Interface Perspectives on Clausal Complementation : The Case of Subjunctive Obviation

Abstract

This book investigates the properties of a phenomenon occurring in the main Western Romance languages, subjunctive obviation, the unavailability of coreference involving the subject of a subjunctive clause and the experience of the main clause.

Full text

Francesco Costantini
Interface Perspectives
on Clausal Complementation
The Case of Subjunctive Obviation

Francesco Costantini, Interface Perspectives on Clausal Complementation.
The Case of Subjunctive Obviation
© 2009 Università Ca’ Foscari Venezia
ISBN 978-88-7543-245-4
In copertina: Alexander Calder, Arc of Petals, 1941, Painted aluminum, approximately 214
cm high, Peggy Guggenheim Collection, Venice 76.2553 PG 137, © Alexander
Calder, by SIAE 2008
Libreria Editrice Cafoscarina
Dorsoduro 3259, 30123 Venezia
www.cafoscarina.it

Contents
Preface............................................................................................................ 9
1 Introduction ............................................................................................... 11
1.1 Preliminary remarks .......................................................................... 11
1.2 Theoretical relevance of the phenomenon.......................................... 13
1.3 Outline of the study ............................................................................ 14
2 Subjunctive obviation: preliminary remarks ............................................. 19
2.1 Introduction ........................................................................................ 19
2.2 Obviation and mood in Italian............................................................ 19
2.3 Obviation comparatively .................................................................... 22
2.3.1 Catalan............................................................................................. 22
2.3.2 French.............................................................................................. 23
2.3.3 Portuguese ....................................................................................... 23
2.3.4 Spanish ............................................................................................ 24
2.3.5 Summary ......................................................................................... 25
2.3.6 An aside on Latin ........................................................................... 25
2.4 Obviation and de se ............................................................................ 27
2.4.1 Castañeda’s examples...................................................................... 27
2.4.2 Subjunctive clauses and the de se reading....................................... 28
2.5 Conclusion.......................................................................................... 30
3 Environments weakening subjunctive obviation...................................... 33
3.1 Chapter outline ................................................................................... 33
3.2 Obviation and subjunctive mood........................................................ 34
3.2.1 Obviation and subjunctive ‘tenses’ ................................................. 34
3.2.1.1 Obviation ‘weakening’ and periphrastic tenses............................ 34
3.2.1.2 Another aside on Latin ................................................................ 37
3.2.2 Passive voice .................................................................................. 38
3.2.3 Modal verbs.................................................................................... 40
3.2.3.1 Romance languages...................................................................... 40
3.2.3.2 Again an aside on Latin............................................................... 41
3.2.4 Other periphrastic forms and restructuring verbs............................ 42
3.2.5 Combining auxiliaries and modals .................................................. 44
3.3 Obviation and arguments.................................................................... 45
3.3.1 Embedded subject............................................................................ 45
3.3.1.1 Theta-role sensitivity.................................................................... 45

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
6
3.3.1.3 Strong pronouns ........................................................................... 47
3.3.1.4 Overlapping reference .................................................................. 48
3.3.2 Embedded object ............................................................................. 49
3.3.3 Matrix arguments ............................................................................ 50
3.3.3.1 Introduction .................................................................................. 50
3.3.3.2 Experiencer arguments ................................................................. 50
3.3.3.3 Arguments of directive verbs ....................................................... 52
3.3.3.4 Possessives and genitives ............................................................. 53
3.4 Double embedding ............................................................................ 54
3.5 Adverbial clauses ............................................................................... 56
3.6 Topics and clause dislocation............................................................. 58
3.6.1 Left dislocated topics....................................................................... 58
3.6.2 Clause dislocation............................................................................ 60
3.7 Coordination....................................................................................... 60
3.8. Conclusions ....................................................................................... 61
4 Theories of subjunctive obviation ............................................................. 63
4.1 Intorduction ........................................................................................ 63
4.2 ‘Competition’ theories....................................................................... 64
4.3 Binding-Theoretical approaches....................................................... 67
4.3.1 Binding Theory Principle B ............................................................ 67
4.3.2 Hypothesis on the ‘binding domain extension’ ............................... 68
4.3.2.1 Tense anaphoricity ....................................................................... 68
4.3.2.2 Williams’ Associations................................................................. 69
4.3.2.3 I-to-C movement .......................................................................... 70
4.3.2.4 Nominative Case marking ............................................................ 71
4.3.2.5 I-to-C and C-to-C movement ....................................................... 71
4.3.2.6 C-to-V movement......................................................................... 72
4.3.2.7 Exceptional Case Marking ........................................................... 72
4.3.2.8 Syntactic dependency ................................................................... 73
4.3.3 Explaining obviation through Binding Theory................................ 73
4.4 Remarks on the theories on obviation ................................................ 76
4.4.1 Integrating ‘competition’ and Binding Theory approaches............. 76
4.4.2 Competition theories: an analysis.................................................... 77
4.4.3 Binding Theory-based theories under scrutiny................................ 81
4.5 Conclusive remarks ............................................................................ 82
5 Subjunctive mood at the interface ............................................................. 85
5.1 Introduction ........................................................................................ 85
5.2 Complementizer Deletion and Double Accessibility Reading ........... 86
5.2.1 Complementizer Deletion descriptively .......................................... 86
5.2.2 Common properties of Complementizer Deletion and obviation.... 90
5.2.3 CP-layer, AgrP and Complementizer Deletion ............................... 91

CONTENTS
7
5.2.4 CP, MOODP, and DAR .................................................................. 92
5.2.5 Subjunctive tenses as anaphors ....................................................... 92
5.3 Long-distance Anaphors..................................................................... 95
5.3.1 The syntax of LDAs ........................................................................ 95
5.3.2 Common properties of LDAs and obviation ................................... 98
5.3.3 LDAs and indexicality..................................................................... 99
5.4 Conclusion........................................................................................ 103
6 An interface proposal for subjunctive obviation .................................... 107
6.1 Introduction: desiderata for a theory of obviation............................ 107
6.2 The hypothesis.................................................................................. 108
6.2.1 Generalities.................................................................................... 108
6.2.2 Obviation weakening..................................................................... 110
6.2.2.1 Non-finite full verbs ................................................................... 110
6.2.2.2 Topic, focus, strong pronouns, dislocated clauses...................... 112
6.2.2.3 First-person pronouns................................................................. 113
6.2.2.4 Directive predicates.................................................................... 114
6.2.2.5 Double embedding ..................................................................... 114
6.2.2.6 Adverbial clauses ....................................................................... 115
6.2.2.7 Coordination and overlapping reference .................................... 118
6.3 Problems and tentative solutions...................................................... 119
6.3.1 Theta-role sensitivity..................................................................... 119
6.3.2 Implications concerning Control ................................................... 122
6.3.2.1 Unsatisfied positions .................................................................. 122
6.3.2.2 Overlapping reference and Partial Control................................. 125
6.4 Conclusions ...................................................................................... 126
References .................................................................................................. 127

Preface
This monograph is a completely revised version of my 2006 doctoral disser-
tation. Although many of the ideas presented here originate in the earlier
work, major changes have been made both with respect to the topic and to
the structure of the dissertation. In sections 3.3, 3.5, 3.6, 3.7, I discuss new
empirical data and in sections 6.2.2.2 and following I provide new analyses.
Many people have contributed to make this book as it is. First, I would
like to express my deepest gratitude to my thesis supervisor, Prof. Alessan-
dra Giorgi, who guided my studies during my years as a Ph.D. student. I
thank her for believing in me, dedicating me much of her time, for the in-
sightful discussions, for carefully reading my papers, for supporting and
helping me. I am grateful to Prof. Guglielmo Cinque, who has always given
me precious advice, and never failed to encourage and support me. I am in-
debted to Prof. James Higginbotham and Prof. Pier Marco Bertinetto for ac-
cepting to be part of my dissertation committee. I thank Prof. James
Higginbotham for the stimulating discussions and his encouragement, and to
Prof. Pier Marco Bertinetto for his thoughtful comments on my thesis. I
would like to express my gratitude to Prof. Tim Stowell for allowing me to
visit the UCLA Linguistics Department. Materials from this book were pre-
sented at IGG 30 (Rome), ICLL 13 (Brussel), IGG 34 (Padua), as well as at
the University of Venice, Padua and Geneva. I am thankful to the audiences
of these conferences and seminars.
I dedicate this book to Elisa and Vesselina.

1 Introduction
1.1 Preliminary remarks
In Italian and in the main Western Romance languages the subject of a su-
perordinate clause and the subject of a subordinate clause cannot refer to the
same individual.
(1) Gianni vuole che _ legga il libro.
Gianni wants that reads.SUBJ.PRES.3SG the book
‘Gianni wants him/her to read the book’.
Intuitively, in sentence (1), the null embedded subject cannot refer to Gianni.
Indicative and conditional embedded clauses behave differently. The ma-
trix subject and the null embedded subject can refer to the same individual if
the embedded verb is in the indicative or in the conditional:
(2) a. Gianni ha detto che _ leggerà il libro.
Gianni has said that reads.IND.FUT.3SG the book
‘Gianni said he will read the book’.
b. Gianni ha detto che _ avrebbe letto il libro.
Gianni has said that
AUX.COND.3SG read the book
‘Gianni said he would read the book’.
Finally, infinitival clauses behave differently both from the subjunctive
and from the indicative and conditional clauses, in that the matrix subject
and the implicit subject of the infinitive must refer to the same individual:
(3) Gianni vuole _ leggere il libro.
Gianni want read.INF.PRES il book
‘Gianni wants to read the book’.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
12
In the contemporary approach to the study of language named Generative
Grammar, the phenomenon illustrated in example (1) has been labeled ‘ob-
viation’ (or ‘subjunctive obviation’), or ‘subject disjoint reference effect’.
The two labels are equivalent. The former is inherited from the American
Indian linguistics. The French term ‘obviatif’, from which the English words
‘obviative’ and ‘obviation’ stem, was coined by the French linguist J. A.
Cuoq (Etudes Philosophiques sur Quelques Langues Sauvages de l’Amé-
rique, 1866) to refer to a grammatical category of the Algonquian languages.
In the Algonquian languages and in some other American Indian languages,
‘animate’ nouns, demonstratives, pronouns in the third person, and verbs can
be categorized as ‘proximate’ or ‘obviative’ (alternative terms are ‘fourth
person’ or ‘second third-personal form’). A constituent belonging to one of
these categories is marked morphologically as proximate in a sentence if it
refers to «the person from whose point of view events are described, the pro-
tagonist in narratives, the focus of the speaker’s empathy» or to the «dis-
course topic» (Mithun 1999). In the same sentence, all the others constitu-
ents belonging to the above categories are marked as obviative. The con-
stituents marked as obviative and the constituent marked as proximate can-
not refer to the very same individuals; they must be referentially disjoint.
In the theoretical framework of the Generative Grammar, the term ‘obvia-
tion’ was introduced by Chomsky (1981). In such a theoretical framework,
indexes have been used as a standard notational device to indicate the refer-
ence of a nominal phrase. Thus, two phrases having the same index refer to
the same entity, whereas two phrases having two different indexes refer to
different entities. Chomsky defines the notion of ‘proximity’ and ‘obviation’
as follows:
We will call a pronoun […] “proximate” when it is coindexed with an ante-
cedent and “obviative” if it is not. (Chomsky 1981: 61)
And again:
Pronouns are “proximate” if they are coindexed with some other element and
“obviative” if not coindexed with any other element (Chomsky 1981: 186)
In example (1), the implicit subject of the embedded clause is obligatorily
obviative with respect to the matrix subject. Adopting the referential in-
dexes, here represented by natural numbers, example (1) may be then repre-
sented as follows:
(4) Gianni1 vuole che _*1 legga il libro.
Gianni1 wants that _*1 reads.SUBJ.PRES.3SG the book
‘Gianni1 wants him*1/her to read the book’.

INTRODUCTION
13
The index preceded by an asterisk indicates that such an indexation gives
rise to ungrammaticality.
The second term – ‘subject disjoint reference effect’ – refers exactly to
the unavailability of coindexation in examples like (1). Formatives having
different indexes are indeed called ‘referentially disjoint’ and such are the
matrix and the embedded subjects in the example discussed.
Following the standard notation introduced by Chomsky (1982), the null
subject in pro-drop languages will be labeled as pro (‘small’ or little ‘pro’), a
phonetically unrealized pronoun that can be assigned Nominative Case. Sen-
tence (1) can accordingly be represented as follows:
(5) Gianni1 vuole che pro*1 legga il libro.
Gianni1 wants that pro*1 reads.SUBJ.PRES.3SG the book
‘Gianni1 wants him*1/her to read the book’.
1.2 Theoretical relevance of the phenomenon
The phenomenon of obviation raised interest in the mid-1980s as it appeared
to challenge the Binding Theory. The Binding Theory rules the distribution
of nominal constituents: full NPs (‘referential expressions’), pronouns, and
anaphors. As for pronouns, the Binding Theory Principle B states that a pro-
noun must be free in its binding domain, where ‘free’ means not bound, and
bound means coindexed with a c-commanding constituent. A clause may be
provisionally taken as a binding domain. In other words, a pronoun cannot
be coindexed with a c-commanding nominal expression occurring within its
clause.
The unavailability of coindexation between the matrix subject and pro in
sentences like (1) appeared surprising in the light of Binding Theory Princi-
ple B. Pro was indeed expected to be optionally coindexed, exactly as it is in
sentences like (2). The fact that it cannot be so, suggested the hypothesis that
the matrix subject and pro occur in the same binding domain, in spite of the
clause boundary existing between the two constituents.
Since the mood of the embedded verb distinguishes the sentence in (1)
from the sentences in (2), subjunctive mood was claimed to be responsible
for the interpretative properties of pro at once: indicative (and conditional)
clauses constitute ‘closed’ (‘opaque’) binding domains, subjunctive clauses
do not constitute closed domains, so that a binding domain for pro in (1) is
the entire sentence, and not only the embedded clause.
Theoretically, this is a step of great importance in at least two respects:
first, mood appears to interact with Binding Theory; second, despite Binding
Theory Principle B rules the distribution of pronominals in view of their in-
terpretation, it is a syntactic module, as it is standardly assumed. Subjunc-

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
14
tive clauses must then be syntactically different from indicative (and condi-
tional) clauses.
Theoretical linguistics within the generative approach of Government and
Binding Theory inaugurated by Chomsky (1981) and within the following
minimalist framework (Chomsky 1995) has struggled to cope with these
facts, which did not appear to be entirely syntactic or entirely semantic, but
seemed to mingle syntactic and semantic aspects. Thus, subjunctive obvia-
tion appeared to be a privileged field of investigation on what has been
dubbed as the syntax-semantics interface, the subcomponent of the computa-
tional system of language that maps syntax into semantics (Logical Form).
The theoretical interest of studying obviation has survived until today.
This is due to various circumstances. First, descriptively, only a core set of
data on obviation has been analyzed, while less obvious aspects of the phe-
nomenon have been generally left outside the scope of inquiry. Moreover,
theoretically, many aspects of the problem, even relevant ones, are not well
understood. Finally, an increasing number of proposals concerning subjunc-
tive mood from a syntactic, semantic, and an interface viewpoint, suggests
that the problem of obviation may be tackled taking full advantage of the
new ideas on subjunctive complementation.
This study is an attempt to fill at least some of these gaps.
1.3 Outline of the study
This book is organized in six chapters (including the present introduction). I
will present here a summary of the content of every chapter.
Chapters 2 and 3 are devoted to the empirical framework of the phe-
nomenon at issue. In chapter 2 I will discuss the core data concerning sub-
junctive obviation in Italian and in the other main Western Romance lan-
guages. Following Schlenker (2005), I will reformulate the notion of obvia-
tion getting rid of the referential indexes (which nonetheless will be kept as a
notational device) and defining obviation as the unavailability of the de se
interpretation (in the sense of Lewis 1979).
In chapter 3 I will develop a more fine-grained analysis of the phenome-
non in order to make an inventory of all the syntactic environments where
obviation appears to be affected. Such an analysis will take into account at
least the following variables:
a) matrix verb: i. it may belong to a class of predicates that select for sub-
junctive clauses (in Italian, volitional, desiderative, epistemic, emotive-
factive, psych-verb, etc.) or to a class of predicates that do not subcategorize
for subjunctive clauses. In this case a polarity item, negation or interrogative
force, can instantiate the subjunctive verb in the embedded clause; ii. tense;
iii. person (agreeing with the matrix subject);

INTRODUCTION
15
b) matrix argument DP: i. it can be external or internal (depending on the
lexical class of the main verb); ii. person;
c) subjunctive clause: i. the subjunctive clause can be an argument of the
matrix predicate – external or internal (dependently on the lexical class of
the matrix verb); ii. the subjunctive clause can be an adjunct clause – ration-
ale clause, concessive clause, before-clause, etc.; ii. structural position: the
subjunctive clause can be in situ or left-dislocated;
d) subjunctive verb: i. type of verb: it can be a lexical or a modal predi-
cate; ii. type of subjunctive (dependent on the matrix verb): it can be inten-
sional (directly selected by the main predicate) or triggered by a polarity
item (negative or interrogative operator); iii. tense agreement (dependent on
matrix tense); iv. tense: it can be a simple or a perfect tense; iv. voice: active
or passive; v. person (dependent on the matrix subject);
f) embedded subject: i. theta-role assigned to it; ii. it can be phonetically
realized (as a strong pronoun) or null (pro) (in French, clitic); iii. person;
g) additional factors: topics within a subjunctive clause, coordination.
In chapter 3 I will investigate whether these variables affect the disjoined ref-
erence effect and I will reach the empirical generalization that in Western Ro-
mance languages pro (or the subject clitic in French) is obligatorily obviative
when the subjunctive morphology is attached to the verbal root. Examples of
such a type constitute the core set of data on obviation, the set of environments
in which obviation occurs consistently and without intralinguisic variability.
In the other cases, the embedded subject may be generally coindexed with
an argument of the matrix clause (with the caveat that intraliguistic variabil-
ity has been recorded as to the availability of coindexation).
Moreover, the phenomenon appears to be sensitive to other syntactic and
semantic properties: information structure factors (strong pronouns may be
coindexed with a c-commanding noun phrase; topicalized constituents
within the embedded clause makes it easier to get a coreferential interpreta-
tion; the dislocation of the subjunctive clause also appears to facilitate the
possibility of coindexation); the class of the matrix verb: volitional verbs
appear to be more rigid in not allowing coindexation than any other class
even in the environments that seemingly allow coindexation. The phenome-
non of subjunctive obviation is then not so black and white as it has often
been assumed. Thus, all things considered, a sentence like the following ap-
pears to be fully acceptable:
(6) Che domani pro1 riesca a superare l’esame, Gianni1 lo spera pro-
prio.
That tomorrow pro1 succeed.SUBJ.PRES.3SG to pass the exam,
Gianni1 it.CL hope really.
‘That he1 succeeds to pass the exam tomorrow, Gianni1 really
hopes so’.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
16
In sentence (6) obviation does not arise. The matrix and the embedded sub-
ject refer to the same individual, in spite of the subjunctive verb.
The literature on subjunctive obviation has focused mainly on the ‘core’
set of data, contrasted with the indicative and infinitival clauses. Some spo-
radic reference has also been made to examples involving a tense auxiliary
or a modal in the subjunctive, in which coindexation is allowed. This is by
and large the empirical base that the existing theories of obviation attempted
to explain. A theory on subjunctive obviation should also be able to predict
the cases in which obviation does not occur in spite of the presence of a sub-
junctive verb. Chapter 4 addresses the existing theories on obviation and il-
lustrates their pros and cons.
There are two main lines of investigation on subjunctive obviation. The
first, which has been supported in most of the literature on obviation, claims
that the subjunctive disjoint reference effect is a consequence of the interac-
tion between the interpretative properties of the subjunctive mood and the
Binding Theory. According to Picallo’s (1985) seminal theory, for instance,
the binding domain of the subject pronoun in a subjunctive clause is sup-
posed to stretch over the matrix clause, due to the ‘anaphoric’ nature of sub-
junctive tense. Binding Theory Principle B rules out coindexation between
the matrix subject and the embedded subject pro. Other theories based on the
Binding Theory (Avrutin 1994, Avrutin and Babyonyshev 1997, Kempchin-
sky 1987, 1997, Manzini 2000, Progovac 1993, 1994, Rizzi 2000, Tsoulas
1996), though involving different linguistic devices are able to explain the
same set of data – basically the ‘core’ set.
The second line of investigation is based on the observation that in West-
ern Romance languages, subjunctive clauses and infinitival clauses are com-
plementary in that subjunctive clauses do not allow coindexation between
the matrix and the embedded subject, while infinitival clauses obligatorily
require it. Bouchard (1984) proposes that such a contrast results from Chom-
sky’s (1981) Elsewhere Principle, which states that a pronoun is unavailable
for coindexation with a c-commanding constituent, if an anaphor is avail-
able. Proposals in the same spirit, yet exploiting different principles of mood
‘competition’, have been worked out by Farkas (1992) and by Schlenker
(2005).
Both Binding Theory-based and ‘competition’ approaches are able to ac-
count for the ‘core’ set of data. A desideratum for a theory of obviation, is
however that it should be able to account for the ‘core’ set of data and for
the ‘peripheral’ data. Chapter 5 is intended to prepare the ground for such a
theory.
To start with, the interface properties of subjunctive clauses will be illus-
trated within the theoretical framework developed by Giorgi and Pianesi
(2001). They maintain that subjunctive ‘anaphrocity’ can be reformulated
stating that the temporal coordinates of the attitude episode (the matrix even-

INTRODUCTION
17
tuality) are obligatorily represented in an embedded clause, whereas the ut-
terance time, ‘now’, is syntactically represented only in the environments in
which a Double Access Reading (‘DAR’, for short) obtains. Embedded
clauses having no utterance coordinates are temporally interpreted only with
respect to the matrix eventuality, whereas embedded clauses containing the
utterance coordinates are temporally interpreted both with respect to the ma-
trix eventuality and with respect to the utterance time – they display a DAR.
Only the former clauses are temporally ‘anaphoric’. Subjunctive clauses be-
long to this type of clause. Indicative clauses, on the other hand, display
DAR. Thus, they must contain both the matrix eventuality coordinates and
the utterance time.
Giorgi and Pianesi claim that the absence of the utterance coordinates is
responsible for a bunch of properties that distinguish subjunctive from in-
dicative embedded clauses: Sequence of Tenses (SOT for short), Comple-
mentizer Deletion (CD), Long-Distance Anaphor (LDA) Binding. I will pur-
sue the hypothesis that subjunctive obviation also follows from the ‘ana-
phoric’ nature of the subjunctive mood as understood by Giorgi and Pianesi.
A theory of obviation in the spirit of Giorgi and Pianesi is worked out in
chapter 6. Following Giorgi’s (2004, 2006, 2007) hypothesis (discussed in
chapter 5) that in an environment devoid of the utterance coordinates, a de se
reading can only be achieved via theta-identification of an unsatisfied posi-
tion (that is, an unassigned theta-role), spelled-out as a LDA, with the agent
of the context, I will attempt to account in the very same way both for exam-
ples exhibiting obviation and for subjunctive clauses that do not exhibit ob-
viation. The hypothesis will be investigated that in the latter, but not in the
former environments, an unsatisfied position occurs, possibly associated
with the non-finite form of the lexical verb, instantiating the de se reading. I
will discuss the theory framed on such hypothesis vis-à-vis the empirical
framework discussed in chapters 2 and 3 and I will show that it appears to be
empirically adequate, opening promising perspective on our understanding
of the interactions between thematic roles and pragmatic roles.

2 Subjunctive obviation: preliminary remarks
2.1 Introduction
In the present chapter I will discuss the basic properties of subjunctive ob-
viation. I will show how the subject disjoint reference effect is sensitive to
the mood of an embedded clause in Italian and in the other main Western
Romance languages (Catalan, French, Portuguese, and Spanish), as it obtains
in subjunctive clauses, though not in indicative, conditional, and infinitival
clauses.
In the last section, I will reformulate the notion of obviation as a condi-
tion of unavailability of the de se reading.
2.2 Obviation and mood in Italian
In Italian, the pronominal subject of a complement clause in the subjunctive
cannot corefer with the matrix subject (it is obviative):
(1) Gianni1 vuole che pro*1/2 legga il libro.
Gianni1 wants that pro*1/2 read.SUBJ.PRES.3SG the book.
‘Gianni1 wants him*1/2/her to read the book’.
Obviation is sensitive to the mood of the embedded clause. The possibil-
ity of coindexation between the embedded and the matrix subject distin-
guishes subjunctive clauses, which do not allow for coindexation, from in-
dicative clauses:
(2) Gianni1 ha detto che pro1/2 leggerà il libro.
Gianni1 has said that pro1/2 read.IND.FUT.3SG the book
‘Gianni1 said he1/2/her will read the book’.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
20
Moreover, coindexation is allowed when the embedded verb is in the
conditional mood:
(3) Gianni1 aveva detto che pro1/2 sarebbe partito il giorno dopo.
Gianni1 had said that pro1/2 AUX.COND.3SG left on the following
day
‘Gianni1 said he1/2/she would have left on the following day’.
Finally, coindexation is obligatory when the embedded verb is an obliga-
tory control infinitive:
(4) Gianni1 vuole PRO1/*2 leggere il libro.
Gianni1 wants PRO1/*2 to read the book.
‘Gianni wants to read the book’.
In Italian, obviation occurs only within subjunctive clauses, no matter
what the typology of the subjunctive clause is. Subjunctive clauses have of-
ten been distinguished into two categories, called ‘intensional’ and ‘polarity’
subjunctive (Stowell 1993, Quer 1997), depending on whether the subjunc-
tive verb is triggered by an intensional verb, or by some operator, like the
negative operator, the interrogative operator, or by left dislocation.
As for the first class, the predicate selecting for a subjunctive argument
can be volitional, epistemic (or dubitative), or emotive-factive (or thematic)1.
No difference in coindexation occurs dependently on the matrix predicate:
(5) a. Gianni1 pensa che pro*1/2 legga molti libri.
Gianni
1 thinks that pro*1/2 read.SUBJ.PRES.3SG many books
‘Gianni1 thinks that he*1/2/she reads many books’.
b. Gianni1 teme che pro*1/2 faccia molti errori.
Gianni1 fears that pro*1/2 make.SUBJ.PRES.3SG many mistakes
‘Gianni1 is afraid he*1/2/she will make many mistakes’.
c. Gianni1 si rammarica che pro*1/2 legga pochi libri.
Gianni1 regrets that pro*1/2 read.SUBJ.PRES.3SG few books
‘Gianni1 regrets that he*1/2/she reads few books’.
To the above class of predicates, one may add interrogative predicates. Indi-
rect questions also trigger obviation:
1 I follow here Wandruszka’s (1991) classification.

SUBJUNCTIVE OBVIATION: PRELIMINARY REMARKS
21
(6) Gianni1 si chiede quale libro pro*1/2 legga.
Gianni1 wonders which book pro*1/2 read.SUBJ.PRES.3SG
‘Gianni1 wonders which books he*1/2/she is reading’.
When the subjunctive is instantiated by the matrix predicate, obviation ap-
pear to be strictly linked to the subjunctive mood. The fact that when a verb
can select either a subjunctive complement clause or an indicative argument,
obviation occurs only within the subjunctive complement clause, enforces
the conclusion that descriptively, obviation depends on the embedded mood.
(7) a. Gianni1 pensa che pro*1/2 legga il libro presto.
Gianni1 thinks that pro*1/2 read.SUBJ.PRES.3SG the book soon
‘Gianni1 thinks he*1/2/she will read the book soon’.
b. Gianni1 pensa che pro1/2 leggerà il libro.
Gianni1 thinks that pro1/2 read.IND.FUT.3SG the book
‘Gianni thinks that he1/2/she will read the book’.
As for polarity subjunctive, although it has sometimes been claimed that
it does not instantiate obviation (Kempchinsky 1987, Quer 1997), evidence
from Italian appears to support the opposite claim. Take for instance the verb
sapere ‘to know’. This verb generally selects for indicative or conditional
clausal arguments:
(8) Gianni sapeva che Maria era/sarebbe/*fosse partita.
Gianni knew that Maria AUX.IND/COND/*SUBJ.PRES.3SG partita.
‘Gianni knew that Maria had left/would leave’.
However, if the embedded clause occurs under the scope of an operator or if
it is left dislocated, it may take a subjunctive argument clause. In this case,
the embedded subject is obviative.
(9) a. Gianni1 non sapeva dove pro*1/2 andasse.
Gianni1 not knew where pro*1/2 go.SUBJ.IMPF.3SG
‘Gianni1 didn’t knew where he*1/2/she was going’.
b. *Dici che tu parta?
Say(2ps) that you leave.SUBJ?
c. Che pro*1/2 partisse il giorno dopo, Gianni1 lo sapeva già.
That pro*1/2 leave.SUBJ.IMPF.3SG the following day, Gianni1 it
knew already
‘Gianni1 already knew that he*1/2/she was leaving on the follow-
ing day’.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
22
In sentence (9)a the subjunctive is triggered by the negation, in (9)b by the
interrogative force, in (9)c by movement to the left periphery of the embed-
ded clause. All three sentences display obviation.
2.3 Obviation comparatively
2.3.1 Catalan
Obviation occurs in subjunctive, but not in indicative clauses (examples
from Picallo 1985):
(10) a. [En Jordi]1 espera que pro*1/2 vingui.
[The Jordi]
1 hopes that pro*1/2 come.SUBJ.PRES.3SG
‘Jordi1 hopes that he*1/2/she will come’.
b. [En Joan]1 ha decidit que pro1/2 telefonarà al Pere.
[The Joan]1 has decided that pro1/2 call.IND.FUT.3SG to-the Pere
‘Joan has decided that he1/2/she will call Pere’.
The co-indexation between the matrix and the embedded subject is available
when the embedded clause is in the indicative, whereas it is not when the
embedded clause is in the subjunctive.
Quer (1997) observes that in Catalan the intensional subjunctive triggers
obviation, whereas the polarity subjunctive does not:
(11) Pro1sg no crec que pro1sg la convidi.
Pro1sg not think(1sg) that pro1sg her invite.SUBJ.PRES.3SG
‘I don’t think I will invite her’.
However, he also observes that there are some cases in which polarity sub-
junctive does trigger obviation2:
(12) *Creus que la convidis?
Believe.IND.PRES.2SG that her invite SUBJ.PRES.2SG?
Lexically selected subjunctive clauses also contrast with infinitival clauses,
in which PRO is controlled by the matrix subject and accordingly corefers
with it:
2 See Quer (1997), p. 38, n. 25.

