an explanation for DOM that relies on a harmonic alignment of two

hierarchies or scales involved in the distribution of DOM. Harmonic Alignment (see

Prince and Smolensky [1993] 2004) is an operation of combining two scales in order to

obtain a set of constraints. The operation connects the highest element on a binary scale

with all the elements in other scale, from the highest to the lowest; in addition, it connects

the lowest element on the binary scale from the lowest to the highest. The scales in

question are given below (from Croft 1988):92

(78) RELATIONAL SCALE: Subject > Object

(79) a. ANIMACY SCALE: Human > Animate > Inanimate

b. DEFINITENESS SCALE: Pronoun > Name > Definite > Indefinite Specific > Non

Specific

After harmonizing (78) and (79a) with (79b), we obtain these sets of Harmonic scales,

respectively:

92
Notice that in the Animacy Scale, “Animate” actually means “Non Human Animate” (otherwise we
would oppose “Animate” to “Human”). With respect to the Definiteness Scale, we have to assume that a
phrase like the cat is “Definite” and also “Specific”, and a phrase like a cat is “Indefinite” but it could be
“Specific” or “Non Specific”. As we saw in section 2.1.1, this by itself constitutes a problem since definite
determiners can also be non-specific, which correlates with the A-marking. Aissen 2003 does not discuss
any of these notions, merely assuming the hierarchies.

145
(80) HARMONIC ANIMACY SCALES

a. Subject/Human ≻ Subject/Animate ≻ Subject/Inanimate

b. Object/Inanimate ≻ Object/Animate ≻ Object/Human

(81) HARMONIC DEFINITENESS SCALES

a. Subject/Pronoun ≻ Subject/Name ≻ Subject/Definite ≻ Subject/Specific ≻

Subject/Non Specific

b. Object/Non Specific ≻ Object/Specific ≻ Object/Definite ≻ Object/Name ≻

Object/Pronoun

The Harmonic scales express the markedness reversal between subject and object: what is

marked for objects is unmarked for subjects and vice versa. In other words, the scales

show the likelihood for a member of the Relational Scale (78) to be placed high or low in

the Animacy Scale (79a) or in the Definiteness Scale (79b). In that sense, the scales are

harmonic (or markedness) hierarchies, which can be treated as Avoid constraints, if we

reverse their rankings:

(82) ANIMACY CONSTRAINTS

a. *Subject/Inanimate >> *Subject/Animate >> *Subject/Human

b. *Object/Human >> *Object/Animate >> *Object/Inanimate

(83) DEFINITENESS CONSTRAINTS

a. *Subject/Non Specific >> *Subject/Specific >> *Subject/Definite >>

*Subject/Name >> *Subject/Pronoun

146
 b. *Object/Pronoun >> *Object/Name >> *Object/Definite >>

*Object/Specific >> *Object/Non Specific

These constraint hierarchies imply that a sentence will be penalized if its subject or its

object does not comply with the markedness reversal. Notice that DOM objects are

“marked” objects, that is, objects that do not comply with the markedness reversal. For

Aissen, the gist of DOM is that “marked” objects are also morphologically marked.93 To

encode this correlation, Aissen (2003: 447) proposes this constraint:

(84) STAR ZERO

*øC : Penalizes the absence of a value for the feature CASE.

*øC is compatible with a situation where all objects have a morphological mark of Case.

But we need to induce Case-marking just for the “marked” objects. It is necessary (i) to

connect *øC with the Animacy Constraints (82) and the Definiteness Constraints (83),

and (ii) to find a way to neutralize *øC in the relevant cases. In order to connect *øC with

(82) and (83), Aissen proposes to apply the operation of Local Constraint Conjunction

(Smolensky 1995). 94 The conjunction between *øC and the constraints (82) and (83)

results in a set of subject and object-oriented constraints that include *øC in each point of

the scale. Here, I will represent only the object-oriented constraints, since they are the

93
The reader should be aware that two senses for marked are being used here. In the sense of the
markedness reversal, “marked object” means that the characteristics of the object are unexpected (for an
object). In the morphological sense, it simply means that there is an overt marking accompanying it
(preposition, case ending, etc).
94
According to this operation, by conjoining two constraints C1 and C2 in a domain D, we obtain the
constraint C1 & C2, which is violated if both C1 and C2 are violated.

147
ones relevant for DOM.95

(85) a. LOCAL CONJUNCTION OF *øC WITH THE ANIMACY CONSTRAINTS

*Object/Human & *øC >> *Object/Animate & *øC >> *Object/Inanimate &

*øC

b. LOCAL CONJUNCTION OF *øC WITH THE DEFINITENESS CONSTRAINTS

*Object/Pronoun & *øC >> *Object/Name & *øC >> *Object/Definite & *øC

>> *Object/ Specific & *øC >> *Object/ Non Specific & *øC

[Aissen 2003: 448]

Since this ranking enforces case marking for all objects, we need a way to neutralize *øC.

To achieve this, Aissen posits an additional constraint:

(86) *STRUCC: Penalizes a value for the morphological category CASE.

[Aissen 2003: 448]

*STRUCC can be inserted at any point in the ranking established by (85), obtaining the

different cutting points that we see in various DOM languages. Obviously, this system is

not yet ready for Spanish DOM, because it can only apply to languages where just one

dimension (Animacy or Definiteness) is used for DOM. To make the system suitable for

two dimensional languages, Aissen conjoins the Animacy Constraints (82) and the

95
Keep in mind, however, that there are languages that exhibit Differential Subject Marking (DSM),
although they are far less common (Aissen 2003: 472-474). Of course, a question arises immediately
regarding why DSM is so rare. Notice that nothing prevents a local conjunction with the subject-oriented
constraints.

148
Definiteness Constraints (83). As in the one-dimensional cases, the resulting constraints

need to be conjoined with *øC, obtaining constraints of this form:

(87) [*Object/Human & Object/Pronoun] & *øC

This yields a huge set of possibilities for a cutting point, that is, the point where *STRUCC

can be inserted. An additional complication arises: according to Aissen 2003, in some

constructions DOM is optional, then, *STRUCC can be reranked at will, at least in some

portions of the ranking. This is then her proposed hierarchy for Spanish:

(88) DOM in Modern Spanish96

[Adapted from Aissen 2003: 472]

According to (88), there are three situations relevant to Spanish DOM. First, A is

96
In this figure, a constraint like:
i. [*Object/Human & Object/Pronoun] & *øC
is abbreviated as follows:
ii. *Oj/Human-Pro & *øC

149
mandatory where the constraints always dominate *STRUCC; in these cases, the object

must be: Human Pronoun, Human Proper Noun, Animate Pronoun (non-human), Human

Definite, Animate Proper Noun (non-human), or Human Specific (non-definite). Second,

the marking is optional when *STRUCC and the constraints can be reranked; in these cases,

the object is Inanimate Pronoun, Animate Definite (non-human), Inanimate Proper Noun,

or Animate Specific (non-human, non-definite). Finally, the marking is impossible if

*STRUCC always dominates the constraints; in these cases, the object should be: Human

Non-Specific, Inanimate Definite, Animate Non-Specific, Inanimate Specific (non-

definite), or Inanimate Non-Specific.

Given that definite determiners can also be non-specific, as discussed in section 2.1.1, the

fact that the A-marker is optional in such cases is not covered by (88); the same happens

with kind-denoting nominals (and these problems cannot be solved by adjusting the

position of *STRUCC: it is necessary to redefine the components of the Definiteness Scale).

Additionally, animate non-specific nominals can receive A optionally, but A is predicted

to be absent by (88) (although in this case it will be enough to adjust the position of

*STRUCC). Furthermore, as discussed in section 2.1.2, with some verbs, inanimate objects

must receive the A-marker, but (88) predicts this situation to be impossible (and there is

no way to encode this information in (88)).

Aissen acknowledges that (88) “oversimplifies [the data] in some respects” (2003:

471)—this situation is not limited to Spanish DOM: her proposal for Hindi has the same

empirical limitations, as she acknowledges. Nevertheless, she thinks that the facts “are

150
consistent with the basic hypothesis: if overt marking is possible with direct objects with

property α, then it is possible with direct objects with property β, where β dominates α [in

(88)]” (2003: 468). However, as should be apparent, this cannot be true if some inanimate

objects (which are in the low portion of (88)) receive a mandatory A, but some definite

objects (located in the high part of (88)) receive an optional A.

In fact, the whole issue of markedness reversal can be called into question, as has already

been noted, even in the OT framework (de Swart 2003), strictly on a empirical basis—see

also Filimonova 2005 for other languages where the marking does not comply with the

expectations of the hierarchies, which means that they will challenge (88) too.

Næss 2004a also points out that, if we follow the logic of markedness reversal, some very

well established mechanisms, like incorporation of objects (Baker 1988 and many others),

are unexpected, since this process usually involves “demotion” of “typical” objects. In

other words, incorporation deprives prototypical objects of “objecthood”, contrary to

what we would expect. In that sense, the notion of “more appropriate object” becomes

meaningless. In fact, Næss also suggests that DOM is not a mechanism to mark that the

object is more definite or more animate, but, rather simply the fact that it is affected. As it

is well-known, affectedness is a property of the participants in an event, which correlates

with animacy and definiteness; this means that DOM has to be understood from the point

of view of the event structure too, a point very often disregarded in the literature—see

however Torrego 1998, Ritter and Rosen 2001, 2005, Martín 2005, 2006.97

97
This is, of course, not true regarding the functional literature (see Næss 2004b for a careful review and
some innovative proposals).

151
Regarding later OT models for DOM (like de Swart 2003 and Morimoto and de Swart

2006), they add to the picture the notion of distinguishability, and propose a constraint

that interacts with Aissen’s system, which we already discussed (see (63), repeated here

as (89)):

(89) DISTINGUISHABILITY

Mark objects that are not outranked by the subject in prominence (animacy,

specificity).

[Morimoto and de Swart 2006: 232]

As we discussed in section 2.1.2, (89) is also empirically wrong for Spanish. The reader

is referred to Carnie and Jelinek 2003, Carnie 2004, 2005, Isaak 2000, Newmeyer 2002,

Haspelmath 2004, among several others for discussion of the empirical limitations of the

functional framework with respect to other DOM languages.

2.2.2 The GB/Minimalist framework

Spanish DOM (and DOM in general) has attracted little attention in the GB/Minimalist

framework. Although the issue is often addressed when dealing with related

phenomena—in particular clitics (Strotzer 1976), clitic doubling (Suñer 1988, Franco

1993, Bleam 1999), ditransitives (Strotzer 1976, Demonte 1987, Cuervo 2003),

information structure (López 2006), subjects (Fernández-Soriano 1989, Contreras 1996),

152
pronouns (García 1975) and reflexives (Schroten 1972)—and it is always noted in broad

reviews of Spanish syntax from a generative perspective (see Demonte 1989 and Zagona

2002, to mention a couple that offer some fresh perspectives on the issue)—there are

actually only three comprehensive studies of Spanish DOM from a formal-theoretical

syntactic perspective: Isenberg 1968, Brugé and Brugger 1996, and Torrego 1998, and

only the last two are from a GB/Minimalist perspective. 98 I have referred to them a

number of times in this chapter, and I will discuss their key theoretical claims in the

following sections. Before that, let me point out that there are actually many important

descriptive and theoretical articles that deal with DOM from a generative point of view,

which have been in fact mentioned earlier in this chapter. However, they do not attempt

to encompass the whole phenomenon.

I should also add that, in recent years, there have been some attempts to address Spanish

DOM from a purely formal semantic perspective, inside the generative tradition (see

Agüero-Batista 2005, Bleam 2005). This is certainly a welcome venue of research, and I

do not aim to contribute anything new to it; however, it must be pointed that so far this

tradition deals only with animate objects where A is mandatory, putting aside the cases

where it is optional (recall from our discussion in the previous sections that A is optional

with animate non-specific objects). The issue of inanimate objects with A is also not

addressed in this line of research.

98
I will not discuss Isenberg 1968, since it is not based on GB (see Schroten 1972, Laca 1995, and Pensado
1995 for some criticism of this approach).

153
2.2.2.1 Brugé and Brugger 1996

Based on the Empty Category Principle (ECP) and on the notion of Partitive Case

(Belletti 1988), Brugé and Brugger 1996 propose that the trigger for DOM is a [+animate]

feature in nominals. They assume that the A-marker is hosted by a functional projection

FP, which is part of the extended projection of the nominal. The A-marker is specified for

[+accusative] and [+animate], which restricts its presence to animate objects. Non-

animate objects cannot receive A in their system; in addition, specificity, according to

them, plays no role in the distribution of Spanish DOM. The structure they propose for

transitive sentences is the following:

(90) …VP

V FP

F’

F DP
(A)

D’

D NP

In addition, Brugé and Brugger 1996 assume the following version of the ECP (from

Cinque 1990: 49):

154
(91) Empty Category Principle

A non-pronominal empty category must be properly head-governed by a head

nondistinct from [+V].

The A-marker does not count as a proper lexical governor, since it is distinct from [+V],

but, if present, it does count as in intervener, preventing the verb from properly governing

an empty D0. This accounts for the absolute lack of A with bare nouns, which we already

saw in section 2.1.1. Since they assume that A is mandatory if F is [+accusative] and

[+animate], animate bare nouns must receive Partitive Case (in the sense of Belletti 1988).

Given that, when modified by other phrases, determinerless nouns can have A, Brugé and

Brugger 1996: 22 assume that the modification strips an empty D0 from its status of

empty category, by supplying some abstract features (as suggested by Longobardi 1994),

or by allowing N-to-D movement (as proposed by Delfitto and Schroten 1991).

Given that theirs is not the most common view on the A-marker, Brugé and Brugger 1996

spend some time showing that A cannot be a marker of specificity, since it is compatible

with [+specific] readings. I have discussed this issue at length in section 2.1.1, adding

some new evidence to this effect. I am thus siding with these authors in this particular

point, although I do not accept that specificity is not a factor in the distribution of Spanish

DOM. As discussed in section 2.1.1, the A-marker can appear with non-specific objects,

but it is optional in those cases, and only in those cases, a generalization for which Brugé

and Brugger 1996 have no account. In addition, they distinguish kind-interpretation from

object-interpretation, claiming that kind interpretation can freely be [± animate], which

155
accounts for the possibility of dropping A with kinds; as we discussed earlier, this idea is

highly counterintuitive.

Although they account for a broad variety of cases, it is important to notice that Brugé

and Brugger 1996 do not discuss cases where inanimate objects do receive A, which is a

major empirical limitation of their study. I will, however, adopt the view that [+animate]

is the initial trigger of Spanish DOM, and I will offer an account for the problematic

cases below. In addition, we have to notice that, once the ECP is out of our theoretical

array (see Chomsky and Lasnik 1993), we lose the main theoretical mechanism of their

account.

2.2.2.2 Torrego 1998

The most comprehensive study of Spanish DOM is, without doubt, Torrego 1998. With

the conception of the Syntax-Semantics interface based on Diesing 1992 and Hale and

Keyser 1993, Torrego 1998 presents the idea, which I will also adopt here (although with

very different motivation and technical details), that A-objects are higher than unmarked

objects. For her, A-marked objects undergo overt raising to the specifier of vP, over base

postion of the subject:

156
(92) vP

OBJECT v’

SUBJECT v

V OBJECT

According to Torrego, this configuration accounts for the properties she attributes to A-

marked objects: they have to be specific and animate, and they delimit the event. For her,

the secondary specifier of small v is a privileged position, in three different senses. First,

it is the locus of specificity (as in Diesing 1992), that is, only specific objects can raise

and any raised object must be specific. Second, extending an idea from Marantz 1984

regarding subjects, she assumes that in this position the object receives a secondary

agentive role, which is compatible only with animates. Finally, adapting suggestions from

Marantz 1990, Travis 1992 and Hale and Keyser 1993, she claims that A-objects have the

ability to delimit the event (telicity), a point to which I will return.

Furthermore, she assumes that v has a D feature that attracts the marked DO, and that A

also has a D feature to check against v. She also indicates, as an alternative, that there is a

[person] feature involved in the checking; although she does not elaborate on the details

of the latter checking, she suggests that the [person] feature may be the reason for the

animacy restrictions.99 Notice that for Torrego, the raising of the object happens for two

99
It is worth noticing that there exists a long standing trend in Spanish traditional grammar that associates
some notion of Person to A-marked objects; in fact, the marker is often called “the personal A” in this

157
different reasons. On one hand, there is a particular position reserved for a subset of

objects, namely, specific and animate objects, which means that the objects in question

must raise there. On the other hand, A-objects have to check a D feature, which requires

them to raise to [Spec, vP]. Both A-marked and unmarked objects check structural

accusative case against small v, but only the former raise.

Additionally, Torrego distinguishes two kinds of A-markers. According to her, affected

A-marked objects receive inherent case from their verbs, in addition to the structural case

they check, that is, they are quirky objects (Torrego 1998: 23-25 and 43), whereas non-

affected A-marked objects receive only structural case. Although this accounts for the

mandatory presence of A with affecting verbs (irrespective of the specificity of the

object), the possibility of inherent case for affected objects must be relativized to animate

nominals, since non-animate objects do not receive A, even if affected, as discussed in

section 2.1.3.

According to Torrego 1998, telicity also plays a role in Spanish DOM.100 She bases her

claim on contrasts like (93), where the sentence with the A-object (93a) is ambiguous

between an accomplishment and an activity, but (93b), without A, can only be an activity:

literature (see Bello 1847, Cuervo 1881, Fernández Ramírez 1986, Pensado 1995 and several others). It is
necessary to point out that “Person” in this tradition means an individual, that is, a particular person or
animate being (although the notion is not fully elaborated). I think Torrego is aiming to link herself with
this line of thought, but she does not make this explicit. As should be evident from chapter 1, I claim that
the feature [person] plays a crucial role in Spanish DOM, which is a way to incorporate this tradition into a
formal framework.
100
As discussed in section 2.1.3, the boundedness of the event (quantization) must not be confused with its
delimitation (telicity).

158
(93) a. Laura escondió a un prisionero durante dos años.

Laura hid A a prisoner during two years

Laura hid a prisoner during two years.

b. Laura escondió un prisionero durante dos años.

Laura hid a prisoner during two years

Laura hid a prisoner during two years.

(93a) is ambiguous between a repetitive act of hiding (an activity) and a unique act of

hiding (an accomplishment). (93b) has only the repetitive reading. I entirely agree with

this judgment (as do other native speakers I have asked); however, I think that there is no

need to invoke telicity to account for the presence of A in (93a). On the activity reading,

the object is [+specific], whereas on the accomplishment reading the object is [-specific].

This is enough to obtain the difference. Note that, since the object in (93a) is ambiguous

between specific and non-specific, this is additional evidence that A cannot be a marker

of specificity, as already concluded in section 2.1.1. It is also worth noticing that Torrego

has to appeal to an additional factor in this case, since for her all A-marked objects must

be specific. Notice also that I do not mean to say that there is no a relation between the A-

marker and a telic difference ((93) shows that), but that this relation is mediated by

specificity. This is by no means a novel idea, since the relation between specificity and

telicity has been recognized for a long time (at least since Verkuyl 1972).

Torrego also links the presence of A in DOM with the dative preposition that Indirect

Objects receive (which is also A). According to her, what unifies the two is the

159
quirky/lexical/inherent nature of Dative case. Notice that she does not mean that DOM-

objects are Dative, only that, like Dative objects, DOM-objects can receive inherent case

(but they don’t all have it). However, she does suggest a relation between the dative

preposition P and small v, by proposing that small v selects an intermediate projection

headed by a null Preposition P, which, in later work (Torrego 2002), she identifies with

the dative preposition that incorporates into BE to create HAVE, discussed by Freeze

1992 and Kayne 1993:

(94) [vP SUBJECT v [PP P [VP V A-OBJECT ] ] ]

[Torrego 2002: 346]

Torrego uses this idea to explain a correlation that she establishes between clitic-doubling

and the alleged behavior of causative constructions in some Spanish dialects. According

to Torrego 1998, 2002, South Cone Spanish speakers do not accept both sentences in (95):

(95) a. La felicitamos a la pianista

CL-her congratulate A the pianist

We congratulated the pianist.

b. Ese guardia hizo al chico guardar la botella.

That policeman made A+the kid to put away the bottle

That policeman made the kid to put away the bottle.

[Torrego 2002: 347-348]

160
The implication is the following: if a speaker can accept the clitic doubling in (95a), then

he/she cannot accept the preverbal causee in (95b). According to Torrego, this difference

is produced because in these dialects, the vP does not select the abstract preposition P, but

a DP that can host the clitic (for this reason, there can be a clitic in these dialects):

(96) [vP SUBJECT v [DP D-clitic [VP V A-OBJECT ] ] ]

However, although Limeño Spanish does not actually count as South Cone (a term used

to refer to both Argentina and Chile), in this dialect sentences like (95a) are also accepted

(as reported by Sánchez 2005: 14, Mayer 2006), but (95b) is equally acceptable. This

means that the fact that some dialects (as Peninsular Spanish) cannot have (95a) must be

explained by appealing to the specific properties of clitic doubling, independently of

DOM. This reinforces my previous conclusion that clitic doubling and DOM are separate

phenomena.

There are additional problems for Torrego’s account. She assumes that all A-marked

objects have to be [+specific], but, as discussed in section 2.1.1, this is not true. Notice

that there is no obvious way in her system to accommodate the fact that A-marked

objects can be non-specific, since she relates the presence of A to the idea that there is a

dedicated position for specific objects.

Additionally, if the reason for the animacy restriction is that there is a secondary agentive

161
role in [Spec, vP], then it is predicted that all A-marked objects must be [+animate],

which is contradicted by the facts (as shown in section 2.1.2), Torrego simply assumes a

[+animate] feature in all marked objects, which means that [± animate] cannot be a

semantic feature (Torrego 1998: 55). Putting aside its circularity, the proposal is not only

counterintuitive, but it also undermines her system, since it breaks the semantic

correlation between the secondary agentive role and the (now non-semantic) [+animate]

feature.

Although I will not adopt Torrego’s account of Spanish DOM, it is important to note that

my approach is Torregian in its essence: DOM objects are also higher than unmarked

objects in my analysis. I will also incorporate an important component of Brugé and

Brugger 1996, namely the assumption that animacy is the initial trigger for DOM. To do

that, I will use the idea developed in chapter 1 that small v is φ-incomplete in all Spanish

transitive sentences, and therefore that φ-complete nominals cannot check case against it.

I will assume that all animate nominals receive a [person] feature, but that this feature is

deleted under certain conditions (depending on the D/*D distinction discussed in section

2.1.1), which will allow me to accomodate the specificity facts discussed earlier in this

chapter. The result of the system is that only φ-complete objects can receive the A-

marker.

162
2.3 Agree and DP structure

2.3.1 The main mechanism: raising by Agree

As outlined in chapter 1, I assume that the main mechanism for DOM is the operation

Agree (Chomsky 2000, 2001, 2004, 2005b, a, 2006, Bošković 2005, 2007b, a). Under

Bošković 2007b, a version of the Agree system, where movement is driven by an

uninterpretable feature (uK) in the goal G, there are two types of movements that are

related to Agree. I will call these movements: Blind Movement and Checking Movement.

(97) Blind movement

If a uK of a phrase XP is not valued/checked in a given linearization domain (the

complement of a phase PH), XP must move to the edge to avoid been frozen from

further movements.

(98) [PH XP [PH’ XP ] ]

uK

(99) Checking movement

If a Probe P matches the interpretable feature (iF) of a Goal G, G having a uK, G

must move to the closest position c-commanding P to check/value its uK.

(100) [ G [ P [ G ]]
iF uF
uK K

163
Notice that Checking Movement implies that, in order to complete the operation Agree,

G and P actually c-command each other (see Bošković 2005, 2007b, a).101 At this point, it

is not important if K in P is interpretable or not. Both alternatives are possible, as long as

we assume that K in P is valued. If there is an iK, this is the standard assumption; but

there is an alternative. Pesetsky and Torrego 2007 show that uninterpretable features can

be valued too; that is, there is no biconditional relation between the interpretability of

features and the valuation of the features; under this view, there can be four types of

features:

(101) a. [uF: ] : uninterpretable, unvalued

b. [uF: α ] : uninterpretable, valued

c. [iF: ] : interpretable, unvalued

d. [iF: α ] : uninterpretable, valued

Since this alternative exists—and it is perfectively compatible with the Agree system, as

highlighted by Pesetsky and Torrego 2007 who offer a variety of phenomena that

instantiate the possibilities in (101)—the only thing needed is that K in (100) is valued; it

can be interpretable or uninterpretable.

At any rate, if K in P is valued, the checking/valuation of uK in G follows the same

procedure as checking/valuation of uF in P, that is, Agree, which means that it should be

subject to the same conditions, which, although not made entirely explicit in Bošković

2007b, a, can be naturally accommodated in his system.

101
Only if P has a uF; without it, this is not possible.

164
I further assume that, for Spanish, the initial trigger for DOM is the animacy of the object.

Animacy, however, is a semantic property of nominals, and by itself does not play a role

in the Case/Agreement system. However, as it has been soundly established in the

literature, there is a correlation between animacy and Case/Agreement—the exact effects

of animacy on the structure will be discussed in the next subsection. I propose to

represent this correlation using the feature [person] for the reasons discussed in chapter

1—in sum, because the elements that uncontroversially hold this feature, namely personal

pronouns, are always animate. In that sense, the proposal is similar to the one made by

Kratzer 2006, since she assumes that phi-features do not actually carry an interpretative

force, but they are in the DP to negotiate Case and Agreement. Since this does not imply

that φ-features cannot have a value, nothing changes if we assume that this [person]

feature is uninterpretable, which would be in line with the system outlined by Pesetsky

and Torrego 2007, who assumes that interpretation and valuation do not correlate with

each other, as just seen in (101). However, I will continue using the traditional notation,

mainly for expository reasons (for instance, a [iperson:-1-2] in the nominal).

I further assume that Agree is governed by a principle that maximizes matching effects

(for discussion of the various effects of this principle see McGinnis 1998, Chomsky 2001,

Béjar 2003, Rodríguez-Mondoñedo 2006d, Bošković 2007b, a, among others):

(102) Maximize matching effects

[Chomsky 2001: 15]

165
It is important to stress that I do not interpret (102) as a preference principle, but as

requiring simultaneous matching of all uninterpretable feature. In that sense, (102) could

be restated in this way:

(103) Match only once.

If P does not match all the uninterpretable feature of G, G must wait to check its features

until all the Ps that match G’s uninterpretable features are inserted into the structure, and

then match all of them at once by Checking Movement (that is, by probing its Ps; G

would then probe all the Ps in (104) only after raising to its final landing site):

(104)

G P3 P2 P1 G
Z Y K uK
uY
uZ

Notice also that, if there is a phase before all the relevant Ps enter the structure, G must

move to the edge of the phase, although it would not check any of its uninterpretable

features (Blind Movement). It would check all its features at once after the relevant Ps

have been introduced (Checking Movement):

166
(105)

G P2 [PHASE G P1 [PHASE COMPLEMENT G ]]

In addition, it is in principle possible that lower P1 (which is valued by G) is not able to

value a single feature of G (under certain conditions to be established). When that is the

case, G must raise out of the phase complement looking for another probe P2 (notice that

the features of P1 can be valued when P1 matched (and probed) G, before G undergoes

movement:

(106)

G P2 [PHASE G P1 [PHASE COMPLEMENT G ]

We will see immediately that Differential Object Marking instantiates the possibility

depicted in (106).

As explained in chapter 1, I assume that with all transitive verbs in Spanish, the small v

has only [number] and no [person]:

(107) v
[number]

That means that in Spanish only nominals that are not specified for [person] will check

167
case against small v. If an internal nominal is specified for person, small v will be able to

probe it and value its own [number] feature, but it will not be able to value the [case]

feature of the object. This comes from (102), or more specifically, from the corollary of

(102) that we discussed in chapter 1 (based on Béjar 2003):

(108) Condition on Case-Valuation

Only a probe P that matches all the relevant features of a goal G can value the

[case] feature of G.

As a result, a [person] nominal cannot value its Case against small v. If a nominal cannot

check its Case inside vP, it must move out to avoid being spelled-out with an unvalued

feature, creating a new specifier of vP to escape (Bošković 2007b, a); here is a schematic

picture of the process, as briefly discussed in chapter 1:102

102
In (109), the discontinuous line indicates valuation under Agree, with the higher element functioning as
the probe. Notice also that the nominal establishes an Agree relation with the Dat head, and moves to c-
command it in order to check case.

168
(109) DATP

A-nominal DAT’
[inum: PLU]
[ipers: 3P]
[ucase: DAT] DAT vP
[unum:PLU]
[upers: 3P ]
[case: DAT] nominal v
[inum:PLU]
[upers: 3P ]
[ucase: ]
v VP
8 [unum:PLU]
[case: ACC]
V nominal
besó [inum: PLU]
CHECKING [ipers: 3P ]
MOVEMENT [ucase: ]

BLIND
MOVEMENT

This is what happens in Spanish: [person] nominals move via [Spec, vP] to an additional

head, which gives them Dative Case, which I assume is manifested by the A-marker.103

Recall for Bošković 2007b, a, there must be a c-command symmetry between probe and

goal in the case under consideration here: the probe must c-command the goal for φ-

valuation, but the goal must c-command the probe for case-checking; more precisely, the

elements in question probe each other (i.e. they are both probes and goals). This implies

that the goal has to move to the specifier of its case-checker. A [person] object then has to

c-command the DAT head. On the other hand, φ-incomplete objects (i.e. the ones that are

not specified with [person]) only need to c-command small v.

103
For the complete details of this process, see the discussion of (126).

169
This analysis maintains a difference in height between the landing sites of the two types

of objects, implementing it by placing [person] objects in the specifier of the DAT head,

while non-person objects are located in [Spec, vP] (so, unlike non-person objects, [person]

objects do not stop in [Spec, vP], but raise further). Notice that (109) instantiates (106), a

situation that is produced by the Condition on Case-Valuation (108), under the

assumption that small v is φ-incomplete. That is, the lower probe must be ignored

because it is not capable of checking the case-feature of the goal.

As mentioned, I interpret the maximization condition in (102) as requiring that Agree

values features of G only when P fully matches the features of G (an interpretation that is

also used in (108), for case-valuation); that is, G can get a value only when G is getting a

value for all its features. Given this, the combination of Blind Movement (97), Checking

Movement (99) and the maximization of matching effects (102) makes two interesting

predictions. Bošković 2007a spells out the first one: if G has more than one unvalued

feature, it must wait until all Ps that hold the relevant uninterpretable features have

entered the structure before undergoing Checking Movement: G gets a value only at the

point at which G probes its Ps, that is, when G undergoes Checking Movement. This is

expressed in (104). As shown in (105), this situation does not affect Blind Movement: if a

phase-level is reached before G can get a value for its features, G will undergo Blind

Movement—this is, in fact, the way successive cyclic movement is implemented in the

system (see Bošković 2007a).

170
Given that the uK of G is valued/checked under Agree, as previously assumed, then if G

has several unvalued features, and, as explained above, moves to check the features, i.e.

undergoes checking for all the features at the same point, a second prediction arises. It is

possible that a higher P2 could have a K feature with a value that is different from the

value of the K feature in a lower P1 (which was the first one that probed G). If this

happens, the prediction is that G will have a value that corresponds to the higher P2, and

not to the lower P1. Let me illustrate this.

Assume the following:

(110) i. one goal G with two unvalued features uK and uH

ii. two probes P1 and P2 with [K: α] and [H: β] respectively

iii. P2 has also a [H: γ]

Filling the valued features gives us the following feature specification for G and Ps:

(111) a. P1 G
[uF: ] [iF: δ]
[K: α ] [uK: ]
[uH: ]

b. P2 P1 G
[uF: ] [uF: ] [iF: δ]
[K: γ ] [K: α ] [uK: ]
[H: β ] [uH: ]

Suppose P1 enters the structure first. Then, under (102), G cannot move to probe P1 (at

point (111a)), because P1 does not match all the features of G. Given this, either G stays

in situ or G undergoes Blind Movement to the specifier of P1, depending on whether or

171
not P1 heads a phase. Notice that P1 does get a value for its uF under Agree (by probing

G), because it does match the features of G maximally, since maximization is a one-way

relation, as discussed in chapter 1:

(112)

P1 …
[uF: δ ]
[K: α ]
G
[iF: δ]
[uK: ]
[uH: ]

Notice that, if G undergoes Blind Movement, G will c-command P1, but it will not probe

P1, because of (102).

