UNIVERSIT DE GENVE - Université de Genève

UNIVERSITÉ DE GENÈVE FACULTÉ DES LETTRES 

THE SYNTAX-INFORMATION STRUCTURE INTERFACE: 

A COMPARATIVE VIEW FROM ROMANIAN 

Thèse de doctorat 

présentée à la Faculté des Lettres de l’Université de Genève 

pour obtenir le grade de docteur ès lettres par 

Gabriela Soare 

2009 

Jury de thèse: Prof. Ur Shlonsky (Université de Genève) - Directeur 

Prof. Eric Wehrli (Université de Genève) - Président 

Prof. Luigi Rizzi (Université de Genève/Sienne) 

Prof. Guglielmo Cinque (Université de Venise) 

Prof. Alexandra Cornilescu (Université de Bucarest) 

Prof. Yoshio Endo (Université de Kanda)

Table of contents 

I. Introduction………………………………………………………………………. 1 

1.1. Kayne’s Antisymmetry Hypothesis……………………………………… 1 

1.2. Cartography of Syntactic Structures……………………………………... 2 

1.3. Syntax-Information Structure Interface…………………………………. 4 

1.4. Criterial Freezing………………………………………………………….5 

1.5. Locality and Rizzi’s (2004b) Relativised Minimality…………………… 5 

1.6. Types of Agree………………………………………………………….. 7 

1.7. The Organisation of the Dissertation…………………………………….. 8 

II. The Romanian Mittelfeld……………………………………………………… 13 

1. Information Structure and Syntax……………………………………… 13 

2. Argument- and Verb-related Projections among Adverb Functional 

Projections……………………………………………………………… 23 

3. Word Orders in Romanian. Romanian VSO and Adverb Placement…… 32 

4. Romanian SVO and Adverb Placement………………………………… 51 

5. Romanian VOS and Adverb Placement………………………………… 75 

6. The Distribution of Adverbs in the Vorfeld…………………………… 84 

7. The Case of Clitic Adverbs……………………………………………… 90 

8. Conclusions……………………………………………………………… 98 

III. The Left Periphery in Romanian: An Overview……………………………. 103 

1. Introduction……………………………………………………………… 103 

2. Topics and Foci………………………………………………………… 105 

3. The Cartographic Approach vs. the Generalised TP Approach………….. 122 

4. Rounding off the Picture of the Romanian Left Periphery………………..130 

IV. Wh-Movement in Romanian 

1. Introduction………………………………………………………………. 131 

2. Previous Approaches to Wh-Movement…………………………………. 131 

2.1. The Adjunction Analysis…………………………………………....133 

2.2. The Wh-Cluster Analysis………………………………………….. 134 

2.3. The multiple Spec analysis………………………………………….136 

2.3.1. The multiple Spec IP analysis……………………………… 136 

2.3.2. The multiple Spec CP analysis……………………………. 137 

2.4. The split CP analysis……………………………………………….. 138 

2.4.1. The split CP analysis: Grohmann (2002)………………….. 138 

2.4.2. The split CP analysis: Krapova & Cinque (2005)…………. 139 

3. The Ordering of Wh-phrases in Romanian………………………………. 140 

3.1. The order of bare wh-phrases in Romanian……………………….. 140 

3.1.1. [+Human] Wh-Subjects and Objects………………………..140

3.1.2. Bare Wh-Subjects and Objects: [+Human] and [-Human] Wh- 

Phrases…………………………………………………...... 141 

3.1.3. The order of wh-adjuncts………………………………….. 144 

3.1.4. [+/-Human ] Wh-Subjects and Objects and Wh-Adjuncts.... 144 

3.2. The order of D-linked and bare wh-phrases……………………….. 150 

4. Wh-chunk Movement in Romanian…………………………………….... 153 

4.1. Krapova and Cinque’s (2005) Analysis……………………………. 153 

4.2. Wh-Chunk Movement……………………………………………....160 

4.3. The Syntax of D-linked Wh-Phrases………………………………. 169 

5. The Case of a Wh-in-situ in Romanian: ‘Why’……………………….. ....172 

5.1. The Position of dacă ’if’ within the Romanian Left Peripehry..........174 

5.2. De ce - Its Properties……………………………………………….. 175 

5.2.1. (Non-)inversion…………………………………………….. 176 

5.2.2. Focus Sensitivity…………………………………………… 178 

5.2.3. Interaction with Topicalisation……………………………...180 

5.2.4. Negation in Long-Distance Dependencies…………………. 180 

5.2.5. Quantificational Adv in Long-distance Dependencies……...181 

5.2.6. Wh-phrases – Main contexts……………………………….. 181 

5.3. The syntax of de ce. Intervention effects with de ce.......................... 183 

6. Conclusion...................................................................................................191 

V. Japanese Wh-scrambling in the Mittelfeld and Vorfeld……………………….195 

1. Introduction………………………………………………………………. 195 

2. A Cartographic Approach to Japanese: Some Implications………………196 

3. Short DP ‘Scrambling’………………………………………………….. 200 

3.1. ‘Scrambling’: A Definition………………………………………… 201 

3.2. Short DP Scrambling in SOV and IS features…………………....... 203 

3.3. Short DP Scrambling in OSV and IS features………………………207 

4. Japanese Wh-Phrases: Mittelfeld-internal ‘Scrambling’………………… 211 

4.1. Are Wh-Phrases Really In-situ or Do They Move?........................... 213 

4.1.1. The Interaction of Wh-Phrases and Adverbs………………..213 

4.1.2. Canonical and scrambled wh-phrases. The issue of 

optionality………………………………………………….. 216 

5. Long ‘Wh-Scrambling’…………………………………………………... 223 

5.1. The ‘undoing’ property and optionality………………………….... 223 

5.2. Approaches to wh in situ…………………………………………… 224 

5.3. Islands in Japanese……………………………………………….... 227 

5.4. Wh-islands and overt long-extraction. Reconstruction in Japanese.. 232 

5.5. Conclusion…………………………………………………………..242 

VI. A Cross-linguistic Typology of Question Formation………………………… 245 

1. Introduction: the Q- and Wh-features………………………………….... 245 

2. Antisymmetry and the Interrogative ka in Japanese…………………….. 249 

3. Languages with an Overt Q Particle. Covert Q-Feature Languages…...... 257 

4. Languages with a (C)overt Wh Feature…………………………………. 264 

5. A Case of Q [+EPP] and Wh [-EPP] Language: Sinhala…………………273

6. A Note on the Locality of Selection…………………………………….. 281 

7. Concluding Remarks……………………………………………………. 283 

VII. Concluding Remarks…………………………………………………………...285 

Bibliography................................................................................................................ 291

ACKNOWLEDGEMENTS 

I have been waiting for this moment – the moment to write the Acknowledgments section 

– for a long time, actually for three years and eight months. I received much help along 

the way from many people and events. Here are the names I am to gratefully name. 

I begin with my committee board. I thank you all for inculcating such enthusiasm 

and desire in all of us, the graduate students, to always move forward in our research. 

I would like to express my gratitude to Ur Shlonsky, director of my dissertation, 

whose precious advice on the many meetings together, patience and good-humoured 

nature have been of immense help. I thank him for taking me on as an assistant on a FNS 

project Micro-parameters in Romance Wh-Movement, from which I could benefit to a 

large extent in my dissertation. I also thank him for the confidence he has shown in 

giving me various classes to teach, from beginners in syntax to advanced students. Ur 

taught me to ask the right questions and to strive to make as many connections as 

possible. 

My gratitude also goes to Eric Wehrli, president of the board, who took me on in 

my first year of assistantship to partly work on the Romanian FIPS (French Interactive 

parsing system) and whom I thank for his trust in me. 

I am also very fortunate to have worked with Luigi Rizzi. He read my handouts 

with meticulous care, going at the depth of every detail I would have put down. His 

comments are extremely insightful and encouraging. After each meeting with him, I 

would make a refinement in the analysis or even a substantial change which would take 

me to a completely different direction. 

My gratitude also goes to Guglielmo Cinque whose comments and discussions were 

of great help to me. His insightful comments broadened my perspective on the 

theoretical issues I was dealing with. 

I am also hugely greatful to Alexandra Cornilescu for her generosity and catching 

enthusiasm in teaching and, more generally, in doing linguistics. I cannot thank her 

vii

enough for the word of advice she has always given me and for having her house more of 

a public linguistics library for those of us not living in Bucharest. 

I owe so many thanks to Yohio Endo for his patience and readiness in clarifying 

things about Japanese for me. An important amount of what I know today about Japanese 

I learnt from him. 

I am hugely grateful to so many other people. 

I am so grateful to Christopher Laenzlinger for his support in lots of forms, precious 

advice and encouragement over the years. Discussions with him managed to soothe my 

fears about parts of my account I would suddenly become unsure of. I cannot thank him 

enough for his support and encouragement at times when I would be discouraged. 

My gratitude also goes to Jacques Moeschler for his advice, generosity and stunning 

enthusiasm in everything he does. His systematic questions of how I would progress in 

the writing process somehow made me more reassured. I have always appreciated his 

sense of confidence in the people who work with him. Many of us duly think of him more 

of a ‘papa des assistants’. 

One big thank you goes to Genoveva Puskas for her kindness and readiness in 

discussing things with me. I really appreciated the help she has proposed to me these past 

months to teach for me. One cannot be luckier than that! Merci, Geno! 

I would also like to thank Eric Haeberli and Paola Merlo for their comments and 

suggestions on different parts of the dissertation that I presented on various occasions. 

My thanks also go to my fellow colleagues, present and former, both in the syntax section 

and the computational linguistics section. I thank Lena Baunaz, Greg Ellison and Nina 

Rojina for the discussions we had together on several occasions. Thanks Greg for being 

such an amazing team worker! Thanks to Goljihan Kachaeva, Eszter Varga, Tabea Ihsane 

and Stephanie Dürrleman; to Joanna Blochowiak, Athina Michou and Antonio Leoni de 

Leon; to Jean-Philippe Goldman (la JP-ette), Luka Nerima, Tanja Samardzić, Violeta 

Seretan Palotta, Lorenza Russo, Yves Scherrer, Alexis Kauffmann, Lonecke van der Plas. 

I also thank Eva Capitao, our secretary, whose generosity and constant support are 

to admire so. 

I am immensely grateful to people who took the time to answer my so many 

questions about Romanian and other languages. One special thank you to Cristina Soare 

viii

and Corina Andone for having been my constant informants from the very beginning and 

for always replying very quickly to my so many emails. Thanks also to Mihaela Dogaru, 

Carmen Miresan, Madalina Tenea, Adriana Spiridon, Oana Lungu, Imola Farkas, Mara 

Radulescu, Daniela Tuchel and Catalina Necula for their help with Romanian at some 

point or other. For help with Spanish, thanks to Vidal Valmala and Antonio Leoni de 

Leon. For French, I thank Christopher Laenzlinger, Jacques Moeschler, Alexis 

Kauffmann, Genoveva Puskas, and Lena Baunaz. For help with Japanese, thanks to 

Yoshio Endo and Izumi Tahara. Thanks to Greg Ellison and Daniel Teylor for help with 

English data. For help with Russian, I thank Nina Rojina and Goljihan Kachaeva. Joanna 

Blochowiak kindly provided me with judgments on Polish and Tanja Samardić on Serbo- 

Croatian. Again, thank you! 

Further, I would like to thank several people who discussed aspects of the materials 

in this dissertation with me at various moments along the way: Adriana Belletti, Anna 

Cardinaletti, Kleanthes Grohmann, Iliyana Krapova, Shigeru Miyagawa (for parts that I 

only partially included here). 

I also benefited from comments, suggestions and criticism on distinct parts of the 

material presented here at several conferences and workshops and I cannot but gratefully 

thank all those audiences. 

There is of course a life outside syntax and linguistics and I have been very lucky to 

have so many very good friends in Geneva and elsewhere and I must thank each and 

every of them for sharing so many precious moments with me: Paola Minen (thanks for 

your thoughtful words of encouragement), Dragana Lukajic (I loved our discussions on 

the philosophical and spiritual side of life), Chiara Fiorentini, Saskia Seeger and Annik 

Baumagartner (thanks esp. for our wonderful Christmas dinners), Joanna Blochowiak, 

Nina Rojina, Goljihan Kachaeva, Aluna Cosbuc, and Julia Giese. 

One huge thank you goes to my family: my sister, Cristina, my brother, Iulian, and 

my parents, Gigi and Victoria, for their love and support, and for their belief in me all 

these years. 

I dedicate this dissertation to Christopher, and to my parents. 

ix

CHAPTER I 

INTRODUCTION 

This dissertation explores a number of issues involving the clausal structure of Romanian, 

in particular the structure of the Romanian IP field, i.e. the Mittelfeld, and the left 

periphery, on a comparative basis with other Romance languages. Regarding this latter 

issue the dissertation insists on wh-phrases proposing an analysis in lines with current 

Minimalist guidelines. Considering that Romanian, a multiple wh-fronting language, is 

situated at one end of the wh-continuum, the dissertation also studies the behaviour of 

wh-phrases in a language situated at the other end, i.e. a ‘scrambling’ language, Japanese. 

The discussion is framed within the idea that syntax may directly access features 

related to Information Structure that may be ‘floating’ on distinct DP- and verb-related 

projections inside the IP-field, or the Mittelfeld, or project themselves inside the CP area 

(i.e. Rizzi 1997, 2001a, 2004a). 

The theoretical implications and predictions related to wh-movement and wh- 

scrambling and lack thereof are all explored by the end of the dissertation where a 

typology of question formation is proposed. 

1.1. Kayne’s Anytisymmetry Hypothesis 

At the basis of the theoretical apparatus to employ lies Kayne’s (1994) antisymmetry 

proposal which unifies hierarchical structure and linear ordering of the constituents of a 

sentence. Following the antisymmetry, specifiers are unique left-branching adjoined 

phrases. In other words, the only case of phrasal adjunction is that of specifiers. Multiple 

1

adjunctions, or multiple specifiers for that matter, are thus banned. Syntactic structure is 

universally of the form Spec-Head-Complement. The complement of a head invariably 

follows the head and the specifier invariably precedes both head and complement. This 

universal order is conceived of as holding of all stages of the derivation, both prior to 

movement and after movement. Antisymmetry is therefore more restrictive than previous 

approaches to syntactic analysis in terms of the theoretical apparatus allowed. 

It obviously poses an interesting challenge for OV languages. Starting from 

Kayne’s (2005) discussion of “Antisymmetry and Japanese”, this dissertation will 

discuss, in Chapter V, some implications for the behaviour of wh-phrases in the Japanese 

Mittelfeld, under the widely-accepted assumption that it is a ‘scrambling’ language, with 

‘wh-in-situ’. This latter notion will be challenged and it will be shown that wh-phrases 

need not remain in situ rather they move to distinct positions that are to be identified 

inside the Mittelfeld. Both the wh-object/subject and their non-wh-counterparts are 

shown to vacate the vP domain. Relying on antisymmetry and making use of a more 

complex theoretical apparatus, Chapter VI proposes a typology of question formation 

which takes into account, OV languages like Japanese, Korean and Sinhala, besides more 

familiar SVO languages. It will be argued that these languages make recourse to piedpiping 

of subclausal structures in forming questions. 

1.2. Cartography of Syntactic Structures 

Instrumental to this study is the cartographic approach to syntactic structures, such as has 

been put forth by Rizzi (1997, 2004a), Cinque (1999, 2002a, 2006), Belletti (2004) a.o., 

where the CP (or the Vorfeld) and IP (or the Mittelfeld) are areas rich in functional 

projections, roughly represented in (1). 

(1) Force Top* Int Top* Foc Mod* Top* Fin Mood … Mod T Mod Asp … Voice … Asp 

The underlying idea is to arrive at a as detailed as possible a map of the structure of the 

clause (and its phrases, i.e. DP) starting form the assumption that all languages share the 

same principles of structure formation and the same functional projections. What varies 

2

among languages is the types of movement operations and the overt realisation of certain 

functional projections. 

As illustrated in (1), both the left periphery of the clause and the IP field are split in 

several projections. One major advance in this direction is Cinque’s (1999) fine-grained 

analysis of the adverb positions inside the Mittelfeld where the appropriate functional 

heads express properties of various types of mood, mode, tense and aspect. Crucially, all 

these functional projections are hierarchically ordered, as in (1). This universal hierarchy 

is argued to be provided by UG. 

(2) 

[Frankly Moodspeech act > [surpisingly/fortunately Moodevaluative > [allegedly Moodevidential > [probably 

Modepistemic > [once Tpast > [then Tfuture> [perhaps Mod(ir)realisis > [necessarily Modnecessity > [possibly 

Modpossibility > [willingly Modvolitional > [inevitably Modobligation > [cleverly Modability/permission > [usually 

Asphabitual> [again Asprepetitive> [often Aspfrequentative > [quickly Aspcelerative> [already Tanterior> [no longer 

Aspperfect> [still Aspcontinuative > [always Aspperfect > [just Aspretrospective > [soon Aspproximative > [briefly 

Aspdurative > [characteristically Aspgeneric/progressive > [almost Aspprospective > [completely AspSgCompl etive(I) > 

[tutto AspPlCompl > [well Voice > [fast/early Aspcelerative(II) > [completely AspSgCompletive(II) > [again 

Asprepetitive(II) > [often Aspfrequentative ]]]]]]]]]]]]]]]]]]]]]]]]]]]]] 

Adverbs thus occur in a Spec-head configuration with the appropriate head where they 

value substantive (interpretable) features. Under this view of adverbs as unique specifiers 

of semantically related projections, the clause structure contains a large number of 

functional projections and even many more on the hypothesis that there is DP- and verbmovement 

among distinct classes of adverbs. This will mainly be the focus of Chapter II 

where the Romanian Mittelfeld is compared to Spanish and Italian. 

Cartography is therefore an articulate model of linguistic research which offers 

solid means to uncover loci of linguistic variation. Within the framework adopted here, a 

study of a particular grammatical system or of several systems can serve to enhance our 

understanding of those systems – Romanian (within Romance), Japanese – and ultimately 

of the underlying principles of UG. 

3

1.3 Syntax-Information Structure Interface 

If adverbs merge in the appropriate specifiers of distinct, rigidly ordered functional 

projections where they can value their features, DPs and verbs can be found interspersed 

among them (Cinque 1999). It will be the main aim of Chapter II to identify the variable 

positioning of the subject and object DPs as well as the verb among the rigid projections 

of adverbs. This is will be done on a comparative basis with other Romance languages. 

In focusing on the structural cartography of the Romanian Mittelfeld, distinct word 

orders are considered and it will be suggested that distinct projections in the Mittelfeld, 

interspersed among rigidly ordered adverb functional projections (Cinque 1999), contain 

features related to Information Structure. The features to consider here are New 

Information Focus, Contrastive Focus, Topic-aboutness, Topic (associated with a 

comment) and Contrastive Topic. In the model of the grammar adopted in this 

dissertation a DP may enter the derivation not only with phi and Case features to get 

valued but also with feature related to IS, which it takes from the Numeration. The view 

to defend is that at least some positions in the Mittelfeld are associated with distinct IS 

features. This leaves open the possibility, largely acknowledged (Rizzi 2006a, 

Cardinaletti 2004), that some DP movement may be triggered purely by phi- (and case-) 

features. It is in this precise way that syntax interfaces with Information Structure. Syntax 

thus directly accesses IS-features. The claim is made that such IS-feature are intimately 

related to EPP. Movement induced by an EPP associated with an IS-feature may be seen 

to derive from a version of Fox’s (2000) principle which basically requires that 

movement have an effect on the output. 

The same line of argumentation is used in Chapter V on Japanese. We follow 

Kayne (1994, 2005) and hypothesise that the OV order is derived, and is the result of all 

V’s arguments having moved into Specs of higher functional projections inside the 

Mittelfeld to have their IS-features (i.e. Topic and Focus) valued. More generally, the 

notion of ‘scrambling’ will be understood as Topic- or Focus-driven movement. 

4

1.4. Criterial Freezing 

Importantly, these IS-features driving movement inside the Mittelfeld are not criterial 

features but are to be understood as EPP-related. They are also some sort of ‘floating’ 

features, floating therefore on Subject- and Object-related projections, themselves, 

interspersed among distinct adverb-related projections. According to the Criterial 

Freezing approach (Rizzi 1997, 2006b, 2007, 2008; Rizzi & Shlonsky 2007), a criterial 

position is a position expressing a scope-discourse property in Chomsky’s (2000) 

terminology and gives rise to a Criterion in the sense of Rizzi (1997). In other words, a 

chain ‘stops’ at a criterial position associated with a particular interpretive property. It 

reads as in (3): 

(3) Criterial Freezing: in a criterial configuration, the Criterial Goal is frozen in place. 

Rizzi 1997, 2006b, 2007, 2008 

One consequence of the criterial approach is that the functional heads in the left periphery 

hosting IS-features (or scope-discourse features) - the Criterial probes – which attract 

specific elements in their specifiers - Criterial goals – are ordered in a fixed way (see (1)). 

It is the focus of Chapter III to investigate how Romanian fits in with this fixed split CP 

component. Any departure from (1) or absence of a certain projection in Romanian but 

present in (1) needs accounting for. Romanian accommodates well with (1) but also 

exhibits specific properties. 

1.5. Locality and Rizzi’s (2004b) Relativised Minimality 

The features in the split CP area, or criterial features, are sensitive to syntactic principles 

like locality principles. This point will be made esp. in Chapter IV and V when discussing 

some cases of (un)grammaticality effects with wh-extraction out of a wh-island in 

Romanian and Japanese. This substantiates the claim made in Starke (2001) and extended 

under Relativised Minimality (Rizzi 2004b) that the feature make-up of both intervenor 

and extractee has an essential role to play. 

5

Therefore I follow in this dissertation Rizzi’s RM. RM takes the relevant intervenor 

to be relativised to the nature of the structural relation to establish between X and Y in the 

relation in (4). 

(4) …X…Y…Z 

Rizzi’s definition of Relativized Minimality, which is a condition on chains (5), is 

defined in terms of Minimal Configuration (6). Local relations must be satisfied in a 

minimal configuration, actually the smallest configuration possible. This configuration is 

defined in (6). 

(5) (A1, …, An) is a chain iff, for 1 ≤ i < n 

(i) A1 = Ai+1 

(ii) Ai c-commands Ai+1 

(iii) Ai+1 is in a Minimal Configuration with Ai. 

(6) Y is in a Minimal Configuration with X iff there is no Z such that 

(i) Z is of the same structural type as X, and 

(ii) Z intervenes between X and Y. 

The ‘structural-type’-based definition of RM is replaced by ‘feature type’, where the 

features relevant for the computation are those of Rizzi (2004b): 

(i) Argumental: person, number, gender, Case, … 

(ii) Quantificational: Wh, Foc, Neg, measure… 

(iii) Modifier: evaluative, epistemic, Neg, frequentative, celerative, measure,… 

(iv) Topic 

RM is thus construed as an economy principle in that it severely limits the portion of the 

structure within which the local relation between X and Y in (1) is computed. RM 

(2004b) is a move forward from RM (1990) in that the former involves a more finegrained 

typology of intervening A’ specifiers which trigger minimality effects on A’ 

chains. In other words, it successfully accounts for selective locality effects, while at the 

6

same time extending to head-XP interactions and some cases of ellipsis. By and large, the 

original RM (1990) version stated that a chain relation fails if a position of the same kind 

as the target (6) intervenes (basically the A/A’ distinction is made). However, this version 

of RM fails to account for some cases of selectivity with certain types of intervenors (i.e. 

the well-known distinction between quantificational adverbs like peu ‘little’ or beaucoup 

‘a lot’ versus non-quantificational adverbs like attentivement ‘carefully’, which belong to 

the modifer class in the 2004 version): 

(7) a. Combien de films a-t-elle peu aimés ? 

‘How many films did she little like?’ 

b. * Combien a-t-elle peu aimé ? 

(8) a. Combien de films a-t-elle attentivement regardés ? 

‘How many films did she carefully watch?’ 

b. Combien a-t-elle attentivement regardé ? 

As mentioned, RM (1990) makes too gross a distinction, i.e. the A/A’ split, and fails to 

account for cases like (7)-(8) which can be fully accounted for under RM (2004). 

1.6. Types of AGREE 

Also part of the theoretical apparatus underlying the dissertation are some Minimalist 

ideas related to feature valuing/checking. Chomsky (2000, 2001) proposes a theory of 

AGREE elaborating ATTRACT of Chomsky (1995). The uninterpretable features of a probe 

α and a goal β are deleted under the structural relation (9), subject to the matching 

Condition, as defined in (10). 

(9) AGREE (Chomsky 2000) 

α β 

AGREE (α, β) 

(10) Match (Chomsky 2000) 

a. Matching is feature identity. 

7

. D(P) is the sister of P. 

c. Locality reduces to ‘closest c-command’. 

In addition to AGREE, the computational system also relies on movement which is internal 

Merge, i.e. the linguistic representation built up is merged with a subpart of itself and the 

result is a copy of the subtree as sister to the root. In the Minimalist Programme, MOVE is 

only instantiated by an EPP-feature present on the probe. Besides the usual single AGREE 

operation, Hiraiwa’s (2000) mechanism of MULTIPLE AGREE defined in (11) is also 

adopted in this dissertation. 

(11) Hiraiwa’s MULTIPLE AGREE (2000:69): 

MULTIPLE AGREE (multiple feature checking) with a single probe is a single simultaneous 

syntactic operation; AGREE applies to all the matched goals at the same derivational point 

derivationally simultaneously. 

I believe there is compelling empirical evidence for (overt) multiple feature 

valuing/checking in Romanian. This will be discussed in Chapters IV, V and VI. The 

advantage of this refined theory of multiple feature valuing/checking is that it does away 

with ‘tucking-in’ which is incompatible with antisymmetry (and, consequently, with 

cartography). Crucially, it takes place derivationally simultaneously. 

The cartographic approach coupled with minimalist ideas provide therefore a 

theoretical apparatus that allows one to raise questions, formulate and test hypotheses and 

predictions, and ultimately discover new facts. 

These ideas underlie the present dissertation which studies a number of issues in the 

syntax of Romanian on a comparative basis with other Romance languages and, from the 

perspective of the behaviour of wh-phrases, with Bulgarian essentially, as well as, 

apparently surprisingly, Japanese. 

1.7 The Organisation of the Dissertation 

The dissertation is organized as follows. One of the major issues addressed in Chapter II 

is the architecture of the Romanian Mittelfeld and the unveiling of micro-parameters 

8

elated to subject-, object- and verb-movement possibilities which distinguish it from 

other Romance languages and in particular from Spanish and Italian. It draws on 

Cinque’s (1999) suggestion who acknowledges the existence of several DP-related 

positions interspersed among the fixed positions of adverbs. I will show that in addition 

to the ‘usual’ subject (at least two for Italian and Romanian; see Cardinaletti 1996, 

Cornilescu 1997) and object positions, which are mainly responsible for Case and phifeature 

valuation/checking, the Romanian Mittelfeld also contains DP-projections 

associated with distinct Information Structure (IS) features (or interpretive properties), 

which are realized on Subj(ect) and Obj(ect) projections. The hypothesis will be 

entertained that the EPP feature inducing movement is closely associated with such an ISfeature. 

In so doing, the rigid placement of adverbs is taken as a crucial criterion in 

testing the movement possibilities of the subject, direct object and verb in the VSO, SVO 

and VOS word orders. These movement possibilities are also tested in Spanish (VSO, 

SVO, VOS) and Italian (VSO, SVO and more restrictively VOS). 

The fact that at least some IS readings are obtained in the Mittelfeld does not imply 

that it contains discrete IS-related projections. Rather, the locus of Topic and Focus 

projections, where scope is calculated, and which are (purely) A’ positions and criterial 

positions, is the left periphery. The Topic- and Focus-related features existing in the 

Mittelfeld are to be viewed more like ‘floating’ non-Criterial features and as parasitic on 

Case- and phi-features giving rise to distinct word orders. 

One major assumption that I will make is that whereas the Nominative of a DP is 

valued via AGREE, the highest position a preverbal subject may fill is Rizzi’s (2006a, 

2007) and Rizzi and Shlonsky’s (2007) Criterial SubjP. It will be shown that, as in 

Italian, Romanian SVO order is a marked order, where the subject is more prominent and 

is attracted to this criterial position where it values/checks a Topic-aboutness feature. 

If adverbs represent a solid testing ground for argument and verb movement 

possibilities, they also come to be under scrutiny. In particular certain classes of adverbs 

are shown to move to the left periphery of the clause (the Vorfeld), where they can 

occupy Modifier, Focus and Topic (under specific discourse situations), being thus 

licensed by the appropriate features on the respective head. When the adverb fills the 

Spec Focus position, the adjacency with the verb is required, this being a more general 

9

phenomenon specific to Romanian (and Spanish). A restricted class of adverbs is 

discussed which has clitic-like behaviour in that they can only fill a preverbal position, 

adjacent to the whole series of elements cliticising on the verb (pronouns and auxiliaries). 

Romanian exhibits another special class of fairly short adverbs, i.e. truly clitic 

adverbs. These adverbs do not constitute a homogenous class, some presenting 

idiosyncratic behaviour with respect to negation in simple and/or compound tenses or as 

far as the position with respect to the auxiliary is concerned. As will be seen, all of them 

are aspectual adverbs belonging to distinct aspect classes in Cinque’s (1999) functional 

hierarchy. 

Chapter III starts with an outline of the structure of the Romanian left periphery 

within the cartographic framework while insisting on the main focus of discussion, whphrases. 

It offers several arguments in favour of this stand, i.e. accommodating Romanian 

to cartography. From this perspective, an analysis of the Wh/Foc-V adjacency, specific to 

Romanian (and Spanish) is forwarded in terms of Rizzi’s Criteria. 

In Chapter IV the ordering of wh-elements is considered in more detail under the 

widely-acknowledged fact that this ordering is quite rigid. At this point Romanian is 

compared to Bulgarian as described by Krapova and Cinque (2005) (henceforth K & C). 

It will be seen that, roughly speaking, both languages make use of a residual hierarchy of 

Animacy, i.e. wh-arguments are ordered in function of the [+human] feature. Once this 

ordering is arrived at, an analysis is suggested that meets the requirement, specific to 

Romance, that there be a single Foc position at the left periphery. It also preserves the 

idea formulated in K & C that multiple wh-phrases at the left periphery is pre-established 

in the Mittelfeld. Essentially, the analysis for multiple wh-phrases to suggest in this 

chapter exploits the mechanism of MULTIPLE AGREE under the assumption that Foc° 

probing for the wh/Foc-feature is marked [+multiple]. 

Besides questions with two or even three wh-phrases, D-linked wh-phrases are 

shown to fill a dedicated position within the wh-domain and are also provided an analysis 

implying AGREE from Foc° and movement to a dedicated Topic-like position associated 

with special interpretive properties (i.e. givenness or topicality). If feature valuation 

involves the wh-feature on Foc°, then this is expected to have consequences on the 

10

positioning of the D-linked wh-phrases with respect to other types of Topics in the left 

periphery. It will be seen that this is indeed the case. 

This chapter ends with a discussion of a ‘wh-in-situ’ in Romanian. This is the case 

of de ce ‘why’, which scopes over the entire clause. The point will be made that though in 

the split CP, it merges fairly low in this area whence it may or must move higher, i.e. to 

Rizzi’s (2001a) Spec IntP. 

Chapter V represents an attempt at analyzing some Japanese facts along the lines of 

Kayne’s antisymmetry hypothesis. More precisely, the same hypothesis, set forth in 

Chapter II, namely that a DP with the appropriate features taken from the Numeration can 

target a position associated with TopP or FocPNew Information features is tested for the 

Japanese Mittelfeld. The hypothesis that both the subject and the object leave the vP 

domain in an OV language is tested with different classes of adverbs. The question will 

also be addressed whether in a matrix context wh-phrases remain in-situ or move. In 

order to answer this question, the test with adverbs will prove most useful. 

Another focus of the discussion of this chapter is represented by indirect questions 

and some cases of extraction out of a wh-island. Before dealing with these topics, an 

analysis of (matrix/indirect) question formation will be provided which takes into account 

the existence of an overt Q-particle, responsible for clausal typing, which is absent in the 

Indo-European languages seen up to this point in the dissertation, coupled with the 

existence of a covert wh-feature on Foc°. The point is made that the absence of an EPP 

related to Foc° accounts for the absence of wh-movement in Japanese. The conclusion to 

reach is that after all, an antisymmetric approach to Japanese seems worth undertaking. 

Given the analysis of wh in a ‘multiple wh-fronting language’ like Romanian and in 

a ‘wh-scrambling language’ like Japanese, as well as the comparative standpoint and 

theoretical apparatus that cartography endows one with, it appears worth looking into a 

typology of question formation. This is the gist of Chapter VI. The typology relies onto 

two features, i.e. the Q-feature and the wh-feature, either plausibly associated with EPP. 

This parameter leads to distinct language types. Another regards the overt/covert 

morphological realization of Q/wh. These combinatorial possibilities account for the 

existence of wh-movement in some languages (like Vata and Tlingit), on the one hand, 

11

and French and Romanian, on the other, and its absence in Japanese, Sinhala, Chinese, 

Tumbuka and French on the ‘in-situ’ strategy. 

12

1. Information Structure and Syntax 

CHAPTER II 

THE ROMANIAN MITTELFELD 

It is a well-known fact that Information Structure affects syntax in some particular ways 

though the interaction between these two modules is still not very well understood. In a 

wide range of languages, information structure is closely interwoven with syntactic linear 

and hierarchical structure, semantic structure, prosodic phonological structure and 

morphological structure (particularly in Japanese and Korean, or Gungbe). 

Two major approaches to Information Structure (henceforth IS) have been proposed 

in the literature: one claiming that such effects have a prosodic trigger (Zubizaretta 1998; 

see also Gussenhoven 1992, Cinque 1993, Selkirk 1995, Ladd 1996, a.o.), or occur in the 

PF component (Holmberg 1999). The other approach argues that certain discourse 

functions, such as Topic and Contrastive Focus may be encoded by movement into 

designated syntactic positions, typically at the left periphery of the clause (cf. Rizzi 

1997), and constitute therefore core syntactic operations. Below a few general remarks 

about both approaches are made. 

The general view for more than two decades has been that syntax and phonology 

interact in a global way: phonological rules apply at the output of a complete derivation 

(Chomsky 1981). In some Minimalist work (Uriagereka 1997, 1999, Chomsky 1998, 

1999) which adopts a multiple Spell-Out architecture of the syntactic component, the 

syntactic structure as has been created up to a certain point in the derivation undergoes 

Spell-Out and is given a phonological representation. The derivation continues with 

13

additional instances of Spell-Out and consequently of phonological representations. More 

recent versions of Minimalism (Chomsky 2001, 2004, 2005) strengthen the notion of 

cyclic derivation, i.e. the derivation must proceed phase by phase, where the two strong 

phases are CP and v*P. Under the cyclic view of the derivation, Fanselow (2004) adopts 

a radically different position by arguing that phonological rules apply prior to movement 

of a constituent and that it is therefore phonological properties that determine which 

categories move. 

Within the approach that IS effects are determined by PF considerations, a 

considerable amount of attention has been given to the relation between focus and 

prosody. For Romance, it is generally claimed that focal stress is determined by some 

version of the Nuclear Stress Rule (NSR), as can be found in work by Cinque (1993) and 

Zubizaretta (1998). Cinque’s theory of phrasal stress is syntax-dependent, and is meant to 

capture the generalisation that nuclear stress falls on the most embedded element on the 

recursive side of the tree. For Zubizaretta, focus is directly identified by prosodic notions, 

such as relative prominence. It is argued that sentence-final prominence is attributed to a 

phrasal stress rule. In a system which puts to work NSR and what she calls the Focus 

Prominence Rule (FPR), the focused constituent is argued to contain the intonational 

nucleus of the intonation phrase, i.e. it must contain the syllable with main phrasal 

prominence. 1 The interplay of both rules determines which element is more prominent 

given two sister nodes. 

The second major approach, which I will adopt in this thesis, is syntactic in nature. 

In Rizzi (1997, 2001, 2004a) the rich articulated left periphery of the clause (CP) contains 

elements expressing scope-discourse semantic properties (see also subsequent work by 

Poletto 2000, Benincà and Poletto 2004, etc). The split CP constitutes therefore the 

domain of anchoring of syntax to the syntax-discourse interface, where discourse may be 

understood as IS. More precisely, the complementizer system is conceived of as a 

structural zone consisting of distinct functional projections, much as the IP system 

1 

Zubizarreta (1998) and others following her (Reglero 2007) attempt to explain distinct syntactic 

phenonema in Spanish by the Nuclear Stress (NS) adopting p(rosodic)-movement of any element to the 

right of the element bearing NS. For counterarguments, see for instance, Etxepare and Uribe-Etxebarria 

(2008). 

14

(Cinque 1999) and the DP system (Cinque 1994, Cinque ed. 2002, Laenzlinger 2005, 

a.o.). This fixed component minimally involves the heads specifying Force and 

Finiteness, and an accessory component involving IS-related elements, such as 

(recursive) Topic(s) and Focus, which are activated when needed. In recent years, the 

idea has been developed that it is not only the CP area that may contain topic and focus 

elements, but also the IP internal low area. Belletti (2001, 2004) argues that in Italian the 

area immediately above VP displays significant resemblance to the left periphery. In 

particular, a Focus projection surrounded by Topic positions is identified. Essentially, 

such positions are associated with interpretations and intonations different from those at 

the left periphery. More on this Focus position will be said in section 2 on a comparative 

basis with another Romance language, Romanian. It has been shown that some SOV 

languages also have an IP-internal Focus position. A case in point is Malayalam, which, 

as Jayaseelan (2004) argues, has a Focus projection immediately above vP, which hosts 

wh-phrases (or pied-piped clauses containing a wh-phrase). Along the same lines, 

Chinese, an SVO language, has been argued to display an IP-internal domain containing a 

Topic and Focus projection above vP (Tsai 1994, Ernst and Wang 1995, Shyu 1995, Paul 

2002). 

In this chapter, focusing on the structural cartography of the Romanian IP-field, or 

‘Mittelfeld’, distinct word orders are considered and it will be suggested that distinct 

projections in the Mittelfeld, interspersed among rigidly ordered adverb functional 

projections (Cinque 1999), contain features related to Information Structure. The model 

of the grammar adopted is that of Chomsky (2000, 2001), which relies on the operations 

of Merge, which takes two objects and combines them into a larger phrase, and Agree 

between a probe and a matching goal. If a probe has an EPP-feature, the goal is piedpiped 

to the specifier of the probe. Thus, a DP enters the derivation with phi and Case 

features which it takes from the Numeration and such features need valuing. Much in the 

spirit of Laenzlinger (to appear) and Aboh (2007), the suggestion is made that in addition 

to these, the Numeration also contains Information Structure-features, where the relevant 

features to consider are Topic(-related) and Focus(-related). Laenzlinger takes the strong 

position and shows that DP movement in the Mittelfeld is triggered by Case coupled with 

some IS-feature valuing. However, the view to defend here is that at least some positions 

15

in the Mittelfeld are associated with distinct IS features. This leaves open the possibility, 

largely acknowledged (Rizzi 2005, Cardinaletti 2004), that some DP movement may be 

triggered purely by phi- (and case-) features. Thus, for instance, as argued for in Rizzi 

(2005), in null subject languages (NSL), the DP subject leaves its merge position and 

moves to a position to check its phi-features, whereas the expletive fills the position 

associated with properties of scope-discourse semantics (hence satisfies Rizzi’s 2005, 

2006, 2007 and Rizzi and Shlonsky’s Subject Criterion 2007). Cardinaletti (2004) also 

argues that there is more than one position that hosts distinct types of subjects: there is an 

EPP position hosting expletives, pro and (certain) weak pronouns, and a higher one 

where full DPs, strong pronouns and special weak pronouns (the Italian egli ‘he’) check 

the Subject-of-predicaiton feature. I will come back to these issues. Whether all 

Mittelfeld-internal DP positions resulting either from internal merge or external merge 

must be associated with an EPP understood as IS-related is a matter of future research. 

The view is upheld here that the specifier formed as a result of second merge may 

thus be associated with a particular interpretation. Movement induced by an EPP 

associated with an IS-feature may be seen to derive from a version of Fox’s (2000) 

principle which basically requires that movement have an effect on the output. 2 The fact 

that least some IS readings are obtained in the Mittelfeld does not imply that it contains 

discrete IS-related projections. Rather, the locus of Topic and Focus projections, where 

scope is calculated, and which are A’ positions, is the left periphery. Put differently, the 

underlying hypothesis is that syntax dictates the positions potentially available for 

arguments and verbs in the Mittelfeld, Vorfeld (left periphery) and Nachfeld (the lowest, 

right periphery of the Mittelfeld) while the information structure assigns a particular 

argument in a particular position at some point in the derivation a certain value. 

2 

Fox (2000) discusses cases of optionality, more exactly of seemingly optional quantifier raising. His 

Scope Economy principle reads as follows: 

A Scope Shifting Operation can move XP1 from a position in which it is interpreted only if the movement 

crosses XP2 and is not scopally commutative (Fox 2000:26). 

However, the core of the principle that there is no such thing as optional operations and that even optional 

QR must lead to new scope relations is commonly interpreted more generally, i.e. movement must have an 

effect on the output (see also Miyagawa 2006a,b). 

16

IS is conceived of as being tightly linked to LF or it may even be the case that it is a 

subcomponent thereof. As mentioned above, Topic- and Focus-related features are part of 

the Numeration and may enter the derivation associated with an EPP to assign a 

particular value (interpretation) to the (attracted) DP. They are non-interpretable on the 

goal and must get deleted at Spell-Out. In this case, the Numeration of a lexical item 

includes phonological, formal, and semantic and IS-features. 

Since certain types of Foci and Topics are associated with particular intonations, or are 

defined in terms of prosodic prominence, IS is also tuned in with PF, possibly in the ways 

described in Cinque (1993) and sketched above. The architecture of the grammar would 

then look like in (1) below. 

(1) Numeration:{Case, phi-features,(Topic), T, 

(Focus), etc.} 

Syntax/Phase 

LF(+Information Structure) Word order PF (prosody) 

Syntax plays with two sets of features, uninterpretable Case and phi-features, on the one 

hand, and uninterpretable IS-features, on the other. The interplay of such features gives 

rise to distinct word orders that the Mittelfeld superficially displays. The orders exhibited 

at the left periphery are not considered here. Once Match and Agree, and/or Move have 

taken place, syntax ‘communicates’ with the LF (+ IS) and PF interfaces. As shown in 

the table, at the LF (+ IS) interface, deletion applies to uninterpretable features, whereas 

PF is responsible, among other things, for the prosodic prominence (in the form of pitch) 

usually associated with Focus and certain types of Topics (for distinct approaches to the 

Syntax-Information Structure interface see Vallduví 1993, Choi 1999). 

With these ideas in mind, I turn to the notions related to Information Structure as 

they will be used throughout this chapter. I suggest that some IS-features are: New 

Information Focus, Contrastive Focus, (Thematic) Topic, Contrastive Topic, Topic- 

17

aboutness (but see Büring 2003 for exclusion of Topic as a category in IS). A few 

remarks concerning the meanings they are used in are provided below. However, these 

are the notions which are relevant for the discussions to follow, and in no way are they 

the only notions relevant to IS. 

Let us begin by briefly considering the notion of focus. Since Chomsky (1972) and 

Jackendoff (1972) introduced it into generative theory, the concept of Focus has received 

an enormous amont of attention (Szabolsci 1981, Rooth 1985, Rochemont 1986, Vallduví 

1992, Lambrecht 1994, Kiss 1998, Molnár and Winkler (eds.) 2006, among many others). 

The notion of Focus has received different interpretations and also different labels: 

Constrastive Focus vs. Presentational Focus, Narrow Focus vs. Wide Focus, 

Identificational Focus vs. Information Focus. The general term ‘Focus’ and its alleged 

correlate, contrast, are often regarded as very closely related concepts of IS (though I do 

not adhere to the view that contrast is a separate IS-feature). Thus, Lambrecht (1994) 

links focus to contrast(iveness) (but see Kiss 1998). To him, focus is intrinsically 

contrastive, where the notion of contrast is used as a synonym for opposition. For 

Lambrecht, contrast arises from particular inferences made on the basis of conversational 

contexts. As opposed to Information Focus, Contrastive Focus is usually associated with 

a limited number of contextually given alternatives (see Molnár 2006 and references 

therein for a view that at least in some languages explicit mentioning of alternatives is 

required). Importantly, the idea of set of alternatives narrows down to that of closed set of 

alternatives. 

Based on data from Hungarian and English, Kiss (1998) distinguishes between 

Identificational Focus and Information Focus. The former expresses “exhaustive 

identification of the subset of the set of contextually or situationally given elements for 

which the predicate phrase can potentially hold” (Kiss 1998:245). Syntactically, this type 

of Focus is an operator which moves to a scope position in the left periphery of the 

sentence and binds a variable. The latter expresses new, non-presupposed information, 

bears a pitch and does not imply a set of alternatives. Therefore, it does not express 

exhaustive identification. As opposed to the Identificational Focus, Information Focus is 

not associated with movement. Consider the examples below taken from Kiss (my fonts). 

18

(2) a. Mari EGY KALAPOT nézett ki magának. 

Mary a hat.Acc picked out herself.ACC 

‘It was a hat that Mary picked for herself.’ 

b. Mari ki nézett magának egy kalapot. 

‘Mary picked for herself a hat.’ Kiss 1998 (8) 

It is argued in Kiss that egy kalapot ‘a hat’ in (2a) involves a subset of the set of some 

given elements and only one element is selected and predicated of, being in this sense 

[+exhaustive] in a well-defined contextual event of shopping. It undergoes 

identificational focus movement to some high functional projection (FP in her terms) in 

the left periphery. The same constituent in (2b) fills a typical in-situ postverbal 

Information Focus position and does not presuppose the existence of any set of 

alternatives. However, for Kiss, the Hungarian Identificational Focus is not necessarily 

[+contrastive]. It is [+contrastive] only if “the contrasting complementary subset” is 

involved and “operates on a closed set of entities whose members are known to the 

participants of the discourse”, and [-contrastive] if it operates on an open set of elements, 

i.e. no clearly identifiable complementary members are given. The relevant example is 

the following. 

(3) A: Ki írta a Háború és békét? 

Who wrote the War and Peace 

‘Who wrote War and Peace?’ 

B: [TopP A Háború és békét [FP Tolsztoj írta]] 

the War and Peace.Acc Tolstoy wrote 

‘It was Tolstoy who wrote War and Peace.’ Kiss 1998 (67) 

Kiss argues that (3B) is a felicitous answer to the question in (3A) and that Tolsztoj is the 

Identificational Focus filling the specifier of the high FP. Crucially, since exhaustivity is 

associated with contrast, and since no closed set of writers to be opposed to is explicitly 

made reference to in the question in A, the Identificational Focus in B [-exhaustive] for 

Kiss. However, as alluded to in the discussion above, one important ingredient to qualify 

19

an element as IF (or CF) is the availability of an explicit set of alternatives or even of an 

implicit set of alternatives as long as the latter is easily retrievable. It may be argued that 

the IF in B does not really entail an open set of writers but rather an implicit closed set 

minimally characterized as ‘writers capable of writing War and Peace’ or else ‘Russian 

writers capable of writing War and Peace’. In this sense, it is still [+contrastive] since it 

implies ‘Tolstoy but not Thackeray’ or ‘Tolstoy but not Dostoyevsky’. Put differently, 

the element War and Peace already introduces a restriction of meaning in the computing 

of alternatives. 

The notions of Foci as used throughout the chapters is that of Information Focus, 

corresponding to Kiss’, on the one hand, and of Contrastive Focus, roughly the 

equivalent of her Identificational Focus, the latter being exhaustive and referring to a 

closed class of alternatives, which may be explicitly mentioned in the previous context or 

easily retrievable in that particular context. 

Another IS-feature to deal with is Topic. It is regarded as the complementary notion 

to focus. The most widely-spread view of topic is that it contains old or given information 

about the utterance. Givenness constitutes a necessary condition for topichood. The view 

of topicality based on what the sentence is about is rejected by Chafe (1976) who defines 

real topics, characterizing topic prominent languages, as frames, meant to set a spatial or 

individual framework within which the main predication holds. Such topics may be called 

frame topics. Since the Prague School it has been common knowledge to view the notion 

of Topic in the sentence bipartition Theme-Rheme, Topic-Comment or Focus 

Background/Presupposition (Halliday 1967, Gundel 1974, Kuno 1980, Prince 1981, 

Reinhart 1982, a.o.). However, Vallduví (1992) like Lambrecht (1994,) takes IS or 

Information Packaging to be an independent component of the grammar which interacts 

with other components. In his system the sentence is partitioned into Focus and Ground 

but crucially the latter is further divided into Link and Tail. The Link elements are leftdetached 

(in the sense of left dislocation) and the Tail is represented by right-detached 

(right dislocated) elements. In other words, Link is the Topic proper, what the sentence is 

about, whereas the Tail is other given information or the rest of the ground information. 

The acceptation of the term Topic is (at least) two-folded. There is a distinction between 

a Topic associated with a Comment and a Topic expressing aboutness. The former is 

20

normally expressed in the left periphery of the clause as Clitic Left Dislocation (Cinque 

1990, Rizzi 1997). An example from French is given below. 

(4) [TopP Jean, [TP il est déjà rentré hier]]. 

Jean, he has already come.back yesterday 

‘Jean already came back yesterday.’ 

The latter type of Topic is the subject of the sentence and satisfies the Extended 

Projection Principle which requires that every clause have a subject (Chomsky 1981). It 

refers to a constituent having the capacity to fill the structural subject position, i.e. Spec 

TP or Spec SubjP (see also Rizzi 2005, 2006, 2007, Rizzi and Shlonsky 2007). The rest 

of the predication is about the element in Spec TP, Jean, in (5). 

(5) [TP Jean est déjà rentré hier]. 

‘Jean already came back yesterday.’ 

Another Topic-related notion pertaining to IS is that of Contrastive Topic (henceforth 

CT). CT is also given, presupposed, or anchored in the speech-situation to a certain 

degree like a non-contrastive Topic. Essentially, the given part of the accommodated part 

must be present in the discourse. In this sense a Contrastive Topic is topical, i.e. it bears 

the feature of topicality. Therefore, the head of CT is topical. Thus, in the example (6) 

below, oft-cited in the literature, HE in the second question is not a pure Focus. 

(6) A: Well, what about FRED? What did HE eat? 

B: FREDB B ate BEANSA. 

Jackendoff 1972 

HE in the second question is a CT as it is one of the people in the context and was 

mentioned or accommodated in the previous question, thus being in the background as 

given. CT displays A-/B-accents in English (Bolinger 1965, Jackendoff 1972, Büring 

2003). Similarly, FRED in (6), is a CT as it is part of a partial answer. 

21

(7) A: What did the students eat? 

B: FRED ate BEANS. 

FRED is part of the relevant set of students given in question A. Crucially, (B) is not 

suitable as a complete answer to the question Who ate what? Incompleteness is 

considered a characteristic property of CT (Lee 1999, Tomioka 2007). A CT typically 

selects only one element out of a set of alternatives. Moreover, since only one element, 

out of the relevant set of students, has been singled out in the answer, a CT is anti- 

exhaustive. The choice of one element from alternatives is in itself focal. It differs from 

Contrastive Focus in that in the case of the latter one other element or alternative is 

explicitly (or implicitly, esp. in the cases with only/even) opposed to, hence its other 

denomination at times as Corrective Focus. Put differently, contrastiveness in the case of 

Focus is associated with exhaustivity (‘A but not B’). By evoking a scalar implicature 

which is conventional (‘but I don’t know about the others’ or ‘but I don’t have X’), a CT 

is anti-exhaustive. Semantically, a CT is taken to generate a set of scalar alternatives, 

ordered in terms of their semantic strength, i.e. among the set of scalar alternatives to a 

sentence α, there must be at least one proposition that is stronger than [[α]] (see discussion 

in Tomioka 2007). The conventional implicature is that speaker of [[α]] believes that the 

stronger propositions are possibly false. Consider (8) below. 

(8) A: Did the students pass the exam? 

B: Mary did. 

In this case, A assumes that B has complete knowledge of the result of the exam. The 

implicature in (8) guarantees that B thinks it possible that students other than Mary, say 

Bill and Jill, have failed. A would thus draw a conclusion which is much stronger than 

that. (The assumption that B knows that result of the exam, and given the Gricean 

principle according to which B should be as informative as possible, leads to the 

conclusion that Bill and Jill failed the exam.) 

Whereas the CT is marked with the B accent in English, in languages like Japanese 

and Korean, the –wa marker in Japanese and the –nun marker in Korean, which are the 

22

typical markers of a Thematic Topic, get a high pitch (Nakanishi 2001, Tomioka 2007, 

Chungmin Lee 1999, 2006). 

(9) Ken-WA ukat-ta. Japanese 

Ken-CT pass-Past 

‘Ken passed.’ Tomioka 2007 (5) 

(10) Yengswu-NUN w-ass-e. Korean 

Yengswu-CT come-Past-Dec 

‘Yengswu came.’ Chungmin Lee 2006 (14) 

The possibility has been considered in the literature that constrast is itself an autonomous 

phenomenon of IS which cuts across and supersedes topic and focus (see Kiss 1995 for 

Finnish). Molnár (2006) goes even further and suggests the presence of a Konstrastive 

projection immediately below ForceP within Rizzi’s (1997) split CP model. Such a 

projection is suggested to host a feature which ensures the left-most position of an 

element coupled with the explicit mentioning of relevant excluded alternatives within a 

closed set. In the case of CT, such a feature ensures givenness or high degree of 

recoverability of the respective constituent. However, such a projection as hosting an 

operator and non-operator element at the same time can be done away with. Besides, as 

argued in Rizzi (1997), there is only one Focus per clause, whereas a Topic may be 

recursive. 

2. Argument- and Verb-related Projections among Adverb Functional 

Projections 

The approach adopted here is in line with Kayne’s (1994) conception of phrase structure 

in which specifiers are unique left-branching adjoined phrases, multiple adjunction being 

banned. The only possible (universal) order is Spec-X-Complement. As already 

mentioned, I also follow Kayne’s (1994) antisymmetry hypothesis and the cartographic 

approach to the clause architecture, such as has been put forth by Belletti (2004), Cinque 

23

(1999, 2002, 2006) and (Rizzi 1997, 2004a), where the CP (the Vorfeld) and IP (the 

Mittelfeld) are areas rich in functional projections, roughly represented in (11). 

(11) [ForceP [TopP* [FocP [TopP* [FinP … [MoodP [ModeP [TP [AspP [VoiceP …]]]]]]]]]] 

Adverbs are adjoined specifiers attached to the left, linearly preceding the head with 

which they are associated. Adverbs must occur in a Spec-head configuration with the 

appropriate head, the features they check are substantive (interpretable) features. If 

adverbs are unique specifiers of semantically related projections, the clause structure 

contains a large number of functional projections, as proposed by Cinque (1999). The 

hierarchy of functional projections is provided in (12) below. 

(12) 

[Frankly Moodspeech act > [surpisingly/fortunately Moodevaluative > [allegedly Moodevidential > [probably 

Modepistemic > [once Tpast > [then Tfuture> [perhaps Mod(ir)realisis > [necessarily Modnecessity > [possibly 

Modpossibility > [willingly Modvolitional > [inevitably Modobligation > [cleverly Modability/permission > [usually 

Asphabitual> [again Asprepetitive> [often Aspfrequentative > [quickly Aspcelerative> [already Tanterior> [no longer 

Aspperfect> [still Aspcontinuative > [always Aspperfect > [just Aspretrospective > [soon Aspproximative > [briefly 

Aspdurative > [characteristically Aspgeneric/progressive > [almost Aspprospective > [completely AspSgCompl etive(I) > 

[tutto AspPlCompl > [well Voice > [fast/early Aspcelerative(II) > [completely AspSgCompletive(II) > [again 

Asprepetitive(II) > [often Aspfrequentative ]]]]]]]]]]]]]]]]]]]]]]]]]]]]] 

Within Kayne/Cinque’s model of phrase and clause structure a number of movement 

operations affecting the arguments and the verb are envisaged with a view to account for 

the right placement of adverbs with respect to the head that licenses them. Given the 

richly articulated structure of the IP field which contains 32 projections dedicated only to 

adverb placing, Cinque suggests that the fact that DP’s and floating quantifiers can be 

found interspersed among them clearly points to the existence of several DP-related 

positions (see Laenzlinger in prep.). In addition to the ‘usual’ subject (at least two for 

Italian and Romanian; see, for instance, Cardinaletti 1996, Cornilescu 1997) and object 

positions, which are mainly responsible for Case and phi-feature checking/valuation, 

Cinque assumes that the split IP-field, or Mittelfeld, contains DP-projections associated 

with distinct interpretive functions. The main aim of this chapter is to identify the 

positions of the arguments in the Romanian Mittelfeld of the sentence taking as a crucial 

24

criterion the rigid placement of the adverbs. As suggested in section 1, such interpretive 

readings, or at least, some of them, will be argued to exist on distinct heads, being closely 

associated with an EPP. 

Besides the Kaynean approach to the phrase structure, I also adopt the Minimalist 

view of feature theory as expounded in Chomsky’s (2000, 2001). Chomsky proposes a 

theory of AGREE elaborating ATTRACT of Chomsky (1995). The uninterpretable features 

of a probe α and a goal β are deleted under the structural relation (13), subject to the 

matching Condition, as defined in (14). 

(13) AGREE (Chomsky 2000) 

α β 

AGREE (α, β) 

(14) Match (Chomsky 2000) 


b. D(P) is the sister of P. 

c. Locality reduces to ‘closest c-command’. 

Once AGREE between α and β is realized, uninterpretable features are deleted. As 

suggested in Chomsky, AGREE exists because it deletes uninterpretable features. The 

deletion operation is a requirement imposed by interface considerations. Feature 

valuation/checking is therefore done via AGREE. In addition to AGREE, the computational 

system also relies on movement which is actually internal Merge, i.e. the linguistic 

representation built up is merged with a subpart of itself and the result is a copy of the 

subtree as sister to the root. In the Minimalist Programme, MOVE is only instantiated by 

an EPP-feature present on the probe. In more concrete terms, for instance, in Chomsky’s 

Derivation by Phase (DbP), the T head is a probe with a complete set of uninterpretable 

phi-features, such as number and person, and it is this phi-bundle that acts as a probe. 3 

3 

In Chomsky (2000, 2001, 2008) it is argued that the phi- and T-features on the T head are derivative, i.e. 

they are determined by C or by the selecting verb and that in the lexicon, T actually lacks these features. 

Though this view of Nominative checking/valuing does not bear on the discussions here, Adger’s (2004) 

25

The goal DP has interpretable phi-features and an uninterpretable structural Nominative 

Case feature, the latter rendering it active. Essentially, structural Case is not a feature of 

the T probe but is assigned a value under AGREE and then gets deleted. As a result of 

Match and AGREE, structural Nominative Case is valued as parasitic on phi-features (for 

some criticism and an alternative account, see Pesetsky and Torrego 2004). If T probes 

and agrees with the goal DP, the latter can remain in situ under AGREE, with all 

interpretable features valued, or can raise to Spec TP, again with all features valued. A 

similar view is expounded in DbP for the Accusative except that it is v that assigns the 

Accusative value to the DP. As mentioned, Case itself is not matched, but deletes under 

matching of phi-features. Since the T head as it is the locus of tense/event structure, also 

has a Tense-feature. The Tense-feature on the T bears a value and v bears an 

uninterpetable tense feature which is unvalued. As a result of Match and AGREE, Tense 

on v gets valued. 4 

Throughout this thesis I adopt the notations used in Laenzlinger (2002): Subj is the 

head of Subj(ect)P which is to be partly understood as the equivalent of the former 

AgrSP. The Subj head has uninterpretable phi- and Nominative Case features as well as 

an EPP-feature, which can be checked either by a lexical DP in Spec SubjP or by an 

expletive (generating the SVO word order). Agreement features (number and person) and 

Nominative case can be checked either in the Spec-head configuration between the DP 

subject and the lexical verb (in the SVO order) or by downward AGREE with a Spec 

SubjP realized lower in the tree (in the VS(O) and V(O)S structure). An expletive fills 

(the highest) Spec SubjP. As suggested, the EPP on the Subj head may be associated with 

an IS-feature and it will be seen that this position is Rizzi’s (2005, 2006) and Rizzi and 

Shlonsky’s (2007) Topic-aboutness characterizing the DP or the expletive in preverbal 

position in the highest Subj projection. 

view is adopted throughout the thesis according to which the T head has a Nominative uninterpretable Case 

feature. Once T and the DP in the specified vP Agree for Case, the latter’s Case feature is valued as [Nom] 

and T°’s EPP feature raises the DP into Spec TP. 

4 

As mentioned, Pesetsky and Torrego propose a system distinct from that of DbP and MI (Minimalist 

Inquiries), based on the separation of interpretability and valuation. Crucially, in their system, the tense 

feature on the T head is unvalued acting therefore as a probe, and it is the T-feature on v that has the value. 

26

Obj is the head of Obj(ect)P which is roughly the equivalent of the former AgrOP. 

The head Obj has a N-feature to check/value and also bears an Accusative feature. It 

enters into an AGREE relationship with the DP object, valuing the Case on the object as 

[Accusative]. In a manner similar to Nominative, either a Spec-head configuration 

obtains, or downward AGREE (in the word order V(S)O). Essentially, as discussed, the 

EPP-feature on both Subj° and Obj° must be understood as being closely associated with 

an IS-feature among the features already defined. Once the SVO order is obtained 

sandwiched in between, say, ModPepistemic and TPpast, a chunk is formed. Within such a 

sysmtem there is a multiplicity of Sub(ject) and Obj(ect) positions. All this is summarized 

in (15) below. 

(15)[ SubjPDP[ObjPDP[ModPepistemicAdv[SubjPDP[ObjPDP[TPpastAdv[SubjPDP[ObjPDP…[ VoiceP Adv[vP… 

Regarding verb movement, Romanian (and Spanish) exhibits the auxiliary-verb 

adjacency and no material, other than clitic adverbs in the case of the former, can disrupt 

this configuration 5 . In order to keep the analysis simple, it may be assumed that the 

auxiliary merges in Subj°, whereas the participial verb moves to the immediately lower 

head on condition that this projection does not project a specifier, much in the spirit of 

Dobrovie-Sorin (1994) (for a more complex analysis which captures long head 

movement, see Laenzlinger in progress). 6 Another solution would be to assume a 

5 

As discussed in Dobrovie-Sorin (1994) and will be seen in section 7, only a restricted class of clitic 

adverbs can disrupt the Aux-V adjacency. 

6 

It is stressed out in Dobrovie-Sorin (1994) and Alboiu (2002) etc. that even the infinitive verb can raise 

quite high in the IP domain. However, considering the test with distinct adverbs in Cinque’s (1999) 

hierarchy, the contrast below shows that on the assumption that the adveb probabil fills its first-merge 

position high in the Mittelfeld, the infinitive verb must move higher than the adverb, plausibly in Fin°. 

Recall that in Romanian (and Spanish) no XP material can break the infinitive/subjunctive marker (or 

auxiliary) – verb adjacency. 

(i) a. A reveni probabil/poate/regulat/adesea în aceste locuri e cel mai indicat pentru ea. 

To come.back probably/perhaps/regularly/often in these places is most indicated for her 

‘To probably/perhaps/regularly/often come back to these places is most indicated for her.’ 

27

dedicated AuxP hosting the auxiliary merged outside vP which has its uInfl(ection) 

feature valued as [present] by a higher Subj head. The value Aux° is actually the default 

[present] in Romanian. As known in the literature on Romanian and as will be see in the 

sections below, Aux° subsequently gets attracted to Subj° and may move even higher. 

Once valued and moved, the auxiliary is spelled out in its present from via the interface 

with morphology. 

It has been suggested for Spanish, Catalan and Romanian that the T head, which 

would correspond to Subj° in the system adopted here, is a syncretic category which can 

host both Topic and Focus/Wh features (Zubizarreta 1998, Barbosa 2001, Hill 2002, 

Alboiu 2002, Cornilescu 2002) and this accounts, for instance, for the strict adjacency 

between a fronted wh-phrase and the verb and the lack of the root-embedded asymmetry 

in this regard. Without discussing these issues here (but see a more thorough discussion 

in chapter III), the cartographic framework which does away with multiple specifiers and 

the view that a head can carry a complex A and A’ feature specification forces a distinct 

treatment of Topic and Focus/Wh features. T° cannot bear features of the Q-type, nor can 

it bear other features of the A’ type. Such features head distinct specific projections in 

syntax which encode different discourse-scope/IS properties. A syncretic view of T 

would have to resort to multiple specifiers and furthermore would need some mechanism 

capable of explaining the particular rigid ordering of Topic, Focus and Wh features. 

In the sections to follow I will mainly concentrate on identifying the possible 

landing sites for argument and verb movement among various adverb classes in the 

Romanian Mittelfeld, while occasionally taking into account other Romance languages, 

like Spanish, Italian and French. English may at times also be considered. 

As known, the clause can be divided into three domains, as indicated in (16) below. 

There is the Mittelfeld which extends from the complementizer position (not included) to 

the VP-shell structure (not included). The Vorfeld is associated with the complementizer 

domain, exhibiting a rich array of functional projections (Rizzi 1997). Finally, the 

Nachfeld refers to the Larsonian VP-shell structure. 

b. * A probabil/poate/regulat/adesea reveni în aceste locuri nu e lucru uşor. 

28

(16) ForceP>TopP>FocP>TopP>FinP>MoodP>NegP>TP>AspP1>AspP2>vP>VP 

VORFELD MITTELFELD NACHFELD 

(Laenzlinger 2002, (11)) 

Whereas the Mittelfeld contains the rigidly ordered first-merge positions of the adverbs, 

and it is the arguments and the verb that may float around these projections, the domain 

of adverb movement is the Vorfeld and some remarks will be made in this respect in 

section 7. The interaction of arguments (and V) with respect to adverbs will be 

investigated in the VSO, SVO and VOS word orders. The Nachfeld will not be dealt 

with. 

With these basic remarks in mind, I now turn to the interactions between arguments 

(and the verb) and distinct adverb classes with the view to detect the exact interspersed or 

floating subject, object and verb positions as well as the possible interpretations 

associated with any one of them. The more general goal is to identify the positions that 

Romance languages make available to subject and object DPs along the hierarchy of 

AdvPs. Before properly embarking upon the analysis, here is the typology of the 

Romanian adverb classes. 

(17) MoodPspeech-act: personal ‘personally’, sincer ‘sincerely’ 

MoodPevaluative: în mod regretabil ‘regrettably’, în mod neaşteptat ‘unexpectedly’, 

din (ne)fericire ‘(un)fortunately’ 

MoodPevidential: evident ‘obviously’, în mod clar ‘clearly’ 

ModPepistemic: probabil ‘probably’, sigur ‘certainly’ 

Tpast: odată/odinioară ‘once’ 

Tfuture: apoi ‘then’ 

MoodPirrealis: poate ‘perhaps’ 

ModPnecessity: în mod necesar ‘necessarily’ 

Asphabitual: de obicei ‘usually’, regulat ‘regularly’ 

AsprepetativeI: iar(ăşi), din nou ‘again’ 

29

AspfrequentativeI: adesea/deseori ‘often’ 7 

Modvolotional: intenţionat ‘intentionally’ 

AspcelerativeI: repede/rapid ‘quickly’ 

Tanterior: deja ‘already’, recent ‘recently’ 

Aspterminative: mai ‘no longer’ 8 

7 Recall that in Cinque some adverbs, such as the frequentative aspect adverb often may be merged in two 

distinct specifier positions, each associated with a specific scope position. The same judgments also hold 

for Romanian. Consider (i) below: 

(i) a. Maria e de obicei adesea/deseori obligată să stea acasă. 

Maria is usually often forced să.Subj stay at home 

‘Maria is usually often forced to stay at home.’ 

b. * Maria e adesea/deseori de obicei obligată să stea acasă. 

As in Italian, though not exemplified here, the higher frequentative adverb adesea/deseori follows the 

habitual adverb de obicei (ia) and precedes the Tanterior adverb deja (iia). The lower adesea/deseori follows 

deja, as in (ib) above, or, as in Italian, it expectedly follows other low adverbs, such as the voice adverb 

atent/bine, as in (iii). 

(ii) a. Când, în sfârşit, crede că a descoperit şi el ceva, se dovedeşte că respectiva descoperire a fost 

făcută adesea deja de altcineva. 

When finally thinks that has descovered he something, se.Refl turns.out that the respective 

descovery was made often already by somebody else 

’When he finally thinks that he has descovered somehting, it turns out that that descovery has 

often already been made by somebody else.’ 

b. Insula aceasta a fost deja descoperită deseori în ultimii ani. 

Island this has been already descovered often these last years 

’This island has already been often descovered these last years.’ 

(iii) A scris atent/bine deseori poezii în tinereţe. 

Has written carefully/well often poems in youth 

’He often well/carefully wrote poems in his youth.’ 