SUBJUNCTIVE OBVIATION: PRELIMINARY REMARKS
23
(13) [En Jordi]1 espera PRO1 venir. (Picallo 1985)
[The Jordi]
1 hopes PRO1 come(inf).
‘Jordi hopes to come’.
2.3.2 French
In French obviation is also sensitive to mood. The subjunctive mood triggers
obviation, whereas the indicative does not.
(14) a. Pierre1 veut qu’il*1/2 parte. (Farkas 1992)
Pierre1 wants that he*1/2 leave.SUBJ.PRES.3SG
‘Pierre1 wants him*1/2 to leave’.
b. Pierre1 a promis qu’il1/2 partira. (Farkas 1992)
Pierre1 has promised that he1/2 leave.IND.FUT.3SG
‘Pierre1 promised that he1/2 to leave’.
In infinitival clauses the embedded subject corefers obligatorily with the ma-
trix subject:
(15) Pierre1 veut PRO1 partir.
Pierre1 wants PRO1 to leave
‘Pierre wants to leave’.
2.3.3 Portuguese
In Portuguese obviation depends on the mood of the embedded predicate, in
the same way as in Italian, Catalan and French (examples from Raposo
1985):
(16) a. [O Manel]1 deseja que pro*1/2 leia mais livros.
[The Manel]1 wishes that pro*1/2 read.SUBJ.PRES.3SG more
books
‘Manel1 wishes that he*1/2/she reads more books’.
b. [O Manel]1 pensa que pro1/2 lê bastanetes livros.
[The Manel]1 thinks that pro1/2 read.IND.PRES.3SG enough books
‘Manel1 thinks that he1/2/she reads enough books’.
c. [O Manel]1 deseja PRO1 ganhar a corrida.
[The Manel]1 wishes PRO1 to win the race
‘Manel wishes to win the race’.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
24
2.3.4 Spanish
Kempchinsky (1987) claims that in Spanish not all subjunctive argument
clauses instantiate the disjoint reference effect. Only the volitional and de-
siderative subjunctives do, whereas the (negative) epistemic subjunctives
and the emotive-factive subjunctives do not:
(17) a. Paco1 quiere que pro*1/2 estudie latín.
Paco
1 wants that pro*1/2 studie.SUBJ.PRES.3SG Latin.
‘Paco
1 wants him*1/2/her to study Latin’.
b. Ana1 duda que pro1/2 sea la persona más apta para el puesto.
Ana1 doubts that pro1/2 be.SUBJ.PRES.3SG the person most fit for
the job
‘Ana1 doubts that she1/2/he is the fittest person for the job’.
c. Ana
1 no creía que pro1/2 fuera la major candidata.
Ana1 not believed that pro1/2 be.SUBJ.IMPF.3SG the best candi-
date
‘Ana didn’t think she1/2/he was the best candidate’.
d. Ana1 lamenta que pro1/2 tenga tanto trabajo.
Ana1 regrets that pro1/2 have.SUBJ.PRES.3SG so-much work
‘Ana regrets that she1/2/he has so much work’.
Luján (1999) shows however that examples involving an emotive-factive
subjunctive do determine obviation:
(18) *(Tú) lamentas que (tú) vengas.
(You) regrets that (you) come.SUBJ.PRES.2SG
Hence, it seems that the claim that in Spanish only the volitional subjunc-
tive triggers obviation must be reconsidered. We will see in chapter 3 that
Kempchinsky’s examples involve some of the factors that weaken obviation,
which may explain the availability of coindexation.
In infinitival clauses, finally, the co-referential reading is not only possi-
ble, but obligatory (example from Kampchinsky 1987):
(19) Paco1 quiere PRO1 estudiar latín.
Paco1 wants PRO1 study.INF.PRES Latin.
‘Paco wants to study Latin’.

SUBJUNCTIVE OBVIATION: PRELIMINARY REMARKS
25
2.3.5 Summary
From the above data concerning the Western Romance languages, it seems
that cross-linguistically, there is a strict relation between the disjoint refer-
ence effect and the subjunctive mood. Since the class of predicates that se-
lects for a subjunctive clause most uniformly across Romance languages is
the class of volitional and desiderative predicates, obviation occurs mainly in
volitional contexts. However, in Italian the entire class of intensional sub-
junctive, including volitional and desiderative, epistemic and emotive-factive
subjunctive, trigger obviation. This seems to be true of the other Romance
languages as well – verbs lexically selecting for subjunctive clauses trigger
obviation. As for polarity subjunctive, it appears that it does instantiate ob-
viation as well, though not as rigidly and as uniformly as ‘intensional’ sub-
junctive does.
2.3.6 An aside on Latin
In Latin, obviation seems to occur in a limited set of contexts, mainly in sub-
junctive clauses selected by volitional and desiderative verbs (‘verba volun-
tatis’).
In Catalan, French, Italian, Portuguese, and Spanish, volitional verbs can
select for two types of clausal arguments: the infinitival clause, and the sub-
junctive clause. In the second type of clause the obviative interpretation of
the subject generally obtains.
Latin verba voluntatis (volo, ‘I want’, nolo, ‘I do not want’, malo, ‘I pre-
fer’, cupio, ‘I desire’, etc.) can select three types of clausal complements: the
infinitival clause (also called ‘infinitival complement’), the structure called
‘accusativus cum infinitivo’ (AcI), and the subjunctive clause3:
(20) a. Volo esse liber. (Plaut. Trin. 440)
Want.
IND.PRES.1SG be.INF free
‘I want to be free’.
3 A subjunctive clause occurs more often when the matrix verb is in the optative subjunctive
(vellem, ‘I would like’, nollem, ‘I would not like’, mallem, ‘I had rather’, ecc.). This embed-
ded clause may be introduced by the subjunctive complementizer ut, but more frequently it is
not. Compare, for instance, sentence (20)c, where the argument clause is introduced by the
subordinating conjunction, with the following example, where the complementizer is omitted:
(i) tu vellem ego vel cuperem adesses (Cic. Att. 2,18,4)
You.NOM want.SUBJ.IMPF.1SG I.NOM or desire.SUBJ.IMPF.1SG come.SUBJ.IMPF.2SG
‘I wish – I desire you to come here’.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
26
b. Senatus te voluit mihi nummos, me tibi frumentum dare. (Cic.
II Verr. 3,197)
Senate you wanted me.DAT money, me.ACC you.DAT wheat
give.INF
‘The senate wanted you to give me money, and me to give you
the wheat’.
c. Volo ut mihi respondeas tu. (Cic. Vatin. 14)
Want.IND.PRES.1SG that me.DAT answer.SUBJ.PRES.2SG
you.NOM
‘I want you to answer me!’.
In the ‘infinitival complement’ the understood subject of the embedded
clause is obligatorily interpreted as coreferential to the matrix subject. In the
‘infinitival proposition’ the subject optionally corefers with the matrix sub-
ject. Finally, in a subjunctive clause the embedded subject is generally ob-
viative with respect to the matrix subject4.
It seems that the volitional subjunctive is the only type of subjunctive
triggering obviation. The epistemic subjunctive do not trigger obviation:
(21) Metuo ne sero veniam (Plaut. Men. 989)
‘I am afraid I will come late’.
Indirect questions do not trigger obviation, as well:
(22) Homo quid ageret, taceret responderet, nesciebat (Cic. Verr. II, 3,
62)
‘That poor man]1 didn’t know what he1/2/she had to do, whether
he1/2/she had to keep quiet or to answer’.
4 A limited series of exceptions will be considered below. Notice that both in AcI structures
and in the subjunctive clause, the embedded subject can be obviative with respect to the ma-
trix subject. It is not clear, however, what the exact difference is between the two type of con-
structions. Moreover, the alternance between AcI and subjunctive was remarked by the Latin
grammarians Priscianus and Macrobius: «sciendum tamen, quod in hoc sensu frequenter in-
venimus pro infinito verbo subiunctivum poni, ut ‘iubeo facias; iubeo dicas; impero venias;
hortor legas’, in quibus deest ‘ut’, quod licet vel addere vel non» (Priscianus, Institutionum
grammaticarum libri, 18, 45 – «it must be known, however, that in this sense we often find
that the subjunctive mood can be put in spite of the infinitive, as ‘iubeo facias; iubeo dicas;
impero venias; hortor legas’, where ‘ut’, which can be adjoined or not, is omitted»). «[Latini]
aliquotiens [infinitum] pro coniunctivo ponunt: Cicero pro Sestio ‘rei publicae dignitas me ad
se rapit et haec minora relinquere hortatur’ pro ‘hortatur ut relinquam’, Vergilius ‘hortor
amare focos’ pro ‘hortor ut ament’» (Macrobius, Excerpta de differentiis et societatibus greci
latinique verbi, 164 – «Latins sometimes put the infinitive in spite of the subjunctive: Cicero
pro Sestio ‘rei publicae dignitas me ad se rapit et haec minora delinquere hortatur’ in spite of
‘hortatur ut relinquam’, Vergilius ‘hortor amare focos’ in spite of ‘hortor ut ament’»).

SUBJUNCTIVE OBVIATION: PRELIMINARY REMARKS
27
Furthermore, emotive-factive predicates generally do not trigger obvia-
tion:
(23) [Haec urbs] mihi laetari videtur, quod tantam pestem evomuerit
(Cic. Cat. 2, 2)
‘It seems to me that [this town]1 rejoices that he/she/it1/2 rejected
such a great disaster’.
Thus, again, the class of volitional predicates appears to be the only one
that requires disjointedness rigidly.
2.4 Obviation and de se
2.4.1 Castañeda’s examples
We have seen that obviation can be defined as a disjointedness condition
between two arguments. This definition can be reformulated in terms of un-
availability of ‘first-person’ interpretation, a notion originally discussed by
Castañeda (1966, 1968), or ‘de se’ interpretation, following Lewis’s (1979)
terminology.
A de se attitude may be defined as a subject’s attitude that is consciously
directed toward a proposition involving the subject herself. To illustrate,
consider the following scenario (see Castañeda 1966, Higginbotham 1992).
A certain war hero suffers from amnesia and does not remember any of his
deeds in wartime. Suppose that this unfortunate person reads a book about
the war he fought in. He also reads about a soldier’s heroic exploits. The
soldier the unfortunate is reading about, is actually himself, but he is amne-
siac and does not realize that it is himself the war hero. He finally comes to
have some belief about the soldier he is reading about and may say: ‘That
soldier is a hero’ – without realizing that the soldier he is talking about is
himself.
In such a scenario, sentence (24)a can be considered true, whereas sen-
tences (24)b and (24)c cannot:
(24) a. The unfortunate man believes he is a hero.
b. The unfortunate man believes that he himself is a hero.
c. The unfortunate man believes himself to be a hero.
Sentences (24)b and (24)c are infelicitous in the above scenario because the
anaphor himself introduces the presupposition that the attitude expressed by
the matrix verb is de se – that is, it is a conscious attitude about the author of
the attitude himself.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
28
In a different scenario, say, one in which the war hero is not amnesiac (he
is quite aware about who he is and about his own past experience in the war-
time, and he may say ‘I am a war hero’), sentences (24)b and (24)c would be
felicitous. Sentence (24)a might also be appropriate. In such a scenario all
three sentences might express that the unfortunate is consciously speaking
about himself (de se).
Higginbotham (1992) notices that the null subject of infinitival clauses,
PRO, typically introduces a de se interpretation. Thus, imagine that in the
above scenario, while still reading about the deeds of the soldier, the unfor-
tunate comes to get some expectations and says: ‘This soldier will turn out to
be a war hero’. We may report the unfortunate’s sentence through (25)a, but
not though (25)b or c:
(25) a. The unfortunate expects he will turn out to be a war hero.
b. The unfortunate expects PRO to turn out to be a war hero.
c. The unfortunate expects himself to turn out to be a war hero.
Sentence (25)a can express a non-de se attitude, thus being compatible with
the described scenario, whereas sentence (25)b and c expresses a de se atti-
tude, and are infelicitous under the described circumstances. Thus, sentence
(25)a can be considered true in the given scenario. Sentences (25)b and (7)c,
cannot.
Moreover, sentence (25)a is ambiguous between the de se interpretation
and the non-de se interpretation, whereas sentences (25)b and c are unambi-
guously de se.
2.4.2 Subjunctive clauses and the de se reading
Schlenker (2005) was the first, as far as I know, who defined obviation in
terms of unavailability of de se reading. He discusses some examples report-
ing an attitude involving subjunctive clauses and claims that a sentence
having a subjunctive clausal complement may be felicitous only when it
expresses a non-de se attitude:
(26) George1 voudrait qu'il1 soit élu.
George1 want.COND.PRES.3SG that he1 AUX.SUBJ.PRES.3SG elected
‘George1 wishes that he1 is elected’.
Sentence (26) is appropriate in a scenario in which George expresses the
wish ‘This candidate should be elected’, though not if George’s wish has the
form ‘I wish to be elected’.
Let us consider the Italian translation of Schlenker’s example:

SUBJUNCTIVE OBVIATION: PRELIMINARY REMARKS
29
(27) Gianni1 vorrebbe che pro1 sia eletto.
Gianni1 want.COND.PRES.3SG that he1 AUX.SUBJ.PRES.3SG elected
‘Gianni1 wishes that he1 is elected’.
Sentence (27) is interpreted as non-de se, thus being appropriate in a sce-
nario in which Gianni has a wish of the form ‘This candidate should be
elected’ – where the candidate is in fact Gianni himself, though he does not
realize it. However, a de se interpretation for sentence (27) is not completely
ruled out. To show this, consider the following sentences:
(28) a. Maria spera che Gianni sia eletto.
Maria hopes that Gianni is.SUBJ.PRES.3SG elected
‘Maria hopes that Gianni will be elected’.
b. E anche Gianni spera che sia eletto.
And also Gianni hopes that is.SUBJ.PRES.3SG elected
‘And Gianni hopes that he is elected too’
Intuitively, sentence (28)b can be an appropriate continuation of sentence
(28)a. Thus, although the preferred reading for (28)b is non-de se, especially
if assessed out of the blue, the de se reading does not seem to be completely
excluded and sentence (28)b appears to be ambiguous between the two read-
ings.
As it will be shown in chapter 3 sentence (27) (and (28)b) shows peculiar
properties. There are indeed some environments that enforce obviation rig-
idly, and environments that do not. Subjunctive clauses having a voice auxil-
iary belong to the latter class. Environments triggering strict obviation, ap-
pear to behave differently as for the de se interpretation.
Let us imagine a scenario in which the unfortunate reads on the newspa-
per that a certain war hero, who is he himself, will be soon awarded a medal
by the president in person. Thus, he may come to think: ‘This war hero will
meet the president’. The unfortunate’s wife knows that her husband will
soon meet the president. She also knows that the meeting will probably take
place, say, in May. She may say ‘I think my husband will meet the president
in May’. Then we may report the unfortunate’s wife’s attitudes as follows:
(29) Lo moglie dello sfortunato crede che suo marito incontri il presi-
dente a maggio.
The unfortunate’s wife thinks that her husband meet.SUBJ.PRES.3SG
the president in May
‘The unfortunate’s wife thinks her husband will meet the president
in May’.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
30
A possible continuation for sentence (29) may be (30):
(30) E anche lo sfortunato crede che incontri il presidente in maggio.
And also the unfortunate thinks that meet.SUBJ.PRES.3SG the presi-
dent in May
‘And the unfortunate also thinks he will meet the president in
May’.
Sentence (30) is clearly non-de se and appear to be appropriate in the de-
scribed scenario. Imagine however that the unfortunate recovers from his
amnesia. He becomes aware that the war hero is in fact himself and that he
will meet the president, probably in May. Then sentence (30) would not be
an appropriate continuation for sentence (29) anymore and an indicative
would be appropriate instead5:
(31) E anche lo sfortunato crede che incontrerà il presidente in maggio.
And also the unfortunate thinks that meet.IND.FUT.3SG the presi-
dent in May
‘And the unfortunate also thinks he will meet the president in
May’.
Thus, sentence (30) cannot be de se, contrasting with the subjunctive
clauses in (27)a and (28)b, which in an appropriate context can be de se. The
subjunctive clause in (30) contrasts also with the indicative clause in (31)
which can be de se, but it can also be non-de se. Thus, the difference be-
tween the two sentences is that the subjunctive clause in (30) is unambigu-
ously non-de se, whereas the indicative clause is ambiguous between the de
se and the non-de se reading.
2.5 Conclusion
In the present chapter I have provided a general characterization of the phe-
nomenon at issue. I have shown that in Italian obviation between the matrix
and the embedded subject obtains in subjunctive, but not in indicative, con-
ditional and in infinitival clauses, no matter what the type of predicate trig-
gering the subjunctive is involved – will, wish, epistemic, emotive-factive,
question verbs all instantiate obviation. Also, both intensional and polarity
subjunctive appear to instantiate obviation.
5 An infinitival clause would also be odd, but we assume this is irrelevant for the present dis-
cussion. We assume that infinitival complements of verbs like credere ‘believe’, pensare
‘think, believe’ are incompatible with eventive predicates, which rules out an alternative in
the infinitive for sentence (30)

SUBJUNCTIVE OBVIATION: PRELIMINARY REMARKS
31
In the other main Western Romance languages – Catalan, French, Portu-
guese, Spanish – the domain of occurrence of obviation appears to overlap
with the one in Italian: obviation obtains in subjunctive clauses, although
some examples from Romance languages apparently challenge the descrip-
tive generalization that polarity subjunctive instantiate obviation. We will
see in the following chapter that such examples involve in fact factors that
appear to ‘weaken’ obviation even in Italian.
Finally, following Schlenker (2005), I have provided a semantic charac-
terization of obviation in terms of unavailability of the de se interpretation.

3 Environments weakening subjunctive obviation
3.1 Chapter outline
Common to all the existing analyses on obviation is the observation that ob-
viation holds between the matrix subject and the subject of subjunctive ar-
gument clauses. The present chapter places under scrutiny this descriptive
generalization. Remarkably, it will turn out to be incomplete.
First, obviation does not arise in all the subjunctive clauses. Ruwet
(1984) observes that in French, the subject of some subjunctive clauses is not
obligatorily disjoint in reference from the matrix subject. His examples have
their counterpart in the other Western Romance languages. Second, Manzini
(2000) observes that the obviative interpretation can occur even in adverbial
clauses – not only in argument clauses. Finally, the subject of an embedded
clause can be obviative even with respect to a matrix direct or indirect ob-
ject.
In order to reach an adequate description of the phenomenon at issue I
will first consider some factors concerning the verbal form carrying subjunc-
tive morphology, which appears to affect the interpretative properties of the
embedded subject: subjunctive tense, passivization, and modalization. Then,
I will analyze the question which arguments and which grammatical proper-
ties of an argument are involved in obviation both in the embedded and in
the matrix clause. I will analyze obviation with respect to double embedding,
adverbial clauses, and peripheral factors such as left dislocation and clause
dislocation. Finally, I will illustrate some data concerning coordination.
This analysis will allow us to formulate a novel descriptive generalization
that states that obviation concerns the embedded subject and the matrix ar-
gument referring to the individual to whom a propositional attitude is attrib-
uted. Following the consuetude in the philosophy of language and semantics,
this argument may be referred to as ‘bearer of the attitude’, ‘context agent’,
or ‘context author’. Such an argument may (it does not need to) be the matrix

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
34
subject. It may also be a direct or an indirect object. This generalization is
just rephrasing the definition of obviation as the unavailability of de se read-
ing, which has been discussed at the end of the previous chapter, since de se
attitudes obviously concern the bearer of an attitude.
A second descriptive statement that will be achieved is that obviation oc-
curs if the verbal form inflected in the subjunctive is a full verb. The dis-
jointedness effect appears to be somehow ‘weakened’ if the verbal form car-
rying subjunctive morphology is an auxiliary or a modal.
Finally, a series of factor appear to affect the interpretation of the embed-
ded subject: the theta-role it discharges, number, its realization as a strong or
weak/clitic (or null) pronoun, the presence of left-dislocated constituent and
clause dislocation. All these factors appear to facilitate the de se reading,
showing a more complex framework.
3.2 Obviation and subjunctive mood
3.2.1 Obviation and subjunctive ‘tenses’
3.2.1.1 Obviation ‘weakening’ and periphrastic tenses
Ruwet (1984) and Raposo (1985) observe that obviation is sensitive to the
“tense” of a subjunctive verb. That is, coindexation appears to be marginally
available if the subjunctive tense is periphrastic.
In Italian the subjunctive mood has four ‘tenses’: presente ‘present sim-
ple’, passato ‘present perfect’, imperfetto ‘imperfect’ and trapassato ‘plu-
perfect’. The distribution of these forms in subordinate contexts depends on
the rules of Sequence of Tenses (SoT). In an environment of primary subor-
dination, SoT depends on the mood and the time of the matrix predicate and
to the time relation between the matrix event and the embedded event:
- simultaneity (the matrix and the embedded eventuality overlap tempo-
rally);
- anteriority (the embedded eventuality precedes the matrix eventuality);
- posteriority (the embedded eventuality follows the matrix eventuality).
The following (simplified) paradigm illustrates the SoT system in Ital-
ian6:
6 The scheme does not include indicative present perfect tense, which is compatible with the
subjunctive present (perfect) and with the subjunctive imperfect or pluperfect.

ENVIRONMENTS WEAKENING SUBJUNCTIVE OBVIATION
35
Matrix verb Subjunctive tense
Mood Tense Simutaneity Anteriority Posteriority
Present
Future
Present Present perf.
(Imperfect)7
Present
Indicative
Imperfect
Present Conditional
Past
Imperfect Pluperfect
(Imperfect)8
Imperfect
If the matrix verb is in the present, the embedded verb in the subjunctive
may be either in the present or in the present perfect. Coindexation between
the matrix and the embedded subject is not available in the former case. For
many Italian native speakers, if the embedded verb is a present perfect sub-
junctive, intuitively coindexation is marginally possible, although the obvia-
tive reading is the more natural option:
(1) a. Gianni1 pensa che pro%?1/2 abbia superato l’esame.
Gianni1 thinks that pro%?1/2 AUX.SUBJ.PRES.3SG passed the exam
‘Gianni1 thinks that he%?1/2/she has passed the exam’.
b. Gianni1 teme che pro%?1/2 non abbia superato l’esame
Gianni
1 fears that pro%?1/2 not AUX.SUBJ.PRES.3SG passed the exam
‘Gianni1 is afraid that he%?1/2/she has not passed the exam’.
c. Gianni1 si rammarica che pro?1/2 non abbia superato l’esame.
Gianni1 regrets that pro%?1/2 not AUX.SUBJ.PRES.3SG passed the
exam
‘Gianni1 regrets that he%?1/2/she has not passed the exam’.
Interestingly, even for the speakers who do not accept coindexation in
sentences like (1)a, there is a contrast between such sentences and sentences
like (1) in chapter 2.
If the matrix verb is in the past, the embedded verb in the subjunctive
may be imperfect or pluperfect. Coindexation is ruled out if the embedded
verb is imperfect subjunctive (exactly as it is when the verb is in the pre-
sent):
(2) a. Gianni1 voleva che pro*1/2 leggesse il libro.
Gianni
1 wanted that pro*1/2 read.SUBJ.IMPF.3SG the book.
‘Gianni1 wanted him*1/2/her to read the book’.
7 The imperfect subjunctive expressing anteriority can be the predicate of a clausal argument
of a verb in the present or in the imperfect. Such an option is restricted to some temporal and
aspectual properties that will be considered later.
8 See note 6.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
36
b. Gianni1 pensava che pro*1/2 leggesse molti libri.
Gianni1 thought that pro*1/2 read.SUBJ.IMPF.3SG many books
Gianni1 thought that he*1/2/she used to read many books.
c. Gianni1 temeva che pro*1/2 facesse molti errori.
Gianni1 feared that pro*1/2 made.SUBJ.IMPF.3SG many mistakes
‘Gianni1 was afraid he*1/2/she was making many mistakes’.
d. Gianni1 si rammaricava che pro*1/2 leggesse pochi libri.
Gianni1 regretted that pro*1/2 read.SUBJ.IMPF.3SG few books
‘Gianni1 regretted that he*1/2/she used to read few books’.
If the embedded subjunctive is pluperfect, however, according to many
Italian speakers, coindexation is marginally available, although the obviative
interpretation is preferred9:
(3) a. Gianni1 pensava che pro%?1/2 avesse letto molti libri.
Gianni1 thought that pro%?1/2 AUX.SUBJ.IMPF.3SG read many
books
Gianni1 thought that he%?1/2/she had read many books.
b. Gianni1 temeva che pro%?1/2 avesse fatto molti errori.
Gianni1 feared that pro%?1/2 AUX.SUBJ.IMPF.3SG made many
mistakes
‘Gianni1 was afraid he%?1/2/she he had made many mistakes’.
c. Gianni1 si rammaricava che pro%?1/2 avesse letto pochi libri.
Gianni1 regretted that pro%?1/2 AUX.SUBJ.IMPF.3SG read.INF
books
‘Gianni1 regretted that he%?1/2/she had read few books’.
The other Western Romance languages parallel the behavior of Italian:
(4) Catalan (Gemma Rigau, p.c.)
a. [En Joan]1 està sorprès que pro?1/2 hagi fet molts errors.
[The Joan]1 stays surprised that pro?1/2 AUX.SUBJ.PRES.3SG
made many mistakes.
‘Joan1 is surprised that he?1/2/she has made many mistakes’.
9 To some Italian native speakers, the sentences in (3) appear to be slightly more acceptable
than sentences in (1). Tense appears to matter here, as in some other cases that will be dis-
cussed.

ENVIRONMENTS WEAKENING SUBJUNCTIVE OBVIATION
37
b. [En Joan]1 estava sorprès que pro?1/2 hagués fet molts errors.
[The Joan]1 stayed surprised that pro?1/2 AUX.SUBJ.IMPF.3SG
made many mistakes.
‘Joan1 was surprised that he?1/2/she had made many mistakes’.
(5) French (Ruwet 1984)
?Je veux (absolument) que je sois parti dans dix minutes.
I want (absolutely) that I AUX.SUBJ.PRES.1SG left in ten minutes.
‘I want to be gone in ten minutes’.
(6) Portuguese (Raposo 1985)
a. [O António]1 receia que pro1/2 tenha bebido a água envenenada.
[The António]1 fears that pro1/2 AUX.SUBJ.PRES.3SG drunk the
water poisoned
‘António fears that he1/2/she has drunk poisoned water’.
b. [A Maria]1 preferia que pro1/2 não tivesse encontrado o Manel.
[The Maria]1 preferred that pro1/2 not AUX.SUBJ.IMPF.3SG met
the Manel
‘Maria1 preferred she1/2/he had not met Manel’.
(7) Spanish (María Martínez Atienza, p.c.)
Juan1 esperaba que pro1/2 hubiera cometido pocos errors (pero en
realidad había cometido muchos).
Juan1 hoped that pro1/2 AUX.SUBJ.IMPF.3SG made few mistakes (but
in fact had made many)
‘Juan1 hoped that he1/2/she had made few mistakes (but in fact
he1/2/she made many)’.
The above data from Italian, Catalan, French, Portuguese, and Spanish
seem to suggest that when the subjunctive verb is in a ‘simple’ tense,
coindexation is unavailable. When it is in a periphrastic tense, coindexation
tends to be possible. Thus, obviation does not simply occur in subjunctive
clauses, but only in the subjunctive clauses in which the verb is in a ‘simple’
tense.
3.2.1.2 Another aside on Latin
In Latin, although the volitional subjunctive does not allow for coindexation
between the matrix and the embedded subject, there are some cases in which

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
38
the coreference is nonetheless admitted. In one of these cases, the subjunc-
tive verb is pluperfect10.
(8) a. Vellem non costituissem me hodie venturum esse L. Aelio
(Cic. De or. 1, 265)
‘I had rather I did not decide to go to L. Aelius today’.
b. Hoc ipsum tacuissem mallem (Cic. Quinct. 30)
‘I had rather I had been silent about this same fact’.
c. Nollem dixissem (Cic. Verr. 4, 43)
‘I would like I did not say it’.
Thus, Romance languages may have developed a tendency whose initial
stage can be traced back to Latin.
3.2.2 Passive voice
As shown by Ruwet (1984), in French the de se reading is marginally possi-
ble when the embedded subjunctive verb is passive. Raposo (1985) shows
that Portuguese parallels French. Italian does so as well, although intralin-
guistic variability holds:
(9) a. Gianni1 spera che pro%??1/2 venga promosso.
Gianni
1 hopes that pro%??1/2 AUX.SUBJ.PRES.3SG promoted
‘Gianni1 hopes that he%??1/2/she will pass the exam’.
b. Gianni1 sperava che pro%?1/2 fosse promosso.
Gianni1 hoped that pro%?1/2 AUX.SUBJ.IMPF.3SG promoted
‘Gianni1 hoped that he%?1/2/she would pass the exam’.
Although the above examples are only marginally acceptable under the
coreferential interpretation, still they are considerably more acceptable than
the example (1) in chapter 2, which is completely ruled out under the same
reading.
10 The manuscript tradition is not uniform with regard to the following examples, and the
examples in which the subjunctive occurs are generally textual variants of forms in the infini-
tive. Example (a) is the only sentence accepted by the most prominent editions. As for exam-
ple (b), both the Les Belles Lettres and the Oxoniensis critic edition accept the textual variant
tacuisse instead of tacuissem. As for example (c), finally, the Oxoniensis edition accept
dixisse instead of dixissem.

ENVIRONMENTS WEAKENING SUBJUNCTIVE OBVIATION
39
Sentence (9)a is intuitively less acceptable than (9)b. This does not seem
to relate with the choice of the auxiliary, since the sentence Gianni spera che
sia promosso (‘Gianni hopes he/she is promoted’) does not seem to display
any appreciable difference in status with respect to sentence (9)a.
Coindexation is more natural if the embedded passive verb is in the past,
and it appears to be even more so if the verb is a pluperfect subjunctive:
(10) a. Gianni1 spera che pro%?1/2 sia stato promosso.
Gianni1 hopes that pro%?1/2 AUX.SUBJ.PRES.3SG been promoted
‘Gianni1 hopes that he%?1/2/she has passed the exam’.
b. Gianni1 sperava che pro(?)1/2 fosse stato promosso.
Gianni1 hoped that pro(?)1/2 AUX.SUBJ.IMPF.3SG been promoted
‘Gianni1 hoped that he(?)1/2/she had passed the exam’.
As mentioned at the beginning of the present section, this case of obvia-
tion ‘weakening’ has been observed in other Romance languages:
(11) a. Catalan (Picallo 1985)11
Que pro1sg sigui amenaçat de mort no m’impressiona.
That pro1sg AUX.SUBJ.PRES.1SG menaced of death not me.CL
impresses.
‘It does not impress me that I am menaced with death’.
b. French (Ruwet 1984)
?Je veux que je sois autorisé à partir demain.
I want that I AUX.SUBJ.PRES.1SG authorized to leave.INF tomor-
row.
‘I want to be allowed to leave tomorrow’.
c. Portuguese (Raposo 1985)
[O Manel]1 deseja que pro1/2 seja admitido no concurso.
[The Manel]1 wishes that pro1/2 AUX.SUBJ.PRES.3SG admitted in-
the contest.
‘Manel1 wishes he1/2she was admitted in the contest’.
d. Spanish (Quer 2005)
Pro1sg espero que pro1sg sea autorizado a ir.
Pro1sg hope that pro1sg AUX.SUBJ.PRES.1SG authorized to
go.INF.
‘I hope I will be allowed to leave’.
11 For obviation from the matrix object, see section 2.4.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
40
The data considered here further show that not all subjunctive clauses
trigger obviation. Only those in which the verb is in the active voice do,
whereas the passive voice allows for the coindexation.
Note that the data described in the previous section and in the present one
share the same property: the verbal inflected in the subjunctive is not the
lexical (or full) verb, but an auxiliary – the tense or the voice auxiliary. In
both cases the full verb is in a non-finite form – the past participle.
3.2.3 Modal verbs
3.2.3.1 Romance languages
Another environment in which obviation tends not to occur is the case when
the form carrying subjunctive morphology is a modal verb. Intuitively,
coindexation between the matrix and the embedded argument is not quite
ruled out in the following examples, at least for many Italian native speakers:
(12) a. Gianni1 spera che pro%?1/2 possa partire domani.
Gianni1 hopes that pro%?1/2 can.SUBJ.PRES.3SG leave.INF tomor-
row
‘Gianni1 hopes that he%?1/2/she will be able to leave tomorrow’.
b. Gianni1 sperava che pro%?1/2 potesse partire il giorno dopo.
Gianni1 hoped that pro%?1/2 can.SUBJ.IMPF.3SG leave.INF the day
after.
‘Gianni1 hoped he%?1/2/she would be able to leave on the follow-
ing day’.
The de se reading appear to be even more acceptable if the matrix predi-
cate is emotive-factive:
(13) a. Gianni1 si rammarica che pro%?1/2 debba partire domani.
Gianni1 regrets that pro%?1/2 must.SUBJ.PRES.3SG leave.INF to-
morrow
‘Gianni1 regrets that he%?1/2/she must leave tomorrow’.
b. Gianni1 si rammaricava che pro(?)1/2 dovesse partire il giorno
dopo.
Gianni1 regretted that pro(?)1/2 must.SUBJ.IMPF.3SG leave.INF the
day after
‘Gianni1 regrets that he(?)1/2/she had to leave on the following
day’.