When P2 enters the structure, G moves to a position c-commanding P2, as a result of

which it also c-commands P1. However, given that P2 also has a K, and K in P2 has a

value (γ) that is different from K in P1 (α), the uK in G will be γ and not α, because P2 is

closer than P1. This is illustrated below:

172
(113) P2 enters the structure

P2 …
[uF: ]
[K: γ ]
[H: β ] P1 …
[uF: δ ]
[K: α ]
G
[iF: δ]
[uK: ]
[uH: ]

As discussed above, if P1 heads a phase, then G will undergo Blind Movement to the

specifier of P1, and the condition on maximization will prevent any feature valuation at

this point:

(114)

G
[F: δ]
[K: ]
[H: ]
8 P1
[uF: δ ]
…

[K: α ] G
[iF: δ]
[uK: ]
[uH: ]

When P2 enters the structure, its uF gets valued by probing G:

173
(115)

P2 …
[F: δ]
[K: γ ]
[H: β ] G
[iF: δ ]
[uK: ]
[uH: ] P1 …
[uF: δ ]
[K: α ]
G
[iF: δ]
[uK: ]
[uH: ]

After (115), G undergoes Checking Movement, and then its uninterpretable features will

get valued by P2 and (crucially) not by P1:

(116) G moves to c-command P2 (Checking Movement)

G
[F: δ]
[K: γ ]
[H: β ] P2 …
[uF: δ ]
[K: γ ]
[H: β ] P1 …
[uF: δ ]
[K: α ]
G
[iF: δ]
[uK: ]
[uH: ]

The theoretical possibility described above has empirical support. I claim that this is

precisely the case of non animate nominals receiving A, and causing the quantization of

174
the event. Recall from the discussion of (72)-(73) that there are some verbs whose object

must quantize the event (one of the arguments concerned the fact that they are not

compatible with adverbs expressing degree of participation). In such cases, regardless of

any other factors, A is mandatory:

(117) a. Un adjetivo califica (# un poquito) *(a) un sustantivo

An adjective qualifies (# a bit) A a noun

An adjective qualifies a noun (# a bit)

b. Los días siguen (# considerablemente) *(a) las noches

The days follow (# considerably) A the nights

The day follow the nights (# considerably)

Adapting ideas from Ritter and Rosen 2001 and Martín 2005 (who in turn use ideas from

Krifka 1989, 1992), I assume that when an event must be quantized, there is a projection

XP whose head X has an interpretable [quantize] feature with a value ‘yes’: [iquant: YES].

I understand this to mean that a head with the feature [iquant: YES] has the ability to

quantize the event using the nominal in its specifier. I further assume that X has a Dative

feature: [case: DAT]. This implies that, with verbs like those in (117), where the event is

always quantized, this XP must always be present. The object must raise to [Spec, XP] to

obtain the quantization. To achieve the raising, I assume that the nominal also has a

[quantize] feature, which is interpretable but unvalued: [iQUANT: ], a possibility

explored by Pesetsky and Torrego 2007 on independent grounds (as we already discussed

175
with respect to (101)).104 Then, if we add the φ-features, the X in XP has the following

feature structure:

(118) X
[uperson: ]
[unumber: ]
[iquant: YES]
[case: DAT ]

In turn, the relevant nominal has the following feature structure (if inanimate):

(119) DP
[inumber: α ]
[iquant: ]
[ucase: ]

This gives us all the ingredients we need to derive sentences like (120), which exhibit the

characteristics observed—see the discussion of (72)-(73):

(120) Los días siguen a las noches

The days follow A the nights

The day follow the nights

In (120), [las noches] is a DP that does not have a [person] feature since it is not animate.

Therefore, in principle, [las noches] should be able to check ACC case against small v.

104
Perhaps [quantize] does not need to be interpretable in the nominal (but it can be, since nominals can be
quantized independently of the event, as argued by Krifka 1989, 1992, and many others); if it does not have
to be interpretable, the nominal will have a more familiar [uquant: ]. As mentioned above, for the system
developed here, it is enough that the feature does not have a value.
We can also ask what [iquant: no] would mean; in particular if it is in any respect different from the plain
lack of the feature. If it is not, then it can be dispensed with, and then [quantize] will be a feature with a
unitary value (namely ‘yes’), a possibility that it is not at all strange in feature systems, and my analysis
would remain unchanged.

176
However, given that it has to quantize the event, [las noches] must have a [iquant: ]

feature. I contend (adapting an idea from Martín 2005) that verbs like the ones in (120)

select a nominal with a [quantize] feature and, consequently, have to have a phrase in

their extended projection (which is XP from above) to value this feature of their object.

Notice that this feature is not freely assigned to nominals, that is, not all nominals have it,

since we do not see its effects everywhere. I assume that this feature is hosted in the

highest functional projection of the nominal. As I will show in section 2.3.2, this

projection is KP. This means that K comes in two flavors: with the [quantize] feature and

without it. When needed, K enters the numeration with a [quantize] feature.

In the framework I have developed so far, this means that [las noches] cannot check ACC

case against small v, because small v cannot value nominal’s [iquant: ] feature.

Remember that we interpret the maximization condition in (102) as requiring that Agree

can value features only when the probe fully matches the features of the goal (see also

(64) in chapter 1). This requirement is not met here: being inanimate, the nominal does

not have [person], but small v is still incomplete with respect to the object since small v

does not carry a [quantize] feature. Since vP is a phase, [las noches] undergoes Blind

Movement to [Spec, vP]. It will further raise to [Spec, XP], when X enters the structure.

From that position, given that X is higher than v, the value that [las noches] will get for

its case feature is DAT and not ACC, which explains the mandatory presence of the A-

marker in (120). A remaining issue is how the [person] feature of X is valued. We have

already encountered this situation in chapter 1.

177
Recall that we noted in chapter 1 that φ-valuation is forced when the G receives a case

value from its Probe. I expressed this with Condition (64) in chapter 1, which I repeat

here:

(121) If a Goal G receives a [case]-value from a Probe P, it must value all the

uninterpretable φ-features that the P is matching.

This allows the valuation of the [number] feature in X; however, it leaves the [person]

feature unvalued. As discussed in chapter 1, a mechanism of default agreement has to be

assumed to repair heads that have incomplete valuation. In other words, if one φ-feature

needs to be valued—in this case, the [number] feature, as required by (121)—the

remaining unvalued φ-features will receive default value. In the case under consideration,

this means that the [person] feature in X will receive default [3p]. This is parallel to what

happens with φ-incomplete nominals when they check case against T, discussed in

chapter 1, which does not cause any mismatch, since [1p] and [2p] nominals are always

φ-complete, so default agreement will not need to be invoked in those cases.

The derivation for (120), including default agreement for [person] in X, is expressed in

the following partial structure:

178
(122)

XP

a-las noches X’
[inum: PLU]
[ucase: DAT]
[uquant: YES] X vP
[upers: 3P ]
[unum: PLU]
[case: DAT] las noches v
[iquant: YES] [inum: PLU]
[ucase: ]
[uquant: ]
8 v VP
[unum: PLU]
[case: ACC]
V las noches
siguen [inum: PLU]
CHECKING [ucase: ]
MOVEMENT [uquant: ]

BLIND
MOVEMENT

It is important to stress that this is exactly the situation predicted by the Agree system,

discussed above regarding (116). Given that the lower probe P1 (in this case v) is

incomplete with respect to the goal G, the valuation of G’s features must wait until P2 (in

this case X) enters the structure; at that point, G raises to the specifier of P2 (X) and

values its features against the higher probe P2, not P1 (given that it is closer). For (120),

this means that G (las noches) will get DAT and not ACC.

The system outlined so far allows us to make another prediction. As discussed above, the

verbs like the ones in (72) or (120) must have an XP projection with a [quantize] feature

179
to account for the mandatory quantization of the event. This also accounts for the

unexpected presence of A with non animate verbs. If there are also verbs that can have

XP optionally, we predict that these verbs will require the quantization of the event if A

is present. That is, the prediction will be the following: the verbs that have the possibility

of having the aforementioned XP projection can have inanimate objects with the A-

marker. Different from the verbs like those in (120), which are just a handful (see

Torrego 1999: 1788), this other group of verbs have XP only optionally; this means that

the presence of A will not be mandatory with these verbs. However, when these verbs do

have XP (and consequently A with inanimate objects), the event must be quantized (in

other words, the XP option will force both quantization of the event and A-marking of the

object, regardless of (in)animacy). This prediction is borne out, as I will now demonstrate.

As discussed earlier, in some cases, A is possible with inanimate objects. This is actually

one of the most troubling facts about Spanish DOM, which has led the researchers either

to give up any generalization or relativize the notion of animacy, concluding that animacy,

as Torrego puts it, “does not seem to be semantic in nature” (Torrego 1998: 55). This

conclusion seems to imply that semantically inanimate objects with A are really animate

in some other sense. It is hard to see what this sense could be.

Some researchers claim that, given this situation, it is impossible to study the

phenomenon relying only on judgments from constructed examples (Laca 1995); that is,

that we are in the unpleasant situation that we cannot trust native speakers intuitions,

given the generalized impression that A is somehow connected to animate nouns (the

180
traditional grammarians even call it “the personal A”, by which they mean the A used for

persons), and the fairly well spread presence of A with inanimate objects in real discourse

(see Fernández Ramírez 1986 and Laca 1995 for some examples from written texts),

which are actually not so hard to construct either (see Luján 1978).

Given this situation, and to be on the safe side, I will use examples taken from collected

corpora, in particular, the Corpus de Referencia del Español Actual (CREA), an online

database of oral and written texts taken from both Peninsular and Latin American

Spanish, maintained by the Real Academia Española, which has more than 160 millions

words.105 It is beyond my goals at this point to conduct a full scale description of DOM in

this corpus (a project that is something to keep in mind for the future). So I have chosen

to search for two inanimate A-phrases: a las paredes (to the walls) and a las columnas (to

the columns).106

The phrase a las columnas had 35 occurrences and only three of them were direct objects;

a las paredes had 232 occurrences, and only seven of them were direct objects. This not

surprising since “a” also has a directional meaning (like English “to”), which is quite

common and is not relevant here. In addition, some verbs also require prepositional

objects with “a”—such cases are different from the cases in (72) because their objects

cannot be cliticized with an accusative clitic (for instance, referirse a (to refer to)).107 As

predicted, all the cases with inanimate direct objects preceded by A were cases with a

105
It can be consulted here: http://corpus.rae.es/creanet.html
106
The plural forms have to be used to properly test for quantization.
107
Through this dissertation, in the text and in the glosses, I have consistently used capital A to indicate the
“a” that is used for DOM objects. When reference to other types of “a” is necessary, I indicate so by using
“a” or “to”. The reader should keep this in mind.

181
quantized event. As illustration, let us consider two examples with the same verb:

(123) a. la matriz […] que rodea a las paredes celulares vegetales

the matrix that surrounds A the walls cellular vegetable

the matrix that surrounds the vegetable cellular walls

b. los mesones que rodean a las columnas

the tables that surround A the columns

the tables that surround the columns

[from REAL ACADEMIA ESPAÑOLA: Banco de datos (CREA) [en línea].

Corpus de referencia del español actual. <http://www.rae.es> [April 27 2007]]

It is not possible to have a non-quantized interpretation in these sentences. In (123a) it

must be the case that the matrix surrounds each of the cellular walls, and not the cellular

walls as a whole. The same holds for (123b): the tables have to be around each column.

So, in both cases there are multiple events of surrounding. What is crucial here is that, if

we drop the A-marker, the quantization becomes optional:

(124) a. la matriz […] que rodea las paredes celulares vegetales

the matrix that surrounds the walls cellular vegetable

the matrix that surrounds the vegetable cellular walls

b. los mesones que rodean las columnas

the tables that surround the columns

the tables that surround the columns

182
That is, it is possible to interpret a unique event of surrounding in both sentences in (124),

even if that interpretation is highly unlikely in (124a). Under the system we have outlined

here, this is predicted. The Spanish DOM system is not sensitive to quantization, that is,

quantization can happen freely to A-marked and unmarked objects (see the discussion of

(69)-(71)). In other words, the semantic property of quantization does not need to have a

morpho-syntactic reflex.108

Under normal circumstances, A-marking is depending on animacy. As discussed above,

some verbs (like those in (120)) have a mandatory XP projection that forces the

quantization of the event and the raising of the object; given that XP has a DAT case, the

objects in these cases must take a DAT value (i.e. A) for its case feature (X being closer to

it after object raising). In addition, verbs like rodear in (123)-(124) can have an optional

XP that, when chosen, triggers the same effect: mandatory quantization and A-marker.

This means that rodear is a regular DOM-marking verb, which marks its object in a

manner consistent with the animacy constraint, and is insensitive to quantization.

However, it can optionally take a quantizing functional category, in which case it would

mark its object with A irrespective of animacy.

We can actually construct minimal pairs that reflect this alternation between A-marking

and quantization:

108
But in some languages (like Finnish), there is a systematic morpho-syntactic correlation with
quantization. See Ritter and Rosen 2001 and the references therein. I suggest that Spanish DOM expresses
the same thing with a subset of verbs.

183
(125) a. El chico abrazó a las columnas

The kid hugged A the columns

The kid hugged the columns.

b. El chico abrazó las columnas

The kid hugged the columns

The kid hugged the columns.

The only meaning that (125a) has is where the kid hugs the columns one by one. Only

(125b) has the meaning where the kid hugs all the columns at once (but the reading where

the kid hugs the columns one by one is available in (125b) too).

Given this, we can safely conclude that Spanish DOM is really sensitive to animacy, and

the A that appears with inanimate nominals with some verbs is a result of independent

factors, given the mechanism explained in (116).109

The remaining question is how specificity enters into the picture. To answer this question,

I will to need address an issue that is often neglected with respect to DOM systems in

general (but see Lidz 2006 for Kannada): how the structure of DP is related to case

checking.

109
A question that remains open is what causes the verbs in question to behave like this (after all, not all
verbs can license A with inanimate objects, as has been widely observed). I will not solve this question here,
and I will merely assume that including an optional quantizing functional projection is a lexical property of
a subset of verbs. Further research with corpora data could help discover a unifying property of these verbs.

184
2.3.2 Inside DP

To address this question, let me assume a structure for Spanish DP along the lines of

Ticio 2003, with three domains, which I will represent by K, D and small n—simplifying

Ticio 2003 just for expository reasons:110

(126) [KP K [DP D [nP n [ NP ] ] ] ]

K is the locus of the [case] feature (along the lines of Bittner and Hale 1996). That is, K

has an unvalued case feature that needs to be checked under Agree. For direct objects, K

can get two values (depending on the head that values the case), which in turn get

different morphological exponents:

(127) a. K ↔ a
[ucase: DAT]

b. K ↔ Ø
[ucase: ACC]

(127a) results from entering into a checking relation with the DAT head, and (127b)

results from entering into a checking relation with the v head.

110
For Spanish DP, Ticio 2003, following Grohmann and Haegeman 2002, proposes that the DP is divided
into three prolific domains (in the sense of Grohmann 2000): a thematic domain, an agreement domain and
a discourse domain. The full display of functional categories is this:

(i) [TopP [DP [AgrP [nP [NP ] ] ] ] ]

Notice that (126) differs from (i), but nothing significant should be attached to this difference. In fact, KP
can be considered the equivalent of TopP, by assuming that TopP is the final host of case and φ-features.
Additionally, (126) conflates AgrP and nP, but nothing will change if we have an AgrP in addition to nP.
The important thing is that, like Ticio’s system, (126) also expresses the idea that there are three different
domains inside DP and, as we will see, crossing from one to the other has some consequences.

185
This means that the A-marker for DOM is not actually a preposition, a fact that has been

very well established in the literature (Schroten 1972, Strotzer 1976, Jaeggli 1982, 1986,

Demonte 1987, Brugé and Brugger 1996, Torrego 1998, among several others).111 This

also means that the Dative case, at least for DOM objects, is structural and not inherent.

Although this may be at odds with some traditional assumptions, according to which

Dative is an inherent case, it has been proposed independently on empirical grounds that

Dative can be structural (see Franks 1994, Bošković 2006b). In fact, there have been

proposals that Dative can be structural even for Spanish argumental indirect objects

(Masullo 1992, Torrego 1998). None of this, of course, rules out the possibility of

inherent Dative for some instances of Dative (e.g. for some verbs in German or Icelandic).

Notice also that (127a) implies that the A-marker for DOM is the same as the regular

Dative for indirect objects—I will discuss evidence for this is chapter 3. For Spanish

DOM, this suggestion was first made in the generative framework by Schroten 1972: 49-

76, who worked on the Case Theory outlined by Fillmore 1968, where this comes out

naturally. However, the idea has also been proposed in different frameworks, and it also

has some support from a historical perspective (see Pensado 1995 for an overview). The

proposal to conflate the case of DOM-objects and Indirect Objects has been criticized on

the grounds that, in Spanish, DOM objects can be passivized but regular IO Datives

cannot. As it should be apparent, this is not a valid objection in a framework that assumes

that grammatical functions are not primitives, nor do they depend on overt case

111
The claim is based on c-command relations inside ditransitive constructions, which have particular
properties that I will address in chapter 3.

186
morphology.

Passives are derived by raising the object to subject position from its based generated

position, which is the standard assumption at least since Chomsky 1957. In the passive,

there is no small v to assign accusative; instead, the internal nominal checks case against

T, receiving nominative. The Dative projection, even if present, plays no role. Different

from other languages, regular Indirect Objects in Spanish cannot get passivized because

the structural conditions do not allow the relevant A-movement.112 In addition, the idea

that Spanish DOM objects have Dative case receives substantial support from the fact

that, even when DOM languages do not correspond to a particular family or area, the

most common marker for DOM objects is the Dative marker (Bossong 1985, 1991),113

which can hardly be a coincidence.

As we have discussed in the previous section, DOM arises from the fact that small v is φ-

incomplete, hence, it cannot value the case feature of φ-complete nominals. This means

that I need to explain how a nominal ends up being φ-complete or φ-incomplete. By

doing so, I aim to substantiate that idea that DOM is a byproduct of the relation between

the structure of DP and the conditions on case-valuation imposed by the Agree system.

We have already discussed some aspects of this issue in chapter 1. Recall that we are

112
See McGinnis 1998, 2001 for the idea that the IO in certain languages (and constructions) cannot
undergo A-movement (and, therefore, cannot get passivized) after the DO leapfrogs over it. If the IO is
base-generated in a position higher than the DO, after leapfrogging, the DO will end up in a position closer
to T, preventing the IO from undergoing A-movement to T. In languages that do passivize the IO, different
structural conditions prevent the DO from leapfrogging over the IO, which is then closer to T.
113
Among the few exceptions are Hebrew, Rumanian and Turkish.

187
assuming that animacy in the nominals is the trigger for DOM (through the [person]

feature). In order to connect the φ-feature [person] to animacy, I postulate the following

condition with respect to the distribution of [person] in Spanish:114

(128) Spanish Distribution of [person]

If [+animate] in N, then [person] in small n.

Notice that (128) is not an arbitrary stipulation. As discussed in chapter 1, nominals that

uncontroversially hold a [person] feature, that is, [1p] and [2p] pronouns, are always

[+animate]. It is natural to assume that a language could link the possibility of having the

[person] feature to one aspect of these pronouns, namely the fact that they are [+animate],

and therefore this language will grant [person] to all [+animate] nominals.

The reason why the [person] feature should appear in n, and not in a higher head in the

structure of DP (cf. (126)) can be reduced to locality: n is the head that immediately c-

commands NP, thus, [+animate] is visible from there. Alternatively, it could be proposed

that [person] actually starts in N, and then it raises—which would make this very similar

to a proposal made by Lidz 2006 with respect to Kannada (another DOM language), and

which is naturally compatible with the subsystem I will outline immediately. I leave open

the choice between the two options noted above, but the reader should be aware that

nothing hinges on it.

As discussed above, Agree is subject to a principle that maximizes matching effects (102).
114
So [animate] is not involved in feature checking; its “reflex”, [person], is.

188
Given this, it seems natural to assume that the relevant features cannot stay scattered

across different heads inside the DP, that is, I assume that the whole DP counts as a

single goal. I thus argue that all φ-features must raise to K, where the [case] feature is

located. This is not an arbitrary condition, but it is motivated by (102).115

The raising of the φ-features can be achieved by using feature percolation, a mechanism

that is familiar in the literature, for which, in what follows, I will offer an implementation.

Under traditional assumptions, an X0 element (a head) is conceived of as a bundle of

features:

(129) [ H]
[person: α ]
[number: β ]
[case: γ ]

Percolation makes possible the creation of associations between a feature F and a head H

that does not originally host F in its lexical entry; that is, a feature F hosted by a head A

becomes part of the feature bundle of another head H (in that sense, it is a sort of re-

bundling):

115
This makes the current system different from the already mentioned mechanism proposed by Lidz for
Kannada, which also raises features across the nominal extended projection invoking “unusual” reasons
(Lidz 2006: 29-30). I can in fact also use (102) to motivate Kannada facts, in accordance with the system
outlined here, and still in Lidz’s line of reasoning. I will come back to this in chapter 4.

189
(130)

[ H]
[ ]

[ A ] …
[ G ]
[ φF ]

After percolation applies in (130), the resulting structure is:

(131)

[H ]
[ φF ]

[ A ] …
[ G ]

It is reasonable to subject this operation to strict locality conditions. The strongest

hypothesis is that we can only associate a feature to the next c-commanding head. This is

expressed here:

(132) A φ-feature F hosted by a head Α can percolate to another head Β iff:

i. Β c-commands Α.

ii. There is no head Γ such that Γ c-commands Α, and Β c-commands Γ.

iii. Β is compatible with F.

190
(132) forces percolation to be bottom-up, and provides a way to stop it based on lexically

defined properties of the c-commanding head: if the next head is not compatible with a φ-

feature, the association cannot proceed.

This is all we need to derive the interactions between specificity and animacy that we

find in Spanish DOM, as I will now demonstrate.

Recall from the discussion in section 2.1, that there are two types of D in Spanish: D and

*D:

(133)
D [+specific] or [-specific]
*D [-specific]

As discussed earlier, the distinction between D and *D cuts across definite determiners,

indefinite determiners, and kinds.

Let us consider first the case of definite DPs. As discussed in section 2.1.1, A-marking

depends on the possibility of having a referential interpretation, which I understood as the

possibility of having a global choice function (a function to pick up the salient individual

in a given context, as defined by Heusinger 2002b). Only if the object has a referential

interpretation, the presence of A is enforced (134); otherwise, it is optional (135):

191
(134) Juan está buscando al decano, es decir, a Smith

John is looking-for A+the dean, is to say, A Smith

John is looking for the dean, namely for Smith.

(135) a. Juan está buscando al/el decano, sea quien sea

John is looking-for A+the/the dean, be-SUBJ who be-SUBJ

John is looking forthe dean, whoever it might be.

b. Está buscando al/el decano más alto

is looking-for A+the/the dean most tall

He is looking for the tallest dean.

I can assume that the object nominal in (134) has a D definite determiner, and that the

object nominals in (135) can have D or *D determiners. Given this, the only thing needed

now to capture the presence/absence of A is to assume that only D is compatible with

[person], that is, that only D can host [person]: *D determiners are incompatible with

[person], which means that *D cannot host [person], that is, [person] cannot percolate to

*D.116

The above predicts that if a [+animate] nominal, that is a nominal that receives a [person]

feature in its small n, has *D as the head of its DP, the [person] feature will get stranded

116
It is possible to stipulate that this is a purely lexical restriction on *D. Another possibility is to suggest
that this restriction is derived from the fact that some uncontroversially [person] nominals, like [1p] and [2p]
pronouns, are prototypically referential. Given than *D cannot be referential, this would account for its
inability to host [person].

192
in small n, which means that it cannot raise to K. I then assume that a stranded feature

must be deleted or the DP could not count as a single goal (which means that it could not

count as a goal at all).

Under the system outlined earlier, a nominal gets A only if it has a [person] feature, since

it then must raise to the Dative phrase, getting a DAT value for its case feature (the A-

marker). This means that *D nominals cannot get A, since they cannot have a [person]

feature in K. Let me illustrate this. The nominal decano (dean) in all the cases in (134)-

(135) starts as an NP with an animate feature [ANI] (only the relevant information is

represented):

(136) NP

N
decano
[ANI]

After small n enters the structure, the condition on the distribution of [person] (128)

applies, and n ends up hosting a [person] feature:

(137) nP

n NP
[person]

N
decano
[ANI]

193
Now, there is an option between D and *D. If *D is chosen, one consequence will be that

the whole definite DP will be non referential, that is, it will not be able to receive a global

choice function. The other consequence is that *D will not be able to host the [person]

feature, and therefore the [person] feature will get stranded, unable to reach K. To allow

for the nominal to count as a single goal, this stranded feature must be deleted:117

(138) KP

K *DP
[case: ]
[num]
*D nP
el

Ψ n NP
[person]

N
decano
[ANI]

This means that K will end up without a [person] feature. Under the current system, this

means that the nominal will be able to check case against small v, as a result of which it

receives ACC, that is, the nominal will not have an A-marker. This is what happens in the

A-less options of (135).

On the other hand, if we choose D, then it becomes possible to assign a global choice

function, that is, the nominal has the possibility to be referential. But D also has the

ability to host a [person] feature. Therefore, the [person] feature will end up in K, after

117
Notice that the original position of the [number] feature is orthogonal to the current system. For the sake
of parallelism, let me assume that [number] starts on n and raises to K under the same procedure discussed
above (since all the heads in its path are presumed to be compatible with this feature).

194
successive percolations (to D and K):

(139) KP

K DP
[case: ]
[num]
[person] D nP
el

n NP

N
decano
[ANI]

In (139), the nominal has a complete set of φ-features, which means that it will not be

able to check case against small v, which is φ-incomplete (specified only with [number]).

Therefore, under the system described in the previous section, the nominal must raise to

the Dative phrase to check DAT, as a result of which it gets the A-marker. This is what

happens in (134) and the A options of (135). Remember too that D does not need to take

a global choice function, that is, it does not need to be referential. This predicts that

nominals can get A without being referential, as in the corresponding versions in (135).

It is important to stress that the system discussed above predicts optionality for the A

marker with non-referential animate nominals, a fact that has been extremely difficult to

capture in other theories (including OT, at least for definite nominals), which routinely

appeal to an idealization of the data, retreating to the partial generalization described in

the beginning of the chapter in (1)-(5).

195
The same system can be straightforwardly applied to indefinites, where the same

predictions arise. Indefinites can also be D or *D. If D is chosen, the [person] feature will

be able to raise to K, which will end up being φ-complete, and the nominal will not be

able to check case against small v. Therefore, after checking their feature against the

Dative phrase, D indefinites will end up being A-marked. *D indefinites, on the other

hand, cannot host a [person] feature; if the nominal has a *D indefinite, the [person]

feature will get stranded (and deleted), resulting in the φ-incompleteness of K, and the

subsequent lack of A-marker (because φ-incomplete nominals can check case against

small v). In other words, D indefinites must be A-marked, and *D indefinites cannot be

A-marked. Since D indefinites can have a local choice function (that is, D indefinites can

be specific), but they do not need to, the system also predicts an optionality of the A-

marker for non-specific indefinites. This is exactly what we have:

(140) Toda cadena que contrata (a) un actor famoso gana dinero con él.

Every network that hires A an actor famous, wins money with him

Every network that hires a famous actor, wins money after him.

As discussed in section 2.1.1, donkey sentences (like (140)) are a test for non specificity

and they are possible with the A marker. The same holds for cleft conditionals:

(141) Si Juan ha conocido (a) una persona interesante en su vida, ha sido María

If John has met A a person interesting in his life, has been Mary

If John has ever met anyone interesting in his life, it is Mary

196
The presence or absence of A in (140)-(141) depends on the choice of D or *D. Notice

that, when D is selected, the lack of a specific reading is forced not by D (which has the

potential to be referential), but by the particular environment in which the object is placed

(donkey sentences, cleft conditionals). On the other hand, if the environment forces

specificity, *D cannot be selected, and the only possibility is D. Under the current system,

the prediction is that objects placed in environments that force specificity must be A-

marked. As discussed in section 2.1.1, the prediction is borne out. A prenominal adjective

forces the specificity of the nominal, and in this case, the nominal must have the A-

marker:

(142) Busco *(a) un famoso actor

look-for A a famous actor

I am looking for a famous actor.

As we also discussed in section 2.1.1, strong quantifiers must have A:

(143) a. Había besado *(a) varias de sus amigas.

Had kissed A several of his girlfriends

He kissed several of his girlfriends.

b. Besó *(a) toda chica con sombrero.

kissed A every girl with hat

He kissed every girl with hat.

197
This means that strong quantifiers are D elements, and therefore they trigger the

procedures described above for D elements (that is, they always end up marked with A).

Weak quantifiers, on the other hand, can be freely D or *D, and the A-marker in these

cases is optional:

(144) Besó (a) algunas chicas

kissed A some girls

He kissed every girl with hat.

Remember also from the discussion in section 2.2.1 that kind-denoting nominals can

optionally have A:

(145) Antoñito buscaba (a) la mujer rica.

Antoñito looked-for A the rich women

Antoñito looked-for the rich women

As mentioned earlier, when confronted with similar sentences, some researchers (for

instance, Brugé and Brugger 1996) even claim that kind-denoting elements can freely be

[±animate], a position that I find counter-intuitive. Under the current system we can

account for the optionality of A with kinds straightforwardly: kind-denoting nominals

(like the one in (145)) are always animate, but they can freely be D or *D. Notice that this

implies that kind-denoting nominals instantiate a reading that is different from merely

198
definite or indefinite DPs (in the spirit of Carlson 1977a). A possible question to ask here

is what difference could it make for a kind-denoting nominal to be D or *D. I contend

that it makes no interpretative difference, and that this is a reflex of the fact that all other

Spanish determiners are D or *D. However, this difference is not totally innocent, as we

will see immediately.

It is worth noticing that, with both definites and indefinites, the *D determiner is the

“marked” form with respect to D. By this I do not mean a morphologically marked form

(D and *D are homophonous), but that the *D form is more specified than D, that is, it

marks the lack of referential reading—whereas, D is, sort of speaking, “unspecified” with

respect to the possibility of referential reading (that is, it can have referential reading or

not).118 Assuming that “unmarked” forms are preferred, we may be able to explain the

well-known preference for A with animate objects.119

In fact, for kind-denoting nominals, the use of A is also preferable. With some verbs, this

preference is even stronger. Brugé and Brugger 1996 notice the following contrast, which

they attribute to the fact that a verb like exterminate can have only objects that denote

kinds:

118
The reader should keep in mind that there is a shift in the use of “marked” here. Crucially, I do not mean
here that “marked” stands for “with the A-marker”. Ironically, the “unmarked” D actually ends up
morphologically marked with A.
119
Notice that this is not an application of the Elsewhere Condition. If this condition were to be invoked
here, D would never surface, leaving these nominals always without A, contrary to the facts. Furthermore,
D and *D are options for the numeration (they are both in the lexicon), not the result of competition
between rules. This means that the Elsewhere Condition does not apply here. The only think that I want to
state here is a general preference for the “unmarked” forms, provided that they are not in competition.

199
(146) a. ? Las enfermedades y la guerra han exterminado el hombre

the illnesses and the war have exterminated the man

the illnesses and the war have exterminated the man

b. Las enfermedades y la guerra han exterminado al hombre

the illnesses and the war have exterminated A+the the man

the illnesses and the war have exterminated the man

[Brugé and Brugger 1996: 6]

This confirms that D is the preferred option with kind-denoting nominals.120

So far, I have accounted for the fact that animate definite nominals that do not receive A

cannot be referential. On the other hand, D nominals (which get [person] in K) can have a

choice function but they do not need to; this accounts for the fact that A-objects can but

do not need to be specific, as discussed in the text. In other words, the system

straightforwardly accounts for the core data of Spanish DOM.121

We can also use the above system to account for a particularly recalcitrant set of data, for

which no theory I know of offers any solution. Consider again the data in (50), repeated

here:

120
It is still unclear why this contrast is not reproduced in other contexts. I have no final account for this.
121
As a matter of fact, I am extending the traditional core data of Spanish DOM by including the optional
cases of A, which are not encompassed by the standard generalization (presented in (5)).

200
(147) a. Vi *(a) alguien en el parque

saw A somebody in the park

I saw somebody in the park

b. No vi *(a) nadie en el parque

No saw A nobody in the park

I saw nobody in the park

c. *(A) quién vi en el parque?

A who saw in the park

Whom did I see in the park?

The problem here is that the above objects, although animate, cannot be specific, which

in our system means they must be *D. This should predict that they cannot get A, but in

fact A is mandatory here.

It must be observed that there is something peculiar about these items: nadie, alguien and

quien (besides their mandatory A). They are amenable to a clear morphological analysis,

on the grounds that they all have a similar ending and a similar interpretation (animate

non-specific).