8 

Mai belongs to the restricted class of clitic adverbs that Romanian exhibits (see Dobrovie-Sorin 1994). As 

will be discussed in section 7, it belongs to two different but contiguous classes, it may thus function as an 

30

Aspcontinuative: încă ‘still’, tot ‘still/on and on’ 9 , mai ‘still’ 

Aspperfect: întotdeauna, mereu ‘always’ 

Aspretrospective: tocmai ‘just’, (de-)abia ‘hardly’ 10 

Aspproximative: imediat ‘immediately’, curând ’soon’ 

Aspprospective: aproape ’almost’ 

AspcompletiveI: complet ‘completely’ 

AspPlCompletive: tot ‘everything’ 

VoiceP: bine ‘well’, atent ‘carefully’, calm ‘calmly’ 

AspcelerativeII: repede/rapid ‘quickly’ 

AsprepetativeII: iar(ăşi), din nou ‘again’ 

AspfrequentativeII: adesea/deseori ‘often’ 

AspcompletiveII: complet ‘completely’ 

Naturally, there are many classes of adverbs in which one adverb may have a 

corresponding PP adverbial counterpart. A case in point is the class of Voice adverbs. 

The first constituent in pairs like atent‘carefully’-cu atenţie‘with care’, calm‘calmly’-cu 

calm‘with calm’, uşor‘easily’-cu uşurinţă‘with ease’, greu‘painfully’-cu greu‘with 

difficulty’ is ambiguous between an adjectival and an adverb reading. However, the class 

of MoodPevaluative adverbs contains only adverbial PPs whose translation is most often 

than not ‘in an X way’. One possible source for the adjective-adverb ambiguity may be 

Aspterminative adverb in the presence of negation, as in (i), or as an Aspcontinuativee, as in (ii). It will be seen that 

it may also be an iterative aspect adverb. 

(i) Nu a mai scris nimic. 

Not has no.longer written nothing. 

‘He hasn’t written anything any longer.’ 

(ii) La vârsta ei, îşi mai cumpăra păpuşi. 

At age her, îşi.Refl still buy dolls 

’At her age she was still buying herself dolls.’ 

9 Besides mai, tot in one meaning is also an adverb clitic in Romanian. 

10 These two retrospective aspect adverbs are restricted, i.e. they can only occur preverbally. 

31

the absence of a specific suffix marking this category, as the suffix -ly for English, -ment 

for French, -mente for Italian and Spanish. I do not discuss the distribution of PP 

adverbials. 

3. Word Orders in Romanian. Romanian VSO and Adverb Placement 

Romanian is an VSO language with relatively free word order, the following orders being 

all possible (18). 

(18) a. Scrie Ion un articol. VSO 

Writes Ion an article 

‘Ion is writing an article.’ 

b. Ion scrie un articol. SVO 

c. Scrie un articol Ion. VOS 

d. Un articol scrie Ion. OVS 

e. Ion, un articol scrie. SOV 

f. Un articol, Ion scrie adesea. OSV 

An article, Ion writes often 

As very general remarks, it may be said that (18a) is the underlying word order in 

Romanian and can be uttered with neutral intonation. (18b) is slightly marked. (18c, d) 

appear rather focalised. The subject gets (narrow or contrastive) focus stress in (18c) and 

the object is either topicalised or focalised in (18d), given the appropriate intonation. In 

(18e), the object bears contrastive Focus, while the subject DP is left dislocated, and an 

intonational break separates it from the object. In (18f), the indefinite DP object is 

topicalized at the left periphery. No clitic doubles the object as it is an indefinite. As in 

the previous case, an intonational break separates the object from the subject. Below, I 

will only focus on the first three orders and only occasionally will remarks be made with 

respect to the other orders. 

32

While it is standard knowledge that the availability of postverbal subjects correlates 

with the pro-drop parameter, there is micro-variation among Romance languages like 

Romanian, Spanish and Italian once the object is present (see, e.g. Roussou and Tsimpli 

2006). Spanish is generally considered an SVO language which nonetheless allows 

subjects to appear postverbally, either before or after objects (Ordóñez 1998, Zubizarreta 

1998, a.o.) 11 By taking into account the hierarchy of adverb-related functional 

projections, a parametrisation of DP- (and V-) movement possibilities is arrived at in 

Romanian, given the orders to consider, and even across some Romance languages. 

I will first concentrate on Romanian. Postverbal subjects were originally analyzed 

as being in Spec VP, the Nominative case position, under government by the verb raised 

to inflection (Motapanyane 1989, Dobrovie-Sorin 1994). For Motapanyane (1995) 

Nominative Case checking of postverbal subjects takes place at LF, i.e. subjects raise to 

Spec TP covertly. In Cornilescu (1997), the postverbal subject can either raise at LF to a 

subject case position (i.e. Spec Agr) or can acquire case in Spec VP by coindexation with 

pro in Spec AgrP. 12,13 Taking the EPP-feature to be a strong V-feature on the I head, 

11 

Zubizarreta (1998) argues that the availability of VSO in Spanish is related to the obligatory presence of 

a preverbal element, be it Topic or Focus. However, Zagona (2002) argues that the acceptability of this 

order varies with speakers. 

12 

Cornilescu (1997) argues for the existence of two argumental postverbal positions in Romanian, i.e. Spec 

AgrSP and Spec VP. She provides morphological and syntactic evidence that that AgrSP, which is 

responsible for Nominative case is below MoodP, the highest inflectional projection in Romanian hosting 

the moved lexical verb. Evidence comes from a comparison of the behaviour of clitic subjects in some 

Northern Italian dialects and clitic doubled subjects in Romanian. Consider the distribution of el in the 

examples below: 

(i) a. Face el ceva Ion. 

Does he something Ion 

V Cl O S 

’Ion must be doing something.’ 

b. Ion face el ceva. S V Cl O 

c. * El face ceva Ion. *Cl V O S 

The constrast indicates that the weak subject pronoun el must only occur postverbally (see however Alboiu, 

30n. 68). See also Hill (2006) for an argument that these are strong subject pronouns (and that Romanian 

exhibits subject pronoun inversion in special contexts). 

33

Alboiu (2002) argues that the postverbal structural Nominative Case is checked as a 

result of AGREE, i.e. the uninterpretable phi-features of I° and the uninterpretable Case 

feature of the subject match and agree, and then get deleted. 

Reconciling Minimalism and cartography, the analysis to suggest here is that the 

Nominative Case and phi-features are checked/valued via AGREE in Romanian. 

Movement of the subject DP to the preverbal position is triggered by Rizzi’s (2005, 2006) 

Topic-aboutness feature on what he labels Subj(ect)P, the highest position in the 

Mittelfeld. Essentially, as argued for in Rizzi (op. cit.) and Rizzi and Shlonsky (2007), 

this feature which attracts a nominal element to its Spec is a criterial position. In Rizzi 

(2006), pro is also used to formally satisfy the Subject Criterion. The definition of a 

Criterion is given in section 4. It suffices to stress out at this point that once an element 

fills a Criterial position, it cannot undergo (further) movement. It will be seen below that 

the Romanian Mittelfeld makes available further positions hosting the DP subject which 

are related to IS features. 

Regarding verb movement, there is a consensus in the Romanian literature that the 

verb raises to the highest inflectional head inside the IP area, be it a T/Agr head, as in 

(Dobrovie-Sorin 1994), or the Mood head, as in Cornilescu (1997), irrespective of the 

whether it is finite or non-finite, or whether there is an auxiliary present. 14,15 A somewhat 

similar analysis is that of Motapanyane (1995) who argues that a distinction must be 

made between structures with an auxiliary and structures with simple tenses. In the case 

of the former, the auxiliary is merged in T° and undergoes short movement to AgrS° and 

the verb raises to T°. In her analysis, the Aux-V adjacency requirement is accounted for 

in terms of the impossibility of subjects to surface in Spec TP, the dedicated preverbal 

subject position being Spec AgrSP. Adopting an approach where the build-up of the 

Romanian IP depends on the type of clause that is involved, Alboiu (2002) argues that the 

Romanian IP domain does not allow for intermediary specifier positions and 

13 

Spanish is also argued to have two different postverbal positions for the subject outside vP (Ordóñez 

2000 and Zubizarreta 1994). 

14 

In Dobrovie-Sorin (1994:21), a Move I to C rule is applied to auxiliaries: the auxiliary moves to Comp 

and subcategorizes for a projection headed by Infl. 

15 

See also footnote 6. 

34

consequently noun phrase movement never targets positions within IP. The postverbal 

subject and object remain in her system in the vP domain. 

VS(O) structures represent the unmarked option in Romanian, being uttered in out- 

of-the-blue contexts. Thus the most natural answer to an out-of-the-blue question like 

‘What happened?’ is an answer taking on the form of the word order VS(O), as illustrated 

below. In this respect, Romanian is partially like Italian (VSO is not accepted in Italian 

unless under special circumstances). 16 

(19) A: Ce s-a întâmplat? 

‘What happened?’ 

B: A căzut copilul(/un copil de pe scară). VS 

Has fallen baby.the(/a baby from onto the ladder) 

’The baby (A baby) has fallen (from onto the ladder).’ 

16 Inosfar as Spanish is concerned, (20) has the counterpart in (i). As mentioned, the VS order is highly 

marginal. 

(i) A: ¿Qué pasaba? 

What happened? 

‘What was going on?’ 

B: Que el niño estaba llorando. 

That the child was crying. 

B’: ?? Que estaba llorando el niño. 

However, if the subject DP is heavier, hence, more specific, both the SV and the VS orders are qually 

acceptable as an answer to A (Vidal Valmala, p.c.) 

(ii) A: ¿Qué pasaba? 

‘What was going on?’ 

B: Que el hijo de Pedro estaba llorando. 

That the child of Pedro was crying. 

‘That Pedro’s son was crying’. 

B’: Que estaba llorando el hijo de Luis. 

35

(20) A: Ce-i zgomotul ăsta? Ce se întâmplă? 

‘What’s this noise? What is going on?’ 

B: Plânge copilul (/un copil în parc). VS 

Cry baby.the (a baby in park) 

‘The baby ( A baby) is crying (in the park).’ 

(21) A: Ai auzit vestea? 

‘Did you hear the news?’ 

B: Se retrag trupele din Irak. VSAdvPP 

Se.Refl withdraw troops.the from Irak 

‘The troops withdraw from Irak.’ 

(22) A: Ce s-a întâmplat? 


B: Şi-a cumpărat Ion maşină nouă (ieri). VSO (Adv) 

Şi.Refl-has bought Ion car new (yesterday) 

‘Ion bought himself a new car (yesterday).’ 

It is to note, however, that an answer in the SVO form is possible but in this case the 

subject is to some extent known or implied in the universe of discussion/discourse and is 

chosen to be asserted or talked about. Consequently it gets prominent (see section 4). 

Whereas Romanian allows Free Inversion, illustrated in (28), in Italian there are some 

constraints on subject inversion, for instance, a lexical direct object cannot occur after the 

subject in the VSO word order unless it is a PP, as shown in (23). 

(23) a. A transcris Maria un curs/cursul. 

Has copied Maria a course/course.the 

‘Maria wrote a course/the course.’ 

b. Au cântat copiii cântece de Crăciun. 

Have.3pl sung children.the songs of Christmas 

‘The chidren sant carols.’ 

36

c. Va explica Maria ecuaţia. 

Will.3sg explain Maria equation.the 

‘Maria will explain the equation.’ 

(24) a. * Capirà/spiegherà Maria completamente (al direttore). 

Will.understand/explain Maria completely (to the director) Belletti 2004 (4a) 

b. (?) Ha telefonato Maria al giornale. 

Has phoned Maria to the newspaper Belletti 2004 (20a) 

As Belletti (2004) argues, there is a minimality violation between the DP subject and the 

DP object regarding the latter’s case assignment in Italian. If the object is a PP, case is 

assigned differently, i.e. by the preposition. Given that Romanian (and Spanish) are cliticdoubling 

languages, following proposals on clitic doubling (Uriagereka 1995, Torrego 

1995, 1998, a.o.), the DP is more than a mere DP, i.e. it is a PP with a dummy preposition 

and an overt clitic if the object is doubled, or an empty preposition and a null clitic if it is 

not doubled. Hence the lack of RM effects in the VSO configuration (as well as the VOS 

one to be dealt with in section 5). 

It is generally considered that Romance languages like Italian express focus in the 

rightmost position in the clause, as attested in many studies (e.g. Calabrese 1992, Belletti 

and Shlonsky 1995, Zubizarreta 1994, 1998, Frascarelli 2000, Belletti 2001, 2004). Based 

on evidence from the interaction of the postverbal subject with low adverbs, the 

impossibility of extraction from a postverbal subject and interpretation facts, Belletti 

(2004) in particular argues that Italian has a low New Information Focus at the left 

periphery of VP. Though allowing free inversion, languages like Italian, Spanish and 

Romanian, to which Greek may be added, differ with respect to the VSO order. On the 

one hand, Italian differs from Spanish, Romanian and Greek in that it does not readily 

permit VSO. Greek differs Spanish (at least the variety described in Zubizarreta 1998) in 

that VSO is more productive in the former, on the other hand. 17 

17 

Roussou and Tsimpli (2006) explain the differences between Italian and Greek in terms of the distinct 

lexicalisation options of the D-system in the two languages. In particular, whereas Greek distinguishes 

between a nominative (‘subject’) and accusative (‘object’) determiner, Italian does not, hence distinct 

37

Let us consider the interaction of the subject with respect to the low singular and 

plural completion elements, AspPSgCompletive and AspPPlCompletive, and the Voice adverb 

calm ‘calmly’ in Romanian. 18 

feature lexicalisation in both languages. Assuming the clasual structure divide, C, T, v, and the presence of 

(recursive) Clitic-shells in the clause structure, they argue that subject and object DPs can appear in the 

same domain in Greek (distinct feature spell-out and distinct grammatical function). Conversely, in Italian, 

DPs spell out the same set of features irrespective of their grammatical feature, hence the absence of VSO. 

Following their line of argumentation, Romanian would be expected to pattern with Italian rather than 

Greek given that its enclitic determiners show similar forms in the Nominative and Accusative. 

Nonetheless, it will be seen that Romanian is not like Italian regarding the VSO order. 

18 

In Cinque’s hierarchy of adverbs there are two distinct completamente ‘completely’ adverbs, one filling a 

position higher than tutto and bene, but lower than quasi ‘almost’, and the other, with a partially different 

interpretation, filling the lowest position. Similarly, the Romanian aproape ‘almost’ in (ia) below, filling 

the specifier of the prospective aspectual projection is higher than the completive aspectual adverb complet. 

The reverse order in (ib) is possible only on the reading where the prospective adverb modifies the 

completive one. 

(i) a. Au dărâmat aproape complet clădirile. 

b. 

Have.3pl pulled.down almost completely buildings.the 

’They almost completely pulled down the buildings.’ 

* Au dărâmat complet aproape clădirile. 

As expected, the prospective aspect adverb precedes the plural completive adverb totul ’everything’ and 

the voice adverb bine, as in (iia). It cannot fill a position lower down, as shown by the ungrammaticality of 

(iib). Recall that the reading where aproape modifies tot(ul) is of no interest here. 

(ii) a. A transcris aproape tot(ul) bine. 

b. 

Has copied almost everything well 

’He almost copied everything well.’ 

* A transcris totul bine aproape. 

Importantly, the adverb complet ’completely’ can also occupy a postobject position, in which case it fills 

the lower completive aspect projection. 

(iii) A transcris totul complet. 

Has copied everything completely 

38

(25) a. Paşea calm Maria. 

Was.walking calmly Maria 

‘Maria was calmly walking.’ 

b. Va explica totul Maria. 

Will.3sg explain everything Maria 

‘Maria will explain everything.’ 

c. S-a diluat complet soluţia. 

S.refl-has dilluted completely solution.the 

’The solution got dilluted.’ 

This kind of distributional evidence concerning the location of the subject and 

Aspcompletive adverb, such as complet ‘completely’ and the quantifier totul ‘everything’or a 

Voice adverb, such as calm ‘calmly’, situated rather low in the clause structure according 

to Cinque’s (1999) hierarchy point to the conclusion that the postverbal subject may fill a 

very low position in the clause as it follows these elements. In the spirit of Belletti (2001, 

2004), it may tentatively be concluded that, like Italian, Romanian also has a very low 

New Information FocusP, situated at the left periphery of VP. However, as opposed to 

Italian, where the counterparts of (26) are ungrammatical, the subject in Romanian can 

appear above the adverbs, as in (26a, c), or the quantifier in (26b). 19 

(26) a. Paşea Maria calm. 

Was.walking Maria calmly 

‘Maria was calmly walking.’ 

b. Va explica Maria totul. 

’He copied everything completely.’ 

19 The Italian counterparts of (23a) and (23b) are marked ‘’?’’ by Belletti (2004) and the contrast with the 

perfect counterpart of (23c) is accounted for in terms of movement of the quantifier to a position above the 

subject DP versus the first merge positions of the adverbs. 

39

Will.3sg explain Maria everything 

‘Maria will explain everything.’ 

c. S-a diluat soluţia complet. 

S.refl-has dilluted solution.the completely 

’The solution got dilluted completely.’ 

The grammaticality of both (25) and (26) clearly suggests that the Romanian Mittelfeld 

has a higher subject position (SubjP) associated with a New Information Focus-feature 

that is not available in Italian. 

In order to show that the postverbal focal subject is not the same as the left 

peripheral subject, the following reasoning is advanced by Belletti, which holds for 

Romanian as well. Consider the examples below: 

(27) A: Cine a venit? 

Who has come? 

‘Who came?’ 

B: * ION a venit. 

’It is Ion who came.’ 

B’: A venit Ion. 

Has come Ion 

‘Ion came.’ 

(28) A: Ce ai citit? 

What have.2sg read 

‘What did you read?’ 

B: * UN ROMAN am citit. 

’It is the book that I read.’ 

B’: Am citit un roman. 

Have.1sg read a novel 

‘I read a novel.’ 

40

Since (27B, 28B) do not constitute appropriate answers to genuine questions eliciting 

information, it is therefore legitimate to conclude that the focus position hosting the 

postverbal subject is distinct from the one located in the left periphery. It is to note 

however that one other possible answer to (27) and (28), though not the most natural one, 

is under the SVO form with some slight stress on the subject or the object, as in (29/30B). 

Crucially this kind of stress is not the contrastive focus stress characterizing the (B) 

examples above. Following Göbbel’s (2003) terminology the subject/object associated 

with New Information Focus in (28/29) has a falling tone, whereas the one in (29/30) is 

associated with a high fall. This is evidence that suggests that Romanian has a preverbal 

position that is potentially associated with New Information Focus. 

(29) A: Cine a venit? 

B: Ion a venit. 

(30) A: Ce ai citit? 

B: Un roman am citit. 

The immediate question that arises is whether the object remains inside VP or moves to 

some position in the Mittelfeld given that the subject can move to distinct Spec SubjP 

positions inside the Mittelfeld. In order to answer this question, the test with the low 

completive aspectual adverb and the voice adverb is made. This is illustrated in (31) 

below. 

(31) a. ? A transcris Maria cursul complet. 

Has copied Maria the course completely 

‘Maria copied the course completely.’ 

b. ? A descris Ion incidentul calm. 

Has described Ion incident.the calmly 

‘Ion described the incident calmly.’ 

41

On the assumption that adverbs fill their external merge positions in the Mittelfeld and 

that it is the arguments (and the verb) that move the data above suggest that both the 

object and subject DPs (as well as the auxiliary and the participial verb) have vacated the 

vP and moved to the dedicated Mittelfeld-internal SubjP and ObjP situated immediately 

above the two “low” adverbs. 

Let us consider more in detail the interaction of the DP subject with distinct adverb 

classes in the Romanian VSO order. The test with high adverbs is made in order to detect 

the possible landing sites of subject argument. 

(32) a. Din fericire, citeşte Ion această carte. 

‘Happily Ion reads this book.’ 

b. Citeşte, din fericire, Ion această carte. 

c. ? Citeşte Ion, din fericire, această carte. 

d. Citeşte Ion această carte, din fericire. 

e. Citeşte, din fericire, Ion această carte. 

The MoodPevaluative adverbial has a parenthetical reading in the above sentences. In 

Romanian the large majority of MoodPevaluative adverbs as well as most MoodPspeech-act 

adverbs are adverbial PPs. Parenthetical expressions will not be insisted on. Besides, they 

have not received a uniform treatment in the literature. 20 

Consider next Spanish: high adverbs are not accepted in a sentence-initial position, 

as shown below: 

(33) a. ?* Afortunadamente leyó Juan este libro. 

Fortunately reads Juan this book 

20 One idea that is largely agreed upon is that parenthetical expressions show quite different properties from 

adverbs. Ernst (2002:33), for instance, assumes that parenthetical expressions are licensed by movement 

after Spell-Out at PF. In order to acount for the sentence-final position of parenthetical expressions, he 

postulates downward movement (at PF) which has the effect of forcing the characteristic comma intonation 

of parenthetical expressions. Cardinaletti (1997) discusses the issue of parentheticals and suggests two 

possible analyses: they are either adjoined to maximal projections, or fill the specifier position of a 

designated functional projection. 

42

. ?* Francamente dirá María la verdad. 

Frankly says Maria the truth 

High adverbs are not accepted in sentence-initial position. 

Let us further consider the ModPepistenic adverb probabil ‘probably’ in simple and 

compound tensed sentences. As illustrated in (34), the subject (and the verbal complex) 

can precede this adverb. On assumption that the DP subject Ion fills the specifier of a 

subject-related projection, SubjP, above the merge position of the high adverb, the verb 

citeşte ‘reads’ moves higher, plausibly to Fin°. On the rather simplistic view adopted here 

that the auxiliary and the participial verb form a complex head uninterrupted by any 

maximal category, the entire complex head also moves to Fin°. 

(34) Citeşte/A citit Ion probabil această carte. Reads/Has read Ion probably this book 

‘Ion (has) probably read this book.’ 

It is expected that the subject also precedes adverbs lower than the ModPepistenic adverb. 

Consider the sentences in (35) below. 

(35) a. Citeşte/A citit probabil Ion această carte. 

b. A spus Ion odată adevărul. Has said Ion once truth.the 

b’. (?) A spus odată Ion adevărul. 

‘Once Ion told the truth.’ 

c. Spune/A spus Ion din nou povestea. Tells/Has told Ion again story.the 

c’. (?) Spune/A spus din nou Ion povestea. 

‘Ion told the story again.’ 

d. (?) Spune/A spus Ion deja povestea. Tells/Has told Ion deja story.the 

‘Ion has already told the story.’ 

As the grammaticality of the sentences above shows, the DP subject can float to 

dedicated positions below the epistemic adverb. The DP subject moves to positions 

interspersed above the Tpast adverb odată ‘once’, the AspPrepetativeI adverb like din nou 

43

‘again’ and the Tanterior adverb deja ‘already’. However, the subject in the VSO order 

cannot occur below the projection of the Tanterior adverb (36a). This is suggested for 

instance by the ungrammaticality or the strong marginality of the cases in (36) where the 

subject follows the Aspcontinuative in (36b), the Aspperfect adverb in (36c), the Aspproximative 

adverb (36d), the Advprospective adverb in (36e). Expectedly, the DP subject cannot remain 

in a position lower than the “low” adverbs, Voice and the AdvSgCompletive, as shown by the 

ungrammaticality of (36f) and (36g). 

(36) a. ?? Spune/A spus deja Ion povestea. Tells/Has told already Ion story.the 

‘Ion has already told the story.’ 

b. * Repetă/A repetat încă Ion cursul. Repeats/Has repeated still Ion course.the 

‘Ion still repeates/repeated the course.’ 

c. ?* Repetă/A repetat mereu Ion cursul. Repeats/Has repeated always Ion 

course.the 

‘Ion always repeates/repeated the course.’ 

d. ??(?) Va rosti curând Maria un discurs. Will deliver soon Maria a speech 

’Maria will soon deliver a speech.’ 

e. * Citeşte/A citit aproape Maria discursul.Reads/Has read almost Maria 

speech.the 

‘Maria almost read the speech.’ 

f. * Transcrie/A transcris complet Maria cursul.Copies/Has copied completely 

Maria course.the 

‘Maria completely copied the course.’ 

g. * Recită/A recitat bine Maria poemul.Recites/Has recited well Maria poem.the 

‘Maria recited the poem well.’ 

So far it has been established that in the VSO order, the DP subject must move to the 

middle space of the Mittelfeld, more exactly above the Tanterior projection. 

Let us focus next on Spanish: 

(37) a. * Destruyeron las tropas probablemente los pueblos. 

44

Destroyed the troops probably the villages 

‘The troops probably destroyed the villages.’ 

b. ?? Dijo María hace mucho la verdad. 

Said Maria once the truth 

‘Maria once said the truth.’ 

c. ?? Copió María completamente la conferencia. 

Copied Maria completely the lecture 

‘Maria copied the lecture completely.’ 

These sentences show that as opposed to Romanian, in the Spanish VSO configuration 

the subject DP can only very marginally raise past the ‘low’ Aspcompletive adverb in (37c) 

and the high T(Past) adverb in (37b) and cannot raise past the even higher Modepistemic 

adverb in (37a). 

There is evidence that the verb/auxiliary too can occupy several different positions 

in the Romanian Mittelfeld. Consider in this respect the examples below. 

(38) a. Probabil citeşte/a citit Ion cursul. Probably reads/has read Ion course.the 

’Ion (has) probably reads/read the course.’ 

b. Apoi va citi/ a citit Ion cursul. Then will read/hs read course.the 

’Then Ion (has) read/reads the course.’ 

c. Poate citeşte/a citit Ion cursul. Maybe reads/has read course.the 

’Maybe Ion reads/read course.the’ 

c. De obice reciteşte Ion cursurile seara târziu. Usually re-reads Ion courses.the 

late in the evening 

’Ion usually rereads his courses late in the evening.’ 

e. ADESEA citeşte/a citit Ion cursul. OFTEN reads/has read Ion course.the 

’It is often that Ion reads/has read the course.’ 

f. BINE redă/a redat Ion cursul. AGAIN reproduces/hasreproducedIon course.the 

’It is well that Ion reproduces/has reproduced the course.’ 

g. UŞOR rezolvă/a rezolvat Ion exerciţiul.EASILY solves/has solved Ion exercice.the 

45 

’Ion solved the exercice easily.’

The cases above show that “high” adverbs can precede the lexical verb/auxiliary. 

Adverbs belonging to the “middle” space can also precede it. However, the “lower” pre- 

VP adverbs, such as Aspfrequentative(II) and Voice adverbs can also precede the 

verb/auxiliary but they bear heavy stress, being focalized. As will be argued for in section 

6, these adverbs move to Spec Foc in the left periphery. Focalisation works well with 

fairly short adverbs, such as bine ‘well’, calm ‘calmly’, while with long Voice adverbs, 

such as inteligent ‘intelligently’, the result is slightly marginal (modulo the high subjectoriented 

interpretation of inteligent). 

Let us next focus on the Spanish cases below: 

(39) a. ?? Copió completamente María la conferencia. 

Copied completely María the lecture 

‘María copied the lecture completely.’ 

b. ?? Dijo hace mucho María la verdad. 

Said once Maria the truth 

‘Maria once said the truth.’ 

c. * Ha leido probablemente Juan este libro. 

Has read probably Juan this book 

‘Juan probably read this book.’ 

In Spanish, the verb/auxiliary can only very marginally move above Aspcompletive, as in 

(39a) and the high T(Past) adverb, as in (39b) and cannot move past the Modepistemic 

adverb, as indicated by the ungrammaticality of (39a). However, if ‘low’ adverbs are 

focalised or topicalised and the verb follows them the sentences become perfect. Thus, in 

(40) below, completamente and fácilmente move to Spec Focus in the left periphery. 

(40) a. COMPLETAMENTE copió María la conferencia. ‘Maria COMPLETELY copied 

the lecture.’ 

b. FÁCILMENTE resolvió Juan el ejercicio. ‘Juan EASILY solved the exercise.’ 

46

In the VSO order, the object DP can appear to the left of Aspproximative adverb curând 

‘soon’, as illustrated in (41) below. 

(41) Va citi Ion discursul curând. Will read Ion speech.the soon 

‘Ion will read his speech soon.’ 

Expectedly, the object can move to dedicated positions higher than this adverb class. 

Consider (42). 

(42) a. (?) Şi-a pierdut Ion capul mereu după ea. Să.refl-has lost Ion head always after 

her 

‘Ion has lost his mind after her.’ 

b. A citit Ion un roman recent. Has read Ion a novel recently 

‘Ion has recently read a novel.’ 

c. Citea Ion un roman de obicei. Read Ion a novel usually 

‘Ion usually read a novel.’ 

(42) shows that the object DO can move to a specified ObjP position above the Aspperfect 

adverb mereu ‘always’, or above the Tanterior adverb recent ‘recently’, or even above the 

Asphabitual adverb de obicei ‘usually’. However, with higher adverbs, either parenthetical 

or focalization readings obtain. In (43a), a parenthetical reading obtains, whereas in 

(43b), the sentence-final position of the Tpast adverb somewhat forces a focusing effect on 

the object. 

(43) a. Povesteşte Ion accidentul, probabil. Tell Ion accident.the probably 

‘Ion is probably describing the accident.’ 

b. Citea copilul romane odată. Read child.the novels once 

‘The child would once read novels.’ 

The object DP can expectedly move higher than a Voice adverb. This is illustrated in 

(44). However, such very low adverbs bear focus stress. As pointed out in Cinque, in the 

47

VS order in Italian the Voice adverb bene ‘well’ can follow the complements of the 

participle if focalised, as in (45). 

(44) A rezolvat Ion exerciţiile bine/calm. Has solved Ion exercises.the well/calmly 

‘Ion solved the exercises well/calmly.’ 

(45) Hanno detto tutto alla maestra BENE. Have said everything to the teacher well 

‘They have said everything to the teacher well.’ 

Cinque 1999, (26c) 

So in the VSO order, the object DP in Romanian can raise past the Voice adverb targeting 

distinct positions till the T(Past) projection. 

Let us consider the Spanish examples below: 

(46) a. ? Destruyeron las tropas el pueblo completamente. 

Destroyed the troops the village completely 

‘The troops destroyed the village completely.’ Valmala (p.c.) 

b. ? Describió Juan el incidente con calma/tranquilamente 

Described Juan the incident with calm/calmly 

‘Juan described the incident calmly.’ Valmala (p.c.) 

c. Leía Juan una novela a menudo. 

Read Juan a novel often 

‘Juan often read a novel.’ Valmala (p.c.) 

d. Leía Juan novelas hace tiempo. 

Read Juan novels once 

‘Juan once read novels.’ Valmala (p.c.) 

e. Ha leído Juan una novela recientemente. 

Has read Juan a novel recently 

‘Juan has recently read a novel.’ Valmala (p.c.) 

As in Romanian (and Italian), the object DP in Spanish moves above low adverbs, as in 

(46a-b) though with more marginal results, and can move higher up in the Mittelfeld, i.e. 

48

it targets distinct specifiers of ObjPs which are available above Aspfrequentative a menudo 

‘often’ and, even higher, above T(Past) adverbs, hace tiempo ‘once’ and recientemente 

‘recently’. As pointed out above, in the VSO order, the adverb in sentence final position 

bears focus stress 

Following the data presented thus far, the distributional properties of the DP subject 

and object as well as of the verb in the VSO order in Romanian can be summarised as 

follows: 

(i) The postverbal subject DP must move to a Spec SubjP position higher than the 

T(Anterior) adverb projection and can optionally move higher to a position above 

habitual, repetitive aspect and even epistemic adverbs; 

(ii) The verb, or the auxiliary (together with its participial verb), must move to a 

position higher than the T(anterior) adverb projection. From this position it can float to 

distinct head positions and can move as high as Fin° (see example 35a); 

(iii) The object DP can move to Spec ObjP past the Voice adverb targeting distinct Spec 

ObjP positions till the T(Past) projection but cannot move any higher; 

(iv) It has been observed that in both Romanian and Spanish, low adverbs, such as 

Voice adverbs and even the frequentative II adverbs in pre-verbal position move to Spec 

Focus in the left periphery. 

All this is represented below (only the non-focalised and non-parenthetical readings are 

put down): 

49

Fin Moodspeech/eval/evid Subj V Obj Modepist Subj V Obj TPast Subj V Obj TAnt Subj V Obj Voice Subj V Obj 

R V √ √ * √ √ √ √ * √ √ 

S * * * * ?? ?? √ ?? ?? √ ?? ?? ? √ 

Table 1 

50

The question arises why the subject DP must vacate the vP in Romanian (as well as in 

Italian and Spanish). The multiplicity of positions available in the Mittelfeld for both the 

subject and object DP (as well as for the verb) are also in need of an explanation. On 

Chomsky’s (2001) view there is only one position available to accommodate movement, 

‘outer’ Spec of vP. However, for the present purposes such a position is not sufficient, it 

cannot host both a DP subject and object unless multiple ‘outer’ specifiers are assumed 

which runs counter the view adopted here. If we follow Cinque (1999), we arrive at the 

fine structure of the Mittelfeld in (47) above. As suggested, subject and object movement 

possibilities are determined by IS-valuing, i.e. Top-feature on the subject DP and 

FocusNew Information on the object DP. The Romanian subject DP must vacate the vP 

because valuation of its IS-feature(s) cannot take place from inside vP. It has been further 

suggested that an IS-feature is associated with EPP which accounts for the vP-vacating 

movements. If such an assumption is on the right track, one expects very subtle IS-related 

readings associated with a given DP moved to a position above or below a particular 

adverb class. I leave this issue open. 

The fact that at least in the VSO order, the Romanian subject must leave the vP and 

move at least as high as the T(Anterior) projection reminds one of Alexiadou and 

Anagnostopulou’s (2001) subject-in-situ generalisation (see Alexiadou and 

Anagnostopulou 2006 for a revised version). According to this generalisation, “vP can 

contain no more than one argument with an unchecked Case feature” by Spell-Out. The 

approach suggested here is stronger and states that besides Case and phi-features, 

whenever there is an IS-feature to value, vP-vacating must take place. 

4. Romanian SVO and Adverb Placement 

The Romanian subject may also surface preverbally. The status of the preverbal subject 

has constituted a matter of debate. It is claimed in Dobrovie-Sorin (1994), Cornilescu 

(1997) that the preverbal subject has been assigned Nominative Case in Spec VP and has 

moved to Spec IP which is seen as an A’ position, i.e. a Topic position, able to host 

51

topicalisations and left dislocations. 21 Cornilescu (2002) argues that such a position is 

accessible to subjects and non-subjects. The canonical A-position for subjects is in their 

accounts Spec VP. Conversely, Motapanyane (1994, 2002) argues that the preverbal 

subject moves to an argumental position, Spec TP. Adopting the view that the T head is a 

syncretic category, she argues that the preverbal subject moves to the outer specifier of T, 

the inner specifier being occupied by Focus. 

Another approach to the Romanian preverbal subject is the one advanced by Alboiu 

(2002) who argues that Romanian DPs check case in merge positions. In her account, 

Case checking is dependent on the nature of the EPP feature. Somewhat along the lines of 

Alexiadou and Anagnostopoulu (1998), she shows that the EPP can be parametrised 

cross-linguistically in the sense that there are languages which require [D], [V], or [T] as 

their EPP feature (though it is not clear to me what the exact difference between the V- 

and T-features really is). In what she calls T-type EPP languages, EPP attracts a predicate 

to Spec IP or to I. In D-type EPP languages, such as English, a DP or an expletive merges 

in Spec IP. Romanian is considered a V-type EPP language, where EPP triggers verb 

movement to I° and never NP movement to Spec IP. The uninterpretable phi-features on 

I° enter into Agree with those on the probe, i.e. the DP subject, Nominative case on the 

goal is thus checked. Since there is no strong [D] feature on the inflection head in 

Romanian, but only a [V] feature, the subject DP never moves to Spec IP and 

consequently no specifier of IP is projected. When in a preverbal position, the subject NP 

21 In Dobrovie-Sorin’s account, the Spec IP position can also be filled by various topicalised elements, such 

as object DPs and adverbials, the latter being exemplified in (i). 

(i) Astăzi vor sosi sportivii. 

Today will arrive sportsmen-the. 

‘Today the sportsmen will arrive.’ 

However, as will be argued, since the highest projection hosting the subject is associated with an EPPfeature, 

the ‘subject-of-predication’ feature, and consequently is Rizzi’s (2005, 2006) and Rizzi and 

Shlonsky’s (2007) Subject Criterial position, the adverbial cannot fill the preverbal subject position 

(Dobrovie-Sorin’s Spec IP or what is labelled here Spec SubjP). Rather, for reasons to be seen below and in 

section 6, the adverbial occupies a position at the left periphery of the clause, Spec Mod(ifier)P or TopP in 

the sense of Rizzi (2004) or some left-peripheral position related to scene-setting. 

52

is a topicalised subject that has moved to a D-linked position, unless it is contrastively 

focused. However, a quantifier of the type cineva ‘somebody’ in Cineva a dormit 

‘Somebody slept’ cannot be analysed as Topic. 

Cornilescu (1997) analyses the preverbal subject position as non-L-related, left- 

dislocated. In Cornilescu (1997, 2002), Romanian is analyzed on a par with Spanish 

(Zubizerreta 1998, Barbosa 2001; see also Hill 2002) in having a generalized TP, i.e. T° 

can check both Case and phi-features and operator features, such as [Focus], [wh]. T° is 

seen therefore as a multiple specifier head. In this account, T° can optionally have a 

strong D, which may be realized in two ways: as [+D, +phi]-features, and in this case a 

dislocated constituent is attracted to Spec TP, or as [+D, -phi]-features, and bare QPs are 

attracted to Spec TP. Crucially, analyzed in this light, inflection is pronominal and acts as 

a clitic with respect to the subject which is to Spec TP. She argues that Spec TP thus 

becomes a position available not only for subjects, but also for any doubled (in the sense 

of clitic doubled but crucially not clitic left-dislocated to the left periphery) DP, DOs and 

IOs alike. Recall from section 3 the availability of preverbal subjects and objects 

associated with New Information Focus in Romanian. Following Cornilescu’s account, 

the PP pe Ion in (48), which is obligatorily clitic resumed due to the presence of the 

accusative preposition pe, and the DP Mariei in (49), also obligatorily clitic resumed 

(proper nouns must be clitic resumed in Romanian), move to Spec TP. The PP and the 

DP are not separated from the rest of the sentence by an intonational break and are 

answers to questions like Pe cine ai văzut până la urmă ieri? ’Who did you finally see 

yesterday?’ and Cui i-ai dat cartea până la urmă ieri? ’Who did you finally give the 

book yesterday?’ 

(48) Pe Ion l-am văzut ieri. 

Pe Ion him-have.1sg seen yesterday 

’I saw Ion yesterday.’ 

(49) Mariei i-am dat cartea ieri. 

Maria.Dat her-have.1sg given book.the yesterday 

’I gave the book to Maria yesterday. 

53

However, when the PP and the DP above are given information about which a comment 

will be made, they move to the left periphery in Spec TopP, in which case an intonational 

break becomes possible. This is the classical CLLD construction (Cinque 1990, Rizzi 

1997, a.o.). More on issues related to topicalisation to the left periphery will be discussed 

in Chapter III. The PP and the DP above fill an A’ position as indicated by the WCO 

effects in the examples below (both sentences are understood as answers Pe cine au văzut 

părinţii ieri? and Cui au dat cartea prietenii ieri?): 

(50) a. ?* Pe Ioni l-au văzut părinţii luii ieri. 

Pe Ion him-have.3rd seen parents lui yesterday 

’*Hisi parents saw Ioni yesterday.’ 

b. ?* Marieii i-au dat cartea prietenii eii ieri. 

Maria.Dat her-have.3rd given book.the friends.the her yesterday 

’*Her friendsi gave the book to Mariai yesterday.’ 

Within the framework adopted here, where T° (or Subj° following the terminology in this 

analysis) has purely A-features and hosts Topic-aboutness sujects, it may be assumed that 

the object DPs above fill a Focus position at the left periphery of the clause. It may be the 

case that Rizzi’s (1997) FocusP can be associated with New Information besides being 

Contrastive in Romanian. I leave this issue open for further research. 

In keeping with cartographic guidelines, a single head cannot carry both purely A 

and operator feature specification. Consequently, the line of thinking to adopt departs 

from this view of the Romanian T°. I will show throughout the thesis that Romanian fits 

in well with the cartographic approach and that operator features, such as Contrastive Foc 

and wh (and other A’-related features), are checked in the left periphery and not in Spec 

TP. One of the arguments frequently used for the generalized T view for Spanish and 

Romanian is that the wh-phrase or a Contrastive Focus must be adjacent to the verb. 

Without insisting on this issue here, one way to account for this is to suggest the absence 

of any intermediate projection between Spec Foc and Fin°, the position of the moved 

verb, and movement of a wh/focalized element to the Spec FocP, the criterial position 

54

(for other counterarguments that Spec TP has mixed properties concerning focus 

movement and wh-phrases, see Chapter III). 

Below I concentrate on the position of the preverbal subject. In Rizzi (2005, 2006, 

2007) and Rizzi and Shlonsky (2007), the subject position, traditionally associated with 

Chomsky’s (1981) extended projection principle (EPP), is analysed in accordance with 

minimalist assumptions (Fox 2000 drawing on Chomsky’s 1986 last resort operation) that 

any movement operation must have an effect on the interfaces, i.e. a movement chain 

must terminate in an interface effect. By way of illustration, in Rizzi (op.cit.) 

interpretively relevant positions, such as the left peripheral positions of relative and 

interrogative operators are treated in terms of Criteria (see also Rizzi 1991, 1997). The 

definition of a Criterion is given in (51), where the relevant configuration here is (51a): 

(51) a. For [+F] a criterial feature, H+F is in a Spec-head configuration with A+F. 

b. For [+F] a criterial feature, H+F is locally c-commanded by A+F. 

F = Q, R, Top, Foc, … 

(Rizzi, 1991, 1997, 2006, 2007) 

These criterial positions, Top, Foc, R(elative) send ‘a message’ to the interfaces that their 

specifier positions be interpreted as Top, Foc or R(elative) operator. Movement to one 

such specifier position creates a chain that terminates in a criterial position. Somewhat 

similar to these left peripheral positions, it is shown in Rizzi (2005, 2006) and Rizzi and 

Shlonsky (op. cit.) that movement to the subject position is also movement to a dedicated 

interpretive position. More precisely, the subject position has a Topic-aboutness 

interpretation. A subject shares with a Topic the fact that it is “prominent”. It is about an 

event, hence Rizzi’s and Rizzi and Shlonsky’s denomination of Topic-aboutness. An 

argument is selected and constitutes the starting point in the description of an event. It 

may thus constitute “given information”. However, it differs from a real Topic in the left 

periphery in that it is not inherently D-linked in the sense of Pesetsky (1987). In the 

above-mentioned work, it is suggested that the subject head, Subj°, is the Mittelfeld- 

topmost projection, at the border between the CP and IP, sharing with the former the fact 

that it is a criterial position and with the latter the fact that it is an obligatory 

55

head/projection within the Mittelfeld/IP functional field. Therefore, the EPP attracting the 

DP subject to Spec SubjP is associated with Topic-aboutness. 

The idea often present in the literature on Romanian that the preverbal subject is a 

Topic or left-dislocated element can actually be recast in terms of the criterial approach: 

the preverbal subject in SubjP is about an event and is therefore interpreted as Topic- 

aboutness. In this sense, the word order SVO is more marked in Romanian if compared to 

the VSO word order, i.e. the postverbal subject cannot be topic-aboutness. In Romanian, 

the word order SVO is most often than not obtained as an answer to the question ‘What 

about Ion?’, encapsulating the idea of ‘aboutness’. In the spirit of Rizzi (2005), the 

following two question-answer pairs may be reproduced in Romanian. 

(52) A: Ce e cu Ion?/Ce a făcut Ion? 

‘What about Ion?’/‘What did Ion do?’ 

B: Ion a condus un camion. 

‘Ion drove a truck.’ 

B’: #Camionul, l-a condus Ion. 

‘The truck, Ion drove.’ 

B’’: #A condus Ion un camion. 

The only possible answer, B, takes over the DP subject already mentioned in the question 

and provides information about this argument (the event of driving a truck). In this sense, 

the subject becomes more salient, bearing a topic-like feature. Conversely, in B’, which is 

a CLLD construction, the dislocated DP camionul is expected to be ‘given’ information 

in A, which is not at all the case, and thus B’ is an infelicitous answer to the question in 

A. Similarly, B’’, in the form of VSO, is quite unnatural as an answer to A as the 

argument about which the predication holds (Ion) breaks the flow of new information ([A 

condusF] [IonTop-about] [un camionF]). 

It was seen in section 3 that the most natural answer to an out-of-the-blue context 

(“What happened?”) is the VSO word order. In such a case, the whole clause is taken to 

be FocusNew Information (alternatively, in VSO, the subject DP can be associated with a 

Topic-given feature and the object DP with a Focus New Information/Contrastive feature). 

56

However, an SV answer to a What happened? question is also possible with all the three 

verb types but in this case the preverbal subject is understood as Topic-aboutness, i.e. it is 

implied in the universe of discourse and selected to be talked about. No such 

interpretation is available to the subject in the VS order. Consider the examples below: 

(53) a. Un copil a căzut de pe scară. 

A child has fallen from onto the ladder 

‘A child has fallen down from onto the ladder’. 

a’. # Un copil a căzut. 

b. Un copil plânge în parc. 

A baby cry in park 

‘A baby is crying in the park.’ 

b’. # Un copil plânge. 

c. Hoţii au spart un butic Bouchard azi noapte. 

Thieves.the have broken.into a boutique Bouchard last evening 

‘The thieves broke into a Bouchard boutique last evening.’ 

In both (53a and b), the PPs bring additional information to the description of an event 

which is about the preverbal subject DP. Lack of this piece of information would result in 

unappropriateness of both sentences (53a’) and (53b’). 

It is generally acknowledged that a preverbal subject is semantically constrained, it 

must be specific. It is to note that there are circumstances where the SV(O) order is the 

only possible order. This is the case of stative verbs and all sorts of generic sentences. 

This is illustrated in (54). 

(54) a. Studentul/un student cunoaşte răspunsul. 

Student.the/a student knows answer.the 

’The student/A student knows the answer.’ 

a’. # Cunoaşte un student răspunsul. 

a’’. ?# Cunoaşte studentul răspunsul. 

b. Nichelul este un metal magnetic. 

57

Nickel.the is a magnetic metal 

’The nickel is a magnetic metal’. 

b’. # E nichelul un metal magnetic. 

c. Lâna se spală uşor. 

Linen se washws easily 

’Linen washed easily’. 

c’. # Se spală lâna uşor. 

(54a) contains a stative verb, (54b) involves a generic subject with a stage-level predicate 

and (54c) is a middle structure also with generic subject. The neutral postverbal subject in 

(54 a’,a’’), (54b’) and (54c’) is infelicitous unless it is contrastively focused. The 

interpretation of the preverbal subject with these predicates is generally taken to depend 

on the temporal and aspectual specification of the predicate. 22 

The postulation of a dedicated criterial subject position is therefore divorced from 

the requirements of the Case-agreement system (see Cardinaletti 2004 and Rizzi 2005, 

2006). One of the arguments used in Rizzi (op. cit.) comes from the existence of quirky 

22 

It is generally acknowledged that bare nominal subjects are not allowed in the preverbal subject position 

either unless under coordination, as in (ia), co-occurring with a quantifier, as in (ib), or in an enumeration, 

as in (ic). 

(i) a. Femei şi copii au părăsit oraşul Gori. 

Women and children have left town.the Gori 

’Women and children left Tbilisi.’ 

b. Multe balene au eşuat pe o plajă din Noua Zeelandă. 

Many whales have drifted ashore on a beach in New Zealand 

‘Many whales drifted ashore on a beach in New Zealand.’ 

c. Tinere gimnaste sunt nemulţumite, antrenori protestează, iar arbitri reclamă criterii de punctaj 

prea severe. 

Young gymnasts are not content, coaches object, and referees cry.out score criteria scales 

too strict 

‘Young gymnasts are not content, coaches object, and referees cry out to strict score criteria 

scales.’ 

58

subjects in Romance. The same argument may be extended to Romanian (and Spanish). 

As in Italian (55), Romanian also has quirky subjects with psych verbs (56). An example 

from Spanish is given in (57). 

(55) a. A Gianni piacciono queste idee. 

To Gianni like.3.pl these ideas 

‘Gianni likes these ideas.’ 

b. Queste idee piacciono a Gianni. 

(56) a. Lui Andrei îi plac mult fetele. 

To Andrei him.Cl like a.lot girls.the 

‘Andrei likes girls a lot.’ 

b. Fetele îi plac mult lui Andrei. 

(57) a. Yo sé que a Ana le gustan elos. 

I know that to Ana them.Cl like they. 

‘I know that Ann likes them.’ 

b. Yo sé que elos le gustan a Ana. 

In both Romanian and Italian, and partially in Spanish 23 , the nominative logical object 

triggering verb agreement can be 3 rd , 2 nd , or 1 st person in the presence of a dative logical 

subject. In Romanian and Spanish, but not in Italian, the dative must be doubled by a 

dative clitic. The dative argument moves to the preverbal subject position to check the 

topic-aboutness feature, or in Cardinaletti’s (2004) terms, the ’subject-of-predication’ 

feature. Following Rizzi’s (2005) analysis, Nominative case and phi-features are checked 

23 Some quirky constructions in Spanish pattern with Icelandic in that they display person restrictions, some 

such verbs are antojar (se) ‘to have a fancy on’, ocurrir (se) ‘imagine, think of’, olvidar (se) ’forget’. Such 

differences are captured by Rivero (2004) by the PCC (Person Case Constraint) which interacts with the 

use of a reflexive Accusative clitic form. If the dative-nominative relation is mediated by the reflexive 

Accusative clitic, no person restriction arises, whereas if this relation is mediated by such a reflexive, 

person restrictions on the inflection occur. 

59

y the postverbal theme which moves together with the verb (VP scrambling) to a 

position above the experiencer DP. No c-command relation obtains between the two DPs, 

as represented below for (56a). The dative experiencer can subsequently move to the 

Criterial Spec SubjP position. 

(58) SubjPCriterial 

2 

lui Andrei2 

Subj …φP 

2 

2 

φ …?P 

2 

2 

V + DP …?P 

2 

AAAGREE tlui Andrei …vP 

… 

Belletti and Rizzi (op. cit.) observe that when in Spec Subj, the dative argument does not 

interfere with movement, whereas when the same argument interferes with movement 

across a subject, the result is degraded. To my ear, such a slight degradation is also 

observed for Romanian. Compare (59) to (60). 

(59) a. Le idee che a Gianni piacciono di più sono queste. 

The ideas that to Gianni like most are these 

‘The ideas that Gianni likes most are these.’ 

b. ?(?) Le idee che a Gianni Maria raccomanda sono queste. 

The ideas that to Gianni Maria recommends are these 

‘The ideas that Maria recommands to Gianni are these.’ 

(60) a. Fetele care lui Andrei îi plac cel mai mult au plecat demult. 

Girls.the that to Andrei him.Cl like the most have left long.ago 

‘The girls that Andrei likes the most left long ago.’ 

b. ? Fetele pe care lui Andrei prietenii i-au prezentat au plecat demult. 

60

Girls.the that to Andrei friends.the him.C-have introduced have left 

long.ago 

‘The girls that to Andrei his friends introduced left long ago.’ 

In this section we will continue to determine the exact slots interspersed among the 

functional projections of the adverbs that the arguments and the verb target with the 

ultimate goal of detecting the distributional differences among Romanian, Spanish and 

Italian. 

Consider the position of the subject in Romanian. As shown in (61), it must precede 

deja ‘already’, and consequently, all adverbs lower than deja. That this is indeed the case 

is shown by the ungrammaticality of (61b-f) below. It is to note that except for the Voice 

adverb in (61f), these are adverbs that cannot prepose to the left periphery. 

(61) a. Ion deja a citit/citeşte câteva romane poliţiste. 

b. * Deja Ion a citit/citeşte romane poliţiste. 

c. 

Already Ion has read/reads novels detective 

’Ion already reads detective novels.’ 

* Incă Ion a citit/citeşte romane poliţiste. 

d. * Aproape Maria a făcut toate lecturile. 

e. * Complet pompierii au stins incendiul. 

f. * Uşor pompierii au stins incendiul. 

The subject DP can also precede aspectual adverbs, i.e. the celerative adverb 

repede/rapid ‘quickly’, as in (62), the frequentative adverb rareori rarely or 

adesea/deseori ‘often’, as in (63) as well as Voice adverbs, as in (64), but in this case the 

subject is topicalised to the left periphery and is separated by an intonation break from the 

adverb which is also moved to the left periphery, i.e. in a (Contrastive) Focus position. 

With neutral intonation the sentences are not very acceptable (a’-b’). 

(62) a. (?) Ion REPEDE/RAPID a recitat poemul. 

Ion quickly has recited poem.the 

61

‘Ion QUICKLY recited the poem.’ 

a’. ?? Ion repede/rapid a recitat poemul. 

(63) a. (?) Ion RAREORI a citit/citeşte romane poliţiste. 

Ion rarely has read/reads novels detective 

’Ion (has) RARELY reads/read detective novels.’ 

a’. ?? Ion rareori a citit/citeşte romane poliţiste. 

b. (?) Ion ADESEA a citit/citeşte romane poliţiste. 

Ion often has read/reads novels detective 

’Ion (has) OFTEN reads/read detective novels.’ 

b’. ?? Ion adesea a citit/citeşte romane poliţiste. 

(64) a. ? Copiii FRUMOS au împachetat cadourile. 

Children.the BEAUTIFULLY have wrapped.up presents.the 

‘The children wrapped up the presents BEAUTIFULLY.’ 

a’. * Copiii frumos au împachetat cadourile. 

With the habitual adverb de obicei ’usually’ the subject can expectedly precede or follow 

it. So is the case of higher adverbs, such as the Moodirrealis adverb poate ’perhaps’, of the 

Tfuture apoi ’then’, the Tpast adverb odată ’once’ and Modepistemic probabil ’probably’. This 

is illustrated in (65). 

(65) a. De obicei Ion a citit/citeşte romane poliţiste. 

Usually Ion ha sread/reads novels detective 

’Ion /has/ usually read/reads detective novels.’ 

a’. Ion de obicei a citit/citeşte romane poliţiste. 

b. Poate Ion a citit/citeşte romane poliţiste. 

Perhaps Ion has read/reads novels detective 

’Perhaps Ion read/reads detective novels.’ 

b’. Ion poate a citit/citeşte romane poliţiste. 

c. Probabil Ion a citit cartea. 

62

Probably Ion has read book.the 

’Ion probably read the book.’ 

c’. (?) Ion probabil a citit cartea. 

Consider further the examples below: 

(66) (Din fericire,) Ion (, din fericire), a citit (, din fericire,) cartea (, din fericire). 

Ion, happily, has read the book. 

’Happily, Ion has read the book.’ 

As was the case with the VSO word order, the MoodPevaluative PP adverbial preserves the 

parenthetical reading. The same result is expected with MoodPspeech-act adverb sincer 

’sincerely’and with the MoodPevidential adverb evident ’evidently, obviously’ in (67a,b). 

(67) a. Eu (, sincer), cred (, sincer,) acest lucru. 

I sincerely believe this thing. 

’Sincerely, I believe this.’ 

b. Ion (, evident,) a citit (, evident,) toată cartea. 

Ion, obviously, has read whole book-the. 

‘Obviously, Ion read the whole book.’ 

Comparative evidence from Italian suggests that such high adverbs do not display a 

parenthetical reading. In Italian, a subject position occurs to the left of the two high 

adverbs in (68a). The subject and the auxiliary verb may be adjacent, filling the Spec and 

head positions of the same projection, with the possibility of an adverb to precede them 

(68b), or another high adverb may intervene between Spec SubjP and the head filled by 

the auxiliary (68c) (these are Cardinaletti’s (2004) examples 123). 

(68) a. Gianni francamete purtroppo si era formato una pessima opinione di voi. 

Gianni frankly unfortunately si.Refl was made a very bad opinion of you 

‘Frankly Gianni unfortunately made himself a bad opinion of you.’ 

63

. Francamente Gianni si era purtroppo formato una pessima opinione di voi. 

c. Francamente Gianni purtroppo si era formato una pessima opinione di voi. 

Italian thus differs from Romanian in that a subject position is projected to the left of the 

highest functional projections associated with an adverb, i.e. there is a Spec-SubjP to the 

left of the MoodPspeech act, no parenthetical reading being involved. As shown in Cinque 

(1999:110), the subject DP in Italian must precede the terminative adverb mica ‘no 

longer’ and consequently all adverbs lower than mica. Some examples are given below: 

(69) a. Maria mica prende il treno. ‘Maria not takes the train.’ 

b. * Mica Maria prende il treno. 

c. * Già Maria è di ritorno, per la una. ‘Already Maria is back, at one o’clock.’ 

Cinque 1999 (3a,b,c): 110 

It is now interesting to see whether Spanish behaves more like Romanian where the set of 

the three high adverbs, speech act, evaluative and evidential, engender a parenthetical 

reading, or like Italian which displays a subject position in the upper Mittelfeld. Consider 

(70): 

(70) a. Afortunadamente (,) Juan leyó este libro. 

Unfortunately Juan reads this book 

b. Juan afortunadamente leyó este libro. 

c. Juan felizmente ha leido este libro. 

Juan happily has read this book 

d. Juan probablemente ha leido este libro. 

Juan probably has read this book 

These examples show that Spanish patterns with Italian in allowing a subject position 

above ‘”high” adverbs, such as the speech act adverb in (70b) and the evaluative adverb 

in (70c). Expectedly, the DP subject can also precede the epistemic adverb 

probablemente ‘probably’. However, Spanish patterns with Romanian in that the subject 

64

cannot precede “low” adverbs but the sentences improve if the latter are focalised. This is 

illustrated below. 

(71) a. * Los estudiantes fácilmente resolvieron los ejercicios. 

The students easily solved the exercises 

a’. ? Los estudiantes FÁCILMENTE resolvieron los ejercicios. 

b. * Los niños maravillosamente envolvieron los regalos. 

The children wonderfully wrapped up the presents 

b’. ? Los niños MARAVILLOSAMENTE envolvieron los regalos. 

Let us consider the French examples below: 

(72) a. * Jean franchement a lu ce livre. 

b. * Jean probablement lit ce livre. 

d. * Jean déjà a lu ce livre. 

e. 

* Jean toujours a lu ce livre. 

Unlike the languages discussed above, French does not allow any adverb class to occur 

between the subject and the auxiliary (or the verb). In other words, the subject and the 

auxiliary/verb are adjacent in French. 

Subject 

positions 

All these subject movement possibilities are summarized in Table 2 below. 

SubjP Moodspeech-act/eval/evid SubjP Modepist SubjP TPast … SubjP TAnt … SubjP Voice 

Rom * √ √ √ 

Sp √ √ √ √ 

It √ √ √ √ 

Fr * Subj-V Subj-V Subj-V Subj-V 

Table 2 

Let us further deal with verb movement in the languages considered. As commonly 

known (Dobrovie-Sorin 1994), in Romanian, no material can intervene between the 

65

subject and the verb/auxiliary except for clitic adverbs (*Ion a probabil/poate sosit ‘Ion 

has probably/perhaps arrived’). Consider first Romanian, illustrated in (73). 

(73) a. Ion a sosit/soseşte probabil. ’Ion probably arrived/arrives.’ 

b. Ion a sosit/soseşte poate. ’Perhaps Ion arrived/arrives. ’ 

c. Ion doarme/a dormit mult de obicei. ’Ion usually sleeps/slept a lot.’ 

d. Ion a luat/ia din nou trenul. ’Ion took/takes the train again.’ 

e. Ion a luat/ia adesea trenul. ’Ion often took/takes the train.’ 

f. Ion a sosit deja. ’Ion has already arrived.’ 

g. ? Ion a ajuns aproape. ’Ion almost arrived.’ 

g. Ion va ajunge/a ajuns imediat. ’Ion will come/has come 

immediately.’ 

h. Ion a darmat/daramat complet casa. ’Ion pulled down/pulls down the 

house completely.’ 

i. Ion a dormit/doarme bine. ’Ion slept/sleeps well.’ 

Except for speech act/evaluative/evidential MoodP adverbs which are parenthetical, the 

verb and the auxiliary in simple and compound tensed sentences may move to distinct 

floating positions, ranging from the low Voice adverb bine ’well’/uşor ’easily’ to a 

position above aspectual and Tense adverbs, and even above the ModP adverb probabil 

’probably’. Whereas the verb/auxiliary may therefore follow all these adverbs, it cannot 

follow adverbs belonging to the ”low space” in the hierarchy, completely and well, as 

suggested by the ungrammaticality of (74). This means that in Romanian, the verb or the 

auxiliary must raise at least to a position to the left of completely and well. 

(74) a. * Ion bine/uşor a rezolvat/rezolvă exerciţiile.’Ion solved/solves the exercises 

well/easily.’ 

b. * Pompierii complet au stins incendiul.’The firemen completely put out the 

66 

fire.’

However, as already observed, the sentences improve considerably if the adverb moves ot 

the left periphery, i.e. to Spec Foc. 

As far as Italian is concerned, Cinque (1999:152; fn.7:215; contra fn. 80:180) shows 

that the finite verb must raise higher than the Voice adverb bene ’well’ and 

completamente ’completely’ and may precede or follow all higher adverbs. The finite 

auxiliary can also move high up, i.e. higher than Moodspeech act adverbs (recall also 

example 68 above). This is illustrated in (74c,d) below. It can therefore float to distinct 

head positions to the right of the subject-oriented adverb saggiamente ’wisely’ in (74d). 24 

(75) a. * Maria bene fece tutti i compiti. 

’Maria well did her homework.’ 

b. * Maria completamente distrusse tutto quello che aveva fatto fino ad allora. 

‘Maria completely destroyed all she had done till then.’ 

Cinque 1999 (fn.7i(l,m): 214) 

c. Mi ero francamente purtroppo evidentemenete formato una pessima opinione 

di voi. 

‘Frankly I unfrotunately had clearly formed a very bad opinion of you.’ 

Cinque 1999, (15a): 49 

d. Gianni saggiamente ha accettato. 

‘Gianni wisely accepted.’ Cinque 1999, (19a): 49 

Insofar as Spanish is concerned, as in Romanian, no material can intervene between the 

auxiliary and the participial verb. In a compound tensed sentence, the auxiliary + 

participial verb moves to a position above the manner adverb, as in (76c) and can move to 

distinct positions, i.e. above Asphabitual adverb class, as in (76b), but crucially not above 

the epistemic adverb, as in (76a). 

(76) a. * Juan ha leído probablemente este libro. 

b. Juan ha leído a menudo este libro. 

24 As opposed to the finite verb, the finite auxiliary in Italian must move higher than Aspperfect sempre 

‘always’. 

67

c. Los estudiantes resolvieron facilmente los ejercicios. 

The students solved easily the exercises 

Let us further identify the possible floating positions of the verb and the auxiliary in 

French. Consider the cases below: 

(77) a. Les pompiers sont, franchement, arrivés très vite. 

‘Frankly the firemen arrived very quickly.’ 

b. Les pompiers sont(,) évidemment(,) arrivés très vite. 

‘Obviously the firemen arrived very quickly.’ 

c. Les pompiers sont probablement arrivés très vite. 

‘Probably the firemen arrived very quickly.’ 

d. Les pompiers ont complètement éteint l’incendie. 

‘The firemen put out the fire completely.’ 

e. Les pompiers ont complètement tout bien éteint. 

The firemen completely put everything well out.’ 

(78) a. Les pompiers éteignent évidemment cet immense incendie. 

‘Obviously the firemen put out this huge fire.’ 

b. Les pompiers éteignent probablement cet immense incendie. 

‘Probably the firemen put out this huge fire.’ 

c. Les pompiers éteignent souvent des immenses incendies. 

‘Firemen often put out huge fires.’ 

d. Les élèves lisent attentivement ce texte en chinois. 

‘The students carefully read this text in Chinese.’ 

The auxiliary in French moves to the head of a projection above the Voice adverb in 

(77e) and floats to distinct positions in the Mittelfeld up to the epistemic projection, as 

indicated by (77c) and may even move to a head position above the evidential, as 

suggested by the optional parenthetical reading of évidemment in (77b). However, it 

cannot move higher, otherwise a parenthetical reading obtains in (77a). Similarly, the 

68

simple verb must move higher than the Voice adverb, as suggested by (78d) and can float 

to a head position higher than aspectual adverbs, as indicated by (78c). It can further 

target the head position above probablement (78b) but cannot move any higher. If it does, 

a parenthetical reading obtains, as with évidemment in (78a). 

I now turn to identifying the floating positions of the object among the adverb 

projections. Let us first concentrate on Romanian. 

(79) a. Ion a adus florile(,) probabil. ‘Ion brought the flowers probably.’ 

a’. Ion a adus probabil florile. 

b. Ion aducea flori odată/odinioară. ’Ion would once bring flowers.’ 

b’. Ion aducea odată/odinioară flori. 

c. Ion a adus florile deja. ‘Ion has already brought the flowers.’ 

c’. Ion a adus deja florile. 

d. Ion a adus florile întotdeauna/mereu. ‘Ion has always brought the flowers.’ 

d’. Ion a adus întotdeauna/mereu florile. 

e. Elevii au dizolvat soluţia complet. ‘The students dissolved the solution 

e’. Elevii au dizolvat complet soluţia. 

completely.’ 

f. Elevii au rezolvat exerciţiile uşor.’ The students solved the exercises easily.’ 

f’. Elevii au rezolvat uşor exerciţiile. 

The examples in (79b-f) show that the DP object in Romanian can float to positions 

ranging from above the Voice adverb to a position quite high in the Mittelfeld, i.e. 

immediately above the projection hosting the past Tense adverb odată/odinioară ’once’. 

When following the object, these all adverbs bear heavy Focus stress. There are therefore 

object-related positions interspersed among the low completive aspect and the high past 

tense adverb projection. The object DP cannot move higher than the epistemic adverb, 

otherwise a parenthetical interpretation of the adverb obtains. As expected, the object can 

follow probabil and any lower adverb, as in (a’-f’). It can presumably remain in the VP, 

as suggested by (79f’). In pre-verbal position, all these adverbs, except probabil, have a 

69

neutral reading. However, the object cannot move to a position higher than the epistemic 

adverb, otherwise a parenthetical reading obtains in (79a). 

Let us next focus on Italian. Consider (80). Examples (80b,c,d) are Cinque’s 

(1999:61) examples (14). 

(80) a. Gianni ha letto il libro, probabilmente.’Gianni has probably read the book.’ 

a’. Gianni ha letto probabilmente il libro. 

b. Gianni ha ricevuto la notizia *(di) gia.’Gianni has already got the news 

b’. Gianni ha ricevuto già la notizia. 

already.’ 

c. Gianni vede Maria ancora. ‘Gianni is still seeing Maria.’ 

c’. Gianni vede ancora Maria. 

d. Gianni ha rifatto i compiti bene. ‘Gianni redid his homework well.’ 

d’. Gianni ha rifatto bene i compiti. 

As in Romanian, the DP object in Italian cannot move higher than the epistemic adverb, 

in which case the adverb has a parenthetical reading in (80a). As seen above for 

Romanian, the DP object can move higher than the Voice adverb and float to distinct 

specifier ObjPs up to the one dominating T (anterior) adverb projection. This is illustrated 

in (b-d). In these cases, the adverbs must bear heavy stress (Cinque 1999:14). The object 

in the SVO order can also appear lower than all the adverbs, as in (a’-d’). (d’) indicates 

that the object remains in the VP. 

As far as Spanish is concerned, the object cannot move higher than the epistemic 

adverb. It can float to distinct Spec ObjP positions among the Aspfrequentative adverb, as in 

(81b), the T(Anterior) adverb, as in (81c), the Aspcompletive adverb, as in (81d) and the 

Voice adverb in (81e). 

(81) a. * Las tropas destruyeron el pueblo probablemente. 

The troops destroyed the village probably 

b. Juan traía flores hace tiempo. 

Juan brought flowers once 

70

c. Juan ha traído las flores ya. 

Juan has brought the flowers already 

d. Las tropas destruyeron el pueblo completamente 

The troops destroyed the village completely 

e. Los estudiantes resolvieron los ejercicios fácilmente. 

The students solved the exercises easily 

As with the languages above, in French the object cannot move higher than the epistemic 

adverb. 

(82) a. Jean a amené des fleurs(,) évidemment. 

Jean has brought flowers obviously 

b. Jean a amené des fleurs probablement. 

Jean has brought flowers probably 

c. * Jean a amené des fleurs déjà. 

Jean has brought flowers already 

c’. Jean a amené déjà des fleurs. 

d. * Jean n’a amené de fleurs plus. 

Jean not has brought flowers any.longer 

d’. Jean n’a amené plus de fleurs. 

e. * Jean a amené des fleurs encore. 

Jean has brought flowers still 

e’. Jean a amené encore des fleurs. 

f. * Jean a amené des fleurs toujours. 

Jean has brought flowers always 

f’. Jean a amené toujours des fleurs. 

g. Les policiers ont détruit l’immeuble complétement. 

Les policemen have destroyed the building completely 

g’. Les policirers ont détruit complétement l’immeuble. 

h. Les élèves ont résolu les exercices attentivement. 

The students have solved the exercises carefully 

71

h’. Les élèves ont résolu attentivement les exercices. 

In French, the high adverbs in (82a-b) can only have a parenthetical reading. With 

adverbs belonging to the middle space of the adverb functional hierarchy, ungrammatical 

results obtain, as in (82c-f). Recast in different terms, the French DP object cannot move 

higher than aspectual adverbs but must remain lower, as in (c’-f’). As in Romanian, 

Italian and Spanish, the adverb in sentence final position in French also bears heavy 

stress. 