ENVIRONMENTS WEAKENING SUBJUNCTIVE OBVIATION
41
The data from the other Western Romance languages are similar to those
from Italian:
(14) a. Catalan (Picallo 1985)
Pro1 sentien que pro1 deguessin produir una falsa impressió.
Pro1 regretted that pro1 must.SUBJ.PRES.3SG produce a false im-
pression
‘They regretted they had to produce a false impression’.
b. French (Ruwet 1984)
?Je veux que je puisse partir dès demain.
I want that I can.SUBJ.PRES.3SG to leave by tomorrow.
‘I want to be able to leave by tomorrow’.
c. Portuguese (Raposo 1985)
[O Manel]1 exige que pro1/2 possa ver o seu advogado.
[The Manel]1 requires that pro1/2 can.SUBJ.PRES.3SG see.INF the
his attorney
‘Manel1 requires that he1/2/she be able to see his/her attorney’.
d. Spanish (Quer 2005)
Pro1sg espero que pro1sg pueda ir.
Pro1sg hope that pro1sg can.SUBJ.PRES.3SG to go
‘I hope to be able to go’.
The data in the present section, as well as the data discussed in the pervi-
ous ones, suggest that obviation is sensitive to the nature of the verb in-
flected in the subjunctive. It appears that only full verbs trigger obviation
strictly, whereas auxiliaries and modals – functional verbs, in the sense of
Cinque (1999, 2004) allow (more or less marginally) the coindexation be-
tween the matrix and the embedded subject.
3.2.3.2 Again an aside on Latin
At least two examples in Latin show that the coreferential interpretation is
possible when the form carrying subjunctive morphology is a modal verb12:
(15) a. Vellem praesentem possem P. Helvidium Rufum nominare.
(Cic. Cluent. 198)
‘I would like to have mentioned P. Helvidius Rufus in front of
himself’.
12 In the main critical editions the modal verb in the subjunctive is accepted. Thus, these ex-
amples appear to be valid.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
42
b. Vellem equidem idem possem gloriari (Cic. Sen. 32)
‘I would like I praised myself about this’.
Thus, the phenomenon in Romance languages may be rooted in Latin.
3.2.4 Other periphrastic forms and restructuring verbs
The examples in the preceding sections show that coindexation appears to be
marginally available when the subjunctive verb is a functional verb and the
full verb is non-finite. I will now consider to what extent this seems to hold.
In Italian, the progressive auxiliary, stare (literally, ‘to stay’) appears to
marginally allow for the de se reading:
(16) a. Gianni1 teme che pro?1/2 stia facendo molti errori.
Gianni1 fears that pro?1/2 stay.SUBJ.PRES.3SG making many mis-
takes
‘Gianni1 is afraid that he?1/2/she is making many mistakes’.
b. Gianni1 temeva che pro(?)1/2 stesse facendo molti errori.
Gianni1 feared that pro(?)1/2 stay.SUBJ.IMPF.3SG making many
mistakes
‘Gianni1 was afraid that he(?)1/2/she was making many mistakes’.
As observed by Picallo (1985) with respect to Catalan, however, restruc-
turing verbs do trigger obviation. In Italian, they do so too:
(17) a. Aspectual verbs
Gianni1 pensa che pro*1/2 cominci a leggere il libro.
Gianni1 thinks that pro*1/2 begin.SUBJ.PRES.3SG to read.INF the
book
‘Gianni1 thinkes that he*1/2/she begins to read the book’.
b. Motion verbs
Gianni1 pensa che pro*1/2 vada a prendere il libro.
Gianni1 thinks that pro*1/2 go.SUBJ.PRES.3SG to take.INF the
book.
‘Gianni1 thinks he*1/2/she goes and gets the book’.
c. Conative verbs
Gianni1 pensa che pro*1/2 provi a leggere il libro.
Gianni1 thinks that pro*1/2 try.SUBJ.PRES.3SG to read.INF the
book.
‘Gianni1 thinks he*1/2/she tries to read the book’.

ENVIRONMENTS WEAKENING SUBJUNCTIVE OBVIATION
43
However, the presence of a tense auxiliary or of a modal verb (preferably
in the imperfect subjunctive) considerably improves the possibility of the
coreferential reading (provided the appropriate scenario):
(18) a. Gianni1 sperava che pro1/2 avesse cominciato a leggere il libro
giusto.
Gianni1 hoped that pro1/2 AUX.SUBJ.IMPF.3SG begun to read.INF
the book right
‘Gianni1 hoped that he1/2/she had begun to read the right book’.
(Scenario: Gianni is not sure about the book he was assigned to
read)
a'. Gianni1 sperava che pro1/2 potesse cominciare a leggere il libro.
Gianni1 hoped that pro1/2 can.SUBJ.IMPF.2SG to began to read the
book
‘Gianni1 hoped that he1/2/she could begin to read the book’.
(Scenario: Gianni is not sure whether he is authorized to start
reading the book)
b. Gianni1 sperava che pro1/2 fosse andato a prendere il libro giu-
sto.
Gianni1 hoped that pro1/2 AUX.SUBJ.IMPF.3SG gone to bring the
book right.
‘Gianni1 hoped that he1/2/she had gone and bring the right
book’.
(Same scenario as a)
b'. Gianni1 sperava che pro1/2 potesse andare a prendere il libro.
Gianni1 hoped that pro1/2 CAN.SUBJ.IMPF.3SG to go to bring the
book.
‘Gianni1 hoped that he1/2/she could go and bring the book’.
(Same scenario as a')
c. Gianni1 sperava che pro1/2 potesse provare a leggere il libro.
Gianni1 hoped that pro1/2 could.SUBJ.IMPF.3SG to try to read.INF
the book.
‘Gianni1 hoped that he1/2/she could try to read the book’.
(Same scenario as a' and b')
According to some Italian native speakers, implicative verbs (riuscire, ‘to
manage’, etc.) are not as strict as aspectual, motion, and conative verbs in
triggering obviation. In the following examples the two subjects may core-
fer, even though the most natural interpretation is the obviative one:

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
44
d. Gianni1 sperava che pro(?)1/2 riuscisse a leggere il libro entro do-
mani.
Gianni1 hoped that pro(?)1/2 manage.SUBJ.IMPF.3SG to read the
book by tomorrow.
‘Gianni1 hoped that he(?)1/2/she would manage to read the book
by tomorrow’.
Thus, implicative verbs appear to behave differently form aspectual, motion,
and conative verbs.
3.2.5 Combining auxiliaries and modals
The combination of at least two of the factors discussed above (tense auxilia-
ries, voice auxiliaries, modal verbs) generally appears to increase the possi-
bility to coindex the matrix and the embedded subject. The possible combi-
nations are the following:
a. Auxtense+Auxvoice +Vparticiple
b. Mod+Auxtense+Vparticiple
c. Mod+Auxvoice+Vparticiple
d. Mod+ Auxtense+Auxvoice+Vparticiple
e. Auxtense+Mod+Vinfinitive
Among the above combinations, the first has already been discussed (sec-
tion 3.2.2).
Let us consider the remaining four combinations, which also appear to
weaken the disjoint reference effect:
(19) a. Mod+Auxtense+Vinfinitive
Gianni1 pensa/pensava che pro1/2 possa/potesse aver fatto molti
errori.
Gianni1 thinks/thought that pro1/2 can.SUBJ.PRES/IMPF.3SG have
made many mistakes
‘Gianni thinks/thought it is/was possible he1/2/she has/had made
many mistakes’.
b. Mod+Auxvoice+Vparticiple
Gianni1 pensa/pensava che pro1/2 possa/potesse essere ammesso
all’università.
Gianni1 thinks/thought that pro1/2 can.SUBJ.PRES/IMPF.3SG be
admitted at the university
‘Gianni1 thinks/thought that he1/2/she can/could be admitted at
the university’.

ENVIRONMENTS WEAKENING SUBJUNCTIVE OBVIATION
45
c. Mod+ Auxtense+Auxvoice+Vparticiple
Gianni1 pensa/pensava che pro1/2 possa/potesse essere stato
ammesso all’università.
Gianni1 thinks/thought that pro1/2 can.SUBJ.PRES/IMPF.3SG be
been admitted at the university
‘Gianni1 thinks/thought that he1/2/she could have been admitted
at the university’.
d. Auxtense+Mod+Vinfinitive
Gianni1 si rammarica/rammaricava che pro1/2 sia/fosse dovuto
partire così presto.
Gianni1 regrets/regretted that pro1/2 AUX.SUBJ.PRES/IMPF.3SG
must.PRT leave so early
‘Gianni regrets/regretted that he1/2/she has/had had to leave so
early’.
Intuitively, obviation does not appear to hold in the above sentences.
Thus, apparently, the more embedded a full verb is, the more available
coindexation will be. Thus, obviation appears to be sensitive to locality re-
strictions.
3.3 Obviation and arguments
3.3.1 Embedded subject
3.3.1.1 Theta-role sensitivity
Ruwet (1984) observes that in French coindexation is marginally acceptable
if the embedded verb is agentless or ‘psychological’:
(20) a. ?Je veux que je guérisse.
I want that I get.SUBJ.PRES.1SG better
‘I want I get better’.
b.
?Je veux que j’amuse ces enfants.
I want that I amuse.SUBJ.PRES.1SG those children.
‘I want I amuse those children’.
Similar examples may be found in Spanish (Suñer 1986):
(21) Pedro1 negó que pro1/2 supiera la verdad.
Pedro
1 denied that pro1/2 konowSUBJ.IMPF.3SG the truth
‘Pedro1 denied that he1/2/she knew the truth’.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
46
In Italian, the same phenomenon is observed only with respect to agent-
less verbs. It appears that a further improvement is achieved if the matrix
verb is epistemic rather than volitional, if emphatic stress is added, and if the
subjects are first person pronouns:
(22) a. ?Pro1ps Spero (proprio) che io guarisca presto.
Pro1ps hope (really) that I recover.SUBJ.PRES.1SG soon
‘I hope I will soon get better’.
b. *Pro1ps Spero (proprio) che io diverta quei bambini.
Pro1ps hope (really) that I amuse.SUBJ.PRES.1SG those children
Thus, the agent theta-role appears to be implicated in obviation to a larger
extent than the theme (or the benefactive).
3.3.1.2 First person pronouns
Obviation appears also to be sensitive to the grammatical person. Intuitively,
first person pronouns appear to facilitate coindexation, but only in the envi-
ronments in which coindexation is already (more or less) marginally avail-
able:
(23) a. *Voglio che io legga il libro.
I-want that I read.SUBJ.PRES.1SG the book.
b.
?Spero (proprio) che io abbia superato l’esame.
I-hope (really) that I
AUX.SUBJ.PRES.1SG passed the exam
‘I (really) hope I have passed the exam’
c. Spero (proprio) che io sia stato ammesso all’università.
I-hope (really) that I AUX.SUBJ.PRES.1SG been admitted at the
university
‘I (really) hope I have been admitted at the university’
d. Spero (proprio) che io possa leggere quel libro.
I-hope that I can.SUBJ.PRES.1SG read that book
‘I (really) hope I can read that book’.
The presence of the first-person pronoun in the subjunctive clause is
obligatory. A sentence like spero che sia stato ammesso ammesso all’uni-
versità can only be interpreted as ‘I hope that he/she has been admitted at the
university’. And a sentence like Gianni spera che sia stato ammesso all’uni-

ENVIRONMENTS WEAKENING SUBJUNCTIVE OBVIATION
47
versità is ungramamatical under the interpretation ‘John hopes that I have
been admitted at the university’. As it has sometimes been noticed, first and
second person present subjunctive behaves as non-pro-drop verbal forms.
The fact that the first (and second) person subjunctive is non-pro-drop does
not appear in itself to be the only reason for the obviation ‘weakening’, as
the data concerning the third person in the following paragraph show.
3.3.1.3 Strong pronouns
Examples involving an explicit third person pronoun – a weak pronoun (see
Cardinaletti and Starke 1999) – do not appear indeed to have a different
status from examples including a phonetically unrealized pronoun pro.
Compare for instance the following examples with the sentences (1), chapter
2, (1)a, (9)a, and (12)a:
(24) a. Gianni1 vuole che egli*1/2 legga il libro.
Gianni
1 wants that he*1/2 read.SUBJ.PRES.3SG the book
‘Gianni
1 wants that he*1/2 reads the book’.
b. Gianni1 pensa che egli%?1/2 abbia superato l’esame.
Gianni
1 thinks that he%?1/2 AUX.SUBJ.PRES.3SG passed the exam
‘Gianni
1 thinks that he%?1/2 has passed the exam’.
c. Gianni1 pensa che egli%?1/2 sia stato promosso.
Gianni
1 thinks that he%?1/2 AUX.SUBJ.PRES.3SG been promoted
‘Gianni
1 thinks that he%?1/2 will pass the exam’.
d. Gianni1 pensa che egli%?1/2 possa partire domani.
Gianni1 thinks that he%?1/2 can.SUBJ.PRES.3SG leave.INF tomor-
row
‘Gianni1 thinks that he%?1/2 will be able to leave tomorrow’.
The presence of a strong pronoun, however, seems to matter, since
coindexation appears to be almost acceptable in ‘weak’ obviative environ-
ments13:
(25) a. Gianni1 pensa che lui(?)1/2 abbia superato l’esame.
Gianni
1 thinks that he(?)1/2 AUX.SUBJ.PRES.3SG passed the exam
‘Gianni
1 thinks that he(?)1/2 has passed the exam’.
13 For the syntactic and semantic differences between strong, weak and clitic/null pronouns,
see Cardinaletti and Starke (1999) and Cardinaletti (2002).

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
48
b. Gianni1 pensa che lui(? )1/2 sia stato promosso.
Gianni
1 thinks that he(?)1/2 AUX.SUBJ.PRES.3SG been promoted
‘Gianni
1 thinks that he(?)1/2 will pass the exam’.
c. Gianni1 pensa che lui(?)1/2 possa partire domani.
Gianni1 thinks that he(?)1/2 can.SUBJ.PRES.3SG leave.INF tomor-
row
‘Gianni1 thinks that he(?)1/2 will be able to leave tomorrow’.
Intuitively, the sentences in (25) appear to be better than the sentences in
(24), under the coreferential interpretation.
Moreover, if the subject pronoun is focused, the coindexation appears to
be possible even if the subjunctive morphology is attached directly to the
verbal free morpheme:
(26) Gianni1 pensa che solo LUI?1/2 parta domani.
Gianni1 thinks that only HIM?1/2 leave.SUBJ.PRES.3SG tomorrow.
‘Gianni thinks he himself only will leave tomorrow’.
3.3.1.4 Overlapping reference
Suñer (1985) observes that obviation concerns the strict coreference between
two arguments, but does not occur if two arguments overlap in reference.
(27) Lía1 animó Julián2 a que pro1+2 escribieran algo juntos.
Lía1 encouraged Julián2 to that pro1+2 wrote.SUBJ something to-
gether.
‘Lía encouraged Julián to write something together’.
In the above sentence, the embedded subject can refer to a set of individuals
that contains at least two elements, one of which can be the individual the
matrix subject denotes – thus, the reference of the matrix and of the embed-
ded subject ‘overlap’.
In Italian and in French (see Schlenker 2005) overlapping reference is
also possible:
(28) a. Gianni1 ha detto a Maria2 che pro1 vuole che pro1+2 partano
prima possibile.
Gianni1 has told Maria2 that pro1 wants that pro1+2
leave.SUBJPRES.3PL as soon as possible
‘Gianni told Maria that he wants they leave as soon as possi-
ble’.

ENVIRONMENTS WEAKENING SUBJUNCTIVE OBVIATION
49
b. Spero che partiamo subito.
Hope.
IND.PRES.1SG that leave.SUBJ.PRES.1PL at once
‘I hope that we leave at once’.
c. Tu voudras que vous vous raisez à 7h.
You.2SG will-want that you.2PL you.2PL shave.SUBJ.PRES.2PL at
7am
‘You will want that you shave at 7am’.
The availability of overlapping reference does not seem to be granted on
the whole. The following sentence allows coindexation very marginally, if
ever:
(29) Il presidente1 si chiede dove pro??1+ si riuniscano.
The chair1 wonders where pro??1+ gather.SUBJ.PRES.3PL gathered
‘The chair1 wonders that they??1+ gathered’.
I assume that the unavailability of overlapping reference is associated with
interrogative predicates.
3.3.2 Embedded object
It is well known that subjunctive obviation never involves the embedded ob-
ject:
(30) Gianni1 desidera che Maria lo1/2 inviti alla riunione.
Gianni1 wishes that Maria him1/2 invite.SUBJ.PRES.3SG to the meeting
‘Gianni1 wishes that Maria invited him1/2 to the meeting’.
An embedded object is free to corefer with the matrix subject. This holds for
the other Western Romance languages as well:
(31) a. Catalan (Picallo 1985)
[En Joan]1 esperava que [en Jordi] l1/2’invités a la reunió.
[The Joan]1 hoped that the Jordi him1/2 invite.SUBJ.IMPF.3SG to
the meeting
‘Joan1 hoped that Jordi would invite him1/2 to the meeting’.
b. French
Jean1 veut que Marie l1/2’invite au colloque.
Jean1 wants that Marie him1/2 invite.SUBJ.PRES.3SG at the meeting
‘Jean1 wants Marie to invite him1/2 to the meeting’.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
50
c. Portuguese (Raposo 1985)
[O Manel]1 deseja que a Maria o1/2 insulte.
[The Manel]1 wishes that the Maria him1/2 insult.SUBJ.PRES.3SG
‘Manel wishes that Maria insulted him’.
d. Spanish (Suñer 1986)
Paco1 quiere que María lo1/2 invite a la fiesta.
Paco wants that Maria him invite.SUBJ.PRES.3SG at the party
‘Paco wants Maria to invite him to the party’.
3.3.3 Matrix arguments
3.3.3.1 Introduction
In the present section I will address the question, which matrix arguments
may be involved in subjunctive obviation. I will consider three types of ar-
guments: object arguments of psych- and epistemic verbs, object arguments
of directive verbs, and prepositional arguments of nouns. The divide be-
tween arguments that may be involved in obviation from arguments that may
not is provided by the Theta-theory. Only arguments that are assigned the
experiencer theta-role, however they surface categorially, must be referen-
tially disjoint from the embedded subject.
3.3.3.2 Experiencer arguments
The embedded subject in subjunctive clauses is generally obviative with re-
spect to the matrix subject, but it can obviate even with respect to a matrix
object. Psych-verbs and epistemic verbs, which assign the experiencer theta-
role to an argument surfacing as the direct object (e.g. preoccupare ‘to
worry’), or as the indirect object (e.g. dispiacere ‘to regret’, sembrare ‘to
seem’, used as an attitude predicate)14, as requiring a subjunctive argument
clause, do trigger the disjoint reference effect between the matrix object and
the embedded subject:
(32) a. Lo1 preoccupa che pro*1/2 parta domani.
Him
1 worries that pro*1/2 leave.SUBJ.PRES.3SG tomorrow
‘He1 is worried that he*1/2/she will leave tomorrow’.
14 See Belletti and Rizzi (1988).

ENVIRONMENTS WEAKENING SUBJUNCTIVE OBVIATION
51
b. A Gianni1 dispiace che pro*1/2 parta domain.
To Gianni1 regrets that pro*1/2 leave.SUBJ.PRES.3SG tomorrow
‘Gianni1 regrets that he*1/2/she will leave tomorrow’.
c. A Gianni1 sembra strano che pro*1/2 parta domani.
To Gianni1 seems strange that pro*1/2 leave.SUBJ.PRES.3SG tomor-
row
‘It seems strange to Gianni1 that he*1/2/she will leave tomorrow’.
The same status holds for the examples containing a psych- or an epis-
temic verb in Catalan (Picallo 1985):
(33) Li1 agradava que pro*1/2 llegís el diari.
[To-Him/her]1 was-pleasant that pro*1/2 read. SUBJ.IMPF.3SG the
newspaper.
‘He/she1 found it pleasant that [he/she]*1/2 used to read the newspa-
per’.
The restrictions we have noticed for the cases of obviation involving the
matrix and the embedded subject are true even of the cases of object-subject
obviation. Thus, if the subjunctive morphology is attached to a tense or pas-
sive auxiliary, or to a modal verb, coreference tends to be acceptable:
(34) a. A Gianni1 sembrava strano che pro?1/2 avesse fatto molti errori.
To Gianni1 seemed strange that pro?1/2 AUX.SUBJ.IMPF.3SG
made many mistakes
‘It seemed strange to Gianni1 that he?1/2/she had made many
mistakes’.
b. Lo1 preoccupava che pro?1/2 non venisse ammesso all’università.
Him1 worried that pro?1/2 not AUX.SUBJ.IMPF.3SG admitted to
the university
It worried him1 that he?1/2/she would not be admitted to the uni-
versity’.
c. Lo1 preoccupava che pro?1/2 dovesse partire così presto.
Him1 worried that pro?1/2 must.SUBJ.IMPF.3SG leave so early
‘It worried him1 that he?1/2/she had to leave so early’.
d. Gli1 dispiaceva che pro?1/2 non fosse stato ammesso all’univer-
sità.
Him1 regretted that pro?1/2 not AUX.SUBJ.IMPF.3SG been admit-
ted to the university
He regretted1 that he?1/2/she was not been admitted to the uni-
versity’.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
52
e. Gli1 dispiaceva che pro1/2 potesse aver fatto molti errori.
Him1 regretted that pro1/2 can.SUBJ.IMPF.3SG have made many
mistakes
‘He1 regretted that he1/2/she may have made many mistakes’.
3.3.3.3 Arguments of directive verbs
Obviation arises in sentences containing a directive predicate between the
matrix subject and the embedded subject, but not between the matrix object
and the embedded subject (such sentences belong to a rather formal register):
(35) Gianni1 chiese a Maria2 che pro*1/2 partisse il giorno dopo.
Gianni1 asked to Maria2 that pro*1/2 leave.SUBJ.IMPF.3SG the day
after
‘Gianni1 asked Maria2 PRO*1/2 to leave on the following day’.
Similar data can be found in the other Romance languages here dis-
cussed:
(36) a. Catalan (Picallo 1985)
[En Pere]1 va convèncer [en Jordi]2 que pro*1/2 anés a Nova
York.
[The Pere]1 goes persuade.INF [the Jordi]2 that pro*1/2
go.SUBJ.IMPF.3SG to New York
‘Pere
1 persuaded Jordi2 PRO*1/2 to go to New York’
b. Spanish (Suñer 1986)
José1 lo2 persuadió a que pro*1/2 apagara la TV.
José1 him2 persuaded to that pro*1/2 turn-off.SUBJ.IMPF.3SG the
TV
‘José1 persuaded him2 PRO*1/2 to turn off the TV’.
In Italian, if the embedded verb is passive, pro can corefer freely with the
matrix subject:
(37) Gianni1 chiese a Maria che pro1/2 fosse autorizzato a partire.
Gianni1 asked to Maria that pro1/2 AUX.UBJ.IMPF.3SG authorized to
leave
‘Gianni1 asked Maria that he1/2 would be authorized to leave’.
The availability of coindexation in the above sentence is somehow ex-
pected, considering how passive voice interacts with obviation.

ENVIRONMENTS WEAKENING SUBJUNCTIVE OBVIATION
53
All in all, the above data show that obviation can involve a matrix (direct
or indirect) object. The embedded subject must obviate from an object, only
if the object is assigned the experiencer theta-role. What sentences in which
object-subject obviation obtains have in common with examples in which
subject-subject obviation obtains, is that the embedded subject must be ref-
erentially disjoint from an argument referring to the individual to which an
attitude towards the propositional content the embedded clause refers to is
attributed. Following a common usage in the philosophy of language, such
an argument may be dubbed as ‘bearer of attitude’ (or ‘attitude holder’), or,
as is habitual in semantic literature, ‘agent (or author) of the context’15.
3.3.3.4 Possessives and genitives
When a subjunctive clause is the argument of a deverbal noun expressing an
attitude (for instance, volontà, ‘will’, desiderio, ‘desire’, speranza, ‘hope’),
the subject of the subjunctive clause can obviate with respect to possessive
pronoun or a ‘genitive’ constituent:
(38) Il {suo1} desiderio {di Gianni1} che pro*1/2 parta domani svanirà
presto.
The [his/her]1 wish that pro*1/2 leave.SUBJ.PRES.3SG tomorrow will
vanish soon.
‘[His/her]1 wish that [he/she]*1/2 left tomorrow will vanish soon’.
The possessive and the genitive refer to the bearer of attitude here. Thus, the
above sentence simply shows that obviation does not depend on the ‘surface’
grammatical category.
Obviation within clausal complements of nouns has the same distribution
of obviation within argument clauses of verbs. First, indicative, conditional,
and infinitive mood do not trigger obviation:
(39) a. La sua1 affermazione che pro1/2 partirà domani ha sorpreso tutti.
The [his/her]1 statement that pro1/2 leave.IND.FUT.3SG tomorrow
has surprised all
‘[His/her]1 statement that [he/she]1/2 will leave tomorrow has
surprised everyone’.
b. La sua1 affermazione che pro1/2 sarebbe partita il giorno dopo
sorprese tutti.
The [his/her]1 statement that pro1/2 AUX.COND.3SG left the day
after surprised all
15 For the present purpose, we will take these terms as synonyms.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
54
‘[His/her]1 statement that [he/she]1/2 would leave on the follow-
ing day surprised everyone’.
c. La sua1 decisione di PRO1 partire sorprese tutti.
The [his/her]1 decision of PRO1 to eave surprised all
‘[His/her]
1 decision to leave surprised everyone’.
Second, if the form carrying subjunctive morphology is an auxiliary or a
modal, coindexation tends to be recovered:
(40) a. La sua1 speranza che pro1/2 avesse fatto pochi errori svanì ben
presto.
The [his/her]1 hope that pro1/2 AUX.SUBJ.IMPF.3SG made few
mistakes vanished very soon
‘[His/her]1 hope that [he/she]1/2 had made few mistakes van-
ished quickly’.
b. La sua1 speranza che pro1/2 fosse (stato) ammesso all’università
fu delusa.
The [his/her]1 hope that pro1/2 AUX.SUBJ.IMPF.3SG (been) admit-
ted at the university was disappointed
‘[His/her]1 hope that [he/she]1/2 would be/have been admitted to
the university was disappointed’.
c. La sua1 speranza che pro1/2 potesse partire il giorno dopo rimase
delusa.
The [his/her]1 hope that pro1/2 can.SUBJ.IMPF.3SG leave.INF the
day after remained disappointed
‘[His/her]1 hope that it was possible for [him/her]1/2 to leave on
the following day was disappointed’.
Since possessives and genitives satisfy the role of the experiencer and de-
note the agent of the context, the facts concerning obviation in a clausal ar-
gument of a noun parallel the observation stated so far with respect to clausal
arguments of verbs.
3.4 Double embedding
The data discussed so far concern only one level of embedding. In such
cases, the embedded subject and a matrix argument cannot corefer. As it has
sometimes been observed in the literature, in double embedding, the same
rule holds locally. That is, the most embedded subject is obviative with re-

ENVIRONMENTS WEAKENING SUBJUNCTIVE OBVIATION
55
spect to the intermediate subject, but it is not obviative with respect to the
matrix subject:
(41) Gianni1 sperava che Maria2 desiderasse che pro1/*2/3 partisse con
lei.
Gianni1 hoped that Maria2 wish.SUBJ.IMPF.3SG that pro1/*2/3 leave.
SUBJ.IMPF.3SG with her
‘Gianni1 hoped that Maria2 wished that he1/3/she*2/3 left with her’.
The co-indexation between the most embedded subject and the matrix sub-
ject is available even when the intermediate predicate is ‘modal’ (like to be
possible, to be necessary, etc.) and has an explective phonetically unrealized
argument:
(42) Gianni1 pensava che fosse indispensabile che pro1/2 telefonasse.
Gianni1 thought that be.SUBJ.IMPF.3SG indispensable that pro1/2
call.SUBJ.IMPF.3SG.
‘Gianni hoped that it was indispensable that he1/2 called.
Other Western Romance Languages behave alike:
(43) a. Catalan (Picallo 1985)
En Pere1 esperava que en Jordi2 volgués que pro1/*2/3 hi anés.
[The Pere]1 hoped that [the Jordi]2 want.SUBJ.IMPF.3SG that
pro1/*2/3 there go SUBJ.IMPF.3SG.
‘Pere1 hoped that Jordi2 wanted him1/*2/3/her to go there’.
b. French (Progovac 1993, 1994)
Jean1 veut qu’il*1/2 desire qu’il1/*2 aime Marie.
Jean1 wants that he*1/2 wish.SUBJ.PRES.3SG that he1/*2 love
SUBJ.PRES.3SG Marie
‘Jean1 wants him*1/2 to wish he1/*2 loved Marie’.
c. Portuguese (Raposo 1985)
[O Eduardo]1 deseja que [o Manel]2 queira que ele1/*2/3 compre
um automóvel novo.
[The Eduardo]1 wishes that [the Manel]2 want.SUBJ.PRES.3SG
that he1/*2/3 buy.SUBJ.PRES.3SG a car new.
‘Eduardo1 wishes that Manel2 wanted him1/*2/3 to buy a new
car’.
From the above data, the conclusion follows that obviation can involve the
embedded subject and the subject (or the object) of the immediately dominat-
ing clause.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
56
3.5 Adverbial clauses
Manzini (2000) observes that in rationale clauses the subject displays all the
interpretative properties it has in clausal arguments, namely, it cannot be
coindexed with the subject of the main clause:
(44) *Pro1sg vengo perché pro1sg ti aiuti.
Pro1sg come.IND.PRES.1SG in-order-that pro1sg you-CL
help.SUBJ.PRES.1SG
According to Manzini, this sentence can be paraphrased as ‘I come because I
want that I help you’. In her view, the fact that rationale clauses include a
implicit will attitude explains the obviative interpretation.
Obviation in rationale clauses is sensitive to all the characteristics we
have observed in complement clauses concerning the presence of auxiliaries
and modals. Thus, the following sentences appear to be almost acceptable:
(45) a. ?Vado a comprare il biglietto perché io possa partire domani.
Go.IND.PRES.1SG to buy the ticket in-order-that I
can.SUBJ.PRES.1SG leave tomorrow
‘I go to buy the ticket so that I can leave toorrow’.
b. ?Ho parlato con il mio avvocato perché fossi rilasciato.
AUX.IND.PRES.1SG talked with my attorney in-order-that
AUX.SUBJ.IMPF.1SG released
‘I talked with my attorney so that I could be released’.
Obviation can be instantiated even in subjunctive adverbial clauses that
do not involve an attitude (at least, not obviously). Before-clauses instantiate
obviation. However, as Manzini (2000) points out, there are some examples
in which coreference is almost legitimate, as (46)b:
(46) a. Gianni1 legge il giornale prima che pro*1/2 faccia colazione.
Gianni1 reads the newspaper before that pro*1/2
make.SUBJ.PRES.1SG breakfast
‘Gianni1 reads the newspaper before he*1/2/she has breakfast’.
b.
?Pro vado prima che pro mi arrabbi.
Go.IND.PRES.1SG before that get-angry.SUBJ.PRES.1SG
‘I go before I get angry’.
In any case, example (46)b may not be actually valid, since agentless verbs
and the first person facilitate coindexation.