(148) a. nad-ie

b. algu-ien

c. qu-ien

201
I propose that ie/ien is the spell out of the [animate] feature. I also contend that the items

in question are complex lexical items, by which I mean that they are formed in the

lexicon (in the spirit of Hale and Keyser 1993). We have been assuming that the basic

components of any nominal are K, D, n and N ((126), based on Ticio 2003). I will

assume the same here. Given that the elements in question are always non-specific, it is

reasonable to assume that the D-component must be *D. The structure of these items

before merging with K should then be the following:

(149) *D

*D n
nad-/alg-/qu

n N
-ie(n)
[ANIM]

In accordance with (128), a [person] feature will be present in n (I will abstract away

from other φ-features):

(150) *D

*D n
nad-/alg-/qu

n N
[person] -ie(n)
[ANIM]

At this point, by hypothesis, *D cannot get associated with [person]. This is so far the

202
situation we find with any *D. Notice, however, that, different from the other *Ds

discussed previously, the structure in (150) is a single complex lexical item, and that the

next item to merge is K, a different lexical item. Being a lexical item entails the atomicity

of the elements in questions. According to DiSciullo and Williams 1987:49, who

implement a version of the Lexical Integrity Hypothesis (Lapointe 1980), a lexical item is

atomic, that is, the hierarchical structure of the features inside of it is not visible from the

outside (but, of course, the features themselves are). In other words, only features that

have percolated to the head of the word are visible from the outside.

In that sense, what we want is that, when K is merged with (150), K could see *D’s

features as a whole, that is, as being part of a single unit, as an atomic element. This will

allow association of [person] to K, skipping the effects of the incompatibility of *D to

host [person], since it means that *D would be flattened, that is, [person] would already

be in the head of the word. This possibility is illustrated here:

(151) KP

K *D
[ ]

[person]
[ANIM]
nad-/alg-/qu+ie(n)

K would then get [person], and these nominals will end up with the A-marker, following

the mechanisms discussed above.

203
For (151) to be possible, it is crucial that the [person] feature is still there at the point at

which K enters the structure. Indeed, there is no need to delete the feature at this point.

As discussed above, stranded features must be deleted to ensure that the whole DP acts

like a single goal. But *D cannot be a goal for Agree relations before K enters the

structure, so [person] is not deleted yet. Given that *D is atomic, when K is merged, the

[person] feature can freely percolate to K.

An interesting confirmation of this mechanism comes from an additional twist in the

behavior of nadie and alguien (reported by Suñer 1988 and López 2006: 149 fn. 155).

These elements have a mandatory A in sentences like (147), that is, when they are bare

(without modifiers), but when they take adjuncts—which I assume are base-generated as

right specifiers (see Ticio 2003 for discussion)—the A becomes optional:

(152) a No encontré (a) nadie que pudiera encargarse

No found (A) nobody that could charge-self

I didn’t find anyone that could take charge.

b. Busco (a) alguien que sepa tres idiomas.

Look-for (A) somebody that knows three languages

I am looking for someone that speaks three languages.

The merge of the relative clause, a phrasal constituent should happen before the merging

of K. Obviously, the resultant node (*DP) cannot be atomic:

204
(153) *DP

*D CP

*D n que pudiera encargarse

nad-

n N
[person] -ie
[ANIM]

If the head of *DP were still be atomic (as in (151)), nothing would change here: K

would still be able to see the [person] feature, which means that the nominal would have

to get A. Notice, however, that the A-marker is optional in cases like (152); therefore, we

have to provide a way to allow K to count the internal structure of *D; since [person]

feature percolation to K would then be blocked, this would yield the option without A.

A question arises regarding how to achieve the flattening of the structure of features. For

the features relevant for this discussion (the φ-features), in the vast majority of cases, this

is vacuous: simple lexical items do not have their φ-features organized in a hierarchical

structure, so they are inherently atomic. But the issue is relevant for complex lexical

items. I assume that complex lexical items enter the numeration with their internal

structure intact, and that this structure must be flattened. 122 For Chomsky 1995: 249,

given bare phrase structure, lexical items are ambiguous between minimal and maximal

122
There is no need to postulate an independent mechanism of flattening (restructuring of the features, for
instance): it could be enough to assume that the internal structure of complex lexical items becomes
inaccessible. I will not discuss the issue further in this dissertation, but it seems apparent that a more
detailed discussion of the morphology-syntax interface (in particular, in relation with atomicity) can
provide a more principled account.

205
projections (see also Bošković 2002b). If a complex lexical item counts as a minimal

projection, it must be atomic, therefore, its internal structure is inaccessible. But if it

counts as a maximal projection, it is accessible.

It is usually assumed that adjuncts can be inserted either cyclically or acyclically. It

seems safe to assume that when *D merges with the clearly phrasal adjunct, *D should be

considered as having inherent structure (which means *DP would be a maximal

projection). In this case, [person] would not be able to percolate to K. However, the

adjunct can also be inserted acyclically, after K merhes with *D. In this respect, since it

is not present in the structure when K and *D merged, it does not interfere with K-*D

relations, and K can receive [person] in the manner discussed above. We then obtain the

desired result.123

There is a final set of data that requires explanation. As discussed in the first section of

this chapter, bare nouns without modifiers cannot receive A, but A is optional if the bare

noun receives a modifier:

123
Another possibility is to assume that when a complex lexical item merges with another head, it must
count as a minimal projection (that is, as a head), but when it merges with a maximal projection, it remains
ambiguous between a maximal and a minimal projection. In that sense, in (151), *D merges with K,
another head; therefore, it counts as a minimal projection, and its internal structure must be flattened (i.e. it
becomes inaccessible). As a result, the [person] feature will percolate to K, which in turn will be φ-
complete, and consequently will have to get the A-marker. On the other hand, in (153), *D merges with a
maximal projection, hence it remains ambiguous between a minimal and a maximal projection. If the
former, its internal structure is inaccessible (so it will end up with A); if the latter, it is still accessible, so
*D will count as an intervener between the [person] feature in small n and K, impeding the percolation (as
a result, the nominal will not get A). If this is correct, this recalcitrant piece of data is not only accounted
for, but becomes evidence in favor of the bare phrase structure system.

206
(154) a. El comité busca (*a) estudiantes

The committee looks-for A students

The committee looks for students.

b. El comité busca (a) estudiantes con buenas calificaciones

The committee looks-for A students with good grades

The committee looks for students with good grades.

The optionality of A in (154b) suggests the familiar D/*D distinction. To explain the lack

of A in (154a) I will assume that there is no KP-DP layers in bare nouns without

modifiers; that is, they are small ns. The consequence will be that they do not have a case

feature to value (and therefore they stay in situ).124 But the KP-DP layers must be present

if a bare noun has a modifier, assuming that the modifier is introduced within those layers.

We thus account for these data without appealing to the ECP (contra Brugé and Brugger

1996), which is a welcome result, because the ECP cannot be naturally accommodated in

the theoretical toolkit of the Minimalist Program, since the ECP is based on the notion of

government, which has been abandoned—for good reasons (see Bošković 1994 for

discussion on this issue).

Further evidence in favor of the above suggestion may come from the fact that we find a

similar contrast with bare nouns in other positions, like Subjects and Indirect Objects

(which are not subject to DOM constraints, so this lack of KP-DP layer is an independent

124
This may be a way to implement processes like semantic incorporation (Van Geenhoven 1998) or
pseudo-incorporation (Massam 2001). I leave this issue open.

207
fact). Bare nouns are not possible without modifiers in those positions—as discussed

early by Suñer 1982a, who coined the term “Naked Noun Constraint” to describe this

situation (see also Pérez Silva 1990 and Martín 1999: 473-478):

(155) a. * Ingenieros construyeron el edificio

Engineers built the building

b. Ingenieros sin preparación construyeron el edificio

Engineers without preparation built the building

Engineers without skills built the building.

(156) a. * Le dieron el projecto a ingenieros

CL-DAT gave the project to engineers

b. Le dieron el projecto a ingenieros sin preparación

CL-DAT gave the project to engineers without preparation

They gave the project to engineers without skills.

The reason is for this constraint is that the naked nominal cannot be in a case-checking

position, since, lacking KP, a naked nominal cannot participate in Agree relations.125

125
Interestingly, with post-verbal subjects, a naked noun is possible:
(i) Llegaron ingenieros.
Arrived engineers
There arrived engineers.

208
2.4 Summary and conclusions

We have reviewed the core set of data relevant to Spanish DOM as well as the

problematic data, such as the optionality of A with non-specific objects, the presence of

A with inanimate objects, and the mandatory A with some non-specific quantifiers (like

nadie and alguien). We have seen that the mechanism of Agree (which we have used in

chapter 1 to account for Spanish existential constructions) is able to account for all of

them, as long as we take into consideration some key aspects of the structure of Spanish

DP. My main conclusion is that the initial trigger for DOM is the animacy of the object,

which requires a [person] feature in small n, a low functional head in the DP. This feature

has to raise to a higher head, subject to conditions imposed on the operation that raise it,

which reflect the peculiarities reported in the phenomenon. In particular, if the [person]

feature cannot be hosted by one of the intermediate heads, we get a φ-incomplete nominal,

which will end up without A.

A novel prediction was tested with respect to the relation between quantization and the

presence of A, which confirms the mechanism of Agree assumed here. I provided an

account for the presence of A with non-animate nominals, without the counterintuitive

stipulations that other GB/Minimalist solutions were forced to make. This result of the

system represents a major advantage over the Functional-OT system of Aissen 2003,

since there is no way to accommodate this property in her system.

Additionally, we have been able to accomodate the cases of optional A with non-specific

209
nominals, which represents a major advantage over Torrego’s 1998 system. My analysis,

however, preserves the fundamental part of her insight, namely, that A-marked objects

are in a higher position than A-stripped objects.

The Spanish DOM phenomena discussed so far have been limited to monoclausal

transitive sentences. In the next chapter I will examine DOM in non monoclausal

sentences and ditransitive constructions, which, I believe, provide further support for

some key aspects of the system presented here.

210
 Chapter 3

Beyond the simple transitive clause

3.1 DOM in ditransitives

One interesting set of properties of Spanish Differential Object Marking concerns its

interaction with the Dative Indirect Object in ditransitive constructions. I will discuss two

aspects of the interaction here. The first one, perhaps one of the most elusive and

controversial aspects of the phenomenon, concerns the deletion of A-marking on the

Direct Object in the presence of an Indirect Object. The second one, where the issues are

better understood, concerns the question of the height of the objects, i.e. their c-command

relations (with the corresponding effects in scope). Both issues have a potential impact in

the system I have developed in the previous chapters. I will dedicate this chapter to them.

3.1.1 Double A

In chapter 2 I proposed that the Spanish DP is a projection headed by K, which hosts the

case feature (following Bittner and Hale 1996, and others, in a way that is compatible

with the structure of DP argued for in Ticio 2003). In addition, in chapter 1, I argued that,

given that Spanish small v is φ-incomplete (it does not have a [person] feature), nominals

with a [person] feature cannot check their case against small v. φ-complete nominals then

must check their case against a Dative head, and consequently, they get Dative, which

accounts for the A-marker with such nominals:

(1) K ↔ a
[ucase: DAT]

The DP in question then ends up with the same structural case as the Indirect Object (see

Franks 1994, Bošković 2006b for the idea that Dative can be structural, a possibility that

has also been pursued for Spanish argumental IO (Masullo 1992, Torrego 1998)). Given

this, a question arises regarding what happens in ditransitive constructions. Under the

system I am developing here, we would expect to have two objects marked with Dative

(i.e. A) if the DO is φ-complete. There is evidence that this is the right conclusion.

With respect to the Dative marker (that is, A), there is a long standing trail of

observations that strongly suggests that there is some competition for the A-marker: one

A can be dropped in a ditransitive sentence when both objects are full DPs (2a), which is

not possible if the Indirect Object is not a full DP (e.g. when it is just a clitic, as in (2b)):

(2) a. Juan le presentó María a Pedro

John DAT-CL introduced Mary to Peter

John introduced Mary to Peter

b. Juan le presentó *(a) María

John CL-DAT introduced A Mary

John introduced Mary (to him/her)

In fact, the presence of an additional A decreases the grammaticality of the ditransitive

212
sentence in (2b):

(3) ?? Juan le presentó a María a Pedro

John CL-DAT introduced A Mary to Peter

John introduced Mary to Peter

On the surface, the contrast between (2a) and (3) represents a problem for the system

developed here, since it seems that the object in (2a) is getting accusative, that is, it is

getting a case value from small v, which I argued is not possible, as discussed in the

previous chapters. However, I will show that this state of affairs actually constitutes

evidence in favor of my analysis.

The observation that the A-marker can be dropped in some cases is very old in the

traditional grammar of Spanish, although it has never been expressed in the way I will

present it. The nineteenth century grammarian Andrés Bello, perhaps the most reputed

traditional grammarian ever, noticed the possibility of dropping the A in ditransitives, and

he offered this sentence:

(4) Presentaron la cautiva Zenobia al vencedor

Introduced the prisoner Zenobia to+the victor

They introduced the prisoner Zenobia to the victor

[Bello 1847: §900]

213
The way this issue has been addressed in the descriptive literature, however, is not

consistent. The possibility of dropping A has been occasionally denied by using examples

that do not allow the dropping (without explaining what happens in other cases). Several

researchers report the dropping with skepticism, but still, it is mentioned quite often—as

Torrego graciously puts it: “the linguistic literature on Spanish is replete with examples

of objects preceded by the dative preposition that mysteriously go awry in ditransitives”

(Torrego 1998: 131).126

The reports about this issue can be divided in different kinds—a split that, as we will see

immediately, can be found even inside the same paper, which enhances the “mysterious”

character of the phenomenon. For some, the preposition just does not drop at all (Pensado

1995: 25, Brugé and Brugger 1996: 11 fn. 14), or they suggest a dialectal split (Brugé and

Brugger 1996: 29 fn. 34); others believe that A drops in all relevant cases (Schroten 1972:

59, Laughren and Eisenchlas 2006); most researchers suspect that A drops only in some

occasions (Demonte 1994: 460-461, Torrego 1998: 131-147, Campos 1999: fn. 77);

many use examples with A-dropping while discussing related issues without mentioning

the lack of A (famously, Strotzer 1976: 206 et seq.); some suggests that the dropping of A

is optional with ditransitives (Suñer 1988: 401). The issue has even made its way to the

prescriptive grammar, where it is recommended to drop the A to avoid a potential

ambiguity with the Indirect Object (Real Academia Española 1931: 191), a reason that

was also invoked by Bello 1847: §900 (and others after him).

126
Despite her observation, which is correct, she does not cite any references, merely stating that her data
come from Strotzer 1976, who does not actually make this claim explicitly (but does provide examples of
this sort in her monumental dissertation about Spanish clitics).

214
I will take the side of the majority in this case, that is, I will suggest that dropping A is

not possible in all occasions, but I will provide motivation that is very different from the

reasons invoked in other literature to explain this dropping. Additionally, I believe that

my analysis can help us explain why there is so much controversy around this particular

issue in the literature.

It is clear that the dative markers cannot drop freely. Only the A-marker that corresponds

to the DO can drop; the A for the IO cannot be deleted:

(5) a. El jefe le presentó [DO el empleado ] [IO a Pedro]

The boss CL-DAT introduced the employee to Peter

The boss introduced the employee to Peter

b. * El jefe le presentó [DO al empleado] [IO Pedro]

The boss CL-DAT introduced A+the employee Peter

The boss introduced the employee to Peter.

Secondly, if the A-marker drops, the A-stripped object (the DO) must precede the IO (cf.

(5a)):

(6) * El jefe le presentó [IO a Pedro] [DO el empleado]

The boss CL-DAT introduced A Peter the employee

The boss introduced the Peter to the employee.

215
This is relevant because, normally, there is a degree of freedom with respect to the word

order between DO and IO:127

(7) a. Juan le dio un libro a María

John CL-DAT gave a book to Mary

John gave a book to Mary

b. Juan le dio a María un libro

John CL-DAT gave to Mary a book

John gave a book to Mary

Third, clitic-doubled IOs seem to allow the dropping more easily than there non-doubled

counterparts, at least for some speakers (Demonte 1994: 460-461, Torrego 1998: 131-

147). A note of caution is necessary here. Besides well-known correlations between

doubled and non-doubled IOs (Strotzer 1976, Masullo 1992, Demonte 1994, Torrego

1998, Bleam 1999, 2003, Cuervo 2003, Montrul and Perpiñán 2006, among others), there

seems to be dialectal or even individual variation with respect to the possibility of

doubling IOs. Clitic-doubling with the IO is far more common that non-doubling (see

Branchadell 1992 and the references therein).128 Then, perhaps it is not a surprise that

there are also mixed judgments with respect to the possibility of dropping the DO A-

marker in ditransitives if there is no clitic-doubling for the IO (some speakers report that

127
V+DO+IO (7a) may be considered the neutral order, but (7b) is not deviant at all.
128
Notice also that there is a difference here with clitic-doubling with the DO, which is much more
restricted (but also subject to dialectal variation).

216
this is not possible, others that it is optional). Given this situation, I will put aside here the

non-doubled cases, restricting the discussion to the doubling structures.

Before presenting my proposal, I will review some of the solutions that have been

presented in the literature, and briefly point out the problems they face.

First, it is unlikely that there is a dialectal split behind the A-dropping for doubling
129
structures (in the relevant cases), as Brugé and Brugger 1996 suggest. The

phenomenon is reported using sources from very different origins. In addition, Brugé and

Brugger 1996 claim that the split is Latin American Spanish versus Peninsular Spanish,

with the first variety allowing the dropping and the second prohibiting it; nevertheless,

they report a personal communication from Violeta Demonte (Brugé and Brugger 1996:

29 fn. 34), who is a linguist born and raised in Spain and allows the dropping (and even

wrote about it: Demonte 1994: 460-461); other speakers of Peninsular Spanish I have

consulted also allow the dropping in the cases discussed here. Furthermore, since these

authors do recognize that the dropping exists at least in one variety, a question remains

regarding how to capture this option; they present no solution. 130 So, pending a full

examination of dialectal variation of Spanish DOM in ditransitives, here I will assume

that the dropping of A does not behave differently in Peninsular and Latin American

Spanish.131

129
Demonte 1994: fns. 30 and 31 cites certain reports according to which proper nouns resist dropping A. If
this is correct, then it may be a dialectal difference, since this is not the case for all Spanish speakers. I have
no explanation for this.
130
In fact, their analysis completely bans the dropping of A, since, as explained in chapter 2, section 2.2.2.1,
they assume that A is mandatory with nominals that are accusative and animate.
131
It is an interesting question to ask if Spanish DOM in general (not only with respect to ditransitives)
shows dialectal variation. Almost all of the few reports on this issue are contradictory and anecdotic: some

217
Secondly, the suggestion that a speaker drops A to avoid ambiguity with the Indirect

Object conflicts with the fact that this dropping does not happen all the time; in fact

sometimes it cannot even take place (even with clitic-doubling structures, as I will show

later). Morimoto and de Swart 2006: fn 6, who, as discussed in chapter 2, propose that a

constraint on distinguishability between subjects and objects is responsible for Spanish

DOM, try to interpret the A-dropping in ditransitives as showing support for their

proposal. They appeal to the idea that two A-markers would introduce ambiguity (this

time with the IO). However, as it should be apparent, the absence of A reintroduces the

lack of distinction with respect to the subject, which the presence of A was supposed to

avoid. This means that the facts in question are actually evidence against the

distinguishability hypothesis.132

Finally, I would like to discuss the most attractive (but ultimately incorrect) solution to

the problem posed by the dropping of A in ditransitives, put forward by Torrego 1998:

131-147. The contrast that Torrego tries to explain is the following:133

researchers claim that Latin American Spanish is more restricted with respect to the use of A, others claim
that it is more liberal (see Pensado 1995: 17 fn. and references therein). As far I know, the only fairly
systematic exploration of Spanish DOM looking for dialectal differences is Heusinger and Kaiser 2005,
who show that Peninsular and Latin American Spanish are quite similar in this respect, except perhaps for
the interaction with clitic doubling, a fact that could be assimilated to the very well-known dialectal
differences between these varieties in the latter domain, which, as we discussed in chapter 2, must be
analyzed independently. If DOM is indeed a stable phenomenon across Spanish dialects, this is a
remarkable situation, given the wide range of differences that Spanish dialects exhibit and the very subtle
(that is, potentially unstable) conditions that govern DOM. This means that DOM is a core property of
Spanish, not a peripheral phenomenon. This strongly points out toward a theory based on core grammatical
devices (like the system I am developing here).
132
It is always possible to propose (especially in the OT framework, which these authors use) that the
distinguishability with the IO outranks the distinguishability with the subject. However, this predicts that A
would have to be dropped in all ditransitive sentences, which is again incorrect. Besides, the
distinguishability with the subject is also in question, as discussed in chapter 2.
133
Torrego uses only one ‘?’ for (8a). I use two ‘??’ because this corresponds better with my judgment
(also expressed in (3)). However, everybody agrees that (8b) is worse than (8a).

218
(8) a. ?(?) Presenté al alumno al profesor

Introduced A+the student to+the professor

I introduced the student to the professor.

b. * Describí al alumno al profesor

Described A+the student to+the the professor

I described the student to the professor.

[Torrego 1998: 134]

According to her, in (8b) the DO cannot raise out of the VP, which is crucial to get an A-

marker. The direct objects in (8) then must be in different case-assignment domains. The

DO in (8a) raises to [Spec, vP] and receives case (i.e. A) because it checks a D-feature

against small v (which is the general solution Torrego applies to Spanish DOM, as

discussed in chapter 2); also in (8a), the IO receives a second instance of accusative in

situ from the v-V compound, when V raises to v. The verb does not allow the object in

(8b) to raise out of VP (for reasons that are not explained, see footnote 134); as a result,

the DO gets case in situ, which means that the A-marker cannot appear:134

134
Notice that this analysis undermines Torrego’s explanation of the DOM system in general, since she is
assuming that the case of [+specific, +animate] is systematically assigned after raising the object. She
recognizes this, but she also “leave[s] open the question of why the derivation of some ditransitives verbs
prevents the accusative object from raising overtly” (Torrego 1998: 179 fn 5). She tries to suggest that
somehow affectedness is related to this difference; however, since it is hard to argue that the verb in (8b)
(to describe) is an affecting predicate, she gives up any semantic characterization, claiming that in (8), there
is a syntactic difference (describir does not allow object raising, but presentar does), “whatever the
semantic difference” is (Torrego 1998: 135).
Torrego also tries to link the dropping of A to a certain interaction between the clitic and the procedure of
Case checking; however, she pays attention solely to the presence of the clitic with DO, putting aside the
fact that it is actually the presence of IO clitic that shows a systematic difference (which is, after all, the
very issue that Strotzer 1976, whose data Torrego relies on, tried to account for).

219
(9) ? Describí el alumno al profesor

Described the student to+the the professor

I described the student to the professor.

There is a problem with the data in (8), however. As mentioned above, IOs in Spanish

require clitic-doubling in several cases, perhaps in most cases (see Strotzer 1976 for a

lengthy overview of different possibilities). Describir (to describe) is a verb whose IO is

better with a doubling clitic, but with presentar (to introduce), the presence of the clitic

makes no difference for the grammaticality; and this is independent of DOM, as we can

see in (10):

(10) a. ?(Le) describió su casa a Juan

CL-DAT described his house to John

He described his house to John

b. (Le) presentó su casa a Juan

CL-DAT introduced his house to John

He showed his house to John.

In other words, ditransitive sentences without IO-clitic doubling are not really suitable for

comparison between verbs (since some verbs prefer the clitic-doubling option for

independent reasons). On the contrary, given that all verbs allow IO clitic-doubling, it is

220
safer to check the relevant property with IO clitic-doubling structures. In fact, if we

restore the clitic in (8), the contrast disappears:

(11) a. ?? Le presenté al alumno al profesor

CL-DAT introduced A+the student to+the professor

I introduced the student to the professor.

b. ?? Le describí al alumno al profesor

CL-DAT described A+the student to+the the professor

I described the student to the professor.

Interestingly, both sentences become better if we drop the A in the DO:

(12) a. Le presenté el alumno al profesor

CL-DAT introduced the student to+the professor

I introduced the student to the professor.

b. Le describí el alumno al profesor

CL-DAT described the student to+the the professor

I described the student to the professor.

Another case that does not fit well with Torrego’s explanation of A-dropping is wh-in-

situ—see Reglero 2004 for a thorough revision of the conditions that govern Spanish wh-

in-situ. In simple transitive sentences, wh-in-situ questions must have the A-marker:

221
(13) a. Juan vio a quién

John saw A who

John saw who.

b. * Juan vio quién

John saw who

John saw who.

However, when there are two wh-objects in situ, the A must be dropped, no matter what

the verb is:

(14) a. Le presentaron (*a) quién a quién

CL-DAT introduced A who to who

They introduced who to whom.

b. Le describieron (*a) quién a quién

CL-DAT described A who to who

They described who to whom.

Notice that (14b) with A is worse than (11b)—I will provide reasons for this later.

So, although the interactions between DOM and clitic-doubling may need further

222
clarification, it seems clear that we need to find an independent source for the dropping of

A. I contend that the reasons for this dropping are not related to case checking, but with

the conditions on the linearization process. There is very strong evidence in favor of this

view, and the data have been, again, noticed for a long time, but the generalization has

remained elusive so far.

When discussing the dropping of A, Andrés Bello, the aforementioned nineteenth century

grammarian, presents an example where, as he said it, “es inevitable la repetición del a”

[it is inevitable to repeat A] (Bello 1847: §900):135

(15) El traidor Judas vendió a Jesús a los sacerdotes y fariseos

The traitor Judas sold A Jesus to the priests and Pharisees

The traitor Judas sold Jesus to the priests and Pharisees.

[Bello 1847: §900]

Although Bello does not conclude anything from his observation, the interesting thing

about this sentence is that one of the objects is heavy. On the other hand, (3), repeated

here as (16), has short objects:

135
A similar sentence is provided by Meyer-Lübke, also in the nineteenth century:

(i) Recomiende usted a mi sobrino al señor director

Recommend you A mi nephew to+the senior director
Recommend mi nephew to the director.
[Meyer-Lübke 1899: §350]

Examples like (i) (see also (42)) are the primary evidence for those who deny that A-dropping can happen
(for instance, Pensado 1995: 25).

223
(16) ?? Juan le presentó a María a Pedro

John CL-DAT introduced to Mary to Peter

John introduced Mary to Peter

It is unlikely that we could explain this difference purely on the basis of a case-theoretical

approach (as in Torrego 1998), which contends that there are different mechanisms of

case-checking related to different verbs, since in this case we have a difference with the

same verb. In fact, we can follow Bello’s intuition further, and find ways to modify (16)

and look for the results. The sentence improves notably with heavy objects (17a), as

expected, but also if we have a pause (17b), or if we displace one of the objects (17c):

(17)a. Juan le presentó a María de las Nieves a Pedro Vargas Prada

John CL-DAT introduced A María de las Nieves to Pedro Vargas Prada

John introduced María de las Nieves to Pedro Vargas Prada

b. Juan le presentó a María, a Pedro

John CL-DAT introduced A Mary to Peter

John introduced Mary, to Peter

c. A Pedro, Juan le presentó a María

to Peter John CL-DAT introduced A Mary

To Peter, John introduced Mary

224
In other words, manipulating the adjacency or the size of the objects blocks the dropping

of the A-marker:

(18)a. *Juan le presentó María de las Nieves a Pedro Vargas Prada

John CL-DAT introduced María de las Nieves to Pedro Vargas Prada

John introduced María de las Nieves to Pedro Vargas Prada

b. *Juan le presentó María, a Pedro

John CL-DAT introduced Mary to Peter

John introduced Mary, to Peter

c. *A Pedro, Juan le presentó María

to Peter John CL-DAT introduced Mary

To Peter, John introduced Mary

This calls for an explanation. The above data resemble the so called syntactic Obligatory

Contour Principle (OCP), and more in general, the idea that adjacency and PF identity

play a crucial role in blocking the pronunciation of otherwise expected elements

(Perlmutter 1971, Stemberger 1981, Menn and MacWhinney 1983, Hoekstra 1984,

Mohanan 1994, Pesetsky 1997, 1998, Yip 1998, Franks 1998, Plag 1998, Anttila and

Fong 2000, Bošković 2001, 2002c, Bobaljik 2002, Neeleman and van de Koot 2004,

Richards 2006, Bošković and Nunes 2007, inter alios.)

225
In fact, this possibility has already been explored by Richards 2006 with respect to the

dropping of the case-marking element in several DOM languages. Richards presents an

analysis of sentences like (16) based on the idea that linearization targets labels and not

terminal nodes, and therefore two DPs in the same linearization domain have to be

distinct enough to get linearized.136 According to Richards, this condition is not satisfied

in (16). Richards bases his analysis on Torrego 1998, and suggests that the case

assignment domain for (16) is also its the linearization domain. Given Torrego’s

suggestion that some verbs do not accept the dropping of A (because they have different

case-assignment domains), in these cases the linearization domains must be different.

Richards proposes the following condition on linearization:

(19) Distinctness

If a linearization statement <α, α> is generated, the derivation crashes.

[Richards 2006: 4]

Distinctness can be understood as targeting the label of the constituents involved in the

linearization, but, as Richards also points out, it can also be understood as targeting the

features of the head. There are also various situations that prevent Distinctness from

crashing the derivation. They involve either modifying one of the relevant constituents

(to make them different) or putting them in a different linearization domains, which will

136
If X merges with YP and projects XP, in the resulting structure [XP X [YP Y Z ] ], X, Y and Z are the
terminal nodes, and XP and YP are the labels. A linearization domain is, in principle, the complement of a
phase.

226
mean that no linearization statement <α, α> will ever be generated, assuming, as it is

standard, that linearization statements are relative to certain domains (see Fox and

Pesetsky 2004 for some discussion and a profitable use of this strategy). Richards

assumes that the drop of the dative marker in DOM languages instantiates the first option

(as in (5a)), repeated here as (20):

(20). El jefe le presentó el empleado a Pedro

The boss CL-DAT introduced the employee to Peter

The boss introduced the employee to Peter

I contend that the lack of dropping in the sentences in (17), which are not discussed by

Richards, are instances of the second option. Of course, I do not claim that Distinctness

drives the operations that are behind the surface order in (17); rather, these operations,

however they are implemented, bleed Distinctness, that is, place the objects in different

linearization domains.

In the case of (17a), one of the objects undergoes Heavy-NP shift (HNPS). Notice that it

is enough that one of the objects is heavy to prevent the A from dropping (as already

shown by Bello’s example in (15), repeated here as (22)):137

(21) Juan le presentó a María a Pedro Vargas Prada

John CL-DAT introduced A María to Pedro Vargas Prada

John introduced María to Pedro Vargas Prada

137
As we will see immediately, the order of the objects does not play a role here.

227
(22) El traidor Judas vendió a Jesús a los sacerdotes y fariseos

The traitor Judas sold A Jesus to the priests and Pharisees

The traitor Judas sold Jesus to the priests and Pharisees.

[Bello 1847: §900]

So, it seems clear that heavy objects prevent the dropping of A. The question is how

HNPS prevents the generation of a linearization statement of the form <α, α>. Notice that

the system outlined so far requires that both objects in ditransitive constructions check

case against the Dative head. The syntax of ditransitive constructions is a topic with

several ramifications, which I would not want to bring up here (see Branchadell 1992,

Masullo 1992, Demonte 1994, 1995, Torrego 1998, Cuervo 2003, among others, for

discussion of this issue in Spanish). I will limit myself to the question of case-checking in

ditransitive construction.

I adopt Cuervo’s 2003 proposal, based on Pylkkänen 2002, that both objects in Spanish

start in an Applicative Phrase below VP:

(23)

VP

V ApplP

IO Appl’

Appl DO

228
Under the system outlined in chapter 2, if the DO is φ-incomplete, it will check case with

small v, moving to [Spec, vP] (it will not get the A-marker, which requires that the DO

raises to DatP). As mentioned in the beginning of this section, I am assuming that the IO

has to check structural case, which means that it has to raise to [Spec, DatP].138 Given

that vP is a phase, the IO has to stop first in [Spec, vP] (this is Blind Movement from

chapter 2, section 2.3). I also assume that the dative clitic is hosted by the corresponding

case-checking head (following Franco 1993, and others), that is, Dat:

(24) Unmarked DO

DatP

IO Dat’

Dat vP
le

IO vP

DO v’

v VP

V ApplP

IO Appl’

Appl DO

138
This does not preclude the possible assignment of inherent dative, as long as we assume that there is also
structural dative (IOs would then be quirky objects), combining structural case and inherent case—see the
discussion in the next page.

229
As mentioned above, there is relatively free order between DO and IO; to account for this

I assume that the specifier in the DatP can be either at the right or at the left.139 Notice

that, in principle, it does not matter in which order DO and IO move to [Spec, vP],

provided that the IO is φ-complete. In (24), the DO undergoes Checking Movement,

which means that it is no longer a goal for higher heads. If the IO is φ-complete, small v

cannot check its case (since, by hypothesis, small v is φ-incomplete, as discussed in

chapter 1), and then the IO has to raise to [Spec, vP], where it is the only possible goal for

higher heads.