To sum up this section, several conclusions have been arrived at: 

(i) It has been shown that, as in Italian, Romanian SVO order is a marked order, where 

the subject is more prominent and is attracted to the criterial SubjP position where 

besides case and phi-features it checks/values a Topic-aboutness feature; 

(ii) In Romanian, the subject and the verb/auxiliary (rather the verbal complex) can be 

part of the same subdomain (structure sandwiched between two contiguous adverbs) and 

move to Spec SubjP and Subj°, respectively, above Modepistemic. The subject cannot 

remain inside VP but must leave the VP and raise at least as high as SubjP above 

TP(Anterior). In this respect the Romanian DP subject patterns with the Italian DP 

subject, i.e. in neither language can the subject remain in VP or in the lower portion of 

the Mittelfeld. The lexical preverbal subject in Italian patterns with the one in Spanish 

moving higher than it does in Romanian, i.e. above the evidential/ evaluative/speech-act 

adverbs; 

(iii) Therefore, in both Romanian and Spanish, Voice adverbs, whether short or long, as 

well as some quantificational adverbs (rareori ‘rarely’, adesea ‘often’) in pre-verb 

position move to Spec Foc in the left periphery; 

(iii) The finite verb must also vacate the VP and raise higher than VoiceP, preceding or 

following adverbs belonging to the ‘middle’ space of the Mittelfeld and can even raise 

above Modepistemic, as mentioned in (i) above. Recall that these options are also available 

in Italian, expect that the finite verb moves even higher as adverbs from the ‘high’ space 

are not parenthetical as they are in Romanian. Conversely, the verb in Spanish targets 

head position lower than the epistemic projection; 

72

(iv) Regarding the DP object in Romanian, it must also leave the VP and move to the 

specifier of an ObjP above VoiceP. It can further raise as high as above TP(Past). In 

French, the lexical object can move to a low portion of the Mittelfeld and cannot raise 

higher than Voice and AspCompletive projections; 

(v) French exhibits the adjacency subject-Aux/V. The object DP can move higher than 

Voice adverbs but cannot move higher. 

All these movement possibilities are provided below (only the non-focalised and 

non-parenthetical readings are put down): 

73

Subj Moodspeech/eval/evid Subj V Obj Modepist Subj V Obj TPast Subj V Obj TAnt Subj V Obj Voice Subj V Obj 

R * √ √ * √ √ √ √ √ √ * √ √ * * √ 

S √ √ * * √ √ √ √ √ √ * * √ 

I √ √ √ * √ √ √ √ √ √ * √ √ √ 

F * S+V * S+V * S+V * S+V √ √ 

Table 3 

74

5. Romanian VOS and Adverb Placement 

Like Spanish and to a lesser degree Italian (but unlike French) 25 , Romanian also has the 

VOS word order. It is argued by Ordóñez (1998, 2000) for Spanish, Cornilescu (1997) 

and Alboiu (2002) for Romanian and Cardinaletti (1997, 2004), Belletti (1999, 2004) for 

Italian (see also Belletti and Shlonsky 1995) that an adjunction to VP/IP analysis is not 

tenable and that postverbal subjects occupy an IP-internal position and the object 

scrambles/moves past the subject. 

For both Spanish and Romanian, some strong evidence in favour of this analysis 

comes from binding. Thus, in Alboiu (2002) one piece of evidence in favour of the object 

movement analysis past the subject DP comes from binding phenomena. In (84a) the 

direct object in the SO structure is felicitously bound by the subject. (84b) showing OS 

order is ungrammatical as the subject DP no longer c-commands the object. In keeping 

with the right-adjunction analysis, the subject fills a higher position, c-commanding the 

objects, thus leading to unaltered binding relations. The grammaticality of (84c) 

displaying quantified object suggests that the direct object felicitously binds the subject to 

its right, which can easily be accounted for under an object raising analysis. (84d) is ruled 

out because, as expected, SO structures exclude subject binding by the object. 

(84) a. L-a strigat [fiecare mamă]i [pe copilul ei]i. 

HimCl-has called every mother pe child.the her. 

’Every mother called her child.’ 

b. * L-a strigat [pe copilul ei]i [fiecare mamă]i. 

c. L-a strigat [pe fiecare copil]i [mama lui]i. 

HimCl-has called pe every child mother his 

d. * L-a strigat [mama lui]i [pe fiecare copil]i. Alboiu 2002 (24,26):128/130 

25 French exhibits Stylistic Inversion triggered by interrogative and subjunctive contexts, so that an VOS 

answer to the question ‘Who read the book?’ is impossible. 

75

The same facts and explanations hold valid for Spanish as well. Below are given for 

illustration Ordóñez’ (2000) examples: 

(85) a. * Aquí presentó sui madre (a) cada niñoi. 

Here introduced his mother each boy 

‘Here, his mother introduced each boy.’ Ordóñez’s (2000)(31d) 

b. Aquí presentó (a) cada niñoi sui madre. 

‘Here, his mother introduced each boy.’ Ordóñez’s (2000)(25d) 

In Romanian any type of DP object can move across the subject: a definite or an 

indefinite DP, a bare plural constituent, and even a bare NP (see examples below). Alboiu 

argues for Romanian that in such a position the object gets a presuppositional reading. I 

do not indulge in a discussion whether this reading can be assimilated to specificity 

which characterizes Spanish object movement past the subject (Ordóñez op. cit.). For the 

present purposes, it suffices to keep in mind that this movement has an effect on the 

outcome. 

VOS is an appropriate answer to a question where the verb and the object are 

repeated. Consider the question-answer pairs below: 

(86) A : Cine a scris un roman/romanul? 

‘Who wrote a/the novel?’ 

B : A scris un roman/romanul Ion. 

Has written a/the novel Ion 

‘Ion wrote a/the novel.’ 

B’: L-a scris Ion. 

(87) A: Cine l-a văzut pe Petre? 

‘Who saw Petre ?’ 

B : L-a văzut pe Petre Ion. 

Him.Cl-has seen pe Petre Ion 

‘Ion saw Petre.’ 

B’: L-a văzut Ion. 

76

Interpretively, the subject DP in the (B) answers is emphatic. It is associated with a Focus 

New Information reading. It will be seen below whether this is the lowest position or 

whether there may be floating positions associated with this particular feature. Whereas 

in Romanian (and Spanish), (B) is a natural answer to (A), in Italian it is not very natural 

(Belletti 2004) 26 . (B) would be fine in Italian only under the Contrastive Focus reading of 

the subject. The three languages are however alike in allowing an answer with the clitic 

form of the object, as in (B’). Recall that clitic doubling of a (preverbal or postverbal) 

DP with the preposition pe is obligatory in Romanian (87B). Alternatively, the postverbal 

subject may function as topic in the context of a question like (A) in (88) below and the 

sentence is associated with an intonational break which becomes even more prominent 

once other material is inserted between the object and the subject, as in (89). 

(88) A: Ce a scris Ion? 

‘What did Ion write?’ 

B: A scris un roman(,) Ion. 

(89) A scris un roman recent(,) Ion. 

Has written a novel recently Ion 

‘In all probabilities Ion wrote a novel.’ 

Given that the subject is old information, i.e. a right-dislocated Topic, the object is Focus 

New Information. Alternatively, VOS is a possible answer to a question on the verb, i.e. 

What did Ion do?, in which case the verb and the object constitute new information and 

the subject has, as above, a dislocated reading. 

Another possible interpretation for the VOS order is that of Constrastive Focus 

associated either with the object, followed by a deaccented DP subject, or with the 

subject. One possible analysis of the contrastively focused postverbal subject is that of 

26 Belletti (2004) stresses out that VOS is perfectly natural in Italian in special registers, such as TV reports 

or sports. 

77

Belletti (2004) where the constituent V+O moves to Spec Top in the left periphery 

whereas the subject targets Constrative Foc. 27 

Much the same way the argument got unfolded in the previous sections, arguments 

and verb movement possibilities are also identified with respect to the fixed positions of 

the adverb projections. 

Let us concentrate first on the position(s) of the object. Consider (90): 

(90) a. Au distrus/Distrug satul complet georgienii. 

Have destroyed/Destroy village.the completely Georgians.the 

‘The Georgians destroyed/are destroying the village completely.’ 

b. Transcrie cursul calm Ion. 

(90) shows that the object in the VOS order moves to a position above the adverbs in the 

very ‘low’ space of the Mittelfeld. The subject is interpreted either as Focus New 

Information or Topic, right-dislocated. In the former case, (90a) is a natural answer to the 

question Who destroyed the village completely?, whereas in the latter case it may be an 

answer to How did the Georgians destroy the village? In this latter interpretation the 

adverb bears focus stress. As shown in (91), the DP object can also move quite high in 

the Mittelfeld, i.e. to a position above Asphabitual adverb, and even (more) marginally 

above Modepistemic, as below. 

(91) a. Bombardau satele de obicei trupele inamice. 

Were.bombarding villages.the usually troops.the ennemy 

b. 

‘The enemies would usually bombard the villages.’ 

Au bombardat satul din nou/iarăşi trupele ruseşti. 

27 

If this is the right analysis for the contrastively focused postverbal subject, it poses a problem for a 

similar analysis of the object given that the left periphery of the Romanian sentence does not have a TopP 

below Contrastive Focus (to be discussed in Chapter III). I leave open the issue of whether the left 

peripheral Focus position is the only one associated with contrastive stress in Romanian or whether it may 

also be associated with New Information. What the properties of the left-peripheral and the right-peripheral 

Contrastive Foci are is an open question. 

78

Have bombarded village.the again troops.the Russian 

‘The Russian troops bombarded the village again.’ 

c. ?(?) Au bombardat satul probabil trupele ruseşti. 

Have bombarded village.the probably troops.the Russian 

‘The Russian troops probably bombarded the village.’ 

As above, in these cases, too, the postverbal subject in sentence-final position fills either 

Spec SubjP associated with New Information Focus (and 91b for instance is an answer to 

Who bombarded the village again?) or is right-dislocated and interpreted as Topic. On 

the first reading, the analysis goes as follows: The constituent representing given 

information is the V+O which in Belletti’s (2004) analysis moves to the specifier of a 

Mittelfeld-internal Top projection. In line with the analysis adopted thus far, the remnant 

VP containing the trace of the already moved subject is attracted to Spec ObjP due to the 

EPP associated with a Topic-feature. The subject DP has previously moved to a lower 

Spec SubjP whose head has an EPP associated with the Focus New Information feature. 

However, as pointed out, the VOS configuration in Romanian (and Spanish) is freer than 

it is in Italian. On the assumption that V and O do not move as remnant VP, the DP object 

in both Romanian and Spanish may be analyzed as a disguised PP. Given that the two are 

clitic-doubling languages, following proposals on clitic doubling (Uriagereka 1995, 

Torrego 1995, 1998), the DP is more than a mere DP, i.e. it is a PP with a dummy 

preposition and an overt clitic if the object is doubled, or an empty preposition and a null 

clitic if it is not doubled. The subject DP moves to the Mittelfeld-internal Spec SubjP 

associated with a New Information Focus feature and the object (a disguised PP) targets 

Spec ObjP and hence the lack of RM effects in the VOS word order. 

Let us further what happens to the object in Spanish. Consider the sentences below: 

(92) a. Destruyeron el pueblo completamente las tropas rusas. 

Destroyed the village completely the Russian troops 

‘The Russian troops destroyed the village completely.’ 

b. Resolvieron los ejercicios fácilmente los estudiantes. 

Solved the exercises easily the students 

79

‘The students solved the exercises easily.’ 

c. Destruyeron el pueblo a menudo las tropas rusas. 

Destroyed the village often the Russian troops 

‘The Russian troops often destroyed the village.’ 

d. * Destruyeron el pueblo probablemente las tropas rusas. 

Destroyed the village probably the Russian troops 

‘The Russian troops probably destroyed the village.’ 

As in Romanian, the DP object in Spanish moves to a position above “low” and “middle 

space” adverbs, as illustrated in (92a-c). Interpretively, either the adverb itself or the 

subject is Focus, in this latter case the subject fills New Information Focus, as in 

Romanian (see discussion on subject movement possibilities below). If the adverb bears a 

focusing reading, the DP subject is right-dislocated, an intonation break setting off the 

two elements. Regarding the object, it cannot move past the Advepistemic projection hosting 

probablemente. 

Let us next focus on the verb. Consider (93): 

(93) a. Probabil au bombardat satele trupele ruseşti. 

Probably have bombarded villages.the troops.the Russian 

‘The Russian troops probably bombarded the villages.’ 

a’. Au bombardat probabil satele trupele ruseşti. 

b. De obicei bombardau satele trupele ruseşti. 

Usually were.bombarding villages.the troops.the Russian 

‘The Russian troops would (usually) bombard the villages.’ 

b’. Bombardau de obicei satele trupele ruseşti. 

c. Distrug complet satele trupele inamice. 

Destroy completely villages.the troops.the ennemy 

‘The enemy troops destoy the villages completely.’ 

c’. ? COMPLET distrug satele trupele inamice. 

d. Rezolvă uşor exerciţiile toţi elevii din clasă. 

Solve easily exercises.the all students in class 

80

‘All students in class solve the exercises easily.’ 

d’. ? UŞOR rezolvă exerciţiile elevii (toţi). 

On the assumption that high adverbs fill their first-merge position and do not move, the 

verb in Romanian can either remain lower than the epistemic adverb, as in (93a) or move 

above, targeting Fin° in (93a’). Expectedly, the verb can float to distinct head positions 

above low adverbs (b’). However, with neutral intonation, the verb must move higher 

than completely and easily, i.e. they must leave the vP domain (c-d). The verb can also 

follow these two low adverbs but if so, the adverbs marginally have contrastive focus 

stress and fill Spec Focus in the left periphery. Alternatively, the adverb may be 

topicalised in the left periphery. 

Next let us focus on the interaction of the verb movement possibilities among 

adverbs in Spanish. Consider the examples below: 

(94) a. ?? Destruyeron probablemente el pueblo las tropas rusas. 

Destroyed probably the villages the troops Russian 

‘The Russian troops probably destroyed the villages.’ 

b. Destruían habitualmente el pueblo las tropas rusas. 

Destroyed usually the villages the troops Russian 

‘The Russian troops usually destroyed the villages.’ 

c. Resolvieron fácilmente los ejercicios los estudiantes. 

Solve easily the exercises the students 

‘The students solve the exercises easily.’ 

d. Destruyeron completamente el pueblo las tropas rusas. 

Destroyed completely the villages the troops Russian 

‘The Russian troops destroyed the villages completely.’ 

Spanish patterns with Romanian in that the verb moves above the ‘low’ Aspcompletive 

projection and the ‘middle’ Asphabitual projections, as shown in (94b-c). However, as 

opposed to Romanian, the verb in Spanish cannot move higher than the Modepistemic 

adverb probablemente (cf. 93a’-94a). The only possible interpretation for the DP subject 

81

is that of New Information Focus in the sense of Belletti (op. cit.). If the verb follows 

‘low’ adverbs, the latter must either be focalised or topicalised. If it is topicalised, an 

intonation break sets if off from the verb. Put differently, in the VOS order, ‘low’ adverbs 

must move either to Spec Foc or Spec Top in the left periphery of the clause and the verb 

remains in a lower position. In this respect Spanish is like Romanian. The relevant 

examples with the focused adverbs are provided below: 

(95) a. COMPLETAMENTE destruyeron el pueblo las tropas rusas. 

COMPLETELY destroyed the villages the troops Russian 

‘The Russian troops COMPLETELY destroyed the villages.’ 

a’. * Completamente destruyeron el pueblo las tropas rusas. 

b. FACILMENTE resolvieron los ejercicios los estudiantes. 

EASILY solve the exercises the students 

‘The students EASILY solve the exercises.’ 

b’. * Facilmente resolvieron los ejercicios los estudiantes. 

Let us finally consider subject movement possibilities among the adverbs’ projections. 

Some relevant examples are given below: 

(96) a. ?(?) A descris/Descrie incidentul Ion calm. 

b. ?* Au distrus/Distrug satul inamicii complet. 

Have destroyed/Destroy village.the enemies.the completely 

‘The enemies destroyed/are destroying the village completely.’ 

c. ?? Iau notiţe elevii mereu. 

Take notes pupils.the always 

‘Students frequently write each other short notes.’ 

d. ?? Îşi scriu bileţele elevii frecvent. 

Refl. write notes pupils.the frequently 

‘Pupils frequently write each other short notes.’ 

e. * Îşi scriu bileţele elevii probabil. 

Refl. write notes pupils.the probably 

82

‘Pupils probably write each other short notes.’ 

These sentences show that the DP subject in the VOS order can only marginally be 

followed by a Voice adverb (96a), other adverb classes producing either more marginal 

results or completely ungrammatical results (96b-e). Not even a focusing reading is 

available to post-subject adverbs. These data indicate that in Romanian the subject in the 

VOS configuration fills Belletti’s (2004) New Information Focus at the periphery of the 

vP. 

Not only in Romanian but also in Spanish, the DP subject in the VOS order cannot 

move to a position above any adverb position (notice however the slightly less 

ungrammatical result with the Voice adverb, the same effect having been observed in 

Romanian (97a)). Thus, Spanish, too, avails itself of Belletti’s New Information Focus at 

the periphery of the vP. 

(97) a. ?? Resolvieron los ejercicios los estudiantes fácilmente. 

Solved the exercises the students easily 

‘The students solved the exercises easily.’ 

b. * Destruyeron el pueblo las tropas rusas completamente. 

Destroyed the village the Russian troops completely 

‘The Russian troops destroyed the village completely.’ 

c. * Destruyeron el pueblo las tropas rusas a menudo. 

Destroyed the village the Russian troops often 

‘The Russian troops often destroyed the village.’ 

d. * Destruyeron el pueblo las tropas rusas probablemente. 

Destroyed the village the Russian troops probably 

‘The Russian troops probably destroyed the village.’ Valmala (p.c.) 

To sum up this section, the distributional properties of arguments and the verb among the 

adverbs considered can be stated as follows: 

(i) The VOS order in Romanian (and Spanish) is freer than it is in Italian and may be 

associated with several IS-values regarding both the subject and the object; 

83

(ii) Ignoring its dislocated Topic reading, the subject DP in both Romanian and Spanish 

cannot be followed by any adverb class and is analyzed as occupying Spec New 

Information Focus in the sense of Belletti (or recast in terms of the analysis adopted 

throughout, the EPP associated with Focus New Information on a low Subj° attracts the 

subject to the specifier position and no other Spec SubjP is available in the structure); 

(iii) On the assumption that the object DP is a disguised PP in Romanian and Spanish, 

neither the VSO configuration, not the VOS configuration induces any RM effects; 

(iv) The DP object moves to floating Spec ObjP positions above the Asphabitual adverb; 

(v) The verb in VOS can move as high as Fin°; 

(vi) Some adverbs (especially low adverbs) can front to Spec Focus in the left periphery 

or even Spec Top. 

6. The Distribution of Adverbs in the Vorfeld 

It has been seen in the sections above that the adverbs in the Mittelfeld fill their merge 

positions and that it is argument DPs and the verb that may float to distinct positions 

interspersed among different adverb-related projections. But adverbs, too, can move. This 

section bears on another aspect of Romanian adverb syntax, namely adverb movement 

possibilities within the left periphery of the clause, or the Vorfeld. 

Following the cartography of structural positions proposed by Rizzi (2004a,b), the 

complementizer system is assigned the structure in (98) for Italian. It is shown in Rizzi 

(op. cit.) that the CP layer contains three potential positions for preposed adverbs to land 

in: the Spec Mod(ifier), the Spec Foc and the Spec Top position. An adverb in the Spec- 

Modif position which is situated lower than the operator area in Italian is licensed by the 

appropriate head whose role is to render the moved adverb prominent, such a property 

being also shared by Topics. However, an adverb moved to Spec Top presupposes strong 

connection with the discourse context. Besides Spec Modif and Spec Top, an adverb can 

also move to Spec Foc and gets the intonational contour of a contrastive focus. 

(98) ForceP > TopP* > IntP > TopP* > FocP > Mod(ifier)P* > TopP* > FinP > IP … Italian 

84

We begin the analysis with adverbs filling the Spec Modif. A general observation may be 

in order here: The adverbs that can move to such a position may express manner, time, 

and also, different types of aspect, as indicated in (99). However, rather low aspectual 

adverbs, such as the retrospective tocmai, ’just’, the prospective (de-)abia ‘hardly’ and 

the continuative tot ‘still’ as well as the completive complet ‘completely’, the repetitive 

iar ‘again’, or the prospective aproape ‘almost’ cannot get preposed into Spec Mod. 

(99) √ odată/odinioară √ apoi √ de obicei √ adesea √ repede √ deja √ calm 

once then usually often quickly already calm 

Consider the cases below. 

(100)Adesea, martorul a spus adevărul. AspPfrequentative 

Often witness.the has told truth.the 

‘Often the witness told the truth.’ 

(101) Rapid, judecătorul a înţeles despre ce este vorba. AspPcelerative 

Quickly judge.the has understood what it is about 

‘Quickly the judge understood what it was about.’ 

(102) Recent, copiii au cântat colinde. TPanterior 

Recently children.the have sung carols 

‘Recently the children have sung carols.’ 

(103) Calm, copiii s-au îndreptat spre ieşire. VoiceP 

Calmly children.the s.Refl-have made for exit 

‘Calmly the children made for the exit.’ 

After having been licensed in the appropriate Spec positions in the Mittelfeld, the 

frequentative, celerative aspect, time and manner adverbs in (100-103) above move to the 

85

left periphery filling the Spec Modif, being licensed by the Modif heads, thus acquiring 

structural prominence in the absence of any connection with the discourse. 

Besides Modif(ier)P, another position available to adverbs at the left periphery is 

that of Spec FocP. Let us examine the cases below. 

(104)ADESEA a spus martorul adevărul. AspPfrequentative 

Often has told witness.the truth.the 

‘It is often that the witness told the truth.’ 

(105)RAPID a înţeles judecătorul despre ce este vorba. AspPcelerative 

Quickly has understood judge.the what it is about 

‘It is quickly that the judge understood what it was about.’ 

(106) RECENT au cântat copiii colinde. 

Recently have sung children.the carols 

TPanterior 

‘It is recently that the children have sung carols.’ 

(107) CALM s-au îndreptat copiii spre ieşire. 

Calmly s.Refl-have made children.the for exit-the 

‘It is calmly that the children made for the exit.’ 

VoiceP 

In the examples above, the aspect, time and manner adverbs bear heavy stress and move 

to the left periphery filling Spec Foc. Evidence that this is FocP comes from the 

incompatibility of any of the above adverbs and a wh-phrase: 

(108) * RAPID/RECENT/CALM cui (i-)ai vorbit? 

Quickly/Recently/Calmly who (i.Cl-)have talked 

‘Who have you quickly/recently/calmly talked to?’ 

86

Adverb movement to such a position triggers obligatory subject-verb inversion. No 

material can intervene between the adverb and the verb. In other words, the required 

Focus-V adjacency is created. 

In his discussion on the typology of the licensing substantive features, Rizzi 

distinguishes the special case of the Topic position that an adverb can occupy at the left 

periphery. Essentially an adverb in this position can evade Relativised Minimality effects 

(Topics fall outside his typology of licensing substantive features: argumental, 

quantificational, modificational). Interpretively, it presupposes strong connection with the 

previous discourse situation, i.e mention of the adverb must have been made in the 

previous discourse. 

Consider the context situations given in (109-110). 

(109) Aceşti copiii sunt cunoscuţi pentru faptul că merg adesea în străinătate să cânte. 

Aş vrea acum să ştiu: recent, ce gen de muzică anume au cântat? 

‘These children are known to often go singing abroad. Now I would like to know: 

recently, what kind of music have they sung?’ 

(110) Petru mi-a spus că şi-a făcut unele teme repede, iar altele încet, în funcţie de cât 

de grele erau. Aş vrea acum să ştiu: repede, ce obiect a terminat? 

‘Petru told me that he had done some of his homework quickly, and some slowly. 

Now I would like to know: quickly, what subject did he finish?’ 

(109)-(110) constitute cases of special contextual circumstances where the adverb is 

either presupposed by the presence of the frequentative often, as in (109), or mentioned in 

the immediately preceding discourse, as in (110). Its being anchored in the previous 

context allows it to have a freer distribution than in simple preposing, undergoing 

movement to the dedicated Spec Top position. It can precede a wh-phrase (it would be 

unnatural to have a preposed adverb in a wh-question). 

As in Italian, movement to Spec Top can cross negation, a hierarchically higher 

adverb and Focus. This is illustrated below: 

87

(111) La actuala ediţie a Jocurilor olimpice echipa de caiac-canoe a obţinut rezultate 

slabe. Federaţia se întreabă dacă curând, la urmatoarele campionate europene şi 

mondiale, caiaciştii vor putea obţine vreo medalie. 

‘In these Olympic games the kayak-canoe team has scored meagre results. The national 

sports committee wonders whether any time soon, in the next European and World 

championships, the kayakers will be able to get any medal.’ 

a. Curând, nu va câştiga nici un canoist nici o medalie căci nu sunt bine pregătiţi. 

Soon, not will get no canoer no medal as are not well trained 

’Soon no canoers will get any medal as they are not well trained.’ 

b. Curând, probabil nu va câştiga nici un canoist nici o medalie căci nu sunt bine 

pregătiţi. 

’Soon no canoer will probably get any medal as they are not trained.’ 

c. Curând, NUMAI LA GIMNASTICĂ se mai speră câte o medalie. 

’Soon ONLY IN GYMNASTICS they hope to get a medal.’ 

No RM effect arises as there is no identical feature chain crossing. Conversely, as 

expected, a quantificational adverb, adesea ‘often’ in (112a) below, blocks wh-movement 

of an adjunct, whereas a non-quantificational adverb, recent ‘recently’, does not, as 

shown in (112b). 

(112)a. * Cum au ajuns adesea copiii la destinatie? 

‘How did the children often reach their destination?’ 

b. Cum au ajuns copiii recent la destinatie? 

‘How have the children recently reached their destination?’ 

As also discussed in Rizzi, adverb movement to Spec Foc is blocked by negation as the 

two elements are quantificational. This is also borne out by the Romanian data below: 

(113)a. * RECENT nu au cantat copiii colinde. 

‘RECENTLY the children haven’t sung carols.’ 

88

. * ADESEA nu a spus martorul adevarul. 

‘OFTEN the witness didn’t say the truth.’ 

There are some adverbs which have rather restricted use: tocmai ‘just’, (de-)abia 

‘hardly’, tot ‘still’. Examples are provided below: 

(114)a. (Ion) tocmai l-a întâlnit (Ion) pe Petre (Ion). AspPretrospective 

Ion just himCl-has met pe Petre 

‘Ion just met Petre.’ 

b. * (Ion) l-a întâlnit (Ion) tocmai pe Petre (Ion). 28 

c. * Tocmai Ion l-a întâlnit pe Petre. 29 

(115)a. (Dan) abia îl aşteaptă (Dan) pe fratele Mariei (Dan). AspPprospective 

Dan hardly himCl waits pe brother MariaDat 

Dan can hardly wait for Maria’s brother. 

b. * Il aşteaptă abia Dan pe fratele Mariei. 

c. * Abia Dan îl aşteaptă pe fratele Mariei. 

(116)a. (Deşi e supărat,) (Diana) tot îl întreabă (Diana) până la urmă ce AspPcontinuative 

s-a întamplat ieri seară. 30 

(Though he is upset) Diana still him finally asks what s.Refl-has happened last 

evening 

‘(Though he is upset) Diana still finally asks him what happened last evening.’ 

b. * (Desi e suparat,) îl întreabă tot Diana s-a întamplat ieri seară. 31 

c. * Tot Diana îl întreabă ce s-a întamplat ieri seară. 32 

28 The function of the retrospective adverb tocmai as a modifier of the DP pe Petre is ignored here. 

29 The adverb as a phrasal modifier of the DP Ion renders the sentence perfetc but this is not of interest 

here. 

30 As will be seen in section 7, tot is ambiguous between a continuative reading and what will be termed a 

continuous reading. Essentially, in the former it is an XP adverb. 

31 The adverb tot can modify subject DP to its right and the sentence become sperfect. Such an 

interpretation does not fall within the scope of this paper. 

89

These adverbs exhibit a constrained behaviour in that they can occur only in the preverbal 

field. Neither verb movement across the adverbs, as in the (b) cases, nor their preposing 

to the left periphery across the subject, as in (c), is available. Though they may resemble 

clitic adverbs, they are not as they must precede a pronominal or the auxiliary. 

To sum up, it has been seen in this section that not only the DP and the V can move 

in the Mittelfeld among the functional projections of the adverbs, but the adverbs can also 

move to the left periphery in one of the three positions available: Modifier, Focus and 

Topic (under the specific discourse situations), being thus licensed by the appropriate 

features on the respective head. When the adverb fills the Spec Focus position, the 

adjacency with the verb is required, this being a more general phenomenon specific to 

Romanian and Spanish. The observation has also been made that there are few adverbs in 

Romanian that can only fill a preverbal position, adjacent to the whole series of elements 

cliticising on the verb (pronouns and auxiliaries). 

7. The Case of Clitic Adverbs 

It has occasionally been pointed out in this chapter that the Auxiliary-Verb adjacency, 

characteristic of Romanian (and Spanish), can only be interrupted by a restricted set of 

short monosyllabic adverbs, called clitic adverbs. Such a class includes the following 

items: şi ‘quickly’, mai ‘again’/’still’/‘no longer’, tot ‘still’/’continuously’, cam ‘rather, a 

little’ and prea ‘really, too…’. This section is rather descriptive and merely suggests an 

analysis of the clitic elements in terms of Kayne’s (1994) antisymmetric theory. 

These adverbs do not constitute a homogenous class, some presenting idiosyncratic 

behaviour with respect to negation 33 in simple and/or compound tenses or as far as the 

32 

See fn. 30. 

33 

Negation is expressed in Romanian by the syntactic clitic nu which precedes the lexical verb/the 

auxiliary. In a string of clitic elements it occupies the leftmost position. Though requiring a syntactic host, 

it does not attract verb incorporation (i.e. it never relies phonologically or morphologically on the verb). 

Romanian also has another negative marker, ne-, whose distribution is restricted only to gerundive clauses 

90

position with respect to the auxiliary is concerned. As will be seen, all of them are 

aspectual adverbs belonging to distinct aspect classes. 

The first adverb to consider is şi ‘quickly’ in the sentences below. Only the 

preverbal position of the subject is given. 

(117) a. (Maria) îi şi divulgă secretul. 

Maria him quickly give.out secret.the 

‘Maria quickly gives out the secret to him.’ 

b. (Maria) i-a şi divulgat secretul. 

Maria him-has quickly given.out secret.the 

‘Maria has quickly given out the secret to him.’ 

(118) a. Compania aeriană i-a şi anunţat deja vestea cea tristă. 

Company air them-has quickly let.know already news.the the sad 

‘The air company has already quickly let them know the sad news.’ 

b. * Compania aeriană deja i-a şi anunţat vestea cea tristă. 

Though this adverb also has the flavour of temporal priority thus belonging to the 

TP(anterior), its more salient reading in (117) and (118) is that the celerative aspectual 

one which indicates that the action has been performed quickly. In (117) Maria was quick 

in giving out the secret and in (118) the company was quick in letting them know the sad 

news. It occurs between a pronominal clitic and the lexical verb in simple tenses or 

between the auxiliary and the participial verb. On the view to adopt later on that such 

clitics materialize heads of distinct adverb functional projections, the contrast in (118) 

shows that şi materializes the head of AspcelerativeP which is immediately higher than 

TP(anterior). 

and deverbal adjective structures. The negative affix is ususally taken to display a morphologically 

subcategorised position to which the verb and only the clitic mai ‘no longer’, if present, must raise, 

resulting in ne+mai+auzind…> nemaiauzind ‘no longer hearing…’ No other clitic adverb has this option 

with ne-. This is expected given that mai can function as a terminative aspect adverb in the scope of a 

negative element. 

91

As far as negation is concerned, şi is incompatible with it in both simple and 

compound sentences, as indicated by the ungrammaticality of (119a,b) in both simple and 

compound sentences. As shown in (119c), a pre-negation position is also ruled out. As 

pointed out in Dobrovie-Sorin, negation is leftmost in the clitic string in Romanian. 

(119) a. * (Maria) nu îi şi divulgă secretul. 

Maria not him quickly give.out secret.the 

‘Maria doesn’t quickly give out the secret to him.’ 

b. * (Maria) nu i-a şi divulgat secretul. 

Maria not him-has quickly given.out secret.the 

‘Maria hasn’t quickly given out the secret to him.’ 

c. * (Maria) şi nu i-a divulgat secretul/şi nu îi divulgă secretul. 

The second adverb to consider is mai. Just like English which has still as the counterpart 

of no longer, or Italian which also exhibits the pair (non…) piu and ancora, Romanian 

has one adverb, mai, that can change meaning according to the presence/absence of 

negation. Thus, the positive adverb mai has the meaning ‘still’ and belongs to the 

continuative class, while the negative adverb means ‘no longer’ and whenever under 

negation it gets a terminative aspect reading. Besides, as a continuative, it may co-occur 

with the full XP adverb încă ‘still.’ Consider the cases below. 

(120) a. (Ion) îi mai cumpără prăjituri copilului (deşi acesta nu are voie să mănânce 

dulciuri.) 

Ion him still buys cakes child.Dat (thouh this one must not eat sweets) 

‘Ion still buys cakes for the child (thouh he must not eat sweets).’ 

b. (Ion) i-a mai cumpărat prăjituri copilului (deşi acesta nu are voie să mănânce 

dulciuri.) 

‘Ion still bought cakes for the child (thouh he must not eat sweets).’ 

c. (Ion) nu îi mai cumpără prăjituri copilului. 

Ion not him any longer buys cakes child.Dat 

‘Ion did not buy cakes for the child any longer.’ 

92

d. (Ion) nu i-a mai cumpărat prăjituri copilului. 

Ion not him-has any longer bought cakes child.Dat 

‘Ion did not buy cakes for the child any longer.’ 

Though it may function either as a positive adverb or as an NPI, it has a regular 

behaviour with respect to negation in both simple and compound tenses, i.e. it appears 

between the pronominal clitic and the verb, as in (120a-c) and between the auxiliary and 

the participial verb, as in (120b-d). 

This adverb is homophonous with the repetitive mai. In (121) it scopes over the 

event of coming over here (several times). Given that it must be higher than the 

T(anterior) deja, it fills the higher AspPrepetitive. 

(121) a. A mai trecut pe aici deja (de câteva ori) de la ultima noastră întâlnire. 

Has again come over here already (several times) since our last meeting 

‘He has come over here again already (several times) since our last meeting.’ 

b. ?? Deja a mai trecut pe aici (de câteva ori) de la ultima întâlnire. 

The third adverb to deal with is tot. As illustrated in (122a), it has a double status, i.e. it 

may function either as full adverb or as clitic adverb but with meaning differences. In one 

meaning in (a) and in its pre-auxiliary position (b), it functions like a full XP adverb and 

relates to the continuative aspect head. However, since it points to the continuity of a 

process, this adverb may also be related to an immediately lower continuous aspect head 

(see also Cinque 1999:96) and in this interpretation it behaves like a clitic. This is 

illustrated in (122a) in one meaning and in (122c) and (123). Notice the co-occurrence 

with the adverbial într-una ‘continuously’ in (123) which indicates that it is a non-stop 

eventuality as opposed to the buying event in (122c) which could have been interrupted 

and the sentence would still be true. 

(122) a. Ion tot cumpără ziare şi reviste. 

Ion on.and.on buys newspapers and magazines. 

Interpretation 1: ‘Ion keeps on buying newspapers and magazines.’/’Ion is 

93

uying newspapers and magazines on and on.’ 

Interpretation 2: ‘Ion still buys newspapers and magazines.’ 

b. (Deşi mai avea doar foarte puţini bani), Ion tot a cumpărat ziare şi reviste. 

(Though he had very little money left) Ion still has bought newspapers and 

magazines 

‘(Though he had very little money left) Ion still bought newspapers and 

magazines.’ 

c. Ion a tot cumpărat ziare şi reviste. 

Ion has on.and.on bought newspapers and magazines 

‘Ion bought newspapers and magazines on and on.’ 

(123) Cât era ziua de lungă, bătrânul tot şedea într-un şezlong pe prispa casei şi tot vorbea 

(într-una) despre pământ. 

All day long the old man on.and.on would.sit in a loungechair on the porch of the 

house and on.and.on talk (continuously) about land 

The homophanous nature of tot is also reflected in the distinct behaviour with respect to 

the pronominal se below: 

(124) Ion (tot) s-a (tot) dus la poliţie. 

Ion still se-has on.and.on gone to police 

Interpretation 1: ‘Ion still went to the police.’ 

Interpretation 2: ’Ion kept on going to the police.’/’Ion went to the police on 

and on.’ 

The clitic adverb follows the pronominal clitic element and the auxiliary in the 

continuous aspect reading, but precedes both in its continuative reading. 

In (125) the behaviour of the adverb with respect to negation is tested. It is expected 

that the continuative tot is not sensitive to negation, whereas the continuous tot is. This is 

indeed borne out by the data below. As the glosses in (125a,b) show, the clitic adverb 

cannot co-occur with negation, it can neither fill a position lower than the negative clitic 

94

element, nor even lower than the auxiliary. Negation can only interact with the full 

adverb tot in the same sentences. The clitic’s incompatibility of occurring with negation 

parallels the behaviour of şi and mai (except for the latter’s terminative aspect reading 

which requires negation). 

(125) a. Ion tot nu (*tot) cumpără ziare şi reviste. 

Ion still not buys newspapers and magazines 

‘Ion still doesn’t buy newspaper and magazines.’ 

b. Ion tot nu (*tot) a (*tot) cumpărat ziare şi reviste. 

Ion still not has bought newspaper and magazines. 

‘Ion still didn’t buy newspapers and magazines.’ 

The fourth adverb, cam, meaning ‘rather, a little’, also breaks the Aux-Verb adjacency. It 

expresses approximation. 

(126) a. (Ion) cam greşeşte exerciţiile. 

Ion rather is.wrong exercises.the 

‘Ion is rather wrong about the exercises.’ 

b. ( Ion) a cam greşit exerciţiile. 

Ion was rather wrong exercises.the 

‘Ion was rather wrong about the exercises.’ 

Given that it appears between the auxiliary and the participial verb, it can safely be 

considered a clitic adverb. 34 It is phrasal restriction adverb. It patterns with şi and both 

tot’s in that it cannot co-occur with negation in simple tenses, but quite paradoxically 

may marginally in compound tenses. 

34 In very colloquial Romanian, the order cam > Aux in (i) may marginally obtain. 

(i) ? Ion cam i-a bombardat cu întrebări. 

‘Ion rather bombarded them with questions.’ 

95

(127) a. * Ion nu cam ştie răspunsul la întrebări. 

Ion not rather knows answer.the to questions 

‘Ion didn’t really know the answer to the questions.’ 

b. ? (Ion) nu a cam ştiut răspunsul la întrebări. 35 

Ion not has rather known answer.the to questions 

‘Ion didn’t really know the answer to the questions.’ 

The fifth clitic adverb in this restricted class is prea. Consider the examples below: 

(128)a. Ion prea exagerează cu întrebările sale despre accident. 

Ion too.much exaggerates with questions.the his about accident 

‘Ion exagerates too much with his questions about the accident.’ 

b. Ion (prea) a ( ?* prea) exagerat cu întrebările sale despre accident. 

Ion too.much has exaggerated with questions his about accident 

‘Ion exaggerated too much with his questions about the accident.’ 

c. Ion nu prea pune întrebări despre accident. 

Ion not too.much/really asks questions about accident 

‘Ion doesn’t really ask question about the accident.’ 

d. Ion nu (prea) a ( ?* prea) pus întrebări despre accident. 36 

Ion not too.much/really has asked questions about accident 

‘Ion didn’t /really/ ask /too many/ questions about the accident.’ 

35 The adverb can move higher than negation in more colloquial Romanian, which gives a more marginal 

result, as in (i). However, such more colloquial uses are disregarded here. 

(i) ?? Ion cam nu a ştiut răspunsul la întrebări. 

36 The presence of prea between the auxiliary and the participial verb may sound however better in other 

contexts: 

(i) In meciul acesta, Federer nu a prea fost Federer. 

In match.the this, Federer not has really been Federer 

‘In this match, Federer hasn’t really been himself.’ 

96

What is striking about its behaviour is the fact that though patterning with mai in 

(positive and negative) simple tensed-sentences (a-c) confirming to the behaviour of any 

clitic adverb, it differs from a prototypical clitic adverb in that it necessarily precedes the 

auxiliary in (positive and negative) compound tenses (b-d). (b) resembles (125b) with tot 

continuative, which, recall, is an XP and seemingly has an XP status but unlike all other 

adverbs considered so far it can follow negation and precede the auxiliary in (d). The 

adverb functions as a degree adverb which takes scope over an event, be it a simple verb 

(a-c) or the Aux-Verb complex (b-d). It is under the scope of negation. 

One further observation regarding its behaviour is that all other adverbs follow a 

pronominal clitic, prea must precede it. This is illustrated below: 

(129) a. Prea îi bombardează cu întrebări. 

Really them bombards with questions 

‘He really bombards them with questions./He bombards them with too many 

questions.’ 

b. * Ii prea bombardează cu întrebări. 

To make a short summary of the ongoing discussion, it has been observed that tot has a 

dual status, it may either be a continuative aspect XP adverb or a clitic continuous aspect 

adverb, thus showing distinct properties with respect to auxiliary and negation. Şi, mai 

(except for the terminative reading requires a c-commanding negation) and continuous tot 

share the same properties, i.e. they follow the auxiliary and are incompatible with 

negation in whatever order. Cam differs from these adverbs in more or less marginally 

co-occurring with negation in compound tenses but paradoxically not in a simple tense. 

Prea has the least properties of a clitic, i.e. it is higher than the auxiliary and even a 

pronoun but is compatible with negation in that it must follow it (simply put, the order 

should be Neg > prea > Pron > Aux). 

In Minimalism, specifier positions are not automatically present but are created by 

an EPP feature. X° categories with no specifier positions are considered however 

maximal categories and consequently are both X° and XP categories. Given the 

observations above that such clitic adverbs have distinct meanings and belong to distinct 

97

aspect classes, it may be suggested that these are syntactic clitics which fill rigidly 

ordered head positions of Spec-less maximal categories. The sequence in (130) provides 

the fixed order of the adverbs. 37 

(130) … maiRepetitive > şi > totContinuative > maiTerminative > totContinuous > cam > prea … 

One more observation is in order here: the order of head adjunction actually reflects the 

mirror image of the aspectual heads in Cinque’s hierarchy. This is reminiscent of Baker’s 

Mirror Principle (1988). If negation and the auxiliary are present, the ordering is the one 

below: 

(131) … totContinuative > Neg > prea > Aux > maiTerminative > cam … 

8. Conclusions 

The hypothesis has been forwarded in this chapter that besides Case and phi-features, the 

Numeration contains features pertaining to Information Structure, such as Topic and 

Focus, both of distinct types, which are realized on Subj(ect) and Obj(ect) projections in 

the Mittelfeld. The movement possibilities of the DP subject and object, and the verb 

37 The ordering of clitic adverbs in Romanian is reminiscent of the ordering of adverbs in Hebrew event 

nominals (Ur Shlonsky, p.c.) Shlonsky (2004) argues that the nominalising head in Hebrew derived 

nominals lies below AspcelerativeII and this accounts for the presence of only some, i.e. lower, adverbs but not 

higher (than AspcelerativeII) adverbs inside such nominals. This is illustrated in (i): 

(i) a. * harisat ha cava ‘et ha kfar legamrey 

destruction the army Acc the village completely 

‘the complete destruction of the village by the army’ 

b. harisat ha cava ‘et ha kfar laxalutin 

destruction the army Acc the village completely Shlonsky 2004 (117) 

Legamrey belongs to AspcompletiveI whereas laxalutin to the lower AspcompletiveII and the nominalising head is 

in between. 

98

have been identified in three configurations, VSO, SVO and VOS by using the test with 

distinct adverb classes. 

Thus, in the VSO configuration, which is the unmarked word order in Romanian, 

the Mittelfeld contains several SubjPs. More precisely, the DP subject is probed and must 

move to the middle space of the Mittelfeld, i.e. to Spec SubjP above the Tanterior projection 

and can also occur in distinct positions interspersed among Tanterior and Modepistemic. In this 

respect Romanian departs from Italian, where in the VS configuration, the Mittelfeld 

exhibits New Information FocusP which is at the vP periphery. Spanish somehow 

patterns with Romanian in that the DP subject can, though only quite marginally, move to 

specifiers positions of SubjP in the fairly high portion of the Mittelfeld, i.e. till T(Past) 

but cannot move any higher. This is schematically represented below (only the landing 

sites are indicated): 

(132) Advepistemic … Adv … AdvT(Past) … AdvT(Anterior) … Adv … Adv 

Rom 

Sp. 

Verb movement possibilities above high adverbs in Romanian lend further support to the 

widely-acknowledged idea that the verb can target Fin°. However, in Spanish the verb 

can only very marginally move till T(Past). 

Regarding the object DP, it can get attracted to distinct specifier positions up to 

T(Past) both in Romanian and Spanish, as indicated below (as above, only the topmost 

landing position is indicated): 

(133) Advepistemic AdvT(Past) … Adv … Adv 

Rom, Sp. 

In the SVO configuration, more marked in Romanian and Italian, the DP subject must 

vacate the vP and move to a position above T(Anterior) and can further move to higher 

positions available in the upper portion of the Mittelfeld, but not above Moodspeech act or 

Mood evidential adverbs, which are parenthetical. The Italian and Spanish subjects are like 

99 

It.

Romanian but also have the option of moving above these adverbs (actually the Italian 

subject must move above mica). 

(134) Advspeech-act/evid. … Advepistemic AdvT(Past) … AdvT(Anterior) … Adv … 

It, Sp 

Concerning verb movement possibilities, the finite verb in Romanian can float to 

positions above aspectual and T adverbs and even higher up, i.e. above Modepistemic, 

whereas the finite verb in Italian must move above Voice adverbs and can move above 

Modepistemic. Similarly, in Spanish, the verb must vacate the vP but cannot move higher 

than Modepistemic. This is shown below: 

(135) Advspeech-act/evid. … Advepistemic AdvT(Past) … AdvT(Anterior) … Adv … 

The object DP in Romanian and Spanish can float to distinct specifiers of ObjPs, up to 

T(Past) but not any higher. In Italian DP object can similarly float to distinct specifier 

ObjPs up to the one dominating T (Past) adverb projection. 

The third configuration considered was VOS. Given that Romanian and Spanish are 

object clitic doubling languages no RM effects arise as the object is suggested to be a 

disguised PP. This captures the intuition that this configuration is freer in Romanian and 

Spanish but more restricted in Italian. Several interpretations are possible with both the 

subject and the object. In Romanian the DP object can also move quite high in the 

Mittelfeld, i.e. to a position above the Asphabitual adverb, and even (more) marginally 

above Modepistemic and in this case the subject fills either Spec SubjP associated with New 

Information Focus or is right-dislocated and interpreted as Topic. Similarly, in Spanish 

the object can float to positions in the middle portion of the Mittelfeld but lower than 

Modepistemic,. Interpretively, either the adverb itself or the subject is Focus, in this latter 

case the subject fills New Information Focus, as in Romanian. If the adverb bears a 

100 

Rom 

It 

Sp 

Rom

focusing reading, the DP subject is right-dislocated, an intonation break setting off the 

two elements. 

The finite verb in Romanian can target as position as high as Fin° and, crucially, 

must leave the vP domain. Spanish patterns with Romanian in that the verb moves above 

the ‘low’ Aspcompletive projection and the ‘middle’ Asphabitual projections, but as opposed to 

Romanian, the verb cannot move higher than Modepistemic. 

The suggestion has been made that subject and object DP movement possibilities 

are determined by IS-feature valuing inside the Mittelfeld. It is this mechanism that 

triggers vP vacating. In considering the three configurations the fine structure of the 

Romanian Mittelfeld has been arrived at on a comparative basis with other Romance 

languages. 

It has further been seen that not only the DP and the V can move in the Mittelfeld 

among the functional projections of the adverbs, but the adverbs can also move to the left 

periphery in one of the three positions available: Modifier, Focus and Topic (under the 

specific discourse situations), being thus licensed by the appropriate features on the 

respective head. The observation has been made that there is a restricted class of adverbs 

in Romanian that can occur only in the preverbal field. Neither verb movement across the 

adverbs, nor their preposing to the left periphery across the subject, is available. Though 

they may resemble clitic adverbs, they are not as they must precede a pronominal or the 

auxiliary. The special case of a five-member adverb class has been further focused on as 

these are truly clitic adverbs in that they typically occur in between the auxiliary and the 

participial verb. Upon closer insight, they do not constitute a homogenous class. Their 

ordering reflects the mirror image of Cinque’s aspectual classes. 

101

CHAPTER III 

THE LEFT PERIPHERY IN ROMANIAN: AN OVERVIEW 

1. Introduction 

This chapter provides an overview of the Romanian left periphery thus laying the ground for 

an extensive discussion of wh-movement in Chapter IV. 

Research on the nature of structural configurations as argued for in cartographic studies, 

such as Rizzi (1997, 2001a, 2004b), Poletto (2000), Rizzi 2004a (ed.), Beninca and Poletto 

(2004), a.o.; see also Munaro 2002, Pollock and Poletto 2004) has shown that the 

complementizer system is conceived of as a structural domain consisting of distinct functional 

heads and their projections. As argued for in Rizzi (op. cit.), the left periphery of the clause is 

a fixed component unfolding the optional Focus head, iterative Topic heads, etc. sandwiched 

in between two obligatory heads: Force, responsible for clause typing (declarative, 

interrogative, exclamative, etc.) and Fin(iteness), the locus of finite, non-finite, or mood 

specification. The map of the left periphery is provided in (1). 

(1) [ForceP [IntP [TopP* [FocP [Q [FinP …]]]]] Rizzi, class lectures 2007 

In this chapter, along these lines, I will briefly explore the structure of the Romanian left 

periphery. In so doing, I adopt three ideas presented in Rizzi (2005, 2006, 2007): 

103

(2) a. Syntactic movement to the left periphery is treated in terms of Criteria, i.e. the 

features on the heads creating a specifier position have well-determined specific 

interpretive import and function as scope markers for any existing quantificational 

phrases in a particular local domain. 

b. A Criterion can be satisfied either in a Spec-head configuration or in a head-head 

configuration: 

a. For [+F] a criterial feature, H+F is in a Spec-head configuration with A+F. 

b. For [+F] a criterial feature, H+F is locally c-commanded by A+F. 

[+F] = Foc, Q, Rel, Top, etc. 

Rizzi 1997, 2006, 2007; Rizzi & Shlonsky 2007 

c. Criterial Freezing: In a criterial configuration, the Criterial Goal is frozen in place. 

A relevant example of Criterial Freezing often given by Rizzi (2006, 2007) and Rizzi and 

Shlonsky (2007) and which may be reproduced for Romanian is provided in (3). 

(3) a. Ion, [_ Rel [nu e încă sigur [[câte traduceri după care] Wh [vor fi publicate tDP]]]] 

Ion, it is not yet sure how many translations by whom will be published 

b. Ion, [după care Rel [nu e înca sigur [[câte cărţi tPP] Wh [vor fi publicate tDP]]]] 

‘Ion, by whom it is not yet sure how many books will be published.’ 

c. * Ion, [[câte traduceri după care] Rel [nu e încă sigur [tDPWh[vor fi publicate tDP]]]]] 

’Ion, how many books by whom it is not sure will be published.’ 

The complex phrase câte traduceri după care ‘how many translations of whom’ cannot be 

pied-piped form the embedded CP to a higher relative C. Given the intermediate structure in 

(3a), the complex phrase satisfies the Criterion, i.e. the Wh-Criterion (or, more precisely, Q in 

the above references; I employ Q in a distinct sense in Chapter V and VI ) at the embedded 

level and gets frozen in place, as shown by the ungrammaticality of (3c). However, in (3b), 

the PP după care can get subextracted as it does not function as the Criterial Goal in the 

indirect question. 

104

2. Topics and Foci 

The existence of the two elements closing off the articulated CP layer manifests itself when an 

element sandwiched in between gets activated. Consider the following cases in Romanian. 

(4) a. Cred că georgienii vor semna acordul de pace. 

Believe.1.sg that Georgians.the will sign agreement.the of peace 

‘I believe that the Georgians will sign the peace agreement.’ 

b. Sper să semneze acordul curând. 

Hope1.sg să.Subj sign.3.Pl agreement.the soon 

’I hope they will sign an agreement soon.’ 

(5) a. * Cred acordul că georgienii îl vor semna. 

Believe.1.sg agreement.the that Georgians.the it.Cl will sign 

‘I believe that the Georgians will sign the peace agreement.’ 

b. Cred că acordul georgienii il vor semna curând. 

(6) a. Sper, ca acordul, să îl semneze curând. 1 

Hope1.sg that agreement.the să.Subj it.Cl sign soon 

’I hope that they will signthe agreement soon.’ 

b. * Sper să accordul semneze curând. 

The indicative complementizer că ’that’ closses off the CP layer, filling Force°, and preceding 

the Clitic Left Dislocated (henceforth CLLD) DP in (5b), whereas să, the marker of the 

subjunctive mood in Romanian, roughly the equivalent of the infinitival Italian di or the 

French de, fills the lower position closing off the CP layer, i.e. Fin°. That this is so is also 

shown by the fact that the topicalised DP must precede the mood marker (cf. 5b-6b). 2 Note 

1 The complementizer ca can be dropped and there is variation in this respect. There are speakers who cannot 

drop it in front of a Topic or Focus. Personally, I can easily drop ca with both, as in (7) and (i). 

(i) Sper UN ROMAN să îmi ofere (şi nu un volum de poezii). 

Hope.1sg A NOVEL să.Subj me offer.3sg (and not a volume of poetry) 

2 Infinitives have restricted use in modern Romanian and are replaced by subjunctives. 

105

that Romanian can lexicalise both the Force and the Fin heads, as illustrated by the example 

in (6a). (6) displays the lexical complementizer ca ‘that’, which whenever present, must co- 

occur with the subjunctive particle să. 3 The DP in between can function either as Topic or 

Contrastive Focus (see below for the property specific to Romanian that a clitic can, and 

sometimes, must co-occur with Contrastice Focus). However, ca can be dropped, as in (7). 

(7) Sper acordul să îl semneze curând georgienii. 

Hope1.sg agreement.the să.Subj it sign soon Georgians.the 

Though there is speaker variability with respect to the use of the exact conditions of 

occurrence of the lexical complementizer ca, it is generally agreed upon that in the absence of 

any material preceding the subjunctive particle, ca is preferably left out, as in (8) below: 

(8) ?? Sper ca să semneze curând un acord georgienii. 

As discussed in Rizzi (1997, 2001, 2004a,b), Topic, Focus and other types of elements may 

occur between Force and Fin. By Topic, a CLLD phrase is meant, expressive of the topic- 

comment articulation. The focus-presupposition articulation is expressed in many Romance 

languages by preposing the focused element and assigning it special stress. In Italian, 

Romanian and other Romance languages Focus in the left periphery is restricted to 

Constrastive Focus (Dobrovie-Sorin 1994, Alboiu 2002 for Romanian; Torrego 1984, 

Contreras 1991, Zubizaretta 1998, etc. for Spanish; Vallduví 1992 for Catalan). 

A number of properties distinguish CLLD from Focus: (i) whereas a resumptive clitic is 

obligatory with a CLLD object, it is impossible with Focus (but see discussion below on 

Romanian); topicalised adverbs form a distinct class and fill a Top position of their own; (ii) it 

does not give rise to Weak Crossover effects; (iii) it is not a quantificational structure; (iv) in 

Italian, Topics are recursive and are not strictly ordered with respect to Focus. The Romanian 

3 Note that the particle să has been the topic of much debate in the literature (Dobrovie-Sorin 1994) since it is 

considered to share the properties of a complementizer as well as of the inflectional head. It shares the properties 

of the latter since no XP material can intervene between să and the verbal cluster. It shares with the former the 

fact that it heads an embedded clause, it is invariable and is the leftmost element of the verbal cluster 

(să>Neg>ClPron>ClAdv>Aux>VPPast Part), hence Dobrovie-Sorin’s proposal that it is in C° and selects a Spec-less 

IP complement. In Cornilescu (1997, 2002), Isac (2001), Alboiu (2002), Barbosa (1995), să is the lexicalisation 

of Finiteness (or Mood). 

106

CLLD shares the same properties, as illustrated by the presence of the clitic and the binding 

relations in (9). 

(9) Pe Ioni, mama luii îli ocroteşte prea mult. 

’Ion, his mother protects him too much.’ 

Conforming to the structure in (1), in Romanian, too, Topic must precede a wh-operator in 

Focus, as in (10). 

(10) a. Lui Ion, despre ce i-ai vorbit? 

‘To Ion, what did you talk about?’ 

b. * Despre ce, lui Ion i-ai vorbit? 

A Topic must precede any other type of element in Focus, as illustrated in (11). Focus is in 

bold cases. 

(11) a. Cred că această melodie ASEARĂ am auzit-o. 

Believe1.sg, that this song, last evening have1.sg heard-it.Cl 

‘I believe that it is last evening that we heard this song.’ 

b. * Cred că ASEARĂ această melodie am auzit-o. 

However, as indicated by the ungrammaticality of (11b), a Topic phrase cannot follow Focus, 

Romanian differing in this respect from Italian. As pointed out in Chapter II, and as 

extensively discussed in the literature (Dobrovie-Sorin, Cornilescu, Alboiu, a.o.) in Romanian 

the Focus element must be adjacent to the verb. Topics may be recursive above Focus, as in 

(12), where a dislocated PP precedes a clitic-left dislocated DP. 

(12) Cred că, cu mare plăcere, această melodie, ASEARĂ am ascultat-o. 

Believe1.sg that with great pleasure this song LAST.EVENING have.1.sg listened-it 

‘I belive that it is last evening that I listened to this song with great pleasure.’ 

Barbosa (1998) observes that among Romance languages, Portuguese has English-type 

topicalisation, illustrated in (13), which co-exists with CLLD. 

107

(13) O teu livro, comprei de certeza. 

‘Your book, I bought for sure.’ Barbosa 1998 (8):6 

Romanian also has English-type Topicalisation. As shown in Cornilescu (2002), only bare 

NPs and weak quantifiers can undergo this type of Topicalisation, as illustrated in (14). In 

neither language is the moved element clitic resumed. Nor does it bear contrastive stress 

although a contrastive reading associated with contrastive stress is available, but in this case 

the NP and the QP would be Contrastive Focus (henceforth CF). 

(14) a. Ciocolată mereu le aduce copiilor. 

Chocolate always them.Cl brings children.Dat 

‘He always brings chocolate to the children.’ 

b. Mulţi studenţi, vinerea la ora 8, vezi la cursul de sintaxă comparată. 

Many students, on Fridays, at 8 o’clock see.2.sg to class.the of syntax 

comparative 

‘Many students on Fridays, at 8 o’clock, you can see in the comparative syntax 

class class.’ 

As shown in Cornilescu (2002), this type of Topicalisation has the properties of syntactic 

operators as it is sensitive to strong islands and may license parasitic gaps. 

(15) a. * Tort musafirii au plecat înainte de a mânca. 

Cake guests.the have left before to eat 

‘Cake the guest left before they could eat.’ Cornilescu (2002), (6) 

b. Ciocolată am cumpărat fără să plătesc prea mult. 

Chocolate have bought without to pay too much 

‘Chocolate I bought without paying too much.’ 

A topicalised phrase can precede a wh-phrase, as in (16a), adverbs and adverbials, as in (16b) 

and a parenthetical, as in (16c). 

(16) a. Mulţi studenţi, cine va avea la examen? 

Many students who will have in exam 

‘Who will have many students in the exam?’ 

108

. Mulţi studenţi, mâine dimineaţă n-ai să vezi la examen. 4 

Many students tomorrow morning not-are.2sg să see in exam 

‘You won’t see many students in the exam tomorrow morning.’ 

c. Mulţi studenţi, după părerea mea, n-ai să vezi la examen. 

Many students in my opinion not-are.2sg să see in exam 

‘In my opinion you won’t see many students in the exam.’ 

As the cases below show, a phrase displaced by the English-type Topicalisation can either 

precede or follow the CLLDed one: 

d. Ciocolată, copiilor mereu le aduce. 

Chocolate children.Dat always them.Cl brings 

e. Copiilor, ciocolată, mereu le aduce. 

More generally, operators, such as relative operators must precede Topics, as in (17), whereas 

a question operator, such as cui, must follow Topics, as in (18) (because of the adjacency with 

the verb). 

(17) a. Matematicianul căruia premiul Abel i-l vor acorda ... 

‘The mathematician to whom the Abel prize they will award…’ 

b. * Matematicianul, premiul Abel, căruia i-l vor acorda ... 

(18) a. * Cui, premiul Abel, vor decerna anul acesta? 

Who the prize Abel will give this year 

‘Who will they give the Abel prize this year?’ 

b. Premiul Abel, cui vor decerna anul acesta? 

4 The reading of (16) is ‘The students whom you will see tomorrow morning at the exam are not many’, where 

the quantifier is under the scope of nu. As extensively discussed in Dobrovie-Sorin, Cornilescu (2000), the use of 

the preposition pe forces the presence of the clitic in (i), which correlates with the reading ‘The students who you 

don’t see tomorrow morning at the exam are many’, under this latter interpretation the quantifier being outside 

the scope of negation. Simply put, this relates to the usual split between specific and non-specific readings (Heim 

1988, Diesing 1992). 

(i) Pe mulţi studenţi, mâine dimineaţă, n-ai să-i vezi la examen. 

Pe many students tomorrow morning not-are să-them.Cl see.2sg in the exam 

109

As argued in Cinque (1990), Rizzi (1997), a.o, movement to Focus has the properties of wh- 

movement: (i) Focus is quantificational; (ii) there are reconstruction possibilities; (iii) it 

involves WCO effects; (iv) successive cyclic movement is involved. A wh-phrase moves to 

Focus in the matrix clause. In (19a) and (19b) the wh-phrase ce ‘what’ and the focalised 

adverbial compete for the same position. By virtue of filling the same position Focus and Wh 

are adjacent to the verb. 

(19) a. * Ce ASEARĂ a ascultat Ion (nu azi dimineaţă)? 

What LAST.EVENING has listened.to Ion (not this morning) 

‘What Ion did listen to LAST.EVENING (not this morning)?’ 

b. * ASEARĂ ce a ascultat Ion (nu azi dimineaţă)? 

c. Ce a ascultat Ion aseară? 

One defining property of Focalisation in the Italian is the absence of the clitic, as in (20). 

Compare it to the Romanian (21)-(22). 

(20) GIANNI vedrò domani (, non Marcelo). 

GIANNI see1.sg tomorrow (, not Marcelo) 

‘It is GIANNI that I will see tomorrow (, not Marcelo).’ 

Romanian differs from Italian in that a clitic may be obligatory with a fronted Focus. The 

focalized person-denoting definite DP in (21a) forces the presence of the clitic, exactly as in 

the CLLD case in (21b). 

(21) a. PE ION *(îl) voi vedea mâine (, nu pe Andrei). 

PE ION him will see tomorrow (, not Andrei) 

b. Pe Ion, cu siguranţă, îl voi vedea mâine. 

Pe Ion, surely, him will see tomorrow 

110

Conversely, the indefinite in (22a) does not force the presence of a clitic. 5 Following 

Cornilescu (2002), bare NPs cannot be reanalysed as DPs, which accounts for the lack of a 

clitic, whether Focus, as in (22a) or Topic, as in (22b). However, both objects, pe Ion (21a) 

and the mass noun ciocolată (22a), share the same focus stress and verb adjacency. The 

suggestion will be made that they rely on two distinct syntactic structures. 

(22) a. CIOCOLATĂ le-am adus-(*o) copiilor (, nu fructe). 

CHOCOLATE them-have.1sg brought-it children.to (, not fruit) 

‘It is chocolate that I brought to the children (, not fruit).’ 

b. Ciocolată, Maria le aduce mereu copiilor. 

Chocolate, Maria them brings always to the children 

Consider another case of indefinite, (23) below, with a moved Accusative DP, either non PE- 

marked, as in (23a,a’) or PE-marked, as in (23b,b’). Example (23a) displays a well-known 

ambiguity characteristic of indefinites, i.e. it may have a specific reading or a non-specific 

reading within a given set of girls. (23b), where the indefinite is PE-marked and clitic 

doubled, has the specific reading only (or the D-linked reading; the two terms are taken to be 

equivalent here). 6 This latter interpretation is also available with Focalisation in (23a’) and 

(23b’). 

(23) a. O fată, am cunoscut ieri. 

A girl, I have met yesterday 

a’. O FATĂ am cunoscut ieri (, nu un băiat) 

‘A GIRL I met yesterday (, not a boy).’ 

b. Pe o fată, am cunoscut-o ieri. 

Pe a girl, have met-her yesterday 

b’. PE O FATĂ am cunoscut-o ieri (, nu pe un băiat) 

PE A GIRL I met-her yesterday (, not pe a boy) 

5 Indefinite direct objects, moved or in-situ, may optionally be preceded by PE and clitic-doubled. This correlates 

with the specific vs. non-specific reading in Romanian (see also Heim 1988 and Diesing 1992). No such 

interpretive ambiguity is available with indirect object indefinites. 

6 

Cornilescu (2000) also relates the specificity of indefinite PE DPs to D-linking. 

111

Though it is not my purpose here to indulge in a theory of doubling related to Left Dislocation 

or other constructions, I limit myself to nothing other contexts of movement to the Romanian 

left periphery where the clitic must or may occur. 

Certain quantifiers can also be Clitic doubled (see discussion in Cinque 1990). It is well- 

known that universal quantifiers are more easily left dislocated than existential and this has to 

do with their interpretation: they are more readily available as [+specific] especially that the 

(distributive) universal is marked by the accusative preposition PE in (24a) by virtue of being 

[+human]. It may be conceivable that specificity is encoded in syntax, as proposed by Diesing 

(1992), a.o. 7 The same doubling possibilities, in the accusative in (24a) and the dative in 

(24b), characterizing CLLD, also occur with focalisation (24a’, 24b’). 

(24) a. Pe fiecare mâine la prima oră *(îl) voi examina. 

Pe everybody tomorrow at first hour *(him) will.2sg examine 

‘Everybody, I will examine first thing tomorrow morning.’ 

a’. PE FIECARE *(îl) voi examina mâine (, nu pe toţi odată). 

EVERBODY I will examine *(him) (all together) 

b. Fiecăruia mâine *(îi) voi oferi un cadou. 

b’. FIECĂRUIA *(îi) voi oferi un cadou. 

TO EVERYBODY I will offer *(him) a present. 

7 Though Romanian does not have subject clitics (but see Cornilescu 1997; see also fn.12:22, Chapter II) but 

rather strong subject pronouns under precise conditions (see a thorough discussion in Hill 2006), the universal 

quantifier may marginally be doubled by a plural pronoun, whereas the existential can be doubled by a singular 

one, as illustrated by the contrasts in (i)-(ii). 

(i) a. ?(?) 

Fiecare, după părerea mea, vor contribui ei cu ceva. 

Everybody, in my opinion, will contribute them something 

b.* Fiecare, după părerea mea, va contribui el cu ceva. 

Everybody, in my opinion, will contribute him something 

(ii) a. Cineva, până la urmă, tot va veni el la mine. 

Somebody, in the end, will still come him over to me 

b.* Cineva, până la urmă, tot vor veni ei la mine. 

Somebody, in the end, will still come them over to me 

Besides the feature [+/-specific], the number feature also has a role to play in the doubling possibilities. 

112

The bare negative quantifier nimeni ‘nobody’ in (25a) which is not associated to a lexical 

restriction and which is necessarily preceded by PE (because of the [+human], [+pronominal] 

features) cannot function as Topic and be clitic doubled. Expectedly it allows focalization, as 

in (25a’). However, its Dative form may optionally be Clitic doubled (25b,b’) but the Dative 

form of the Clitic is irrelevant for the variable status of the trace/copy (see an extensive 

discussion on the variables and Case in quantification and CLLD in Romanian in Dobrovie- 

Sorin 1994). 8 

(25) a. Pe nimeni nu (*l)-am întâlnit in parc. 

I didn’t see anybody in the parc 

a’. PE NIMENI nu (*l)-am întâlnit in parc. 

NOBODY I saw in the parc 

b. Nimănui nu (i)-am vorbit zilele astea. 

Nobody.Dat have I been talking to (him) these days 

b’. NIMĂNUI nu (i)-am vorbit zilele astea (, nu unor poliţişti) 

NOBODY have I been talking to (him) these days (, not to any policemen) 

Weak NPs may optionally be preceded by PE and be clitic-doubled, this optionality having to 

do with the non-specific (26a) or specific reading (26b) in focalization. As discussed in 

relation to examples like (14)-(16), (26a) has an English-like Topic counterpart which is 

(26a’), associated with the absence of the clitic and functioning like a syntactic operator, 

whereas (26b) has the counterpart (26b’) which is a left-dislocated structure obligatorily 

doubled. 