ENVIRONMENTS WEAKENING SUBJUNCTIVE OBVIATION
57
Relative clauses also display obviation:
(47) Gianni1 sta cercando un esperto con cui pro*1/2 parli del problema.
Gianni1 is looking for an expert with whom pro*1/2
talk.SUBJ.PRES.3SG about the problem
‘Gianni1 is looking for an expert he*1/2/she can talk about the prob-
lem with’.
A modal verb improves the sentence to full grammaticality:
(48) Gianni1 sta cercando un esperto con cui pro1/2 possa parlare del
problema.
Gianni1 is looking for an expert with whom pro1/2
can.SUBJ.PRES.1SG talk about the problem
‘Gianni1 is looking for an expert he1/2 can talk about the problem
with’.
Other subjunctive adverbial clauses seem to be immune from obviation.
Apparently, in clauses expressing a comparison, in conditional, and in con-
cessive clauses coindexation is not ruled out:
(49) a. Gianni1 è più intelligente di quanto pro1/2 non pensi.
Gianni
1 is more smart than pro1/2 not think.SUBJ.PRES.3SG
‘Gianni
1 is smarter than he1/2/she thinks’.
b. Gianni1 si divertirebbe, se pro1/2 venisse.
Gianni
1 would-amuse, if pro1/2 go.SUBJ.PRES.3SG
‘Gianni
1 would amuse himself, if he1/2/she went’.
c. Gianni1 perderà il treno nonostante pro1 prenda un taxi.
Gianni1 will lose the train despite pro1 take.SUBJ.PRES.3SG a taxi
‘Gianni will lose the train despite he will take a taxi’.
d. Gianni1 mi potrà parlare purchè prima pro1 mi chieda scusa.
Gianni to-me will be able to talk only if before pro1 to-me
apologize.SUBJ.PRES.3SG
‘Gianni will be able to talk to me only if he apologizes to me
before’.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
58
3.6 Topics and clause dislocation
3.6.1 Left dislocated topics
In Italian the insertion of a topic appears to allow coindexation more easily.
For the native speakers who do not exclude coindexation in the former of the
following sentences, the coreferential interpretation is even more acceptable
if a left dislocated topic occurs:
(50) a. Gianni1 spera che pro%?1/2 abbia fatto pochi errori all’esame di
linguistica.
Gianni1 hopes that pro%?1/2 AUX.SUBJ.PRES.3SG made few mis-
takes at-the exam of linguistics
‘Gianni1 hopes that he%?1/2/she has made few mistakes at the
exam of linguistics’
b. Gianni1 spera che, all’esame di linguistica, pro%1/2 abbia fatto
pochi errori.
Gianni1 hopes that at-the exam of linguistics pro%1/2
AUX.SUBJ.PRES.3SG made few mistakes
‘Gianni1 hopes that at the exam of linguistics, he%1/2/she has
made few mistakes’
This phenomenon is observed in Catalan as well (Gemma Rigau, p.c.,
Feldhausen 2008):
(51) a. [En Joan]1 no es pensa que pro%?1/2 hagi fet molts errors a
l’examen.
[The Joan]1 not it thinks pro%?1/2 that AUX.SUBJ.PRES.3SG made
many mistakes at the exam
‘Joan1 doesn’t think he%?1/2/she has made many mistakes at the
exam’
b. [En Joan]1 no es pensa que [a l’examen de lingüística
computacional] pro1/2 hi hagi fet molts errors.
[The Joan]1 not it thinks that at the exam of computational lin-
guistics pro%?1/2 AUX.SUBJ.PRES.3SG made many mistakes
‘John doesn’t think he pro%?1/2/she has made many mistakes at
the exam of computational linguistics’.
Particularly, Feldhausen (2008) observes that such an effect is achieved
only if a circumstantial is left dislocated (LD). If it is right dislocated (RD),

ENVIRONMENTS WEAKENING SUBJUNCTIVE OBVIATION
59
the coreferential reading is odd (as in the above examples) and disjointed-
ness is enforced:
(52) a. Gianni1 pensa che pro%?1/2 sia andato nel 1991 ad Amburgo (ma
non si ricorda l’anno esatto)
Gianni1 thinks that pro%?1/2 AUX.SUBJ.PRES.3SG gone in 1991 to
Hamburg (but he doesn't remember the precise year)
‘John thinks that he%?1/2/she has gone to Hamburg in 1991 (but
he doesn’t remember the precise year)’.
b. Gianni1 pensa che [ad Amburgo], pro1/2 ci sia andato nel 1991
(ma non si ricorda l’anno esatto). (LD)
c. Gianni1 pensa che pro%?1/2 ci sia andato nel 1991, [ad Amburgo]
(ma non si ricorda l’anno esatto). (RD)
The same effect holds in Catalan:
(53) a. [En Joan]1 espera que pro%?1/2 no digués res mal dit al congrés
de la societat sociològica.
[The John]1 hopes that pro%?1/2 not say.SUBJ.IMPF.3SG nothing
bad say at.the congress of the society sociology
‘John1 hopes that he%?1/2/she has said nothing bad at the con-
gress of the society of sociology.’
b. [En Joan]1 espera que [al congrés de la societat sociològica]
pro1/2 no hi digués res mal dit. (LD)
c. [En Joan]1 espera que pro%?1/2 no hi digués res mal dit, [al
congrés de la societat sociològica]. (RD)
Apparently, even ‘core’ instances of obviation appear to be influenced by the
presence of a topic. The former of the following sentences appears to be
more acceptable than the second, which is out, under the relevant indexation:
(54) a. Gianni1 sperava che all’esame di linguistica, pro??1/2 facesse po-
chi errori.
Gianni1 hoped that at the exam of linguistics pro??1/2
make.SUBJ.IMPF.3SG few mistakes
‘Gianni1 hoped that at the exam of linguistics, he??1/2/she would
make few mistakes’.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
60
b. Gianni1 sperava che pro*1/2 facesse pochi errori.
Gianni
1 hoped that pro*1/2 make.SUBJ.IMPF.3SG few mistakes.
‘Gianni hoped that he*1/2/she would made few mistakes’.
3.6.2 Clause dislocation
The dislocation of a subjunctive clause appears to affect the interpretative
properties of its subject. The Italian speakers who marginally accept sen-
tence (50)a under the coreferential interpretation, consider such a reading
even more acceptable if the subjunctive clause is left dislocated as a topic or
a focus:
(55) a. Che pro%(?)1/2 abbia fatto pochi errori all’esame di linguistica,
Gianni1 lo spera proprio.
That pro%(?)1/2 AUX.SUBJ.PRES.3SG made few mistakes at-the
exam of linguistics, Gianni1 it hopes really.
‘Gianni1 really hopes that he%(?)1/2/she has made few mistakes at
the exam of linguistics’.
b. Che pro%?1/2 abbia fatto POCHI ERRORI, Gianni1 pensa.
That pro%(?)1/2 AUX.SUBJ.PRES.3SG made FEW mistakes, Gianni1
thinks.
‘Gianni1 thinks that he%(?)1/2/she has made FEW mistakes’.
We note that if the focus is in situ (or in the right periphery), it brings about
no change in status:
(56) Gianni1 pensa che pro%?1/2 abbia fatto POCHI errori.
Gianni1 thinks that pro%?1/2 AUX.SUBJ.PRES.3SG made FEW mis-
takes
‘Gianni1 thinks that he%?1/2/she has made FEW mistakes’.
Thus, contrastive focus appears to affect obviation no matter where the fo-
calized constituent occurs.
3.7 Coordination
Ruwet (1984) and Schlenker (2005) observe that when a subjunctive clause
is coordinated, coindexation is marginally possible. This can be observed
even in Italian, but only if the subject of the subjunctive clause is explicit:

ENVIRONMENTS WEAKENING SUBJUNCTIVE OBVIATION
61
(57) a. Gianni1 spera che Maria parta e che lui1/2 resti.
Gianni1 hopes that Maria leave.SUBJ.PRES.3SG and that he?1/2
stay.SUBJ.PRES.3SG
‘Gianni hopes that Maria leaves and that he stays.
b. Gianni1 spera che Maria parta e che pro*1/2 resti.
Gianni1 hopes that Maria leave.SUBJ.PRES.3SG and pro1/2
stay.SUBJ.PRES.3SG
Intuitively, here pro can only be interpreted as the subject of the coordinated
clause.
Ruwet’s (1984) original example from French sounds as follows:
(58) ?Je veux que tu partes et que je reste.
I want that you leave.SUBJ and that I stay.SUBJ.
‘I want you to leave and me to stay’
3.8. Conclusions
In this chapter I have tried to define the syntactic environments in which ob-
viation occurs. Many factors appear to influence the interpretative properties
of the embedded subject, so that not in all subjunctive clauses the subject
displays obviation. I have reached the descriptive conclusion that in the main
Western Romance languages, a subject is obviative with respect to the agent
of the context only in clauses in which the subjunctive morphology is at-
tached to the full verb. This represents a restricted subset of the subjunctive
environments. In other contexts obviation appears to be ‘weaker’, if any.
Such ‘weakening’ depends on a series of factors: tense – the imperfect
subjunctive appears to weaken obviation more than the present subjunctive,
provided that other condition, such as the presence of an auxiliary or the
presence of a modal, are satisfied; the morpheme to which subjunctive mor-
phemes are attached – auxiliaries and modals partially allow for coindexa-
tion, whereas full verbs generally do not; the theta-role assigned to the sub-
ject also appears to affect the possibility of coindexation; obviation appears
to be sensitive to locality, as the data from double embedding show; in ad-
verbial clauses, subject obviation does not arise in all clauses requiring the
subjunctive mood: it does only in rationale clauses, before-clauses and rela-
tive clauses; finally, left dislocated topics, topicalization, focusing and coor-
dination appear to improve the degree of acceptability of sentences in which
coindexation is at most marginally possible. All these factors appear to
strengthen or weaken obviation:

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
62
(59) a. Matrix verb
i. ‘Simple’ tenses > perfect tenses
ii. Present subjunctive > imperfect subjunctive
iii. Subjunctive full verb > Subjunctive modal
b. Embedded subject
i. Agent > Theme, Experiencer
ii. Weak, clitic subject pronoun > strong subject pronoun
iii. 3rd ps > 1st ps
iv. Strict coreference > overlapping reference
c. Matrix Experiencer > Matrix Theme, Benefactive
d. No dislocated XP > topic, focus XP
e. No dislocated subjunctive CP > topic, focus CP
f. Rationale, before-, relative clauses > conditional, concessive,
comparison clauses
g. No coordinated clause > coordinated clause
As involving many grammatical properties, obviation does not appear to
be an all-or-nothing phenomenon. Rather, the environments I have discussed
may be ordered in a continuum, the extremes of which are given by ‘future-
oriented’ clausal complements, in which coindexation is ruled out, and, to
remain within the domain of subjunctive clauses, clauses having a subjunc-
tive modal verb followed by a past passive infinitive and a left dislocated
topic, in which, baroque though it may be, coindexation appears to be fully
legitimate:
(60) a. *Gianni1 vuole che pro1 superi l’esame.
Gianni1 wants that pro1 pass.SUNJ.PRES.3SG the exam
b. Gianni1 sperava che l’esame di linguistica computazionale, pro1
lo potesse aver superato con il massimo dei voti.
Gianni1 hoped that the computational linguistic test, pro1 it.CL
may.SUBJ.IMPF.3SG have passed with full marks.
‘Gianni hoped he may have passed the computational linguis-
tics test with full marks’.
Between these two environments, a series of intermediate degrees holds,
including sentences in which the availability of coindexation appears to be
intralinguistically variable and, if any, more or less marginal.

4 Theories of subjunctive obviation
4.1 Intorduction
Two kinds of hypothesis have been generally pursued to account for the
facts concerning obviation. Both analyze the ‘core’ set of data concerning
obviation, paying few attention, if any, to the nuances the phenomenon dis-
plays.
The first type of approach stems from the tenet that obviation is the con-
sequence of the ‘competition’ between the subjunctive and the infinitive
mood. This viewpoint has been maintained by Bouchard (1983, 1984), Far-
kas (1992), and more recently by Schlenker (2005).
The second type of approach builds on the idea that the binding domain
of the embedded subject includes the matrix subject. Principle B of Binding
Theory prevents pro from being coindexed with the subject of the main
clause and obviation arises. The causes of the binding domain extension are
generally attributed to the properties of the subjunctive. However, different
machineries have been implemented in order to account for such an exten-
sion. The ‘standard’ hypothesis (Picallo 1985, Raposo 1985, Rizzi 1986,
Progovac 1993, 1994) argues that the binding domain extension is due to the
subjunctive inflection, which is claimed to be ‘anaphoric’ and must be bound
by the matrix tense. Everaert (1986), Kempchinsky (1987, 1998), Raposo
(1985), Avrutin (1994), Tsoulas (1996) and Avrutin and Babyonyshev
(1997), Manzini (2000) adopt different strategies, which also build on the
properties of the subjunctive mood.
In what follows I will be briefly illustrate the two types of approach and
then discuss the advantages and the disadvantages of each.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
64
4.2 ‘Competition’ theories
Competition theories are based on the descriptive observation that obviation
occurs in the syntactic environments in which an infinitive is available to
convey a coreferential reading.
Bouchard (1983, 1984) claims that the Elsewhere Principle rules the dis-
tribution of anaphors and pronouns. It states that, in a given environment, an
anaphoric interpretation cannot be obtained by means of a pronoun if it can
be obtained by means of an anaphor:
(1) Elsewhere Principle
Don’t put a pronoun in a position where an anaphor is possible,
that is, in a position where the pronoun will be interpreted as co-
referential with an NP that can bind it.
Bouchard argues that the Elsewhere Principle can be applied to the obvia-
tive examples, as the following sentence in French:
(2) *Je veux que j’aille voir ce film.
I want that I go.SUBJ.PRES.1SG to see this movie.
He observes that in this sentence the matrix and the embedded subjects can-
not be coreferential. They must be coreferential, however, when the argu-
ment clause is infinitive:
(3) Je veux PRO aller voir ce film.
I want PRO go.INF see this movie.
‘I want to go and see this movie’.
The hypothesis that the subject position of a subjunctive or of an infiniti-
val clause is subjected to the Elsewhere Principle, predicts that a pronoun
cannot convey an anaphoric reading when appearing in such a position, since
an anaphor, PRO, is available. Thus, the pronoun must be referentially dis-
joint from the matrix subject.
Bouchard’s theory also predicts that when PRO is pronominal, the Else-
where Principle does not rule out a pronoun. In Bouchard’s view, this pre-
diction is borne out, since the following sentences are fully acceptable:
(4) a. PRO1 d’être menacé de mort ne me1 fera pas changer d’idée.
PRO1 be.INF menaced with death not me1 makes NEG change
of idea
‘Being menaced with death will not make me change my mind’.

THEORIES OF SUBJUNCTIVE OBVIATION
65
b. Que je sois menacé de mort ne me fera changer d’idée.
That I be.SUBJ.PRES.1SG threatened with death not me makes.fut
NEG change.INF of idea
‘That I have been threatened with death will not make me
change my mind’.
Bouchard claims that in the former sentence PRO is pronominal, since it is
not bound by a c-commanding NP. In the latter, the embedded subject is
pronominal as well. Since the subject of the embedded clause is pronominal
in both sentences, there is no contrast between the two in terms of the Else-
where Principle and coreference is not ruled out in the subjunctive clause.
Farkas (1992) proposes an analysis in the same spirit. She maintains that
obviation follows from the competition between the subjunctive and the in-
finitive moods. She observes that obviation occurs only in those languages in
which the infinitive competes with the subjunctive, such as Western Ro-
mance languages. In Romanian as well as in the languages included in the
so-called Balkan Sprachbund, there is no such modal competition and no
disjoint reference effect arise16.
The relevant generalization arising from the data she analyses is that ob-
viation occurs (i) between the matrix and the embedded subject, if the latter
is assigned the agent theta-role, (ii) in volitional and desiderative environ-
ments. Such a generalization predicts, beside Bouchard’s data, examples in
which the embedded subject is not assigned the agent role and obviation
does not occur in spite of the subjunctive mood, and despite an infinitival
‘competitor’ is available. The following sentences, originally discussed by
Ruwet (1984), illustrate the point:
(5) a. Je veux que je puisse partir.
I want that I can.SUBJ.PRES.1SG leave.
‘I want I can leave’.
b. Je veux que je sois autorisé à partir tôt.
I want that I be.SUBJ.PRES.1SG authorized to leave soon
‘I want to be authorized to leave soon’
c. Je veux que je guérisse aussi tôt.
I want that I recover.SUBJ.PRES.1SG soon
‘I want to recover soon’.
16 Terzi (1992) and Krapova (1998) claim that subjunctive complements in Greek and Bulgar-
ian are indeed ambiguous (see section 3.4 for a brief discussion of her idea). For a more recent
analysis in the same spirit, applied also to Hebrew data, see Landau (2004).

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
66
Moreover, the subjunctive-infinitive ‘rivalry’ is neutralized in sentences
in which the embedded subject corefers with a matrix argument that is not
assigned the agent role as in the following sentences from Spanish:
(6) a. Forcé a Ana1/La1 forcé a que pro1 visitara al medico.
I forced Ana1/Her.CL1 forced to that pro1 visit.SUBJ.IMPF.3SG
the doctor.
‘I forced Ana/her to go to the doctor’
b. Forcé a Ana1/La1 forcé a PRO1 visitar al medico.
I forced Ana1/her.CL1 forced to PRO1 visit.INF the doctor.
‘I forced Ana/her to go to the doctor’
Farkas’ hypothesis is that in the contexts described, the infinitival clause
‘blocks’ the subjunctive clause to express coreference between the matrix
and the embedded subjects, where the term ‘blocking’ is used as in lexical
semantics and morphology (see Aronoff 1976): a more ‘specific’ construc-
tion, excludes a more general one in a determined environment. Infinitive
structures are taken to be more ‘specific’ than subjunctive in that the subject
of an infinitive inherits the referential properties of another argument as a
lexical requirement, whereas such a restriction does not hold for the subject
pronoun in subjunctive clauses17.
Schlenker (2005) builds on Farkas’ (1992) theory to extend the idea that
obviation is due to the competition between subjunctive and infinitive. He
argues that subjunctive mood is a semantic default – that is, it has vacuous
semantics, bearing no presuppositions. Due to the pragmatic principles
‘Maximize Presupposition!’ and ‘Prefer de se!’, subjunctive must be used
only in case its competitor, the infinitive mood, as far as we are concerned,
causes a semantic failure – that is, when the presupposition introduced by the
infinitive is infelicitous. Since the infinitive is a semantically non-null ele-
ment – it bears the presuppositions that the propositional attitude is individ-
ual and event first-personal or de se (in the sense of Castañeda 1966, 1968
and Lewis 1979)18, a semantic failure may obtain if this presupposition is not
satisfied – say, if the situation that is being reported is not de se. If the situa-
tion that is being reported is compatible with a de se logical form (LF), by
17 To be more precise, Farkas states that obviation follows from a pragmatic principle stating
that «in world-dependent complements [such are the complements of volitional and desidera-
tive predicates, but not, crucially, of epistemic and emotive-factive predicates] that conform to
the canonical control case [the configuration in which both the matrix subject and the ‘subject
dependent’ – referentially dependent from the matrix subject – embedded subject are respon-
sible for bringing about the situation denoted by the complement], the form used to mark
subject dependency [the infinitive] blocks the form used for world dependency [the subjunc-
tive]».
18 For a discussion on event de se attitudes, see sections 4.4.2 and 6.3.1.

THEORIES OF SUBJUNCTIVE OBVIATION
67
‘Maximize Presupposition!’ and by ‘Prefer de se!’ the infinitival structure is
to be chosen, whereas the subjunctive is infelicitous. Subjunctive can be
used in non-de se situations, where a semantic failure rules the infinitive out
(as in the sentences in (5), where the event variable is non-de se), or in the
environments in which the infinitive is ruled out on syntactic grounds, as in
Ruwet’s (1984) examples concerning coordination19:
(7) a. ?Je veux que tu partes et que je reste.
I want that you leave.SUBJ.PRES.2SG and that I
stay.SUBJ.PRES.1SG
‘I want you to leave and me to stay’
b. *Je veux que tu partes et rester.
I want that you leave.SUBJ.PRES.2SG and that I stay.INF
The restriction that two coordinate clauses must be isomorphic rules the
latter sentence out. The only option available is then the former sentence.
Notice that under this interpretation of the facts concerning the subjunc-
tive and the infinitive, there is nothing in itself that prevents pro from being
coindexed with a matrix argument, and, as a matter of fact, subjunctive
clauses may even be de se. We refer to section 4.4 for a discussion on this
point.
4.3 Binding-Theoretical approaches
4.3.1 Binding Theory Principle B
Obviation in subjunctive clauses has been often claimed to follow from Prin-
ciple B of Binding Theory:
(8) Binding Theory Principle B
A Pronominal is free in its Governing Category
The terms ‘free’, ‘bound’, and ‘Governing Category’ (GC) are defined as
follows:
(9) a. Binding
α is bound by β iff α and β are co-indexed and β c-commands α;
α is free iff it is not bound (Chomsky 1980, 1981).
19 See section 3.7.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
68
b. Governing Category (Chomsky 1981)
β is the GC for α iff β is the minimal category containing α, a
governor of α, and a SUBJECT accessible to α,
where SUBJECT is agreement in tensed clauses, the subject of
infinitival clauses, NPs or small clauses.
Obviation is expected to obtain under the hypothesis that the governing
category for the subject position in subjunctive clauses includes the matrix
subject. As for the reason why such a governing category extension occurs, a
general consensus has not been reached and different strategies have been
pursued. I will illustrate here some of them.
4.3.2 Hypothesis on the ‘binding domain extension’
4.3.2.1 Tense anaphoricity
Picallo (1985), Raposo (1985), Rizzi (1986), Progovac (1993, 1994) and
Manzini (2000) claim that the subjunctive mood is ‘anaphoric’ in that it is
assigned a temporal value in relation to the time frame of its superordinate
clause20. Lacking autonomous tense specification, subjunctive tenses must be
‘bound’ by tenses having an autonomous time specification. Picallo (1985)
calls this binding relation a ‘tense chain’ and hypothesizes that a binding
domain for a constituent x may be defined as the minimal domain of a tense
chain containing x and an accessible (c-commanding) subject.
Similarly, Raposo (1985) claims that in Romance languages a binding
domain can be either the c-commanding domain of a subject or the c-
command domain of a ‘verbal operator’, raised to C at LF. Such are the op-
erator [+Tense], which conveys a ‘deictic’ time reference in indicative
clauses, modal verbs, and even auxiliaries.
Rizzi’s (1986) definition of governing category is substantially equivalent
to Picallo’s:
(10) Z is a governing category for X iff Z is the minimal category with a
subject containing X, a governor G for X, and where the binding
requirements of X and G are satisfiable.
The governor of the embedded subject is the subjunctive inflection. Thus,
the governing category for the embedded subject includes the matrix subject,
since the binding requirements for the subjunctive inflection are satisfied by
the matrix verbal inflection.
20 See section 5.2.5 for a discussion of this property in Italian.

THEORIES OF SUBJUNCTIVE OBVIATION
69
Progovac’s (1993, 1994) definition of binding domain is also substan-
tially equivalent to Picallo’s, Raposo’s, and Rizzi’s. She accepts Chomsky’s
definition of governing category and proposes that the subjunctive C/I sys-
tem is deleted at LF as temporally uninterpretable, so that a subjunctive
clause is not a Governing Category for pro.
A series of theories slightly departs from the above ones, adopting differ-
ent mechanism to explain obviation, which is however assumed to be a
Binding Principle B violation21.
4.3.2.2 Williams’ Associations.
Everaert (1986) proposes an account for obviation that relies on Binding
Theory and on William’s (1985) Association theory.
According to Williams, a verb specifies the referential ‘associations’
among thematic roles. To illustrate, consider the following sentence:
(11) *John1 performed his1 autopsy
Here the agent role of the main verb and the patient role of autopsy, which is
assigned to the argument his, cannot be ‘associated’ – the agent of perform is
rather associated with the ‘actor’ of autopsy. Since associated roles must be
uniformly indexed, whereas non-associated roles cannot, the sentence is
ruled out. Thus, Binding Theory is sensitive to associations.
Everaert (1986) applies Williams’ association theory to the data on obvia-
tion from Dutch. He compares the following sentences:
(12) a. Jan1 wil dat hij*1/2 het boek leest.
Jan1 wants that he*1/2 the book reads
‘Jan1 wants him*1/2 to read the book’.
b. Jan1 wil van Karel2 dat hij*1/2 het boek leest.
Jan1 wants of Karel2 that he*1/2 the book reads
‘Jan1 wants Karel to read the book’.
He assumes that in the former example the verb willen ‘to want’ has an im-
plicit argument, which is made explicit in the former sentence by the prepo-
sitional phrase. He then hypothesizes that the verb willen specifies an asso-
ciation between the van-argument and the agent of the embedded predicate.
21 For two different proposals that I will not illustrate here, see Suñer (1986) and Hornstein
and San Martin (2005). Suñer’s account of obviation builds on the interpretative properties of
subjunctive clauses, which are encoded within the subjunctive complementizer. Hornstein and
San Martin (2005) hypothesize that a mechanism of ‘anti-control’ determines obviation.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
70
Concerning theta-roles, this association holds even when a theta-role is not
assigned to an explicit argument. Thus, in sentence (12)a, the implicit argu-
ment of the matrix predicate and the agent of the embedded predicate are
associated, implicitness notwithstanding. As for the theta-role assigned to
Jan, it is not associated with the embedded agent so that Jan and hij cannot
be coindexed.
Everaert’s theory predicts that if the embedded subject is not assigned the
agent role, no disjoint reference effect should arise. He assumes that modal
verbs assign a ‘secondary theta-role’ to their subject and that an association
is sensitive to it. Hence, if the verb of the embedded clause is modal, obvia-
tion is not expected to arise. In his view, this idea explain the following con-
trast:
(13) a. Peter1 eiste van Jan2 dat hij*1/2 meewerkt.
Peter
1 demanded from Jan2 that he*1/2 cooperates
‘Peter1 asked Jan2 that he*1/2 would cooperate’.
b. Peter1 eiste van Jan2 dat hij1/?*2 mee mocht werken.
Peter1 demanded from Jan2 that he1/?*2 can cooperate
‘Peter1 asked Jan2 that he1/?*2 would be allowed to cooperate’.
In the former sentence hij ‘he’ cannot be coreferent with Peter, due to the
associations required by the matrix verb, whereas in the latter sentence, the
matrix and the embedded subjects corefer.
4.3.2.3 I-to-C movement
Kempchinsky (1987) maintains that the binding domain extension follows
under the hypothesis that volitional verbs select for a subjunctive operator in
C, which is conveyed by the subjunctive morphology. The subjunctive I
must raise to C at LF in order to satisfy the subcategorization requirements
of the matrix predicate. Defining a governing category for a constituent x as
the least Complete Functional Complex, the minimal category in which all
grammatical functions are satisfied (Chomsky 1986), containing x and the
governor of x, she claims that due to I-to-C movement, the governing cate-
gory for the embedded subject at LF is no longer the embedded clause, since
the governor of the subject, I, has moved higher up, but the dominating
clause. Thus, due to Binding Principle B, obviation obtains.

THEORIES OF SUBJUNCTIVE OBVIATION
71
4.3.2.4 Nominative Case marking
In further research, Kempchinsky (1998) hypothesizes that obviation is ex-
plained by hypothesizing, as Watanabe’s (1993) and Koizumi’s (1995) do,
(i) that C is the ultimate licenser of Nominative Case, (ii) that the functional
structure of subjunctive clauses lacks the CP level, and, finally, (iii) that a
binding domain may be defined as the Nominative Case-checking domain.
Since the Nominative Case on the embedded subject must be checked by the
matrix C, being the embedded C missing, the binding domain for the em-
bedded subject includes the arguments of the matrix clause, thus instantiat-
ing obviation.
4.3.2.5 I-to-C and C-to-C movement
Avrutin (1994) and Avrutin and Babyonyshev (1997) build a theory of ob-
viation discussing data from Russian. These data parallel the data from Ro-
mance languages. The paradigm they analyze is the following:
(14) a. Volodja1 xočet čtoby on*1/2 pocelovala Nadju.
Volodya
1 wants that.SUBJ he*1/2 kissed Nadya
‘Volodja wants to kiss Nadya’.
b. Volodja1 skazal čto on1/2 poceloval Nadju.
Volodya1 said that he1/2 kissed Nadya
‘Volodja1 said that he1/2 kissed Nadya’.
c. Volodja1 xočet čtoby Nadja pocelovala ego1/2.
Volodya1 wants that.SUBJ Nadya kissed him1/2
‘Volodja1 wants Nadja to kiss him1/2’.
d. Volodja1 xočet čtoby emu1 bylo veselo.
Volodya1 wanted that.SUBJ him.DAT1 was.SG.NEUT fun
‘Volodya wants to be having fun’.
e. Volodja ugovoril Nadju1 čtoby ona1 poexala v Evropu.
Volodya persuaded Nadya1 that.SUBJ she1 went to Europe
‘Volodya persuaded Nadya to go to Europe’.
In Russian, obviation obtains between the nominative matrix and the
nominative embedded subject of ‘subjunctive’ clauses (introduced by the
complementizer čtoby). It does not occur if the embedded clause is in the
indicative (introduced by the complementizer čto – see example (14)b), be-

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
72
tween the matrix subject and the embedded object (see example (14)c), be-
tween the matrix subject and the embedded quirky subject (see (14)d), and
between the matrix object and the embedded nominative subject (see (14)e).
They assume that V moves to I (that is, to T and to AgrS) and (in non-V2
languages, covertly) to C, with no cross-linguistic variation. Moreover, they
take that since volitional verbs are ‘future oriented’, an event operator, which
they hypothesize is the subjunctive compelmentizer čtoby, co-binds the
events of the embedded and of the matrix clause, determining the temporal
ordering. To do so, it is assumed to move at LF to the matrix C, from where
it can c-command and bind both the matrix VP and the embedded VP, pass-
ing through any head between its base position and the matrix C. Finally,
they assume that Agr heads are pronominal and that they are coindexed with
a constituent standing in a specifier-head relation with them.
They show that this theoretical apparatus and Binding Principle B are
able to account for obviation, which is predicted to take place at LF between
the matrix and the embedded AgrS°, and not directly between the matrix and
the embedded subject22. Avrutin (1994) claims that independent evidence for
the mechanism proposed is provided from data on language acquisition.
4.3.2.6 C-to-V movement
Tsoulas (1996) implements an account of obviation in minimalist terms. Fol-
lowing Chomsky (1995), he claims that a binding domain may be defined as a
‘minimal domain’ – that is, the minimal subset of categories locally related to
a head. A minimal domain is defined derivationally, not representationally: if
a head moves, the minimal domain is defined with respect to the resulting
chain. He then hypothesizes that at LF an embedded C having subjunctive
mood features moves to the matrix V to get its features checked (Tsoulas
1994a, b). The binding domain for the embedded subject thus includes the
trace of the matrix subject in the specifier of VP, determining obviation.
4.3.2.7 Exceptional Case Marking
Luján's (1999) hypothesizes that control and obviation are a result of Case
marking. She assumes that pronouns universally undergo LF-movement to C
in order to define their reference, and that clausal complements are marked
with Case (through Exceptional Case Marking). Due to these principles, at
LF the subject pronoun of a subjunctive clause lays on the head carrying the
Case features assigned by the matrix verb. Thus, it has to be interpreted in
22 See Avrutin (1994) and Avrutin and Babyonyshev (1997) for a detailed discussion.