A potential locality problem arises with respect to IOs that are φ-incomplete, since now

small v should be able to check its case. Given that the IO is closer to v than the DO, the

question is what would prevent the IO from receiving accusative from small v. I will

adopt a solution provided by Cuervo 2003, with a slight change. Following Cuervo 2003

I assume that the Appl head assigns inherent dative to the IO; different from her, however,

I also assume that the IO is a quirky object, that is, that the IO must also check structural

case—for a discussion regarding the need to combine inherent and structural case to

account for quirky case see Nomura 2005:19-21 and the references therein.

In order to implement this idea, I propose that the Appl head values the case feature of

the nominal, but, given the quirky nature of the dative case feature, this valuation does

not prevent the nominal from undergoing Checking Movement to check its dative feature

139
Right specifiers have been used to account for word order in other Spanish constructions; for instance,
postverbal subjects (Zubizarreta 1994, 1999) and NP modifiers (Ticio 2003).

230
with a matching head (in fact, such movement is necessary).140 As a result, the IO will

still be a goal visible for heads that match its dative case feature. Given that, as is well-

known, nominals with quirky case can check structural case against a head with a value

different than the value of its inherent case, a question still remains regarding why it

cannot check structural accusative against small v. However, now the competition

between the two objects for small v is unbalanced, that is, the IO has already a value.

Recall also that the actual valuation happens by Checking Movement, that is, when the

goal c-commands the probe (i.e., when the goal probes its probe).

We can assume that there is a preference principle for quirky case: nominals with quirky

case prefer, if possible, to check their structural case against probes with the same value

than its inherent case. Give this, a quirky case nominal would withhold checking its

structural case if the probes it c-commands (after Checking Movement) does not match

the value of its inherent case, waiting for another goal that satisfies this preference. If the

probe does appear in the same phase, the quirky case nominal checks structural case

against this new matching head. If a matching head is not merged during the same phase,

the nominal in question checks structural case with the remaining head. In (24), there is

additional head, namely, the Dat head. Therefore, the IO will not check structural case

against small v, but against Dat. The problem noted above then disappears.

140
Notice that the above may be interpreted as implying a difference between the valuation process
(equivalent to case-assignment) and the checking process (under matching). In other words, for regular
objects Checking Movement values the case feature of the object, whereas for quirky objects Checking
Movement merely checks it (provided that the checking head has a matching case-value). For an argument
in favor of a purely checking approach (as opposed to case-assignment) see Bošković 2006b. See also
Torrego 1998 for the assumption that IOs are quirky elements. Nomura 2005:53-54 also claims that
inherent case is valued case.

231
With respect to φ-complete DOs, small v cannot value their case, as also discussed in

chapter 2, section 2.3. Therefore, the DOs must undergo Blind Movement to [Spec, vP]

and from there, Checking Movement to DatP (together with the IO).

(25) A-marked DO

DatP

A-DO DatP

IO Dat’

Dat vP
le

IO vP

DO v’

v VP

V ApplP

IO Appl’

Appl DO

In (25), Dat is checking the case of two goals. We could assume that, in order to do so,

Dat must attract its goals at the same time, which would mean that it can attract them in

any order—in a way similar to the analysis provided by Bošković 1999, 2002c for focus

232
movement in Bulgarian. As mentioned above, the specifiers of Dat can be on the left or

on the right. The combination of these assumptions predicts the following configurations

for ditransitive sentences with A-marked objects:141

(26) a. DatP b. DatP

A-DO DatP DatP A-DO

IO Dat’ IO Dat’

Dat vP Dat vP

(27) a. DatP b. DatP

IO DatP DatP IO

A-DO Dat’ A-DO Dat’

Dat vP Dat vP

This state of affairs predicts ambiguous scope between IO and A-marked in ditransitive

constructions: in (26) the DO c-commands the IO but in (27) the IO c-commands the A-

DO. As we will see in the next section, in ditransitive sentences where the IO is doubled

by a clitic (the only structures we are dealing with in this dissertation, as mentioned

above), the scope relation predicted by (27) are not obtained, that is, the A-marked DO

141
Notice that for (26b) and (27b) the lower specifier can be on the right too, but this would not produce a
change in word order (remember also that V raises to T, taking all the heads in its path with it).

233
apparently would be higher that the IO (independently of word order):142

(28) a. Le enviamos a un profesor famoso a todas las facultades de medicina

CL-DAT sent A a professor famous to all the schools of medicine.

We sent a famous professor to all the schools of medicine.

b. Le enviamos a todas las facultades de medicina a un profesor famoso

CL-DAT sent to all the schools of medicine A a professor famous.

We sent a famous professor to all the schools of medicine.

In (28), the DO has scope over the IO, and the inverse scope cannot be obtained—this is

an old observation with respect to the height of the objects in ditransitive constructions

(Demonte 1987, Leonetti 2003, 2004, López 2006). I will present a more detailed

discussion in next section, but for now (28) is enough to establish the need to block (27).

Here I will present a mechanism to that effect, which will be based on my previous claim

that IOs also receive inherent case.

142
I must stress that I am putting aside the structures without clitic doubling, where both the dropping of A
and the scope relations have been argued to follow different patterns from the corresponding doubling
structures, independently of DOM; some researchers (see Masullo 1992:60) even argue that the IO that is
not doubled by a clitic is not actually a dative argument, but a PP modifier. I must also point out that the
literature on scope relations between DO and IO in Spanish (regardless of DOM) is not in complete
agreement (compare, for instance, Demonte 1994 with Cuervo 2003 and Montrul and Perpiñán 2006). The
discussion here and in the next section should then be taken carefully, since it is limited to ditransitive
sentences with A-marked DO, where the IO is doubled by a clitic. The situation is further complicated by
the fact that clitic-doubling is subject to dialectal variation. A full scale exploration of scope relations
between DO and IO is necessary but I leave it for future research. See Strotzer 1976, Masullo 1992,
Demonte 1994, Torrego 1998, Bleam 1999, 2003, Cuervo 2003, Montrul and Perpiñán 2006, Leonetti in
press-a for extensive discussion. It is worth noticing, however, that the analysis of the relations between
DO and IO in the literature rarely includes discussion of DOM—for exceptions, see Torrego 1998 and
Leonetti 2003, 2004.

234
Recall that, following Cuervo 2003, I have assumed that the Appl head assigns inherent

case to the IO. I have also proposed that IOs are quirky objects, by which I meant that

they receive an inherent dative value, but they still need to check this dative value against

a Dative head. Recall that the Dat head attracts both objects; however, by hypothesis, the

IO already has a value, but the DO has not. I contend that this asymmetry causes the IO

to raise first, assuming that a purely checking movement (see footnote 140) must happen

as soon as possible. The DO, which requires valuation, raises second. The possibility in

(27) then never arises.143

There is a final question that the issue of case-checking for the dative brings up. We

could ask if it is possible for the IO to receive just inherent case and remain in situ (as

Cuervo 2003 suggests). Since this would mean that the IO would remain inside vP and

that the DO would raise out of vP, the consequence would be that IO and DO would

always be in different linearization domains: the IO would stay inside the complement of

small v, and get linearized first. This, however, would bleed Distinctness, and the A

would never drop. Since the A does drop, I conclude that both IO and DO raise out of vP

and therefore they are in the same linearization domain. This reinforces the idea that

Dative is structural case, which is the position I took in chapter 2.

From the previous discussion, it should be apparent that both objects end up in the same

143
Notice that I am not assuming tucking-in (Richards 2001), which means that the DO will move on top of
the IO. Notice too that the system predicts a number of lower copies for both DO and IO (see (25)), which
could potentially create different scope relations under reconstruction. I assume, following Chomsky 1995:
326-327, that there is no A-reconstruction, at least not for scope reasons (Lasnik 1999b has even suggested
that A-movement does not leave traces)—for further discussion on this issue see Boeckx 2001 and Lasnik
2003b.

235
domain, which means that, if they are linearized together (as for example in a

linearization statement like (31), to be discussed immediately), Distinctness (i.e. (19))

will be violated. The A-marker is then dropped to avoid a violation of Distinctness. As

mentioned above, if the linearization domains change, Distinctness is also respected.

Turning now to HNPS instances, I would like to propose that HNPS—which has received

several treatments in the literature (see Nissenbaum 2000: 59-94 for discussion of some

alternatives)—should be treated as a result of a constraint on linearization. This makes

HNPS a purely spell-out phenomenon.144 Let me briefly show how HNPS could work. If

the first linearization cycle is the vP phase, the first linearization domain is then the

complement of v, given Chomsky’s proposal that what is sent to Spell-Out is the

complement of the head phase. For our proposes, this is irrelevant, since in the relevant

cases (see (25)), there are only copies inside [Comp, vP]—the linearization algorithm

does not see copies (see Fox and Pesetsky 2004). The objects must get linearized in the

next linearization cycle, ZP, which includes DatP in its linearization domain:145

(29) A-marked object (from (25))

[ZP … Z [DatP A-OD IO [vP … ] ] ]

This means that there will be two possible linearization statements (see the discussion

regarding (26)):

144
There are a number of additional questions that this speculation brings up, which I have to leave for
future research.
145
I do not want to speculate about what ZP could be. It may be CP or IP. It has in fact been proposed that
linearization domains are not the same across languages.

236
(30) a. A-DO > IO [from (26a)]

b. IO > A-DO [from (26b)]

If the DO and the IO are both Dative, as I have assumed in this dissertation for the A-

marked object, the statements in (30) are affected by Distinctiness (19), since it implies

that both are a KP with the same case value:

(31) KP > KP

As a result, one A-marker must be deleted. Recall that we saw in the beginning of this

section that the A-marker for DOs is the only one that can be deleted (see the discussion

regarding (5)). Following the standard assumption that inherent case must be preserved

(Freidin and Sprouse 1991), I suggest that this is a consequence of the IO receiving

inherent case, as I discussed above—recall that IOs are quirky objects, i.e. IOs can

receive inherent and structural case. Let’s assume that the process of A-deletion can only

target the first A it encounters. If the process of A-deletion cannot target inherent case

markers, then it is predicted that the A corresponding to IO, which is both a structural and

an inherent case marker, cannot be deleted:

(32)* El jefe le presentó [IO Pedro] [DO al empleado]

The boss CL-DAT introduced Peter A+the employee

The boss introduced the employee to Peter.

237
In addition, we predict that it would not be possible to have the following word order if

the DO is φ-complete:

(33) * V IO DO

(if DO is φ-complete)

The reason for this is that in this context no A-marker can be deleted. The DO is not first,

so its A is not visible for the deletion. The IO is inherent, which makes it also undeletable.

This is exactly why (6), repeated here as (34), is ungrammatical:

(34)* El jefe le presentó [IO a Pedro] [DO el empleado]

The boss CL-DAT introduced A Peter the employee

The boss introduced the employee to Peter.

Given that no A-marker can be deleted, both A-markers stay, and the sentence becomes

ungrammatical, since it violates Distinctness:

(35)* El jefe le presentó [IO a Pedro] [DO al empleado]

The boss CL-DAT introduced A Peter A+the employee

The boss introduced the employee to Peter.

Although the idea that Dative can be inherent in addition to structural might have more

238
consequences that the ones discussed here, I will put them aside. I want to point out,

however, that this does not affect Distinctiness, which can be understood as targeting

only the values of KP, irrespective of their origin (structural or inherent)—see Richards

2006 for other uses of this assumption.

Now, suppose that one of the KPs is heavy (it does not matter here if it is the DO or the

IO). Let’s mark it with a subscript H (only for expository reasons):

(36) [ZP … Z [DatP KP KPH [vP … ] ] ]

The only thing that a HNPS rules needs to do is to be able to linearize DatP and the heavy

KPH, putting the heavy element in the end. In this way, KPH will be preceded by all the

elements of DatP (including the other KP):

(37) DatP > KPH

Notice that (37) is not affected by Distinctness. (37) linearizes a KP and a DatP, not two

KPs (the other KP gets linearized as a result of being the one left inside DatP). As a result,

there is no need to delete any A-marker.

Notice further that there is an additional possibility here. Suppose that (37) is not the only

linearization order. Suppose that (38) is also possible, at least in Spanish:

239
(38) KPH > DatP

This means that the effects of HNPS are neutralized, that is, a heavy object will precede a

non-heavy one (which is possible in Spanish). But (38) will not be affected by

Distinctness; hence, here we should not be able to delete the A-marker in this

configuration either. This prediction is borne out.

(39) a. *Juan le presentó María de las Nieves a Pedro

John CL-DAT introduced María de las Nieves to Pedro

John introduced María de las Nieves to Pedro.

b. *Juan le describió los alumnos y profesores a Pedro

John CL-DAT introduced the students and teachers to Pedro

John described students and teachers to Pedro.

(40) a. Juan le presentó a María de las Nieves a Pedro

John CL-DAT introduced A María de las Nieves to Pedro

John introduced María de las Nieves to Pedro.

b. Juan le describió a los alumnos y profesores a Pedro

John CL-DAT introduced A the students and teachers to Pedro

John described students and teachers to Pedro.

240
I also want to point out that, although I have implemented HNPS as an effect on

linearization (which makes it naturally compatible with the Distinctness condition), that

is, as a post-syntactic rule, the analysis can be maintained if HNPS is a syntactic rule. It

will be enough to assume that the landing site of HNPS is the specifier of ZP (in (36)),

and that this can be a right specifier (for (37)) or a left specifier (for (38))—provided that

[Comp, ZP] is still the domain of linearization.

Finally, given that HNPS has some degree of optionality (it can even be applied to some

light DPs), we may be able to account for the skepticism that some speakers (including

several researchers, as discussed above) exhibit with respect to the A-dropping. Some

speakers may be able to apply HNPS more freely than others, thus allowing the “double

A”.

Let’s return now to the other cases in (17), that is, (17b)-(17c), repeated here as (41a) and

(41b):

(41) a. Juan le presentó a María, a Pedro

John CL-DAT introduced A Mary to Peter

John introduced Mary, to Peter

b. A Pedro, Juan le presentó a María

to Peter John CL-DAT introduced A Mary

To Peter, John introduced Mary

241
In the case of (41a), we must assume that a pause indicates different linearization

domains; this sentence can in fact be a case of right dislocation. It is obvious that in (41b),

where the one of the objects has moved to the beginning of the sentence (the sentences

involves left dislocation), the objects are in different linearization domains.

Perhaps a case that is less clear is the following—modeled from a sentence provided by

Meyer-Lübke 1899: §350, which is used sometimes to deny the existence of A-dropping

(Pensado 1995: 25):

(42) Juan le presentó a María en París a Pedro

John CL-DAT introduced to Mary in Paris to Peter

John introduced Mary to Peter in Paris

Here, there is a constituent intervening between the objects. However, it seems

reasonable to assume that adjuncts are in a different linearization domain from

complements (after all, they may even merge acyclically (Lebeaux 1988, Stepanov

2001a)); given that one of objects is on the right edge, it is possible to suggest that we

may also have right dislocation here. If this is correct, (42) can be unified with (41). This

means that dislocation is the phenomenon that is saving these constructions from

violating Distinctness.

Now, suppose that there is some context that allows us to neutralize the effects of HNPS

242
or dislocation; the prediction will be that all speakers would reject this kind of examples.

I believe that wh-in-situ is such a context and that this is the reason why (14), repeated

here as (43), is categorically bad with both As:

(43) a. * Le presentaron a quién a quién

CL-DAT introduced A who to who

They introduced who to whom.

b. * Le describieron a quién a quién

CL-DAT described A who to who

They described who to whom.

I do not believe that there are any Spanish speakers who accept (43) but not (44)—to the

extent that they accept wh-in-situ questions in the first place (see Reglero 2004 for

discussion):

(44) a. Le presentaron quién a quién

CL-DAT introduced who to who

They introduced who to whom.

b. Le describieron quién a quién

CL-DAT described who to who

They described who to whom.

243
Under the system outlined in this dissertation, in (43) the nominals are identical (two KPs

with the same values and the same PF). Therefore it is possible to apply the same

criterion that Bošković 2002c uses for multiple wh-fronting in Bulgarian, Russian, and

Romanian, that is, that sequences of homophonous wh-phrases must be avoided.146 The

strategy that Spanish uses is to delete the first A-marker, which makes the wh-KPs non-

homophonous. Notice further that failing to drop the A-marker with wh-in-situ questions

is worse than failing to drop it in declarative contexts—as expressed by the ‘*’ in (43)

and the ‘??’ in the other cases (see (3), repeated here as (45)):

(45) ?? Juan le presentó a María a Pedro

John CL-DAT introduced A Mary to Peter

John introduced Mary to Peter

I contend that the reason for this contrast is that (43) violates both Distinctness and the

constraint on sequences of homophonous phrases, whereas (45) only violates Distinctness.

The above data provide strong evidence in favor of the idea that a Dative head licenses

the case of DOM objects in Spanish, since the analysis give above depends on both DO

and IO having the same marker (e.g. A) which is licensed by the same head. It is also

worth noticing that the form [a] has other values in Spanish; in particular, it can be a full

preposition (then, heading a PP, not a KP); when this happens we do not get A-dropping:

146
These languages front all wh-phrases, but fronting is not possible if it would result in a sequence of
homophonous wh-phrases.

244
(46) Juan le presentó a María a las tres

John CL-DAT introduced A Mary to the three

John introduced Mary at 3.

This means that we are not dealing here merely with a constraint on phonetic form.

Note also that Richards 2006 observes that dropping similar to Spanish A-dropping is

attested in Chaha, Hindi and Miskitu, all of them DOM languages that also have Dative

as the marker for DOM-objects; see also Patel 2007: 31-32 for Kutchi Gujarati, another

DOM language with a Dative DOM-marker. In all these languages, the DOM-marker is

dropped in ditransitive sentences, as a result of Distinctness.

Another argument that suggests that the A-marker is the same for DO and IO may be

provided by the fact that a single A-marked relativized element can be related to both DO

and IO—as observed by García 1975 (who quotes real texts). This shows that the dative

marker can serve as a marker for both objects, even at the same time, which may be used

as evidence that the case value is the same:

245
(47) No tenían ningún respecto por las mujeres, muchísimo menos por las jovencitas,

They have no respect for women, far less for the young ones,

a las que

A the that

to whom/whom

levantaban las polleras y maltrataban con las injurias más ofensivas

lifted the skirts and mistreated with the slurs most

offensive

they lifted the skirts and mistreated with the most offensive slurs.

[García 1975: 93]

In this sentence, we have two coordinate relative clauses with a common antecedent

(jovencitas) and a common relative item (a las que). The gaps in the clauses are the

Indirect Object and the Direct Object. The example is a case of Across-the-Board (ATB)

movement, and its structure is as follows (abstracting away from other properties):147

(48) [KP a las que] [levantaban las polleras [IO a las que] y [maltrataban [DO a las que]] ]

to whom lifted the skirts to whom and mistreated whom

147
See, however, Franks 1993 for the idea that morphological case is not enough to license ATB
dependencies.

246
To summarize, we have seen in this section that there is evidence in favor of the DOM A-

marker as Dative, which is licensed by the same head that licenses regular Dative object

(i.e. Indirect Object).

3.1.2 The height of the objects

The situation with respect of the exact position of the object with respect to the verb has

never been clear in the literature of Spanish. Torrego even suggested that there is no

“direct empirical evidence determining where the object raises…The target of object

raising in Romance languages has to be determined on purely theory internal grounds”

(Torrego 1998: 6). For simple transitive constructions this may be true. Since the verb

always raises to T, as it is generally assumed for Spanish (and other Romance

languages)—at least since Emonds 1978 (see Zagona 2002: 164-168 for specific

discussion of Spanish)—we cannot use relative position to the verb to determine object

height.148 Also, the position of the adverbs is extremely free in Spanish,149 which makes it

also very difficult to draw conclusions from the position of the object relative to the

adverbs.150

148
A possible argument from word order is provided by Bošković 1997a: 243-244, using the fact that
Spanish violates the Superiority Condition. According to Bošković, this is possible because the DO has to
leave the VP and check case in [Spec, AgrOP], over the subject, which stays in [Spec, VP], thus being
closer to the final landing site.
149
Spanish adverbs can be virtually any place in the simple clause: x S x V x O x
However, some adverbs cannot appear before a preverbal subject, and a few others cannot appear between
the preverbal subject and the verb (see Zagona 2002: 162-164 for some details).
150
Most adverbs can freely appear between the verb and the objects. A few adverbs (like allá “there”) seem
to require to be in a intonational phrase that does not contain the object (Zagona 2002: 163), but even if this
is violated, the deviance is just minimal (the brackets in (i-iii) represent intonational phrases):

(i) [Conoció allá] [a una mucama]

Hired there A a maid
He hired a maid there.

247
The scope of adverbs can be more informative, though. It has been standardly assumed

that these sentences show evidence for the direction of adverb adjunction:

(49) a. John [intentionally [twice [knocked on the door]]] intentionally > twice

b. John [[[knocked on the door] intentionally] twice] twice > intentionally

In (49a), where intentionally scopes over twice, there is only one intention, which was to

knock twice. In (49b), John has two instances of intentional knocking on the door. This

contrast has been used as evidence that in the first sentence adverbs are left-adjoined,

since in this way intentionally, being higher, can have scope over twice; in the second

sentence, which requires the opposite, the adverbs are right-adjoined (see Andrews 1983,

Pesetsky 1989: 19, Bošković 1997b: 121-122, among others). This contrast has been used

as a test to determine if the object has left the VP (Bošković 1997b: 121-122, Stjepanović

1999: 83-84, Reglero 2004: 65-66). For Spanish, the results do not distinguish between

A-marked and unmarked objects: 151

(ii) ? [Contrató allá a una mucama]

Hired there a maid
He hired a maid there.

Interestingly (at least for my judgment) this requirement is waived if the object is unmarked:

(iii) [Contrató allá una mucama]

Hired there a maid
He hired a maid there.
Although this suggests a difference between the objects (A-marked vs. unmarked), which may be related to
height, I will leave the issue for future research, because the deviance, which is minimal, is related to
intonational phrasing, so it is necessary to explore the correlations between syntax and intonation, an issue
that is beyond the goals of this dissertation.
151
Notice that the possibility of having a single constituent [intentionally twice] does not arise in Spanish,
since in this language twice is actually a DP (dos veces, “two times”).

248
(50) a. Juan golpeó la puerta deliberadamente dos veces

John knocked the door intentionally twice

b. Juan besó a María deliberadamente dos veces

John kissed A Mary intentionally twice

Both sentences in (50) are ambiguous, having the readings in (49), that is, with

intentionally > twice and twice > intentionally. In order to obtain the reading in which

intentionally scopes over twice, the adverbs must be left-adjoined; then, the object must

be out of the VP in both sentences. Interestingly, as noticed by Reglero 2004, the scope is

also ambiguous with ditransitive constructions; thus the reading with intentionally >

twice is evidence that both objects leave the VP:

(51) Juan le enseñó el libro a María intencionalmente dos veces

John CL-DAT showed the book to Mary intentionally twice

[Reglero 2004: 66]

This discussion shows that all objects leave the VP: both marked and unmarked DOs, as

well as IOs (see Torrego 1998, Suñer 1999, 2000, Cuervo 2003, De Pedro Munilla 2004

for similar conclusions). However, it does not show us anything about the relative

position of A/non-A marked objects. As discussed above, in the system I developed in

chapter 2, the A-marked object is higher than the unmarked object, i.e. the DOM objects

must be in a Dative projection above vP, and unmarked objects are in [Spec, vP]—see

249
(24)-(25). Notice that this means that both objects, marked and unmarked, are outside of

VP, which captures the above scopal data. However, as noted above, the data in question

do not tell us anything about the relevant position of the A/non-A marked objects.

Ditransitive sentences, however, are more informative in this respect.

As noticed by Leonetti 2003, 2004 and others, the A-marked DO in ditransitive

constructions behaves as if it is higher than the unmarked object in these constructions.

This conclusion comes from asymmetries with respect to scope. In (52a), the unmarked

DO must have narrow scope with respect to the IO, that is, the sentence is interpreted as

saying that a different expert was sent to each department; in (52b), the DO has wide

scope with respect to the IO, that is, in these case the same expert was sent to all the

departments:152

(52) a. Le enviamos un especialista a todos los departamentos afectados.

CL-DAT sent an expert to all the departments affected.

We sent an expert to all the departments affected.

b. Le enviamos a un especialista a todos los departamentos afectados.

CL-DAT sent A an expert to all the departments affected.

We sent an expert to all the departments affected.

[Adapted from Leonetti 2004: 102]

152
Notice that in (52b) the IO undergoes HNPS. Under the view advocated in the previous section, HNPS
is a PF rule, that is, it does not changes scope relations. This is confirmed by the fact that in (52b) the
HNPS-IO cannot scope over the DO. As discussed in the previous section (see (28)), the scope relations in
(52) hold with the V IO DO order, which may be evidence that the difference in word order is just a matter
of the direction of the specifier, not a different hierarchical position.

250
Recall that, according to the system presented in chapter 2, the A-marked object can

optionally receive a choice function, but it does not have too. Then, (52b) should also

have a reading similar to (52a), that is the narrow scope reading, but it does not have such

a reading. I then conclude that this happens because in this case, the DO is higher than

the IO.

Recall also that A-marked objects can receive choice-functions, but do not need to. This

means that the mandatory wide reading of the DO in (52b) cannot be a result of a

mandatory choice function assignment but of the fact that the DO is higher than the IO.153

Notice also that in (52a), where the animate object is unmarked, the DO cannot receive a

choice function, according to the system we developed in chapter 2, where unmarked

objects have a *D, an element that cannot receive choice functions and that is

incompatible with [person] features (so the [person] feature associated with animate

nominals has to be deleted). Note that Distinctness is not relevant to the D/*D difference,

since it targets the highest projection, which is the traditional noun phrase. So, the

contrast between (52a-b) is consistent with the system developed in chapter 2.

The system makes an additional prediction, with respect to the difference between

animate and inanimate objects. Animate nominals that are combined with D must receive

A; having D entails they can receive an optional choice function, as explained. On the

other hand, if they have a *D, they cannot receive a choice function, and they cannot

153
In that sense, this is a case that corresponds to the so called “positional specificity” rather the “inherent
specificity”, which corresponds to nominals that receive choice functions—see Lidz 2006 for discussion of
these notions with respect to Kannada, another DOM language.

251
receive A either, because *D is incompatible with the [person] feature needed to force the

nominal to check case with the Dative projection. But inanimate nominals do not receive

a [person] feature ever (and if they receive A, it is only by accident, as explained in

chapter 2). This means that inanimate nominals are free to have D or *D, which is fine,

because they can receive a “specific” interpretation too (that is, a choice function). The

prediction is then that inanimate objects could optionally receive a choice function

interpretation in ditransitive constructions, which will have the effect of wide scope

(independent of DO-IO order, for the reasons discussed above). This prediction is borne

out:

(53) a. Le construimos un edificio a todos los departamentos de idiomas

CL-DAT built a building to all the departments of languages

We built a building to all language departments.

b. Le construimos a todos los departamentos de idiomas un edificio

CL-DAT built to all the departments of languages a building

We built a building to all language departments.

(53a) and (53b) are ambiguous. The sentences can mean that a single building was built

for all language departments or that a different building was built for each one of them.

This ambiguity does not obtain with unmarked animate objects, as discussed above. This

means that the DO in (53) is still lower than the IO, but it can optionally receive a choice

function, which accounts for the wide reading—see also footnote 153.

252
Similar conclusions arise from other quantifiers. The distributive quantifier cada (each)

forces a distributive reading on the indefinite it c-commands. This reading is blocked

with A-marked DOs, but allowed with unmarked DOs (also independently of DO IO

order):

(54) a. Le devolvieron un prisionero enfermo a cada tribu.

CL-DAT returned a prisoner sick to each tribe.

They returned a sick prisoner to each tribe.

a’. Le devolvieron a cada tribu un prisionero enfermo.

CL-DAT returned to each tribe a prisoner sick.

They returned a sick prisoner to each tribe.

b. ?? Le devolvieron a un prisionero enfermo a cada tribu.

CL-DAT returned A a prisoner sick to each tribe.

They returned a sick prisoner to each tribe.

b. ?? Le devolvieron a cada tribu a un prisionero enfermo.

CL-DAT returned to each tribe A a prisoner sick.

They returned a sick prisoner to each tribe.

[Adapted from Leonetti 2004: 102]

This means that in (54a) the IO c-comands the DO, but that is not the case in (54b); in

253
other words, in the latter, the A-marked DO is in a position higher than the position

occupied by the unmarked DO in (54a). Interestingly, with inanimate nominals the

wide/narrow ambiguity in (53) disappears, as expected:

(55) Le construimos un edificio a cada departamento de idiomas

CL-DAT built a building to each department of languages

We built a building to each language department.

This again means that the ambiguity in (53) depends on the possibility of choice function

assignment, which, as is well known, cannot be done under the scope of a distributive

quantifier (that is, in (55)). This also confirms that the unmarked DO is lower than the IO.

This discussion does not tell us what the positions in question are, but only that A-marked

objects are higher than unmarked ones, a position that is in agreement with other

accounts of Spanish DOM (see Torrego 1998, Leonetti 2003, 2004, among others).

Given this state of affairs, we predict that short DOs without the A in ditransitive

constructions—which, as discussed in the previous section, can drop the A because of

constraints on the Syntax-PF interface, in particular, Distinctness, which affects

linearization (recall that they are already located in [Spec, DatP])—should behave

ambiguously with respect to scope facts. Since they do raise higher than the IO, they

should be able to scope over it, although they end up without the A for different reasons.

This prediction is borne out:

254
(56) Le enviamos un experto a todos

CL-DAT sent an expert to everybody

We sent an expert to everybody

Sentence (56), where the A is excluded for the reasons discussed, is ambiguous. Both

scope relations are possible: it can be a unique expert that was sent to everybody, or it can

be one different expert for each one. In the first reading, the A is dropped because of

Distinctness: DO and IO are in the same linearization domain (both are in [Spec, DatP]),

as discussed in the previous section (in other words, this reading has the same derivation

than (52b), except that the A drops). In the second reading, the A was never there; the DO

is at [Spec, vP] and the IO is at [Spec, DatP].

Interestingly, if we change the DO-IO order the ambiguity disappears:

(57) Le enviamos a todos un experto

CL-DAT sent to everybody an expert

We sent an expert to everybody

(57) only means that a different expert was sent to each one. This is predicted under the

current system. Since A-dropping targets only the first A and the IO-A cannot be deleted,

it follows that A-dropping happens only in the DO-IO order. Given that (57) is

grammatical, the only possibility is that A was never there, which means that the DO has

255
a *D determiner (in other words, the A was never there).

The above data provides strong evidence that the A-dropping does not have a purely

syntactic origin, but is a by-product of the so called Syntactic OCP, as discussed above.

3.2 DOM, Small Clauses and Possessor Raising

There is another type of construction that exhibits a contrast between A-marked and

unmarked objects: a clause that embeds a Small Clause inside. It is a long-standing

observation that a Direct Object inside a Small Clause can be optionally marked with A.

The basic paradigm is represented by the verb tener (to have):154

(58) a. Tiene (*a) un hermano

has A brother

S/he has a brother

b. Tiene (a) un hermano en la cárcel

has A brother in the jail

S/he has a brother in jail

We can be sure that (58b) contains a Small Clause because the object can be cliticized

independently of the PP:

154
I will indicate, as is traditional, the impossibility of using A with (*a). This means that in (58a) the A-
marker is not possible. Additionally, I will indicate the lack of optionality for A (i.e. obligatory A) with
*(a).

256
(59). Lo tiene en la cárcel

CL-ACC has in the jail

S/he has him in jail

It is worth noticing, however, that there is a difference between (58a) and (58b). The first

sentence, where A is not possible, must be interpreted as involving inalienable possession

(with respect to the subject). This interpretation is not mandatory in (58b). We can

confirm this with a different possessor:

(60) a. * Ella tiene un hermano de Juan

She has a brother of John

b. Ella tiene *(a) un hermano de Juan en la cárcel

She has A a brother of John in the jail

Notice that in (60b) A becomes mandatory: if the interpretation with inalienable

possession is not possible, A must be present. This suggests that what blocks the presence

of A in (58a) is the mandatory inalienable possession interpretation. If fact, we can

confirm this with other verbs, in particular, with the so called dative external possessor

that is possible in some ditransitive constructions (independently of the DO-IO order, as

witnessed by (61a-b)),155 as noted by Brugé and Brugger 1996. They attempt to explain

155
The term dative external possessor construction is used because the possessor (which has dative case) is
not inside the phrase that contains the possesse. As witnessed by (61), the possessor (el hijo de Paco

257
the lack of A by claiming that in cases like (61) the possesse must have the feature [-

animate]:

(61) a. Juan le robó (*a) la novia al hijo de Paco

John CL-DAT stole A the fiancé to+the son of Paco

John stole Paco’s son’s fiance.

b. Juan le robó al hijo de Paco (*a) la novia

John CL-DAT stole to+the son of Paco A the fiancé

John stole Paco’s son’s fiance.