(26) a. MULŢI TINERI (*i)-am văzut la teatru. 

MANY YOUNG.PEOPLE *(them)-have seen at theatre 

‘It is many young people that I saw at the theatre.’ 

8 As originally discussed in Dobrovie-Sorin (1994), non-Clitic doubled dislocated Accusative QPs (i.e. PE- 

marked) leave behind a variable and are necessarily associated with a [non-specific] reading (and have narrow 

scope). Conversely, if a Clitic is present with these dislocated Accusative QPs, they are [specific] (and have 

wide-scope reading). However, the optionality of the Clitic with indirect objects does not correlate with 

interpretive ambiguity. She relates this to the inherent Case which is relevant for NPs/DPs, but not for variables. 

113

a’. Mulţi tineri, din ce în ce mai des, (*ii) poţi vedea la teatru. 

Many young people, ever more often, one can see (*them) at the theatre 

b. PE MULŢI TINERI *(i)-am văzut la teatru (, nu pe puţini). 

b’. Pe mulţi tineri, din ce in ce mai des, *(îi) poti vedea la teatru. 

Many young people, ever more often, one can see *(them) at the theatre 

Romanian, a clitic doubling language, and Italian, a non-clitic doubling one, share the fact 

that Contrastive Focus is purely quantificational, but in addition the former presents cases 

where it includes a Topicality feature, hence the presence of an obligatory (or optional in the 

case of Dative) Clitic. Italian consistently distinguishes between Focus and TopCLLD in terms 

of syntax, prosody and interpretation. Although with respect to indefinites (the non-PE 

marked ones) and certain quantifiers Romanian patterns with Italian, we have seen evidence 

that Romanian differs in one major respect: it differs in that any definite DP and certain other 

types of quantifier(-like) elements fronted to the left periphery, and characterized by the 

prosody and interpretation of Contrastive Focus, must be resumed by a clitic. Below an 

analysis is suggested to account for a case like (27a). The DP bearing contrastive stress is an 

instance of Contrastive Focus and must be clitic resumed in Romanian, as in (27a) compared 

to (27b). If an indefinite DP (without PE) is Contrastive Focus, no clitic is involved, as shown 

by the contrast (27c) and (27d). 

(27) a. APARTAMENTUL l-am cumpărat anul trecut (, nu maşina). 

THE FLAT it-have bought last year (, not the car) 

b. * APARTAMENTUL am cumpărat anul trecut (, nu maşina). 

c. UN APARTAMENT am cumpărat anul trecut (, nu o maşină). 

A FLAT have bought last year (, not a car) 

d. * UN APARTAMENT l-am cumpărat anul trecut (, nu o maşină). 

I label the type of Contrastive Focus, which shares the properties of both CF and CLLD, Left- 

Dislocated Contrastive Focus (henceforth LDCF). Interpretively, LDCF shares with both 

CLLD and CF the fact that it must be ‘given’ in the context and the presence of a clitic is 

associated with D-linking (see 23b’ or 26b). Prosodically, it patterns with CF, and 

syntactically, with CLLD, due to the presence of the obligatory clitic and the fact that, like 

114

CF, LDCF must also be adjacent to the verb. 9 Therefore LDCF involves both a Topic-related 

feature and a CF feature. The account to suggest is similar to Rizzi’s (2001b) analysis of 

extraction of D-linked wh-phrases from wh-islands. I suggest that a minimal c-commanding 

head-head criterial configuration is involved with LDCF: the Contrastive Focus head bearing 

the Focus feature moves to the immediately c-commanding head, Topic (in the sense of 

CLLD) bearing the Top feature, and a complex head is formed. This is represented below: 

(28) Force 

3 

…TopCLLD 

3 

3 

Foc°+Top° Foc 

3 

Foc Fin 

… 

The resulting complex head, TopCLLD+Foc, attracts the DP apartamentul in (27) to its 

specifier and by Spec-head agreement the DP inherits the discourse properties of the complex 

feature head. In this way the interpretive properties, the focal stress and the presence of the 

clitic are accounted for and the Criterial Freezing in (2) is satisfied. Following the “big DP” 

hypothesis, the clitic is merged with the DP and is further attracted to and cliticises onto the 

raised verb, whereas the lexical DP part enters the derivation with an (uninterpretable valued, 

in Pesetsky and Torrego’s 2004 terms) CF feature and gives its value to the (interpretable 

unvalued) Foc by moving into Spec FocusP. 

This analysis captures the long-standing observation that Contrastive Focus in 

Romanian involves either a quantificational operator or a non-quantificational anaphoric 

operator (Comorovski 1996, Dobrovie-Sorin 1994, Alboiu 2002, a.o.) 

9 

Though it might be envisageable that this is Contrastive Topic (CT), it is not. Firstly, in Romanian the CT and 

CF have distinct stress and the example above does not have the stress characteristic of CT but that of CF (the 

stress associated to CT resembles more if not perfectly to that of Topic). Secondly, recall from Chapter II that 

one major property of CT is incompleteness in the accommodated or presupposed contrast, or the contrast 

involved in LDCF, similar to CF, is complete or exhaustive. It emerges quite clear therefore that LDCF involves 

both a Topic-related feature and a CF feature. 

115

Although a lot more restricted, the same construction also appears in Italian, as discussed by 

Bocci (2007) who provides a somewhat similar analysis for the following examples 10 : 

(29) A : Ha detto che il tappeto l’ha comprato l’anno scorso. 

[He] said that the carpet [he] it-bought last year. 

(…) 

B: Non, ti sbagli. Ha deto che LA POLTRONA l’ha comprato l’anno scorso (, non il 

tappeto)! 

No, you are wrong. [He] said that THE ARMCHAIR [he] it-bought last year (, not the 

carpet)! Bocci 2007(18a-c) 

As in Romanian, a definite DP object bearing contrastive focus must be clitic resumed. 

However, in Italian the occurrence of LDCF is more constrained, i.e. it occurs only in 

correction contexts where the alternative is explicitely given and it is subject to a definiteness 

requirement, indefinite or specific indefinites resist this construction. 

In what follows, emphasis is laid on another quantificational operator, the wh -operator. 

In order to account for the linear adjacency of the wh-operator and the inflected verbal 

element in English and Italian and, more generally, in Romance languages, Rizzi (1991) 

proposes the Wh-Criterion, stated in (30). 11 

(30) The Wh-Criterion 

A: A Wh-Operator must be in a Spec-head configuration with an X°[+wh] 

B: An X°[+wh] must be in a Spec-head configuration with a Wh-Operator. 

In (31a), the [+wh] feature is on I° and I°-to-C°, which is necessary in order to satisfy the Wh- 

Criterion, has not taken place, hence ungrammaticality is produced. 

10 

Bocci (2007) puts forth an analysis in terms of Top° movement to Foc° but, essentially, Top below Foc is 

associated with a definiteness feature in Italian. For independent reasons, Romanian does not exhibit a Top 

position below Focus. The only available Top position associated with definiteness or doubling is immediately 

above the Foc position. 

11 

Despite the pervasive role of AGREE in feature matching, the original Spec-head Wh-Criterion can be 

understood as more than AGREE, wh-feature valuing (matching) + EPP, such a Criterial configuration (following 

Rizzi 2007) being also related to specific interpretive properties. This is all the more obvious in French, where 

the interpretations of the moved wh seem to be different from those of the wh-in-situ (Mathieu 2004). 

116

(31) a. * Chi Gianni ha invitato? 

Who Gianni has invited 

In (31b), even if I°-to-C° has taken place, no grammaticality is obtained with the DP in Italian 

and the same holds true of other Romance languages. 

b. * Chi ha Gianni invitato? 

In order to account for the ungrammaticality of (32b), Rizzi suggests an independent 

explanation, in case-theoretic terms, based on Rizzi and Roberts’ (1989) analysis of French, 

where I°-to-C° destroys the Spec-head Nominative case assignment configuration, in (32a). 

The clitic in (32b) is allowed since it incorporates into the inflected verb in C and therefore 

does not rely on case assignment. 

(32) a. * Où est Marie allée? 

Where is Marie gone 

‘Where did Marie go?’ 

b. Où est-elle allée ? 

Regarding the general impossibility of the Romance subject DP to occur between the wh- 

phrase and the verb, it is argued in Cardinaletti (2007) that this is not a property of NSL and 

that it also occurs in non-NSL. She suggests that the restriction against subjects correlates 

instead with verb movement and that though the verb in Italian raises higher in interrogatives 

than in declaratives, it does not reach the C layer. Based on various empirical evidence 

(parentheticals, ‘high’ adverbs, specific pronoun uno ‘one’, etc.) she also shows that after all 

the wh-phrase and the verb do not make up the Spec-head of the same projection. Without 

entering into a discussion of the Italian data, it will be seen below that in Romanian the 

fronted wh-phrase moves to fill the Criterial position, i.e. Spec FocP, and the verb fills Fin 

(under the assumption that in Romanian declaratives and interrogatives the verb fills the same 

position; see for instance Cornilescu 1999, 2002). 

Consider first the distribution of parenthetical phrases in (33). 

(33) a. ?? Pe cine, după părerea ta, vor invita? 

117

Who in your opinion will.3Pl invite 

b. * Ce, după părerea ta, au semnat? 

What in your opinion did they sign 

As illlustrated in (33), a parenthetical is not generally allowed between the wh-phrase and the 

verb though there is speaker variability, i.e. some speakers may marginally accept a 

parenthetical. Even for these speakers, a parenthetical between ce and the verb is completely 

ungrammatical as opposed to a parenthetical between cine and the verb. Such a contrast, 

which is sharper in Italian (Cardinaletti 2007) and French (see a discussion in Shlonsky 2008) 

can be accounted for in terms of the weaker nature of ce. It will be seen in chapter IV that ce 

fills the lowest argumental position within the hierarchy of wh-phrases as it is the lightest 

element in point of feature content (i.e. besides being [wh], it is marked by default as [- 

Human]). 12 Importantly, data like (33) clearly suggest that a wh must be adjacent to the verb 

in Romanian. 

Consider next adverbs. The following examples confirm that nothing can intervene 

between the wh-phrase and the verb. 

(34) a. * Pe cine, din fericire, a chemat profesorul? 

Who happily has called the professor 

b. * Pe cine, probabil, a chemat profesorul? 

Who probably has called the professor 

c. ?* Pe cine poate va invita Maria? 

Who parhaps will invite Maria 

d. * Pe cine, repede, a îmbrăţişat Maria? 

Who quickly has kissed 

e. * Pe cine, de obicei invita Maria la ea acasa? 

12 The same contrast is preserved with modified cine ‘who’ and ce ‘what’ in (i), which further reinforces the idea 

that a wh-phrase in Romanian must be adjacent to the verb. 

(i) a. ?* Pe cine naiba, după părerea ta, vor invita? 

Who the hell in your opinion will (they) invite 

b. * Ce naiba, după părerea ta, au semnat? 

What the hell in your opinion did they sign 

118

Who usually invited Maria at her place 

f. * Pe cine deja a invitat Maria la ea acasă? 

Who already has invited Maria at her place 

g. * Cui, politicos, a înmânat premiul Maria? 

Whom politely has handed in the prize Maria? 

The data in (34) shows that whether an adverb belonging to the high portion of the Mittelfled 

is involved, or a lower one, the wh-phrase must remain adjacent to the verb. However, as 

already discussed in Chapter II, there are some aspectual adverbs which have restricted use 

tocmai ‘just’, (de-)abia ‘hardly’, tot ‘still’, i.e. they can only occur in preverbal position. 

Precisely these adverbs can break the adjacency, as shown in (35): 

(35) a. Pe cine tocmai a invitat Maria la ea acasă? 

Who just has invited Maria at her place 

b. Pe cine (de-)abia aşteaptă Maria la ea acasă? 

Who hardly waits Maria at her place 

c. (?) Pe cine tot a întrebat până la urmă ce s-a întâmplat aseară? 

Who still has asked in the end what had happened the evening before 

Due to their restricted use, these adverbs were suggested in Chapter II to exhibit clitic-like 

behaviour (see also Alboiu 2002). 

Though there is hardly any empirical evidence that the verb moves as high as Foc°, 

empirical evidence clearly suggests that it moves at least as high as Fin° (see also Chapter II). 

The suggestion to make is that the Wh-Criterion and V-movement are divorced. In Romanian 

the verb moves higher than it does in Italian, both in declaratives and in interrogatives (see 

discussion in Cardinaletti 2007). We suggest that the wh-phrase moves to the criterial 

position, Spec FocP, and the verb targets Fin. Essentially, FinP does not project a specifier. 

Thus, the wh-V adjacency has to do with verb movement, on the one hand, and the lack of 

any Top, Adv, etc. This is represented below. 

119

(36) …FocusP 

2 

2 

Foc FinP 

[+wh/Foc] 2 

Fin2 

… 

2 

wh 

V 

Romanian is known to present another type of wh-structure, one that allows Clitic doubling 

(see, for instance, extensive discussion in Dobrovie-Sorin 1994; see also Comorovski 1996; 

Cornilescu 2002). Thus, Romanian exhibits quantification wh-structures like those seen so far 

and wh-structures, headed by care ‘which’ which occur with an obligatorily Clitic, as in (38a) 

or an optional one, as in (38b). Compare the paradigm in (38) to (39) where a clitic is absent 

due to the purely quantificational structure of cine ‘who’ and ce ‘what’. 

(38) a. Pe care (gimnast) *(îl) cunoşti? 

‘Pe which (gymnast) do you know him?’ 

b. Cui (i) te-ai adresat până la urmă? 13 

‘Who did you finally talk to (him)?’ 

(39) a. Pe cine (*îl) cunoşti? 

‘Who do you know (*him)?’ 

b. Ce i-ai spus gimnastului până la urmă? 

‘What did you finally tell the gymnast?’ 

13 Another example of optionality of wh-quantifier is given and discussed in Dobrovie-Sorin: 

(i) a. Câţi studenţi ai examinat? 

‘How many students did you examine?’ 

b. Pe câţi studenti i-ai examinat? 

‘Pe how many students did you examine them?’ Dobrobvie-Sorin (1994), (20):208 

Whereas the Accusative wh-phrase in (ia) is purely quantificational, the PE-marked one in (ib) is not. This 

contrast correlates with a distinct interpretation, i.e. câţi studenţi in the former case is [-specific]/[-D-linked] (it 

involves a quantitative reading), the PE-marked constituent in the latter is [+specific]/[+D-linked] (see also 

Rizzi 2001b). 

120

Doubled wh-phrases do not give rise to WCO effects, nor do they license parasitic gaps. 

Interpretively, they correlate with D-linking and such care wh-phrases are actually inherently 

D-linked. 14 Recall that it was seen with examples like (24) above that a Clitic is optional with 

a Dative QP and that no [specific] reading was registered. Thus, with the Dative form cui in 

(38b) the clitic optional and no such reading obtains. 

As opposed to bare wh-phrases, D-linked care wh-phrases need not be V-adjacent, 

which means that they fill a higher position than the former (see Chapter IV for a detailed 

discussion and analysis). 

To summarize so far, the Wh/Foc-V adjacency in Romanian is suggested to be 

accounted for in terms movement of wh/Foc to the criterial position and of the verb to Fin. 

FinP does not project a specifier, nor are there any other available projections in Romanian. 

Under the cartographic approach the structure of the left periphery in Romanian is given in 

(40). 

(40) …TopCLLD/Enlish-like > TopCLLD > Foc > Fin … 

14 According to Comorovski (2005), the Romanian care in its bare form, i.e. without a following N, may also be 

non-D-linked or weakly D-linked, such cases arising in one particular context, namely copular questions, which 

she analyses as a predicate nominal as below (her examples 4a and 5c). 

(i) Care e capitala Moldovei? 

What is capital.the Moldova.Gen 

‘Which is the capital of Moldova?’ 

(ii) Care e singurul naist român cunoscut în străinătate? 

What is only.the pan-piper Romanian known abroad 

‘Which is the only Romanian pan-piper known abroad?’ 

I abstract away from these complications and stick to the bare vs. D-linked distinction. 

121

3. The Cartographic Approach vs. the Generalised TP Approach 

The discussion so far has shown that Romanian fits well into a cartographic approach and that 

its specific properties can be put down to the ways of realization of the Criteria in the left 

periphery: movement of wh/Foc-element to the Criterial position, and Foc-to-Top (followed 

by Spec-head agreement). However, an article by Barbosa (2001) (see also Uribe-Etxebarria 

1991, Contreras 1991, Zubizarreta 1998 on Iberian Spanish, Vallduví 1992 on Catalan, 

Motapanyane 1998, Hill 2002, 15 Alboiu 2002, Cornilescu 2002 16 on Romanian, among many 

others) have adopted a generalised TP approach – presupposing a certain amount of feature 

syncretism on the T head, which can value/check both head-related features and operator 

features like [Foc] and [wh] - and have shown that a cartographic approach may not be 

tenable for Romance languages as opposed to English. More precisely, Barbosa argues that 

the obligatory adjacency between the wh-phrase and inflection in Romance cannot be 

assimilated to English Subject-Aux inversion. She proposes that inside the Romance group, 

there are languages like Iberian Spanish, Catalan, Portuguese and Romanian which have only 

one operator position, namely Spec TP, and languages like Italian, which have two operator 

positions, Spec TP and a higher left peripheral Spec FocusP. 

One of their arguments for postulating that the Inflection does not raise to C in 

Germanic languages and that consequently the Wh-Criterion is satisfied in TP is the absence 

of matrix-embedded asymmetry. This is illustrated for Romanian below: 

15 Under a multiple specifier analysis, Hill (2002) argues that the [Focus] feature enters the grammar by merging 

with the T head as long as the T also has Tense and Case features. 

16 Cornilescu (2002) adopts a multiple specifier generalized TP approach and argues that the T head may 

check/value both Case and operator features, such as [Focus], [wh], etc. Thus, Spec TP has an optionally strong 

D feature, realized as [Cl+phi] features, and thus dislocated constituents get attracted (and may move higher up), 

or realized as [+D, -phi], in which case bare QPs are attracted to this position. Besides Spec TP, which is 

considered to be an A’ position, the lowest position of the Romanian left periphery is argued to consist of (at 

least) one more projection, which she identifies as Top, whose head check/value the [N]-feature. Spec Top 

accommodates therefore NPs (weak NPs and bare NPs). In a nutshell, Spec Top hosts constituants that have a 

presuppositional reading, whereas Spec TP is associated with non-presuppositional readings. 

I believe that one needs also account for the case of CLLD which is necessarily [+D] and fills Spec TopP in the 

left periphery. This means that it is not only the T head which (optionally) has this feature but so does the Top 

head. 

122

(41) a. Pe cine va invita Maria? 

Pe who will invite Maria 

‘Who will Maria invite?’ 

b. * Pe cine Maria va invita? 

(42) a. Nu ştiu pe cine va invita Maria. 

Not know pe who will invite Maria 

b. * Nu ştiu pe cine Maria va invita. 

The lack of root-embedded asymmetry with regard to the adjacency of the moved wh-element 

and the verb has led the above-mentioned authors to suggest that Spec TP (in their analysis) is 

an A’ position in NSLs. However, as was seen in Chapter II, Romanian avails itself of several 

subject preverbal (as well as postverbal) positions interspersed among distinct adverb 

projections and it was also shown that the highest such position – situated above the adverb 

probabil ‘probably’ - fills Rizzi’s Subject Criterion, the DP subject being necessarily 

associated with a Topic-aboutness feature. Cardinaletti (1997, 2004) provides a fine-grained 

hierarchy of structural preverbal subject positions in Italian, the highest subject position being 

the Criterial position, which check the Topic-aboutness feature, whereas the lower Spec TP 

hosts pronominal subjects which only check/value phi-features. Given this proliferation of 

preverbal subject positions, each associated with a discrete A-related and/or special 

interpretive/informational features (Topic-aboutness, phi-feature valuation, etc.) it would then 

be hard to understand why the highest Spec position of the Mittelfeld (their Spec TP) should 

host a purely A’ position while all other preverbal positions are purely A positions. This is a 

language-internal argument against the idea that Spec SubjP is a left-dislocated A’ position. 

Let us consider then some language-external arguments. 

One such case comes from the so-called Doubly Filled Comp Filter, which is a point of cross- 

linguistic variation. As (43) shows, the Dutch overt complementizer of may have a wh-phrase 

in its specifier. 

(43) Ik weet niet wie of Jan gezien heeft. 

I know not who C Jan seen has 

‘I don’t know who Jan saw.’ 

123

Couching this filter in cartographic terms, it may be suggested that the complementier fills Fin 

and the wh-subject moves to Spec Foc. What is of interest here is that the wh-phrase 

undoubtedly occupies a left-peripheral position. 

Some languages like Lele, a Chadic language, and Tlingit, a Na-Dene language, show 

morphological expression of Focus under the form of particles (see Chapter V and references 

therein). This is illustrated in (44) and (45). 

(44) Me ba gol di gà? 

What Foc see 3sgM Int 

‘What did he see?’ Aboh 2007 (18c) 

(45) Daa sáwé i éesh al’óon? 

What Q.foc.part your father he.hunts.it 

‘What is your father hunting?’ Dauenhauer & Dauenhauer 2000 

This indicates that the wh-object in (44) and in (45) fill the dedicated left-peripheral position. 

This is a strong argument that goes in the direction of postulating uniformity of whmovement 

to the left periphery of the clause rather than movement to Spec TP in some cases 

(say, Romanian (41)) and to the left periphery in some others (Lele (44) and Tlingit (45)). 

For Barbosa (2001), the preverbal subject is left-dislocated or Topic and this would 

explain the ungrammaticality below: 

(46)*Pe cine Ion a invitat? 

Besides left-dislocated elements precede wh-phrases, as in (47), and thus (46) and (47b) are 

ruled out on similar grounds. 

(47) a. Pe Ion când l-ai invitat? 

b.* Cand pe Ion l-ai invitat? 

Such an analysis is not tenable on the grounds seen above, i.e. violation of the Criterial 

configuration. 

124

Though it is uncontroversial that the DP subject Ion in (48) is left-dislocated, and that (48) 

and (47) are the same structure, the left-dislocation view fails to account for other facts. A 

case in point is the paradigm with the negative quantifier in (49). 

(48) Ion, pe cine a invitat? 

‘Ion, who did he invite?’ 

(49) a. Nimeni nu l-a invitat pe director. 

‘Nobody invites the director.’ 

b. Nimeni nu l-a invitat pe director? 

‘Did nobody invite the director?’ 

c. * Nimeni, pe cine nu a invitat? 

‘Who did nobody invite?’ 

The left-dislocation analysis fails to account for the contrast between (49a,b) and (49c). The 

negative quantifier can occur in the preverbal position in both declaratives and yes-no 

questions but cannot precede a wh-phrase. According to the approach adopted so far, the 

subject in (48) and (49c) is left-dislocated. (49c) is ungrammatical because the negative 

quantifier cannot be left-dislocated (Cinque 1990, Rizzi 1997). In (49a,b) the subject occurs in 

the preverbal, non-dislocated, subject position, Spec SubjP. 

Here are other arguments that the preverbal subject does not fill Topic. 

Firstly, as observed by Alboiu (2000), bare nouns are able to appear as postverbal 

subjects, as in (50a), but not as preverbal subject, as in (50b). 

(50) a. Cântă privighetori în mai toate parcurile de aici. 

Sing nightingales in almost all parcs here 

b.* Privighetori cântă în mai toate parcurile de aici. 

As discussed above, Romanian has English-like Topicalisation, in which exactly this type of 

bare nouns appears, as in (51). It is therefore unlikely that preverbal subjects and Topics share 

the same position. 

(51) Privighetori, mai toată ziua, bătrânul stă şi ascultă în parc. 

Nightingales, almost all day long the old man is sitting down listening to in the parc 

125

Secondly, within Rizzi’s system, a Topic may precede a wh-phrase in Focus. Example (52a) 

clearly shows that a negative quantifier cannot occur in Spec TopP. Thus, (52b) is unexpected 

under the Top-analysis of preverbal subjects. 

(52) a. * Nimeni ce veste nu a auzit? 

Nobody what news not has heard 

‘What news did no one hear?’ 

b. Ce veste nimeni nu a auzit? 

This contrast, again, strongly suggests that preverbal subjects are not in Topic position. 

Thridly, another contrast between Topics and preverbal subjects is that the former but 

not the latter create an island for wh-movement. This is illustrated below: 

(53) a. ?* Cui crezi că premiul l-a decernat directorul? 

Who believe.2sg that prize.the it-has awarded director.the 

‘Who do you think that the prize the director awarded to?’ 

b. Cui crezi că directorul a decernat premiul? 

In (53a), where the subject DP, directorul, is postverbal and does not interfere with the 

movement of the wh-phrase, Top in the left periphery of the embedded clause blocks 

movement of cui ‘to whom’ to the matrix clause. Either an RM effect under feature sharing 

(i.e. of the A’-feature) may be invoked, or the barrierhood of Top may be given as an 

argument of the ungrammaticality of (53a). Conversely, in (53b) there is no Topic (in the left 

periphery) and the subject is preverbal, therefore not in a A’ position. 17 

17 

Similar effects obtain with French (Christopher Laenzlinger, p.c.). As (i) shows, movement of the wh across 

the embedded Top results in a high degree of marginality. The same explanation holds true of French. No 

barrierhood is involved in (ii). 

(i)??/*A qui crois-tu que ce prix on l’a donné? 

Who do you think that this prize they gave it to 

(ii) Je me demande à qui, ce prix, tu l’as donné. 

I wonder who, this prize, you gave it to 

126

These are all arguments that the Topic and the preverbal subject are in distinct positions. 

More precisely, the data above show that the preverbal subject is in a position lower than the 

Topic, i.e. Spec SubjP. 

Fourthly, facts like (54) have been taken to suggest that the preverbal subject and the 

wh-phrase compete for the same position. Consider (54) and (55) below, where Focus and wh 

create islands for further wh-movement though the effect is slightly less ungrammatical with a 

wh. 

(54)* Cui crezi că UN PREMIU a decernat directorul festivalului? 

Who believe.2sg that A PRIZE has awarded director.the festival.Gen 

‘Who do you think that A PRIZE the director of the festival awarded?’ 

(55) ?? Cui vrei să ştii ce premiu a decernat directorul festivalului? 

Who want.2sg sa know what prize has awarded director.the festival.Gen 

‘Who do you want to know what prize the director of the festival awarded?’ 

A violation of Rizzi’s (2004b) Relativised Minimality can account for these facts. As seen in 

(53b) above, unlike a Foc/wh-element, the subject does not create islands. This strongly 

suggests that preverbal subjects fill a position lower than Spec FocP. 

Fifthly, as argued in Chapter II, the Romanian preverbal subject fills the Criterial 

Subject position. That this is so is shown by the subextraction possibilities in (57) below: 

(56) Traducerea (ta) după poem a câştigat premiul cel mare. 

Translation.the (yours) of poem has won prize.the the big 

‘(Your) translation of the poem won the big prize.’ 

(57) a. Poemul, după care cred că [traducerea (ta) _] Subj° a caştigat 

premiul cel mare, … 

Poem.the of which believe.1sg that translation (your) has won 

prize.the the big 

b. * Poemul, [traducerea (ta) dupa care] cred că [[_] Subj° a caştigat 

premiul cel mare], … 

c. ?(?) Poemul, [traducerea (ta) după care] Subj° a caştigat premiul cel mare], e acum 

foarte cunoscut. 

127

Poem.the translation.the (yours) of which has won prize.the the big is now very 

well-known. 

On the assumption that in (57) the phrase traducerea (ta) după poem fills the Criterial Spec 

SubjP, where it is attracted to check its [N] and [Topic-aboutness] features, wh-subextraction 

in (57a) is allowed as it is the subpart marked as [+N], traducerea (ta), which carries the 

criterial feature and thus satisfies the Subject Criterion. Conversely, movement of the whole 

constituent violates the Criterion in (57b). 

Let us go back to examples (41)-(42), repeated below. Contrast (42) to (58). The latter 

example is used in the generalised TP approach as an argument to further reinforce the idea 

that a DP may incorporate a [Foc] feature at various locations and consequently that there is 

hardly a dedicated FocusP for languages like Romanian (and Spanish). 

(41) a. Pe cine va invita Maria? 

Pe who will invite Maria 

‘Who will Maria invite?’ 

b. * Pe cine Maria va invita? 

(42) a. Nu ştiu pe cine va invita Maria. 

Not know pe who will invite Maria 

b. * Nu şitu pe cine Maria va invita. 

(58) a. Nu ştiu DIRECTORUL ce le va explica angajaţilor (, nu şeful 

executiv.) 18 

18 Rizzi (1997) observes that in embedded contexts a wh-element is marginally compatible with a focalized 

element, as in (i). Moreover, this is possible if the focalised element is a PP and the wh is a DP. 

(i) a. Mi domando A GIANNI che cosa abbiano detto (non a Piero). 

‘I wonder TO GIANNI what they have said (not to Piero).’ 

b.*? Mi domando che cosa A GIANNI abbiano detto (non a Piero). 

It is to note that in Romanian no such categorical distinction is needed, as exemplified in (ii). 

(ii) a. Mă intreb ION pe cine a văzut la concert (, nu Maria). 

‘I wonder ION who he saw at the concert (, not Maria).’ 

128

Not kwow.1sg DIRECTOR.THE what them will explain employees.Dat (, not the 

executive-in-chief) 

b. * Nu ştiu ce DIRECTORUL le va explica angajaţilor (, nu şeful executiv). 

Following Cardinaletti (2007), the contrast in (41a,b) has to do with verb movement and the 

impossibility of certain types of subjects in Italian, but not in Romanian, to occur between the 

wh and the verb. 19 This strict adjacency characterising both matrix (41) and embedded 

contexts (42) has been accounted for in terms of a particular Criterial configuration, a 

property specific to Romanian (and possibly Spanish, too). 

It remains to account for (58). The generalised TP approach, where Spec IP is an A’ 

position hosting the wh-phrase, would have two options to get round: (i) either it is forced to 

admit that Focus is part of the left periphery and in such case it runs into more conceptual 

problems, for instance, why both a projection, part of the CP layer, and another, part of the TP 

functional architecture, are quantificational in nature (recall that the [Focus] feature is on T°). 

In other words, under such an analysis [Focus] is a parasitic feature; (ii) or if Focus does not 

project, it is not clear why only certain heads may optionally have a Focus feature and under 

which precise circumstances this takes place. 

One suggestion to make for (58) is that embedded contexts are related to selection and 

involve a mechanism of fission (Ur Shlonsky, p.c.), i.e. Focus splits itself into Focus and Wh 

(more precisely, FocWh) so that the latter can attract the wh-element, ce ‘what’. Alternatively, 

the order Foc > wh in embedded contexts may be attributed to feature precompilation. 

Romanian and Italian behave alike in this respect (see Shlonsky 2006). However, Romanian 

presents one more complication in the sense that even in these embedded contexts the wh-V 

adjacency is required, whereas in Italian a subject may follow the wh-phrase. This particular 

order may come from the fact that Romanian makes use of selection for the Q-feature (on 

Force; see also Chapter VI on feature distribution) coupled with movement to the lower 

criterial FocWh, which explains the wh-V adjacency even in embedded contexts (the same 

mechanism would be involved in Spanish as well). 

b. Aş vrea să ştiu PE ION ce îl vor intreba la examen (, nu pe Maria). 

‘I would like to know ION what they will ask him in the exam (, not Maria).’ 

19 

Cardinaletti (2007) shows that in Italian the restriction against subjects to appear in between the wh and the 

verb affects only the highest subjects in the hierarchy of Cardinaletti (1997, 2004), namely full DPs and strong 

pronouns. 

129

Evidence has been given so far that a preverbal subject may fill Spec of the Criterial Subject 

position where it is necessarily interpreted as Topic-aboutness. This position has been shown 

to be distinct from the left-dislocated Topic position and is identified to be Rizzi’s criterial 

Spec SubjP. It has further been seen that a wh-phrase fills the dedicated Spec FocP in the left 

periphery. For all these reasons, which among others, can easily account for strict orderings, 

the generalized TP approach can be done away with. 

4. Rounding off the Picture of the Romanian Left Periphery 

Taking into account the discussion so far, one arrives at the following (preliminary) structure 

of the Romanian left periphery in embedded clauses: 

(59) Force Top*CLLD/English-like Top*CLLD Focus FocusWh Fin (preliminary) 

130


CHAPTER IV 

WH-MOVEMENT IN ROMANIAN 

Romanian is a multiple wh-fronting language and this chapter will propose an analysis of 

wh-movement in Romanian. 

2. Previous Approaches to Multiple Wh-Movement 

We begin by reviewing the place Romanian occupies among other so-called multiple whfronting 

languages. 

It is standardly assumed that there are several language types with respect to 

possibilities for wh-fronting in multiple questions. There is (i) the English-type, which 

moves only one wh-phrase, the other(s) remaining in situ, including what is called French 

on the movement strategy in (1); (ii) the Chinese type illustrated in (2), where all whphrases 

(arguably) remain in situ, Hindi, Japanese, French on the in-situ strategy, being 

some other exemplifications; and (iii) multiple wh-fronting (MWF) languages like Slavic 

languages and Romanian, where all wh-phrases move to clause-initial position. Within the 

group of MWF languages, Richards (2001), drawing on Rudin (1988), makes a distinction 

between the so-called IP-absorption languages 1 like Serbo-Croatian, Polish and Czech, 

exemplified in (3), and CP-absorption languages like Romanian and Bulgarian, illustrated 

in (4). The difference between the two groups resides in movement of all wh-phrases to 

1 

Rudin (1988)’s original terminology distinguishes [-Multiply Filled Spec] languages, like Serbo-Croatian, 

Polish and Czech, from [ + Multiply Filled Spec] languages, like Bulgarian and Romanian (see section 2.1). 

131

multiple specifiers of CP for the latter group, and movement to one or more IP projections 

for the former group. 

(1) a. Who gave the book to whom? 

b. Qui a donné le livre à qui? French on the movement strategy 

‘Who gave the book to whom?’ 

(2) a. Dare-ga nani-o tabeta no? Japanese 

Who-Nom what-Acc ate Q 

‘Who ate what?’ 

b. John gei-le shei shenme? Chinese 

John give-Perf. who what 

‘What did John give to who?’ 

c. Il a donné quoi à qui? French on the in-situ strategy 

He gave what to whom 

‘What did he give to whom?’ 

(3) a. Ko koga/Koga ko voli? Serbo-Croatian 

Who whom loves 

'Who loves whom?' 

b. Kto jak/Jak kto zareagował na nowosci z Kijawa? Polish 

Who how reacted on news from Kiev 

‘Who reacted how to the news from Kiev?’ 

c. Kdo koho/Koho kdo vybral do dalšiho kola? Czech 

Who who.Acc/ Who.Acc who chose to next round 

‘Who elected whom for the next round?’ 

(4) a. Koj kakvo dade? Bulgarian 

Who what gave 

‘Who gave what?’ 

b. Cine ce a scris? Romanian 

Who what has written 

‘Who wrote what?’ 

132

This chapter focuses on wh-movement in Romanian and Bulgarian, insisting though on the 

former. On the basis of Krapova & Cinque’s (2005) hypothesis of order parallelism 

between the left periphery and the Mittelfeld, an alternative derivational device will be 

suggested for Romanian, consisting of MULTIPLE AGREE followed by movement of a 

clausal sub-structure, called “chunk”, containing only wh-elements. 

In the following section, an overview of the major analyses to multiple whmovement 

proposed in the literature is provided. 

2.1. The Adjunction Analysis 

In her 1998 seminal paper, Rudin makes a split among multiple wh-fronting languages, 

showing that based on several criteria, such as the position of parentheticals among whphrases, 

Superiority effects and clitic insertion, Bulgarian and Romanian pattern alike, 

whereas languages like Serbo-Croatian, Polish and Czech form a distinct group. Adopting 

Aloun et al.’s (1987) account of ECP, she proposes an analysis in terms of right-adjunction 

to multiple specifiers of CP to account for the behaviour of wh-movement for the former 

group. Regarding the latter group, she argues in favour of movement of the highest whelement 

to Spec CP, whereas the other(s) occupy an IP-adjoined position(s). The relevant 

representation for the former case is provided below. 

(5) CP 

3 

Spec IP 

2 5 

Spec wh dade 

2 kakvo 

Spec wh 

kogo 

wh 

koj 

whoNom whomDat what gave 

In Rudin’s analysis, the order of the fronted wh-elements is obtained by successive rightadjunction 

to the specifier of CP (whence Multiple filled Spec languages). 

As for the other group of languages (IP-absorption languages), one wh-element 

moves to Spec CP (a unique constituent), while the other wh-phrases freely left-adjoin to 

IP (i.e. in any order). Contrary to Romanian and Bulgarian, the first wh-element can be 

133

separated from the other wh-phrases by clitics and parentheticals, hence the division 

between CP- and IP- attachment, as represented in (6) for Czech. 

(6) CP 

3 

kdo C’ 

who+Nom 3 

C IP 

3 

(clitic + parenth)kdy IP 

when 3 

koho …pozval 

whom invited 

The three wh-elements in (66) can be rearranged in any order. 

For independent empirical and conceptual reasons, an ECP account of Superiority effects 

is obsolete in today’s theoretical thinking. Besides if ECP had the explanatory power to 

account for the ordering restrictions in Romanian and Bulgarian, it failed to accout for the 

trigger of overt movement of multiple wh-phrases in these languages and overt movement 

of one wh-phrase in English. 

2.2. The Wh-Cluster Analysis 

Another line of approach is that of cluster formation. Thus, Grewendorf (2001) (see also 

Sabel 2001, Grewendorf and Sabel 1999) argues that the three wh-elements in example (5) 

above form a cluster by adjunction of one to the other(s), such movement being driven for 

feature checking purposes. In keeping with Cheng (1991) he assumes that the formation of 

wh-clusters is morphologically driven in the sense that the quantificational force of a wh- 

phrase is provided by a null determiner that bears a wh-feature ([DP[DØwh[NP wh-word]]). 

He proposes the Wh-Cluster Hypothesis in (7), according to which the direct object kakvo 

right-adjoins to the indirect objet kogo, and the resulting cluster thus formed further right- 

adjoins to the subject koj in Spec IP. The wh-cluster moves as one constituent to Spec CP, 

the target position for wh-movement, as represented in (8). 

(7) Wh-Cluster Hypothesis (Grewendorf 2001:94) 

A particular feature of a wh-element acts as a checker for other wh-elements 

(wh-arguments as well as wh-adjuncts) 

134

(8) CP 

3 

C’ 

3 

C[Q] IP 

3 

koj I’ 

2 3 

koj kogoj I VP 

2 5 

kogo kakvok tj tk 

More technically, Grewendorf (2001) proposes that the uninterpretable Q-feature on the 

indirect object wh-phrase attracts the direct object wh-phrase which right-adjoins to the 

former, thus forming a cluster. Q-feature matching results in deletion of the uninterpetable 

Q-feature on the probe and in the unintepretable wh-feature on the goal. Next, the 

interpretable Q-feature on the newly-formed cluster matches the corresponding 

uninterpretable Q-feature of the subject wh-phrase resulting in the Q-feature deletion of koj 

and in the uninterpretable wh-feature deletion on the goal. The cluster is thus right- 

adjoined to the wh-subject. Finally, the uninterpretabe Q feature of C matches the 

interpretable Q feature on the three-element cluster which thus moves to Spec CP. 2 As in 

Rudin (1988), the order of the wh-elements depends on the successive operations of (right) 

adjunction of one wh-phrase to the other. 

Such an analysis fails to capture the ‘right’ order and seems to involve a 

parametrisation, i.e. right-adjunction vs. left-adjunction. It is not very clear what would 

prohibit left-adjunction (one would expect to find a language which is the mirror image of 

Romanian and Bulgarian). 3 Besides, an analysis in terms of right-adjunction poses 

problems for the antisymmetric approach (Kayne 1994). In addition to this, contrary to 

standard assumptions of feature distribution the D-head of a wh-phrase bears both an 

interpretable and an uninterpretable Q-feature (see however Grewendorf for an 

explanation). 

2 Without entering into details here, he assumes that there are two exceptions to the wh-cluster formation, 

wh-phrases in Spec CP and wh-adjuncts are shown not to be adjoinable to. 

3 I thank Ur Shlonsky for this comment. 

135

2.3. The multiple Spec analysis 

Within the Minimalist framework, multiple wh-fronting is often analysed in terms of 

multiple specifiers. Such an approach may branch into multiple Spec IP and the multiple 

Spec CP analysis, respectively. 

2.3.1. The multiple Spec IP analysis 

As far as Romanian is concerned, some authors (Alboiu 2002, Motapanyane 1998, 

Cornilescu 1999) propose that, once the first wh-phrase (e.g. the subject) has moved and 

merged as Spec IP (or Spec MoodP in Cornilescu’s analysis), the other wh-phrases (objects 

or adjuncts) tuck in underneath this ‘outer’ specifier. The “tucking in” process of merging 

specifiers is taken to account for superiority effects (the order Subj > Obj(s)). The structure 

in (9) shows that the wh-subject cine first merges as the outer specifier of IP, and then the 

two wh-objects tuck in any order. 

(9) IP 

3 

Whsuj I’ 

3 

WhIO/DO I’ 

3 

Wh DO/IO I’ 

3 

I 6 

Cine pe cine cui a prezentat 

cui pe cine 

‘who whomacc/dat whomdat/acc has introduced’ 

As discussed in Chapter III, the IP projection in Romanian is argued to be a discourserelated 

projection and the I head is argued to be a syncretic head hosting a wh-feature 

(Alboiu 2002 and Cornilescu 1999, 2002). I follow the idea that overtly moved whelements 

target the specifier of FocP, following the structure of the Romanian left 

periphery provided in Chapter III, sections 2 and 3. 

136

2.3.2. The multiple Spec CP analysis 

The multiple Spec CP analysis is proposed by Bošković (1998, 2002), Richards (1997, 

2001), Pesetsky (2000), etc. Bošković (1998, 2002) argues that in multiple wh-fronting 

languages, the first moved wh-element (e.g. the subject in (10)) is attracted to C to check 

one-F(eature) wh-feature, while the other wh-elements (the objects in (10)) are attracted to 

C to check all-F(eature) focus feature. Bošković argues that C in Bulgarian (and 

Romanian) is a sort of syncretic head, having both an Attract one-F wh-feature and an 

Attract all-F focus feature, economy motivations driving movement of the highest wh- 

phrase to this head followed by movement in any order of the remaining wh-elements in 

order to check their strong focus-feature. After the wh-subject has merged as the external 

specifier of C, the wh-objects tuck in as inner specifiers of C. 

(10) CP 

3 

Wh1[+wh] C(P) 

koj 3 

Wh2[+F] C(P) 

kogo 3 

Wh3[+F] C(P) 

kakvo 3 

C IP 

one-Wh 6 

all-Foc dade 

‘who+Nom whom+Dat what gave’ 

However, under such an account, there are several focus positions per clause, which runs 

counter well-established facts for Romance (Rizzi 1997). 

Richards (1997, 2001) relates the impossibility of wh-movement out of wh-islands to 

the non-availability of multiple specifiers of CP and argues that, based on several criteria, 

i.e. the appearance of wh-island effects, scrambling, Superiority effects, WCO effects, the 

interaction of wh-movement with quantifiers, multiple wh-fronting languages branch into 

‘IP-absorption’ and ‘CP-absorption languages’. More exactly, Romanian and Bulgarian are 

held to be alike in that all wh-phrases move to fill multiple Spec CP, whereas languages 

like Serbo-Croatian have a single element in Spec CP. However, though not properly 

discussed here, Romanian (and Bulgarian) do exhibit island effects. 

137

Within a minimalist vein, Pesetsky (2000) adopts overt wh-phrase movement for Bulgarian 

and languages that are alike. His analysis relies not only on distinct pronunciation rules 

which distinguish English from Bulgarian, but also on the fact that languages would 

dispose of a certain type of Complementizer, of a C0-spec for wh-in-situ languages, of a C1- 

spec for English and of a Cm-spec for Bulgarian (and Romanian), the availability of a 

particular type of complementizer depending on the resources of the lexicon of that 

language. His proposal gives one the multiple vs. Non-multiple distinction but fails to 

account for the order of th-elements in a multiple wh-fronting language. 

2.4. The split CP analysis 

The analyses mentioned above, which make use of right adjunction, as proposed by Rudin 

(1998) and Grewendorf (2002), and multiple specifiers, as advanced by Bošković (1998), 

Richards (1997, 2001) and Pesetsky (2000), are incompatible with Kayne’s (1994) 

antisymmetry. Below I review some analyses that have been proposed along the lines of a 

fine articulated left periphery (for multiple questions) in German and Bulgarian. 

2.4.1. The split CP analysis: Grohmann (2002) 

Grohmann (1998, 2001) discusses the phenomenon of multiple wh in German (restricted to 

binary questions). According to Grohmann, the German left-periphery contains a FocP 

hosting one wh-phrase and a lower projection he labels FP hosting the second wh-phrase. 

One strong piece of evidence in favour of the split CP is the possibility for topicalisable 

elements to fill a position between the two wh-phrases, as illustrated in (12). 

(12) a. [FocP Wer hat [TopP alle/viele/die meisten Bücher [FP wo gekauft ]]]? 

‘Who bought all/many /most books where?’ 

b. * Wer hat kein Buch/wenige höchstens drei Bücher wo gekauft? 

‘Who bought no book/few/at most three books where?’ 

Grohmann takes these data to show that both wh-elements move overtly to FocP and FP, 

and that anything in between must fill Rizzi’s (1997) TopP. This accounts for the fact that 

whereas the quantificational elements in (12a) are topicalisable, the negative 

138

quantificational phrases kein N ‘no N’, wenige N ‘few N’, in (12b) are not. In his account 

German is after all like Bulgarian. 

2.4.2. The split CP analysis: Krapova & Cinque (2005) 

Krapova and Cinque (2005) provide an analysis of multiple wh fronting in Bulgarian that 

relies on Rizzi’s (1997) split-CP hypothesis and Rizzi’s (2004b) Relativised Minimality. 

The authors first establish a hierarchy of wh-phrases occurring in the CP-domain. Their 

data are summarised in table 1 below. 

Clitic resumed 

Topic wh 

(na) 

koj/koja/koe/koi 

(N) 

‘which N’ 

D-linked wh Non-D-linked wh 

(na) 

koj/koja/koe/koi 

(N) 

‘which N’ 

kogo (kogo) 

na kogo (na kogo) 

*?kakvo kakvo(marked) 

kâde/koga 

(marked) 

koj 

‘who’ 

kogo 

‘whom’ 

na 

kogo 

‘to 

whom’ 

Table 1 

koga 

‘when’ 

kâde 

‘where’ 

kakvoSubj 

‘what’ 

kolkoSubjN 

‘how 

many’ 

kakvoObj 

‘what’ 

kolkoObjN 

‘how 

many’ 

kak 

‘how’ 

The clitic resumed D-linked wh-elements fill the highest space at the left periphery being 

followed by D-linked wh-elements. Next come ordered wh-arguments (Subj > Object(s)), 

the lowest in the wh-hierarchy being the adjuncts, also observing Cinque’s (1999) 

hierarchy of adverbs (Time > Loc > Manner), and the quantificational phrases. 

In Krapova & Cinque (2005)’s analysis, each class of wh-phrases in table 1 is 

suggested to target a specific wh-position in the left-periphery (recursive Foc positions) 

(see also Krapova 2002). The details of their analysis, as it may apply to Romanian, are 

presented in section 4.1. The core idea is that the order of wh-elements in the left-periphery 

is dependent upon their Mittelfeld order prior to wh-movement. They make recourse to a 

version of Relativised Minimality in which only a whole chain, not a chain link, counts as 

an intervenor. This is a way of accounting for Superiority effects. Consider the sentences 

below (their sentences (13a, b)). 

(13) a. [CP Kakvo [CP kak [IP šte napraviš t t ]] What how you will do? 

139

. * [CP Kak [CP kakvo [IP šte napraviš t t ]] How what you will do? 

K&C (2005), (59) 

Chain crossing in (13a) reflects the Mittelfeld order Subject > Adjunct in the left periphery. 

The order of the two wh-phrases will not change upon movement, yielding the effect of 

Superiority. In other words, only a chain link intervenes between the adjunct kak ‘how’ and 

its trace, which does not give rise to a Relativised Minimality violation. Similarly, a chain 

link intervenes between the wh-argument and its trace. In (13b), a whole chain intervenes 

between the argument/adjunct and its trace and the sentence is thus ruled out. I will return 

to this analysis and consider how it fares in Romanian. It is to note that a case like (13a), 

involving crossing chains violates Pesetsky’s (1982) Path Containment Condition. 

3. The Ordering of Wh-phrases in Romanian 

Before proposing an analysis of wh-movement in Romanian, the particular ordering of wharguments 

with respect to wh-adjuncts and their internal make-up will be considered more 

in detail. K & C emphasize the role the feature [human] plays in accounting for the 

positioning of wh-subjects and objects. It will be shown below that this rigid ordering 

accounts of the formation of a wh-chunk inside the Mittelfeld. The discussion will also 

show that, like Bulgarian, Romanian exhibits a residual Animacy hierarchy. The aim of 

this section is to arrive at a close comparison between Romanian and Bulgarian, as 

described by K & C. 

3.1. The order of bare wh-phrases in Romanian 

The description below follows the description provided for Bulgarian by K & C. 

3.1.1. [+Human] Wh-Subjects and Objects 

The examples below test the order of wh-subjects and objects specified for the feature 

[+human]. For comparative purposes, the reader is referred to Table 1. 

140

(14) a. Cine pe cine a întâlnit? 

Who pe who has met 

‘Who met who? 

b. * Pe cine cine a întâlnit? 

(15) a. Cine cui (i)-a vorbit? 

Who whom him-has talked 

‘Who talked to whom?’ 

b. * Cui cine i-a vorbit? 

(16) a. Pe cine cui (i)-ai prezentat? 

b. 

Pe who who.Dat him-have2.sg introduced 

’Who did you introduce to whom?’ 

Cui pe cine ai prezentat? 

(14) and (15) show that the wh-subject cine ’who’ in Romanian, like koj in Bulgarian, 

positively specified for the feature [+human], necessarily precede the (direct/indirect) 

objects. While multiple questions containing two [+human] wh-objects must show the 

order Direct Object > Indirect Object, the reverse order, as in (16b), is also possible, but 

the dative object gets a D-linked interpretation. 

Descriptively, a long-acknowledged constraint like Subject first is operative with 

subjects and objects. i.e. the leftmost wh-phrase which is the subject was the highest before 

movement. 

The hierarchy obtained so far for Romanian is given in (17). 

(17) Cine > pe cine > cui 

3.1.2. Bare Wh-Subjects and Objects: [+Human] and [-Human] Wh-Phrases 

The examples below test the order of wh-subjects and objects differently marked for the 

feature [human]. 

(18) a. Cine ce a prezentat? 

Who what has presented 

141

‘Who presented what?’ 

b. * Ce cine a prezentat? 

(19) a. Pe cine ce ai întrebat ? 

Pe who what have.2.sg asked 

‘Who did you ask what?’ 

b. * Ce pe cine ai întrebat? 

(20) a. Cui ce (i)-ai arătat? 

Who.Dat what him-have.2.sg showed 

‘Whom did you show what?’ 

b. * Ce cui i-ai arătat? 

The [+human] subject and objects must precede the [-human] object, Romanian patterning 

with Bulgarian. The single interrogative pronoun ce ‘what’ lacks the feature [human], i.e. it 

is a ‘pure’ wh-element but once it is followed by a ceratin type of NP, it may become 

[+wh, +thing] or [+wh, +human]. 

Consider next the ordering of the inanimate Nominative subject ce ‘what’ with 

respect to the argumental [+human] wh-elements in (21) and (22). 

(21) a. Cui ce i s-a întâmplat? 

Whom what him.Cl s.refl-has.3sg happened 

‘What happened to whom?’ 

b. * Ce cui i s-a întâmplat? 

(22) a. Pe cine ce a surprins? 

Pe who what has.3sg surprised 

’What surprised whom?’ 

b. * Ce pe cine a surprins? 

Such cases may be taken as a violation of the Superiority condition as the inherently 

[+human] Dative and Accusative must precede the [-human] Nominative ce ’what’. This is 

reminiscent of the Animacy hierarchy strictly observed in languages like Navajo (Willie 

1991, Hale 1973) (see also section 4.1). 

142

Other wh-phrases to consider are cât ’how much’ and ce ’what’, both lacking the feature 

[human]. 

(23) a. ? Câţi copii ce colinde au cântat de sărbători? 

How many children what carols have.3P.pl sung on holidays 

‘How many children sang what carols on holidays?’ 

b. ??/ * Ce colinde câţi copii au cântat de sărbători? 

(24) a. ? Ce banchieri cât au investit în piaţa imobiliară? 

What bankers how.much have.3.pl invested on the housing market 

‘What bankers invested how much on the housing market?’ 

b. ?? Cât ce bănci au investit în piaţa imobiliară? 

(25) a. Pe cine câtor studenţi vei prezenta la petrecere? 

Pe who how.many.Dat students will.2sg introduce at party 

’Who will you introduce to how many students at the party?’ 

b. ?(?) Câtor studenţi pe cine vei prezenta la petrecere? 

As expected, in (23), the [+human] Nominative câţi copii ‘how many children’ precedes 

the [-human] Accusative ce colinde ’what carols’. Similarly, the [+human] Nominative ce 

banchieri ’what bankers’ in (24) precedes the non-human Accusative. Conversely, the 

[+human] Accusative in (25) precedes the [+human] Dative câtor studenţi ’to how many 

students’. However, if the order is reversed, as in (25b), a slightly marginal result obtains. 

The dative is understood as D-linked. The observation is that with two objects, the 

Animacy hierarchy ’decides’ which one is higher. 

To sum up, the relative order among the subjects and objects in Romanian is given in 

(26). 

(26) Cine > pe cine > cui > ce/câtNom > ce/câtObj 

The observation to make is that there is a high space at the Romanian left periphery filled 

by bare wh-arguments specified for the feature [human] followed by wh-subjects and 

objects which are either negatively specified or underspecified for the same feature. 

143

3.1.3. The order of wh-adjuncts 

As shown in (27) through (29), wh-adjuncts do not exhibit free order: când ‘when’ 

necessarily precedes unde ‘where’, which in its turn precedes cum ‘how’. By transitivity, 

când ‘when’ should precede cum ‘how’ and this is indeed the case in (29). The hierarchy 

of wh-adjuncts is given in (30). 

(27) a. Când unde l-ai văzut pe Ion? 

When where him-have2.sg seen pe Ion 

‘When did you see Ion where?’ 

b. ?/?? Unde când l-ai văzut pe Ion? 

(28) a. Unde cum i-ai reparat maşina? 

Where how him-have2.sg repaired car.the 

‘When did you repair the car how?’ 

b. * Cum unde i-ai reparat maşina? 

(29) a. ? Când cum i-ai reparat maşina? 

When how have2.sg repaired car.the 

’When did you rapair his car how?’ 

b. * Cum când i-ai reparat maşina? 

(30) Când > unde > cum 

The wh-adverbs observe therefore Cinque’s (1999) hierarchy of adverbs. 

3.1.4. [+/-Human ] Wh-Subjects and Objects and Wh-Adjuncts 

Subjects and objects with different feature specification show different orderings with 

respect to wh-adjuncts. Thus, the [+human] subject cine in Romanian precedes unde, când, 

cum, as shown by the contrasts in (31) through (33). 

(31) a. Cine unde va dormi? 

Who where will sleep? 

144

‘Who will sleep where?’ 

b. ?* Unde cine va dormi? 

(32) a. Cine când va sosi? 

Who when will come? 

‘Who will come when?’ 

b. ?* Când cine va sosi? 

(33) a. Cine cum va rezolva acest exerciţiu? 

Who how will solve this exercise? 

‘Who will solve this exercise how?’ 

b. * Cum cine va rezolva acest exerciţiu? 

Conversely, both the subject and the object ce (or ce N) ’what’ (’what N’) follow time and 

locative adjuncts, as in (34a) and (35a). 

(34) a. ? Când/unde ce s-a întâmplat? 

When/where what s.refl-has happended 

‘What did you see when/where?’ 

b. ?* Ce când/unde s-a întâmplat? 

(35) a. ? Când/unde ce i-ai spus despre mine? 

When/where what him-have said about me 

‘What did you tell him about me when/where?’ 

b. ?* Când/unde ce i-ai spus despre mine? 

Similarly, cât (N), whether subject (36) or object (37), follows time and locative adjuncts. 

Reversing the order leads to rather ungrammatical results. 

(36) a. Când/unde câţi invitaţi vor veni? 

When/where how many guests will arrive 

’How many guests will arrive?’ 

b. ?* Câţi invitaţi când/unde vor veni? 

145

(37) a. Când/unde câţi lei ai cheltuit iarăşi? 

When/where how much lei have spent again 

’How much lei did you spend when/where again?’ 

b. ?* Câţi lei când/unde ai cheltuit iarăşi? 

As exemplified below, both ce ’what’ and cât ’how much’ must precede, though with 

slightly marginal results, the voice adverb cum ’how’. This suggests that both arguments 

fill a position higher than the voice wh-adjunct but lower than the temporal and locative 

adverbs. 

(38) a. ? Cât cum a mâncat la prânz (de se simte atât de rău acum)? 

How.much how has eaten for lunch (that se.Refl is.feeling so bad now) 

‘How much has he eaten how at noon (that has made him be feeling so bad 

now)?’ 

b. ?? Cum cât a mâncat la prânz (de se simte atât de rău acum)? 

(39) a. ? Ce cum îi vei spune până la urmă fetei? 

What how her will tell finally girl.Dat 

‘What will finally you tell the girl how?’ 

b. ?* Cum ce îi vei spune până la urmă fetei? 4 

The low position of cum ‘how’ with respect to the human or non-human ce ‘what’ and cât 

‘how much’ leads to certain expectations. If a strict hierarchical order is assumed, the voice 

adjunct should be preceded by whatever matrial precedes ce and cât. The human objects 

(and subject) are expected to occur to the left of cum. This is indeed borne out by the data 

below. 

(40) a. Pe cine cum vei ajuta? 

Pe who how will.2sg. help 

‘Who will you help how? 

b. ?? Cum pe cine vei ajuta? 

146

(41) a. Cui cum vei anunţa această veste? 

Pe who how will.2sg.break.out this news 

‘Who will you break out the news? 

b. ?? Cum cui vei anunţa această veste? 

Similarly, other [+human] wh-arguments precede the time and locative adjuncts, as shown 

below. 

(42) a. Pe cine când/unde ai văzut ultima oară? 

Pe who when/where have2.sg. seen last 

‘Who did you see where last?’ 

b. ?? Unde pe cine ai văzut ultima oară? 

(43) a. Cui când (îi) vei vorbi depsre accident? 

Whom when (him) will talk about accident 

’Who will talk about the accident when?’ 

b. ?? Când cui (îi) vei vorbi depsre accident? 

However, there are some speakers for whom the temporal/locative > [+human] 

Accusative/Dative order is fine even under a non-Dlinked reading of the adjunct. For some 

speakers within this group, a human Dative/Accusative wh-argument and the nonhuman/animate 

ce ‘what’ seem to also be acceptable, again under the non-D-linked reading 

of the adjunct. In this respect, Romanian patterns differently from Bulgarian. These 

possibilities are exemplified below: 

(44) a. Unde pe cine ai văzut ultima oară? 

Where Pe who have.2.sg. seen last? 

‘Who did you see where last?’ 

b. ?(?) Pe cine unde ai văzut ultima oară? 

(45) a. Când pe cine vei suna? 

When Pe who will.2.sg. call up? 

‘Who will you call up when?’ 

b. ?(?) Pe cine când vei suna? 

147

(46) a. Cum pe cine vei ajuta? 

How Pe who will.2.sg help? 

‘Who will you help how?’ 

b. ?? Pe cine cum vei ajuta? 

(47) a. Când cui (i)-ai dat vestea cea bună? 

When whom (i)-have.2.sg given news.the the good 

‘Whom did you break the good news?’ 

b. ?(?) Cui când i-ai dat vestea cea bună? 

(48) a. Unde cui (i)-ai vorbit? 

Where whom (i)-have.2.sg spoken 

‘To whom did speak where?’ 

b. ?(?) Cui unde (i)-ai vorbit? 

(49) a. Cum cui (i) te-ai adresat? 

How whom (i) yourself.refl-have.2.sg speak.to 

‘Whom did you speak how?’ 

b.* Cui cum (i) te-ai adresat? 

Expectedly then a non-human wh like ce ’what’ in this variety marginally follows cum 

’how’. 

(50)a. ? Cum ce le va spune? 

How what them will say 

‘What will he tell them how?’ 

b. ?? Ce cum le va spune? 

It is also to note that in most cases the reverse order, esp. with the locatie and temporal 

adjuncts, results in slightly more marginality. I will henceforth label this variation, as 

instantiated in (40)-(43), G(rammar)1. The other group of speakers have G(rammar)2 and 

148

is exmplified by (44)-(50). Such data need however further investigation. 5 In the remainder 

of this dissertation I will mak use of the data specific to G1. 

To summarize thus far, the ordering of wh-adjuncts and arguments taking into 

account the two groups of Romanian speakers is provided in tables 2 and 3. They are to be 

compared to the data summarised in table 1 for Bulgarian. 

[+Human] Adjunct [-Human] Adjunct 

cineNom 

’who’ 

pe cineAcc 

’who’ 

cuiDat 

‘whom’ 

când 

’when’ 

unde 

’where’ 

ceNom/câtNom 

‘what’’how 

much’ 

Table 2: G1 - Romanian speakers 

ceAcc/câtAcc 

[+Human] Adjunct [+Human] [-Human] 

cineNom 

’who’ 

când 

’when’ 

unde 

’where’ 

cum 

‘how’ 

pe cineAcc 

’who’ 

cuiDat 

‘whom’ 

Table 3: G2 - Romanian speakers 

The summarised data point to the following generalisations: 

ceNom/câtNom 

‘what’’how 

much’ 

cum 

‘how’ 

ceAcc/câtAcc 

• The bare [+human] cine moves to the highest position in the wh-space, i.e. above 

the space occupied by the [+human] wh-objects for G1 speakrs or above the space 

filled by the adjuncts for the G2 speakers; 

• Therefore, the bare [+human] wh-objects, i.e. pe cine, cui move to a space above 

the adjunct space for G1, but below it for G2, a micro-parameter being thus set in. 

Bulgarian shares with Romanian only the first alternative; 

• The bare wh-subjects negatively specified or underspecified for the feature 

[+Human], i.e. ce ‘what’ and cât ’how much’ fill a position quite a low position in 

the wh-space. This is attributed therefore to their poor feature make-up; 

• Within G1, the lowest wh-element is the Voice djunct cum ’how’. 

• Both Grammars share the [+Human] > [-Human] split. 

5 I have been unable to make a correlation of these data to the origin of the speakers. It woud also be 

interesting to see whether this issue correlates with some other facts. I leave this as an open question. 

149

3.2. The order of D-linked and bare wh-phrases 

As discussed by Comorovski (1996), Cornilescu (2002), Alboiu (2002), D-linked wh- 

phrases in Romanian must precede a bare wh-phrase and are known not to show 

Superiority effects. Consider the cases below: 

(50) a. Care student ce i-a spus profesorului? 

Which student what him-has told professor.Dat? 

‘What did which student tell the professor?’ 

b. * Ce care student i-a spus profesorului? 

(51) a. Pe care student cine l-a examinat? 6 

Pe which student who him-has examined 

‘Who examined which student?’ 

b. ?(?) Cine pe care student l-a examinat? 7,8 

If ce ‘what’ (50a) and cine ‘who’ (51a) are understood as bare, non-D-linked wh-phrases, 

then they must follow the subject D-linked wh in (50) and the D-linked Accusative wh in 

(51). Note that the subtle contrast between (50b) and (51b) may be attributed to the poorer 

featural nature of ce ‘what’ (i.e. it is a ‘pure’ wh, having the default interpretation [- 

human]). 9 (51a) is a case of Superiority violation. 

Interpretively, D-linked wh-phrases are associated with givenness, being analysed by 

Cornilescu and Alboiu as Topics. 

Further examples of the rigid ordering of D-linked wh-phrases with respect to bare 

Nominative wh, and adjuncts, are provided below. 

6 Recall from the discussions in Chapter III that D-linking correlates with Clitic doubling in Romanian. 

7 Some speakers may find this sentence more unacceptable and mark it as ?? or ?* It does not sound very bad 

to me under the non-D-linked interpretation of cine. 

8 Such an order becomes acceptable under the D-linked reading of cine ‘who’ if an appropriate context is 

given: I know this is the lingustics exam day and that Professor Smith will examine some of the students and 

that Professor Johnson will examine the remaining ones. So I want to know … 

9 The fact that ce is light regarding its feature make-up is also supported by the fact that not even in echo 

questions, known not to observe Superiority, can it be the first wh in any sequence of wh’s, as observed by 

Comorovski (1996). 

150

(52) a. Cărei fete câţi băieţi i s-au adresat? 

Which.Dat girl how.many boys her s.refl-have spoken? 

‘How many boys spoke to which girl?’ 

b. ??/?* Câţi băieţi cărei fete i s-au adresat? 

(53) a. Pe care (student) când l-ai examinat? 

Which (student) when him-have.2sg examined 

‘When did you examine which (student)?’ 

b. ?? Când pe care (student) l-ai examinat? 

(54) a. ? Pe care maşină cum ai reparat-o? 

Pe which car how have.2sg repaired-it? 

‘Which car did you repair how?’ 

b. * Cum pe care maşină ai reparat-o? 

Another characteristic property of D-linked wh in Romanian is that, as opposed to 

Bulgarian (K&C identifies two spaces dedicated to D-linked wh-elements), in which Dlinked 

wh-elements observe the order subject > objects, they can be freely interchangeable. 

Thus, in (55), the inanimate Accusative wh can precede the human Nominative wh (see 

also footnote 7). Similarly, the D-linked voice adverb can precede the D-linked locative 

one though the sentence sounds a bit too ‘heavy’. The use of the conjunction şi ‘and’ 

would make the sentence sound very natural. Therefore, the Animacy hierarchy observed 

with bare wh is inoperative with D-linked wh. As suggested in section 4.3, this is related to 

their Topic-like nature. 

(55) a. La care articol care asistent lucrează zilele acestea? 

At which article which assistant is.working days.the these 

‘Which assistant is working on which article these days?’ 

b. Care asistent la care articol lucrează zilele acestea? 

(56) a. In ce mod anume în care loc s-a petrecut schimbul de 

ostateci? 

In what way exactly in which place s.refl-has taken.place exchange.the of 

hostages 

151

‘Which place did the hostage exchange take place in which particular way?’ 

b. In care loc în ce mod anume s-a petrecut schimbul de ostateci? 

The freely interchangeable nature of D-linked wh affects the modes of answering. (57a) 

below is a felicitous answer to (55a) and (57b) to (55b). Answering (57b) to a question like 

(55a) is infelicitous. (57a) is an infelicitous answer to (55b). 

(57) a. Zilele acestea, la articolul pe semantica lui why lucrează Joanna, la articolul pe 

probabilităţi lucrează Tanya, la articolul pe dialecte de germană lucrează Yves, 

... 

Days.the these, at article.the on semantics.the of why is.working Joanna, at 

article.the on probabilities is.working Tanya, at article.the on dialects of 

German is.working Yves, ... 

‘These days, Joanna is working on the article on the semantics of why, Tanya 

on the article on probabilities, Yves, on the article on German dialects, …’ 

b. Zilele acestea, Joanna lucrează la un articol pe semantica lui why, Tanya la un 

articol pe probabilităţi, Yves la un articol pe dialecte de germană, ... 

Days.the these, Joanna is.working at an article on semantics.the of why, Tanya 

is worksing at an article on probabilities, Yves is.working at an article on 

dialects of German, ... 

‘These days, Joanna is working on an article on the semantics of why, Tanya on 

the article on probabilities, Yves, on the article on German dialects, …’ 

The orderings in the tables 2 and 3 coupled with the data above give us the general picture 

of the Romanian wh-domain in table 4. 

152

D-linked Non-D-linked 

care 

‘which’ 

cărui(a) 

‘whose’ 

cine 

‘who’ 

Etc. 

Human G1:Human Adjunct - 

Cine 

‘who’ 

Pe cine 

whoAcc 

Când 

When 

Cui 

WhoDat 

Human 

Când 

When 

Adjunct -Human Adjunct 

Unde 

where 

CeNom/câtNom 

what/how 

much 

CeAcc/câtAcc 

what/how 

much 

G2:Adjunct +Human -Human 

Unde 

where 

Cum 

how 

Pe cine 

whoAcc 

Cui 

WhoDat 

Table 4 – General picture of the Romanian wh-space 

CeNom/câtNom 

what/how 

much 

In the next section we present Krapova & Cinque’s (2005) analysis in more details and 

consider how it applies to the Romanian data. 

4. Wh-chunk Movement in Romanian 10 

4.1. Krapova and Cinque’s (2005) Analysis 

Before proposing an analysis for Romanian, let us first assess how Krapova & Cinque’s 

(2005) analysis couched in terms of chain crossing, proposed for Bulgarian, would fare in 

Romanian. 

Cum 

how 

CeAcc/câtAcc 

Krapova & Cinque propose an analysis of multiple wh fronted elements in Bulgarian 

within a cartographic approach to the clause (Belletti 2004, Cinque 1999, 2003, Rizzi 

1997, 2004a, etc) relying on the following assumptions: (i) multiple wh-elements move 

individually to projections in the left periphery, and (ii) Superiority effects are obtained 

due to chain crossing. 