THEORIES OF SUBJUNCTIVE OBVIATION
73
the domain of the main clause. Unlike subjunctive clauses, which are as-
sumed to have a simple CP, indicative clauses assumingly have a double CP.
A subject pronoun moves to the lower C, whereas Case is assigned to the
higher C. Thus, the binding domain for the pronoun does not extend to the
matrix clause. Infinitival clauses, finally, have a simple CP structure, like
subjunctive clauses. But since their subject, PRO, is anaphoric, coindexation
can (and must) take place.
4.3.2.8 Syntactic dependency
Manzini (2000) defines a binding domain relying on the notion of ‘syntactic
dependency’, and claims that (as in Avrutin’s and Avrutin and Babyony-
shev’s theories) obviation occurs between the matrix and the embedded
agreement morphology, rather than between the subjects in themselves.
Building on the observation that in Romance languages the embedded I tense
specification depends on the temporal properties of the matrix I, she assumes
that, since I contains [Tense] and [Agr], the matrix and the embedded [Agr]
belong to the same syntactic ‘dependency’. Moreover, [Agr] is taken to be
pronominal. Hypothesizing that a dependency is a binding domain, Binding
Principle B rules out coindexation between agreement heads occurring
within the same dependency and obviation follows.
4.3.3 Explaining obviation through Binding Theory
All these theories can be considered as extensions of the Binding Theory in
order to account for the data concerning subjunctive clauses. Thus, while
making the same predictions as Chomsky’s original theory about the distri-
bution of anaphors and pronouns in a local domain, they predict that the
binding domain for the subject of a subjunctive clause should include in
some sense the matrix subject.
They explain a series of facts:
i) While subjunctive clauses trigger obviation, indicative clauses do not.
Indicative verbs are not ‘anaphoric’, so they do not need to be part of a tense
chain with a c-commanding verb (in Picallo’s terms) – or, equivalently, they
do not need to be bound:
(15) a. Subjunctive: [X1 I2 [Y*1 I2]
b. Indicative: [X1 I2 [Y1 I3 ]
Whatever mechanism one chooses, the binding domain for the embedded
subject is extended to the matrix clause if the embedded tense is ‘anaphoric’,
otherwise it corresponds to the embedded clause itself.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
74
ii) The object in a subjunctive clause is free to corefer with the matrix
subject:
(16) [X1 [Y2 Z1]]
The embedded subject is an accessible subject for an embedded object and,
whatever binding requirements the embedded verb may have, the matrix
subject is not contained in the binding domain for the embedded object.
iii) The subject of a doubly embedded clause in the subjunctive can be
coindexed with the matrix subject, but must be obviative with respect to the
intermediate subject:
(17) a. [X1 [Y2 [Z1 ]]]
b. [X1 [Y2 [Z*2 ]]]
The minimal domain including an accessible subject for the most embedded
subject contains the intermediate, but not the matrix subject, whatever bind-
ing requirements the embedded verb may have.
iv) Obviation may involve a matrix object, proviso that the notion of ‘ac-
cessible subject’ is substituted with the notion of accessible ‘argument’. As
Picallo (1985) points out, not all object arguments can be considered as ‘ac-
cessible’ for the embedded subject. Only those that belong to the thematic
structure of ‘ergative-type’ verbs (please, regret, etc.) appear to have this
property. On the other hand, ‘transitive’ predicates (persuade, convince, etc.)
do not instantiate obviation between the embedded subject and a matrix ob-
ject, which is then inaccessible for the embedded subject, although they do
instantiate obviation between the embedded and the matrix subjects (exam-
ples from Picallo 1985):
(18) a. En Pere va convèncer [en Jordi]1 que pro1 anés a Nova York.
The Pere persuaded [the Jordi]1 that pro1 go.SUBJ.IMPF.3SG to
New York
‘Pere persuaded Jordi to go to New York’
b. *[En Pere]1 va convèncer en Jordi que pro1 anés a Nova York.
[The Pere]1 persuaded the Jordi that pro1 go.SUBJ.IMPF.3SG to
New York
Here Picallo claims that the embedded clause is extraposed23. Kempchinsky
(1987) also adopts a similar solution. The embedded subject is then c-
23 Due to Stowell’s (1981) Case Resistance Principle, which states that Case cannot be as-
signed to a category bearing a Case-assigning feature, like complement clauses, and requires
that a clausal argument be removed from a Case-assignment position, like, Picallo assumes,
the VP-internal position of the object clause.

THEORIES OF SUBJUNCTIVE OBVIATION
75
commanded by the matrix subject but not by the matrix object. Thus, the
matrix subject, but not the matrix object is an accessible argument for pro,
which consequently is obviative with respect to the former, but it may be
proximate with respect to the latter.
(19) a. [X1 [Y2] [Z2 ]]
b. [X1 [Y2] [Z*1 ]]
Avrutin’s (1994), Avrutin and Babyonyshev’s (1997) and Manzini’s
(2000) accounts of examples like 0a differ from Picallo’s and Kempchin-
sky’s in that they resort subject agreement. Since in Romance languages
(and in Russian) an object does not agree with a verb, there is no restriction
on the referential properties of the embedded subject, which is allowed to
corefer with the matrix object.
v) Picallo (1985) and Raposo (1985) also observe that modal verbs de-
termine ‘opacity’, so that the subject of a subjunctive clause can be
coindexed with the matrix subject (examples from Picallo 1985):
(20) a. Pro1 sentien que pro*1 produissin una falsa impressió.
Pro1 regret.IND.IMPF.3PL that pro*1 produced.SUBJ.IMPF.3PL a
false impression
‘They1 regretted that they*1 give a false impression’
b. Pro1 sentien que pro1 deguessin produir una falsa impressió.
Pro1 regretted.IND.IMPF.3PL that pro1 must.SUBJ.IMPF.3PL pro-
duce.INF a false impression
‘They
1 regretted that they1 must give a false impression’
Picallo hypothesizes that sentential operators, such as [+Tense] for in-
dicative clauses, or modal verbs, must raise to C, and in doing so, they de-
limit a binding domain, thus eliminating the disjoint reference effect. Such a
hypothesis has also been maintained by Raposo (1985), who includes among
the ‘verbal operator’ even the auxiliaries to account for examples like the
following:
(21) a. [O Manel]1 deseja que pro1/2 seja admitido no concurso.
[The Manel]1 wishes that pro1/2 AUX.SUBJ.PRES.3SG admitted in-
the contest
‘Manel1 wishes that he1/2/she was admitted in the contest’.
b. [A Maria]1 preferia que pro1/2 não tivesse encontrado o Manel.
[The Maria]1 preferred that pro1/2 not AUX.SUBJ.IMF.3SG met the
Manel.
‘Maria1 wished she1/2/he had not met Manel’.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
76
Thus, the structural configuration of sentences involving an auxiliary or a
modal verb corresponds to (15)b rather then to (15)a.
4.4 Remarks on the theories on obviation
4.4.1 Integrating ‘competition’ and Binding Theory approaches
I take that the ‘Competition’ and the Binding Theory-based approaches have
gained important insights into the phenomenon of obviation, and although
they have often been considered as antithetical – the evidence against the
one has often motivated the attempt to pursue the other, they may indeed be
seen as complementary or even be equivalent, as Luján (1999) suggests.
Picallo (1985) and Rizzi (1986) claim indeed that if the binding domain
for the subject position of an embedded clause includes the matrix subject
(or the object of a psych-verb), pronouns are expected to be referentially dis-
joint from this matrix argument, but anaphors may be bound by them. This
expectation is incorrect, however. Take for instance Picallo’s example:
(22) *[En Pere] espera que si mateix arribi.
[The Pere] hopes that himself arrive.SUBJ.PRES.3SG
As both Picallo and Rizzi point out, however, the ungrammaticality of
this sentence may be due to independent reasons. As an empirical generali-
zation, anaphors cannot be in the Nominative. This may be due to the licens-
ing of lexical anaphors in the subject position of a finite clause, or perhaps to
Case checking.
In any case, if, more in general, one takes the binding domain for the sub-
ject of a clause having anaphoric tense to include a matrix argument, since
under Binding Theory-based accounts an anaphor would be legitimate in the
embedded subject position, the distribution of PRO, which is anaphoric, fol-
lows straightforwardly. Note that infinitival clauses share the temporal prop-
erties of subjunctive clauses, in that their interpretation depends on the ma-
trix tense (see section 6.2.1). Thus, this solution may be pursued for free.
This notwithstanding, the two types of approach have been generally re-
garded as mutually exclusive. In what follows I will review the two types of
approaches and show that, attractive though a unified approach may be, it is
not feasible, and that in spite of some limitations, competition theories ap-
pear to be more promising.

THEORIES OF SUBJUNCTIVE OBVIATION
77
4.4.2 Competition theories: an analysis
Common arguments against the competition theories are that the comple-
mentary distribution of subjunctive and infinitival clauses does not always
hold – a claim that is meant to show that it is not always the case that a ‘pro-
noun cannot be put in a position where an anaphor is possible’, to repeat the
basic tenet of competition theories. Discussing Bouchard’s (1983, 1984)
proposal, Picallo (1985) observes that in Catalan the subjunctive-infinitive
opposition is neutralized in directive environments – see for instance exam-
ple 0a, here repeated as 0a, contrasted with 0b:
(23) a. En Pere va convèncer [en Jordi]1 que pro1 anés a Nova York.
The Pere persuaded [the Jordi]1 that pro1 go.SUBJ.IMPF.3SG to
New York
‘Pere persuaded Jordi that he should go to New York’.
b. En Pere va convèncer [en Jordi]1 de PRO1 anar a Nova York.
The Pere persuaded [the Jordi]1 of PRO1 go.INF to New York
‘Pere persuaded Jordi to go to New York’.
Farkas (1992) rejects the above data as a piece of evidence against a
competition approach in itself, restricting the empirical field where competi-
tion is at work to sentences conforming to what she dubs ‘canonical control
case’ – volitional environments involving a controllee satisfying the agent
role (see note 17 of the present chapter). But such a step appears to be justi-
fied only on an empirical ground24.
Suñer (1986) puts forward the same objection as Picallo, observing that,
in Spanish, not only directive verbs do not instantiate obviation, but denial,
dubitative, and emotive-factive predicates do so as well, despite the avail-
ability of a ‘rival’ in the infinitive:
(24) a. Pedro1 negó que pro1/2 supiera la verdad.
Pedro
1 denied that pro1/2 konowSUBJ.IMPF.3SG the truth
‘Pedro1 denied that he1/2/she knew the truth’.
a'. Pedro negó PRO1 saber supiera la verdad.
Pedro1 denied PRO1 know.INF the truth
‘Pedro1 denied knowing the truth’.
b. Dudo que pro1/2 lo hubiera pagado.
Doubt.IND.PRES.1SG that pro him AUX.SUBJ.IMPF.1SG paid
‘I doubt having that I/he/she had paid him’.
24 For a discussion on directive predicates and competition theories of obviation, see also
Feldhausen (2007).

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
78
b'. Dudo PRO haberlo pagado.
Doubt.IND.PRES.1SG have(inf)-him paid
‘I doubt having paid him’.
c. Yo sentí mucho que pro no lo haya visto.
I regret much that pro not him.CL AUX.SUBJ.PRES.1SG seen
‘I deeply regret that I/he/she have/has not seen him’.
c'. Yo sentí mucho PRO no haberlo visto.
I regret much PRO not AUX.INF-him.CL seen
‘I deeply regret not having seen him’.
One may even add examples in which coindexation appears to be possible:
(25) a. Gianni1 pensava che pro1/2 potesse essere ammesso all’univer-
sità.
Gianni1 thought that pro1/2 can.SUBJ.IMPF.3SG be admitted at the
university
‘Gianni1 thought that he1/2/she could be admitted at the univer-
sity’.
b. Gianni1 pensava di PRO1 poter essere ammesso all’università.
Gianni1 thought to PRO1 can.INF be admitted at the university
‘Gianni1 thought he1 could be admitted at the university’.
Here again, Farkas’s theory would strike the above examples from the
‘canonical control cases’ – somehow stipulatively, however, since there is no
obvious sense in which one is not responsible for paying or for seeing
(though there may be some sense in which one is not responsible for know-
ing the truth or for being admitted at the university).
It is perhaps here the case to note that some of the examples involving
multiple functional verbs may even give support to the competition theories,
since infinitival clauses appear to be rather degraded:
(26) a. Gianni1 pensava che pro1/2 potesse aver fatto molti errori.
Gianni1 thought that pro1/2 can.SUBJ.IMPF.3SG have made many
mistakes
‘Gianni thought he1/2/she might have made many mistakes’.
a'.
??Gianni1 pensava di poter aver fatto molti errori.
Gianni1 thought C/P PRO1 can.INF have made many mistakes

THEORIES OF SUBJUNCTIVE OBVIATION
79
b. Gianni1 pensava che pro1/2 potesse essere stato ammesso all’uni-
versità.
Gianni1 thought that pro1/2 can.SUBJ.IMPF.3SG been been admit-
ted at the university
‘Gianni1 thought that he1/2/she could have been admitted at the
university’.
b'. ??Gianni1 pensava di PRO1 poter essere stato ammesso
all’università.
Gianni1 thought C/P PRO1 can.INF be been admitted at the uni-
versity
Compelling evidence against ‘competition’ theories could only involve a
volitional verb and an embedded subject satisfying the agent role. Another
potential falsifying evidence may be provided by non ‘canonical control’
examples in which obviation does occur.
Prima facie, examples of the former sort – concerning a tense auxiliary in
the subjunctive, appear to correspond to this description. The following ex-
ample from Italian reproduces one of Ruwet’s (1984) examples from French:
(27) ?Voglio (assolutamente) che io sia partito entro dieci minuti.
I want (absolutely) that I AUX.SUBJ.PRES.1SG left in ten minutes.
‘I want that I am gone in ten minutes’.
An alternative option in the infinitival is available:
(28) Voglio (assolutamente) essere partito entro dieci minuti.
I want (absolutely) AUX.INF left in ten minutes.
‘I want to be gone in ten minutes’.
It seems however that the above sentences have different truth conditions.
Intuitively, the former is understood to mean that I want that the condition
for me to leave must be brought about in ten minutes in order for me to be
gone by that time, but I cannot do anything myself to achieve this. By con-
trast, the latter sentence is understood to mean that I can do something in
order to leave in ten minutes. Thus, in a scenario in which I got stuck in a
traffic jam, the former sentence would be felicitous, whereas the latter would
be infelicitous (for a discussion on alike examples, see section 6.4.1).
Farkas’ theory would probably rule out sentence (27) from the ‘canonical
control case’. In Schlenker’s (2005) theory, although the two sentences are
individual de se, the latter, but not the former, is ‘event de se’ – that is, the
propositional attitude in the latter sentence is about the first person (individ-
ual de se) and an event that is brought about by the first person (event de se).

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
80
To illustrate, a sentence like John wishes to be elected is individual de se,
but it is not event de se, since the event of being elected is not brought about
by John himself.
Thus, bearing no event de se presupposition, the subjunctive can be used
if the infinitive would give rise to a semantic failure. In the traffic jam sce-
nario, the infinitive does give rise to a semantic failure, since the event de se
presupposition is incompatible with the context set. Thus, the default mood,
the subjunctive, is an available option.
As for the second type of evidence, in Italian obviation occurs even in
epistemic and emotive-factive environments, which do not conform to the
‘canonical control’ cases and do not appear to be event de se – the leaving
event may not be brought about by Gianni (he may be compelled or re-
quested to leave):
(29) a. *Gianni1 teme che pro1 parta domani.
Gianni
1 fears that pro1 leave.SUBJ.PRES.3SG tomorrow
b. Gianni1 teme di PRO1 partire domani.
Gianni
1 fears C/P PRO1 leave.INF tomorrow
‘Gianni is afraid to leave tomorrow’.
Farkas’ theory may accommodate this example by modifying the notion
of ‘canonical control’ as to include epistemic predicates. Schlenker’s theory
would claim that the pragmatic principle Maximize presupposition! forces to
choose the latter sentence in a situation compatible with a de se interpreta-
tion, facing no problem at all.
Such an answer, however, requires to scrutiny more in detail the follow-
ing sentences:
(30) a. Spero che io possa vivere a lungo.
Hope.
IND.PRES.1SG that I can.SUBJ.PRES.1SG live long
‘I hope I can live long’.
b. Spero di poter vivere a lungo.
Hope.
IND.PRES.1SG C/P can.INF live long
‘I hope I can live long’.
Intuitively, such sentences are individual de se, but obviously not event de
se. Thus, in a context compatible with a de se interpretation, both sentences
are felicitous. This fact is surprising, since by Maximize presupposition!, in
such a context the latter sentence should be chosen. On the other hand, the
former sentence should be felicitous only in contexts in which an infinitive
determines a semantic failure, but apparently this is not the case, since the
two sentences are appropriate in the same relevant context.

THEORIES OF SUBJUNCTIVE OBVIATION
81
All in all, competition theories appear to be promising in order to explain
obviation, but something more should be said about how competition works.
4.4.3 Binding Theory-based theories under scrutiny
Common objections to the Binding Theory approaches have laid stress on
the fact that in languages where the infinitive is missing in the relevant con-
texts, subjunctive clauses do not trigger obviation. Farkas, for instance, dis-
cusses the following sentence from Romanian:
(31) Ion1 vrea pro1/2 să plece.
Ion1 wants pro1/2 PART.SUBJ leave.3SG
‘Ion wants to leave/that he/she leaves’.
Kempchinsky (1987) argues that data like the one above are not compel-
ling, maintaining that the subjunctive particle să, and not the full verb, which
morphologically do not differ from the indicative form, raises to C at LF.
Thus, the binding domain for pro remains unchanged. Alternatively, one
may follow Terzi’s (1992) hypothesis that in Romanian as well as in other
Balkan languages, the absence of disjoint reference effects in subjunctive
clauses is only apparent. Subjunctive complements are assumed to be am-
biguous since their subject may be PRO or pro:
(32) a. O Yiannis theli na fai to rizogalo.
The Yiannis wants PRT eats the rice pudding
‘John wants (him/her) to eat the rice pudding’.
b. [O Yiannis]1 theli [CP [C [MP PRO1 [M na fai
c. [O Yiannis]1 theli [CP [C na fai [MP pro*1/2 [M …
Evidence for the existence of the two structures is established independently
and correlates with the interpretation of tense in the two types of clause. As
for the interpretative properties of the embedded subject, in the former struc-
ture, it must be interpreted as coreferential with respect to the matrix subject,
whereas in the second structure, it is obviative25. Thus, the absence of obvia-
tion in the systems in which there is no mood competition is apparent and
cannot be taken in itself as a proof to exclude Binding Theory-based ap-
proaches.
25 We refer to Terzi (1992) for a detailed discussion on the subject.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
82
A more substantial objection to Binding Theory approaches is discussed
by Schlenker (2005). He observes that Binding Condition B prohibits not
only coreference, but also overlapping reference:
(33) a. #Tu vous admireras. (French)
You.2SG you.CL.2PL will-admire.
b. *I admire us.
If obviation were a consequence of Binding Theory, one would expect
that the prohibition against overlapping reference should occur even between
the matrix and the embedded subject in subjunctive clauses. But crucially,
this prediction is not borne out26:
(34) a. #Tu voudras que tu te rases à 7h.
You.2SG will-want that you.2SG you.2SG shave.SUBJ.PRES.2SG
at 7am
b. Tu voudras que vous vous raisez à 7h.
You.2SG will-want that you.2PL you.2PL shave.SUBJ.PRES.2PL at
7am
‘You will want that you shave at 7am’.
Schlenker takes these facts as strong evidence against the Binding Theory-
based analysis of obviation.
If this is correct, competition and Binding Theory-based approaches can-
not be claimed to be equivalent, as I have hypothesized at the beginning of
this section, since examples like (34)b, which can be explained by ‘competi-
tion’ as not belonging to the ‘canonical control case’ or not involving a de se
reading, under a Binding-Theoretical approach remains unaccounted for27.
4.5 Conclusive remarks
Everything considered, from a descriptive viewpoint, ‘competition’ theories
appear to be more promising than Binding-Theoretical accounts of obvia-
26 See also the examples in section 3.3.1.4.
27 Additionally, Binding Theory-based proposals are descriptively problematic in that, first,
either they restrict the domain of obviation to volitional contexts only, or they generalize it to
all subjunctive clauses – but some subjunctive clauses do in fact allow for coindexation, and,
second, they depict the phenomenon as a prohibition against coindexation between the matrix
and the embedded subject, which does not appear to be quite accurate. Explicatively, many of
them employ rather stipulative theoretical principles.

THEORIES OF SUBJUNCTIVE OBVIATION
83
tion. Even the empirical generalization on the availability of indexation be-
tween a matrix argument and the embedded object, or between a matrix ar-
gument and the doubly embedded subject, or, finally, between the object of a
directive verb and the matrix subject are easily accounted for by the most
recent competition theory: object pronouns are ambiguously de se, non-de se
and de re, thus coreference is expected. De se attitudes involve a contextual
author and a propositional content, so that doubly embedded clauses are ex-
pected not to be de se with respect to the matrix author – the locality restric-
tions on obviation then follows straightforwardly. Finally, directive proposi-
tions are not first-personal (although they may be second-personal, see
Schlenker, to appear).
It must be noted, however, that many of he factors described in chapter 2
must be accommodated within a ‘competition’ theoretical framework (and
they should be even within a Binding Theory-based approach). Specifically,
the asymmetries (a) between deficient (weak and clitic) subject pronoun and
full subject pronoun (especially if focused), (b) between first and third per-
son, (c) between clauses having a dislocated constituent and clauses having
no such a constituent, and (d) between sentences where the subjunctive
clauses is in situ and sentences where it is dislocated, are all facts that must
be explained. As for the facts in (c) and in (d) one may note that infinitival
clauses do not admit a left dislocated topic or focus and that they themselves
cannot be dislocated:
(35) a. ?*Gianni spera, l’esame di linguistica, di averlo superato.
Gianni hopes, the exam of linguistics, C/P AUX.INF-it.CL passed
b. *Gianni spera di, l’esame di linguistica, averlo superato.
Gianni hopes C/P, the exam of linguistics, AUX.INF-it.CL passed
(36) ??Di aver superato l’esame, Gianni lo spera proprio.
C/P
AUX.INF-it.CL passed the exam, Gianni it.CL hopes really
Thus, the fact that obviation is weakened in these environments is somehow
predicted.
The data concerning the contrasts between first vs. third person and fo-
cused vs. non-focused pronouns cannot be explained through a competition
strategy straightforwardly. As for the first one, both subjunctive clauses in-
volving the first person (in which obviation appears to be weaker if com-
pared with sentences containing a third person pronoun) and those involving
the third person pronoun ‘compete’ with an alternative in the infinitive:
(37) a. ?Spero (proprio) che io abbia superato l’esame.
I-hope (really) that I
AUX.SUBJ.PRES.1SG passed the exam
‘I (really) hope I have passed the exam’

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
84
a'. Spero (proprio) di aver superato l’esame.
I-hope (really) C/P
AUX.INF passed the exam
‘I (really) hope to have passed the exam’
b. Gianni1 spera che pro%?1/2 abbia superato l’esame.
Gianni
1 hopes that pro%?1/2 AUX.SUBJ.PRES.3SG passed the exam
‘Gianni1 hopes that he%?1/2/she has passed the exam’.
b'. Gianni1 spera di aver superato l’esame.
Gianni
1 hopes C/P AUX.INF passed the exam
‘Gianni1 hopes to have passed the exam’.
These examples arise two questions. First, a de se reading is available in
(37)a, and, at least for some speakers, in (37)b, despite the rivalry of an in-
finitival clause. This is a crucial question, the endowment of which is proved
by other examples as well – see for instance the examples in (30), and re-
quires careful consideration. Second, sentence (37)a is intuitively more ac-
ceptable than sentence (37)b. This fact is also to be explained.
As for the second contrast (focused vs. non-focused pronouns), if a fo-
cused pronoun occurs, the subjunctive and the infinitive options are both
syntactically possible and, as far as interpretation is concerned, they are both
de se:
(38) a. Gianni1 spera che parta solo LUI1/2, domani.
Gianni
1 hopes that leave.SUBJ.PRE.3SG only HIM1/2, tomorrow
‘Gianni hopes that he himself only will leave tomorrow’.
b. Gianni1 spera di partire solo LUI1, domani.
Gianni
1 hopes C/P leave.INF only HIM1 tomorrow
‘Gianni hopes to leave himself only tomorrow’.
(Since left dislocated focuses are ruled out in infinitival clauses, the contrast
has been illustrated locating the focused constituent at the right of the verb).
Here again, competition theories should be refined in order to account for
this fact.
In the following chapters I will address these questions.

5 Subjunctive mood at the interface
5.1 Introduction
In the present chapter I will present some phenomena concerning subjunc-
tive mood in Italian, whose properties appear to recall the facts related to
obviation. Although the similarities with obviation are only partial, I will
nonetheless take them to be substantial and not accidental, so that a theory
that account for the ones, must also be able to account for obviation.
The first of the phenomena that I will discuss is Complementizer Dele-
tion (henceforth, CD), the optional omission of the complementizer che,
which obtains in syntactic contexts that appear to partly overlap with those
in which obviation obtains. Connected with CD is the second phenomenon
that I will illustrate, Double Accessibility Reading (DAR for short), which
never obtains in environments in which CD is possible.
The third phenomenon is Long-Distance Anaphors (for short, LDA) bind-
ing, which obtains in subjunctive clauses (and in conditional as well), and
conveys the de se reading – so that in the very environment in which a sub-
ject cannot be de se (obviation), a LDA must be so.
I will assume that these properties may be accounted for through the
same underlying principles. Thus, we will present Giorgi and Pianesi’s
(1997, 2004) theory of CD, which crucially relies on their (2001) theory of
Sequence of Tenses (SoT) and of Double Accessibility Reading (DAR), and
Giorgi’s (2004, 2006, 2007) theory of LDA, which is also based on the the-
ory of SoT.
I will assume that this theoretical apparatus is what is needed to imple-
ment an account for the facts concerning obviation presented in chapters 2
and 3. Such an account will be the topic of chapter 6.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
86
5.2 Complementizer Deletion and Double Accessibility Reading
5.2.1 Complementizer Deletion descriptively
Giorgi and Pianesi (1997, 2004) analyze in detail the environments allowing
CD. Some of the syntactic properties characterizing the environments allow-
ing for CD appear to correlate to some of the environments in which obvia-
tion occurs. A unified explanation for these two phenomena is then desir-
able.
Descripiely, Giorgi and Pianesi observe the following distributional
properties of CD28:
a) CD is optionally possible only in subjunctive clauses29:
28 For completeness, I note that beside the properties I am going to discuss here, Giorgi and
Pianesi single out the following properties affecting CD, which I think are not relevant for the
present purpose: CD clauses have a peculiar distribution of the subject (and, at least for some
Italian native speakers, the occurrence of a preverbal full subject within the embedded is ruled
out); CD is possible in root subjunctive clauses having an optative or imperative meaning; CD
displays ‘first-person effect’. Although obviation is sensitive to the first person, I will take
that this phenomenon roots on different principles and has more to do with the phenomenon
of first-person ‘blocking’ in LDA-Binding (see section 5.3.3) than with the first-person effect
in CD.
29 The only exceptions concern the future indicative and the past conditional (future in the
past) in an embedded clause, which do allow for CD, as Poletto (1995) and Giorig and Pianesi
(1997, 2004) observe:
(i) a. Crede (che) verrà.
Believe.IND.PRES.3SG (that) come.IND.FUT.3SG
‘She/He believes s/he will come’.
b. Sperava (che) sarebbe venuto.
Hope.IND.IMPF.3SG (that) AUX.COND.PRES.3SG come
She/He hoped s/he would come’.
To my intuitions, however, the above sentences are very marginal. Switching the third to the
first person improves them:
(ii) a. Credo (che) verrà
Believe.IND.PRES.3SG (that) come.IND.FUT.3SG
‘I believe s/he will come’.
b. Speravo (che) sarebbe venuto.
Hope.IND.IMPF.1SG (that) AUX.COND.PRES.3SG come
I hoped s/he would come’.
In such sentences the verbs are epistemic and desiderative respectively – that is, predicates
which can select for a subjunctive clause. We assume that this someway counts. CD is indeed
unavailable if the matrix predicate is a verbum dicendi – a predicate which cannot select for a
subjunctive clause – even if the embedded predicate is future:
(iii) a. Ha detto *(che) verrà/sarebbe venuto.
Has said *(that) come.IND.FUT.3SG/AUX.COND.PRES.3SG come
‘She/he has said s/he will come/would come’.
b. Ho detto *(che) verrà/sarebbe venuto.
Have said *(that) come.IND.FUT.3SG/AUX.COND.PRES.3SG come
‘I have said s/he will come/would come’.
We will not consider these facts further here.

SUBJUNCTIVE MOOD AT THE INTERFACE
87
(1) a. Crede (che) sia partito.
Thinks (that) AUX.SUBJ.PRES.3SG left
‘Gianni thinks he left’.
b. Ha detto *(che) è partito.
Has said *(that) AUX.IND.PRES.3SG left
‘Gianni said he left’.
However, not all subjunctive clauses admit CD. Clausal arguments of vo-
litional, desiderative, and epistemic predicates do, whereas emotive-factive
predicates, which select for a subjunctive clause as well, do not30:
(2) Gianni si rammarica *(che) sia partito.
Gianni regerets *(that) is(subj) left
‘Gianni regrets that ha left’.
b) Dislocated subjunctive clauses do not allow for CD, either:
(3) a. *(Che) fosse partito, lo credeva.
That AUX.SUBJ.IMPF.3SG left, it.CL believed
‘That he had left, he believed’.
b. *(CHE) FOSSE PARTITO, credeva.
That
AUX.SUBJ.IMPF.3SG left, believed.
‘HE HAD LEFT, he believed’.
c. *(Che) sia già partito, è probabile.
That
AUX.SUBJ.PRES.3SG already left, is probable
‘That he has already left, is probable’.
c) CD is not allowed in ‘Double Accessibility Reading’ (DAR, for short)
contexts – that is, in contexts in which an embedded eventuality is inter-
preted twice, with respect to the attitude eventuality and with respect to the
30 As in other environments, if the matrix verb is in the first person, CD is not quite ruled out:
(i) a. *Gianni rimpiange tu abbia perso il treno.
Gianni regrets you have lost the train
a'. *A Gianni dispiace tu non abbia vinto.
To Gianni regrets you not have won
b. (?)Rimpiango tu abbia perso il treno.
I regret you have lost the train
‘I regret you have lost the train’
b'. (?)Mi dispiace tu non abbia vinto.
To me regret you not have won
‘I regret you haven’t won’

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
88
utterance time. To show this property, Giorgi and Pianesi underline that the
indicative tenses (excluding the imperfect), which do not allow for CD, typi-
cally display the DAR.
(4) Gianni ha detto che Maria è incinta.
Gianni has said that Maria be.IND.PRES.3SG pregnant
‘Gianni has said that Maria is pregnant’.
This sentence can be considered true if the Maria’s state of pregnancy holds
both at the time of Gianni’s saying eventuality and at the utterance time,
‘now’. Thus, the following formula may express the appropriate logical form
for sentence (4)31:
(5) ∃e∃e' say(G, [be-pregnant(M, e')], e) & t(e) < now & t(e') ≈ t(e) ≈
now
Indicative tenses trigger the DAR, whereas subjunctive tenses apparently
do not:
(6) Gianni pensava che Maria fosse incinta.
Gianni believed that Maria was(subj) pregnant
‘Gianni thought Maria was pregnant’.
This sentence can be considered true if the state of pregnancy holds at the
time of Gianni’s saying (or if the state of pregnancy was anterior to Gianni’s
saying). But there is no time relation specified between the embedded even-
tuality and the utterance time.
However, in some cases subjunctive verbs do instantiate DAR. In these
cases, CD is unavailable. To illustrate, verbs like ipotizzare (‘to hypothe-
size’), which select for a subjunctive clausal argument even when employed
to express a speech act, can convey a DAR32:
(7) a. Gianni ha ipotizzato che Maria fosse incinta.
Gianni has hypothesized that Maria AUX.SUBJ.IMPF.3SG preg-
nant
‘Gianni hypothesized that Maria was pregnant’.
31 In the following formula t(e) and t(e') indicate the time interval of the eventualities e and e'.
The symbol ‘≈’ indicates the overlapping relation.
32 In fact this property appears to relate the matrix verb in the present perfect rather than a
specific class of predicates selecting for subjunctive clauses. Another class of verbs requiring
a subjunctive complement though triggering the DAR and disallowing CD are the directive
predicates, like dire ‘to say’ under the meaning ‘to order’, ordinare ‘to order’, etc.