I reject the claim that the possessee must be [-animate] in these sentences, for obvious

reasons. In fact, I think that the system I am developing here predicts these facts, when

we introduce some fairly common assumptions about the syntax of possession, as we will

see below.

The syntax of external possession is a hotly debated topic in Syntax. There are all kind of

proposals regarding this issue, ranging from movement and control to treating it as a

semantic or pragmatic effect (see Landau 1999, Payne and Barshi 1999, Coene and

D’hulst 2003, Vermeulen 2005, among others, for extensive cross-linguistic discussion

and references). For Spanish (and other Romance languages), there is a solid tradition

that assumes a movement analysis (Hornstein et al. 1994, Uriagereka 1998, Bleam 1999,

Cuervo 2003), where the possessor starts as a constituent of a lower phrase and moves

“Paco’s son”) is in a different phrase than (la novia “the fiancé”).

258
out to obtain Dative case. According to Bleam 1999: 94-112 and Uriagereka 1998, both

possessor and possessee are based generated in a Small Clause inside a DP. In the

internal possessor construction (62), the possessee raises to an internal agreement

projection, whose head is spelled-out as de (of); on the other hand, in the external

possessor construction (61), in addition to the possessee raising to [Spec, Agr], the

possessor raises to [Spec, DP] and from there, unable to receive Case, has to raise further

to check Dative.

(62) Juan robó la novia del hijo de Paco

John stole the fiancé of+the son of Paco

John stole Paco’s son’s fiance

This is illustrated here:

(63) a. Internal Possession

DP

D AgrP
la

novia Agr’

Agr SC
de

novia el hijo de Paco

259
 b. External Possession156

DP

el hijo de Paco D’

D AgrP
la
IT MOVES OUT
TO CHECK CASE novia Agr’

Agr SC

novia el hijo de Paco

I will assume this structure here, and in particular Uriagereka’s 1998 analysis for the

syntax of possession in Spanish—which follows similar ideas first suggested by

Szabolcsi 1983 and Kayne 1993. In other words, (63b) can also be the syntax for (58a),

mutatis mutandi, as Uriagereka proposes. The difference is that in this case, with tener (to

have), the possessor raises to subject position and gets Nominative. If this is correct, it is

crucial that tener does not have a Dative case to assign. Before coming back to (58), let

us make sure that this is indeed what happens.

Recall from chapter 1 that I assume that haber does not allow a Dative head. It is very

well known that haber and tener belong to the same family of verbs (together with

ser/estar (to be)). A relation between the verbs that instantiate have and be has been

156
Notice that the possessor raises to [Spec, DatP], which can be to the right or to the left, accounting for
the word order in (61a-b).

260
proposed by Freeze 1992,157 who shows that these verbs are related, suggesting a unique

deep structure for all of them in the relevant respects. Also, Kayne 1993 uses Freeze’s

account (combined with ideas from Szabolcsi 1983) to explain several properties of the

possessive—for similar ideas in the context of Romance languages see also Guéron 1995,

Longa et al. 1998, among others. A comparison between Spanish and English is

extremely revealing in this respect. We can trace a very complex scenario here. In table

(64) I intend to represent some of the basic properties that relate the verbs in question,

without the intention of being exhaustive. As we can see, given the set of properties

associated with be and have, Spanish has a more finer-grained distinction than English:

(64)

Properties Spanish English

a. Passive
b. Individual-level predicates Ser
c. Locative (for events)
d. Progressive Be
e. Stage-level predicates Estar
f. Locative (for individuals)
g. Existential
Haber
h. Auxiliary
Have
i. Possessor Tener

The following sentences exemplify the cases in table (64):

(65) a. El ratón era perseguido por el gato PASSIVE

The mouse was chased by the cat

157
This follows an idea that can be traced back to Benveniste 1966.

261
 b. El gato era inteligente INDIVIDUAL-LEVEL

The cat was intelligent

c. La clase es en este edificio LOCATIVE (EVENTS)

The class is in this building

d. El gato estaba persiguiendo al ratón PROGRESSIVE

The cat was chasing the mouse

e. El gato estaba cansado STAGE-LEVEL

The cat was tired

f. El gato estaba en la casa LOCATIVE (INDIVIDUALS)

The cat was in the house

g. Había un gato en la casa EXISTENTIAL

was a cat in the house

There was a cat in the house

h. El gato ha muerto PERFECT

The cat has died

i. El gato tiene un collar POSSESSIVE

The cat has a collar

It is quite possible that each one of the “properties” in (64) has a different surface

syntactic configuration and that such “properties” can be grouped in a different fashion in

262
different languages. To account for all these properties is beyond the scope of this

dissertation. However, I want to highlight a correlation between Possessor and Existential

(expressed in English by have and be respectively):

(66)

Properties Spanish English

Possessive (64i) Tener Have

Existential (64g) Haber Be

Following Freeze 1992 and Kayne 1993, I assume that possessive tener (have) is derived

from the incorporation of an abstract Dative preposition into an abstract verb BE. I

further assume that this abstract Dative preposition is actually the same as the Dat

projection I have been using so far (which, in fact, is a way to express a proposal made by

Torrego 2002, as discussed in chapter 2, section 2.2). If Dat incorporates into BE, it can

no longer serve as a probe, therefore it cannot check Dative case.158 As also suggested by

Freeze 1992, Spanish haber, although expressed by English be, is actually a HAVE-type

existential verb—for systematic differences between BE-type and HAVE-type

existentials see also Svenonius 2002: 5-9 and Schoorlemmer 2005—and consequently,

Dative case assignment is also excluded.

Evidence for this claim comes from the fact that purely BE verbs, for instance, ser, which

do not incorporate a Dative preposition, do not exclude Dative case:

158
In a system where movement is driven by a feature in the moving element (Bošković 2007b), there
would need be a feature on the Dat head indicating the need for incorporation. We would then know that
Dat needs to incorporate even before BE enters the structure, which is when the Dat head needs to serve as
a probe.

263
(67) Tu amor le es indiferente a María

Your love CL-DAT is indifferent to Mary

Mary is indifferent to your love.

Even more interestingly, this is true even when we have a possessive reading, that is, we

can have Dative external possessors with BE-verbs:

(68) La novia le es indiferente a Juan

The fiancé CL-DAT is indifferent to John

John is indifferent to his fiancé.

(69) La mano le fue cortada a Juan

The hand CL-DAT was cut to John

John’s hand was cut.

Notice further that BE-verbs cannot check Accusative case for nominals. Therefore, in

(68)-(69) the possessee has to raise to [Spec, TP] and gets Nominative. The possessor, as

explained, raises to a Dative head.

It is worth noticing that verbs that include a Small Clause (SC) do show the presence of

ACC (as can be inferred from the cliticization):

264
(70) a. Juan es [SC [Juan] [un buen abogado]]

John is John a good lawyer

John is a good lawyer.

b. Juan lo es

John CL-ACC is

There is, however, a very important difference between this cliticization and the one

regarding ACC objects. The clitic does not substitute for the nominal in (70). I will assume

a structure with an internal SC for these sentences. Then, the nominal starts in the SC and

raises to [Spec, TP]:159

(71) a. [TP [VP [SC [nominal] [predicate] ] ] ]

b. [TP [nominal]i [VP [SC ti [predicate] ] ] ]

Given this, it seems reasonable to assume that the predicate has been substituted by the

clitic in these cases.160 This is confirmed by the fact that no agreement arises in these

contexts: the ACC clitic always appears in the invariant form lo, which I assume is φ-less.

Forms with φ–features are ungrammatical, which indicates that there is no agreement

159
In Chomsky’s 2000 system, this happens to satisfy the EPP. However, as discussed in chapter 1, it has
been argued that there is no EPP. There are alternative explanations for raising of subjects. See Epstein and
Seely 1999, Boeckx 2000, Grohmann et al. 2000, Bošković 2002a, 2005, 2007b, Epstein and Seely 2006,
Wurmbrand 2006, among others.
160
In fact, clitics like the one in (70) are usually called “predicate clitics” (see Sportiche 1995 for some
discussion in the context of a Romance language).

265
with the nominal:

(72) a. María es una buena abogada

Mary is a good lawyer-FEM

b. María lo es

Mary CL-ACC is

c. *María la es

Mary CL-FEM-ACC is

(73) a. María está muy ocupada

Mary is very busy-FEM

b. María lo está

Mary CL-ACC is

c. *María la está

Mary CL-FEM-ACC is

If verbs in the ser-estar-haber family indeed have an underlying common structure, as

the relevant literature suggests, then we may have an explanation for the apparently

strange small v in Spanish existential constructions that I have postulated in chapter 1. It

comes from the same small v that is necessary to license SC with these verbs (see

discussion below regarding SC licensing). The difference between the two small vs is that

266
with haber, the head v has φ-features, whereas with estar/ser it doesn’t—for proposals

that different kinds of small v produce different syntactic and semantic effects, see

Boeckx 2003, Legate 2003, Folli and Harley 2004.

Having φ-features or not has a very important consequence for case-valuation. The v with

haber is able to value ACC Case because it is able to establish an Agree relation with the

nominal, given that it has φ-features. This means that the small v in haber-sentences is

licensing both the nominal and the SC. It seems plausible that the small v has an

additional set of features (let’s call them Predicate-features) that can establish an Agree

relation with the head of the SC—a PredP category, as it is sometimes proposed (Bowers

1993).

With ser/estar, however, the small v can only license the SC, because it does not have φ-

features. The [case]-features of the nominal must be valued by T; therefore, with

ser/estar, the nominal will get NOM, whereas with haber it will get ACC. This is illustrated

below:

(74) [TP T-φ [vP v-φ [VP haber [SC [nominal] [predicate] ] ] ] ]

Agree

(75) [TP T-φ [vP v [VP ser/estar [SC [nominal] [predicate] ] ] ] ]

Agree

Recall from chapter 1 that in (74), T-φ values its φ-features by probing v-φ (in SII, the

267
Spanish dialect with agreement) or by getting a default value (in SI, the Spanish dialect

that does not show agreement with the internal nominal). Since v is φ-less in (75), it does

not interfere with the relation between the nominal and T-φ. If this is correct, the family

of verbs ser, estar, haber (that is, the BE-HAVE family) share something special: they all

require a small v to license its internal SC. In the case of ser and estar, this small v is φ-

less, so the nominal needs to value its Case by undergoing Agree with T, receiving NOM.

In the case of haber, this small v has φ-features, so the nominal can receive ACC—and T

either probes small v (SII) or receives a default value (SI), as discussed earlier.

This means that the idea of a special small v (a small v without external argument) is not

an ad hoc device for explaining the properties of haber-sentences, but that its presence

may be motivated by the need to license the internal SC that all verbs of this class have. It

is of course not clear why a SC needs to be licensed in this way. But, as mentioned above,

we could be dealing here with a general property of embedded clauses. Even full

embedded clauses trigger the presence of an ACC clitic, which, I assume indicates the

presence of small v. It is also possible to suggest that it is a property of clauses in general

(at least in Spanish), that is, that clauses need case (see Bošković 1995 and, for Spanish,

Velásquez 1991).

(76) a. Juan dijo que volverá

John said that come-FUT

John said he will come

268
 b. Juan lo dijo

John CL-ACC said

Significantly, if a verb does not have accusative, it cannot have a clause as an object

either. In Spanish, filosofar (to think philosophically) is incompatible with an accusative

clitic, and it does not have the possibility of taking a clausal complement:

(77) a. * Lo filosofa

CL think-philosophically

b. * Filosofa que la vida es bella

Think-philosophically that the life is beautiful

c. Filosofa sobre el hecho de que la vida es bella

Think-philosophically about the fact that the life is beautiful

A deeper exploration of these issues is needed, which I will leave for future research.

Having established that possessive tener (to have) cannot have a Dative case, let’s return

to the contrast in (58), repeated here as (78). Remember that A is not possible under

inalienable possessor interpretation:

269
(78) a. Tiene (*a) un hermano

has A brother

S/he has a brother

b. Tiene a un hermano en la cárcel

has A brother in the jail

S/he has a brother in jail

In fact, A becomes mandatory if the interpretation with inalienable possession is not

possible—as in (79), repeated from (60b):

(79) Ella tiene *(a) un hermano de Juan en la cárcel

She has A a brother of John in the jail

We can capture this distinction by assuming that Determiners that trigger inalienable

possession are always *D and never D, that is, only *D determiners can have the syntax

described in (63b).

Actually, this proposal is not really new. Vergnaud and Zubizarreta 1992, discussing

inalienable possession in French, have suggested that the definite Determiner in these

constructions is what they call an “expletive article”. We can expresses the same idea by

using *D. This not only captures the above distinctions with respect to tener, but also the

Dative external possessor construction—I repeat (61a) here:

270
(80) Juan le robó (*a) la novia al hijo de Paco

John CL-DAT stole A the fiancé to+the son of Paco

John stole Paco’s son’s fiance

If the Determiner in (80) is *D and not D, the object will never receive A: its [person]

feature will be deleted because it will get stranded since *D cannot be associated with it.

The same happens in (78a). On the other hand, if the Determiner is D, then the object

must have A, but it does not demand an inalienable possession interpretation, as (79)

shows. Then, we do not need to adopt the rather counterintuitive idea that the possessee

in the external possessor construction is [-animate], as Brugé and Brugger 1996

suggested.

An additional prediction arises. Since definite articles can be *D too (as discussed in

chapter 2, and shown in (80)), we predict that we can have A-less definite articles with

tener and that they will receive an inalienable possession interpretation. This prediction is

borne out:

(81) Ella tiene el hermano en la cárcel

She has the brother in the jail

She has her brother in jail.

Interestingly, (81) also contains a Small Clause. This means that it would be incorrect to

271
make A-marking dependent on the presence of a Small Clause in the cases discussed

earlier (see (58)). The A-marking is dependent on the properties of the DP in combination

with the Agree system. There is nothing special about this situation, which is not

exceptional but regular, under the system developed here.

3.3 DOM and coordination

It is well known that A-marked objects cannot be coordinated with unmarked objects:

(82) *? Mencionaron a Juan y el libro

Mentioned A John and the book

They mentioned John and the book.

[Camacho 1999: 2646]

This is usually derived from the ban on coordination of dissimilar constructions. In that

sense, at the first sight, (82) does not seem to be particularly interesting, since any theory

that postulates a formal difference between the objects in question can rule it out (it may

suffice to assume two different kinds of Accusative, for instance, provided that the

languages count them as dissimilar). We will see, however, that this cannot be the case.161

There is a contrast, however, with coordination that involves two objects that are

supposed to be marked:

161
In addition, it must be noticed that languages do admit coordination of DPs with different case-values:
They and me left.

272
(83) a. Visitaron a Juan y María

Visited A John and Mart

They visited John and Mary.

b. Visitaron a la profesora y la alumna

Visited A the professor and the student

They visited the professor and the student

Interestingly, these sentences have a particular meaning: they are only grammatical if the

event of visiting happened to both individuals at the same time. They do not have the

reading where the visiting occurs in at different times; so, in (83) there cannot be two

events of visiting, that is, the event cannot be quantized (see also the discussion in

chapter 2, section 2.1.3). This is not the case if we put A in both nominals:

(84) a. Visitaron a Juan y a María

Visited A John and A Mart

They visited John and Mary.

b. Visitaron a la profesora y a la alumna

Visited A the professor and A the student

They visited the professor and the student

273
The sentences in (84) are ambiguous: there can be one event of visiting or two. We can

represent the difference by saying that in (83) there is only one KP projection, with two

coordinated DPs inside, whereas in (84) the coordination is between two different KPs:

(85) a. [KP K [BP [DP ] and [DP ] ]]

b. [BP [KP ] and [KP ]]

Under the assumption that the BP can quantize the event, but it cannot do it if it is not the

maximal projection, we capture the above ambiguity. In (83), the BP is inside the KP (see

(85a); in (84), it is the maximal projection (see (85b). Note also that, if the Boolean

Phrase is headed by or, and not by and, it is not possible to have a single A-marker:

(86) a. *? Visitaron a Juan o María

Visited A John or María

They visited John and Mary.

b. *? Visitaron a la profesora o la alumna

Visited A the professor or the student

They visited the professor and the student

This is, of course, expected, given that with or it is normally not possible to have a

274
single-event interpretation.162

Now, if (85) is a possible scenario, then we could ask why (85a) cannot be a possible

structure for the object in (82), allowing it to bypass the ban on coordination of dissimilar

constituents:

(87) *? Mencionaron [KP a [BP [Juan] y [el libro] ] ]

Mentioned A John and the book

After all, it seems that animate nominals can be coordinated with inanimate ones (so

animacy by itself cannot be seen as making dissimilarities relevant for coordination):

(88) Juan y el libro son incompatibles

John and the book are incompatible.

This means that (87) is actually more interesting than we originally thought. Recall that

in the system I have developed in chapter 2, in order for a KP to be able to get an A, it

must have a [person] feature, since only this way small v (the probe), which has only

[number], will be incomplete with respect to the goal. However, [person] starts inside the

DP, and it must find its way up to K. In a structure like (87), the only path to K is through

the head of BP. It is natural to assume that the Boolean element requires that both

coordinated DPs have the [person] feature. Notice that this does not affect the

162
It is worth noticing that there is a marginal reading with a single-event interpretation that is possible for
(86), that is, the reading on which John and Mary, and the professor and the student, are the same person.
This reading, however marginal, confirms the intuitions about the unity of the nominals with a single A.

275
grammaticality of (88), since there we can assume that the coordinate element has a

structure like (85b), that is, with two different KPs (both in Nominative) and BP as its

maximal projection:

(89) [BP [KP Juan] y [KP el libro] ] son incompatibles

John and the book are incompatible.

In fact, (89) may be the only possibility for subjects, given the well known fact that

subjects always quantize the event, and the assumption that BP cannot quantize the event

if it is embedded in other structure.

This means that (87) is actually ruled out by the traditional ban on dissimilar

constructions, which, however, does not refer to case, but to feature specification in D.

These data, then, provide crucial evidence to distinguish the purely semantic [±animate]

feature from the φ-feature [person].163 A purely semantic ban on coordination of animate

nominals cannot account for both (87) and (88), but a ban on coordination of DPs with

different φ-features (i.e. with or without [person]) can.

Another issue regarding the interaction between DOM and coordination concerns

pseudogapping. If coordinated sentences do not both have the same kind of object (A-

marked or unmarked), pseudogapping is not possible, a matter that has never been

discussed in the DOM literature on Spanish:

163
This sets the basis to consider [person] an interface feature (in the sense of Svenonius 2006), that relates
structure with meaning.

276
(90) a. * Juan besó a María y Pedro la pared

John kissed A Mary and Peter the wall

b. * Juan besó la pared y Pedro a María

John kissed the wall and Peter A Mary

c. Juan besó a María y Pedro a Carolina

John kissed A Mary and Peter A Caroline

John kissed Mary and Peter did Caroline

d. Juan besó la fotografía y Pedro la pared

John kissed the picture and Peter the wall

John kissed the picture and Peter did the wall

Under the assumption that pseudogapping involves VP-ellipsis (Lasnik 1995c, 1999a),

and given that, in the system that I am developing here, both objects leave the VP, the

question is what prevents a derivation like the following, with VP ellipsis:

(91) * Juan besó [DATP a María [vP [VP ] ]] y Pedro [vP la pared [VP ] ]

John kissed A Mary and Peter the wall

Notice that appealing to a vP ellipsis may not help either, since in (90b) we would have a

277
vP as the antecedent:

(92) * Juan besó [vP la pared [VP ] ] y Pedro [DATP a María [vP [VP ] ]

John kissed the wall and Peter A Mary

We can consider the possibility that the VPs involved are not really equivalent. Under

Merchant 2001 theory of ellipsis, VP-ellipsis would be blocked if these VPs do not have

the same meaning (to be more accurate, if the entailments of the antecedent are not the

same as the entailments of the target). There seems to be no independent evidence for this

assessment, however. Therefore, it has to remain a stipulation that a VP that has been

extended with a Dative Phrase has some semantic property that makes it different from

other VP that has not been extended in this way.

Another possibility is to assume that the parallelism requirement does not hold only for

the deleted part, but also for what is not deleted (following Fox and Lasnik 2003, Park

2005, among others). This will also give us a difference, since, in the system I have

developed here, the A-marked object lands in [Spec, DatP], but the unmarked object

lands in [Spec, vP], as shown in (92), which breaks the parallelism (in other words, we

have a DatP only in one conjunct).

278
3.4 Remaining issues

There are a couple of issues that I want to address before closing this chapter. I must

stress that both of them will be briefly discussed below, in spite of their obvious

importance, because they involve external factors that would lead me too far if I were to

integrate them into the discussion. The issues in question involve DOM inside

nominalizations and the interplay between DOM and causative constructions.

It has been claimed that nominalizations do exhibit DOM (Torrego 1998: 38-41, Richards

2006). I contend that this is not correct. At the first sight, it seems quite clear that there is

no DOM in nominalizations: an object inside these nominals gets the usual de (of) case-

marker (which, in addition, is common for theme and agent) and not A:

(93) a. La acusación del profesor

The accusation of+the professor

b. El asesinato de Juan

The murder of John

c. La eliminación de los competidores

The elimination of the competitors

However, as observed by Torrego 1998: 38-39, for some nominalizations, when the agent

279
is realized with a possessive determiner, the case marker shifts to A:

(94) Su acusación al/*del profesor

His accusation A+the/of+the professor

But this is not the case for all nominalizations:

(95) a. Su asesinato *a/de Juan

His murder A/of John

b. Su eliminación *a/de los competidores

His elimination A/of the competitors

Torrego 1998: 171 fn 29 claims that the impossibility of shifting is restricted to non-

affected objects, the implication being that all affected objects inside nominalizations

must shift the marker to A if the agent is expressed by a possessive determiner.164 As

shown by (95), this is not really true. The possibility of using A seems to be idiosyncratic,

depending on the verb that is nominalized (but it is still true that non-affected objects do

not shift). Torrego also suggests that process nominals (in the sense of Grimshaw 1990)

trigger this shift. Notice, however, that the nominals in (95) are process nominals too.

164
There is also a question regarding the issue of defining affectedness. For instance, we could ask if to
accuse (the verb relevant in (94)) is an affecting predicate or not. Torrego includes “psychological change”
in her definition of affectedness (which presumably includes to accuse). Since to murder and to eliminate
(where Torrego’s generalization does not apply) are clearly affecting predicates, I will not discuss the issue
here. See also the discussion in section 2.2.2.

280
For these reasons, I reject the claim that we have DOM inside nominalizations. A

question remains, of course, with respect to the reasons of the attested shift to A (see

example (94)). I will leave this for future research, speculating now that some

nominalized heads can have an optional inherent Dative which surfaces when there is an

additional Genitive constituent; that is, it seems that some verbs are developing an

aversion to double Genitive (as a result of which a Genitive is replaced by Dative).

The final issue I want to address, also briefly, concerns causative constructions. This is a

vast topic, with a large body of research, which includes detailed empirical observations

and theoretical discussions that go well beyond DOM (for Spanish and other Romance

languages see Kayne 1975, Strotzer 1976, Zubizarreta 1985, 1987, Reed 1992, Treviño

1994, Guasti 1996, Cerbasi 1997, Wunderlich 1997, Torrego 1998, Jiménez Peña 2001,

Cépeda 2006, Folli and Harley 2007, among many others).

A crucial fact in Spanish causative constructions is that the causee always has an A-

marker if the infinitive is transitive:

(96) a. Hicieron pasar la noticia a la televisora

Made broadcast the new A the network

b. Hicieron vender la casa a Juan

Made sell the house A John

281
However, as the cliticization shows, it is not really clear that we have a DOM-object in

all these cases. In (96a) the causee is replaced by the IO-Dative clitic LE, whereas in

(96b), the form LO (which is never used for IOs) is possible:

(97) a. Le hicieron pasar la noticia

CL-DAT made broadcast the new

b. Lo hicieron vender la casa

CL-ACC Made sell the house

So, the actual case of the causee is a legitimate issue to examine. Is it Dative or

Accusative? This issue is particularly important in the current system, since I am

assuming that the case of the DOM object is actually Dative. So we would then expect

that in those cases where the A is mandatory, it must be Dative.

A clarification is in order with respect to the form of the clitic. It is well-known that

Spanish is split in (at least) two dialects with respect to the form of the masculine clitic

for Direct Objects (see Gutiérrez Ordóñez 1999 for an overview). Generally speaking, the

dative form LE can be used for some Direct Objects in one dialect (called the leísta

dialect), but the form LO must be used in the other dialect for the same objects.

Interestingly, LE is used in the leísta dialect with DOM objects, and it is not used for

unmarked objects. Although this provides support for the system we are using here, a

question remains about the dialects that only have LO.

282
Notice that, as it is standardly assumed, in Spanish the verb must raise to T, taking with it

all the heads on its path. Assuming that clitics are hosted by their corresponding

functional heads (Dat and small v), we can postulate that the difference between leísta

and non-leísta dialects resides in the way the complex v+Dat is interpreted

morphologically. In leísta dialects it is interpreted as LE; and in non-leísta dialects as LO.

In both dialects, however, a single Dat is always LE, and a single v is always LO. This is

illustrated below:

(98)
DATP

DAT’

DAT vP
LE

v’

LO v VP
LE LO

n
o
p

That is, we have the following correlations between the dialects:

283
(99)

vn DAT n
Non-leísta
LO LE
Leísta

(100)

v+DAT
Non-leísta LO o
Leísta LE p

In other words, LO is the morphological realization of v in both dialects, and LE is the

morphological expression of DAT in both dialects. But when we have a complex head

v+DAT, which is the case with DOM objects, the leísta dialect expresses this with LE,

whereas the non-leísta dialect expresses it with LO.165 The consequence of this state of

affairs is that, in the non-leísta dialect, LE can only be a reflex of DAT, that is, it indicates

that the nominal is a regular Dative and not a DOM-object. Notice that this implies that

clitics are the heads of small v and Dat, that is, probes in the process of structural case

checking (as in Franco 1993, Linares 2005, among others).

Therefore, for non-leísta dialects, the case of the causee in (96a) must be a regular Dative,

not a DOM-object, whereas the causee must be a DOM-object in (96b). As has been

noticed for a long time (see Strotzer 1976 for a detailed discussion), sentences like (96b)

also have a LE counterpart:

165
Notice that ditransitive constructions require both heads to be expressed separately, so option n prevails.

284
(101) Le hicieron vender la casa

CL-DAT made sell the house

In other words, for a significant number of verbs, the causee can be a regular Dative or a

DOM-object. This is not without consequences. The LO version is usually understood as

being more agentive than the LE version, that is, the higher subject is understood more

like an agent than simply a cause. However, this agentive reading is not absent in (101),

at least for non-leísta dialects.

Given the system outlined so far, it is unlikely that we can derive these differences purely

on the basis of Case/Agreement relations (for an account based on a different case

assignment mechanism, see Torrego 1998). For this reason, I will put causative

constructions aside. Further research on the semantic properties of these constructions

and their relation with clitics and clitic climbing will allow a better understanding of their

interaction with DOM. I leave that for the future.

285
 Chapter 4

A brief cross-linguistic exploration of DOM

A fundamental question that should be addressed when dealing with DOM is the

following: why do we have two different ways to mark the direct object in so many

languages? The number of languages that have different morphological markers for the

direct object is more than 300 (Bossong 1985, 1991); if we add the languages that have

different agreement markers corresponding to different types of objects, we have a

mechanism that is a good candidate for a universal. Moravcsik 1978 presents a range of

languages with differential marking and concludes that there exist a considerable cross-

linguistic convergence in the ways the marking mechanism works in these languages (see

also Bossong 1982, 1983-1984, 1985, 1988, 1991, 1997, Aissen 2003, de Swart 2003,

Lima 2003, 2006, among others, just to mention a few works that take a cross-linguistic

perspective).

The need to come up with an integrated solution to different DOM systems is urged by

the fact that, in the languages where the acquisition of DOM has been investigated—for

instance, Spanish (Rodríguez-Mondoñedo 2006b) and Turkish (Ketrez 2003)—children

seem to acquire the conditions to tease apart the two types of marking of DOM objects

very quickly and with no errors, despite the fact that the conditions that govern this

phenomenon are not simple, as we have seen in the previous chapters. It is important to

stress that, as shown in Rodríguez-Mondoñedo 2006b, this errorless performance of

children is not limited to commission, but it includes omission as well; that is, it is not
only the case that children do not use the marker when they should not, but they also do

not fail to use it when they should use it. This is relevant because it has been noticed that,

although children tend to have few errors of commission in general, they do make errors

of omission (Snyder in press). This clearly suggests the idea that some core mechanism

of the grammar is responsible for the distinction in question, i.e. the distinction is not a

peripheral or language-particular device.

It was Bossong 1982, 1985, 1991 who coined the term Differential Object Marking

(DOM) to refer to languages that have the split in question. I have adopted this term en

general, but I need to point out that, at least in its normal use, the term has some

limitations. For instance, it does not include the so-called head-marking languages (which

exhibit what we could call Differential Object Agreement), which are also sensitive to the

same kind of constraints.166

One important problem of both traditional and non traditional treatments of DOM

languages is that the property makes the languages in question somehow special with

respect to others languages. This is consistent with a narrow view of the phenomenon,

according to which it takes place in a subset of languages. However, no DOM-parameter

has been ever proposed, to my knowledge. I think the reason for this is two fold. First, all

the elements usually involved in the phenomenon (case marking, agreement, person

hierarchies, functional hierarchies, markedness) have to be universal, and in fact they are

allegedly present in all languages, but not all languages have (a narrow) DOM; a certain

166
In these languages, the agreement markers appear in the corresponding head depending on conditions
similar to DOM languages (animacy and specificity).

287
degree of arbitrariness in selecting the type of language with DOM then has to be

admitted. Second, under a narrow view of the phenomenon, there are very few

consequences of being a DOM or a no DOM language. So, at the first hand, DOM looks

like an issue were syntax has little to say; I think, however, that this is not correct. Using

Spanish in the previous chapters, I have tried to make a case that DOM is consequence of

a broader phenomenon, namely, the way Case and Agreement interact with each other. In

this chapter, I will explore the nature of DOM from a cross-linguistic perspective,

keeping an eye on the aspects of the system I have developed for Spanish. The goal of

this chapter is not to achieve a comprehensive solution for DOM, but to point out some

aspects in the properties of other DOM languages that suggest that the system outlined

for Spanish is not a language specific-artifact but it has some potential to be used cross-

linguistically. I leave for future research a full scale cross-linguistic exploration of DOM.

4.1 Is DOM universal?

Let’s assume, as a working hypothesis, that universally there are at least two different

positions that license the Direct Object, as we found to be the case in Spanish. Although

this idea has strong similarities with long standing intuitions in this respect, stemming

from Diesing 1992 seminal work, it aims to suggest something else: that the reason for

the different positions is actually syntactic, and not just semantic. If this is correct, it

means that all languages have an object split; what is different is the types of object that

are licensed in each position and the means that languages use to mark this licensing. For

many analyses (for instance, Aissen ’s 2003 OT system discussed in chapter 2), DOM is

288
limited to languages where one type of objects is morphologically marked and the others

are unmarked. This excludes, for instance, Finnish or Scottish Gaelic, where there are

two morphological object-markers, and a semantic difference between the objects

correlates with the markers—for Finnish see Holmberg and Nikanne 1993, Kiparsky

1998, 2001, 2005, Csirmaz 2005, among others; for Scottish Gaelic see Ramchand 1997.

It also excludes languages like German or Icelandic where all objects receive the same

morphological marker, but there is a semantic split between the objects that correlates

with their position in the sentence (object-shift)---see Holmberg 1986, Diesing and

Jelinek 1993, Holmberg and Platzack 1995, Bobaljik 1995, Thráinsson 2001, among

others. And of course, it excludes English, where all objects are morphologically

unmarked.

If we were to propose a system where all these possibilities are instances of the same

phenomenon, the first picture that emerges from these considerations is the following (A

and B indicate different morphological objects and Ø the lack of a morphological marker):

(1)

DO Position 1 DO Position 2
a. Finnish, Scottish Gaelic A B
b. German, Icelandic A A
c. Spanish, Turkish A Ø
d. English Ø Ø

In other words, there are two universal positions to license the Direct Object. Finnish and

Scottish Gaelic mark each of them with a different marker, German and Icelandic with

289
the same one, Spanish and Turkish only mark one position, and English does not mark

any of them.