Insofar as the first assumption is concerned, they tacitly assume the existence of 

several Foci at the Bulgarian left periphery (Krapova 2002). The second assumption 

regards the order of multiple wh-constituents, which is constrained by Superiority. They 

10 This section is a revised version of Laenzlinger and Soare 2005. 

153 

what/how 

much

explore the idea that the order of wh-elements at the left periphery reflects their order 

before the derivation reaches CP. To this effect, they modify Chomsky’s (2000) principle 

of what counts as a blocker for the Minimal Link Condition with A chains (i.e. only the 

head of an A-chain) and adapt it to A’ chains. It reads as follows: 

(58) Only a whole chain, not just a link of a chain, counts as an intervenor. 

K&C (2005), (58) 

Superiority is subsumed under Rizzi’s (2001) Relativised Minimality. (58) says that no 

chain can intervene between the trace and the target if the intervening chain and the target 

share the same feature(s). Consider (59) as an exemplification: 

(59) a. [CP Kakvo [CP kak [IP šte napraviš t t ]] What how you will do? 

b. * [CP Kak [CP kakvo [IP šte napraviš t t ]] How what you will do? 

K&C (2005), (59) 

In their account, the manner wh-adjunct merges as Spec VoiceP. Irrespective of whether 

kakvo ‘what’ moves to a case position or not, it comes to be higher than the adverb. Chain 

crossing in (59a) reflects the Mittelfeld order Object > Adjunct in the left periphery, 

yielding the effect of Superiority. In (59a) only a chain link intervenes between the adjunct 

kak ‘how’ and its trace. Similarly, a chain link intervenes between kakvo ‘what’ and its 

trace. Chain crossing confirms therefore to RM. Conversely, in (59b), a whole chain 

intervenes between the adjunct and its trace in violation of (58). Besides, both chains are 

marked [+wh]. Chain nesting may however obtain with cases where the intervening chain 

(i.e. marked [+wh]) does not bear the same feature(s) as the target (i.e. [+D-linked, +wh]). 

Below we explore K & C’s idea that the order of the wh-phrases in the left periphery 

reflects their order prior to wh-movement. We make the assumption that this Mittelfledinternal 

order is also obtained with their non-wh counterparts. The relevant order of the wh 

and of their non-wh counterparts is VS(O). Consider the cases below: 

154

(60) Cui când (îi) va vorbi Ion? 

Whom when (him) will.3sg speak Ion 

’Whom will Ion speak when?’ 

The wh-adverb merges as Spec TP(future). The indirect object, cui, the verb and the 

subject Ion vacate the vP for phi feature valuation/checking. Once these movements have 

taken place, the wh-argument comes to precede the wh-adjunct’s merge position. The 

merge, non-wh canonical counterpart of (120) is the structure in (121a): 

(61) a. … [T(future) mâine [vP Ion [vorbi [Mariei]]]]. 

tomorrow Ion speak Maria.Dat 

In keeping with the above assumption, once both the subject and objects DPs have moved 

to Spec SubjP and ObjP, respectively, and the verb has raised to Subj° (61b), the relevant 

structure to be mirrored by the wh is therefore 61b, i.e. the VSO order (irrelevant steps are 

omitted). 

b. Îi va vorbi Ion Mariei [T(future) mâine [vP ]] VSOAdv 

As seen in section 3.1.3. the order of wh-adjuncts reflects Cinque’s (1999) hierarchy of 

adverbs, namely TimeP > LocP > VoiceP. On the assumption that Romanian is like 

Bulgarian in having several landing sites for adjuncts of the same nature, then as (62) 

shows, the hierarchy realized in the Mittelfeld and is retrievable through chain crossing at 

the left periphery. 

155

(62) ForceP 

h 

...FocP+time 

2 

FocP+location 

2 

FocP+manner 

2 

…IP 

TimeP 

2 

când LocP 

2 

unde VoiceP 

2 

cum …vP 

It is to mention that the order of wh-adverbs in (62) is exactly the same as the non-wh- 

adjunct (and adverb) counterparts, as (63) indicates. 11 The same holds valid for French in 

(64). 12 

(63) Am dormit azi noapte în sufragerie foarte bine. 

Have.1sg slept last night in dining-room very well 

’I slept last night in the dinnig-room very well.’ 

(64) J’ai dormi toute la nuit dans le salon dans un état de grande fatigue. 

I have slept all night long in the room in a state of great tiredness 

‘I was sleeping all night long in the room in a state of great tiredness’ 

11 One more argument supporting the hierarchy TimeP > LocP > VoiceP is provided by the order of 

existential wh-indefinites in German (Frey 2000, Grohmann p.c.). 

(i) Hans hat wann wo wie geschlaffen. 

Hans has whenever wherever however slept 

‘Hans slept whenever wherever however.’ 

12 In both Romanian and French, adjuncts can reorganize in the postverbal domain, giving rise to distinct 

orders. Cinque (2002b) proposes movement of a sub-structure containing one or more adjuncts past one or 

more adjuncts. 

156

Insofar as the order of argumental wh-phrases is concerned, several remarks are in order at 

this point. First, the subject c-commands the argumental objects, which reflects the order in 

the VP-shell and the neuter one in the Mittelfeld, as indicated in (65a,b). Since the order of 

the wh reflects the order of their (non-)wh counterparts in VSO, then the relevant one is 

thus (65b). 

(65)a. Cine ce a citit? 

Who what read 

b. A citit Ion cartea. VSO 

Has read Ion book.the 

‘Ion read the book.’ 

Secondly, the order among objects varies according to their grammatical function and also 

according to their being marked for the feature [+/-human]. In a three wh-phrase sequence 

the order among the second and the third wh-objects is relatively free (Rudin 1988, 

Richards 1997, 2001, Pesetsky 2000, a.o.). This is shown in (66). 

(66) a. Cine pe cine cui a prezentat ? 

Who pe who whom has introduced 

‘Who introduced who to whom?’ 

b. Cine cui pe cine a prezentat? 

Such freedom of ordering among the two [+human] wh-objects must be related to the 

ordering of non wh-phrases in the Mittelfeld. This is illustrated in (67). 

(67) a. Decanul l-a prezentat pe noul profesor studenţilor în dimineaţa aceasta. 

Dean.the him-has introduced pe new.the professor students.to this 

morning 

‘The dean introduced the new professor to the students in the class this 

morning.’ 

b. Decanul l-a prezentat studenţilor pe noul profesor în dimineaţa aceasta. 

157

The direct object and indirect object movement possibilities evinced in (67) can thus 

account for the order variation in (66). In (67a) the [+human] accusative PP precedes 

[+human] dative DP, whereas in (67b) the order is reversed. 

It was seen above that for one group of Romanian speakers(G1), all [+human] wh- 

arguments fill a space which is higher than the wh-adjuncts in the left periphery, as 

illustrated below. 

(68) Cine pe cine când a prezentat studenţilor? 

Who pe who when has introduced students.Dat 

‘Who introduced who to the students when?’ 

Here too, the order of the wh-phrases is derivable from their Mittelfeld-internal order. 

After having merged in the vP, the arguments land in SubjP and ObjP positions in the 

Mittelfeld. The time adjunct occupies its merge position, Spec TP (Past). As already seen 

in Chapter II, the (direct) object non-wh DP in Romanian can move as high as T(Past). 

Another possibility exists: as illustrated in (69) below, the [+human] object fills a 

position below the adverb. Under the hypothesis that the orders at the left periphery are 

derivable from those found in the Mittelfeld, the order in (69) is accounted for in terms of 

object DP remaining in a lower ObjP position among the many available ones below the 

TP hosting the adverb. 

(69) Cine când pe cine a prezentat studenţilor? 

Who when pe who has introduced students.Dat 

‘Who introduced who to the students when?’ 

Such considerations receive therefore support from the non-wh counterparts of (68) and 

(69) provided in (70). 

(70) a. Decanul l-a prezentat pe noul profesor în dimineaţa aceasta 

studenţilor. 

Dean.the him-has introduced pe new.the professor this morning 

students.Dat 

’The dean introduced the new profesor to the students this morning’. 

b. Decanul l-a prezentat în dimineaţa aceasta pe noul profesor studenţilor. 

158

The empirical data in the above section has shown that in both Romanian and Bulgarian 

wh-argument ordering exhibits the Animacy hierarchy at the left periphery, i.e. first come 

the [+human] arguments and then the [-human] ones. As discussed in many typological 

works (for instance, Comrie 1981), the animacy hierarchy is overtly expressed in the 

Mittelfeld of some languages such as Navajo (Hale 1973, Willie 1991). In Malayalam, a 

language with differential object marking, only animate and human objects are marked 

with the accusative case, but not inanimate objects (de Swart, Lamers and Lastrade 2008). 

This is illustrated in (71) below. 

(71) a. Avan oru pa∫uvine vaŋŋi. 

He a cow.Acc buy.Pst 

‘He bought a cow.’ 

b. Avan pustakam vaayiccu. 

He book read.Pst 

‘He read the book.’ Asher and Kumari 1997 

Only the [+animate] direct object in (71a) but not the [-animate] one in (71b) gets 

accusative case marking. As reported by de Swart et al. (2008), in Malayalam the subject 

has to be at least as animate as the object (see de Swart et al. for a collection of articles on 

on the role of animacy in argument structure and argument encoding, comprehension, etc.) 

Some facts from German are also reminiscent of such a hierarchy. One case in point 

is the derived positions of the indefinite/existential quantifiers in the German Mittelfeld. 13 

(72) a. Peter hat wem was gekauft. 

Peter has someone.Dat something bought 

‘Peter bought something to someone’ 

b. * Peter hat was wem gekauft. 

13 Ur Shlonsky (p.c.) points out that floating quantifiers in French also display the [+human/animate]>[- 

human/animate] hierarchy, as seen in (i) as opposed to (ii). 

(i) Je leur ai à toutes tout dit. 

I them.Dat have to all all said 

‘I said everything to all of them’ 

(ii) * Je leur ai tout à toutes dit. 

159

The human Dative wem must precede the non-human/animate was. 

To sum up, an analysis in terms of chain crossing which is proposed for Bulgarian 

may be tenable for Romanian. Yet, for reasons to be considered in the next section, we will 

adopt a different analysis for Romanian. 

4.2. Wh-Chunk Movement 

In this section an alternative analysis to wh-movement is suggested for Romanian while 

preserving the idea that the order of the wh-phrases at the left periphery reflects their order 

prior to wh-movement. 

One major parameter distinguishing Romanian from Bulgarian regards the so-called 

penetrability test. Bošković (1998) mentions that the distinction between Bulgarian and 

Serbo-Croatian w.r.t. the penetrability of the wh-cluster is not clear cut as some Bulgarian 

speakers allow the split. Indeed Lambova (2001), Krapova (p.c.) allow a split between the 

first wh-phrase and the second. The former provides evidence for this split coming from 

different types of particles, such as the vocative feminine ma (non-standard), and 

masculine be (non-standard), mari or mori (dialectal), bre or ba (dialectal), etc, from 

intervening parentheticals such as sâs sigurnost ‘for sure’, etc, from null subject clauses 

with (non-)bridge verbs such as kazvam ‘say’, iskam ‘want’, etc. and from adverbs. Some 

examples are given below. 

(73) Koj, čunkim, kakvo mi e dal? 

Who dub.prt what me.Cl Aux given 

Lambova 2001 (21) 

‘Who, for God’s sake, has given me what (for free)?’ 

(74) a. Koj, spored tebe, kakvo pie? 

Who according-to you what drinks 

‘Who, according to you, is drinking what?’ 

b. Koj, iskaš da znaeš, kakvo šte kaže? 

Who want.2.sg DA know what will say 

‘Who, do you want to know, will say what?’ Lambova 2001 (22) 

(75) Koj, navjarno, kâde koga šte porâča tortata? 

160

Who perhaps where when will order cake.the 

‘Who will perhaps have the cake made where and when?’ Lambova 2003 (32) 

Such data can be tied in with K&C’s analysis seen above in which the wh-phrases 

individually move to the appropriate positions at the left periphery, i.e. to several FocP’s. 

As far as Romanian is concerned, the penetrability test constitutes a strong argument 

in favour of the formation of what will be dubbed a wh-chunk, a subclausal portion 

containing all and only wh-phrases (Laenzlinger and Soare 2005). As the examples below 

show, neither particles, like the vocative măi (colloquial) and bre (colloquial and dialectal), 

nor parentheticals, like după tine ‘according to you’, după părerea ta ‘according to you’ 

can appear between the wh-elements, so muh the less adverbs of any type, as exemplified 

below (recall discussions in chapter III). 

(76) a. * Cine, măi, ce ţi-a spus? 

Who, you, what you-has told 

approx. ‘Hey you, who told you what?’ 

b. * Cine, bre, ce ţi-a spus? 

Who, you, what you-has told 

approx. ‘Hey you, who told you what?’ 

c. * Cine, după părerea ta, ce-a văzut? 

Who, acording to you, what-has seen 

‘Who saw what according to you?’ 

d. * Cine probabil/recent/calm ce a afirmat? 

Who probably/recently/calmly what has asserted 

‘Who has asserted probably/recently/calmly what?’ 

(76a,b) become fine if the vocative particles are fronted before the wh’s, or are clause-final 

separated by an intonational break. Similarly, (76c,d) become grammatical if the 

parenthetical phrase and the adverbs are clause-initial or clause-final. 

These data may be taken as strong evidence that as opposed to Bulgarian and Serbo- 

Croatian, wh-elements in Romanian do not target several positions in the left periphery. 

Recall also the complementary distribution of wh-phrases and focalised phrases in matrix 

context. This is illustrated in (77). 

161

(77) (*MARIEI) pe cine (*MARIEI) ai prezentat? 

Maria.Dat pe who have introduced 

‘Who did you introduce TO MARIA?’ 

Within a system which does away with multiple specifiers and right adjunction and where 

neither merge of the two constituents (Rudin, Grewendorf) nor individual movement of the 

wh-phrases (K&C) can account for wh-movement, it is suggested that it is a subpart of the 

Mittelfeld structure containing all and only wh-phrases that moves (Laenzlinger and Soare 

2005). This is schematically represented below. 

(78) [ForceP [FocP [XP cine ce] [FinP [XP ] … a spus …]]] 

who what has said 

The moved phrasal structure is labelled XP in (78) and it corresponds to a minimal 

Mittelfeld substructure to be identified in the following subsections. This type of 

movement makes recourse to two types of operations extensively used in the 

antisymmetric and cartographic approaches, namely heavy pied-piping and remnant 

movement. 

In the literature on pied-piping it has been pointed out that specifiers are potential 

pied-pipers (the whose picture type of pied-piping; see Cinque 2005) and that pied-piping 

is considered as a repair strategy, a Last Resort mechanism (see Heck 2004 for a thorough 

discussion on this phenomenon). The mechanism of feature percolation has been 

formalised in terms of the principle of feature inheritance by Pearson (2000). 

(79) Feature Inheritance Principle (Pearson 2000: 339). 

a. A functional projection FP is categorically non-distinct from a lexical 

head L iff it has inherited a lexical feature from L. 

b. A functional projection FP inherits a feature φ from a lexical category L iff: 

(i) An X°-projection containing φ adjoins to F°, or 

(ii) An XP-projection containing φ enters into a Spec-Head configuration with 

F°. 

162

The mechanism of extended pied-piping is a Last Resort strategy meant to satisfy interface 

properties while avoiding a locality principle. One needs one more clause under (144b), 

namely the picture of whom type of pied-piping (Cinque 2005). 

(iii) An XP-projection containing φ enters into a Compl-Head configuration with 

F°. 14 

Koopman & Szabolcsi (2000) introduce the mechanism of heavy pied-piping to treat cases 

of verb clustering in Dutch and Hungarian in terms of extended verb projection movement. 

Laenzlinger (2004, 2005) also makes use of heavy pied-piping that applies to substructures 

of the clause and the noun phrase, notably to account for the reverse order of 

modifiers and complements in the postverbal/nominal domain in Romance (see also Aboh 

2004, Shlonsky 2004, Cinque 2002b, among many others). 

In addition to the “heavy” pied-piping mechanism, the antisymmetric/cartographic 

approach also makes use of the remnant movement (see Kayne 1994, 2002, Belletti 2004, 

Cinque 2002b, Benincà and Poletto 2005, Munaro and Pollock 2005, a.o.). As far as the 

verb is concerned, remnant VP movement is proposed by Mahajan (2000) and Koopman & 

Szabolcsi (2000). Originally remnant XP movement has been proposed by den Besten & 

Webelhuth (1990) to account for cases of remnant VP-topicalisation in German, as in (80). 

(80) a. [CP [XP Das Buch nicht gelesen] [C hat] Hans gestern. 

The book not read has Hans yesterday 

b. [CP [XP Gelesen] [C hat] Hans das Buch nicht gestern. 

Read has Hans the book not yesterday 

‘Read the book Hans did not yesterday’ 

It is clear that the main verb and possibly its complements move to Spec CP (or Spec 

TopP). The verb alone has not moved as a head but as VP in order to observe structure 

preservation (Spec=XP). The nominal complement of the verb has left the VP. This 

remnant VP movement is licensed by interface requirements, more precisely by 

topicalisation of the lexical verb only in (80b) and of its complement and the negation in 

(80a). 

14 We leave aside the parametrisable restrictions that apply to the three clauses in (79b) crosslinguistically. 

163

We propose an analysis of multiple wh-fronting which relies on the mechanism of heavy 

pied-piping. At the Mittelfeld level, a subclausal structure made up of all and only wh- 

phrases is formed, i.e. a wh-chunk, which gets probed by the Foc head. Prior to this 

movement, the wh-phrases in a multiple wh-question need have their features valued. I 

suggest that in a multiple question, multiple feature valuing/checking is done via Hiraiwa’s 

(2000) mechanism of MULTIPLE AGREE. I adopt his mechanism of MULTIPLE AGREE 

originally proposed as an alternative to Chomsky’s (2000, 2001) AGREE unable to account 

for (covert) multiple feature-checking under the elimination of equidistance in multiple 

specifiers. I adapt his definition of MULTIPLE AGREE, applied to covert feature checking in 

Japanese, to overt multiple wh-fronting in Romanian. This mechanism reads as follows: 

(81) Hiraiwa’s MULTIPLE AGREE (2000:69): 




This definition relies on Chomsky’s (2000) Matching Condition: 

(82) Match (Chomsky 2000:122) 

(83) α > β > γ 


b. D(P) is a sister of P. 

c. Locality reduced to ‘closest c-command’. 

(AGREE(α, β, γ), where α is a probe and both β and γ are matching goals for α.) 

Once the probe P has been merged, it does SEARCH for the closest matching goal feature 

within its c-command domain and matches with the closer goal β but no AGREE takes 

place. Being [+multiple], P searches down for all matching goals, in this case for γ, until 

the matching is completed within an ‘accessible’ domain. At this point of the derivation, 

AGREE applies to all the matched goals derivationally simultaneously. 

All agreed wh-phrases must move to the left periphery but recall from Chapter III 

that Foc° has an EPP and that only one Spec FocP position is available in Romanian. The 

164

grammar needs therefore to find a way to move the wh’s. Chunk movement, relying on 

remnant movement and pied-piping, is a way to solve this tension. It is to point out that 

two or more wh-phrases in Romanian value/check the same feature on the probing head. 

This way, the derivation does away with tucking-in (as a result of MULTIPLE AGREE and 

movement of the chunk, the c-command relations among the wh-phrases are preserved). 

Let us see how this system works. Consider (84) below. 

(84) Cui ce spui? 

Who.Dat what [you] say 

‘Who are you saying what?’ 

The two wh-elements leave the VP-shell to reach Spec ObjIO and Spec ObjPDO, 

respectively, in the Mittelfeld, where they can value/check their case- and phi-features. As 

the [+/-human] feature is part of the feature make-up of wh-elements, the human dative is 

higher than the accusative. The verb moves high, i.e. to Fin°. 

(85) ForceP 

FocPwh 

3 

Spec 3 

Foc FinP 

uwh 3 

+EPP V … ObjPIO 

+multiple spui 3 

(2) cui+wh ObjPDO 

iwh 3 

(1) ce 

iwh vP 

(1) 

The feature system in the derivation in (85) involves an uninterpretable wh/Foc-feature on 

Foc which needs deleting/erasing. Foc° also has an EPP feature. The probe feature 

searches down for the closest mathichg goal feature within its c-command domain. It 

‘meets’ cui ‘whom’ and MATCH takes place. This does not result in immediate AGREE. 

Being [+multiple], the probe feature continues to probe for the next closest goal, resulting 

in matching with ce ‘what’. At this point, the probing head multiply agrees with both goals 

165

derivationally simultaneously. 15 At the point in the derivation where Foc° is merged the 

iwh/Foc-feature on both goals is active which makes it possible for both goals to be agreed 

with. The wh-chunk is formed and contains only the two wh-elements. By the mechanism 

of feature percolation, the wh/Foc-feature on the highest wh, i.e. cui, is transmitted to the 

entire functional category, i.e. ObjP. The EPP-feature can now attract the wh-chunk to 

Spec FocP. 

Let us further consider the case below containing besides the two wh-elements an 

adverbial PP. 

(86) Cine cui ce ai dat cu atenţie? 

Who whom what have.2sg given with care 

‘Who gave what to whom carefully?’ 

After merge of the arguments with V and v in the VP-shell, each targets an A-position in 

the Mittelfeld for phi/Case feature-checking. According to Cinque’s (1999) hierarchy, the 

manner adjunct is lower than the argumental positions in the Mittelfeld (see Cinque 2006). 

The derivation cannot proced as above, i.e. the pied-piping strategy allowing movement of 

SubjP to Spec FocP cannot take place as this domain also contains a manner adjunct, a 

non-wh element. In order for the derivation not to crash, the adjunct cu atenţie must leave 

this domain. By analogy with Diesing’s (1992) existential closure, where arguments in a 

non-existential reading, i.e. specific reading, must leave the verbal domain (this is 

applicable to Germanic languages), a similar mechanism, of wh-closure, may be suggested. 

As a result, the wh-domain must be vacated of all non-wh phrases. Once this is done, 

MULTIPLE AGREE from Foc° applies to the three matched goals derivationally 

simultaneously. This is represented below. 

15 

Somewhat the same idea is present in Pesetsky and Torrego (2004) who replace Chomsky’s definition of 

AGREE (2000, 2001) by a feature sharing view of AGREE. They state that once AGREE has applied between 

two (unvalued) occurrences of F(eature) at two locations (α and β), a later operation AGREE applies between 

one of these two instances of F and a distinct (valued) F at a distinct location (γ) and the result will be a 

valued feature F present at three locations (my italics). 

166

(87) …FocPwh 

3 

SubjP+wh 3 

Foc 

uwh FinP 

+EPP 3 

+multiple V …FP 

a dat 3 

PP SubjP +wh 

3 

cine+wh 

(2) ObjP+animate 

2 

cui ObjP-animate 

2 

(2) ce VoiceP 

(3) 2 

(1) cu atenţie vP 

2 

VP 

2 

V 

2 

V 

Remnant movement (of any kind), preceded by evacuation of the chunk/domain by some 

constituent(s), raises a problem for the Probe-Goal system. It can rightly be supposed that a 

constituent must leave a domain (existential, wh, Topic, Focus…) as it would be an 

illegitimate object in that domain and the derivation would crash. The constituent must 

target a specifier position outside the respective domain by being attracted by the head of 

the projection containing this specifier position. Related to this, it is to note however that 

an analysis which makes recourse to remnant movement lacks, to my knowledge, a sound 

motivation for why (some) elements vacate the remnant. This is a more general problem of 

this approach. 

One advantage of the analysis in terms of chunk movement proposed here is that it 

preserves K & C’s hypothesis that multiple wh-phrases at the left periphery is preestablished 

in the Mittelfeld. For example, in (88), the order of the argumental and adjunct 

wh-phrases is established in the Mittelfeld in such a way that the wh-chunk (=SubjP) 

reflects the order at the left periphery. It is to mention that wh-closure requires that the 

dative non-wh argument move out of this domain, as represented in (88). 

167

(88) Cine pe cine când a prezentat Mariei ? 

(89) 

Who what when has introduced Mary.Dat 

‘Who introduced whom where when how to Mary?’ 

ForceP 

....FocPwh 

2 

FinP 

2 

a prezentat FP 

2 

Mariei …SubjP 

2 

cine+wh ObjPanimate 

2 

pe cine ObjPanimate 

2 

Mariei …TimeP 

2 

când … 

Given that extraction of two wh-elements out of a wh-island is made by chunks, which 

constitutes a non-argumental category, Rizzi (p.c.) points out that such an analysis makes 

the prediction that such extraction of a wh-chunk should be worse than movement of a 

single argument. In other words, the status of (90a) should be somewhat similar to that of 

(90b) rather than (90c). Though judgments are very subtle, it seems to me that this 

prediction is indeed borne out. 

(90)a. ** Cine pe cine ştii când a întâmpinat? 

Who pe whom know.2.sg how has welcomed 

‘Who do you know welcomed whom how?’ 

b. * Când ştii pe cine a întâmpinat? 

c. ?? Pe cine ştii când a întâmpinat? 

‘Who do you know how he welcomed?’ 

168

It has been shown in this section that multiple feature-valuing/checking can be accounted 

for in terms of MULTIPLE AGREE and heavy pied-piping of a wh-chunk containing all and 

only wh-elements. I turn to D-linked wh-phrases. 

4.3 The Syntax of D-linked Wh-Phrases 

As already discussed in section 3.2 and as summarised in table 4, D-linked wh-phrases in 

both Romanian and Bulgarian must precede bare wh-elements at the left periphery. 

Interpretively, they resemble Topics in that they contain the notion of givenness. 

(91) a. Pe care fată când ai văzut-o ultima dată? 

Pe which girl when have.2.sg seen-her last time 

’When did you see which girl last?’ 

b. ?? Când pe care fată ai văzut-o ultima dată? 

Whereas Bulgarian exhibits two distinct spaces for D-linked wh-phrases, a higher one 

filled by clitic-resumed wh (na koj prijatel in (91)), a lower one not resumed by a clitic 

(koja kniga in (92)), Romanian does not display such a distinction. It departs therefore 

from Bulgarian in having only one space D-linked wh-phrases fill. Recall that any nonsubject 

D-linked wh-phrase in Romanian is obligatorily clitic resumed. Compare (92) to 

(93). Another characteristic property is that D-linked wh-phrases in Romanian do not 

observe the Subject > Object order and are freely interchangeable, as in (93). 

(92) Na koj prijatel, spored tebe, koja kniga da mu (*ja) dam? 

To which friend, according to you, which book to him (it) give-I 

’According to you, to which friend should I give which book?’ K&C (42c) 

(93) a. Cărui prieten pe care carte *(i)-*(o) vei da? 

To which friend pe which book him-it (you) will give 

‘To which friend will you give which book?’ 

b. Pe care carte cărui prieten *(i)-*(o) vei da? 

As far as CLLD is concerned, D-linked wh tend to follow them. This is illustrated below: 

169

(94) a. Cadoul, cărei fete i-l vei oferi? 

Present.the which.Dat girl her-it (you) will offer 

‘The present, which girl will you offer?’ 

b. ?? Cărei fete cadoul i-l vei oferi? 

(95) a. Mâine, pe care fată o vei vedea la cocktail? 

Tomorrow pe which girl her (you) will see at cocktail 

’Tomorrow which girl will you see at the cocktail party?’ 

b. ? * Pe care fată mâine o vei vedea la cocktail? 

I suggest that the wh/Foc-feature on Foc° probes and agrees with a D-linked wh which 

subsequently moves to Spec Foc. In keeping with Rizzi’s (2001b) analysis, since a D- 

linked wh also has a Top-like feature, it further moves to the specifier of a projection 

labelled here Top+WhP. The more marginal status of (94b) and (95b) above may be taken 

to result from a violation of the hierarchy in (98) below. However, the gradient in 

(un)grammaticality displayed by (94b) and (95b) may be put down to the Topic + 

Argument nature of both DPs in the former case vs. the Topic (or even Modifier) nature 

only of the adverbial in the latter. This further suggests that the ordering of Topics in the 

Romanian left periphery is not completely free and that at least D-linked wh-elements must 

be lower than other Topics. 

Let us see how the derivation works for a sentence like (96) below: 

(96) a. Cadoul, cărei fete când i-l vei oferi? 

Present.the which.Dat girl when her-it (you) will offer 

‘The present, when will you offer to which girl?’ 

b. ?(?) Cărei fete, cadoul, când i-l vei oferi? 

c. * Cărei fete când cadoul i-l vei oferi? 

As above, the most natural order is the one in (96a), with the non-wh Topic preceding the 

D-linked Topic-like DP followed by the bare wh. (96b) displaying the reverse order is 

slightly marginal, whereas (96c) is ungrammatical as it is a violation of the wh-V 

adjacency (or of the Foc-Fin criterial configuration). 

Let us see the mechanics of the derivation. I suggest that in (96a), the probing 

wh/Foc-feature on Foc° searches down for matching features and ‘locates’ the wh/Foc- 

170

feature on the closest goal, cărei fete ’to which girl’. AGREE does not take place at this 

point. Since the probe is [+Multiple], it does SEARCH again and matches with the wh- 

feature on când ’when’. At this point MULTIPLE AGREE takes place as a single syntactic 

operation. This is shown in (97) by the undotted lines (1). Since the D-linked wh also has a 

Topic-like feature it cannot get attracted to Spec Foc as it would get frozen in the criterial 

Spec Foc position and would remain with the Top-feature unvalued. Consequently, EPP on 

Foc° attracts only the bare wh (the D-linked wh is now inactive and does not function as a 

blocker). This is shown by the dotted line (2). The Top feature on Top+Wh° probes and 

agrees with the corresponding feature on the D-linked element. Due to EPP, the wh-DP 

moves to Spec Top+WhP (the dotted line (3)). Upon merge of Top° associated with 

CLLD, the probing Top-feature does SEARCH and agrees with the full DP cadoul ‘the 

present’ which then moves to the specifier position (the dotted line (4)). All this is 

represented below. 

(97) 

...TopP*CLLD 

3 

3 

+Top TopP+Wh* Cl-D-linked 

+EPP 3 

3 

+Top FocPWh 

+EPP 3 

3 

+wh FinP 

+EPP 3 

+multiple …ObjPDO 

3 

cadoul ObjPIO D-linked-wh 

(4) 3 

cărei fete T(Future) 

(3) 3 

(1) Mult.Agree când … 

(2) 

(1) Mult.Agree 

As there is no feature-identity, no RM effect obtains among the Top-, Top+Wh- and Focchains. 

To sum up, the analysis in terms of chunk movement of the multiple wh-phrases 

suggested for Romanian relies on K & C’s idea that the order of wh-phrases at the left 

171

periphery in Romanian/Bulgarian mirrors their order in the Mittelfeld. However, the 

analysis differs from that K & C’s in that it implies only one Foc position for the bare whphrases. 

It involves heavy pied-piping of a substructure of the Mittelfeld containing only 

wh-phrases once MULTIPLE AGREE from the probe, Foc, has taken place (I have adopted 

Hiraiwa’s mechanism and thus explored its implementation for multiple wh-phrases in 

Romanian). As opposed to Bulgarian which has two spaces for (inherently) D-linked whphrases 

(K&C 2005:186), one for the clitic-resumed, the other for the non-Cl-resumed whelements, 

Romanian inherently (care ‘which’) D-linked wh-phrases fill must be clitic Dlinked. 

It has been suggested that given that such a D-linked wh bears a Top-like feature, it 

gets attracted to the appropriate position, i.e. Spec Top+WhP above FocP. In the light of 

these considerations, the left periphery of the Romanian matrix clause is provided below: 

(98) Force … Top*CLLD/English-like … Top*CLLD … Top+Wh* … Foc … Fin … (revised I) 

5. The Case of a Wh-in-situ in Romanian: ‘Why’ 

This section is part of ongoing work with Ur Shlonsky. 

It has been seen so far that wh-phrases in Romanian merge inside vP and get attracted to 

the Mittelfeld to value their DP- or adverb-related features only to move further up in Spec 

Foc. As shown by Rizzi (1990, 2001a), why, an adverb scoping over the proposition in its 

immediate scope, is directly merged fairly high in the left periphery in a dedicated position 

labelled IntP. The section investigates the cartographical issue of structural placement of 

the Romanian why, i.e. de ce, on a comparative basis with the Italian perché, as analysed 

by Rizzi. Once this issue is dealt with, the behaviour of de ce is explored with respect to 

Intervention Effects. 

It is common knowledge that adjuncts like why (and how) in English cannot remain 

in situ in (99a,b) and must move hence the possible co-occurrence of why with another whphrase, 

as in (99c) (Bollinger 1978, Bromberg 1987, Collins 1990, Rizzi 2001a; for the 

idea that why cannot remain in situ but must merge high in CP in the so-called wh-in-situ 

languages see Tsai 1994, 2008 for Chinese; Ko 2005 for Korean and Japanese). 

(99) a. * Who left why? 

172

. * Who fixed the car how? 

c. Why did John eat what? 

Similarly, in French, which has the in-situ strategy, pourquoi ‘why’ cannot remain in situ 

but must move. If in situ, it can only have the purpose meaning (Aoun 1986, Rizzi 1990). 

This is shown in (100). 

(100) a. * Tu es venu pourquoi? 

You have come why 

b. Tu es venu pour quoi? 

a. Pourquoi tu es venu? 

Why is in itself ambiguous between a reason interpretation and a purpose interpretation. 

This is exemplified in (101) below. In Stepanov and Tsai’s (2008) and Tsai’s (2008) 

analysis this structurally correlates with the CP-merge position of reason why and with the 

vP periphery-merge position of purpose why. 

(101) A: Why did John bring the ladder into the conference room? 

B: i. In order to fix the light bulbs. 

B: ii. Because he wanted/needed to fix the light bulbs. Tsai (2008) 

The same meaning ambiguity is present with Romanian de ce in (102). 

(102) De ce a plecat Ion? 

Of what has left Ion 

‘Why did Ion leave?’ 

Answer 1: Pentru că era foarte obosit. de ce R(eason) 

‘Because he was very tired.’ 

Answer 2: Ca să prindă trenul. de ce P(urpose) 

‘To catch the train.’ 

This meaning ambiguity is rendered in some languages by distinct lexical items. Thus, 

Russian distinguishes between začem ‘for which purpose’ and počemu ‘for which reason’, 

and Polish has po co purpose ‘why’ and dlaczego reason ‘why’ (though the latter can still 

be ambiguous between purpose and reason, Joanna Blochowiak, p.c.). Serbo-Croatian also 

173

has two distinct lexical items: zbog čega reason ‘why’ (though the latter can still be 

ambiguous between purpose and reason, Tanja Samardzic, p.c.) and zasto purpose ‘why’. 

Chinese also has an optional aspect marker with purpose wei(-le)shenme, whereas reason 

why cannot have it, wei(-*le)shenme. Besides the intrinsic meaning ambiguity of why, 

Romanian has several PPs to express purpose why: pentru ce for what (‘what for’) and la 

ce at what (‘what for’) 16 . The PP de ce is lexicalised behaving as a real Adv. 

5.1 The position of dacă ’if’ within the Romanian Left peripehry 

Embedded yes/no questions in Romanian are introduced by dacă. Its Italian counterpart se 

shares some properties with the declarative complementizer che, i.e. both are compatible 

with a focused phrase and only se can be preceded and followed by topicalised phrase 

(Rizzi 2001a). As will be seen below, Romanian dacă behaves in a similar manner. 

Consider the interaction of the complementizer că and Focus below. Focus must 

follow but not pecede the declarative complementizer. 

(103) a. Cred că UN MACINTOSH îşi va cumpăra. 

‘I believe that A MACINTOSH he will buy himself.’ 

b. * Cred UN MACINTOSH că îşi va cumpara. 

Similarly, Focus must follow dacă, otherwise ungrammaticality obtains. 

(104) a. Mă întreb dacă UN MACINTOSH îşi va cumpăra. 

‘I wonder if A MACINTOSH he will be himself.’ 

b. * Mă întreb UN MACINTOSH dacă îşi va cumpăra. 

Whereas a Topic must follow că, it can precede or follow dacă, as illustrated by the 

contrast below. 

(105) Cred, (*lui Ion), că (lui Ion) îi vor oferi o sticlă de şampanie Moet. 

’I believe that to Ion they will offer a bottle of Moet champagne.’ 

16 Note, however, that provided an appropriate context, a causal reading is not excluded with la ce. 

174

(106) a. Nu ştiu, (lui Ion), dacă (lui Ion) îi vor oferi o sticlă de şampanie Moet. 

’I don’t know if to John they will offer a bottle of Moet champagne.’ 

As was proposed for Italian, in Romanian, too, there is an Int(errogative) projection high in 

the left periphery, sandwiched between the two available Topics. 

(108) Force (Top*) IntP (Top*) Foc Fin IP (revised II) 

Colloquial Romanian has the option of expressing both the Force head and the Int head in 

this order (just like Spanish does). This is illustrated below. 

(109) a. Mă tot întreba că dacă nu vreau să merg cu ei toţi în vacanţă. 

Mă.refl still asked that whether not want să.Subj want with them on holidays 

‘He kept asking me whether I wanted to join them all on holidays.’ 

b. * Mă tot întreba dacă că nu vreau să merg cu ei toţi în vacanţă. 

Recall that in embedded clauses a focused phrase, such as Maria in (110), co-occurs with 

the wh-phrase ce ‘what’ in this fixed order. 

(110) a. Nu ştiu MARIA ce găseşte pe birou, la servici, dar eu găsesc mereu câte un plic. 

’I dont know MARIA what finds on their desk at work, but I always find some 

envelope’. 

b. * Nu ştiu ce ALŢII găsesc pe birou, la servici, dar eu găsesc mereu câte un plic. 

Putting all these facts together, one arrives at the following structure of the left periphery 

of the Romanian embedded clause: 

(111) Force (Top*) IntP (Top*) Foc FocWh Fin (revised II) 

5.2 De ce - Its properties 

Below we discuss some properties characteristic of this wh-phrase (see also Soare 2008). 

175

5.2.1 (Non-)inversion 

As pointed out by Rizzi, whereas wh-phrases, arguments or adverbs, require verb- 

adjacency, perché ‘why’does not, as illustrated by the contrast (112)-(113) below. 

(112) a. Dove è andato Gianni? 

‘Where did Gianni go?’ 

b. * Dove Gianni è andato? Rizzi 2001a (16) 

(113) Perché Gianni è venuto? 

Why Gianni has come 

‘Why did Gianni come?’ Rizzi 2001a (21) 

Similarly, in Romanian, the verb adjacency need not occur with de ce, as opposed to cum 

‘how’ in (115). There is a clear contrast between (114b) and (115b). 

(114) a. De ce urăsc copiii gradiniţa la început? 

Why hate children.the kindergarden in the beginning 

‘Why do children hate kindergarden in the beginning?’ 

b. (?) De ce copiii urăsc gradiniţa la început? 

(115) a. Cum a ajuns Ion la aeroport? 

How has arrived Ion at airport 

‘How did Ion arrive at the airport?’ 

b. * Cum Ion a ajuns la aeroport? 

Though there is some speaker variability as to the judgments in (116), (117), (118) below 

and (114b) above, some speakers find the co-occurrence of de ce and subject DP either 

perfect or slightly degraded, whereas others seem not to accept such sentences. This may 

be related to the VS/SV character of Romanian (Ur Shlonsky, p.c.) As I am part of the first 

group, the analysis to suggest will take into account the grammaticality of such sentences. 

(116) 

De ce miniştrii nu au participat şi la alte mitinguri? 

Why ministers.the not have participated also at other meetings 

176

‘Why didn’t the ministers take part in other meetings, too?’ 

(117) De ce guvernul îşi face capital electoral pe banii contribuabilului? 

Why government.the makes itself polling advertisement on the tax payers’s 

money 

‘Why does the government make itself polling advertisement on the tax 

payers’s money?’ 

(118) De ce studenţii nu se implică mai mult în politică? 

Why students not se.refl get involved more in politics 

‘Why don’t students get more involved in politics?’ 

Evidence coming form verb interaction with the epistemic adverb and from subjunctive 

(and infinitival) has shown that the verb in Romanian may move quite high, i.e. it can 

move as high as the Fin head in the VS(O) order, whereas the subject DP fills a high 

position in the Mittelfeld. In the SV(O), the verb remains lower, i.e it may fill Subj° of the 

Criterial postion, whereas the subject fills the specifier of this position. Since the subject 

DPs in the examples above do not fill Spec TopP but Spec SubjP, where they get a Topicaboutness 

interpretation, the variety of Romanian where (114) as well as (116-118) are fine 

is thus straightforwardly explained. The adjunct de ce does not require inversion and 

consequently the verb need not move to Fin. In other words, the Wh-Criterion does not 

obtain with de ce. Rizzi attributes the lack of I-to-C requirement to the fact that the Int 

position is intrinsically endowed with the wh-feature. 

The possibility of merging de ce with no adjacency requirements for the verb also 

leads to the expectation that other material than subjects may occur between the wh-phrase 

and the verb. The nature of these potential elements will be discussed in the sections 

below. I content myself with noting that an embedded clause may do so. Consider the 

contrast below: 

(119) a. De ce [când vine la Geneva] îşi cumpără mereu ciocolată? 

Why when (he) comes to Geneva himself buys always chocolate 

‘Why does he always buy himself chocolate when he comes to Geneva?’ 

b. * Ce [când vine la Geneva] îşi cumpără mereu? 

What when (he) comes to Geneva himself buys always 

‘What does he always buy himself when he comes to Geneva?’ 

177

’. Ce îşi cumpără mereu când vine la Geneva? 

What himself buys always when (he) comes to Geneva 

‘What does he buy himself when he comes to Geneva?’ 

The contrast (119a-b) shows that an embedded clause may separate de ce and the main 

verb and that this is not licensed for ce ‘what’. In this latter case, the embedded când- 

clause must follow the main verb, as in (119b’). 

Consider next the possible interpretations with a two-clause sentence: 

(120) a. De ce spune secretara că mulţi din angajaţi vor demisiona? 

Why says secretary.the that many of the employees will resign 

Why does the secretary say that many of the employees will resign? 

[Ambiguous reading] 

b. De ce secretara spune că mulţi din angajaţi vor demisiona? 

[Matrix reading] 

The observation is made at this point that with inversion in the matrix clause both the 

matrix and the embedded reading are obtained. Though judgments are very subtle, the 

matrix reading is strongly preferred if not the only available one with non-inversion. 17 

These facts will receive an account in section 5.3. 

5.2.2.Focus Sensitivity 

If de ce is higher than subjects and Topics, it must be higher than Foci. That this is indeed 

the case is shown by the grammaticality of the examples below. 18 As pointed out in 

chapter III, other wh-phrases cannot co-occur with a focused element. 

(121) De ce FOTBALIŞTII primesc sume imense de bani (dar nu şi alţi sportivi)? 

17 

If inversion occurs both in the matrix and in the embedded clause, the same ambiguity as in (120a) obtains. 

If inversion occurs in the embedded clause but not in the matrix one, the matrix reading is preferred, as in 

(120b). 

18 

Both groups of speakers consider such senteces either perfectly grammatical or mark them with a single ‘?’ 

question mark. 

178

Why footballers.the get sums huge of money (but not other sportsmen) 

’Why do THE FOOTBALLERS get huge sums of money (but not other 

sportsmen)?’ 

(112) De ce LUI ION i s-a cerut demisia (dar nu şi colegilor săi)? 

Why TO ION him s-has asked resignation.the (but not to his colleagues) 

‘Why was ION asked his resignation (but not his colleagues)?’ 

The argument that why is sensitive to Focus goes back to Bromberger (1992). Consider the 

pairs (123) and (124). 

(123) a. Why did ADAM eat the apple? 

b. Why did Adam eat the APPLE? 

c. Why did Adam EAT the apple? 

d. Why did Adam eat the apple? 

(124) a. When did ADAM eat the apple? 

b. When did Adam eat the APPLE? 

c. When did Adam EAT the apple? 

d. When did Adam eat the apple? 

Focusing distinct constituents in (123) changes the type of the answer. Thus, as shown in 

Stepanov and Tsai (2008), the following are the four answers to (125). 

(125) a. Because he (Adam) was the one that Eve worked on 

b. Because it (the apple) was the only food around 

c. Because he couldn’t think of anything else to do with it 

d. Because God intended that to happen. Stepanov and Tsai (2008), (28) 

Conversely, only one answer is possible to the four questions in (126). 

(126) At 4 p.m. on July 7 th . Stepanov and Tsai (2008), (29) 

This contrast shows that why but not when can scope over the focus operator. 

179

5.2.3 Interaction with Topicalisation 

In Italian, a Topic element can either precede or follow perché (Rizzi). The same holds 

true of Romanian. Consider the examples below: 

(127) a. (Dar) celorlalţi de ce nu le-ai vorbit? 

(But) the.others.Dat why not them-have talked.to 

‘(But) the others why didn’t you talk to?’ 

b. De ce celorlalţi nu le-ai spus nimic? 

Why the.others.Dat not them-have said nothing 

‘Why didn’t you say anything to the others?’ 

(128) a. De ce întotdeauna numai eu fac astfel de greşeli? 

Why always only I make this kind of mistakes 

‘Why it is me that I always make mistakes?’ 

b. * Intotdeauna de ce numai eu fac astfel de greşeli? 

In (127a), the positive quantifier is clitic left-dislocated to a position above IntP, the 

position filled by de ce, whereas in (127b) the quantifier fills a lower TopP. The contrast in 

(128) shows that IntP is higher than Mod(ifier)P. 

5.2.4 Negation in Long-Distance Dependencies 

Let us further consider the interaction of why with negation. 

(129) De ce nu crede secretara că (unii angajaţi) vor demisiona (unii angajaţi)? 

‘Why doesn’t the secretary believe that some employees will resign?’ 

[Matrix reading] 

With negation in the matrix clause, only the matrix reading of de ce obtains. The presence 

of negation in the matrix clause has a blocking effect on the embedded reading. 

5.2.5 Quantificational Adv in long-distance dependencies 

180

The examples below contain a Voice and a quantificational adverb, respectively. Whereas 

in the former case, ambiguity can obtain, in the latter, the embedded reading is harder to 

get. 

(130) De ce anunţă calm că (unii angajaţi) vor demisiona (unii angajaţi)? 

‘Why does he calmly announce that some employees will resign?’ 

[Matrix+embedded] 

A1: Pentru că nu îi place să adopte un ton ferm în astfel de condiţii. 

‘Because he doesn’t like to adopt a firm tone of voice in such conditions.’ 

A1: Pentru că sunt prost plătiţi. 

‘Because they are badly paid.’ 

(131) De ce anunţă mereu că (unii angajaţi) sunt pe punctul de a demisiona (unii 

angajaţi)? 

‘Why does he always announce that some employees are about to resign?’ 

[Matrix] preferred 

A1: Pentru că îi place să ne ţină la curent. 

‘Because he likes to keep us posted.’ 

A2:# Pentru că sunt prost plătiţi. 

‘Because they are badly paid.’ 

5.2.6 Wh-phrases – Main contexts 

The data discussed in the sections above confirm Rizzi’s assumptions made for Italian, i.e. 

de ce merges in a position above subjects, Topics and Foci. Let us further consider the 

interaction of de ce with wh-phrases given that Romanian is a multiple wh-fronting 

language. 

(132) a. Cine de ce a plecat? 

Who why has left 

‘Who left why?’ 

b. Cine cum a reparat maşina? 

181

Who how has repaired car.the 

‘Who repaired the car how?’ 

c. (?) Cui de ce (i) te-ai adresat ieri? 

Whom why him you-have spoken.to yesterday 

‘Why did you speak to whom yesterday?’ 

d. ? Ce de ce i-ai spus ieri? 

What why him-have told yesterday 

‘Why did you tell him what yesterday?’ 

Contrary to expectations, de ce follows a wh-argument. As the examples below show, why 

also follows a wh-adjunct. 19 However, with the lowest element in the wh-hierarchy 

19 Not only Romanian, but also Bulgarian seems to present the same pattern (though things may be more 

complicated as Bulgarian prefers the coordination pattern, Iliyana Krapova, p.c.). Thus, in (i) zašto must 

follow the wh-subject. 

(i) a. Koj zašto kupil kniga? Bulgarian 

Who why bought book 

b. * Zašto koj kupil kniga? Stepanov and Tsai (2008), (7/94) 

Polish (ii)/(iii) and Serbo-Croatian (iv)/(v) pattern alike in that the both reason and purpose lexicalised items 

can be switched in the presence of another wh-element, this free order engendering distinct presuppositions 

(Joanna Blochowiak, for Polish, p.c.) 

(ii) a. Kto dlaczego umarl? Dlaczego R 

Who why died 

‘Why did who die?’ 

b. Dlaczego kto umarl? Joanna Blochowiak, p.c. 

(iii) a. Po co kto przyjechal? Po co P 

For what who came 

‘Why did who come?’ 

b. Kto po co przyjechal? Joanna Blochowiak, p.c. 

(iv) a. Ko je zbog čega umro? zbog čega R 

Who Aux why died 

‘Why did who die?’ 

b. Zbog čega je ko umro? Tanja Samardzic, p.c. 

182

established in section 3.1.4, i.e. cum ‘how’, there is incompatibility between the two 

adjuncts. The result is even worse if de ce precedes cum. As it emerges clear from the 

proposal to be advanced below, this very subtle judgment is expected under a feature-based 

view of RM (Rizzi 2004b). 

e. (?) Când de ce i-ai vorbit ultima dată? 

When why him-have talked to last time 

e. ?* Cum de ce a reparat maşina? 

How why has repaired car.the 

f. * De ce cum a reparat maşina? 

The data seen in the sections above suggest that de ce is merged very high in the clause, 

i.e. above subjects, Topics and Foci. However, the data from wh-phrases indicates that de 

ce is merged in a position lower than that the landing site of wh-phrases. This position is 

labelled ReasonP by Shlonsky and Soare (in progress) (see also Shlonsky 2005). This term 

will continue being used in the sections below. 

5.3. The syntax of de ce. Intervention effects with de ce 

In Rizzi’s proposal, there is an asymmetry involving perché between matrix and embedded 

clauses. Thus, in matrix questions, perché merges in Spec IntP, with no obligatorily 

accompanying verb movement. This correlates with the availability of Foci, Topics, 

(v) a. Ko zasto mnogo trenina? zasto P 

Who why trains a.lot 

‘Why does who train a lot?’ 

b. Zasto ko mnogo trenina? Tanja Samardzic, p.c. 

Given that both Polish and SC pattern alike with respect to the free ordering of wh-phrases in a binary 

question, it may not be surprising that the same free ordering appears with reason (and purpose) why. I do 

not investigate all this any further. 

183

embedded clauses, etc. Conversely, in tensed embedded clauses, perché is moved from the 

embedded clause to matrix Spec FocP, and thus behaves like a normal wh-phrase. Rizzi 

discusses cases like (133) where perché can also be merged in Spec IntP of the matrix 

clause and only the local reading obtains. The asymmetry is thus related to the nature of 

positions, whether they are movement or merge positions. 

(133) Perché QUESTO avremmo dovuto dirgli, non qualcos’altra? 

‘Why THIS we sould have said to him, not something else?’ Rizzi 2001a (23a) 

It is suggested here that the merge position of de ce is low in the CP-layer, Spec ReasonP 

and that it either moves to Spec IntP or AGREE from Int takes place. The properties of de 

ce/why which are distinct from other wh-phrases are put down to its purely wh-nature 

(Rizzi 2001a), whereas all other wh-elements have a [Focus] feature in addition to the [wh] 

feature. Spec FocP is thus a landing-in position, whereas Spec ReasonP is a merge 

position. Under this line of approach Spec IntP may also be a landing-in position. This 

short movement leaves behind a variable. This may make sense given that it takes the 

underlying event as its internal argument, much like a generalised quantifier. 20 

(134) [Int … [Reason [Fin … 

Then the final structure of the left periphery of the Romanian embedded clause looks like 

(135) below 21 : 

20 

The idea that the event itself is treated as a variable goes back to Reichenbach (1947) and was further 

extended by Davidson (1980) and Higginbotham (1995). 

21 

Though the sentence sounds cumbersome, in an embedded context with both Focus and wh in addition to 

de ce, the most natural order is (ia). This is further distributional evidence that in the presence of a wh-phrase, 

the reason adjunct must remain in its merge position. 

(i) a. Vreau să ştiu UN MACINTOSH cui de ce i-ai oferit (şi nu UN PC). 

I want to know A MACINTOSH to whom why you gave (not A PC) 

b. * Vreau să ştiu de ce UN MACINTOSH cui i-ai oferit (şi nu UN PC). 

In embedded contexts, as in matrix ones, de ce is followed by Focus, as in (ii) and is preceded by a whphrase, 

as in (iii). 

184

(135) Force (Top*) IntP (Top*) Foc FocWh ReasonP Fin (final) 

Given these assumptions, let us consider the derivation of (136) repeated below: 

(136) a. De ce LUI ION i s-a cerut demisia (dar nu şi colegilor săi)? 

Why TO ION him s-has asked resignation.the (but not to his colleagues) 

‘Why was ION asked his resignation (but not his colleagues)?’ 

In the absence of a wh-phrase, de ce cannot remain in-situ and must raise to Spec IntP, 

which following Rizzi is intrinsically endowed with the wh-feature. The DP lui Ion gets 

attracted to Spec Foc and no feature clash arises. More generally, when Foc° is the probe in 

the case of wh-phrases which are marked [+wh/Foc], Move is subject to the usual RM 

effects (Rizzi 2004b). However, not having a Focus-feature, but being marked only [+wh], 

de ce undergoes some sort of ‘pure’ wh-movement to Spec IntP across Focus, as shown in 

(136b) for (136a). 

b. … Int … Foc … Reason … [ … ObjIO … vP] 

mvt for Foc/wh 

mvt for wh 

As seen in section 5.2.6, de ce can co-occur with a wh-phrase in this fixed order. Let us 

reconsider the examples below. 

(137) Cine de ce a plecat? 

Who why has left 

(ii) a. Vreau să ştiu de ce SECRETAREI i-a vorbit despre sustragerea de fonduri (şi nu direct 

şefului)? 

I want to know why TO THE SECRETARY he talked about the flight of funds (and not to his boss) 

b. ?? Vreau să ştiu SECRETAREI de ce i-a vorbit despre sustragerea de fonduri (şi nu direct 

şefului)? 

(iii) a. Vrea să ştie cui de ce (i)-a oferit Andrei şampania Moet? 

I want to know whom why (him)-has offered the Moet champagne 

b. ?? Vrea să ştie de ce cui (i)-a oferit Andrei şampania Moet? 

185

(138) 

?* Cum de ce a reparat masina? 

De ce merges in Spec ReasonP. Upon merge of Foc, the wh-phrase cine in (137), is probed 

and moves into Spec FocP. Although the two elements are wh-phrases, no RM effect arises 

as cine is marked [+wh/Foc] and in addition is also an argumental, whereas de ce is only 

[+wh]. Conversely, de ce blocks movement of cum in (138) as both share the [–Arg] 

feature. 

(139) Force 

2 

Int 

2 

Foc 

2 

[uFoc]Reason 

2 

de ce SubjP 

+wh 2 

-Arg cine 2 

+Foc 2 

+wh 2 

+Arg VoiceP 

2 

√ cum vP 

+Foc 

+wh 

-Arg 

* 

The lack of obligatory inversion with de ce (section 5.2.1) follows from the fact that the 

highest position that the Romanian may verb move to is Fin°. Given that in the absence of 

a wh-phrase de ce moves from Spec ReasonP to Spec IntP it follows straightforwardly that 

it comes to precede Topics, Foci and subjects. Recall that in the SV order, the highest 

position a subject DP may fill in Romanian is at the border of the CP-layer and the 

Mittelfeld, where it gets the Topic-aboutness interpretation. This acconts for their 

occurrence below de ce (Soare 2008). 

In the light of these theoretical assumptions let us further provide an account for the 

observations made in section 5.2 related to two-clause de ce sentences. 

In (140), the left periphery of the embedded clause contains a focused DP and the 

copy of de ce. 

186

(140) De ce i-ai spus secretarei [că DIRECTORUL a demisionat]? 

‘Why did you tell the secretary that THE DIRECTOR resigned?’ 

Though judgemnts are very subtle, both a matrix and an embedded reading are available. 

This shows that no intervention effects occur with the lower Focus in long-distance 

construals. This is expected: under the matrix reading, de ce is merged in matrix ReasonP 

and moves to IntP. The verb fills Fin. Hence the apparently obligatory inversion 

phenomenon. In the case of the embedded reading, matrix Foc° probes for de ce which 

moves from embedded Spec ReasonP (possibly to the embedded IntP) across Foc - and for 

the reasons seen above no RM violation is involved – and finally lands in matrix Spec 

FocP. This latter derivation is schematically rendered in (141). 

(141) ForceP 

2 

IntP 

2 

FocP 

2 

2 

[uFoc] ReasonP 

2 

SubjP 

2 

…vP 

2 

ForceP 

2 

IntP 

(3) 2 

FocP 

2 

ReasonP 

2 

(2) SubjP 

2 

(1) …vP 

In (142) Focus is in the matrix clause. This example resembles Rizzi’s (133) discussed in 

section 5.3. As in Italian, embedded scope is excluded. 22 

22 The same interpretation obtains for those speakers not allowing Focus between de ce and the verb. 

187

* 

Foc 

(142) De ce SECRETAREI i-ai spus că directorul va demisiona? 

‘Why did you tell THE SECRETARY that the director will resign?’ 

The reason adjunct cannot move from the embedded clause to matrix Spec FocP as it 

would have to cross the focused object. In other words, matrix Focus has a blocking effect 

on the long-distance movement of de ce. No such effect obtains at the level of the matrix 

clause where the adjunct can move to Spec IntP across Focus. 

Similarly, intervention effects arise with a wh-element in long-distance construals such as 

(143) and (144). The only available reading is the matrix reading of de ce. 

(143) De ce vrei să ştii [cine [a demisionat]]? 

‘Why do you want to know who resigned?’ 

(144) De ce vrei să ştii [cui [i s-a cerut demisia]]? 

‘Why do you want to know who was asked to resigned?’ 

As shown in the derivation below, given that the lower de ce is probed by Foc°, it cannot 

move across cine and cui at the embedded level. The only possible derivation is that under 

which de ce is merged in matrix ReasonP and moves to Spec IntP. 

188

(145) ForceP 

2 

IntP 

2 

FocP 

2 


2 

de ce …SubjP 

2 

…vP 

2 

ForceP 

2 

IntP 

2 

FocP 

2 

cine ReasonP 

2 

de ce FinP 

2 

* SubjP 

… 

In order to get the embedded reading, de ce must surface in the embedded clause thus 

scoping over the embedded event and cine moves to the matrix clause, as in (146a). As 

already explained, movement of cine across de ce is possible as the two are of distinct 

nature (i.e. features). 

(146) Cine vrei să ştii [ de ce [ a demisionat ]]? 

‘Who do you want to know why resigned?’ 

Under the theory developed here it is expected that extraction of a wh out of a chunk 

(containing a wh-adverb) moved to Spec FocP be worse than simple extraction of a wh 

(when de ce is present) (Ur Shlonsky, p.c.). This is indeed borne out by the data below: 

(147) a. ?* Pe cine nu ştii [unde a sărutat Ion aseară]? 

‘Who don’t you know where Ion kissed last evening?’ 

b. ?(?) Pe cine nu ştii [de ce a sărutat Ion aseară]? 

‘Who don’t you know why Ion kissed last evening?’ 

189

(147a) is a case of utter marginality as multiple agree and chunk movement applies, i.e. 

once the chunk in Spec FocP, the wh-argument is further extracted to the matrix CP. 

(147b) is better as the embedded CP, the wh-argument and de ce do not form a chunk. I 

argued elsewhere that Romanian displays a gradient of (un)grammaticality with wh- 

arguments and adjuncts extracted out of a wh-island (Laenzlinger and Soare 2005). 

One of its well-known properties is that the reason adjunct is insensitive to negation. 

This is consistent with the analysis suggested here: Romanian de ce is merged in ReasonP 

from where it raises to IntP. However, negation has a blocking effect in long-distance 

construals. In (148), only the matrix reading obtains. It is virtually impossible to get an 

answer scoping over the embedded clause: Pentru că este foarte bine plătit. ‘Because he is 

very well paid.’ 

(148) De ce nu crede secretara [că (directorul) va demisiona (directorul)]? 

‘Why doesn’t the secretary believe that the director will resign?’ 

As shown below, given that the lower de ce is probed by Foc, it behaves like any whadjunct 

and thus becomes sensitive to negation. The only convergent derivation is the one 

under which it merges in matrix ReasonP. This is schematically rendered below. 

(149) ForceP 

2 

IntP 

2 

FocP 

2 

2 


2 

SubjP 

2 

…NegP 

2 

…vP 

2 

ForceP 

2 

IntP 

* 2 

FocP 

2 

ReasonP 

2 

SubjP 

… 

190

It has been shown thus far that intervention effects arise on the long-distance construal 

with de ce and they confirm to Rizzi’s distinct nature of the IntP-FocP positions, i.e. merge 

vs. movement positions, which consequently behave differently with respect to movement. 

The distinct nature of these positions also correlates with the inversion – non inversion 

phenomenon. It has been claimed that Spanish why behaves the same way as Italian why 

(Contreras 1989 cited by Rizzi 2001a). In this respect the three languages come to be alike. 

To sum up this section, based on evidence from the order of the wh-phrases with 

respect to de ce, the (non-)availability of embedded readings with Neg, Foc and Wh, etc.. it 

has been suggested that de ce is merged low in the split CP, namely in Spec ReasonP, of 

the clause it modifies. From this position it moves to Spec IntP over Foc and no RM effect 

arises as this is an instance of pure wh-movement which does not interact with Focus 

movement. In long-distance construal, de ce behaves like regular wh-phrases, giving rise to 

the usual RM effects. 

6. Conclusions 

This chapter has shown that Romanian fits in well with the cartographic approach. Within 

this framework the left periphery has been investigated. It contains a series of recursive 

Topics, of which Top+WhP is the lowest. Next comes FocP followed by FinP and, 

essentially, no TopP is available between the two. In the proposed analysis wh-phrases 

(focused elements) value their features by moving to the criterial Spec FocP and the Foc°- 

Fin° criterial configuration is formed, thus accounting for the absence of any material 

intervening between V in Fin° and the wh in Spec FocP. The same mechanism is involved 

in the case of (non-wh) focalised elements. One other characteristic property of Romanian 

is the presence of what has been labelled LDCF (left dislocated Contrastive Focus). It is 

suggested to involve movement of Foc° to Top°CLLD which accounts both for its 

interpretation and the obligatory presence of a clitic. 

Another focus of this chapter has been multiple wh-phrases in Romanian and their 

rigid hierarchy. Whereas languages like Navajo or Malayalam pervasively express 

Animacy in their grammars, the Argument > Adjunct split in the Romanian bare whphrases 

has been seen to also follow a residual Animacy hierarchy, [+human] > [-human]. 

More precisely, wh-arguments divide into [+human] and [-human], the latter occupying the 

191

lower portion of the wh-space. The discussion has also unveiled some micro-parameters 

distinguishing Romanian from Bulgarian without pursuing however a deeper understading 

of why this is so: (i) whereas Romanian patterns with other Romance languages in having 

only one Focus position in the left periphery, Bulgarian seems to display more than one; 

(ii) whereas in Romanian, parentheticals, all sorts of particles, etc. cannot occur between 

the wh-phrase and the verb, they can in Bulgarian; (iii) Romanian has one specific position 

to host D-linked wh, which are obligatorily clitic doubled, whereas Bulgarian has two, one 

for the clitic doubled ones, the other for the non-clitic doubled ones; (iv) Romanian allows 

free permutation of D-linked wh’s, whereas Bulgarian does not, i.e. they observe the 

Subject > Object order. An analysis has been suggested both for the bare wh and the Dlinked 

wh. Insofar as the former are concerned, the analysis relies on remnant movement 

and heavy pied-piping of a wh-chunk made up of wh-elements only. The wh-phrases, part 

of the chunk, are multiply agreed derivationally simultenously from Foc°. The chunk 

analysis, which is a last resort operation, can thus account for the multiple wh-movement 

to a sole Foc position in the Romanian left periphery. Within such an analysis, in 

Romanian (and presumably Hungarian) all wh/Foc-features on the wh-phrases have a 

strong value and thus they must move as a chunk to the single Foc position endowed with 

an EPP. Conversely, in Bulgarian all wh-phrases have a similar strong value but move to 

independent Foci. In French on the in-situ strategy wh-phrases do not have a strong 

wh/Foc-feature value and AGREE, with no MOVE, suffices for feature valuation/checking. 

In languages like English and French on the movement strategy the highest wh-element 

has a strong wh/Foc-feature value which overtly moves to Foc endowed with an EPP, the 

other wh’s having a weak value and consequently no overt movement takes place. In 

languages like Serbo-Croatian and Polish one wh-element has a strong wh/Foc-feature 

value hence it moves to Spec Foc, the other wh-elements having a distinct strong feature 

value and move as a chunk to a lower position in the split CP, if one follows Bošković’s 

(1998, 2001) analysis. A language like Japanese (and possibly Chinese) will be argued to 

nonetheless display movement/scrambling of the wh inside the Mittelfeld. This is 

summarized in table 6. 

192

Single overt 

movement to a 

single FocP 

Multiple overt 

movement to 

several FocP’s 

Chunk movement 

to FocP 

Wh-‘scrambling’ 

in the Mittelfeld 

or AGREE 

Romanian Bulgarian French 

in-situ 

English 

Frenchm 

vt 

Serbo- Croatian 

Polish 

Japanese 

(possibly 

Chinese) 

+ + - + + - 

- + - - - - 

+ - - - - - 

- - + - - + 

Table 6 

In the last part of the chapter, the special case of a wh-in-situ with a multiple wh-fronting 

languages like Romanian has been dealt with: this is de ce ’why’. Distinct kinds of 

evidence show that it merges low in the split CP, i.e. in ReasonP, and in the absence of a 

wh, raises to Rizzi’s (2001a) IntP. This movement, accounts for the lack of obligatory verb 

inversion with de ce (recall that the highest position that the Romanian may verb move to 

is Fin°) and the fact that it may precede Topics, Foci and subjects. 

193


CHAPTER V 

JAPANESE WH-SCRAMBLING IN THE MITTELFELD AND 

VORFELD 

Japanese, Korean, Malayalam, and other South Asian languages, are OV languages with 

relatively free word order which have been analysed in the literature as head final languages 

(Saito 1985, 1992, etc., Miyagawa 1997, 2001, etc., Tada 1993, a.o., on Japanese; Choe 1995, 

on Korean, a.o.; Jayaseelan 2004 on Malayalam). However, following the Kaynean (1994) 

antisymmetry hypothesis, such a conception of the clause structure of these languages is 

misleading. This chapter mainly studies the phenomenon of wh-scrambling in Japanese in 

terms of the antisymmetry approach. Before properly dealing with this issue, a few 

observations will be made regarding the Japanese Mittelfeld. Relying on Miyagawa’s (2001, 

2006) and Endo’s (2007) feature-triggered view of (short) ‘scrambling’, it will be seen that 

the Japanese Mittelfeld is a domain rich in functional projections which contain Topic and 

Focus features which are parasitic on phi-features driving movement of DPs and wh-phrases. 1 

1 It is noteworthy that Japanese does not exhibit articles (Bošković and Takahashi 1998). Nonetheless, I continue 

using the term DP throughout but one should keep this idea in mind. 

195

2. A Cartographic Approach to Japanese: Some Implications 

Within the theoretical framework adopted throughout this paper, i.e. the cartographic 

approach to the clause structure, previous analyses of the Japanese clause structure (Saito 

1989, 1992, Miyagawa 1996, 2001, Ura 1996, among many others) are incompatible with 

Kayne’s (1994) antisymmetry theory. To my knowledge, few attempts have been proposed in 

the literature to analyze Japanese in terms of Kayne’s antisymmetry. One case in point is 

Whitman’s (2001) account of negation in Japanese and Korean in terms of verb raising to 

adjoin to Neg, whose complement VP moves beyond NegP in Japanese. Hoshi (2005) 

proposes pied-piping and remnant movements for his analysis of association with Focus in 

Japanese. Endo (2007) provides a detailed analysis of the Japanese clause structure in terms of 

the cartography. Another analysis along the antisymmetry hypothesis is that of Koopman’s 

(2005) who gives a syntactic account of Korean and Japanese morphology by showing that 

inflected words are derived from head-initial structures by phrasal movement and that 

agreement can be triggered under pied-piping. 

In what follows, I mention some implications of the antisymmetric view of Japanese as 

they have been discussed in the literature and which prove relevant to the discussions to 

follow. 

Kayne (2005) suggests that as far as the object position is concerned in an OV language, 

it must have moved to some specifier position higher than V and most plausibly higher than 

certain classes of adverbs in Cinque’s (1999) hierarchy. What these positions are will be the 

purpose of section 3. Therefore, following Kayne’s antisymmetry hypothesis, the structure of 

Japanese clause is the one given in (1). 