SUBJUNCTIVE MOOD AT THE INTERFACE
89
b. Gianni ha ipotizzato che Maria sia incinta.
Gianni has hypothesized that Maria AUX.SUBJ.PRES.3SG preg-
nant
‘Gianni has hypothesized that Maria is pregnant’.
In sentence (7)a, the embedded eventuality is interpreted as simultaneous (or
anterior) with respect to the attitude episode (John’s hypothesizing). In sen-
tence (7)b, the embedded eventuality is interpreted as simultaneous with re-
spect to the attitude episode, but it must also be interpreted as holding ‘now’,
that is, it must be simultaneous with respect to the utterance time. For sen-
tence (7)b to be true, it must be the case that the embedded eventuality
(Maria’s pregnancy) holds at the time when Gianni hypothesizes that p and
at the time of the utterance (see (8)b). The truth conditions of sentence (7)a,
on the other hand, do not include the reference to the utterance time (see
(8)a)33:
(8) a. ∃e∃e' hypothesize(G, [be-pregnant(M, e')], e) & t(e) < now &
t(e) ≈ t(e')
b. ∃e∃e' hypothesize(G, [be-pregnant(M, e')], e) & t(e) < now &
t(e') ≈ t(e) ≈ now
The interpretation sentence (7)b displays DAR, since the embedded even-
tuality is temporally evaluated both with respect to the matrix eventuality,
and with respect to the utterance time – there is a double ‘access’ to the tem-
poral information of the embedded eventuality.
Giorgi and Pianesi show that only in example (7)a is CD available:
(9) a. Gianni ha ipotizzato (che) fosse incinta.
Gianni has hypothesized (that) AUX.SUBJ.IMPF.3SG pregnant
‘Gianni hypothesized (that) she was pregnant’.
33 These two types of interpretation arise not only with hypothesize, but also with the other
verbs selecting for subjunctive clauses in the present perfect – conveying an ‘aoristic’/’in-
cohative’ interpretation:
(i) a. Quando Gianni ha visto Maria, ha pensato che sia incinta.
When Gianni has seen Maria, he has thought that be.SUBJ.PRES.3SG pregnant.
‘When Gianni has seen Maria, he thought she is pregnant’
b. Quando Gianni ha visto Maria, ha pensato che fosse incinta.
When Gianni has seen Maria, he has thought that be.SUBJ.PRES.3SG pregnant.
‘When Gianni has seen Maria, he has thought she was pregnant’.
The former sentence, which involves a present subjunctive, is felicitous if the state of being
pregnant overlaps both to John’s thinking and now, thus displaying DAR, whereas the latter
sentence, which involves an imperfect subjunctive, must be simultaneous to John’s thinking,
but not necessarily with the utterance time:
(ii) a. ∃e∃e' think(G, [be-pregnant(M, e')], e) & t(e) ≈ t(e') ≈ now
b. ∃e∃e' think(G, [be-pregnant(M, e')], e) & t(e) ≈ t(e')

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
90
b. Gianni ha ipotizzato *(che) sia incinta.
Gianni has hypothesized *(that) AUX.SUBJ.PRES.3SG prgnant
‘Gianni has hypothesized that she is pregnant’.
Thus, CD cannot obtain if DAR does, and vice versa.
d) The occurrence of a topic or a focus within the embedded clauses de-
grades CD:
(10) a. Gianni crede che la mela, Maria l’abbia mangiata.
Gianni believes the apple, Maria it.CL has eaten
‘Gianni believes Maria has eaten the apple’.
a'. (?)?Gianni crede la mela, Maria l’abbia mangiata.
Gianni believes the apple, Maria it.CL has eaten.
b. Gianni crede che LA MELA, Maria abbia mangiato, non la pe-
ra.
Gianni believes that THE APPLE, Maria has eaten, not the pear
‘Gianni believes that Maria has eaten the APPLE, not the pear.
b'. ?*Gianni crede LA MELA, Maria abbia mangiato, non la pera.
Gianni beleves THE APPLE, Maria has eaten, not the pear
The presence of a right dislocated phrase, however, does not affect the
status of a sentence:
(11) a. Gianni crede tu l’abbia mangiata, la mela.
Gianni believes you it.CL has eaten, the apple
‘Gianni believes you have eaten the apple’.
b. Gianni crede tu abbia mangiato LA MELA, non la pera.
Gianni believes you have eaten the APPLE, not the pear
‘Gianni believes you have eaten the APPLE, not the pear’.
5.2.2 Common properties of Complementizer Deletion and obviation
Comparing the above survey of the distribution of CD phenomena with the
syntax of obviation, it becomes clear that the two phenomena are affected by
the same syntactic properties. They both occur in subjunctive clauses only;
they are both sensitive to the presence of a topic, of a focus, and to clause
dislocation – both these factors rule CD out and weaken obviation. As for
the relation between DAR and obviation, see section 5.2.6.

SUBJUNCTIVE MOOD AT THE INTERFACE
91
5.2.3 CP-layer, AgrP and Complementizer Deletion
Building on the properties of CD in Italian, Giorgi and Pianesi (1997) pro-
pose that the Italian complementizer che lexicalizes two different sets of fea-
tures: those of the indicative complementizer, C, and those of the subjunctive
complementizer, which they dub ‘MOOD’:
(12) a. … [CP [C che] [AgrP [Agr … ] …
b. … [MOODP [MOOD che] [AgrP [Agr … ] …
They propose that only the complementizer che heading MOODP can be
‘deleted’. ‘Deletion’ is in fact a syntactic mechanism stating that more fea-
tures are lexicalized in a single morpheme. This mechanism is an implemen-
tation of a grammatical principle, which they call ‘Feature Scattering Princi-
ple’, according to which each feature can head a projection, though it needs
not. When it does not head a projection, it is realized in another head, which
‘syncretically’ encodes at least two features.
They hypothesize that CD is a case of syncretism: subjunctive mor-
phology may lexicalize both the feature of the MOOD head, and the features
of the subject agreement head ‘Agr’:
(13) … [MOOD/AgrP [MOOD/Agr … ] …
This option may be chosen, if there are no additional, non-inherent features,
encoded in the complementizer che, which makes it necessary to prefer the
non-syncretic option.
Under Giorgi and Pianesi’s (1997) theory, the facts discussed in the pre-
vious section are accounted as follows. First, the contrast between indicative
and subjunctive is due to the fact that MOOD may not head a maximal pro-
jection, whereas C must do so. Thus, the indicative C can never be deleted,
that is, realized syncretically with agreement features.
The contrast between volitional and epistemic subjunctive clauses and
emotive-factive subjunctive clauses follows from the hypothesis that factiv-
ity is encoded in C, which cannot be deleted. Under this hypothesis, the left
periphery of emotive-factive subjunctive clauses should contain both C and
MOOD (syncretically realized as MOOD/Agr):
(14) [CP [C che[+fact]] [MOOD/AgrP …
In dislocated clauses, MOOD also carries additional, non-intrinsic fea-
tures – the features connected to the semantics of the dislocation, which can-
not be realized syncretically, thus disallowing the omission of C.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
92
The presence of additional, non-intrinsic features also explains why topi-
calization and focusing rule CD out: MOOD carries non-intrinsic features
that cannot be lexicalized through a syncretic category; moreover, CP must
be split à la Rizzi (1997) to make room for a topic or a focus XP34:
(15) [CP [C che] [TopicP [FocusP [MOODP [AgP…
5.2.4 CP, MOODP, and DAR
In further research, Giorgi and Pianesi (2001, 2004) propose that the inter-
face properties of the two complementizer C and MOOD are even responsi-
ble for the facts concerning DAR, and for the fact that in embedded clauses
Sequence of Tenses (for short, SoT) – that is, the morphological agreement
between the matrix and the embedded tense – obtains.
Since C cannot be omitted in contexts instantiating the DAR, Giorgi and
Pianesi claim that some of the features conveying the DAR must reside in C.
Particularly, following Higginbotham’s (1995, 2001) idea that the attitude
episode must be represented within the content of the attitude itself in order
to anchor it and that SoT is the morpho-syntactic counterpart of this seman-
tic requirement, they hypothesize that at the syntax-semantics interface, C
contains the temporal features of the indicative tenses, which deliver the ut-
terance time, whereas T or, for the subjunctive, MOOD contain the coordi-
nates of the bearer of the attitude:
(15) a. [CP [C-utterance che] [AgrP [TP [T-attitude … ] …
b. [MOODP [MOOD-attitude che] [AgrP [TP [T … ] …
Thus, C provides the temporal relation between the interval of the em-
bedded eventuality t(e') and now represented in the logical forms discussed
above, whereas T and MOOD provide the temporal relation between the in-
tervals of the embedded eventuality t(e') and that of the attitude episode t(e).
5.2.5 Subjunctive tenses as anaphors
The accounts of obviation based on the Binding Theory have often built on
the tenet that subjunctive tenses are ‘anaphoric’. The interface properties of
subjunctive clauses as here analyzed help in defining more appropriately this
notion.
34 The fact that the two complementizers cannot be both lexicalized is assumed to be a phono-
logical requirement of Italian. It does not hold in some Northern Italian dialects (see Poletto
1995, Paoli 2003).

SUBJUNCTIVE MOOD AT THE INTERFACE
93
The pieces of evidence that has been given in support of the claim that
subjunctive tenses are anaphoric are mainly two: first, subjunctive cannot
occur in a root clause having declarative force and, second, SoT is enforced
in subjunctive clauses:
(16) a. *Gianni parta domani.
Gianni leave.SUBJ.PRES.3SG tomorrow
b. *Gianni pensa che Maria partisse domani.
Gianni thinks that Maria leave.SUBJ.IMPF.3SG tomorrow
ok: parta (leave.SUBJ.PRES.3SG)
c. *Gianni pensava che Maria parta domani.
Gianni thought that Maria leave.SUBJ.PRES.3SG tomorrow
ok: partisse (leave.SUBJ.IMPF.3SG)
The first evidence does not seem to be compelling, however. Subjunctive
verbs do occur in a root clause, on condition that the clause has an impera-
tive or optative force:
(17) a. (Che) partano immediatamente!
(That) leave.SUBJ.PRES.3PL at once
‘They leave at once!’
b. Se solo fosse qui Gianni!
If only be.SUBJ.IMPF.3SG here Ganni
‘If only Gianni were here’.
Thus, the fact that subjunctive verbs do not occur in declarative root sen-
tences appears to be linked to the modal semantics of the subjunctive, rather
than to its temporal interpretation in itself.
The facts concerning SoT appear to be more substantial, since the sub-
junctive does in fact contrast with the indicative in this respect. The tense of
an indicative clause is independent of the matrix tense:
(18) a. Gianni sostiene che Maria partiva domani.
Gianni claims that Maria leave.IND.IMPF.3SG tomorrow
‘Gianni claims that Maria was expected to leave tomorrow’.
b. Gianni sosteneva che Maria parte domani.
Gianni claimed that Maria leave.IND.PRES.3SG tomorrow
‘Gianni claimed that Maria is expected to leave tomorrow’

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
94
These examples contrast minimally with examples (16)b and c. In (18)a and
(16)b, the SoT present-imperfect occurs; the former sentence is grammatical,
the latter is unacceptable. In (18)b and (16)c, the SoT imperfect-present
holds; again, only the former is grammatical. The grammatical sentences, to
be sure, contain an indicative embedded clause. Thus, the second piece of
evidence in support to the claim that subjunctive tenses are anaphoric ap-
pears to be uncontroversial35.
Three counterarguments to the claim that subjunctive tenses are ana-
phoric may be rejected. The first one is the fact that embedded clauses in
which the subjunctive verb is not instantiated uniquely by the matrix tense,
do exist. The examples in (7) are a case in point, although it must be noted
that they involve a present perfect indicative attitude predicate, a tense that
in the Northern varieties of Italian, is ambiguous between a present (perfect)
and a real past. Under the former option, the present tense agreement is in-
stantiated, whereas under the latter, the imperfect tense agreement is instan-
tiated. Thus, the argument that DAR obtains in some cases does not appear
to be valid.
The second (apparent) piece of evidence against the claim that strict SoT
is enforced in subjunctive clauses is discussed in Giorgi (2009) and Costan-
tini (2007):
(19) Il testimone crede che ieri alle 5 l’imputato fosse a casa.
The witness believes that yesterday at 5 the defendant
be.SUBJ.IMPF.3SG at home.
‘The witness thinks that yesterday at 5 the defendant was at home’.
35 The argument Giorgi and Pianesi (1997, 2001) adopt to claim that subjunctive tenses are
anaphoric – that is, the coordinate of the utterance are not represented in a subjunctive clause
– do not seem to be fully convincing. They take as evidence for this claim the fact that sub-
junctive clauses are interpreted with respect to the attitude time frame, though not with respect
to the utterance time:
(i) Gianni pensava che Maria fosse incinta.
Gianni believed that Maria be.SUBJ.IMPF pregnant
‘Gianni thought Maria was pregnant’.
This sentence can be considered true if the state of pregnancy holds at the time of Gianni’s
saying (or if the state of pregnancy was anterior to Gianni’s saying, if the context provides a
salient time reference). But there is no time relation specified between the embedded eventu-
ality and the utterance time. Thus, Giorgi and Pianesi conclude that a subjunctive clause does
not give access to the speaker’s point of view.
However, the lack of a deictic anchoring is a property of the imperfect tense, rather than
of the subjunctive mood. The imperfect indicative does not instantiate DAR either:
(ii) Gianni ha detto che Maria era incinta.
Gianni has said that Maria be.IND.IMPF.3SG pregnant
‘Gianni has said that Maria was pregnant’.
Here too, the state of pregnancy needs not stretch over the utterance time, though it must
overlap with the Gianni’s speech act.

SUBJUNCTIVE MOOD AT THE INTERFACE
95
Here the matrix verb is present indicative, the embedded verb is imperfect
subjunctive, which apparently violates a strict SoT. Costantini (2007) shows
however that the imperfect subjunctive can occur under a present matrix
only if the embedded eventuality is stative, either lexically, or as a result of a
type-shift through adverbial or aspectual modifiers, and is shifted in a time
interval prior to the attitude episode. When the embedded predicate denotes
an event, the strict SoT is enforced – the perfect aspect possibly sets the
eventuality in an interval preceding the attitude eventuality:
(20) *Il testimone crede che ieri l’imputato partisse.
The witness thinks that yesterday the defendant leave.SUBJ.IMPF
Finally, if the matrix verb is present, the embedded eventuality is tempo-
rally accessed from the attitude episode time and from the utterance time,
which coincide. However, it is not possible to establish whether the DAR
obtains or whether the only available access is that from the attitude episode,
which indirectly sets the embedded eventuality with respect to the utterance
time.
To conclude, subjunctive tenses may be truly considered ‘anaphoric’,
keeping this term as a label underlying the unavailability of DAR.
5.3 Long-distance Anaphors
5.3.1 The syntax of LDAs
Giorgi (2004, 2006, 2007) investigates the phenomenon of long-distance
binding of anaphors. She proposes that their distribution follows from the
same apparatus ruling the temporal interpretation of subjunctive sentences.
In Italian, LDAs are indeed sensitive to the mood of the embedded clause
and are subject-oriented, or rather de se.
As for the first property, in Italian the LDA proprio ‘one’s own’ can ap-
pear in subjunctive (and conditional) clauses, but it is ruled out in indicative
clauses (‘verbal blocking effect’, Giorgi 2006. See also Giorgi 1983):
(21) a. [Quel dittatore]1 spera che i notiziari televisivi parlino a lungo
delle proprie1 gesta.
[That dictator]1 hopes that the TV news talk.SUBJ.PRES.3PL for a
long time about self’s deed.
‘[That dictator]1 hopes that the TV news will talk about his1
deeds for a long time’.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
96
b. *[Quel dittatore]1 ha detto che i notiziari televisivi hanno parla-
to a lungo delle proprie1 gesta.
[That dictator]1 has said that the TV news AUX.IND.PRES.3PL
talked for a long time about self’s deed.
Giorgi (2006) observes that LDAs are sensitive to some nominal phrases
as well. First and second person pronouns degrade a sentence in which a
LDA occurs (‘nominal blocking effect’):
(22) a. Gianni1 pensa che tutti siano innamorati della propria1 moglie.
Gianni1 believes that everybody be.SUBJ.PRES.3PL in love with
self’s1 wife
‘Gianni1 believes that everybody is in love with his1 wife’.
b. Gianni1 pensa che Mario sia innamorato della propria1 moglie.
Gianni1 believes that Mario be.SUBJ.PRES.3SG in love with his1
wife
‘Gianni1 believes Mario is in love with his1 wife’.
c. ?*Gianni1 pensa che io sia innamorato della propria1 moglie.
Gianni1 believes that I be.SUBJ.PRES.1SG in love with his1 wife
‘Gianni1 believes I am in love with his1 wife’.
d. ?*Gianni1 pensa che tu sia innamorato della propria1 moglie.
Gianni1 believes that you be.SUBJ.PRES.2SG in love with his1
wife
‘Gianni1 believes you are in love with his1 wife’.
Examples (22)a and b contrast with examples (22)c and d in that the embed-
ded subject is in the third person and LDA binding is possible. In the latter
sentences, the embedded subject is in the second and in the first person, re-
spectively, and long distance binding is almost unacceptable.
Another well-known property of LDAs is that they are subject-oriented
(Giorgi 1983, 2004, Pica 1987, Sells 1987, Cole, Hermon and Sung 1990).
Thus, in the following sentence proprio can be bound by Gianni, though not
by Maria:
(23) Gianni1 ha detto a Maria2 che i notiziari televisivi avrebbero parla-
to del proprio1/*2 libro.
Gianni1 has told to Maria2 that the TV news would-have talked of-
the self’s1/*2 book.
‘Gianni1 told Maria2 that the TV news would talk of his1/*2 book’.

SUBJUNCTIVE MOOD AT THE INTERFACE
97
The above sentence contrasts minimally with the following with respect to
the interpretation of proprio:
(24) Gianni1 ha detto a Maria2 che i notiziari televisivi avrebbero parla-
to del suo1/2 libro.
Gianni1 has told to Maria2 that the TV news would-have talked of-
the his1/2 book.
‘Gianni1 told Maria2 that the TV news would talk of his1/2 book’.
In this sentence both Gianni and Maria may serve as antecedents for the
pronominal suo, whereas in sentence (23) the only available antecedent for
proprio is the matrix subject.
LDAs may be bound by arguments that do not serve as subject. Consider
for instance the following sentence:
(25) Che la propria1 figlia sia andata in campeggio da sola preoccupa
Gianni1.
That the self’s daughter is gone to camping by herself worries
Gianni
‘That his1 daughter is camping by herself worries Gianni1’.
LDAs can be bound by the object argument of psych-verbs. Hence, although
LDAs are more often subject-oriented, they are not always so. In chapter 3
we have noticed that the individual argument of psych-verbs refers to the
bearer of attitude, the individual to which an attitude, possibly de se, is at-
tributed. LDAs, rather than being subject-oriented, have indeed been shown
to be de se (Chierchia 1989, Pan 1998, 2001, Huang-Liu 2001)36.
36 To show this, Chierchia (1989) considers the following sentence (originally discussed by
Kaplan 1989):
(i) John believes that his pants are on fire.
Sentence (i) is ambiguous. It can be de se – the speaker might be saying that John has a belief
about himself; but it can also be considered true as non-de se – in a scenario in which John is
looking at a mirror and is seeing a person whose pants are on fire. The person he is seeing is
in fact John himself, but he does not realize it. Of course, a third interpretation is also avail-
able, namely, John is quite aware of who is who and has a de re attitude which does not in-
volve himself.
In Italian there are two possible sentences corresponding to sentence (i):
(ii) a. Gianni crede che i suoi pantaloni siano in fiamme.
Gianni believes that the his pants are on fire
‘Gianni believes his pants are on fire’.
b. Gianni crede che i propri pantaloni siano in fiamme.
Gianni believes that the self’s pants are on fire
‘Gianni believes that his own pants are on fire’.
Sentences (ii)a and b contrast minimally whereby the former is ambiguously de se or non-de
se. Gianni might have two different beliefs: he might be completely aware that the person
whose pants are on fire is himself; or, he might not be aware that the object of his belief is

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
98
Finally, LDAs can appear in adverbial clauses:
(26) Il primo ministro1 sperava che il dittatore2 partissse prima che i ri-
voluzionari sequestrassero il proprio1/*2 patrimonio.
The PM1 hoped that the dictator2 left before that the revolutionaries
sequestered self’s1/*2 patrimony
‘The PM1 hoped that the dictator2 would leave before the revolu-
tionaries sequestered his own1/*2 patrimony’.
In this sentence the matrix subject, though not the subject of the argument
clause, is available as an antecedent for proprio37.
5.3.2 Common properties of LDAs and obviation
LDAs appear to share some properties with obviation. They concern the dis-
tribution of the two phenomena, although the behavior of the two in the
same environment is the opposite. Thus, they are both sensitive to mood, in
that they are instantiated in subjunctive clauses, though not in indicative
clauses. The presence of a first (or second) person embedded subject appears
to block slightly both LDA-binding and obviation. They can obtain in adver-
bial clauses.
Moreover, they have contrasting properties as far as their interpretation is
concerned: LDAs are de se, whereas obviation has been defined as the un-
availability of this interpretation. In a way, this observation comes out as no
surprise, since LDAs are anaphors, whereas obviation concerns (subject)
pronouns. Thus, in the light of Binding Theory, the fact that in the same en-
vironment they display complementary properties is expected.
himself. On the other hand, sentence (ii)b is unambiguously de se. This observation leads
Chierchia to conclude that proprio patterns just like the null subject of infinitival clauses,
PRO, and might be viewed as a phonologically realized counterpart of PRO.
This claim seems however to be too strong, as Giorgi (2004) points out, since it suggests that
proprio is inherently de se. But this does not seem to be the case. Consider the following
example:
(iii) Gianni spera che Maria recuperi i propri soldi.
Gianni hopes that Maria recovers.SUBJ the self’s money
‘Gianni hopes Maria recovers his money’/’Gianni hopes Maria recovers her money’.
If proprio were intrinsically de se, it might not corefer with Maria. Hence, either proprio is
necessarily de se only when it is bound outside its clausal domain (LDA), though not in a
local domain, or there are two proprio – one locally bound, the other long-distance bound.
Only the latter is de se.
37 Irrelevantly, the subject of the adverbial clause, as a coargment, can bind proprio.

SUBJUNCTIVE MOOD AT THE INTERFACE
99
5.3.3 LDAs and indexicality
Giorgi (2006) hypothesizes that the speaker’s coordinates delimits the do-
main of LDA binding. More in general, she postulates a ‘blocking condition’
that states that an event located with respect to the speaker’s coordinate must
be fully saturated – where saturation is to be intended in a Fregean sense.
Following Higginbotham’s (1997), she hypothesizes that LDAs are im-
plicit or mixed anaphora – that is, unsatisfied theta-positions that must be
assigned through theta-identification. Before illustrating Giorgi’s hypothesis,
let us consider Higginbotham’s notion of theta-identification and implicit
anaphora.
Higginbotham (1985) observes that a series of examples show that part b
of Chomsky’s (1981) Theta-criterion is simplistic:
(27) Theta-criterion
a. Every argument is assigned one and only one thematic role;
b. Every thematic role is assigned to one and only one argument.
Consider now the following example:
(28) The boat was sunk [PRO to collect the insurance] (Manzini 1983)
It means, by and large, that whoever sank the boat, did it in order to collect
the insurance. The predicate sink specifies two theta-roles, agent and theme,
of which only the theme is satisfied (it is discharged by the subject the boat).
The agent role is not assigned to an argument (although it is active as the
controller of PRO). The fact that the agent is not assigned to an argument
violates part b of the Theta-criterion.
Consider now the following example:
(29) the dog
In many language nouns can serve as predicates and in semantics nouns are
predicates of the type <e,t> (Heim and Kratzer 1997). Thus, the noun dog is
supposed to have a theta-grid in its lexical entry, containing one position,
which Williams (1980) calls ‘R’ (referential). When the noun heads an NP,
that position is not assigned to any argument. This also seems to violate part
b of the Theta-criterion.
Finally, consider the following example:
(30) This is a big butterfly
It does not simply mean ‘this is a butterfly and this is big’. Rather, it can be

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
100
paraphrased as ‘this is a butterfly and this is big for a butterfly’. The adjec-
tive big grades the noun butterfly with respect to butterflies, not in general
terms (even a big butterfly is not a big thing). Thus, it is supposed to have
two thematic positions, one of which must be satisfied by the attribute the
adjective grades (the other is the theta-position any predicate has, ‘R’). Part
b of the Theta-criterion does not help in explaining how the attribute theta-
role is assigned.
To fill these gaps in the Theta-criterion, Higginbotham proposes that
theta-marking is not the only way of discharging a theta-role. He argues that
there are three more ways of discharging a theta-role: theta-binding, theta-
identification, and ‘autonymous’ theta-marking.
Example (28) illustrates the first procedure. The determiner the is sup-
posed to ‘bind’ the theta-position of the noun and, in this way, to discharge
the theta-position. Example (29) illustrates the second and the third way of
discharging a theta-role. Both the adjective and the noun can be predicates,
and must accordingly have a theta-position. However, the whole constituent
a big butterfly is a predicate and, consequently, must have a unique theta-
position. Therefore, the theta-position of the noun and that of the adjective
are supposed to be ‘identified’ and passed to the whole constituent38. More-
over, the adjective has a second theta-position, because it grades the noun it
modifies with respect to the attribute the noun denotes. Hence, the second
theta-position of the adjective is discharged by the noun it modifies.
In the light of these data, Higginbotham substitutes the second statement
of the Theta-criterion with the following, more general, one:
(31) Every thematic role is discharged.
In further research, Higginbotham (1997) discusses the properties of ‘im-
plicit apnaphora’ and ‘mixed anaphora’. The former term refers to the case
in which the antecedent and the anaphora are both unsatisfied theta-position
or implicit arguments (that is, a theta-position that is not satisfied through
obvious theta-marking). Consider for instance the following example:
(32) A self-starting motor
The theta-grid of the adjective self-starting contains two theta-positions
(agent and theme) that are identified. Thus, a self-starting motor is ‘an x such
that x is a motor and x starts x’.
As for the term ‘mixed anaphora’, it refers to the case when a binding rela-
tion holds between an explicit antecedent and an unsatisfied theta-position. He
38 Theta-assigning is assumed to hold under sisterhood. This condition may be satisfied by
means of percolation (Williams 1987) – that is, non-discharged positions are passed up to the
containing XP, if that XP is a predicative expression, until they are discharged.

SUBJUNCTIVE MOOD AT THE INTERFACE
101
claims that Obligatory Control relations may be analyzed as instances of mixed
anaphora, taking PRO to be the spell-out of an unsatisfied theta-position.
Let us now go back to Giorgi’s theory of LDA. Giorgi’s hypothesis states
that LDAs are unsatisfied theta-positions that must be assigned through
theta-identification. Her hypothesis states that they may be theta-identified
either with an argument of the embedded clause (a co-argument) or with the
author of the context, which must be locally represented (within MOODP)39.
Thus, theta-identification brings the de se reading about. Moreover, due to
the ‘blocking condition’, theta-identification must obtain within the domain
of the speaker’s coordinates, otherwise the embedded predicate would be
unsaturated. Finally, the theta-position so satisfied is spelled out as the adjec-
tive proprio40. Thus, a theory of LDAs can be summarized as follows:
(33) Giorgi (2004, 2006, 2007) theory of LDAs
A. LDA-binding
a. LDA is the spell-out of an unsatisfied position;
b. A LDA can be satisfied either
i. by a co-argument, or
ii. by the bearer of the attitude;
B. Blocking condition
An event located with respect to the speaker’s coordinate must
be fully saturated.
Giorgi claims that under this theory, the distribution of LDAs follows
straightforwardly. First, LDAs must be de se. Consider the following con-
figuration:
(34) ca … V [MOODP [XP … [Th …
The unsatisfied theta-position Th may be identified with the local antecedent
XP; however, if XP is not the intended antecedent, MOODP is unsaturated
and Th may be identified with the bearer of the attitude ca. Finally it is
spelled-out as proprio. This is what happens in the following sentence:
(35) Gianni1 crede che Mario2 odi la propria1/2 moglie.
Gianni1 believes that Mario2 hates self’s1/2 wife.
‘Gianni1 believes that Mario2 hates his1/2 wife’.
39 In Cinque’s (1999) hierarchy of verbal functional projections (concerning the higher IP-
field), the highest projections, speech act mood, evaluative mood, and evidential mood, are
related with the pragmatic roles associated with a sentence – speaker and attitude bearer. This
is also what Speas (2002) suggests. According to these points of view, the speaker and the
attitude bearer should be represented in one of these heads.
40 Implicit here is Halle and Marantz’s (1993) Distributed Morphology framework.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
102
The anaphor may be bound either by Gianni or by Mario. Mario is indeed a
co-argument of the (clause-bound) anaphor (see clause (b)i in (33)). Though,
if this does not convey the appropriate meaning, the unsatisfied theta-
position may even be discharged through theta-identification with the bearer
of the attitude.
In multiple embedding, this mechanism works recursively:
(36) ca … V [MOODP ca' … [MOODP XP … [Th …
Here, Th can be identified with XP, with ca', or with ca, depending on the
intended meaning. Thus, the following sentence is also expected:
(37) Gianni1 supponeva che Mario2 pensasse che Paolo3 odiasse la pro-
pria1/2/3 moglie.
Gianni1 supposed that Mario2 believed that Paolo3 hated the
self’s1/2/3 wife.
‘Gianni1 supposed that Mario2 believed that Paolo3 hated his wife’.
The anaphor proprio may be bound by Gianni, Mario, or Paolo. Paolo is
locally available as an antecedent. If it is not the intended antecedent, the
anaphor may be theta-identified with the attitude bearer Mario. If Mario is
not the intended antecedent either, the anaphor may be identified with the
matrix argument referring to the attitude bearer.
Second, the asymmetry between indicative and subjunctive clauses fol-
lows from (33). LDAs cannot occur in indicative clauses, since the speaker’s
coordinate intervenes (in C) and, by the blocking condition, the embedded
predicate must be fully saturated when C is merged:
(38) ca … V [CP [XP … [Th …
Thus, in the following sentence the anaphor proprio can only be locally
bound:
(39) Gianni1 ha detto che Maria2 ama la propria*1/2 madre.
Gianni1 has said that Maria2 loves the self’s*1/2 mother
‘Gianni1 said Maria2 loves her own mother’.
Maria is an available antecedent, whereas Gianni is not, because the
speaker’s coordinate in the embedded C requires that all theta-position be
satisfied. This may be done only if theta-identification with a co-argument
discharges an otherwise unsatisfied theta-position. The contrast between the
above sentence and the following ones is also predicted:

SUBJUNCTIVE MOOD AT THE INTERFACE
103
(40) a. Gianni1 ha detto che Mario2 pensa che Paolo3 odi la propria*1/2/3
moglie.
Gianni1 has said that Mario2 think.IND.PRES.3SG that Paolo3
hate.SUBJ.PRES.3SG the self’s*1/2/3 wife
‘Gianni1 has said that Mario2 thinks that Paolo3 hates his*1/2/3
wife’.
b. Gianni1 ha detto che Mario2 sa che Paolo3 odia la propria*1/2/3
moglie.
Gianni1 has said that Mario2 know.IND.PRES.3SG that Paolo3
hates.IND.PRES.3SG the self’s*1/*2/3 wife
‘Gianni1 has said that Mario2 knows that Paolo3 hates his*1/*2/3
wife’.
In sentence (40)a, Mario and Paolo are available antecedents of proprio,
whereas Gianni is not: the speaker’s assignment sequence intervenes in the
intermediate C and ‘closes’ the binding domain for the anaphor proprio. In
sentence (40)b, Paolo is the only available antecedent for proprio – the most
embedded clause contains the utterance coordinates and the unassigned
theta-position can only be discharged by a co-argument.
Finally, the facts concerning adverbial clauses are also accounted for:
(41) ca … V [MOODP [XP … [AdvCP YP … [Th …
Here XP is not the agent of the context, nor is it a coargument. Thus, the
only possible antecedent for Th is the matrix subject. This accounts for the
peculiar interpretation of proprio in examples like (26).
5.4 Conclusion
In this chapter we have shown the similarities and the differences between
CD and obviation and LDA-binding and obviation. Obviation, CD, and
LDAs appear to be sensitive to the same syntactic and semantic factors. The
descriptive properties of CD and LDAs appear to follow from the interface
properties of subjunctive clause. Assuming that the similarities (and the dis-
similarities) between obviation and the other phenomena are not random, we
will pursue the hypothesis that the same machinery ruling CD and LDAs
(but DAR and SoT as well) is also responsible for the characteristics of ob-
viation.
As a starting point, we summarize the generalization concerning the phe-
nomena discussed in the present chapter in the following table:

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
104
(42) DAR CD LDA Obviation
V [C … ok * * *
V [C [MOOD … ok * ok41 ok
V [MOOD … * ok ok (%)ok/?/??
The de se reading of the embedded subject is unavailable in environments
involving MOOD, the subjunctive complementizer, no matter whether C
occurs as well. Thus, its distribution overlaps with that of LDAs42. CD and
obviation are also affected by the presence of C43.
Apart from the syntactic properties analyzed here, however, obviation
appears to be sensitive to other characteristics, which do not seem to depend
on mood selection – the presence of a ‘functional’ verb in the subjunctive,
for instance, the theta-role assigned to the embedded subject, which seem-
ingly require an independent account.
41 The distribution of LDAs is not sensitive to the presence of C (here indicated by the pres-
ence of a topic):
(i) [Quel dittatore]1 spera che questa sera i notiziari televisivi parlino a lungo delle proprie1
gesta.
[That dictator]1 hopes that this evening the TV news talk.SUBJ.PRES.3PL for a long time
about self’s deed.
‘[That dictator]1 hopes that this evening, the TV news will talk about his1 deeds for a long
time’.
42 Though, while LDAs (as well as CD) can occur in an embedded clause in the conditional,
obviation is not required in this type of sentence.
43 I note however that the relations between obviation and CD are not completely clear. Con-
sider again the examples in (7). The first one involves an imperfect subjunctive, which is in-
terpreted only with respect to the attitude episode, the second example involves a present
subjunctive, which enforces the DAR. If the behavior of obviation parallels that of CD, we
would expect that obviation should obtain in the latter case, but not in the former. However,
unlike CD, the de se reading is more acceptable in sentences having an imperfect rather than a
present subjunctive (as we have noted in section 3.2.1):
(i) a. Maria1 ha ipotizzato che pro(?)1/2 fosse incinta.
Maria1 has hypothesized that pro(?)1/2 AUX.SUBJ.IMPF.3SG pregnant
‘Maria hypothesized that she was pregnant’.
b. Maria1 ha ipotizzato che pro??1/2 sia incinta.
Maria1 has hypothesized that pro??1/2 AUX.SUBJ.PRES.3SG pregnant
‘Maria has hypothesized that she is pregnant’.
Apparently, these two sentences have a different status, which suggests that DAR affects even
obviation. The point is how it does affect the possibility of coindexation.
Note that coindexation is even more acceptable if CD obtains:
(ii) Maria1 ha ipotizzato pro?1/2 fosse incinta.
Maria1 has hypothesized that pro(?)1/2 AUX.SUBJ.IMPF.3SG pregnant
‘Maria hypothesized that she was pregnant’.
Thus, clauses displaying the DAR, while ruling CD out, do require obviation (that is, they
prevent the de se reading), whereas clauses that are devoid of the deictic anchoring, license
the de se reading (and CD) more easily. This is surprising, since other syntactic characteristics
such as topicalization, focalization and clause dislocation, affect CD and obviation in a paral-
lel way.