In the system I am trying to push here, objects must hold some dependency with the head

that licenses their case (as discussed in chapters 1 and 2). In this scenario, the difference

in the cutting point for DOM-objects, the exact types of objects that are marked or

unmarked, depends on the lexical properties of the corresponding elements (the head and

the nominal object), and the conditions imposed on the dependency between those heads

(by Agree, in our system). Notice the above this would greatly improve the child’s ability

to acquire DOM, since it implies that actually all languages have it; the only thing that

the child must learn is how to mark these positions, that is, under which conditions s/he

should use A, B or Ø.

This picture, however attractive, faces some serious challenges, as we will see.

Nevertheless, I think that we can learn a lot about the nature of the phenomenon by

entertaining (1)—even if we ultimately have “to throw away the ladder after having

climbed up”, to borrow Wittgenstein’s metaphor. Pursuing that goal, this chapter takes (1)

as the point of departure.

Table (1) essentially predicts that the difference is arbitrary: there is nothing fundamental

about being (1a) or (1c), just the object markers, merely a morphological difference.

However, this is plainly false. There are significant differences between each of the

language-types predicted by (1). Languages like (1a) exhibit a high sensitivity to the

290
structure of the event in correlation with the markers (in particular to quantization, see

Ritter and Rosen 2001, 2005, among others). Languages like (1b) exhibit Object-Shift,

that is, a displacement of the object that correlates with informational structure (see

Diesing 1992, Meinunger 2000, López 2006, among many others). 167 (1c) is the

traditional DOM. English (a (1d)-type language) exhibits at least one form of sensibility

to conditions that are related to DOM: it is not possible to do wh-extraction from definite

and specific objects (Chomsky 1973, Fiengo and Higginbotham 1981). This may be an

independent effect, but, as Stepanov 2001b suggests, if specific nominals (but not non-

specific nominals) are assumed to move out of VP, the effect can be accounted for by the

ban on movement out of moved elements.

It is possible to suggest that all the phenomena in (1) may actually be present in Spanish,

in different degrees, which would cast some doubts on the idea that (1) represents a

unified phenomenon. This still leaves out a crucial question with respect to DOM

(understood as (1c)): Turkish, Spanish and other DOM languages do not use the marker

under the same conditions. Therefore, we have to provide at least a suggestion regarding

how these differences could be described in the system I have developed. I will do that in

the next section.

Before doing that, let me point out that (1) allows us to raise the issue of markedness,

which is commonly associated with DOM. I will address this question now, extending it

to all types in (1) (exploring the possibility of a unification).

167
Following Bošković and Takahashi 1998, Bošković 2006a, and others, I take “object shift” to cover also
what others call scrambling in German.

291
4.1.1 The question of markedness reversal

Position 1 in (1) is usually associated with a particular interpretation, which I will call

INT (which is variable across languages). INT is compatible (but not identical) in

Position 1 across languages. By compatible I mean that the objects with an INT in one

language may be a subset of the objects with the INT in another language.

For languages like the ones in (1c), this compatibility is usually deduced by assuming a

universal hierarchy inside the semantic categories that interact with DOM (as explained

in chapter 2, when we discussed Aissen’s 2003 system; see also Isaak 2000, Newmeyer

2002, Carnie and Jelinek 2003, Carnie 2004, Haspelmath 2004, Carnie 2005). These

hierarchies are taken from the typological literature (Silvernstein 1976, Comrie 1989,

Croft 1990, 2003). The functional literature is also relevant to this topic; in particular, the

prolific line of research stemming from the notion of degree of transitivity Hopper and

Thompson 1980)—for Spanish, see the papers in Clements and Yoon 2006. In this

respect, we have discussed the OT model for DOM, which aims to integrate the

functional-typological findings in a formal fashion (Aissen 1997, 1999, 2003, de Swart

2003, among others).

Now, under standard assumptions, Position 2 is associated with the complement of INT,

which I will call INT’. INT’ varies across languages in a way that mirrors the variation in

INT, that is, the different INT’s should also be compatible in Position 2 cross-

292
linguistically. Given this state of affairs, (1) is saying that an object in Position 2 cannot

be marked if an object in Position 1 is also not marked. This is the common wisdom

about DOM.

Under this view, the key mechanism behind DOM is the notion of markedness reversal,

according to which there is an imbalance between syntactic functions, such that there is

subset of nominals (with INT interpretation) that are more appropriate to be subjects, and

therefore less appropriate to be objects; DOM, under this approach, is a mechanism to

mark these “less appropriate objects”. 168 Obviously, this point of view will be

compromised if we can find situations that violate the markedness reversal. There are

indeed cases where some objects in Position 2 are marked, but some objects in Position 1

are not.

We have seen cases of this violation in Spanish (see the discussion in chapter 2). In

sentences like (2), both the subject and the object have the same degree of definiteness

and animacy, but the marker is not possible, which means that markedness reversal does

not apply:

(2) El huracán provocó (*a) la tormenta

The hurricane caused the storm

This is not unique to Spanish. Two cases in point involved the antipassive construction in

ergative languages and the ba construction in Chinese.

168
The properties of INT are normally associated with subjects.

293
In the antipassive construction (Kalmár 1979, Bittner 1987, Baker 1988, Bok-Bennema

1991, Johns 1992, Bittner 1994, Schmidt 2003, Wharram 2003, among others), the object

receives an INT’ interpretation, but, very frequently, it also receives a morphological case

marker (generally, oblique). In fact, the object with the INT interpretation must be

unmarked, as we can see in (3b), which includes an antipassive:

(3) a. Kaali-p nanoq toqup-paa West Greenlandic

Kaali-ERG bear kill-IND.SING

Kaali killed the polar bear.

b. Kaali nannu-mik toqut-si-voq

Kaali bear-INST kill-ANTIPASSIVE-IND.SING

Kaali killed a polar bear.

[Schmidt 2003: 390]

Even more interestingly, this is not the case for all languages with antipassive. For

instance, in Nez Perce (Rude 1985, Deal to appear, among others), the antipassive is

morphologically unmarked, and the object in the transitive construction is

morphologically marked:

294
(4) i. Iin-im ciq’áamqal hi-p-teetu núkt (ANTIPASSIVE)

1SG-GEN dog 3SUBJ-eat-HAB meat

‘My dog eats meat’ [roughly, ‘my dog does meat-eating’]

ii. Iin-im ciq’aamqal-nim pée-p-teetu nukú-ne (TRANSITIVE)

1SG-GEN dog-SUBJ 3/3-eat-HAB meat-OBJ

‘My dog eats meat’

[Deal to appear: 2]

An interesting question that arises here is why Nez Perce should have an antipassive

construction at all (as (4) shows, there is not even an antipassive morpheme). Deal (to

appear), following Rude’s 1985 initial proposals, convincingly shows that the semantic

conditions that apply to regular antipassive languages (Bittner 1994, Wharram 2003) also

apply to Nez Perce. However, one could still ask why Nez Perce is not like Turkish:

Öztürk 2005 also treats the morphologically unmarked objects in Turkish DOM as

“demoted” objects—more precisely, as undergoing pseudo-incorporation (in the sense of

Massam 2001)—and the semantic effects are very similar to the ones we find in the

antipassive. Of course, Turkish, in contrast to Nez Perce, is not an ergative language, and

the common wisdom is that antipassive is linked to ergativity. At the very least, the

antipassive construction still shows a phenomenon that does not comply with markedness

reversal.

295
The ba construction in Chinese, which is sensitive to both animacy and definiteness (see

Zou 1993, Li 2001, van Bergen 2006 and several others), also provides evidence against

the view of DOM that relies on markedness reversal. Adding more interest to the picture,

ba constructions seem to show an interplay between differential marking and object-shift

(but see the following sections). As is well known, word order in Chinese is usually strict;

however direct objects can (and some times must) undergo object-shift. In situ objects are

never marked, but most shifted objects must be marked by the pre-nominal element ba:

(5) a. Ta *(ba) yi-ge pingguo chi le

He BA one-CL apple eat PRT

‘He ate an apple’

b. Ta *(ba) wo da le

He BA I hit PRT

‘He hit me’

[van Bergen 2006: 7-9]

Sometimes, however, ba is optional:

(6) Ta (ba) na-ge pingguo chi le.

He BA that-CL apple eat PRT

‘He ate that apple.’

[van Bergen 2006: 7-9]

296
Notice that, from the point of view of a definiteness/specificity scale (a degree scale from

[+definite] to [-definite]), (6) is in the middle of (5a) and (5b). Too put it in terms of

markedness reversal, (5a) is a “more appropriate” object than (6); but still (5a) is

obligatorily marked and (6) is not. As van Bergen 2006 observes, following similar

concerns pointed out by de Swart 2003, this situation requires a departure from the

traditional conception of DOM. Indeed, de Swart 2003, also working in an OT model,

proposes a constraint of minimal semantic distinguishability, according to which the

arguments of a transitive clause must be minimally distinct, that is, the subject must be

higher than the object on the relevant semantic scale; when this constraint is violated, the

object is marked, giving rise to a DOM system. As we discussed in chapter 2 (when we

discussed sentences like (2), this way of reinterpreting DOM does not really work.

Næss 2004a also points out that, if we follow the logic of markedness reversal, some

well-known grammatical process that affect objects become unexpected. For instance,

incorporation of objects (Baker 1988 and many others); since this process usually

involves the “demotion” of “typical” objects (with INT’), Næss wonders why are

precisely the typical objects the ones that are target by this process. In other words,

incorporation deprives prototypical objects of “objecthood”, leaving non prototypical

ones as the only remaining objects. In that sense, the notion of “more appropriate object”

becomes meaningless, independently of DOM systems.

297
With respect to the marker, the picture that emerges from the previous considerations is

something along these lines:

(7)

DO Position 1 DO Position 2
a. Finnish, Scottish Gaelic A B
b. German, Icelandic A A
c. Spanish, Turkish A Ø
d. West Greenlandic Ø A
e. English Ø Ø

In other words, what is relevant is the fact that there are two positions available for the

object, not how the object itself is marked (or if it is marked at all).

4.1.2 Object Shift is not DOM

The general idea that I have developed for DOM is the following. The initial case-

licensing head (small v) is unable to check accusative case of certain nominals, because

of the way Agrees operates (more precisely, Agree fails to value the case feature of the

nominals in question). As a result, these nominals need to look for another way of

checking their case feature. In Spanish and, presumably in the vast number of languages

where the differential marker is the same as the dative marker (Bossong 1991), this head

is a dative head; the object has to raise to be in the checking domain of this other head.

Notice that this does not have to be the only way to repair the failure of Agree; in fact,

different mechanisms may be available, giving rise to the possibilities in (7). In this

298
scenario, it is tempting to propose that all the slots in (7) emerge from an interaction

between case and agreement; when there are two different ways to license different kinds

of objects, each way corresponding to one position. In particular, in line with the

mechanisms discussed in the previous chapters, we could imagine that the feature

structure of one type of object makes it impossible to license the object in the first

position it encounters (what I am calling here DO Position 2). Therefore, the object must

raise to get licensed upstairs, with an additional mechanism, which does not need to be

the same across languages.

A major set back for this picture is the behavior of Object-Shift languages, that is, the slot

in (7b). There is a strong correlation between informational structure and Object Shift. In

languages like Icelandic and German, shifted objects act like internal topics. That is, there

is a discourse-motivated reason to shift the object. It is reasonable to assume that a

[+topic] feature drives the movement (as, for instance, in Meinunger 2000 for German),

or, alternatively, that there is a topic-like interpretation assigned to the internal periphery

of the clause. Taking this second position, López 2006 argues that DOM-objects do not

need to have a topic interpretation, but (as also observed by Diesing 1992) OS-objects

must be interpreted as topics—see also the discussion in chapter 2. Notice also that it is

not enough for Icelandic OS-objects to be definite or specific; if an object is focused, it

cannot shift:

(8) Hvað las Jón ekki?

what read Jón not

299
 a. Hann las ekki bækurnar

he read not books-the

b. # Hann las bækurnar ekki.

he read books.the not

[López 2006: 161]

An OS-object is only felicitous if it is discourse anaphoric:

(9) [Jón went into town a month ago and bought things for 20000 kroner: books,

clothes, CDs. When he came home, he listened to all of the CDs that he bought

but …]

… hann las bækurnar ekki fyrr en miklu seinna. [Ice]

… he read books.the not before than much later.

[López 2006: 161]

These conditions do not apply to DOM-objects, which can be focused or not, and they do

not need to be discourse anaphoric.169 In that sense, as anticipated in chapter 2, Object-

Shift and internal topicality in general, must be considered independent of DOM.

169
I am putting aside here the alleged optionality of Object-Shift (at least for full DP/NP, Holmberg and
Platzack 1995: 162-165). Bobaljik 1995: 346-351 argues, however, that Object-Shift always happens, but
sometimes the lower copy needs to be pronounced to avoid a disruption of adjacency between the verb and
an inflexional morpheme, in the context of Holmberg’s Generalization—see also Bobaljik 2002 and
Bošković and Nunes 2007 (Diesing 1996 also argues that Object-Shift in Icelandic is actually mandatory).

300
4.1.3 The Specificity Condition does not drive DOM

As noted above, English objects are subject to a constraint that resembles DOM

phenomena: in general, it is not possible to extract from definite or specific objects. This

constraint is sometimes called the Specificity Condition (Chomsky 1973, Fiengo and

Higginbotham 1981, Mahajan 1992 among others), which simply states that specific DPs

cannot hold traces inside:170

(10) a. * What did you buy the books about t?

b. What did you buy a book about t ?

In Spanish, the Specificity Condition holds—subject to further constraints, see Ticio

2003, 2005 for discussion and references:

(11) *¿De qué autor has leído [los libros t ]?

of which author have-you read the books

[Ticio 2005: 238]

However, the specificity effect is independent of DOM. (11) presents a case where an

unmarked object cannot hold a trace inside. DOM objects, under the same conditions, do

not allow extraction either:

170
See Yoshida 2003 and others for some limitations on this characterization of the Specificity Condition.

301
(12) *¿De qué universidad has contrado [a los ingenieros t ] ?

of which university have-you hired A the engineers

In addition, according to Mahajan 1992, in Hindi the Specificity Condition does not hold

(that is, wh-elements can be extracted from specific objects). However, as is well known,

Hindi is a DOM language. This confirms that the Specificity Condition is not related to

DOM, and therefore it cannot be used to characterize (for instance) English as a DOM

language.171

We have seen in this section that, even when there are several phenomena that show

some resemblance with the conditions associated with DOM, they cannot be

mechanically incorporated into the DOM system. This does not necessarily mean that it

may not be possible to come up with a mega-system from which all of the phenomena

discussed above would emerge, but it does mean that DOM has its own properties, that is,

that the options in (7) are not merely superficial differences behind a single phenomenon.

4.2 The DP-vP connection: Kannada

If we consider that DOM emerges from an interaction between case and agreement, and

(perhaps more generally) the licensing of nominals, it should not be surprising that there

is a wealth of interfering factors.172 These mechanisms interact with other components of

171
It is worth noticing, however, that Stepanov 2001b argues that specific and non-specific objects in
English are in different positions; if this is correct, it would mean that this difference belongs to (1)
172
Notice further that this opens the possibility that Position 1 and Position 2 could be collapsed in some
languages, under the assumption that both kinds of objects are licensed with different mechanisms, but

302
the grammar. To see an example of how far these interactions go, consider the case of

Kannada, a language where DOM-marking is sensitive to animacy and specificity, but

where all plural objects must be marked (regardless of their animacy or specificity):173

(13) a. * Pratiyobba vidyaarthi eraDu pustaka-gaLu huDuk-utt-idd-aane

every student two book-PLU look.for-NPST-be-3S.M

Every student is looking for two books.

b. Pratiyobba vidyaarthi eraDu pustaka-gaL-annu huDuk-utt-idd-aane.

every student two book-PLU-ACC look.for-NPST-be-3S.M

Every student is looking for two books.’

[Lidz 2006: 25]

This state of affairs asks for an integrated solution, which should include the relation

between the structural case position of the nominal (in the clause) and the position of the

case feature and the plural feature inside the nominal (NP/DP). I claim that we can

capture the apparent anomaly illustrated (13) by using the idea that different feature

structures in the nominal have effects on the case/agreement system, also the gist of my

analysis for Spanish in chapter 2.

without displacement (this means that no EPP feature on the target or a relevant uninterpretable feature on
the above elements would be invoked in Chomsky’s/Bošković’s systems respectively).
173
Interestingly, this situation is not a peculiarity of Kannada, but is present in other DOM languages with
overt case marking too—for instance, Hup (Epps 2005). It is sometimes related to the well known
phenomena associated with split number marking, which is also sensitive to animacy and definiteness—see
Corbett 2000: 54-132 for an overview.

303
The facts in Kannada are a little bit more complicated than the description above

indicates. According to Lidz 2006, the following table captures the relevant

generalizations regarding Kannada DOM:

(14)

ANIMATE INANIMATE

MARKED de dicto / de re n de re p

UNMARKED *o de dicto / de re q
[Lidz 2006: 12]

Remember that de dicto readings correspond to non-specific interpretations, whereas de

re readings correspond to specific interpretation. This means that box n is saying that all

animate objects are marked, regardless of specificity; also, as boxes p-q show,

inanimate objects can be all unmarked, but they can optionally have a case-marker if they

are specific. As noted above, we must add to (14) the information that all plural objects

are marked, irrespective of their animacy or specificity. Notice that (14) requires that all

animates must be marked too, regardless of their specificity.

Recall from chapter 2 that a [person] feature must be present if a nominal is animate. In

turn, the [person] feature will get the object morphologically marked (with the marker A,

in Spanish), under the Agree system developed here. We can extend this analysis to

Kannada straightforwardly capturing n and o.

Interestingly, in Kannada the DOM marker (-vannu) is not Dative but Accusative (the

304
Dative marker is -age). Let’s assume that, in Kannada, different from Spanish, small v is

complete, that is, it has both [person] and [number]:

(15) v
[person]
[number]

At first sight, this seems to be odd, since this will mean that a [person] object will have

no problems checking its case against small v. However, I claim that this is exactly what

happens: Kannada [person] nominals are able to check case against small v and they get

Accusative.

However, there are two additional properties of Kannada nominals that we have to

consider. First, Kannada does not have overt article; following Bošković 2006a, let me

assume that this means that it does not have DP (so the D/*D distinction that was used to

capture specificity distinctions in Spanish has to be expressed in another way)—see

Chierchia 1998 for interpretation of nominals that do not have DP). Secondly, Lidz 2006

suggests that Kannada only projects a [number] phrase with plural nominals.174 Let me

interpret this fact by assuming that only plural nominals have a [number] feature. This

means that the relevant feature specifications of Kannada nominals (represented here by

G) are as follows:

174
This may be related to the previous property—Lidz does not go that far, though (he actually assumes a
DP projection).

305
(16)

Singular Plural

Animate G G
[person] [person]
[number]

Inanimate G G
[number]

Note that inanimate singulars do not have either [person] or [number]. Now, recall from

chapter 1 that the Condition on Case-valuation requires that the Probe match all the

features of the Goal:

(17) Condition on Case-valuation

Only probes P that match all the relevant features of a goal G can value the [case]

feature of G

However, the condition does not require the Goal to match all the features of the Probe.

This means that if small v is complete, it can value the case-feature of all the nominals in

(16), except inanimate singulars, since inanimate singulars do not have any φ-feature to

match. This straightforwardly accounts for the fact that all plural objects receive the

Accusative marker. It also accounts for boxes n, o and q in (14). n gets marked

because animates objects have [person]; therefore, there cannot be any unmarked animate

(as *o expresses). On the other hand, q cannot get marked because singular inanimate

nominals are φ-featureless; so small v does not match anything in the Goal G (which in

306
turn cannot act as a probe to check its case). The only remaining set of data is p (for

singulars).

Here we can adopt a solution already suggested by Lidz 2006, adapting it to our

framework. Lidz suggest that some Kannada nominals can be “inherently specific”,

which he understands as the need to receive a choice function. I have already used a

similar device for Spanish in chapter 2—although it had an opposite sign: *D can be

understood as a sort of “inherently non-specific”. Recall also that Kannada does not have

overt Determiners, so we cannot use a D/*D distinction as we did for Spanish. Let’s

assume that the locus of the distinction is K, that is, we have a K/*K distinction in

Kannada, which merely translates Lidz’s solution to our framework:

(18)

K de dicto / de re

*K de re (i.e. it must receive a choice function)

The only thing that we need now is the following set of morphological exponents:

(19) a. K : -vannu
[φ-F: α]
[K: ACC]

b. K : Ø
[K: ACC]

c. *K : -vannu
[K: ACC]

307
This means that if a Kannada object has φ-features, any type of φ-features, [person] or

[number], it will get -vannu, under the mechanism just explained (19a). But if an object

does not have φ-features, there are two options: if it must receive a choice function, then

it will get -vannu (19c); if it does not need to receive a choice function, it will be

unmarked (19b).

The remaining issue is how the nominal in (19b) and (19c) gets the case value. It cannot

get the value from a small v represented in (15), for the reasons discussed above (there

are no φ-features to match). There are two options here. One is that there is an additional

projection that also checks ACC, and the relevant objects, that is, p and q (for singulars)

in (14) move there under the same conditions explained for Spanish (the combination of

Blind Movement and Checking Movement in chapter 2). This second projection would

have to be a kind of small v without φ-features. The other possibility is that (19b), which

corresponds to q, is actually default case (that is, no ACC), whereas (19c), which

corresponds to p, is inherent accusative. I do not have sufficient information on Kannada

to tease apart these options at this point. I leave the issue open.

I would like to stress that I have accounted for Kannada DOM (to the extent that the

generalization in (14) is correct) without adding any substantial pieces to the system

developed for Spanish. The only relevant differences are lexical. This can be interpreted

as an indication that the DOM system being developed here, which is based on the

connection between the feature specification of DPs and the conditions on Agree, is on

the right track. Additional Kannada data with respect to other correlations still needs to

308
be checked, but the initial picture seems to give us a promising result.

It should also be added that the Kannada generalization in (14) provides evidence against

an OT approach to DOM, since it directly contradicts the entailments based on constraint

reranking (see chapter 2). That is, an OT-based model cannot account for Kannada

without significant modifications, while the current approach is able to account for both

Spanish and Kannada with essentially the same system.

4.3 Systematic DOM differences

A significant merit of Aissen’s 2003 OT system (and related work) is that it offers a

principled way to describe fine-grained differences that are attested across different DOM

languages. In fact, as pointed out by Aissen, this achievement provides a possible

standpoint to criticize Torrego’s 1998 account of Spanish DOM: there is little space in

her system to describe cross-linguistic differences. Recall however that the force behind

the OT system for DOM does not actually belong to OT itself: it is borrowed from the

typological and functional literature (the Definiteness and Animacy hierarchies), as

discussed in chapter 2. Torrego does not try to run her system against the hierarchies, but

it is not impossible that such an attempt could gain some cross-linguistic confirmation for

her system (even if further assumptions are needed). It is not our place to check this.

However I do have the obligation to show that the system I have developed here has

some potential to at least describe DOM systems cross-linguistically. The fact that we can

309
account for Kannada already gives us some advantage.

Although I have presented some critical considerations that are problematic not only for

the OT system, but also for the hierarchies on which it is based (recall the discussion of

(7)), let me try to check how the mechanism I have developed here fares with respect to

the hierarchies. This attempt should not be considered an endorsement of the hierarchies;

much more discussion and empirical exploration needs to be done to actually do that (or

the opposite). This is just an attempt to see how the differences that the hierarchies

predict can be accommodated in the system I have developed. In that sense, the following

discussion should be considered merely descriptive.

The central idea is that the interpretation of the feature [person] varies according to the

hierarchies. Taking as a point of departure the set of languages invoked by Aissen, and

the parameters used by her, DOM languages belong to three different groups:

(20) a. Languages in the Definiteness Scale

b. Languages in the Animacy Scale

c. Two-dimensional languages

Running the interpretation of the feature [person] against the Definiteness Scale, we

obtain the following values for the feature [person] for each of the languages.175

175
The characterization of the particular languages that are mentioned in (21) and (22) is based on the
information provided by Aissen 2003.

310
(21)

Catalan Pitjantjatjara Hebrew Turkish Written

Japanese

Pronoun [person]
[person]
Name [person]
[person]
Definite [person]

Indef Specific

NonSpecific

Running the interpretation of the feature [person] against the Animacy Scale obtains the

following correlation:

(22)

Yiddish Dhargari Dhalandji

Human [person]
[person]
Animate [person]

Inanimate

Under the assumption that in the corresponding languages there is also a small v that has

only [number] and not [person], the correlations in (21) and (22) predict the type of

object that will get marked for each language. For two-dimensional languages it will be

necessary to combine the values in (21) and (22).176

176
Interestingly, this will predict more possible languages than are actually attested, which is also a
problem in Aissen’s system (see Rodríguez-Mondoñedo 2006b for further discussion on this issue).

311
As witnessed by the situation faced in Kannada and Spanish, this matching is far from

being perfect. In fact, (21) and (22) inherit the empirical problems from Aissen 2003,

which have been discussed in several places through this dissertation. A more

comprehensive discussion of particular languages is needed to achieve a better system. It

may even be the case than the (21)-(22) correlations are entirely wrong, that is, that the

cuts they predict do not actually fit the differences between marked and unmarked objects

that we find in these languages. Interestingly, if this were the case, Aissen’s system

would likely collapse, but not the system that we have developed here, as also witnessed

by our account of Spanish and Kannada, languages that do not fit (21)-(22)—nor the

combination of them, as discussed above.

Let’s provide an additional piece of evidence in favor of the system I have developed.

Turkish is usually assumed to be a well-behaved DOM language—by which I mean a

language that does conform to (21) or the corresponding generalizations in different

frameworks. However, this language also exhibits a type of “exceptional” data that we

find in Spanish.

Recall that when we discussed the quantifiers nadie (nobody) and alguien (somebody) in

chapter 2, I noted that they constitute a prima facie evidence against the idea that A is a

marker of specificity. Although animate, nadie is non-specific, but it still receives a

mandatory A marker:

312
(23) a. Vi *(a) alguien en el parque

saw A somebody in the park

I saw somebody in the park

b. No vi *(a) nadie en el parque

No saw A nobody in the park

I saw nobody in the park

As noticed by Brugé and Brugger 1996, the same holds for Turkish. The quantifiers

kimse (nobody) and birisi (somebody) must have the accusative marker that characterizes

[+specific] objects:177

(24) a. Kimse-yı görmedim

nobody-ACC saw

I saw nobody

b. Ali birisi-nı gördü

Ali somebody-ACC saw

Ali saw somebody

(25) a. * Kimse görmedim

nobody saw

I saw nobody

177
They provide no account for these facts (or the corresponding Spanish data). Persian, also a DOM
language, exhibits similar behavior in this respect.

313
 b. * Ali birisi gördü

Ali somebody saw

Ali saw somebody

[Brugé and Brugger 1996: 47]

This suggests that the Spanish “exception” cannot be treated like a language-specific

quirk, but as a regular consequence of the DOM system. It also provides some reason to

believe that the system we used to explain the Spanish data is not really an ad hoc

solution.

As I said in the beginning, much more discussion and cross-linguistic comparison is

needed to actually claim victory over the DOM puzzle. In this dissertation DOM is a

result of an interaction between case and agreement, in particular, it arises because the

initial case-checking head, small v, is unable to check accusative case of certain nominals

because of the way Agree operates. I believe this idea points directions that may allow us

some new ways to look at DOM.

314
References

Abbott, Barbara. 1992. Definiteness, Existentials, and the ‘List’ Interpretation. In SALT II,
1-16.
Adger, David, and Harbour, Daniel. 2003. The Syntax and Syncretisms of the Person Case
Constraint. Ms., (To appear in MIT Working Papers in Linguistics).
Agüero-Batista, Calixto. 2005. Noun Phrase Interpretation and Differential Object
Marking. Paper presented at Sinn und Bedeutung 10.
Ahern, Aoife, and Leonetti, Manuel. 2004. The Spanish subjunctive: Procedural semantics
and pragmatic inference. In Current Trends in the Pragmatics of Spanish, eds.
Rosina Márquez-Reiter and María Elena Placencia, 35-56. Amsterdam: John
Benjamins.
Aissen, Judith. 1997. On the Syntax of Obviation. Language 73:705-750.
Aissen, Judith. 1999. Markedness and Subject Choice in Optimality Theory. Natural
Language & Linguistic Theory 17.
Aissen, Judith. 2003. Differential object marking: Iconicity vs. economy. Natural
Language & Linguistic Theory 213:435-448.
Alexiadou, Artemis, and Anagnostopoulou, Elena. 1998. Parametrizing AGR: word order,
V-movement and EPP checking. Natural Language & Linguistic Theory 16:491-
539.
Alexiadou, Artemis, and Anagnostopoulou, Elena. 2004. From hierarchies to features:
person splits and direct-inverse alternations. Ms., University of Stuttgart and
University of Crete.
Anderson, Mona. 1979. Noun Phrase Structure, University of Connecticut: Ph.D.
Dissertation.
Anderson, Mona. 2006. Affectedness. In The Blackwell Companion to Syntax, eds. Martin
Everaert and Henk Van Riemsdijk. Malden, MA: Blackwell
Andrews, Avery. 1983. A note on the constituent structure of modifiers. Linguistic Inquiry
14:695–697.
Anttila, Arto, and Fong, Vivienne. 2000. The partitive constraint in Optimality Theory.
Journal of Semantics 17:281-314.
Ariza Viguera, Manuel. 1989. La preposición A de objeto: Teorías y panorama. Lexis
13:203-222.
Baker, C. Leroy. 1966. Definiteness and Indefiniteness in English, University of Illinois:
Master’s Thesis.
Baker, Mark. 1988. Incorporation: a theory of grammatical function changing. Chicago:
University of Chicago Press.
Barbosa, Pilar. 1995. Null Subjects, MIT: Ph.D. Dissertation.
Barraza Carbajal, Georgina. 2003. Estructura y evolución de la frase nominal de objeto
directo inanimado en español. Medievalia 35:155-182.
Barwise, Jon, and Cooper, Robin. 1981. Generalized quantifiers and natural language.
Linguistics and Philosophy 4:159-219.
Béjar, Susana. 2003. Phi-Syntax: A theory of Agreement, University of Toronto: PhD
Dissertation.
Béjar, Susana, and Rezac, Milan. 2003. Person Licensing and the Derivation of PCC
Effects. In Romance Linguistics. Theory and Acquisition. Selected papers from the

315
 32nd Linguistic Symposium on Romance Languages (LSRL), Toronto, April 2002,
eds. Teresa Pérez-Leroux and Yves Roberge, 49-62. Amsterdam: John Benjamins.
Belletti, Adriana. 1988. The Case of unaccusatives. Linguistic Inquiry 19:1-34.
Belletti, Adriana. 2001. Inversion as focalization. In Subject Inversion in Romance and the
Theory of Universal Grammar, eds. Aafke Hulk and Jean-Yves Pollock, 60-90.
Oxford: Oxford University Press.
Belletti, Adriana. 2004. Aspects of the low IP area. In The Structure of IP and CP. The
Cartography of Syntactic Structures, ed. Luigi Rizzi, 16-51. New York: Oxford
University Press.
Belletti, Adriana. 2005. Extended doubling and the VP periphery. Probus 17:1-35.
Bello, Andrés. 1847. Gramática de la lengua castellana destinada al uso de los
americanos. Anotada por Rufino José Cuervo, ed. crítica de Ramón Trujillo.
Madrid: Arcolibros, 1988.
Bentivoglio, Paola. 1989. Haber. ¿Un verbo impersonal? Paper presented at 45 Congreso
Internacional de Hispanistas. Estudios sobre el español de América y lingüística
afroamericana, Bogotá: Instituto Caro y Cuervo.
Benveniste, Émile. 1966. Problèmes de linguistique générale. Paris: Gallimard.
Bhatt, Rajesh. 2005. Long-Distance Agreement in Hindi-Urdu. Natural Language and
Linguistic Theory 23:757-807.
Bittner, Maria. 1987. On the semantics of the Greenlandic antipassive and related
constructions. International Journal of American Linguistics 53:194-231.
Bittner, Maria. 1994. Case, scope, and binding: Studies in natural language and linguistic
theory v. 30. Dordrecht ; Boston: Kluwer Academic Publishers.
Bittner, Maria, and Hale, Ken. 1996. The Structural Determination of Case and Agreement.
Linguistic Inquiry 27:1–68.
Bleam, Tonia M. 1999. Leísta Spanish and the Syntax of Clitic Doubling, University of
Delaware: Ph.D. Dissertation.
Bleam, Tonia M. 2003. Properties of the Double Object Construction in Spanish. In A
Romance Perspective on Language Knowledge and Use, eds. Rafael Núñez-
Cedeño, Luis López and Richard Cameron, 233-252. Amsterdam: John Benjamins.
Bleam, Tonia M. 2005. The Role of Semantic Type in Differential Object Marking.
Belgian Journal of Linguistics 19:3-27.
Bobaljik, Jonathan David. 1995. Morphosyntax: The syntax of verbal inflection, MIT:
Ph.D. Dissertation.
Bobaljik, Jonathan David. 2002. A-Chains at the PF-Interface: Copies And 'Covert'
Movement. Natural Language & Linguistic Theory 20:197–267.
Bobaljik, Jonathan David. 2006. Where's Phi? Agreement as a Post-Syntactic Operation.
Ms., University of Connecticut. Storrs.
Boeckx, Cedric. 2000. EPP eliminated. Ms. University of Connecticut.
Boeckx, Cedric. 2001. Scope Reconstruction And A-Movement Natural Language &
Linguistic Theory 19:503-548.
Boeckx, Cedric. 2003. Intricacies of Icelandic Agreement. Ms.
Bok-Bennema, Reineke. 1991. Case and agreement in Inuit. Berlin; New York: Foris
Publications.
Bonomi, Andrea. 1995. Transparency and specificity in intensional contexts. In On Quine,
eds. P. Leonardi and M. Santabrogio, 164-185. Cambridge: Cambridge University