(1) TopP 

2 

FocP 

2 

… 

SubjP 

2 

ObjP 

2 

…. 

vP 

2 

DPSubj VP 

2 

V DPObj 

196

Naturally, positing a universal SVO merge implies that not only the object but also the subject 

vacate the vP domain and get attracted to the Mittelfeld licensing phi-, case- and IS-related 

features. Further, as suggested by Kayne (1994:54), in a wh-question, the whole IP, 

corresponding to SubjP in our analysis moves to Spec ForceP (his Spec CP) and thus the final 

position of the question particle no is accounted for, the generalisation being that languages 

with clause-final question particle lack visible wh-movement. This is roughly shown in (2). I 

will come back to this analysis in section 3. 

(2) a. 

b. 

John-ga nani-o katta no? 

John-Nom what-Acc bought Q 

‘What did John buy?’ 

ForceP 

2 

no+EPP FocP 

2 

wh TP 

2 

(2) 2 

ta SubjP 

2 

(1) John-ga ObjP 

2 

nani-o vP 

AGREE 2 

kat … 

Aside from the question marker no, Japanese has many other particles. 2 Kayne (1994, fn 

3:143) suggests that the Japanese particles wa ((non-)contrastive Top/Focus marker) and ga 

(generally, Nominative marker) 3 (and o, Accusative case marker) are overt heads whose overt 

complements follow them. It means then that these heads have an EPP property associated 

with them that create a specifier attracting the DPs in (3). As pointed out by Whitman (2001), 

2 

Besides the question particle no, Japanese has another question particle, ka. Miyagawa (2001, fn.27:332) states 

that as far as no is concerned, there is some debate, i.e. it may be the case that it involves an empty question 

particle. I do not explore this hypothesis here. 

3 

Ga can also be a Focus marker when suffixed to an adverbial and thus able to create RM effects (Endo 2007, 

Vermeulen 2005). 

197

the fact that no other particles, such as wa, mo, or sae in (3) can follow ga and wa strongly 

suggest that these are indeed clausal heads. Conversely, other phrasal particles, such as the 

locative/Dative ni in (4a) can be followed by the association-with-focus particle sae. Another 

association-with-focus particle dake ‘only’ in (4b) can also be followed by the Nominative 

Case particle ga or the Top particle wa. 

(3) a. John-ga-*wa/*mo/*sae kita. 

John-Nom Top also even came 

b. John-wa-*ga/*mo/*sae kita. Whitman 2001 (15) 

(4) a. John-ni-wa/ mo/ sae 

John-Dat-Top also even 

b. John-dake-ga/wa kita. 

John-only Nom Top came taken from Whitman 2001 (16) 

Though the particle ga is generally regarded as the marker for Nominative case in Japanese, it 

may have an additional effect on the interpretation of the phrase to which it is suffixed under 

certain contexts. 

Next I consider one such special use of ga, as illustrated in (5). As thoroughly discussed 

by Vermeulen (2005), among many others, Japanese exhibits possessive and adjunct multiple 

nominative constructions. In (5) and (6), the marker ga is suffixed on the two constituents and 

according to Vermeulen the first one must be obligatorily focused. 

(5) Usagi-ga mimi-ga naga-i. ⇒ possessive multiple Nominative construction 

Rabbit-ga ear-ga long-pres 

‘It is rabbits which have long ears.’ Vermeulen 2005, (1) 

(6) Ano mise-ga gakusee-ga yoku hon-o kau. ⇒ adjunct multiple Nominative construction 

That shop-ga student-ga often book-Acc buy 

‘It is that shop that students often buy books.’ Vermeulen 2005, (2) 

I only focus on the possessive construction here and present some important properties as are 

described by Vermeulen with a view to suggest that the first ga-marked phrase in (5) may 

actually not be Focus but rather Rizzi’s Criterial Subject position. 

198

The DP/NP usagi-ga in (5) is interpreted as a possessor of the following ga-marked 

phrase and the two phrases have a fixed ordering (obviously no such reading is available in 

(6)). Vermeulen claims that the particle ga contains case features, which are relevant to the 

syntax whenever ga marks an NP bearing a theta role, but crucially it may also be interpreted 

as a Focus marker on condition it appears on the first c-commanding NP/DP (the same Focus 

reading is obligatory with the adverbial suffixed by ga; see also a discussion in Endo 2007). 

Furthermore, she argues that the possessive ga-phrase, merged in a high Spec TP behaves like 

a subject and the clause to its right as a predicate, the second ga-phrase filling a lower Spec 

TP. 4 This predication relation in this construction has been described in the literature as an 

aboutness relation (Fukuda 1991). She enumerates some properties that the higher possessive 

subject shares with normal subjects: (i) when embedded, the possessive ga-phrase in 

ECM/control structures can appear with the Accusative o; (ii) when the possessive is 

[+Human] and the second ga-phrase is [-Animate], and if the speaker shows respect, subject 

honorification appears on the verb. These properties show that both ga-phrases are inside the 

TP domain/Mittelfeld. 

In the light of these syntactic and interpretive properties, I would like to suggest that the 

ga-head in examples like (5) is the realization of Rizzi’s (2004) Subject head. Given that the 

higher ga-marked phrases in both (5) and (6) behave differently from each other though 

sharing the same interpretation, and that it is only the ga-phrase in (5) but not in (6) that 

creates a subject-predicate relation, it may be the case that they fill distinct positions, i.e. leftperipheral 

FocP for the adverbial and SubjP for the NP/DP. It may be interesting to note that 

under embedding the exhaustive reading of matrix clauses is no longer the only available one. 

In (7), besides Vermeulen’s Foc interpretation, another is readily available: ‘Even kids know 

that about rabbits, they have long ears.’ The higher ga-phrase is selected or accommodated 

out of context and is asserted something about. 

(7) Usagi-ga mimi-ga naga-i koto-wa kodomo-demo sittairu. 

Rabbit-ga ear-ga long-pres fact-Top kids-even know 

‘Even kids know the fact that rabbits have long ears.’ 

4 She analyses such constructions on a par with the English tough construction. 

199

This latter interpretation, more prone to the Topic-aboutness interpretation of the higher ga- 

phrase, becomes the more easily obtainable interpretation once an adverb is inserted between 

the two DPs. This is illustrated below: 

(8) Usagi-ga chinamini mimi-ga naga-i. 

Rabbit-ga incidentally ear-ga long-pres 

‘Incidentally, rabbits have long ears.’ 

If the same adverb occurs in the adjunct multiple Nominative construction, no change in 

interpretation obtains, i.e. only the Focus one is available (I thank Yoshio Endo for providing 

me with this argument on this contrast). 

(9) Kono mise-ga chinamini gakusei-ga hon-o yoku kau 

This store-Nom incidentally student-Nom book-Acc often buy 

‘Incidentally at this store (, not at other stores) students often buy books.’ 

Once merged in the criterial position, the DP usagi-ga ‘rabbits-ga’ gets frozen in place and 

can no longer move. This correlates with Kuno’s (1973) and Saito’s (1985) observation that 

ga-marked subjects cannot undergo scrambling, especially that subjects suffixed by other 

particles can scramble. 

(10) a. Taroo-ga [usagi-ga mimi-ga naga-i to] omotte iru (koto). 

Taro-ga [rabbit-ga ear-ga long-pres C] thinking is (fact) 

‘Taro thinks that tt is rabbits which have long ears.’ 

b. * Usagi-gai Taroo-ga [ti mimi-ga naga-i to] omotte iru (koto). 

Thus, if this is correct, the ga-head is the realization of both Case and the Criterial Subject. 

3. Short DP ‘Scrambling’ 

In this section the notion of short ‘scrambling’ or A-scrambling i.e. ‘scrambling’ inside the 

Mittelfeld, is briefly investigated and the divide in the approach to this phenomenon is 

presented. The goal is to seek to identify DP ‘short scrambling’ with IS-triggered movement 

200

in the Mittelfeld. Put differently, the attempt is made to put a label on the fairly vague notion 

of ‘scrambling’. 

3.1. ‘Scrambling’: A Definition 

Japanese (and Korean) scrambling is known to have characteristic properties: (i) scrambling 

need not have an effect on the interpretation (Kuroda 1988, Saito 1989, 2004, Fukui 1993, 

Saito and Fukui 1998, a.o., but see discussions below); (ii) it freely applies long-distance 

(long-scrambling), but does not establish an operator-variable relation; (iii) as exemplified in 

(11) below, a phrase moved clause internally can serve as an A-binder but one moved by 

long-scrambling cannot; (iv) there is an A/A’ asymmetry with clause-internal scrambling 

(short scrambling) and long-scrambling (Mahajan 1990, Tada 1990, Saito 1989 1992, contra 

Saito 2003); (v) long-scrambling exhibits the so-called radical reconstruction property, i.e. a 

moved phrase is interpreted in its original merge, i.e. it is ‘undone’ at LF. 

Japanese A-‘scrambling’ is known to be locality-free. As shown in (11), the object 

lexical anaphor karera-o ‘they’ can A-scramble over the subject and serves as an A-‘binder’. 

The object DP thus c-commands otagai ‘each other’ with which it is coindexed. 

(11) ?[Karera-oi [[otagai-no sensei]-ga ti hihansita]] (koto). 

They-Acc each other-Gen teacher-Nom criticized (fact) 

‘Themi, [each other’s teachers] criticized ti.’ Saito 2003, (9) 

However, distinct analyses have recently challenged the notion of optionality, characteristic of 

Japanese scrambling (Miyagawa 2005, 2006a, 2006b, Ishihara 2000, Bošković & Takahashi 

(henceforth B&T) 1998, Bošković 2004, etc). Ishihara (2000), for instance, claims that 

scrambling has a semantic/pragmatic contribution, i.e. it creates new focus domains that are 

not available with a different word order. The syntactic component yields a syntactic 

derivation to two interface levels, PF, the level of stress assignment, and LF, the locus of 

calculation of the focus domain in a scrambled or non-scrambled case. Adopting a head-final 

analysis of the Japanese phrase structure, Miyagawa (2001) claims that scrambling is a 

syntactically obligatory operation driven by the EPP feature on T and assumes following 

Chomsky (1998) that case features are checked in situ by virtue of AGREE. Thus both the 

subject and the object can stay in situ. However, the optionality of scrambling is reduced to 

201

optionality of either DP, subject or object, to move for the EPP on T. 5 Consider the examples 

below, which are Miyagawa’s (2001) examples (11): 

(12) a. Zen’in-ga sono tesuto-o uke-nakat-ta (yo/to omou). 

All-Nom that test-Acc take-Neg-Past (Excl/Comp think) 

‘(I think that)All did not take that test.’ 

*not>all, all>not 

b. Sono tesuto-o zen’in-ga t uke-nakat-ta (yo/to omou). 6 

That test-Acc all-Nom t take-Neg-Past 

‘All did not take the test.’ 

not>all, (all>not) 

The subject, zen’in-ga in (12a) scopes over negation and fills Spec TP (in his analysis). When 

the object is scrambled and the OSV order obtains, the subject zen’in-ga in (12b) need not 

move, occupying Spec vP. On the assumption that Japanese has V-to-T raising, Miyagawa 

explains this subject-object asymmetry in terms of the equidistance of both DPs with respect 

to T. Note, however, another reading is possible, that in which the subject ‘all’ is outside the 

scope of negation. This reading results when the subject moves to Spec TP (in his analysis) to 

satisfy EPP, and the object then undergoes A’-scrambling for focus. 

In a later study and in a vein similar to Ishihara (2000), Miyagawa (2005) claims that the 

driving force of short overt scrambling in Japanese is focus, i.e. scrambling alters the focus 

potential of a sentence. More generally, he claims that languages are parametrised as to 

agreement/focus, i.e. there are agreement-prominent languages, like English – EPP on T 

raises an agreeing phrase to Spec TP - and focus-prominent languages, like Japanese - EPP on 

T raises a DP with focus to Spec TP. 

Endo (2007) shows that, when scrambling is focus-driven, it does not target Spec TP, 

but Rizzi’s (1997) FocusP in the left periphery. He further shows that it is also for the Subject 

of predication feature at the upper Mittelfeld border that DPs scramble. One more alternative 

is that a DP may move to TopicP in the left periphery. 

5 

Miyagawa (op. cit.) actually argues that even a wh-PP (in the wh SV order) can satisfy EPP. 

6 

In a footnote, Miyagawa states that for some speakers the distinction in scope interpretation becomes neater if 

the tense is non-past. 

202

3.2. Short DP Scrambling in SOV and IS features 

An alternative analysis following the anti-optionality trend is suggested in this section. Given 

the antisymmetric SVO merge order followed by subsequent subject and object movement to 

the Mittelfeld, the claim to defend is that the EPP feature driving (full) DP movement is 

related to IS-feature valuation. More precisely, it is claimed that in Japanese where IS is 

closely interwoven with morphological and prosodic phonological structure, the Mittelfeld 

contains Topic and Focus features, which are parasitic on distinct phi/Case projections. 

Within the model adopted here, in the Numeration, besides Case and phi-features an object 

DP, for instance, also has an (interpretable) IS-feature, be it Top or Foc. We also suggest that 

the Mittelfeld contains an ObjP whose head has Case and phi features as well as an OCC/EPP 

feature which selects a Goal with the appropriate feature make-up, i.e. Accusative case, phi 

and an iIS-feature and attracts it to its specifier position. 

There are two main claims concerning the surface position of the subject in Japanese. 

One is that subjects never move/scramble from their merge position, and remain in situ in 

syntax (Saito 1985, Kuroda 1988, Fukui 1986). The other, on the contrary, is that subjects do 

move/scramble to, i.e. Spec IP/Spec TP, in syntax (Ueda 1993, Miyagawa 2001, Endo 2007, 

a.o.; see also Ko 2007 for Korean but with implications for Japanese). Relying on questionsanswer 

possibilities and DP-adverb interactions the latter claim will be upheld here. 

Consider the SOV order below: 

(13) Taroo-ga hon-o katta. 

Taroo-Nom book-Acc bought 

‘Taroo bought a book.’ 

(13) is a possible answer to the questions in (14) below (Miyagawa 2005, Ishihara 2000). 

(14) a. Nani-o atta no? 


b. Taro-ga nani-o sita no? 

‘What did Taro do?’ 

c. Taro-ga nani-o katta no? 

‘What did Taro buy?’ Ishihara 2000 (26) 

203

The question possibilities in (14) to which (13) is an answer may be reinterpreted the 

following way: firstly, given that (13) constitutes a possible answer to (14a), then the latter 

constitutes Focus, which necessarily contains the main stress of the sentence (Ishihara 2000). 

It is plausible to analyse the subject DP Taro-ga as filling Rizzi’s (2004) criterial Subject 

position, i.e. it functions like Topic-aboutness. The object DP in (13) occupies ObjP. 

Secondly, (13) viewed as an answer to (14b), contains the VP [hon-o katta] as the focus 

domain which includes the main stress in the sentence. Structurally, one analysis is that the 

ObjP containing the both the object DP and the verb pied-piped to the higher position in the 

Mittelfeld associated with FocPNewInformation. Thirdly, (13) also constitutes a possible answer to 

(14c) and the object DP in his analysis bears a special stress, typical of narrow focus. 

Couched in cartographic terms, the DP Taro-ga is some sort of given information and fills 

therefore some Spec SubjP, whilst the object DP moves to SpecObjP, which is necessarily 

associated with FocPNewInformation in the sense of Belletti (2001, 2004) though a Voice adverb 

may follow the object. 7,8 

7 An alternative explanation in terms of Chomsky’s (1998, 2001) analysis of Object Shift has been proposed for 

Japanese scrambling (Ishihara 2000). Whether the object shifts or not depends on whether v has an EPP feature 

to check, as in the case of Object Shift. However, such an approach with multiple specifiers is antiantisymmetric. 

Besides, as will be seen in section 3, the object may move quite high in the Mittelfeld. 

8 

Partly based on the contrast in (i), Miyagawa (1997) identifies a Contrastive Focus position in the Mittelfeld. 

The object in (i) marked with the contrastive particle wa must be higher than the adverb quickly. It is focus that 

is the trigger of object movement out of vP/VP. 

(i) a. ?? John-ga isoide hon-WA katta. 

John-Nom quickly book-Contrast bought 

‘John quickly bought A BOOK.’ 

b. John-ga hon-WA isoide t katta. 

Importantly, an object may also move to this position even if it is not wa-marked (Miyagawa 1997, fn. 18). One 

can be more precise as to the exact position of Contrastive Focus. Consider in this respect (ii) below: 

(ii) a. ?? Hanako-ga umaku/tyuuibukaku hon-WA yonda. 

b. 

Hanako-Nom well/carefully book-WA read 

Hanako-ga saikin hon-WA yonda. 

Hanako-Nom recently book-WA read Endo (p.c.) 

Examples (iia,b) and (ia) above suggest that the Japanese Mittelfeld exhibits a Mittelfeld-internal CF position 

which is in its low portion, more specifically above low adverbs like celeretive II and voice adverbs. This 

204

Evidence that the subject and object DP leave the vP domain in Japanese and target 

distinct positions in the Mittelfeld comes from the interaction of a subject/object DP and 

distinct classes of adverbs within Cinque’s (1999) functional hierarchy. In order to show that 

the subject DP in the SOV order fills a higher position in the Mittelfled, not its merge 

position, Ono (2001) cites Ueda (1993) and Miyagawa (1989) who show that Mood, T (and 

locative) adverbs may intervene between the subject and its quantifier and that internal 

arguments, and crucially for us, Voice adverbs cannot. 9 This is illustrated below: 

(15) a. Gakusei-ga unwaruku sannin sono ziko-ni atta. 

Students-Nom unfortunately three that accident met 

‘Unfortunately, three students were involved in that accident.’ 

b. Gakusei-ga kinoo sannin hon-o yonde-ita. 

Students-Nom yesterday three book-Acc were reading 

‘Yesterday, three students were involved in that accident.’ Ono2001,(7) 

c. ??/* Gakusei-ga isoide sannin pizza-o tabeta. 

Students-Nom quickly three pizza-Acc eat-Past 

‘Three students quickly ate pizza.’ Endo (p.c.) 

In other words, this paradigm shows that the subject DP may move at least to a position above 

the celerative adverb in (15c). The representation of the subject DP floating possibilities is 

provided in (16). The double bar indicates the impossibility of a DP to occur in that particular 

position. 

explains the grammaticality of the object DP to the right of recently in (iib). It is noteworthy that the wa-marked 

object may move across the ga-subject to Spec FocP in the left periphery (see discussion in Endo 2007). 

Vermeulen (op.cit, fn9:1336) makes a similar point. 

9 Kuroda (1980) and Haig (1980) claimed that the subject classifier and its associate DP require strict locality, 

i.e. the classifier cannot be separated from the DP. One of several counterexamples is the presence of an adverb 

between the two. 

205

(16) SubjP 

3 

DP MoodP 

3 

Unwaruku SubjP 

3 

DP TP 

3 

Kinoo SubjP 

3 

DP AspPcelerative 

3 

Isoide vP 

… 

Let us further consider the interactions between adverbs and the object DP in (17): 

(17) a. Gakusei-ga sono mondai-o subayaku toita. 

A student-Nom the problems-Acc quickly solved 

b. Gakusei-ga sono mondai-o umaku toita. 

A student -Nom the problems-Acc well solved 

c. ? Gakusei-ga sono mondai-o saikin toita. 10 

Student-Nom the problem-Acc recently solved 

d. Gakusei-ga sono mondai-o kinoo toita. 

Student-Nom the problem-Acc yesterday solved 

e. * Gakusei-ga sono mondai-o koounnimo toita. 

Student-Nom the problem-Acc fortunately solved Endo (p.c.) 

The paradigm in (17) shows that the object moves to a position above the celerative adverb 

subayaku ‘quickly’ in (17a) and the manner adverb in (17b) and can move even higher, i.e. 

above T adverbs in (17c,d) but cannot move above the high evaluative adverb, as shown in 

(17e). Movement of the definite object above these adverb classes may be suggested to be 

driven by specificity. If the same tests are made with an indefinite object, i.e. sono is dropped, 

there is a degradation with (18a,b). On a par with (17e), (18c) is independently ruled out. I 

thank Yoshio Endo for pointing this out to me. 

(18) a. ? Chomsky-ga ronbun-o saikin kaita. 

Chomsky-Nom paper-Acc recently wrote 

10 The slight marginality of (17c) does not reproduce with other types of predicates. 

206

‘Chomsky has recently written a paper.’ 

b. ? Chomsky-ga ronbun-o kinoo kaita. 

Chomsky-Nom paper-Acc yesterday wrote 

‘Chomsky wrote a paper yesterday.’ 

c. * Chomsky-ga ronbun-o koounnimo kaita. 

Chomsky-Nom paper-Acc fortunately wrote 

‘Fortunately Chomsky wrote a paper.’ 

If no movement of the indefinite object takes place, such sentences become perfect, as 

illustrated below. This shows that object positions in the SOV order are not always the same. 

(19) a. Chomsky-ga saikin ronbun-o kaita. 

Chomsky-Nom recently paper-Acc wrote 

‘Chomsky has recently written a paper.’ 

b. Chomsky-ga kinoo ronbun-o kaita. 

Chomsky-Nom yesterday paper-Acc wrote 

‘Chomsky wrote a paper yesterday.’ 

c. (Koounnimo) Chomsky-ga (?koounnimo) ronbun-o kaita. 

(Fortunately) Chomsky-Nom (?fortunately) paper-Acc wrote 

‘Fortunately Chomsky wrote a paper.’ 

3.3. Short DP Scrambling in OSV and IS features 

Let us next consider the scrambling possibilities in the OSV order. Consider example (20). 

(20) Hon-o Taro-ga katta. 

(20) constitutes an appropriate answer to the questions in (21) following Miyagawa (2005) 

and Ishihara (2000): 

(21) a. Dare-ga hon-o katta no? 

‘Who bought the book?’ 

b. Nani-ga atta no? 


207

c. * Taro-ga nani-o sita no? 

‘What did Taro do?’ 

In (21a), the DP hon-o represents given information, and thus fills an ObjP associated with 

the old information, some sort of Mittelfeld-internal Top-feature in the sense of Belletti 

(2004). The subject DP Taro-ga moves to a low SubjP position necessarily associated with 

FocNewInformation in the Mittelfeld. What is interesting is that OSV may also function as an 

answer to an out-of-the-blue question in (21b). However, OSV is more natural if mention of 

the object has been made in the previous context or whether it functions as a Hanging Topic 

and therefore an intonation break cuts it off from the rest of the sentence (Yoshio Endo, p.c). 

This latter reading is not insisted upon here. To illustrate the topic-like character of the 

scrambled object consider the contexts below. 

Context: Mary and her friend know that there is going to be a race car today in which several 

cars will take part. Mary has to leave before the race starts and later on she asks John. 

(22) M: Nani-ga atta no? 

What-Nom happened Q 

J: ?*% Ferrari-o Sauber-ga oikosita. 

Ferrari-Acc Sauber-Nom overtook 

‘The Sauber overtook the Ferrari.’ Endo (p.c.) 

J’s answer sounds strange and it is the SOV word order which sounds natural. However, OSV 

may sound acceptable to some other speakers simply because the Ferrari car is evoked in the 

car racing context and Ferrari is a famous racing car. Even in this case, there is therefore a 

presupposition of cars (and of possible accidents taking place). Therefore, the object DP has a 

Topic-feature. 

Consider next the following example (23). 

Context: Mary and her friend John meet up in the street, John looks a bit pale. 

(23) Q: Doo sita nodesu ka 

How did polite Q 

‘What happened (to you)?’ 

A: ?* $300,000 syain-ga otosita. 

$300,000 office.staff-Nom lost 

Endo (p.c.) 

208

‘One of the staff member lost $300,000.’ 

Even within an accommodating context, a cardinality answer like (23), which is incompatible 

with presuppositionality, sounds strange (Yoshio Endo, p.c.). Similarly within the context 

below, a cardinality answer under the OSV form sounds strange, whereas the SOV order 

would be prefect. 

Context: Three kids who are working on their homework in the room and their father comes 

in and asks: 

(24) Q: Nani yatteru no? 

What doing Q 

A: ?* Yatto 30-page-o Taro-ga owatta. 

Finally page 30-Acc Taro-Nom finished 

‘Taro finally finished with the problem on page 30.’ Endo (p.c.) 

However, the OSV in (24) is fine in the context in (25) which includes presuppositionality, 

(i.e. problems from page 30 to page 40): 

Context: The three kids have problems to solve from page 30 to page 40 and the father asks: 

(25) Q: Nani yatteru no? 

What doing Q 

A: Yatto 30-page-o Taro-ga owatta. 

Finally page 30-Acc Taro-Nom finished 

Thus OSV cannot be an answer to (21b) unless some presuppositionality is involved. 

Let us go back to (20) repeated below: 

(20) Hon-o Taro-ga katta. 

Besides the interpretations above, (20) may have others under the appropriate stress 

conditions. Thus, the object DP may also function as a Topic, thus filling Rizzi’s Spec Top in 

209

the left periphery (Endo p.c.). In such a case, an intonational break follows the DP with the 

remaining part of the sentence constituting the comment regarding the book. Furthermore, if 

the object DP is read with a distinct specific stress, it can undergo movement to a Contrastive 

FocusP in the left periphery (Endo p.c.). In all these last three cases, the subject DP fills SubjP 

associated with new information Focus. 

(20) with the moved object cannot be a felicitous answer to (21c) which requires that all 

elements associated with New Information must be adjacent and not separated by another 

which is given information. The appropriate answer would be one under the form of SOV. 

Insofar as the positions of the object DP in the OSV order is concerned, we make 

recourse to the same test by using distinct classes of adverbs, i.e. an AdvevaluativeMood like 

unwaruku ‘unfortunately’, AdvepistemicMode like tabun/osoraku ‘perhaps’, an AdvTime like kinoo 

‘yesterday’, and saikin ‘recently’ and an AdvceleretiveAsp like isoide ‘quickly’. 

(26) a. ??(?) Sono ziko-ni unwaruku gakusei-ga ate. 

That accident unfortunately students-Nom met 

‘Unfortunately the students were involved in that accident. 

b. ??(?) Sono ziko-ni tabun/osoraku gakusei-ga atta. 

That accident perhaps students-Nom met 

‘Perhaps the students were involved in that accident.’ 

c. ? Sono ziko-ni kinoo gakusei-ga ate. 

That accident yesterday students-Nom met 

‘Yesterday the students were involved in that accident.’ 

d. Sono ziko-ni saikin gakusei-ga ate. 

That accident recently students-Nom met 

‘Recently the students have been involved in that accident.’ 

e. Sono cake-o isoide gakusei-ga tabe-ta. 

The cake-Acc quickly student-Nom ate 

‘The students quickly ate the cake.’ Endo (p.c.) 

In (26a) the DP moves across the fixed position of the evaluative adverb and the sentence 

sounds strange. Similarly, (26b) indicates that the DP cannot move higher than an epistemic 

adverb. However, with lower adverbs in Cinque’s hierarchy the DP can float among such 

positions (26c,d,e). This is given in (27), where the double bar points to the impossibility of a 

DP to occur in that particular position. 

210

(27) ObjP 

3 

DP MoodP 

3 

unwaruku ObjP 

3 

DP ModePepistemic 

3 

tabun ObjP 

3 

DP TP 

3 

kinoo ObjP 

3 

DP AspPcelerative 

3 

isoide vP 

… 

To summarize, based on question-answer possibilities and the interpretation possibilities, on 

the one hand, and relying on subject and object interactions with distinct adverb classes, this 

section has suggested that the Japanese Mittelfeld contains Topic and New Information Focus 

features responsible for driving argument movement. The subject in the SOV order can move 

to a position as high as TP and mood adverbs although it cannot sit immediately above a 

celerative aspect adverb. The object in SOV can also vacate the vP and move to a position 

below high adverbs but not higher. In the OSV order it ‘floats’ to distinct positions inside the 

Mittelfeld but must remain lower than high adverbs. 

4. Japanese Wh-Phrases: Mittelfeld-internal ‘Scrambling’ 

This section turns to the behaviour of Japanese wh-phrases. It is a common property of many 

SOV languages to have wh-in-situ. In some, a wh-phrase must show up in a position 

immediately to the left of the verb, linearly speaking. Thus, Jayaseelan (2004) argues that in 

Malayalam, another superficial SOV language, a wh-phrase fills Spec FocP immediately 

above vP. This is illustrated below: 

(28) a. Nin-ne aarə talli? Jayaseelan 2004 (1) 

You-Acc who beat(Past) 

211

‘Who beat you?’ 

b. * Aarə nin-ne talli? 

The wh-phrase aarə ‘who’ must move to a position immediately at the periphery of vP, 

namely Spec FocusP. In his analysis which incorporates Chomsky’s (2001) PIC, movement to 

Spec FocP is driven by the need of a question operator to access the wh-phrase. It is quite 

generally acknowledged that in OV languages the position to the left of the verb is a Focus 

position. It has been seen in section 3 above that in non-wh-clauses, Japanese exhibits a 

projection which is associated with a New Information Focus feature. 

Another OV language which does not overtly move the wh-phrase to the left periphery 

is Bangla (Bengali), an Asian Indo-Aryan language. Bhattacharya and Simpson (2003) argue 

in favour of an underlying SVO analysis and claim that the classical wh-in-situ cases should 

be reanalysed as instances of overt movement below CP, more precisely, below the subject 

position. Thus, in (29), movement is argued to take place overtly: the embedded CP[wh] is 

pied-piped from its postverbal base position to the post-subject position. (To them this is 

actually wh-movement to the post-subject position). 

(29) JOn [ke cole gEche] bollo. 

John [who left gone] said 

Bangla 

‘Who did John say left?’ Bhattacharya & Simpson 2003, (7) 

The assumption has been made in Chapter III that wh-movement is uniformly and universally 

to the left periphery in languages like English and Romanian. The claim to entertain with 

respect to the so-called wh-in-situ languages is that languages like Japanese (and Bangla for 

that matter) allow short movement of the wh-phrase not to the left periphery, but inside the 

Mittelfeld. As will be seen in section 5, at least as far as Japanese is concerned, a wh-phrase 

can also target a position in the left periphery but this is not wh-movement. It is movement 

driven by topicality. 

4.1. Are Wh-Phrases Really In-situ or Do They Move? 

212

One aim of this section is to provide some arguments in favour of the idea that in matrix 

clauses a wh-phrase can move/scramble Mittelfeld-internally. This is therefore A-movement. 

Another is to answer the question of whether ‘scrambling’ of a lower wh-phrase above 

another results in any systematic semantic or pragmatic effect. The investigation starts from 

Miyagawa’s (2006a) example (27) who shows that when the object wh-phrase is ‘scrambled’ 

over the subject wh-phrase, the most natural way to interpret the object is to presuppose a “set 

of objects, and for each object, one is supposed to return the answer of who bought it”. In 

other words, the ‘scrambled’ object gets a D-linking reading. 

(30)? Nani-oi John-ga [Taroo-ga ti katta ka] siritagatteiru? 

What-Acc John-Nom Taro-Nom t bought Q want.to.know 

‘John wants to know what Taro bought.’ Miyagawa 2006a, (27) 

4.1.1 The Interaction of Wh-Phrases and Adverbs 

As argued for in Kayne (2005), the antisymmetry claim holds that Japanese objects must end 

up in or within a position c-commanding the verb. The test with Cinque’s (1999) classes of 

functional adverbs proves relevant in identifying these landing sites. 

(31) a. Hanako-ga [subayaku] nani-o sita no? 

Hanako-Nom quickly what-Acc did Q 

‘What did Hanako quickly did?’ 

b. Hanako-ga [umaku] nani-o honyakusita no? 

Hanako-Nom well what-Acc translated Q 

‘What did Hanako well translated?’ 

c. Hanako-ga [sibasiba/hinpanni] nani-o sita no? 

Hanako-Nom often what did Q 

‘What did Hanako often do?’ 

d. Hanako-ga [saikin] nani-o sita no? 

Hanako-Nom recently what-Acc did Q 

‘What has Hanako recently done?’ 

e. Hanako-ga [kinoo] nani-o sita no? 

Hanako-Nom yesterday what-Acc did Q?’ 

‘What did Hanako do yesterday?’ 

213

f. Hanako-ga [tabun/osoraku] nani-o sita no? 

Hanako-Nom probably what-Acc did Q 

‘What did Hanako probably do?’ 

g. Hanako-ga [kooun-ni-mo] nani-o sita no? 

Hanako-Nom fortunately what-Acc did Q 

‘What did Hanako fortunately do?’ 

The sentences in (31b-g) indicate that the object wh-phrase remains below all these adverb 

classes and that the subject wh-phrase may move very high, i.e. above the evaluative adverb, 

as in (31g). On the antisymmetric analysis adopted here, even nani-o in (31a) may be 

analyzed as having vacated the vP domain and having moved to an Object-related position at 

the vP border. Based on the data below, subayaku ‘fast/early’ fills Cinque’s lower celerative 

Aspect projection. 

(32) a. John-wa tyuuibukaku subayaku toori-o watatta. 

John-Top carefully fast street-Acc crossed 

‘John crossed the street carefully fast.’ 

b. * John-wa subayaku tyuuibukaku toori-o watatta. 

Given that only the order in (32a) is the only available one, subayaku must fill the lower 

celerative AspP, i.e. the adverb quantifies over the process of crossing the street. This adverb 

may co-occur with hayaku ‘quickly’, and as the contrast in (33) shows, it fills the higher 

celerative AspP 11 

(33) a. Kyoo John-ga hayaku subayaku okita. 

Today John-Top quickly early woke up. 

‘Today John woke up early quickly.’ 

b. * Kyoo John-wa subayaku hayaku okita. 

Let us consider the cases below showing movement of the object across the same adverbs. 

11 Both subayaku and hayaku may be translated as ‘quickly’ but for expository reasons we provide glosses 

corresponding to their structural positions. 

214

(34) a. (?) Hanako-ga nani-o subayaku kaiketusita no? 

Hanako-Nom what-Acc quickly/early solved Q 

‘What did Hanako quickly deal with?’ 

b. (?) Hanako-ga nani-o umaku kaiketusita no? 

Hanako-Nom what-Acc well solved Q 

‘What did Hanako well deal with?‘ 

c. (?) Hanako-ga nani-o sibasiba/hinpanni sita no? 

Hanako-Nom what often did Q 

‘What did Hanako often do?’ 

d. (?) Hanako-ga nani-o saikin sita no? 

Hanako-Nom what-Acc recently did Q 

‘What has Hanako recently done?’ 

e. (?) Hanako-ga nani-o kinoo sita no? 

Hanako-Nom what-Acc yesterday did Q 

‘What did Hanako do yesterday?’ 

f. ?* Hanako-ga nani-o tabun/osoraku sita no? 

Hanako-Nom what probably did Q 

‘What did Hanako probably do?’ 

g. ?* Hanako-ga nani-o kooun-ni-mo sita no? 

Hanako-Nom what-Acc fortunately did Q 

‘What did Hanako fortunately do?’ 

h. ?* Hanako-ga nani-o sinsetu-ni-mo sita no? 

Hanako-Nom what-Acc kindly did Q 

‘What did Hanako kindly do?’ 

The sentences (34a-e) suggest that scrambling of the object wh-phrase targets distinct 

positions above the lower celerative Asp adverb but it cannot move very high in the 

Mittelfeld, i.e. it cannot move above epistemic and evaluative adverb projections (34f) and 

(34g). Expectedly, a subject-oriented adverb like sinsetu-ni-mo ‘kindly’ cannot be crossed 

over. The object wh-phrase exhibits the same movement possibilities as the non-wh object 

DP. 

4.1.2 Canonical and scrambled wh-phrases. The issue of optionality 

215

As already mentioned, I follow the feature-driven view of short scrambling, i.e. scrambling in 

the Mittelfeld. Miyagawa (2005) shows that whereas short scrambling of DPs may affect the 

focus potential of the sentence, short scrambling of wh-phrases leads to new interpretations, 

i.e. a D-linking reading of the scrambled wh-phrase. Therefore, at least with short scrambling 

there is no optionality in Japanese, contrary to Saito (1989). In what follows I bring more 

evidence that this is a systematic fact and that the Mittelfeld contains a high D-linked 

projection hosting scrambled wh-elements. 

I will first consider the ordering of wh-adjuncts. 

(35) a. Itu doko-de Hanako-ni atta no? 12 

When where-at Hanako-Dat saw Q 

‘When did you see Hanako where?’ 

b.?? Doko-de itu Hanako-ni atta no. Ok under a D-linked reading 

(36) a. Itu donoyoo-ni Hanako-ga hataraita no? 

When how-Dat Hanako-Nom worked Q 

‘When did Hanako work how?’ 

b.?? Donoyoo-ni itu Hanako-ga hataraita no. Ok under a D-linked reading 

(37) a. Doko-ni donoyoo-ni Hanako-ga tyoosya sita no? 

Where-to how Hanako-Nom parked-his-car Q 

‘Where did Hanako park his car?’ 

b.?? Donoyoo-ni doko-ni Hanako-ga tyuusyasita no. Ok under a D-linked reading 

The sentences in (35a), (36a) and (37a) show that Cinque’s (1999) hierarchy of adverbs is 

observed in a natural way of asking the question corresponding to their canonical order. It is 

to observe that adverbs like doko-de ‘where’ and donoyoo-ni ‘how’ are adverbial PP’s. 

However, scrambling of the adverbs in the (b) sentences provides somewhat awkward 

sentences but, importantly, they become acceptable under a D-linked interpretation of the 

locative and of the manner adverbs. 

12 Here and elsewhere examples using both doko-ni and doko-de are given. Whatever the particle used, the 

judgments remain the same. 

216

The canonical order of the wh-adverb and adverbials is therefore summarized in (38): 

(38) itsu ‘when’ > doko-ni/doko-de ‘where’ > donoyoo-ni ‘how’ 

The examples below test the order of wh-arguments. 

(39) a. Dare-ga dare-ni atta no? 

Who-Nom whom-Dat met Q 

‘Who met whom?’ 

b. ?(?) Dare-ni dare-ga atta no? Ok under a D-linked reading 

(40) a. Dare-ga dare-o tataita no? 

Who-Nom who-Acc hit Q 

‘Who hit who?’ 

b. ?(?) Dare-o dare-ga tataita no? Ok under a D-linked reading 

(41) a. Dare-ni dare-o Ken-ga syookaisita no? 

Whom-Dat who-Acc Ken-Nom introduced Q 

‘Who did Ken introduce to whom?’ 

b. ?(?) Dare-o dare-ni Ken-ga syookaisita no? Ok under a D-linked reading 

The sentences in (b) above do not represent a natural way of asking a question in an out of the 

blue context. However, if an appropriate context is provided, they improve considerably, 

hence their D-linked interpretation under scrambling. 

(42) dare-ga ‘who.Nom’ > dare-ni ‘whom’ > dare-o ‘who.Acc’ 

Similar observations can be made with respect to the order of wh-arguments marked for the 

features [+/-Animate]. More exactly, if the Accusative [-Animate] nani-o in (43b) scrambles 

above the Nominative [+Animate] dare-ga, the sentence sounds rather unnatural but it 

becomes acceptable under a given context where speaker and hearer share a set of known, 

say, papers presented at the conference they attended, i.e. the scrambled nani-o has a D-linked 

reading. Given that dare-ga precedes dare-ni, it is expected that the latter also precedes nani- 

217

o, and this is the case in (44). (43b) and (44b) become acceptable under a D-linked 

interpretation of the scrambled wh-argument. 

(43) a. Dare-ga nani-o kaigi-de happyoo sita no? 

Who-Nom what-Acc presented at conference Q 

‘Who presented what at the conference?’ 

b. ?(?) Nani-o dare-ga kaigi-de happyoo sita no. Ok under a D-linked reading 

(44) a. Dare-ni nani-o kimi-ga tanonda no ? 

Whom-Dat what-Acc you-Nom asked Q 

‘Whom did you ask what?’ 

b. ?(?) Nani-o dare-ni kimi-ga tanonda no. Ok under a D-linked reading 

Therefore, the [-Animate] nani-o follows the sequence of [+Animate] wh-arguments: 

(45) dare-ga ‘who’ > dare-ni ‘whom’ > dare-o ‘who.Acc’ > nani-o ‘what.Acc’ 

Further, let us consider the order of [+/-Animate] arguments wrt adjuncts. As was the case 

before, scrambling of the adverb across the [+Animate] argument in (46b), (47b), (48b) and 

(49b) leads to awkward results unless an appropriate context is involved, in which case the 

scrambled element gets a D-linked reading. Even in the case of the D-linked donoyoo-ni in 

(49b), one may imagine a context where ‘you can meet person X at the airport emotionally, 

nervously, joyfully, unhappily, etc’. 

(46) a. Dare-ga itu neru no? 

Who-Nom when will.sleep Q 

’Who will sleep when?’ 

b. ?(?) Itu dare-ga neru no. Ok under a D-linked reading 

(47) a. Dare-ga doko-de neru no? 

Who-Nom where-at will.sleep Q 

’Who will sleep where?’ 

b. ?(?) Doko-de dare-ga neru no. Ok under a D-linked reading 

(48) a. Dare-ni itu sono yoi sirase-o osieta no? 

218

Whom-Dat when the good news gave Q 

‘Whom did you give the good news?’ 

b. ?(?) Itu dare-ni sono yoi sirase-o osieta no. Ok under a D-linked reading 

(49) a. Dare-ni donoyoo-ni kuukoo-de au no ? 

Whom-Dat how-Dat airport-at see Q 

‘Whom will you meet how at the airport?’ 

b. ?(?) Donoyoo-ni dare-ni kuukoo-de au no. Ok under a D-linked reading 

Let us consider [-Animate] wh-arguments. The time adverb in (50) precedes the low [- 

Animate] Nominative wh-argument in a natural way of asking the question but if scrambling 

of the latter applies, it gets a D-linked reading. Given the wh-adverb order in (38) above, and 

the fact that nani-ga ‘what’ fills a low position in the wh-domain (possibly competing for the 

same position as its accusative counterpart), it is expected that the wh-adjunct doko-de 

‘where’ precede the argument. This is shown in (51). However, both the nominative wh- in 

(42) and accusative wh-arguments in (53) precede the manner adjunct donoyoo-ni ‘how’. As 

was systematically observed before, scrambling of the wh-argument across the adjunct 

induces a D-linked reading of the argument. 

(50) a. Itu nani-ga okotta no? 

When what-Nom happened Q 

‘What happened when? ’ 

b. ?(?) Nani-ga itu okotta no. Ok under a D-linked reading 

(51) a. Doko-de nani-ga okotta no? 

Where-at what-Nom happened Q 

‘What happened where?’ 

b. ?(?) Nani-ga doko-de okotta no . Ok under a D-linked reading 

(52) a. Nani-ga donoyoo-ni okotta no? 13 

What-Nom how-Dat happened Q 

13 

One may argue that the order in (52) is biased as the adverbial-PP donoyoo-ni ‘how-Dat’ contains dono 

‘which’. However, if the test is done with the proper wh-adverb doo ‘how’, the same judgments obtain (Endo 

p.c.). 

219

‘What happened how?’ 

b. ?? Donoyoo-ni nani-ga okotta no. Ok under a D-linked reading 

(53) a. Nani-o donoyoo-ni Taro-ga sakuban sita no? 

What-Acc how-Dat Taro-Nom did last evening Q 

‘What did Taro do how last evening?’ 

b. ?? Donoyoo-ni nani-o Taro-ga sakuban sita no? Ok under a D-linking reading 

The hierarchy of the canonical order of the wh-phrases in the Japanese Mittelfeld is given in 

Table 1. 14 

Scrambling Canonical ordering of wh-phrases 

Any wh-phrase 

+Animate Arg Adverbs/Adverbials -Animate Arg Adv 

dare-ga > dare-ni > dare-o 

who whom who.Acc 

itsu> doko-ni 

when where 

Table 1 

nani-ga/nani-o 

what.Nom/Acc 

donoyoo-ni 

Movement of a wh-phrase over another in Japanese is A-movement and it necessarily induces 

a D-linked interpretation. To some extent such an effect is reminiscent of multiple whfronting 

languages like Romanian and Bulgarian, though in such case it is an A’ phenomenon 

and is very restricted. As in these languages, the same ordering which takes into account the 

make-up of the wh-element, i.e. +/-Animate, has a role to play: first come the [+Animate]marked 

wh-elements, and then the [-Animate]-marked ones (see Laenzlinger & Soare 2005). 

As in Romanian (one variety) and Bulgarian, the lower space of wh-ordering is filled by 

‘what’ and ‘how’. 

The conclusion to draw so far is that all these cases reinforce Miyagawa’s idea that 

there is no optionality with ‘short-scrambling’ and that ‘short-scrambling’ affects 

interpretability. 

14 

Interpretively, the scrambled D-linked wh-phrase in (39b), for instance, refers to a presupposed set of persons, 

and for each person, one is expected to answer who met him/her. In the unscrambled version in (39a), both a pair 

list and a single pair interpretation are possible. However, Kitagawa, Roehrs & Tomioka (2004) consider that it 

is harder to get a PL reading with scrambling. Irrespective of the exact reading, I suggest that, recast in terms of 

information structure, the scrambled wh-phrase becomes more prominent. 

220 

how

The question to tackle with is the exact position of the moved D-linked element. In his 

analysis of what elements satisfy the EPP feature on T, Miyagawa (2001) argues that a wh-PP 

can do so and provides the following examples. 

(54) Doko-no disuko-de i zen’in-ga t i odora-nakat-ta no? 

Where-Gen disco-at all-Nom t dance-Neg-Past Q 

‘At which disco, all didn’t dance?’ Miyagawa 2001, (52) 

Not>all, (all>not) 

(55) Dare-tai zen’in-ga t i asoba-nakat-ta no? 

Who-with all-Nom t play-Neg-Past Q 

‘With whom, all didn’t play?’ Miyagawa 2001, (53) 


In his account where Japanese is analyzed as being head-final, the scope relations between 

negation and subject quantifier are indicative of the position of the wh-PP. A NegP projection 

is assumed to be situated below SubjP, headed by -nakat- to which the lexical verb adjoins. 

On the basis of the scope relations noted, though he admits that the reverse scope relations are 

obtainable, he concludes that EPP feature on T is satisfied by the wh-PP. However, it seems 

that the cases above are instances of D-linked wh-phrases. If an unaggressively D-linked 

element, such as ittai ‘the hell’/’in the world’, is used, as in (56-57) below, the two cases still 

sound strange on the not>all interpretation, the predominant one being all>not (Yoshio Endo, 

p.c.). 

(56) Ittai doko-no disuko-dei zen’in-ga ti odora-nakat-ta no? 

In.the.world where-Gen disco-ati all-Nom ti dance-Neg-Past Q 

‘At which disco in the world, all didn’t dance?’ 

All>not, ???not>all 

(57) Ittai dare-tai zen’in-ga ti asoba-nakat-ta no? 

In.the.world who-withi all-Nom ti play-Neg-Past Q 

‘With whom the hell, all didn’t play?’ 

All>not, ???not>all 

221

The scope relations in (54) and (55) also obtain for a D-linked non-PP wh-element, as in (58). 

(58) Dono gakusei-oi zen’in-ga ti mi-nakat-ta no? 

Which student-Acc all-Nom ti see-Neg-Past Q 

‘Which studenti all didn’t see ti?’ 


Irrespective of the position of negation and of the point of the derivation where c-command 

relations obtain, the example below, where (59a) sounds more natural than (59b), may further 

indicates that there is a higher projection hosting the scrambled D-linked wh-phrase (Yoshio 

Endo, p.c.). 

(59) a. Dono disuko-de ittai dare-ga odotta no? 

Which disco-at the hell who dance-past Q 

‘Who the hell danced at which disco?’ 

b. ?(?) Ittai dare-ga dono-no disuko-de odotta no? 

Movement of the D-linked wh-phrase, whether a PP or a DP, is a case of local scrambling 

which is known to exhibit A-properties (Mahajan 1990). That the proposed Dlinked-WhP is 

not part of the left periphery is indicated by an example like (60). Scrambling of the D-linked 

wh object over the Nominative otagai-phrase results in a grammatical reciprocal reading 

(recall the classical case (11) discussed above). This shows that the movement position of the 

karera-phrase is a c-commanding A-position. 

(60) [Karera-no (uti) dono hito]-oi [otagai-noi sensei]-ga hihansita no? 

They-Gen (among) which one-Acc each other teacher-Nom criticized Q 

‘Which one (of them), each other’s teacher criticized?’ 

This means that scrambling of a DP or PP wh-phrase over another wh-element targets a 

position responsible for the D-linked reading of the short-scrambled wh-phrase. 

To sum up this section, it has been observed that that there is a preference for the adverb 

to precede the wh-object. The reversed order must be informationally more marked. It may be 

assumed that very subtle interpretive differences arise from wh-object movement across 

different classes of adverbs. I do not investigate this any further. A hierarchy of wh-ordering 

222

in the Japanese Mittelfeld has been arrived at where the lowest wh-elements are nani-o ‘what’ 

and donoyoo-ni ‘how.’ Short wh-scrambling is equated to movement to a position inside the 

Mittelfeld responsible for the D-linked interpretation of the moved element. 

5. Long ‘Wh-Scrambling’ 

This section concentrates on ‘long-scrambling’ of wh-phrases. One defining property of 

Japanese scrambling in general, long scrambling in our case, is that it is optional (Saito 1989; 

contra Saito 2007). This issue will be discussed below. 

5.1 The ‘undoing’ property and optionality 

Since Saito (1985, 1989), scrambling has been considered as a strictly optional operation 

(Kuroda 1988, Fukui 1993, Saito and Fukui 1998; cf. Saito 2007). At least insofar as A’- 

scrambling is concerned, it has been considered to be semantically vacuous though Saito 

(2007) shows that it affects scope and discourse interpretation. His examples are given below: 

(61) a. John-ga [Mary-ga dono hon-o yonda ka] siritagatteiru (koto) 

John-Top [Mary-Nom which book-Acc read Q know.to.want (fact) 

‘John wants to know [Q Mary bought which book].’ 

b. ? Dono hon-oi John-ga [Mary-ga ti yonda ka] siritagatteiru (koto) 

This type of scrambling has been associated by Saito (1989, 2004) with the property of radical 

reconstruction. In (61b), the wh-phrase dono hon-o ‘which book’ has been scrambled from the 

indirect question to the matrix clause, a position that cannot license a wh-phrase since it is a 

declarative clause, i.e. no Q-particle is present. Despite scrambling to the matrix clause, the 

wh-phrase is interpreted in its original position, in Saito’s terms it is “undone” at LF and such 

movement is necessarily semantically vacuous. This LF “undoing” property of Japanese 

scrambling has later come to be called its radical reconstruction property. Crucially, this 

movement does not violate the Proper Binding Condition (Fiengo 1977, May 1977), 

according to which a trace must be bound. 

There is nonetheless the opposing view, to which we subscribe in this chapter, that 

scrambling is feature-driven (see Miyagawa 1997, 2001, 2006a,b; Ishihara 2000; Grewendorf 

223

& Sabel 1999; Endo 2007). In Miyagawa (2006b), for instance, A’-scrambling is an optional 

overt operation subject to more general conditions on movement, namely to Fox’s (2000) 

Scope Economy Principle which is reinterpreted as below: 15 

(62) Interpretation Economy 

The movement is licensed in the new position iff it alters the interpretation of the string. 

This is held to be a universal requirement imposed on optional movement, i.e. 

movement must have an effect on the output (see also Reinhart 1995, Chomsky 2001). 

In what follows, I follow the idea that at least as far as movement of a wh-phrase is 

concerned, intermediate movement is triggered by a formal (Focus) feature. The reconstructed 

reading will be argued to obtain at the lower substantive Focus head though movement of the 

wh-phrase is further triggered by an EPP-feature on the higher non-substantive Focus head. 

5.2. Approaches to wh in situ 

Before properly dealing with the issue of wh-islands and reconstruction effects, it appears 

important to summarize the approaches proposed to the non-scrambled wh-phrases, i.e. to wh- 

in-situ. The analysis that I will suggest is somewhat similar to Richards’(2000). 

Huang (1982) treats Chinese wh-in-situ in terms of LF-movement, thus paralleling the 

effects of wh-movement in English, as can be seen from the examples below. (64b) is the LF 

representation for (64a). 

(63) John wonders [whati Mary bought ti]. 

(64) a. Zhangsan xiang-zhidao [Lisi mai-le shenme]. 

Zhangsan wonder Lisi bought what 

‘Zhangsan wonders what Lisi bought.’ 

b. Zhangsan xiang-zhidao [shenme [Lisi mai-le ti]]. 

15 The Scope Economy principle as formulated by Fox reads as follows: 

A Scope Shifting Operation can move XP1 from a position in which it is interpreted only if the movement 

crosses XP2 and is not scopally commutative. (Fox 2000:26) 

224

Nishigauchi (1990), Choe (1987) and Pesetsky (1987) reinforce Huang’s analysis by claiming 

that LF movement parallels overt movement in being subject to Subjacency. However, what 

distinguishes the grammatical status of the Japanese CNP in (65) from the ungrammatical 

status of its overtly moved English counterpart (*Whoi is John reading a book that criticizes 

ti?) is that what undergoes LF-movement in the former is not the bare wh-phrase but the piedpiped 

constituent in (65c). These are Nishigauchi’s examples (57) and (61). For the apparent 

complex NP island violation in (65a), Nishigauchi proposes that its LF-representation is not 

(65b) which does exhibit a CNP violation, but rather (65c), where the complex NP that 

contains the wh moves to Comp, as well as the wh itself, which moves within the complex 

NP. 

(65) a. Kimi-wa [dare-ga kai-ta hon-o] yomi-masi-ta ka? 

You-Top who-Nom wrote book-Acc read-polite Q 

‘You read books that who wrote?’ 

b. You read [[x wrote] books] [Comp whox ka] Nishigauchi 1990, (58) 

c. [You read y] [NP[[x wrote] WHOx] books]y 

Nishigauchi 1990, (61) 

Essentially, in Nishigauchi’s analysis, neither the wh-phrase which moves within the complex 

NP, nor the CNP itself which moves to Comp violate Subjacency. 

Assuming that there is no phrasal LF-movement at all (Chomsky 2000), on the one hand 

and that Japanese LF movement obeys the same restriction as English overt movement, i.e. 

Subjacency, on the other hand, Watanabe (1992) argues that in Japanese as well wh-questions 

involve movement in the overt syntax of an invisible operator from the specifier position of 

the wh-element to Spec CP, as represented in (66b) for (66a). 

(66) a. Boku-wa [John-ga nani-o katta ka] siritai. 

I-Top John-Nom what-Acc bought Q want.to.know 

‘I want to know what John bought.’ 

b. Boku-wa [Opi [John-ga [ti nani]-o katta] ka] shiritai. 

225

He adduces one piece of evidence in favour of such an analysis from the fact that an 

additional wh-phrase situated outside the wh-island cancels the wh-island effect. This is 

shown by the contrast in (67), where according to Watanabe’s analysis the wh-phrase nani-o 

‘what’ inside the embedded clause in (67a) taking matrix scope is incompatible with kadooka 

‘whether’ and induces a wh-island effect. Conversely, in (67b), the null operator originating 

in Spec DP of the wh-phrase dare-ni ‘to whom’, situated outside the island, moves to Spec CP 

overtly, and the sentence is grammatical. 

(67) a. ?? John-wa [Mary-ga nani-o katta kadooka] Tom-ni tazuneta no? 

John-Top Mary-Nom what-Acc bought whether Tom-Dat asked Q 

‘What did John ask Tom whether Mary bought?’ 

b. John-wa [Mary-ga nani-o katta kadooka] dare-ni tazuneta no? 

John-Top Mary-Nom what-Acc bought whether who-Dat asked Q 

‘Who did John ask whether Mary bought what?’ 

In more recent work, Watanabe (2001) suggests the possibility that what undergoes overt 

movement is not the operator, bur the wh-feature on the wh-phrase. 

However, recent work in psycholinguistics (Deguchi & Kitagawa 2002, Ishihara 2002, 

Kitagawa, Roehrs and Tomioka 2004, Kitagawa & Fodor 2006) casts doubt on facts like 

(67a), i.e. on the Additional Wh-effect and on the very existence of the wh-island effect with a 

wh-argument. I will come back to such issues in the sections below. 

I would like to suggest that in languages like Japanese (and Chinese) which realize a 

question particle in clause-final position, there is a Q-feature on the Force head in the left 

periphery which has an EPP-feature. This Q-feature realized as no/ka is responsible for 

interrogative clause typing. The discussion of these features will be developed in Chapter VI. 

The Focus head in the left periphery contains an invisible wh-feature which is marked [–EPP], 

hence the absence of wh-movement in Japanese (and Chinese). The subclausal complement of 

Focus, containing the already scrambled wh-phrase(s), overtly moves into Spec Force. It is 

thus in this respect that this analysis somewhat resembles Richards’ and Pesetsky’s but 

crucially differs in that I argue in favour of overt movement of a larger pied-piped constituent 

containing the wh-phrase(s). 

5.3 Islands in Japanese 

226

This section deals with both strong islands (SI) and wh-islands. 

As well-known, Japanese does not exhibit strong island sensitivity with either a wh- 

adjunct or argument. This is examplified 68-73 below. Here I do not consider the behaviour of 

naze ‘why’, which shows properties different from the other wh-adjuncts. I content myself 

with pointing out that naze ‘why’ is sensitive to any type of island, as illustrated in (68c) and 

(69c). 16 

Doo/itsu/doko-ni; naze in a CNP island 

(68) a. Kimi-wa [John-ga doo nagutta hito]-o hihansi-masita ka? 

You-Top John-Nom how hit person-Acc reproach-Polite Q 

‘Howi did you reproach [the person John hit ti]?’ Ura 1993, (8a) 

b. ? Kare-wa [Taroo-ga itu/doko-de katta hon]-o yonde-iru no? 

He-Top Taro-Nom when/where bought book read-prog Q 

‘Wheni/wherei is he reading [a book that Taro bought ti]?’ Endo, p. 

c. * Kimi-wa [John-ga naze nagutta hito]-o hihansi-masita ka? 

You-Top John-Nom why hit person-Acc reproach-Polite Q 

’Why i did you reproach [the person John hit ti]?’ Ura 1993, (7a) 

Doo/Doko-ni; naze in an Adjunct island 17 

(69) a. Kimi-wa [John-ga doo Mary-o kizutuketa noni] kare-o kabai-masita ka? 

You-Top John-ga how Mary-Acc hurt though he-Acc defend-Polite Q 

’Howi did you defend him though John hit Mary ti?’ Ura 1993, (8c) 

b. ? Taroo-ga [doko-ni itta kara] umaku itta no? 

Taro-Nom where-Dat went because well went Q 

‘Wherei did things go well because Taro went ti?’ Endo, p.c. 

c. * Kimi-wa [John-ga naze Mary-o kizutuketa noni] kare-o kabai-masita ka? 

You-Top John-ga why Mary-Acc hurt though he-Acc defend-Polite Q 

’Why i did you defend him though John hit Mary ti?’ Ura 1993, (7c) 

16 The wh-adjunct naze ‘why’ can be construed across a clause as long as there is no island; unlike other bare 

wh-phrases, it can follow a quantifier phrase, and must follow other wh-phrases (the so-called anti-superiority 

effect). 

17 As pointed out by Richards (2000), not all Adjunct islands have the same status. For instance, doo in an 

adjunct island introduced by atode ‘after’ gives an ungrammatical result (Yoshio Endo, p.c.) 

227

Doo in a Manner-of-Speaking Verb island/Non-bridge Verb island 

(70) Kimi-wa [kimi-no haha-ga doo kimi-mo koibito-o nagutta to] sasayaki- 

masita ka? 

You-Top you-Gen mother-Nom how you-Gen lover-Acc hit C whisper- 

Polite Q 

‘Howi did you whisper [that your mother hit your lover ti]?’ Ura 1993, (b) 

Dare-o in a Manner-of-Speaking Verb island/Non-bridge Verb island 

(71) Kimi-wa [kimi-no haha-ga dare-o nagutta to] sasayaki-masita ka? 

You-Top you-Gen mother-Nom who-Acc hit C whisper-Polite Q 

‘Whoi did you whisper [that your mother hit ti]?’ 

Dare-o in a CNP island 

(72) Taroo-wa [sono yakuza-ga dare-o korosita tatemono-o] katta no? 

Taro-Top that gangster-Nom who-Acc killed building-Acc bought Q 

‘Whoi did Taro buy a building where that gangster killed ti?’ Richards 2000 (1) 

Dare-o in an Adjunct island 

(73) [Taroo-ga dare-o Amerika-ni yatta kara] umaku itta ka? 

Taro-Nom who-Acc America-to sent because welll went Q 

‘Whoi did things go well because Taroo because Taro sent ti to America?’ 

The fact that doko-ni ‘where’ in (68b) and (69b) gives slightly marginal results is not due to 

its adverbial PP status (note that the real adverb itsu ‘when’ also gives rise to the same kind of 

effect), but seems to be due to speaker variation. For those speakers who find a wh-adjunct in 

a strong island slightly marginal, a D-linked adjunct provides perfect results. This is 

exemplified below. 

D-linked wh-adjunct in a CNP island 

(74) Kare-wa [Taroo-ga dono hi-ni katta hon]-o yonde-i-masu ka? 

He-Top Taro-Nom which.day-on bought book read-prog Q 

‘Which dayi is he reading a book that Taro bought ti?’ Endo p.c. 

228

D-linked wh-adjunct in a Adjunct island 

(75) [Taroo-ga dono basyo-ni iki-masita kara] umaku iki-masita ka? 

Taro-Nom which.Place-to went-polite because well went-polite Q 

‘Which placei did go things go well because Taro went ti?’ Endo p.c. 

Notice that for those speakers grading a wh-adjunct in a strong island as slightly marginal, a 

lexical non-wh adjunct also results in grammaticality, as shown below. 

Non-wh adjunct in the CNP island 

(76) Kare-wa [Taroo-ga kinoo katta hon]-o yonde-i-masu. 

He-Top Taro-Nom yesterday bought book-Acc read-prog-polite 

‘He is reading the book that Taro bought yesterday.’ Endo p.c. 

Non-wh adjunct in the Adjunct island 

(77) [Taroo-ga Paris-ni itta kara] umaku iki-masita. 

Taro-Nom Paris-to went because well went-polite 

‘Things go well because Taro went to Paris.’ Endo p.c. 

Two groups of approaches have been proposed in the literature in order to account for the lack 

of SI effects. Under one view, wh-phrases are not subject to the relevant island constraints 

either because they hold of overt movement only (Huang 1982) or because these wh-phrases 

never undergo movement at any point in the derivation (Reinhart 1995, Cole and Hermon 

1998). Under another view, covert movement is subject to island effects, but the island may 

be circumvented if movement affects not the wh-phrase but the island containing it 

(Nishigauchi 1990, Pesetsky 1987, Richards 2000). 

Our analysis is somewhat similar to the second kind of approaches. We propose that a 

Q-feature fills the Force head in the left periphery being responsible for clause-typing and that 

dependent on the language Q may be endowed with EPP. The Foc head has a wh-feature, also 

possible associated with EPP. Based on this feature split, we propose that in Japanese that the 

Q-feature on the Force head in the left periphery has an EPP-feature and thus its complement, 

Spec Foc, containing the already scrambled wh-phrase(s) overtly moves into Spec Force. The 

wh-feature on Foc does not have an EPP, which accounts for the lack of wh-movement in 

Japanese. Wh-feature checking/valuing at a distance via AGREE. Such an analysis somewhat 

resembles Richards’ and Pesetsky’s but crucially differs in that we argue in favour of overt 

229

movement of a larger pied-piped constituent containing the wh-phrase(s). Overt movement of 

the heavy pied-piped constituent cuts across an island and a mere clausal substructure. 

Evidence for overt movement to the dedicated Force position comes from the scope 

effects of the wh-phrase pied-piped with an entire structure. However, the status of the pied- 

piped clausal structure bears on extraction. Extraction of a wh-phrase from a (strong) island is 

prohibited. Conversely, wh-subextraction out of the non-island pied-piped clausal constituent 

is permitted and no Criterial Freezing (Rizzi 2006, Rizzi & Shlonsky 2006) violation is 

involved. I come back to these issues in the section below. Interpretive properties lend further 

support for such movement. The cartography offers a rich articulation of the clausal 

projection, where the subextracted wh-element moves to a higher TopP. Such an analysis is at 

work in the section below. 

The prediction that the overt pied-piping analysis suggested makes is that there should 

be no SI/WI effects. Let us see how this works. 

Recall that since Huang (1982), LF movement does not obey Subjacency in Japanese 

(Lasnik and Saito 1984, Nishigauchi 1990, etc). 

Nani-o in a CNP island 

(78) John-wa [nani-o katta hito]-o sagasite iru no? Watanabe 1992 (5) citing 

John-Top [what-Acc bought person]-Acc looking for Q Lasnik & Saito 1984 

‘Whati is John looking for the person who bought ti?’ 

Nani-o in a wh- island 

(79) ?/?? John-wa [Mary-ga nani-o katta kadooka] siritagatte iru no? 

John-Top [Mary-Nom what-Acc bought whether know.want Q 

‘Whati does John want to know whether Mary bought ti?’ 

Whereas a wh-argument does not violate Subjacency in the case of the CNP island in (78), the 

status of the wh-island effect with the same wh-argument in (79) varies with authors. Thus, 

Nishigauchi (1990) and Lasnik and Saito (1984) mark it with a ‘?’ mark, while Watanabe 

(1992, (14)), Ura (1993 (5b)) and Tanaka (1999 (4)) judge it as more marginal, being marked 

by ‘??’. A wh-adjunct also exhibits the wh-island effect, though with a marked contrast. 

Compare (79) to (80). 

Doo in a wh-island 

230

(80)* Kimi-wa [John-ga doo Mary-o nagutta kadooka] sitte-imasu ka? 

You-Top John-Nom how Mary-Acc hit whether know-Polite Q 

‘Howi do you know whether John hit Mary ti?’ Ura 1993 

The status of (79) has been challenged by recent work in psycholinguistics which takes into 

account the role played by prosody and its interaction with syntax and processing. For 

exemplification, we present here some major ideas in this respect. Thus, Deguchi and 

Kitagawa (2002), Ishihara (2002), Kitagawa, Roehrs and Tomioka (2004) and Kitagawa and 

Fodor (2006) (henceforth K&F) challenge the claim that wh-in-situ in Japanese exhibits whisland 

effects. Deguchi and Kitagawa (2002), for instance, point out (i) that wh-questions are 

accompanied by some special intonation, called Emphatic Prosody (henceforth EPD), 

consisting of an emphatic accent on the focused wh-element followed by deaccenting, or postfocal 

eradication, of all lexical items up to some C, and (ii) that the domain of EPD coincides 

with the scope domain of the wh-phrase, i.e. it depends at which C[+wh] the EPD ends (but see 

Ishihara 2004 for a different claim). They also show that there is a correlation between the 

extent of the prosodic deaccenting and the extent of the semantic scope of the wh-phrase. 

Subordinate scope (indirect question) is associated with what they call Short-EPD, i.e. EPD 

ending at C[+wh] of the embedded clause, whereas matrix wh-scope is associated with Long- 

EPD, i.e. EPD extending to the matrix C[+wh]. 

Thus, these studies show that if pronounced with Long-EPD, i.e. deaccenting continues 

all the way until the matrix no, a sentence like (79) is acceptable and has the matrix 

interpretation of the wh-phrase. If pronounced with Short EPD, it is unacceptable. This is 

represented below. 

(79’) a’. John-wa [Mary-ga nani-o katta kadooka] siritagatte iru no↑? 

a’’. John-wa [Mary-ga nani-o katta kadooka] siritagatte iru no↑? 

K&F account for this ungrammaticality on two reasons: either the hearer attempts to interpret 

the sentence with matrix wh-scope and would judge the prosody stopping at kadooka 

‘whether’ inappropriate (a case of violation of the prosody-scope match) or s/he judges it as a 

yes/no question but the subordinate complementizer is ungrammatical, i.e. it is incompatible 

231

with C[+wh]. 18 It is therefore the awkwardness resulting from the misinterpretation of prosody- 

scope relations that has been assessed in the literature as the Subjacency effect. 

Therefore such recent facts substantiate the prediction made above, namely that under 

AGREE from Foc and the pied-piping analysis there are no SI/WI effects in Japanese. 19 

5.4 Wh-islands and overt long-extraction. Reconstruction in Japanese 

In this section the analysis suggested above is put to work. Hiraiwa’s (2000) mechanism of 

MULTIPLE AGREE will be introduced and slightly modified to account for what is commonly 

designated as the Additional Wh-effect. The section ends with a short discussion of radical 

reconstruction, and following Miyagawa (2006b), it is shown that a long-distance scrambled 

wh-phrase fills a Top+Wh projection in the left periphery. 

Consider the example in (81) with the wh-argument inside the island. 

(81) Naoya-wa [Mari-ga nani-o nomiya-de nonda ka] oboeteru? 

Naoya-Top Mari-ga what-Acc bar-Loc drank Q remember 

‘Naoya still remembers what Mary drank at the bar.’(slightly modified from Ishihara 

2004(4b)) 

Phonologically, pitch falls on the wh-phrase and Ishihara’s (2002, 2004) post-Focus 

reduction, PFR, (corresponding to Deguchi & Kitagawa‘s deaccenting phenomenon or to 

18 In the literature on Japanese, some complementizers are ambiguous (see fn. 2). Thus, ka can function either as 

wh-scope marker in any clause, or as ‘whether’ in subordinate clauses, or else as a yes/no question marker in 

matrix clauses. Its yes/no marker interpretation in matrix clauses as well as its ‘whether’ interpretation are not 

considered here. Similarly, kadooka is ambiguous for some speakers. For most speakers, it is Cwhether though for 

only some of them it functions as a wh-scope marker in an embedded clause (Kitagawa & Fodor 2006:fn2). 