SUBJUNCTIVE MOOD AT THE INTERFACE
105
In the following chapter I will attempt to account for all the properties
obviation displays. In so doing, I will exploit the theoretical apparatus de-
scribed in the present chapter, resorting additional principles to account for
the properties that do not depend on mood.

6 An interface proposal for subjunctive obviation
6.1 Introduction: desiderata for a theory of obviation
In this chapter I will work out a proposal to account for the facts I discussed
in chapters 2 and 3 in view of the theoretical notions discussed in chapter 5.
As we have seen, obviation – that is, the unavailability of the de se read-
ing – obtains only in subjunctive clauses – indicative, conditional, and infini-
tival clauses do not trigger it. It involves only the subject (pro, in Italian) of a
subordinate clause and the argument of the superordinate clause referring to
the person to whom a certain attitude toward a propositional content is at-
tributed. It obtains strictly if the form carrying subjunctive morphology is a
full verb, whereas if the form carrying subjunctive morphology is a tense
auxiliary, a passive auxiliary, or a modal verb, obviation tends not to occur.
Additional factors, such as theta-role, number, implicitness or explicitness of
the embedded subject, intervening material in the left periphery of the sub-
ordinate clause, clause dislocation, coordination, affect the availability of the
de se reading.
A theory of obviation in subjunctive clauses should be able to account for
all these facts. Particularly, it should be able to answer he following ques-
tions:
a. Why are subjunctive clauses different from indicative, conditional and
infinitive clauses with respect to the referential properties of their subject?
b. Why are functional verbs (that is, tense and voice auxiliaries, modal
verbs) different from full verbs with respect to the referential properties of
the subject of a subjunctive clause?
c. Why does obviation involve only matrix arguments referring to the
bearer of the attitude?
d. How do the other factors contribute the interpretative properties of the
embedded subject? Particularly, (i) how is theta-theory concerned? (ii) How

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
108
is the phonological typology of the subject pronoun (iii) How is number in-
volved? (iii) And how is information structure involved?
To answer these questions, I will build on the conceptual apparatus
worked out by Giorgi and Pianesi (2001) and Giorgi (2004, 2006, 2007),
discussed in chapter 5, here summarized:
(1) Giorgi and Pianesi’s (2001) theory of CD, DAR, and SoT:
a. Indicative clause: [CP [C-utterance che] [AgrP [TP [T-attitude … ] …
b. Subjunctive clause: [MOODP [MOOD-attitude che] [AgrP [TP [T … ] …
(2) Giorgi’s (2004, 2006, 2007) theory of LDAs
A. LDA-binding
a. LDA is the spell-out of an unsatisfied position;
b. A LDA can be satisfied either
i. by a co-argument, or
ii. by the bearer of the attitude;
B. Blocking condition
An event located with respect to the speaker’s coordinate must
be fully saturated.
In chapter 5 I have discussed some phenomena, CD, DAR, SoT, LDA-
binding, which correlate with obviation as for the environment in which they
occur and for the interpretative properties. In previous research, it has been
shown that these phenomena are accountable under a theory that includes the
above principles. The attempt to extend the above principles to the data con-
cerning obviation appears to be desirable, if not necessary.
6.2 The hypothesis
6.2.1 Generalities
Building on the above theoretical apparatus I propose that the following hy-
pothesis, which is an extended version of Giorgi’s LDA binding, is able to
explain the facts concerning obviation:
(3) Hypothesis
The de se interpretation obtains if and only if an unsatisfied posi-
tion is theta-identified with the agent of the context.
I assume that the above hypothesis is restricted by the Blocking Condition
(2)b. This hypothesis correctly predicts the distribution of LDAs and, as
Giorgi (2006, 2007) points out, even the interpretation of the null subject of

AN INTERFACE PROPOSAL FOR SUBJUNCTIVE OBVIATION
109
infinitival clauses, PRO, which as Higginbotham (1997) shows, is semanti-
cally compatible with an unsatisfied position (see section 5.3.3).
Furthermore, both LDAs and PRO occur in environments that are devoid
of the utterance temporal coordinates, so that the blocking condition does not
apply. This property has already been discussed with respect to subjunctive
clauses in chapter 5, but it can be easily shown that infinitival clauses do not
license the DAR as well:
(4) Maria credeva di essere incinta.
Maria believed C/P be.INF pregnant
‘Maria believed she was pregnant’
This sentence can be considered true if the state of pregnancy holds at the
time of Mary’s belief, though not necessarily at the utterance time. Thus, the
temporal coordinate of the speaker is not represented within an infinitival
clause – only the temporal coordinate of the attitude bearer is.
The above hypothesis then accounts for the fact that infinitival clauses are
de se straightforwardly. The speaker’s coordinate does not occur in an infini-
tival clause. Thus, the blocking condition does not apply and the unsatisfied
position within the infinitival clause can be theta-identified with the bearer
of the attitude, thus conveying the de se reading.
As for subjunctive clauses, the hypothesis above predicts that these
clauses cannot be de se. Small pro does not indeed spell out an unsatisfied
position. It is standardly assumed to saturate a predicate whereas PRO can-
not:
(5) a. Pro legge molti libri.
Pro reads many books
‘He/she reads a lot of books’.
b. *PRO leggere molti libri.
PRO read.
INF many books
Since pro is a syntactic formative satisfying a theta-role through obvious
theta-marking, if no implicit argument occurs in a subjunctive clause, such a
clause cannot be de se by the hypothesis proposed.
Moreover, conditional clauses do not include the coordinate of the utter-
ance. Thus, a de se reading can be achieved only if an unsatisfied position
occurs. Typically, when conveying a future in the past reading, the tense of a
conditional clause is periphrastic, composed by an auxiliary bearing condi-
tional morphology and by full verb in the past participle. Following
Higginbotham’s (1997), Roeper’s (1993), and Safir’s (1991) I propose that
the external argument of non-finite lexical verbs is not assigned a theta-role

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
110
via obvious theta-marking. If so, by hypothesis the unassigned theta-role can
be theta-identified with a co-argument or with the bearer of the attitude. In
this case, a de se reading obtains.
Note that if the full verb carries conditional morphology, a sentence may
be de se as well:
(6) Gianni ha detto che pro partirebbe (se solo pro potesse).
Gianni has said that pro leaves.COND, if only pro could.SUBJ
‘Gianni said that he would leave (if only he could)’.
Notice, however, that full verbs carrying conditional morphology are part
of an implicit conditional. In such a structure, the conditional must include
the coordinate of the utterance, since it can be uttered by itself:
(7) Gianni partirebbe (se solo pro potesse).
Gianni leaves.COND, if only pro could.SUBJ
‘Gianni would leave (if only he could)’.
This fact explains the apparent counterevidence to the hypothesis here
discussed. The domain of application of the above hypothesis includes in-
deed the clauses devoid of the coordinate of the speaker – subjunctive and
infinitival clauses. As for the clauses including the coordinate of the speaker
(indicative and present conditional clauses), the blocking condition states
that an event must be saturated if the coordinate of the utterance is intro-
duced. All theta-positions must be satisfied and the de se reading cannot be
achieved via theta-identification of an unsatisfied position. Thus, this reading
can be conveyed even by an explicit formative, like pro.
The above hypothesis thus appears to explain straightforwardly the facts
discussed in chapter 2.
6.2.2 Obviation weakening
6.2.2.1 Non-finite full verbs
Although the prediction that subjunctive clauses cannot be de se is true for
many subjunctive clauses, this is not true for all subjunctive clauses. In gen-
eral, subjunctive clauses cannot be de se if the morphology of the subjunc-
tive mood is attached to the full verb (and not to an auxiliary or to a modal
verb). Moreover, many additional factors have to be considered. Let us first
examine the case involving a full verb in the subjunctive vis-à-vis the case
involving a functional verb in the subjunctive.

AN INTERFACE PROPOSAL FOR SUBJUNCTIVE OBVIATION
111
Treating the case of conditional clauses, I have hypothesized that the ex-
ternal argument of non-finite lexical verbs is not assigned a theta-role via
obvious theta-marking. That is, if in a subjunctive clause the lexical verb is
in a non-finite form, the subject argument can only be satisfied via theta-
identification. The mechanism of LDA-binding can here be assumed to hold,
so that the unsatisfied position can be theta-identified either with a co-
argument or with the attitude bearer. I propose that this is exactly what hap-
pens in the instances of obviation ‘weakening’, which crucially involve non-
finite forms of lexical verbs.
The following structures illustrate the contrast between non-de se (obvia-
tive) subjunctive clauses and (optionally) de se subjunctive clauses:
(8) a. DP1 … [MOODP [ pro*1 V.SUBJ [pro1 V
b. DP
1 … [MOODP [pro*1 F.SUBJ [x V.INF/PART/GER
In (8)a there is no unsatisfied position. Thus, by (3) the de se reading is ruled
out. In (8)b, one of the arguments (notated as a variable x) is not satisfied
through theta-marking, because the lexical verb is non-finite. By the hy-
pothesis above, this argument is satisfied through theta-identification either
with a co-argument (pro, which cannot be de se), or with the agent of the
context. If the first option holds, the subjunctive clause is non-de se, other-
wise, it is de se. This appears to be correct, since a de se subjunctive clause
is not obligatorily de se.
Note that the hypothesis under investigation predicts that infinitival and
subjunctive clauses should be complementary (as competition theories do)
only if the verbal form carrying the morphology of the infinitive or of the
subjunctive is a lexical verb. If the verbal form carrying the morphology of
the infinitive or of the subjunctive is a functional verb, infinitive and sub-
junctive clauses are no longer in complementary distribution, since in both
clauses an unsatisfied position occurs, licensing the de se reading.
Thus, the hypothesis investigated here accounts for the mood ‘competi-
tion’ in a novel way, building on interface rather than simply on semantic
properties, and accounting for the facts concerning obviation ‘weakening’,
which represented a challenge to Farkas’ and Schlenker’s competition theo-
ries (see 3.4.2). Particularly, the following couple of examples appeared to
be problematic for a competition account on obviation:
(9) a. Spero che io possa vivere a lungo.
Hope.
IND.PRES.1SG that I can.SUBJ.PRES.1SG live long
‘I hope I can live long’.
b. Spero di poter vivere a lungo.
Hope.
IND.PRES.1SG C/P can.INF live long
‘I hope I can live long’.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
112
To summarize the point at issue, we have seen that intuitively, both sen-
tences are de se, which appears to be unpredicted. Since the condition of fe-
licity for these sentences are the same, by the pragmatic principle Maximize
presupposition! (Schlenker 2005), the infinitival clause should be the only
form appropriate in a de se scenario, whereas the subjunctive clause should
be ruled out as bringing about a semantic failure. This is not the case, how-
ever, since the subjunctive clause is appropriate in the same relevant context
as the infinitival clause.
Under the hypothesis I have proposed here, both sentences are expected
to be de se. In both sentences an unsatisfied position occurs, instantiating the
first personal interpretation: in the former sentence, the full verb is non-
finite, whereas in the latter both the modal auxiliary and the full verb are
non-finite44. Thus, in both a theta-position can be theta-identified with the
bearer of the attitude.
6.2.2.2 Topic, focus, strong pronouns, dislocated clauses
In chapter 5 we have seen that topic and focus block CD, a fact that, as
Giorgi and Pianesi (1997, 2004) maintain, is due to the activation of the CP-
field.
Let us assume that the CP introduces a reference to the utterance, since it
interfaces syntax with pragmatics and links a clause to the conversational
background – the set of propositions that two interlocutors share (see Chier-
chia and McConnell-Ginet 1990 and Chierchia 1997). Thus, it requires that
all theta-positions must be satisfied, by the Blocking Condition. If so, a
clause containing a topic or a focus should be weakly obviative. As we have
seen in section 3.6.1, the data are compatible with this proposal.
The case is on the whole similar to indicative clauses: no unsatisfied posi-
tion can occur within clauses having a topic or a focus, so the de se reading
must be conveyed through alternative strategies than theta-identification of
an unsatisfied position. This implies that the presence of a topic or of a focus
should improve the availability of a de se reading even if no additional fea-
tures weakening obviation occurs. This prediction appears to be correct:
(10) a. Gianni1 sperava che pro*1 rispondesse a tutte le domande.
Gianni1 hoped that pro*1 answered to all the questions
‘Gianni1 hoped he*1 answered all the questions’.
44 Note also that in the above example the first person is concerned. As we have seen (section
3.3.1.2) this appears to matter. I will show in section 6.3.2.3 that the hypothesis discussed
here is able to account for this fact quite naturally.

AN INTERFACE PROPOSAL FOR SUBJUNCTIVE OBVIATION
113
b. Gianni1 sperava che domani, all’esame di linguistica computa-
zionale, pro??1 rispondesse a tutte le domande.
Gianni1 hoped that tomorrow, at the computational linguistics
test, pro?1 answered to all the questions
‘Gianni hoped that tomorrow, at the computational linguistics
test, he?1 answered all the questions’.
A similar solution may be proposed with respect to strong pronouns. Car-
dinaletti (2004) shows that strong pronouns (together with full DPs) occupy
a higher position than weak pronouns. Moreover, she maintains that this po-
sition is located within the CP. (Under this hypothesis, the fact that a group
of speakers do not accept CD when the embedded subject is a DP or a strong
pronoun, follows straightforwardly, assuming that the subject position is
within the CP and that the CP is deleted).
If a strong pronoun occurs, the CP field is activated. No unsatisfied posi-
tion can occur and the strategies to achieve a de se reading do not involve
theta-identification of such a position.
Note that the sentences in (10), the CP does not introduce the temporal
coordinate of the speaker: the sentence does not convey the DAR, indeed.
Thus, the reference to the utterance, rather than the utterance temporal coor-
dinate, appears to be responsible for the ‘blocking condition’ in this case.
All this considered, it comes as no surprise that dislocated clauses facili-
tate the de se reading (see section 3.6.2). If a clause is dislocated, the infor-
mation structure it carries makes it unavailable the application of the hy-
pothesized principle, carrying a reference to a specific conversational back-
ground.
6.2.2.3 First-person pronouns
The above hypothesis also accounts for the fact that first-person pronouns
appear to facilitate the de se reading.
We have seen (section 5.3.1) that a first-person pronoun ‘blocks’ the
availability of binding a LDA (nominal blocking condition). Giorgi (2006)
shows that this is predicted by the blocking condition discussed above, as-
suming that first person pronouns introduce a reference to the utterance con-
text.
The fact that obviation is ‘weakened’ if the first person is involved may
then follow from the blocking condition. The first person pronoun introduces
a reference to the utterance. The predicate must be saturated – all the theta-
positions must be satisfied and the de se reading cannot be conveyed via
theta-identification of an unsatisfied position with the bearer of the attitude.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
114
As for the case of strong pronoun above, the prediction arise that even in
‘core’ obviation cases the first-person pronoun is more acceptable than a
third-person (null) pronoun. For a discussion on this topic, see section 6.3.1
below.
6.2.2.4 Directive predicates
In section 3.3.3.1 I have shown that a clausal argument of a directive verb is
obviative with respect to the matrix subject, but it can be proximate with re-
spect to the matrix object, I assume that orders and requests can be viewed as
a attitudes involving the will of the person who expresses them, the attitude
bearer. If so, the fact that the embedded proposition cannot be de se follows
straightforwardly under the above hypothesis. Pro can however refer to the
indirect object in the matrix clause, since it does not refer to the attitude
holder.
Note that if the embedded form carrying subjunctive morphology is a
voice auxiliary, the embedded clause may be first-personal:
(11) Gianni1 chiese al rettore che pro1 fosse ammesso all’università.
Gianni asked to the chancellor that pro1 was.SUBJ admitted at the
university
‘Gianni asked the chancellor that he was admitted at the univer-
sity’.
In the above sentence an unsatisfied position is associated with the non-finite
form. By the hypothesis here proposed, it can be theta-identified with the
attitude bearer, conveying the de se reading.
6.2.2.5 Double embedding
The data concerning double embedding discussed in the section 3.4. are also
accounted for by the mechanism hypothesized here. As it has often been
shown in the literature concerning obviation, a doubly embedded subject is
obviative only with respect to the intermediate subject (or object), not with
the matrix. A de se reading involves an attitude and the object of that atti-
tude. In the double embedding cases, the most embedded clause does not
refer to the object of a propositional attitude by the matrix author. Rather, an
attitude toward the content of the most embedded clause is attributed to the
intermediate author. Thus, the most embedded subject cannot be de se with
respect to the intermediate author (unless an unsatisfied position occurs

AN INTERFACE PROPOSAL FOR SUBJUNCTIVE OBVIATION
115
within the most embedded clause), though it can be with respect to the ma-
trix45.
6.2.2.6 Adverbial clauses
In section 3.5 I have shown that obviation obtains even in some adverbial
clauses in the subjunctive. Particularly, in rationale, before-, and relative
clauses the matrix and the embedded subjects cannot be coindexed if the
lexical verb of that clause is inflected in the subjunctive, whereas in conces-
sive, conditional and comparison clauses they can be coindexed.
The hypothesis under investigation appears to be correct with respect to
rationale clauses. As Manzini (2000) and von Fintel and Iatridou (2005)
suggests, in rationale clauses the complementizer introduces a volitional atti-
tude controlled by the matrix predicate. Thus, under the hypothesis here pro-
posed, the de se reading is expected to obtain only if the predicate of the ra-
tionale clause is non-finite. This appears to be correct, since, as I have shown
in section 3.5, the occurrence of a tense, voice, modal auxiliary appear to
improve the coreferential reading, and an infinitival clause grants the de se
reading:
(12) a. *Pro vengo perché pro ti aiuti. (= 3.5, (44))
Pro.1SG come.IND.PRES.1SG in-order-that pro.1SG you-CL
help.SUBJ.PRES.1SG
b. Pro vengo per PRO1 aiutarti. (= 3.5, (44))
Pro.1SG come.IND.PRES.1SG in-order-to PRO1 help.INF you-CL
‘I come to help you’.
45 If the most embedded clause were the object of the attitude on the part of the matrix subject,
some morpho-syntactic implication would be expected. If the matrix verb selected for a sub-
junctive clause, the most embedded clause would be in the subjunctive as well, whatever
mood the intermediate predicate selected for. Hence, if the matrix predicate selected for a
subjunctive clause, and the intermediate predicate selected for an indicative clause (when non-
embedded), the verb of the most embedded clause should be in the subjunctive, too. Vice
versa, if the matrix predicate selected for an indicative clause, and the intermediate predicate
selected for a subjunctive clause, the verb of the most embedded clause should be able to be
in the indicative. But this does not seem to be the case:
(i) a. Gianni sperava che Maria dicesse che pro era/?*fosse partito.
Gianni hoped that Maria told.SUBJ that pro was.IND/?*SUBJ left
‘Gianni hoped that Maria told that he/she had left’.
b. Gianni ha detto che Maria voleva che pro *partiva/partisse presto.
Gianni has said that Maria wanted.IND that pro left.*IND/SUBJ soon
‘Gianni said that Maria wanted him/her to left soon’.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
116
I assume that an implicit attitude occurs in relative clauses as well. Under
this assumption, the hypothesis here proposed appears to be adequate, since
it correctly predicts the status of clauses in which the full verb bears the
morphology of the subjunctive vis-à-vis the status of clauses in which the
full verb is non-finite:
(13) a. *Gianni1 sta cercando un esperto con cui pro*1 parli del proble-
ma. (= 3.5, (47))
Gianni1 is looking for an expert with whom pro1
talk.SUBJ.PRES.3SG about the problem
b. Gianni1 sta cercando un esperto con cui PRO1 parlare del pro-
blema.
Gianni1 is looking for an expert with whom PRO1 talk.INF about
the problem
‘Gianni is looking for an expert to talk about the problem with’.
c. Gianni1 sta cercando un esperto con cui pro1/2 possa parlare del
problema. (= 3.5, (48))
Gianni1 is looking for a n expert with whom pro*1/2
can.SUBJ.PRES.1SG talk about the problem
‘Gianni1 is looking for an expert he1/2 can talk about the prob-
lem with’.
As for the before-clauses, the assumption that an implicit attitude occurs
does not appear to be feasible, since the relation the embedded eventuality
has with respect to the matrix eventuality is purely temporal. Whereas ra-
tionale and relative clauses may be de se, before-clauses cannot be consid-
ered de se, since no attitude is reported. Thus, the hypothesis in (3) is not
concerned in before-clauses.
However, since the modal ‘competition’ between subjunctive and infini-
tive holds, a mechanism like Bouchard’s (1984) Elsewhere Principle would
be adequate:
(14) a. *Gianni1 legge il giornale prima che pro1 faccia colazione. (=
3.5, (46)a)
Gianni1 reads the newspaper before that pro*1
make.SUBJ.PRES.1SG breakfast
b. Gianni1 legge il giornale prima di PRO1 fare colazione.
Gianni1 reads the newspaper before PRO1 make.INF breakfast
‘Gianni reads the newspaper before having breakfast’.

AN INTERFACE PROPOSAL FOR SUBJUNCTIVE OBVIATION
117
Since an anaphoric relation cannot be expressed through a pronoun if in the
same environment an anaphor is available, sentence (14)a is predicted to be
ungrammatical, since sentence (14)b expresses an anaphoric relation em-
ploying an anaphor (PRO).
Note that this mechanism would be able to generate the subjunctive ad-
verbial clauses in which obviation does not occur. As we have seen in sec-
tion 3.5, in comparison, conditional, and concessive clauses obviation is not
triggered. The Elsewhere Principle predicts that PRO is not available. This
appears to be correct:
(15) a. Gianni1 è più intelligente di quanto pro1/2 non pensi.
Gianni
1 is more smart than pro1/2 not think.SUBJ.PRES.3SG
‘Gianni
1 is smarter than he1/2/she thinks’.
a'. *Gianni1 è più intelligente di quanto PRO1 pensare.
Gianni
1 is more smart than PRO1 not think.INF
b. Gianni1 si divertirebbe, se pro1/2 venisse.
Gianni
1 would-amuse, if pro1/2 go.SUBJ.PRES.3SG
‘Gianni
1 would amuse himself, if he1/2/she went’.
b'. *Gianni1 si divertirebbe, se PRO1 venire.
Gianni
1 would-amuse, if PRO1 go.INF
c. Gianni1 sa guidare nonostante pro1 non abbia ancora la patente.
Gianni1 can drive despite pro1 not AUX.SUBJ.PRES.3SG yet the
driving license
‘Gianni can drive the car despite he has not had the driving li-
cense’.
c'. *Gianni1 sa guidare nonostante PRO1 non avere ancora la paten-
te.
Gianni1 can drive despite PRO1 not have.INF yet the driving li-
cense
d. Gianni1 mi potrà parlare purchè prima pro1 mi chieda scusa.
Gianni to-me will be able to talk only if before pro1 to-me
apologize.SUBJ.PRES.3SG
‘Gianni will be able to talk to me only if he apologizes to me
before’.
d'. *Gianni1 mi potrà parlare purchè PRO1 chiedermi scusa.
Gianni to-me will be able to talk only if before PRO1 to-me
apologize.INF

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
118
To conclude, while the hypothesis in (3) appears to be adequate in order
to explain obviation in attitudinal environments, an additional rule is needed
to account for obviation in non-attitudinal environments, as before-clauses.
To a first, sketchy, analysis, the Elsewhere Principle appears to be able to
undertake this task.
6.2.2.7 Coordination and overlapping reference
The proposal here investigated is able to make correct predictions even with
respect to the data concerning coordination (illustrated in 3.7). Due to the
syntactic properties of coordination, an infinitival clause cannot be coordi-
nated with a preceding subjunctive clause (while a subjunctive clause can be
coordinated with a preceding infinitival clause):
(16) a. *Gianni spera che Maria parta e di venire.
Gianni hopes that Maria leave.SUBJ.PRES.3SG and to come.INF.
b. Gianni spera di venire e che Maria parta.
Gianni hopes to come.INF and that Maria leave.SUBJ.PRES.3SG
‘Gianni hopes to come and that Maria leaves’.
The hypothesis in (3) and the blocking condition in (33)A state that in an
environment devoid of references to the utterance a de se reading is achieved
via theta-identification of an unsaturated position with the agent of the con-
text. This implies that if a clause containing an unsaturated position and a
clause that do not contain are both available, the former, but not the latter,
must be used to express a de se attitude.
Let us suppose, however, that the clause containing the unsaturated posi-
tion is unavailable due to independent reasons. In this case, the clause that
does contain the unsatisfied position can be interpreted de se as well. Indica-
tive clauses are such a case, to the extent to which I have characterized them,
since due to their interface properties, an unsatisfied position is unavailable
and the de se reading must be achieved in a different way than via theta-
identification of an unsatisfied position with the bearer of the attitude.
As for the examples involving coordinated subjunctive clauses, an ac-
count may proceed as follows. Due to the syntax of coordination, an infiniti-
val clause cannot be coordinated with a subjunctive clause preceding it, as
example (16)a shows. Thus, since an infinitival clause, which contain an un-
saturated position, is unavailable, the de se reading may be conveyed even
by an explicit formative. This appears to be correct46:
46 I take that the strong pronoun is responsible for the strong marginality, though not complete
ungrammaticality, of a de se reading. The null pronoun is not available in construals like (17)
(see section 3.7).