316
 Press.
Bošković, Željko. 1994. D-structure, Theta Criterion and movement into theta positions.
Linguistic Analysis 24:247-286.
Bošković, Željko. 1995. Case Properties of Clauses and the Greed Principle. Studia
Linguistica 49:32-53.
Bošković, Željko. 1997a. On Certain Violations of the Superiority Condition, Agro, and
Economy of Derivation. Journal of Linguistics 33:227-254.
Bošković, Željko. 1997b. The syntax of nonfinite complementation: an economy approach:
Linguistic inquiry monographs, 32. Cambridge, Mass: MIT Press.
Bošković, Željko, and Takahashi, Daiko. 1998. Scrambling and Last Resort. Linguistic
Inquiry 29:347-366.
Bošković, Željko. 1999. On multiple feature checking: multiple Wh-fronting and multiple
head-movement. In Working Minimalism, eds. Samuel David Epstein and Nolbert
Hornstein, 159-187. Cambridge, MA: MIT.
Bošković, Željko. 2001. On the nature of the syntax-phonology interface: Cliticization
and related phenomena. Amsterdam: Elsevier Science.
Bošković, Željko. 2002a. A-Movement and the EPP. Syntax 5:167-218.
Bošković, Željko. 2002b. Clitics as non-branching elements and the Linear
Correspondence Axiom. Linguistic Inquiry 33:329-340.
Bošković, Željko. 2002c. On multiple wh-fronting. Linguistic Inquiry 33:351-383.
Bošković, Željko. 2005. On the Locality of Move and Agree: Eliminating the Activation
Condition, Generalized EPP, Inverse Case Filter, and Phase-Impenetrability
Condition, eds. Simona Herdan and Miguel Rodríguez-Mondoñedo: University of
Connecticut Occasional Papers in Linguistics, 3.
Bošković, Željko. 2006a. What will you have, DP or NP? Paper presented at NELS 37.
Bošković, Željko. 2006b. Case and agreement with genitive of quantification in Russian.
In Agreement Systems, ed. Cedric Boeckx, 99–120. Amsterdam: John Benjamins.
Bošković, Željko. 2007a. On Successive Cyclic Movement and the Freezing Effect of
Feature Checking. In Sounds of Silence. Empty Elements in Syntax and Phonology
eds. Jutta M. Hartmann, Veronica Hegedus and Henk van Riemsdijk. Amsterdam:
Elsevier.
Bošković, Željko. 2007b. On the locality and motivation of Move and Agree: An even
more minimal theory. Linguistic Inquiry 38.4.
Bošković, Željko, and Nunes, Jairo. 2007. The Copy Theory of Movement: A view from
PF. Ms. University of Connecticut/Universidade de São Paulo.
Bosque, Ignacio. 2001. Adjective Positions and the Interpretation of Indefinites. In
Current Issues in Spanish Syntax and Semantics, eds. Javier Gutiérrez-Rexach and
Luis Silva-Villar, 17-37. Berlin & New York: Mouton de Gruyter.
Bossong, Georg. 1982. Der prapositionale Akkusative im Sardischen. In Festschrift für
Johannes Hubschmid zum 65. Geburtstag, 579-599. Bern: Francke.
Bossong, Georg. 1983-1984. Animacy and Markedness in Universal Grammar.
Glossologia 2-3:7-20.
Bossong, Georg. 1985. Empirische Universalienforschung. Differentielle
Objektmarkierung in der neuiranischen Sprachen. Tübingen: Narr.
Bossong, Georg. 1988. Differential Object Marking in Romance and Beyond. Paper
presented at XVII Linguistic Symposium on Romance Languages. New analysis in

317
 Romance linguistics., Urbana Champaign.
Bossong, Georg. 1991. Differential object marking in Romance and beyond. In New
Analyses in Romance Linguistics, Selected Papers from the XVIII Linguistic
Symposium on Romance Languages 1988, eds. D. Wanner and D. Kibbee, 143-170.
Amsterdam: Benjamins.
Bossong, Georg. 1997. Le Marquage Différentiel de L'Objet dans les Langues d'Europe.
In Actance et Valence dans les Langues d'Europe, ed. J. Feuillet, 193-258. Berlin:
Mouton de Gruyte.
Bowers, John. 1993. The Syntax of Predication. Linguistic Inquiry 24:591-656.
Branchadell, Albert. 1992. A Study of Lexical and Non-lexical datives, Universitat
Autónoma de Barcelona: Doctoral Dissertation.
Brugé, Laura, and Brugger, Gerhard. 1996. On the Accusative a in Spanish. Probus 8:1-51.
Brugè, Laura. 2000. La realizzazione morfologica del caso accusativo. In Categorie
funzionali del nome nelle lingue romanze, 93-304. Milano: Cisalpino.
Burzio, Luigi. 1986. Italian Syntax. Dordrecht: Reidel.
Burzio, Luigi. 2000. Anatomy of a Generalization. In Arguments and Case: Explaining
Burzio's Generalization, ed. Eric Reuland, 195-241. Amsterdam: John Benjamins.
Butt, Miriam. 1995. The structure of complex predicates in Urdu. Stanford: CSLI
Camacho, José. 1999. La coordinación. In Gramática descriptiva de la lengua española,
eds. Ignacio Bosque and Violeta Demonte, 2633-2694. Madrid: Espasa-Calpe.
Campos, Héctor. 1999. Transitividad e Intransitividad. In Gramática Descriptiva de la
Lengua Española, eds. Ignacio Bosque and Violeta Demonte, 1519-1574. Madrid:
Espasa Calpe.
Cardinaletti, A., and Guasti, T. 1995. Small Clauses: Syntax and Semantics 28. New York:
Academic Press.
Carlson, Greg. 1977a. Reference to kinds in English, University of Massachussetts: Ph.D.
Dissertation.
Carlson, Greg. 1977b. Amount Relatives. Language 58:520-542.
Carnie, Andrew, and Jelinek, Eloise. 2003. Argument hierarchies and the mapping
principle. In Formal Approaches to Function in Grammar. In honor of Eloise
Jelinek, eds. Andrew Carnie, Heidi Harley and MaryAnn Willie, 265–296.
Carnie, Andrew. 2004. Some Remarks on Markedness Hierarchies: A Reply to Aissen
1999 and 2003. Ms., University of Arizona.
Carnie, Andrew. 2005. A Phase-geometric approach to multiple marking systems. In
Perspectives on Phases, eds. Martha McGinnis and Norvin Richards, 49:87-102.
Cambridge, Ms.: MIT Working Papers in Linguistics.
Cecchetto, Carlo. 2000. Doubling structures and reconstruction. Probus 12:93–126.
Cépeda, Paola. 2006. Causativos léxicos y sintácticos en castellano: una aproximación
minimista, Pontificia Universidad Católica del Perú: M.A. Thesis.
Cerbasi, Donato. 1997. Las construcciones causativas del tipo hacer + infinitivo en
español, portugués e italiano. Lingüística Española Actual XIX:155-171.
Chierchia, Gennaro. 1998. Reference to kinds across languages. Natural Language
Semantics 6:339-405.
Chomsky, Noam. 1957. Syntactic structures. The Hague: Mouton.
Chomsky, Noam. 1973. Conditions on transformation. In A Festschrift for Morris Halle,
eds. Anderson and Paul Kiparsky, 232-286. New York: Holt, Rinehart and

318
 Winston.
Chomsky, Noam, and Lasnik, Howard. 1993. The theory of principles and parameters. In
Syntax: An international handbook of contemporary research, eds. Joachim Jacobs,
Arnim von Stechow, Wolfgang Sternefeld and Theo Vennemann, 506-569. Berlin:
Walter de Gruyter.
Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: The MIT Press.
Chomsky, Noam. 2000. Minimalist inquiries: the framework. In Step by step: essays on
minimalist syntax in honor of Howard Lasnik, eds. Roger Martin, David Michaels
and Juan Uriagereka, 89-155. Cambridge, Mass: MIT Press.
Chomsky, Noam. 2001. Derivation by Phase. In Ken Hale: A Life in Language, ed.
Michael Kenstowicz, 1- 52. Cambridge, Mass.: MIT Press.
Chomsky, Noam. 2004. Beyond Explanatory Adequacy. In Structures and Beyond. The
Cartography of Syntactic Structures, ed. Adriana Belletti, 104 – 131. Oxford:
Oxford University Press.
Chomsky, Noam. 2005a. Three Factors in Language Design. Linguistic Inquiry 36:1–22.
Chomsky, Noam. 2005b. On phases. Ms. MIT.
Chomsky, Noam. 2006. Approaching UG from below. Ms., MIT. Cambridge, MA.
Chung, Sandra, and Ladusaw, William. 2004. Restriction and Saturation. Cambridge:
MIT Press.
Cinque, Guglielmo. 1990. Types of A'-dependencies: Linguistic Inquiry monographs 17.
Cambridge, Mass.: MIT Press.
Clements, J.Clancy, and Yoon, Jiyoung. 2006. Functional Approaches to Spanish Syntax.
Lexical Semantics, Discourse and Transitivity. New York: Palgrave.
Coene, Martine, and D’hulst, Yves eds. 2003. From NP to DP. Volumen 2: The
Expression of Possession in Noun Phrases. Amsterdam: John Benjamins.
Company Company, Concepción. 2002. El avance diacrónico de la marcación prepositiva
en objetos inanimados. In Actas del II Congreso de la SEL. Presente y Futuro de la
Lingüística en España, 146-154. Madrid: Consejo Superior de Investigaciones
Científicas.
Comrie, Bernard. 1989. Language universals and linguistic typology: syntax and
morphology. Oxford: Blackwell.
Contreras, Heles. 1976. A theory of word order with special reference to Spanish.
Amsterdam: North-Holland.
Contreras, Heles. 1986. Spanish bare NPs and the ECP. In Generative studies in Spanish
syntax, eds. Ivonne Bordelois, Heles Contreras and Karen Zagona, 25–50.
Dordrecht: Foris.
Contreras, Heles. 1991. On the position of subjects. In Sintax and Semantics 25, 63-79.
Contreras, Heles. 1996. Economy and Projection. In Aspects of Romance Linguistics:
selected papers from the Linguistic Symposium on Romance Languages XXIV
March 10-13, 1994, eds. Claudia Parodi, Carlos Quicoli, Mario Saltarelli and
Maria Luisa Zubizarreta, 149-160. Washington DC: Georgetown University Press.
Corbett, Greville. 2000. Number. Cambridge: Cambridge University Press.
Croft, William. 1988. Agreement vs. Case Marking and Direct Objects. In Agreement in
Natural Language: Approaches, Theories, Descriptions, eds. M. Barlow and C.
Fergunson, 159-179. Stanford: CSLI.
Croft, William. 1990. Typology and Universals. Cambridge: Cambridge University Press.

319
Croft, William. 2003. Typology and Universals. Cambridge: Cambridge University Press.
Csirmaz, Aniko. 2005. Semantics and Phonology in Syntax, MIT: PhD Dissertation.
Cuervo, María Cristina. 2003. Datives at large, MIT: Ph. Dissertation.
Cuervo, Rufino. 1881. Notas a la Gramática Castellana de Andrés Bello. Paris: Roger y
Chernoviz.
Czinglar, Christine. 2001. Decomposing existence. In Issues in Formal German(ic)
Typology, eds. Werner Abraham and C. Jan-Wouter Zwart, 85-126. Amsterdam:
John Benjamins.
De Hoop, Helen. 1992. Case configuration and noun phrase interpretation,
Rijksuniversiteit Groningen: PhD. Dissertation.
De Hoop, Helen, and Lamers, Monique. 2005. Incremental Distinguishability of Subject
and Object. Ms., Radboud University. Nijmegen.
de Jong, Jelly Julia. 1996. The Case of Bound Pronouns in Peripheral Romance,
Groningen University: PhD Dissertation.
De Mello, George. 1991. Pluralización del verbo haber impersonal en el español hablado
culto de once ciudades. Thesaurus. Boletín del Instituto Caro y Cuervo XLVI:445-
471.
De Mello, George. 1994. Pluralización del verbo haber impersonal en el español hablado
culto. Studia Neophilologica 66:77-91.
De Pedro Munilla, Mónica. 2004. Dative Doubling Structures in Spanish: Are They
Double Object Constructions. In Proceedings of the 23rd West Coast Conference
on Formal Linguistics, eds. Vineeta Chand, Ann Kelleher, Angelo J. Rodríguez
and Benjamin Schmeiser, 168-181. Cascadilla Press.
de Swart, Peter. 2003. The Mirror Case, University of Nijmegen: Master Thesis.
Deal, Amy Rose. 2006. The origin and content of expletives: evidence from “selection”.
Ms. University of Massachussetts, Amherst.
Deal, Amy Rose. to appear. Case-marking and object interpretation in Nez Perce.
Proceedings of the 42nd Annual Meeting of the Chicago Linguistic Society.
Delbecque, Nicole. 1998. Why Spanish has two transitive construction frames. Leuvense
Bijdragen 87:387-415.
Delfitto, Denis, and Schroten, Jan. 1991. Bare plurals and the number affix in DP. Probus
3:155-185.
Demonte, Violeta. 1987. C-command, prepositions, and predications. Linguistic Inquiry
18:147-157.
Demonte, Violeta. 1989. Teoría sintáctica. De las estructuras a la rección. Madrid:
Síntesis.
Demonte, Violeta. 1994. La ditransitividad en español: léxico y sintaxis. In Gramática del
español, ed. Violeta Demonte, 431-470. México: El Colegio de México.
Demonte, Violeta. 1995. Dative alternation in Spanish. Probus 7:5–30.
Díaz-Campos, Manuel. 1999-2000. La Pluralización del verbo haber en dos áreas
dialectales de Hispanoamérica. Anuario de Lingüística Hispánica 15-16:235-245.
Díaz-Campos, Manuel. 2003. The Pluralization of Haber in Venezuelan Spanish: A
Sociolinguistic Change in Real Time. IU Working Papers in Linguistics 03-05.
Díaz, Miriam. 2004. Las oraciones existenciales con haber en español en el marco
minimalista, University of Arizona: M.A. Thesis.
Diesing, Molly. 1992. Indefinites. Cambridge, Massachusetts: MIT Press.

320
Diesing, Molly, and Jelinek, Eloise. 1993. The Syntax and Semantics of Object Shift.
Working Papers in Scandinavian Syntax 51:1-54.
Diesing, Molly. 1996. Semantic variables and object shift. In Studies in Comparative
Germanic Syntax, eds. Höskuldur Thráinsson, Samuel David Epstein and Steve
Peter, 66–84. Dordrecht: Kluwer.
DiSciullo, Anna Maria, and Williams, Edwin. 1987. On the Definition of Word.
Cambridge, MA: MIT.
Domínguez, Carmen Luisa, Guzmán, Blanca, Moros, Luis, Pabón, Maryelis, and Vilaín,
Roger. 1998. Personalización de HABER en el español de Mérida. Lengua y Habla
3:23-36.
Emonds, Joseph. 1978. The verbal complex V’-V in French. Linguistic Inquiry 9:151-175.
Enç, Mürvet. 1991. The semantics of specificity. Linguistic Inquiry 22:1-26.
Epps, Patience. 2005. A Grammar of Hup, University of Virginia: PhD Dissertation.
Epstein, Samuel David, and Seely, T. Daniel. 1999. SPEC-ifying the GF "subject";
eliminating A-chains and the EPP within a derivational model. Paper presented at
LSA summer institute workshop on grammatical relations.
Epstein, Samuel David, and Seely, T. Daniel. 2006. Derivations in Minimalism.
Cambridge: Cambridge University Press.
Estrada García, Varinia. 2005. Perfilamiento del Mundo del Hablante en el Complemento
Directo Preposicional Inanimado. Ms., Universidad Nacional Autónoma de México.
Mexico.
Farkas, Donka. 2002. Specificity Distinctions. Journal of Semantics 19:213-243.
Farkas, Donka, and Heusinger, Klaus von. 2003. Referential stability and differential case
marking in Romanian. Ms.
Felber, Sarah. 2002. Two Types of Expletive Constructions. Ms. University of
Connecticut.
Fernández-Ordoñez, Inés. 1999. Leísmo, laísmo y loísmo. In Gramática Descritiva de la
Lengua Española, eds. Ignacio Bosque and Violeta Demonte, 1317-1398. Madrid:
Espasa Calpe.
Fernández-Soriano, Olga. 1989. Rección y Ligamiento en Español: Aspectos del
Parámetro del Sujeto Nulo, Universidad Autónoma de Madrid: Doctoral
dissertation.
Fernández-Soriano, Olga. 1999. Two types of impersonal sentences in Spanish: Locative
and dative subjects. Syntax 2:101-140.
Fernández-Soriano, Olga, and Táboas-Baylin, Susana. 1999. Construcciones impersonales
no reflejas. In Gramática descriptiva de la lengua española, eds. Ignacio Bosque
and Violeta Demonte, 1721-1778. Madrid: Espasa-Calpe.
Fernández Ramírez, Salvador. 1986. Gramática española 4. El verbo y la oración. Madrid:
Arco/Libros.
Fiengo, Robert, and Higginbotham, James. 1981. Opacity in NP. Linguistic Analysis
7:347-373.
Filimonova, Elena. 2005. The noun phrase hierarchy and relational marking: Problems
and counterevidence. Linguistic Typology 9:77–113.
Fillmore, Charles. 1967. The Syntax of Preverbs. Glossa 1:91-125.
Fillmore, Charles. 1968. The Case for Case. In Universals in Linguistic Theory, eds.
Emmon Bach and Robert Thomas Harms, 1-88. New York: Holt, Rinehart, and

321
 Winston.
Fodor, Janet. 1970. The Linguistic Description of Opaque Context, MIT: PhD.
Dissertation.
Folli, Raffaella, and Harley, Heidi. 2004. Flavors of v: Consuming results in Italian and
English. In Aspectual Inquiries, eds. Roumyana Slabakova and Paula
Kempchinsky, 95-120. Dordrecht: Kluwer.
Folli, Raffaella, and Harley, Heidi. 2007. Causation, Obligation, and Argument Structure:
On the Nature of Little v. Linguistic Inquiry 38:197–238.
Fox, Danny, and Lasnik, Howard. 2003. Successive Cyclic Movement and Island Repair:
The Difference Between Sluicing and VP Ellipsis. Linguistic Inquiry 34:143-154.
Fox, Danny, and Pesetsky, David. 2004. Cyclic Linearization of Syntactic Structure.
Theoretical Linguistics 31:1-46.
Franco, Jon. 1993. On object agreement in Spanish, University of Southern California:
Ph.D. Dissertation.
Frank, Robert. 2005. Phase theory and Tree Adjoining Grammar. Ms. Johns Hopkins
University.
Franks, Steven. 1993. On parallelism in across-the-board dependencies. Linguistic Inquiry
24:509–529.
Franks, Steven. 1994. Parametric properties of numeral phrases in Slavic. Natural
Language and Linguistic Theory 12:570-649.
Franks, Steven. 1998. Clitics in Slavic. Paper presented at Comparative Slavic
Morphosyntax Workshop, Spencer Creek, Indiana.
Freeze, R. 1992. Existentials and other locatives. Language 68:553-595.
Freidin, Robert, and Sprouse, Rex A. 1991. Lexical case phenomena. . In Principles and
Parameters in Comparative Grammar, ed. Robert Freidin, 392-416. Cambridge:
MIT.
García, Erica. 1975. The Role of Theory in Linguistic Analysis. The Spanish Pronoun
System. Amsterdam: North-Holland Publishing Company.
García, Erica. 1993. Syntactic diffusion and the irreversibility of linguistic change:
personal a in Old Spanish. In Sprachwandel und Sprachgeschichte: Festschrift für
Helmut Lüdtke, eds. Jürgen Schmidt-Radefeldt and Andreas Harder, 33-50.
Tübingen: Gunter Narr Verlag.
García, Erica. 1995. Relevancia expresiva vs. Desambiguación. In El complemento directo
preposicional, ed. Carmen Pensado. Madrid: Visor.
García García, Marco. 2005. Differential object marking and informativeness. In
Proceedings of the Workshop “Specificity And The Evolution Emergence Of
Nominal Determination Systems In Romance”. Arbeitspapier 119. Konstanz:
Universität Konstanz.
García Llebra, Valentín. 1983. ¿Complemento Directo o sujeto con las formas
unipersonales de haber? Revista de Filología Española LXIII:33-71.
García Martín, José María. 1988. Objetos directos preposicionales no personales en la
prosa castellana de los siglos XIII y XIV. In Actas del I Congreso Internacional de
Historia de la Lengua Española, eds. A. Ariza, A. Salvador and A. Viudas, 363-
376.
García Martín, José María. 1992. Evolución del objeto directo preposicional en la
tradición textual de algunas obras del siglo XIII. Cahiers de Linguistique et de

322
 Civilisation Hispaniques Médiévales 17:47-86.
García Martín, José María. 1996. Sobre la localización originaria del uso del objeto
directo preposicional peculiar del castellano: el testimonio de los documentos
notariales prealfonsíes. In Actas del III Congreso Internacional de Historia de la
Lengua Española: Salamanca, ed. Alegría Alonso González, 281-292. Madrid:
Arco.
García Martín, José María. 1998. Factores diafásicos en el uso del objeto directo
preposicional en las lenguas románicas: Estado de la cuestión. In Estudios de la
Universidad de Cádiz ofrecidos a la memoria profesor Braulio Justel Calabozo,
eds. J. Martín Castellanos, F. Velázquez Basanta and J. Bustamante Costa, 263-
272.
Givón, Talmy. 1978. Definiteness and referentiality. In Universals of Human Language,
ed. Joseph Greenberg, 291-330. Cambridge: MIT.
Grimshaw, Jane. 1990. Argument structure. Cambridge, MA: MIT Press.
Groat, E. M. 1999. Raising the case of expletives. In Working Minimalism, eds. Samuel
David Epstein and Norbert Hornstein, 27-44. Cambridge: MIT.
Grohmann, Kleanthes. 2000. Prolific Peripheries: A radical view from the left, University
of Maryland, College Park: PhD dissertation.
Grohmann, Kleanthes, Drury, John , and Castillo, Juan Carlos. 2000. No more EPP. Paper
presented at WCCFL 19.
Grohmann, Kleanthes, and Haegeman, Liliane M. V. 2002. Resuming Reflexives. In 19th
Scandinavian Conference in Linguistics. ms.
Guasti, Maria Teresa. 1996. Semantic Restrictions in Romance Causatives and the
Incorporation Approach. Linguistic Inquiry 27:294-313.
Guéron, Jacqueline. 1995. On HAVE and BE. Paper presented at NELS 28.
Gutiérrez Ordóñez, Salvador. 1994. Estructuras ecuandicionales. In Gramática del
español, ed. Violeta Demonte, 363-384. Mexico: Colegio de México.
Gutiérrez Ordóñez, Salvador. 1999. Los dativos. In Gramática descriptiva de la lengua
española, eds. Ignacio Bosque and Violeta Demonte, 1855-1930. Madrid: Espasa-
Calpe.
Haider, Hubert. 1985. The Case of German. In Studies in German Grammar, ed. J. Toman,
65-101. Dordrecht: Foris.
Hale, Ken, and Keyser, S. J. 1993. On argument structure and the lexical expressions of
syntactic relations. In The view from building 20: Essays in Linguistics in honor of
Sylvain Bromberger, eds. K. Hale and S. J. Keyser, 53-109. Cambridge, MA: MIT
Press.
Hannay, Michael. 1985. English Existentials in Functional Grammar. Dordrecht: Foris.
Hanssen, Federico. 1945. Gramática Histórica de la Lengua Castellana. Buenos Aires: El
Ateneo.
Harley, Heidi. 1995. Subjects, Events and Licensing, MIT: PhD Dissertation.
Harley, Heidi, and Ritter, Elizabeth. 2002. Person and Number in Pronouns. Language
78:482-526.
Haspelmath, Martin. 2004. Against markedness (and what to replace it with). Ms., Max-
Planck-Institut für evolutionäre Anthropologie.
Heim, Irene. 1990. E-type pronouns and donkey anaphora. Linguistics and Philosophy
13:137–177.

323
Heim, Irene. 1999. Notes on superlatives. Ms. MIT.
Herdan, Simona, and Sharvit, Yael. 2006. Definite and Nondefinite Superlatives and NPI
Licensing. Syntax 9:1–31.
Herdan, Simona. 2007. A superlative theory of amount relatives. Paper presented at
WCCFL 26. University of California, Berkeley.
Herrera Caso, Consuelo. 1997. Acerca de la preposición en el objeto directo y otras
cuestiones. In Contribuciones al estudio de la lingüistica hispánica. Homenaje al
profesor Ramón Trujillo, eds. Manuel Almeida and Josefa Dorta Luis, 197-208.
Santa Cruz de Tenerife y Barcelona: Cabildo de Tenerife y Montesinos.
Heusinger, Klaus von. 2002a. Specificity and definiteness in sentence and discourse
structure. Journal of Semantics 19:245-274.
Heusinger, Klaus von. 2002b. Choice Functions and the Anaphoric Semantics of Definite
NPs. In Proceedings of the Workshop “Choice Functions and Natural Languages
Semantics” Arbeitspapier 110, eds. Klaus von Heusinger, Ruth Kempson and
Wilfried Meyer-Viol, 63-83. Konstanz: Universität Konstanz.
Heusinger, Klaus von, and Kaiser, Georg A. 2003. Animacy, Specificity, and Definiteness
in Spanish. In Proceedings of the Workshop Semantic and Syntactic Aspects of
Specificity in Romance Languages. Arbeitspapier 113, eds. Klaus von Heusinger
and Georg A. Kaiser, 41-65. Konstanz: Universität Konstanz.
Heusinger, Klaus von, and Kaiser, Georg A. 2005. The evolution of differential object
marking in Spanish. In Proceedings of the Workshop “Specificity And The
Evolution Emergence Of Nominal Determination Systems In Romance”.
Arbeitspapier 119, eds. Klaus von Heusinger, Georg A. Kaiser and Elisabeth Stark,
33-70. Konstanz: Universität Konstanz.
Hills, Elijah Clarence. 1920. The accusative "A". Hispania 3 216-222.
Hoekstra, Teun. 1984. Transitivity. Dordrecht: Foris.
Holmberg, Anders. 1986. Word order and Syntactic Features in the Scandinavian
Languages and English, University of Stockholm: PhD Dissertation.
Holmberg, Anders, and Nikanne, Urpo. 1993. Case and other functional categories in
Finnish syntax: Studies in generative grammar 39. Berlin ; New York: Mouton de
Gruyter.
Holmberg, Anders, and Platzack, Christer. 1995. The role of inflection in Scandinavian
Syntax. . Oxford: Oxford University Press.
Hopper, Paul J., and Thompson, Sandra A. 1980. Transitivity in grammar and discourse.
Language 56:251-300.
Hornstein, Nolbert. 1999. Movement and control. Linguistic Inquiry 30:69-96.
Hornstein, Nolbert. 2000. Existentials, A-chains and reconstruction. DELTA, 16:45-79.
Hornstein, Nolbert. 2001. Move! A Minimalist Theory of Construal. Malden, MA:
Blackwell.
Hornstein, Nolbert. 2007. A Very Short Note on Existential Constructions. Linguistic
Inquiry 38:410-411.
Hornstein, Norbert, Rosen, Sara Thomas, and Uriagereka, Juan. 1994. Integrals. In
University of Maryland Working Papers in Linguistic, ed. Jairo Nunes, 70-90.
Volume 2.
Hornstein, Norbert, Rosen, Sara , and Uriagereka, Juan. 2000. Integrals. Ms., University of
Maryland, University of Kansas.