19 If, following such recent developments, Japanese does not exhibit the wh-island effect with a wh-argument, 

then it comes to have the argument-adjunct asymmetry (79 vs. 80). In order to account for this asymmetry, it 

may be suggested that AGREE is subsumed under Rizzi’s (2004b) RM. More precisely, kadooka ‘whether’ is 

decomposed into ka Q + doo how + ka Q, with the higher ka fills Force° whereas the lower ka heads a projection 

of its own, similar to the Italian si or the English if called IntP (Rizzi 2001) in whose specifier doo, the yes/no 

operator, sits. At the Spell-Out to the phonological component, the three elements are interpreted as a 

phonological unit. If the wh-feature on Foc probes over –doo- (i.e. [-Arg]) in search of a wh specified [+Arg], 

then RM arises. If, as in (80), it probes for an adjunct, then RM arises. 

232

K&F’s Short-EPD) stops at the end of the embedded clause, which corresponds to the Q- 

particle ka (there is a match between the focus intonation domain and the wh-scope). 

For reasons to discuss in Chapter VI, I follow Pesetsky and Torrego’s (2004) system of 

feature valuation. Thus, derivationally, the embedded Focus head hosts an interpretable 

unvalued wh-feature which AGREES with the wh-phrase. Lack of an EPP associated with the 

wh-feature explains the absence of wh-movement in Japanese. Force° has an interpretable 

substantive Q-feature with an EPP-feature which attracts its pied-piped FocusP complement 

into its specifier. At this point of the derivation, the wh-phrase gets embedded scope. 

Assuming matrix Obj in (81) has an EPP-feature, the larger ForceP structure moves to matrix 

SpecObjP. 

Consider next the case of a wh-island. 

(82) John-wa [Mary-ga dare-o hometa ka] siritagatteiru no? 

John-wa [Mary-Nom who-Acc admired Q] want.to.know Q 

‘Does John want to know whom Mary admired?’ 

‘Whom does John want to know whether Mary admired?’ 

One the embedded scope interpretation of the wh, post-focus reduction (PFR) continues until 

the embedded C and stops at this point. Besides embedded scope, the sentence has a yes/no 

interpretation. On the second interpretation, if PFR is extended to the matrix C, matrix scope 

obtains and the sentence is interpreted as a direct wh-question. I suggest the following 

derivations to account for the two possible interpretations. 

Consider (83a): On the embedded scope reading, the iwh-feature on the embedded 

Focus AGREES with the wh-phrase, as indicated by the dotted line (2). Then the clausal SubjP 

portion, which contains the wh moved from vP to ObjP, raises to Spec of a projection headed 

by the past tense morpheme –ta- (this projection is labelled here TP). This is represented by 

the dotted line (4). Next iQ on Force° attracts the FocP containing the wh-phrase into its 

specifier. Given that embedded Focus˚ has a substantive wh-feature, the scope of the whphrase 

gets frozen at this level. As represented in (83b), once the entire clausal structure in 

Spec Force, the edge of the phase, it can get attracted further up. The clausal structure ForceP 

moves to matrix Spec Obj, as indicated in (5). 20 Subsequently, as shown by (6), the matrix 

20 More generally, we assume that in Japanese EPP is blind to category. So not even DP objects but also CPs 

may ‘scramble’. 

233

SubjP gets attracted to matrix Spec Force due to the EPP associated with Q (other derivational 

steps are ignored). The wh-feature on the matrix Force is of a different nature from the one on 

the embedded Force. It is labelled a yes/no feature here. This derivation with scope at the 

embedded level is represented below. The dotted line represents the AGREE mechanism. 

(83a) …Force 

2 

2 

ka FocusP 

iQ 2 

2 

(4) iwh TP 

2 

2 

-ta SubjP 

(3) 2 

Mary-ga ObjP 

2 

dare-o vP 

(2)AGREE 2 

home- 

(1) 

(83b) ForceP 

2 

2 

no FocP 

(6) 2 

2 

uwh TP 

2 

SubjP 

2 

John ObjP 

2 

[… dare-o…] vP 

2 

siritagatte- ForceP 

(5) 2 

… 

Let us further consider how the matrix scope reading obtains. As opposed to the above 

derivation, the embedded Foc has an uwh feature, and thus no scope freezing is realized at 

this level. As willl explained more in detail below, the relevant phase (I assume it is the 

complement of Force) is not yet sent to Spell-Out as the wh-element still bears its 

234

uninterpretable wh-feature. The large pied-piped constituent (ForceP) moves to matrix Spec 

Obj. The iwh-feature on the matrix Foc° AGREEs with its counterpart on dare-ni, which is 

included into the moved pied-piped structure. As a result, deletion of the uninterpretable wh- 

feature takes place. The larger FocP structure further gets probed and attracted by the 

substantive Q-feature on the matrix Force°. Matrix scope is thus obtained at the matrix level 

where iwh finds itself. 

One prediction that the pied-piped clausal analysis makes is that two wh-phrases should 

have the same scope, either embedded or matrix scope. This prediction is borne out, as 

reported in the literature (see also discussion in Richards 2000). Consider (84) below. 

(84) Keesatu-wa [ano-ban dare-ga dare-to atteita-ka] minna-ni tazuneta-no? 

Police-Top that night who-Nom who-with seeing-Q everyone-Dat asked- Q 

‘Did the police ask everyone who was with whom that night?’ 

‘Whoi is such that the police asked you whether hei was with whom that night?’ 

In order to account for cases with two wh-phrases in their canonical order, and not scrambled 

one over the other, some theoretical remarks are in order here. 

I adopt Hiraiwa’s (2000) theory of MULTIPLE AGREE meant to capture (covert) multiple 

feature-checking. This type of AGREE has already been presented when discussing cases of 

multiple wh-movement on Romanian in Chapter III. It is repeated below. 

(85) MULTIPLE AGREE Hiraiwa’s (2000:69) 




This definition relies on Chomsky’s (2000) Matching Condition: 

(86) Match (Chomsky 2000:122) 


b. D(P) is a sister of P. 

c. Locality reduced to ‘closest c-command’. 

235

MULTIPLE AGREE as a single simultaneous operation in Hiraiwa: 

α > β > γ 

(AGREE(αβγ), where α is a probe and both β and γ matching goals for α) 

In his system, once the probe P has been merged, it searches down for the closest matching 

goal feature within its c-command domain and matches with the closer goal β but not AGREE 

takes place. Being [+multiple], P searches down for all matching goals, in this case for γ, until 

the matching is completed within an ‘accessible’ domain. At this point of the derivation, 

AGREE applies to all the matched goals derivationally simultaneously. 

Here is an illustration of how MULTIPLE AGREE works. Consider (35) repeated below. 

(35) Dare-ga dare-ni atta no? 

Who-Nom whom-Dat met Q 

‘Who met whom?’ 

The Focus head in our system has an interpretable wh-feature and probes into its domain and 

enters into Match with the two goals, the first wh-phrase dare-ga and the second dare-ni 

simultaneously. At this point, the uninterpretable wh-features on the wh-phrases are valued 

and deleted. Thus, MULTIPLE AGREE is applied to the two goals. Since Force° has a Q-feature, 

the pied-piped constituent containing the two goals is attracted to Spec Focus. 

Let us return to example (84) above. On the embedded wh-interpretation in (87a), the 

embedded Foc has an iwh-feature that enters into MULTIPLE AGREE with both Goals 

derivationally simultaneously. Consequently, the wh-features on both goals get valued and 

deleted. Since the embedded Force° has iQ with an EPP-subfeature, FocP is further attracted 

to Spec ForceP. However, the scope of the wh-elements is fixed at the embedded Focus level 

since it has an iwh-feature. The large ForceP structure successively moves through Spec Obj 

and gets attracted to matrix Spec Force (what gets attracted is actually the matrix SubjP 

containing the ForceP structure already moved to Spec Obj). A yes/no feature is assumed on 

the matrix Foc. 

(87a) …ForceP 

2 

2 

ka FocusP 

236

uQ 2 

(4) 2 

uwh TP 

2 

2 

(3) -ta SubjP 

2 

dare-ga ObjPPP 

2 

dare-to vP 

2 

(2) 2 

(1) attei- 

Let us consider the derivation on which the matrix wh-interpretation obtains. The embedded 

Foc is [-multiple] and has formal features. We suggest that at this later in the derivation the 

ForceP structure is not sent to the Spell-Out. We suggest that Spell-Out is delayed until merge 

of the matrix iwh-feature. In other words, Spell-Out is delayed until Matching is completed. 

The derivation proceeds as usual. At he embedded level, the two wh’s leave the vP and move 

to Spec Subj and Spec ObjP, respectively. Next, the entire SubjP portion is pied-piped to Spec 

TP, headed by the past tense morpheme, -ta. Subsequently, due to EPP on Q, realised as ka, 

FocP is pied-piped to Spec ForceP. I propose a modification of Hiraiwa’s MULTIPLE AGREE: 

assume that the probe on the matrix Foc with iwh-feature is [+multiple] and consequently 

does MULTIPLE AGREE with the wh-phrases contained in the pied-piped ForceP clausal 

structure which has moved to Spec ObjP. Scope is computed or ‘frozen’ therefore at the 

matrix level. Once MULTIPLE AGREE has been done, the EPP associated with Q on matrix 

Force drives successive-cyclic movement of the TopP structure, as shown in (87b) below. 

Before the entire TopP clausal structure moves to Spec ForceP, the DP keesatu ‘police’ moves 

to the left-peripheral TopP from its Spec SubjP. This is schematically represented below 

(irrelevant steps omitted): 

(87b) ForceP 

3 

no TopP 

iQ+EPP 2 

2 

wa FocP 

2 

2 

[+mutiple] SubjP 

237

iwh 2 

keesatu ObjPDO 

2 

[…dare-ga dare-to…] ObjPIO 

2 

minna-ni vP 

2 

tazune- ForceP 

3 

... 

Let us see how such an analysis works for cases of long-scrambled wh-phrases and whether 

this is indeed long wh-scrambling. Consider (88) below: 

(88) a. Nani-o John-wa [Mary-ga t tabeta-ka] siritagatteiru-no? 

What-Acc John-Top Mary-Nom t ate-Q wants.to.know-Q 

a. Preferred reading: ‘What does John want to know whether Mary ate?’ 

b. Dispreferred reading: ‘Does John want to know what Mary ate?’ 

Though there has been disagreement in the literature over the scope of the scrambled wh- 

element, there is a preference for the matrix scope of the wh (see experimental studies 

referred to above). 

However, the double status of (88) receives an immediate explanation along the lines of 

the analysis suggested here. If the embedded Foc does not have a substantive wh-feature, but 

a purely formal one, Spell-Out is delayed and the pied-piped clausal structure containing the 

wh-element is further probed by the higher wh-feature and therefore interpreted in this latter 

higher position, hence the matrix scope interpretation of the wh-phrase. No Criterial Freezing 

applies at this level. The wh-element gets extracted and moves higher to a Top+Wh 

projection, above the FocP, hosting only non-wh topics in order to value its Top-feature. 

Conversely, if the embedded Foc has an iwh-feature, the wh-phrase receives the embedded 

interpretation on the lower Foc and the clausal substructure is further attracted to matrix Force 

due to the presence of an EPP on Force°. 

It is important to note that the object wh-phrase is not frozen in place as it is contained 

into a larger pied-piped constituent and thus, following Rizzi (2006), it can get extracted and 

moves to check a Top-like feature on a higher complex Top+Wh projection in the left 

periphery, i.e. it has a D-linked reading. The derivation on the matrix reading is shown in (89) 

below. 

238

(89) ForceP 

2 

Top+Wh 

3 

nani-o Top 

3 

Foc 

3 

3 

iwh SubjP 

3 

John ObjP 

3 

[…nani-o…] vP 

AGREE 3 

ForceP 

3 

3 

ka FocP 

uQ 3 

+EPP 3 

uwh …SubjP 

3 

Mary-ga ObjP 

3 

nani-o vP 

… 

Conversely, no such D-linked interpretation is available on the embedded scope reading 

(Endo, p.c.) 

Let us turn to an example of ‘radical reconstruction’ in Saito’s (1989) original terms: 

(90) ? Nani-o Mary-ga [John-ga t katta ka] siritagatteiru. 

What Mary-Nom John-Nom t bought Q wants.to.know 

‘Mary wants to know what John bought.’ 

Although Saito marks the sentence as slightly marginal, as indicated by the single question 

mark, Miyagawa (2005, 2006b) and Endo (p.c.) judge it perfectly grammatical. This is the 

kind of example discussed in Miyagawa (2005, 2006b) and argued to not necessarily display 

the reconstruction effect. 21 He argues that in this position, the wh-phrase has a D-linking 

21 Ishihara (2004) argues in favour of a Multiple Spell-Out approach to the phonology-semantics 

correspondence. For cases of long-scrambled wh-phrases, he argues that a moved wh-phrase lands outside the 

domain (i.e. phase) where it would be assigned scope. The wh-phrase nani-o in (90) should be assigned scope in 

239

eading (“the wh-phrase nani-o ‘what’ is most easily interpreted as ‘among the things we are 

talking about, John wants to know which of them Taro bought’. No such reading is available 

- or at least requires – in the nonscrambled version” Miyagawa 2006b:620). In line with the 

theoretical assumptions made so far, in (91a) both the subject and the object wh vacate the vP 

domain and move to Spec SujP and Spec ObjP, respectively. Then the clausal SubjP portion 

raises to Spec TP headed by the past tense morpheme –ta-. This is represented by the dotted 

line (3). Next EPP on Force° attracts the FocP containing the wh-phrase into its specified (3). 

Given that embedded Focus° does not have a substantive wh-feature, the scope of the wh- 

phrase does not get frozen at this level. AGREE does not take place from the uninterpretable 

wh-phrase and the wh-element does not have its feature checked/valued. Consequently Spell- 

Out is delayed. If the matrix Force has an invisible Q with an EPP-feature and matrix Foc has 

an interpretable wh-feature, the derivation proceeds as in (91b): The larger clausal structure 

ForceP subsequently passes through Spec Obj. At this point the iwh-feature on Foc AGREES 

with its uninterpretable counterpart on nani-o which gets deleted. Since the wh-element still 

bears a Top-like feature, it is probed and attracted to Spec Top+Wh. This is how the D-linked 

reading of the wh-DP is accounted for. 

(91a)…Force 

2 

2 

ka FocusP 

2 

2 

(3) wh TP 

2 

its embedded clause, marked by the embedded C[+wh] scope marker. However, he argues that the domain where 

the wh-phrase takes scope, i.e. what he calls the domain of focus intonation, roughly corresponding to Deguchi 

& Kitagawa‘s deaccenting domain, is postponed to a later Spell-Out cycle. Consequently, it becomes larger than 

the actual wh-scope. 

240

2 

-ta SubjP 

(2) 2 

John-ga ObjP 

2 

nani-o vP 

2 

kat- 

(1) 

(91b)ForceP 

3 

Top+Wh 

3 

nani-o FocP 

3 

iwh SubjP 

3 

ObjP 

3 

vP 

AGREE 3 

ForceP 

3 

… 

So far it has been observed that with a long-distance moved wh-element, a D-linked reading is 

involved which is best analyzed as some sort of Topicalisation in the sense of Rizzi (2001). 

The generalisation to make from these examples is that long wh-scrambling is equated 

to movement to a higher Top+WhP in the left periphery, the wh-phrase having a D-linked 

reading (this is some sort of topicalisation). 

To sum up this section, it has firstly been suggested that matrix scope in a complex 

sentence obtains by delaying Spell-Out until the matrix clause where a Focus head with an 

interpretable wh-feature is merged. 

Secondly, the ambiguous reading of a long-distance moved wh-phrase has also been 

dealt with. The matrix scope reading of the wh-element obtains when the heavy clausal piedpiped 

structure containing the wh-element lands in the matrix Spec Force position after 

AGREE from the matrix Focus head, having substantive wh-features, has applied. The lower 

Focus has only purely formal features. The generalisation is that scope freezing takes place on 

the Focus that presents interpretive/substantive features. 

241

Thirdly, building on Miyagawa’s (2006b) observation that ‘radical reconstruction’ does 

not necessarily take place it has been seen that even such a case represents an instance of 

topicalisation of the moved wh-element. 

Fourthly, in order to account for cases of two canonically ordered wh-phrases in a 

matrix clause or inside a weak wh-island and for cases where one wh-element is inside a whisland 

whereas the other is outside, the mechanism of MULTIPLE AGREE has been introduced. 

Essentially, a Focus head marked [+Multiple] probes down into its domain and values the whfeatures 

on the two goals. 

6. Conclusion 

This chapter represents an attempt at analyzing Japanese along the lines of Kayne’s 

antisymmetry hypothesis. It has shown that the Japanese Mittelfeld contains Topic and Focus 

features parasitic on phi/Case features. More precisely, it has been shown that a DP with the 

appropriate features taken from the Numeration can target a Mittelfeld-internal position 

associated with TopP or FocPNew Information features, close to Belletti (or Jayaseelan) though 

such a position may be followed by distinct adverb classes. The hypothesis that both the 

subject and the object leave the vP domain receives support from the tests with the different 

classes of adverbs. 

The paper has also addressed the question whether wh-phrases remain in-situ or move 

and the same tests have pointed out that they move to distinct positions in the Mittelfeld. 

Following Kayne’s (1994, 2005) insights, it has been observed that the wh-object in the SOV 

order can float to positions ranging from the lower celerative aspect adverb projection to the 

time adverb projection, but cannot move above high adverbs like probably and fortunately. 

The suggestion has also been put forth that such movement possibilities in the Mittelfeld 

induce very subtle interpretive differences which thus do away with the idea that (short) 

scrambling is optional. Similarly, based on Miyagawa (2005), it has also been argued that Ascrambling 

of a wh-phrase over another systematically induces a D-linked reading. 

The analysis in terms of (sub)clausal pied-piping also accounts for the matrix vs. 

embedded scope reading of the wh-phrase in terms of the substantive vs. formal wh-feature on 

the Focus. It has been shown that matrix scope obtains when there is Spell-Out delay until the 

matrix Focus. Embedded scope obtains on the embedded Focus which has substantive 

features. 

242

Hiraiwa’s (2000) mechanism of MULTIPLE AGREE, slightly modified, has been adopted 

here in order to account for cases with two wh-elements and their sychronised matrix scope. 

As adopted here, this mechanism presupposes AGREE at a distance from a single probe (the 

Focus) to the two matched goals derivationally simultaneously. 

The cartography coupled with the non-application of Criterial freezing (due to rules 

regulating pied-piping) provides an explanation for extraction of the wh-phrase (from the 

heavy pied-piped clausal structure) to a higher Top+Wh projection, responsible for the D- 

linked reading observed in such cases. Movement to this position concurs with Criterial 

freezing. 

An antisymmetric approach to Japanese seems after all worth undertaking. 

243

CHAPTER VI 

A CROSS-LINGUISTIC TYPOLOGY OF QUESTION FORMATION 

1. Introduction: the Q- and Wh-features 

In this chapter I propose a typology of question formation strategies in line with the 

antisymmetry hypothesis articulated in Kayne (1994), (2005). I hope to show that head-final 

languages like Japanese possess left peripheral Force and Focus heads, each with a potential 

EPP feature. As a result, the complement of Force 0 is pied-piped to its specifier, giving rise to 

a linear head-final configuration. 

There is a group of analyses of a non-antisymmetric nature that assumes some kind of 

feature attraction to a right-headed C. Among them are Miyagawa’s (2001), Watanabe’s 

(1992) and Hagstrom’s (1998, 2004). 

Hagstrom (1998) and Miyagawa (2001) propose that wh-question formation in several 

descriptively head-final wh-in-situ languages involves movement of a Q particle to C. This 

operation, labelled Q-Movement, is illustrated schematically for Japanese in (1). 

(1) CP 

3 

TP C 

3 

vP T 

6 AAgree and/or movement 

wh-phrase-ka 

Watanabe (1992) argues that Japanese wh-questions involve movement of a phonologically 

empty operator to Spec CP in overt syntax and in this way Japanese parallels English. 

245

However, under such an analysis English and Japanese differ minimally in that in the former, 

it is the entire wh-phrase that is attracted to C, whereas in the latter it is an empty operator 

originating in Spec DP that raises to Spec CP. The head C is filled by the Q-particle. In more 

recent work, Watanabe (2001) suggests the possibility of wh-feature movement in overt 

syntax for Japanese. 

In Hagstrom’s (1998) original proposal, the Q-particle originates with the wh-element, 

thus accounting for the ambiguous status of the wh-element, i.e. either an existential 

quantifier or a wh-phrase. In a wh-question, the Q-feature on the C head attracts the 

Q(uestion)-particle which raises to C, whereas the wh-phrase remains in-situ. 

With Miyagawa’s proposal (2001), the two features relevant to wh-question functioning 

are the Q-feature and the wh-feature. In English, the two features situated on C determine 

pied-piping of the wh-phrase to Spec CP, whereas in Japanese the two features are 

morphologically separated. In this latter case, the Q-feature on C agrees with the Q-particle 

and the EPP associated with Q-feature triggers raising of the Q-particle to C. In Miyagawa’s 

approach, the wh-feature is not on C but on T and since T also has an EPP, then a wh-PP 

moves to Spec TP. 

Therefore, the above-mentioned analyses postulate movement of the Q-particle into C, 

which may be covert, as in Sinhala, or overt, as in Japanese. 

What I retain from these analyses is the idea that there are Q- and wh-features but I will 

deploy them in a different, i.e. antisymmetric framework. Within the cartographic approach 

(Belletti 2004, Cinque 1999, 2002, 2006, Rizzi 1997, 2004a) adopted here, I suggest a 

typology of question formation based on the abstract morphological split between the Qfeature 

and the wh-feature. I assume that there is a general principle according to which the 

Q-feature is associated with a clausal category (clause typing) with which it agrees. I adopt 

Chomsky’s (2000, 2001) definition of AGREE, stated in (2) below, which is a syntactic 

feature-checking relation operation which eliminates the ‘feature-movement’ part of 

ATTRACT. 

(2) AGREE 


b. D(P) is the sister of P. 

c. Locality reduces to “closest c-command.” (Chomsky 2000:122) 

246

I assume that the Q-feature is universally merged in the Force head. Some arguments in 

favour of this will be given in section 2 below. 

Given these two principles, one parameter that distinguishes among languages has to do 

with the fact that the Q-feature is some languages can be overtly realized as a particle, or can 

be covert in others. Another parameter holds that the Q-feature can be associated with an 

EPP-feature, in which case it triggers MOVE in addition to AGREE. AGREE+MOVE thus triggers 

movement of the “complement” (clausal) category to Spec ForceP. 

Concerning the wh-feature, one principle the typology relies on holds that the whfeature 

is associated with an XP+wh with which it agrees. Following standard assumptions, I 

assume that the wh-feature is realized on Focus°. In cases of multiple wh-phrases, one further 

principle that may be involved regards Hiraiwa’s (2000) mechanism of MULTIPLE AGREE 

discussed in Chapter IV. 

Based on these principles, the parameters related to the wh-feature involve the 

possibility of its being overt or covert. Like the Q-feature, it can be associated with an EPPfeature, 

in which case it triggers MOVE in addition to AGREE. As mentioned, Hiraiwa’s (2000) 

mechanism of MULTIPLE AGREE is considered, an additional parameter may be involved: the 

existence of single wh-MOVE or ‘multiple wh-MOVE’. 1 

Therefore, under such theoretical assumptions, a wh-question may the structural 

character represented below, thus comprising at least steps (3a) and (3c). 

(3a) … ForceP 

3 

(Q-Prt) …FocusP 

Q-feature 3 

(Wh-Prt) …SubjP 

Wh-feature 3 

ObjP 

Agree 3 

wh-phrase vP 

3 

1 

‘Multiple wh-MOVE’ is a cover term to refer to several types of movement possibilities that characterize the 

so-called multiple wh-fronting languages. Recall that, on the one hand, in Serbo-Croatian the first wh-phrase is 

attracted to a wh-related position in the left periphery, whereas the other wh-phrase(s) are attracted to focus 

related position (Bošković 1998, 2002). Bulgarian, on the other hand, is suggested to exhibit several Foci in the 

left periphery (Krapova and Cinque 2005). Furthermore, Romanian was analysed in Chapter IV as involving 

movement of a wh-chunk containing all ordered wh-phrases. 

247

(3b) … ForceP 

3 


Q-feature 3 

3 


wh-feature 3 

ObjP 

3 

Move wh-phrase vP 

3 

(3c) … ForceP 

3 


Q-feature 3 


wh-feature 3 

ObjP 

Agree Agree and/or 3 


3 

(3d)… ForceP 

3 

3 


Q-feature 3 


Wh-feature 3 

ObjP 

Move Agree and/or 3 


3 

The six-way combinatorial possibilities are summarized in Table 1. They give rise to four 

language types which constitute the subject matter of sections 3 and 4. 

248

Q-feature Wh-feature 

+EPP/Agree+Move 

-EPP/Agree 

+Overt 

-Overt 

Table 1 

2. Antisymmetry and the Interrogative ka in Japanese 

I follow Kayne’s (1994) antisymmetry hypothesis and the cartographic approach to the clause 

architecture, such as has been put forth by Belletti (2004), Cinque (1999, 2002, 2006) and 

(Rizzi 1997, 2004a), where the CP (the Vorfeld) and IP (the Mittelfeld) are areas rich in 

functional projections, roughly represented in (4). 

(4) [ForceP [TopP* [FocP [TopP* [FinP … [MoodP [ModeP [TP [AspP [VoiceP …]]]]]]]]]] 

It is obvious that previous analyses of the Japanese clause structure (Saito 1989, 1992, 

Miyagawa 1996, 2001, Ura 1996, etc.) are incompatible with Kayne’s (1994, 2005) 

antisymmetry theory. To my knowledge, rather few attempts have been proposed in the 

literature to analyze Japanese in terms of Kayne’s antisymmetry. One case in point is 

Whitman’s (2001) account of negation in Japanese and Korean in terms of verb raising to 

adjoin to Neg, whose complement VP moves beyond NegP in Japanese. Hoshi (2005) 

proposes pied-piping and remnant movements for his analysis of association with Focus in 

Japanese and Endo (2007) accommodates Japanese to the framework of the cartography. 

Another analysis along the antisymmetry hypothesis is that of Koopman (2005) who provides 

a syntactic account of Korean and Japanese morphology by showing that inflected words are 

derived from head-initial structures by phrasal movement and that agreement can be triggered 

under pied-piping. 

More generally, other SOV languages have been analysed along the antisymmetric 

hypothesis. Thus, Bhattacharya and Simpson (2003) argue in favour of an underlying SVO 

analysis of Bangla thus providing empirical support for a Kaynean account of strongly-headfinal 

languages. 

249

Kayne (1994) argues against the existence of syntactic projections whose complements 

intervene between head and specifier, as has long been assumed for languages like Japanese. 

With regard to wh-movement, Kayne’s view is that languages with clause-final question 

particles lack visible wh-movement. More precisely, in a wh-question, the whole IP moves to 

Spec CP, to the left of the Japanese no/ka question particle. As he states it, this approach can 

capture Bach’s (1971) correlation between lack of wh-movement and head-final languages. 

In keeping with the theoretical assumptions in section 1, Japanese merges no/ka in the 

Force head which takes a Focus complement and attracts it into Spec ForceP due to the 

presence of an EPP feature associated with the particle. I argue that since there is no EPP 

feature associated with the wh-feature present on the Focus head able to trigger XP 

movement, Japanese does not display wh-movement. The wh-phrase is in fact included into a 

larger pied-piped subclausal constituent wh ich gets attracted to Spec ForceP. 

The antisymmetric analysis of Japanese has the implication that the notion of ‘final 

head’ characterizing Japanese needs reinterpreting and that, more 

generally, languages are not 

consistent in being only ‘final head’ or ‘initial 

head’. Thus, Kayne (1994:143, fn.3) proposes 

that particles 

like ga and wa are actually ‘initial’ heads whose overt complements follow 

them and attract a DP/PP. More precisely, he suggests that wa may be a Top head in Rizzi’s 

(1997) sense (see Endo 2007 for an argument that wa may represent the head of distinct 

projections). A ‘final’ head is reinterpreted as a head whose entire complement has moved 

past it to a higher position. The Q-particles ka/no are therefore derived ‘final’ heads. 

Whitman (2001) has expanded on Kayne (1994) and shown that, for instance, the genitive 

particle 

no in a multiple genitive structure illustrated in (5) is the result of the merge of no 

into the D head with a lower DP projection, the argument DP moving into Spec of no (this is 

Whitman’s example 10 bracketed here). 

(5) [DP[DP yuubokumini [D no [DP tosij [D no [ti tj [NP hakai]]] 

nomad Gen city Gen destruction 

‘the nomads’ destruction of the city’ 

Coming back to wh-questions, exemplified in (6) below, no is the overt realization of the Q- 

feature on the Force head in the left periphery of the clause. The head of FocusP hosts a wh- 

feature. The wh-object nani-o ‘what’ vacates the VP domain and so does the subject DP. A 

TP headed by the past tense morpheme -ta is assumed to be above SubjP. Since the high T 

head has an associated 

EPP feature, the large constituent SubjP containing the wh-phrase 

250

undergoes 

pied-piping to Spec TP. This step is represented by the dotted line in (1). 

Subsequently the larger FocusP structure is attracted to Spec ForceP by virtue of the EPP 

associated with Force°, as in (2). All this is shown in (7). 

(6) 

John-ga nani-o katta no? 


'What did John buy?' 

(7) ForceP 

2 

no+EPP FocP 

2 

wh TP 

2 

(2) 2 

ta SubjP 

2 

(1) John-ga ObjP 

2 

nani-o vP 

AGREE 2 

kat … 

Recall that SubjP and ObjP as they are used throughout this thesis are to be understood as the 

equivalent of the former AgrSP and AgrOP to which an Information Structure-feature/value 

corresponds, which renders them interpretable at the syntax-IS interface. 

A point is worth making about the Japanese ka which holds valid of its counterparts in 

other languages to discuss in the following sections, Tumbuka, Sinhala and Lele. The 

Japanese ka particle, labelled here as the Q-particle, which is taken to be prototypical of wh- 

questions, is homophonous with the clause-final ka in yes/no questions, as illustrated in (8). 2 

(8) Taroo-ga sono hon-o yomimasita 

ka? 

Taro-Nom that book-Acc read.Pol Q 

‘Did Taro read that book?’ 

2 

The particle –no which appears in matrix wh-questions, 

is considered a reduced form of -no desu ka, consisting 

of 

a clausal nominalizer -no, ‘be’, and the question marker –ka (Hagstrom 1998, Miyagawa 1998). 

251

As is well known, in a wide variety of languages, question words and quantifier expressions 

show morphological correlations. Thus, in Japanese, the question word suffixed with the 

particle –ka yields an existential quantifier: 3 

(9) Dare-ka-ga hon-o katta. 

Who-Q-Nom book-Acc bought 

Japanese 

‘Someone bought books.’ Kuroda 1965:97 

The same morpheme also occurs with phrasal disjunction, 

as in (10). 

(10) John-ka Bill-(ka-)ga hon-o 

katta. 

John-Q Bill-(Q-)Nom book-Acc bought 

Japanese 

‘John or Bill bought 

books.’ Kuroda 1965:97 

Since the k yes/no questions and of (constituent) wh-questions filling a descriptively 

clause-final position is responsible for clause typing, and since it is the Force head that 

e, interrogative, relative, comparative, different 

7, a.o.), it becomes natural to assume that ka is 

erged in the highest projection of the split CP domain. 4 

a of 

distinguishes various clause types (declarativ 

types of adverbial clauses, etc.; see Rizzi 199 

m 

3 

Japanese use s a distinct marker to get a universal quantifier. 

Thus, -mo contributes universal quantification in 

(i) and existential quantification in the 

NPI in (ii). 

( i) Dare-mo-ga kita. 

What-mo-Nom came 

‘Everyone came.’ 

(ii) Dare-ni-mo aw-a-na-katta. 

Who-Dat-mo meet-not-Past 

‘I did not meet anybody.’ Nishigauchi 1990 (3b):117 

4 

Cable (2006) also suggests an analysis in terms of the wh-feature and universal overt/covert Q-feature and 

argues that cross-linguistically a Q-particle ‘accompanies’ the wh-phrase covertly, as in English, or overtly as in 

Japanese, Sinhala, etc. Crucially, in his approach, the Q-particle is either adjoined to a wh-phrase, as in 

Japanese, or takes it as complement, as in English, Sinhala and Tlingit. Consequently, when the C head probes, 

it encounters QP projected by the Q-particle with whose Q-feature it agrees. Depending on the selecting 

252

A somewhat similar idea is argued for by Jayaseelan (2001) for Malayalam, an SOV 

language, which except for correlative clauses, 

behaves like Japanese. He claims that 

universally, questions contain a disjunction operator 

generated as the Force head which is 

also available to existential quantifiers 

(and in correlatives). In his analysis the disjunction 

operator in Force, a truth-functional operator of logic, ‘types’ the clause interrogative by the 

mechanism by association with Focus (Rooth 1992). 5 

head/adjunction position of the Q-particle, eith er the Q-particle alone gets moved to C, 

or the entire QP 

containing the wh-phrase, that is there is no syntactic relat ionship between the interrogative 

C in Cable’s 

analysis and the wh-phrase itself. However, I fond that there are several problems with such an analysis: first, 

why, if Sinhala and Tlingit have the same structure, it is the latter only that has wh-movement; second, why the 

Q-particle 

should be adjoined to the projection containing the wh-element in some languages, while in others it 

should take it as its complement; third, as discussed in Nishigauchi (1990), Japanese 

presents cases where ka 

follows the case 

marker of a DP; fourth, he argues that when C agrees with the Q-particle, it is the whole 

projection 

QP containing Q and the projection of the wh-phrase that moves. If so, either feature percolation or 

the pied-piping mechanism should be involved, which is exactly what his analysis explicitly rejects. 

5 

In Malayalam the question word suffixed with the particle –oo yields an existential quantifier (i). Like 

Japane se, a distinct morpheme, -um, is used for univers al quantifi cation. Similar to Japanese, 

the same 

morpheme 

–oo occurs in phrasal disjunction, as illustrated in (ii). 

(i) ñaan aar-e-(y)oo kaNDu. 

I who-Acc-oo saw 

‘I saw somebody.’ Jayaseelan 2001 (18) 

(ii) John-ine-(y)oo Bill-ine-(y)oo 

John-Acc-oo Bill-Acc-oo 

‘John or Bill’ Jayaseelan 2001 (4c):65 

However, as opposed to Japanese, Malayalam does 

not exhibit this marker in (constituent) questions except for 

those 

with a cleft. Jayaseelan argues in favour of the existence a covert –oo marker in Malayalam questions. 

Give n that yes/no questions involve 

(at least) an implicit disjunction and its negation (Larson 1985), the particle 

–oo is treated as a disjunction morpheme, like the English or. 

Jayaseelan goes a step 

further and argues in favour of the idea that the interpretation of question should be 

assimilated to the interpretation of disjunction, 

which in its turn is treated similarly to the focusing particles only 

and even (Rooth 1992, a.o.). Both disjunction 

and the two particles involve focused elements and so does a whinterrogative. 

Another shared property 

is that a single disjunction operator can be associated with or can license 

two or more variables (question words). The same again is true for wh-interrogatives. As was 

seen in Chapter V, 

the 

same configuration obtains for Japanese as well. 

253

Although existentials (9) and disjuncts (10) fall outside the domain of interest here some 

remarks need however be made. It may be the case that they should be treated differently. 

In the case of the existential quantifier in (9), it may firstly be suggested that it enters 

the derivation as a whole, i.e. [wh-form + Q]. 

Secondly, wh-words in Japanese have been analyzed on a par with Chinese, i.e. as pure 

variables that can be bound by non Q operators. For instance, Nishigauchi (1990) argues that 

the operators ka (and mo) are adverbial quantifiers that can unselectively bind variables made 

available by indeterminate phrases. His analysis has 

been criticized for deriving incorrect 

interpretations for questions and universal quantification structures (see von Stechow 1996, 

Ohno 

1989; cf Shimoyama 2006 for a purely semantic account). 

The disjuncts in (10) may also be suggested to involve a disjunction operator scoping 

over all disjuncts within its c-command domain much in the spirit of Jayaseelan (2001). 

Alternatively, it may be assumed in line with Kayne (1994, 1998) that the first disjunct is 

located in the specifier position of the first ka ‘or’ and that John-ka Bill fills the specifier of 

the second ka. 

Since existentials (or disjuncts) do not constitute the domain of investigation here, I do 

not take a position as which analysis seems most appropriate. 

Coming back to the Q-particle, as mentioned in section 1, it has been analyzed as 

merged clause-internally. Thus, Hagstrom (1998:30) argues that in languages like Sinhala, to 

be dealt with in section 3, and Japanese, the Q-particle ‘must be as close (hierarchically) to 

the wh-word as it can be without being separated from the associated interrogative clause 

periphery by an island boundary.’ 6 In other words, ka is ‘adjacent’ to the wh-DP/PP/AdvP 

and moves overtly in Japanese to the periphery of the clause (Hagstrom remains agnostic as 

to the exact nature of the hosting projection). However, such an analysis where the question 

particle moves via head movement 

to the left periphery is in conflict with locality constraints, 

7 

i.e. the Head Movement Constraint. Furthermore, if ka is “merged with 

the wh-word”, it 

then 

excorporates and one needs posit some constraints on this type of movement which 

leaves behind the wh-part. 

In his analysis, the wh-word is a focused variable in the c-command domain of a disjunction operator 

merged in 

the head Force. From this position it applies to question words 

by association with focus, yielding the question 

interpretation. 

6 Hagstrom replaces the notion of ‘close to’ by ‘as sister of’: Q actually starts as sister of the (lowest) wh-phrase. 

7 The solution that Hagstrom (1998) proposes is to claim that the HMC does not actually apply in this case and 

adopts feature-sensitive head movement. 

254

To resume, out of the descriptively ‘adjacent’ vs. long-distance cases with ka that 

Japanese exhibits, it is only (12b) that is of interest here. (12c) and (12d) are given for 

comparative 

purposes. 

(12) 

a. wh-ka c. wh-mo 

b. [Clause wh …-ka] 

b’.[ CNP/Adjunct Island wh …-ka] d. [CNP/Adjunct Island wh …-mo] 

Therefore, in the approach suggested here, clausal typing arises as a result of Agree between 

the Force head and its complement, whereas wh-feature valuation 

is a result of Agree 

between 

the wh-feature on the Focus head and its counterpart on the wh-element. Such an 

approach departs from the movement analysis proposed by Nishigauchi 1990, Watanabe 

1992, von Stechow 1996, a.o. 

I follow Pesetsky and Torrego’s (2004) system of features and their view of valuation 

and interpretability. 

Crucially, their system does away with Chomsky’s (2001) view of 

Valuation/Interpretability, 

which says that a feature F is uninterpretable iff F is unvalued. 

They argue in favour of a feature-sharing view of AGREE ( informally put, valuation of a 

feature 

F2 by another F1 creates a link which is accessible to subsequent processes, such as 

another instance of AGREE) coupled with the idea that syntax identifies a feature as probe not 

because it is uninterpretable but because it is unvalued. Thus, they suggest two novel 

combinations of properties: uninterpretable valued features (uF[val]) and interpretable 

unvalued features (iF[ ]). The existence of such probing features in P and T’s system makes 

the Q- and wh-features in the MI/DbP approach redundant. The analysis suggested here is 

cartographic in nature, i.e. each feature on a projection is responsible for a particular 

discourse-semantic interface interpretation. For this reason, I adopt the MI/DbP distinction Q- 

Wh but I recast the two features on two distinct heads, i.e. Force, hosting Q, and Focus, 

hosting 

Wh. 

Therefore, following Pesetsky and Torrego, one may posit that nani in (6) repeated 

below comes from the Numeration with an uninterpretable valued wh-feature. I abstract away 

from how the Case particle gets adjoined. The Focus head has an iwh[ ] feature – 

interpretable but unvalued – which acts as a probe and receives its value from the 

uninterpretable counterpart uwh val on the wh-word. Example (13) illustrates this. I follow 

their notation: a feature F that has not participated 

in Agree is indicated by an empty pair of 

brackets, F [ ] if unvalued, or F val [ ] otherwise. 

255

(6) John-ga nani-o katta no? 



AGREE 

(13) 

… Focus … wh-elementObj/Subj …. ⇒ …Focus … wh-phraseObj/Subj 

iwh[ ] uwh val iwh [2] uwh [2] 

It is as a result of Agree that the wh-word is actually interpreted as a proper 

wh-phrase. The 

fact that nani must have a wh-feature is in accordance 

with the Matching condition of the 

definition of Agree in (2) above. 

8, 

9 

A similar mechanism is envisaged for 

the Q-particle in 

order to distinguish between an interrogative and yes/no interpretation. The head Force may 

be suggested to have an interpretable unvalued interrogative Force feature which enters into 

Agree with its complement bearing a counterpart uninterpretable valued feature. 10 It may be 

the case that this AGREE relation is prompted by selection between Force and Focus. 

The idea that the Q-feature 

is relevant to typing the sentence interrogative is highly 

reminiscent 

of the Clausal Typing Hypothesis proposed by Cheng (1991). Cheng’s hypothesis 

is accomplished either by wh-movement or by a question particle. Put differently, a language 

will have either wh-movement or a question particle. No language will have both or neither. 

As argued by Bruening (2007), these predictions are not correct in the sense that there is no 

direct relation between question particles and wh-in-situ, on the one hand, and between wh- 

indefinites and wh-in-situ, on the other. Further, it will be shown in the sections below that 

the existence of languages with both a Q particle and wh-movement emerges from the system 

to be proposed. 

8 Chomsky’s (1998) feature system in wh-questions is fairly unclear in the relation Match and convergence. The 

uninterpretable wh-feature on the wh-phrase is automatically deleted as a side effect in the process of Match and 

Agree with uQ on C. 

9 

I do not indulge into the semantics for wh-words and wh-indefinites in any of the languages discussed in this 

paper. See Reinhart (1997), Hagstrom (1998), Kratzer and Shimoyama (2002), Beck (2006), among others. 

10 

Likewise, it may be assumed that the (indefinite/indeterminate) existential interpretation results from a similar 

process of Agree between an appropriate existential feature on a head X - interpretable but unvalued – which 

probes and receives its value from an uninterpretable 

valued counterpart on the wh-element. 

256

In what follows this typology will be discussed in detail with exemplifications from each 

language type. 11 

3. Languages with an Overt Q Particle. Covert Q-Feature Languages 

Within a framework relying on Q-Wh feature split, on AGREE and AGREE + MOVE, and 

potential EPP, the parametrisation regarding 

the Q-feature is given below. 

A. Q-feature [-EPP] 

languages 

In the absence of an EPP-feature associated 

with the Q-feature on Force, a language L has 

only AGREE between Q and its c-commanded (clausal) complement; a 

Tumbuk a (particle-initial language). 

(i) overt initial particle: Tumbuka 

(ii) covert particle: French on the movement/in-situ strategy, English, 

Romanian/Bulgarian, Tlingit 

B. Q-feature [+EPP] languages 

case in point is 

If an EPP-feature is associated with the Q-feature, a language L has AGREE+ MOVE, i.e. the 

selected complement moves to Spec ForceP; these are Japanese, Korean and Chinese; 

Sinhala; Vata; Lele. Sinhala will be discussed more in detail in section 5. 

(i) overt (final) particle: Japanese, Korean, Chinese, Vata, Lele, Sinhala 

(ii) covert particle: (Chinese) 

Let us take some examples and discuss these parameters in more detail. 

One language type 

is represented by Tumbuka, a Bantu language spoken primarily in 

Malawi. Consider the examples below. 

11 The discussions to follow and, more generally, the contents of this chapter are a revised version of Soare 

(2007a,b). 

257

(14) Kasi Suzo a-ka-p h ik-a vici 

Q Suzo SM-pst-cook-fv what 

‘What did Suzo cook?’ Kimper 2006, (4) 

(15) Kasi u-ka-rut-a koci 

Q SM-pst-go-fv where 

‘Where did he go?’ Kimper 2006, (6) 

s discussed by Kimper (2006), Tumbuka exhibits the Q-particle kasi at the left periphery of 

the cla adjunct in (15) remains in situ, neither can undergo 

h-movement. 12 A 

use and the wh-object in (14) or 

w 

It patterns with Japanese and other languages in that the Q-particle kasi is 

12 

In Tumbuka, the Q-feature can occur on the matrix Force head and the wh-phrase is located in the embedded 

clause, 

as in (i). Under the assumption that AGREE is cyclic, the embedded Focus head agrees with the whphrase. 

(i) 

Kasi 

Q SM-pres-think-fv 

‘Where do you think he went?’ Kimper 2006, (7) 

n k h -u-ghanaghan-a kuti u-ka-rut-a koci 

C SM-pst-go-fv where 

In addition, Tumbuka appears to be a mixed language 

in that it shows AGREE, without MOVE, with an object and 

an adjunct, i.e. no wh-movement, but optionally has wh-movement with a subject, as in (ii) and (iii). I leave the 

question 

open as to why this is possible only with subjects. 

(ii) njani ( 

‘Who (do you think) made that noise?’ Kimper 2006, (8, 11) 

n k h -u-ghanaghan-a) o-pang-a congo ico 

Who SM-pres-think-fv SM-make-fv 7-noise 7-dem 

(iii) 

kasi ( 

ake-fv 7-noise 7-dem who 

‘Who (do you think) made that noise?’ Kimper 2006, (12, 13b) 

n k h -u-ghanaghan-a kuti) o-pang-a congo ico njani 

Q SM-pres-think-fv C SM-m 

When there is wh-movement of the subject, the Q-particle is absent. Conversely, when the wh-subject is final, 

the Q-particle is obligatory. One may speculate that the absence of the Q-particle on the movement option is 

related to economy reasons. 

258

also used in yes/no questions. The Q-particle fills the Force head and c-commands its selected 

complement which contains the wh-phrase. The derivation for (14) is given in (16). 

(16) ForceP 

2 

kasi 

FocusP 

2 

wh SubjP(=A grSP) 

2 

Suzo vP 

2 

vici 

In (16), the null iwh [ ] on the Focus hea d – interpretable unvalued – acts as a probe and 

receives its value from the uninterpretable valued counterpart on the wh-object vici ‘what’. 

Consequently the wh-interpretation of vici obtains and its uninterpretable feature gets deleted. 

At the next step of the derivation, the Force head selects its complement, FocusP, and an 

AGREE relation obtains. Since Force in Tumbuka is marked –EPP, no MOVE component is 

needed. 

a-ka-p h ik-a 

AGREE 

Consider next examples of French on both the movement and the in-situ strategy. 

(17) a. Qui as-tu vu? 

Who have-you seen 

‘Who did you see?’ 

b. Tu as vu qui? 

You have seen who 

‘Who did you see?’ 

(18) Who did you see? 

French exhibits the optionality of moving a wh-phrase 

to the left periphery of the clause or 

leaving it in situ. There are several analyses of French 

wh-in-situ, describing (at least) two 

varieties. One is that of Mathieu (2004) who shows that French exhibits no optionality 

at all 

and 

that the EPP feature on C is always strong. Rather, optionality resides with the fronting of 

the nominal part with which the phonologically null wh-operator 

is associated in a split DP 

analysis. In his analysis, the moved wh-element correlates with 

one reading, i.e. specificity in 

259

a prominent context, whereas the in-situ wh-element may, but need not, correlate with 

another. 13 

Concentrating more on the mapping between the syntax, prosody and semantic 

interpretations of the wh-in-situ, Baunaz (2008), partly relying on Starke (2001), argues that 

there are three types of wh- in-situ in French, each correlating with three distinct intonations, 

and having distinct semantic readings. o not enter into the complex issue of the possible 

14 I d 

interpretations associated with 

the moved wh, and whether there is a one-to-one mapping 

between the syntax and interpretation 

of the in-situ wh-phrase and of the moved one. It may 

be suggested that both English and French 

wh-interrogatives contain a covert Q-feature. Such 

an 

idea has already been forwarded for French by Cheng and Rooryck (2000). More exactly, 

the intonation in yes/no questions is tightly linked to that of wh-in-situ questions. They 

postulate an underspecified (yes/no or wh) Q-morpheme in the overt syntax, which can be 

disambiguated by LF wh-feature movement to this C position. Consequently, in wh- 

questions, the value of Q is set to [wh]. Naturally, more needs to be said about the 

correlations between this left-peripheral Q and the distinct prosodic contours of wh-in-situ 

identified by Baunaz. I leave this issue open. 

The suggestion that there is a covert Q-feature is tenable for multiple wh-fronting 

languages 

like Romanian (and Bulgarian). 

(19) Cine va cumpăra maşina asta veche? 

Who will buy car.the this old 

‘Who will buy this old car?’ 

I have analyzed Romanian 

as involving wh-chunk movement (i.e. movement of a 

substructure 

containing all and only wh-phrases) to the sole Spec FocusP available in the left 

periphery (see also Grewendorf 

and Sabel 1999, Grewendorf 2001 and Sabel 2001 and the 

Wh-Cluster 

Hypothesis). Empirical evidence that such (multiple) wh-movement languages 

have a Q-particle comes from Romanian and Bulgarian. Thus, Romanian has a particle 

corresponding to the Japanese ka, also used in unmarked yes/no questions, as in (20). 

13 Interpretively, Mathieu’s approach to (a certain) in-situ variety of French goes against that of Chang (1997) 

and Cheng and Rooryck (2000). 

14 Baunaz (2008) arrives at the conclusion that the interpretations induced by wh-in-situ are retrievable with 

moved wh-phrases, ultimately 

suggesting that such optionality may have to do with register (i.e. colloquial 

French 

vs. formal French). 

260

(20) Oare va spune ceva directorul de la Dacia? 

Q will say something director.the at Dacia 

‘(I wonder) will the director at Dacia say anything?’ 

‘Should the director at Dacia say anything?’ 

(21) a. Oare cine (oare) va cumpăra (oare) maşina asta veche ( 

Q who Q will buy car.the this old 

‘(I wonder) who will buy this old car?’ 

b. Oare Mariei (oare) i-a dat (oare) cartea (oare)? 

? oare)? 

I will not enter into a discussion of the distributional properties of oare in Romanian but 

content myself with noting that it may occur in several positions. Thus, in (21a), Q-particle 

may be in clause initial position and scopes over the whole proposition. This is expected 

under the assumption that it is the realization of the Force head. It may also follow the whelement 

and emphasis lays on cine (Who it is that will buy…?) The Q-particle may also 

follow the verb and even the whole clause. Without proposing any analysis of the behaviour 

of oare, I only suggest that it is a head that may optionally raise to several head positions in 

the left periphery. Similarly, in (21b), if the DP Maria is Topic, then oare may also raise to 

several head positions, the highest one, being Force°. It may also move from a position under 

the Contrastive Focus if Mariei is interpreted as Focus. 

Somewhat similarly, Bulgarian exhibits 

the particle li in yes/no questions which is 

enclitic 

and subject to the second position requirement (Bošković 2001, a.o.). It is analyzed as 

an interrogative complementizer to which a (complex) X-element(s) head-adjoins. A case in 

point is (22). 

(22) 

Čete li knige? 

Reads Q books 

‘Is s/he reading books?’ Bošković 2001 (31a) 

Given that li is subject to the second position requirement (there is no such requirement for 

the Romanian oare), it cannot 

occur clause-initially, but follows the first wh-phrase, as in 

(23) below. 

261

(23) Koj li kakvo kupuva? 

Who Q what buys 

‘Who buys what?’ Bošković 2001 (63) 

Without insisting on this kind of structures, we follow Bošković in assuming that li heads the 

projection in whose specifier koj ‘who’ has moved, possibly 

Spec FocusP. Under his 

prono unce-a-copy analysis the second wh-phrase is pronounced in a lower position of the 

chain 

created by its movement to the left periphery. So due to language-internal factors, i.e. 

the second position requirement, this particle can only be the realization of the wh-feature on 

Focus in wh-questions. 

Altho ugh enumerated among these languages, Tlingit will be dealt with in section 4. 

In additi on to the above-mentioned language type where the 

Q-feature does not have an EPP, 

there 

is another language type where the Q-feature has an EPP-feature. The expectation is that 

languages belonging to this group have (sub-)clausal pied-piping movement. We will 

consider the case of Japanese, Chinese and Sinhala and leave languages like Vata and Lele 

for a more thorough discussion in section 4. 

Let us consider the Japanese example (6) repeated below. 




It has been discussed in Chapter IV and in section 2 of this chapter that the Q-feature on the 

Force head in the Japanese left periphery has an EPP-feature that attracts its complement, i.e. 

FocusP, containing the wh-phrase nani-o, into Spec ForceP. Movement of the entire structure 

thus accounts for the final sentence position of the Q-particle. 

Chinese may also be considered a language allowing the Q-wh feature split. As pointed 

out 

in Li (1992), the presence of a question marker requires the co-occurrence of a wh- 

element. lustrated below, where the Q-particle ne co-occurs with the subject wh- 

15 This is il 

15 

However, there are cases where the question particle is absent. In the matrix clause (i) below and the 

embedded context selected by wonder, the wh-word s can on ly have wh-interpretation. It may be 

assumed that 

even 

in such cases an invisible Q-particle is merged in the Force head. The issue of the restrictions under which 

the interrogative interpretation obtains in the absence of an overt Q-particle falls outside the scope of this paper. 

262

phrase shei in (24a) and with the object wh-phrase shenme in (24b). Importantly, Chinese 

differs from Japanese, Tumbuka and Romanian (and Bulgarian) but, as will be seen, patterns 

with Tlingit, in exhibiting the special Q-particle, ne, which appears only in wh-interrogatives. 

It 

has a distinct particle, ma, in yes/no questions (Ta xihuan shenme ma? ‘Does he like 

anything?’) 

(24 ) a. Shei xihuan ni ne? 

Who like you Q 

‘Who likes you?’ Li 1992, (36b) 

b. Hufei chi-le shenme ne? 

Hufei eat-Asp what Q 

‘What did Hufei eat?’ Cheng 1991, 112 (1) 

Though Chinese uses wh-words as indefinites which require a licenser (a nonfactive verb, 

negation, a yes/no question 

particle, a modal or a conditional), the presence of the question 

particle ne forces the wh-interpretation. This is further illustrated in (25) and (26). 

(25) a. Ta yiwei shei xihuan shenme. 

He think who like what 

‘He thought somebody liked something.’ 

b. Ta yiwei shei xihuan shenme ne? 

‘Who(x), what(y), he thought x likes y?’ 

(26) a. Yaoshi shei xihuan shenme, ta jiu gaoxing le. 

If who like what he then 

happy 

‘If somebody likes something, 

he would be happy.’ 

b. Yaoshi shei xihuan shenme, ta jiu gaoxing ne? 

(i) Shei/Shenme ren xihuan shenme? 

Who/What man like what 

‘Who likes what?’ Li 1992 (7) 

(ii) Wo xiang-zhidao shei xihuan ni. 

I wonder who like you 

‘I ownder who likes you.’ Li 1992 (37b) 

263

‘Who(x), what(y), if x likes y, he would be happy?’ Li 1992, (33) 

In (25a), the presence of the non-factive verb determines the indefinite interpretation of the 

wh-words. The presence of an overt Q-particle forces the wh-interpretation in (25b). 

Similarly, in (26a), both wh-words in the conditional clause are wh-indefinites but get the whinterpretation 

in (26b) due to the Q-particle ne. 16 In technical terms, the wh-interpretation is 

the 

result of the AGREE relation between the interpretable unvalued wh-feature on Focus° and 

its uninterpretable valued counterpart 

on the wh-element. Since ne occupying Force° has an 

EPP, the complement Focus P is attracted to Spec ForceP. This movement accounts for the 

final position of the interrogative particle. The representation of (23a) is provided below. 

(27) 

ForceP 

2 

2 

ne FocusP 

2 

wh SubjP 

2 

shei vP 

AGREE 2 

xihuan ni 

4. Languages with a (C)overt Wh 

Feature 

So far we have arrived at a parametrisation 

regarding the Q-feature which is expressed by 

points A and B. In what follows, the wh-feature is 

considered. 

C. Wh-feature [-EPP] languages 

If a language has only AGREE between Focus and the wh-phrase within SubjP/IP, it has ‘wh- 

in-situ’, a case in point being Japanese, 

Chinese; Sinhala; Tumbuka; French on the in-situ 

strategy (all non-wh movement languages). 

16 

For such cases with two wh-phrases in their canonical order, the mechanism of MULTIPLE AGREE is assumed to 

be 

involved. I leave open the issue of how the wh-phrases in the if-clause take matrix scope. 

264

(i) overt particle: none (possibly Lele) 

(ii) covert particle: Japanese, Korean, Chinese, 

Sinhala, French on the in-situ 

strategy, Tumbuka (in-situ languages) 

Importantly, the analysis here concentrating on ten languages 

seems to suggest that there is a 

strong correlation between the overt wh-feature and the EPP featu re. 

D. Wh-feature [+EPP] languages 

If a language has AGREE+MOVE, it has wh-movement of a wh-XP to Spec FocusP, as in Lele, 

Tlingit, Vata, Romanian, etc. 

(i) overt particle: Tlingit (multiple wh-movement language), Lele 

(ii) covert particle: Vata (single wh-movement language) 

French (on the movement strategy), English (single wh- 

movement languages) 

Romanian/Bulgarian (multiple wh-movement languages) 

We first discuss the language type 

at point C. As argued in section 2, in Japanese (and 

Korean), the Focus head has a covert 

interpretable unvalued wh-feature that must enter into 

Agree with its counterpart on the wh-word 

for reasons of wh-feature valuation. The fact that 

the wh-feature does not have an EPP feature accounts for the lack of wh-movement in 

Japanese 

(and Korean). 

Sinhala patterns with Japanese in exhibiting a covert wh-feature without an EPP feature 

and thus absence of wh-movement. An analysis 

along similar lines accounts for the Chinese 

data 

considered in the section above. Likewise, French on the non-movement strategy is 

another case in point (28a). The Focus head has a covert wh-feature which agrees with the 

wh-phrase inside SubjP/IP. below. 

17 This is schematically represented 

(28) 

a. Tu as parlé à qui? 

AGREE 

b. Force° … Focus° … [SubjP/IP …… à qui] 

17 Belletti (2006) suggests that wh-in-situ in French “stops in the low VP peripheral focus position”. 

265

Tumbuka patterns with these languages. 18 Consider (14) repeated below. 

(14) Kasi Suzo a-ka-p ik-a vici 


‘What did Suzo cook?’ 

h 

At face va lue, Tumbuka is the counterpart of French with a realized Q-particle, kasi, 

in 

Force°. As in the case of Fr ench, a covert wh-feature fills Focus° and agrees with the wh- 

element 

inside SubjP/IP. 

As opposed to the cases showing no EPP feature, there is a language type which has an 

overt particle with an associated EPP feature. As shown at point 

D, such a particle, 

substantiating 

the wh-feature, may be either overt or covert. 

Thus, Tlingit, belonging to the Na-Dene 

language family, spoken primarily in British 

Columbia, 

is a predominanty SOV language which freely allows the VO order. Not only does 

it overtly realize Focus° as a particle (following our assumptions) but it also has whmovement 

to the left periphery. 

As discussed in Cable (2006), a wh-question in Tlingit must 

contain the particle sá. If the particle is absent, the sentence is ill-formed. Two relevant 

examples 

are shown in (29). 

(29) a. Daa *(sá) aawaxáa i éesh? 

What wh he.ate.it your father 

‘What 

did your father eat?’ 

b. Goodéi *(sá) kkwagóot? 

Where.to wh I.will.go 

‘Where will I go?’ 

The wh-phrase must be followed by the particle. Tlingit patterns with some of the languages 

discussed above in that wh-words 

can function as indefinites and the presence of the particle 

is obligatory (Gungbe also exhibits an obligatory particle, the realisation of the wh-feature; 

see Aboh 1999). 

18 

Recall from footnote 12 that Tumbuka optionally has wh-movement of the subject. It would be interesting to 

see whether there is any detectable semantic or pragmatic difference between the in-situ and the movement 

option. 

266

(30) Tlél ggdéi *(sá) xwagoot. 

Not where.to wh I.went 

‘I didn’t go anywhere.’ 

However, it differs from Japanese in that sá occurs neither in yes/no questions, nor in 

disjunctions. Without entering into a detailed discussion here, 

this may actually not be 

accidental. For this reason, here sá is not glossed as Q, as Cable does, but 

as the wh-feature 

proper. As will be seen in (34) and (35) below, one further distributional argument comes 

from m ultiple wh-fronting in multiple wh-questions. 

Turning to wh-questions, the cases below follow naturally from the analysis adopted. 

(31) a. Daa sá i éesh aawaxáa? 

What wh your father he.ate.it 

‘What did your father eat?’ 

b. * Sá i éesh daa aawaxáa? 

Cable 2006, (60b) 

The wh-feature with an EPP on the Focus head attracts the wh-DP into its Spec in (31a). This 

is wh-movement. The AGREE relation between the wh-feature 

on Focus° and its counterpart 

on the 

wh-phrase across the DP subject i éesh ‘your father’ does not suffice, as suggested by 

the ungrammaticality of (31b). A Spec-head configuration must obtain. This is represented in 

(32). 

(32) ForceP 

2 

Q-EPP FocusP 

2 

daa 2 

sá+EPP SubjP 

2 

i éesh ObjP 

2 

AGREE+ t vP 

MOVE 2 

aawaxáa 

267

Cable’s argument that the wh-phrase moves to the left periphery of the clause comes from the 

fact that wh-words in wh-questions 

must precede the main predicate of the clause (his 

examples 

(59)). 

(33) a. [[Goodéi] i sá [has uwajée [ti woogootx] i shagóonich]]? 

Where.to wh they.think he.went your parents.erg 

‘Where 

do your parents think he went?’ 

b. * [Goodéii [has uwajée [ti sá woogootx] i shagóonich]]? 

where.to they.think wh he.went your parents.erg 

c.*[Goodéii [has uwajée [ti woogootx sá] i shagóonich]]? 

where.to they.think he.went wh your parents.erg Cable 2006, (59) 

The EPP associated with the wh-feature on Focus attracts an XP-wh, as in (33a). A wellformed 

question must have the wh-particle in its left periphery, more precisely in the matrix 

Focus. (33b,c) are thus ruled out. 19 

Tlingit allows multiple wh-phrases and the expected 

Superiority effects arise, as in (34) 

and (35). 

(34) a. Aa sá daa sá aawaxáa? 

Who wh what wh they.ate.it 

‘Who ate what?’ 

b. * Daa sá aa sá aawaxáa? Cable 2006, 

(24) 

19 As in the Sinhala, in Tlingit, sá can also be preceded by the entire embedded clause, i.e. it is no longer 

adjacent to the wh-element, as seen in (i). 

(i) [[Goodéi woogootx sá]i [has uwajée ti i shagóonich]]? 

where.to he.went Q they.think your parents.erg 

Where do your parents 

think he went? Cable 2006, (59) 

In (i), the wh-phrase moves to the edge 

of the clause which subsequently raises to the matrix Spec FocusP in 

whose head the particle sits. By allowing 

fronting of the embedded clause containing the wh-phrase, Tlingit 

becomes the counterpart of Basque and Quechua 

without an overt Q-particle. 

268

(35) a. Aa sá goodéi sá woogoot? 

Who wh where.to wh they.went 

‘Who went where?’ 

b. * Goodéi sá aa sá woogoot? 

Cable 2006, (25) 

An argument wh-phrase precedes an adjunct wh-phrase and the wh-particle immediately 

follows the wh-phrase. If sá were the Q-feature located in Force°, one would have a difficult 

time accounting for its reduplication. 

In order to account for the cases of multiple sá with 

multiple 

wh-elements, it would be interesting to investigate whether generally Tlingit has 

multiple Foci. In the absence of further empirical evidence, data like (34) 

and (35) may be 

suggestive of the presence of two Focus positions (although this runs counter Rizzi’s 1997 

‘one Focus per clause’ 

condition). 20 

These facts make Tlingit look like the counterpart of multiple wh-movement 

languages 

like Romanian and Bulgarian, which do not realize the wh-particle (besides, 

Tlingit also 

exhibits island effects). 

(36) 

a. * [Waa] sá [[kligéiyi] xáat] i tuwáa sigóo? 

b. 

How Q it.is.big.REL fish your spirit it.is.glad 

‘How big a fish do you want?’ 

(A fish that is how big do you want?) 

[[Waa kligéiyi] xáat] sá i tuwáa sigóo? Cable 2006, (67-68) 

The analysis suggested correctly 

predicts the contrast between (36a) and (36b). The 

impossibility of (36a) is a straightforward result 

of the fact that the particle is merged in 

Focus° and attracts the entire island containing the wh-element into its specifier. Extraction of 

the wh-phrase alone results in ungrammaticality because of Barriers/Subjacency. 

Another language that overtly realizes the wh-feature which has an EPP is Lele. It is a 

Chadic language, which is SVO. It patterns with Japanese 

in that both yes/no questions and 

wh-questions require the presence of a clause-final 

particle, gà. 

20 

In Cable’s analysis where sá is Q and where the wh-phrase is part of a projection sister to Q, tucking-in is 

responsible for multiple wh-fronting. For theoretical reasons discussed in detail in Chapter III, this mechanism is 

not adopted here. Besides, as stressed out, in the analysis proposed, sá is not 

a Q-particle, but rather a wh- 

particle. 

269

(37) a. 

Kiya hàb kùlbá ke-y gà? 

Kiya find cow Gen.3sg[M] 

Q 

‘Did Kiya find his cow?’ Aboh 2007 (18a) 

b. Mè ày 

wéy gà? 

2sgF marry who Q 

‘Who did you marry?’ 

Aboh 2007 (18b) 

As stressed out by Aboh (2007), Lele dispays 

both the in-situ and the movement strategy with 

a wh-phrase. Importantly, on the movement option, a focus marker, ba, appears. 

This means 

that 

the wh-particle does not start out as part of a more complex form, wh-DP + particle. 

This is illustrated in (37c) below. 

c. Me ba gol di gà? 

What Foc see 3sgM Int 

‘What did he see?’ Aboh 2007 (18c) 

Aboh proposes the following analysis for (51c): the wh-phrase is probed by the overt whfeature 

which by virtue of having an EPP attracts me to its specifier. Since the Force head, 

realized as gà, also has an EPP, its entire complement, FocusP, is pied-piped to Spec ForceP. 

This is shown in (38). 

(38) ForceP 

2 

2 

gà FocusP 

2 

me 2 

ba … SubjP/IP 

2 

… 

gol di tme 

It has by now become clear that languages like Tlingit, Lele, and Vata to consider below, are 

in obvious violation of Cheng’s (1991) typology. Tlingit and Lele have an overt wh-particle 

270

in addition to obligatory overt fronting of the wh-phrases, whereas Vata has wh-movement in 

addition to an overt Q-particle. 

In addition to languages that realize a wh-feature which has an EPP-feature, there are 

languages that exhibit wh-movement to the left-peripheral Focus but such a feature is covert. 

This is the case of French (on the movement strategy) and English, 

on the one hand, and of 

Romanian 

and Bulgarian, on the other. 

(39) Qui as-tu vu? 

Who have you seen 

‘Wh o did you see?’ 

(4 0) Who did you see? 

(41) Pe cine ce ai întrebat? 

Pe who what have asked 

’Who did oyu ask what?’ 

(42) 

Kogo kakvo e pital Ivan? 

Whom what is asked Ivan 

‘Who did Ivan ask what?’ 

In all these languages, the covert wh-feature on the Focus head agrees with the wh-phrase 

inside IP/SubjP prior to its movement to the left periphery. Due to the EPP associated with 

the wh-feature, the wh-phrase moves to Spec FocusP in French and English. 

As far as Romanian is concerned, it was seen in Chapter IV that wh-phrases form a whchunk 

inside the IP/SubjP. 

Due to the EPP associated with the wh-feature, the chunk moves 

to Spec FocusP. As for Bulgarian, as seen in Chapter III, the two wh-phrases in (42) get 

probed by distinct 

features on two features on distinct heads (a wh-feature and a focus 

feature, following Bošković 1999). 

Another language 

which does not realize the wh-feature but has wh-movement is Vata. 

It is a Kru language spoken 

in the Ivory Coast which exhibits wh-movement in addition to a 

p article fillin g the clause-final position 

(Koopman 1984). Consider (43) below. 

271

(43) a. àlÓi Kòfí yE ti yé là 

Who Kofi saw Prt Q 21 

‘Who did Kofi see?’ Koopman 1984, (45a) 

b. àlÓ Ò nÙ mÍ là 

Who he did it Q 22 

‘Who did it ?’ Koopman and Sportiche 1986, (4a) 

We take the particle là to be the realization of Force. Within 

the analysis suggested here, a 

covert wh-feature on Focus agrees with a wh-phrase inside SubjP/IP. Since, as assumed, it 

has an EPP-feature, MOVE then triggers attraction of the goal, the object wh-phrase àlÓ 

‘who’, into Spec FocusP at the left periphery of the clause in (43a). Such movement is 

followed by movement of FocusP into Spec ForceP, in whose head the Q-particle là, 

associated with 

an EPP, has been merged. This is shown in (44) below. 