AN INTERFACE PROPOSAL FOR SUBJUNCTIVE OBVIATION
119
(17) Gianni1 spera che Maria parta e che lui1 venga.
Gianni1 hopes that Maria leaves and that he1 comes
‘Gianni hopes that Maria will leave and he will come’.
Vice versa, since an infinitival clause can be coordinated with a subjunctive
clause following it (as in (16)b), the de se reading can be conveyed by the
infinitive. A de se subjunctive clause should then be ruled out. This appears
to be correct:
(18) *Gianni1 spera che pro1 venga e che Maria parta.
Gianni1 hopes that pro1 comes and that Maria leaves.
If the unavailability of a de se competitor makes it possible a de se read-
ing for a subjunctive clause, the fact that a de se reading of an embedded
pronoun can be achieved when such a pronoun is coordinated, is expected.
Coordination concerns indeed syntactic formatives. Thus, the de se interpre-
tation cannot be achieved by an unsatisfied position, which is not a forma-
tive. This prediction appears to be correct:
(19) Gianni1 spera che lui1 e Maria escano insieme.
Gianni1 hopes that he1 and Maria go outside together.
‘Gianni hopes that he and Maria go outside together’.
These facts shed light on the data concerning overlapping reference (see
section 3.3.1.4). If the subject of the subjunctive clause is intended to refer-
entially include, though not to exhaust, the author of the context, an unsatis-
fied position is not available, since it would convey a strictly de se reading.
Thus, again, a formative can convey this reading. For a more in depth analy-
sis of overlapping reference data, see section 6.3.3.2 below.
6.3 Problems and tentative solutions
6.3.1 Theta-role sensitivity
I have shown that the hypothesis in (3) is able to account for most of the
facts concerning obviation that I have discussed in chapters 2 and 3. How-
ever, a principled account appears to be missing on some data and on some
implication of the proposal here investigated. I will briefly discuss these is-
sues, proposing some tentative solution.
Let us consider first the empirical issues concerning the hypothesis pro-
posed. Descriptively, we have seen (section 3.3.1.1) that obviation is sensi-
tive to the theta-roles of the embedded clause, in that the de se reading is

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
120
harder to get if the embedded subject discharges the agent theta-role. To il-
lustrate, consider the examples in (22), section 3.3.1.1, repeated here:
(20) a. Pro1ps Spero (proprio) che io guarisca presto.
Pro1ps hope (really) that I recover.SUBJ.PRES.1SG soon
‘I hope I will soon get better’.
b. *Pro1ps Spero (proprio) che io diverta quei bambini.
Pro1ps hope (really) that I amuse.SUBJ.PRES.1SG those children
The former sentence is intuitively less degraded (and in fact it is almost fully
acceptable) than the latter. The hypothesis in (3) does not discern this differ-
ence.
As a preliminary remark, it must also be observed that the first-person
pronoun vis-à-vis the third person pronoun improves the acceptability of the
former sentence, though not of the latter:
(21) *Pietro1 spera (proprio) che pro1 guarisca presto.
Pietro1 hopes (really) that pro1 recover.SUBJ.PRES.3SG soon
The contrast between (20)a and (21) can be accounted along the lines
traced in section 6.2.2.3. What is interesting, however, is that the first-person
pronouns improves sentence (20)b much less than sentence (20)a.
A possible solution to the question raised by the sentences in (20) may be
to claim, in the spirit of Schlenker’s (2005), that the notion of de se in the
hypothesis in (3) is to be defined as individual and event de se. By event de
se attitude I mean here an attitude towards an eventuality that assumingly is
consciously brought about by the very same individual to which an attitude
towards a proposition is attributed. Thus, in an individual de se eventuality,
the agent of the context takes a role in that eventuality; in an event de se
eventuality, the agent of the context is the ‘initiator’ (in the sense of Farkas
1988) of the eventuality.
To illustrate, in uttering the sentence I want to leave, I am reporting an at-
titude toward a leaving eventuality that involves myself. Intuitively, in utter-
ing this sentence, I presuppose that I can bring about the eventuality of leav-
ing. We then may define this attitude report as event de se, since the initiator
is the first person, and individual de se, since the argument of leave is the
first person.
If the presupposition that I have the opportunity to bring about the event
of leaving cannot be considered true, the sentence I want to leave determines
a semantic failure (alternatively, the sentence can be accommodated within
the context set as an emphatic expression of the wish to leave). This is what
would happen in a scenario in which I am not allowed or I am not physically

AN INTERFACE PROPOSAL FOR SUBJUNCTIVE OBVIATION
121
able to leave: I might not properly want to leave, though I might wish to
leave. For instance, if I got stuck in a traffic jam, to mention a scenario al-
ready discussed (in section 4.4.2), I may say I wish to leave, but I may not
say I want to leave, since the latter sentence presupposes that I can bring
about the conditions to leave by a personal resolution, which is not felicitous
in such a scenario. In other words, an event de se attitude report is inappro-
priate in such a scenario, since I cannot be the initiator of the event of leav-
ing.
As Farkas (1988) claims (and as the above example shows), the agent and
the initiator need not coincide. If they do, an attitude report is individual and
event de se. If they do not, the attitude report can be individual de se, though
not event de se, or the other way around – individual non-de se and event de
se. When the agent and the initiator do not coincide, the attitude report may
be individual de se, but if so, the initiator cannot be de se as well. Vice versa,
if the initiator is de se, the argument of the embedded predicate cannot be de
se as well. Moreover, if an eventuality is agentless, the initiator does not co-
incide with the subject argument (it may be taken to be as a quasi-argument,
as the subject of weather predicates).
If the hypothesis in (3) refer to both the individual and event de se read-
ing, it may be paraphrased as stating that both the initiator and the agent-role
must be unsatisfied in order to achieve the de se reading (the blocking condi-
tion remains unaltered). This predicts that if the agent and the initiator are
both formatives (pro), the attitude is event and individual non-de se. If the
agent and the initiator do not coincide, there are two options: If the agent is
not assigned via theta-marking while the initiator is, the attitude is individual
de se and event non-de se. I assume that if agentless predicates (like recover)
are a subcase of this option. Vice versa, if the agent is assigned and the ini-
tiator is not, the attitude will be individual non-de se and event de se.
Given these assumptions, let us consider the contrast in (20). In sentence
(20)a the embedded predicate is agentless. Thus, the initiator does not coin-
cide with the subject argument, is satisfied by a pro argument, thus convey-
ing the non-de se event reading. Moreover, if the embedded subject is as-
signed a theta-role, the attitude is individual non-de se (unless additional fac-
tors intervenes); if this theta-role remains unassigned, the attitude is individ-
ual de se. This may be achieved through an infinitival clause:
(22) Gianni spera di guarire presto.
Gianni hopes to recover soon.
‘Gianni hopes to recover soon’.
In sentence (20)b the embedded predicate assigns the agent (or rather, the
causer) theta-role to the subject. Since in this case the initiator and the agent
generally coincide, either they are both unsatisfied through theta-marking,

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
122
conveying the individual and event de se reading, or they are assigned to a
formative, conveying the individual and event non-de se reading. This ap-
pears to be correct, since as far as interpretation goes, sentence (20)b reports
the speaker’s hope that he complies with a request to amuse the children.
Thus, while in sentence (20)a the only violation of the hypothesis pro-
posed concerns the individual de se reading – the matrix subject is a forma-
tive, a violation that is remedied by the first-person pronoun, in sentence
(20)b there is a double violation – with respect to the individual de se read-
ing and with the event de se reading.
The idea here pursued also allows us to predict that subjunctive clauses
containing a stative predicate are more easily de se than subjunctive clauses
containing an eventive predicate that assigns the agent role:
(23) Maria1 supponeva pro?1 fosse incinta.
Maria1 supposed pro?1 was pregnant.
‘Maria1 supposed she?1 was pregnant’.
Stative predicates are indeed agentless. The subject argument and the ‘initia-
tor’ do not coincide. Thus, the de se reading may be marginally achieved
even by pro, because the violation to the hypothesis in (3) concerns only the
requirements concerning the individual de se reading.
6.3.2 Implications concerning Control
6.3.2.1 Unsatisfied positions
The hypothesis proposed above assumes Higginbotham’s (1997) claim that
Obligatory Control can be restated in terms of theta-identification between
two theta-roles, one of which, the theta-role assigned to the subject argument
of an infinitive verb, cannot be satisfied through obvious theta-marking (sec-
tion 5.3.3). In Higginbotham’s view, PRO might be merged only if the theta-
role it discharges has already been theta-identified with a matrix argument –
the controller.
This idea raises questions concerning a syntactic implementation of
Obligatory Control. Recent theories on Obligatory Control have tried to get
rid of an autonomous module of Control resorting general principles. Horn-
stein (1999), Manzini and Roussou (2000) have proposed that Obligatory
Control can be restated in terms of raising. Landau (2000) has claimed that
Obligatory Control can be explained in terms of abstract Agree (Chomsky
1998). Wurmbrand (2000) and Cinque (2004, 2006), treating restructuring
predicates as raising predicates, claim that some structures traditionally con-

AN INTERFACE PROPOSAL FOR SUBJUNCTIVE OBVIATION
123
sidered as Control structures can be considered as restructuring structures –
thus involving raising.
The raising analysis of some Control predicates, which appears to be
valid (as also the data discussed in Costantini 2008 and to appear show),
contrasts with the idea that Obligatory Control involves an unsatisfied posi-
tion. The latter idea states that this position is discharged via theta-
identification with another theta-position. This view is in itself incompatible
with the raising analysis of Control. If Obligatory Control is raising, the
higher position is not assigned a theta-role, because a movement chain can
contain only one theta-position47. Since movement cannot target a position in
which a theta-role is discharged, the locus of theta-assignment in infinitival
clauses involving Obligatory Control must be within the embedded VP. This
implies that the embedded subject is assigned a theta-role through theta-
marking.
This observation may be even addressed to the idea, developed in 6.3.2.1,
that in modal constructions and in periphrastic tenses the theta-role that is
generally assigned to the subject argument is satisfied through theta-
identification rather than through theta-marking. As for modal constructions,
movement appears to be involved. As for periphrastic tenses, movement is
standardly taken to occur.
At present, the hypothesis that Obligatory Control involves theta-
identification is not able to cope with these theoretical shortcomings. This
also weakens the hypothesis here proposed.
Moreover, assuming a mechanism of theta-identification for the cases of
obviation weakening concerning embedded clauses having a non-finite full
verb leaves in the obscurity the status of the subject of the embedded
clauses, pro. Under the hypothesis proposed, it is not merged within the VP
and it is not assigned a theta-role by the functional verb, because functional
verbs presumably do not assign theta-roles (see Cinque 2004). Thus, it must
be analyzed as an expletive. However, differently from expletives, we have
assumed that it has an autonomous reference, since de se subjunctive clauses
are ambiguous between the de se and the non-de se reading. Moreover, if it
has an autonomous reference, it cannot be de se by hypothesis. As a tenta-
tive, though not quite satisfactory, answer, we may take subjunctive clauses
having a non-finite full verb to be the spell-out of two different structures,
one having a position that is theta-identified with the bearer of the attitude
and having pro as a (non-referential) expletive (or having no pro at all48),
and the other having a referential pro, generated within the VP:
47 Hornstein (1999) claims that A-chains can contain more than one theta-position. Such a
claim is however controversial.
48 For recent discussions on pro see Alexiadou and Anagnastopoulou (1998), Uriagereka
(1999), Cardinaletti (2002), Holmberg (2005). Alexiadou and Anagnastopoulou (1998) and
Uriagereka (1999) claim that pro can be dispensed with in a minimalist framework.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
124
(24) a. DP1 … [ pro*1 F.SUBJ [pro*1 V.INF/PART/GER (non-de se)
b. DP
1 … [proexpl F.SUBJ [x V.INF/PART/GER (de se)
This solution does not appear to be satisfactory, however, since expletives
pronouns are standardly taken not to occur in environments as the ones at
issue.
A final question is raised by Giorgi (2007). As we have seen in 5.3.3,
Giorgi (2007) proposes that the mechanism of theta-identification is respon-
sible for the de se reading of both PRO (in Obligatory Control) and LDAs.
Giorgi (2007) argues that some control facts may be accounted for under
this analysis, but others cannot. In particular, she observes that while Ex-
haustive Control (Landau 2000, EC for short) can be treated as involving
that-identification (PRO is indeed referentially identical to the cntroller), the
phenomenon of Partial Control (Landau 2000, henceforth, PC) does not
seem to be compatible with the hypothesis that PRO is the spell-out of an
unsatisfied position, since its reference includes the author of the context, but
can include some other referentially salient individual. Thus, PC PRO is not
strictly de se.
To illustrate, consider the following examples:
(25) a. *Mary thought that John managed to [PRO meet at 6].
b. Mary thought that John wanted [PRO to meet at 6].
The embedded predicate in the examples above is a collective verb. Its sub-
ject cannot be singular:
(26) *John met at 6.
In both the examples in (25) the controller is singular. This notwithstanding,
sentence (25)a is ungrammatical, whereas sentence (25)b is acceptable
Landau takes collective predicates (gather, meet, etc.) as a diagnostics to
divide all instances of Obligatory Control into the classes of EC and PC. The
former class of predicates requires that the reference of the controller and of
PRO should be strictly identical, whereas the latter class does not require
strict coreference between the controller and the controllee. Though, the ref-
erence of the controller must be included within the reference of PRO.
EC involves the predicates belonging classes of the implicative, aspec-
tual, and modal predicates. Since these predicates are not attitude report
predicates – PRO is not de se, strictly speaking, one may claim that they do
not raise substantial questions to the hypothesis in (3), because it states a
principle concerning attitude predicates (which EC predicates are not) and
attitudes de se.

AN INTERFACE PROPOSAL FOR SUBJUNCTIVE OBVIATION
125
PC involves desiderative, epistemic, factive predicates – they are attitude
predicates. The question then arises, how the non-exhaustiveness reading is
achieved, assuming that theta-identification (which implies strict identity
between two theta-roles) is concerned.
A possible solution may resort Landau’s (2000, 2004) theory of PC, ac-
cording to which PC PRO carries semantically plural features (or the feature
[+mereogical], see Landau 2004). He claims that semantic plurality is lexi-
cally encoded in PRO. Since PRO can be understood as a consequence,
rather than the causer of control – that is, PRO is supposed to be «selected to
occur in certain configurations where control [i.e., theta-identification] has
already been established» (Higginbotham 1997: 192), the fact that the PC
reading holds does not challenge the view on Obligatory Control here pur-
sued.
6.3.2.2 Overlapping reference and Partial Control
Given the above view on PC and theta-identification, a problem arises, how-
ever, for the hypothesis here proposed. Since PC is de se (see Landau 2000),
the hypothesis predicts that the overlapping reference relative to the subject
of a subjunctive clause should be ruled out.
As has been shown in section 3.3.1.4, this is contrary to the expectation.
In section 6.2.2.7, I have proposed a solution based on the argument that
syntax excludes that an unsatisfied position can refer to a set of individuals
that includes the author of the context. Implicit in this argument is the idea
that theta-identification instantiates strict coreference.
However, the facts concerning PC appear to challenge this view. Con-
sider the following examples:
(27) a. Il presidente1 crede che pro1+ si siano riuniti inutilmente la notte
scorsa.
The chair1 believes that pro1+ AUX.SUBJ.PRES.3PL gathered in
vain the night passed
‘The chair1 believes that they1+ gathered in vain last night’.
b. Il presidente1 crede di PRO1+ essersi riuniti inutilmente la notte
scorsa.
The chair1 believes P/C PRO1+ AUX.INF gathered in vain the
night passed
‘The chair believes to have gathered in vain last night’.
Sentence (27)a is ambiguous between the de se and non-de se reading,
whereas sentence (27)b can only be de se. Thus, in a scenario in which the

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
126
chair is amnesiac and does not remember that, for instance, the company
council of directors of which he is the chair gathered the preceding night
without reaching an agreement, sentence (27)a may be considered true,
whereas sentence (27)b may not.
The point is that the former sentence may be considered true even in a
scenario in which the chair is completely aware of whom she is – it can be
de se. Under the hypothesis in (3), this is unexpected, since the de se reading
may only be conveyed by theta-identification of an unsatisfied position,
which cannot be generally spelled out by pro (reformulating what has been
claimed in section 6.3.2), though it can be by PRO.
We observe however, that sentence in (27)a involve an environment in
which obviation is weakened, which may be treated along the lines drawn in
section 6.3.2.1. Once we consider a sentence having no non-finite null verb,
the de se reading appears to be very degraded as example (29), section
3.3.1.4, shows:
(28) Il presidente1 si chiede dove pro??1+ si riuniscano.
The chair1 wonders where pro??1+ gather.SUBJ.PRES.3PL gathered
‘The chair1 believes that they??1+ gathered’.
Thus, although some examples appear to show that overlapping reference
is possible, some others makes the claim questionable whether this reading is
possible. Further analysis appears to be needed in order to reach an adequate
descriptive generalization.
6.4 Conclusions
In the present chapter I have presented the hypothesis that an attitude is de se
if the content of the attitude is unsaturated. The theta-role that remains unas-
signed via obvious theta-marking is theta-identified with the bearer of the
attitude, giving rise to the de se interpretation. This principle is sensitive to
Giorgi’s (2006) Blocking Condition, which restricts the distribution of unsat-
isfied positions to the environments devoid of a reference to the utterance
situation.
I have shown that the hypothesis proposed is able to account for most of
the data concerning obviation and obviation weakening. Some problematic
issues have been accommodated within the theory. The only real limitation
of the theory concerns the theoretical status of unsatisfied positions, which at
present appears to be problematic. The wide range of phenomena accounted
for by the hypothesis proposed suggests however to provisionally keep this
theoretical notion, relying on future research to shed more light on the non-
trivial relations existing between theta-roles, pragmatic roles, and arguments.

References
Abusch, Dorit, 1997. Sequence of Tense and Temporal De Re. Linguistics and Phi-
losophy 20: 1-50.
Alexiadou, Artemis, and Elena Anagnostopoulou, 1998. Parametrizing Agr: word
order, V-movement and EPP-checking. Natural Language and Linguistic Theory
16: 491-539.
Aronoff, Mark, 1976. Word Formation in Generative Grammar. Cambridge
(Mass.): MIT Press.
Avrutin, Sergey, 1994. Psycholinguistics investigations in the theory of reference.
Ph.D. diss., MIT.
Avrutin, Sergey, and Maria Babyonyshev, 1997. Obviation in subjunctive clauses
and Agr: evidence from Russian. Natural Language and Linguistic Theory 15:
229-267.
Belletti, Adriana, and Luigi Rizzi, 1988. Psych-verbs and θ-theory. Natural Lan-
guage and Linguistic Theory 6: 291-352.
Bouchard, Denis, 1983. The Avoid Pronoun Principle and the Elsewhere Principle.
In Proceedings of ALNE 13/NELS 13, Peter Sells, Charles Jones (eds.): 29-36.
Amherst (Mass.): GLSA.
Bouchard, Denis, 1984. On the Content of Empty Categories. Dordrecht: Foris.
Brody, Michael, and Rita Manzini, 1991. On Implicit Arguments. In Mental Repre-
sentations. The Interface between Language and Reality, Ruth Kempson (ed.):
105-130. Cambridge: Cambridge University Press.
Cardinaletti, Anna, 2002. Towards a Cartography of Subject Positions. In The Struc-
ture of CP and IP: The Cartography of Syntactic Structures vol. 2, Luigi Rizzi
(ed.): 115-165. New York: Oxford University Press.
Cardinaletti, Anna, and Michal Starke, 1999. The Typology of Structural Defi-
ciency. A Case Study of Three Classes of Pronouns. In Clitics in the Languages
of Europe, EALT-EUROTYP 20-5, Henk van Riemsdijk (ed.): 145-233. Berlin-
New York: Mouton de Gruyter.
Castañeda, Hector-Neri, 1966. He*: a Study in the Logic of Self-Consciousness.
Ratio 8: 130-157.
Castañeda Hector-Neri, 1968. On the Logic of Attributions of Self-Knowledge to
Others. The Journal of Philosophy, 65/15: 439-456.
Chierchia, Gennaro, 1989. Anaphora and Attitude de se. In Semantics and Contex-
tual Expressions, Renate Bartsch, Johan van Benthem, Peter van Emde (eds.):
1-31. Dordrecht: Foris.
Chierchia, Gennaro, 1997. Semantica. Bologna: Il Mulino.
Chierchia, Gennaro, and Sally McConnell-Ginet, 1990. Meaning and Grammar. An
Introduction to Semantics. Cambridge: MIT Press.
Chomsky, Noam, 1980. On Binding. Linguistic Inquiry 11: 1-46.
Chomsky, Noam, 1981. Lectures on Government and Binding. Dordrecht: Foris.
Chomsky, Noam, 1982. Some Concepts and Consequences of the Theory of
Government and Binding. Cambridge (Mass.): MIT Press.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
128
Chomsky, Noam, 1986. Knowledge of Language: its Nature, Origin and Use. New
York: Praeger.
Chomsky, Noam, 1995. The Minimalist Program. Cambridge (Mass.): MIT Press.
Chomsky, Noam, 1998. Minimalist Inquiries: The Framework. In Step by Step:
Essays in Minimalist Syntax in Honor of Howard Lasnik, Robert Martin, David
Michaels, Juan Uriagereka (eds.): 89-155. Cambridge: MIT Press.
Chomsky, Noam, 1999. Derivation by Phase. In Ken Hale: A Life in Language,
Michael Kenstowicz (ed.): 1-52. Cambridge: MIT Press.
Cinque, Guglielmo, 1999. Adverbs and Functional Heads. A Cross-Linguistic
Perspective. New York: Oxford University Press.
Cinque, Guglielmo, 2004. Restructuring and Functional Structure. In Structures and
Beyond. The Cartography of Syntactic Structures, vol 3, Adriana Belletti (ed.):
132-191. New York: Oxford University Press.
Cole, Peter, Gabriella Hermon, and Li-May Sung, 1990. Principles and Parameters
of Long-Distance Reflexives. Linguistic Inquiry 21: 1-22.
Costantini, Francesco, 2005. On Obviation in Subjunctive Clauses: The State of the
Art. Annali di Ca’ Foscari 44: 97-132.
Costantini, Francesco, 2006. Obviation in Subjunctive Argument Clauses and the
First-Personal Interpretation. In Phases of Interpretation, Mara Frascarelli (ed.).
Berlin: Mouton de Gruyter.
Costantini, Francesco, 2007. On Tense Mismatch and a Morphosyntactic Theory of
Sequence of Tenses, Rivista di Grammatica Generativa 32, pp. 39-64.
Costantini, Francesco, 2008. Movement or Control? Some Hints from Floating
Quantification, University of Venice Working Papers in Linguistics 18: 131-144.
Costantini Francesco, to appear. On infinitives and Floating Quantification. Linguis-
tic Inquiry.
Costantini, Francesco, and Luigi Zennaro, to appear. Some Aspects of the Syntax
and Semantics of Verba Voluntatis, in Proceedings of the 13th International
Colloquium on Latin Linguistics, Bruxelles, April 4-9, 2005.
Everaert, Martin, 1986. Long Reflexivization and Obviation in the Romance Lan-
guages. In Formal Parameters of Generative Grammar, Peter Coopmans, Ivonne
Bordelois, Bill Dotson Smith (eds.): 51-71. Dordrecht: Foris
Farkas, Donka, 1988. On Obligatory Control. Linguistics and Philosophy 11: 27-58.
Farkas, Donka, 1992. On Obviation. In Lexical Matters, Ivan A. Sag, Anna
Szabolcsi (eds.): 85-109. Stanford University: CSLI.
Feldhausen, Ingo (2007). The Relation of Obviation and Control: Subjunctives and
Infinitives in Spanish. Ms., ZAS Berlin.
Feldhausen, Ingo (2008). The Prosody-Syntax Interface in Catalan. Ph.D. diss.,
Universität Potsdam.
Giorgi, Alessandra, 1983. Toward a Theory of Long Distance Anaphors: A GB Ap-
proach. The Linguistic Review 3: 307-361.
Giorgi, Alessandra, 2004. Long Distance Anaphors and Temporal Relations. In
WCCFL 23 Proceeings, Vineeta Chand, Ann Kelleher, Angelo Rodríguez, Ben-
jamin Schmeiser (eds.): 223-236. Sommerville, MA.: Cascadilla Press.
Giorgi, Alessandra, 2006. From temporal anchoring to long distance anaphors.
Natural Language and Linguist Theory 24: 1009-1047.
Giorgi, Alessandra, 2007. On the nature of long distance anaphors. Linguistic
Inquiry, 38: 321-342.

REFERENCES
129
Giorgi, Alessandra, 2009. Toward a syntax of the subjunctive mood. Lingua.
Giorgi, Alessandra, and Fabio Pianesi, 1997. Tense and Aspect. From Semantics to
Morphosyntax. New York: Oxford University Press.
Giorgi, Alessandra, and Fabio Pianesi, 2001. Tense, Attitudes, and Subjects. In
SALT XI: Proceedings from Semantics and Linguistic Theory XI, Rachel Hast-
ings, Brendan Jackson, Zsofia Zvolenszky (eds.). Cornell University: CLC Pub-
lications.
Giorgi, Alessandra, and Fabio Pianesi, 2004. Complementizer Deletion in Italian.
The Structure of CP and IP. The Cartography of Syntactic Structures. Volume 2.
Luigi Rizzi (ed.): 190-210. New York: Oxford University Press.
Halle, Morris, and Alec Marantz. 1993. Distributed Morphology and the Pieces of
Inflection. In The View from Building 20, Kenneth Hale and S. Jay Keyser (eds.):
111-176. Cambridge (Mass.): MIT Press.
Heim, Irene, and Angelika Kratzer ,1997. Semantics in Generative Grammar. Ox-
ford: Blackwell.
Higginbotham, James, 1985. On Semantics. Linguistic Inquiry 16: 547-593.
Higginbotham, James, 1989. Elucidations of Meaning. Linguistics and Philosophy
12: 465-518.
Higginbotham, James, 1991. Belief and Logical Form. Mind and Language 6: 344-
369.
Higginbotham, James, 1992. Reference and Control. In Control and Grammar,
Richard K. Larson, Sabine Iatridu, Utpal Lahiri, James Higginbotham (eds.): 79-
108. Dordrecht: Kluwer Academic Publishers.
Higginbotham, James, 1995. Tensed Thoughts. Mind and Language 10: 226-249.
Higginbotham, James, 1997. A Plea for Implicit Anaphora. In Atomism and Binding,
Hans Bennis, Pierre Pica, Johan Rooryck (eds.): 182-203. Dordrecht: Foris.
Higginbotham, James, 2003. Remembering, Imagining, and the First Person. In
Epistemology of Language, Alex Barber (ed.): 496-533. Dordrecht: Foris.
Holmberg, Anders, 2005. Is There a Little Pro? Evidence from Finnish. Linguistic
Inquiry 36: 533-564.
Hornstein, Norbert, 1999. Movement and control. Linguistic Inquiry 30: 69-96.
Hornstein, Norbert, and Itziar San Martin, 2005. Obviation as Anti-Control. Interna-
tional Journal of Basque Linguistics and Philology (ASJU).
Huang, C.T. James, and Luther Liu, 2001. Logophoricity, Attitudes, and ziji at the
Interface. In Long Distance Reflexives, Peter Cole, Gabriella Hermon, C.-T.
James Huang (eds.): 141-195. New York: Academic Press.
Kaplan, David, 1989. Demonstratives: an Essay on the Semantics, Logic, Metaphys-
ics, and Epistemology of Demonstratives and Other Indexicals. In Themes from
Kaplan, Joseph Almog, John Perry, and Howard Wettstein (eds.): 481-563. New
York: Oxford University Press.
Kempchinsky, Paula, 1987. The subjunctive disjoint reference effect. In Studies in
Romance Linguistics, Carol Neidle, Rafael Nuñez Cedeño (eds.): 123-140.
Dordrecht: Foris.
Kempchinsky, Paula, 1998. Mood Phrase, Case Checking and Obviation. In Ro-
mance Linguistics: Theoretical Perspectives, Armin Schwegler, Bernard Tranel,
Myriam Uribe-Etxebarria (eds.): 143-154. Amsterdam/Philadelphia: John Ben-
jamins.
Koizumi, Masatoshi, 1995. Phrase structure in Minimalist syntax. Ph.D. diss., MIT.

INTERFACE PERSPECTIVES ON CLAUSAL COMPLEMENTATION
130
Krapova, Iliyana, 1998. Subjunctive Complements, Null Subjects and Case Check-
ing in Bulgarian. University of Venice Working Papers in Linguistics 8/2: 73-93.
Landau, Idan, 2000. Elements of Control: Structure and Meaning in Infinitival
Constructions. Dordrecht: Kluwer.
Landau, Idan, 2002. Movement out of control. Linguistic Inquiry 34: 471-498.
Landau, Idan, 2004. The Scale of Finiteness and the Calculus of Control. Natural
Language and Linguistic Theory 22: 811-877.
Lewis, David, 1979. Attitudes de dicto and attitudes de se. Philosophical Review 88:
513-543.
Luján, Marta, 1999. A Unified Approach to Control and Obviation. In Grammatical
Analyses in Basque and Romance Linguistics, Jon Franco, Alazne Landa, Juan
Martín, 105-130. Amsterdam/Philadelphia: John Benjamins.
Manzini, Maria Rita, 1983. On Control and Control Theory. Linguistic Inquiry 14:
421-446.
Manzini, Maria Rita, 2000. Sentential Complementation. The Subjunctive. In Lexi-
cal Specification and Insertion, Peter Coopmans, Martin Everaert, Jane Grim-
shaw (eds). Amsterdam/Philadelphia: John Benjamins.
Manzini, Maria Rita, and Anna Roussou, 2000. A Minimalist Theory of
A-movement and Control. Lingua 110: 409-447.
Marianne Mithun, 1999. The Languages of Native North America. Cambridge,
Cambridge University Press.
Pan, Haihua, 1998. Closeness, Prominence, and Binding Theory. Natural Language
and Linguistic Theory 16: 771-815.
Pan, Haihua, 2001. Why the Blocking Effect? In Long Distance Reflexives, Peter
Cole, Gabriella Hermon, C.-T. James Huang (eds.): 279-316. New York: Aca-
demic Press.
Paoli, Sandra, 2003. COMP and the Left Periphery: Comparative Evidence from
Romance. Ph.D. diss., University of Manchester.
Pesetsky, David, and Esther Torrego, 2000. T-to-C Movement: Causes and Conse-
quences. In Ken Hale: A Life in Language, Michael Kenstowicz (ed.): 355-426.
Cambridge: MIT Press.
Pica, Pierre, 1987. On the Nature of the Reflexiviztion Cycle. In Proceedings of the
17th Annual Meeting of the North-East Linguistic Society. NELS 17, Joyce
McDonough, Bernadette Plunkett (eds.): 483-499. Amherst: GLSA.
Picallo, Carme, 1985. Opaque domains. Ph.D. diss., CUNY.
Progovac, Ljiljana, 1993. Subjunctive: the (Mis)behavior of Anaphora and Negative
Polarity. The Linguistic Review 10: 37-59.
Progovac, Ljiljana, 1994. Negative and Positive Polarity: a Binding Approach.
Cambridge: Cambridge University Press.
Quer, Josep, 1997. Mood at the Interface. LOT Dissertation Series.
Quer, Josep, 2005. Subjunctives. In The Syntax Companion, Martin Everaert and
Henk van Riemsdijk (eds.), vol. IV: 660-684. Oxford: Blackwell.
Raposo, Eduardo, 1985. Some Asymmetries in the Binding Theory in Romance”.
The Linguistic Review 5: 75-110.
Rizzi, Luigi, 1986. Null Objects in Italian and the Theory of pro. Linguistic Inquiry
17: 501-557.
Rizzi, Luigi, 1997. The Fine Structure of the Left Periphery, in Elements of Gram-
mar, Liliane Haegeman (ed.): 281-337. Dordrecht: Kluwer.

REFERENCES
131
Rizzi, Luigi, 2000. On the Anaphor-Agreement Effect. In Luigi Rizzi, Comparative
Syntax and Language Acquisition: 158-173. London/New York: Routledge.
Roeper, Thomas, 1993. Explicit Syntax in the Lexicon. In Semantics and the Lexi-
con, James Pustejovsky (ed.): 183-220. Dordrecht: Kluwer.
Ruwet, Nicolas, 1984. Je veux partir / *Je veux que je parte: on the Distribution of
Finite Complements and Infinitival Complements in French”. Cahiers de Gram-
maire 7: 75-138. Reprinted in Nicolas Ruwet, 1991. Syntax and Human Experi-
ence. Chicago/London: The University of Chicago Press.
Safir, Ken, 1991. Evaluative Predicates and the Representation of Implicit Argu-
ments. In Principles and Parameters in Comparative Grammar, Robert Freidin
(ed.): 99-131. Cambridge (Mass.): MIT Press.
Schlenker, Philippe, 2005. The Lazy Frenchman’s Approach to the Subjunctive
(Speculations on Reference to Worlds, Presuppositions, and Semantic Defaults
in the Analysis of Mood). In Romance Languages and Linguistic Theory 2003,
Twan Geerts, Ivo van Ginneken, Haike Jacobs (eds.): 269-309. Amster-
dam/Philadelphia: John Benjamins.
Schlenker, Philippe, to appear. Indexicality and De Se Reports, in Handbook of Se-
mantics, Klaus von Heusinger, Maienborn and Paul Portner (eds.), Berlin/New
York: Mouton de Gruyter.
Sells, Peter, 1987. Aspects of Logophoricity. Linguistic Inquiry 18: 445-479.
Speas, Margaret, 2004. Evidentiality, Logophoricity and the Syntactic Representa-
tion of Pragmatic Features. Lingua 114: 255-276.
Stowell, Tim, 1993. Syntax of Tense. Ms., UCLA.
Suñer, Margarita, 1986. On the Referential Properties of Embedded Finite Clause
Subjects. In Generative Studies in Spanish Syntax, Ivonne Bordelois, Helas
Contreras, Karen Zagona (eds.): 183-196. Dordrecht: Foris.
Terzi, Arhonto, 1992. PRO in Finite Clauses: a Study of the Inflectional Heads of
the Balkan languages. Ph.D. diss., CUNY.
Tsoulas, George, 1994a. Indefinite Clauses. Some Notes on the Syntax and Seman-
tics of Subjunctives and Infinitives. In WCCFL 23 Proceedings. Sommerville,
MA.: Cascadilla Press.
Tsoulas, George, 1994b. Subjunctives as Indefinites. In Teoria del linguaggio e ana-
lisi linguistica, Gianluigi Borgato (ed.). Padova: Unipress.
Tsoulas, George, 1996. The Nature of the Subjunctive and the Formal Grammar of
Obviation. In Grammatical Theory and Romamnce Languages, Karen Zagona
(ed.): 293-306. Amsterdam/Philadelphia: John Benjamins.
von Fintel, Kai, and Sabine Iatridou, 2005. Anatomy of a Modal. Ms., MIT.
Wandruszka, Ulrich, 1991. Frasi Subordinate al Congiuntivo. In Grande grammati-
ca italiana di consultazione. II. I sintagmi verbale, aggettivale, avverbiale. La
subordinazione, Lorenzo Renzi, Giampaolo Salvi, Anna Cardinaletti (eds.): 415-
482. Bologna: Il Mulino.
Watanabe, Akira, 1993. Agr-based Case Theory and its Interaction with the A-bar
System. Ph.D. diss., MIT.
Williams, Edwin, 1980. Predication. Linguistic Inquiry 11: 203-238.
Williams, Edwin, 1985. PRO in NP. Natural Language and Linguistic Theory 3:
277-295.
Williams, Edwin, 1987. Implicit Arguments, the Binding Theory, and Control.
Natural Language and Linguistic Theory 5: 151-180.

Stampato in Italia presso
Laser Copy S.r.l., Via Livraghi 1, Milano
Prima edizione ottobre 2009