324
Hulk, Aafke, and Pollock, Jean-Yves. 2001. Subject Inversion in Romance and the Theory
of Universal Grammar. Oxford: Oxford University Press.
Hurtado, Alfredo. 1984. La hipótesis de la discordancia. Los clíticos del español y la
gramática universal. Supplement of the Revista Argentina de Lingüística:56- 77.
Isaak, André. 2000. Split Case Marking and Prominence Relations, University of
Massachusetts, Amherst: Ph.D. Dissertation.
Isenberg, Horst. 1968. Das Direkte Objekt im Spanischen: Studia Grammatica 9. Berlin:
Akademie-Verlag.
Jackendoff, Ray. 1983. Semantics and Cognition. Cambridge, MA.: MIT Press.
Jaeggli, Oswaldo. 1982. Topics in Romance Syntax. Dordrecht: Foris.
Jaeggli, Oswaldo. 1986. Three issues in the theory of clitics: case, doubled NPs, and
extraction. In The Syntax of Pronominal Clitics, ed. H. Borer, 15-42. Orlando:
Academic Press.
Jiménez Peña, Silvia. 2001. El papel temático de causa en los predicados de cambio de
estado. Ms. Universidad Autónoma de Barcelona.
Johns, Alana. 1992. Deriving ergativity. Linguistic Inquiry 23:57-87.
Kalmár, Ivan. 1979. The antipassive and grammatical relations in Eskimo. In Ergativity :
towards a theory of grammatical relations, ed. Frans Plank, 117-143. London:
Academic Press.
Kany, Charles. 1951. American-Spanish syntax. Chicago: University of Chicago Press.
Kato, Mary Aizawa. 1999. Strong and Weak pronominals in the null subject parameter.
Probus 11:1-37.
Kato, Mary Aizawa. 2000. The partial Prodrop Nature and the restricted VS order in
Brazilian Portuguese. Ms., UNICAMP.
Kayne, Richard. 1975. French Syntax. The Transformational Cycle. Cambridge: MIT
Press.
Kayne, Richard. 1993. Toward a modular theory of auxiliary selection. In Parameters and
Universals, ed. Richard Kayne (2000), 107-130. New York: Oxford University
Press.
Keenan, Edward. 1987. A Semantic Definition of "Indefinite NP". In The Representation
of (In)definiteness, eds. Eric Reuland and Alice ter Meulen, 286-317. Cambridge,
MA: MIT.
Keenan, Edward. 2003. The Definiteness Effect: Semantics or Pragmatics? . Natural
Language Semantics 11:187-216.
Ketrez, F. Nihan. 2003. Children’s accusative-marked indefinites. Paper presented at
Workshop on Altaic in Formal Linguistics, Massachusetts Institute of Technology,
Cambridge, MA.
King, Larry D. 1984. The Semantics of Direct Object "A" in Spanish. Hispania 67:397-
403.
Kiparsky, Paul. 1998. Partitive case and aspect. In The Projection of Arguments: Lexical
and Compositional Factors, eds. Miriam Butt and Wilhelm Geuder, 265-308.
Stanford: CSLI.
Kiparsky, Paul. 2001. Structural case in finnish. Lingua 111:315-376.
Kiparsky, Paul. 2005. Absolutely a matter of degree: The semantics of structural case in
Finnish. In CLS 41.
Kiss, Katalin É. 1998. Identificational Focus versus Information Focus. Language 74:245-

325
 273.
Kliffer, Michael. 1982. Personal a, kinesis and individuation. In Papers from the XII
Linguistic Symposium on Romance Languages, ed. P. et al Baldi, 195-216.
Amsterdam: John Benjamins.
Koizumi, Masatoshi. 1993. Object Agreement Phrases and the Split VP Hypothesis. In
MIT Working Papers in Linguistics 18, eds. Jonathan David Bobaljik and Colin
Phillips.
Koizumi, Masatoshi. 1995. Phrase structure in minimalist syntax, MIT: PhD Dissertation.
Koopman, Hilda, and Sportiche, Dominique. 1988. Subjects. Ms., UCLA, Los
Angeles,California.
Kratzer, Angelika. 1998. Scope or pseudo-scope? Are there wide-scope indefinites? In
Events in Grammar, ed. Susan Rothstein, 163-196. Dordrecht: Kluwer.
Kratzer, Angelika. 2006. Minimal Pronouns. Ms. University of Massachusetts at Amherst.
Krifka, Manfred. 1989. Nominal Reference, Temporal Constitution and Quantification in
Event Semantics. In Semantics and Contextual Expressions, eds. Renate Bartsch,
Johan van Benthem and Peter van Emde Boa, 75-115. Dordrecht: Foris.
Krifka, Manfred. 1992. Thematic relations as links between nominal reference and
temporal constitution. In Lexical Matters, eds. Ivan Sag and Anna Szabolcsi, 29-53.
Stanford: CSLI.
Krifka, Manfred. 1998. The origins of telicity. In Events and Grammar, ed. Susan
Rothstein, 197-236. Dordrecht: Kluwer.
Laca, Brenda. 1987. El acusativo preposicional en español. Romanisstisches Jahrbuch
38:209-312.
Laca, Brenda. 1995. Sobre el uso del acusativo preposicional en español. In El
complemento directo preposicional, ed. Carmen Pensado, 11-59. Madrid: Visor.
Laca, Brenda. 2006. El objeto directo. La marcación preposicional. In Sintaxis histórica de
la lengua española. Primera parte: La frase verbal, ed. Concepción Company,
423-478. México: Fondo de Cultura Económica / UNAM.
Landau, Idan. 1999. Possessor Raising and the Structure of VP. Lingua 107:1-37.
Landau, Idan. 2001. Elements of Control: Structure and Meaning in Infinitival
Constructions. Dordrecht: Springer.
Lapesa, Rafael. 1980. Historia de la lengua española. Madrid: Gredos.
Lapesa, Rafel. 1968. Sobre los orígenes y evolución del leísmo, laoísmo y loísmo. In
Festschrift von Wartburg. Vol I, 523-551. Tübigen.
Lapointe, Stephen. 1980. A Theory of Grammatical Agreement, University of
Massachusetts: PhD. Dissertation.
Lasnik, Howard. 1992. Case and expletives: Notes toward a parametric account. Linguistic
Inquiry 23:381-405.
Lasnik, Howard. 1995a. Case and expletives revisited: On greed and other human failings.
Linguistic Inquiry 26:615-633.
Lasnik, Howard. 1995b. Case and Expletives Revisited - on Greed and Other Human
Failings. Linguistic Inquiry 26:615-633.
Lasnik, Howard. 1995c. A note on pseudogapping. In MIT Working Papers in Linguistics
27, 143-163.
Lasnik, Howard. 1999a. Pseudogapping puzzles. In Fragments: Studies in Ellipsis and
Gapping eds. Shalom Lappin and Elabbas Benmamoun, 141-174. Oxford Oxford

326
 University Press.
Lasnik, Howard. 1999b. Minimalist analysis. Malden: Blackwell.
Lasnik, Howard. 2003a. Minimalist investigations in linguistic theory. London; New York:
Routledge.
Lasnik, Howard. 2003b. On a Scope Reconstruction Paradox. In Minimalist investigations
in linguistic theory, 118-124. London; New York: Routledge.
Laughren, Mary, and Eisenchlas, Susana. 2006. The role of animacy and definiteness in
the clitic-DP nexus. In Selected Papers from the 2005 Conference of the Australian
Linguistic Society: Edited by Keith Allan.
Lazard, Gilbert. 1989. Transitivity and markedness: the antipassive in accusative
languages. In Markedness in synchrony and diachrony, ed. Olga Mišeska Tomić,
309-331. Berlin: Mouton de Gruyter.
Lebeaux, David. 1988. Language acquisition and the form of the grammar, University
of Massachusetts, Amherst: PhD. Dissertation.
Legate, Julie Anne. 2003. Some Interface Properties of the phase. Linguistic Inquiry
34:506-516.
Legate, Julie Anne. 2005. Phases and cyclic agreement. In Perspectives on Phases: MIT
Working Papers in Linguistics, 49, eds. Martha McGinnis and Norvin Richards,
147-156.
Leonetti, Manuel. 1999. El artículo. In Gramática Descriptiva de la Lengua Española, eds.
Ignacio Bosque and Violeta Demonte, 787-890. Madrid: Espasa-Calpe.
Leonetti, Manuel. 2003. Specificity and Object Marking: the Case of Spanish a. In
Proceedings of the Workshop "Semantic and Syntactic Aspects of Specificity in
Romance Languages". eds. Klaus von Heusinger and Georg A. Kaiser, 67-101.
Fachbereich Sprachwissenschaft: Universität Konstanz.
Leonetti, Manuel. 2004. Specificity and Differential Object Marking in Spanish. Catalan
Journal of Linguistics 3:75-114.
Leonetti, Manuel. 2005. Definiteness Effects and the Role of the Coda in Existential
Constructions. Ms.
Leonetti, Manuel. in press-a. Sobre la relación entre doblado de clíticos y movimiento de
objetos. Cuadernos de Lingüística (Instituto Universitario Ortega y Gasset) 14.
Leonetti, Manuel. in press-b. Clitics do not Encode Specificity. In Proceedings of the
Workshop «Definiteness, Specificity and Animacy in Ibero-Romance Languages».
Arbeitspapier-Fachbereich Sprachwissenschaft der Universität Konstanz, eds.
Georg A. Kaiser and Manuel Leonetti. Konstanz: Universität Konstanz.
Li, Y.-H. Audrey. 2001. The ba construction. Ms.
Lidz, Jeffrey. 2006. The Grammar of Accusative Case in Kannada Language 82:10-32.
Lima, Ananda. 2003. Differential Object Behaviour, University of Queensland: Honors
Thesis.
Lima, Ananda. 2006. A Minimalist View on Differential Object Marking for Specificity.
In 36th Annual Meeting of the Michigan Linguistics Society. Oakland University.
Linares, Carlo. 2005. Elementos para una teoría de clíticos y concordancia, Pontificia
Universidad Católica del Perú: Tesis de Licenciatura.
Lois, Ximena. 1989. Aspects de la syntaxe de l'espagnol et theorie de la grammaire,
Université de Paris, VIII: Thèse de doctorat nouveau régime.
Longa, Víctor, Lorenzo, Guillermo, and Rigau, Gemma. 1998. Subject clitics and clitic

327
 recycling: locative sentences in some Iberian Romance languages. Journal of
Linguistics 34:125-164.
Longobardi, Giorgio. 1994. Proper names and the theory of N-movement in syntax and
logical form. Linguistic Inquiry 25:609-665.
López, Luis. 2001. On the (Non)complementarity of Theta-Theory and Checking Theory.
Linguistic Inquiry 32:694–716.
López, Luis. 2006. A Derivational Syntax for Information Structure. Ms., University of
Illinois-Chicago. To appear in the ZASPiL series.
Luján, Marta. 1978. Direct object nouns and the preposition "a" in Spanish. Texas
Linguistic Forum 10:30-52.
Luque Moreno, Jesús. 1978. En torno al sintagma "haber impersonal + sustantivo" y sus
orígenes latinos. Revista Española de Lingüística 8:125-147.
Lyons, Christopher. 1999. Definiteness. Cambridge: Cambridge University Press.
Mahajan, Anoop. 1992. The Specificity Condition and the CED. Linguistic Inquiry
23:510-516.
Manzini, Rita, and Savoia, Leonardo. 2002. Parameters of Subject Inflection in Italian
Dialects. In Subjects, Expletives and the EPP, ed. Peter Svenonius, 157-199.
Oxford: Oxford University Press.
Marantz, Alec. 1984. On the Nature of Grammatical Relations. Cambridge, MA: MIT.
Marantz, Alec. 1990. Implications of Asymmetries in Double Object Constructions. Ms.
University of North Carolina at Chapel Hill.
Marantz, Alec. 1991. Case and Licensing. Paper presented at ESCOL.
Martín, Juan. 1999. The Syntax and Semantics of Spanish Accusative a. In Advances in
Hispanic Linguistics: Papers from the 2nd Hispanic Linguistics Symposium, eds.
Javier Gutiérrez-Rexach and Fernando Martínez-Gil, 469-485: Cascadilla Press.
Martín, Juan. 2005. Aspectual Quantization and [±] Accusative Case Checking in
Romance. In Romance Languages and Linguistic Theory 2003, eds. Twan Geerts,
Ivo van Ginneken and Haike Jacobs, 177–196. Amsterdam: John Benjamins.
Martín, Juan. 2006. Las construcciones de participio absoluto y su relevancia para la
Teoría del Caso. In Actas del XXXV Simposio Internacional de la Sociedad
Española de Lingüística, ed. Milka Villayandre Llamazares. León: Universidad de
León, Dpto. de Filología Hispánica y Clásica.
Martin, Roger. 1992. Case theory, A-chains, and expletive replacement. Ms., University of
Connecticut.
Massam, Diane. 2001. Pseudo Noun Incorporation in Niuean. Natural Language and
Linguistic Theory 19:153-197.
Masullo, Pascual. 1992. Incorporation and case theory in Spanish. A crosslinguistic
perspective, University of Washington: PhD. Dissertation.
Mayer, Elisabeth. 2006. Optional direct object clitic doubling in Limeño Spanish. In The
11th International LFG Conference. University of Konstanz.
McGinnis, Martha. 1998. Locality in A-movement, MIT: PhD Dissertation.
McGinnis, Martha. 2001. Variation in the phase structure of applicatives. In Linguistic
Variation. Yearbook 1, 105–146.
McNally, Louise. 1997. A Semantics for the English Existential Construction. New York:
Garland
Meinunger, André. 2000. Syntactic Aspects of Topic and Comment. Amsterdam: John

328
 Benjamins.
Menn, Lise, and MacWhinney, Brian. 1983. The repeated morph constraint: Toward an
explanation. Language 60:519-541.
Merchant, Jason. 2001. The syntax of silence: Sluicing, islands, and the theory of ellipsis.
Oxford: Oxford University Press.
Meyer-Lübke, Wilhelm. 1899. Grammaire des langues romanes. Vol. III. Paris: Welter.
Miles, Cecil. 1990. About the Spanish verb haber. Canadian Modern Language Review
46:317-323.
Milsark, Gary. 1974. Existential Sentences in English, MIT: Ph.D. dissertation.
Milsark, Gary. 1977. Toward an explanation of certain peculiarities of the existential
construction in English. Linguistic Analysis 3:1-29.
Mohanan, Tara. 1994. Case OCP: A constraint on word order in Hindi. In Theoretical
perspectives on word order in South Asian languages, eds. Miriam Butt, Tracy
Holloway King and Gillian Ramchand, 185-216. Stanford: CSLI.
Molho, Maurice. 1958. La question de l'objet en espagnol. Vox Romanica XVII:209-219.
Montalbetti, Mario. 1984. After Binding. On Interpretation of Pronouns, MIT: Ph.
Dissertation.
Montes, José Joaquín. 1982. Sobre el sintagma "haber + sustantivo". Thesaurus. Boletín
del Instituto Caro y Cuervo XXXVII:383-385.
Montrul, Silvina, and Perpiñán, Silvia. 2006. On Binding Asymmetries in Dative
Alternation Constructions in L2 Spanish. In Selected Proceedings of the 7th
Conference on the Acquisition of Spanish and Portuguese as First and Second
Languages, eds. Carol A. Klee and Timothy L. Face, 135-148. Somerville, MA:
Cascadilla Press.
Moravcsik, Edith. 1978. On the case marking of objects. In Universals of Human
Language, eds. Joseph Greenberg, Charles Fergunson and Edith Moravcsik, 249-
289. Stanford: Stanford University Press.
Morimoto, Yukiko, and de Swart, Peter. 2006. Language Variation and Historical Change:
The Spread of DOM in Spanish. In Romance Languages and Linguistic Theory
2004. Selected papers from ‘Going Romance’, Leiden, 9–11 December 2004, eds.
Jenny Doetjes and Paz González, 225–245 Amsterdam: John Benjamins.
Næss, Ashild. 2004a. What Markedness Marks: The Markedness Problem with Direct
Objects. Lingua 114:1186–1212.
Næss, Ashild. 2004b. Transitivity: From Semantics to Structure, Universiteit Nijmegen:
Doctoral Dissertation.
Navarro Correa, Manuel. 1991. Valoración social de algunas formas verbales en el habla
de Valencia (Venezuela). Thesaurus. Boletín del Instituto Caro y Cuervo
XLVI:304-315.
Neeleman, Ad, and van de Koot, Hans. 2004. Syntactic OCP Effects. Ms. UCL.
Nevins, Andrew. 2007. The Representation of Third Person and its Consequences for
Person-Case Effects. Natural Language & Linguistic Theory 25:273-313.
Newmeyer, Frederick J. 2002. Optimality and Functionality: A Critique of Functionally-
Based Optimality-Theoretic Syntax Natural Language & Linguistic Theory 20:43–
80.
Nicita, Linda M. 1997. Impersonal Haber and the road to copula function. Colorado
Research in Linguistics 15:32-46.

329
Niculescu, Alexandru. 1959. Sur l’objet direct prépositionnel dans les langues romanes. In
Recueil d’études romanes, IXème Congrès International de linguistique romane à
Lisbonne, 167-185. Bucarest.
Nissenbaum, Jon. 2000. Investigations of covert phrase movement, MIT: PhD dissertation.
Nomura, Masashi. 2005. Nominative Case and AGREE(ment), University of Connecticut:
PhD Dissertation.
Núñez del Prado, Zelmira. 1997. Spanish Accusative Clitics a Strong determiners. Cornell
Working Papers in Linguistics 15:281-261.
Obediente, Enrique. 1984. La personalización de haber en el habla culta de Caracas. Actas
del VII Congreso de la ALFAL. Santo Domingo, RD:51-61.
Olarrea, Antxon. 1996. Pre- and Postverbal Subject Position in Spanish: A Minimalist
Account, University of Washington: PhD Dissertation.
Ordóñez, Francisco. 1997. Word Order and Clause Structure in Spanish and Other
Romance Languages, City University of New York: Doctoral Dissertation.
Ordóñez, Francisco, and Treviño, Esthela. 1999. Left dislocated subjects and the pro-drop
parameter: A case of study of Spanish. Lingua 107:39-68.
Ortega-Santos, Iván. 2004. Impersonal Constructions and the EPP in Spanish. Paper
presented at West Coast Conference on Formal Linguistics 23.
Ortega-Santos, Iván. 2006. On New Information Focus, Sentence Stress Assignment
Conditions and the Copy Theory: A Spanish Conspiracy. In University of
Maryland Working Papers in Linguistics 14, eds. Nina Kazanina, Utako Minai,
Philip Monahan and Heather Taylor, 88-212. College Park, MD: UMWPiL.
Öztürk, Balkiz. 2005. Case, Referentiality and Phrase Structure Amsterdam John
Benjamins.
Park, Bum-Sik. 2005. Identity and Locality in Ellipsis, University of Connecticut: Ph.D.
Dissertation.
Patel, Pritty. 2007. Split Ergativity & the Nature of Agreement, UCL: MA Thesis.
Payne, Doris L., and Barshi, Immanuel eds. 1999. External Possession. Amsterdam: John
Benjamins.
Pensado, Carmen. 1995. El complemento directo preposicional: estado de la cuestión y
bibliografía comentada. In El complemento directo preposicional, ed. Carmen
Pensado, 11-59. Madrid: Visor.
Pérez Saldanya, Manuel. 1999. El modo en las subordinadas relativas y adverbiales. In
Gramática Descriptiva de la Lengua Española, eds. Ignacio Bosque and Violeta
Demonte, 3209-3251. Madrid: Espasa.
Pérez Silva, Jorge. 1990. Sobre la frase nominal castellana, Pontificia Universidad
Católica del Perú: BA Thesis.
Perlmutter, David M. 1971. Deep and surface structure constraints in syntax:
Transatlantic series in linguistics. New York,: Holt Rinehart and Winston.
Pesetsky, David Michael. 1987. Wh-in Situ: Movement and Unselective Binding. In The
Representation of (In)definiteness, eds. Eric Reuland and Alice ter Meulen, 98-
129. Cambridge, Mass.: MIT Press.
Pesetsky, David Michael. 1989. Language-Particular Processes and the Earliness Principle.
Ms. MIT.
Pesetsky, David Michael. 1997. Optimality theory and syntax: Movement and
pronunciation. In Optimality theory: An overview, eds. Diana Archangeli and

330
 Terence Langendoen, 134-170. Malden, Mass.: Blackwell.
Pesetsky, David Michael. 1998. Some optimality principles of sentence pronunciation. In
Is the Best Good Enough, eds. Pilar Barbosa, Danny Fox, Paul Hagstrom, Martha
McGinnis and David Michael Pesetsky, 337-383. Cambridge, Mass.: MIT Press.
Pesetsky, David Michael, and Torrego, Esther. 2007. The syntax of valuation and the
interpretability of features. In Phrasal and Clausal Architecture. Syntactic
derivation and interpretation. In honor of Joseph E. Emonds, eds. Simin Karimi,
Vida Samiian and Wendy Wilkins, 262–294. Amsterdam: John Benjamins.
Picallo, M. Carme. 1998. On the Extended Projection Principle and null expletive subjects.
Probus 10:219-241.
Plag, Ingo. 1998. Morphological haplology in a constraint-based morpho-phonology. In
Phonology and Morphology of the Germanic Languages, eds. Wolfgang Kehrein
and Richard Wiese, 199-215. Tübingen: Niemeyer.
Prince, Alan, and Smolensky, Paul. [1993] 2004. Optimality Theory: Constraint
Interaction in Generative Grammar. Malden, MA: Blackwell.
Pylkkänen, Liina. 2002. Introducing Arguments, MIT: Ph.D. Dissertation.
Quer, Josep. 1998. Mood at the Interface. The Hague: Holland Academic Graphics.
Quine, Willard Van Orman. 1960. Word and object. New York: Wiley
Ramchand, Gillian. 1997. Aspect and predication. Oxford: Clarendon Press.
Rando, Emily, and Napoli, Donna Jo. 1978. Definites in THERE-sentences. Language
54:300-313.
Real Academia Española. 1931. Gramática de la Lengua Española. Madrid: Espasa-Calpe.
Rebollo Torío, Miguel Angel. 1976/1977. Algunas observaciones sobre la
presencia/ausencia de a + complemento directo. Revista de lingüística teórica y
aplicada 14-15:49-58.
Reed, Lisa. 1992. Remarks on Word Order in Causative Constructions. Linguistic Inquiry
23:164-172.
Reglero, Lara. 2004. A’-dependencies in Spanish and Basque, University of Connecticut:
PhD dissertation.
Reinhart, Tanya. 1992. Wh-in-situ. An apparent paradox. In Proceedings of the Eight
Amsterdam Colloquium, eds. P. Dekker and M. Stokhof, 483-492. Amsterdam:
ILLC, University of Amsterdam.
Reinhart, Tanya. 1997. Quantifier Scope: How Labor is Divided between QR and Choice
Functions. Linguistics and Philosophy 20:335-397.
Reinhart, Tanya. 2006. Interface Strategies. Cambridge, MA: MIT Press.
Reuland, Eric ed. 2000. Arguments and Case. Explaining Burzio's Generalization.
Amsterdam: John Benjamins.
Richards, Norvin. 2001. Movement in Language: Interactions and Architectures. Oxford:
Oxford University Press.
Richards, Norvin. 2006. A Distinctness Condition on Linearization Ms. MIT.
Rigau, Gemma. 1994. Catalan Presentational Sentences and the properties of AGR. In
Paths Towards Universal Grammar. Studies in Honor of Richard S. Kayne., eds.
Guglielmo Cinque, Jan Koster, J. Y. Pollock, Luigi Rizzi and Rafaella Zanuttini,
343-360. Washington D.C.: Georgetown University Press.
Rigau, Gemma. 2001. Temporal existential constructions in Romance. In Romance
Languages and Linguistic Theory 1999, eds. Yves D'hulst, Johan Rooryck and Jan

331
 Schroten, 307-334. Amsterdam: John Benjamins.
Ritter, Elizabeth, and Rosen, Sara Thomas. 2001. The Interpretive Value of Object
Agreement. Language Sciences 23:425-451.
Ritter, Elizabeth, and Rosen, Sara Thomas. 2005. Topic or Aspect: Functional heads,
features and the grammaticization of events. In Aspectual Inquiries, eds. Paula
Kempchinsky and Roumyana Slabakova, 21-39. Netherlands: Springer
Rivero, Maria Luisa. 1977. Specificity and Existence: A Reply. Language 53:32-48.
Rodríguez-Mondoñedo, Miguel. 2006a. A restriction on the Definiteness Effect in Spanish.
In NELS 37. University of Illinois at Urbana-Champaigne.
Rodríguez-Mondoñedo, Miguel. 2006b. The Acquisition of Differential Object Marking in
Spanish. Ms. University of Connecticut.
Rodríguez-Mondoñedo, Miguel. 2006c. An interpretable [person] in T and a phi-
incomplete small v for Spanish. Paper presented at Western Conference on
Linguistics (WECOL) 2006, California State University, Fresno.
Rodríguez-Mondoñedo, Miguel. 2006d. Spanish existentials and other accusative
constructions. In Minimalist Essays, ed. Cedric Boeckx, 326–394. Amsterdam:
John Benjamins.
Roegiest, Eugeen. 1979. A propos de l'accusatif prépositionnel dans quelques langues
romanes. Vox Romanica 38:37-54.
Roegiest, Eugeen. 1998. Variación del objeto directo español y dinamicidad verbal. In
Estudios en honor del profesor Josse de Kock eds. Nicole Delbecque and Christian
de Paepe, 469-488. Leuven: Leuven University Press.
Rude, Noel. 1985. Studies in Nez Perce grammar and discourse, University of Oregon:
Ph.D. dissertation.
Runner, Jeffrey. 1992. Expletives, 'replacement', and economy. In Catalan Working
Papers in Linguistics, ed. A. Branchadell et al., 281-317. Barcelona: Universitat
Autònoma de Barcelona.
Ruys, Eduard G. 1992. The Scope of Indefinites, Utrecht University: PhD. Dissertation.
Saab, Andrés Leandro. 2006. Accusative Case, Subject Agreement: the Latin American
Spanish existential construction with haber. Paper presented at Romania Nova.
Universidade Federal de Alagoas.
Safir, Kenneth. 1985. Syntactic chains. Cambridge: Cambridge University Press.
Saito, Mamoru, and Hoshi, Hiroto. 1994. Japanese light verb construction and the
Minimalist Program. Ms., University of Connecticut.
Sánchez, Liliana. 2005. Clitic Doubling and the Checking of Exhaustive Identificational
Focus in Limenian Spanish. Ms., Rutgers University. New Brunswick.
Schmidt, Bodil Kappel. 2003. West Greenlandic antipassive. Proceedings of SCL 19.
University of Tromsø Working Papers on Language and Linguistics 31:385-399.
Schoorlemmer, Erik. 2005. Agree and Existentials. Paper presented at Second Syntax AiO
Meeting (SAM 2), Utrecht.
Schroten, Jan. 1972. Concerning the Deep Structures of Spanish Reflexive Sentences. The
Hague: Mouton.
Selkirk, Elisabeth. 1977. Some Remarks on Noun Phrase Structure. In Formal Syntax, eds.
Peter W. Culicover, Thomas Wasow and Adrian Akmajian. New York: Academic
Press.
Siewierska, Anna. 2004. Person. Cambridge: Cambridge University Press.

332
Silva-Corvalán, Carmen. 2001. Sociolingüística y pragmática del español. Washington:
Georgetown University Press.
Silvernstein, Michael. 1976. Hierarchy of features and ergativity. In Grammatical
categories in Australian languages, ed. Robert M. W. Dixon, 112-171. Camberra:
Australian Institute of Aboriginal Studies.
Smolensky, Paul. 1995. On the internal Structure of the constraint component Con of UG.
Ms. ROA-86-000. http://roa.rutgers.edu.
Sportiche, Dominique. 1995. French predicate clitics and clause structure. In Syntax and
Semantics, eds. Anna Cardinaletti and Maria Teresa Guasti, 287-324. New York:
Academic Press.
Stemberger, Joseph. 1981. Morphological haplology. Language 57:791-817.
Stepanov, Arthur. 2001a. Late Adjunction and Minimalist Phrase Structure. Syntax 4:94-
125.
Stepanov, Arthur. 2001b. Cyclic Domains: A Minimalist Study on Adjunction and
Extraction, University of Connecticut: PhD Dissertation.
Stepanov, Arthur. 2002. Derivational Properties of Inherent Case. Ms., Universität
Potsdam.
Stjepanović, Sandra. 1999. What do Second Position Clitization, Scrambling and Multiple
Wh-Fronting Have in Common?, University of Connecticut: PhD Dissertation.
Stjepanović, Sandra. 2003. A Word Order Paradox Resolved by Copy Deletion at PF. In
Linguistic Variation Yearbook, ed. Pierre Pica, 139-177. Amsterdam: John
Benjamins.
Stowell, Tim. 1981. Origins of Phrase Structure, MIT: Doctoral dissertation.
Stowell, Tim. 1983. Subjects across categories. The Linguistic Review 2:258-312.
Strotzer, Judith. 1976. Clitics in Spanish: PhD dissertation, UCLA.
Suñer, Margarita. 1982a. Syntax and Semantics of Spanish Presentational Sentence-Types.
Washington, D.C.: Georgetown University Press.
Suñer, Margarita. 1982b. Impersonal haber in Non Standard Spanish. Cornell Working
Papers in Linguistics:94-107.
Suñer, Margarita. 1988. The Role of Agreement in Clitic-Doubled Constructions. Natural
Language & Linguistic Theory 6:391-434.
Suñer, Margarita. 1999. Clitic-doubling of strong pronouns in Spanish: an instance of
object shift. In Grammatical analyses in Basque and Romance linguistics, eds. Jon
Franco, Alazne Landa and Juan Martín, 233-256. Amsterdam: John Benjamins.
Suñer, Margarita. 2000. Object Shift: Comparing a Romance Language to Germanic.
Probus 12:261-289.
Svenonius, Peter. 2002. Subjects, expletives, and the EPP: Oxford studies in comparative
syntax. Oxford ; New York: Oxford University Press.
Svenonius, Peter. 2006. Interpreting Uninterpretable Features. Ms. To appear in Linguistic
Analysis.
Szabolcsi, Anna. 1983. The possessor that ran away from home. The Linguistic Review
3:89- 102.
Szabolcsi, Anna. 1986. Comparative superlatives. MIT Working Papers in Linguistics
8:245-265.
Tenny, Carol. 1994. Aspectual Roles and the Syntax-Semantics Interface. Dordrecht:
Kluwer.

333
Thráinsson, Höskuldur. 2001. Object shift and scrambling. In A handbook of
contemporary syntactic theory, eds. Mark Baltin and Chris Collins, 148-202. New
York: Blackwell.
Ticio, Emma. 2003. On the structure of DPs, University of Connecticut: Ph.D.
Dissertation.
Ticio, Emma. 2004. On the position of subjects in Puerto Rican Spanish. University of
Connecticut Working Papers in Linguistics 12:78-92.
Ticio, Emma. 2005. Locality and Anti-Locality in Spanish DP. Syntax 8:226-286.
Torrego, Esther. 1984a. Determinerless NPs. Ms., University of Massachusetts. Boston.
Torrego, Esther. 1984b. Algunas observaciones sobre las oraciones existenciales con
haber en español. Estudis gramaticals 1:329-339.
Torrego, Esther. 1998. The dependencies of objects: Linguistic inquiry monographs 34.
Cambridge, Mass.: MIT Press.
Torrego, Esther. 1999. El complemento directo preposicional. In Gramática Descriptiva
de la Lengua Española, eds. Ignacio Bosque and Violeta Demonte, 1779-1806.
Madrid: Espasa.
Torrego, Esther. 2002. Aspects in Prepositional System of Romance. In Current Issues in
Romance Linguistics, eds. Teresa Satterfield, Christina Tortora and Diana Cresti,
337-358. Amsterdam: John Benjamins.
Travis, Lisa. 1992. Inner Aspect and the Structure of VP. Ms. McGill University.
Treviño, Esthela. 1994. Las causativas del español con complemento infinitivo. México:
El Colegio de México.
Treviño, Esthela. 2003. On the Subjecthood Issue of Existential Haber. In Linguistic
Theory and Language Development in Hispanic Language, eds. Silvina Montrul
and Francisco Ordóñez, 178-192. University of Illinois at Urbana-Champaign:
Cascadilla Press.
Uriagereka, Juan. 1998. From Being to Having: Questions about ontology from a
Kayne/Szabolcsi Syntax. In Romance Linguistics: Theoretical Perspectives, eds.
Armin Schwegler, Myriam Uribe-Etxebarria and Bernard Tranel, 283-306.
Amsterdam/Philadelphia: John Benjamins.
Vainikka, Anne, and Maling, Joan. 1996. Is Partitive Case Inherent or Structural? In
Partitives. Studies on the distribution and meaning of partitive expressions, ed. J.
Hoeksema, 179-208. Amsterdam: Mouton de Gruyter.
van Bergen, Geertje. 2006. To ba or not to ba. Differential Object Marking in Chinese,
Radboud University Nijmegen: Master Thesis.
Van Geenhoven, Veerle. 1998. Semantic Incorporation and Indefinite Descriptions:
Semantic and Syntactic Aspects of Noun Incorporation in West Greenlandic.
Stanford: CSLI.
Velásquez, Héctor. 1991. Subordinación argumental, Pontificia Universidad Católica del
Perú: BA Thesis.
Vergnaud, Jean, and Zubizarreta, Maria Luisa. 1992. The definite determiner and the
inalienable constructions in English and French. Linguistic Inquiry 23:595-652.
Verkuyl, Henk. 1972. On the Compositional Nature of the Aspects. Dordrecht: Reidel.
Vermeulen, Reiko. 2005. The Syntax of External Possession: Its Basis in Theta-theory
University College London: PhD Dissertation.
Vikner, Sten. 1995. Verb movement and expletive subjects in the Germanic languages:

334
 Oxford studies in comparative syntax. New York: Oxford University Press.
von Fintel, Kai, and Heim, Irene. 2003. Beyond de re–de dicto: The Third reading. Ms.,
MIT.
Weissenrieder, Maureen. 1985. Exceptional uses of the Accusative a. Hispania 68:393-
398.
Weissenrieder, Maureen. 1990. Variable Uses of the Direct-Object Marker a. Hispania
73:223-231.
Weissenrieder, Maureen. 1991. A Functional Approach to the Accusative a. Hispania
74:146-156.
Westphal, German F. 1980. Subjects and Pseudo-Subjects in Spanish: The Verb
Agreement Question in the Impersonal Se Construction: Current Inquiry into
Language and Linguistics Monograph # 31. Edmonton-Carbondale: Linguistic
Research, Inc.
Wharram, Douglas. 2003. On the interpretation of (un)certain indefinites in Inuktitut and
related languages, University of Connecticut: PhD Dissertation.
Woolford, Ellen. 2003. Burzio's Generalization, Markedness, and Locality Constraints on
Nominative Objects. In New Perspectives on Case Theory, eds. Ellen Brandner
and Heike Zinsmeister, 299-327. Stanford: CSLI.
Wunderlich, Dieter. 1997. Cause and the structure of verbs. Linguistic Inquiry 28:27-68.
Wurmbrand, Susi. 2006. License Case. Journal of Germanic Linguistics 18:174-234.
Yip, Moira, Maling, Joan, and Jackendoff, Ray. 1987. Case in Tiers. Language 63:217-
250.
Yip, Moira. 1998. Identity avoidance in phonology and morphology. In Morphology and
its relation to phonology and syntax, eds. Steven Lapointe, Diane Brentari and
Patrick Farrell, 216-246. Stanford: CSLI.
Yoshida, Masaya. 2003. The Specificity Condition: PF-Condition or LF-Condition? In
WCCFL 22 Proceedings, eds. G. Garding and M. Tsujimura, 547-560. Somerville,
MA: Cascadilla Press.
Zagona, Karen T. 2002. The Syntax of Spanish. Cambridge: Cambridge University Press.
Zou, Ke. 1993. The syntax of the Chinese ba construction. Linguistics 31:715-736.
Zubizarreta, María Luisa. 1985. The Relation between Morphophonology and
Morphosyntax: The Case of Romance Causatives. Linguistic Inquiry 16:247-289.
Zubizarreta, María Luisa. 1987. Levels of representation in the lexicon and in the syntax:
Studies in generative grammar 31. Dordrecht: Foris.
Zubizarreta, María Luisa. 1994. El Orden de Palabras en Español y el Caso Nominativo.
In Gramática del Español, ed. Violeta Demonte, 21-49. Mexico: El Colegio de
México.
Zubizarreta, María Luisa. 1999. Word Order in Spanish and the Nature of Nominative
Case. In Beyond Principles and Parameters. Essays in memory of O. Jaeggli, eds.
K. Johnson and I. Roberts, 223-250. Dordrecht: Kluwer Academic Press.
Zucchi, Alessandro. 1995. The ingredients of the definiteness effect. Natural Language
Semantics 3:33-78.

335