(44) ForceP 

2 

2 

Q-fEPP FocP 

là 2 

àlÓ 2 

Foc SubjP/IP 

wh-fEPP … 

wh-mvt 

To 

conclude, the typology of question formation comprising ten languages, which relies on 

Q- and wh-feature feature split, on their covert 

or overt nature, and a potential EPP feature 

associated with either or both, is summarized in Table 2. 

21 

Koopman (1984) and Koopman and Sportiche (1986) gloss là as wh to indicate that the clause is a whquestion. 

This corresponds to the Q-particle typing the clause interrogative. 

22 

Note that with subjects, Vata shows a resumptive pronoun. With objects, it shows a trace. This does not affect 

the analysis proposed. 

272

Languages – matrix contexts 

Featu Realiza EPP Japane Chines Sinhala Lele Vata Tumbu French(m Romanian Tlin French 

re tion 

se e 

ka vt)/ Bulgarian git (‘in- 

English situ’) 

Q Overt +EPP ∨ ∨ ∨ ∨ ∨ 

-EPP ∨ 

Covert +EPP (∨) 

-EPP ∨ ∨ ∨ ∨ 

Wh Overt +EPP 

-EPP 

(∨) ∨ 

mult 

iple 

Covert +EPP ∨ ∨ ∨ 

multiple 

-EPP ∨ ∨ ∨ (∨) 

Table 2 

∨ ∨ 

5. A Case of Q [+EPP] and Wh [-EPP] Language: Sinhala 

Another language that falls within the typolo gical group of languages which do not exhibit 

wh-movement 

and display an overt particle, the realisation of the Q-feature, is Sinhala, an 

SOV language, spoken in Sri Lanka. In many respects question formation in Sinhala is very 

similar to question formation in Japanese though there are some differences. As Japanese, 

Sinhala employs a Q-particle, də, in both yes/no questions (45) and in wh-questions (46). 

(45) Chitra ee potə kieuwa də? 

Chitra that book read Q 

‘Did Chitra read that book? ’ Hagstrom 1998 (14a) 

(46) Siri 

mokak də keruwe? 

Siri what Q did-E 

‘What did Siri do?’ 

Hagstrom 1998 (12) 

In yes-no questions, 

the ‘Q’ morpheme də can appear clause finally, as shown in (45). As in 

Japanese, in wh-interrogatives (46), the object wh-element mokak ‘what’ remains in situ (or 

may 

move inside the IP/Mittelfeld; this issue does not bear on the analysis here). Crucially, as 

opposed to Japanese, the Q-particle must appear adjacent to the wh-element (at least in a 

matrix context) and in this case the verb has a special suffix, glossed as ‘E’. Though not 

shown here, any material between the wh-phrase and the 

Q-particle would lead to 

ungrammaticality. As pointed out by Hagstrom (1998), in wh-questions, ‘E’ suffixing 

contributes scope marking 

for the wh-phrase. Again, as in Japanese (see (9) above), the Q- 

273

particle adjacent to the wh-element gives rise to an existential reading. An example is given 

in (47). In existentials the verb does not bear the E-suffix. 

(47) Mokak də wć tuna. 

What Q fell 

‘Something fell.’ Hagstrom 1998, (19b) 

Furthermore, 

like the Japanese ka, the Sinhala də also occurs in disjunction, as in (48). 

(48) Mahatte atə tee də koopi də oonə? 

Gentleman.Dat tea Q cof fee Q necess ary 

‘Do you want tea or coffee?’ 

This is the kind of comparative data that leads Hagstrom to consider that the Sinhala particle 

is a Q-partic le. 

Let us focus on wh-questions in order to understa nd how Sinhala fits 

in the typology 

g iven. Consider more examples in (49) below. The Q-particle must 

follow the wh-word and 

the 

E-suffix appears on the verb. Sinhala ‘E’ is argued by Hagstrom and Kishimoto to 

indicate an unchecked feature which is then checked by covert movement 

of the Q-particle. 

(49) a. Chitra [mokak] də kieuwe? 

Chitra what Q read-E 

‘What did Chitra read?’ Hagstrom 2004, (4) 

b. Ranjit [kaa-ge potə] də kieuwe? 

Ranjit who-Gen book Q read-E 

‘Ranjit read whose book?’ Kishimoto 1992, (1) 

c. Chitra [kohee indan] də enne? 

Chitra where from Q come-E 

‘From where did Chitra come?’ Kishimoto 1992, 

(5) 

I suggest that, on analogy with 

the analysis proposed in Chapter V for Japanese, the Sinhala 

də, exactly like the Japanese 

ka, is the realization of the Force head in the left periphery. 

Consider the derivation of ( 49a), whose mechanics also holds of (49b) and (49c). On 

assumption it has an EPP, də triggers remnant movement of the SubjP containing the wh- 

274

phrase into Spec Focus. This is indicated by the dotted line in (2). Prior to remnant SubjP 

movement, the covert wh-feature of the Focus head agrees with wh-feature on the wh- 

element. As shown by (1), the verb moves to a V-related projection in the left periphery 

labelled FP, situated between FinP, which may be filled by the complementizer (as will be 

seen below), and the FocusP. 23, 24 The derivation of (49a), whose mechanics also holds of 

(49b) and (49c) is given below. 

(50) ForceP 

3 

Chitra mokak 3 

də FocusP 

3 

23 

(2) 

wh-f … FP 

3 

enne 

SubjP 

3 

Chitra ObjP 

3 

vP 

3 

(1) 

The need to postulate an independent V-related FP projection 

in the left periphery of the Sinhala clause is 

independently motivated by data like (i). 

(i) [Kau də aawe iiye kiyəla] Ranjit dannəwa. 

Who Q came-E yesterday that Ranjit know 

‘Ranjit knows 

who came.’ Hagstrom 2004 (44a) 

The presence of an adverb between the verb and the complementizer indicates that the verb cannot fill FinP as 

might have been envisaged. The complementizer kiyəla ‘that’ fills FinP and the adverb is in Spec ModifP in the 

sense of Rizzi (2002). 

24 

As discussed in Hagstrom (1998, 2004) and Kishimoto (1997, 2005), the Sinhala ‘E’ suffix occurring on the 

verb is an unchecked feature that gets deleted once movement of də has taken place. However, the approach 

suggested here brings Sinhala in line with other languages (French past participle 

agreement; wh-agreement 

languages like Chamorro) in that movement results 

in suffixation on the verb. In the matrix clauses in (49) and 

as in the representation (50), the verb leaves the IP/SubjP domain and moves to the projection labelled FP in the 

left periphe ry . Subsequent remnant SubjP movement across the verb induces E-morphology. This may be a 

result of SubjP transiting through Spec FP in its way to Spec 

Force. 

275

In the case of an island, such as the CNP island in (51) below, the Q-particle cannot appear 

inside (51b) but at the edge of the clause (51a) (Kishimoto 

1997, 2005, Hagstrom 1998). 

Hagstrom argues that in the case of strong islands, 

there is covert movement of the də-particle 

from the base-generated position as a sister to t he wh-word to a position external to the 

island. 

This movement is what he labels covert ‘Q-migration’. In the analysis suggested here, 

there is no covert or overt Q-movement. The Q-particle is merged in the Force head. 

(51) a. Oyaa [Chitra kaa-te dunnə potə] də kieuwe? 

You Chitra who-Dat gave book Q read-E 

‘You read the book that Chitra gave to whom?’ Kishimoto 1997, 33 

‘To whom did you read the book that Chitra gave t?’ 

b.*Oyaa [Chitra kaa-te də dunnə potə] kieuwe? 

(our glosses) 

Sinhala islands behave like Japanese islands. The analysis suggested here predicts that the 

entire island is pied-piped to the 

embedded Spec ForceP. This large pied-piped structure 

needs 

further move to matrix Spec ObjP. Such cases can be accounted for in terms of 

Chomsky’s (2000, 2001) 

phase-based theory: in order for some element to be extracted out of 

a phase, it must be located at the edge of that phase, either by merging into that position or by 

moving there. I suggest that in Sinhala the ForceP clausal structure is forced to move to the 

edge of the matrix vP phase and that this movement triggers agreement on the verb signalled 

by the E morphology. This E-suffixing 

also correlates with the fact that the wh-phrase inside 

the moved 

island has matrix scope. Movement to the edge of the phase determined by the 

EPP-feature on v° is represented below: 

(52) 

…[ObjP Obj° [vP v°+E [ForceP Force° [FocusP Focus° … 

276

One special case discussed in Hagst rom where movement of də is optional and overt, though 

generally, he argues i a wh-island under the factive verb dannəwa ’know’. 25 

t is covert is with 

Consid er the examples below. 

(53) a. Ranjit [kau 

də 

aawe kiyəla] dannəwa. 

Ranjit who 

Q came-E that 

know 

‘Ranjit knows who 

came.’ 

b. Ranjit [kaur u aawa də 

kiyəla] dannəwa. Kishimoto 1997, (6-7) 

The Q-particle which appears adjacent to the wh-word in (53a) overtly ‘migrates’ to the 

clause periphery of the embedded clause in (53b) if we are to use Hagstrom’s terms. Note that 

in (53b) it is no longer the E-morphology that marks the embedded scope of the wh-phrase 

but the position of the Q-particle. Under his analysis, the Q-particle undergoes head- 

movement into the interrogative C head. Such movement is in contradiction with 

strict 

locality 

conditions generally imposed on head movement (going back to Travis’ 1988 HMC). 

Cast in terms of the analysis suggested here, 

Hagstrom’s optionality of overt də 

‘migration’ reduces to the type of the moved constituent, 

i.e. an XP or a (sub-)clausal 

structure. Consequently, no stipulation about the head-particle feature insensitivity is needed 

to avoid the violation of strict locality. As the derivation in (54) given for (53a) shows, verb- 

movement to the left-peripheral FP takes place (1) and subsequent remnant SubjP is attracted 

to 

Spec Force (2). This derivation resembles that in (50). As in (50), remnant SubjP 

movement to Spec ForceP results in E-morphology on the moved verb aawe ‘came-E’. Next, 

ForceP, the complement of the matrix verb, pied-pipes to matrix Spec ObjP passing through 

the edge of vP. 26,27 

25 

As first discussed in Kishimoto (1997) and then in Hagstrom (1998), other verbs displaying the same 

selectional properties include doubt and look into but exclude ask and say. 

26 

If one adopts Shlonsky’s (2006) idea that all heads and specifiers within CP are computed as part of the CP 

phase edge, then remnant SubjP movement in matrix cases like (49)/(53), or the island examples like (51), all 

involve the E agreement marker on V moved to the left periphery. 

27 

I have considered here that the ForceP structure passes through the edge of the phase, matrix Spec vP. This 

position 

is viewed as an outer specifier of v. This is not incompatible with the antisymmetry as this specifier 

may not actually count as it is a phase-derived position. Spec ObjP may 

alternatively be viewed as allowing 

movement out of the phase given that, according to the data presented, it fills a position 

above vP. 

277

(54) … SubjP 

2 

Ranjit ObjP 

2 

vP 

2 

2 

dannəwa ForceP 

2 

kau 2 

də FocusP 

(3) 2 

wh-feature FP 

2 

aawe FinP 

2 

kiyəla SubjP 

(2) 2 

kau vP 

(1) 2 

aaw 

(53b) is represented in (55). No verb movement takes place and SubjP gets attracted to the 

left periphery. Then ForceP, complement of the matrix verb, pied-pipes to matrix Spec vP 

and then to ObjP. 

(55) … SubjP 

2 

Ranjit ObjP 

2 

vP 

2 

2 

dannəwa ForceP 

2 

2 

də FocusP 

2 

wh-feature FinP 

2 

kiyəla SubjP 

2 

kau vP 

2 

aaw 

278

On a Kaynean analysis of the clause structure, Sinhala has a low that complementizer which I 

assume to fill Rizzi’s 

(1990, class lectures 2007) FinP position. 

As illustrated 

in (56), Sinhala also exhibits the co-occurrence of a declarative 

complementizer and an interrogative one but significantly, as opposed to (53b) above, də 

fo llows the complementizer 

kiyəla. The co-occurrence of the two complementizers is highly 

reminiscent of Dutch, illustrated in (57) where a wh-phrase can 

co-occur with the 

interrogative complementizer of 

and the declarative one dat. 

(56) a. ? Ranjit [Chitra monəwa kieuwa 

də-nć ddə kiyəla] də danne? Kishimoto 1997 

Ran jit C hitra what read whether that Q know-E 

‘Ranjit knows whether Chitra read what?’ 

b.?* Ranjit [Chitra monəwa də kieuwa 

də-nć ddə kiyəla] danne? 

(57) Ik vraag me af wie of dat er morgen 

komt. 

I wonde r who if that there tomorrow 

comes 

‘I wonder who comes tomorrow.’ Koopman 1996 (20a) 

Both the Sinhala case in (56a) and the Dutch example in (57) are accounted for by positing a 

richer left periphery (see Koopman 

1996). Consider the representation of (56a) in (58). 

279

(58) …SubjP 

2 

Ranjit ObjP 

2 

vP 

2 

(4) 2 

danne Force 

2 

(3) 2 

də FocusP 

2 

IntP 

2 

də-nć ddə 2 

FinP 

2 

kiyəla SubjP 

2 

(2) Chitra ObjP 

2 

monəwa 

vP 

2 

kieuwa 

In th e derivation in (58), the wh-object monəwa ‘what’ va cates the vP and moves to Spec 

ObjP. 

kiyəla ‘that’ heads its own projection, FinP. I suggest that the interrogative 

complementizer də-nć ddə ‘whether’ occupy the specifier of a dedicated interrogative 

projection, IntP, in the sense of Rizzi (2001a). As shown by (2), the entire SubjP gets pied- 

piped to Spec 

Force due to the EPP associated with də on Force° (some speculations on this 

type of selection at a distance will be made in section 6). Movement of the ForceP structure to 

the edge of the phase, triggered by the EPP on v°, 

induces E-morphology on the matrix verb 

( 1) 

danne ‘know-E’ that agrees with the moved structure. 

(59) 

Sinhala also allows multiple questions. This is exemplified in (59). 

a. Kauru mokak də kieuwe? 

Who what Q read-E 

‘Who read what?’ 

b. * Kau də mokak kieuwe? Hagstrom 1998, (3) 

280

As described in Hagstrom, the Q-particle must follow the lowest wh-element. Such cases can 

be 

analysed exactly as the simple questions in (49), namely, as verb movement to the V- 

related 

projection FP followed by movement of the remnant SubjP containing the two wh- 

elements to Spec 

Force. As above, movement of the remnant across the moved verb results in 

E-suffixing. That the 

verb in Sinhala moves to a fairly low position in the left periphery of the 

clause is also suggested 

by Hagstrom (2004:18) within a head-final analysis of Sinhala: 

“…the verb moves up in standard fashion to a position quite close to the Focus° head, close 

enough that the feature on the Focus° head responsible for the E-morphology can be spelled 

o ut as a suffix on the verb when 

present…” 

6. A Note on the Locality of Selection 

A few remarks about the locality of selection in the cartographic model are in order at this 

point. Obviously, it cannot be straightforwardly expressed in terms 

of sisterhood or local 

government. The analysis so far h as relied in many cases 

on the selection of the complement 

of a head, say, Force, and on movement to the specifier of Force if there was an EPP. 

The 

more general implications of the strict 

locality conditions of 

such movement are left out. 

Three distinct patterns may have emerged obvious in the discussion so far. One is the pattern 

Sinhala exhibits. Recall example (49) repeated below. 

(49) 

Chitra mokak də kieuwe? 

Chitra what Q read-E 

‘What did Chitra read?’ 

If Force° occupied by də selects its sister, it should select FocusP. However, if the analysis 

proposed is correct, it selects and attracts SubjP/IP (Chitra mokak ) into its specifier 

across Focus and other potential intervening projections. The question is how selection works 

in this case. A second general pattern exemplified by the Tumbuka case (14) repeated below 

regards the way Force° filled by kasi somehow ‘communicates’ with its complement 

FocusP 

without attracting into its specifier. 

281

(14) Kasi Suzo a-ka-p h ik-a vici 


‘What 

did Suzo cook?’ 

The 

third pattern is exemplified by Japanese. In (6), the Q-feature on Force° attracts its 

complement FocusP into its specifier. More generally, all left-peripheral heads show strict 

locality conditions in moving their complements into the specifier position. 




In (4) repeated below, the highest projection in the split CP domain is ForceP, while the wh- 

operator sits in a lower Spec FocusP and the question one has to answer is how the locality of 

semantic or categorical 

selection across intervening categories is ensured. As observed by 

Rizzi 

(1997), the minimal contents of the CP layer are the specification of Force and its 

paired finiteness feature. 

(4) [ForceP [TopP* [FocP [TopP* [FinP … ]]]]]] 

C is thus a Force feature paired with a finiteness feature. When C is merged, both Force and 

Fin enter into the syntactic computation and must be assigned a value. A solution to the 

problem raised is to resort to Shlonsky’s (2006) mechanism of copy and remerge of C. It 

holds that the initially merged Force° is associated with an ordered set of precompiled 

features which correspond to Rizzi’s 

Focus, Topic, Fin, etc. In order to 

activate 

one in the set, C is remerged. Essentially, as he argues, information about the feature 

contents travels up and down the extended C projection independently 

of the particular order 

of the projected features. 

In a similar fashion it may 

tentatively be suggested that C and its ordered set of features 

also 

have an EPP. The selectional feature EPP is transmitted all the way up to C via copy and 

remerge and must be satisfied at each step, i.e. upon insertion/activation of an (ordered) 

feature. Thus, in Japanese, the EPP feature is satisfied immediately, i.e. upon insertion of 

each ordered feature in the left periphery (this is related to its head-final character). 

Conversely, in Sinhala, EPP is satisfied on the highest copy of C. Concretely, də on this 

282

highest C can attract SubjP at the last step of the mechanism, when the rightmost feature of 

the set, i.e. Force, entails copy and remerge of the highest (i.e. last) occurrence of C. No 

problem 

is posed by the intervention of a FocusP or a FP hosting the moved verb, as in (49). 

Information 

related to EPP travels up and down the extended C projection independently of 

the particular order of features. Whereas 

Japanese satisfies EPP upon every copy and remerge 

of C and Sinhala awaits upon insertion of the highest C, Tumbuka satisfies the selectional 

requirements of C via Agree only. 

7. Concluding Remarks 

In this chapter I have suggested an antisymmetric typology of question formation based on 

the cartographic framework. Some conceptual implications regard the fact that the 

antisymmetric-cartographic approach has 

the advantage of the comparative standpoint: (i) a 

s trict universal order of the phrase 

structure; (ii) it presupposes universal merge of a Q-feature 

and 

a wh-feature in the left periphery identified with distinct projections, ForceP and FocusP; 

(iii) all parameters are termed in the overt syntax (and morphology) in the movement/nonmovement 

distinction and further in the type of the moved constituent. So far the parametric 

model for wh-construction stands entirely in the overt syntax (no phrase structure parameter, 

no overt/covert movement and no PF computation, i.e. linear ordering). 

One major principle of the typology is that two types of features, i.e. a Q-feature and a 

wh-feature, 

in the interrogative construction enter into AGREE with a matching goal. One of 

the parameters distinguishing among languages 

reads as follows: ‘MOVE if Q on Force or wh 

on 

Focus is associated with an EPP’. The other regards the overt/covert morphological 

realization of Q/wh. These combinatorial possibilities account for the existence of whmovement 

in languages like Vata and Tlingit (which overtly realize one feature), on the one 

hand, and French and Romanian (no overt realization), on the other, and its absence in 

Japanese, Sinhala, Chinese, Tumbuka and French on the in-situ strategy. 

We suggested that Sinhala exhibits an agreement marker induced by movement to the 

edge of the phase. 28 

28 

The following generalisations can be made: (i) when Q is overt, it tends to have an EPP 

See also Rackowski and Richards (2005) for a discussion of Tagalog which has cases of agreement 

morphology that agrees with a CP complement. 

283

feature, whereas when covert, it tends to be –EPP. This correlates with the oft-mentioned 

observation that overt morphology is the trigger of movement; (ii) when the wh-feature is 

overtly realized, it tends to have an EPP-feature, whereas when covert, both options occur; 

(iii) apparently there is a gap which consists in the non-existence (rather rare existence?) of 

an overt wh-particle with a –EPP feature. This last fairly premature generalisation needs 

further investigation. 

284

CHAPTER VII 

CONCLUDING REMARKS 

This dissertation has first explored several issues involving the structure of the Romanian 

Mittelfeld and of the left periphery, on a comparative basis with other Romance languages. 

Secondly, it has also focused on the behaviour of wh-phrases and wh-movement in Romanian 

and the lack thereof in Japanese, the choice of the two languages having to do with the fact 

that both are situated at the opposite ends of what may be called the wh-continuum. This 

comparative study has then constituted the subject matter of the last chapter which sets forth a 

typology of question formation mainly relying on the overt/covert existence of the Q- and wh- 

particle and EPP associated with either or both. 

To summarize, here are the main results and some of their implications. 

I follow Cinque’s (1999) suggestion that the space between TP and vP/VP contains more 

functional architecture in addition to the fixed, rigidly ordered hierarchy of AdvPs. It is due to 

the finer articulation of this space – ‘the fine structure of the Mittelfeld’ – that we have arrived 

at unveiling (micro-)parameters regarding to subject, object and verb movement possibilities. 

I have departed from the view that there is a parallelism between Rizzi’s (1997) ‘fine 

structure of the left periphery’ and the vP periphery and suggested instead that at least as far 

as Romanian is concerned, the fine structure of the Mittelfeld contains ‘floating’ Topic- and 

Focus-related features on SubjPs and ObjPs interspersed among distinct Adv-related 

projections. 

The Romanian Mittelfeld in the VSO configuration has a focus-related SubjP position 

that is not available in Italian. To be more precise, we have identified several such positions, 

floating positions, that occur above the Tanterior projection and even in the upper portion of the 

285

Mittelfled, i.e. among Tanterior and Modepistemic. Verb movement possibilities in Romanian lend 

further support to the widely-acknowledged idea that the verb can target Fin°. 

The preverbal subject the more marked SVO configuration in Romanian has been shown 

not to fill a truly A’-position but rather fill the Criterial Spec SubjP where it values its Topic- 

aboutness feature. The DP subject must vacate the vP and move to a position above TAnterior 

and can further move to higher positions available in the upper portion of the Mittelfeld (but 

not above Moodspeech act or Moodevidential adverbs, which are parenthetical), whereas Italian and 

Spanish subjects have the option of moving above these adverbs. 

In the VOS configuration the subject DP moves to a low Subj° situated at the vP 

periphery, whose EPP is associated with New Information Focus. The object DP may function 

either as Topic (together with the verb) or as New Information Focus. It can move quite high 

in the Mittelfeld, i.e. to a position above the Asphabitual adverb, and even (more) marginally 

above Modepistemic. 

Another topic addressed regarding the Romanian Mittelfeld is the ordering of clitic 

adverbs which are shown not to constitute a homogenous class. A closer investigation of their 

properties has revealed that their ordering reflects the mirror image of Cinque’s aspectual 

classes. 

It is also in the finer articulation of the Mittelfeld that a solution has been found to the 

optionality problem of clause-internal ‘scrambling’ in Japanese. I have argued that the 

occurrence of a phrase to the left of its canonical position is motivated by Topic- and Focus- 

features intimately associated with the EPP present on Subj° and Obj° and inducing therefore 

certain interpretive effects. 

Thus the DP subject in the SOV configuration, which is a derived word order, must 

vacate the vP domain and can move as high as TP and mood adverbs but cannot sit 

immediately above a celerative adverb projection, whereas the object can move below high 

adverbs but not any higher. The subject can occupy the Criterial SubjP and the object DP can 

fill a position associated with New Information Focus. 

In OSV, the object DP fills ObjP associated with a Topic-feature, while the subject 

occupies SubjP whose head hosts a New Information Focus-feature. Under the appropriate 

intonation contours, the object has been identified to fill Top in the left periphery, while the 

subject remains lower, i.e. in Spec Subj associated to a New Information Focus-feature. A 

third possibility occurs the object, namely to move to Spec Foccontrastive in the left periphery. 

Similarly, wh-phrases do not remain in situ and do move inside the Japanese Mittelfeld. In 

286

particular, in SOV, the object wh-phrase exhibits the same movement possibilities as the its 

non-wh counterpart. 

The ordering of the wh-phrases in the Japanese Mittelfeld, an instance of A-movement, 

reflects the ordering of the wh-phrases in the Romanian left periphery, an instance of A’movement. 

The highest space a wh can fill in both languages correlates with the D-linked 

interpretation. This provides further support to the claim that the order of the wh-phrases in 

the left periphery of Romanian (and Bulgarian) reflects their order prior to wh-movement. 

If in addition to purely phi- (and Case) feature purposes, subject and object DPs leave 

the vP/VP domain for such IS-related feature parasitic on phi-features, I then adopt the (fairly 

strong) view that it is not that at least one DP must remain inside vP (Alexiadou & 

Anagnostopoulou 2001, 2006) but rather both subject and object DPs must vacate vP. This 

view receives even stronger support from Japanese, where I hope to have shown that DP 

clause-internal leftward scrambling is after all an instance of movement to a projection 

associated with a topic feature (of a certain type) or to another associated with a focus feature 

(of a certain type). 

This dissertation has also dealt with the question of locality in syntactic operations. The 

discussion focused on the locality constraints that hold in the process of feature valuation, on 

the one hand, and in the operation of movement and chain formation, on the other. The 

valuation procedure is subject to a configurational constraint. The valuation procedure in the 

case of Criterial features can only apply to two configurations: the head-head configuration 

and the Spec-head configuration. 

A criterial Spec-head Foc configuration has been argued to satisfy the Wh-Criterion 

(Rizzi 1990) in Romanian, the same configuration also holding of focalised elements. More 

generally, the Spec-head configuration results both in A’-feature valuation, i.e. criterial 

feature valuation, and A-feature valuation, i.e. Mittelfeld-internal, non-criterial feature 

valuation. 

The approach adopted which combines cartography and MULTIPLE AGREE makes 

recourse to pied-piping of subclausal structures. Two pied-piping mechanisms have been at 

stake: a wh-chunk pied-piping to Spec FocP in Romanian, and overt movement of ever larger 

pied-piped portions of subclausal structure, from complement to specifier position, as has 

been proposed for Japanese and other languages discussed in Chapter VI. This latter type of 

movement, which exhibits strict locality conditions needs further looking into. 

287

This type of analysis coupled with the type of wh-feature, i.e. substantive or purely 

formal, can account for the matrix vs. embedded scope reading of a wh or of wh’s in 

Japanese. 

We have carried the analysis of Japanese, an OV language with an overt Q-particle in 

sentence-final position, one step further: we have proposed a typology of question formation 

based upon Q- and wh- feature split, on their (c)overt realisation and their possibly being 

associated with an EPP-feature. This typology relies on AGREE (or MULTIPLE AGREE) and /or 

MOVE from the wh-feature on Focus°. All this leads to distinct types of movement 

possibilities. Besides, these theoretical assumptions can account for the existence of wh- 

movement in languages like Vata and Tlingit (which overtly realize both the Force and Foc 

features in the case of the former, and only the wh-feature in the latter), on the one hand, and 

French and Romanian (no (obligatory) overt realization), on the other, and its absence in 

Japanese, Sinhala, Chinese, Tumbuka and French on the ‘in-situ’ strategy. 

All parameters are termed in the overt syntax (and morphology) in the movement/non- 

movement distinction and in the type of the moved constituent. The parametric model for wh- 

construction as proposed in this dissertation stands entirely in the overt syntax (no phrase 

structure parameter, no overt/covert movement and no PF computation, i.e. linear ordering). 

I hope to have shown in this study that the issue of word order variation, overt wh- 

movement in a multiple fronting language and wh-topicalisation and wh-focalisation can be 

adequately handled by means of a few principles and assumptions that concern the theory of 

movement and its locality conditions, and the theory of feature valuation/checking and its 

conditions of application in a model of grammar in which the Syntax-Information Structure 

interface plays crucial role. 

These discussions are framed in a model of grammar where syntax interfaces with these 

IS-features which the Numeration contains in addition to Case and phi-features. Since (certain 

types of) Foci and Topics are associated with particular intonations, or are defined in terms of 

prosodic prominence, IS is also tuned in with PF. The architecture of the grammar as has been 

conceived of in this dissertation is the following: 

288

(1) Numeration:{Case, phi-features, Topic, T, 

Syntax/Phase 

LF(+Information Structure) Word order PF (prosody) 

Focus, etc.} 

Syntax therefore plays with two sets of features, uninterpretable Case and phi-features, on the 

one hand, and interpretable IS-features, on the other. The interplay of such features gives rise 

to distinct word orders that the Mittelfeld displays. Such feature can also come to be criterial 

features in the left periphery and are precompiled. Once Match and Agree, and/or Move have 

taken place, syntax ‘communicates’ with the LF and PF interfaces. We have assumed in this 

work that IS is a subcomponent of LF, where semantic interpretation of such a 

valued/checked feature takes place. One may also wonder whether IS could also be viewed as 

an autonomous component. I leave this as an open question. It is also to note that in some way 

or another (LF +) IS also ‘communicates’ with PF. The exact ways this is done remains a 

question for future research. 

289

BIBLIOGRAPHY 

Aboh, E. (1999) From the syntax of Gungbe to the Grammar of Gbe. Edition à la Carte SA. Sierre, 

Suisse. 

Aboh, E. (2004) “Topic and Focus within D.” In Linguistics in the Netherlands 2004, 1-12. 

Aboh, E. (2007) “Information Structuring begins with the Numeration.” Ms. Universiteit van Amsterdam. 

Alboiu, G. (2002) The Features of Movement in Romanian. Bucharest University Press, Bucharest. 

Alexiadou, A. and E. Anagnostopoulou (1998) “Parametrizing AGR: Word order, V-movement and EPP- 

Checking.” Natural Language and Linguistic Theory 16, 491-539. 

Alexiadou, A. and E. Anagnostopoulou (2001) “The Subject-in-Situ Generalization, and the Role of Case 

in Driving Computations.” Linguistic Inquiry 32:193-231. 

Alexiadou, A. and E. Anagnostopoulou (2006) “The Subject-In-Situ Generalization Revisited.” In U. 

Sauerland & H-M Gärtner (eds) Interfaces + Recursion = Language? Chomsky's Minimalism and 

the View from Syntax and Semantics. Mouton de Gruyter, Berlin. 

Aoun J. et al. (1987) “Two Types of Locality.” Linguistic Inquiry 18, 537-577. 

Aoun J. (1986) Generalized binding: The syntax and logical form of wh-interrogatives. Foris 

Publications, Dordrecht:. 

Barbosa, P. (1995) Null Subjects. Ph.D. dissertation, MIT. 

Barbosa, P. (2001) “On inversion in wh-questions in Romance.” In A. Hulk and J-Y Pollock (eds.), 

Subject Inversion in Romance and the Theory of Universal Grammar. Oxford University Press, 

Oxford. 

Baunaz, L. (2008) DP-Splitting: a Syntactic Approach to French quantification, PhD dissertation, 

University of Geneva. 

Beck, S. (2006) “Intervention effects follow from focus interpretation.” Natural Language Semantics 14, 

1-56. 

Belletti A. (2001) “Inversion as focalization.” In Subject Inversion in Romance and the Theory of 

Universal Grammar, A.Hulk and J.Y.Pollock (eds.), 60-90. Oxford University Press, New York. 

Belletti, A. (2004) “Aspects of the low IP area.” In L. Rizzi (ed.) The Structure of IP and CP. The 

Cartography of Syntactic Structures, vol.2. 16-51. Oxford University Press, New York. 

Belletti, A. (2006) “Clefts and wh in situ: some notes.” Lisbon COST-meeting. 

Belletti, A. and U. Shlonsky (1995) “The Order of Verbal Complements: a Comparative Study.” Natural 

Language and Linguistic Theory, Vol. 13, 489-526. 

Benincà, P. and C. Poletto (2004) “Topic, Focus, and V2: Defining the CP sub-layers.” In Rizzi, L. (ed.), 

The Structure of CP, Oxford University Press, Oxford. 

Benincà, P. and C. Poletto (2005) “On Some Descriptive Generalisations in Romance.” In Cinque, G. and 

R. Kayne (eds) The Oxford Handbook of Comparative Syntax. 221-259. Oxford University Press. 

Bhattacharya, T. and A. Simpson. (2003) “Obligatory overt wh-movement in a wh-in-situ language.” 

Linguistic Inquiry 34, 127-142. 

Bocci, G. (2007) “Criterial Positions and left Positions in Italian.” In Nanzan Linguistics, Special Issue3. 

Bolinger, D. (1965) Forms of English: Accent, Morpheme, Order. Edited by Isamu Abe and Tetsuya 

Kanekiyo. Cambridge, Mass.: Harvard University Press. 

Bolinger, D. (1978) "Asking More than One Thing at a Time." In H. Hiz (ed.) Questions. 107-150. 

Dordrecht, Reidel. 

Bošković, Z. (1998) “Multiple wh-fronting and economy of derivation.” In E. Curtis, J. Lyle, 

and G. Webster (eds.). The proceedings of the Sixteenth West Coast Conference on Formal 

Linguistics, 49-64. Stanford, Calif.: Stanford University, Center for the Study of language and 

Information. 

Bošković, Ž. (2002) “On multiple wh-fronting”. Linguistic Inquiry 33:351-383. 

Bošković, Ž. and D. Takahashi (1998) “Scrambling and Last Resort.” Linguistic Inquiry 29:347-366. 

291

Bošković, Z. (2001) On the Nature of the Syntax-Phonology Interface: Cliticization and Related 

Phenomena. Elsevier Science, Amsterdam. 

Bošković, Ž. (2004) “Topicalization, focalization, lexical insertion, and scrambling.” Linguistic Inquiry 

35:613-638. 

Bromberg, S. (1987) “What we know when we don’t know why.” In N. Rescher (ed.) Scientific inquiry in 

philosophical perspective, 75-104. Lanham, MD, University Press of America. 

Bruening, B. (2007) “Wh-in-Situ Does not Correlate with Wh-Indefinites or Question Particles.” Linguistic 

Inquiry 38, 139-166. 

Büring, D. 2003 “On D-Trees, Beans, and B-Accents". Linguistics & Philosophy 26:5. 511-545.. 

Cable, S. (2006) “Q-Particles and the nature of Wh-Fronting.” In Matthewson, L. (ed.) Quantification: 

Universals and Variation, Elsevier. 

Calabrese, A.. (1992) “Some Informal Remarks on Focus and Logical Structures in Italian.” In Kuno S., 

Thrainsson, H., (eds.), Harvard Working Papers in Linguistics, Vol. 1. 

Cardinaletti, A. (2004) “Towards a cartography of subject positions.” In L. Rizzi, (ed.). The structure of 

IP and CP. The Cartography of Syntactic Structures, vol. 2, 115-165. Oxford University Press, New 

York. 

Cardinaletti, A. (1997) “Subjects and clause structure.” In L. Haegeman The New Comparative Syntax, 

33-63. Longman. 

Cardinaletti, A. (2007) “Subjects and Wh-Questions. Some New generalisaitons.” Ms. Università 

Ca’Foscari, Venezia. 

Chang, L. (1997) “Wh-in-situ phenomena in French”, Master’s thesis, University of British Columbia. 

Cheng, L. (1991) On the typology of wh-questions. PhD dissertation, MIT, Cambridge, Mass. 

Cheng, L.-S. and J. Rooryck (2000) “Licensing Wh-in-Situ.” Syntax 3, 1-19. 

Choe, J.-W. (1987) “LF Movement and Pied-Piping.” Linguistic Inquiry 18, 348-353. 

Choe, H.-S. (1995) “Focus and Topic Movement in Korean and Licensing.” In K. E. Kiss (ed.) Discourse 

Configurational Languages, 269-335. Oxford University Press, New York/Oxford. 

Choi, H.-W. (1999) Optimizing Structure in Context: Scrambling and Information Structure. Stanford, 

CSLI Publications. 

Chomsky, N. (1973) “Conditions on transformations. A festschrift for Morris Halle.” In S. R. Anderson & 

P. Kiparsky (eds) A Festschrift for Morris Halle, 232-286. Holt, Rinehart and Winston. New York. 

Chomsky, N. (1981) Lectures on Government and Binding, Foris, Dordrecht. 

Chomsky, N. (1995) The Minimalist Program, MIT Press, Cambridge, MA. 

Chomsky, N. (1998) “Some observations on economy in generative grammar.” In P. Barbosa et al. (ed.) Is 

the best good enough? Optimality and competition in syntax, 115-127. MIT Press, Cambridge,Mass. 

Chomsky, N. (2000) “Minimalist Inquiries: The framework.” In R. Martin, D. Michaels & J. Uriagereka 

(eds) Step by Step: Essays on minimalist syntax in honour of Howard Lasnik. 89-156. MIT Press, 

Cambridge, Mass. 

Chomsky, N. (2001) “Derivation by phase.” In M. Kenstowicz (ed.) Ken Hale: a Life in Language, 1-52. 

MIT Press, Cambridge, Mass. 

Chomsky, N. (2004) “Beyond Explanatory adequacy.” In Belletti A. (ed.) Structures and Beyond: The 

cartography of syntactic structure. 104-131. Oxford, Oxford University Press. 

Chomsky, N. (2005) “On Phases.” Ms., MIT, Cambridge, Mass. 

Cinque, G. (1990) Types of A-bar dependencies, Linguistic Inquiry Monographs 17, MIT. 

Cinque, G. (1993) “A null theory of phrase and compound stress.” Linguistic Inquiry 24:239-267. 

Cinque, G. (1994) "On the evidence for partial N movement in the Romance DP." In Cinque, G, J. Koster, 

J.-Y. Pollock, L. Rizzi & R. Zanuttini (eds.) Paths Towards Universal Grammar, 85-110. 

Georgetown University Press, Georgetown. 

Cinque, G. (1999) Adverbs and functional heads: A cross-linguistic perspective, Oxford 

University Press, New York. 

Cinque, G. (ed.) (2002) Functional Structure in IP and DP: The cartography of Syntactic Structures, vol 1, 

Oxford University Press, New York. 

292

Cinque, G. (2002b) “Complement and Adverbial PPs: Implications for Clause Structure”. Talk given at 

GLOW 25, Amsterdam, April 10. 

Cinque, G. (ed.) (2006) Restructuring and Functional Heads. The Cartography of Syntactic Structures, 

vol.4, Oxford University Press, New York. 

Cinque, G. and R. Kayne (eds) (2005) The Oxford Handbook of Comparative Syntax. Oxford University 

Press. 

Cole, P. and G. Hermon (1998) “The Typology of WH Movement.” Syntax 1, 221-258. 

Collins, C. (1990) “Why and how come”, in H. Demirdache & L.L. Cheng (eds.) Papers on wh- 

movement, Cambridge, MA, Distributed by Department of Linguistics and Philosophy, MIT, 31-45. 

Comorovski, I. (1996) Interrogative Phrses and the Syntax-Semantics Interface. Kluwer Academic 

Publishers, Dordrecht. 

Comorovski, I. (2005) “Intensional Subjects and Indirect Contextual Anchoring”. Paper Presented at the 

International Round Table on Time and Modality, Paris, December 8-10. 

Comrie, B. (1981) Language Universals and Linguistics Typology. Syntax and Morphology. Basil Blackwell, 

Oxford. 

Contreras, H. (1989) “Closed domains.” Probus 1, 163-180. 

Contreras, H. (1991) “On the position of subjects.” In Susan Rothstein (ed.) Perspectives on Phrase 

Structure: Heads and Licensing (Syntax and Semantics 25), 63-79.Academic Press, New York. 

Cornilescu, A. (1997) “Some notes on the syntax of the subject.” In Revue Roumaine de Linguistique 

XLII, 101-147, Bucharest, Editura Academiei Romane. 

Cornilescu, A. (2002) “The Double subject construction in Romanian.” In V. Motapanyane (ed.) 

Comparative Studies in Romanian Syntax, 83-133. 

Cornilescu, A. (2000) “Notes on the Interpretation of the Prepositional Accusative in Romanian.” 

Bucharest Working Papers in Linguistics, 91-107. University of Bucharest Review. 

Dauenhauer, N. M and R. Dauenhauer 2000. Beginning Tlingit. Sealaska Heritage Institute. Juneau, AK. 

Davidson, D. (1980) Essays on Actions and Events. Oxford: Clarendon Press. 

Deguchi, M. and Y. Kitagawa. (2002) “Prosody and wh-questions.” In M. Hirotani (ed.) Proceedings of 

the North East Linguistic Society 32, 73-92. GLSA Publications, Amherst, Mass. 

Diesing, M. (1992) Indefinites. MIT, Cambridge. 

Dobrovie-Sorin, C. (1994). The Syntax of Romanian, Mouton de Gruyter, Berlin. 

Endo, Y. (2007) Locality and Information Structure: A Cartographic Approach to Japanese, John 

Benjamins, Amsterdam. 

Ernst, T. (2002) The Syntax of Adjuncts. Cambridge University Press, Cambridge. 

Ernst, T. and C.C. Wang (1995) “Object preposing in Mandarin Chinese.” Journal of East Asian 

Linguistics 4:3, 235-60 

Etxepare, R. and M. Uribe-Etxebarria (2008) “The Syntax-Semantics Interface of Focus in Negative 

Sentences.” Paper presented at Incontro di Grammatica Generativa, Padova, February 21-23. 

Fanselow, G. (2004) “The MLC and Derivational Economy.” In A. Stepanov, G. Fanselow & R. Vogel 

(eds). The Minimal Linik Condition, 73-124.Mouton de Gruyter. 

Fiengo, R. (1977) "On Trace Theory." Linguistic Inquiry 8, 35-61. 

Fox, D. (2000) Economy and Semantic Interpretation. MIT Press, Cambridge, Mass. 

Frascarelli, M. (2000) The Syntax-Phonology Interface in Focus and Topic Constructions in Italian. 

Kluwer Academic Publishers, Dordrecht. 

Frey, W. (2000) “Syntactic Requirements on Adverbs”. In Fabricius-Hansen, C., Lang, E. & Maiernborn, C. 

(eds.) Approaching the Grammar of Adjuncts (ZASPIL 17), 107-134. ZAS, Berlin. 

Fukuda, M. (1991) “A movement approach to multiple subject constructions in Japanese.” Journal of 

Japanese Linguistics 13, 21–51. 

Fukui, N. (1993) “Parameters and Optionality.” Linguistic Inquiry 24, 399-420. 

Göbbel, E. (2005) "Focus and Marked Positions for VP Adverbs". In S. Winkler & K. Schwabe (eds.) On 

293

Information Structure, Meaning and Form. S. Winkler & K. Schwabe (eds.). John Benjamins, 

Amsterdam/Philadelphia. 

Grewendorf, G. (2001) "Multiple wh-fronting". Linguistic Inquiry 32, 87-122 

Grewendorf, G. and J. Sabel 1999”Scrambling in German and Japanese: Adjunction versus 

Multiple Specifiers.” Natural Language and Linguistic Theory 17:1–65. 

Grohmann, K. (2001) “Multiple Wh-Fronting and the Left Periphery: German = Bulgarian + Italian“. In 

Simon M. & J. Mittelstaedt (eds.) Georgetown University Working Papers in Theoretical 

Linguistics 2, 83-115. 

Grohmann, K. (1998) ‘Syntactic Inquiries into Discourse Restrictions on Multiple Interrogatives’. In W. 

Abraham (ed.) Groninger Arbeiten zur germanistischen Linguistik 

Gussenhoven 1992 Sentence accents and argument structure. In I. M. Roca (ed.) Thematic structure: its 

role in Grammar. Foris, Berlin. 

Gundel, J. (1974) “The role of topic and comment in linguistic theory.” PhD dissertation, University of 

Texas, Austin. 

Hagstrom, P. (1998) Decomposing Questions. PhD dissertation, MIT. 

Hagstrom, P. (2004) “Particle movement in Sinhala and Japanese.” In V. Dayal and A. Mahajan (eds) 

Clause Structure in South Asian Languages, 227-252. Kluwer, Dordrecht. 

Hale, K. (1973) “A Note on Subject-Object Inversion in Navajo.” In B. B. Kachru et al. (eds), Issues in 

Linguistics: Papers in Honor of Henry and Renée Kahane, 300-309. 

Halliday, A.K. (1967) “Notes on transitivity and theme in English – parts 1 and 2.” Journal of 

Linguistics 3/1-2: 37-81/199-244. 

Heck, F. (2004) A Theory of Pied Piping. PhD thesis, Universitat Tubingen. 

Heim, I. (1988) The Semantics of Definite and Indefinite Noun Phrases The Semantics of Definite and 

Indefinite Noun Phrases. Garland, New York. 

Higginbotham, J. (1995) Sense and Syntax. Oxford: Clarendon Press. 

Hill, V. (2002) “Adhering focus.” Linguistic Inquiry 33 (1), 164-172. 

Hill, V. (2006) “Stylistic Inversion in Romanian.” Studia Lingusitica 60 (2). 156-180. 

Hiraiwa, K. (2000) “Multiple Agree and the Defective Intervention Constraint in Japanese.” 67-80. 

MIT Working Papers in Linguistics 40. 

Holmberg 1999 “Remarks on Holmberg’s Generalization”, Studia Linguistica 53.1, 1-39. 

Hoshi, K. (2005) “Deriving Association with Focus in Japanese within the Single-Cycle System.” Ms. 

Keio University. 

Huang, J. (1982) Logical relations in Chinese and the theory of grammar. PhD dissertation, MIT. 

Isac, D. (2001) “The Force of Negative Moods.” Ms. UQAM. 

Ishihara, S. (2000) “Stress, focus, and scrambling in Japanese”, MIT Working Papers in Linguistics 39, 

151-185. 

Ishihara, S. (2002) “Invisible but audible wh-scope marking: Wh-constructions and deaccenting in 

Japanese.” In L. Mikkelsen and C. Potts Proceedings of the 21st West Coast Conference on Formal 

Linguistics (WCCFL 21), 180–193.Cascadilla Press, University of California, Santa Cruz 

Somerville, MA 

Ishihara, S. (2004) “Prosody by Phase: Evidence from Focus Intonation-Wh-scope Correspondence in 

Japanese.” In S. Ishihara, M. Schmitz & A. Schwarz (eds), Interdisciplinary Studies on Information 

Structure 1, 77-119. Universitätsverlag Potsdam, Potsdam 

Jackendoff R.S. (1972) Semantic Interpretation in Generative Grammar. MIT Press, Cambridge, Mass. 

Jayaseelan, K. A. (2001) “IP-internal Topic and Focus Phrases.” Studia Linguistica 55. 39-75. 

Jayaseelan, K. A. (2004) “Question Movement in Some SOV Languages and the Theory of Feature 

Checking.” Language and Linguistics 5, 5-27. 

Kayne, R. (1994) The Antisymmetry of syntax. MIT Press, Cambridge, Mass. 

Kayne, R. (2002) "On Some Prepositions that Look DP-internal: English of and French de". Catalan 

Journal of Linguistics vol.1. 71-116. 

294

Kayne, R. (2005) “Antisymmetry and Japanese.” In R. Kayne (ed.) Movement and Silence, 215-240. 

Oxford University Press, Oxford. 

Kimper, W. A. (2006) “Question Formation in the Karonga Dialect of Tumbuka.” In O. F. Arasanyin & 

M. A. Pemberton (eds) Selected Proceedings of the 36 th Annual Conference on African Linguistics, 

Cascadilla Proceedings Project, MA Somerville. 

Kishimoto, H. (1992) “LF Pied Piping: Evidence from Sinhala”, Gengo Kenkyuu 102, 46-87. 

Kishimoto, H. (1997) “Wh-in-situ and null operator movement in Sinhala questions.” Ms. Hyogo 

University of Teacher Education. Draft 10/21/97. 

Kishimoto, H. (2005) 

Kiss , K. E.1(998) “Identificational focus versus information focus.” Language 74. 245-273. 

Kitagawa, Y., D. Roehrs and S. Tomioka (2004) “Multiple Wh-Interpretations.” Proceedings of GLOW 

Asia 2003. 

Kitagawa Y. and J. D. Fodor (2006) “Prosodic Influence on Syntactic Judgments.” In G. Fanselow et al., 

(ed.) Gradience in Grammar: Generative Perspectives, 336-358. OUP, Oxford 

Ko J. (2007) “Asymmetries in scrambling and cyclic linearization.” Linguistic Inquiry 37.1, 49-83. 

Koopman, H. (1984) The Syntax of Verbs: from Verb Movement rules in the Kru Languages to Universal 

Grammar. Foris Publications, Dordrecht. 

Koopman, H. (1996) “The Doubly filled Comp filter, the Principle of Projection Activation and Historical 

Change.” In F. Dupuis & M. Duchaine (eds) Proceedings of DIGS IV Cambridge University Press. 

Koopman, H. (2005) “Korean and Japanese Morphology from a Syntactic Perspective.” Linguistic Inquiry 

36, 601-633. 

Koopman, H. and D. Sportiche (1986) "A Note on Long Extraction in Vata and the ECP." In Natural 

Language and Linguistic Theory, vol 4, no.3. 357-374. 

Koopman, H. and A. Szabolcsi (2000). Verbal Complexes. Current Studies in Linguistics, 34. MIT Press, 

Cambridge, Mass. 

Krapova, I. (2002) “On the Left Periphery of the Bulgarian sentence.” University of Venice Working 

papers in Lingusitics (12). 107-129. 

Krapova, I. and G. Cinque (2005) “On the Order of Wh-phrases in Bulgarian Multiple Wh-fronting.” vol. 

15, 171-197.University of Venice Working papers in Lingusitics. 

Kratzer, A. and J. Shimoyama (2002) “Indeterminate Pronouns: The View from Japanese.” In Y. Otsu 

(ed.) The Proceedings of the Third Tokyo Conference on Psycholinguistics (TCP 2002), 

1‐25, Hituzi Syobo, Tokyo. 

Kuno 1980 “The scope of the question and negation in some verb-final languages.” CLS 16, 155-169. 

Kuroda S.-Y. (1988) "Whether we agree or not: A comparative syntax of English and Japanese." 

Linguisticae Investigationes 12, 1-47. 

Kuroda S.-Y. 1980. Bun kouzou no hikaku [The comparison of sentence structures]. In Kunihiro, Tetsuya 

(ed.), Niti-Eigo Hikaku Kouza 2: Bunpou [Lectures on Japanese-English Comparative Studies 2: 

Grammar], 23-61. Tokyo: Taisyukan. 

Ladd, D. R. (1996) Intonational Phonology. Cambridge University Press. 

Laenzlinger, C. (2002) “A Feature-based Theory of Adverb Syntax.” Volume 3, 67-105. Université de 

Genève. 

Laenzlinger, C. (2004) ”The Feature-based Theory of Adverb Syntax.” In Austin, J. R., S. Engelberg, G. 

Rauh (éds.). Adverbials. The Interplay between Meaning, Context and Syntactic Structure. 205-252. 

Benjamins, Amsterdam, Philadelphia. 

Laenzlinger, C. (2005) “French Adjective Ordering: Perspectives on DP-internal Movement Types.” 

Lingua. 115/5, 645-689. 

Laenzlinger, C. (in progress) Elements of Comparative Generative Syntax: A Cartographic Approach. 

Laenzlinger, C. and G. Soare (2005) “Multiple Wh-Fronting in Romanian: A Cartographic Approach.” 

Bucharest Working Papers in Linguistics 1, 23-60. 

Lambova, M. (2001) “On A-bar movements in Bulgarian and their interaction”. The Linguistic Review 18, 

327-374. 

295

Lambrecht, K. (1994) Information Structure and Sentence Form. Cambridge, CUP. 

Larson, R. K (1985) “On the Double Object Construction.” Linguistic Inquiry 19: 335-391. 

Lasnik, H. and M. Saito (1984) “On the nature of proper government.” Linguistic Inquiry 15, 235-289. 

Lee, C. (1999) "Contrastive topic: A locus of the interface." In K. Turner (ed)The Semantics/Pragmatics 

Interface from Different Points of View 1, 317-41. Elsevier, London 

Lee, C. (2006) "Contrastive (Predicate) Topic, Intonation, and Scalar Meanings." In C. Lee, M. Gordon 

and D. Buring (eds) Topic and Focus: Crosslinguistic Perspectives on Meaning and Intonation , 

151-175. Springer. 

Li, Y.-H. A. (1992) “Indefinite Wh in Mandarin Chinese.” Journal of East Asian Linguistics 1, 125–155. 

Mahajan, A. (1990) The A/A-bar Distinction and Movement Theory. PhD dissertation, MIT. 

Mathieu, E. (2004) “The Mapping of Form and Interpretation: The Case of Optional WH movement in 

French.” Lingua 114, 1090-1132. 

May 1977 The Grammar of Quantification. Doctoral dissertation, MIT, Cambridge, Mass. 

Miyagawa, S. (1989) Syntax and Semantics 22: Structure and case marking in Japanese. Academic Press. 

San Diego. 

Miyagawa, S. (1996) ”Word Order Restrictions and Nonconfigurationality”, Proceedings of Formal 

Approaches to Japanese Linguistics 2, MIT Working Papers in Linguistics 29, 117-142. 

Miyagawa, S. (1997) “Against optional scrambling.” Linguistic Inquiry 28:1-26. 

Miyagawa, S. (2001) “EPP, scrambling, and wh-in-situ.” In M. Kenstowicz (ed.) Ken Hale: A Life in 

Language, 293-338. MIT Press, Cambridge, Mass. 

Miyagawa, S. (2005) “EPP and Semantically Vacuous Scrambling”, in J. Sabel & M. Saito (eds) The Free 

Word Order Phenomenon: Its syntactic Sources and Diversity, 181-220. Mouton de Gruyter. 

Miyagawa, S. (2006a) ”Moving to the Edge.” Proceedings of the 2006 KALS-KASELL International 

Conference on English and Linguistics, Pusan National University, Busan, Korea, 3-18. 

Miyagawa, S. (2006b) “On the ‘Undoing’ Property of Scrambling: a Response to Bošković.” Linguistic 

Inquiry 37, 607-624. 

Molnár, V. (2006) ”On Different Kinds of Contrast.” In V. Molnár, & Winkler, S. (eds), Architecture of 

Focus. 197-233. De Gruyter, Berlin. 

Molnár, V. and S. Winkler (2006) (eds.) Architecture of Focus. 197-233. De Gruyter, Berlin. 

Motapanyane, V. (1989) ”La position du sujet dans une langue à l’ordre SVO/VSO.” Rivista di 

Grammatica Generativa 14, 75-103. 

Motapanyane, V. (2002) ”Adhering Focus.” Linguistic Inquiry 33, 164-172. 

Motapanyane, V. 1994. “An A-Position for Romanian Subjects.” Linguistic Inquiry 25, 729-734 

Motapanyane, V. (1995) Theoretical Implications of Complementation in Romanian. Unipress, Padova. 

Munaro, N. (2002) "Splitting up subject-clitic verb inversion." In Beyssade, C., R.Bok-Bennema, F. 

Drijkoningen, and P. Monachesi (eds.) Romance Languages and Linguistic Theory 2000, 233-252. 

John Benjamins, Amsterdam/Philadelphia 

Munaro, N. and J.Y. Pollock (2005) “Wh-est-ce Questions, Subject Extraction, and Negation in NE 

Romance.” In Cinque, G. and R. Kayne (eds) (2005) The Oxford Handbook of Comparative Syntax. 

542-607. Oxford University Press. 

Nakanishi, K. (2001) “Prosody and Information Structure in Japanese: A Case Study of Topic Marker 

Wa.” In Noriko Akatsuka and Susan Strauss (eds.), Japanese/Korean Linguistics 10:434-447. 

Nishigauchi, T. (1990) Quantification in the theory of grammar, Kluwer, Dordrecht. 

Ohno, Y. (1989) “Mo.” In E. Bach et al. (eds.), Papers on Quantification, NSF Grant Report, Department 

Of Linguistics, UMass, Amherst, GLSA 224-250 . 

Ono, H. (2001) “EPP-driven XP Movement in Japanese.” In Proceedings of WECOL '99. Department of 

Linguistics, California State University, Fresno. 

Ordonez, F. (1998) “Post-verbal asymmetries in Spanish.” Natural Language and Linguistic Theory 16, 

313-346. 

Ordonez, F. (2000) The Clausal Structure of Spanish: A Comparative Study. Garland, New York 

Publishing House. 

296

Paul, W. 2002 “Sentence-internal topics in Mandarin Chinese: the case of object preposing.” Language 

and Linguistics 3:4 , 695-714. 

Pearson, M. (2000) "Two types of VO languages". In Svenonius, P. (ed.) The derivation of VO and OV. 

327-363. John Benjamins, Amsterdam. 

Pesetsky, D. (1982) Paths and Categories. Ph.D. Dissertation, MIT. 

Pesetsky, D. (1987) “Wh-in-situ: Movement and unselective binding.” In E. Reuland & A. ter Meulen 

(eds) The representation of (in)definiteness, 98–129. MIT Press. Cambridge, Mass. 

Pesetsky, D. and E. Torrego (2004) “The syntax of valuation and the interpretability of Features.” In S. 

Karimi, V. Samiian & W. Wilkins (eds) Clever and Right: a Festschrift for Joe Emonds. 

Pesetsky, D. (2000) Phrasal Movement and its Kin. MIT Press. Cambridge, MA. 

Poletto, C. (2000) The Higher Functional Field: Evidence from Northern Italian Dialects, Oxford 

University Press. New York & Oxford. 

Poletto, C. and J-Y. Pollock (2004) "On the Left Periphery of Some Romance Wh-Questions." In Rizzi, L. 

(ed.) The structure of CP and IP: The Cartography of Syntactic Structures. 251-296. New York & 

Oxford: Oxford University Press. 

Prince 1981 “Language and the law: a case for linguistic pragmatics.” Working papers in Sociolinguistics, 

112-160. Southwest Educational Development Laboratory, Austin. 

Rackowski, A. and N. Richards (2005) “Phase Edge and Extraction: A Tagalog Case Study. ” Linguistic 

Inquiry 36.4:565-599. 

Reichenbach, H. (1947) Elements of Symbolic Logic. New York: The Free Press. 

Reinhart, T. (1995) Interface Strategies. Utrecht: OTS Working Papers. 

Reinhart, T. (1997) “Quantifier-Scope: How labor is divided between QR and choice functions”, 

Linguistics and Philosophy 20, 335-397. 

Reinhart, T. (1981) Pragmatics and Linguistics: An Analysis of Sentence Topics. Philosophica 27: 53-94. 

Richards, N. (1997) What Moves Where When in Which Language. Doctoral Dissertation, MIT, 

Cambridge. 

Richards, N. (2001) Movement in language: Interactions and architectures. Oxford University Press, 

Oxford. 

Richards, N. (2000) “An Island Effect in Japanese”, Journal of East Asian Linguistics 9, 187-205. 

Rivero, M. L. (2004) “Datives and non-active voice/reexive clitic in balkan languages.” Balkan Syntax 

and Semantics, pages 237-267. 

Rizzi, L. (1990) Relativized minimality. MIT Press, Cambridge, Mass. 

Rizzi, L. (1991) Residual Verb Second and the Wh-Criterion.” Technical Reports in Formal and 

Computational Linguistics 2, University of Geneva 

Rizzi, L. (1997) “The fine structure of the left periphery.” In L. Haegeman (ed.) Elements of 

Grammar, 281-337. Kluwer, Dordrecht. 

Rizzi, L. (2001a) “On the position "Int"(errogative) in the Left Periphery of the Clause.” In 

G. Cinque & G. Salvi (eds) Current studies in Italian syntax: Essays offered to Lorenzo Renzi, 287- 

296. Elsevier, New York. 

Rizzi , L. (2001b) “Extraction from Weak Islands, Reconstruction, and Agreement.” Ms., University of 

Siena. 

Rizzi, L. (2004a) The structure of CP and IP: The cartography of syntactic structures, vol. 2, Oxford 

University Press, New York. 

Rizzi, L. (2004b) “Locality and the left periphery.” In A. Belletti (ed.) Structures and Beyond. 

The Cartography of Syntactic Structures, vol. 3, 104-131. Oxford University Press, New York. 

Rizzi, Luigi. (2005) On some properties of subjects and topics. In Brugé, L., Giusti, G., Munaro, N., 

Schweikert, W. and Turano, G. (eds.). Proceedings of the XXX Incontro di Grammatica Generativa. 

Venezia, Cafoscarina. 

Rizzi, L. (2006) “On the Form of Chains: Criterial Positions and ECP Effects.” In Cheng & N. Corver 

(eds), Wh-movement: Moving on, 97-134. MIT Press Cambridge, Mass. 

297

Rizzi, L. (2007) “On Some Properties of Criterial Freezing.” In CISCL Working Papers - STiL - Studies 

in Linguistics, Vol.1, 145-158. 

Rizzi, L. (2008) Some Consequences of Criterial Freezing. Ms, University of Siena. 

Rizzi, L. & U. Shlonsky (2007) “Strategies of subject extraction.” In H. M. Gärtner & U. Sauerland (eds) 

Interfaces + Recursion = Language? Chomsky's Minimalism and the View from Syntax-Semantics. 

115-160. Mouton de Gruyter, Berlin. 

Rochemont, M. S. (1986) Focus in generative grammar. Benjamins, Amsterdam. 

Rooth, M. (1985) “Association with focus.” Graduate Linguistics Students Association. 

Rooth, M. (1992) “A Theory of Focus Interpretation.” Natural Language Semantics 1.75-116. 

Roussou, Anna and Tsimpli Ianthi-Maria (2006) “On Greek VSO Again!” Journal of Linguistics 42, 317- 

354. 

Rudin, C. (1988) "On Multiple Questions and Multiple Wh-Fronting". Natural Language and Linguistic 

Theory 6, 445-501. 

Sabel, J. (2001) “Deriving multiple head and phrasal movement: the cluster hypothesis.” Linguistic 

Inquiry 32, 532-547. 

Saito, M. (1989) “Scrambling as semantically vacuous A’-movement”, in M. Baltin and A. Kroch (eds) 

Alternative Conceptions of Phrase Structure, 192-200. University of Chicago Press, Chicago. 

Saito, M. (1992) “Long distance scrambling in Japanese.” Journal of East Asian Linguistics 1, 69-118. 

Saito, M. (2003) 

Saito, M. (2004) “Scrambling and the functional interpretation of wh-phrases.” In Y.-S. Kim, B.-C. Lee, 

K.-J. Lee, H.-K. Yang, and J.-Y. Yoon (eds)Explorations in generative grammar: festschrift for 

Dong-Whee Yang, Y.-S. Kim, B.-C. Lee, K.-J. Lee, H.-K. Yang, and J.-Y. Yoon (eds), 571-588. 

Hankuk, Seoul. 

Saito, M. (2007) “Radical Reconstruction and the First-Position Effects.” Presentation Worksop on 

Romance-Japanese: Comparative Syntax and Language Acquisition, Siena, Italy, May5-6 

Saito, M. and N. Fukui (1998) “Order in phrase structure and movement.” Linguistic Inquiry 29, 43-474. 

Selkirk, E. (1995) “Sentence prosody: Intonation, stress and phrasing.” In J. Goldsmith (ed.) The 

Handbook Of Phonological Theory. 550-569. Blackwell, Cambridge. 

Shimoyama, J. (2006) “Indeterminate Phrase Quantification in Japanese.” In Natural Language 

Semantics. 

Shlonsky, U. (2004) “The form of Semitic nominals.” Lingua 114. 1465-1526. 

Shlonsky, U. (2005) “Where’s ‘why’ and how it gets there”, paper presented at the GeneWHa 

Workshop, December 16. 

Shlonsky, U. (2006) “Projection étendue et cartographie de SC.” Nouveaux cahiers de linguistique 

française 27, 83-93. 

Shlonsky, U. (2008) “Notes on wh in situ in French. ” To appear in Brugè, L. et al. A Festschrift for 

Guglielmo Cinque, Oxford University Press, Oxford. 

Shlonsky, U. and G. Soare (in progress) “Where’s Why?” 

Shyu S.-I. (1995) The syntax of focus and topic in Mandarin Chinese. PhD dissertation, University of 

Southern, California. 

Soare, G. (2007a) “A Cross-linguistics Typology of Question Formation and the Antisymmetry 

Hypothesis.” In G. Campbell Ellison & G. Soare (eds) Generative Grammar in Geneva 107-133. 

Soare, G. (2007b) “A Cross-linguistics Typology of Question Formation and the Antisymmetry 

Hypothesis.” Presented at TEAL-4, Hong Kong, December 30. 

Soare, G. (2008) ‘De ce’ in the Hierarchy of Wh-Phrases. Some Notes on the French ‘Pourquoi’. 

Presented at the 10 th Bucharest Conference of the English Department Bucharest 

June 5-7. 

298

Starke, M. (2001) Move Dissolves into Merge. PhD dissertation, University of Geneva. 

Stepanov, A. and Tsai, W.-T. D. (2008) “The hows and whys of how and why: Unselective binding and 

merger of wh-adjuncts.” NLLT 26 (3), 589-638. 

Szabolcsi, A. (1981) “The Semantics of Topic-Focus Articulation.” In J.A.G. Groenendijk et al. (eds) 

Formal Methods in the Study of Language, Mathematisch Centrum, Amsterdam. 

de Swart, P. M. Lamers and S. Lastrade (2008) “Animacy, argument structure, and argument encoding.” 

Lingua 118, 131-140. 

Tada. H. (1993) A/A’ partition in derivation. PhD dissertation, MIT, Cambridge, Mass. 

Tanaka, H. (1999) “LF Wh-Islands and the Minimal Scope Principle.” Natural Language and Linguistic 

Theory 17, 371-402. 

Tomioka, S (2007) “Contrastive Topics, Speech Acts, and (Anti-)Exhaustivity.” Talk given at Stony 

Brook University, February 16 th , 2007; available at 

http://www.ling.udel.edu/stomioka/papers/tomioka_stony_brook.pdf 

Torrego, E. (1995) “On the nature of clitic doubling.” In H. Campos & P. Kempchinsky (eds.), Evolution 

and revolution in linguistic theory, 399-418. Georgetown University Press, Washington. 

Torrego, E. (1998) “Nominative subjects and pro-drop INFL.” Syntax 1, 206–219. 

Torrego, E. (1984) "On Inversion in Spanish and Some of Its Effects." Linguistic Inquiry 15, 103-129. 

Tsai W.-T. D. (1994) On Economizing the Theory of A’-dependencies, PhD dissertation, MIT, Cambridge, 

MA. 

Tsai, D. (1994) On Economizing the Theory of A-bar Dependencies, PhD Dissertation, MIT, Cambridge, 

MA. 

Tsai, D. (2008) “Left periphery and how-why alternations”, Journal of East Asia Linguistics 17, 83-115. 

Ura, H. (1993) “Extraction of Doo and Its Implications.” Journal of Japanese Linguistics 15, 139-150. 

Ura, H. (1996) Multiple Feature-checking: A Theory of Grammatical Function Splitting. PhD dissertation, 

MIT. 

Uriagereka, J. (1995) “Aspects of the syntax of clitic placement in Western Romance.” Linguistic Inquiry 

26: 79-123. 

Uriagereka, J. (1995) “An F Position in Western Romance.” In K. Kiss (ed.) Discourse configurational 

Languages, 153- 175. Oxford University Press, Oxford and New York. 

Uriagereka, J. (1997) “Formal and Substantive Elegance in the Minimalist Program (On the Emergence of 

Some Linguistic Forms).” In M. Bierwisch et al. (eds.) The role of economy principles in linguistic 

theory, 170-204. Akademie Verlag, Berlin. 

Uribe-Etxebarria, M. (1991) “On the structural positions of the subject in Spanish, their nature and their 

consequences for quantification.” Ms. Univ. of Connecticut. 

Vallduví, E. (1992) The Informational Component. New York, Garland. 

Vermeulen, R. (2005) Possessive and adjunct multiple nominative construction in Japanese. Lingua 115: 

1329-1363. 

Zagona, K. (2002) The Syntax of Spanish. Cambridge University Press, Cambridge. 

Zubizarreta, M. L. (1994) “On Some Prosodically Governed Syntactic Operations.” In Cinque, G., Koster, 

J., Pollock, J.Y., Rizzi, L., Zanuttini R., (Eds.), Paths Toward Universal Grammar. Studies in 

Honor of Richard S. Kayne, 473-485. Georgetown University Press, Wash. 

Zubizarreta, M. L. (1998) Prosody, Focus, and Word Order. MIT Press, Cambridge, MA. 

Watanabe, A. (2001) “Wh-in-situ Languages”, in M. Baltin and C. Collins (eds) The Handbook of 

Contemporary Syntactic Theory, 203-225. Blackwell, Oxford. 

Watanabe, A. (1992) “Subjacency and S-Structure movement of wh-in-situ.” Journal of East Asian 

Linguistics 1, 255-291. 

Whitman, J. (2001) “Kayne 1994:P.143, Fn. 3.” In G. Alexandrova (ed.) The Minimalist Parameter, 77- 

100. John Benjamins, Amsterdam. 

Willie, M. A. (1991) Pronouns and Obviation in Navajo. PhD dissertation, University of Arizona. 

299

UNIVERSIT DE GENVE - Université de Genève

Create successful ePaper yourself

Delete template?

Save as template?